争取选票：对1991年－1995年间秘鲁社会基金（FONCODES）支出的政治经济学研究

6) s C.41 1.0 POLICY RESEARCH WORKING PAPER 2166 Seeking Votes As the literature on political influences on the allocation of discretionary funds predicts, The Political Economy of spending by the Peruvian Expenditures by the Peruvian Social Fund, FONCODES, Social Fund (FONCODES), 1991-95 increased significantly before elections. FONCODES projects were also dirmected Norbert R. Schady at provinces where the marginal political impact of expenditures was likely to be greatest. The World Bank Poverty Reduction and Economic Management Network Poverty Division H August 1999  oi:1( Y RLSE&ARCH- WORKING P aFR 21 66 Summary fi'ndilngs r:siPcemt Albcrro Fujimlori created the Peruvian Social * FONCODES projects were directed at poor FLlind (FONCODES) in I 99 1 with the stated objectives provinces, as well as provinces in which the mar-ginal of gencratil- cnioptoviient helping to alleviate poverty, political impact of expenditures was likely to be greatest. and i.mrprovmgo access to social services. The results are robust to many specifications and ScLhadY USes province-level data on monthly controls. The Peruvian data thus support predictions cxpenleturcs, socioeconomic indicators, and electoral made in the literature on political business cycles as well outcomes to analvze political influences on the timing as the literature on political influences on the allocation an g' ographic distribut;on of FONCODES of discretionary funds. expenditures between 1991 and 1995. He finds that: CONCODDES expenditures mcreased significantly .-efoTe electiotns. This -pper - a product of the Poverty Division, Poverty Reduction and Economic Management Network - is part of a laracr I rfo;r im the netwvork to understand the functioning and impact of social funds. Copies of the paper are available free fion' the World Bank, 1818 H Street N-W, Washington, DC 20433. Please contact Norbert Schady, room MC4-629, t eph oru e 202-45 8-8247, fax 202-222-3283, Internet address nschady(@worldbank.org. Policy Research Working Papers ale a,)0 potsed on the Web at http:/","ww.worldbank.org/html/dec/Publications/Workpapers/home.html. August 1999. (30 p i"'s? ihc Se ollc Resea,ch kVorkXkng Paper Series disserninates the findings of work in progress to encourage the exchange of ideas aboot oIeuesopmZent issiues. Ani objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. 'Te s ape1s coyrr tibe n,amr,es ot the authors and should be cited accordingly. The findings, interpretations, and conclosions expressed in this p.per are eni!reiv those of the authors. They do vot necessarily represent the viewJ of the World Bank, its Executive Directors, or the corrI0tr!es t"ly rresent. Produced by the Policy Research Dissemination Center  Seeking Votes: The Political Economy of Expenditures by the Peruvian Social Fund (FONCODES), 1991-95 Norbert R. Schady World Bank* Norbert Schady is an economist in the Poverty Group, Poverty Reduction and Economic Management Vice- Presidency at the World Bank, 1818 H Street NW, Washington, DC 20433, tel. (202) 458-8247, e-mail nschady@worldbank.org. An earlier version of this paper was presented at the Northeast Universities Development Conference held at Yale University on October 16-17, 1998. 1 would like to thank Christina Paxson for much useful advice throughout this research, and Helge Berger, Jean Boivin, Anne Case, Angus Deaton, Avinash Dixit, Susan Stokes, Chang Tai-Hsieh, Mark Thomas, John Waterbury, participants at seminars at Princeton University and the World Bank, and two anonymous referees for helpfuil comments.   1. Introduction Do incumbents use social programs to get themselves re-elected? Much theoretical and empirical work on public expenditure patterns in democratic regimes argues that politicians manipulate the timing, composition, and geographic distribution of expenditures to maximize the likelihood of remaining in office. Increases in public expenditures are timed to coincide with upcoming elections; budgetary trade- offs are made to satisfy important constituencies; and expenditures are directed towards areas of the country that are considered critical for an incumbent's re-election bid. This paper discusses the relationship between elections and the expenditures made by the Peruvian Social Fund (FONCODES) during President Alberto Fujimori's first tern in office (1990-1995). FONCODES was created in 1991, during a severe recession which followed the adoption of stringent stabilization and structural adjustment measures by the Fujimori government. Its stated objectives were to generate employment, help alleviate poverty, and improve access to social services (World Bank 1998). By 1993, the Peruvian economy had begun a remarkable recovery, and FONCODES was being hailed by govermnent officials and foreign donors alike as an important component of a medium-term strategy to reduce poverty. Scholars in Peru and abroad have generally been more guarded in their praise. Critics argue that social funds like FONCODES undermine line ministries and local governments. And questions about the "politicization" of FONCODES have bedeviled the program from its creation (Graham 1994, Kay 1995, Ball6n and Beaumont 1996). Yet there are no satisfactory studies of the impact of elections on FONCODES expenditures. This paper attempts to fill this gap using highly disaggregated data on expenditures, socio- economic indicators, and electoral outcomes. I use econometric techniques to analyze the timing and geographic distribution of FONCODES expenditures, and reach two main conclusions. First, FONCODES expenditures increased significantly prior to national elections. Second, FONCODES projects were directed to provinces in which the marginal political