ensayos de política económica - año...

Ensayos de Política Económica – Año 2010

ensayo 2010 (formulitas).indd 3 5/2/11 8:42:19 AM

4 •

universidad católica argentina RectoR: PbRo. DR. VíctoR Manuel FeRnánDez

facultad de ciencias económicas Decano: DR. José MaRía Dagnino PastoRe

escuela de economía francisco valsecchi

DiRectoR: DR. PatRicio Millán sMitMans

Ensayos de Política Económica – Año 2010

Ensayos de Política Económica es una publicación anual de la Escuela de Economía Francisco Val-secchi de la Facultad de Ciencias Económicas de la Universidad Católica Argentina.

Editores: Francisco Ciocchini, Ernesto A. O’Connor, Marcelo F. Resico

Consejo Consultivo: Dagnino, Pastore José MaríaFerreres, OrlandoGonzález, Fraga JavierMachinea, José LuísMillán Smitmans, PatricioNogués, JulioPrat Gay, AlfonsoTami, FelipeVillanueva, Javier

Asistentes de edición: Estefanía Pozzi

Escuela de Economía Francisco ValsecchiFacultad de Ciencias EconómicasAv. Alicia Moreau de Justo 1400, 4º piso, oficina 54 (C1107AFB), Buenos Aires, Argentina.E-mail: [email protected] Teléfono: 4338-0834Para consultar la versión electrónica: http://www.uca.edu.ar

El contenido del presente informe es responsabilidad de sus autores y no compromete la opinión de la Pontificia Universidad Católica Argentina, se autoriza su reproducción citando la fuente.Los autores ceden sus derechos, en forma no exclusiva para que se incorpore la versión digital de los artículos al Repositorio Institucional de la Universidad Católica Argentina, o a otras bases de datos que considere de relevancia académica.

Esta publicación fue realizada en Noviembre de 2010.

Consejo Editorial: Fracchia, EduardoMonat, PabloNicchi, FernandoGrandes, MartínJacobo, AlejandroSánchez, GabrielSanchís Muñoz, GerardoGasparini, LeonardoMontoya, SilviaGarcía Cicco, JavierSabbioni Pérez, GuillermoCaballero, Alicia


• 5f a c u l t a d d e c i e n c i a s e c o n ó m i c a s

Ensayos de Política Económica - Año 2010

Índice

Editorial 6

Does spatial clustering help explaining differences in the inequality of income distribution? Evidence from Argentina 7

Alejandro Cañadas

Unemployment and private returns to higher education in Argentina (1974-2002) 32

Cecilia Adrogué

Estimación de una función de producción agregada: Argentina 1975-2006 54Fernando Suárez

Did bankruptcy reform lead to looser mortgage lending standards? Evidence from the U.S. mortgage market 2000-2007 88

Prajakta Bhide, Lucia Fiorito, Zachary Noteman y Kunal Sawardekar

Fiscal Imbalances, Inflation and Sovereign Default Dynamics 108Ramiro Sosa Navarro

Reseña Bibliográfica: ‘‘ El Regreso de Keynes”, de Robert Skidelsky 143Alejandro D. Jacobo

Reseña Bibliográfica: “Populismo o mercados. El dilema de América Latina”,de Sebastián Edwards 145

Eduardo Luis Fracchia

Presentación y selección de trabajos 147


6 •

e d i t o r i a l

Ensayos de Política Económica - Año 2010

Objetivos y cobertura temática

La “Revista Ensayos de Política Económica” es una publicación de la Escuela de Economía Fran-cisco Valsecchi de la Facultad de Ciencias Económicas de la Universidad Católica Argentina. Su primer número fue publicado en 2007.

La cobertura temática abarca la política económica y social, cubriendo áreas tales como macroeco-nomía, economía internacional, economía monetaria, economía financiera, políticas fiscales, crecimiento, desarrollo, historia de la política económica, instituciones, regulación económica, organización industrial, políticas sociales, mercado de trabajo, pobreza y distribución del ingreso, entre otros.

La publicación apunta a un balance entre cuestiones positivas y normativas de la política econó-mica, desde diversos enfoques, considerando particularmente dentro de las cuestiones normati-vas aquellos temas relevantes para la Doctrina Social de la Iglesia.

La Revista esta abierta a distintas colaboraciones. La selección de los trabajos se encuentra bajo la responsabilidad del Consejo Editorial, integrado por reconocidos investigadores de la Universidad y de otros Centros de Estudios. Los artículos recibidos son sujetos a un referato anónimo.

Editorial - Año 2010

Con esta nueva publicación se da a conocer el cuarto número de la Revista Ensayos de Política Económica. En el presente número, los artículos pueden agruparse a partir de las temáticas pre-sentadas. En primer lugar, dos artículos que tratan aspectos sociales como la desigualdad en la distribución del ingreso, desempleo y educación. Luego, se presenta un artículo sobre la temática del crecimiento con una estimación de la función de producción agregada para Argentina. El siguiente artículo trata temas financieros vinculados a la crisis de las hipotecas. El último artículo se ocupa de interrelaciones macroeconómicas entre déficit, inflación y deuda. La revista concluye con la sección dedicada a las reseñas bibliográficas, que analizan dos recientes obras. La primera, se refiere al legado de Keynes y su aplicación a la crisis económico-financiera actual. La otra, al dilema entre Populismo y Mercado en America Latina.

Por otra parte, informamos que la revista ha sido incorporada a la base de datos de EBSCO.


E n s A y o s d E P o l í T i C A E C o n ó M i C A – A ñ o 2 0 1 0


Does spatial clustering help explaining differences in the inequality of income distribution? Evidence from Argentina

Alejandro Cañadas1

AbstractThis paper analyzes the relationship between the spatial clustering of income distribution and inequality in the provinces of Argentina. The goal of the paper is to use spatial techniques to ana-lyze to what extent the spatial clustering of income distribution affects the inequality of income distribution in a regional context of Argentina. In general, the literature of inequality implicitly considers each region or province as an independent entity and the potential for observational interaction across space has often gone ignored. However, spatial autocorrelation occurs when the spatial distribution of the variable of interest exhibits a systematic pattern. I compute three measures of global spatial autocorrelation: Moran’s I, Geary’s c, and Getis and Ord’s G, as the degree of provincial clustering between 1991 and 2002. The paper’s main conclusion is that there is evidence that relatively high (low) unequal provinces tend to be located nearby other high (low) unequal provinces more often than would be expected due to random chance. Therefore each province should not be viewed as an independent observation, as it has been implicitly assumed in previous studies of regional income inequality.

Resumen

Este artículo analiza la relación entre la agrupación espacial de la distribución del ingreso y la desigualdad en las provincias de Argentina. El objetivo de este trabajo es usar técnicas espacia-les para analizar hasta que punto la agrupación espacial de la distribución del ingreso afecta la desigualdad de la distribución del ingreso en un contexto regional de Argentina. En general, la literatura de desigualdad implícitamente considera a cada región o provincia como una entidad independiente y el potencial para la observación de la interacción a través del espacio a menudo se ha ignorado. Mientras tanto, la autocorrelación espacial ocurre cuando la distribución espacial de la variable de interés exhibe un patrón sistemático. Yo computo tres medidas de autocorrela-ción espacial global: La I de Moran, c de Geary, y G de Getis y Ord, como grado de CLUSTERING provincial entre 1991 y 2002. La principal conclusión del trabajo es que hay evidencia que provin-cias con desigualdad relativamente alta (baja) tienden a ser localizadas cerca de otras provincias con alta (baja) desigualdad más a menudo de lo esperado debido al azar. Por ende cada provincia no debería ser vista como una observación independiente, como ha sido supuesto implícitamente en estudios previos sobre la desigualdad de ingresos regional.

Keywords:Inequality in income distribution, spatial autocorrelation, Argentina

JEL Classification: [D31] [R11] [R12]

1.- Alejandro Cañadas is Assistant Professor of Economics at Mount St. Mary’s University, Emmitsburg, MD. Address 16300 Old Emmitsburg Road, Emmitsburg, MD, [email protected]


8 •

1. Introduction

This paper deals with the research question: of whether spatial clustering helps in explaining dif-ferences in the inequality of the distribution of per capita income across regions in Argentina.

The paper has three main purposes. First, it shows the importance of studying inequality, particu-larly in its relationship with poverty and growth. Second, it assesses the income distribution of Ar-gentina and it compares it to that of other Latin American countries. Third, it analyzes Argentina’s regional differences in income distribution inequality using spatial econometrics tools.

Most economic analysis is concerned with inequality in the distribution of some measure of indi-vidual well-being. Inequality generally refers to a measure of dispersion in the distribution. Most measures used are consistent with certain desirable attributes, known as axioms of inequality measurement (Atkinson, 1970; Cowell and Jenkins, 1995; Cowell, 1998).

2. Importance of the Study of Inequality

As de Ferranti et al. (2004) conclude, inequality is pervasive. It characterizes every aspect of life, including access to education, health, and public services. It prevents access to land and other assets, and it affects the functioning of credit and formal labor markets. It excludes people from the attainment of political voice and influence (inequality of agency). Inequality in Latin America has been rooted in exclusionary institutions that have been perpetuated ever since colonial times. It has reduced the impact of economic growth on poverty reduction and, inequality is associated with a greater prevalence of social conflict and violence and it may impair an economy’s ability to respond effectively to macroeconomics shocks (de Ferranti et al., 2004).

3. Relationships among Poverty-Growth-Inequality

Bourguignon (2004) describes changes in poverty in a given period as reflecting growth in mean income and changes in the distribution of relative income. The decomposition illustrated in Fig-ure 1 in the Appendix corresponds to an identity described as the “Poverty-Growth-Inequality Triangle.”

A change in the distribution of income can actually be decomposed into two effects. First, there is the effect of a proportional change in all incomes that leaves the distribution of relative income unchanged (growth effect). Second, there is the effect of a change in the distribution of relative incomes (which, by definition, is independent of the mean), known as a distributional effect.

In Figure 1 in the Appendix, the poverty headcount is the area under the density curve to the left of the poverty line (here set at US$1 a day).2 The movement from the initial to a new distribution goes through an intermediate step, namely the horizontal shift of the initial density curve to curve (I). Because of the logarithmic scale, this change corresponds to the same proportional increase of all incomes and thus stands for the pure “growth effect.” Then, the movement from curve (I) to the new distribution occurs at constant mean income and it corresponds to the “distribution”

2.- This figure shows the density of the distribution of income, which is the share of the population at each level of income, where income is represented on a logarithmic scale on the horizontal axis.




effect. Both growth and inequality changes thus play a role in generating changes in poverty. The shaded areas to the left of the poverty line show these changes.

4. Relationship between Inequality and Growth

Beyond the Bourguignon (2004) identity, former levels of inequality may have an influence on future growth. The literature on inequality and economic growth is indeed rich. However, the empirical evidence about the relationship is mixed. On the one hand, the literature that uses OLS regressions over a cross-section of nations generally finds that initial inequality is negatively related with future growth, when considering over 30 years (Alesina and Perotti, 1994). On the other hand, the literature using panel data over shorter periods generally finds a positive inequal-ity-growth relationship (Li and Zou, 1998; Forbes, 2000).

In particular, Forbes (2000) suggests that one reason for the conflicting inequality-growth results in the literature is that the relationship may differ for short and for long periods (for example 5-10 versus 25-30 years). Forbes also notes that panel techniques, such as fixed-effect estimators, cap-ture how time-series changes in inequality within a country (or state) affect changes in its growth rate over a short period. In contrast, Barro (2000) argues that OLS models capture how persistent cross-sectional differences in inequality affect long-run growth rates, which is more relevant to understanding growth disparities. Therefore, the two methods may reflect different responses.

Recently, the influences of the New Economic Geography and spatial econometrics have shed light on the relationship between regional inequality and economic growth. There are a few studies of US regional growth, such as Partridge (1997). Using pooled OLS models, this author finds that inequality is positively related to growth. Panizza (2002), using panel data with fixed-effect models, finds that small specification changes can turn around these results. Thus, cross-state results can suffer from the same sign changes that characterize cross-country studies, when switching from OLS to panel approaches. Again, these results may suggest different short-term and long-term influences. This calls for careful specifications of the relationship.

An important advantage of these types of studies is that regions or states can be used as good laboratories to examine inequality-growth issues. For example, Partridge (1997, 2005) and Paniz-za (2002) both note that many of the hypotheses about these relationships should apply to states, because they are essentially small open economies with distinct histories and institutions. These authors also claim that, among states, there appears to be sufficient variation in income distribu-tion to produce differential outcomes, due to large factor flows across states. In contrast, greater legal and informational barriers would limit the flow of resources among countries, especially for low-income economies. This, in turn, would reduce the factor flows that produce larger growth rate differentials. Consequently, any income distribution-growth relationship should be much easier to detect using data for states rather than countries (Siebert, 1998).

Partridge (2005) studies these issues using data for 48 US states, over the 1960-2000 period, and he finds that inequality is positively related to long-run growth. In his paper, Partridge makes four important contributions. First, regarding the ambiguous findings in the literature when moving from cross-sectional to panel data methods, Partridge suggests that, instead of considering them as conflicting, these results should be considered as complements in the analysis. In general, con-flicting results from various methodologies may not be a signal of lack of robustness, if there are separate long-run and short-run linkages. The results support Temple (1999), who argued that a variety of cross-sectional and panel approaches are necessary to fully understand the determi-


10 •

nants of growth. Partridge’s conclusion is that “by examining separate short-run and long-run models, researchers can gain a more complete picture of transitory and dynamic responses and a better understanding of how policy affects economic processes” (p.389).

Second, after allowing for short-run and long-run effects, Partridge (2005) controls for different effects for the tails and the middle of the distribution. In this, he follows Easterly (2001), who argues that a middle-class consensus promotes growth by encouraging stability, mass education, and better public services. In effect, Partridge finds that a more vibrant middle class, measured by the middle-quintile income share (Q3), is positively related to growth.

Third, following Kaldor (1956), Partridge (2005) argues that income inequality generates incen-tives for resources to be channeled into more efficient uses and is conductive to saving and capi-tal accumulation. This may explain the positive inequality-growth relationships found. However, these hypotheses assume that there is sufficient factor mobility in a given society, which may not be true for some developing countries. In fact, Partridge (2005) makes it clear that these results are derived from the experience of advanced economies. It is thus interesting to test these hy-potheses for developing countries, where high inequality and slow growth have been present.

5. Inequality in Argentina

In order to assess the dynamics of income inequality in Argentina, I will use data from the SED-LAC, which is a database of socio-economic statistics constructed from microdata coming from the Latin American and Caribbean (LAC) household surveys and developed by CEDLAS (Univer-sidad Nacional de La Plata) and The World Bank’s LAC poverty group (LCSPP). All estimates are computed from the Encuesta Permanente de Hogares (EPH). This survey has been carried out by the Instituto Nacional de Estadística y Censos (INDEC) since the early 1970s in the Greater Buenos Aires area and since the 1980s in most large cities (with over 100,000 inhabitants), in two rounds: May and October.

During 2003, a major methodological change was implemented by INDEC, including changes in the questionnaires and in the timing of the survey visits. The new survey (known as EPH Conti-nua or EPH-C) is now conducted over the whole year. INDEC also started to provide population weights that take the income non-response problem into account. To assess the impact of these methodological changes, I present three sets of statistics for 2003 in most tables: one computed from the EPH carried out in May, and two computed from the EPH-C of the second semester of 2003. One of them is generated with the old weights (ignoring income non-response) and the other two use the new weights.

The EPH-C covers 28 conglomerates or large urban areas, which are home to around 70 percent of the Argentine urban population. Since the share of urban areas in Argentina is 87 percent of the total population, the sample of the EPH represents around 60 percent of the total popula-tion of the country. Household surveys in Argentina cover only urban areas (the same problem is found in Uruguay). However, both Argentina and Uruguay are two of the most urban countries in the world, with over 85 percent of the population living in cities.

In Argentina, like in many Latin American countries, household surveys have experienced signifi-cant improvements. In particular, major changes have been implemented since the early 1990s. Although these changes are very welcome, they pose significant problems for comparison pur-poses within countries over time. This is one reason why I decided to present data for Argentina




since 1992. Other reasons are the incorporation of major cities in 1998 and the change in meth-odology.

Each decile in Table 1 in the Appendix includes an equal number of individuals (not households). The income ratio 10/1 is the mean income of decil ten divided by the mean income of decil one. The ratio 90/10 is the mean income of percentile 90 divided by the mean income of percentile ten. Finally, the ratio 95/80 is the mean income of percentile 95 divided by the mean income of percentile 80.

The richest 10 percent of the population earned up to 40 percent of the total income, a peak reached in the first semester of 2002. By the second semester of 2005, this share had declined lo levels similar to those for a decade before. However, the post-crisis shares are higher than the pre-crisis shares of the Menem’s period (1989-1995 and 1995-1999). In contrast, the poorest 10 percent of the population earned as little as 1 percent of total income (2001 to 2003). These extremes coincided with the financial crisis in 2001. Afterwards, however, the share of the poor-est 10 percent never recovered to its 1992 level. This share was higher in the pre-crisis period. Thus, the rapid growth and stability of the pre-crisis era seem to have been associated with less inequality.

Table 2 in the Appendix shows several inequality indices related to the distribution of per capita household income: the Gini coefficient, the Theil index, the Coefficient of Variation (CV), the At-kinson index with parameters 0.5, 1 and 2, and the generalized entropy index, with parameters 0 and 2 (the Theil index is the entropy index with parameter 1). A simple correlation analysis among these indices shows that the Gini coefficient is highly correlated with the Theil index (95 percent), the Atkinson index with parameters 0.5, 1 and 2 (99 percent, 99 percent, and 96 percent, respec-tively), and with the generalized entropy index with parameter 0 (99 percent). Therefore, I can analyze the inequality in the distribution of income by just looking at the Gini coefficient. All the inequality indices were calculated from the various editions of the household surveys (Encuesta Permanente de Hogares, EPH). Table 2 compares all the indices taking care of all the modifications in the EPH between 1992 and 2005.

Figure 2 in the Appendix shows the Gini coefficient for Argentina. A major increase in inequality took place in the country since 1992 (when the Gini was 0.45 percent). After the Argentinean crisis in December of 2001, the Gini jumped to 0.533, and then it reached its maximum level dur-ing the first half of 2003, at 0.537. Since then, the level of inequality has declined slightly.

6. Inequality in Latin America

It is known that Latin America has been having high and persistent level of inequality. What explains the high level of inequality observed in Latin America? To a large extent, most inter-pretations pursue the colonial inheritance argument, together with the persistence of the initial institutions. Among others, de Ferranti et al. (2004) highlight the combined role played by fac-tor endowments and institutions. These authors explain that factor endowments, technology, and the relative scarcity of resources had important implications for the initial inequality. In Latin America, the characteristics of the colonies favored the establishment of large plantations (such as sugar) and mining activities that employed forced labor. As a result, a social structure emerged where a privileged few were in control of the most profitable activities and where most of the population was excluded from access to land, education, and political power. In contrast, the co-lonial powers in North America soon learned that there was no gold, there were few indigenous


12 •

peoples to exploit, and soils and climates would not support the production of crops based on large slave plantations. Interestingly, Argentina is very different compared to most Latin Ameri-can countries. In Argentina, there were no large plantations and mining activities that employed forced labor. Like in North America, land was cheap and labor scarce, while fertile soil and good weather conditions attracted migrants. These conditions might have explained the success of Argentina up to 1913.

Why did inequality persist over time? In answering this question, de Ferranti et al. (2004) argue that the persistence of inequality during the colonial and early independence period took place because the initial “nexus” of institutions continued to exist, as did the justification for these institutions. The elites that had benefited from colonial disparities were able to quickly gain effec-tive control of the independent countries and determine the general structure of the institutions in ways that favored their interests (Engerman and Sokoloff, 2006).

For many scholars, explaining the persistence of inequality over the 20th century is more chal-lenging, because significant social, economic, and political changes occurred during the 1900s. Moreover, for some authors the increase in urbanization rates should have somewhat mitigated the relevance of the highly inegalitarian pattern of land ownership and its impact on income inequality. Also, modernization moved most of the Latin American countries in the direction of more open and democratic societies. However, de Ferranti et al. (2004) believe that the most important causes for the persistence of inequality over the 20th century are the low quality of edu-cation, a development strategy based on import substitution and isolation from world markets, and imperfect financial markets that may have prevented the poorest from taking advantage of economic opportunities, by restricting their access to credit.

Unfortunately, there is no quantitative estimation of long-run inequality authenticating these arguments for Latin America. A good example is provided by Bourguignon and Morrisson (2002), who investigated the historical trends in world income inequality. In their studies, conventional wisdom and lack of empirical evidence led them to assume that no changes in income distribu-tion had taken place in Latin America from independence to the mid-20th century.

Some authors claim that it is possible to infer the evolution of inequality since 1950 on the basis of direct income distribution observations. Table 3 in the Appendix reports Gini coefficients for several Latin American countries. The table indicates that inequality continued to be essentially constant from 1950 to 2000, with a Gini between 0.51 and 0.55. There is, however, significant country heterogeneity. For instance, the Gini coefficient noticeably increased in Argentina, from 0.396 to 0.477 between 1950 and 1990, but it may have declined in Venezuela, from a high of 0.613 in the mid-20th century to 0.459 in 1990. Likewise, El Salvador may have experienced a major deterioration in inequality over the 1960–1990 period, while Peru saw some progress.

For the pre-1950 period, data availability prevents direct inequality assessments. One can still empirically investigate the evolution of income inequality using indirect indicators and ranges of country studies follow this approach. Bértola (2005) provides rough estimates of income distribu-tion and Gini coefficients for Uruguay that go back to the late 1800s. Williamson (1999) looked at the consequences for inequality of the early phase of globalization (1870–1914). This author showed an increase of within-country inequality for Argentina and Uruguay over that period, on the basis of the evolution of the wage–land rental ratio. Bértola and Williamson (2003) claim that inequality trends reverted in the interwar period, when the observed abrupt decline in the wage-rental ratio stopped. This ratio increased somewhat after the 1930s. Calvo, Torre, and Szwarcberg (2002) suggest that the level of inequality changed little during the 20th century in Argentina,




while Londoño (1995) claims that the inequality levels observed in Colombia during the 1990s were probably similar to those observed in 1938.

Prados de la Escosura (2005) builds on Williamson (2002) to explore the historical evolution of the ratio of GDP per worker to the unskilled wage between 1850 and 1950 (or earliest possible date) for Argentina, Brazil, Chile, Mexico, and Uruguay. The justification for this selection is that such a ratio compares the returns to unskilled labor with the returns to all factors of production. Since unskilled labor is the more equally distributed factor of production in developing countries, an increase in the ratio suggests that inequality is rising.

On this basis, Prados de la Escosura (2005) infers that in Argentina, Chile, and Uruguay income inequality does not seem to have changed much over the period, whereas Brazil and Mexico may have suffered some worsening in the distribution of income. On the whole, the evidence that emerges from these studies indicates that, on average, Latin America started the 20th century with a very high level of inequality, which continued for the rest of the century, regardless of large variations by country in special periods.

Table 4 in the Appendix compares the changes in inequality measured by the Gini coefficient, using household surveys for 18 Latin American countries. By focusing on the performance of in-equality between 1989 and 2004, Gasparini, Gutierrez and Tornarolli (2007) find that Argentina, Costa Rica, the Dominican Republic, Uruguay, and Venezuela consistently rank as the most equal economies in the region, while Bolivia, Brazil, Ecuador, Panama and Paraguay occupy the last po-sitions in the inequality ladder. However, Argentina and Colombia stand out as the countries that experienced the largest increases in inequality, with Gini changes of around 6 percentage points. Brazil and Mexico are the only countries that have experienced a drop in income inequality.

How do these tendencies contrast to those observed in developed economies? Spain, for in-stance, experienced an important decline in income inequality between the 1970s and the 1990s, when the Gini coefficient went down by more than 10 percentage points (Table 3 in the Appen-dix). Prados de la Escosura (2005) finds indirect indicators that suggest that income inequality has been declining in Spain since the 1950s, when Spain may have had inequality levels comparable to (if not higher than) those observed in Latin America. For 1950, Prados de la Escosura (2005) estimates a Gini coefficient for Spain above 0.50.

In the United States, at the beginning of the century inequality was very high, with a Gini of ap-proximately 0.60 in 1920 (Plotnick et al., 1996). Inequality reached its pre-World War II high in 1929. Then, from 1929 to 1951, income inequality fell dramatically, to a Gini of about 0.40.

The United Kingdom experienced a similar pattern. Acemoglu, Johnson, and Robinson (2002) show that the Gini coefficient for the United Kingdom might have been more or less 0.55 in the 1890s. After that, for most of the 20th century, inequality seems to have weakened. The authors also conclude that most of the decline of the United Kingdom’s inequality took place between 1940 and the late 1970s. Atkinson (2003) relies on income tax statistics to show that, in the early 1900s, the richest 1 percent in the United Kingdom shared almost 20 percent of total personal income; in the late 1970s, this group got 6 percent of this total.

Inequality in France evolved in about the same way. In the early 1900s, the share of income of the wealthiest percentile in France was about 20 percent, whereas in the 1980s it was approximately 7 percent. The main difference between France and the United Kingdom is that most of the de-cline in French income inequality took place between the 1920s and 1950.


14 •

Thus, while inequality in Latin America has been persistent and steady over the last century, inequality in Europe and the United States seems to have declined significantly over the 20th century. If other countries have managed to break with their histories on both the growth and income distribution fronts, then why Latin America cannot also break with its history? This ques-tion goes beyond the scope of this paper.

7. Regional Inequality in Argentina

Figure 3 in the Appendix shows the Gini coefficient for 23 provinces in Argentina, for the period 1991-2002. This coefficient ranges between 0.40 and 0.50. Figure 4 in the Appendix compares the regional Gini coefficient for six regions in Argentina (See Maps 1 and 2 in the Appendix). Two interesting conclusions emerge. First, during the 1991-2002 period, the regions that experienced some positive per capita GDP growth (Pampeana and Pa-tagonia) showed relatively less inequality, while the regions that experienced negative per capita GDP growth (Northwest, Northeast, and Cuyo) showed relatively higher inequality levels. Second, the capital of Argentina, Greater Buenos Aires (BA) showed the highest per capita growth rate and also the highest level of inequality (0.482).

To test for differences in the Gini coefficient among regions, I use the General Linear Model (GML) procedure that tests the null hypothesis that there is no difference in the mean of the Gini coefficient among the provinces in each region. The test rejects the null at the 1 percent level. So, after finding that differences exist among the means of the Gini coefficient across regions, I use the Bonferroni’s and Tukey’s tests to determine which means differ.3 Table 5 in the Appendix shows that both tests found that there are differences among Gini coefficient among regions. So, I can consider regions 2, 6, and 3 (Pampeana, Patagonia and Cuyo) as one cluster with lower Gini coefficients, and regions 5, 1 and 4 (Northwest, Capital City, and Northeast) as another cluster of higher Gini coefficients.

In order to analyze the development of inequality within each region in Argentina, Table 6 in the Appendix presents the percentage change of four measures related to the income distribution. Column 1 compares the percentage change in the Gini coefficient for the six regions of Argen-tina between 1991 and 2002. It shows that the Pampeana Region had the largest increment in inequality (19.3 percent, which represents a 0.079 points increment), followed by Greater Buenos Aires (17.1 percent, which represents a 0.078 points increment). Patagonia showed the smallest increase in inequality. Column 2 shows the changes in the share of the third Quintile (Q3), which accounts for a “middle class consensus” and the role of the median voter.4

Partridge (2005) explains that the Gini is used to control for the overall distribution, while the share of the third Quintile (Q3) can be used to account for that specific group in the population.

3.- The Bonferroni test, based on Student’s t statistic, adjusts the observed significance level for the fact that multiple comparisons are made. Tukey’s honestly significant difference test uses the Studentized range statistic to make all pairwise comparisons between groups and sets the experiment wise error rate to the error rate for the collection for all pairwise com-parisons. When testing a large number of pairs of means, Tukey’s honestly significant difference test is more powerful than the Bonferroni test. For a small number of pairs, Bonferroni is more powerful.4.- A quintile is any of the four values which divide the sorted data set into five equal parts, so that each part represent one fifth or 20 percent of the sample population. The third quintile represents the group of population between the 40 and 60 percent of income levels.




Comparing across the regions of Argentina, Table 3 indicates that the share of total income that the “middle class group” earned during this period has been getting smaller in all regions. The decline amounts between 6 percent (0.059 points) in Cuyo to 15 percent (0.067 points) in the Northeast region of Argentina.

Finally, I compare the changes in the shares of the richest 10 percent of the population and poor-est 20 percent of the population, in each region. In column 3, with the exception of Pampeana, in all the other regions the richest 10 percent of the population gained, up to additional 14.6 percent (4.8 percentage points) of total income in the Northeast. In contrast, in all the regions, the share of the poorest 20 percent of the population declined. The worst negative effect was suffered by Buenos

8. Spatial Autocorrelation of Income Inequality

Now I want to test if spatial autocorrelation characterizes the measures of inequality among the provinces of Argentina. Figure 5 in the Appendix displays the Moran’s I statistic for the provincial Gini coefficients in Argentina between 1991 and 2002. 5 It shows that the Moran’s I statistic has been fluctuating during this period.

Table 7 in the Appendix presents the estimates for the Moran’s I statistic. For the 1991-2002 pe-riod, I estimated the coefficients using the EPH. The table shows that there is evidence of spatial dependence, as the statistics are highly significant during this period.

The Moran’s I statistic corroborates that positive spatial autocorrelation exists. The same results are found using the Geary’s c, and Getis and Ord’s G statistics. That is, the value taken by the Gini coefficient at each province i tends to be similar to the values taken by the Gini coefficient at spatially contiguous provinces.

8.1 Local Spatial Autocorrelation for Income Inequality

Figures 6 and 7 in the Appendix offer a more disaggregated view of nature of the spatial au-tocorrelation for the initial (Figure 6) and final (Figure 7 and 8) years. Each figure contains a Moran scatterplot for the Gini coefficient. The slope of the regression line equals Moran’s I = 0.015 for 1991, and Moran’s I = -0.111 for 2002. The Moran scatterplot is divided into four quadrants:

The upper right quadrant represents spatial clustering of a high-Gini province with high-Gini •neighbors (HH-quadrant I). In general, these locations are associated with positive values of the local Moran Ii.The upper left quadrant represents spatial clustering of a low-Gini province surrounded by •high-Gini neighbors (LH-quadrant II). In general, these locations are associated with negative values of the local Moran Ii.The lower left quadrant represents spatial clustering of a low-Gini province surrounded by •

5.- The reason why I am considering the period 1991-2002 in order to calculate the Moran’s I statistic for the provincial Gini Coefficient is because only during those years the “Encuesta Permanente de Hogares” (EPH) includes 23 provinces from a total of 24 provinces, and only the province of Rio Negro is not included in the sample.


16 •

low-Gini neighbors (LL-quadrant III). In general, these locations are associated with positive values of the local Moran Ii ; and The lower right quadrant represents spatial clustering of a high-Gini province with low-Gini •neighbors (HL-quadrant IV). In general, these locations are associated with negative values of the local Moran Ii .

Viewing Figures 6 and 7 together corroborates the lack of stability in the measures of local spatial dependence for the Gini coefficient. While in Figure 6 there is no sign of local spatial autocorrela-tion at all in 1991, in Figure 7 there seems to be some negative spatial autocorrelation in 2002. However, the local Moran’s I in those years, 1991 and 2002, is not statistically significant. Then the local Moran’s I was considered in 2001. In summary, the relationship is not stable.

From Figures 6 and 7 it can be concluded that there is not a clear pattern of clustering for the Gini coefficient. More specifically, only in half of the years, a pattern given by a positive spatial autocorrelation for the provincial Gini coefficient can be observed. Thus, only for the years 1992 to 1994, 1997 and 1998, and 2001, provinces have the local indicators that significant fall in either quadrant I or III of the scatterplot, reflecting HH and LL clustering, respectively.

Concentrating only on those years where the Moran’s I shows statistically significant local spatial autocorrelation, two clusters were identified. First, there is the cluster of high-Gini province with high-Gini neighbors, represented mainly by the provinces of the Northeast region like Chaco, Formosa, Misiones and some provinces from the Northwest region, including Catamarca, Jujuy, Tucuman and Santiago del Estero, each of which appears in quadrant I. The other main cluster of a low-Gini province surrounded by low-Gini neighbors (LL) includes provinces from the Pampeana region, Cuyo and Patagonia, such as Buenos Aires, Capital City, and La Pampa (Pampeana); Men-doza, San Luis, and San Juan (Cuyo); Chubut, Santa Cruz and Tierra del Fuego (Patagonia), all of which fall in quadrant III, the vast majority of the years. These results corroborate the findings using Bonferroni and the Tukey’s tests (see Map 3 in the Appendix).

Finally, as a measure of robustness of these results, the global and local measures of spatial auto-correlation were estimated while changing the W matrix. All the previous results were obtained using wij as an element of a spatial weights matrix W such that wij = distance in kilometers between each provincial capital city (location i) to all the others provincial capital cities (location j), using a cutoff point of 800 km and the actual routes available in Argentina. Alternatively, I calculated wij as an element of a spatial weights matrix W such that wij = number of hours that it takes to drive from location i to location j, using seven hours as the cutoff point. The same significant results were obtained.

9. Conclusion

Latin America is characterized by high and persistent level of inequality. The high level of in-equality observed in Latin America could be explained by a social structure emerged where a privileged few were in control of the most profitable activities and where most of the population was excluded from access to land, education, and political power. The persistence of inequality over the 20th century is due to the low quality of education, a development strategy based on import substitution and isolation from world markets, and imperfect financial markets that may have prevented the poorest from taking advantage of economic opportunities, by restricting their access to credit.




Argentina is a very special case; it has also experienced the second largest increase in inequality in the region with Gini changes of around 6 percentage points between 1992 and 2004.

So, when spatial clustering of income distribution was tested using the Gini coefficient, there is some evidence that at least 50 percent of the time the Moran’s I shows statistically significant lo-cal spatial autocorrelation. Concentrating on those years, two clusters were identified. First, there is the cluster of high-Gini province with high-Gini neighbors, represented mainly by the provinces of the Northeast region like Chaco, Formosa, Misiones and some provinces from the Northwest region, including Catamarca, Jujuy, Tucuman and Santiago del Estero, each of which appears in quadrant I. The other main cluster of a low-Gini province surrounded by low-Gini neighbors (LL) includes provinces from the Pampeana region, Cuyo and Patagonia, such as Buenos Aires, Capital City, and La Pampa (Pampeana); Mendoza, San Luis, and San Juan (Cuyo); Chubut, Santa Cruz and Tierra del Fuego (Patagonia), all of which fall in quadrant III, the vast majority of the years. These results corroborate the findings using Bonferroni and the Tukey’s tests for differences in the Gini coefficient among regions for the whole sample (see Map 3 in the Appendix).

Therefore, there is evidence to show that in Argentina between 1991 and 2002 relatively high (low) unequal provinces tend to be located nearby other high (low) unequal provinces more often than would be expected due to random chance. Therefore each province should not be viewed as an independent observation.

