highways and productivity in urban and rural locations...1 highways and productivity in urban and...
TRANSCRIPT
1
Highways and productivity in urban and rural locations
Adelheid Holl1
Work in progress - February, 2014
Abstract:
I estimate the effect of access to highways on firm-level productivity using a micro level panel
data set for Spanish manufacturing and business service firms from 1997 to 2007. To estimate
the causal relationship between firm level productivity and access to highways, I use
instrumental variables estimation together with panel data techniques. The results show a
significant positive effect of access to highways on firm-level productivity for both
manufacturing and business service firms. However, the results also show that the gains from
access to new highways are not evenly distributed across space and type of firms. Firms that
benefit most are very small manufacturing firms and larger service sector firms located in
urban core areas as well as suburban manufacturing firms and to some extent also larger rural
manufacturing firms.
Keywords: highways, transport infrastructure, firm-level productivity, urban, suburban, rural.
JEL: C23, D24, R12, R3, R4;
1 Institute of Public Goods and Policy (IPP), CSIC – Spanish National Research Council, c/ Albasanz 26-28, 28037 Madrid, Spain; e-mail: [email protected] Acknowledgements: This research has received financial support from the projects SEJ2006-08063 and ECO2010-17485 (Ministerio de Ciencia e Innovacio) and CSIC [200910I105]. I thank Sabine D'Costa, Alexander Lembcke and Ricardo Mora for their useful comments and suggestions. I am also grateful to conference and seminar participants at the Urban Economics Association Meeting 2013 and the 2013 Barcelona Workshop on Regional and Urban Economics. .
2
1. Introduction
The effect of infrastructure investment on economic outcomes is of central concern to policy
makers and academics as well. The subject has attracted increasing attention over the last
decades. Traditionally the emphasis has been on macro-economic studies (for a recent review
see Melo et al. 2013), but recently the focus is shifting to micro-economic analyses. However
the exact effects, and most importantly, whether there are causal effects still remains a matter
of debate (Funderburg et al. 2010; Crescenzi and Rodríguez-Pose 2012; Leduc and Wilson
2012).
Using data for Spain from 1997 to 2007, I focus on the direct effect of access to highways
and its improvement on firm-level productivity of manufacturing and business service firms.
A better accessible location can provide important transport and travel cost savings leading to
improved market access, and thereby allowing greater exploitation of economies of scale and
specialisation. Highways not only reduce direct transportation costs but also time costs (such
as, for example, delivery uncertainty) and facilitate travel for labour and information flows.
Together, this can create opportunities for new forms of organization of production,
improvements to supply chains and client services leading also to greater productivity via the
optimisation of production and input and output market relations. Shirely and Winston
(2004) show, for example, that highways reduce firms’ inventories and consequently logistics
costs. Similar evidence is found in Datta (2012) and Li and Li (2013) for India and China
respectively. This is consistent with transport investment induced changes in the organisation
of production towards an increasing reliance on transport that substitutes for traditional
inventory holding (Hesse and Rodrigue 2004). Transport, travel cost, and time savings from
improved access to highways and the resulting changes in production organisation will reflect
on firms’ productivity.
The impact of access to transportation networks has been studied by Banerjee et al. (2012)
for county economic outcomes in China. They find a small positive effect on GDP per capita
in levels but no effect on GDP per capita growth. They also find that an increase in distance
reduces average firm profits and the number of firms in counties. Faber (2014) finds that the
expansion of China’s National Trunk Highway System has reinforced concentration of
economic activity in central nodes with negative growth effects in peripheral counties along
3
the new routes. Ghani et al. (2012) find for India's Golden Quadrilateral Program, a major
highway project, higher entry rates and increases in average plant productivity for non-nodal
districts within a distance of 10km from the project compared to non-nodal districts further
away. There are also studies that have focused specifically on the role of access to transport
infrastructure for firm location (Guimarães et al. 1998; Coughlin and Segev 2000; Gabe and
Bell 2004; Holl 2004a, 2004b), firm relocation (Holl 2004c; de Bok and van Oort 2011) or
firm level exports (Volpe Martincus et al. 2012; Albarran et al. 2013). 2
Few studies have investigated the impact of transport infrastructure on firm-level
productivity. Lall (2004), Graham (2007a, 2007b), Gibbons et al. (2012) and Holl (2012) are
recent examples. However, these studies use area market potential measures to capture road
improvement effects. In this paper, I use the variation in the distance from each firm to its
nearest limited access highway (autovías and autopistas) and pay particular attention to the
heterogeneity of impacts.
My study is close in spirit to Faber (2014), Banerjee et al. (2012), and Ghani et al. (2012).
However, these studies analyse the effects of access to transportation networks on county and
district outcomes. Ghani et al. (2013) further presents aggregate impacts separated between
urban and rural parts of districts. My study differs in so far that I estimate the effect of access
to highways at the firm level. This allows controlling for heterogeneity across firms.
Furthermore, I directly explore the time variation in the construction of a major limited
access highway network and therefore I am able to use distances that vary over time.
