On a Highway to Where? The Political Economy of the Distribution of Public Infrastructure in Developing Federal. Democracies

On a Highway to Where? The Political Economy of the Distribution of Public Infrastructure in Developing Federal Democracies Lucas González (UCA-UdeSA-UNSAM), Marcelo Leiras (UdeSA), Ignacio Mamone (UCA) (Draft. Do not cite without authors permission) Abstract: Redistribution in very unequal developing countries is a divisive issue. Most researchers recognize a crucial role of the state in redistributive politics. Despite this, there is little we know about the factors that shape government redistribution. The mechanisms to reallocate wealth from one group in society or region in a country to others can take varied forms, from subsidies to particular industries to construction projects in some districts. This paper studies political and economic factors that affect the allocation of interregional redistributive transfers in Argentina, a highly unequal country in Latin America (the world s most unequal region). The focus is on funds with high redistributive impact and over which the central government may have large discretion: public infrastructure. Using original data for the 24 districts in the federation (2000-2009), this paper shows how political factors linked to the incentives presidents face in relation to the formation and maintenance of electoral coalitions and legislative majorities are crucial to explain the regional distribution of public infrastructure. These factors are more relevant than programmatic ones associated to equity or efficiency criteria. We also show that the legislative does not influence the outcome, contrary to what several authors found for the American case. Overrepresentation does not seem to influence results either, contrary to what other scholars found for the distribution of structural funds in the European Union. The main argument is that presidents in Argentina tend to be safe political investors, inefficient economic agents, and poor redistributive actors. It is only when the poorer districts are core political allies and represent a safe political investment that the redistributive impact is significant. We discuss the implications of these findings for the comparative debate. Prepared for delivery at the International Conference on Social Cohesion and Development, OECD, Paris 20-21 January 2011. 1

On a Highway to Where? The Political Economy of the Distribution of Public Infrastructure in Developing Federal Democracies Introduction Redistribution in very unequal developing countries has historically been a sensitive and divisive issue. How much redistribution can be tolerated by the richest strata of society? How to achieve it? These have been long-standing questions for which available answers still generate sharp disagreements among scholars and policymakers. Most researchers recognize a crucial role of the state in redistributive politics. Despite this general appreciation, there is still little we know about the factors that shape government redistribution. In Lindbeck and Weibull (1987, p.273) words, [t]he driving forces behind government-induced redistributions of income and wealth are still not well understood. The government can redistribute wealth to compensate the effects of an uneven distribution of wealth among individuals, economic or social groups, or across regions in a territory of a country. In almost every country, but especially in federal ones, the central government transfers large amounts of funds across regions. Sometimes institutions specifically designed for that purpose regulate these transfers. These institutions are relatively stable over the years and, for that reason, the total amount and regional share of these funds tend to remain reasonably fixed for large periods of time, without changing much from year to year. This is the case of revenuesharing schemes in most federal countries and in some unitary cases. Some other times, regional transfers are instead a by-product of a general government program or policies (Persson and Tabellini, 1996, p.980). These mechanisms to reallocate wealth from one group in society or 2

region in a country to others can take a variety of forms, from subsidies or tariffs to particular industries, to construction projects in some districts (Dixit and Londregan, 1996, p.1133). The discretion of the central government, for instance over the territorial allocation of infrastructure investment, tends to be larger than in the other schemes: it is usually more straightforward to reallocate highway or road funds from one region to another than it is to redistribute wealth through public consumption or employment policies (Sollé Ollé, 2010, p.297). The main goal in this paper is to study the political and economic determinants that affect the allocation of regional redistributive transfers. The focus is on funds over which the central government may have large discretion and high redistributive impact: public infrastructure. 1 The regional allocation of infrastructure investment is a way to redistribute money from some regions (those that pay taxes to finance these funds) to others (those in which the investment is actually made) (Sollé Ollé, 2010, p.299). Exploring the dynamics affecting this distribution is crucial to explore the regional impact of public investment to foster growth and development. Several authors claim infrastructure transfers should be allocated to those districts with the greatest relative need (according to either efficiency or equity criteria), but several others claim that in many cases they have been distributed according to political considerations (Holcombe and Zardkoohi, 1981, p.393). The main question in this work is which are the political and economic determinants in the distribution of public infrastructure funds in Argentina between 2000 and 2009? We study the politics of redistribution in a highly unequal Latin American case (the latest value of the Argentine gini coefficient is.51 in 2005; WIID, 2008); a region that is, in fact, the most unequal 1 Dixit and Londregan (1996, p.1135) call them as soft money, or funds that can be allocated either to the political machine in Chicago or to swing states, such as California during the Clinton years. 3

