ESTUDIOS SOBRE LA ECONOMIA ESPAÑOLA

EEE 141

June 2002

FEDEA Fundación de Estudios de Economía Aplicada

http://www.fedea.es/hojas/publicado.html

Joan Costa Font a,b

Eduardo Rodriguez-Oreggia

^aLondon School of Economics, London UK.

^bDepartament de Teoria Econòmica,Universitat de Barcelona, Spain.

Correspondence Author: Joan Costa i Font, London School of Economics, Cowdray House, Houghton St., London WC2A 2AE. Email: j.costa-font@lse.ac.uk

Abstract

This paper empirically examines the hypothesis of public investment influence on regional dispaities channeled by trade. The application focuses on two countries: Mexico and Spain, both involved as “followers” of trade integration arrangements, but differing in their degrees of regional integration and development. Results show that whereas, in Spain, public investment acts as a substitute for private investment, in Mexico public investment instead plays a complementary role to private investment. Findings indicate that regional inequalities in Mexico were significantly driven by differences in the export capacity whilst inequalities in Spain were mainly explained by previous endowments and private capital formation.

Key words: regional inequalities, public investment, trade integration and human capital. JEL: F1,O1, H4

1. Introduction

The allocation of public investment as an instrument to reduce regional disparities does not appear to be grounded on robust empirical evidence. European Union (EU) public investment policies, although aiming to reduce regional disparities, do not seem to be influential despite the development of a common regional policy in the EU periphery. In examining the explanatory motives of this feature, a relevant aspect often dissmissed is the influence that trade generation (resulting from regional integration) might have in enhancing the level of income inequalities. The connection between the between trade liberalisation within and the emergence of regional inequalities still lacks acomprehensive empirical and theoretical analysis.

Trade integration alone may be expected to increase disparities between regions (Hall, 1983). Regional specificities (e.g., human capital availability and private capital) might result in a heterogeneous propensity to export, thus enhancing trade to concentrate in a small set of regions (e.g., in the centre and the periphery), which in turn might give rise to regional inequalities. In the EU context, Puga (1999) shows that under regional integration and agglomeration economies, regional inequalities - when interregional wage differentials are significant and labour mobility is small- might shrink. However, when public investment is introduced in the analysis as influencing transaction costs and trade, then results are expected to show an opposite pattern (Martin and Rogers, 1995; Martin,1999). Public investment is classified in a first sort of projects that encourage trade and a remaining set which exercise no influence. Industrial localisation may be significantly influenced by the fist sort of projects as reducing transaction costs. Reduction in transaction costs contains the incentive of firms to operate from low-income regions, which may in turn result in a rise in regional disparities. Hence, the effect of transaction costs on income inequalities is indirectly channelled by the previous effect of public investment as encouraging regional trade.

The EU as is an economic integration arrangement that goes beyond trade liberalisation. Additionally, in another reference area and exclusively grounded on promoting trade in North America, the North America Free Trade Agreement (NAFTA), has been in place effectively from the mid nineties onwards. Both arrangements offer an interesting application field for comparing how the regional incomes of regions involved evolved as a result of public investment and trade. In these study we have chosen a country of each regional area , Spain and Mexico.

This paper empirically examines the role of trade as channelling regional disparities. Two countries have been chosen to undertake the empirical application, Mexico and Spain, both being followers¹ of trade integration arrangements and showing large and persistent regional inequalities (De la Fuente and Vives, 1995), but differing in the degree of regional integration and development. Trade integration arrangements -especially when developing countries are included– involve supply side public policies to encourage trade (e.g., public investment allocation) and additionally the allocation of private investment². The central idea of this paper is the introduction of the role of trade in shifting the patterns that explain regional inequalities in developed and developing countries. We argue that whereas public investment tends to be allocated to trade intensive regions in developing countries, this is not the case in developed countries, where trade intensive regions capture a larger share of private capital and where income may be more polarised (Quah, 1993, 1996). Additionally, we test whether the (indirect) effect of public investment through trade is significant in Spain as a member of the EU, and in Mexico, currently involved in the NAFTA. Empirical analysis undertaken here employs a 2SLS that accounts for the influence of public investment, human capital and trade after controlling for geographical characteristics.

When comparing inequalities between Mexico and Spain we should bear in mind that the EU provides public assistance to less developed regions whereas NAFTA does not. As a result, 58% of the Spanish population benefits from the major source of EU funding, the Structural Fund target, called Objective 1. However, even though one might suggest that this reduces regional inequalities arising as a result of free trade, prior evidence suggests (in the convulsively???) the impact of regional funds on economic convergence and on the reduction of economic disparities is not significant (De la Fuente and Vives, 1995; Boldrin and Canova, 2001). Therefore, national public policies may have the role of reducing income disparities. That is, the effect of public investment still has to be channelled by the national states rather than the EU.

¹ The term follower in this paper recalls the concept involving periphery areas, typically relatively poor and with less power to influence the development of the integration process.

