Introduction

Since the mid-1990s, the Mexican economy has been exposed to significant supply and demand shocks, first with the start of NAFTA in January 1994, and then with the devaluation of the Mexican peso in December of that same year. On the positive side, as a result of NAFTA, Mexico's total trade to GDP ratio increased from 34% in 1993 to 64% in 2000. Foreign Direct Investment (FDI) inflows went from 4.3 billion U.S. dollars in 1993 to 10.9 billion U.S. dollars in 1994. Between 1994 and 2006, Mexico received more than 217 billion U.S. dollars in FDI (totaling around 40% of its GDP). Nevertheless, following the devaluation of the Mexican peso, in 1995 inflation surpassed 50% and real GDP fell by 6%. In terms of international migration, and as a result of considerable job losses, the net outflow of people rose from 412 thousand in 1993 to 625 thousand by the year 2000; this is roughly about 0.5% of Mexico's total population each year.¹ All these shocks have had positive and negative effects on the aggregate economy which are considerably difficult to manage at the regional level.

According to the so-called “compensation hypothesis” proposed by ^{Rodrik (1998)}, a more open economy requires a larger government capable of attenuating the effects of externally generated economic instability, which in turns prompts the demand for social security. In spite of the importance that this relation has for developing countries, only a few studies have examined it at the subnational level or by employing alternative non-trade-related openness measures more suitable for developing economies. This paper revisits the role of openness in the size of government across Mexican states and proposes the use of alternative measures of international capital and labor mobility as proxies for openness rather than more traditional trade-related measures.

We feel that the assessment proposed here is important for economic development and contributes to the existing literature in at least three important ways. First, ever since the seminal papers by ^{Rodrik (1998)} and ^{Alesina and Wacziarg (1998)}, the literature has mostly employed total trade and trade volatility measures as proxies for openness in order to assess their relation to government size. While these measures are important in assessing the degree of openness of any economy, developing countries have been exposed to openness—traditionally and in recent years—in different ways than industrial economies.²

Historically, many developing countries have been significantly more exposed to the openness of their labor market, by means of legal and illegal immigration, than to trade. In Mexico, for instance, trade openness started in 1985 when the country joined the General Agreement on Tariffs and Trade (GATT). However, labor openness, in the form of international legal and illegal migration, came into existence much earlier.³ Throughout international migration, the labor market has worked as a buffer or shock-absorber device for many developing economies engaged in market liberalization, including Mexico (^{Robertson, 2000}). Moreover, trade openness is usually accompanied by financial openness and, in recent years, emerging markets have benefitted greatly from the latter. Particularly since the early 1990s, foreign direct investment has played a major role in the development of several developing economies.⁴ For instance, in the case of Mexico, ^{Ramirez (2006)} documents gains in labor productivity arising from FDI flows.

While the studies relating government size to capital flows are not many, recently ^{Gemmell, Richard, and Sanz (2008)} have employed the inward stock of FDI as a proxy for openness across a sample of OECD economies. They explore the effects of these flows on government spending and find that inward FDI does not affect the size of government spending but that it affects its composition by shifting it toward social spending. Also, ^{Kimakova (2009)} explores this alternative dimension of openness across a sample of 87 emerging and industrial economies. She finds that trade alone does not capture some of the important elements that characterize the modern international economy and observes a statistically significant and strong positive relation between government size and financial openness.

In the absence of reliable trade statistics at the state level for Mexico, we employ two alternative proxies of openness: FDI flows and international migration rates. As far as we know, no proxies for openness associated with the labor market have been previously employed in the literature that studies the determinants of government size. In our analysis states with higher migration flows are not necessarily more open to trade or FDI. Instead, the idea is that more open states are more exposed to foreign shocks and, as a result, are the source of more migration. While different in nature to FDI and other previous proxies for openness employed in the literature, in our model variations in the rate of international migration captures the effects that openness has on, for instance, employment in Mexican regions. Henceforth, along with FDI flows, international migration controls for the effects of shocks linked to distortions affecting the capital and labor markets.