For future research these results should be analyzed together with the real per capita growth in GDP in order to discover any pattern of spatial interaction between inequality and growth that would help us understand such complex relationship a little better.

10 References

Acemoglu, D., S. Johnson, and J. Robinson (2002). Reversal of Fortune: Geography and Institu-tions in the Making of the Modern World Income Distribution. The Quarterly Journal of Eco-nomics, 117(4), 1231-1294.

Alesina, A., and R. Perotti. (1994). The Political Economy of Economic Growth: A Critical Survey of the Literature. World Bank Economic Review, 3, 351–371.

Atkinson, A. (1970). On the Measurement of Inequality. Journal of Economic Theory, 2, 244-263.

Atkinson, A. (2003). Top Incomes in the United Kingdom over the Twentieth Century. U.K.: Uni-versity of Oxford.

Barro, R. J. (2000). Inequality and Growth in a Panel of Countries. Journal of Economic Growth 5(1), 5-32.

Bértola, L. (2005). A 50 años de la curva de Kuznets, una reivindicación sustantiva: distribución del ingreso y crecimiento económico en Uruguay y otros países de nuevo asentamiento desde 1870. Instituto Laurentano Figuerola de Historia Económica, Working Paper Series 05-04(Universidad Carlos III de Madrid).

Bértola, L., and J.G. Williamson (2003). Globalization in Latin America before 1940. National Bureau of Economic Research, Working Paper 9687. Cambridge, MA: National Bureau of Economic Research..

Bourguignon, F. (2004). The Poverty-Growth-Inequality Triangle. The World Bank Policy Research Working Paper (28102).

Bourguignon, F., and C. Morrisson (2002). Inequality among World Citizens: 1820-1992. The American Economic Review, 92(4), 727-744.


18 •

Calvo, E., J. Torre, and M. Szwarcberg (2002). The New Welfare Alliance. Buenos Aires: Universi-dad di Tella, Department of Political Science.

Cowell, F. (1998). Measurement of Inequality. Paper presented at the London School of Econom-ics and Political Science, London.

Cowell, F. A., and A. Jenkins. (1995). How Much Inequality Can We Explain? A Methodology and an Application to the United States. Economic Journal, 105, 421-430.

De Ferranti, D., G. E. Perry, F. Ferreira, and M. Walton. (2004). Inequality in Latin America & the Caribbean: Breaking with History? Washington D.C.: World Bank.

Easterly, W. (2001). The Middle Class Consensus and Economic Development. Journal of Develop-ment Economics, 6, 317–335.

Engerman, S., and K. Sokoloff. (2006). The Persistence of Poverty in the Americas: The Role of Institutions. In S. Bowless, S. Durlauf, and K. Hoff (ed.), In Poverty Traps: Forthcoming.

Forbes, K. J. (2000). A Reassessment of the Relationship Between Inequality and Growth Ameri-can Economic Review, 90, 869–887.

Gasparini, L., F. Gutierrez, and L. Tornarolli. (2007). Growth and Income Poverty in Latin America and the Caribbean: Evidence from Household Surveys. Review of Income and Wealth, 53(2), 209-245.

Kaldor, N. (1956). Alternative Theories of Distribution. Review of Economic Studies, 23 83–100.Li, H., and Hengfu Zou. (1998). Income Inequality is Not Harmful for Growth: Theory and Evi-

dence. Review of Development Economics, 2, 318–334.Londoño, J. L. (1995). Distribucióon del Ingreso y Desarrollo Económico: Colombia en el Siglo

XX. Bogotá.Panizza, U. (2002). Income Inequality and Economic Growth: Evidence from American Data. Jour-

nal of Economic Growth 7(1), 25-41 Partridge, M. (1997). Is Inequality Harmful for Growth? Comment. The American Economic Re-

view, , 87, 1019–1032.Partridge, M. (2005). Does Income Distribution Affect U.S. State Economic Growth? Journal of

Regional Science, 45(2), 363–394.Plotnick, B., E. Smolensky, E. Evenhouse, and S. Reilly. (1996). The Twentieth-Century Record of

Inequality and Poverty in the United States. In G. Engerman and R.Gallman (ed.), The Cam-bridge Economic History of the United States (Vol. II). Cambridge, U.K.: Cambridge University Press.

Prados de la Escosura, L. (2005). Growth, Inequality, and Poverty in Latin America: Histori-cal Evidence, Controlled Conjectures. Economics History and Institutions Working Papers, 54104(Universidad Carlos III, Departamento de Historia Económica e Instituciones, Madrid).

Siebert, H. (1998). Commentary: Economic Consequences of Income Inequality. Paper presented at the Symposium of the Federal Reserve Bank of Kansas City on Income Inequality: Issues and Policy Options.

Temple, J. (1999). The New Growth Evidence. Journal of Economic Literature, 37, 112–156.Williamson , J. G. (1999). Real Wage Inequality and Globalization in Latin America before 1940.

Revista de Historia Económica, 17, 101-142.Williamson , J. G. (2002). Land, Labor, and Globalization in the Third World, 1870-1940. Journal

of Economic History, 62(1), 55-85.




11 Appendix

Source: http://209.15.138.224/argentina_mapas/m_rArgeninaPolitic.htm


20 •

Map 1-Map of the provinces of Argentina

Note: Metropolitana is the Ciudad Autonoma de Buenos Aires, Noroeste is Northwest, Nordeste is Northeast.

Source: http://209.15.138.224/argentina_mapas/m_rArgentinaZonas.htm




Map 2-Map of the regions of Argentina


22 •

Map 3- Two main clusters for the Gini in the provinces of Argentina using the Moran I, 1991-2002

Source: Bourguignon (2004).




Share of deciles Income ratios

1 2 3 4 5 6 7 8 9 10 10/1 90/10 95/80

EPH-15 cities

1992 1.8 3.0 4.1 5.1 6.2 7.6 9.4 12.0 16.5 34.1 19.0 7.9 2.0

1993 1.7 3.0 4.1 5.2 6.4 7.9 9.6 12.3 16.6 33.1 19.9 8.1 1.9

1994 1.7 2.9 4.0 5.1 6.3 7.7 9.5 12.1 16.4 34.2 19.7 8.2 1.9

1995 1.4 2.7 3.7 4.8 5.9 7.3 9.1 11.6 16.7 36.7 25.8 9.6 2.1

1996 1.4 2.6 3.6 4.7 5.9 7.3 9.2 11.9 17.0 36.5 26.5 10.1 2.0

1997 1.4 2.6 3.6 4.7 6.0 7.3 9.2 12.0 17.2 36.1 26.7 10.5 2.1

1998 1.2 2.4 3.4 4.5 5.7 7.0 9.0 12.0 17.1 37.7 30.2 11.2 2.1

EPH - 28 cities

1998 1.3 2.4 3.4 4.5 5.7 7.1 9.0 11.9 16.9 37.8 29.9 11.1 2.1

1999 1.3 2.5 3.5 4.6 5.8 7.3 9.2 12.0 17.0 36.8 28.0 10.9 2.1

2000 1.2 2.3 3.3 4.4 5.6 7.2 9.1 12.2 17.4 37.4 32.3 11.9 2.1

2001 1.0 2.1 3.1 4.1 5.4 6.9 9.0 12.0 17.4 39.0 40.0 13.9 2.2

2002 1.0 2.0 3.0 4.1 5.4 6.9 8.7 11.6 17.2 40.3 39.4 14.3 2.3

2003 1.1 2.1 3.0 4.0 5.2 6.8 8.8 11.9 17.3 39.8 34.8 13.5 2.2

EPH-C

2003-II 1.0 2.1 3.1 4.1 5.3 6.7 8.8 11.9 17.1 39.8 38.1 13.7 2.2

2004-I 1.2 2.3 3.3 4.3 5.5 7.1 9.0 11.9 16.8 38.6 32.7 11.8 2.1

2004-II 1.1 2.3 3.3 4.4 5.7 7.2 9.1 12.0 17.0 37.9 33.0 12.0 2.0

2005-I 1.2 2.4 3.4 4.4 5.7 7.3 9.1 11.9 16.9 37.8 32.5 11.7 2.1

2005-II 1.2 2.3 3.4 4.5 5.8 7.3 9.1 11.9 16.8 37.6 32.7 11.8 2.1

Note: Income distribution for the population in major urban cities of Argentina.

Source: Constructed by the author using Socio-Economic Database for Latin America and the Caribbean (CEDLAS and The World Bank).

Figure 1- Decomposition of a change in distribution and poverty into growth and distributional effects.


24 •

Gini Theil CV A(.5) A(1) A(2) E(0) E(2)

EPH-15 cities

1992 0.450 0.370 1.101 0.165 0.299 0.510 0.355 0.606

1993 0.444 0.359 1.077 0.162 0.297 0.517 0.352 0.580

1994 0.453 0.378 1.112 0.168 0.303 0.510 0.361 0.618

1995 0.481 0.430 1.205 0.190 0.340 0.569 0.416 0.726

1996 0.486 0.442 1.260 0.194 0.349 0.607 0.429 0.793

1997 0.484 0.422 1.146 0.190 0.346 0.586 0.424 0.656

1998 0.502 0.471 1.300 0.207 0.369 0.608 0.461 0.845

EPH - 28 cities

1998 0.502 0.472 1.307 0.207 0.368 0.605 0.458 0.854

1999 0.491 0.443 1.213 0.197 0.356 0.606 0.440 0.735

2000 0.504 0.464 1.231 0.208 0.377 0.647 0.474 0.757

2001 0.522 0.497 1.264 0.224 0.404 0.675 0.517 0.798

2002 0.533 0.530 1.356 0.233 0.412 0.657 0.530 0.920

2003 0.528 0.519 1.343 0.227 0.401 0.637 0.512 0.902

EPH-C

2003-II (*) 0.537 0.625 3.056 0.244 0.417 0.673 0.539 4.671

2003-II 0.529 0.532 1.457 0.231 0.407 0.672 0.522 1.061

2004-I 0.510 0.507 1.714 0.216 0.380 0.621 0.478 1.469

2004-II 0.506 0.499 1.550 0.213 0.379 0.624 0.476 1.201

2005-I 0.502 0.473 1.306 0.208 0.373 0.624 0.466 0.853

2005-II 0.501 0.480 1.418 0.209 0.373 0.624 0.467 1.005

Note: (*) this calculation uses the EPH weights corresponding to the 28 major provincial cities. CV=coefficient of variation. A(e) refers to the Atkinson index with a CES function with parameter e. E(e) refers to the generalized entropy index with parameter e. E(1)=Theil.

Source: Constructed by the author using Socio-Economic Database for Latin America and the Caribbean (CEDLAS and The World Bank).

Table 1- Distribution of household per capita income in Argentina (deciles shares and income ratios), 1992-2005.




Note: (*) this calculation uses the EPH weights corresponding to the 28 major provincial cities.

Source: Constructed by the author using EPH.

Table 2- Inequality Indices from household surveys in major provincialcities in Argentina, 1992-2005


26 •

Country 1950 1960 1970 1980 1990

Argentina 0.396 0.414 0.412 0.472 0.477

Bolivia 0.53 0.534 0.545

Brazil 0.57 0.571 0.571 0.573

Chile 0.482 0.474 0.531 0.547

Colombia 0.51 0.54 0.573 0.488 0.503

Costa Rica 0.5 0.445 0.485 0.46

Dominican Republic 0.455 0.421 0.481

El Salvador 0.424 0.465 0.484 0.505

Honduras 0.618 0.549 0.57

Mexico 0.55 0.606 0.579 0.509 0.531

Panama 0.5 0.584 0.475 0.563

Paraguay 0 0.451 0.57

Peru 0.61 0.485 0.43 0.464

Uruguay 0.37 0.428 0.436 0.406

Venezuela 0.613 0.462 0.48 0.447 0.459

LAC 4 0.505 0.532 0.531 0.491 0.507

LAC 6 0.548 0.548 0.532 0.542

LAC 15 0.539 0.519 0.532

Spain 0.457 0.363 0.347

Note: LAC 4 = population-weighted average of Brazil, Chile, Mexico and Venezuela. LAC 6 = population-weighted average of LAC 4 + Argentina and Uruguay. LAC 15 = population-weighted average of LAC 6 + Colombia, Cuba, Ecuador, Peru, Costa Rica, El Salvador, Guatemala, Honduras, and Panama.

Source: Constructed by the author using Perry (2006); Altimir (1987); Lodoño and Szekely (2000).

Figure 2- Gini Coefficient for Argentina, from the distribution of per capita household income, 1992-2005.




Country Period Change in Country Period Change in Gini points Gini points

Argentina 1992-1998 0.05 El Salvador 1991-2003 -0.02

1998-2002 0.03 Honduras 1997-2003 0.01 2002-2004 -0.02 Jamaica 1990-1999 -0.02 1992-2004 0.06 1990-2002 0.02Bolivia 1993-1997 0 Mexico 1992-1996 -0.02 1997-2002 0.03 1996-2002 -0.03 1993-2002 0.02 1992-2002 -0.04Brazil 1990-1995 -0.01 Nicaragua 1993-1998 -0.02 1995-2003 -0.02 1998-2001 0 1990-2003 -0.03 1993-2001 -0.02Chile 1990-1996 0 Panama 1995-2002 0.01 1996-2003 0 Paraguay 1997-2002 0.01 1990-2003 -0.01 Peru 1997-2002 0.01Colombia 1992-2000 0.07 Uruguay 1989-1998 0.02 2000-2004 0 1998-2003 0.01Costa Rica 1992-1997 0 Venezuela 1989-1995 0.04 1997-2003 0.04 1995-2003 0 1992-2003 0.04 1989-2000 0.02Dominican Republic 2000-2004 -0.01 1989-2003 0.04Ecuador 1994-1998 0.02

Source: Constructed by the author using Gasparini, Gutierréz and Tornarolli (2007).

Table 3- Inequality in Latin America between 1950 and 2000. Measured by Gini coefficients.


Table 4- Changes in inequality measured by percentage points of Gini Coefficient using household surveys in each country.


28 •

Notes: Means for groups in homogeneous subsets are displayed. Based on Type III Sum of Squares. The error term is Mean Square(Error) = .001.a Uses Harmonic Mean Sample Size = 32.727.b The group sizes are unequal. The harmonic mean of the group sizes is used. Type I error levels are not guaranteed.c Alpha = .05.Regions: 1) Capital City, 2) Pampeana, 3) Cuyo, 4) Northwest, 5)Northeast and 6) Patagonia.

Figure 3- Provincial Gini coefficients for Argentina. Averages for 1991-2002

Subset

Región N 1 2 3

2 60 .44047753

6 48 .44218363

3 36 .45430459 .45430459

Tukey HSD(a,b,c) 5 72 .46939949 .46939949

1 12 .48243993

4 48 .48440541

Sig. .418 .317 .323

Figure 4- Regional Inequality in Argentina, as shown by Gini coefficients. Averages for 1991-2002.





Table 6- Changes in Gini coefficient, third quantile (Q3), top 10 percent and bottom 20 percent shares in income of the population by region, between 1991 and 2002

(percentage).

Change Change Change Top Change BottomRegion Gini Q3 10% 20% 91-02 91-02 91-02 91-02

Buenos Aires 17.06 -11.26 12.52 -44.32

Pampeana 19.28 -10.42 -2.59 -29.71

Cuyo 14.21 -6.23 12.45 -28.39

Northeast 15.23 -15.05 14.59 -33.54

Northwest 11.94 -7.49 12.32 -19.21

Patagonia 9.32 -12.07 9.77 -18.61

Argentina 16.22 -7.82 15.30 -35.56


Table 5- Bonferroni and the Tukey’s tests to determine means differ in Gini coefficient among regions in Argentina, 1991-2002.


30 •

Year Moran’s I E(I) sd(I) z p-value*

1991 0.015 -0.045 0.164 0.372 0.71

1992 0.325 -0.045 0.164 2.25 0.024

1993 0.25 -0.045 0.169 1.75 0.08

1994 0.246 -0.045 0.167 1.746 0.081

1995 0.057 -0.045 0.168 0.608 0.543

1996 0.105 -0.045 0.168 0.894 0.371

1997 0.333 -0.045 0.169 2.244 0.025

1998 0.327 -0.045 0.168 2.221 0.026

1999 0.152 -0.045 0.166 1.188 0.235

2000 0.082 -0.045 0.16 0.795 0.427

2001 0.59 -0.045 0.167 3.801 0.000

2002 -0.114 -0.045 0.165 -0.417 0.676

*Two-tail test

Source: Author’s calculation using the EPH.

Figure 5- Moran’s I statistic for the provincial Gini coefficients of Argentina, 1991-2002.


Table 7- Estimates of the Moran’s I statistic for the provincial Gini coefficients of Argentina, 1991-2002.





Figure 6- Local Moran’s I statistic for the Gini coefficients provincial in 1991.


Figure 7- Local Moran’s I statistic for the Gini coefficients in 2002.


32 •

Unemployment and private returns to higher education in Argentina (1974-2002)

Cecilia Adrogué 1*

Abstract

The paper analyzes the private returns to higher education in Argentina during the period 1974–2002. The main conclusion is that returns to education are positive and increase once corrected by the level of unemployment. As a consequence, when analyzing whether to invest in educa-tion, one should not only consider as benefits the differential in earnings, but also the higher probability of having a job that comes with attaining more education. This is particularly relevant in a country like Argentina which had unemployment rates of 5% during the eighties and started to have unemployment rates in the double digits by the end of the 20th century and the begin-ning of the 21st century.

Resumen

El trabajo analiza la evolución de los retornos privados a la educación superior en Argentina du-rante el período 1974–2002 y cómo éstos se vieron afectados por el desempleo. La conclusión es que los retornos a la educación son mayores si se los corrige teniendo en cuenta el desempleo para cada nivel educativo, ya que a mayor nivel, menor tasa de desempleo. Al evaluar invertir en educación no se debería considerar simplemente el diferencial de ingresos sino también la mayor probabilidad de tener un trabajo. Esto es relevante en un país como Argentina que pasó de tener tasas de desempleo cercanas a 5% en la década del ochenta a tener tasas de dos dígitos a fines del siglo XX y comienzos del XXI.

Keywords: Returns to higher education, unemployment

JEL Classification: [I21] [J24] [J31] [J60]

1.- Licenciada en Economía (UCA), Doctorado San Andrés, CONICET, [email protected], [email protected]* I am very grateful to Jorge Paz, Juan J. Llach, Ricardo and Jill Adrogué, Marcos Dal Bianco as well as seminar participants at AAEP, CEMA University and IAE, Austral University. All errors are mine.




1. Introduction

Education can be considered an investment in human capital. That is the way it has been evalu-ated by a vast branch of the economic literature which started with Gary Becker´s work (1964). As with any other investment, education should only be carried out if the benefits are larger than the costs. However, as the investment and the benefits do not take place at the same mo-ment in time, both should be expressed in a homogenous measure. This is done by discounting future flows by an inter-temporal discount factor or interest rate. If the inter-temporal interest rate were zero, which would mean that one peso today is the same as one peso tomorrow, then it would be enough to add up the costs and compare them with the sum of benefits. But if the inter-temporal interest rate were positive, which would mean that one peso today is preferable to one peso tomorrow, the costs and benefits flows should be discounted in order to make them homogeneous and comparable.

A problem that arises when evaluating investment in education is that there is not an obvious interest rate to homogenize the flows, which makes it very difficult to calculate the present value of education, as well as almost impossible to find a price against which it can be compared. This is because there are no comparable investments regarding the risk and other unique characteristics of investing in education.

The relevant literature has typically used the internal rate of return on education, that is, the in-terest rate that makes costs equal benefits. One major advantage of this measure is that, unlike the net present value, it is not expressed in monetary terms, i.e. it does not lose relevance as time goes on, and it is also an intuitive measure of profitability. This internal rate of return could then be compared with the rate of return of other investments in order to decide whether to carry it out. Given these considerations, we believe it to be a worthwhile effort to quantify the benefits provided by education without losing sight of the fact that many of them are very difficult to measure.

2. Literature View

Starting with Gary Becker, Jacob Mincer and Theodore Schultz’s contributions, the demand for education has been studied as an investment in human capital, and its returns estimated (Har-mon et al., 2003). The two most common methods used to estimate returns to education have been the Mincer´s equation and the calculation of the internal rate of return. Both methods are equivalent under the following two assumptions: a) the only cost of studying is the opportunity cost of not being in the labor market and, b) the wage differentials among workers with different schooling are stationary (Margot, 2001).

This theory of human capital assumes that the amount of education, s, is chosen so as to maxi-mize the expected future value of the stream of future incomes w, up to retirement date T, net of the costs of education cs. The optimal s is the one for which the marginal income of an additional year of schooling equals the marginal cost of that year.

At the optimum: (1)


34 •

Where r is the rate of return. If T is large, then the left hand side of the equilibrium relationship can be approximated so that the equilibrium condition becomes:

(2)

If cs is sufficiently small, we can rearrange the expression to give:

(3)

Where ≈ means “approximately equal to...”

One could then estimate the returns to s by analyzing how the log of earnings varies with changes in s. Mincer (1974), who did one of the first empirical studies analyzing the returns to education, used an equation that relates income (wi) with years of education (si), experience (xi), squared experience (xi

2 ) and other observed variables that affect income, different from experience and education (Xi). The term referred to as the “squared experience” was introduced to capture the concavity of the earnings profile, as can be observed in Figure 1.

Figure 1- Income per Hour by Age for Men and Women, for All the Educational Levels. In pesos of 2002.

Source: Author’s calculation based on the EPHs.

(4)

The term ui is a random error term or disturbance and represents other forces which may not be explicitly measured and that also affect the individual’s earnings. When studying the returns to




education there could be the risk that the disturbance term may be related to some of the explan-atory variables and with the explained variable. In the case at hand, both the years of education (explanatory variable) and the earnings (explained variable) depend on the abilities of the person which, since it is not an explanatory variable, is contained in the disturbance term. In this situation problems of endogeneity arise and the estimations by Ordinary Least Squares regressions are not reliable. That is, a phenomenon not attributable to education is captured by the equation.

The literature refers to the reality that higher earnings could actually be due to the fact that those who pursue higher education are generally cleverer as “screening hypothesis” or “sheepskin hypothesis,”2 which may be the reason behind the positive correlation between earnings and education. In other words, there could be a problem of endogeneity and schooling may not be exogenous (Glewwe 2002). In line with the screening hypothesis, Hungerford and Solon (1987) demonstrate the existence of nonlinearity, a wage premium over the average return to schooling for fulfilling a particular year of education, for example the last year of college or high school. This shows that schooling can be used as a way to send a signal to the job market that one is smart. Following on Hungerford and Solon (1987), Spence (1973) analyzes the allocative efficiency of the job market and stresses the role of schooling as a signal.

Layard y Psacharopoulos (1974) found, however, that the rates of returns of uncompleted courses were as high as those of completed courses, that standardized educational differentials rise with age even though employers have better information about older employees´ abilities, and con-cluded that if screening is the main function of education, it could be done more cheaply by simpler testing procedures.

A number of studies have found that higher earnings are due to the fact that the individuals that continue studying acquire more cognitive skills. One of them is Boissiere et al. (1985), who examined urban wage earners in Kenya and Tanzania and found that education raises wages by providing workers with cognitive skills. Their data does not support the alternative hypothesis that education primarily reflects innate ability or sheepskin effects.

Although these estimates of the returns to education have been the most commonly used, they leave out spillover effects. Education has multiple effects on the individual’s life and on society as a whole, over and above the monetary ones (Glewwe, 2002), effects that are excluded in the type of analysis discussed above. Recognizing these shortcomings, several studies have found evidence that education also contributes to economic growth (Krueger y Lindahl, 2001) and raises productivity (Sianesi y Reenen, 2003). Even more, schooling has been shown to be a key variable to determine the wage differentials among the population. It could either narrow the gap and improve the income distribution, in the case of equal educational opportunities, or widen it and make the situa-tion worse, if those who have the possibility to study are just a few. This does not seem to happen only within countries, but also across countries, although evidence of the latter is still sparse.

3. Returns to Education

This paper intends to study the return to schooling through the calculation of the private benefits that derive from it, those that are appropriated directly by the student. In particular, we will ana-

2.- The screening effects are increases in income solely due to the possession of a diploma or other certificate, different from the ones derived from the skills acquired during the schooling process that the certificate or diploma represents.


36 •

lyze how those benefits are modified when one uses expected earnings instead of simple earn-ings in the forecasting equations. The expected earnings are calculated by multiplying the income data times the probability of having a job. The latter is a conditional probability that takes into ac-count the age, gender and the schooling level when evaluating the incidence of unemployment. As can be observed in Figures 2 and 3, the unemployment is higher for the less educated.

Figure 2- Unemployment Rate for Men, for Each Educational Level.





Figure 3- Unemployment Rate for Women, for Each Educational Level.


The methodology being proposed is not exempt from endogeneity problems. As the effect of the innate abilities cannot be isolated from the earnings, part of the returns attributed to education, may actually be due to ability. However, certain elements mitigate this problem. On the one side, more capable individuals tend to have higher income and therefore tend to produce an overes-timation of the returns, but, at the same time, they have higher opportunity costs of studying, which partially compensates for the income effect (Harmon et al., 2003). Moreover, in Argentina the sheepskin effect is partially mitigated by the often called “brain flight” (“fuga de cerebros”), that is, the tendency of smarter Argentines to migrate because they have better opportunities abroad.

In order to calculate the return to the investment in education it is essential to know its costs and benefits. Among the former affecting net returns there are direct costs, such as public expendi-ture in education, private donations, tuition costs, books, materials and transport costs paid by the student, as well as indirect costs, essentially the opportunity cost of not being in the labor force. Though when considering this cost, one should take into account the probability of get-ting a job instead of studying. As can be seen in Figure 4, unemployment is particularly high for people under 22 years old.


38 •

Figure 4- Unemployment Rate by Age for Men and Women, for all the Educational Levels.


Costs could also be classified as social or private, depending on whether the one who pays for them is the person who is being educated or not. That is to say, even if public education is free from an individual standpoint and considering direct costs alone, it does not mean that it is costless, as the State is the one that provides for the costs; therefore, these direct costs should be taken into ac-count in order to calculate the social returns, though not when estimating the private ones.

Similarly, benefits can be classified as private and social. Private benefits are those that accrue to the educated person, while social benefits, also called externalities, refer to the gain to society which results an individual’s education. Private benefits are very important for the family’s deci-sions, while the latter are the ones that should guide the governmental decisions, as they capture the effect of having a more educated population.

Recently in the literature, a new type of costs and benefits have been incorporated, those called fiscal costs and benefits (OECD, 2005). The former include public direct and indirect expenditures on education, as well as lost income tax revenues on students’ foregone earnings. The benefits include increased revenues from income taxes on higher wages.3 These are particularly useful so

3.- In practice, the achievement of higher levels of education will give rise to a complex set of fiscal effects on the benefit side, beyond the effects of wage-based revenue growth. For instance, better educated individuals generally experience superior health status, lowering public outlays on the provision of health care. And, for some individuals, achieving higher levels of educational attainment may lower the likelihood of committing certain types of crime, which, in turn, would reduce public expenditure. However, tax and expenditure data on such indirect effects of education are generally unavailable.




as to have a precise notion of how much the State is recovering from its investment in education since they allow the calculation of the fiscal internal rate of return.

There are basically two methods to estimate the returns to education. The first one is the static method that uses data of a certain moment in time, which means cross section information, in order to infer the earning profile of an individual from the earnings of other people with similar characteristics. If data are only available for a unique moment in time, the static method is the only alternative to get the earnings profile of the people in the sample. The static method was the one used by Mincer. The second method is the dynamic method that uses time series in order to have the earnings profile of a certain individual derived from the observed earnings for that person analyzed at different points in time. The main advantage of this way of calculating the returns is that there is no need to infer the earnings, as in the previous method. But a major disadvantage is that it is more prone to suffer endogeneity, that is, that some characteristics of the individual which are difficult to isolate econometrically may affect the estimations. These two methods could also be seen as estimations ex ante and ex post. If one needs to decide whether to invest or not, there is no alternative than to take the expected return, but, if one is interested in the returns one got with a certain investment, one should do the calculation ex post, knowing exactly how much it cost and how much was recovered.

Figure 5- Earnings profile, comparison of the dynamic and static method

As could be seen in Figure 5, the static and dynamic estimations can differ. In the case of Argen-tina using the dynamic method is not an option, making the inference unavoidable. This is due to the limitations of the available surveys. Specifically, the period for which there is information does not allow for the construction of a whole income profile for an individual and it is not necessarily the same person that is interviewed in the different surveys since half of the sample is replaced each time. For this reason we have decided to use the static method, although we recognize its limitations as ignoring changes in the earnings profile due to economic growth, technological


40 •

change, commercial openness, or other. In Argentina the earnings profile is descending, which would suggest that the static method might overestimate the returns. The 2001 and 2002 crisis depressed earnings quite substantially but the recovery in incomes that followed should have dampened the overestimation.

Figure 6- Income per Hour for Men, for each educational level.In pesos of 2002





Figure 7- Income per Hour for Women, for each educational level. In pesos of 2002


4. Data and Methodology

The database used is the Permanent Household Survey (EPH, Encuesta Permanente de Hogares) which is undertaken by the National Institute of Statistics (INDEC). It contains educational, work-ing, and socioeconomic information of the people living in the city of Buenos Aires and Greater Buenos Aires.4 Information is available for the months of October 1974, 1980, 1986 and 1992-2002. For the intermediate years no information for the people less than 25 years of age was available, and after 2002 the database has changed.5

The internal rate of return was estimated from information corresponding to individuals between 18 and 64 years of age with different schooling levels (secondary complete, university dropout and university graduate). Even though results were obtained for men and women, we consid-ered the information for men to be more reliable, since many women are not in the labor force

4.- The only information available for the period was the one corresponding to Buenos Aires and Greater Buenos Aires. The rest of the urban areas of the country were progressively introduced to the survey since 1992 but, in order to make com-parisons, it was necessary to use a homogeneous data base. Anyway, the area of Buenos Aires and Greater Buenos Aires represents more than 50% of the urban country population, therefore, it has a great weight in the whole average. Notwith-standing that, recent studies including the rest of the country regions indicate that the internal rates of returns to education of the area of Buenos Aires may overestimate the corresponding ones to Argentina, though they present the same trend. 5.- The EPH used to be carried out twice a year, in May and October, but during 2003 a major methodological change was implemented by INDEC, including changes in the questionnaires and in the frequency of the survey visits. Because of this change, the research was done up to the year 2002.


42 •

probably due to non-economic factors (such as maternity), exogenous to what we are trying to measure. As a result, women’s returns would be biased downwards relative to those of men as their earnings become spotty. Harmon et al. (2003) found evidence supporting this assertion; countries with the highest rates of female participation have the lowest differences in schooling returns while countries with the lowest participation rates have amongst the largest.

Another area where information was considered to be unsatisfactory was that of university drop-outs. Even if the estimates for the incomplete university level were calculated, the results were less robust than those corresponding to the complete level.6 We therefore centered the analysis solely on the latter.

The evolution of the returns to education is analyzed during a period of deep economic and social transformations in Argentina (1974-2002). During this period Argentina experienced a number of crises and institutional changes. Non-democratic governments (1976-1983) were followed by democratic ones. During 1989-1990 the country suffered from hyperinflation which was fol-lowed by a period of economic stability and increasing commercial openness that ended with the 2001-02 depression. The unemployment rate varied widely during these times. Some periods were characterized by low unemployment rates and others by very high ones, allowing the period of study to be divided in two sub-periods. The first one, characterized by low unemployment rates, lasted from 1974 until the beginning of the 1990s. The second one, when unemployment rates were above 10%, began in 1993. With these macroeconomic factors in mind we now pro-ceed to analyze the returns to education empirically.

Figure 8- Unemployment Rate in Argentina. 1974-2002.

Source: INDEC.

6. One should bear in mind that this group includes either those who finished only one subject of the career as those that did not finish because of one exam.




5. Rates of Return to Education, 1974-2002

a. Costs Estimation

No reliable estimates of the direct costs of education exist for the period under analysis. Expen-diture surveys are recent and there have been large variations in relative prices during the period being analyzed. For these reasons we have left aside direct expenditures on education and have focused on the indirect costs alone, under the assumption that those who study are not part of the labor force, and therefore the opportunity cost they have is given by the income that a person their age and previous educational level that is in the labor force receives. As such, the indirect cost for a university student is the forgone earnings of a worker that has completed secondary school.

Working age was assumed to start at age 18, normalizing to zero the opportunity cost prior to that age. Also, achieving a university degree was assumed to take six years; for estimation pur-poses, university dropouts were arbitrarily assigned four years of university schooling.

Although the assumption that those who study do not work may seem too strong, it is not so once one takes into account that those who work and at the same time study, generally take lon-ger to graduate and earn lower salaries than after graduation. The income differential and extra time in school make our estimate a good approximation for the actual opportunity cost.

b. Benefits Estimation

The principal benefit that derives from having a higher level of schooling is the earnings differ-ential. The net benefits are thus considered to be the difference between the earnings of two consecutive schooling levels.


44 •

Table 1- Earnings Differentials per Hour for Men and Women, for Each Educational level. In pesos of 2002.

Secondary Secondary University University incomplete complete incomplete complete vs. primary vs. primary vs. secondary vs. secondary complete complete complete complete

Men Women Men Women Men Women Men Women

1974 1.88 2.04 3.49 4.64 1.2 -3.44 8.64 3.11

1980 1.72 0.78 3.74 2.88 4.16 -0.03 5.69 5.52

1986 1.43 1.46 4.18 3.6 1.64 1.1 13.51 2.33

1992 1.02 0.64 3.01 3.44 0.53 2.02 5.64 2.31

1993 0.79 0.73 2.28 0.95 3.18 2.22 7.11 3.98

1994 1.68 0.13 2.12 1.19 1.39 2.85 5.91 2.28

1995 0.96 1.28 2.89 1.43 1.81 1.97 8.31 4.8

1996 0.41 0.56 1.91 1.31 4.96 1.1 6.69 6.82

1997 0.94 0.04 1.81 1.34 2.93 1.26 8.22 2.79

1998 0.77 0.78 2.15 1.27 3.08 1.2 9.65 5.86

1999 0.75 0.09 1.32 2.1 3.19 0.96 9.37 4.34

2000 1.03 0.48 2.16 2.27 1.88 1.09 6.44 4.18

2001 0.84 0.88 1.64 2.32 2.44 0.41 8.96 3.62

2002 0.32 1.24 1.76 1.19 1.51 0.09 6.62 3.45

General

Average 1.04 0.79 2.46 2.14 2.42 0.92 7.91 3.96

Average

1974-1986 1.67 1.43 3.8 3.71 2.33 -0.79 9.28 3.65

Average

1992-1999 0.91 0.53 2.19 1.63 2.63 1.7 7.61 4.15

Average

2000-2002 0.73 0.86 1.85 1.93 1.94 0.53 7.34 3.75


This paper takes this simple calculation one step further by also considering expected earnings adjusted for the incidence of unemployment. Two estimations will be performed. One considers actual earnings of employed workers. The other one adjusts these by the probability of being employed; that is multiplying actual earnings by the conditional employment rate given age, gender and schooling level.