² Under these circumstances, exporting regions would tend to increase their income and therefore regional disparities may tend to rise. Furthermore, in a context of economic integration, proximity to foreign markets can stimulate trade in border regions as well as on regions holding previous advantages in agglomeration economies.

The paper is organised as follows. Section two summarises the main theoretical background. Section three compares the context of the Mexican and Spanish economies immersed in international agreements, to help further analysis of regional inequalities. Section four underlines the empirical methodology and section five deals with the results yield and a discussion about outcome and theory. Finally, section five draws some conclusions.

2. Theoretical background

This section revises three main theoretical explanations of regional inequalities that derive from recent literature on growth theory, trade theory, and new economic geography. Some of these theoretical backgrounds are extremely linked to each other although they differ essentially in identifying the variable that is supposed to channel regional income.

A first approach, as noticed, is connected with economic growth fundamentals, as growth and regional development theories are closely linked. Regional inequalities have shown to be dependent on supply side determinants e.g., human capital accumulation and public investment (De la Fuente and Vives, 1995 and De la Fuente, 1996). Accumulation of human capital is among the main explanatory variables for differences in economic growth between regions, preventing the marginal product of physical capital from falling (Lucas, 1993). Regions concentrating a large share of skilled workers and infrastructures may be those where economic activity is more dynamic, and in turn attracting private capital as well (Mulligan and Sala i Martin , 1993)³. However, interactions between public capital and income are even more complex. Some approaches tend to rely exclusively on the net contribution of public investment in reducing regional disparities (Borts 1960; Siebert, 1965). Other contributions move a step forward and examine the mechanisms explaining allocation of public investment at the regional level (Bosch and Escribano, 1988; De la Fuente, 1996). The primary reference following this second approach is Hirschman (1958), who proposed a three-stage theory. In the first stage, investment is allocated to develop urban regions, facilitating economic activities. The second stage involves the allocation of investment to less developed urban regions. Finally, in a third stage, public investment is equalised among regions.

A second sort of regional inequality determinant emphasises the role of trade and geography. The role of trade flows of recent years is stressed as an import factor in determining regional income (Irwin and Terviö, 2000). Therefore, it is important to account for the possible effects of trade in the development of regional infrastructure.

³ They discuss the inverse influence of private and public investment ratio on growth, and show that public investment should be assigned to regions where the ratio of public capital to private capital is small.

Increasing exports of poorer countries to their wealthier partners and vice versa seems to be associated with an increase in the convergence rate between those countries (Ben-David and Kimhi, 2000). Nonetheless, Venables (1999) shows that trade agreements between developing countries may possibly lead to income divergence, where the wealthier partner benefits from the poorer. The opposite holds when developing countries trade with high income countries, in such a case its more likely to experience income convergence than divergence.

Studies using a similar framework show that a country's geography and local levels of infrastructure are significant and important determinants of both transport costs and bilateral trade flows (Limao and Venables, 1999). For instance, landlocked regions or regions surrounded by others with deficient infrastructure may be at a disadvantage. If this holds, a core-periphery structure can come forward, reductions in transaction costs would lead to spatial concentration of increasing return industries in the "core" of the trade areas, whereas the periphery would specialise in constant return to scale industries (Krugman, 1991). Then, increasing returns would cause cumulative growth divergence between regions (Faini, 1984). The decline in trade costs produces in a first stage inequality among regions although in a second stage inequalities would come down (Fujita, Krugman and Venables, 1999).

Finally, a different approach relies more specifically on the role of industrial localisation (Martin and Rogers, 1995). They analysed the impact of public infrastructure on industrial location in an international context of regional development. Assuming that trade integration encourages firms to locate in countries with attractive domestic infrastructure, regional policy in low industrial concentration regions is based on infrastructure facilitating domestic rather than international trade. Improving the access for poor countries (and typically at the periphery) to core countries (typically, the centre) is likely to accelerate the processes of divergence and industrial concentration. In the same context of economic liberalisation, Martin (1999) remarks that transaction costs exist between regions and also inside regions, both being affected by public infrastructure. The equilibrium location of industries impacts the common rate of innovation because of local technology spillovers. He develops a model in which policies directed towards attracting firms in the poorest regions through the improvement of infrastructure may not generate a favourable geography for growth. This is due to the presence of local spillovers in industrial concentration which leads to lower costs of innovation, and a dispersion of the industry will increase such costs, affecting the rate of growth in the whole set of regions. This would create a trade-off between what kind of policy must be favoured: policies towards geographical equity or those that intend to achieve economic growth (or aggregate efficiency).

