abstract - university of oxford
TRANSCRIPT
Regional inequality in Tunisia:
Is it a geographic misallocation or spatial nature effects?
Jmaii Amal1
Abstract
Using detailed household-expenditure level data from Tunisia, we examines urban-rural inequality in
Tunisia. We use Bayesian regression-based decomposition method to dissect the gap between the two
areas. Results of the present study suggest that the difference between rural poor households and urban
poor households is due essentially to characteristic effects; while for wealthier households both
characteristic and returns to these characteristic effects (for example, efficiency of educational system)
are in charge of this gap. Additionally, the results demonstrate that this is an issue of value, and, more
specifically, an economic development fairness conflict. It is suggested that policy makers should
address a positive discrimination programs in favor of marginalized region.
Keywords: regional inequalities; Bayesian unconditional quantile regression; RIF decomposition;
micro data, Tunisia.
JEL Classification: I31, I24, P25, C14
1 PhD economics sciences, university of Tunis el Manar, Tunis
I. Introduction
“I have cherished the idea of a democratic and free society in which all persons will live
together in harmony and with equal opportunities, ... It is an ideal for which I hope to live
and to see realized ... if needs be, it is an ideal for which I am prepared to die.”
-Nelson Mandela –
Inequality is no longer seen as an inevitable prerequisite for growth, but rather as an obstacle to
both economic growth and poverty reduction (Bourguignon, 2003). A fundamental question in
the field of regional inequality and development is why some regions are rich and others
vulnerable/poor. One explanation is that differences arise from spatial concentrations of
households/individuals with personal characteristics inhibiting growth in their living standards.
Indeed, inequality can efficiently refer to the overall trend of regional development; it does not
completely explain whether certain households are different in their welfare due to geographic
localization, or due to high social/economic disparities within the regions. In such a case,
decompose inequality, which divide aggregate inequality to analyze between-regions
variability, is required for a positive assessment of inequality and for the purposes of effective
social policies. In a quantitative investigation of the impact of changing in economic structure
and urbanization on inequality in Asia, Kanbur and Zhuang (2013) show that economic
structure and urbanization have a positive and significant effect in inequality. Chambers and
Krause (2010) used semiparametric methods to analyze the relationship between income
inequality and economic growth. The authors found that higher income inequality is harmful
for economic growth. Son and Kakwani (2008) found that initial levels of income inequality
and economic development have a positive and significant influence on poverty reduction. In a
study based on a new data set on inequality in the distribution of income, Deininge and Squire
(1996) revealed a systematic link between growth and aggregate inequality variation in parallel
with a significant positive relationship between growth and poverty alleviation.
There has been a wide proliferating literature on inequality decomposition measurement
between regions and its application to micro-data surveys in the developing countries context.
Several studies have been published addressing the source of inequality between rural and urban
households. Nguyen et al. (2007) discuss the welfare inequality between urban and rural areas
from 1993 to 1998 in Vietnam using Machado-Mata decomposition (2005). In this study, they
conclude that inequality differences between the two regions were due to three factors namely
education, ethnicity, and age. Based on a dataset from 65 countries, Young (2013) found that
the gap represents 40% of the average country inequality. In addition, Sicular et al. (2007) used
data from two years respectively 1995 and 2002, to analyze the rural-urban income inequalities
in China. Studying deeply persistent gap between rural and urban sectors can put more emphasis
on the principal factors contributing to this gap and thus improve the decisions required to be
undertaken by policymakers to support regional development.
Poverty and regional inequalities contributed to the rise of the revolution in Tunisia. The
increase in per capita income underestimated persistent regional inequality and poverty,
misrepresenting the situation in the interior where the revolution began. The concentration of
economic activities investment and services in the east regions has increased poverty and
unemployment rate in the west, particularly for youth (Gana, 2012; Boughzala, 2013; Arieff, 2012).
In the spirit of this debate, a legitimate question arises as whether regional inequality in Tunisia
reflects a geographic misallocation problem/a bad government decisions or it’s rather about the
structure and nature of some regions.
To response to this question, this study is organized as follows. First, we began by a Tunisian
background information. Section three presents the data and the model. Section four provides
the results of the proposed methodology. Finally, we conclude and give some policies
recommendations.
II. Background: Tunisia five years after the revolution
1. Spatial inequalities: History and evolution
Income distribution is an important indicator for analyzing poverty and economic
development in a country. A better understanding of the pattern and drivers of regional
inequality is critical for enhancing social cohesion and inclusive growth in the region.