To establish a causal effect it is necessary to address the potential endogeneity of access to
highways. Endogeneity can stem from either the sorting and selection of firms via their initial
location decision or from government policy to specifically allocate new infrastructure to
places with higher or lower expected productivity growth. Governments may respond to
higher productivity growth in some areas or help to stimulate growth in others. There may
also be omitted variables that explain both highway access and firm level productivity. To
2Another related literature studies the effects of highways on growth (Chandra and Thompson, 2000) and the labour market (Michaels, 2008) in rural counties, on urban growth (Duranton and Turner, 2012), the effect of access to railroads on urbanisation (Atack et al., 2010), and the effect of highways on population suburbanization (Baum-Snow, 2007, Baum-Snow and Turner, 2012, Baum-Snow et al. 2013).
4
address these issues, I rely on different fixed effects specifications, instrumental variable
estimation and panel data techniques.
I find that firms’ productivity is indeed significantly influenced by access to highways but
there is also important heterogeneity in impacts across firms. As suggested by economic
geography (NEG) as well as some previous empirical studies (Rephann and Isserman 1994;
Funderberg et al. 2010) effects can be different in urban and rural locations. My preferred
estimates show that improving access to highways benefits mostly very small manufacturing
firms and larger service sector firms located in urban core areas as well as suburban
manufacturing firms and to some extent also larger rural manufacturing firms.
The paper contributes to a better understanding of effects of transport infrastructure
investment on economic outcomes in two important ways. First, the results suggest a
significant role for highway access in shaping the economic geography of a country through
facilitating specialisation in different productive activities. This can improve national growth
performance and lead to important welfare gains wherever such changes result in a more
efficient spatial organisation of production. Second, the paper provides new insights into the
heterogeneity of productivity impacts of major highway projects. Improving our knowledge
of transport infrastructure investment is crucial to transport policy, for ensuring efficient
allocation of resources, and for improving our understanding of the development of cities
and regions.
The paper is organised as follows. Section 2 covers data. Section 3 presents the estimation
approach and variable definition. Section 4 presents the results of the empirical estimations.
Section 5 concludes.
2. Data
2.1. Firm-level data
The firm-level dataset used to calculate total factor productivity (TFP) is the SABI data base.
SABI (Sistema de Análisis de Balances Ibéricos) is generated by INFORMA and Bureau Van
Dyck and contains financial accounts of Spanish companies. I use data for firms in
manufacturing and the business service sectors in mainland Spain. The database contains
5
exhaustive balance sheet information as well as other firm characteristics since 1997. Most
importantly, the data base includes the geographic co-ordinates of firms. I use this
information to relate firms to the highway network.
The information available from the SABI database does not permit outputs and inputs to be
assigned correctly among plants of multi-plant companies. Thus, I have applied two filters.
First, I use only firms with unconsolidated accounts. Second, I have dropped all firms that
report delegations. In also exclude firms that have relocated over the period of analysis and
firms that have changed their industrial sector. To identify those firms, I have extracted
information on firms’ previous location and sector from all earlier editions of the SABI
database. This way, I have identified 8,569 firms that changed location beyond the city level
over the period of analysis and 7,418 firms that changed their industrial sector.
After cleaning the data set, the final sample consists of an unbalanced panel of 126,820 firms.
Table A1 in the Appendix provides information on the distribution of firms by year in the
final sample used for estimation.
2.2. Limited Access Highways
The limited access highway network in Spain has been extended drastically over the last
decades making it the longest network among European Union member countries (Holl
2011). At the same time roads are the dominant transport mode in Spain. The share of roads
in total inland freight transport has actually been growing steadily and today accounts for well
over 90%.3 This is significantly above EU average and indicates the important role of
highways for freight transport in Spain. Roads also represent about 90% of overall passenger
transport in Spain.
To calculate access to highways, I use geographical information systems (GIS). First, I create
digital vector maps of the Spanish road network for the period 1997-2007. I use detailed
information obtained from the Ministry of Public Works regarding the opening of new
3 EU statistics on inland freight transport exclude air and maritime freight. Their weight is greater in international trade flows, but still roads are the dominant mode for exports in Spain and roads account for over 80% of total freight movements.
6
limited access highway segments together with the annual official roadmaps published by the
Ministry of Public Works. Figure 1 shows the highway network for 1997 and for 2007.
Second, I overlay the digital vector maps with a point layer of the geographic coordinates of
all manufacturing and business service firms included in the analysis. With these two layers, I
calculate the distance from each company location to its nearest limited access highway.
Figure 2 shows the limited access highway network for 2007 and its relation to the 2007
sample of manufacturing and business service firms. Since I have information on the year of
opening of new highways, I can calculate this distance for each year of the period of analysis.
This allows me to analyse the relationship between highway access and firm level productivity
in a cross-sectional as well as in a panel data setting.
2.3. Urban and rural areas
Mainland Spain is divided into nearly 8,000 municipalities constituting separate political and
administrative units. Unlike other countries, Spain has no official definition of urban and rural
areas. Some studies distinguish urban and rural areas according to population thresholds (e.g.
municipalities with less than 2,000 or 10,000 inhabitants are sometimes classified as rural).