Table 2- Probability of Employment for Men and Women, for Each Educational Level. In Percentage

Year

Primary Secondary Secondary University University complete incomplete complete incomplete complete

Men Women Men Women Men Women Men Women Men Women

1974 99.0 96.9 98.6 94.8 97.8 96.3 99.3 92.8 100.0 98.0

1980 98.2 97.9 97.1 97.6 98.3 96.2 99.2 98.1 98.7 99.1

1986 96.0 96.2 96.3 94.0 95.2 97.0 97.6 97.7 98.0 98.9

1991 94.9 95.2 96.8 93.0 93.2 94.4 95.5 96.8 97.9 93.7

1992 93.3 94.0 94.1 89.9 95.2 92.3 93.3 96.0 98.3 96.9

1993 90.6 84.9 93.4 87.5 94.1 90.5 95.6 87.2 94.3 95.9

1994 88.5 85.3 90.3 77.9 89.7 85.7 94.2 84.5 93.8 95.0

1995 83.3 80.2 87.0 75.5 85.6 78.7 92.3 74.8 95.0 91.1

1996 81.7 75.5 84.5 77.8 87.2 81.1 86.5 80.4 91.6 90.2

1997 85.8 84.1 89.3 81.4 90.3 82.2 88.5 82.1 94.9 90.8

1998 86.9 78.9 89.3 82.3 91.5 88.7 93.8 85.3 97.2 90.3

1999 86.4 79.3 86.7 79.9 87.9 82.4 88.0 82.2 89.9 91.2

2000 84.3 76.5 87.1 80.5 89.2 83.9 87.8 85.0 96.6 94.2

2001 74.9 78.7 80.4 77.3 78.9 78.1 82.3 81.9 93.9 92.3

2002 75.2 83.1 76.4 76.5 83.2 82.4 83.8 82.0 95.8 83.5

General

Average 87.0 85.1 89.0 83.8 90.0 86.8 92.0 86.4 96.0 93.4

Average

1974-1986 96.3 96.4 96.7 94.8 95.6 95.9 97.4 97.5 98.2 97.2

Average

1992-1999 87.1 82.8 89.3 81.5 90.2 85.2 91.5 84.1 94.4 92.7

Average

2000-2002 78.1 79.4 81.3 78.1 83.8 81.4 84.6 83.0 95.4 90.0


c. Private rates of return

As stated before, private rates of return are calculated leaving aside the externalities produced by education. Even though a number of studies have analyzed the social rates of return to edu-cation by using the earnings before taxes as opposed to the private returns that consider the disposable income (after taxes), the Permanent Household Survey database used in this study has information on disposable income (Llach, 1996) alone. Moreover, the costs analyzed here exclude public expenditure in education and private donations, only viewing the opportunity cost to the student.

Finally, the results we present should be viewed as the marginal benefits to education, only valid at an individual level. This means that they are not necessarily true if all the people with a certain schooling level decided to continue studying to the following level. It may happen that as the sup-ply of trained labor force increases, and the demand for them stays constant, the level of earnings may actually fall, and possibly, the unemployment rate for people with that level of schooling may


46 •

rise. As Ashenfelter and Ham (1979) state “Our results suggest that the excess of the marginal effects of schooling and experience on earnings over their effects on wage rates is due almost entirely to the effect of schooling and work experience in reducing measured unemployment...Of course, this does not imply that increased educational attainment will necessarily reduce ag-gregate unemployment, because the effect we observe may come merely from a redistribution of unemployment among workers.”

Though Ashenfelter and Ham’s (1979) assertion should not be taken lightly, it could also hap-pen that when a large share of the population studies, innovation increases and the ability of the enterprises to incorporate new technology also rises, pushing up economic growth. This may produce an even greater increase in the demand for the more educated than the increase in supply pushing earnings higher instead of lower. The evidence that human capital increases productivity is compelling. Studies that analyze education as a signal, such as Spence (1973), do not deny positive effects on productivity. Sianesi and Van Reenen (2003) found that a one year increase in average education is found to raise the level of output per capita by between three and six percent, and increase one percentage point the growth rate.

d. Estimating the internal rate of return to education

The annual internal rate of return was calculated from the following equation applied to each educational level:

(5)

r: IRR, what we want to calculate.

t: Age of the individual.

T: 64 years old, age at which the person retires.

C: cost of education, in this case it is the opportunity cost.

w: Earnings obtained by an individual with a certain educational level (j).

e: age at which a certain educational level is started. In the case of higher education, it is equal to eighteen.

E: age at which a certain educational level is completed (in this paper we have assumed 18 years of age for a high school graduate, 22 for university dropout and 24 for university graduate).

J: educational level attained.

As the information about earnings for individuals older than 18 years old is more reliable than for those aged between 12 and 17, only the opportunity cost for university students, either those who graduated or those who dropped out was calculated, but not for secondary graduates.

In table 3 the values of the internal rate of return for both men and women are presented, and it




can be seen that the latter are always lower than the former; except for the year 2002 in which the IRR of women was 16% while the corresponding to men was 15%. It should be recalled, though, that the data for men is more reliable because it does not have the noise of entry and exit from the labor force that characterizes the information for women. As shown in the table as well, the return for the complete university level is always higher than the corresponding one for the incomplete level, for all the years studied.

Table 3- Internal Rates of Return to Education*. In Percentage

Men Women

Year University University University incomplete complete complete

1974 5.0 11.1 3.7 1980 10.9 14.1 9.5 1986 6.8 11.9 8.4 1992 7.5 13.7 7.4 1993 10.2 14.9 9.2 1994 14.1 15.9 8.7 1995 7.3 14.7 12.3 1996 8.0 16.2 10.6 1997 8.9 15.7 12.7 1998 10.0 15.7 10.5 1999 12.8 14.4 12.5 2000 8.5 14.3 11.5 2001 16.5 18.2 11.7 2002 8.7 15.0 16.3

General Average 9.7 14.7 10.4

Average 1974-1986 7.6 12.4 7.2 Average 1992-1999 9.8 15.2 10.5 Average 2000-2002 11.3 15.8 13.2

*Internal Rates of Returns for Women for University Incomplete have been omitted because of the fact that the lack of information weakened the results. Source: Author’s calculation based on the EPHs.

Next we studied the internal rate of return corrected for the probability of having a job, that is, both the costs and benefits were corrected for the probability of being employed. This corrected rate of return was calculated by multiplying the earnings differentials by one minus the probabil-ity of being unemployed being of a certain age and having achieved a given educational level. The following equation was solved for each educational level analyzed.

(6)

: Unemployment rate by age (t) and educational level (j y j-1)rc: IRR corrected by the probability of having a job, which affects both the costs and the ben-efits.


48 •

Table 4- Internal Rate of Return of University Studies for Men and Women. Comparison between the Values Revised and Not Revised. In Percentage

Men Women

Resume

University University University incomplete complete complete

Year IRR IRR revised IRR IRR revised IRR IRR revised

1974 5.0* 6.3* 11.1 12.0 3.7 4.6 1980 10.9* 11.9* 14.1 13.1 9.5 9.7 1986 6.8* 9.1* 11.9* 12.8* 8.4* 9.2* 1992 7.5* 9.9* 13.7 15.2 7.4* 7.8* 1993 10.2 11.4 14.9 15.7 9.2 11.5 1994 14.1 20.5 15.9 20.2 8.7* 13.6* 1995 7.3 9.6 14.7 16.3 12.3 19.2 1996 8.0 9.6 16.2 20.1 10.6 13.9 1997 8.9 9.1 15.7 17.0 12.7 18.6 1998 10.0 11.1 15.7 17.7 10.5 13.0 1999 12.8 15.0 14.4 16.7 12.5* 15.6* 2000 8.5 10.4 14.3 17.4 11.5 18.0 2001 16.5 22.3 18.2 26.7 11.7 18.5 2002 8.7 10.6 15.0 18.8 16.3 19.8

General Average

9.7 11.9 14.7 17.1 10.4 13.8

Average 1974-1986

7.6 9.1 12.4 12.6 7.2 7.8

Average 1992-1999

9.8 12.0 15.2 17.4 10.5 14.1

Average 2000-2002

11.3 14.4 15.8 21.0 13.2 18.8

* With a confidence of 95% we cannot say that the variances of the earning profiles are different. Source: Author’s calculation based on the EPH’s.

As could be observed in table 4, the differences between the IRR and the IRR revised were not significant for the years in which the rate of unemployment was low (1974, 1980, 1986 and 1992), and except for the years 1994 and 1999 for women, were significantly different for the period 1993-2002, which was characterized by high rates of unemployment.

Although the IRRs are quite volatile, an ascending trend is apparent, at the same time that the earnings for all the educational levels fall.




Figure 9- Internal Rate of Return to Education for Men


Figure 10- Internal Rate of Return to Education for Women



50 •

The results of the analysis highlight the importance of using expected earnings vis-à-vis actual earnings and the relevance of unemployment particularly during the 1990’s. Since the rate of unemployment has been inversely related to education attainment, using expected returns raises the internal rate of return in every year except 1980 for male university graduates by increasing the expected benefits of higher education vis-à-vis the expected costs.

This conclusion is driven because the opportunity cost falls, while the expected differentials in earnings could increase, decrease or remain unchanged. In the particular case of study, the inter-nal rate of return that zeroes equation (6) is higher than the one that zeroes equation (5) because the incidence of unemployment weights more heavily on the costs (C) than the benefits (w).

Table 5- Difference of the IRR Because of Unemployment. In Percentage

Men Women

Year – Resume University University University incomplete complete complete

1974 1,4 0,9 0,9 1980 0,9 -1,0 0,2 1986 2,2 0,9 0,9 1992 2,5 1,5 0,4 1993 1,2 0,9 2,4 1994 6,4 4,3 4,9 1995 2,2 1,5 6,9 1996 1,6 3,9 3,3 1997 0,3 1,2 5,9 1998 1,1 2,0 2,4 1999 2,2 2,3 3,1 2000 1,9 3,2 6,5 2001 5,8 8,5 6,8 2002 1,8 3,9 3,5

General Average 2,2 2,4 3,4

Average 1974-1986 1,5 0,2 0,6 Average 1992-1999 2,2 2,2 3,7 Average 2000-2002 3,2 5,2 5,6


The table shows how the differential between the internal rate of return and the one corrected for unemployment was always positive, except for the year 1980 the one corresponding to male university graduates, and has risen during the period under analysis. This result should not sur-prise the reader, since the unemployment rate of university graduates has been lower than that of high school graduates (Figures 2 and 3), and this differential has increased during the period we studied. This effect has been compounded by the greater incidence of unemployment among the youth, both men and women (Figure 4).

As can be seen in table 5, the difference in the IRR because of unemployment represented less than 1% for university graduates for the period 1974-1986, rose to 2,2% and 3,7% for men and




women respectively between 1992 and 1999, and during the crisis (2000-2002), this differential increased even more, and was above 5%, both for graduate men and women.

Finally, it is worth mentioning that the calculated rates of return must not be understood as the return to an additional year of education, but as the annual return of reaching a certain educa-tional level.

6. Conclusions

After having calculated the returns to higher education in Argentina, it can be inferred that a uni-versity education is a profitable investment, not only for men but for women as well. The average rates of return are 15% and 10% respectively.7

Additionally, an element that must be taken into account is the different rate of unemployment by education levels. Unemployment is higher among the young which reduces the opportunity cost of studying, because the relevant measure is not the income that a person of the same age can earn with a high school education. This is true because the probability of finding a job is rather small, an important factor which needs to be taken into account. As shown, the difference between the traditional IRR and the corrected by unemployment gets bigger during the period analyzed. It could also be seen that it raises the return to education in a significant way, and the hypothesis that an important benefit of studying is the increase in the probability of having a job could not be rejected. The average IRR’s for women and men rise from 10% to 14% and from 15% to 17% respectively.

These findings raise the following question: Why if education is a profitable investment, which not only has positive effects on the person that studies, some of which are not quantifiable, such as the opportunity to be more cultured, but also has effects over the whole society, such as economic growth and more productivity, do so many individuals choose not to continue studying further than the secondary level? According to the Census that took place in the year 2001, only 17% of the people older than 15 years old planned to continue or continued their studies further than the secondary level and the gross university schooling rate was 25% that year.8

This high rate of return and this low university schooling rate are indicative of a market failure. As Harmon et al. (2003) point out, this could be one of the reasons why individuals do not decide optimally and underinvest.

The market failure in the Argentine education system consists of individuals that would be able to continue their studies, getting a great return, and do not do so. The most likely explanation is that this is due to a lack of liquidity which the traditional financial system does not wish to cover because of the absence of a guarantee and the high risk. At the same time, it could be corrobo-rated that in the case of continuing studies, the costs can be easily repaid due to the high income differentials among the people with different educational levels.

7.- Although with greater uncertainty in the case of the data on women. 8.- The gross university schooling rate is calculated as the sum of the population that attends university, independently of his age, over the total amount of people aged between 18 and 24 years.


52 •

7. References

Ashenfelter, Orley and John Ham (1979), “Education, Unemployment and Earnings” in Journal of Political Economy 87 (5) II pp. 99-116.

Barro, Robert J. (1989), “Economic Growth in a Cross Section of Countries” WP 201, Harvard University.

Becker, Gary (1964), Human Capital, NBER.Boissiere, Maurice, John Knight and Richard Sabot (1985), “Earnings, Schooling, Ability and Cog-

nitive Skills” American Economic Review Vol. 75 (5), pp. 1016-1030.Encuesta Permanente de Hogares, INDEC. Several Waves.FIEL (2002), Competitividad, “Capital humano y educación para el crecimiento”, Buenos Aires.Glewwe, Paul (2002), “Schools and Skills in Developing Countries: Education Policies and Socio-

economic Outcomes”, Journal of Economic Literature Vol. 15, June, pp. 436-482.Hansen, Gary (1985), “Indivisible labour and the business cycle” Journal of Monetary Economics,

16 pp. 309-325.Harmon, Colm, Hessel Oosterbeek and Ian Walker (2003), “The Returns to Education: Microeco-

nomics”. Journal of Economic Surveys, Vol. 17, pp. 115-156.Hungerford, T and G. Solon (1987), “Sheepskin Effects in the Return to Education”, Review of

Economics and Statistics, Vol. 69, pp. 175-177.Krueger, Alan B. and Mikael Lindahl (2001): “Education for Growth: Why and for Whom?”, Jour-

nal of Economic Literature Vol. 39 (4) pp. 1101-1136.Layard, Richard and George Psacharopoulos (1974), “The Screening Hypothesis and the Returns

to Education”, Journal of Political Economy, 82 (5) pp. 985-998. Lee Hansen, G (1963), “Total and Private Rates of Return to Investment in Schooling”, Journal of

Political Economy, April, Vol. 71 (2), pp. 128-140.Llach, Juan J. and Silvia Montoya (2000), Educación para todos, IERAL.Llach, Lucas (1996), “Beneficios de la educación en presencia de alto desempleo: El caso de Ar-

gentina”, Universidad Torcuato Di Tella, May.Margot, Diego (2001), “Rendimientos a la educación en Argentina: Un análisis de cohortes”. WP

No. 33, Facultad de Ciencias Económicas Universidad Nacional de La Plata, July.Meghir, Costas and Marten Palme (2004), “Educational Reform, Ability and Family Background”,

The Institute for Fiscal Studies, WP04/10, 21st September.Mincer, Jacob (1974), Shooling, Experience and Earnings, New York, NBER. Ministerio de Educación Ciencia y Tecnología, Secretaría de políticas universitarias, Anuario 1999-

2003 Estadísticas Universitarias.OECD (2005), Education at a Glance, Paris.Paz, Jorge A. (2004), “Education, Gender and Youth in the Labor Market in Argentina”, WP Nº.

272, CEMA, September.Pessino, Carola (1995), “Returns to Education in Greater Buenos Aires 1986-1993: From Hyper-

inflation to Stabilization”, WP Nº.104, CEMA, June. Sianesi, Barbara and John Michael Van Reenen (2003), “The Returns to Education: Macroeco-

nomics”. Journal of Economic Surveys, Vol 17, pp. 157-200.Spence, A. Michael (1973), “Job Market Signaling”, Quarterly Journal of Economics, Vol. 87 (3),

pp. 355-374. Solow, Robert (1980), “On theories of Unemployment”, American Economic Review, 70(1) pp.

1-11.Sosa Escudero, Walter (2005), “Aproximaciones económicas y econométricas para la problemáti-

ca educativa”, WP N°.17, Escuela de Educación de la Universidad de San Andrés, August.




Streb, Jorge Miguel (2002), “Job Market Signaling under Two-dimensional Asymmetric Informa-tion”, Mimeo. July.

Tobias, Justin L. and Li, Mingliang (2004), “Returns to Schooling and Bayesian Model Averaging: A Union of Two Literatures”. Journal of Economic Surveys, Vol. 18 (2) pp.153-180.


54 •

Estimación de una función de producción agregada: Argentina 1975-2006.Aplicaciones al crecimiento económico

Fernando Suárez1

Resumen

El presente trabajo estima distintas funciones agregadas de producción para la Argentina entre los años 1975 y 2006, con el objetivo de determinar la que resulta más apropiada para describir el proceso productivo argentino. Una vez obtenidas dichas funciones de producción, las mismas se aplican al estudio de la produc-tividad total de los factores, a la determinación del producto potencial y a establecer la contribu-ción al crecimiento de los factores productivos involucrados.Este estudio revela el paulatino deterioro de la productividad total de los factores en dicho perío-do, a pesar de haber atravesado períodos de crecimiento en la convertibilidad de los `90 y la recuperación posterior a la crisis de 2002. Se verifican varios episodios de recalentamiento de la economía, donde el producto observado supera al potencial, y pone en evidencia una matriz de crecimiento basado en acumulación de factores y no en productividad e innovación.

Abstract

This paper estimates different aggregated production functions for Argentine between the years 1975 and 2006, with the object to determine which of them are the most suitable for describing the productive process of the country.Once this goal is achieved, the functions are applied to study the total factor productivity, to determine the potential output, and to establish the growth contribution of the factors invol-ved.This study reveals the gradual deterioration of the total factor productivity in that period, despite of crossing periods of growing in the `90 convertibility and the following recovery to the 2002 crisis. Several episodes of heating of the economy are verified, where the actual output overcome the potential output, and highlights a growth matrix based in factor accumulation instead of productivity and innovation.

Keywords: Función de producción, Productividad de factores, Producto potencial

JEL Classification: [C10] [F43] [O47] [O54]

1.- Ingeniero, Profesor de estadística en Facultad de Ingeniería (UCA), MA en economía aplicada (UCA), [email protected]




1. Introducción

Las altas tasas de crecimiento del producto experimentadas por Argentina en los últimos años, unidas al hecho de una inusual prolongación de las mismas en el tiempo, han vuelto una vez más a plantear el debate respecto al origen de dicho crecimiento y su sustentabilidad en el largo plazo.

Asimismo la gran recuperación de la economía luego de la profunda crisis sufrida por el país en 2002, como consecuencia de la caída del régimen de convertibilidad, es otro hecho que invita al análisis respecto a la naturaleza del mismo.

Estos fenómenos nos remiten a preguntarnos por los factores que intervienen en este proceso, para determinar si el mismo puede continuar en los mismos términos, o bien si es de naturaleza “extensivo”, entendiendo por esto que el crecimiento del producto se debe al aumento de la cantidad de los factores empleados, o “intensivo”, es decir, si el mismo es originado por un mejor aprovechamiento de los factores empleados debido a la utilización de una tecnología más eficien-te. Otra cuestión no menor además de las precedentes es la de determinar si dicho crecimiento es debido a un cambio estructural o bien forma parte de un consabido proceso cíclico que en algún momento invertirá su sentido.

El presente trabajo pretende en primer lugar estimar una función de producción agregada de la economía Argentina que sirva como un instrumento de predicción, y mediante la utilización de la misma aplicarla a responder a los interrogantes planteados.

A diferencia de otros planteos que postulan a priori una determinada función de producción a estimar, la opción elegida para este trabajo consiste en probar distintas funciones de producción, con distintos supuestos para cada una de ellas y luego determinar cual de ellas resulta más apro-piada a los efectos del desarrollo del trabajo.

El período elegido para este estudio que se extiende de 1975 a 2006 está dividido en cuatro etapas signadas por diferentes contextos políticos y económicos: a partir de 1975 hasta 1981, etapa caracterizada por una moderada apertura que se cierra con la guerra de Malvinas; el perío-do siguiente signado por el cierre de la economía, que comienza en 1982 y se cierra con la crisis hiperinflacionaria de 1989-1990; los años que van desde 1991 hasta 2001, marcados por la con-vertibilidad, la apertura de la economía y la crisis del Tequila en 1995, y que se cierran con la crisis de 2001; y el tramo final que va de 2002 a 2006 caracterizado por la devaluación, recuperación y alto crecimiento.

La elección de dicho período permite comparar la heterogeneidad de políticas y situaciones coyunturales dentro de los límites en los cuales se puede contar con una tecnología dada, dispo-niendo así con una cantidad de años suficientes para poder lograr una estimación con un ajuste aceptable dadas las variables explicativas utilizadas.

Tras la obtención de la función de producción agregada el trabajo puede extenderse entonces al análisis de la productividad de los factores, a determinar el producto potencial y la brecha del producto, y a verificar la contribución al crecimiento de cada uno de los factores. Asimismo es posible predecir el producto potencial para algunos años siguientes al período de estudio, con el objeto de inferir la posible presión inflacionaria.


56 •

2. Experiencias anteriores

Han habido en el transcurso de estos años distintos intentos por estimar una función de produc-ción agregada para Argentina, en donde algunos de ellos coinciden parcialmente con el período pretendido por el presente estudio.

Existen numerosos trabajos que estiman una función de producción agregada de tipo Cobb-Do-uglas, en donde los coeficientes de participación del trabajo son tomados de Cuentas Nacionales para un año dado, con el objetivo de determinar el producto potencial y la brecha de producto. Los mismos serán revisados no aquí, sino en el acápite correspondiente a las estimaciones del producto potencial.

Otros intentos apuntan a estimar una función de producción para un sector dado de la economía (generalmente el agropecuario), un ejemplo precursor de esto lo encontramos en Elías (1983) para ese sector en cuestión.

El mismo Elías (1989) estima también una función de producción de tipo Cobb-Douglas para un grupo de siete países de América Latina. En el mismo sentido Hofman (1990) aplica el mismo concepto para estudiar el crecimiento económico y el desempeño de nueve países de América Latina.

Sin embargo, dada la naturaleza del presente trabajo, citaremos a continuación aquellas iniciati-vas que se centraron en una estimación econométrica exclusivamente para la Argentina, median-te la utilización de series de PBI, stock de capital y empleo.

En su ensayo, Aldabe (1965) estima una función de producción agregada de tipo Cobb-Douglas para los años que van de 1947 a 1961, derivando del mismo algunas conclusiones relativas a la verificación de las propiedades teóricas de la función de producción obtenida y a la naturaleza del desarrollo experimentado por el país durante esos años.

Con el objetivo de lograr una interpretación de largo plazo del proceso de crecimiento industrial argentino de posguerra, Katz (1969) estima una función de producción de tipo CES para el perío-do de 1946 a 1961, centrando su estudio a las fuentes del crecimiento y discutiendo el rol de la acumulación del capital como factor del cambio tecnológico.

A partir de la evidencia empírica, Dichiara (1980) efectúa distintas estimaciones de funciones de producción, a saber, una de tipo Cobb-Douglas y otra de tipo CES para la economía argentina en los años que van de 1950 hasta 1973. Intenta fundamentalmente estudiar las elasticidades de escala y de sustitución, como asimismo la naturaleza y el comportamiento del progreso téc-nico.

En una interesante aplicación donde se comparan las tasas de crecimiento de la productividad total de los factores de Argentina con las de algunas economías recientemente industrializadas del este de Asia, Lanteri (1999) estima una función de producción translogarítmica para los años que van de 1977 a 1998, con la que logra determinar cuál es la contribución de dicha producti-vidad total de factores al crecimiento económico.

Quizás el trabajo más conocido y típicamente citado es el de Meloni (1999), quien reproducien-do para la Argentina un intento de Roldós (1997) para el caso chileno, estima varias funciones de producción de tipo Cobb-Douglas para los años que van de 1980 hasta 1997, las cuales son




aplicadas al cálculo del producto potencial, el estudio de la productividad total de los factores y la contribución al crecimiento.

3. Metodología aplicada

A continuación describiremos la metodología utilizada para la obtención de las series necesarias para la estimación de la función de producción agregada de la economía Argentina, a saber, el producto y los insumos empleo y stock de capital.

Las fuentes de las restantes series necesarias para el cálculo de los distintos índices de calidad son mencionadas en las secciones correspondientes. La metodología aplicada para la estimación de las distintas formas propuestas para la función de producción corresponderá analizarla en parti-cular para cada una de ellas.

3.1 Serie de PBI real

La serie de PBI expresada en pesos constantes del año base 1993 fue obtenida de Maia y Nichol-son (2001). Dicho trabajo se encuentra actualizado hasta 2004, por lo cual los años faltantes (2005 y 2006) se obtuvieron de la Dirección Nacional de Cuentas Nacionales.

Gráfico 1

3.2 Serie de empleo

La serie de empleo expresada en miles de puestos de trabajo fue obtenida de Maia y Nicholson (2001). Debido a una actualización parcial de dicho trabajo en 2005 tenemos la ventaja de contar con una serie correspondiente al total del país, y no solo de población urbana, en base a estima-ciones propias de los autores.


58 •

Para los años faltantes (2005 y 2006) se estimó el empleo correspondiente al total del país mediante la tasa de empleo urbana y la población total del país , estimada esta última según proyecciones construidas con datos del Censo Nacional de 2001, suponiendo que dicha tasa se comporta de la misma manera a nivel nacional. Todos estos datos se obtuvieron del INDEC.

3.2.1 Índice de calidad del empleo

Para no introducir distorsiones que se trasladarían en la estimación de la función de producción a la productividad total de los factores, resulta de suma importancia ajustar la serie de empleo con un índice de calidad.

Dicho índice se puede obtener teniendo en cuenta varios procedimientos y variantes derivadas de la contabilidad del crecimiento pero que tienen origen en un principio fundamental: el producto debe equiparar a lo que se paga por los factores empleados para producirlo.

Es decir:

(3.1)

Si calculamos la tasa de cambio , donde representa la derivada de Y respecto al tiempo, esta tasa de cambio se puede descomponer en dos componentes correspondientes a los insu-mos empleo y stock de capital, y en dos componentes correspondientes a la calidad de dichos insumos

Para el factor empleo exclusivamente, el componente de calidad resulta:

(3.2)

en donde representa el salario promedio de la economía, y representa la par-

ticipación de cada categoría laboral en el empleo total de la economía.

Este índice de calidad del empleo tiene en cuenta la heterogeneidad del trabajo según las distin-tas habilidades y por extensión refleja las diferencias de productividad de cada categoría laboral. En esencia es un promedio ponderado del cambio en las participaciones de cada categoría de empleo, donde los factores de ponderación surgen de los salarios relativos de cada categoría laboral respecto al promedio de salarios de toda la economía.

Resta elegir una variable que nos permita categorizar el empleo de forma conveniente, y que cada categoría capture de forma conveniente las distintas productividades. La variable más con-veniente para el caso es la educación.

Siguiendo a Meloni (1999), se empleó la Encuesta Permanente de Hogares (EPH) brindada por el INDEC, analizando los ingresos de los ocupados del Gran Buenos Aires (GBA) como muestra representativa para todo el país, según cuatro categorías definidas del siguiente modo:




• Categoría1:personassininstrucciónoconestudiosprimariosincompletos.• Categoría2:personasconestudiosprimarioscompletosoconestudiossecundariosincom-

pletos (incluye Nacional, Comercial o Técnico).• Categoría 3: personas con estudios secundarios completoso superior incompleto (incluye

Universitario o no Universitario).• Categoría4:personasconeducaciónsuperiorcompleta.

El índice puede ser calculado a partir de la ecuación (3.2) sin inconvenientes. Sin embargo sólo se dispone de las EPH en formato digital de los años 1974, desde 1980 hasta 1982, y desde 1985 hasta 2002. De 1984 sólo se cuenta con la versión preliminar, y desde 2003 hasta 2006, la encuesta no es puntual sino continua.

Para resolver este problema se consultó material de archivo en el INDEC, con lo cual se pudieron obtener datos de las participaciones de cada categoría laboral para los años 1975, 1977, 1983 y 1984. Con estos datos se pudo calcular el índice para los años que van de 1975 a 1980, inter-polando los años faltantes y manteniendo sin variación los salarios relativos. De forma similar se calcularon los índices para los años 1983 y 1984.

Se calculó sin problemas el índice para los años 1981 y 1982 con la ecuación antes mencionada. Asimismo se replicaron para los años desde 1985 hasta 2002 los valores del índice obtenidos por Maia y Kweitel (2003).

Debido al cambio de metodología introducido a partir de 2003 con la EPH continua, se estimaron los años desde 2003 a 2006 mediante una proyección lineal del logaritmo natural del índice de los años anteriores.

Gráfico 2


60 •

De 1975 a 1985 el índice no representa un factor de ajuste importante. A partir de 1985 el índice comienza a crecer, probablemente a causa del establecimiento del ingreso universitario irrestric-to. Si bien mantiene siempre su tendencia positiva, el crecimiento de la productividad laboral es mayor entre los años 1995 y 1999.

Dado que hemos elegido a la variable educación para categorizar las habilidades laborales, la mejora en la instrucción recibida captura de forma aproximada las mejoras en productividad.

Del análisis de la EPH para el Gran Buenos Aires podemos cuantificar las participaciones de cada categoría laboral en el total del empleo.

En 1975 el 25,1 % de los trabajadores pertenecía a la primera categoría (sin instrucción o con estudios primarios incompletos) y el 5,58 % de los asalariados formaba parte de la última cate-goría (estudios superiores completos).

En 1990 el 8,69 % de los asalariados tenía educación correspondiente a la primera categoría, mientras que las personas con educación superior completa ascendían al 11,59 %.

Finalmente, en 2006, el 7,39 % de los asalariados pertenecía a la primera categoría y el 18,08 % poseía algún título correspondiente a educación superior.

3.3 Serie de stock de capital reproductivo

Para nuestro propósito de estimar una función de producción agregada para la economía Argen-tina, el stock de capital relevante no es el total, sino el stock de capital reproductivo, es decir, aquel que se utiliza para producir otros bienes.

Las series de stock de capital total y de stock de capital reproductivo expresadas en millones de pesos constantes del año base 1993 fueron obtenidas de Maia y Nicholson (2001), quienes las generan con el Método de los inventarios permanentes. Del mismo también se utilizaron las series de depreciación, calculadas con el método de depreciación geométrico.

Como ya lo mencionamos anteriormente, dicho trabajo se encuentra actualizado hasta 2004, por lo cual se calcularon los años faltantes (2005 y 2006) aplicando el mismo método, a partir de la siguiente ecuación:

(3.3)

La serie de inversión bruta interna fija se obtuvo de la Dirección Nacional de Cuentas Nacionales. Corresponde aclarar que uno de los componentes más importantes de la misma, como lo es la construcción, no discrimina entre construcción no residencial (que corresponde a capital repro-ductivo) de la construcción residencial (que no corresponde a capital reproductivo).

Para poder obtener la inversión correspondiente a construcción no residencial se consultaron en el INDEC los permisos de edificación otorgados y superficie cubierta autorizada por destino de obra, siendo el dato relevante en este caso no la cantidad de permisos sino la superficie cubierta.

Esta relación se cotejó con los valores históricos de los últimos años disponibles (2003 y 2004), que-dando establecido que el porcentaje de construcción no residencial respecto del total es del 37 %.




3.3.1 Índice de calidad del capital

De la misma forma como se procedió para el empleo, resulta necesario ajustar la serie de stock de capital reproductivo con un índice adecuado, para no introducir distorsiones en la producti-vidad total de los factores y en los coeficientes de las estimaciones de la función de producción agregada.

Para el factor stock de capital reproductivo exclusivamente, el componente de calidad resulta:

(3.4)

en donde representa el retorno real del capital promedio de la economía, y

representa la participación de cada categoría de capital en el stock de capital reproductivo total

de la economía.

rj representa el retorno real de capital de una categoría cualquiera, es decir el precio del servicio de capital de la misma si tuviéramos que salir a alquilar dicho bien al no disponer de él.

Lo ideal es disponer de los precios de mercado, pero al no poder contar con ellos, podemos estimarlos de acuerdo a la metodología de la contabilidad del crecimiento, con la siguiente expre-sión:

(3.5)

Donde Rt representa la tasa de interés de toda la economía, pj,t es el precio del stock de capital de la categoría j y corresponde a la depreciación. El único dato faltante para poder calcular el índice del capital es la tasa de interés de toda la economía, Rt.

Para que el precio del stock de capital de la categoría correspondiente refleje el correcto costo de uso del bien, debemos de excluir de la anterior expresión las ganancias de capital originadas por la variación interanual de los precios, con lo cual dicha expresión deviene en:

(3.6)

La OECD (2001) en su manual para la medición del capital sugiere dos métodos para una esti-mación de Rt, utilizando un promedio de tasas de interés reales de mercado, o calculándola indirectamente a través del excedente bruto de explotación. Ambas alternativas presentan incon-venientes de aplicación y cálculo para nuestro país.

Descartamos el promedio de tasas reales de mercado debido a los ya conocidos episodios inflacio-narios y de regulación de tasas de interés que Argentina ha sufrido, y que conducen a resultados erróneos y distorsivos.

Se optó en cambio por el cálculo indirecto a partir del excedente bruto de explotación, es decir, si obtenemos el excedente bruto de explotación, descontamos el impuesto a las ganancias y lo dividimos por el stock de capital, obtendremos una aproximación de la tasa de retorno real de la economía. Los valores utilizados deben ser calculados a precios corrientes.


62 •

Para los años que van de 1993 a 2006 los datos de la generación del ingreso total de la economía se obtuvieron de la Dirección Nacional de Cuentas Nacionales. Para los restantes años no dispo-nemos de datos consistentes.

Los últimos datos oficiales disponibles de la distribución funcional del ingreso figuran en un tra-bajo del Banco Central de la República Argentina llamado “Sistema de Cuentas del Producto e Ingreso de la Argentina” que abarca el período desde 1950 hasta 1973.

Para poder estimar la retribución al capital para los años faltantes se recurrió al trabajo de Lin-denboim, Graña y Kennedy (2005), quienes presentan distintos resultados para el mismo corres-pondientes a varias fuentes. Del mismo se extrajeron datos que permitieron obtener el ingreso capitalista, suponiendo por nuestra parte una participación del ingreso mixto del 20 %, de acuer-do a valores históricos.

Luego de deducir el impuesto a las ganancias, aproximado al 33 %, se obtuvo el excedente de explotación neto de impuesto a las ganancias, que al dividirlo por el stock de capital resulta la tasa de retorno real de la economía buscada, completando así la totalidad de los datos para el cálculo del índice de calidad del capital.