Considerable work has been undertaken on regional inequalities related to developing countries.
In a dynamic analysis of the patterns of household welfare in Jordan, Mansour (2012)
highlighted a slight decline in inequality during 2002–10 mostly driven by a regional catching-
up effect. Laithy, Abu-Ismail, and Hamdan (2008) found that despite the moderate national
levels of poverty and inequality in Lebanon, there are striking regional income disparities. In
an analysis of the urban-rural gap in North African countries, Boutayeb and Helmert (2011)
show that, these countries has experienced a considerable development in social, economic and
health indicators. Unfortunately, regions/socio-economic groups of the same country have not
benefited by these improvements equally. Bibi and Nabli (2009) found that MENA region has
relatively a higher level of income/expenditure inequality, compared to other regions. Adams
and Page (2003) revealed that compared to other regions, the MENA has a lower income
inequality and poverty rate due to public sector employment and international
migration/remittances. Regional gap in education has been also of concern in some studies on
Arab countries. In an empirical analysis of inequality of education opportunities in MENA,
Salehi-Isfahani, Belhaj Hassine, and Assaad (2013) show that, inequality in educational
achievements is mainly explained by inequality of opportunities. Likewise, Krafft and Assaad
(2016) highlighted that inequality of opportunity- unequal resources allocations on the basis of
circumstances independent from individuals’ control- can led offends people’s sense of fairness,
causing anger and frustration among marginalized groups.
Tunisia, as a developing country, has implemented since the early years of statehood,
several national programs to reduce poverty and regional disparities. The country has indeed
shown some progress in many areas. The per capita income has raised, public services have
developed, and health indicators have improved and the demographic trends have been
favorable. In fact, the country has opted for an open trade policy since the beginning of the
1990s. This strategy was preceded by the implementation of a Structural Adjustment Plan.
While this policy has led, over the last twenty years, to an increase in growth and income, the
gains from free trade are not equally distributed within the. Economic restructuring,
globalization of capital markets and structural adjustment are synonymous of drawdown of the
permanent workers number, the subcontracting with resort to temporary and seasonal work and
the reduction of costs through deregulation of the labor market.
The deterioration of living conditions among rural population, the increase in the
unemployment rate and the inflation in the consumer prices for basic goods associated to
climate change and water scarcity problems, all threatened the sustainability of growth in these
areas, which has provoked the uprising of the people in the interior regions leading to the 2011
Tunisian revolution. Indeed, after the revolution, Tunisia is boldly emerging from the recession
considered as the longest period of economic downturn ever since the establishment of
statehood. In this paper, we emphasized the failure of agricultural, trade and social policies and
the insufficiency of the measures taken to deal with the vulnerability of the rural population.
More than 800,000 people were unemployed in 2014 and the poverty rate reached 15%
according to data released by the INS. Poverty rate is correlated with many factors such as
unemployment rate, illiteracy, access to health services and especially education (African
Development Bank, 2013). Therefore, poverty in Tunisia is mainly concentrated in rural areas
and in some regions of the country, particularly the Central West, a region that first witnessed
the first sparkles of the revolution. In addition, the lack of an infrastructure in these areas, such
as roads and communication facilities, may limit poor people’s access to information or to the
labor markets.
A strong variation in poverty rates between regions may be the cause of social instability and
population movement. Households with higher level of poverty rate are more concentrated in
the interior regions of the country than in the inland ones. Indeed the measurement of poverty
at the regional level allows a better definition of the priorities for regional development. In fact,
the decrease in the poverty rate compared to the higher consumption disparities with economic
inequalities assert that the GDP growth is biased towards the non-poor.
2. Inequality in the Tunisian post-revolution context
Regional inequality is a driving factor behind Tunisia’s economic/political crisis. So long as
the west Tunisia remains ignored compared to its more-developed east, discontent and unrest
will plague its political and economic recovery. Statistics show that during the past two
decades, poverty rates have declined and the overall economic situation has improved.
However, large parts of the country have been neglected and as a result, regional disparities
has been exacerbated. For example, the gap in poverty rates between the capital and the rest of
the country shows that the regional variation in terms of living standards increased between
2010 and 2015 (figure 1).
Up to now, the adopted economic and social development does not correspond to good regional
governance objectives that Tunisia has to achieve. Hence, the disappointment expressed by of
many Tunisians as they had better expectations after the revolution. Many Tunisians has
expressed feelings of distrust towards public policies that devoted little regard for regional
inequalities.