Second, there are only a few studies analyzing the effects of openness on subnational government size. To the best of our knowledge, this relation has only been observed in the case of a few industrial economies. ^{Pickup (2006)}, for instance, examines the relation between trade openness and the magnitude of government spending in Canadian provinces conditional on variables such as the ideology of the government and the degree of provincial unionization. He finds that, while differences arise according to those two conditional variables, the traditional compensatory response to trade openness has been the norm across Canadian provinces. ^{Balle and Vaidya (2002)} explore the relation between trade openness and government size across US states. They find that spending in public welfare and health services is positively affected by the degree of the state's openness. Similarly, ^{Fatas and Mihov (2001)} examine the relation between government size and output volatility across US states and find that states with larger governments are exposed to lower output volatility. As far as we know, our study is the first that analyzes the relation between government size and openness across the states of a developing economy.⁵

The analysis of developing countries is particularly relevant because they are fundamentally different from industrial economies. In general, developing economies possess smaller governments and are more open to trade and capital flows than developed economies (^{Mollick, Cabral, & Carneiro, 2011}). Fiscal policies in developing countries have also different characteristics than in industrial economies. In general, governments in developing countries have a smaller tax base, tend to be more centralized and to depend more on central government transfers. In Mexico, for instance, between 2005 and 2010 states received more than 85% of their total revenue in the form of intergovernmental transfers and their own revenue was less than 7% of their total revenue (^{Abbott, Cabral, Jones, & Palacios, 2015}).

Third, this study builds on recent literature that not only looks at the government size-openness relation but also accounts for the link between these two variables and the size of a country, as proposed by ^{Alesina and Wacziarg (1998)}. In his seminal paper ^{Rodrik (1998)} argues that more open economies need larger governments in order to respond to ‘external shocks’ and stabilize their economies. ^{Alesina and Wacziarg (1998)} cast doubt on Rodrik's argument and point out two important causal relations linking openness to government size. According to them, due to economies of scale, the larger a country, the lower its per capita government expenditures; consequently, small economies need larger government spending. At the same time, the smaller a country, the greater the incentives for it to be and remain open in order to gain access to larger markets abroad. As a result of these two underlying forces, more open economies tend to have bigger governments. We start our empirical analysis following the estimation of a parsimonious set of equations proposed by ^{Ram (2009)} to evaluate those three causal relations anticipated by ^{Alesina and Wacziarg (1998)}. Thereafter, we build a more comprehensive model to explore the relationship between government size and openness controlling for several important factors including country size.

Having those underlying relations in mind, we employ a panel comprising all the 32 Mexican subnational entities over the period 1996-2006 and make use of two alternative measures of international capital and labor mobility to assess the relation between openness and government size. Following the estimations, we find some interesting results. First, using the parsimonious set of equations proposed by ^{Ram (2009)}, we confirm the negative association between country size and both government size and openness. We also find systematic positive effects of our openness measures on the size of the states’ total government spending. Second, despite accounting for country size and other traditional determinants of government size, a more comprehensive model confirms the positive association between the latter variable and openness. These two results suggest that while the underlying relations pointed out by ^{Alesina and Wacziarg (1998)} exist, they are not producing the expected positive relation between openness and government size. Third, focusing on three subcategories of spending associated with social welfare (education, health and poverty alleviation programs), we observe that FDI flows are only a weak determinant of states’ social spending, while labor openness, in the form of international migration, has a significant and even more important impact on some of these categories than on total spending.

The rest of this paper proceeds as follows. The following section describes the data employed in the empirical analysis and presents some descriptive statistics. The third section explains the empirical methodology used and the fourth section reports our empirical results. Final section presents the main conclusions and implications of this paper.