impact of expenditures was likely to be largest. The results are robust to a large number of specifications and controls. The Peruvian data thus support predictions made in the literature on "political business cycles" (Nordhaus 1975, Rogoff 1990) as well as the literature on political influences on the allocation of discretionary funds (Wright 1974, Cox and McCubbins 1986, Lindbeck and Weibull 1987, Dixit and Londregan 1996, Case 1997). The rest of the paper proceeds as follows. Section 2 provides a brief description of Peruvian politics between 1990 and 1995, and FONCODES. Section 3 describes data requirements and sources for the paper. Section 4 estimates the effect of the electoral calendar on the timing of FONCODES I Moncada (1996) and Graham and Kane (1998) discuss the timing and distribution of FONCODES expenditures, respectively. As I discuss below, however, both papers have serious limitations. 2  expenditures. Section 5 estimates the effect of the 1990 presidential election on the distribution of expenditures. Section 6 concludes with some general thoughts on the use of social expenditures for electoral purposes. 2. The Fujimori years and FONCODES The election of the dark horse candidate Alberto Fuji:mori to the Peruvian presidency in 1990 marked the beginning of a fundamental transformation of Peruvian politics. Fujimori, an independent with no political party to speak of, defeated the coalition of center-right parties headed by the well-known novelist Mario Vargas Llosa. Fujimori's party did not, however, win a majority in either house of the Peruvian congress. Blaming Peru's "corrupt" institutions for his inability to introduce a coherent economic program or combat the terrorist insurgencies of Shining Path and the Tupac Amaru, Fujimori used the military to shut down congress and the judiciary in April 1992. Under international pressure to restore democracy, he drafted a new constitution which was approved by a small margin in a referendum held in October 1993, and won a landslide re-election to the presidency and a majority for his party in the new uni-cameral congress in April 1995. The academic literature on Peru argues that traditional sources of authority and intermediation between citizens and the state have been weakened during F'resident Fujimori's period in office (McClintock 1993, Cameron 1994, Mauceri 1995, Conaghan 1996, Seddon 1997). The "historical" political parties have all but disappeared.2 The legislative and judiciary branches of government have not acted as an effective check on the executive.3 There have also been important changes in the relationship between various levels of government, and in the conduct of social policy. Decree 776, passed in December 1993, abolished most sources of tax revenue traditionally used by municipal governments. Meanwhile, responsibility for the administration of numerous social and basic infrastructure programs has been transferred from local governments and line ministries in education, health, and transportation to the Ministry of the Presidency. By 1995, this Ministry controlled more than 20% of the central government 2 The four major parties which dominated civilian politics in Peru for decades have become irrelevant. These four parties--the conservative Popular Christian Party, the center-right Popular Action, the center-left APRA, and the left- wing United Left-jointly garnered 82.2% and 97% of the vote in the presidential elections of 1980 and 1985, respectively. In the 1995 presidential election they won a mere 6.3% of the vote (Tuesta, 1994; Jurado Nacional de Elecciones, 1995). Cambio 90-Nueva Mayoria, president Fujimori's party, has no outstanding leaders other than the president himself. Independent candidates with no affiliation to any party have dominated local politics. In the 1993 municipal elections, for example, independent candidates receivel almost two-thirds (64.7%) of the votes cast (Tuesta, 1995, p. 106). 3 The legislative, where Fujimori's party has had an absolute majority since 1992, rubber-stamps the President's initiatives; on the only important occasion on which the Supreme Court contravened Fujimori's wishes, ruling that a draft law which would allow Fujimori to run for a second re-election in the year 2000 was unconstitutional, the three members who had voted with the majority opinion were dismissedl from the bench. 3  budget, and oversaw thirteen programs involved in, amongst other things, nutrition, education, health, water, sanitation, and housing (World Bank 1996, p.2). FONCODES is one of the new programs created by President Fujimori. Its charter establishes it as an "autonomous body" outside the traditional line ministries. By law, the FONCODES executive director reports directly to a five-member Board of Directors appointed by the President. This institutional set-up raised concerns that FONCODES could easily become a means for dispensing pork, and these misgivings were reinforced when Luz Salgado, a former head of President Fujimori's party, was appointed as FONCODES' first executive director (Graham 1994, p. 108). Subsequent directors have been recruited from the private sector (Arturo Woodman, Alejandro Afuso) and from other branches of government (Manuel Estela, Manuel Vara Ochoa), and this has helped dispel some of the initial concerns. Over time, FONCODES has also improved coordination in the planning, execution, operation and maintenance of projects with the line ministries and, to a lesser extent, with local governments and NGOs. Between December 1991 and May 1995, FONCODES funded almost 16,000 community-based projects, for a total budget of about 465 million soles.4 At the 1992 exchange rate, this was equivalent to roughly US $ 285 million. The average project