Here, I use the information from the open data project AUDES (Áreas Urbanas de España;
see Ruiz 2010) which provides a typology of Spanish municipalities.4 For mainland Spain,
AUDES defines 129 urban areas. These include a densely populated central city and its
adjacent municipalities which are selected based on land use continuity and commuting data
(totalling 1,357 municipalities). The remaining municipalities are defined as rural. 5
Figure 3 shows the AUDES urban areas defined according to Ruiz (2010) and their relation
to the 2007 highway network. It is well documented that urban areas are major contributors
to national economies but rural areas are not neglectable either. In my sample, nearly 30% of
4 See, http://alarcos.esi.uclm.es/per/fruiz/audes/ 5 Recently, the OECD in collaboration with the EU has developed a new approach to classifying functional urban areas (OECD, 2012). For mainland Spain, their classification defines 101 functional urban areas with hinterlands made up by 2,378 municipalities. Using this definition, results are qualitatively similar and available upon request.
7
the manufacturing firms are located outside urban areas. Although, in the case of business
service firms this is just about 8%.
For transport planners, urban planners, and regional policy makers, an important question is
whether there are spatial variations in highway improvement effects on firm level productivity
and whether or not urban and rural areas are affected equally. There is evidence in the
literature that effects may depend on the geographical context (Funderberg et al. 2010), such
as the degree of prior urbanisation (Rephann and Isserman 1994). New economic geography
(NEG) shows that in the presence of increasing returns to scale, the spatial distribution of
benefits from falling trade costs between asymmetric markets depends on the initial size of
the costs (Puga 1999). With high initial transport costs, new highways can reinforce the
concentration of economic activity (see, Faber 2014; Roberts et al. 2012, for recent empirical
evidence on the Chinese highway system) but as shown in Kilkenny (1998) at relative low
average industrial transport costs, further reductions can also positively relate to rural
development. For the organized manufacturing sector in India, Ghani et al. (2013) find a
higher effect on entry rates in rural parts of districts that were in a distance of up to 10 km
from the highway upgrades compared to the urban parts of the districts. Regarding
productivity effects they find however very similar results in both rural and urban locations.
Rural and urban studies have largely developed separately with a focus on different issues
(Kilkenny 2010). But as Partridge et al. (2008) show, urban and rural areas should not be
viewed in isolation. Since highways can re-enforce rural-urban interdependences, there are
insights to be gained from approaches that consider both types of areas and the links between
them.
3. Estimation
I estimate a reduced-form equation where the logarithm of total factor productivity (TFP) of
firm i in sector s, province p, and year t is a function of the distance to the nearest highway
disit, firm-specific characteristics cit, industry-year fixed effects st, and province-year fixed
effects pt.
8
isptptstititispt ecdisTFP 21log (1)
To calculate firm-level TFP, I apply the standard approach of Levinson and Petrin (2003).
This approach uses intermediate inputs as proxy for unobserved productivity shocks to
account for the possible endogeneity arsing from such unobserved shocks. I use the Stata
routine levpet provided by Petrin et al. (2004) and estimate firm-level production functions
separately for 23 manufacturing and business service sectors.6
The main variable of interest is the log distance to the nearest limited access highway (disit).
This distance measure is open to different endogeneity sources. First, higher productivity
firms may self-select into cities and areas with better access to highways through their initial
location decision (Baldwin and Okubo 2006; Nocke 2006). Second, governments may place
new limited access highways specifically in higher productivity areas to respond to the needs
of future growth or the government may place new highways in lower productivity areas to
stimulate growth. As highways in Spain where partly financed with the EU Structural Funds
and the Cohesion Fund, new highway placement could have been biased to poorer regions
and specifically to their urban areas. Moreover, the main productive areas in Spain are in the
centre of the country and along the north and eastern coastal regions. By linking those areas,
the new highways inevitably crossed regions with lower economic activity so that an
important part of the new infrastructure investments also went to areas of lower economic
development. Third, there may be omitted variables that explain both highway access and
firm-level productivity. I address potential endogeneity with different fixed effects
specifications, instrumental variables, and IV panel data regression.
First, I use industry-year and province-year fixed effects. The industry-year fixed effects
absorb industry-specific dynamic shocks to productivity. The province-year fixed effects
capture time-varying specificities at the province level and thus time varying productivity
advantages of some areas over others. In particular, Spain’s provinces relate fairly closely to
labour market areas and the province-year fixed effects should therefore capture any labour
market factors influencing firm-level productivity. More generally, the province-year fixed
6 Results from the estimation of the production function are not included here, but are available upon request.
9
effects control for regional dynamic factors that might have a simultaneous effect on the
placement of new highways and firm-level productivity.
Second, to deal more generally with the potential two-way causality between productivity and
highway access I further use instrumentation. In choosing the instruments for estimation, I
follow the recent related literature which has used historical road networks as instruments to
account for the potential endogeneity of current highway improvements (see, Baum-Snow et
al. 2013; Duranton and Turner 2012; and Volpe Martincus et al. 2012, amongst others).