3. Preliminary evidence

In recent years regional policies in Europe and in North America have been largely influenced by a tendency to invest in infrastructure. The logic of this position relies on the comprehensible direct beneficial impact of regional productivity on regional growth. We argue that this phenomenon might be fuelled by trade effects. Empirical evidence shows that assuming free trade to be exogenous, increasing the volume of trade between liberalising countries reduces income gaps and improves economic growth (Ben-David and Loewy, 1998). From a specific country perspective, a movement towards free trade is likely to require infrastructure improvements; presumably it is the result of the need to take advantage of possible infrastructure improvements rather than an exogenous feature. There is no clear-cut study that analyses the link between infrastructure improvements and regional inequality under free trade conditions. It is noticeable that in the long run, income gaps tend to shrink as a result of persisting free trade (Ben-David and Loewy, 1998).

Although differing in their objectives and integration speed, the European Union (EU) and more recently the North American Free Trade Agreement (NAFTA) are the two main regional integration arrangements where at least, in their origin, the objective was mainly to promote trade. In the case of the EU the objective of reducing disparities between regions has been explicitly emphasised since its foundation. In 1987 after the approval of the Single European Act, the EU developed specific regional polices aimed at reducing regional differences. In the case of the NAFTA, this refers exclusively to an agreement to set up an are of free trade (Bosworth, Collins and Lustig, 1997). The NAFTA does not have any compensatory mechanism at the regional level, but the basic aim that justifies the arrangement is the elimination of barriers to internal trade to facilitate cross border movement of goods and services between the member states, as well as to substantially increase investment opportunities in their territories.

Insert Table 1 about here

Certainly, central mechanisms to close(use “eliminate” or “decrease” en vez de “close”) differences may play an important role. The aim of the European regional policy is to stimulate economic activity in less favoured regions in order to reduce existent disparities by supplementing private sector investment through public support coming from both the states and the Community. The EU has at its disposal four Structural Funds through which it channels financial assistance to address structural economic and social problems in order to reduce inequalities between different regions and social groups. Funds are allocated according to its main objectives. About 30% of the fund is spent on infrastructure investment, including telecommunications, and energy, justified by the fact that disparities in infrastructure are greater than in income (Martin, 1998). Another 30% of the fund is committed to support education, training systems and labour market policies. The remaining 40% is devoted to subsidising industries. In the case of Spain, these kind of funds are likely to have an impact on regional disparities (De la Fuente and Vives, 1995), accounting for the Spanish regions with a sigma coefficient of 0.22 in 1998 (Table 1). Graph 1 plots the per capita GDP relative to the national average in 1991 against its value in 1998, with a 45-degree line to facilitate comparison. From this graph it can be suggested that, in general, high income regions have improved their position, while the remain of regions experience mixed results. From all southern European countries, only Spain shows a very weak reduction of regional income inequalities (Boldrin and Canova, 2001), although with a slight increasing in the last two years of the sample.

Insert Graph 1 about here

In the case of the NAFTA it can be argued that it is too early to judge its effects, in part because its provisions have not taken effect yet (Krueger, 2000). While some tariffs and other barriers were eliminated immediately, others will be phased out gradually until 2008. However, the pattern of disparities can be traced in order to identify structural problems and take action to correct them. In the Mexicans’ case, the peso’s devaluation during NAFTA’s infancy in late 1994 plunged the country into a severe recession and sharply altered trade flows⁴. In this context of recurrent crises since middle 1980s, the Mexican regions have increased their differences, polarising the standards of living, especially between North and South (Juan-Ramon and Rivera-Batiz, 1996), displaying a sigma coefficient of 0.45 in 1998, which is the double the coefficient for Spanish regions (Table 1).

Insert Graph 2 about here

Northern Mexico, close to the US border, has largely benefited from US foreign direct investment. Unlike the EU, the NAFTA does not include any programme of structural funds to alleviate differences in regions. Thus, unlike the EU, regional policy relies exclusively on the Mexican Federal Government. Graph 2 plots per capita GDP in 1993 relative to the national average, against the same variable for year 1998, and draws a 45 degrees line for comparison. Although most of the states stay close the 45 degrees line it is noticeable that regions with higher per capita GDP in 1993 are slightly above the 45- degree line, while regions with lower per capita GDP in 1993 tends to stay over the 45 degrees line or under the line. However, this should be analysed in more detail using regression analysis, as it will enable control by the evolution of relevant explanatory variables.

⁴ Overall U.S.-Mexico trade has increased over the last two decades. Since 1993, annual bilateral trade has grown from $81.5 billion to $128.1 billion for 1996.

4. The model and data

4.1 The model for disparities

In this section we underline the characteristics of the empirical model that picks up the theoretical background discussed in prior sections. The model aims to estimate the impact of public investment and other relevant determinants on regional income. We control for schooling and trade (see Appendix 1 for description of variables included in the regression). Thus, relevant explanatory variables are public investment (PUBINV), private investment (PRIVINV), trade (TRADE) and population schooling attainment (SCHOOL). We should expect public investment to reduce regional inequalities if public investment is allocated to poorer regions. However, if public investment is allocated to encourage trade - and, as we show, if trade propensity is more intense in relatively richer regions - then it might be reasonable to expect an opposite effect.