Cabe aclarar que dicha tasa de retorno real no es la tasa de interés real de la economía, dado que tiene incorporada la depreciación promedio de todos los bienes de capital. Para poder contar con el valor de la tasa de interés real de la economía, y así calcular luego el costo de uso para categoría de bienes de capital en particular, debemos despejarla haciendo uso de las ecuaciones de retorno promedio, y de costo de uso (3.6).

El índice de calidad del stock de capital tiene en cuenta la heterogeneidad de los bienes producti-vos según los distintos orígenes y destinos, y por extensión refleja las diferencias de productividad y retorno del capital según cada categoría. En esencia es un promedio ponderado del cambio en las participaciones de cada categoría de bienes productivos, donde los factores de ponderación surgen de los retornos relativos de cada categoría respecto al retorno del capital promedio de toda la economía.

Se trató en el presente trabajo mantener el mayor grado de apertura posible de las categorías de bienes de capital. Así las mismas resultan:

•Categoría1:construcciónnoresidencial.•Categoría2:maquinariayequiponacional.•Categoría3:maquinariayequipoimportado.•Categoría4:equipodetransportenacional.•Categoría5:equipodetransporteimportado.

El índice puede ser calculado a partir de la ecuación (3.4) de forma directa.




Gráfico 3

De 1975 a 1981 el índice no representa un factor de ajuste importante. A partir de 1982, el stock de capital reproductivo comienza a decrecer lentamente, al mismo tiempo que el índice de calidad del capital declina debido al drástico cierre de la economía como consecuencia de la Guerra de Malvinas.

Luego de la crisis hiperinflacionaria de 1989-1990 y como consecuencia del cambio de política, el stock de capital crece rápidamente, al mismo tiempo que el índice de calidad del capital también crece dada la incorporación de bienes de capital importado más productivos, como consecuencia de la apertura de la economía.

La crisis del Tequila en 1995 no afecta de manera pronunciada al stock de capital reproductivo (fijo por naturaleza). Como producto de la última gran crisis, el stock de capital reproductivo se ve muy afectado al mismo tiempo que el índice de calidad del capital, recuperándose ambos a partir de 2003.

Del análisis de la serie desagregada de stock de capital reproductivo vemos que de todas formas el ajuste no compromete grandes variaciones, como en el caso del empleo.

En 1975 la participación de la construcción no residencial era del 63,96 %, la maquinaria y equi-po representaba un 29,53 %, y el equipo de transporte el 6,51 % restante.

En 1990, el 76,65 % correspondía a construcción no residencial, la maquinaria y equipo ascendía a un 20,44 %, y el equipo de transporte cubría el 2,91 % final.

Por último, en 2006, la construcción no residencial participaba con un 73,09 %, la maquinaria y equipo constituían el 21,69 %, y por último el equipo de transporte representaba el restante 5,22 %.


64 •

3.3.2 Utilización de la capacidad instalada

Si bien es cierto que para nuestro propósito el stock de capital relevante es el reproductivo, no debemos olvidarnos que el mismo sólo nos brinda información de la capacidad instalada. La misma es de característica contable, pero para poder estimar la función agregada de producción debemos utilizar el stock de capital efectivamente comprometido en la producción, es decir de la capacidad instalada utilizada.

Existen diferentes métodos para estimar este factor de utilización de la capacidad instalada (UCI), como ser la tasa de ocupación laboral, las horas trabajadas, el consumo de energía, y otros más.

Para nuestro caso utilizaremos el coeficiente de utilización de la capacidad instalada suministrado por FIEL, que lo viene proporcionando desde 1970.

Si bien el mismo surge a partir de una encuesta efectuada trimestralmente en la industria manu-facturera, los sectores alcanzados son muy amplios y el coeficiente manifiesta una fuerte correla-ción con las variaciones de la demanda agregada, lo cual justifica su utilización.

Gráfico 4

4. Estimación de una función de producción agregada para la economía Argentina

Al disponer de los datos de producto, empleo, stock de capital reproductivo y los índices de calidad y UCI correspondientes, estamos en condiciones de llevar a cabo las estimaciones de la función de producción agregada para la economía Argentina.

En esencia el presente trabajo pretende probar distintas alternativas de funciones de producción, para verificar cuál de las mismas refleja con mayor propiedad el proceso productivo argentino.

La metodología para poder obtener las distintas estimaciones se referirá en cada sección particu-lar, sin embargo conviene hacer una aclaración en general.




Los datos de producto, empleo y stock de capital corresponden a series de tiempo, en las cuales los valores del pasado influyen en los valores futuros. Por lo cual primero es necesario verificar si dichas series son estacionarias o no, para no caer en el caso de realizar regresiones espurias.

La verificación arroja que dichas series no son estacionarias. No tener en cuenta este hecho nos llevaría a obtener coeficientes incorrectamente estimados, aunque nos den significativos y con muy buen ajuste del modelo.

Para corregir dicho inconveniente vamos a recurrir a la utilización de variables dummies y de tendencia para modelizar la influencia de la tecnología en el cambio de nivel y tendencia de las series.

Del análisis del gráfico del producto podemos definir las variables dummies y de tendencia de la siguiente forma:

• unavariabledetendencia:t.• unavariabledetendenciaquereflejeuncambioenelaño1990debidoalafuerteinversión

durante la convertibilidad: t90.• unavariabledetendenciaentrelosaños1999y2002quetengaencuentalacaídadelpro-

ducto: trec.• unavariabledummyquecaptureuncambiodenivelenelaño1990debidoalacrisishiper-

inflacionaria: d90.• unavariabledummyqueindiqueuncambiodenivelenelaño1995causadoporlacrisisdel

Tequila: d95.• unavariabledummyquerecojauncambiodenivelenelaño2002originadoporlaúltima

gran crisis: d02.

Estas variables que creamos serán utilizadas en todas las regresiones, aunque no necesariamente deberán ser todas significativas.

4.1 Estimación de una función de producción Cobb-Douglas

Esta función de producción fue introducida en 1928 y es una de las más difundidas y aplicadas. Surgió cuando el Senador Paul Douglas le pide al matemático Charles Cobb que le sugiera una función que describa la relación entre las series de producción manufacturera, trabajo y capital para Estados Unidos entre 1899 y 1922.

Su expresión original está dada por:

(4.1)

Esta función es homogénea de grado 1, es decir que exhibe rendimientos constantes a escala. El factor es positivo, menor que 1 y corresponde a la elasticidad del factor trabajo, y por su parte el factor corresponde a la elasticidad del factor capital.

Una de las características que justifica su amplia utilización es que representa bien la constancia de las relaciones producto-trabajo y producto-capital que presentaban las series originales estu-diadas.


66 •

4.1.1 Función de producción Cobb-Douglas bajo la forma intensiva

Manteniendo el supuesto de retornos constantes a escala, si dividimos ambos miembros de (4.1) por L y tomamos logaritmos, la expresión a estimar deviene en:

(4.2)

En la anterior especificación econométrica, el término representa las perturbaciones aleatorias o error del modelo. Es decir, el mismo resume el error cometido si las condiciones de competencia perfecta, vaciamiento de los mercados y los supuestos de los problemas de agregación antes mencionados no se cumplen de forma plena en la realidad.

Esta forma de estimar, que es conocida como la forma intensiva de la función de producción, presenta varias ventajas. En primer lugar la misma resulta ser lineal en sus logaritmos naturales, con lo cual puede ser estimada mediante MCO.

El hecho de tomar logaritmos evita el problema de la heterocedasticidad, y además se reduce la multicolinealidad ocasionada por la falta de independencia entre K y L.

La desventaja de este enfoque es el supuesto de retornos constantes a escala y la constancia de las participaciones de los insumos en el tiempo.

Se presentan dos variantes de la relación capital-trabajo, una en donde los insumos están

ambos ajustados por calidad y el capital por utilización; y otra en donde los insumos solamente

están ajustados por calidad.

Los coeficientes de las variables son todos significativos y presentan en todas las regresiones el signo esperado con muy buena bondad de ajuste. El mejor ajuste corresponde a la regresión (1), en la cual el valor del coeficiente estimado para la razón capital-trabajo ajustados ambos

por calidad y utilización de la capacidad instalada, , es de 0,5509 y el R2 ajustado de 0,98.

En esta regresión se incluyó un término autoregresivo de orden 1, AR(1), para corregir el problema de la autocorrelación presente por ser datos en serie de tiempo.

En la tabla 1 se presentan las distintas estimaciones de esta forma de función de producción. Se incluyen las variables de tendencia t y t90, y la variable dummy d02. Las variables dummies d90 y d95 resultaron en todos los casos no significativas.




Tabla 1

Variable dependiente: Período 1975-2006

Variables Regresiones2

(1) (2) (3) (4)

1,3316 1,3108 1,0264 1,1905

(0,1821) (0,1716) (0,1388) (0,5545)

0,5509 0,5311 0,6272

(0,0554) (0,0605) (0,0496)

0,5084

(0,1753)

t -0,0088

(0,0027)

t90 0,0150

(0,0037)

d02 -0,0765 -0,0653 -0,0862

(0,0207) (0,0231) (0,0408)

AR(1) 0,4574 0,8465 0,6153 0,8067

(0,2013) (0,1195) (0,1611) (0,1358)

Resumen de estadísticos

R2 0,9833 0,9758 0,9706 0,9269

R2 ajustado 0,9800 0,9731 0,9685 0,9188

F 295,2453 362,1397 461,5856 114,0881

Test de Breusch-Godfrey 2,0171 0,8174 0,5201 1,3437

Test de White (sin términos cruzados) 0,7649 1,0048 0,6822 0,2549

Jarque-Bera 0,1674 0,5661 0,5664 1,7382

Tests de raíz unitaria

ADF -4,1293 -3,5898 -4,1942 -3,0311

Phillips-Perron -4,0770 -4,0834 -4,8319 -3,8822

2.- El error estándar figura entre paréntesis.


68 •

4.1.2 Función de producción Cobb-Douglas bajo la forma directa

Si relajamos el supuesto de retornos constantes a escala, la expresión funcional de la Cobb-Douglas se expresa de la siguiente forma:

(4.3)

Como solamente podemos tomar logaritmos en ambos miembros de (4.3), la ecuación a estimar resulta:

(4.4)

Al igual que en la anterior especificación econométrica, el término representa las perturbacio-nes aleatorias o error del modelo. La misma resulta ser lineal en sus logaritmos naturales, con lo cual también puede ser estimada mediante MCO.

Análogamente el hecho de tomar logaritmos evita el problema de la heterocedasticidad, y ade-más se reduce la multicolinealidad ocasionada por la falta de independencia entre K y L.

Para la correcta estimación del modelo se recurrirá a la utilización de las anteriormente definidas variables de tendencia y dummies, y los insumos ajustados por calidad y utilización, o sólo por calidad. En este caso representa el stock de capital ajustado por calidad y utilización, K el stock de capital ajustado por calidad, y L el empleo ajustado por calidad.

Incluiremos además de los contrastes anteriormente realizados, el test de Wald de restricción de parámetros.

Los resultados de las estimaciones realizadas se presentan resumidos en la tabla 2.

Los coeficientes de las variables son todos significativos y presentan en ambas regresiones el signo esperado con muy buena bondad de ajuste. El mejor ajuste corresponde a la regresión (1), en la cual el valor del coeficiente estimado para el capital ajustado por calidad y utilización de la capacidad instalada, , es de 0,5605, el del empleo ajustado por calidad, L, de 0,2426 y el R2 ajustado de 0,9868.

En esta regresión la suma de los coeficientes se aproxima a uno, y el test de Wald arroja que no podemos rechazar la hipótesis nula de que los coeficientes suman uno.3

Procediendo análogamente como en el acápite anterior, se incluyó un término autoregresivo de orden 1, AR(1) para corregir el problema de la autocorrelación.

3.- Estos elementos, analizados en conjunto con el resultado obtenido bajo la forma intensiva, abonan el hecho de estar en presencia de retornos constantes a escala.




Tabla 2

Variable dependiente: 1n(Y ) Período 1975-2006


(1) (2)

1n(A) 3,0492

(0,9874)

1n(KUCI ) 0,5605

(0,0583)

1n(K ) 0,5664

(0,1252)

1n(L) 0,2426 0,5384

(0,0905) (0,1658)

t90 0,0115

(0,0040)

d02 -0,0737 -0,0924

(0,0229) (0,0380)

AR(1) 0,6303 0,8021

(0,2121) (0,1239)


R2 0,9890 0,9623

R2 ajustado 0,9868 0,9581

F 449,4101

Test de Breusch-Godfrey 1,5271 1,1183

Test de White (sin términos cruzados) 0,8243 0,1749

Test de Wald 4,1380 6,6372

Jarque-Bera 1,2169 2,0059


ADF -4,2602 -3,1782

Phillips-Perron -4,0807 -3,9695



70 •

4.2 Estimación de una función de producción CES

La función de producción de elasticidad de sustitución constante (CES) fue derivada por K. J. Arrow, H. B. Chenery, B. S. Minhas y R. M. Solow y difundida en una publicación de 1961.

En un estudio de corte transversal de 24 diferentes industrias correspondientes a 19 países entre los años 1949 y 1955, ellos encuentran una alta correlación significativa en todas las industrias para la regresión de la productividad laboral y el costo laboral por trabajador.

Tomando logaritmos en la expresión original resulta:

(4.5)

En este caso b representa la elasticidad de sustitución entre capital y trabajo. La evidencia empí-rica encontrada a través de la estimación de la anterior ecuación los alentó a encontrar una fun-ción que tuviera las siguientes propiedades:

• homogeneidad.• elasticidaddesustituciónconstanteentrecapitalytrabajo.• laposibilidaddecontarcondiferenteselasticidadesdesustituciónparadiferentesindustrias.

En su expresión original, dicha función resultó:

(4.6)

En donde actúa como un parámetro de eficiencia, es una transformación de la elasticidad de sustitución, y nos referiremos a él como el parámetro de sustitución, y por último representa al parámetro de distribución.

Para el caso de la función de producción CES la elasticidad de sustitución entre los insumos queda expresada por la siguiente relación:

(4.7)

Cuando la función de producción CES se convierte en una función de producción lineal (sustitución perfecta, ), cuando la función CES deviene en una función de produc-ción Leontief (insumos complementarios, ), y por último cuando estamos en el caso de una función de producción Cobb-Douglas (sustitución unitaria, ).

Si a esta función la generalizamos agregándole un parámetro , entonces le permitimos a la misma exhibir rendimientos constantes, crecientes o decrecientes a escala:

(4.8)




Como podemos apreciar la misma es no lineal en los parámetros, por lo que no es posible esti-marla por MCO. Dentro de las distintas alternativas de estimación resulta atractivo el método uti-lizado por Kmenta (1967), el cual aplica un desarrollo en series de Taylor en torno al punto tomando sólo los términos de primer orden, y hace posible así una estimación de la expansión resultante mediante MCO.

Lamentablemente esta opción tuvo que ser descartada por entregar coeficientes no significativos y de signo contrario al esperado en la participación del trabajo. Sin estar exento de problemas5 se optó por un procedimiento de estimación no lineal, aplicado a la siguiente transformación que surge de tomar logaritmos de (5.8) en ambos miembros:

(4.9)

Para facilitar la estimación se impuso por razones prácticas la condición de retornos constantes a escala ( ).6

Los coeficientes resultan todos significativos y presentan en todas las regresiones el signo espe-rado con muy buena bondad de ajuste. Para las mismas se incluyeron además la variable dummy trec y tres términos autoregresivos, AR(1), AR(2) y AR(3) para corregir el problema de la autoco-rrelación presente por ser datos en serie de tiempo.

El mejor ajuste corresponde a la regresión (1), en la cual el valor estimado para es de -0,6947, queda establecido en 0,3358 y el R2 ajustado de 0,9913. Con los valores obtenidos la elastici-

dad de sustitución resulta ser de 0,59, lo cual define a la función de producción obtenida como una forma intermedia entre una función de producción Cobb-Douglas ( ) y una función de producción Leontief ( ).

Los resultados de las estimaciones realizadas se presentan resumidos en la tabla 3.

5.- Estos pueden ser de varios tipos. En primer lugar problemas de no convergencia (overflows o también la matriz deviene en singular) por ser los valores iniciales incorrectos o por contar con pocas observaciones.Otro inconveniente es que la convergencia puede ser lograda en forma extremadamente lenta (usualmente debería lograrse en menos de 10 iteraciones). Pueden encontrarse errores cuando la cantidad de parámetros a estimar son demasiados (se deben fijar algunos como constantes de acuerdo a un buen criterio). Por último en regresión no lineal es posible arribar a una solución (se encuentra convergencia) que no sea la mejor posible por haber logrado un mínimo local.6.- Se tomó esta decisión dadas las dificultades de estimación inherentes al modelo no lineal anteriormente citadas. De todas maneras, la estimación de la función de producción Cobb-Douglas bajo la forma directa aporta elementos suficientes sobre lo acertado de este supuesto.


72 •

Tabla 3

Variable dependiente: 1n(Y ) Período 1975-2006


(1) (2) (3)

Variables independientes

1n(A) 3,0924 2,7681 2,5967

(0,4688) (0,6239) (0,6692)

-0,6947 -0,7399 -0,7720 (0,0582) (0,0795) (0,0925)

0,3358 0,2851 0,2296 (0,0537) (0,0925) (0,1095)

t90 0,0162 0,0131 0,0125

(0,0022) (0,0030) (0,0038)

d02 -0,0901 -0,0749 -0,0769

(0,0134) (0,0201) (0,0233)

trec -0,0167

(0,0038

AR(1) 0,7177 0,5671

(0,2037) (0,2153)

AR(2) -0,4881 -0,5085

(0,1875) (0,2324)

AR(3) -0,4326

(0,1874)


R2 0,9935 0,9902 0,9883

R2 ajustado 0,9913 0,9876 0,9860

Test de White (sin términos cruzados) 1,1282 0,5259 0,9381

Jarque-Bera 6,4426 1,1335 1,4650


ADF -3,3537 -2,7560 -4,2635

Phillips-Perron -4,3475 -5,5551 -3,8610





4.3 Algunos hechos estilizados en torno a las funciones de producción obtenidas

Tal como se mencionó en la introducción, y como se ha venido desarrollando el presente trabajo, la clave del mismo radica en la estimación de distintas funciones de producción, con distintos supuestos para cada una de ellas y luego determinar cual de ellas resulta más apropiada para representar la tecnología de producción de Argentina para el período considerado.

Por lo tanto corresponde ahora realizar la valoración crítica de los resultados obtenidos. Para lle-var adelante esta evaluación procederemos de forma consistente con el proceso de obtención de las funciones de producción que hemos llevado a cabo, es decir, la teoría de los modelos lineal y no lineal, la teoría económica y el contraste con la realidad.

Para ello se considerarán los siguientes criterios, a saber:

• elR2 ajustado.• lasignificacióndeloscoeficientesylacorrespondenciaconsusignoesperado.• laindependenciadeestimacionesadicionales.• lacoherenciaconexperienciasanteriores.• laaplicaciónprácticadelasfuncionesobtenidas.

La función de producción Cobb-Douglas, estimada bajo la forma intensiva, posee elasticidad de sustitución entre insumos unitaria y además le hemos impuesto la restricción de retornos constan-tes a escala. La misma presenta un buen ajuste (R2 ajustado = 0,98), y la participación del capital obtenida es similar a otros intentos anteriores.8 Tiene la ventaja de ser muy simple y útil en cuanto a derivación de aplicaciones al crecimiento económico y el estudio de la productividad total de los factores, por lo que es la elegida para ese propósito en lo que resta del presente estudio.

La función de producción Cobb-Douglas, estimada bajo la forma directa, sigue presentando elasticidad de sustitución entre insumos unitaria pero esta forma de estimación relaja la restricción de presentar retornos constantes a escala, exhibiendo un muy buen ajuste (R2 ajustado = 0,98).

A la función de producción CES, si bien puede exhibir retornos no constantes a escala, se le ha preferido imponerle la restricción de retornos constantes para facilitar su estimación, dado que la misma es no lineal. Presenta un excelente ajuste (R2 ajustado = 0,9913) y una participación del capital consistente con experiencias realizadas en otros países.9 A nuestro juicio es la que mejor representa la tecnología de nuestro país para los años estudiados.

A continuación resumimos las tres tecnologías más representativas:

Cobb-douglas (forma intensiva)

Cobb-douglas (forma directa)

CES

8.- Un ejemplo concreto es el de Meloni (1999).9.- En general la participación del capital oscila entre 0,30 y 0,45. Consultar las distintas referencias bibliográficas.


74 •

5. Aplicaciones de las funciones de producción obtenidas

En el acápite anterior establecimos las tres tecnologías más representativas para caracterizar al proceso productivo argentino. Sin embargo, las mismas no comparten el mismo grado de practi-cidad en cuanto a las distintas aplicaciones que podemos esperar para cada una de ellas.

La función de producción CES es quizás la más apropiada a los efectos de predecir el PBI espe-rado, como así también un valor particular del mismo, para valores dados de stock de capital y empleo. Este tipo de aplicación no será llevada adelante en el presente trabajo, dado que solo basta con calcular el error de predicción y luego determinar el intervalo de predicción correspon-diente a los valores de stock de capital y empleo dados.

La utilización de una función de producción cobra especial relevancia cuando la destinamos al análisis de la productividad total de los factores, a la determinación del producto potencial y a la contabilidad del crecimiento. Otra aplicación posible se remonta a la estimación del balance estructural de una economía, pero la misma depende de forma indirecta, ya que para su cálculo es necesario determinar primero el producto potencial.

Estas son las tres aplicaciones fundamentales que desarrollaremos en las secciones siguientes. Ahora bien, la forma funcional elegida para esto será la función de producción Cobb-Douglas bajo la forma intensiva. La misma presenta rendimientos constantes a escala, además que al tomar logaritmos en ambos miembros, la misma se reduce a una expresión algebraica sencilla y fácil de operar.

5.1 Análisis de la productividad total de los factores

La forma funcional elegida permite aplicar la metodología de las fuentes del crecimiento econó-mico para descomponer al mismo en la contribución que cada factor, es decir, stock de capital, empleo y la productividad total de los factores, hace al producto.

La productividad total de los factores (PTF) puede ser calculada por diferencia, por lo que se la conoce también como residuo de Solow, y representa en gran medida el aporte del progreso tecnológico.Partimos de la expresión conocida:

(6.1)

Tomando logaritmos en ambos miembros:

(6.2)

Ahora si despejamos a 1n(A) nos queda:

(6.3)

De esta forma podemos calcular la tasa de cambio de la productividad total de los factores en función de la tasa de cambio del producto y las tasas de cambio del stock de capital y del empleo. Si le calculamos el antilogaritmo, podemos obtener la serie de la productividad total de los fac-tores para el período dado.




Dicha serie fue obtenida haciendo uso de la regresión (1), en la cual la participación del capital es de 0,55. La misma puede ser observada en el gráfico 5:

Gráfico 5

Tomando como año base 1975=100, se construyó un índice para dicha serie el cual puede ser observado en el gráfico 6:

Gráfico 6

Tanto a la serie original como al índice construido se les aplicó el filtro de Hodrick-Prescott, a los efectos de descomponer las mismas en tendencia y ciclo. La serie de PTF filtrada será utilizada más adelante para las estimaciones del producto potencial.


76 •

En la Etapa I, que corresponde a los años 1975 a 1981, la PTF aumentó un 1,1 %. Esto es equiva-lente a un crecimiento interanual promedio del 0,16 %. El PBI creció durante este período a una tasa interanual del 0,93 %.

La Etapa II comienza con la guerra de Malvinas y se extiende entre los años 1982 hasta 1990. En este período la PTF experimenta una caída del 9,4 %, equivalente a una tasa interanual promedio del –1,07 %. El producto experimenta durante este lapso de tiempo una caída interanual del 0,21 %.

La Etapa III está caracterizada por la convertibilidad y se extiende desde 1991 hasta 2001, período en el cual la PTF experimenta una fuerte recuperación y crece un 5,7 %, correspondiente a una tasa interanual promedio del 0,49 %. El PBI también se recupera a una tasa interanual del 2,32 %.

Por último, la etapa IV corresponde a la recuperación posterior a la crisis del 2001, signada por la devaluación y alto crecimiento. En este período la PTF profundiza su crecimiento, el cual es del 7,6 %, correspondiente a un crecimiento interanual promedio del 1,45 %. El PBI crece en esta etapa a una tasa interanual del 7,04 %.

Si consideramos todas las etapas en su conjunto, es decir, el período de 1975 a 2006, la PTF experimenta una caída del 3,8 %, que equivale a una tasa interanual promedio del –0,12 %. En este período el producto creció a una tasa interanual del 1,86 %.

Para completar este análisis corresponde estudiar el comportamiento cíclico de la PTF en corres-pondencia con el PBI, hecho que puede ser observado en el gráfico 7:

Gráfico 7

-2%

0%

2%

4%

6%

8%




Como podemos apreciar la PTF presenta un marcado comportamiento procíclico, es decir, que variaciones interanuales de la PTF se corresponden con variaciones del mismo signo del PBI.

Teniendo en cuenta las consideraciones ya mencionadas, y de la observación atenta de los gráfi-cos 5 a 7, podemos adelantar las siguientes conclusiones preliminares:

• entrelosaños1982y1990laPTFexperimentaunprocesodeprolongadodeterioro,coinci-dente con una sostenida caída del producto.

• enlosañoscaracterizadosporlaconvertibilidadyunafuerteinversión,laPTFlograunafuerterecuperación, de la misma forma que el PBI.

• apartirde2002yhasta2006laPTFincrementasuritmoderecuperación,lograndosutasade mayor crecimiento, al igual que el PBI.

• apesardelasmejorasconcretadasenestasdosúltimasetapas,marcadasporlacrisishiperin-flacionaria y la de la caída de la convertibilidad, la PTF no logra recuperar el nivel de los años 1975 a 1981.

• noobstantedelcarácterprocíclicodelaPTF,tomandoenconsideracióntodaslasetapasenconjunto, en el período 1975 a 2006 el PBI creció pero la PTF disminuyó.

• teniendoencuentalacontribucióndelaPTFalcrecimientodelproducto,podemosadelantarque tanto en el período de la convertibilidad, como en la recuperación posterior a 2002, el patrón de crecimiento ha sido más bien “extensivo” que “intensivo”.10 Este hecho será ana-lizado con mayor profundidad en la sección correspondiente a contabilidad del crecimiento.

Por lo tanto estamos en condiciones de anticipar que el crecimiento logrado a partir de 1990 no está basado en una mejora sustancial de la PTF sino en acumulación de factores.

5.2 Estimación del producto potencial

Una de la aplicaciones más difundidas de la función de producción agregada es quizás la deter-minación del producto potencial y la brecha de producto, entendida como el producto actual menos el producto potencial.

Además del trabajo anteriormente mencionado de Meloni (1999), quien luego de estimar una función de producción agregada de tipo Cobb-Douglas la aplica al cálculo del producto potencial, citaremos a continuación algunos intentos similares.

Estos tienen en común que logran la estimación del producto potencial mediante una función de producción, pero que no ha sido obtenida econométricamente, sino que los coeficientes de participación del capital y el empleo fueron extraídos de la información brindada por Cuentas Nacionales.

Maia y Kweitel (2003) estiman el producto potencial de nuestro país para los años 1960 a 2000, utilizando datos de Cuentas Nacionales para el año base 1993 con una función de producción Cobb-Douglas que exhibe rendimientos constantes a escala.

En un trabajo que cuenta con antecedentes en 2004, Elosegui, Garegnani y otros (2006) utilizan como una de las alternativas posibles una función de producción de tipo Cobb-Douglas con

10.- Encontramos conclusiones opuestas en los trabajos oficiales de Meloni (1999), Maia y Nicholson (2001) y Maia y Kweitel (2003). Sin embargo conclusiones similares a nuestro análisis pueden encontrarse en Coremberg (2004) y (2006).


78 •

rendimientos constantes a escala, para estimar el producto potencial de Argentina para los años 1980 a 2005.

La función de producción Cobb-Douglas estimada anteriormente bajo la forma intensiva, será aplicada a la estimación del producto potencial y la brecha de producto (GAP) utilizando la ten-dencia de la serie de PTF aportada por el filtro de Hodrick-Prescott.

Se calculará el producto potencial considerando el pleno uso de los factores productivos. Para el stock de capital potencial esto significa igualarlo al stock de capital actual. Para el Empleo poten-cial debemos considerar a la población económicamente activa (PEA), también conocida como fuerza de trabajo, menos una cierta cantidad que supone el desempleo friccional.

Para el caso del Empleo potencial es posible considerar dos aspectos diferentes, lo cual llevaría a dos estimaciones distintas del producto potencial, según sean los supuestos utilizados. A conti-nuación el producto potencial será calculado con ambos métodos.

5.2.1 Estimación del producto potencial bajo el enfoque de la tasa natural de desem-pleo (TND)

Para proceder al cálculo del producto potencial se utilizará la ecuación:

(6.4)

Se parte de la serie de PTF obtenida con el filtro de Hodrick-Prescott. El stock de capital también es filtrado con el mismo método para separar el componente cíclico.

En cuanto al Empleo, este enfoque de orientación Keynesiana supone al Empleo potencial como el resultado de la PEA (fuerza de trabajo), menos una tasa natural de desempleo (TND) de carácter friccional.11

Si bien no existen estimaciones acerca de esta tasa para nuestro país, hemos supuesto la misma en 5 % para los años 1975 a 1990, y 7 % para los años 1991 a 2006. El incremento intenta recoger los cambios en el mercado de trabajo debidos a las grandes transformaciones sufridas a partir de los ´90.

Luego de su cálculo, el Empleo potencial también fue ajustado por calidad.

Los resultados obtenidos para el producto potencial y la brecha de producto, junto con el PBI observado, pueden ser apreciados en el gráfico 8:

11.- Este enfoque es consistente con el pleno uso de los recursos productivos.




Gráfico 8

La serie de producto potencial 1 (TND) muestra un comportamiento suavizado debido al filtrado previo de la PTF y el stock de capital.

El crecimiento del producto potencial comienza a tomar relevancia a partir de 1990, producto de la fuerte inversión en stock de capital. La crisis de 2002 apenas lo afecta.

Todas las veces en que el producto observado supera al producto potencial, podemos interpretar-las con episodios de recalentamiento de la economía. Los más cercanos que encontramos son los de 1994, 1997-1998, y más recientemente en 2006.

La brecha de producto (GAP) mantiene en general un comportamiento estacionario, con excep-ción de los años 2002-2003, debidos a la crisis posterior a la caída de la convertibilidad.

5.2.2 Estimación del producto potencial bajo el enfoque del suavizamiento del empleo mediante el filtro de Hodrick-Prescott (HP)

Para calcular el producto potencial utilizaremos la misma ecuación que en la sección anterior. Igual procedimiento se sigue con la serie de PTF y con la serie de stock de capital, a las cuales se les quita el componente cíclico con el filtro de Hodrick-Prescott.

La diferencia principal de este enfoque respecto del anterior reside en el tratamiento a la serie de empleo, para obtener el Empleo potencial. En este enfoque, de orientación Neoclásica, se aplica también a la serie de empleo el filtro de Hodrick-Prescott.12

12.- Según este enfoque el producto potencial se origina en perturbaciones de productividad de la oferta agregada, que determinan el crecimiento de largo plazo.

PBI oBservado y Producto PotencIal 1(tnd)

Mill

ones

de

$ de

199

3

Mill

ones

de

$ de

199

3


80 •

El objetivo que se persigue con esta metodología es capturar las posibles variaciones de la tasa natural de desempleo, al mismo tiempo que se pretende también capturar la componente de largo plazo de la participación de la fuerza de trabajo.

Si bien la utilización de este enfoque está plenamente justificada, el mayor cuestionamiento reside en que el filtro de Hodrick-Prescott suaviza excesivamente los episodios de cambios estruc-turales, o cambios bruscos debidos a las frecuentes crisis que ha atravesado nuestro país.

No obstante estas diferencias metodológicas derivadas de distintas visiones macroeconómicas, el patrón obtenido con este enfoque es muy similar al obtenido en el acápite anterior.

Los resultados obtenidos para el producto potencial y la brecha de producto, junto con el PBI observado ser observados en el gráfico 9:

Gráfico 9

La serie de producto potencial 2 (HP) muestra un comportamiento suavizado debido al filtrado previo de la PTF, el stock de capital y la serie de empleo.

El crecimiento del producto potencial también comienza a tomar relevancia a partir de 1990, producto de la fuerte inversión en stock de capital. Debido al filtrado de la serie de empleo y el enfoque de largo plazo, la crisis de 2002 no impacta en el producto potencial.

En general, en los últimos años el producto potencial se mantiene por encima del producto obser-vado. Los años más cercanos en los cuales esto no ocurre son 1997-1998 y 2006, aunque con brechas positivas menores que las registradas con el método anterior. En estos años se registran entonces episodios de recalentamiento.

PBI oBservado y Producto PotencIal 2(hP)

Mill

ones

de

$ de

199

3

Mill

ones

de

$ de

199

3




La brecha de producto (GAP) mantiene en general un comportamiento estacionario, con excep-ción de los años 2002-2003, en los que la misma es mayor a la obtenida con el enfoque ante-rior.

5.3 Contabilidad del crecimiento

Por último, cerrando el apartado de aplicaciones, podemos realizar la contabilidad del crecimien-to de la economía de Argentina para el período dado.

La idea subyacente en esta metodología consiste en descomponer el crecimiento económico para determinar la contribución de cada factor al mismo, lo que nos posibilitará reconocer el patrón de crecimiento de la economía.

Así calcularemos el aporte de dicho crecimiento interanual de los insumos al crecimiento del PBI, lo que comúnmente se denomina contribución al crecimiento (también sería equivalente denomi-narla fuentes del crecimiento). Para ello afectaremos al crecimiento interanual del stock de capital por su participación, que es de 0,55, y lo mismo haremos con el empleo, cuya participación es de 0,45. Estos valores han sido tomados de la ecuación ya conocida:

(6.5)

Tabla 8

PBI Capital Empleo PTF

Etapa Período Variación Contribución Contribución Contribución

interanual interanual interanual interanual [%] [%] [%] [%]

I 1975 a 1981 0,93 0,26 0,51 0,16

II 1982 a 1990 -0,21 -0,53 1,40 -1,07

III 1991 a 2001 2,32 1,10 0,72 0,49

IV 2002 a 2006 7,04 3,25 2,27 1,45

Todas 1975 a 2006 1,86 0,75 1,24 -0,12

La tabla 8 recoge las tasas de variación interanual del capital, el empleo y la PTF en comparación con el PBI de cada uno de los períodos considerados, afectados por sus respectivas participacio-nes.

El gráfico 10 nos muestra las fuentes del crecimiento, reflejando en el eje de ordenadas la contri-bución al crecimiento por etapas para cada factor, tras haber sido afectados el stock de capital y el empleo por sus respectivas participaciones:


82 •

Gráfico 10

En la Etapa I, la economía creció a una tasa interanual del 0,93 %. La contribución del stock de capital fue del 0,26 % y la del empleo 0,51 %. El aporte de la PTF alcanza el 0,16 %.

La etapa II refleja un decrecimiento del PBI a una tasa interanual del 0,21 %. En dicho lapso de tiempo el capital contribuye con –0,53 % y el empleo aporta el 1,40 %. La contribución de la PTF fue de –1,07 %.