Specifically, as in Holl (2012) I use the 1760 Spanish postal route network. The Spanish
highway network presents a strongly radial outlay that dates back to the 18th century when
king Carlos III (1716-1788) ordered the construction of radial roads linking “as direct as
possible” Madrid to Valencia, Andalusia, Catalonia and Galicia. Later, routes to Extremadura
and the Basque Country were added, the latter linking on to France. The road network of that
time is reflected in Thomas Lopez’s map of postal routes, which he draw in 1760 for the
book by Pedro Rodríguez Campomanes entitled “Itinerario Real de Postas de dentro y fuera de
España” (Postal route itinerary in Spain and outside Spain) and published in 1761 (see also,
Menéndez-Pidal 1992). Thomas Lopez distinguished two types of postal routes: “montadas”
(or “sobre ruedas” - referring to routes for transport on wheels) and “no montada” (or “a la ligera”
- referring to routes for transport on horseback). The former are the precursors of the six
main radial highways and together with the latter they are good predictors of an important
part of the current Spanish highway network. I use this map and the itineraries described in
Campomanes (1761) to create a geo-referenced map of both types of 1760 postal routes
(Figure 4). Based on this map, I calculate the distance from each firm to the nearest of the
two types of 1760 postal routes.
The instrumental variable seeks to eliminate the bias from reversed causality. The instrument
is valid as long as it predicts recent highway improvements, but is otherwise uncorrelated with
current firm-level productivity. While it is plausible, that the 1760 postal routes were not
designed in response to current productivity patterns, there could still be factors that
influenced their location in the past and that continue to exert and influence on modern
10
productivity (e.g. geographical characteristics or institutions). Such enduring location
advantages could cause correlation between the 1760 postal routes and factors that cause
modern productivity to be higher. The province-year fixed effects included in all estimations
help to avoid any correlation between the 1760 postal routes and firm level productivity other
than related to new highway construction.
Third, to further control for the existence of unobserved firm-level characteristics that might
jointly determine highway access and firm-level productivity, I also estimate IV panel data
models. For these within regressions, the instrument needs to have a time dimension. Here, I
follow in the spirit of Hornung (2012) and introduce postal routes within a 10km corridor
from the new highways that opened to traffic during that year. The IV panel data estimates
further mitigate bias stemming from omitted firm-level variables and the potential self-
selection of firms through their initial location decision.
4. Results
OLS estimation results
Results in Table 1 are based on OLS estimation of equation (1) where I regress firm-level
TFP on distance to the nearest highway together with firm level controls, industry-year, and
province-year fixed effects. In these estimations, the effect of highway access is driven by
differences across firms within province-year and industry-year combinations. Panel A shows
results for manufacturing. Panel B shows results for business service firms.
The elasticity between distance to highways and firm-level TFP is statistically significant and
negative (-0.013) for the pooled sample of manufacturing firms (column 1). Column 2 shows
the estimated effect for manufacturing firms in urban areas. Column 3 and 4 distinguish
further between micro firms with less than 10 employees and small- and medium-sized firms
with 10 to 250 employees. The Spanish productive system is characterised by its small firm
structure. According to the “Directorio Central de Empresas” (DIRCE- Central Directory of
Companies from the National Institute of Statistics) 99.8% of Spanish companies have less
11
than 250 employees and 94% of all firms (88% of firms with paid employees) have less than
10 employees. Firms with more than 250 employees are somewhat overrepresented in the
SABI data base but still only account for approximately 0.5% of total observations in my
sample. I therefore do not report results separately for this size category. Columns 5 to 7
show the respective results for rural manufacturing firms. The estimated effect is always
statistically significant and higher in rural areas than in urban areas. The strongest effect is
observed for small rural manufacturing firms.
For the pooled sample of business service firms, the elasticity between distance to highways
and firm-level TFP is also statistically significant and negative (-0.007). However, estimations
for firms of different size in urban and rural areas show only significant results for rural firms,
regardless of size.
OLS pooled cross section estimations could, however, be biased whenever access to highways
is endogenous. For example, if more productive firms locate closer to highways or when the
government places new highways in response to productivity differences across locations,
reverse causation could be in place.
IV estimation results
Results of IV estimations exploiting cross sectional variation are presented in Table 2 for
manufacturing and for business service firms respectively. Table 2 also reports the
Kleibergen–Paap F test statistics for weak instruments in the presence of robust clustered
standard errors (Kleibergen and Paap 2006). The statistics show that the instrument is very
strong and that weak-instrument bias is not a problem.
The IV point estimate in column (1) for the pooled sample of manufacturing firms is again
negative and significant (panel A) and of the same magnitude as the corresponding OLS
coefficient. However, important differences arise when the sample is separated into urban
and rural firms. For urban manufacturing firms, TFP appears to be negatively affected by
improving highway access (column 2-4). In contrast, reducing the distance to the nearest
12
highway has positively affected firm-level productivity for manufacturing firms in rural
locations. Here the IV point estimates are significant and larger in absolute value than the
OLS estimates: -0.023 for the pooled sample of rural manufacturing firms (column 5), -0.023
for micro rural manufacturing firms (column 6) and -0.017 for small to medium sized rural
manufacturing firms (column 7).
Panel B shows the corresponding results for business service firms. For the pooled sample
the coefficient is significant and markedly larger than the OLS coefficient (-0.043).