Public investment might play a double role, either complementary or substitutive. On one hand, it might be a complement to private investment, and therefore should follow the same trend. On the other hand, private investment might be complementing the lack of public investment, and thus its expected to show an opposite sign to the one displayed by public investment.

Educational attainment influences industrial localisation as Krugman (1991b) shows. Regions where there is a high education achievement should show a larger economic activity and in turn a large regional income. Furthermore, a set of regional specific characteristics (Z) are included in equation (1):

\[y _ {i t} = y (P U B I N V, T R A D E, P R I V I N V, S C H O O L, Z)\tag{1}\]

where denotes regional income obtained using per capita GDP, where i refers to the geographical unit and t to time. It can be immediately shown that results obtained using regional income as a dependent variable are equivalent to using any inequality gap measure.

The catch-up process is a extensively proven phenomenon in the growth literature. Part of the variability of regional income is captured by the initial income levels, and thus there is catching up if the "poor grow more rapidly than the rich". Empirically, this feature is measured using a variable that takes the value of the initial regional GDP. However, when using databases that cover large periods of time - as we do here- , initial period fixed does exert some influence. Consequently, we have decided to use a five year GDP lag (GDP5lag) as an alternative variable that measures the catching up effect without being influenced by the use of a specific initial period.

Regional income may be influenced by some regional specific individual effects, e.g., unobserved characteristics of the areas influencing the process and do not change over time⁵. Since specific individual effects might bias the estimates, we allow for regional specific fixed effects and time effects in order to control for variables that might have common effects on the regions in a year, such as business-cycles, etc. Time effects were included as measuring the effect that public investment may introduce an additional concern in that it may take some time to have its effects on income. To this extent we introduced lag variables, although as we show later, including trade as a covariate smoothed the temporal effects of public investment. The empirical model specified is the following:

\[\begin{array}{l} y _ {i t} = \alpha + \beta_ {1} E X P O R T _ {i t} + \beta_ {2} P U B I N V _ {i t} + \beta_ {3} P R I V I N V _ {t} + \beta_ {4} S C H O O L I N G _ {i t} \\ + \beta_ {5} D U M M Y + \beta_ {6} G D P 5 l a g + u + \varepsilon_ {i t} \end{array} \tag {2}\]

The model estimated in equation 2 includes trade as an explanatory variable. As economic theory predicts, trade is commonly determined by income and thus, the above specification could bring endogeneity concerns (Frankel and Romer, 1999; Irwin and Terviö, 2000). To deal with this problem we estimate equation (2) using two-stage leastsquares (2SLS) where geographical characteristics were instruments to proxy export determinants, as used in Frankel and Romer (1999) and Irwin and Terviö (2000). Given that geographical characteristics are a powerful determinants of trade flows (Venables and Limao, 1999), it seems to be a good instrument. Moreover, it shows the characteristics not being correlated with income or government policies (Frankel and Romer, 1999). Results from this specification may allow for the obtainment of estimates of structural parameters of the inequality equations. When these parameters are exactly identified, then it can be shown its identical to instrumental variable estimation (Pindyck and Ribinfeld, 1991).

⁵ In order to control for these effects, and following the patterns of disparities we first use a set of dummy variables to isolate the effect of, in the case of Mexico, being an oil producer (OIL), a region in the (NORTH) or in the (CENTRE) of the country, in the case of Spain dummies employed were being an island (ISLAND) and being in the north border (North Border) to account from the larger economic dynamism that border regions might have.

4.2 The model for trade propensity determinants

An additional purpose here is to measure to what extent public investment has an impact on the export levels on regions. Geographical characteristics as proximity to the border (the border effect) have been recently examined by McCallum (1995) and Crucini et al (1999). McCallum shows, using data on the value of exports and imports, that trade between Canadian provinces was 2200% larger than between Canadian provinces and US locations at similar distance. Therefore, the border effect is still very significant. However, Peach and Adkinson (2000) analysed the border effect in the opposite side of the NAFTA, that is the frontier between US-Mexico. They show that for US border regions the NAFTA agreement has not yet increased economic activity of border regions. Unlike the evidence from NAFTA, Crucini et al (1999) show that in Europe, borders matter less than in the US. Other determinants included in the model are public investment, as promoting access ( by means of reducing transaction costs) associated with trade. We specify a model for exports depending on the following: geographical characteristics (GEO), public investment (PUBINV), and a set of regional characteristics (Z):

\[E X P O R T _ {i t} = T (G E O, P U B I N V, Z)\tag{3}\]

where Z includes a dummy variable (DUMMY) for regional specificities and the initial GDP (INI) that accounts for a "catch up" pattern. The final specification is then (4):

\[\begin{array}{l} E X P O R T _ {i t} = \alpha + \beta_ {2 1} G E O _ {i t} + \beta_ {2 2} D U M M Y _ {i t} + \beta_ {2 3} I N I + \beta_ {2 4} P U B I N V + \\ \beta_ {2 5} P U B I N V _ {t - 1} + \varepsilon_ {i t} \quad (4) \end{array}\]