Durante la etapa III, el producto crece a una tasa interanual de 2,32 %. El stock de capital aporta el 1,10 % y el empleo contribuye con un 0,72 %. La contribución de la PTF es del 0,49 %.

Por último, en la etapa IV, la economía crece a una tasa interanual de 7,04 %. Durante este perío-do el capital contribuye con el 3,25 % y el empleo con el 2,27 %. La PTF aporta el 1,45 %.

Si consideramos todas las etapas, es decir de 1975 a 2006, el PBI creció a una tasa interanual del 1,86 %. En ese período el stock de capital contribuye con el 0,75 % y el empleo con un 1,24 %. La PTF contribuye con –0,12 %.

Podemos apreciar que la contribución de la PTF en los períodos en los cuales la economía crece, es baja en comparación con las contribuciones del stock de capital y del empleo.

En cambio cuando el PBI se contrae, la contribución de la PTF a dicha disminución es comparable a los aportes del capital y del empleo, tomados estos en valor absoluto dado que tienen signos distintos.

Los resultados de este análisis recurriendo a las fuentes del crecimiento son consistentes con las conclusiones anticipadas en la sección en que analizamos la PTF.

Podemos afirmar entonces que el patrón de crecimiento de nuestro país es más bien de tipo “extensivo”, es decir basado en la acumulación de factores, que de tipo “intensivo”, que corres-pondería a una mejora sustancial de la PTF.




A pesar de las mejoras evidenciadas, esta conclusión se aplica también a la etapa III, caracterizada por la convertibilidad, y a la etapa IV marcada por la recuperación posterior a la crisis de 2002.

Otro indicio del crecimiento basado en acumulación de factores, es que en todo el período de estudio, la PTF sufrió un deterioro a una tasa interanual del 0,12 %.

6. Conclusiones

Después de haber revisado la historia de otros intentos por estimar para nuestro país funciones de producción agregadas, nuestro esfuerzo se centró en estimar una función de producción agrega-da para la economía Argentina para el período 1975 a 2006.

Al contar con las series correspondientes fue posible entonces proceder a la estimación de las distintas funciones de producción, teniendo en cuenta la utilización de variables dummies y de tendencia para modelizar la influencia de la tecnología en el cambio de nivel y tendencia de las series.

Este resultó necesario dado que disponemos de datos en series de tiempo y debíamos evitar el caso de realizar regresiones espurias.

De acuerdo a los criterios fijados en el acápite correspondiente, de la totalidad de las funciones de producción agregadas estimadas, no todas se adecuan de la misma manera a la descripción del proceso productivo de Argentina.

Las funciones de producción obtenidas que mejor logran su cometido, son la Cobb-Douglas, estimada tanto bajo la forma intensiva como bajo la forma directa y la CES.

Teniendo en cuenta estas tres últimas funciones de producción, podemos apreciar que los supues-tos establecidos facilitan las estimaciones, aunque también introducen variaciones en los coefi-cientes de participación obtenidos. La mejor estimación posible se logra cuando los supuestos establecidos se ajustan bien a la realidad y permite obtener coeficientes realistas.

Tal es el caso de la función de producción CES, a la cual le imponemos retornos constantes a escala, pero le permitimos una elasticidad de sustitución entre insumos constante pero diferente de uno, hechos que caracterizan de forma adecuada al proceso productivo Argentino. En este caso hemos obtenido el mejor ajuste posible.

La función de producción Cobb-Douglas estimada bajo la forma intensiva proporciona también un muy buen ajuste, además que resulta extremadamente útil para derivar aplicaciones diversas.

Como ya hemos anticipado en la sección correspondiente, un adecuado enfoque situacional determinará la función de producción más adecuada a utilizar en cuanto a las aplicaciones al crecimiento económico.

En cuanto a dichas aplicaciones, en primer lugar se obtuvo la serie de productividad total de los factores, con un doble propósito; el del análisis en sí mismo, y la posterior aplicación a la deter-minación del producto potencial.

Del análisis de la PTF, podemos apreciar en la misma un deterioro tendencial, que se hace pronun-


84 •

ciado a partir de 1982, año que marca el cierre de la economía. La década de la convertibilidad hace un gran aporte para lograr un cambio de tendencia, pero los niveles registrados no alcanzan a una plena recuperación de la PTF. La recuperación de la economía posterior a la crisis del 2002 y la devaluación aportan una gran mejora de la PTF, aunque sin poder recomponer los niveles de comienzo de la serie.

Si consideramos el período de 1975 a 2006, la PTF experimenta una caída del 3,8 %, que equi-vale a una tasa interanual promedio del –0,12 %.

La etapa de 1975 a 1981 registra una PTF significativamente distinta del resto de las etapas posteriores. Ahora bien, si nos atenemos al período que llega a 1990, y el período posterior, no encontramos evidencia de una diferencia significativa de la PTF.

Podemos concluir entonces a partir de estos resultados que el crecimiento de nuestra economía no está basado en una mejora de la PTF, sino en acumulación de factores, aún cuando la inversión y modernización de los bienes de capital han sido evidentes como en el caso de la convertibili-dad.13

Otro hecho que resulta significativo es que la PTF se incrementa cuando la economía permanece abierta, pero su deterioro es mucho mayor en períodos en los cuales la economía permanece cerrada. Además muestra un marcado comportamiento procíclico, si nos atenemos a las etapas ya mencionadas en cuestión.

La segunda parte de las aplicaciones se centró en la determinación del producto potencial. Par-tiendo de la serie de PTF anteriormente obtenida, se calcularon la PTF potencial y el stock de capital potencial, ambos a partir de la utilización del filtro de Hodrick-Prescott.

La particularidad introducida en el producto potencial proviene del criterio para determinar el empleo potencial, originados en distintas visiones macroeconómicas.

El enfoque de la tasa natural de desempleo, de orientación Keynesiana, nos obliga a establecer un supuesto del comportamiento de la misma, para restárselo a la PEA y hallar así el empleo potencial.

El enfoque que surge de suavizar la serie de empleo con el filtro de Hodrick-Prescott para obtener la serie de empleo potencial es de orientación Neoclásica, y persigue el objetivo de capturar su componente de largo plazo y sus posibles variaciones, habida cuenta de que no existen hasta el momento buenas estimaciones para la misma en nuestro país.

Los resultados obtenidos bajo ambos enfoques son muy parecidos y muestran que el produc-to potencial presenta un comportamiento suavizado y en general es superior al observado. En cuanto a la brecha de producto (GAP), la misma evidencia un comportamiento estacionario, con excepción de los años 2002-2003, años de grandes cambios estructurales debidos a la gran crisis de la caída de la convertibilidad.

13.- Tal como anticipamos se encuentran conclusiones opuestas en los trabajos oficiales de Meloni (1999), Maia y Nicholson (2001) y Maia y Kweitel (2003), en los cuales, o bien utilizan otro método de depreciación, o bien no ajustan insumos por calidad o por utilización. En Coremberg (2004) y (2006) se pueden encontrar conclusiones similares a las que hemos arribado en nuestro análisis, con la utilización de índices óptimos.




Sin embargo, en la etapa que va de 1982 a 1990, encontramos varios episodios en los cuales el producto observado supera al producto potencial; en los últimos años, estos han sido también manifiestos en 1994, 1997-1998, y más recientemente en 2006.

Estos casos nos muestran que en esos períodos la economía puede presentar ese fenómeno a costa de una mayor presión inflacionaria y recalentamiento de la misma.

Finalmente cerramos la sección de aplicaciones con un análisis de la contabilidad del crecimiento, para lo cual recurrimos a la función de producción Cobb-Douglas bajo la forma intensiva para que nos aporte los coeficientes de participación del capital y el trabajo en el PBI.

Hemos podido apreciar que la PTF contribuye escasamente al crecimiento del PBI, en compara-ción con el aporte del capital y el trabajo. En la etapa I el capital y el trabajo explican el 83 % del crecimiento del PBI. En la etapa III ambos insumos explican el 78 % del crecimiento del producto, y en la etapa IV el capital y el trabajo conjuntamente explican también el 78 % del crecimiento de la economía.

Sin embargo en la etapa II, en la cual el PBI experimenta una contracción a una tasa interanual del 0,21 %, el deterioro de la PTF (-1,07 %) es el doble que la disminución del stock de capital (-0,53 %), es decir, que en períodos de achicamiento de la economía la PTF explica en una gran medida dicha disminución.

Podemos concluir entonces que el patrón de crecimiento de Argentina es “extensivo”, es decir, basado en la acumulación de factores, y no “intensivo”, cuando es la PTF la protagonista principal del crecimiento.

Esta conclusión se hace extensiva al período de la convertibilidad, donde la inversión en bienes de capital fue muy grande, lo que sugiere que el crecimiento de la productividad de la economía fue tan solo aparente e invita a un análisis de dicho período más a fondo.

Por otra parte esta conclusión se puede aplicar también al período de la recuperación posterior a la crisis de 2002, para determinar que el crecimiento a tan altas tasa interanuales experimentado en esos años, no es sustentable a largo plazo sin presión inflacionaria.

8. Bibliografía

Aldabe, H. (1965): “Ensayo de Establecimiento de una Función de Producción”, Desarrollo Eco-nómico, Vol. 5, Nº 17/19, Tomo 2 (Apr. – Dec., 1965), pp. 371-388.

Arrow, K. J., Chenery, H. B., Minhas, B. S. and Solow, R. M. (1961): “Capital-Labor Substitution and Economic Efficiency”, The Review of Economics and Statistics, Vol. 43, Nº 3 (Aug., 1961), pp. 225-250.

Barro, Robert J. (1998): “Notes on Growth Accounting”, Harvard University, Mimeo (Dec., 1998).

Butera, Mauricio y Kosacoff, Luciano (1997): “Evolución del stock de capital en la Argentina: el proceso de acumulación en las últimas tres décadas”, Anales de la AAEP, Bahía Blanca 1997.

CEP (1997): “Evolución del Stock de Capital en Argentina”, Centro de Estudios de la Produc-ción, Secretaría de Industria, Comercio y Minería, Ministerio de Economía y Obras y Servicios Públicos.

Cobb, Charles W. and Douglas, Paul H. (1928): “A Theory of Production”, The American Eco-


86 •

nomic Review, Vol. 18, Nº 1, Supplement, Papers and Proceedings of the Fortieth Annual Meeting of the American Economic Association (Mar., 1928), pp. 139-165.

Coremberg, Ariel A. (2004): “TPF Growth in Argentina during the 1990`s. A Stylized Fact?”, International Association for Research in Income and Wealth 28th Conference.

Coremberg, Ariel A. (2004): “Estimación del Stock de Capital Fijo de la República Argentina 1990-2003. Fuentes, Métodos y Resultados”, Dirección Nacional de Cuentas Nacionales.

Coremberg, Ariel A. (2004): “El crecimiento de la productividad de la economía argentina duran-te la década de los noventa: “mito o realidad””, Instituto de Desarrollo Económico y Social, Cuadernos del IDES Nº 3.

Dichiara, Raul Oscar (1980): “Funciones de Producción para la Economía Argentina”, Anales de la AAEP, Mar del Plata 1980.

DNCN (1999): “Sistema de Cuentas Nacionales. Argentina - Año base 1993”, Ministerio de Eco-nomía, Dirección Nacional de Cuentas Nacionales.

Elías, Víctor J. (1983): “Funciones de Producción de Productos Múltiples”, Universidad Nacional de Tucumán, Anales de la AAEP, Tucumán 1983.

Elías, Víctor J. (1989): “Estimación de la Función de Producción Agregada con Datos de Corte Transversal de Países”, Universidad Nacional de Tucumán, Anales de la AAEP, Rosario 1989.

Elosegui, Pedro, GAREGNANI, Lorena, LANTERI, Luis, LEPONE, Francisco y SOTES PALADINO, Juan (2006): “Estimaciones Alternativas de la Brecha del Producto para la Economía Argentina”, BCRA, Subgerencia General de Investigaciones Económicas.

FIEL (1980): “Indicadores de Coyuntura”, Números varios, Fundación de Investigaciones Econó-micas Latinoamericanas.

FIEL (2002): “Productividad, Competitividad, Empresas. Los Engranajes del Crecimiento”, Funda-ción de Investigaciones Económicas Latinoamericanas.

Goldberg, Samuel y Ianchilovici, Beatriz (1988): “El stock de capital en la Argentina”, Desarrollo Económico, Vol. 28, Nº 110, (Jul. – Sep., 1988), pp. 281-304.

Granger, C. W. J. and Newbold, P. (1974): “Spurious Regressions in Econometrics”, Journal of Econometrics, Nº 2 pp. 111-120.

Graña, Juan M., Kennedy, Damián, Lindenboim, Javier y Pissaco, Carlos (2005): “La distribución funcional del ingreso en Argentina: Incidencia de los precios relativos en la última década”, Proyecto UBACyT E-003, Mimeo.

Greene, William H. (1998): “Análisis Econométrico”, Tercera Edición, Prentice Hall.Hoch, Irving (1962): “Estimation of Production Function Parameters Combining Time-Series and

Cross-Section Data”, Econometrica, Vol. 30, Nº 1 (Jan., 1962). pp. 34-53.Hodrick, Robert and Prescott, Edward (1997 ): “Postwar U.S. Business Cycle: An Empirical Investi-

gation”., Journal of Money, Credit and Banking, Vol. 29, Nº 1 (Feb., 1997), pp. 1-16.Hofman, André A. (1990): “Economic Growth and Performance in Latin America”, Serie Refor-

mas Económicas Nº 54, ECLAC, UN, Marzo.INDEC (1987): “Encuesta Permanente de Hogares. Presentación de Datos del Gran Buenos Aires

– Abril 1987”, Instituto Nacional de Estadística y Censos, Secretaría de Planificación.Jehle, Geoffrey A. and Reny, Philip J. (2001): “Advanced Microeconomic Theory”, Second Edition,

Addison Wesley.Jorgenson, Dale W. and Grilliches, Zvi (1967): “The Explanation of Productivity Change”, The

Review of Economic Studies, Vol. 34, Nº 3 (Jul., 1967), pp. 249-283.Katz, Jorge M. (1969): “Una interpretación de largo plazo del crecimiento industrial argentino”,

Desarrollo Económico, Vol. 8, Nº 32, Argentina: Estrategias de Desarrollo (Jan. – Mar., 1969), pp. 511-542.

Lanteri, Luis N. (1999): “Fuentes de crecimiento en la Argentina y en los países recientemente industrializados del Este de Asia ¿Podría pensarse en un milagro del crecimiento económico Argentino?”, Documento de trabajo Nº 6, Mimeo.




Lindenboim, Javier ,Graña, Juan M. y Kennedy, Damián (2005): “Distribución funcional del ingre-so en Argentina. Ayer y hoy”, Universidad de Buenos Aires, CEPED, Documento de trabajo Nº 4.

Maia, José Luis y Nicholson, Pablo (2001): “El Stock de capital y la Productividad Total de los Fac-tores en la Argentina”, Dirección Nacional de Coordinación de Políticas Macroeconómicas, Secretaría de Política Económica y Regional, Ministerio de Economía.

Maia, José Luis and Kweitel, Mercedes (2003): “Argentina: Sustainable Output Growth after the Collapse”, Dirección Nacional de Políticas Macroeconómicas, Mimeo.

Meloni, Osvaldo (1999): “Crecimiento Potencial y Productividad en la Argentina: 1980-1997”, Universidad Nacional de Tucumán y Ministerio de Economía, Obras y Servicios Públicos, Mimeo.

Miller, Edward (1990): “Can a Perpetual Inventory Capital Stock Be Used for Production Function Parameter Estimation?”, Journal of the Review of Income and Wealth, Series 36, Nº 1 (Mar., 1990).

Motulsky, Harvey J. and Ransnas, Lennart A. (1987): “Fitting curves to data using nonlinear regression: a practical and nonmathematical review”, University of California, Department of Pharmacology, FASEB J. pp. 365-374.

OECD (2001a): “Measuring Capital. OECD Manual. Measurement of Capital Stocks, Consump-tion of Fixed Capital and Capital Services”, OECD Publications, France.

OECD (2001 b): “Measuring Productivity. OECD Manual. Measurement of Aggregate and Indus-try-Level Productivity Growth”, OECD Publications, France.

Roldós, Jorge (1997): “Potential Output Growth in Emerging Market Countries: The Case of Chile”, International Monetary Fund, Working Paper 97/104.

Solow, Robert M. (1956): “The Production Function and the Theory of Capital”, The Review of Economic Studies, Vol. 23, Nº 2 (1955-1956), pp. 101-108.

Solow, Robert M. (1957): “Technical Change and the Aggregate Production Function”, The Review of Economics and Statistics, Vol. 39, Nº 3 (Aug., 1957), pp. 312-320.

Walters, A. A. (1963): “Production and Cost Functions: An Econometric Survey”, Econometrica, Vol. 1/2 (Jan. – Apr., 1963), pp. 1-66.

Wolkowitz, Benjamin (1973): “Estimation of a Set of Homothetic Production Functions: A Time Series Analysis of American Postwar Manufacturing”, Southern Economic Journal, Vol. 39, Nº 4 (Apr., 1973), pp. 626-637.

Young, Alwyn (1995): “The Tyranny of Numbers: Confronting the Statistical Realities of the East Asian Growth Experience”, The Quarterly Journal of Economics, Vol. 110, Nº 3 (Aug., 1995), pp. 641-680.


88 •

Did bankruptcy reform lead to looser mortgage lending standards? Evidence from the U.S. mortgage market 2000-2007

Prajakta Bhide,1 Lucia Fiorito,2 Zachary Noteman3 and Kunal Sawardekar4

Abstract

This paper seeks to find an exogenous cause for deterioration in mortgage lending standards since 2005 that contributed to the subprime mortgage crisis in the U.S. We find that the new means test provision in the Bankruptcy Abuse Prevention and Consumer Protection Act (BAPCPA) of 2005 was such an exogenous shock in the mortgage market. We show that the means test, which makes Chapter 7 bankruptcy unavailable to relatively better off borrowers, caused a shift in the supply of mortgage credit from better off borrowers to relatively poorer borrowers. Simul-taneously, we found borrowers being charged higher interest rates, for all classes of income. Our findings imply that BAPCPA may be a contributing factor toward the deterioration of lending standards in the U.S. mortgage market.

Resumen

El presente trabajo intenta encontrar una causa exógena al deterioro, a partir de 2005, en los estándares de crédito hipotecario que contribuyeron a la crisis subprime en los Estados Unidos. Sostenemos que la nueva provisión de la prueba de medios de la ley Bankruptcy Abuse Preven-tion and Consumer Protection Act (BAPCPA) de 2005 fue dicho shock exógeno en el mercado hipotecario. Mostramos que la prueba de medios, que impide solicitar la bancarrota bajo Chapter 7 a los deudores con mayores ingresos relativos, causó un desplazamiento de la oferta de cré-dito hipotecario de deudores con mayores ingresos a deudores con menores ingresos relativos. Simultáneamente, observamos que todos los deudores debieron pagar tasas de interés más altas, independientemente del nivel de ingresos. Nuestros resultados implican que la ley BAPCPA podría ser un factor que contribuyó al deterioro en los estándares de crédito en el mercado hipotecario de los Estados Unidos.

Keywords: Bankruptcy, means test, mortgage default, lending standards

JEL Classification: [G01] [G21] [G33] [K35]

1.- MA in Economics, New York University, [email protected] MA in Economics, New York University, [email protected] MA in Economics, New York University, [email protected] MA in Economics, New York University, [email protected]




1. Introduction

After a period of rapid expansion beginning since the mid 1990s, the U.S. subprime mortgage market began showing signs of stress in mid-2006. By 2007 it was evident that trouble in the subprime mortgage market had reached crisis proportions, as seen from record mortgage delin-quency rates and foreclosures. In the following months, the crisis has become more widespread, with foreclosures spreading across the near prime and prime market categories.

The turmoil in the U.S. mortgage market has been the subject of recent research. Gerardi et al. (2008) and Mayer et al. (2008) suggest that the decline in home prices beginning mid-2006 played an important part in driving foreclosures and mortgage delinquencies. Dell’Ariccia et al. (2008) link record delinquency rates to a trend of deteriorating lending standards appar-ent through higher loan to income ratios and declining mortgage denial rates during the rapid expansion of credit to the subprime market segment. Mian and Sufi (2008) show an expansion in the supply of mortgage credit to cause a decline in lending standards. Keys et al. (2008) find that securitization of mortgages affects the incentives of lenders to screen borrowers, leading to riskier loans being made.

Some researchers have noted a change in the mortgage market around 2005. Dell’Ariccia et al. (2008) observe, “While denial rates in the prime mortgage market closely mimic the evolution of interest rates in the U.S., with denial rates increasing sharply in 2005 compared to 2004, this is not the case for the subprime market, where denial rates do not increase in 2005 compared to 2004.” Baily et al. (2008) observe “a deterioration in lending standards generally dated to 2004 or 2005.” Haughwaut et al. (2008) observe, “beginning with the 2005 vintage the performance of nonprime mortgage loans became notably worse than previous vintages. The performance of the 2006 vintage deteriorated even further. By 12 months following origination, the 2005 vintage had a 90 day or more delinquency rate that was not reached by the 2003 vintage for 20 months...”

Cagan (2006) shows the cumulative home equity for mortgages by year of origination. Of all mortgages originated in 2003, 13.1% mortgages homeowner equity in the home under 5%. By 2004, 17.6% of homeowners had equity below 5%; in 2005 the number jumped to 38.1% of all mortgages.

Other than falling house prices, securitization and easy availability of credit as sources for looser lending standards, we investigate for an exogenous shock that could have resulted in a significant deterioration in lending standards in 2005. Our paper examines the Bankruptcy Abuse Prevention and Consumer Protection Act (BAPCPA) of 2005 as a possible exogenous cause that contributed to the loosening of lending standards in the subprime mortgage market.

Prior to BAPCPA, any household in financial distress could file for a Chapter 7 Bankruptcy and obtain discharge of unsecured debt, thus freeing up income to save their home from foreclosure, a fact noted by Berkowitz and Hynes (1998), Lin and White (2001) and Jacoby (2007). BAPCPA introduced a means test by which better off filers are compelled to file for Chapter 13, and consequently cannot obtain complete relief from their unsecured debt. This implies that better off filers were more likely to default on their mortgages after BAPCPA. As noted by Morgan et al. (2008), BAPCPA effectively transferred credit default risks from unsecured lenders to secured (mortgage) lenders.


90 •

Using HMDA data for mortgages originated in 2000-2007, this increase in expected mortgage default resulted in the reducing the supply of credit to better off borrowers, i.e. the borrowers af-fected by BAPCPA. Our results suggest that BAPCPA led to contraction in the supply of mortgage credit for borrowers with incomes above their state’s median income- both in terms of lower mortgage application acceptance rates as well as higher spreads on mortgages. Our results also show a change in the supply of mortgage credit to borrowers with incomes below their state’s median income, in terms of higher application acceptance rates, but we also find evidence of low income borrowers being charged higher spreads than before. We speculate that low income bor-rowers represented ‘latent demand’ as per Mian and Sufi (2008), and were induced to absorb the credit at higher spreads through innovative mortgage products and high loan to value ratios.

This paper is not the first to link BAPCPA to the subprime mortgage crisis. Focusing on foreclo-sures, Morgan et al. (2008) observe that by preventing better off filers from getting relief from unsecured debt under Chapter 7, BAPCPA led to a substantial increase in subprime foreclosures — over 32,000 more subprime foreclosures nationwide per quarter. Our paper analyses the ef-fect of BAPCPA at the time of mortgage origination. Indeed, our findings suggest that credit was transferred to low income borrowers, who were more likely to default on their mortgages especially when home prices stopped rising. Moreover, it is likely that the increase in interest rates charged also had the effect of raising the net likelihood of mortgage default.

The rest of this paper is organized as follows: Section II provides a brief outline of bankruptcy law provisions and existing literature on the effect of bankruptcy laws on secured credit markets, Section III provides an analytical framework to predict the effect of BAPCPA on mortgage lenders, Section IV describes the dataset and introduces the econometric model, Section V reports the empirical results and Section VI concludes.

2. Bankruptcy Law in the United States: A Brief Overview

Historically, bankruptcy laws in the U.S. have been the most pro-debtor laws amongst developed nations. Bankruptcy filings in the U.S. steadily climbed from about 300,000 in 1980 to over 1.5 million in 2004, as personal bankruptcy filings increased from 1.4 per thousand of the working age population in 1970 to 8.5 in 2002 (Livshits et al. 2007). Figure 3 shows the annual non-business bankruptcy filings in the United States during 1980-2004.

As per U.S. bankruptcy provisions, an individual bankruptcy is filed under Chapter 7 or Chapter 13, while Chapter 11 is primarily used by corporations. In the years leading up to the bankruptcy law reform in 2005, around 71% of non-business bankruptcy filings were filed under Chapter 7.

Chapter 7 vs. Chapter 13

Under Chapter 7, the debtor’s non-exempt assets are liquidated and the proceeds are sold by the bankruptcy trustee to pay the debtor’s unsecured creditors. The debtor’s future income is not used to repay creditors under Chapter 7 and the debtor is allowed to retain certain assets as exempt from liquidation, depending on state or federal exemption limits, as applicable. If the debtor’s equity in the asset is less than the applicable exemption limit, the debtor retains the as-set, and the creditors do not receive anything. As of mid 2002, about 96% of Chapter 7 cases were closed without any funds distributed to creditors (Flynn et al., 2002). Under Chapter 13, a debtor’s assets are protected from liquidation; however the debtor enters into a repayment plan




to pay all or part of unsecured debt over a period of three to five years. Moreover, creditors have to receive at least as much in Chapter 13 as they would under Chapter 7. Thus the debtor cannot obtain discharge of unsecured debt under Chapter 13.

Chapter 7 as a Strategy to Save the Home from Foreclosure

For a debtor with both unsecured and mortgage debt, the ability to file for Chapter 7 has pro-vided a significant ‘escape route’ in financial distress. A debtor with both unsecured debt and a mortgage may not find it possible to borrow additional funds to tide over periods of sudden increases in expenditure or loss of income. If the debtor would find it difficult to meet monthly payments both the unsecured debt and the mortgage, filing for Chapter 7 would enable the debtor to get complete relief from unsecured debt, freeing up money to stay current on mort-gage payments and thereby save the home from foreclosure.

The use of a Chapter 7 bankruptcy as a way to save a home from foreclosure has been discussed widely. Jacoby (2007) explains, “Chapter 7, the more frequently utilized bankruptcy option for individuals, is also relevant to housing policy even though its primary function-the discharge-relates to unsecured debt. Consider a borrower who has not (yet) defaulted on a mortgage but is having serious financial trouble, or a borrower who has defaulted on a mortgage but has reached an agreement with her mortgagee. These borrowers might file chapter 7 to discharge unsecured debt, leaving them with more available income to make their mortgage payments and less likely to have their homes encumbered by judgment liens.” Berkowitz and Hynes (1999) observed that the benefit of using Chapter 7 to save a home from foreclosure was higher if the debtor’s state allowed for high or ‘unlimited’ exemptions for homestead equity and personal property. Berkow-itz and Hynes found that for states with high equity exemptions, mortgage supply was relatively greater — in terms of both lower interest rates and higher probability of a mortgage application being accepted.

The Bankruptcy Abuse Prevention and Consumer Protection Act (BAPCPA) was passed by Con-gress and signed into law on April 20, 2005. BAPCPA marked the first substantial overhaul of the Bankruptcy code in the United States since the Bankruptcy Reform Act of 1978, and was passed after a considerable lobbying effort by credit card companies as observed by White (2008). BAP-CPA sought to address the escalating bankruptcy filings and resultant unsecured creditors’ losses. It sought to make filing for bankruptcy more expensive in order to serve as deterrent and repeat bankruptcy filers. Among the various changes in bankruptcy law provisions as per BAPCPA, the introduction of a new means test to determine eligibility to file for Chapter 7 is important for our study.

Means Test for Eligibility to File Chapter 7

Post October 2005, debtors are no longer able to choose to file for bankruptcy under Chapter 7; eligibility to file for Chapter 7 is now established based on the outcomes of a ‘means test’. As per the first stage of the means test, a debtor whose income is less than the median income for the state of residence will be eligible to file for Chapter 7 bankruptcy. Debtors whose income is above the state’s median income are subject to the second stage of the means test; if the debtor’s income less certain allowable expenditures is below $167 per month, the debtor qualifies for Chapter 7. If the bankruptcy petitioner ‘fails’ either stage of the means test, he or she is com-pelled to file for bankruptcy under Chapter 13.


92 •

In 2004, non business bankruptcy filings were about 1.5 million. In 2005 they rose to over 2 million, as people ‘rushed to file’ for bankruptcy before the BAPCPA came into effect in 2005. Consequently, in 2006 there was an abrupt fall in filings to 0.5 million. In 2007 filings climbed back up to 0.8 million and were about 1 million in 2008, indicating that the BAPCPA produced a downward shift in the trend in total bankruptcy filings in the United States.

BAPCPA took away the option of filing for Chapter 7 for those who failed the means test, i.e. people with income higher than their state’s median income. While debtors who failed the means test could no longer use Chapter 7 as an escape route to save their homes from foreclosure, because they could no longer get rid of their unsecured debt, to free up income to continue repaying mortgages. Hence, post BAPCPA, ‘better off’ borrowers were more likely to default on their mortgages instead.

3. Hypothesis BAPCPA and the Effect on the Demand and Supply of Mortgage Credit

If BAPCPA raised the probability of mortgage default for those applicants who failed the means test, we can expect to witness a contraction in the supply of credit to such borrowers. We argue that while we expect a supply side effect, we do not expect the BAPCPA to produce a simultane-ous demand shift. As is observed by White (2007), since mortgage payments are ‘deductible’ ex-penses in calculating the bankruptcy petitioners’ disposable income for the means test, it may be that an individual filing for bankruptcy would find it beneficial to get a new mortgage simply to pass the means test. However, such ‘bankruptcy planning’ will only affect an individuals’ demand for a mortgage at the point of filing for bankruptcy, and we have no reason to believe that BAP-CPA in any way raised actual filing rates; in fact, BAPCPA reduced overall bankruptcy filings by raising filing costs. We stress that the new means test raised the probability of mortgage default given bankruptcy post-2005, which is of concern to the mortgage lender. We expect a change in the supply of credit to be apparent in two ways- in terms of a change in ‘quantity’ and change in ‘price’, given demand does not change.

We examine whether the probability of an individuals’ mortgage application being accepted 1. (quantity of mortgages supplied) is lower post-BAPCPA, for borrowers who would be ex-pected to ‘fail’ the means test in the event of bankruptcy. We examine whether borrowers who fail the means test are charged a higher interest post- 2. BAPCPA.

With regard to the borrowers who would be still eligible for Chapter 7, we do not hypothesize a direct effect of BAPCPA on credit supply. However, we believe that if supply of mortgage credit to above-median income borrowers was reduced, below-median income borrowers would receive more mortgage credit than before; however, we speculate that as the underly-ing risk characteristics of the below-median borrowers would be unchanged, lenders would compensate for the risks of any such increase in supply of credit to low income borrowers by charging higher interest rates. Figures 4 and 5 provide a graphical depiction of our predicted outcomes in the mortgage market.




4.

Empirical Methodology

In this section we test for any effect of BAPCPA on mortgage credit allocation. We believe that any general change in the supply of credit by mortgage lenders would be sufficiently captured by examining the effects of the first stage of the BAPCPA means test. The outcome of the second stage of the means test, which involves the process of deduction of certain expenditures from the debtor’s monthly income, would vary largely depending on a debtor’s specific economic situation. The bias in examining our hypothesis solely on the outcome of the first stage of the means test works against us, as those who fail the first stage of the test will be a subset of all those who actually fail the test.

Description of the Data

Our primary source of Data is the HMDA Loan Application Registry from the years 2000 through 2007. The dataset includes individual level economic and demographic information about mort-gage applicants that mortgage lenders are required to collect under the Home Mortgage Disclo-sure Act. From this dataset, we removed applications with missing data for dates of application or action, income or loan amount (applications for manufactured homes drop out at this stage because they do not report income). We then eliminated applications with loan amounts smaller than $1,000, as loan amounts are expressed in terms of thousands of dollars and such applica-tions show up in the dataset with a loan amount of 0. We also removed applications where the action was the loan being purchased, following Dell’Aricca et al. (2008) who note that such applications are reported by both the originating and purchasing institution and including them would amount to double-counting. Applications that are classified as federally insured are also eliminated because of their distinct risk profile. Finally, we removed applications for Home Im-provement or Refinancing, as these are not relevant to our hypothesis.

In addition to this application level data, we also used state-level data on economic indicators from various sources. This includes quarterly state unemployment data from the Bureau of Labor Statistics (BLS), quarterly state per capita income from the Bureau of Economic Analysis (BEA) and quarterly state Housing Price Indices from the Office of Federal Housing Enterprise Oversight (OFHEO) and annual data on the number of bankruptcy filings by state from the American Bank-ruptcy Institute.

The estimates of median state income used in U.S. individual bankruptcy proceedings for means testing are published by the U.S. Census Bureau, post October 2005. These median estimates vary as per the size of the bankruptcy petitioner’s family. The HMDA dataset does not contain information on the loan applicant’s family size. We approximate for this lack of information by using the median four- person family income by state as published by the Census Bureau, in order to apply the means test. The drawback of this approximation is that there may be several mortgage applicants of family size less than four, who will pass the means test with our generous threshold, but would fail it in a bankruptcy court. However, if this approximation results in biased estimates, again the bias works against us, as our hypothesis is concerned with people who fail the approximated means test, which is a subset of the people who would fail actual means test. For dates prior to October 2005 we again use 4 person family median income data from the Census Bureau.


94 •

Table 1 describes the data descriptions and sources used in the empirical estimation, while Table 2 gives the summary statistics for the variables.

Test 1

We investigate the effect of BAPCPA on mortgage originations by analyzing the probability of a mortgage application being accepted with the following logit specification:

probappi = si + ti + qi + β1*Ii + β2*Ei + β3*reld2 + β4*reld3 + β5*reld4 + β6*reld5 + β7*bar2dummy + β8*reld2bar2 + β9*reld3bar2 + β10*reld4bar2 + β11*reld5bar2 + εi

The dependent variable probapp is a binary variable that takes the value 1 if the mortgage ap-plication is approved and the value 0 if the application is denied.

To incorporate the means test in our model, we divide the loan application observations into 5 groups according to the relationship between the applicant’s income and the median income of the state where the property was located. We create 4 dummy variables to represent the classification of observations in the 5 groups. We refer to the control group as reld1 (for which there is no dummy variable) comprising those applications for which the applicant’s income is lower than 0.5 times the state median. The dummy variable reld2 takes the value 1 for the applicants whose income is between 0.5 times and 1 time the state median, the dummy va-riable reld3 takes the value 1 if the applicant’s income is between 1 and 1.25 times the state median, the dummy variable reld4 takes the value 1 if the applicant’s income is between 1.25 and 2 times the state median and the dummy reld5 takes the value 1 if the applicant’s income greater than 2 times the state’s median income. Thus the dummies reld3, reld4 and reld5 comprise those ‘above state median income’ observations that will serve as a rough estimate of applicants who will ‘fail’ the means test. We introduce a dummy variable bar2dummy for the bankruptcy reform which takes the value 1 if the loan application date was after BAPCPA was passed, i.e. after April 20, 2005, the value 0 otherwise. We introduce interaction terms reld2bar2, reld3bar2, reld4bar2 and reld5bar2 which are the product of bar2dummy and each reld* dummy variable.