Controlling for endogeneity also results in a markedly larger coefficient for business services
compared to manufacturing. Separate estimations for firms of different size in urban and
rural areas show that the firms that benefit the most are those in urban areas. Moreover, there
is no statistically significant effect for rural business service firms once endogeneity has been
controlled for via instrumentation.
Recent empirical evidence also suggests that highways may change the spatial organization of
economic activity within urban areas (Baum-Snow 2014). In Table 3, I extend my IV analysis
by separating between the central-city-municipality of my urban areas and the sub-urban
parts. I find that the negative productivity effect in Table 2 for urban manufacturing firms is
only observed for those firms that are located in central city municipalities. In contrast, for
those firms located in suburban parts of urban areas, particularly the medium-sized firms,
improved highway access also shows a significant positive effect on their firm-level
productivity. In contrast, for business service firms the separation into urban core and sub-
urban parts yield weak results with only a marginally significant productivity effect for firms
in central city locations (see column 1, panel B).
IV panel data estimation results
The panel structure of the data allows investigating the effect of access to highways on firm-
level productivity taking into account time-invariant unobserved firm heterogeneity. The
within IV regressions explore short-run variations of variables and thus can be interpreted as
short-run effects of improving access to highways on firm-level productivity.
13
Table 4 shows the estimation results for manufacturing and for business service firms, with
industry-year, and province-year fixed effects. The within IV results differ form the cross-
sectional ones. For the manufacturing firms, the coefficient for access to highways is now
higher (-0.022) but only marginally statistically significant at the 10% level (column 1).
Improving access to highways now shows a stronger effect for urban firms, particularly for
very small ones (column 3), while the coefficient for mediums sized urban manufacturing
firms is not significant. The significantly positive coefficients of distance to highways in
columns (2) to (4) of Table 2 (panel A) are therefore driven by differences across urban firms.
In rural areas only the medium sized firms benefit from highway access improvements
(column 7), but there is no statistically significant effect on very small rural manufacturing
firms (column 6) once unobserved firm-level heterogeneity is also controlled for.
For business services firms (Table 4, panel B), I only find a significant positive productivity
effect for medium sized firms in urban areas (column 4). The within IV point estimate is large
(-0.140) and highly significant. In contrast, for very small business service firms, the increase
in accessibility seems to have a negative effect on productivity. Improved access can increase
competition from distant firms that also experience improved accessibility and those
competition effects can dominate own gains in market access. Making cities more accessible
may also raise urban costs. Improved highway access may lead to a rise in agglomeration and
with it to increased congestion or rising land prices and wages that are bearing more heavily
on small firms.
Table 5 shows the within IV regressions separating between firms located in central cities and
in suburban areas for manufacturing and for business services. These estimations indicate that
the results in Table 4 for urban areas are again driven by those firms located in central cities.
Improved highway access impacts differently on suburban firms. In suburban locations, all
manufacturing firms, regardless of size, witness increases in TFP when access to highways
improves. For suburban business service firms the estimated coefficients indicate negative
productivity effects. However, for these estimations, the Kleibergen- Paap F-statistic suggests
that the time-varying instrument is weak.
14
Overall, the within IV regression results suggest that, at least in the short run, new highways
benefit more firms in urban areas, with the exception of very small business service firms in
central cities. For larger rural manufacturing firms, the results also show positive productivity
effects, albeit somewhat lower. Improved highway access may have enabled these firms to
expand their market areas. In contrast, for business services and very small manufacturing
firms in rural areas the benefits may be off-set in the short run from resulting increased
competitive pressure as new highways integrate markets and not only reduce transport costs
from rural locations but also to rural locations (Kilkenny 1998). 5. Conclusion
The contribution of transport infrastructure to productivity is still a debated issue. I analyse
the effect of access to highways on firm-level productivity. The results show that firm-level
productivity is significantly related to highway access. An important policy concern is the
distribution of benefits from highway investment. The data and setting in this paper makes it
possible to directly analyse whether and to what extent highways affect productivity in urban,
suburban, and rural locations. The results show important heterogeneity of highway impacts.
Not all firms benefit equally from improved highway access. The preferred estimates, that
take into account firm-level unobserved heterogeneity as well as endogeneity of highway
improvements, show that firms that benefit most, at least in the short run, are very small
manufacturing firms and larger service sector firms located in urban core areas as well as
suburban manufacturing firms and to some extent also larger rural manufacturing firms. This
is an important finding for a better understanding of distributional effects of transport
infrastructure investment.
Improved highway access leads to change in urban and rural areas as well as to change within
urban areas and to change in the relationship between the areas. Together, the results suggest
a significant role for highways in shaping the economic geography of a country. Finally, the
results also highlight that transport improvements interact with agglomeration economies and
suggest that they can change the balance between urban costs and benefits.
15
References:
Albarran, P., R. Carrasco, and A. Holl (2013) Domestic transport infrastructure and firms' export market participation, Small Business Economics 40 (4): 879-898.
Atack, J., F. Bateman, M. Haines and R. Margo (2010) Did Railroads Induce or Follow Economic Growth? Urbanization and Population Growth in the American Midwest, 1850-60, Social Science History 34 (2): 171-197.