4.3 The data and empirical strategy

Equations (2) and (4) were estimated employing panels of Mexican and Spanish regions. A first panel of Mexican regions included data for the period 1993 to 1998 (the only period where data on exports is available) and the second panel refers to Spanish regions from 1991 to1998. Data was collected from different sources. Data from Mexico was principally provided by the Instituto Nacional de Estadistica, Geografia e Informatica (INEGI, or National Institute of Statistics, Geography, and Information Systems). Data on Spanish regions was obtained from ICEX and Base de Datos de Conocimiento Regional BBV (appendix 1 explain in detail the sources and limitations of data employed). The variables are inserted in logarithm form except when where expressed as index or as a percentage. OLS and 2SLS estimated are presented in Table 2 and Table 3 for equation (3) and Table 4 for equation (5).

5. Results

This section provides empirical results of this study as follows: results from equation (3) are shown in Table 2 when estimated by OLS and in Table 3 for 2SLS estimates (we report different specifications that appear to be relevant). Table 4 contains estimates of OLS regional trade propensity determinants (auxiliary equations) to enable the understanding of interactions in the two-stage estimation procedure. Hypothetically, variables employed may raise concerns on the existence of endogeneity, heteroscedasticity and multicollinarity issues and on results hypothetically being not robust. To account for all of these issues, we use respectively the Hausman test (Hausman, 1978) for endogeneity, the White test for heterocedastsicity (White, 1980) and we employed the White-Sandwich-Hubert robust standards errors when required. Multicollinearity has been tested through the variation inflation factor (Chaterjee and Price, 1991) . In Table 3, we provide the value of the Hausman test and standard errors. Regressions show a large explanatory power, in all cases was above the 95%. Goodness of the fit increases when fixed effects are included as it is common in the literature. In Table 2, we include some additional diagnostic tests, including the Wald test that checks whether all coefficients equal zero and a Likelihood Ratio F test to account for the random effects fixed effects. Estimates for Mexico account for 32 regions and 6 years, and estimates for Spain include the 17 regions (Autonomous Communities) and 7 years.

The effect of public investment on regional income is significantly different between Mexico and Spain. In the Mexican case, the effect is positive and significant for the public and private investment. However, when fixed effects are included, this effect declines significantly (see table 2).Furthermore, when estimates were corrected by the possible endogeneity of trade, public investment turned out to be significant rather and private investment was not significant anymore (Table 3). Therefore, we can conclude that public investment increased regional disparities in Mexico and that private investment played a complementary role captured by the influenced of public investment . This tendency completes previous results showing that the regional allocation of public investment has a regressive impact on regional inequality (Rodriguez-Oreggia and Costa-Font, 2001). As noticed, public investment evolved in coordination with private investment, and thus both display the same sign before correcting by the endogeneity of trade. Regions attracting private investment benefit from public investment as well.

However, once we correct by possible trade channelling of region income, estimates show that the influence of private investment on regional income turns out to be insignificant whereas public investment maintains a similar coefficient. These results support previous theoretical research indicating that public investment allocation in a process driven by trade integration objectives might improve the core-periphery structures ( e.g., Martin (1999) and Venables (1999)).

Tables 2 and 3 show that unlike in Mexico, in Spain public investment exercised no effect on regional income (a neutral effect), even when estimates were corrected by regional specific fixed effects and when the model was corrected by the existing edogeneity of exports. This result is consistent with most of previous studies . In the Spanish case, the lack of significance of this variable is in tune with the findings of De la Fuente and Vives (1995), who suggested that the small impact on regional disparities during the 1980s is explained by the small redistributive effort undertaken by central government. However, other reasons are that public investment in Spain has intended to overcome the limitations of some regions in attracting private funds (a substitutive role), what may in turn perpetuate income differentials across regions if public investment does not follow private funds (Mulligan and Sala i Martin , 1993). The explanation seems to be the more plausible in Spain from the result obtained. Private investment is a significant and robust determinant of regional income. A 10% increase of private investment rises income in a 1% (see Tables 2 and 3).

Insert Table 2 here.

The effect of trade (EXPORT) is significant and positive both using the OLS procedure for Mexico and for Spain. When we corrected by possible endogeneity of trade in Table 3, exports turned out to be endogenous (as the Hausman test shows) although exports still explained regional income in Mexico. Nonetheless in Spain, once accounting for the endogeneity of trade (Table 3) exports turn out to be non-significant predictors of regional income. Spanish regions that were exporting more intensively during the period examined did not experience similar increases in income per capita. This implies than in Spain, regional inequalities were not significantly driven by regional export capacity as in Mexico, but that other variables might be explanatory, such as private investment and schooling. Actually, when no fixed effects are included, there is a slightly negative effect (significant only at a 10% level) of export in explaining regional income ( see Table 2) as private investment and schooling.