The vector Ii includes individual characteristics at the loan level that have been shown to be good predictors of lending decisions: income denotes the income of the applicant, income2 denotes squared income and loanamt denotes the loan amount of the mortgage. To account for race and sex of the applicant, we include blackdummy, a dummy variable that takes the value 1 if the ap-plicant is black, the value 0 otherwise and femdummy is a dummy variable that takes the value 1 if the applicant is female, the value 0 otherwise.

Controlling for changes in the economic environment at the state level that may affect lenders’ behaviour, the vector Ei includes the variable stunemp, which is the unemployment rate of the applicant’s state of residence and stpci which is the state per capita income. To control for home price appreciation, we include stHPIch, the change in the home prices for the state where the property is located.

In addition, we control for time invariant state specific factors and for time variant national fac-tors, by including state and time fixed effects. Therefore, we are allowing for systematic differ-ences in credit acceptance in addition to those related to the regressors. Because states are likely to experience other differential influences, we estimate a fixed-effects model which allows the




intercept to vary across states. We also allow intercepts to vary over time and quarters. To correct for within state and time dependence we use robust standard errors estimations.

We augment the base model one at a time, with the variables loan to income ratio (lti), the number of bankruptcy filings in the previous year for the applicant’s state (bfl) and the quarterly growth rates in the volume of loan applications for each income group (appgrowth), that cap-tures any effect of changes in the application volumes across income groups that may influence the lender’s decision. While we use April 20, 2005 in our model as the day BAPCPA was signed into law, we also estimate the model using October 17, 2005 as the date when BAPCPA came into effect.

For loan applications received before April 20, 2005, the coefficients on each of the reld* dummy variables will indicate any variation in the probability of an individual’s loan application being ac-cepted relative to the control group reld1. We expect positive coefficients for each of the reld* variables, as we expect applicants from the reld* groups to have a higher chance of securing a mortgage, as they have higher incomes relative to the control group.

The coefficient on bar2dummy indicates the estimated change in the probability of application acceptance for the control group reld1 due to the bankruptcy reform. If our hypothesis implies that credit was diverted from higher income to lower income borrowers post BAPCPA, we expect a positive coefficient on bar2dummy.

The coefficients on reld2bar2, reld3bar2, reld4bar2 and reld5bar2 represent the difference in the probability of application acceptance due to BAPCPA for each group relative to the change in probability for the control group reld1. Since we are interested in looking at whether the ap-plication acceptance probability for each reld* group has changed specifically post- BAPCPA, we need to account for change across groups and across time implied in each reld*bar2 coefficient. Therefore rather that looking at the signs of the reld*bar2 dummy variables in isolation, we are interested in the sum of the coefficients on the dummy variables bar2dummy the relevant reld*bar2. If BAPCPA resulted in reduction in the supply of mortgage credit to the borrowers who were expected to fail the means test, we expect the sum of the coefficients of bar2dummy and the relevant reld*bar2 to be negative for reld3, reld4 and reld5. Consequently, for group reld2, which comprises below median income borrowers, we would expect the sum of the coefficients of bar2dummy reld2bar2 to be positive.

Test 2

As we stated in our hypothesis we need to analyze a change in mortgage lenders’ behaviour in not only in terms of a change in ‘quantity’ but also a change in ‘price’. For this purpose, we run an OLS regression for the rate spread of mortgage applications:

spreadi = si + ti + qi + β1*Ii + β2*Ei + β3*reld2 + β4*reld3 + β5*reld4 + β6*reld5 + β7*bar2dummy + β8*reld2bar2 + β9*reld3bar2 + β10*reld4bar2 + β11*reld5bar2 + εi

The dependent variable spread is the difference between the annual percentage rate (APR) on the originated loan and the rate on Treasury securities of comparable maturity. The estimation period in this case is 2004-2007, since there is no information about spreads before 2004. Moreover, the dataset only includes loans having spreads in excess of 3%. The variables in the spread regres-sion are the same as included in the logit regression. We however add a new variable to account for the applicant’s ethnicity, which is recorded in the HMDA dataset post 2004; latdummy is a


96 •

dummy variable that takes the value 1 if the applicant ethnicity is Hispanic, 0 otherwise. We did not use this variable in Test 1, as data on ethnicity is not reported prior to 2004. We estimate alternative models by adding the variables on loan to income ratio, bankruptcy filings and group-wise application growth rates. We use robust standard errors to correct for within state and time dependence.

The coefficients for the dummy variables reld2, reld3, reld4 and reld5, reflect the change in the spread for each reld* group relative to the spread for the control group before BAPCPA. Here, we expect the coefficients to be negative as higher income borrowers are likely to be charged lower spreads. If we expect a contraction in the supply of credit to above median income bor-rowers, post-BAPCPA, we expect higher mortgage spreads fo these borrowers. Thus, we expect that the sum of the coefficients on the relevant reld*bar2 dummy variable and the bar2dummy dummy variable to be positive in the case of reld3, reld4 and reld5. Since the underlying risk characteristics of the low income borrower pool remain unchanged post-BAPCPA, any increase in higher mortgage application acceptance rates implied by our hypothesis would be expected to be accompanied by an an increase the interest rate spread. Thus we would expect the sum of the coefficients of reld2bar2, and bar2dummy to be positive.

5. Empirical Results

Table 3 reports the results of the logit model for the probability of acceptance for a mortgage application. Model 1 is the base model. Model 2 includes information about the loan to income ratio lti. Model 3 adds the lagged bankruptcy filings bft for the state where application was received. Model 4 includes these last two variables and also controls for the demand effect by including appgrowth. Table 8 is estimated using October 17, 2005 as the date for BAPCPA,

The coefficients on reld2, reld3, reld4 and reld5 are also significant with the expected positive sign, meaning that before BAPCPA, applicants in any of the four income bands had a higher probability of acceptance than those applicants in the lower income group, reld1. The coefficient on bar2dummy is significant and positive in all four models, implying that for those applicants in the control group reld1, the bankruptcy reform increased the probability of applications being accepted after controlling for all other factors. Also, the sum of the coefficients for bar2dummy and the reld*bar2 variables, (as reported in Table 4 with corrected standard errors), is positive and significant for group reld2 and negative and significant for reld3, reld4 and reld5. This confirms our hypothesis, that BAPCPA led to an increase in the probability of a loan application acceptance for below-median income borrowers and a decrease in the application acceptance probability for above-median income borrowers, after controlling for other factors.

The results also show that the probability of acceptance decreases for black people, while the probability of acceptance increases for female applicants. Contrary to our expectations however, Income has a negative and significant coefficient; moreover, the loan to income variable lti is significantly positive in models 2 to 4, both contrary to what would be expected. However, we be-lieve this could be indicative of the change in lenders’ reaction to income after BAPCPA in 2005. Of the economic controls, house prices have an expected positive significant coefficient – an in-crease in house prices raises the probability of application acceptance. The bankruptcy filing vari-able, bfl, is positive and significant. The variable appgrowth is significant and positive; suggesting that growth in volumes of applications received by lenders affects the mortgage supply decision. State per capita income has a negative impact on acceptance, while the state unemployment rate appears to have positive impact on the acceptance.




Table 5 shows the results on the regression model for the rate spread. The coefficients on reld2, reld3, reld4 and reld5 are in this case negative and highly significant, implying that before BAP-CPA, applicants in any of the four (higher) income bands were able to get a lower mortgage rate than those applicants in the lower income group, reld1. The coefficient on bar2dummy is sig-nificant and positive in all four models, implying that for those applicants in the subgroup reld1, the bankruptcy reform increased the rate spread that they had to pay on their loans. This finding confirms our hypothesis of a shift along the supply curve: when the underlying risk characteristics of the below-median borrowers are unchanged, an increase in the quantity supplied must be accompanied by higher interest rates. The four interaction terms are also significant and have a positive sign, meaning that increase in the spread after BAPCPA was higher for applicants from the groups reld2, reld3, reld4 and reld5, relative to those applicants in group reld1. We add the estimated coefficients for bar2dummy and the estimated coefficients of the interaction terms (reld2bar2 to reld5bar2) and we report the coefficients and corrected standard errors in Table 6. All are positive and significant, confirming our hypothesis that BAPCPA had an effect in mortgage lenders’ behaviour not only in terms of a change in ‘quantity’ but also in a change in ‘price’ for above median income borrowers. The four models on the rate spread show that the mortgage spreads are higher for black, female and Hispanic applicants. The coefficient for home prices is negative and significant, suggesting that rising home prices result in smaller spreads. Appgrowth has a negative impact on the spread. Income variables are not significant in any of the four mod-els. The coefficient for stpci is positive and the coefficient for state unemployment has a negative coefficient, contrary to expectations while the bfl variable is not significant.

6. Conclusion

This paper sheds light on the effect of a change in bankruptcy laws on the secured mortgage market. Specifically, we show that the means test provision of the Bankruptcy Abuse Prevention and Consumer Protection Act caused a reduction in the supply of credit to relatively higher-income borrowers, through a decline in loan application acceptance rates and higher spreads. Further, we find that borrowers with income lower than their state’s median income saw an increase in the supply of mortgage credit, accompanied with an increase in interest rates. We speculate that low income borrowers absorbed the increas in supply of credit at higher spreads through incentives like low loan to value ratio requirements, innovations like hybrid adjustable rate mortgages with ‘teaser rates’, and expectations of future house price appreciation. The transfer of credit from relatively better off borrowers to poorer borrowers due to the ‘exogenous shock’ of BAPCPA appears to be a contribution factor to the deterioration of lending standards in the U.S. mortgage market.


98 •

7. Appendix

Figure 1- Annual Bankruptcy Filings in the United States, 1980-2004.

Source: American Bankruptcy Institute

Figure 2- Inward shift in the supply of mortgage credit for borrowers with income above state median.




Figure 3- Shift along the supply curve for borrowers with income less that the state’s median income.


100 •

Table 1- Data Description and Sources

Description

Dummy variable taking the value 1 if the loan is originated and 0 otherwiseSpread between mortgage rate and Treasury security of comparable maturity (only reported if the spread is greater than 3%).Dummy variable taking the value 1 if the applicant is black and 0 otherwiseDummy variable taking the value 1 if the applicant is of Hispanic ethnicity and 0 otherwiseDummy variable taking the value 1 if the applicant is female and 0 otherwisePrincipal amount of the application (in thousands of dollars)

Applicant’s income (in thousands of dollars)

State unemployment rate

State per capita income

Change in state HPI since last period

Bankruptcy filings by state, lagged one period

Growth in applications in the period of the action since the last period. Not defined for 2000 Q1 as application data for the preceding quarter is incomplete.

Dummy variable taking the value 1 if the applicant’s income in the quarter of the action is less than half his state’s median income and 0 otherwise.

Dummy variable taking the value 1 if the applicant’s income in the quarter of the action is between his state’s median income and half his state’s median income and 0 otherwise.

Dummy variable taking the value 1 if the applicant’s income in the quarter of the action is between his state’s median income and 1.25 time his state’s median income and 0 otherwise.

Dummy variable taking the value 1 if the applicant’s income in the quarter of the action is between 1.25 times and twice his state’s median income, and 0 otherwise.

Dummy variable taking the value 1 if the applicant’s income in the quarter of the action is more than twice his state’s median income, and 0 otherwise.

Dummy variable taking the value 1 if the action for the application occurred after 20th April 2005, and 0 otherwise

Variable

Probapp

Spread

Blackdummy

Latdummy

Femdummy

Loanamt

Income

Stunemp

Stpci

stHPIch

Bfl

Appgrowth

Reld1

Reld2

Reld3

Reld4

Reld5

Bar2dummy

Type

Loan level

Loan level

Loan level

Loan Level

Loan Level

Loan Level

Loan Level

Quarterly

Quarterly

Quarterly

Annual

Quarterly

Loan Level

Loan Level

Loan Level

Loan Level

Loan Level

Loan Level

Source

HMDA

HMDA

HMDA

HMDA

HMDA

HMDA

HMDA

BLS

BEA

OFHEO

ABI

HMDA

MDA, Census Bureau

HMDA, Census Bureau

HMDA, Census Bureau

HMDA, Census Bureau

HMDA, Census Bureau

HMDA




Table 2- Summary Statistics

2000 2001 2002 2003 2004 2005 2006 2007

Variable Means

probapp 0.58 0.63 0.70 0.70 0.66 0.63 0.63 0.62

spread - - - - 4.97 5.36 5.77 4.98

blackdummy 0.07 0.06 0.05 0.07 0.08 0.10 0.11 0.09

latdummy - - - - 0.12 0.15 0.16 0.12

femdummy 0.24 0.23 0.24 0.27 0.29 0.30 0.31 0.29

loanamt

[‘000) 117.71 127.99 152.24 166.82 175.92 181.47 188.14 209.63

income

[‘000) 75.87 80.75 90.11 93.84 95.95 102.01 112.94 118.64

effmedian 49899.2 51289.5 52927.2 53371.2 54882.6 64759.8 66725.9 69355.2

stunemp 4.04 4.78 5.86 5.99 5.49 5.04 4.59 4.60

stpci 25297.6 26181.5 27369.8 28209 29801.4 30781.9 32329.9 33633.1

stHPIch 0.04 0.05 0.05 0.06 0.11 0.17 0.06 -0.01

lti 1.83 1.89 2.04 2.18 2.26 2.15 2.06 2.27

bfl 51272.4 50216.39 60982.49 62609.33 65811.31 60938.48 79320.34 20694.33

appgrowth -9.12 0.83 -0.37 5.60 10.47 5.93 -1.67 -17.34

Actions 62923 58034 48214 55524 84230 102521 104679 64701


102 •

Table 3- Determinants of loan application approval

VARIABLES Probapp (1) Probapp (2) Probapp (3) Probapp (4)

blackdummy -0.613*** -0.612*** -0.614*** -0.614*** [0.0101] [0.0101] [0.0101] [0.0102]femdummy 0.0174*** 0.0175*** 0.0179*** 0.0183*** [0.00640] [0.00640] [0.00640] [0.00648]loanamt 8.58e-05*** 1.15e-05 1.93e-05 2.49e-05 [2.12e-05] [3.21e-05] [3.21e-05] [2.96e-05]income -0.000159*** -9.65e-05** -9.86e-05** -0.000101** [3.72e-05] [4.27e-05] [4.26e-05] [4.15e-05]income2 8.19e-09 2.13e-09 2.30e-09 2.48e-09 [5.66e-09] [5.95e-09] [5.95e-09] [5.88e-09]stunemp 0.0177*** 0.0178*** 0.0162*** 0.0172*** [0.00629] [0.00629] [0.00629] [0.00645]stpci -7.39e-05*** -7.38e-05*** -6.57e-05*** -6.18e-05*** [4.63e-06] [4.63e-06] [4.68e-06] [4.79e-06]stHPIch 0.361*** 0.357*** 0.295*** 0.367*** [0.0523] [0.0523] [0.0525] [0.0528]reld2 0.691*** 0.704*** 0.702*** 0.693*** [0.0142] [0.0147] [0.0147] [0.0152]reld3 1.027*** 1.044*** 1.041*** 1.020*** [0.0163] [0.0171] [0.0171] [0.0175]reld4 1.250*** 1.271*** 1.267*** 1.240*** [0.0151] [0.0164] [0.0164] [0.0166]reld5 1.350*** 1.377*** 1.370*** 1.340*** [0.0167] [0.0188] [0.0188] [0.0188]bar2dummy 0.275*** 0.272*** 0.267*** 0.275*** [0.0249] [0.0250] [0.0250] [0.0252]reld2bar2 -0.209*** -0.206*** -0.203*** -0.193*** [0.0215] [0.0215] [0.0215] [0.0219]reld3bar2 -0.383*** -0.379*** -0.371*** -0.348*** [0.0244] [0.0245] [0.0245] [0.0248]reld4bar2 -0.524*** -0.520*** -0.510*** -0.486*** [0.0223] [0.0224] [0.0224] [0.0227]reld5bar2 -0.614*** -0.609*** -0.595*** -0.574*** [0.0230] [0.0231] [0.0231] [0.0234]lti 0.0122*** 0.0121*** 0.00954*** [0.00424] [0.00422] [0.00365]bfl 2.39e-06*** 2.19e-06*** [2.07e-07] [2.13e-07]appgrowth 0.00348*** [0.000347]Constant 2.414*** 2.374*** 2.151*** 2.163*** [0.198] [0.198] [0.199] [0.204]

Observations 580826 580826 580826 565778Pseudo R square 0.0342 0.0343 0.0345 0.0323

Correctly classified 65.61% 65.61% 65.65% 65.67%

Note: Logistic regression on dummy variable taking value 1 if the loan was originated and 0 otherwise. All regressions include state, time and seasonal fixed effects (not reported). For variable definitions, see Table 1. Robust standard errors are in brackets. * denotes significance at 10%; **significance at 5%; *** significance at 1%. Model 4 has fewer observations that the others as app growth is not defined applications with action taken in 2000 Q1.




Table 4- Net Effect of BAPCPA for each income group

VARIABLES (1) (2) (3) (4)

probapp probapp probapp probapp

reld2bar2 0.066*** 0.066*** 0.064*** 0.082***

[0.0187] [0.0187] [0.0187] [0.0188]

- - - -

reld3bar2 0.108*** 0.107*** 0.104*** 0.073***

[0.0220] [0.0220] [0.0220] [0.0221]

- - - -

reld4bar2 0.249*** 0.248*** 0.243*** 0.221***

[0.0196] [0.0197] [0.0197] [0.0198]

- - - -

reld5bar2 0.339*** 0.337*** 0328*** 0.299***

[0.0205] [0.0205] [0.0205] [0.0206]

Note: Models 1, 2, 3 and 4 correspond to the same models in Table 3. Corrected robust standard errors are in brackets. *denotes significance at 10%; **significance at 5%; *** significance at 1%.


104 •

Table 5- Determinants of Spreads

(1) (2) (3) (4)VARIABLES spread spread spread spread

blackdummy 0.328*** 0.345*** 0.345*** 0.344*** [0.0212] [0.0213] [0.0213] [0.0213]latdummy 0.134*** 0.150*** 0.150*** 0.149*** [0.0155] [0.0158] [0.0158] [0.0158]femdummy 0.0582*** 0.0620*** 0.0621*** 0.0617*** [0.0146] [0.0145] [0.0145] [0.0145] - - - -loanamt 0.00462*** 0.00267*** 0.00267*** 0.00268*** [0.000138] [0.000384] [0.000384] [0.000384]income 0.000863*** -7.95e-05 -7.96e-05 -8.79e-05 [0.000195] [0.000216] [0.000216] [0.000216]income2 -5.47e-08 3.60e-08 3.61e-08 3.74e-08 [4.44e-08] [3.47e-08] [3.47e-08] [3.46e-08]stunemp -0.0486** -0.0513*** -0.0528*** -0.0440** [0.0200] [0.0196] [0.0196] [0.0196]stpci 3.67e-05** 3.63e-05** 3.63e-05** 3.54e-05** [1.46e-05] [1.44e-05] [1.44e-05] [1.44e-05]stHPIch -0.662*** -0.555*** -0.544*** -0.590*** [0.108] [0.110] [0.109] [0.110]reld2 -0.0634 -0.258*** -0.258*** -0.329*** [0.0581] [0.0671] [0.0671] [0.0680]reld3 0.0163 -0.274*** -0.274*** -0.374*** [0.0624] [0.0811] [0.0811] [0.0825]reld4 0.0766 -0.276*** -0.276*** -0.399*** [0.0617] [0.0888] [0.0888] [0.0906]reld5 0.224*** -0.225** -0.225** -0.358*** [0.0748] [0.110] [0.110] [0.111]bar2dummy 0.397*** 0.410*** 0.410*** 0.309*** [0.0707] [0.0712] [0.0712] [0.0734]reld2bar2 0.203*** 0.193*** 0.192*** 0.261*** [0.0717] [0.0721] [0.0721] [0.0733]reld3bar2 0.232*** 0.219*** 0.218*** 0.316*** [0.0769] [0.0771] [0.0772] [0.0794]reld4bar2 0.250*** 0.221*** 0.220*** 0.348*** [0.0728] [0.0735] [0.0735] [0.0774]reld5bar2 0.218*** 0.145* 0.144* 0.294*** [0.0783] [0.0792] [0.0793] [0.0842]lti -0.240*** -0.240*** -0.240*** [0.0470] [0.0470] [0.0470]bfl -7.89e-07 -5.68e-07 [5.90e-07] [5.89e-07] -appgrowth 0.00581*** [0.000938]Constant 3.668*** 4.191*** 4.167*** 4.042*** [0.615] [0.616] [0.616] [0.616]

Observations 54211 54211 54211 54211R-squared 0.159 0.182 0.182 0.183

Note: OLS regression on Spreads over Treasury of comparable maturity for years 2004-2007. All regressions include state, time and seasonal fixed effects (not reported). For variable definitions, see Table 1. Robust stan-dard errors are in brackets. * denotes significance at 10%; **significance at 5%; *** significance at 1%.




Table 6- Corrected Coefficients for spreads from Table 5

spread spread spread spreadVARIABLES (1) (2) (3) (4)

reld2bar2 0.600*** 0.603*** 0.602*** 0.570***

[0.0337] [0.0326] [0.0327] [0.0330]

reld3bar2 0.629*** 0.629*** 0.628*** 0.625***

[0.0416] [0.0405] [0.0404] [0.0404]

reld4bar2 0.647*** 0.631*** 0.630*** 0.657***

[0.0362] [0.0355] [0.0356] [0.0361]

reld5bar2 0.615*** 0555*** 0.554*** 0.603***

[0.0466] [0.0469] [0.0470] [0.0480]

Note: Models 1, 2, 3 and 4 correspond to the same models in Table 5. Corrected robust standard errors are in brackets. * denotes significance at 10%; **significance at 5%; *** significance at 1%.


106 •

Table 7- Test 1 Results with the date of BAPCPA October 2005.

Variables Variables probapp(1) spread (1) blackdummy -0.613*** blackdummy 0.342*** (0.0102) (0.0213)latdummy N/A latdummy 0.149*** N/A (0.0159)femdummy 0.0175*** femdummy 0.0624*** (0.00648) (0.0146) -loanamt 1.93e-05 loanamt 0.00266*** (3.00e-05) (0.000384) -income 0.000116*** income -7.45e-05 (4.16e-05) (0.000216)income2 4.53e-09 income2 3.55e-08 (5.90e-09) (3.42e-08)stunemp 0.0185*** stunemp -0.0819*** (0.00643) (0.0195) -6.40e-stpci 05*** stpci 2.12e-05 (4.78e-06) (1.44e-05)stHPIch 0.248*** stHPIch -0.412*** (0.0531) (0.112)bardummy 0.217*** bardummy 0.279*** (0.0273) (0.0696)reld2bar -0.161*** reld2bar 0.0985 (0.0230) (0.0684)reld3bar -0.296*** reld3bar 0.218*** (0.0262) (0.0745)reld4bar -0.434*** reld4bar 0.121* (0.0239) (0.0699)reld5bar -0.514*** reld5bar 0.0267 (0.0245) (0.0732)lti 0.00989*** lti -0.241*** (0.00375) (0.0471)bfl 2.41e-06*** bfl -9.24e-07 (2.12e-07) (5.90e-07) -appgrowth 0.00223*** appgrowth 0.00408*** (0.000348) (0.000855)Constant 2.266*** Constant 4.793*** (0.203) (0.613)




8. References

Baily, Martin Neil, Litan, Robert E., and Johnson, Matthew S. 2008. “The Origins of the Financial Crisis”, The Brookings Institution Initiative on Business and Public Policy.

Berkowitz, J. and R. Hynes, 1999. “Bankruptcy exemptions and the market for mortgage loans.” Journal of Law and Economics, 42(2), 809–830.

Cagan, Christopher L. 2007. “Mortgage Payment Reset: The Rumor and the Reality.” First Ameri-can CoreLogic.

Dell’Ariccia, Giovanni, Igan, Deniz and Laeven, Luc A. 2008. “Credit Booms and Lending Stan-dards: Evidence from the Subprime Mortgage Market”, IMF Working Paper No 08/106

Flynn, Ed, Bermant, Gordon and Hazard, Suzanne. 2002. “Bankruptcy by the Numbers: Chapter 7 Asset Cases”, US Trustee Program Library of Bankruptcy Articles http://www.usdoj.gov/ust/eo/public_affairs/articles/docs/abi122002.htm

Gerardi, Kristopher S., Shapiro, Adam Hale and Willen, Paul. 2007. “Subprime Outcomes: Risky Mortgages, Homeownership Experiences, and Foreclosures”, Federal Reserve Bank of Boston Working Paper No. 07-15

Haughwout, Andrew, Peach, Richard and Tracy, Joseph. 2008. “Juvenile delinquent mortgages: Bad credit or bad economy?” Journal of Urban Economics, 64, 246–257

Jacoby, Melissa B. 2007. “Bankruptcy Reform and Homeownership Risk”, University of Illinois Law Review, 2007; UNC Legal Studies Research Paper No. 918006

Keys, Benjamin J., Mukherjee, Tanmoy K., Seru, Amit and Vig, Vikrant. (2008), “Did Securitization Lead to Lax Screening? Evidence from Subprime Loans”, EFA 2008 Athens Meetings Paper

Lin, Emily Y. and White, Michelle J. 2001. Bankruptcy and the Market for Mortgage and Home Improvement Loans. Journal of Urban Economics, 50, pp 138-162.

Livshits, I., MacGee, J. and Tertilt, M. 2003. “Consumer Bankruptcy: A Fresh Start” Federal Re-serve Bank of Minneapolis Working Paper No. 617.

Mayer, Christopher J., Pence, Karen M. and Sherlund, Shane M. 2008. “The Rise in Mortgage Defaults,” Finance and Economics Discussion Series 2008-59, Board of Governors of the Federal Reserve System

Mian, Atif R. and Sufi, Amir. 2008. “The Consequences of Mortgage Credit Expansion: Evidence from the U.S. Mortgage Default Crisis” Forthcoming, Quarterly Journal of Economics.

Morgan, Donald P., Iverson, Benjamin Charles and Botsch, Matthew. 2008. “Seismic Effects of the Bankruptcy Reform”, Federal Reserve of New York Staff Report No. 358

White, Michelle J. 2007. “Abuse or Protection: The Economics of Bankruptcy Reform under BAP-CPA”, University of Illinois Law Review, vol. 2007:1, pp. 275-304

White, Michelle J. 2008. “Bankruptcy: Past Puzzles, Recent Reforms and the Mortgage Crisis”, National Bureau of Economic Research Working Paper. 14549


108 •

Fiscal Imbalances, Inflation and Sovereign Default Dynamics

Ramiro Sosa Navarro1

Abstract

The central question this paper seeks to answer is how monetary policy might affect the equilib-rium behavior of default and sovereign risk premium. The paper is based on a “one-interest-rate” model. Public debt becomes risky due to an active fiscal policy, as in Uribe (2006), reflecting the fiscal authority’s limited ability to control primary surplus. The insolvency problem is due to a string of bad luck (negative shocks affecting primary surplus). But in contrast to Uribe’s results, as the sovereign debt cost increases (which result from weak primary surplus), default becomes anticipated and reflected by a rising country risk premium and default probability. The default is defined as reneging on a contractual agreement and so the decision is set by the fiscal authority. However, conflicting objectives between fiscal and monetary authority play an important role in leading fiscal authority to default on its liabilities. The characteristic of the government policy needed to restore the equilibrium after the default is also analyzed.

ResumenLa cuestión central que este artículo busca responder es como la política monetaria puede afectar el comportamiento de equilibrio de primas por riesgo soberano y cesación de pagos. El artículo se basa en el modelo de “una-tasa-interés”. La deuda pública se hace riesgosa a causa de una política fiscal activa, como en Uribe (2006), reflejando la habilidad limitada de la autoridad fiscal para controlar el superávit primario. El problema de insolvencia es debido a una oleada de mala suerte (shocks negativos que afectan el superávit primario). Pero en contraste a los resultados de Uribe, a medida que aumenta el costo de la deuda soberana (que resulta de un excedente primario débil), la cesación de pagos se anticipa y es reflejada por una creciente prima de riesgo en el país y una probabilidad de cesación de pagos. La cesación de pagos se define como un incumplimiento de un acuerdo contractual y por ende la decisión es tomada por la autoridad fiscal. Mientras tanto, objetivos conflictivos entre la autoridad monetaria y fiscal juegan un rol importante en llevar a la autoridad fiscal a la cesación de pagos sobre sus pasivos. La característica de la política del gobierno necesaria para restaurar el equilibrio después de la cesación de pagos también es analizada.

Keywords: Sovereign risk Premium, default, fiscal imbalances

JEL Classification: [E4] [E6] [E31] [E63] [G32] [H63]

1.- Licenciado en Economía y MA en Finanzas (UTDT). MA Macroeconomics and Econometrics and Ph.D Economics (Univ. Paris-Evry). Eiffel Excellence Scholarship granted by the Foreign Affairs Ministry of the Republic of France, [email protected]




1. Introduction

Interactions between fiscal and monetary policy in the determination of the price level have been the object of a great debate in monetary theory during years. Sargent and Wallace (1975) argue that if the monetary authorities adopt a policy rule for the interest rate (rather than the money stock) the equlibrium outcome leads to price level indeterminacy. However, the Sargent and Wal-lace result is not entirely general. McCallum (1981) firstly accounts for the following well-known result in the literature. Monetary policy feedback rules, linking the nominal interest rate to en-dogenous variables such as the price level, permit to rule out the classical problem of price level indeterminacy advocated by Sargent and Wallace. Following Taylor´s (1993) stimulating article, the so-called `Taylor rules´ have received growing attention in recent years. According to this type of rule, the central bank´s interest rate target is set as an increasing function of the inflation rate and the output gap.2 In order to rule out multiple equilibria, theoretical studies3 suggest that the monetary authority has to responds to increases in inflation with a more than one-to-one increase in the nominal interest rate. In terms of Leeper (1991), this monetary policy rule is known as an àctive´-otherwise, it is called `passive´.

However, this literature does not account for the fiscal policy behavior. It means that the `Ricard-ian equivalence´ proposition applies; then, a comprehensive study of the implications of govern-ment deficits (and public debt) over the link between interest rate rules and price stability is not possible. All the same, there are important implications to consider within the relation between the monetary and fiscal policies. In the recent macroeconomic debate, it is argued that the lack of sound fiscal policy undermines the objective of price stability.4

The seminal contribution of Leeper (1991) made also an important distinction between àctive´ and `passive´ fiscal policy. It defines a fiscal policy as áctive´ when taxes respond only weakly to public debt levels and ´passive´ ones when taxes respond strongly to debt levels.5 In a standard model the research showed that two combinations, either (i) active monetary and passive fiscal policy or (ii) active fiscal and passive monetary policy yield determinacy, a unique stationary ratio-nal expectations equilibrium. In case (i) the usual monetarist view that inflation depends only on monetary policy is confirmed. However, case (ii) is fiscalist in the sense that fiscal policy, in addi-tion to monetary policy, has an effect on the inflation rate. Leeper (1991) also showed that the steady state is indeterminate, with multiple stationary solutions, when both policies are passive, while the economy is explosive when both policies are active.

Thus, the so-called `fiscal Theory of the Price Level´ (FTPL), has emerged.6 This well-known theo-retical framework enables to capture the effects of fiscal policy on the dynamic behavior of nomi-nal variables, like price level.

The FTPL asserts that fiscal variables can fully determine the price level independently of monetary variables. More specifically, when fiscal solvency is not ensured for each sequence of the price level, fiscal variables uniquely determine the equilibrium level of nominal variables. This extreme

2.- See, among others, Clarida et al., 1998, 2000 which provide empirical evidence to the view that Taylor-type rules describe consistently the behavior of several central banks.3.- See for instance, Taylor 1999 and Woodford 20034.- The Maastricht Treaty and the Stability and Growth Pact in the European Union which set quantitative limits on fiscal deficits and public debt for the Member States is based on this argument.5.- Later on, Woodford (1995) will identify this type of policies as a non-Ricardian and Ricardian fiscal policy.6.- The main contributors are Woodford, 1994, 1995, 2001, Sims, 1994 and Cochrane, 1999.


110 •

result is the polar opposite of the monetarist statement that the price level and the inflation rate depend primarily on monetary variables. Not surprisingly, the Fiscal Theory approach has trig-gered critics and controversy.7

The controversy concerns the nature of the intertemporal budget constraint of the government. In different papers Buiter argues that FTPL confuses the roles of budget constraints and equilibrium conditions in models of a market economy. But more interesting, Buiter (2002) criticizes FTPL as a theory of price level determination because it explicitly rules out default. Equilibrium price-level changes each period in response to the (stochastic) fiscal shocks. And with price level changes in each period providing the capital gains and losses on public debt level necessary for equilibrium, default is never necessary. Once the possibility of explicit default is properly allowed for, non-Ricardian regimes become Ricardian regimes and the Fiscal Theory of the Price Level vanishes. Buiter shows that under a non-Ricardian fiscal-monetary programme with an exogenous nominal interest rate rule, the equilibrium conditions are the same as under the Ricardian fiscal-monetary programme without contract fulfillment and with an exogenous nominal interest rate rule.

Uribe (2006) presents a dynamic FTPL model of default in which he allows limited inflation rate flexibility. When a shock is so large that limited inflation rate flexibility cannot provide the neces-sary capital gain or loss on government debt, then the government either devalues or revalues its debt. Default is a reduction in debt below its contractual value. This is an interesting application of the FTPL to the problem of default, but it neither exhibits an increasing probability of default nor a positive expected default rate as empirical evidence suggests.

The main objective of this chapter is to analyze the price stability and sovereign default risk issue. The model is grounded on a micro-founded equilibrium model with infinitely lived private agents that allow deviations from the Ricardian equivalence. This framework is particularly suitable to study the interactions between monetary and fiscal policy and its effect over both price stability and sovereign risk premium. It is shown that active interest rate rules, overreacting to inflation, are neither necessary nor sufficient to guarantee a unique stable solution for the price level without defaulting. Furthermore, in some cases, even `passive´ interest rate rules might drive the economy to an unsustainable path without defaulting. These results suggest that monetary policy matters being able to worsen a given scenario. Then, sovereign default is required to restore fiscal solvency and price stability. But the default rate must be high enough to ensure that the economy reaches a stable equilibrium in the post-default dynamics.

The rest of the chapter is organized in seven sections. Section II presents the Model. Section III describes the three possible scenarios for this economy and section IV, the inflation and default dynamics. Section V explicitly calculates the expected recovery rate and sovereign risk premium. Section VI provides further Reseach showing how detailed specifications of the monetary rule affect the equilibrium dynamics. Finally, in section VII, the conclusion.