2010 2015
Fig 1: Poverty rate of Tunisia 2010 vs 2015
III. Methodology
1. Data
This paper uses the consumption expenditure survey (CES)2, this survey is conducted by the INS every
five years and provides socio-demographic and economic characteristics of households and individuals.
Indeed, for 2010, it takes a representative sample of 11281 households with 50371 individuals. The data
includes information concerning the head of the household, composition; area of household’s location
and total household’s consumption expenditures. The choice of explanatory variables is based on the
literature (Nguyen et al. (2007); Sicular et al. (2007); Skoufias and Katayam , 2011) and is validated by
the AIC and BIC criteria1. Therefore, we use household size, the proportion of children under 15 years
2 The ideal is if we can made comparative analysis between household situation of 2010 and 2015. Unfortunately, the last household survey of consumption and expenditure, Tunisia 2015 – micro data – is not yet available for researcher.
old in each household and the gender of the household head and the age. As far as the household
education and employment characteristics, we have included the variable of the schooling of household
head: illiterate (as reference), primary, secondary and higher level. For employment variable, we select
four sectors respectively governmental sector (as reference), private sector, self-employed and
agricultural sector. Since the frequency of transfers from foreign countries, we use a dummy variable
indicating depending on whether a household had received transfers from abroad or not.
2. Choice of the model
The principal question this paper tries to answer is whether spatial factors has any effect on
well-being controlling for observable non-geographic characteristics of the households and
whether access to public and private assets compensates for the effects of an adverse
geography. To address this question, we have used Bayesian regression-based decomposition
method.
a) Bayesian unconditional quantile regression
Introduced by Koenker and Bassett (1978), quantile regression models aim at modeling the
effect of the explanatory variables on the conditional distribution of the outcome variable. They
have been increasingly used in empirical labor market studies, to describe parsimoniously the
entire wage conditional distribution (see e.g. Buchinsky 1994, Chamberlain 1994, Machado and
Mata 2001). While the quantile regression models are useful, they have some restrictive.
b) Recentered Influence Function (RIF)
RIF-regression (Fortin, Lemieux, and Firpo, 2011) is a convenient tool to conduct OB type
decomposition for other method besides the mean such as quantiles regression. Therefore, when
we perform quantile model, RIF-regression will be considered as rescaled linear model.
According to this definition, rescaling factor relies on the estimate of the interest quantile
density:
𝑅𝐼𝐹(𝑦, 𝑄𝑞) = 𝑄𝑞 +𝑞 − 1{𝑦 ≤ 𝑄𝑞}
𝑓𝑦(𝑄𝑞)
The distributional statistic can be written based on the conditional expectations of its recentered
influence function: We, then, can perform an OB decomposition using the RIF as response
variable (dependent variable).
IV. Empirical Results 1. Source of rural-urban inequality
When the models are estimated separately for rural and urban areas, it is possible to see the
differences in the factors affecting consumption expenditures. The rural variable was added (as
a dummy variable) in the pooled model, and its coefficient is statistically significant at a level
of 1% with a negative sign. Most variables of both rural and urban models are significant. The
quantile regression results for both rural and urban areas are given in the table 4 and table 5.
[Place table 1 here]
The finding demonstrates that gender variable is statistically significant in the urban model for
the means regression as well as for the quantile estimates. From table 4, we reveal that men’s
expenditures are consistently lower than those of women, but these differences are much greater
in the upper quantiles. We note that this variable is not significant in the 10th percentile.
Although, for the rural model this variable is not significant in the mean and quantile regression,
except in the 25thand 50thpercentiles.
[Place table 2 here]
Results also show that the variable “size of the household” is found to be significant. Ceteris
paribus, with an unit change in the household size, the level of expenditure of household’s
declines by approximately 23.2% for urban areas against 18.4% for rural areas. The breakdown
by age of household heads leads to interesting results in term of social policies. In general,
young people are more exposed to poverty compared to other categories, in both means and
quantiles regression. When looking for the class-age of 31-45 we observe similar results
between rural and urban areas. In most quantiles, this category is not significant. This result
confirms many facts related to the economic/social environment in Tunisia. The increase in the
unemployment rate, in rural area, mainly due to climate change (drought) associated with an
absence of labor market, forced people to internal migration to urban areas. In parallel, we
witness the urbanization of poverty and unemployment caused by the increasing number of
graduate and the inability of the labor market to absorb them. In addition, results show that
categories of age 76 and above have no significant effects or have negative significant effects
of welfare in rural area compared to younger category and compared to urban subgroup. In fact,
for Tunisian case, the rural population is not homogeneous. There is a two-pronged migratory
movement: the departure of skilled young people and the arrival of elderly close to retirement
(Jelili and Mzali, 1998). This phenomenon of older age groups having considerable within-
cohort inequality is an important acknowledged fact and can be clearly observed in the Tunisian
data. These results confirm the urgency of the issue of rural old-age vulnerability in Tunisia.