region in the world. 2 Argentina is the case that suffered the sharpest deterioration in income distribution of the region between 1976 and 2003, 3 partially recovering between 2003 and 2010. 4 Federal grants in Argentina have been an important way to shifting a significant amount of resources from some regions to others (Porto and Sanguinetti, 2001, p.238). The total amount of public infrastructure has increased from AR$1 billion in 2000 to AR$ 11.5 billion in 2009 (or a 1150% increase during the period), becoming a key and among the most important redistributive tools in the hands of the central government. The redistributive dynamics behind these changes in income distribution have, as it will be demonstrated, large implications for the politics of redistribution in this case and other federal countries. The literature on the distribution of public funds has mainly concentrated on the effects of programmatic determinants based either on efficiency or equality criteria. Only more recently, some authors have begun to explore the effects of political factors. This paper is an empirical contribution to this debate because it provides further evidence on the relevance of these programmatic, political, and economic determinants for other cases not yet studied. But it is also a theoretical contribution, as it introduces more specification and precision to the argument. The main claim in this study is that presidents use redistributive transfers as a tool to build up political support both during times of electoral campaigns and when they need to pass their reform agenda in Congress. Those districts that are part of the core of the presidential coalition are more likely to get more infrastructure grants than swing districts. The least favored provinces are those controlled by the opposition. In getting electoral and legislative support, presidents also 2 The average gini coefficient to measure income distribution in Latin America was.52 in 2008, the largest compared to other regions in the world (ECLAC, 2009). 3 In Argentina, the average gini for the period 1972-2005 is.44; with a minimum value of.35 in 1976, reaching a maximum value of.53 in 2003 (WIID, 2008). 4 Gini in Argentina recovered from.533 in 2003 to.51 in 2005 (this is the last year for which there is data available from the WIID, 2008), and official figures indicate that income inequality has been reduced since then and until 2010. 4

care about the capacity of the district to contribute politically during election times and legislative negotiations: they tend to support core districts in which governors have more political support and are more likely to continue in office. As far as we know, none of the studies in the topic has taken into consideration the role of the political power of governors as a critical factor to influence the outcome. This paper makes two main contributions. One, regarding the case analyzed, it shows that presidents in Argentina tend to be safe political investors, inefficient economic agents, and poor redistributive actors. It is only when the poorer districts are core political allies and represent safe political investment that the redistributive impact is significant. Second, and in relation to the comparative literature, it reveals that the politics of infrastructure redistribution can be fundamentally determined by executive politics. We found poor evidence to support the relevance of the legislative in influencing the distribution of infrastructure funding. This is contrary evidence to what other scholars found in the United States (see for instance, Holcombe and Zardkoohi, 1981, p.399). In addition, we found that institutional rules, mainly legislative overrepresentation, is not a relevant factor to explain redistribution, contrary to what several authors found for this case and others, especially the European Union (Rodden, 2002, 2010). We organize the paper as follows. First, we discuss the theoretical literature on the topic and, based on this review, present the main theoretical claim. Second, we operationalize the variables and provide the data sources for the main and other competing hypotheses. Third, we introduce the methodological approach selected to analyze the data. In the fourth section we put forward the empirical findings and discuss them in the last one. 5

The Political Economy of Redistributive Transfers State intervention on the economy can help promoting development (Wade, 1990; Evans, 1995; Kohli, 2004) or hinder it (Scott, 1998). Several scholars have extensively debated on what kind of state intervention is necessary or desirable to foster growth, development, and redistribution (Dixit and Londregan, 1996; Persson and Tabellini, 2000; Lindbeck and Weibull, 1987; Cox and McCubbins, 1986). Part of the literature on this topic has claimed that infrastructure investment is a very important redistributive tool in the hands of the state. Some of these authors concentrated efforts in identifying the determinants for the regional distribution of these funds (Dixit and Londregan, 1998; Randolph and Hefley, 1996; Rodden, 2002, 2010; Castells and Solé Ollé, 2005; Bosch et al., 2010; Sollé Ollé, 2010), assuming that those regions that receive more funds are in better conditions to prosper. These determinants have been clustered into structural (geographic structure of the district), economic (level of development, population density, urbanization, labor force participation rate), fiscal (size of the government budget deficit), political or electoral (partisan alignment, incumbent margin of victory in the last election), institutional (malapportionment, midterm elections, electoral timing), and ideological factors (government programmatic commitment to poverty alleviation). Institutional arguments have developed an important body of research supported by some empirical evidence. Bennett and Mayberry (1979) and Holcombe and Zardkoohi (1981) developed early studies claiming that more overrepresented states tend to receive more federal grants per capita. Atlas et al. (1995) and Lee (1998) found evidence that expenditures and net transfers per capita are significantly greater in smaller and overrepresented states. Rodden (2010, p.204) also found data to support the claim that over-represented states receive disproportionate 6