7.- It has been mainly questioned by Buiter (1999, 2001 and 2002) as well as McCallum (2001) and Niepelt (2004).




2. The Model

2.1 The Households

Consider a closed economy inhabited by a large number of identical infinitely-lived households. Preferences are described by,

(1)

where ct denotes household’s consumption level of a perishable good in period t, is the single-period utility function assumed to be increasing, strictly concave and continuously differen-tiable, denotes the subjective discount factor and Et is the mathematical expectation operator conditional on period .

Each period, households are assumed to have access to a one-period nominal government bond, denoted Bt. This bond offers, in period t + 1 a contractual gross nominal interest rate Rt. However, the fiscal authority may default on its debt and in each period it repays a fraction ht of its liabilities. Therefore, household investment in sovereign bonds in period t is given by Bt whereas the earnings from the last-period investment is expressed as ht Rt–1 Bt–1. This expression is called the recovery value of the sovereign debt whereas ht (0,1) represents the recovery rate.

In our notation, Buiter (1999, 2001 and 2002) do not restrain ht assuming that both ht < 0 and ht > 1 are possible options. The former assumption -- ht < 0 --implies that the sovereign can be a net creditor. This seems unrealistic --particularly, in developing countries-- and not so relevant in a model focused to analyse scenarios of sovereign debt crisis and default. The last assumption -- ht > 1 -- also adopted by Uribe (2006), implies that any surplus resources over the contractual value of the outstanding debt are shared out equally among the holders of the contractual government debt. However, this excess of resources should not be interpreted as a govern-ment subsidy because in general they are allocated to tax payers; not to bondholders. Buiter names these transfers `super-solvency premium’. But even more important, government bonds are fixed-income securities as opposite to any other variable return security, such as stocks. In a more realistic approach we propose constraint ht as (0,1).

Besides, in each period t households have also the opportunity to invest in a complete set of nominal state-contingent assets. The total investment, in nominal terms, can be expressed as EtQt,t+1Dt+1 where Qt,t+1 denotes the stochastic nominal discount factor of an asset with a ran-dom nominal payment Dt+1. The time revenue from the investment made in the previous period, is denoted as Dt.

Finally, households are endowed with a constant and exogenous amount of perishable goods denoted by y and they pay real lump-sum taxes . Their flow budget constraint can be written as,

(2)

where Pt denotes the price level in period t.


112 •

Then, the household is subject to an appropriate set of borrowing limits which prevents “Ponzi Games”. In the absence of financial market frictions, the borrowing constraint takes the form:

(3)

where and .

The representative household maximizes its lifetime utility (1) subject to its flow budget con-

straint (2) and to its borrowing limits (3) by choosing taking as given the set of

processes and the initial values D0 and B–1. In addition to equation (2)

holding with equality, the first order conditions are given by,

(4)

(5)

(6)

where denotes the Lagrangian multiplier in period t.

Equation (4) states that the marginal utility of consumption must equal the marginal utility of wealth, , for all time t. Equation (5) represents the standard pricing equation for each one-period forward nominal contingent asset and equation (6) represents the pricing equation for the case of the risky sovereign bonds between period t and t + 1.

The transversality condition for the financial assets is written:

(7)

2.2 The Monetary and Fiscal Authorities

The fiscal authority levies lump-sum taxes, Pt , which are assumed to follow an exogenous, sto-chastic process. Recalling that fiscal authority issues nominal bonds, Bt, with a contractual gross nominal interest rate, Rt, but may default on its outstanding debt and repays a fraction ht of its liabilities Rt–1 Bt–1 the sequential budget constraint8 is given by,

(8)

where Bt–1 Rt–1 is given in period t and the recovery rate satisfies ht ∈(0,1).

8.- For sake of simplicity, in this paper, we ignore money and seigniorage revenues.




2.2.1 The Monetary Rule

Following Uribe (2006), we suppose that the monetary policy takes the form of an interest-rate feedback rule whereby the short-term nominal interest rate is set as a function of inflation. But while Uribe uses a simple linear Taylor rule, active in the sense of Leeper (1991), and with an ex-plicit inflation targeting objective. We wish to consider a slightly different, asymmetric, monetary regime. The central bank behavior can be expressed as,

(9)

where is the stationary value of the gross nominal interest rate associated to the in-flation target and represents an inflation threshold. It will be useful to define: about which we make the following assumption:

Assumption 1: or, equivalently: .

The monetary rule (9) implies that if current inflation increases beyond the inflation threshold the central bank reacts actively:9 . Otherwise, the central bank pegs10 the current interest rate to its target which is associated to an inflation target lower than . Note that central bank is more concerned about tackling high inflation levels than dealing with scenarios domi-nated by low inflation and by deflation. In most developing countries, high inflation is a relatively frequent phenomena whereas deflation is quite rare and not so deep. Stylized facts on inflation rates in these countries shape an asymmetric behavior. So it seems to be reasonable to suppose an asymmetric behavior of the central bank. This monetary policy can be called “monitoring policy of current inflation”.

In developed countries, much debate has been devoted to the suitability of the Taylor rule in characterizing the behaviour of central banks, especially in abnormal times. Rabanal (2004), for instance, presents evidence that Taylor rule coefficients changed significantly both with time and economic conditions in the United States between 1960 and 2003 using quarterly data. The qualitative interpretation is that the US Federal Reserve places much more weight on inflation stabilization during expansion periods, while it shifts its focus to output stabilization when in recessions. Analogous reasoning applies to the monetary rule (9). In developing countries, Brazil constitutes a successful example of inflation targeting. After being forced to abandon the crawl-ing peg to the US dollar, Brazil adopted an inflation targeting regime in July 1999 which brought annual inflation down to one-digit figures in less than three years.

2.2.2 The Debt Recovery Rule

Given that the fiscal authority does not control the primary surplus, it is useful to suppose the existence of a rule H(.) which specifies how the fiscal authority chooses the recovery rate ht. We will suppose that such a rule is a (non increasing) function of the nominal interest rate, denoted

, to be determined by the monetary authority in the No-Default case. Then,

9.- This condition is identical to that which led Leeper to describe the monetary rule as “active”.10.- Actually, as demonstrated by Uribe (2006), a forward-looking rule of type: will lead to the same results as our simpler pegging rule.

ˆ

ˆ

ˆ ˆ ˆ

ˆ ˆ

ˆ


114 •

represents the potential cost of honoring the whole debt in the future. More precisely, the fiscal authority’s behavior is supposed to be defined by:

(10)

where the threshold denotes the maximum nominal interest rate that the fiscal authority will accept on its new issued debt without defaulting on its current liability. Finally, denotes the fraction of the sovereign debt honored by the fiscal authority in case of default as a decreasing function of function which will be specified later on.

Assumption 2:

This assumption implies that the fiscal authority is more tolerant - said, lax - than its monetary counterpart in terms of equilibrium inflation and interest rates.

However, it is important to point out that the main objective of the central bank is to monitor inflation whereas the fiscal authority only cares about the cost of its debt. Then, note that in order to control current inflation the central bank uses the current interest rate affecting, in this way, the cost of the sovereign debt. Consequently, a conflict of interests between both authorities may arise defining the equilibrium outcome. For instance, an aggressive central bank fighting against inflation may trigger the sovereign default as well as affect its size.

2.2.3 Market Clearing

At equilibrium, the goods market must clear: ct = y meaning that the consumption level is constant along time t. Thus, from equation (4) it turns out that the marginal utility of consumption, , is also constant. Equation (5) becomes . Applying conditional expectations operator Et to the last expression, we obtain where is the the gross rate of inflation and Et Qt,t+1 denotes the nominal price of a risk-free portfolio which pays one unit of currency in all states of the nature. Consequently, the risk-free interest rate can be expressed as,

(11)

Using the constancy of , equation (6) becomes,

(12)

Finally, given that all households are assumed to be identical, at equilibrium, there is no borrowing or lending among them, i.e. Dt = 0 ∀t. Thus all the assets held by private agents are in the form of government debt. Using this result and, again, , the transversality condition can be rewritten (in real terms):

ˆ




(13)

where bt = Bt/Pt.

The sovereign debt dynamics, described by equation (8), can also be written in real terms as:

(14)

Therefore, the equilibrium can be defined as follows:

Definition 1 A rational expectations competitive equilibrium is defined as a set of processes

satisfying equations (11), (12), (13), (14), the monetary rules (9), the debt

recovery rule (10), and the exogenous process for the primary surpluses where R–1b–1 are

given and the recovery rate satisfies ht ∈ (0,1).

Using equations (12), (13), (14), and some algebra - see Appendix - we obtain:

(15)

where Tt is the discounted value of present and future primary surpluses. Note that fiscal surpluses are discounted by the gross real risk-free interest rate given by - see equation (11).

Under this form, (15) is the key equation of the debate between the advocates11 of the Fiscal The-ory of the Price Level determination (FTPL) and its detractors12. If the fiscal authority is committed to honour the whole of its liabilities - and so ht = 1 - then the current inflation rate, , becomes determined according to the FTLP. This is because Tt is exogenous and Rt–1 bt–1 is predetermined in period t. On the contrary, if ht is allowed to be less than unity, then the current value of Tt may affect both current inflation and recovery rate. This may lead to the Buiter’s conclusion that any path for ht and satisfying equation (15) could be considered as an equilibrium outcome.

Using (15) to eliminate from equation (14) we get:

(16)

The real equilibrium value of the public debt is necessarily equal to the present value of future discounted real fiscal surpluses. Now, when t > 0 replaces bt–1 by this equilibrium valuein t –1 into equation (15) we obtain:

(17)

11.- See Leeper (1991), Sims (1994), Woodford (1994, 1995) and Cochrane (1999), among others.12.- Buiter (1999, 2001, 2002) for instance.


116 •

where

is the innovation in percentage points on the present discounted value of primary surpluses. Thus, if the discounted value of present and future primary surpluses is higher than the value ex-

pected for this variable in period t –1. Otherwise, becomes either negative or null and void.

Equation (17) can receive the same interpretation than equation (15).13 Particularly, one may conclude, as Buiter (1999, 2001), that any path for ht and satisfying equation (17) could be considered as an equilibrium outcome. But this is not the case because equation (17) is not the only equilibrium restriction to be satisfied by both and ht. The monetary rule (9) and, espe-cially, the debt recovery rule (10) also affect the equilibrium outcome. Thus, the objective of the next section is to analyze the extent to which each of these variables may react after a shock to Tt. Note however that, whatever the monetary and the recovery rules, the ratio is uniquely determined by equation (17).

3. Three Scenarios for One Economy

The asymmetric form of both equations (9) and (10) may potentially imply the existence of four regimes, but assumptions 1 and 2 permit to exclude the case where the central bank naturally14

pegs the interest rate to leading the fiscal authority to default on its outstanding debt. Indeed, this scenario would require which violates the condition . Three scenarios are left.

The two first scenarios correspond to the No-Default case -where ht = 1- satisfying . Under these scenarios the fiscal authority considers that the potential cost of servicing the whole debt is affordable and so it honors its entire liabilities. The first scenario is characterized by a relatively low current inflation -say, - and so the central bank behaves passively by pegging current interest rates to the level . This type of periods are usually called “Tranquil Times”. The second scenario is characterized by a relatively high current inflation -say, - where the central bank behaves actively by increasing current interest rates. This scenario corresponds to “Infla-tion Times” described by Loyo (1999). The third one is the scenario of Sovereign Default -where

satisfying . In this case, the fiscal authority finds that the potential cost of servicing its whole debt is unaffordable. Consequently, it defaults on its liabilities by honoring only a fraction of its financial obligations.

Both “Tranquil Times” and “Inflation Times” are characterized by the absence of sovereign de-fault. Then, the equilibrium level of inflation and interest rates are given by equations (17) and (9) with ht =1:

(18)

13.- In period t =0 equation (17) becomes where .14.- By “naturally”, we mean: “considering the inflation rate which would be realized in the case of No Default”.

ˆ

ˆ

ˆ




(19)

Equation (18) expresses that current inflation is determined by the current fiscal shock, as predicted by the Fiscal Theory of Price Level (FTPL). And equation (19) expresses that in both No-Default scenarios, the current nominal interest rate is determined by the current inflation level.

3.1 “Tranquil Times”

When the value of the inflation rate (18) satisfies the condition equation (19) implies that the central bank pegs the interest rate to . So, these periods are characterized by both low current inflation and interest rates. We have:

(20)

(21)

where denotes current inflation rate during Tranquil Times and denotes the risky gross nominal interest rate paid by the fiscal authority during Tranquil Times.

Time-t equilibrium is determined as the FTPL determination asserts (See Woodford 1995). The central bank pegs the nominal interest rate to its target and the equilibrium price level is that level that makes the real value of nominally denominated government liabilities equal to the present value of the expected future government budget surpluses.

Both equations (20) and (21) are satisfied on condition that which implies:

(22)

remembering that ; or, equivalently:

(23)

where is defined as:

(24)

Note that, if this scenario applies in period t – 1 we have and the condition (23) can be simplified as: with In this case, Tranquil Times are driven by either positive or not so negative fiscal shocks. It is worth noticing that the negative fiscal shocks must be rather soft. In the case where Rt–1 verifies and especially when a positive fiscal shock may be necessary to restore a period of “Tranquil Times”.

The deterministic steady state associated to (20)-(21) is given by: and Of course, it verifies and . This implies that the steady state inflation level is low enough to let the central bank behave passively, while the low steady state level of the interest rate enables the fiscal authority to honor the entire sovereign debt.

ˆ

ˆ

ˆ

ˆ

ˆ ˆ

ˆ

ˆ

ˆ ˆ

ˆ ˆ

ˆ

ˆ


118 •

Starting from the steady state, the current equilibrium characterizing a Tranquil Time is described by equations (20) and (21) on condition that the economy were not hited by hard negative shocks. Then, if in the next period fiscal shock is void, the economy returns to its steady state.

3.2 ‘‘Inflation Times”

Compared to the previous case, these periods are characterized by both higher current interest rates and inflation levels. This is linked to the fact that the economy is hit by harder negative fis-cal shocks. The current inflation remains defined like in the previous case but it now exceeds the inflation threshold and the central bank behaves actively by increasing current interest rates:

(25)

(26)

where denotes current inflation rate during infaltion times and denotes the risky gross nominal interest rate paid by the fiscal authority during inflation times.This equilibrium is satisfied on condition that and which implies:

(27)

using again or, equivalently:

(28)

with

(29)

and where the function is defined by:

(30)

Condition (28) expresses that a period of inflation is driven by a strictly negative shock which is no longer soft, given the level of Rt–1. The shock is rather hard but not enough to drive the economy into default.

The deterministic steady state is easily obtained by putting and in equation (25) and (26). We obtain:

(31)

(32)

ˆ

ˆ

ˆ

ˆ

ˆ ˆ

ˆ

ˆ ˆ

ˆ

ˆ




This deterministic steady state equilibrium exists on condition that and or, equivalently:

The left-hand side of the previous inequality is always verified under Assumption 1, and the right-hand side requires the following condition:

Assumption 3:

!

R > R + R" R ( )/ #$ "1( ) !

Assumption 3 implies that must be high enough to satisfy . This condition is needed to ensure the existence of a deterministic steady state under a period of inflation.

The (partial) dynamics of these two scenarios is represented on Figure 1 in the case :

Figure 1: The No-Default Case

It is worth noticing that, while is locally stable, given that , RI is an unstable steady state equilibrium. This means that, depending on the previous value of the nominal inter-est rate -at the left or at the right from RI - the current interest rate will converge to (if is void or small enough), or increase toward . Unless a big positive fiscal shocks occurs, the latter scenario inevitably leads to a sovereign default.

In the scenario of “Inflation Times”, a previous value of the nominal interest rate higher than RI cause the financial wealth of private agents to grow faster in nominal terms, which calls for higher inflation. Monetary authority responds to higher inflation with sufficiently higher nomi-nal interest rates forming a vicious circle. Usually, hyperinflation is interpreted as a result of the monetary financing of serious fiscal imbalances. However, in this case a fiscalist alternative is presented in which inflation explodes because of the fiscal effects of monetary policy. Most of the

ˆ

ˆ

ˆ ˆ


120 •

action concentrates on the interest rate pays on the government debt and debt rollover instead of seigniorage. This phenomena is known as ‘fiscalist hyperinflation’ and is the case of Brazil in the late 1970s and early 1980s (See Loyo 1999).

3.3 ‘‘Sovereign Default Time”

According to the fiscal authority, the potential cost of servicing the whole debt becomes too high when which implies:

(33)

or, equivalently,

(34)

This condition shows that for a given level of Rt–1, a scenario of Default can be triggered by a hard negative shock or, for a given shock by a high level of the previous nominal interest rate.

As a consequence, the fiscal authority defaults on its debt by honoring only a fraction ht <1 of its liabilities. From equations (17) and (9), current inflation and interest rate become:

(35)

(36)

Note that without specifying the recovery rule - - the equilibrium in period t remains undetermined and defined by equation (35) and (36). There is a continuum of recovery rate determining the equilibrium inflation rate and so the nominal interest rate. This result is in line with Buiter’s criticism.15 In order to avoid this indeterminacy the fiscal authority has to specify a recovery rule.

Before introducing such a recovery rule, let us rewrite the system in a simplified form, using (18) equations (35) and (36) can be rewritten as:

(37)

(38)

where the last terms have been obtained by inverting the monetary rule (9).

We now can make the following assumption about the recovery function

15.- See Buiter (1999), pp. 50, Proposition 5.

ˆ




(39)

Equation (39) shows that the higher is , the potential cost of honoring the entire debt, the smaller is the recovery rate. Using the recovery function (39) and the monetary rule (9) in equa-tions (37) and (38), the equilibrium values of and become:

Thus, the recovery rule (39) allows the economy, by defaulting on its financial obligations, to reach the stable steady state equilibrium16 in the same period t. The equilibrium value of the recovery rate is:

(40)

4. Inflation and Default Dynamics

This section illustrates the economy dynamics in two different cases of default. In the first one, the current fiscal shock is small but the initial value of the nominal interest rate, R–1 is high. The economy jumps into an inflation episode which leads the central bank to rise its interest rate and, after three periods, the fiscal authority to default. In the second case, the initial interest rate is at its “Tranquil Times” stationary value: but the economy experiences a big negative fiscal shock17 which leads very rapidly to a sovereign default.

Figure 2: A small fiscal shock Figure 3: A big negative fiscal shock

16.- Besides, this recovery rule minimizes the probability of default after the sovereign default. See the next sections on Expected Recovery Rate and Sovereign Risk Spread.17.- Figure 3 and especially 4 are only illustrative because we have to expect that a negative shock - an innovation - has no reason to repeat.


122 •

Deciding to default on the government liabilities is a difficult decision for policy makers. This may explain why, in the data, the actual value of the interest rate is greater than what one might expect. This seems to be the case of Argentina in 1989. At the end of 1989, the year in which Argentina defaulted on its debt, the inflation rate had reached a shocking 4923,6%. Then Argen-tina had gradually converged to its steady state equilibrium (See Table 1):

Table 1: Argentinean Default, 1989

Year 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995

Inflation (%) 81,9 174,8 387,7 4923,6 1343,9 84 17,5 7,4 3,9 1,6

Source: Indec

According to our model, the fiscal authority was both too tolerant and patient, i.e. was too high. Moreover, Argentina in 1989 could minimize the recovery rate on its debt in order to reach faster the (without inflation) steady state equilibrium. To explain the gradual decline of inflation, it is necessary to modify the recovery rule slightly. Suppose that the rule is now defined by:

(41)

with .

The recovery rule (41) and the condition are well specified in order to ensure a post-default equilibrium which drives progressively the economy to “Tranquil Times”, on condition that future fiscal shocks are small enough.

This case is represented on Figure 4:

Figure 4: Argentinean soft-landing

ˆ

ˆ




This policy has the double advantage of reducing the size of the sovereign default necessary to restore the public solvency and to smooth the return toward price stability. On the other hand, this recovery rule does not minimize the probability of a new default after the first sovereign default. For the sake of simplicity, we will adopt in the rest of the paper the simpler assumption

.

5. Expected Recovery Rate and Sovereign Risk Premium

In this section, we make simplifying assumptions on the fiscal shock distribution and we show

that, once the Fiscal Default Rule is known,18 the one-period Expected Recovery Rate, ,

and the Relative Sovereign Risk Premium, , can be explicitly calculated. Note that the

period-t probability of default in t + 1 is simply given by: .

The three scenarios previously described are summarized by the following table:

Table 2: The Three Scenarios

Tranquil Times Inflation Times Sovereign Default Time

ht : 1 1

:

Rt :

The conditions that determine the current regime are entirely defined by the couple of states

variables Let defines the distribution function of the fiscal shocks, and

and , respectively, the upper and lower bound of the compact set on which the shock is

distributed. The Figure 5 summarizes these conditions:

18.- Which is summarized by the choice of and (= in our case).

ˆ ˆ

ˆ


124 •

Figure 5: The Regime Determination

Note that the period-t – 1 probability of default in t is simply given by: the probability

of a “Tranquil Times” period by: , and the probability of an “Inflation Times” episode

by: Then, if Rt–1 < b the ex ante probability of a sovereign default in period

t is null and, when Rt–1 < a the probability of a “Tranquil Times” period equals unity.

5.1 Expected Recovery Rate

Using Table 1, the one-period Expected Recovery Rate can be written:

(42)

Notice that the default probability, is strictly positive (Resp. null) and that is

strictly negative (Resp. null) if (Resp. ). One can conclude that the one-period

expected recovery rate verifies: for and otherwise. Starting from the

ˆ ˆ




“Tranquil Times” steady state, i.e. we have and (42) can be

simplified as:

which then verifies for and otherwise.

5.2 Sovereign Risk Premium

From (11) and (12), the relative sovereign risk premium can be defined by:

Using again the results in Table 1, this expression becomes:

(43)

with when and otherwise. One concludes that the

relative sovereign risk premium is strictly positive for and null otherwise. At the

“Tranquil Times” steady state, i.e. equation (43) simplifies to:

which is strictly positive for and null otherwise.

ˆ


126 •

5.3 Calibration and Simulation

For sake of simplicity, we will assume that fiscal (relative) innovations are uniformly distributed: . The one-period Expected Recovery Rate is given by equation (42) which can be

rewritten, for .

and the one-period Sovereign Risk Premium is:

We can easily illustrate our results by adopting the following annual calibration for the model’s parameters:

Table 3: Calibration

Definition Parameter Value

Discount Factor: 0,95

Taylor Coefficient: 1,50

Interest rate target: 1,05

Monetary threshold: 1,10

Fiscal threshold: 1,50

Upper bound of the distribution function: 0,15

We can firstly calculate the lower threshold value of Rt for which .The solution is 1.177 which is superior to R=1.05 (and even superior to R =1.10. This implies that, starting from the “Tranquil Times” Steady State in period t, the probability for the Government to default on its debt in t +1 is always null. After this calibration, a Sovereign Default cannot be observed without a period of growing inflation. Consequently, an aggressive central bank fighting against inflation does not go without costs. It increases the fiscal burden of the government debt as well as the sovereign risk of default. A higher current interest rate increases the current probability of default which is captured by the current sovereign risk premium.

The resulting values for the fiscal default threshold, the default probability the Expected Recovery Rate, and the one-period Sovereign Risk Premium are rep-resented as functions of the current interest rate Rt:

ˆ




Graph 1: Default Threshold Graph 2: Probability of Default

Graph 3: Expected Recovery Rate Graph 4: Relative Risk Premium

These Graphs show that when Rt ≤ 1.177 even the hardest the negative fiscal shock -say - does not drive the economy into default. Thus, the Probability of Default is null, the

Expected Recovery Rate is equal to one, and so the Relative Risk Premium is void. Otherwise, when Rt ≤ 1.177 there are (negative) values for the fiscal shock that might drive the economy into default. Thus, the Probability of Default becomes positive, the Expected Recovery Rate becomes lower than the unity, and the Relative Risk Premium positive. The higher is Rt the higher are both the Probability of Default and Relative Risk Premium and the lower the Expected Recovery Rate.

This finding contrasts with that of Uribe (2006)’s and is in line with the empirical evidence and estimates presented on Chapter 1.

6. Further Research

This section presents a work in progress. It provides interesting findings and contributes to the discussion over the monetary policy on inflation as well as sovereign default dynamics.

As before, assume that the monetary policy takes the form of an interest-rate feedback rule


128 •

whereby the short-term nominal interest rate is set as a function of inflation. But unlike the pre-vious case, assume that the interest rate controlled by the central bank is the risk-free nominal interest rate, and not the interest rate paid on government debt, Rt.19 Then, the central bank behavior can be expressed as,

(44)

where is the stationary value of the gross nominal interest rate associated to the in-flation target and represents an inflation threshold. It will be useful to define: about which we make the same assumption as before: or, equivalently: .Therefore, the equilibrium can be defined as:

Definition 2 A rational expectations competitive equilibrium is defined as a set of processes

satisfying equations (11), (12), (13), (14), the monetary rules (44), the debt

recovery rule (10), and the exogenous process for the primary surpluses where R–1b–1 are

given and the recovery rate satisfies ht ∈(0,1).

6.1 The Three Scenarios for this Economy

The asymmetric form of both equations (44) and (10) may potentially imply the existence of four regimes, but assumptions 1 and 2 permit to exclude the case where the central bank naturally20 pegs the interest rate to , leading the fiscal authority to default on its outstanding debt. Three scenarios are left.

Both “Tranquil Times” and “Inflation Times” are characterized by the absence of sovereign de-fault. Then, the equilibrium level of inflation and interest rates are given by equations (17) and

(44) with ht=1:

(45)

(46)

The third one is the scenario of Sovereign Default - where ht<1. In this case, the fiscal authority finds that the potential cost of servicing its whole debt is unaffordable. Consequently, it defaults on its liabilities by honoring only a fraction of its financial obligations.

19.- This assumption seems to be more in accordance with the cashless economy framework that we have chosen. Indeed, this framework does not explicitly take into account the open market interventions by the central bank on the government securities market and does not facilitate an explaination of the control of the interest rate Rt by monetary authorities.20.- By “naturally”, we mean: “considering the inflation rate which would be realized in the case of No Default”.

ˆ

ˆ

ˆ ˆ ˆˆ ˆ

ˆ

ˆ




6.1.1 “Tranquil Times”

When the value of the inflation rate (45) satisfies the condition equation (46) implies that the central bank pegs the risk-free interest rate to . Also, these periods are characterized by both low current inflation and interest rates. We have:

(47)

(48)

where denotes the current risk-free nominal interest rate in “Tranquil Times”.

Both equations (47) and (48) are satisfied on condition that which implies:

(49)

remembering that ; or, equivalently:

(50)

where is defined as:

(51)

6.1.2 ‘‘Inflation Times”

Compared to the previous case, these periods are characterized by both higher current interest rates and inflation levels. And this is linked to the fact that the economy is hit by harder negative fiscal shocks. The current inflation remains defined like in the previous case but now it exceeds the inflation threshold and the central bank behaves actively by increasing the current interest rate:

(52)

(53)

where denotes the current risk-free nominal interest rate in “Inflation Times”.

This equilibrium is satisfied on condition that and which implies:

(54)

using again ; or, equivalently:

(55)

ˆ

ˆ

ˆ ˆ

ˆ

ˆ

ˆ

ˆ ˆ

ˆ

ˆ

ˆ

ˆ ˆ

ˆ ˆ

ˆ


130 •

with

(56)

and where the function is defined by:

(57)

Condition (55) expresses that a period of inflation is driven by a strictly negative shock which is no longer soft, given the level of . The shock is rather hard but not enough to drive the economy into default.

6.1.3 ‘‘Sovereign Default Time”

According to the fiscal authority, the potential cost of servicing the whole debt becomes too high

when i.e., which implies:

(58)

or, equivalently,

(59)

This condition shows that for a given shock , a scenario of Default can be triggered by a high level of the previous nominal interest rate or, for a given level of by a hard negative fiscal shock.

As a consequence, the fiscal authority defaults on its debt by honoring only a fraction ht < 1 of its liabilities. From equations (17) and (44), current inflation and risk free interest rates become:

(60)

(61)

where is always defined by (45) as the equilibrium inflation rate in the no default case.

Note that without specifying the recovery rule - - the equilibrium in period t remains undetermined and defined by equation (60) and (61).21 There is a continuum of recovery rate determining the equilibrium inflation rate and so the nominal interest rate. This result is line

21.- We suppose that such a rule is a (non increasing) function of the nominal interest rate, denoted to be determined by the monetary authority in the No-Default case. Then, represents the potential cost of honoring the whole debt in the future.

ˆ

ˆ




with Buiter’s critic.22 In order to avoid this indeterminacy the fiscal authority has to specify a recov-ery rule. We now can make the following assumption about the recovery function :

(62)

where verifies:

Assumption 4:

Equation (62) shows that, for a given value of , the higher is the potential inflation rate , the smaller is the recovery rate. Using the recovery function (62) and the monetary rule (44) in equa-tions (60) and (61), the equilibrium values of and become:

Thus, the recovery rule (62) allows the economy, by defaulting on its financial obligations, to reach a less inflationary equilibrium in the same period t. Note that if the monetary rule (44) implies .

Using (45) in equation (62), the equilibrium value of the recovery rate is:

(63)

Assumption 4 ensures that this recovery rate is always inferior to unity.

6.2 Expected Recovery Rate, Sovereign Risk Premium and Interest Rates

It is possible to express as an invertible function of and hence as a function of . Let denote the gross sovereign risk premium. From (11) and (12), Pt can be defined by:

(64)

Now lets make simplifying assumptions about the fiscal shock distribution and we will see that,

once the Fiscal Default Rule is known - which is summarized by the choice of and -the

22.- See Buiter (1999), pp. 50, Proposition 5.

ˆ

ˆ


132 •

sovereign risk premium, and the one-period expected recovery rate, can be

calculated. Let us begin by summarizing our results with the following table:

Table 4: The three scenarios

Tranquil Times Inflation Times Sovereign Default Time

ht : 1 1

:

Rt :

where and and with and .

6.2.1 Expected Recovery Rate

Using the results presented in the first row of Table 1, the one-period Expected Recovery Rate can be written:

(65)

where the last term after the sign of subtraction is positive under Assumption 4 for and null otherwise.

One can conclude that the Expected Recovery Rate verifies for and

otherwise. Identically, let us define such that According to equation

(54), we have :

ˆ ˆ

ˆ

ˆ ˆ ˆ ˆ




and we know that if .

6.2.2 Sovereign Risk Premium

Using now the second row of Table 4 and equation (64), the gross sovereign risk premium - or country risk spread - can be written:

(66)

where the last term after the sign of addition is positive under Assumption 4 for and null otherwise. So, we can conclude that if .

We can easily obtain the derivative of the function for .

where we have used equation (57) and (56). Assumption 4 insures that this derivative is always positive.

Using again the definition of the gross sovereign risk premium: and equation (66), one can establish a link between and :

(67)

where the function has a derivative given by:

which verifies: if and otherwise.

ˆ


134 •

Equation (67) implicitly permits to determine the interest rate on government securities as a func-tion of the riskless interest rate set by the central bank. By inverting the function G(.) one finds:

(68)

where the function verifies: if and otherwise.

6.2.3 Riskless and Risky Interest Rates

Using finally the last row of Table 4 and equation (68), one can express the risky sovereign debt interest rate:

(69)

Let us define such that: and suppose that

By evaluating the derivative for when we obtain:

We now can represent and as function of in the case :

ˆ

ˆ ˆ ˆ

ˆ

ˆ ˆ ˆ




Figure 6

Note that the assumption is sufficient but unnecessary to guarantee .

Corollary 3 In presence of a default risk, the monetary policy can be active even in the case .

7. Conclusion

The main goal of this chapter is to characterize the way in which monetary policy affects the equilibrium behavior of recovery rate and sovereign risk premiums. This is an issue which has been fairly disregarded by recent monetary theory. The framework of analysis proposed in this chapter offers an additional perspective to discuss the possible interrelations between monetary and fiscal policy and provides supplementary advantages as regards other settings. It allows to overcome somes difficulties like the negative default rate which arises as a consequence of posi-tive fiscal shocks -recall the `super-solvency premium’ in terms of Buiter. This model characterizes the way in which monetary policy affects the equilibrium behavior of price level, recovery rate and sovereign risk premiums. Indeed, in some cases, even a ‘passive’ interest rate rule might drive the economy to an unsustanable path without defaulting. It means that in presence of a default risk, the monetary policy can be active even in the case where . It turns out that monetary policy plays a significant role in shaping the equilibrium behavior of default and risk premiums. Both the Probability of Default and Sovereign Risk Premium are consistent with the empirical estimates presented in the previous chapter.

It also underlines the fact that the size of the equilibrium default rate matters for the post-equi-librium dynamics. The size of the equilibrium default rate cannot be so high in order to ensure a post-equilibrium dynamics without defaulting. This theoretical result is consistent with the argu-ment presented on section 1.4.3 of the previous chapter as to the assesment of the Argentine Debt Haircut after the last event of default on December, 2001. The model explicitly emphasizes the role of the government (the fiscal authority) in resolving the financial crisis.


136 •

Even though the current framework can be extended in different directions, these have been left aside to simplify the exposition. For instance, it can be assumed that a fraction of the public debt is indexed. High inflation economies tend to develop an extensive system of indexed contracts. It is worth noticing that bonds linked to price indices are not `real’ bonds because sampling and computing price indices involve time. The nominal value of indexed bonds is typically adjusted ac-cording to lagged inflation rates. Otherwise, it coud be assumed that public debt is denominated in foreign currency. These are important characteristics of actual emerging economies that would be worthwhile incorporating.

8. Appendix

Multiplying both sides of equation (8) by as .

.

Then, iterating the last expression j times, it results in:

(70)

Dividing both sides of equation (70) by -see that can also be written as

- it turns out that,

(71)

Applying the conditional expectations operator Et, equation (71) becomes written,

(72)




Next, by applying the law of iterated expectations and using equation (12) equation (72) remains expressed as,

(73)

Dividing both sides of equation (73) by and then taking the limit for , it turns out,

(74)

See that defining , the left hand-side of equation (74) can be expressed as

. Then, multiplying this expression by it remains expressed as equation

(13) which is equal to zero. So equation (74) results in,

9. References

Aghion, P., A. Banerjee and T. Piketty (1999), “Dualism and macroeconomic volatility”, Quarterly Journal of Economics, Vol. 114, No 4, p. 1359-97.

Altman, E. and B. Pasternack (2006), “Defaults and Returns in the High Yield Bond Market”, Journal of Applied Research in Accounting and Finance (JARAF), Vol. 1 No. 1, pp. 3-29.

Altman, E., A. Resti and A. Sironi (2004), “Default Recovery Rate in Credit Risk Modeling: A Review of the Literature and Empirical Evidence”, Economic Notes by Banca dei Monte dei Paschi di Siena, Vol. 33.

Arena, M., C. Reinhart and F. Vázquez (2007), “The Lending Channel in Emerging Economies: Are Foreign Banks Different?”, Working Paper, No. 48, International Monetary Fund, Wash-ington, DC.

Barajas, A. and R. Steiner (2002), “Credit stagnation in Latin America”, Working Paper No. 53, International Monetary Fund, Washington D.C.