[Place table 3 here]
Results also show that the agricultural sector has a positively significant coefficient in rural area
at all quantiles excluding the 25thand the 90th. For both urban and rural areas, welfare is
determined by governmental and private sectors, since they are statistically significant with a
positive sign. In the urban model, the effect of these variables is higher in the upper quantile. In
the rural regression, they are much higher at the bottom of the distribution. As indicated in
Tables 5 and 6, returns to education are statistically significant across all quantiles. The findings
show that, for entire distribution, consumption expenditures for people with higher education
levels are higher than illiterate people and those who have just acquired primary education and
secondary education level. In addition, results show that west region is more vulnerable and
have lower welfare than East region. In fact, employments are paying off in the East relatively
to the West and probably for the same reasons: the East gets the lion's share of market’s activity
in Tunisia (manufacturing and services).
2. Counterfactual decomposition results
the overall poverty gap between poor rural and poor urban households may be explained by
the constant term (table 4) of the coefficient effect (14.8%). This means that, even with urban
characteristics, poor rural households will still suffer from a poverty gap of around 9.36%
compared to urban households. This is due to the geography effect, which assigns persistent
spatial differences in living standards to difference in geographic characteristics such as
roads/transport, water, etc.
[Place table 4 here]
Results reveal that differences in the type of employment and education explain respectively
4.58% and 18.32% of the poverty gap between rural and urban areas for the poorest
households (Table 4). Similarly, if the conditions of the labor market in rural areas were
similar to those in the urban areas, the poverty gap could be reduced by 3 points.
Results of the decomposition method revealed that both covariate and returns effects are larger
within higher quantiles, resulting in a wider urban–rural gap. This means that, the principal
reasons of the difference between rural wealthiest households and urban wealthiest households
were twofold: the first one, due to characteristic effect such as education, employment and
household size; the second reason for this difference, is due to the return effect of these
characteristics. Apart from this result, we observe a dominance of characteristics effects at
lower quantile, which attest that differences in characteristics between urban poor households
and rural poor households are more important than differences in returns to those characteristics
(figure 2). This may reflects that the poor earns a wage slightly above the minimum level,
therefore urban-rural change in market yields does not matter at the tail of the distribution.
[Place figure 2 here]
3. Conclusion and recommendations
We start by this study by posing the following question: what about regional inequalities in
Tunisia, is it a spatial misallocation or bad choices? To response to this question we used
Bayesian semi-parametric regression based decomposition method.
objective of this study was to dissect regional inequalities in Tunisia. For this purpose, we used
semi-parametric regression based decomposition methods to assess the source of welfare gap
between urban and rural areas. Results of the study are in agreement with the type of policies
that have been established in Tunisia. For several decades, the government has implemented
reforms that promote education in the rural area especially the west region. Nevertheless, it did
not take into consideration the quality of the education program, as a result educated youth, in
these areas, do not have the capacity to succeed in their national competition and are unable to
compete with other graduates in the private market. Indeed, the lack of a good educational level
may limit the opportunities for these individuals to find a decent job. This may explain the fact
that the gap between the two areas is still deep. The issue of equitable access to education,
especially to the higher level, is probably a real issue that must be addressed as part of an overall
review. In other hand, the breakdown by age of household heads has confirmed the urgency of
the issue of rural old-age vulnerability in Tunisia. Effective policies must take into consideration
the heterogeneous nature of elderly-group, associating both personal characteristics (sex,
education...) and the varying contexts in which they live.
The fight against regional inequalities is not just a matter of equality, but also a question of
economic development fairness. We should know that the development of social groups and
deprived regions inevitably involves a strategy to reduce regional inequality. In addition, the
Tunisian state is not only required to build universities and hospitals (etc.), but also to provide
the needed resources to get the rural areas to the level of the urban ones. We recommend a
positive discrimination programs in favor of marginalized region. For example, the government
can start by sending most competent teachers and doctors to work in these areas, and provide
the necessary resources (good infrastructure, access to quality schools, internet, software ...).