expenditures and intergovernmental grants per capita in Argentina, Brazil, the European Union, Germany, Australia, and the United States. The same finding has been supported by Gibson et al. (2004) for expenditures in Brazil and Argentina. The determinants of redistribution, according to Dixit and Londregan (1998, p.154; see also 1996, pp.1132-33), can be grouped into two main dimensions: ideological (egalitarian or programmatic) and tactical (electoral) dimension. We call them programmatic and discretionary, respectively. Programmatic Redistribution The central government redistributes programmatically when it follows certain ideological beliefs about equality or efficiency. These redistributive programs are usually general and they hand out transfers to all members of some broad socioeconomic group, such as the old, the sick, or the unemployed (Persson and Tabellini, 2000, p.115). Under this type of redistribution, infrastructure investment would favor certain kinds of districts, for instance lower income provinces or the most populated districts, following fixed or objective criteria (such as income per capita, poverty levels or total population). Under those criteria, none of the qualifying districts can be excluded from the government allocation. Some authors divide objective criteria for distribution into two main categories: efficiencyoriented and equity-oriented. In the first one, the most favored regions are those where projects have higher impact, that is, areas with more infrastructure users (measured in numbers of cars or urban density, for instance) or the more developed regions (in GDP terms); in the second group, the favored districts are those with low output levels or the less developed areas (Sollé Ollé, 2010, p.297). In this case, federal transfers are distributed to compensate the effects that an uneven distribution of wealth across a territory of a given country would generate on the 7

geographical distribution of public services (Buchanan, 1950; Musgrave, 1959; Oates, 1972; quoted in Porto and Sanguinetti, 2001, p239). Under these schemes, a government committed to maximizing a nationwide social welfare function allocates grants among states to correct for interjurisdictional externalities or to provide for those that are especially in need (Grossman, 1992, p.295). Discretionary Redistribution The central government redistributes discretionarily when it follows electoral and political considerations that would help the incumbent to retain and consolidate in power. Therefore, we expect this investment to be larger in districts were the government expects electoral benefits and returns in terms of support from the district s governor or her party s delegations in congress. Those districts that are not expected to generate electoral or political returns will be excluded from this type of investment. Discretionary distribution has also been called pork-barrel or machine politics (Dixit and Londregan, 1996, pp.1133-34). Lindbeck and Weibull (1987, p.289) following Hotelling s (1929) principle, argue that in cases of voters with identical consumption preferences but with observed differences in party preferences between groups, parties in a two party system will favor groups with weak party preferences, i.e. marginal voters. An implication of this claim is that, under the abovementioned conditions, parties will spend funds in swing districts (those with a high proportion of relatively unattached voters or in which the incumbent won or lost by a narrow margin) (Sollé Ollé, 2010, p.300). Similarly, for Persson and Tabellini (2000, ch.6), swing regions have larger electoral power than secure ones. Hence, redistribution for them will be larger if the low income regions are swing regions. 8

For Cox and McCubbins (1986, pp.370, 385), by contrast, the optimal strategy for risk-averse candidates is to redistribute to their reelection constituency (that is, those districts in which their core voters are) and over-invest in their closest supporters to maintain existing political coalitions. In other words, politicians will adopt strategies in which they invest little (if at all) in opposition groups, somewhat more in swing groups, and more still in their support groups (Cox and McCubbins, 1986, p.373). This paper is an empirical contribution to this debate as it provides further empirical evidence to support either claims of secure or pivotal redistribution. But it is also a theoretical contribution as it provides more specification in the theoretical argument. A limitation in the literature on political redistributive transfers is that it has not differentiated between the dynamics presidents follow to get territorial support depending on the differential political strength governors have across time. In other words, this paper claims that presidents will have different incentives depending on the territorial strength governors have and the contribution they can make to their party or coalition. Another shortcoming in the literature is that it has concentrated either on executive or legislative politics as the key political determinants for redistribution. Electoral calculations have been among the most relevant factors for the first group; while the distribution of state delegations in key committees has been one of the most relevant for the second one. To our knowledge, no paper has yet integrated these two approaches to explore their relative relevance in a given institutional framework. We do this in a different institutional context aiming to discuss the theoretical implications of the results for the comparative literature in the topic. The Main Argument 9

A key assumption in this work is that politicians, in general, adopt redistributive strategies to serve their electoral objectives. The incumbent will redistribute if that would help her staying in office (Lindbeck and Weibull, 1987; Cox and McCubbins, 1986); 5 the opposition will adopt a redistributive discourse if that will be conducive to getting hold of government positions. Citizens, in turn, vote for those political candidates that give them or promise them greater utility (Cox and McCubbins, 1986, p.373; Downs, 1957). The main argument in this paper is that presidents allocate funds to core districts to consolidate their electoral power and congressional support. Ceteris paribus, we expect that presidents will invest little or nothing in opposition provinces, somewhat more in swing districts, and more still in their support groups (as claimed by Cox and McCubbins, 1986, p.373). In other words, politicians will distribute grants to get access to two main resources: the political capital of state politicians and the support of state voters (Grossman, 1992, p.296). Despite these general claims, we can further specify the dynamics under which politicians will be more likely to transfer more redistributive funds. First, we claim that the political power of the governor is crucial in shaping the outcome. We expect presidents to invest more money in districts where the governors have strong political control (larger share of votes and seats) and their tenure potential is larger. The main reason is that these districts represent a safer and more significant electoral investment for the president than those were the governors face more competition and where they may even lose the next election. This is also the case for governors who have to abandon their positions due to term restrictions (this is captured by tenure potential). Second, we test whether the federal legislative may also have some influence over the outcome. Congressional power depends on the influence key committees, senior members of 5 Politicians, once in office, use several strategies and resources to remain in power. For Cox and McCubbins, (1986, p.383) patronage is the most visible and obviously redistributive strategy employed by politicians. 10