Baldwin, R.E. and T. Okubo (2006) Heterogeneous firms, agglomeration and economic geography: spatial selection and spatial sorting, Journal of Economic Geography 6 (3): 232-346.
Banerjee, A., E. Duflo and N. Qian (2012) On the road: access to transportation infrastructure and economic growth in China, NBER Working Paper 17897.
Baum-Snow, N. (2007) Did highways cause suburbanization?, The Quarterly Journal of Economics 122: 775-805.
Baum-Snow, N., (2014) Urban Transport Expansions, Employment Decentralization, and the Spatial Scope of Agglomeration Economies. mimeo, Brown University.
Baum-Snow, N., and M. Turner (2012) Transportation and the Decentralization of Chinese Cities. mimeo, Brown University.
Baum-Snow, N., Brandt, L., Henderson, J., Turner, M., and Zhang, Q. (2013). Roads, railroads and decentralization of Chinese cities. mimeo, Brown University.
Campomanes, P. R. (1761) Itinerario Real de Postas de dentro y fuera de España, publicado por orden del Rey en 1761.
Chandra, A., Thompson, E. (2000) Does public infrastructure affect economic activity? Evidence from the rural interstate highway system, Regional Science and Urban Economics 30: 457-490.
Coughlin, C. C. and E. Segev (2000) Location Determinants of New Foreign-Owned Manufacturing Plants, Journal of Regional Science 40: 323–351.
Crescenzi, R. and A. Rodríguez-Pose (2012) Infrastructure and regional growth in the European Union, Papers in Regional Science 91 (3): 487–513.
Datta, S. (2011) The Impact of Improved Highways on Indian Firms. Journal of Development Economics 99(1): 46–57.
de Bok, M., and van Oort, F. (2011). Agglomeration economies, accessibility, and the spatial choice behavior of relocating firms. The Journal of Transport and Land Use, 4 (1): 5–24.
Duranton, G., Turner, M.A. (2012) Urban growth and transportation, Review of Economic Studies 79 (4): 1407-1440.
Faber, B. (2014) Trade Integration, Market Size, and Industrialization: Evidence from China’s National Trunk Highway System, University of California Berkeley, mimeo.
Funderburg, R., H. Nixon, M. G. Boarnet, and G. Ferguso (2010) New highways and land use change: Results from a quasi-experimental research design, Transportation Research Part A 44: 76–98.
16
Gabe, T., and K.P. Bell (2004) Tradeoffs between local taxes and government spending as determinants of business location. Journal of Regional Science 44(1): 21-41.
Ghani, E., A. Grover Goswarni, and W. R. Kerr (2012) Highway to Success: The Impact of the Golden Quadrilateral Project for the Location and Performance of Indian Manufacturing, NBER Working Paper No. 18524.
Ghani, E., A. Grover Goswarni, and W. R. Kerr (2013) The Golden Quadrilateral Highway Project and Urban/Rural Manufacturing in India, World Bank Policy Research Working Paper 6620.
Gibbons, S., Lyytikäinen, T., Overman, H., and R. Sanchis-Guarner (2012) New Road Infrastructure: the Effects on Firms. SERC Discussion Paper 117, London School of Economics.
Graham, D. J. (2007a) Agglomeration, Productivity and Transport Investment, Journal of Transport Economics and Policy 41 (3): 317-343.
Graham, D. J. (2007b) Variable returns to agglomeration and the effect of road traffic congestion, Journal of Urban Economics 62: 103-120.
Guimarães P, Rolfe RJ, Woodward D (1998). Regional Incentives and Industrial Location in Puerto Rico, Int. Reg. Sci. Rev. 21: 119-138.
Hesse, M. and J. P. Rodrigue (2004) The transport geography of logistics and freight distribution. Journal of Transport Geography 12 (3): 171-184.
Holl, A. (2004a) Manufacturing location and impacts of road transport infrastructure: Empirical evidence from Spain. Regional Science and Urban Economics 34 (3): 341-363.
Holl, A. (2004b) Transport infrastructure, agglomeration economies, and firm birth. Empirical evidence from Portugal. Journal of Regional Science 44 (4): 693-712.
Holl, A. (2004c) Start-ups and relocations: Manufacturing plant location in Portugal. Papers in Regional Science 83: 649-668.
Holl, A. (2011) Factors Influencing the Location of New Motorways: Large Scale Motorway Building in Spain, Journal of Transport Geography 19(6): 1282–1293.
Holl, A. (2012) Market potential and firm-level productivity in Spain, Journal of Economic Geography 12 (6): 1191–1215.
Hornung, E. (2012) Railroads and Micro-regional Growth in Prussia, Ifo Working Paper 127.
Kilkenny, M. (1998) Transport Costs and Rural Development, Journal of Regional Science 38: 293-312.
Kilkenny, M. (2010) Urban/regional economics and rural development, Journal of Regional Science 50 (1): 449–470.
Kleibergen, F., and R. Paap (2006) Generalized reduced rank tests using the singular value decomposition. Journal of Econometrics 133 (1): 97–126.
Lall, S., Z. Shalizi and U. Deichmann (2004) Agglomeration Economies and productivity in Indian Industry, Journal of Development Economics (73): 643-673.