As we show in Table 4, trade is significantly associated with regional specificities and geography. In Mexico, northern latitude (Latitude) and being a centre region (centre) were significant predictors for regional exports. We test as well for the endogeneity of public investment and trade although this was rejected using a Hausman test..

Insert Table 3

Per capita GDP lagged 5 years has a positive and significant effect on regional disparities in Spain and negative in Mexico. In Mexico the evidence indicates the greater the income in a region at the beginning of the period of analysis, the more likely the region is to remain at the top of the income distribution during the period, which is congruent with plots presented in Graph 2. Among Spanish regions, this variable is consistent with Graph 1 that showed that richer regions seem to benefit from regional integration. Since richer regions are the ones exporting more as well, GDP5Ylag captures the variability that would correspond to exports if the former variable was excluded from the analysis.

Human capital exercised a very strong influence on regional inequalities. The variable for schooling in Mexico and Spain was significant in all regression albeit displaying a very small effect and a very important effect in Mexico. Human capital differences are still high and they exert anplaces a significant effect on disparities in Mexico. The ratio that accounts for the distance between the maximum and the minimum regions in educational achievement terms has moved from a value of 2 to 1.8 from 1993 to 1998. Nonetheless, we should bear in mind the long run effects of human capital investment. In the Spanish case education between regions is more homogenous as the mean ratio of schooling between the maximum and the minimum is only 1.3, and the average years of schooling are higher than in the Mexican regions.

Insert table 3

6. Conclusions

This paper shows that the allocation of public investment as a supply side instrument to reduce regional disparities might be ineffective as a result of the counteracting effects of trade integration arrangements. Inequalities are sensitive to the effects enhanced by trade liberalisation. In Spain, public investment seems to act as a substitute of private investment and places no influence on regional income and, as a result, on regional inequalities. Conversely, in Mexico, the effect of public investment increases income disparities. This result may be explained in the Hirschman (1958) theoretical framework. Mexico may be in the first stage of the public investment process, and thus public investment complements private investment.

The role of education as explaining income inequalities was significant in both countries although more important in Mexico than in in Spain. Lower levels of educational attainment in developing countries tend to foster localisation of economic activity to in regions where skilled workers are available . Therefore, implications of this study show that Mexico should re-address public investment policies and invest more in reducing disparities in human capital in order to reduce regional disparities at a national level. Human capital may be less explanatory of regional disparities within developed countries, since differences in educational achievement are smaller.

These results are consistent with NAFTA being purely a trade promoting arrangement and the EU promoting (although unsuccessfully) regional cohesion and additionally, they show that the stage and the development of regional trade integration matters along with the combined role of public and private investment. Internal trade polices should guarantee that all regions are to participate in the regional integration process if regional inequalities are to be kept to a minimum. Finally, propensity to trade is mainly explained by the border effect. Therefore, even though countries analysed are participating in trade integration arrangements, political boundaries still remain between countries. However, the deeper the trade integration arrangement the lower the border effect.

What can we learn from the theoretical discussion and the empirical application? First, the discussion by Martin (1999) is better suited to an international context than explaining regional disparities within countries. Secondly, the argument seems to apply to developed countries but more work needs to be done to fit developing countries needs. Martin (1999) states that a public policy directed towards decreasing the costs of innovation can attain both objectives of growth and more equity. As human capital is needed to carry out innovative production with increasing returns, to what extent will policies which decrease the costs of innovation be effective? Disparities between Mexican regions are impacted more by the effect of disparities on human capital. Therefore, public policy that pursues this this objective would enforce concentration of increasing returns industries in the core of trade areas, while periphery or less developed, with less human capital, will specialise in constant returns to scale industries as in Krugman (1991). The statement of Lucas (1993) that human capital has the main role in explaining differences between regions is reinforced here as well.

Even though Martin includes education infrastructure and labour markets in subsidies to reduce innovation costs, the lack of backward and forward linkages would not attract manufacturing to the less developed regions (Fujita, Krugman and Venables, 1999), and equalisation of income would come from the migration effect. In the case of policies directed to the enhancement of infrastructure facilitating domestic rather than international trade, as in Martin and Rogers (1995), it is difficult to elucidate their effect from the empirical evidence. However, if domestic infrastructure were going to be suggested as prevailing policy, much work should be done in order to elucidate to what extent benefits from spillovers would be greater than benefits from relocation of industry in low development zones, the operation of the effect in the opposite direction, and favouring concentration again. In the Spanish case, Martin's proposed policies seem more reasonable. Regions already show large human capital availability - and as some economic effects depend of the level of human capital offered in the region (e.g., Borensztein et al., 1998)- a drop of innovation costs could attract manufacture and increase the standards of living. Thus, less developed Spanish regions could support to some extent increasing returns industry.