congress, and congressmen from the president s party can exert over the federal executive to alter the allocation of funds and favor their provinces. More specifically, we will evaluate whether those provinces with more representatives in core committees, those that have senior and influential congressmen, and those with the larger delegations from the president s party are more likely to receive more funds (Holcombe and Zardkoohi, 1981, p.397). Including these variables, we will study whether the institutional structure in congress and its political composition have an influence over the distribution of funds or whether federal and provincial executives have more influence over the final outcome. Data and Variables We use data on federal government infrastructure spending between 2000 and 2009 from the National Accounts (National Budget Office, Ministry of Economy). Total infrastructure funds include transfers from the central government to the provinces from eighteen budget programs of the Ministry of Federal Planning, Public Works, and Services. These programs are classified into funds for housing, community services, public infrastructure, and economic affairs. Funds for housing and community services include those allocated for housing development and water supplies. Expenditure in public infrastructure consist of transactions in fixed assets for purposes other than housing, i.e., buildings such as schools and hospitals, roads and highways, bridges, waterways, dams and pipelines, public monuments, and urban infrastructure. We have deliberately excluded from the analysis all the Ministry of Federal Planning subsidies to the public sector and private industries. All valued are reported in thousands (current) Argentine pesos. 11

In addition to total infrastructure grants, we also use two of the largest infrastructure funds in Argentina for the covered period: road construction and sanitation funds. We use them to check whether there are specificities in the distribution of funds according to the type of grant. Main and Competing Hypotheses Programmatic redistribution According to equity-oriented arguments, relatively deprived districts will be more likely to receive more infrastructure funds. Hence, we expect that the lower the GDP per capita, and the higher the poverty level in the district (measured in terms of the number of households with unsatisfied basic needs), the more infrastructure funds the district will get. For efficiencyoriented claims, funds will flow to those districts in which infrastructure projects relative impact is higher. Therefore, provinces with larger urbanization rate, population density, numbers of cars, and more developed (larger GGP), are more likely to receive more funds. In the same way, the economic structure of the district may also play a role: we can expect more infrastructure investment in those areas in which industrial production is larger (as this activity requires more infrastructure for its expansion). The equation is, infrastruc tureinvestment i 1( GDPpc i ) 2 ( povertyi ) 3( urbani ) 4 ( populi ) Where GDPpc i is district s i per capita GDP, poverty i is the poverty rate, urban i is the percentage of the urban population, and popul i is the population share, all of them in district i. Discretionary redistribution According to discretionary redistribution arguments, more infrastructure investment will be allocated to those provinces that can help presidents stay in office and get more power in Congress. These are the districts in which governors are politically allied with the president. Districts are classified into those belonging to the opposition (opposition; which are expected to 12

receive few funds, if any), swing districts (swing; 6 they are expected to receive somewhat more money), and support districts, or those aligned in partisan terms (which are expected to receive the largest share of funds). Partisan alignment is divided according to two criteria: core coalition (core) if governors and presidents are in the same party, and allied (allied) if they are in the same electoral coalition but from different parties. The equation is: infrastruc ture investment i 1( core) 2 ( allied) 3( swing) 4 ( opposition) The political power of governors (govpower) is also expected to influence the outcome. A given district will receive more funds when the partisan power the governor (the larger the value in the index of partisan power of the governor 7 ), the share of votes the governor got in the last election (govvotes), and the share of seats she controls in the state legislature (percdepseats; Chamber of Deputies in bicameral provinces) are larger. The equation is: infrastruc ture investment i 1( core) 2 ( allied) 3( swing) 4 ( opposition) infrastruc ture investment i 1( govpower i ) 2 ( govvotesi ) 3( govseatsi ) Where govpower i is the partisan power of the governor; marg ini is the governor margin of victory in the last election; govvotes i is the share of votes the governor got in the last election; and govseats i is the share of seats the governor controls after the last election, all of them in district i. 6 The variable swing is calculated as the difference between the share of votes received by the incumbent and the share of votes of the main opposition party; the smaller the difference between the two, the more likely for the district to swing 7 The index of partisan power of the governor is composed of two main dimensions: a) the power of governors in their districts (which includes the electoral support share of votes for the governor; whether the main party in the legislature is the party of the governor, coded as 1 in case they are the same, 0 otherwise; and the governor s party share of seats in the state legislature); and b) governors influence over the federal government or how politically linked governors are to it (here, I include a dummy variable for cases in which presidents and governors are in the same governing coalition; coded as 1 in case they are politically allied, 0 otherwise). The index is a composed measure of all the aforementioned shares and dummies (which contribute.5 points to the index in case they are coded as 1, to balance the effect of each measure). The maximum possible value is 3; the minimum is 0 (González, 2010). We calculated the average value for each year and for all governors in Argentina during the period under analysis. 13