By decomposing the gap between urban and rural area, this study has shown that the problem
is about not only equality but also it is an equity issue. Such equity implies that resources should
be allocated equally between areas with regard to the quality. Although, for a social
phenomenon, such as inequality, a dynamic analysis is required. This type of inequality is often
known as income variation/mobility and poverty dynamics and is important, when viewed from
the social welfare and political perspectives, to understand policy changes, an issue left for
future research.
References
Arieff, A. (2012). Political transition in Tunisia. Current Politics and Economics of Africa
5(2), 278.
Boughzala, M. (2013). Youth employment and economic transition in Tunisia. Brookings
Institution.
Bourguignon, F. (2003). The growth elasticity of poverty reduction: explaining heterogeneity
across countries and time periods. In T. Eicher, and S. Turnovsky (Eds.), Inequality and
Growth: Theory and Policy Implications. Cambridge, MA: Massachusetts Institute of
Technology (MIT) Press.
Fortin N., Lemieux T. and Firpo S. (2011) Decomposition Methods in Economics in
Ashfenfelter O.
and Card D. (eds.) Handbook of labor economics, volume 4A, Amsterdam, Elsevier: 1-
122.
Gana, A. (2012). The rural and agricultural roots of the Tunisian revolution: When food
security
matters. International Journal of Sociology of Agriculture and Food, 13-p.
Kanbur, R., and Zhuang, J. (2013). Urbanization and inequality in Asia. Asian Development
Review.
30(1): 131-147.
Koenker R.and Bassett G. (1978). Regression Quantiles, Econometrica, 46(1): 33 - 50.
Melly B. (2005). Decomposition of Difference in Distribution using Quantile Regression,
Labour Economics, 12(4): 577 - 590.
Mansour, W. (2012). The patterns and determinants of household welfare growth in Jordan
2002–2010. The World Bank Policy Research Working Paper 6249.
Miller P.W. (1987).The Wage Effect of the Occupational Segregation of Women in Britain,
Economic Journal, 97(388): 885-896.
Neuman S. and Oaxaca R. (2004) Wage Decompositions with Selectivity-Corrected Wage
Equations: A Methodological Note, Journal of Economic Inequality, 2(1): 3-10.
Nguyen B.T., Albrecht J.W., Vroman S.B. and Westbrook M.D. (2007). A quantile
regression decomposition of urban-rural inequality in Vietnam, Journal of Development
Economics, 83(2): 466-490.
Oaxaca R. (1973). Male Female Wage Differentials in Urban Labour Markets, International
Economic Review 14: 693–709.
Parente, P. M., & Santos Silva, J. (2016). Quantile regression with clustered data. Journal of
Econometric Methods, 5(1), 1-15.
Salehi-Isfahani, D., Belhaj Hassine, N., & Assaad, R. (2013). Equality of opportunity in
educational achievement in the Middle East and North Africa. Journal of Economic Inequality.
http://dx.doi.org/10.1007/ s10888-013-9263-6.
Sicular, T., Y. Ximing, B. Gustafsson, and Shi L. (2007). The urban- rural income gap and
inequality in china. Review of Income and Wealth 53 (1), 93-126.
Skoufias E. and Katayam R.S. (2011) Sources of welfare disparities between and within
regions of Brazil: evidence from the 2002-2003 household budget survey (POF), Journal of
Economic Geography, 11: 897-918.
Son, H. H., and Kakwani, N. (2008). Global estimates of pro-poor growth. World
Development, 36(6), 1048-1066.
Young, A. (2013). Inequality, the urban-rural gap, and migration. The Quarterly Journal of
Economics, 128(4), 1727-1785.