APPENDIX 1: VARIABLE DEFINITION AND SOURCES

Sources of Mexican data

LGDP. Data for GDP was obtained from the Instituto Nacional de Estadistica, Geografia e Informatica (National Institute for Statistics, Geography and Information Systems, or INEGI) webpage, available at www.inegi.gob.mx. Data for population was extracted from the Statistical Annexes of the Presidential Address to the Nation, various years, and INEGI database.

TRADE: Ratio of state's export to the state's GDP, calculated with export data obtained from the Secretaria de Comercio y Fomento Industrial (Secretary of Commerce and Industry), and with data for state’s GDP based on INEGI database.

DUMMY: 1 for Campeche and Tabasco (states with higher share of oil production), 0 for other states.

PUBINV: Log of the per capita public investment in the state. Data for population as in LGDP, data for public investment extracted from the Statistical Annexes of the Presidential Address to the Nation, various years.

SCHOOL: Log of the average years of schooling of population aged 15 and over. Data was obtained from the Statistical Annexes of the Presidential Address to the Nation, various years.

PRIVINV: Gross capital formation distributed according to the regional share of state’s total credits.

GDP5Ylag Log of the per capita GDP lagged 5 years.

Sources of Spanish data

LGDP and population data were collected from Fundación BBV http://bancoreg.fbbv.es:1268/menu.html. Base de datos de conocimiento regional.

TRADE: is the ratio of state's export to the state's GDP, calculated with export data obtained from ICEX and Contabilidad Regional de España (INE)

DUMMY: 1 for islands (Islas Baleares and Islas Canarias) and 0 for the rest of the states such as another dummy for north border regions.

PUBINV: Obtained from Fundación BBV and Contabilidad Regional de España (INE)

SCHOOL: Log of the average years of school. Data obtained from IVIE (Instituto Valenciano de Investigaciones Económicas) and Bancaja, available at http://www.ivie.es/bdatos/.

GDP5Ylag Log of the per capita GDP lagged 5 years.

NOTES

1.By the time NAFTA went into effect January 1st, 1994, the United States and Canada already had five years of experience with their Free Trade Agreement.

2.Tariffs on other products fall over a longer period of time. For example, U.S. automotive parts entering Mexico see a 75 percent reduction in duties over the first five years of NAFTA, with the rest phased out over 10 years.