Barajas, A., R. Steiner and N. Salazar (2000), “Foreign investment in Colombia’s financial sector”, in Stijn Claessens and Marion Jansen (Eds), The internationalization of financial services: Issues and lessons for developing countries, Boston.

Barro R.J. (1974), “Are Government Bonds Net Wealth?” The Journal of Political Economy, Vol. 82, No. 6, pp. 1095-1117.


138 •

Berger, A., L. Klapper and G. Undell (2001), “The Ability of Banks to lend to Informationally Opaque Small Businesses”, Journal of Banking and Finance, Vol. 25, pp. 2127-2167.

Bhanot, K. (1998), “Recovery and Implied Default Brady bonds”, The Journal of Fix Income, Vol. 14, No. 1, pp. 47 --51.

Biuter, W. (1999) “The Fallacy of the Fiscal Theory of the Price Level”, NBER Working Paper No. W7302.

Biuter, W. (2001), “The Fallacy of the Fiscal Theory of the Price Level, Again”, Bank of England Working Paper Series No. 141.

Biuter, W. (2002), “The Fiscal Theory of the Price Level: A Critique”, The Economic Journal, Vol. 112, pp. 459-480.

Black, F. and Scholes, M. (1973), “The Pricing of Options and Corporate Liabilities”, Journal of Political Economics, Vol. 81, No. 3, pp. 637-654 .

Boyd, J.H. and G. de Nicoló (2005), “The Theory of Bank Risk-Taking and Competition Revisited”, Journal of Finance, Vol. 60, No. 3, pp. 1329-43.

Breitung, J. and M.H. Pesaran (2006), “Unit Roots and Cointegration in Panels”, in L. Matyas and P. Sevestre (Eds.), The Econometrics of Panel Data: Fundamentals and Recent Developments in Theory and Practice, Kluwer Academic Publishers, forthcoming.

Breusch, T.S. and A.R. Pagan (1980), “The Lagrange Multiplier Test and Its Applications to Model Specification in Econometrics”, Review of Economic Studies, Vol. 47, No. 1, pp. 239-53.

Buch, C. (2000), “Why Do Banks Go Abroad? Evidence from German Data”, Financial Markets, Institutions and Instruments, Vol. 9, No. 1, pp. 33-67.

Calvo, G. and C. Reinhart (2002), “Fear of Floating”, Quarterly Journal of Economics Vol. 117, pp.379-408.

Calvo, G.A. and F. Coricelli (1993), “Output collapse in Eastern Europe: the role of credit”, IMF Staff Reports, Vol. 40, pp. 32-52.

Caprio, G., D. Klingebiel, L. Laeven and G. Noguera (2005), “Banking Crisis Database”, in P. Honohan and L. Laeven (Eds.), Systemic Financial Crises, Cambridge University Press, Cam-bridge.

Carstens, A., D. Hardy and C. Pazarbasioglu (2004), “Avoiding Banking Crises in Latin America”, Finance & Development, September Issue, pp. 30-33.

Choi, I. (2001), “Unit Root Tests for Panel Data”, Journal of International Money and Banking, Vol. 20, pp. 249-72.

Claessens, S. and M. Jansen (2000), “The Internationalization of Financial Services: Issues and Lessons for Developing Countries”, Boston, Mass.: Kluwer Academic Press.

Clarida R., J. Gal and M. Gertler (1998), “Monetary Rules in Practice: Some International Evi-dences”, Working Paper No. 6254 NBER.

Clarida, R., G. Jordi and M. Gertler (2000), “Monetary Policy Rules and Macroeconomics Stability: Evidence and Some Theory”,Quarterly Journal of Economics, MIT Press, vol. 115(1), pages 147-180.

Cochrane, J. H. (1999),. “Portfolio advice of a multifactor world,” Economic Perspectives, Federal Reserve Bank of Chicago, issue Q III, pages 59-78.

Corvoisier, S. and R. Gropp (2002), “Bank Concentration and Retail Interest Rates”, Journal of Banking and Finance, Vol. 26, No. 11, pp. 2155-89.

Cowan, K., E. Levy-Yeyati, U. Panizza and F. Sturzenegger, (2006), “Sovereign Debt in the Ameri-cas: New Data and Stylized Facts,” Public Debt in Americas. Working Papers 577, Inter-Amer-ican Development Bank, Washington D.C.

Crystal, J., G. Dages and L. Goldberg (2002), “Has Foreign Bank Entry Led to Sounder Banks in Latin America?”, Current Issues in Economics and Finance, Federal Reserve Bank of New York, Vol. 8, No. 1.

Dages, B.G, L.S. Goldberg and D. Kinney (2000), “Foreign and Domestic Bank Participation in




Emerging Markets: Lessons from Mexico and Argentina”, Economic Policy Review, September Issue, pp. 17-36, Federal Reserve Bank of New York, NY.

Dages, G., L. Goldberg and D. Kinney (2000), “Foreign and Domestic Bank Participation in Emerg-ing Markets: Lessons from Mexico and Argentina”, Policy Review Vol. 6, No. 3, pp. 17--36, Federal Reserve Bank of New York Economic.

Dahl, D. and R. Shrieves (1999), “The Extension of International Credit by U.S. Banks: A Disag-gregated Analysis, 1988-1994”, Journal of International Money and Finance, Vol. 18, pp. 153-167.

de Haas, R.T. (2006), “Monitoring Costs and Multinational-Bank Lending”, DNB Working Paper No. 88, De Nederlandsche Bank, Amsterdam.

de Haas, R.T. and I.P. van Lelyveld (2004), “Foreign bank penetration and private sector credit in Central and eastern Europe”, Journal of Emerging Market Finance, Vol. 3(2), pp. 125-151.

de Haas, R.T. and I.P. van Lelyveld (2006), “Foreign Banks and Credit Stability in Central and Eastern Europe: A Panel Data Analysis”, Journal of Banking and Finance, Vol. 30, No. 7, pp. 1927-52.

Demirgüç-Kunt, A., L. Laeven and R. Levine (2003), “The Impact of Bank Regulations, Concen-tration, and Institutions on Bank Margins”, Policy Research Working Paper, No. 3030, World Bank, Washington, DC.

Demirgüç-Kunt, A., R. Levine and H.G. Min (1998), “Opening to foreign banks: issues of stability, efficiency and growth”, in Seongtae Lee (Eds.), The Implications of Globalization of World Financial Markets, Seúl, Banco de Corea.

Denizer, C., M. Iyigun and A. Owen (2002), “Finance and Macroeconomic Volatility”, Contribu-tions to Macroeconomics, Vol. 2, No. 1, Berkeley Electronic Press, Berkeley, CA.

Detragiache, E. and P. Gupta (2004), “Foreign banks in emerging market crises: evidence from Malaysia”, IMF Working Paper, Washington D.C.

Detragiache, E. and P. Gupta (2006), “Foreign Banks in Emerging Market Crises: Evidence from Malaysia”, Journal of Financial Stability, Vol. 2, No. 3, pp. 217-42.

Detragiache, E., T. Tressel and P. Gupta (2006), “Foreign banks in poor countries: Theory and evidence”, IMF Working Paper No. 18.

Dewatripoint, M. and E. Maskin (1995), “Credit Efficiency in Centralized and Decentralized Econ-omies”, Review of Economic Studies, Vol. 62, No. 4, pp. 541-556.

Diamond, D. W. (1984), “Financial Intermediation and Delegated Monitoring”, The Review of Economic Studies, Vol. 51, No. 3, pp. 393-414.

Dornbusch R., F. Sturzenegger and H.C. Wolf (1991) “Extreme Inflation: Dynamics and Stabiliza-tion”, NBER Working Paper No. R1584.

Dornbusch, R. (1983), “Real Interest Rates, Home Goods, and Optimal External Borrowing”, Journal of Political Economy, Vol. 91, No. 1, pp. 141-53.

Duffie, D. (1998), “Defaultable Term Structure Models with Fractional Recovery of Par”, Unpub-lished manuscript, Graduate School of Business, Stanford University.

Duffie, D. and K.J. Singleton (1999), “Modeling the Term Structures of Defaultable Bonds”, Re-view of Financial Studies, Vol. 12, No. 4, pp. 687-720.

Duffie, D. and K.J., Singleton (1999), “Simulating Correlated Defaults”, Unpublished manuscript, Graduate School of Business, Stanford University.

Elmendorf, D. and N.G. Mankiw (1999), “Government Debt,” in Handbook of Macroeconomics, edited by J.B. Taylor and M. Woodford, Eslevier Science B.V., pp. 1615-1669.

Escudé, G., T. Burdisso, M. Catena, L. D’Amato, G. McCandless and T.E. Murphy (2001), “Las MIP y MES y el mercado de crédido en la Argentina”, Documento de Trabajo No. 15, Banco Central de la Republica Argentina, Buenos Aires.

Evanoff, D. and E. Ors (2002), “Local Market Consolidation and Bank Productive Efficiency”, Working Paper, No. 25, Federal Reserve Bank of Chicago, Chicago, IL.


140 •

Fons, J. (1987), “The default premium and corporate bond experience”, The Journal of Finance, Vol. 42, No. 1, pp. 81-97.

Frankel, J.A. and A.K. Rose (1996), “Currency Crashes in Emerging Markets: Empirical Indica-tors”, Discussion Paper, No. 1349, Centre for Economic Policy Research, London.

Galindo, A., A. Micco and A. Powell (2004) “Loyal Lenders or Fickle Financiers: Foreign Banks in Latin America”, Working Paper 08-2004, Universidad Di Tella

Galindo, A., A. Micco and A. Powell (2005) “Loyal lenders or fickle financiers: Foreign banks in Latin America”, Working Paper No. 529, Inter-American Development Bank, Washington D.C.

Giannetti, M.A. and Ongena (2005), “Financial integration and entrepreneurial activity: Evidence from foreign bank entry in emerging markets”, Working Paper Series No. 498, European Central Bank.

Goldberg, L.S (2001), “When is US bank lending to emerging market volatile?”, NBER Working Paper No. 8209, National Bureau of Economic Research, Cambridge, MA.

Goldberg, L.S. (2002), “When is U.S. Bank Lending to Emerging Markets Volatile?”, Working Paper, No. 8209, National Bureau of Economic Research, Cambridge, MA.

Goldberg, L.S. (2007), “Financial Sector FDI and Host Countries: New and Old Lessons”, Eco-nomic Policy Review, Vol. 13, No. 1, pp. 1-17.

Haouat, M., D.N. Moccero and R.S. Navarro (2006), “Foreign Banks and Economic Stability in Emerging Countries: A Theoretical and Empirical Survey”, mimeo, Université d’Evry val d’Essonne.

Haouat, M., D.N. Moccero and R.S. Navarro (2008), “Foreign Banks and Credit Volatility: The Case of Latin American Countries”, Working Paper No. 61 , The Working Paper Series at The American University of Paris, published and sponsored by the Trustee Fund for the Advance-ment of Scholarship (TFAS).

Hausmann, R., U. Panizza and E. Stein (2001), “Why Do Countries Float the Way They Float?” Journal of Development Economics, Vol. 66, pp.387-414.

Hernandez, L. and H. Rudolph (1995), “Sustainability of private capital flows to developing coun-tries- is a generalised reversal likely?”, Working Paper No. 1518, World Bank Policy Research, World Bank, Washington D.C.

Holmström, B. and J. Tirole (1997), “Financial Intermediation, Loanable Funds, and the Real Sec-tor”, Quarterly Journal of Economics, Vol. 112, No. 3, pp. 663-691.

IADB (2005), “Unlocking Credit: The Quest for Deep and Stable Bank Lending”, Economic and Social Progress Report

IADB (2007), “Living with Debt: How to Limit the Risk of Sovereign Finance”, Economic and Social Progress Report

Im, K., M.H. Pesaran and Y. Shin (2003), “Testing for Unit Roots in Heterogeneous Panels”, Jour-nal of Econometrics, Vol. 115, pp 53-74.

Jarrow, R.A. and S.M. Turnbull (1995), “Pricing Derivatives on Financial Securities Subject to Cred-it Risk”, Journal of Finance, Vol. 50, No. 1, pp. 53-85.

Jarrow, R.A., D. Lando, and S.M. Turnbul (1997), “A Markov Model for the Term Structure of Credit Risk Spreads”, The Review of Financial Studies, Vol. 10, No. 2, pp. 481-523.

Jenneau, S. and M. Micu (2002), “Determinants of international bank lending to emerging mar-ket countries”, BIS Working Paper No. 112, Bank of International Settlements, Basel.

Kiyotaki, N. and J. Moore (1993), “Credit Cycles”, mimeo, London Scholl of Economics.Kraft, E. (2002), “Characteristics and behaviour of foreign banks during and after the banking cri-

sis”, paper prepared for the 5th Conference on Financial Sector Reform in Central and Eastern Europe: The Impact of Foreign Banks Entry, Tallinn, Estonia, 26--27 April.

Leeper, E. (1991), “Equilibria Under `Active’ and `Passive’ Monetary and Fiscal Policies” Journal of Monetary Economics, Vol. 27, pp. 129-147.




Levin, A., J.C. Lin and C.-S. Chu (2002), “Unit Root Tests in Panel Data: Asymptotic and Finite-Sample Properties”, Journal of Econometrics, Vol. 108, pp. 1-24.

Levine, R. (1997), “Financial Development and Economic Growth: Views and Agenda”, Journal of Economic Literature, Vol. 35, pp. 688-726.

Levine, R. (1999), “Foreign bank entry and capital control liberalization: effects on growth and stability”, Manuscript, Carlson school of Management, University of Minnesota, Minneapo-lis.

Loyo, E. (1999), “Tight money paradox on the loose: a fiscalist hyperinflation”, Manuscript, Har-vard University.

Martinez Peria, M.S., A. Powell and I. Vladkova (2005), “Banking on Foreigners: The Behaviour of International Bank Lending to Latin America, 1985-2000”, IMF Staff Papers, Vol. 52, No. 3, International Monetary Fund, Washington, DC.

Martinez Peria, S., A. Powell and I. Vladkova (2002), “Banking on foreigners: The behaviour of international bank lending to Latin America, 1985-2000”, World Bank Working Paper No. 2893, World Bank, Washington D.C.

McCallum, B. T. (1981), “Price level determinacy with an interest rate policy rule and rational expectations,” Journal of Monetary Economics, Elsevier, Vol. 8 No. 3, pages 319-329.

McCauley, R.N., J.S. Ruud and P.D. Wooldridge (2002), “Globalising international banking”, BIS Quarterly Review.

McGuire, P. and N. Tarashev (2005a), “The International Banking Market,” BIS Quarterly Review (June), pp. 15--30.

McGuire, P. and N. Tarashev (2005b), “The International Banking Market,” BIS Quarterly Review (September), pp. 15--30.

Merrick, J.J. Jr. (2001), “Crisis Dynamics of Implied Default Recovery Ratios: Evidence from Russia and Argentina”, Journal of Banking & Finance, Vol. 25, No. 10, pp. 1921-1939.

Merrill Lynch (2000), “Recovery Rates: The Search of Meaning”, High Yield, March 2000.Merton, R. C. (1974), “On the Pricing of Corporate Debt: The Risk Structure of Interest Rates”,

The Journal of Finance, Vol. 29, No. 2, pp. 449-470.Micco, A. and U. Panizza (2004), “Bank ownership and lending behaviour”, Working Paper No.

520, Inter-American Development Bank.Micco, A. and U. Panizza (2006), “Bank Ownership and Lending Behaviour”, Economics Letters,

Vol. 93, No. 2, pp. 248-54.Micco, A., U. Panizza, and M. Yanez (2004), “Bank Ownership and performance: Are public

banks different?”, Mimeographic Document, Inter-American Development Bank, Washing-ton, D.C.

Mishkin, F. (1999), “Financial Consolidation: Dangers and Opportunities”, Journal of Banking and Finance, Vol. 23, pp. 675-91.

Morgan, D. and P.E. Strahan (2003), “Foreign bank entry and business volatility: Evidence from US states and other countries”, NBER Working Paper No. 9710.

Morgan, D.P., B. Rime and P.E. Strahan (2002), “Bank Integration and State Business Cycles”, SSRN Electronic Paper Collection

Morgan, D.P., B. Rime and P.E. Strahan (2004), “Bank integration and state business volatility”, Quarterly Journal of Economics, Vol. 119(4), pp. 1555-85.

Moshiriam, F. (2001), “International Investment in Financial Services”, Journal of Banking and Finance, Vol. 25, pp. 317-37.

Niepelt, D. (2004), “The Fiscal Myth of the Price Level” The Quarterly Journal of Economics, MIT Press, vol. 119(1), pages 276-299.

Peek, J. and E. Rosengren (1997), “The International Transmission of Financial Shocks: The Case of Japan”, American Economic Review, Vol. 87, pp. 495-505.

Peek, J. and E. Rosengren (2000a), “Implication of the Globalisation of the Banking Sector: The


142 •

Latin American Experience”, New England Economic Review, September-October Issue, pp. 45-62.

Peek, J. and E. Rosengren (2000b), “Collateral Damage: Effects of the Japanese Bank Crisis on the United States”, American Economic Review, Vol. 90, No. 1, pp. 30-45.

Pesaran, M.H. (2007) “A Simple Panel Unit Root Test in the Presence of Cross Section Depen-dence”, Journal of Applied Econometrics, Vol. 22, No. 2, pp. 265-312.

Pyle, D. (1971), “On the theory of financial intermedaition”, Journal of Finance, Vol. 26(3), pp. 737-47.

Rabanal, P. (2004), “Monetary Policy Rules and the U.S. Business Cycle: Evidence and Implica-tions” IMF Working Paper No. 164.

Sargent, T.J and N. Wallace (1975), “Rational Expectations, the Optimal Monetary Instrument, and the Optimal Money Supply Rule” The Journal of Political Economy, Vol. 83, No. 2 (Apr., 1975), pp. 241-254

Sims, C. (1994), “A Simple Model for Study of the Determination of the Price Level and the Inter-action of Monetary and Fiscal Policy”, Economic Theory Vol. 43, No. 3, pp. 381-399.

Sturzenegger, F. (2004), “Tools for the Analysis of Debt Problems”, Journal of Restructuring Fi-nance, Vol. 1, No. 1, pp. 1-23.

Sturzenegger, F. and J. Zettelmeyer (2006) “Debt Defaults and Lessons from a Decade of Crises” MIT Press

Taylor, J. (1993),.” Discretion versus policy rules in practice”. Carnegie-Rochester Conference Series on Public Policy 39, pp. 195--214.

Taylor, J. B. (1999), “Monetary Policy Rules”, (Ed.), Chicago: University of Chicago Press.Uribe, M. (2006), “A Fiscal Theory of Sovereign Risk”, Journal of Monetary Economics Vol. 53,

pp. 1857-1875.van Rijckeghem, C. and B. Weder (2001), “Sources of contagion: Finance or trade?”, Journal of

International Economics, Vol. 54, pp. 293-308.van Rijckeghem, C. and B. Weder (2003), “Spillovers through banking centers: A panel data anal-

ysis of bank flows”, Journal of International Money and Finance, Vol. 22(4), pp. 483-509.Verdier, T. and C. Winograd (1998), << Privatisation de masse et macroeconomie. Un modele the-

orique d’une petite economie ouverte >> Revue économique, Vol. 49, No. 1, pp. 257-275.Woodford, M. (1994), “Monetary Policy and Price Level Determinacy in a Cash-in-Advance Econ-

omy”, Economic Theory Vol. 43, No. 3, pp. 345-380.Woodford, M. (1995), “Price Level Determinacy without Control of a Monetary Aggregate”,

Carnegie-Rochester Conference Series on Public Policy Vol. 43, pp. 1-46.Woodford, M. (2001) “Fiscal Requirements for Price Stability,” Journal of Money, Credit and

Banking, Blackwell Publishing, vol. 33, No. 3, pp. 669-728.




Reseña Bibliográfica: “El Regreso de Keynes”, de Robert Skidelsky

Alejandro D. Jacobo1

Skidelsky, Robert: El regreso de Keynes, Crítica, Barcelona, 2009, 249 páginas, (ISBN 978-84-9892-033-8).

El maestro ha vuelto con otro de sus libros: Robert Skidelsky acaba de publi-car un nuevo volumen, en el que reivindica la vigencia de la obra de Keynes como la respuesta adecuada a la actual crisis económico-financiera.

Catedrático emérito de Economía Política en la Universidad de Warwick, Miembro de la Cámara de los Lores y Fellow de la Academia Británica —entre muchas otras distinciones—, Skidelsky es mundialmente cono-cido por su magistral biografía de John Maynard Keynes (1833-1946); circunstancia que posiblemente lo convierte en el exégeta de esta figura

fundamental de la economía y el pensamiento del siglo XX; figura invocada y aclamada por unos, pero denostada y demonizada por otros; figura que, después de más de medio siglo de su muerte, continua siendo una de las más controvertidas.

“El regreso de Keynes” comienza con una pequeña reseña sobre la déringolade financiera, la que presenta lo necesario sobre la anatomía de la crisis reciente, y en la que Skidelsky se las ingenia relativamente bien con lo que él denomina el “juego de las culpas”. En este juego, se termina por inculpar a las ideas cuando resulta evidente que aquellos responsables actuaban sobre lo que creían ser unos sanos principios. Después de todo —según la conocida opinión de Keynes—, “las ideas de los economistas y los filósofos de la política, correctas o incorrectas, resultan más poderosas de lo que se suele suponer. De hecho, por poca cosa más se rige el mundo”. Así, las prácticas de los banqueros, reguladores, gobiernos, sean escandalosas o no, se remontan a las ideas de los economistas y filósofos. En razón de que —según Skidelsky— “la crisis actual es, en gran medida, el fruto del fracaso intelectual de la profesión del economista”, entonces hay que ir tras sus ideas, comenzando por las que están de moda.

Acto seguido, Skidelsky repasa el estado actual de la economía y aprovecha para explayarse sobre las tres premisas imperantes en la macroeconomía clásica del presente: la Hipótesis de las Expec-tativas Racionales, la Teoría del Ciclo Económico Real y la Teoría del Mercado Eficiente; hecho que lo conduce a señalar que la economía dominante de los últimos 30 años promovió un sistema en el que los vendavales financieros —como la crisis reciente— podrían ocurrir. De acuerdo al autor, sucedieron a partir de la creencia —errónea— de que se podría evaluar correctamente todo ries-go y que, por tanto, los mercados se autorregulaban de manera óptima. Keynes nunca hubiera aceptado la idea de que la vida económica podría reducirse a un riesgo calculable y predecible.

Aunque algunos de los capítulos del texto no presentan nada que no conozcan quienes tienen familiaridad con la biografía de las diferentes vidas de Keynes, o con aspectos de su teoría, o con la revolución y contrarrevolución keynesianas, se debe reconocer que Skidelsky es impecable en el tratamiento de estos asuntos en los que se mueve con extrema holgura. Es interesante su señalamiento: la corriente dominante de la economía actual que se mencionó —desarrollando las

1.- Pontificia Universidad Católica Argentina y Universidad Nacional de Córdoba, profesor-investigador.


144 •

matemáticas y abandonando el sentido común— está bastante alejada de la economía de Keynes y entonces se hace necesario volver a él. Y es aquí cuando Keynes, efectivamente, regresa.

El planteo de Keynes y la ética propuesto por Skidelsky obedece a que el enfoque moralista de Keynes —o por lo menos parte de su enfoque— proporciona consideraciones que han adqui-rido relevancia en el contexto de la crisis actual, por lo que no debe ser tomado como un mero discurso. Keynes, quien creía en un óptimo de Pareto-ético, previó una economía capitalista moderna gobernada por un ideal platónico y unos códigos de comportamiento correctos. Pero en la modernidad, cuando el genio capitalista sale de la botella no se lo puede poner al servicio de la ética premoderna vinculada a la buena vida y a unos códigos de comportamiento que ya no estarán. Si la buena vida, en sentido clásico, asume que el deseo humano tiene algún fin último, en la modernidad —teoría económica mediante— presupone que es insaciable.

Aún corresponde ir más lejos y hacer hincapié en que la economía de Keynes tenía un impulso filosófico en el que regía un sistema racional de ética y conducta. Cabe recordar que su énfasis en el razonamiento intuitivo en economía y su hostilidad hacia la econometría tenían un funda-mento filosófico y no eran sólo temperamentales. En todo esto cabe recordar, también, que fue el primer economista que situó la incertidumbre en el corazón del problema económico y planteó la cuestión del ámbito y significado de la racionalidad en economía. Acaso: ¿Es posible la racionali-dad en un mundo incierto? Y si lo es: ¿Cómo debe especificarse? Lo anterior es importante, pues el fracaso de la teoría económica por tomarse la incertidumbre en serio subyace a la continua sucesión de crisis financieras

Tras la decepción, quizás lo más interesante es la reconstrucción de la economía, la que, según el autor, compete a todos. Al respecto, si cualquier gran fracaso lleva a considerar las ideas funda-mentales, va de suyo que éstas están en la universidad y entonces surge una reflexión oportuna e importante que —en parte— se reproduce seguidamente.

Para equipar a los economistas a fin de que entiendan el objeto y el método adecuados de su disciplina, son necesarias dos reformas en la manera de enseñar en las universidades. Primera, “los cursos de la licenciatura en economía deben tener una base amplia. Tendrían que adoptar como lema la sentencia de Keynes de que la economía es una ciencia moral y no una ciencia natural; que el economista tiene que ser matemático, historiador, hombre de Estado y filósofo… en alguna medida, y que ninguna parte de la naturaleza humana o de sus instituciones debe caer enteramente fuera de su consideración. Las licenciaturas en economía deben contener, por tanto, no sólo los cursos estándar de macro y microeconomía (que requieren algunas matemáticas), sino historia económica y política, historia del pensamiento económico, filosofía moral, sociología y política. Aunque podría permitirse alguna especialización durante el último año del grado están-dar, hay que reducir al máximo la importancia de los matemáticos en la ponderación del grado. Esto evitaría el absurdo de que un estudiante pueda alcanzar las máximas calificaciones (matrí-cula de honor o summa cum laude) en economía simplemente por ser bueno en matemáticas”. Segunda, similar consideración cabe en los programas conducentes al Máster, particularmente en los de macroeconomía, en los que “todo lo que se enseña debe ser grados conjuntos, con un componente no económico igualmente ponderado” (p. 217).

No importa cuán de acuerdo se esté con todo esto. La opinión coincidirá en señalar que el libro resulta atractivo, conciso, de lectura amena y sencilla, y en el que Skidelsky demuestra su talento impar. La ilustración de la cubierta de la edición en español de Keynes en su domicilio londinense —seguramente el 46 de Gordon Square, en el estimulante y extravagante Bloomsbury, donde vivió desde 1916 hasta 1946— contribuye a la sobriedad que caracteriza al estilo del autor.




Reseña Bibliográfica: “Populismo o mercados. El dilema de América Latina”, de Sebastián Edwards

Eduardo Luis Fracchia11

Edwards, Sebastián: Populismo o mercados. El dilema de América Lati-na, Editorial Norma, 2009, 330 págs.

El primer capítulo del libro de Sebastian Edwards se titula “Latin Ame-rica: The Eternal Land of the Future”, anticipando una de las principa-les ideas fuerza del trabajo. Latinoamérica ha sido históricamente tierra de oportunidades, de grandes promesas, de inmenso potencial. Pero, en particular, desde la segunda mitad del siglo XX, esas promesas se convirtieron en desilusiones y fracasos. La reciente crisis internacional ha puesto en evidencia que se inicia un inusitado período de auge de los emergentes. Latinoamérica vuelve a presentarse como una región prometedora. Es oportuno por tanto reflexionar en las causas de los fracasos pasados, para no desaprovechar esta ocasión histórica.

El libro repasa la evolución económica de largo plazo de Latinoamérica enfatizando la debilidad institucional, las recurrentes crisis y el drama de la pobreza y la desigualdad en la distribución del ingreso. La Guerra Fría halló a la región en medio de la puja entre las potencias hegemónicas. La revolución cubana impulsó en cierto modo el diseño de la Alianza para el Progreso. En un marco de efervescencia ideológica, los conflictos internos se radicalizaron y las democracias se convirtieron en la excepción más que la regla. A la inestabilidad política se sumaron las ineficien-cias asociadas con el proteccionismo y los excesos en materia fiscal y monetaria que condujeron a sucesivas crisis de balanza de pagos y a instalar a la inflación como un problema cotidiano. La década de los ochenta, “la década perdida”, representó el punto álgido de esta decadencia con las crisis masivas de deuda.

Llegó así la hora del Consenso de Washington. La región se embarcó entonces en programas comprometidos de reformas de mercado con fijación del tipo de cambio, privatizaciones y reduc-ción del tamaño del sector público. Se le ganó la batalla a la inflación. Pero apareció un nuevo problema, el alto desempleo. Asimismo los países de la región quedaron muy expuestos a los cambios de humor de los mercados financieros. Las crisis mexicana, rusa, brasileña, del sudeste asiático, argentina y uruguaya lo evidencian. En suma, las reformas de mercado habían quedado inconclusas porque no se había avanzado en la consolidación de las instituciones y en el respeto a la ley y los derechos de propiedad.

Un ejemplo distinto lo ofrece Chile. La dictadura de Augusto Pinochet encaró reformas de mer-cado a cargo de un grupo de economistas a los que se apodó “Chicago Boys” por su formación académica. La liberalización de la economía permitió la inserción internacional del país sobre la base de la competitividad y la eficiencia. Con el retorno de la democracia, la Concertación de centro-izquierda que se hizo con el poder mantuvo y fortaleció las políticas de Estado abordando con pragmatismo el manejo de la economía. Chile demostró que es posible crecer y avanzar en el combate de la pobreza sobre una base de estabilidad, de seguridad jurídica y con un sector público prudente.

1.- Área de Economía IAE, Director.


146 •

Las situaciones traumáticas producidas por las reformas inconclusas de los noventa sembraron la semilla del populismo encarnado, por ejemplo, en el gobierno de Hugo Chávez en Venezuela. Gracias al boom del precio del petróleo, Chávez vio ampliadas sus posibilidades económicas y sobre ellas basó su discurso antiimperialista y nacionalista. El Estado venezolano fue creciendo en magnitud, funciones, ineficiencia y corrupción. Nacionalizaciones, mercado de dólares dual, des-pilfarro de recursos. Sus políticas económicas atrasan tanto como su discurso. Como resultado, la economía venezolana se halla en un escenario de profunda ineficiencia, altamente dependiente del petróleo y la sociedad radicalizada.

En contraposición, la administración de Lula de Silva en Brasil representa la antítesis del gobierno chavista. A pesar de los temores iniciales, Lula no desechó las importantes reformas desarrolladas por sus predecesores. Fortaleció las políticas de Estado, apuntando al combate de la inflación y a la sustentabilidad fiscal reconociendo que ninguna política distributiva tendría éxito sin antes alcanzar la estabilidad macroeconómica. El pragmatismo de Lula logró posicionar a Brasil en el escenario internacional como uno de los líderes de las naciones emergentes, conjugando los objetivos de crecimiento con equidad. Queda mucho trabajo por hacer en Brasil, pero ya brilla con luz propia.

El desafío es entonces tomar los ejemplos chileno y brasileño, en lugar de caer en el populismo venezolano, condenado al fracaso en una etapa que vuelve a ofrecer múltiples oportunidades a Latinoamérica. Fortalecer las instituciones, combatir la corrupción y la discrecionalidad, abrirse al mundo e implementar las reformas de mercado necesarias pero con objetivos honestos y apli-cación prudencial. Algunas naciones ya caminan este sendero, otras aún dudan. El desafío de aprovechar esta oportunidad histórica queda planteado.

El libro está escrito por un economista con mucha cercanía al terreno y a la vez con un conoci-miento teórico de típicos problemas de la región como crisis bancarias o cambiarias. Evidente-mente la región no ha seguido un escenario de convergencia como en el caso del sudeste asiático y está muy atrás en los rankings de competitividad de mayor difusión, sin embargo existen funda-mentos en la macroeconomía de la mayoría de los países que sumados a un contexto institucio-nal más previsible dan signos esperanzadores para el crecimiento. Siempre está al acecho según Edwards la tentación del populismo, en mayor medida cuando hay ciclos de precios favorables a los commodities. Esta divergencia en la región llama la atención y plantea un desafío de madurez democrática. Es un texto que desde la historia económica de políticas públicas aplicadas anima a construir una agenda de desarrollo para la región o para los países en particular.




Presentación y selección de trabajos

A. Información general

La “Revista Ensayos de Política Económica” es una publicación de la Facultad de Ciencias Eco-nómicas de la Universidad Católica Argentina. Su primer número fue publicado en 2007, y tiene una periodicidad anual. La Revista esta abierta a distintas colaboraciones, en el ámbito de la política económica y social cubriendo áreas como ser macroeconomía, economía internacional, economía monetaria, economía financiera, politicas fiscales, crecimiento, desarrollo, instituciones, regulaciones, organización industrial, politicas sociales, mercado de trabajo, demografía y distri-bución del ingreso, entre otros.

La selección de los trabajos publicados en la “Revista Ensayos de Política Económica” se encuentra bajo la responsabilidad de un Consejo Editorial, integrado por reconocidos investigadores de la Universidad y de otros Centros de Estudios. El Consejo Editorial cumple la función de realizar referatos anónimos de los trabajos presentados.

B. Instrucciones para la presentación de los trabajos

Los trabajos con pedido de publicación deben ser enviados en versión digital vía email a:

[email protected] o [email protected]

Los considerandos de los Ensayos son los siguientes:

1.- Se consideran trabajos tanto en idioma español como en idioma inglés.

2.- El manuscrito será escrito de la siguiente manera:

En primera hoja:

Título del artículo

Nombre y Apellido del autor

Resumen (en español e inglés) Caracteres con espacio, máximo 700.

Palabras clave.

Clasificación JEL.

Dirección de contacto: con asterisco (*) en el nombre del autor, al final de la hoja, pertenencia institucional, dirección, teléfono, email.

Extensión máxima: 35 páginas.

Caracteres con espacio: 50.000 (máximo).


148 •

Letra: Verdana 10.

Interlineado: sencillo.

Gráficos y cuadros: sin restricción, incluidos dentro de las 35 páginas.

Bibliografía

Al final del trabajo, incluida dentro de las 35 páginas, ordenada alfabéticamente por apellido y nombre del autor:

Artículos: nombre del artículo entre comillas, seguido del nombre de la publicación donde esta incluido en itálica, volumen, número, año y páginas

Libros: nombre del libro en itálica, lugar de edición, editorial y año.Ejemplos:

R. Dornbusch; S. Fischer; P. A. Samuelson, “Comparative Advantage, Trade, and Payments in a Ricardian Model with a Continuum of Goods”, The American Economic Review, Vol. 67, N°. 5, Dec. 1977 , pp. 823-839

Aghion, Philippe; Howitt, Peter, Endogenous Growth Theory, Cambridge, Massachusetts, The MIT Press, 1998

3.- No se admiten modificaciones ni agregados una vez iniciado el proceso de edición.

4.- La Revista no se hace responsable por trabajos originales no publicados ni por su devolución en caso de no ser solicitados.


ensayos de política económica - año...

Documents