Leduc, S. and D. Wilson (2012) Should Transportation Spending be included in a Stimulus Program? A Review of the Literature, Federal Reserve Bank Oof San Francisco, Working Paper Series 2012-15.
17
Levinsohn, J, and A. Petrin (2003) Estimating Production Functions Using Inputs to Control for Unobservables, Review of Economic Studies 70(2): 317-341.
Li, H. and Z. Li (2013) Road investments and inventory reduction: Firm level evidence from China, Journal of Urban Economics 76: 43–52.
Melo, P. C., D. J. Graham, and R. B. Noland (2009) A meta-analysis of estimates of urban agglomeration economies, Regional Science and Urban Economics 39: 332-342.
Melo, P., Graham, D. and R. Brage-Ardao (2013) The Productivity of Transport Infrastructure Investment: A Meta-Analysis of Empirical Evidence, Regional Science and Urban Economics 43: 695–706.
Menéndez-Pidal, G. (1992) España en sus caminos. Caja de Madrid, Madrid.
Michaels, G. (2008) The effect of trade on the demand for skill: evidence from the interstate highway system, The Review of Economics and Statistics 90: 683-701
Nocke, V. (2006) A gap for me: Entrepreneurs and Entry, Journal of the European Economic Association 4 (5): 929-956.
OECD (2012) Redefining “Urban”: A New Way to Measure Metropolitan Areas, OECD Publishing.
Partridge, M. D., Rickman, D.S., Ali, K. and M. R. Olfert (2008) Lost in space: population growth in the American hinterlands and small cities, Journal of Economic Geography 8: 727–757.
Petrin, A., Levinsohn, J., and B. Poi (2004) Production Function Estimation in Stata Using Inputs to Control for Unobservables, The Stata Journal 4(2): 113-123.
Puga, D., 1999. The rise and fall of regional inequalities. European Economic Review 43: 303-334.
Rephann, T. and A. Isserman (1994) New highways as economic development tools: An evaluation using quasi-experimental matching methods, Regional Science and Urban Economics 24: 723-751.
Riley, S. J., DeGloria, S. D. and Elliot, R. (1999) A terrain ruggedness index that quantifies topographic heterogeneity, Intermountain Journal of Science 5 (4): 23-27.
Roberts, M., U. Deichmann, B. Fingleton, T. Shi (2012) Evaluating China's road to prosperity: A new economic geography approach, Regional Science and Urban Economics 42 (4): 580–594.
Shirely, C. and C. Winston (2004) Firm inventory behavior and the returns from highway infrastructure investments. Journal of Urban Economics 55: 398-15.
Stock, J. and M. Yogo (2005) Testing for weak instruments in linear IV regression, in Andrews, D. and Stock, J. (eds.), Identification and inference for econometric models: Essays in honor of Thomas Rothenberg. Cambridge University Press. Cambridge, MA.
Volpe Martincus, C., J. Carballo, and A. Cusolito (2012) Routes, Exports, and Employment in Developing Countries: Following the Trace of the Inca Roads, Georgetown University mimeo.
18
Table 1: OLS estimates of the effect of highway access on firm TFP
All firms Urban firms Rural firms all <10 10-250 all <10 10-250 (1) (2) (3) (4) (5) (6) (7) Panel A: manufacturing
log (distance to nearest highway) -0.013***
(0.001) -0.006**
(0.001) -0.005*** (0.002)
-0.004** (0.002)
-0.015*** (0.002)
-0.018*** (0.003)
-0.008*** (0.003)
R2 0.39 0.38 0.30 0.41 0.41 0.33 0.41 Observations 385244 277864 160686 115803 107380 64274 42758 Panel B: business services
log (distance to nearest highway) -0.007***
(0.002) 0.001
(0.002) 0.001 (0.002)
-0.007 (0.005)
-0.021*** (0.007)
-0.019*** (0.007)
-0.053*** (0.020)
R2 0.39 0.38 0.30 0.41 0.41 0.33 0.41 Observations 385244 277864 160686 115803 107380 64274 42758 Note: Robust standard errors corrected for clustering at the firm level are reported in parenthesis. Significant coefficients are indicated by ***, **, *, for significance at the 1%, 5% and 10% level, respectively. Firm level controls include foreign ownership status, age of the company, and dummies for five size-categories based on total deflated assets. All estimations include industry-year and province-year fixed effects.
19
Table 2: IV estimates of the effect of highway access on firm TFP
All firms Urban firms Rural firms all <10 10-250 all <10 10-250 (1) (2) (3) (4) (5) (6) (7) Panel A: manufacturing
log (distance to nearest highway) -0.013***
(0.003) 0.013**
(0.006) 0.018** (0.008)
0.017* (0.009)
-0.023*** (0.004)
-0.023*** (0.006)
-0.017*** (0.006)
Kleibergen–Paap F-statistic 7641.3 2059.4 1654.7 645.9 3444.9 2318.9 1396.8 Observations 385244 277864 160686 115803 107380 64274 42758 Panel B: business services
log (distance to nearest highway) -0.043***
(0.008) -0.057***
(0.014) -0.052*** (0.015)
-0.065* (0.035)
-0.019 (0.013)
-0.015 (0.014)
-0.068 (0.045)
Kleibergen–Paap F-statistic 2775.6 1107.2 1089.5 140.8 743.4 698.7 49.7 Observations 178778 164600 130143 33818 14178 12778 1393 Note: Robust standard errors corrected for clustering at the firm level are reported in parenthesis. Significant coefficients are indicated by ***, **, *, for significance at the 1%, 5% and 10% level, respectively. Firm level controls include foreign ownership status, age of the company, and dummies for five size-categories based on total deflated assets. All estimations include industry-year and province-year fixed effects.