TABLE 1 - OLS and Quantile Regression of Total Consumption Expenditure Per Capita
Variables Means
Quantile regression
5th percentile
10th percentile
25th percentile
50th percentile
75th Percentile
90th Percentile
95th percentile
Intercept 15.109* (0.085)
13.56* (0.166)
13.72* (0.197)
13.99* (0.085)
14.31* (0.129)
14.96* (0.148)
15.48* (0.204)
15.79* (0.524)
Gender -0.106* (0.015)
-0.025 (0.030)
-0.033 (0.026)
-0.046** (0.020)
-0.098* (0.014)
-0.164* (0.020)
-0.204* (0.032)
-0.134* (0.094)
household size -0.212* (0.003)
-0.224* (0.0059)
-0.229* (0.006)
-0.221* (0.0036)
-0.219* (0.004)
-0.209* (0.005)
-0.202* (0.007)
-0.176* (0.018)
% children (under 15 old)
-0.674* (0.029)
-0.522* (0.056)
-0.525* (0.059)
-0.609* (0.025)
-0.664* (0.024)
-0.761* (0.039)
-0.809* (0.050)
-0.843* (0.178)
Age 0.0144* (0.0027)
0.016* (0.005)
0.017* (0.006)
0.018* (0.0031)
0.020* (0.004)
0.007** (0.004)
0.001 (0.006)
0.004 (0.016)
Age2*100 -0.009* (0.002)
-0.011* (0.004)
-0.012** (0.005)
-0.01* (0.002)
-0.01* (0.003)
-0.002 (0.003)
0.003 (0.006)
0.005 (0.014)
Tenant of house
Free housing
-0.059* (0.019) -0.131* (0.025)
-0.084** (0.037) -0.133* (0.049)
-0.092* (0.028)
-0.114** (0.056)
-0.075* (0.023) -0.116* (0.027)
-0.084* (0.023) -0.140* (0.029)
-0.046** (0.021) -0.106* (0.024)
-0.013 (0.030) -0.180* (0.039)
0.082 (0.117) 0.007* (0.157)
Primary
Secondary level
Higher level
0.180* (0.014) 0.455* (0.016) 0.876* (0.024)
0.196* (0.027) 0.444* (0.032) 0.846* (0.048)
0.187* (0.019) 0.442* (0.024) 0.868* (0.044)
0.178* (0.018) 0.424* (0.021) 0.866* (0.019)
0.184* (0.018) 0.458* (0.019) 0.868* (0.026)
0.187* (0.023) 0.476* (0.023) 0.890* (0.047)
0.166* (0.025) 0.474* (0.042) 0.884* (0.051)
0.124* (0.086) 0.529* (0.101) 1.039* (0.152)
sector_Gov 0.104* (0.017)
0.145* (0.034)
0.138* (0.028)
0.128* (0.025)
0.103* (0.027)
0.099* (0.022)
0.085** (0.036)
0.107 (0.107)
sector_Priv
0.116* (0.020)
0.162* (0.039)
0.146* (0.023)
0.139* (0.021)
0.111* (0.019)
0.103* (0.025)
0.103** (0.045)
0.112* (0.122)
sector_Agri 0.014 (0.016)
0.050 (0.032)
0.036 (0.023)
0.030 (0.020)
0.005 (0.020)
0.017 (0.021)
-0.029 (0.030)
-0.051 (0.101)
self 0.072* (0.015)
0.076** (0.030)
0.082* (0.025)
0.085* (0.016)
0.065* (0.016)
0.097* (0.024)
0.069** (0.029)
0.126 (0.094)
West -0.274* (0.0108)
-0.305* (0.021)
-0.308* (0.018)
-0.299* (0.014)
-0.276* (0.015)
-0.254* (0.012)
-0.242* (0.023)
-0.109** (0.066)
Rural -0.266* (0.012)
-0.331* (0.049)
-0.293* (0.019)
-0.270* (0.012)
-0.261* (0.018)
-0.231* (0.016)
-0.218* (0.028)
-0.246* (0.073)
Foreign Transfer 0.188* (0.041)
0.134** (0.080)
0.259* (0.082)
0.179* (0.035)
0.160** (0.071)
0.183** (0.088)
0.265* (0.082)
0.237 (0.254)
Number of observation Pseudo R2
11281
0.572
11281
0.358
11281
0.356
11281
0.359
11281
0.359
11281
0.352
11281
0.336
11281
0.288
Chow Test [K, N-2*K]=
104.70a
Quantile regression and OLS estimates of total observation, with standard deviation in parentheses.
Significance levels are respectively 1% (*) and 5% (**)
a: p-value > F(10, 11281) = 0.000
TABLE 2 - Quantile Regression of the Urban Consumption Expenditure Per Capita
Quantile Regression Estimates Of Urban Observation, With Standard Deviation In Parenthesis.
Significance Levels Are Respectively 1% (*) And 5% (**).