We also expect legislative politics to influence the outcome. Provinces with more representatives in core committees (committee) 8 and those with the larger delegations from the president s party (delegation) 9 would be more likely to receive more funds. The equation is: infrastruc ture investment i 1( committee i ) 2 ( delegationi ) Institutional arguments For institutional claims, districts will be more likely to receive more funds the larger their overrepresentation (overrepres). The reason is that the political benefits from a marginal dollar of increased grants to a small and overrepresented state are greater than a marginal dollar of increased grants to a large state in which the per capita impact is smaller (Grossman, 1992, p.298). Samuels and Snyder (2001; see also Calvo and Murillo, 2005, p.216) calculate legislative overrepresentation using the Loosemore Hanby index of electoral disproportionality as follows: MAL = (1/2) Σ si vi, where si is the percentage of all seats allocated to district i, and vi is the percentage of the overall population residing in district I (values range between.64 and 19.12, mean value 1.97 and standard deviation 2.20). In addition, provinces will be more likely to receive more investment during election times (electionyr; both executive presidential and gubernatorial, and legislative elections). The equation is: infrastruc ture investment i 1( overrepres i ) 2 ( electionyri ) Where overrepres i is the overrepresentation index in district i; electionyr i is whether there were (executive or legislative) elections in the district during a given year. 8 The variable committee reports the number of deputies a given province has in the Public Works Committee in the Chamber of Deputies and in the Infrastructure, Housing, and Transportation Committee in the Senate. 9 The variable delegation is the percentage of congressmen in the Chamber of Deputies who are members of the majority party. 14

Method and Data We test the effects of the two main models (Programmatic Distribution and Discretionary Distribution Models), first by using ordinary least square (OLS) regressions. Second, because the data are cross sectional and time serial, I also run a regression taking into consideration random and fixed effects by generalized least squares (GLS) to correct for heteroskedasticity. With time series it is also sensible to execute a first-order autocorrelation correction. This test takes into account the fact that the data are correlated with themselves, and thus the error terms are correlated. In order to control for autocorrelation, the paper runs a Prais-Winsten regression iterated estimates, to correct for first-order autoregressive errors. Finally, in order to avoid overconfidence in the standard errors using GLS, this work performs an OLS regression with panel corrected standard errors (Beck and Katz 1995). The number of observations oscillates between 329 and 74 (according to the models and due to missing data), for the 24 districts in Argentina (23 provinces and the federal capital), for the years 2000-2009. Data on total federal infrastructure investment from National Bureau of Statistics and Census (INDEC) raises the number of cases up to 329 (24 districts between 1993 and 2006). Table 1: Descriptive Statistics, Comparison of Core and Opposition Districts, 2000-2009 In current pesos Core Districts Mean Std. Dev. Min Max Total Transfers in fixed assets (minplanb) 264,033,000 390,723,000 4,074,000 2,057,220,000 Total Transfers per capita (minplanb) 384 736 2 4,471 Transfers Public Works (Source: INDEC) 38,394,000 89,960,000 489,000 742,619,000 Transfers Public Works per capita (Source: INDEC) 63 172 1 1,430 Funds for road construction 81,405,000 144,927,000 0 810,555,000 Funds for road construction per capita 190 595 0 3,588 Funds for sanitation 21,900,000 35,586,000 0 149,280,000 15

Funds for sanitation per capita 30 68 0 419 Funds for housing 32,514,000 112,690,000 0 1,263,702,000 Funds for housing per capita 102 113 1 673 Opposition Districts Mean Std. Dev. Min Max Total Transfers in fixed assets (minplanb) 74,915,000 88,017,000 4,906,000 469,899,000 Total Transfers per capita (minplanb) 120 121 2 551 Transfers Public Works (Source: INDEC) 19,226,000 22,424,000 228,000 166,867,000 Transfers Public Works per capita (Source: INDEC) 42 63 1 516 Funds for road construction 17,003,000 40,544,000 0 260,028,000 Funds for road construction per capita 23 43 0 252 Funds for sanitation 6,720,000 17,923,000 0 76,249,000 Funds for sanitation per capita 4 10 0 46 Funds for housing 24,669,000 77,659,000 0 1,140,629,000 Funds for housing per capita 67 72 1 360 Empirical Findings Empirical results support the main theoretical expectations. First of all, Table 1 reports some basic descriptive statistics for the period under study, dividing districts into those belonging to the core presidential coalition and those in the opposition. Opposition districts received, on average, 28 percent of the total infrastructure funds that those districts in the core coalition received (or 31 percent in per capita terms). The difference in the distribution of road construction funds is even larger: opposition provinces received only 20 percent (or 12 percent in per capita terms) of the amount core districts got. Second, and getting into the regression results, redistribution does not seem to follow equity or efficiency criteria. The main predictors according to programmatic redistribution do not seem to get enough empirical support. More populated districts (in absolute terms or in terms of population density) and those where industrial production is larger (measured as industrial GGP) 16