Data

Our data set comprises information of all the 32 Mexican subnational entities (31 states and its capital) for the years between 1996 and 2006. The sample looks at a relatively stable period, it avoids using the years prior to 1995, in which an important domestic crisis took place, as well as the global financial crisis years, which also had important repercussions for the Mexican economy. Government expenditure is taken from the database Estadísticas de las Finanzas Públicas Estatales y Municipales published by Mexico's National Institute of Geography and Statistics (INEGI). Only the subcategories of expenditure, employed in the robustness check section, are available since 1998, for 3 of the categories, and since 1999, for one of them. In building output, density, dependency ratios and urban population, the main source of our data is INEGI. For international migration rates and population figures we use data from the National Population Council (CONAPO). Finally, FDI figures were obtained from the Mexican Ministry of Economy.

Tables 1 and 2 present some descriptive statistics and correlations for the main variables of our model. The average government spending to GDP ratio is 0.118, the highest ratios are in the southern states of Tabasco (0.25), Oaxaca (0.24) and Chiapas (0.21), while the lowest ratios are observed in the capital of Mexico, Distrito Federal (0.052), Coahuila (0.068) and Nuevo Leon (0.069). Output per capita, the first control variable in the model, observes a strong negative correlation with government expenditure (correlation coefficient of −0.57). Indeed, we observe that the two entities with the lowest government spending to GDP ratios are also those that possess the highest GDP per capital, Distrito Federal (36,000) and Nuevo León (25,641) and that two of the three entities with the highest GDP ratios are among those with the lowest GDP per capita, Oaxaca (6163) and Chiapas (6237).⁶ These figures might suggest that government spending has been oriented toward the improvement of social welfare in relatively less developed states. Figure 1 shows the geopolitical allocation of government expenditure. It is clear from the picture that states with a lower than average expenditure is mainly located in the south and center of Mexico. Interestingly, none of the states on the northern border poses a higher than average expenditure.

Note: Elaborated with data from the Mexico's National Institute of Geography and Statistics (INEGI).

Figure 1 Geographical allocation of government expenditure as a proportion of GDP.  

In the absence of reliable trade figures at the state level, we measure openness by employing two different alternative indicators: FDI flows and net international migration. For the period observed, FDI represents on average around 1.57% of GDP across all 32 states. The largest receptor of FDI is the capital of Mexico, Distrito Federal, 8.75%, followed by the northern border states of Baja California and Nuevo Leon, with about 5.1% and 4.8% of their respective GDP each. The states with the lowest FDI reception are the southern states of Chiapas and Oaxaca, followed by the central state of Michoacán (less than 0.2% of their respective GDP in all three cases). Figure 2 shows the geographical location of states above and below the average annual FDI to GDP ratio. The concentration of the highest receivers of FDI along the northern border and the center of Mexico is evident from this diagram.

Note: Elaborated with data from the Ministry of Economy, FDI National Registry (RNIE).

Figure 2 Geographic location of FDI across Mexican states.  

The international migration rate is defined as the outflows of people leaving a state to live abroad minus the inflows of people arriving from abroad divided by each state's total population. We observe that only three states present a net inflow of people: Quintana Roo (0.75%), Baja California (0.19%) and the capital of Mexico, Distrito Federal (0.05%). These entities typically receive a considerable number of foreign immigrants due to tourism and business activities. Among the states from which more migrants go abroad are Michoacán (1.63%), Zacatecas (1.5%) and Nayarit (1.32%). From our two openness indicators only international migration observes a positive correlation with the government spending to GDP ratio (correlation coefficient of 0.34), suggesting that greater international migration positively influences the size of the government. The geographical pattern of international migration is shown in Figure 3. Clearly, states with more than average net migration rates extend all along Mexico from Chihuahua in the northern border to Oaxaca, with no particular geographical concentration in the south or the north.

Figure 3 Geographical location of international migration net flows across Mexican states. 