20
Table 3: IV estimates of the effect of highway access on firm TFP in urban agglomerations
Central city location Suburban location all <10 10-250 all <10 10-250 (1) (2) (3) (4) (5) (6) Panel A: manufacturing
log (distance to nearest highway) 0.040***
(0.008) 0.036*** (0.010)
0.054*** (0.012)
-0.025*** (0.009)
-0.017 (0.012)
-0.028** (0.014)
Kleibergen–Paap F-statistic 1432.1 1086.1 491.4 763.2 604.9 243.2 Observations 129884 79206 50050 147980 81480 65753 Panel B: business services
log (distance to nearest highway) -0.034*
(0.020) -0.032 (0.021)
-0.036 (0.047)
-0.001 (0.017)
0.002 (0.020)
-0.001 (0.036)
Kleibergen–Paap F-statistic 591.9 576.6 80.7 549.9 509.8 87.5 Observations 131573 103519 27475 33027 26624 6343 Note: Robust standard errors corrected for clustering at the firm level are reported in parenthesis. Significant coefficients are indicated by ***, **, *, for significance at the 1%, 5% and 10% level, respectively. Firm level controls include foreign ownership status, age of the company, and dummies for five size-categories based on total deflated assets. All estimations include industry-year and province-year fixed effects.
21
Table 4: Panel IV estimates of the effect of highway access on firm TFP
All firms Urban firms Rural firms all <10 10-250 all <10 10-250 (1) (2) (3) (4) (5) (6) (7) Panel A: manufacturing
log (distance to nearest highway) -0.022*
(0.011) -0.049**
(0.025) -0.109*** (0.038)
-0.002 (0.031)
-0.021 (0.015)
0.024 (0.017)
-0.069*** (0.021)
Kleibergen–Paap F-statistic 1547.1 489.1 325.0 169.7 489.1 325.0 169.7 Observations 388176 279937 159973 112936 279937 159973 112936 Panel B: business services
log (distance to nearest highway) 0.024
(0.020) 0.081
(0.053) 0.136** (0.065)
-0.140*** (0.054)
0.017 (0.048)
0.020 (0.051)
-3.123 (10.898)
Kleibergen–Paap F-statistic 885.5 193.8 121.0 48.0 207.3 211.6 0.06 Observations 164059 150659 117867 28875 13400 12007 1208 Note: Robust standard errors are reported in parenthesis. Significant coefficients are indicated by ***, **, *, for significance at the 1%, 5% and 10% level, respectively. All estimations include industry-year and province-year fixed effects.
22
Table 5: Panel IV estimates of the effect of highway access on firm TFP in urban agglomerations
Central city location Suburban location all <10 10-250 all <10 10-250 (1) (2) (3) (4) (5) (6) Panel A: manufacturing
log (distance to nearest highway) -0.020
(0.034) -0.106** (0.054)
0.068 (0.046)
-0.090*** (0.032)
-0.113** (0.047)
-0.091** (0.039)
Kleibergen–Paap F-statistic 376.0 211.5 201.8 121.2 95.3 36.9 Observations 130541 78702 48645 149396 81271 64291 Panel B: business services
log (distance to nearest highway) 0.055
(0.052) 0.115* (0.063)
-0.146*** (0.054)
0.623** (0.320)
0.622* (0.328)
-4.455 (3.162)
Kleibergen–Paap F-statistic 240.4 144.6 47.8 3.8 3.6 1.8 Observations 120147 93662 23287 30512 24205 5588 Note: Robust standard errors corrected for clustering at the firm level are reported in parenthesis. Significant coefficients are indicated by ***, **, *, for significance at the 1%, 5% and 10% level, respectively. All estimations include industry-year and province-year fixed effects.
23
Table A1: Sample: number of firms by year and total number of observations
Year Number of firms
1997 14716
1998 18384
1999 37591
2000 42980
2001 54543
2002 62757
2003 66313
2004 68049
2005 71863
2006 76051
2007 69850
Total no of observations 583097
24
Figure 1: The mainland Spanish highway network in 1997 and in 2007
Source: own elaboration.
25
Figure 2: SABI companies and the highway network in 2007
Source: own elaboration.
26
Figure 3: Urban and rural areas in mainland Spain
Source: own elaboration, with urban areas based on Ruiz (2010).
27
Figure 4: Thomas Lopez’s map of 1760 postal routes
Source: own elaboration based on the map of Thomas Lopez “Mapa de las carreras de Postas de España” (1760) - Real Academia de la Historia, and
Campomanes (1761).