do not seem to get more infrastructure investment. Only number of cars in 2007 seems to reach the standards for statistical significance but the coefficient goes in the opposite direction than theoretically expected (See Model 5). Similarly, provinces with the largest share of poor households or those with lower average income per capita do not appear to receive more investment. On the contrary, most regressions indicate that districts with larger average income per capita seem to have received more funds (See Model 5). These results are in line with those of Grossman (1992, p.301) who found that higher income states in the US tended to receive more money rather than the opposite. The author explains this by claiming that these states have greater influence on grant distribution. Out of the results presented here, neither efficiency nor equity criteria seem to be relevant to explain the allocation of infrastructure funds. Third, the data seems to support the theoretical arguments of discretionary models presented before. More specifically, presidents tend to allocate more infrastructure funds (infrastructure investment in general, as well as specific ones such as road construction and sanitation funds) to districts controlled by partisan allies. These provinces received substantially more funds than swing or opposition districts. Allied provinces receive, on average, $419 per capita more than those not allied in total infrastructure funds, $245 per capita more in road construction, and $34 per capita more in sanitation funds. Opposition provinces, on the contrary, received on average $153 per capita less than the rest of the districts in total infrastructure funds, $95 less in road construction, and $15 per capita less in sanitation funds (See Model 1). Provinces are more likely to get more funds if they are electorally secure and not swing districts. More specifically, they tend to get more funds when the difference between the share of votes the governor got and the share of votes of the main party in the opposition is larger (that is, 17

when the value of the variable swing increases). Coefficients are robust and significant for road construction funds and not significant for total infrastructure funds. The partisan power of the governor is also a relevant factor explaining the distribution of infrastructure funds. In line with our theoretical expectations, more funds are allocated to those districts in which governors have more partisan powers. A one point increase in the partisan power of the governor is associated to a $326 increase in the total infrastructure funds per capita the province receives; or an increase in $250 and $38 per capita in road construction and sanitation funds, respectively (See Model 2). Presidents also allocate more funds to provinces in which governors have larger tenure potentials; that is, districts in which the governor is expected to stay in office for longer periods of time. The governors electoral power (share of votes) is especially relevant to explain funds allocation, more so than the share of seats governors control in the provincial legislature (although both variables show robust and statistically significant coefficients) (See Model 2). Fourth, the main institutional arguments do not receive empirical support in the regressions performed. Overrepresentation does not seem to be a significant predictor. The coefficients are neither robust nor statistically significant. 10 This finding is contrary to what several authors reported in their studies, including Rodden (2010) for the European Union. Furthermore, election years for federal and provincial executives and legislative do not seem to contribute explaining the allocation of infrastructure investment in the federation. None of these regressions reach the standards of statistical significance (See Model 3). Fifth, results seem to indicate that infrastructure distribution in Argentina is mainly decided by the executive and not the legislative. The number of legislators from a province in the relevant infrastructure congressional committee does not have an influence over the amount of funds the 10 The coefficient is significant but very weak only for sanitation data. 18

province receives. However, provinces tend to receive more funds when the state legislative delegations of the governing party are larger (See Model 4). These results hold for total infrastructure and sanitation funds per capita; it is not the case for road construction data. These findings are in line with those Grossman (1992, p.299) got: for him, the larger the legislative majority of the Democratic party (the party in government at the federal level) was empirically associated to larger grants. I control for problems of multicolinearity and heteroskedasticity and substantive results hold. Discussion The distribution of infrastructure funds in Argentina does not seem to be determined by equity or efficiency criteria. Empirical results do not seem to offer support for programmatic arguments. This may have profound implications for development strategies and in terms of redistribution and inter-regional inequality. Infrastructure funds are one of the most important tools in hands of the federal government to redistribute wealth across regions in the country, but they are not used to diminish inequality in Argentina. The main reason is that presidents seem to allocate funds according to discretionary considerations: they distribute more funds to allied districts compared to swing or opposition districts. Governors are more likely to receive more funds when they are powerful in partisan and electoral terms and when their tenure potential is longer. The legislative do not seem to shape the outcome as previous research reported in the United States; and the distribution of infrastructure funds does not appear to be affected by the overrepresentation of the districts, as scholars have found in the European Union. Electoral politics and the politics of legislative coalition formation and maintenance are important factors influencing the distribution of federal infrastructure. This is an important finding. The key question is how these politics may have more redistributive impact. When are 19

districts that need more investment more likely to receive it? This is likely to occur only when the key coalition partners presidents have predominantly ruled in the least developed districts. This is the moment in which federal politics and redistribution coincide. 20