Population in our model is a proxy for size. In Table 1 this is given in thousands of people. The state of Mexico is the one with the highest average population (13.1 million), followed by Distrito Federal (8.7 million) and Veracruz (7.1 million). The states with the lowest population are Baja California Sur (453 thousand), Colima (536 thousand) and Campeche (716 thousand). The last two columns in Table 1 contain two other control variables commonly employed in the literature that analyze the role of openness in government size: population density and urban population. The average population density in Mexico is 267 people per square kilometer. The entities with the highest density are Distrito Federal (5700), the state of Mexico (609) and Morelos (312), while those with the lowest density are Baja California Sur (6), Durango (12), Campeche (13) and Chihuahua (13). In terms of the proportion of urban population, the figures suggest that on average, 65% of the Mexican population lives in urban areas. Finally, the entities with the highest urban population proportions are Distrito Federal (100%), Nuevo Leon (92%) and Coahuila (90%). The states with the lowest levels of urban population are Oaxaca (27.5%), Chiapas (31.6%) and Guerrero (37%).

The model

We start our analysis by estimating ^{Ram's (2009)} parsimonious set of equations that look at the relation between government size, country size and openness as proposed by ^{Alesina and Wacziarg (1998)}. The three regression equations are:

(1)

(2)

(3)

where γ_i is a set of state specific effects, μ_t is a groups of time specific effects, G/Y_it is the logarithm of government spending to GDP ratio in state i at time t, Size refers to the logarithm of the state's size, which is proxied by the log of population, Y/L is the GDP per capita and Openness is the log of our non-trade measures of openness.

Following ^{Alesina and Wacziarg (1998)}, we expect the coefficients α₁ and β₁ in Eqs. (1) and (2), respectively, to be negative. This would suggest that (1) sharing government spending among a larger population reduces per capita expenditure (α₁ < 0); (2) market size enhances productivity through economies of scale; hence, relatively smaller states are more eager to become and remain open in order to gain access to larger markets (β₁ < 0). Given the negative relation of country size with government size and with openness, in Eq. (3) we expect the coefficient for openness (δ₁) to be positive.

After the estimation of this parsimonious set of equations, we verify the robustness of the results by estimating an augmented version of Eq. (3) which controls also for factors such as country size, income per capita, urbanization rates and population density. This model is akin to previous estimates by ^{Rodrik (1998)} and by ^{Benarroch and Pandey (2008)}. We first estimate the model for total government spending, then revisit the augmented model for alternative categories of government spending. The augmented model is as follows:

(4)

where Urban is the urbanization rate across states (i.e. the proportion of urban population living in urban areas) and Density the logarithm of the number of people per square kilometer. With respect to previous studies employing panel data methods (see, for instance, ^{Rodrik, 1998}; ^{Garen & Trask, 2005}; ^{Benarroch & Pandey, 2008}, among others), the model proposed in equation (4) offers the advantage of explicitly controlling for country size as done by ^{Ram (2009)} in his revision of the causality suggested by ^{Alesina and Wacziarg (1998)}.

Results

Concluding remarks

This paper has examined the relation between openness and government size across a sample of 32 Mexican entities over the period 1996-2006 employing two alternative new measures of openness linked to international capital and labor mobility: FDI flows and the net international migration rate. Using a set of parsimonious equations, the study confirms the existing negative relation between country size and both government size and openness, as well as the positive relation between government size and openness, as proposed by ^{Alesina and Wacziarg (1998)}. A more comprehensive model that controls for country size and other important determinants of government spending, confirms the positive association between government size and openness. All these results are observed for our two measures of capital and labor openness.

A disaggregation of total government spending to concentrate the analysis on social spending allocated to education, health and poverty alleviation programs renders different results for our two measures of capital and labor openness. We find no significant effect of capital flows on the size of government for education and poverty alleviation programs, but a weakly significant negative effect on health expenditure. We argue that this has to do with the nature of capital flows which exert little influence on these social categories of expenditure. For international migration we observe positive and strong effects of migration on health and poverty alleviation programs but not on education. These results are consistent with the idea that after household heads migrate, the need for programs that target health and poverty alleviation increases. This could also be true for education; however, we do not find significant evidence that states’ spending has favorably responded to such needs.

As for our proposed measure of openness, the results suggest that, in the absence of trade-related openness measures, the international migration rate is fully capable of capturing the effects of external shocks affecting the Mexican economy. Further investigation on how international migration could capture overall macroeconomic instability across more diverse, less homogenous economies is needed.