Urban
Variables 5th Percentile
10th Percentile
25th Percentile
50th Percentile
75th Percentile
90th Percentile
95th Percentile
Intercept 13.97* (0.201)
14.13* (0.247)
14.34* (0.125)
14.81* (0.131)
15.41* (0.140)
15.69* (0.215)
16.31* (0.819)
Men -0.036 (0.036)
-0.047** (0.021)
-0.060* (0.022)
-0.111* (0.031)
-0.171* (0.022)
-0.149* (0.041)
-0.037 (0.149)
Household Size -0.244* (0.007)
-0.241* (0.007)
-0.235* (0.005)
-0.240* (0.005)
-0.235* (0.005)
-0.228* (0.008)
0.008 (0.018)
% Of Children (Under 15 Years)
-0.425* (0.068)
-0.406* (0.073)
-0.505* (0.063)
-0.557* (0.051)
-0.613* (0.046)
-0.674* (0.073)
-0.081 (0.123)
Age 0.015** (0.006)
0.016** (0.007)
0.019* (0.004)
0.017* (0.0041)
0.009** (0.005)
0.0002 (0.007)
-0.002 (0.012)
Age2*100 -0.008 (0.005
-0.01 (0.006)
-0.01* (0.003)
-0.01* (0.003)
-0.002 (0.004)
0.002 (0.006)
0.003 (0.011)
Homeowner (Ref) Tenant Free Housing
-0.101* (0.037) -0.126** (0.058)
-0.098* (0.021) -0.137* (0.039)
-0.092** (0.027) -0.136* (0.033)
-0.094* (0.021) -0.151** (0.061)
-0.053** (0.025) -0.103* (0.038)
0.004 (0.038) -0.05 (0.057)
0.109 (0.089) 0.167 (0.145)
Illiterate (Ref) Primary Level Secondary Level Higher Level
0.233* (0.035) 0.486* (0.038) 0.913* (0.051)
0.205* (0.031) 0.465* (0.032) 0.877* (0.046)
0.230* (0.022) 0.479* (0.025) 0.907* (0.039)
0.209* (0.024) 0.510* (0.022) 0.863* (0.031)
0.218* (0.023) 0.524* (0.028) 0.913* (0.030)
0.17* (0.039) 0.432* (0.04) 0.835* (0.053)
0.003 (0.078) 0.002 (0.086) 0.038 (0.087)
Sector_Gov 0.142* (0.038)
0.115* (0.025)
0.102* (0.025)
0.089* (0.019)
0.072* (0.019)
0.307* (0.04)
-0.088 (0.155)
Sector_Priv 0.184* (0.043)
0.126* (0.035)
0.115* (0.024)
0.088* (0.023)
0.084* (0.029)
0.313* (0.046)
0.017 (0.178)
Sector Agri 0.013 (0.057)
-0.049 (0.049)
-0.056 (0.048)
-0.036 (0.043)
-0.042 (0.038)
-0.028 (0.073)
-0.299 (0.231)
Self 0.070** (0.034)
0.031 (0.025)
0.030 (0.024)
0.0208 (0.023)
0.036** (0.019)
0.011 (0.033)
0.049 (0.138)
Center-West -0.334* (0.025)
-0.299* (0.014)
-0.303* (0.019)
-0.287* (0.014)
-0.265* (0.020)
-0.246* (0.039)
0.022 (0.105)
Foreign Transfer 0.114 (0.092)
0.246** (0.123)
0.183* (0.056)
0.123** (0.071)
0.081 (0.059)
0.061 (0.096)
-0.329* (0.177)
Number Of Observation Pseudo R2
7261 0.315
7261 0.316
7261 0.330
7261 0.343
7261 0.340
7261 0.330
7261 0.254
15
TABLE 3 - Quantile Regression of the Rural Consumption Expenditure Per Capita
Quantile Regression Estimates Of Rural Observation, With Standard Deviation In Parenthesis.
Significance Levels Are Respectively 1% (*) And 5% (**).