References Atlas, Cary, Thomas Gilligan, Robert Hendershott, and Mark Zupan, Slicing the Federal Net Spending Pie: Who Wins, Who Loses, and Why?, American Economic Review 85, 1995, pp.624 9. Bennett, James and Eddie Mayberry, Federal Tax Burdens and Grants Benefits to States: the Impacts of Imperfect Representation, Public Choice 34, 1979, pp.255-269. Beramendi, Pablo, Inequality and the Territorial Fragmentation of Solidarity, International Organization 61(4), 2007, pp.783-820. Castells, Antoni and Albert Solé Ollé, The Regional Allocation of Infrastructure Investment: The Role of Equity, Efficiency, and Political Factors, European Economic Review, Volume 49, Issue 5, 2005, pp. 1165-1205. Cox, Gary W. and Mathew McCubbins, Electoral Politics as a Redistributive Game, Journal of Politics, 48 (2), 1986, pp.379 89. Dasgupta, Sugato, Amrita Dhillon, and Bhaskar Dutta, Electoral Goals and Centre State Transfers in India, unpublished manuscript, Indian Statistical Institute, 2001. Dixit, Avinash and John Londregan, Fiscal Federalism and Redistributive Politics, Journal of Public Economics 68 (2), 1998, pp.153 80. Dixit, Avinash and John Londregan, The Determinants of Success of Special Interests in Redistributive Politics, Journal of Politics 58 (4), 1996, pp.1132 55. Gibson, Edward and Ernesto Calvo, Federalism and Low-Maintenance Constituencies: Territorial Dimensions of Economic Reform in Argentina, Studies in Comparative International Development 35(3), 2000, pp.32 55. Gibson, Edward, Ernesto Calvo, and Tulia Falleti, Reallocative Federalism: Overrepresentation and Public Spending in the Western Hemisphere, in Edward Gibson (ed.) Federalism and Democracy in Latin America (Baltimore, MD: Johns Hopkins University Press, 2004). González, Lucas, Primus contra Pares: Presidents, Governors, and the Struggles over the Distribution of Power in Federal Democracies, Unpublished PhD Dissertation, Department of Political Science, University of Notre Dame, IN, United States, 2010. Holcombe, Randall and Ashgar Zardkoohi, The Determinants of Federal Grants, Southern Economic Journal 48, 1981, pp.393-399. Lee, Frances, Representation and Public Policy: The Consequences of Senate Apportionment for the Geographic Distribution of Federal Funds, Journal of Politics 60(1), 1998, pp.34 62. 21

Lee, Frances, Senate Representation and Coalition Building in Distributive Politics, American Political Science Review 94(1), 2000, pp.59 72. Lindbeck, Assar and Jörgen Weibull, Balanced-Budget Redistribution as the Outcome of Political Competition, Public Choice 52 (3), 1987, pp.273 297. Persson, Torsten and Guido Tabellini, Federal Fiscal Constitutions, Risk Sharing, and Redistribution, Journal of Political Economy 104 (5), 1996, pp.979 1009. Persson, Torsten and Guido Tabellini, Political Economics. Explaining Economic Policy (Cambridge, MA: MIT Press, 2000). Porto, Alberto and Pablo Sanguinetti, Political Determinants of Intergovernmental Grants: Evidence from Argentina, Economics and Politics 13(3), 2001, pp.237-256. Randolph, Susan and Dennis Hefley, Determinants of Public Expenditure on Infrastructure: Transportation and Communication (Washington, DC: The World Bank, 1996). Samuels, David and Richard Snyder, The Value of a Vote: Malapportionment in Comparative Perspective, British Journal of Political Science 31(4), 2001, pp.651 71. Solé Ollé, Albert, The Determinants of Regional Allocation of Infrastructure Investment in Spain in Bosch, Núria, Marta Espasa, and Albert Solé Ollé, The Political Economy of Inter- Regional Fiscal Flows: Measurement Determinants and Effects on Country Stability (Studies in Fiscal Federalism and State-local Finance) (Cheltenham: Elgar Publishing, 2010). WIID (UNU-WIDER World Income Inequality Database), Version 2.0c, May 2008. 22

Tables Model 1. Table 1 Total Infrastructure Federal Infrastructure INDEC (per capita).061** (.024).001*** -.119 (.085).001 (.001) 6.34*** (1.05) -049 (.030) Road Construction Sanitation Core Coalition.419*** (.106).245** (.087).034** (.012) Tenure.003***.002***.000*** Potential (.001) Share -.569 -.191 -.037 Population (.345) (.283) (.037) Percentage.004.001.000 Poor (.004) (.003) Per capita.000***.000**.000* GDP (3.24) Constant -.165 -.134 -.008 (.118) (.098) (.014) Number of 190 329 185 97 cases R-Squared 0.26 0.20 0.18 0.24 23

Model 1. Table 2 Total Infrastructure Federal Infrastructure INDEC (per capita) -.031** (.012).001*** -.132 (.085).001 (.001).000*** -.028 (.030) Road Construction Sanitation Opposition -.153** (.050) -.095** (.041) -.015** (.006) Tenure.003***.003***.000*** Potential (.001) (.001) Share -.473 -.137 -.028 Population (.348) (.283) (.037) Percentage.003.001.000 Poor (.004) (.003) Per capita.000***.000**.000** GDP Constant -.042 -.061 -.000 (.120) (.099) (.015) Number of 190 329 185 97 cases R-Squared 0.24 0.20 0.16 0.22 24