References

1. Banamex (1999). State GDP: modification of the base year, Review of the Economic Situation of Mexico, vol. LXXIV, no. 873, 207-214.
2. Ben-David, D. (2000). Trade and the rate of income convergence, National Bureau of Economic Research Working Paper, no. 7642.
3. Ben-David, D. and Loewy, M. B. (1998). Free Trade, growth and convergence, Journal of Economic Growth, vol. 3, 143-170
4. Boldrin , M and Canova, F (2001). Inequality and convergence: reconsidering European regional policies. Economic Policy 16:207-253.
5. Borts, G.H (1960). The Equalisation of returns and regional economic growth. American Economic Review, 50: 319-347.
6. Borensztein E., De Gregorio, J. and Lee, J-W. (1998). How does foreign direct investment affect economic growth? Journal of International Economics, vol. 45, 115-135.
7. Bosch, N and Escribano, C (1988). Regional allocation of public funds: an evaluation index. Environment and Planning C: Government and Policy 20:1323-33.
8. Bosworth, B., Collins S. M. and Lustig N. C. (1997). Introduction. In Bosworth, B., Collins S. M. and Lustig N. C. (Eds.) Coming together? Mexico-United States relations. Brooking Institution Press: Washington.
9. Chatterjee S, Price B (1991) Regression analysis by example, 2^nd edition. John Wiley & Son, New York.
10. Crucini, M; Telmer, C and Zachariadis, M (2000). Dispersion in real exchange rates. University of Pennsylvania Working Papers, 2000.
11. De la Fuente, A (1996). Inversión pública y redistribución regional: el caso de España en la década de los ochenta. Papeles de Economía Española, 67: 238-51.
12. De la Fuente, A. and Vives, X. (1995). Infrastructure and education as instruments of regional policy: evidence from Spain, Economic Policy, vol. 20, 11-40.
13. Faini, R. 1984. Increasing returns, non –traded inputs and regional development, The Economic Journal, vol. 94, 308-23.
14. Frankel, J. A. and Romer, D. (1999). Does trade cause growth? American Economic Review, vol. 89, no. 3, 379-399.
15. Fujita, M., Krugman, P. and Venables, A. J. 1999, The spatial economy: cities, regions, and international trade, Cambridge, MIT Press.
16. Hall, P (1993). Forces shaping urban Europe. Urban Studies 30: 893-898.
17. Hausman, J (1978). Specification tests in econometrics. Econometrica 52:1219-1240.
18. Hirschman, A. (1958). The strategy of economic development, New Haven, Yale University Press.
19. Irwin, D. A. and Terviö, M. (2000). Does trade raise income? Evidence from the twentieth century, National Bureau of Economic Research Working Paper, no. 7745.
20. Juan-Ramón, V. H. and Rivera-Batiz, L. A.. 1996. Regional growth in Mexico: 1973-93, International Monetary Fund Working Paper, no. 96/92.
21. Krueger, A. O. (2000). NAFTA's effects: a preliminary assessment, The World Economy, vol. 23, no. 6, 761-775.
22. Krugman, P. (1991). Increasing returns and economic geography, Journal of Political Economy, vol. 99, 483-499.
23. Krugman, P. (1991). Geography and trade. Cambridge, MIT Press.
24. Limao, N. and Venables, A. J. (1999). Infrastructure, geographical disadvantage, and transport cost, The World Bank Policy Research Working Paper, no. 2257.
25. Lucas, R. E. Jr. (1993). Making a miracle, Econometrica, vol. 61, no. 2, 251-272.
26. Martin, P. 1998.Can regional policies affect growth and geography in Europe? The World Economy, vol. 21, no. 6, 757-774.
27. Martin, P. (1999). Public policies, regional inequalities and growth, Journal of Public Economics; vol. 73, no. 1, 85-105.
28. Martin, P. and Rogers, C. A. (1995) Industrial location and public infrastructure, Journal of International Economics, vol. 39, 335-351.
29. McCallum, J (1995). National borders matter: Canada-US regional trade patterns. American Economic Review, 85:615-23.
30. Mulligan, C.B and Sala i Martin, X (1993). Transitional dynamics in two sector model of endogenous growth. Quarterly Journal of Economics, 108:737-73.
31. Peach, J.T and Adkinson R.V (2000). NAFTA and economic activity along US-Mexico border. Journal of Economic Issues, 24:481-489.
32. Puga, D (1999). The rise and fall of regional inequalities. European Economic Review, 43:303-334.
33. Quah, D (1993). Galton´s fallacy and the convergence hypothesis. Scandinavian Journal of Economics, 95: 427-443.
34. Quah, D (1996a). Empirics of economic growth and convergence. European Economic Review, 40:1353-75.
35. Quah, D (1996b). Regional convergence clusters across Europe. Working paper 1286, Centre of Economic Performance, London.
36. Rodriguez-Oreggia, E. and Costa-Font, J. 2001. Are regional inequalities decreasing with public investment? Evidence from Mexico, London School of Economics, mimeo.
37. Siebert, H (1969). Regional economic growth: theory and policy, Scranton, Intranat, Textbook Co.
38. Venables, A. J. 1999. Regional integration agreement, a force for convergence or divergence? The World Bank Policy Research Working Paper, no. 2260.
39. Venables, A. J. and Limao, N. (1999). Geographical disadvantage: a Heckscher-Ohlin-von Thunen model of international specialization, The World Bank Policy Research Working Paper, no. 2256.
40. White, H (1980). A heterocedasticity -consistent covariance matrix estimator and a direct test for heterocedasticity. Econometrica, 48:817-830.

Table 1. Sigma coefficient for convergence across Mexican and Spanish regions.

Source: Own calculations

Graph 1. Per capita GDP relative to the 1991-1998 average. Spanish regions (Mean=1)

Table 2. Determinants of regional income inequality in Mexico and Spain (OLS) (Dependent variable: regional income disparities)

Notes: Hubert/White/Sandwich standard errors in parentheses. ***, **, and * denote significance at 1, 5 and 10% respectively. All regressions include time effects. Mex years 94- 95

Table 3. Estimates of the determinants of regional income in Mexico and Spain (2SLS). (Dependent variable: regional income disparities)

Note: Hubert/White/Sandwich standard errors in parentheses. ***, **, and * denote Significance at 1%, 5% and 10% respectively. All regressions include time effects. Instrumental variables are latitude north, lags of public investment and same variables for remaining concepts. a. The Hausman endogeneity test allows us to choose between the OLS and the alternative estimation where instruments were used to account for the hypothetical endogeneity of the relevant variable. Formally, the tests can be written as:

\[q = (\hat {\beta} _ {O L S} - \hat {\beta} _ {I V}) [ V a r (\hat {\beta} _ {I V}) - V a r (\hat {\beta} _ {O L S}) ] ^ {- 1} (\hat {\beta} _ {O L S} - \hat {\beta} _ {I V}) \approx \chi_ {p} ^ {2}\]

Where q refers to the degrees of freedom. It q is large enough to reject the null hypothesis of endogeneity the OLS estimates are accepted and endogeneity is not a problem, whereas if the coefficient is small then an alternative two stage least squares (2SLS) estimate should be computed.

Table 4. Regional trade (export) propensity determination. OLS estimates(Dependent variable: regional export/GDP)

Note: Hubert/White/Sandwich standard error. Standard errors in parentheses. ***, **, and * denote significance at 1, 5 and 10% respectively. All regressions include time effects. a The null hypothesis was assuming endogeneity between public investment and trade.