Rural
Variables 5th Percentile
10th Percentile
25th Percentile
50th Percentile
75th Percentile
90th Percentile
95th Percentile
Intercept 13.40* (0.331)
13.92* (0.308)
14.24* (0.203)
14.46* (0.190)
15.19* (0.180)
15.52* (0.228)
15.88* (0.462)
Men -0.026 (0.064)
-0.053 (0.055)
-0.037 (0.036)
-0.106** (0.043)
-0.115* (0.033)
-0.187** (0.074)
-0.084 (0.162)
Household Size -0.203* (0.011)
-0.205* (0.013)
-0.197* (0.007)
-0.193* (0.006)
-0.176* (0.006)
-0.175* (0.007)
0.020 (0.027)
% Of Children (Under 15 Years)
-0.684* (0.122)
-0.724* (0.092)
-0.804* (0.053)
-0.824* (0.059)
-0.962* (0.054)
-0.903* (0.074)
0.289 (0.251)
Age 0.024** (0.010)
0.011 (0.010)
0.017** (0.006)
0.015* (0.005)
0.001 (0.005)
0.004 (0.005)
-0.007 (0.027)
Age2*100 -0.022** (0.009)
-0.009 (0.009)
-0.01** (0.005)
-0.01** (0.004)
-0.0624 (0.005)
-0.003 (0.005)
0.003 (0.023)
Homeowner (Ref) Tenant Free Housing
0.058
(0.212) -0.058 (0.112)
-0.082 (0.137) -0.054 (0.059)
-0.018 (0.166) -0.070 (0.044)
0.065
(0.125) -0.119* (0.040)
0.057
(0.072) -0.121** (0.064)
-0.144 (0.109) -0.234* (0.054)
-0.392 (0.297) 0.095
(0.226)
Illiterate (Ref) Primary Level Secondary Level Higher Level
0.122** (0.055) 0.334* (0.077) 0.350** (0.203)
0.144* (0.040) 0.331* (0.061) 0.749* (0.127)
0.162* (0.028) 0.315* (0.048) 0.794* (0.099)
0.166* (0.021) 0.362* (0.042) 0.805* (0.070)
0.132* (0.030) 0.355* (0.048) 0.725* (0.074)
0.088** (0.034) 0.389* (0.062) 0.762* (0.113)
0.144
(0.109) 0.332** (0.167) 0.293
(0.484)
Sector_Gov 0.270* (0.090)
0.296* (0.057)
0.205* (0.044)
0.149* (0.039)
0.122** (0.048)
0.098 (0.080)
-0.384 (0.277)
Sector_Priv 0.174** (0.102)
0.147** (0.081)
0.179** (0.069)
0.182* (0.050)
0.094** (0.059)
0.092** (0.042)
0.662** (0.175)
Sector Agri 0.097** (0.052)
0.131* (0.032)
0.084* (0.022)
0.035 (0.024)
0.037** (0.017)
-0.027 (0.022)
-0.261* (0.121)
Self 0.203* (0.077)
0.250* (0.061)
0.252* (0.040)
0.220* (0.040)
0.192* (0.031)
0.166* (0.035)
0.034 (0.160)
West -0.280* (0.044)
-0.270* (0.021)
-0.275* (0.021)
-0.239* (0.022)
-0.233* (0.027)
-0.192* (0.028)
-0.014 (0.096)
Foreign Transfer 0.061 (0.193)
0.089 (0.202)
0.153** (0.072)
0.221* (0.082)
0.572* (0.203)
0.568* (0.139)
-0.228 (0.243)
Number Of Observation Pseudo R2
4020 0.270
4020 0.279
4020 0.294
4020 0.288
4020 0.276
4020 0.265
4020 0.242
16
TABLE 4 - Urban Rural gap (RIF-regression)
Author’s computing from 2010 National Survey on Households’ Budget, Consumption and Standard of Living
Standard errors are in parentheses and (*** significant at 1%, ** significant at 5%,*significant at 10%)
10th percentile 50th percentile 90th percentile
Estimated std Estimated std Estimated std
Reference group: Urban Coef
Estimated log expenditure gap:
E[RIFq (ln(Expu))]-
E[RIFq(ln(Expr))]
-0.633***
(0.023)
-0.564***
(0.018)
-0.614***
(0.025)
Composition effects attributable to
Age, gender, household size,
foreign transfer and logement
-0.069*** (0.021) -0.071*** (0.015) -0.075*** (0.021)
West region -0.070*** (0.011) -0.077*** (0.007) -0.056*** (0.011)
Education -0.116*** (0.020) -0.117*** (0.013) -0.156*** (0.020)
Employment -0.029** (0.013) -0.034** (0.009) -0.035** (0.014)
Agregated characteristics
effects
-0.298*** (0.030) -0.315*** (0.021) -0.337*** (0.025)
Regional structure effects attributable to
Age, gender, household size,
foreign transfer andlogement
0.494 (0.319) -0.079 (0.224) -0.036 (0.353)
West region 0.031** (0.013) 0.003 (0.009) 0.017 (0.014)
Education -0.116*** (0.020) -0.084** (0.033) -0.148*** (0.051)
Employment 0.075** (0.031) 0.041* (0.021) 0.040 (0.033)
Constant -0.094*** (0.330) -0.073 (0.232) -0.117 (0.366)
Agregated coefficient effect -0.335*** (0.036) -0.249*** (0.024) -0.277*** (0.036)