Model 1. Table 3 Total Infrastructure Federal Infrastructure INDEC (per capita).074* (.043).001*** -.128 (.087).000 (.001).000*** -.037 (.030) Road Construction Sanitation Swing Districts.273 (.180).312** (.152).023 (.022) Tenure.003***.002**.000** Potential (.001) (.001) Share -.438 -.155 -.023 Population (.356) (.286) (.039) Percentage.002 -.001 -.000 Poor (.004) (.003) Per capita.000***.000**.000 GDP Constant -.096 -.088 -.002 (.121) (.098) (.015) Number of 189 325 184 97 cases R-Squared 0.21 0.20 0.16 0.18 25

Model 2. Table 1. Total Infrastructure Gubernatorial.326*** Power (.065) Share -.629* Population (.354) Percentage.003 Poor (.004) Per capita GDP.000*** Constant -.650*** (.188) Number of cases Road Construction.250*** (.054) -.270 (.286).001 (.004).000*** -.514*** (.153) Sanitation.038*** (.008) -.046 (.036).000.000** -.081*** (.024) 176 171 94 R-Squared 0.21 0.16 0.24 26

Model 2. Table 2 Total Infrastructure Road Construction Sanitation Share of Gubernatorial Vote.838*** (.253).782*** (.208).083** (.028) Share Population -.593* (.358) -.214 (.283) -.029 (.038) Percentage Poor -.001 (.004) -.002 (.003) -.000 Per capita GDP.000***.000**.000 Constant -.356** (.176) -.360** (.141) -.031 (.021) Number of 189 184 97 cases R-Squared 0.15 0.12 0.12 27

Model 2. Table 3 Total Infrastructure Share of.268* Deputies (.138) Share -.623* Population (.366) Percentage -.001 Poor (.004) Per capita.000 GDP Constant -.084*** (.145) Number of cases Road Construction.201* (.114) -.264 (.292) -.002 (.003).000** -.081 (.116) Sanitation.032* (.017) -.033 (.040) -.000 (.001).000 -.006 (.018) 192 187 97 R-Squared 0.12 0.07 0.07 28

Model 3. Table 1 Total Infrastructure Overrepresentation.033 (.021) Share Population -.756 (.556) Percentage Poor.001 (.004) Per capita GDP.000*** Constant -.050 (.148) Road Construction.008 (.016) -.298 (.428) -.001 (.004).000** -0.29 (.116) Sanitation.017** (.005).010 (.088).000 (.001).000 -.018 (.021) Number of cases 168 163 74 R-Squared 0.21 0.09 0.22 29

Model 3. Table 2 Total Infrastructure Election Year.058 (.065) Share -.805** Population (.381) Percentage.002 Poor (.004) Per capita.000*** GDP Constant -.065 (.128) Number of cases Road Construction.049 (.050) -.370 (.288) -.001 (.003).000**.029 (.099) Sanitation.006 (.006) -.048 (.039) -.000 (.001).000.010 (.016) 192 187 97 R-Squared 0.24 0.06 0.05 30

Model 4. Table 1 Total Infrastructure Committee -.017 (.036) Share -.570 Population (.711) Percentage.003 Poor (.005) Per capita.000*** GDP Constant -.065 (.142) Number of cases Road Construction -.016 (.028) -.121 (.548) -.001 (.004) -.000**.027 (.109) Sanitation -.004 (.005) -.001 (.086) -.001 (.001).000.017 (.019) 168 164 81 R-Squared 0.25 0.06 0.05 31

Model 4. Table 2 Total Infrastructure District s.452** share of seats (.208) Share -.800** Population (.356) Percentage -.002 Poor (.004) Per capita.000** GDP Constant -.012 (.126) Number of cases Road Construction.183 (.170) -.376 (.288) -.002 (.003).000* -.005 (.104) Sanitation.059** (.021) -.044 (.038) -.000 (.001) -.000.004 (.015) 192 187 97 R-Squared 0.13 0.06 0.11 32

Model 5. Table 1 Total Infrastructure Federal Infrastructure Road Construction Sanitation INDEC (per capita) Core Coalition.490** (.026).062*** (.124).271** (.098).041** (.014) Swing District.282 (.045).070 (.209).290* (.175) -.002 (.028) Tenure Potential.003**.001** (.001).002 (.015).007 (.006) Overrepresentation.023 (.005).001 (.020).002** (.001).000* Share Population -.348 (.098) -.111 (.519).023 (.408) -.016 (.083) Percentage Poor.003 (.001).001 (.005).000 (.004).000 (.001) Per capita GDP.000***.000***.000**.000 Constant -.304** -.061** (.142) -.195* (.114) -.019 (.020) Number of 165 307 160 74 cases R-Squared 0.36 0.22 0.24 0.35 33