size was therefore quite small--about US $ 18,000. These community-based projects include nutrition and family planning projects, rotating credit schemes, and projects for the construction or rehabilitation of schools, health posts, water and sanitation systems, rural roads, electrification schemes, and small-scale irrigation works. In addition, FONCODES has executed a series of centrally-designed "special" projects which have included, inter alia, a school breakfast program, the distribution of uniforms for schoolchildren, shovels for farmers, and motorized canoes for isolated jungle communities. Between December 1991 and May 1995, FONCODES spent about 135 million soles (about US $ 83 million) on these special projects. In 1993, FONCODES expenditures accounted for roughly 0.55% of GNP--more or less in the same order of magnitude as the Mexican Solidarity program, which accounted for about 0.71% of GNP in 1992. FONCODES has much in common with other social funds in the region (Glaessner et. al. 1994). The bulk of FONCODES projects are demand-driven and targeted. FONCODES' community-based projects (but not the special projects) are demand-driven in that communities themselves choose a project from a menu and prepare a proposal for funding. FONCODES then functions as a financial intermediary: rather than execute projects itself, it approves proposals and releases funds to the nucleo ejecutor--a group of community members elected specifically for this purpose. FONCODES also targets its investments- 4 All expenditures are in constant 1992 soles, unless otherwise indicated. 4  first, by using a "poverty map" to assign resources to small geographic areas, and then conducting an informal on-site assessment of the "poverty" of the community requesting a project.5 3. The data set In 1997, Peru was divided into 13 administrative regions, 24 departments, 194 provinces, and 1812 districts (Webb and Femrndez Baca 1997, p. 112). Regions and departments are administrative arms of the central government, which appoints a governor for every region. Provincial and district-level mayors are chosen at municipal elections every three years. Data on FONCODES expenditures, socio-economic characteristics, and voting patterns is available at different levels of geographic aggregation. FONCODES itself keeps monthly records on the number of projects and aggregate amounts spent in each district. Three points are worth noting about these data. First, because expenditures are recorded in the month in which a project is approved, there is a lag of about two weeks before disbursement of the first installment of funds, and of several months before the second (and final) disbursement. Second, only expenditures on projects are included, and not administrative costs or expenses for general overhead. Third, district-level information on the distribution of expenditures is only available for the cornmunity- based projects. There is no information on the allocation formulas or the actual distribution of the special projects. The Peruvian National Statistical Institute (INEI) has estimated population size, the proportion rural, and a host of socio-economic indicators at the district level. Most of these indicators are based on a population and housing census conducted in 1993. The one important exception is the rate of chronic malnutrition, which is based on a census of height and weight amongst schoolchildren also conducted in 1993. The Jurado Nacional de Elecciones, the body which supervises elections in Peru, keeps records on electoral outcomes for elections held since 1990. Some of these data have also been compiled and published (Tuesta 1994, Jurado Nacional de Elecciones). Results are generally available for all provinces in the country, and for all districts in Metropolitan Lima. The level of aggregation used for the analysis in this paper was dictated by theoretical considerations and the availability of data. Section 4 uses national aggregates of FONCODES expenditures to analyze the relationship between the timing of expenditures and the electoral calendar. Sources of funding and the administration of the community-based and special projects are different, so the analysis is broken down by project type. FONCODES has assigned resources to individual provinces since 1993, and to individual districts since 1996. Section 5 uses a province-level data set to analyze the 5 Note, however, that even without political interference actual expenditures in a given province would rarely be equal to allocations because of the demand-driven nature of the program. 5  influence of electoral outcomes on the geographic distribution of FONCODES expenditures for community-based projects. To avoid having one data point account for 5.78 million inhabitants--more than a quarter of the population of the entire country--the province of Metropolitan Lima has been further divided into its districts. The term "province" is used throughout the paper to refer to districts in Metropolitan Lima and provinces elsewhere. 4. The timing of FONCODES expenditures This section discusses the possible effect of the electoral calendar on the timing of FONCODES expenditures. I briefly review the literature, discuss econometric specifications, and present results. There is a long-standing debate about the extent to which politicians can successfully manipulate the economy to coincide with elections. The traditional political business cycle literature predicts increases in expenditures before an election (Nordhaus, Tufte 1978). Although the assumptions underlying these models were brought into question with the advent of rational expectations theory, Rogoff shows that political business cycles could be expected to occur even if voters are fully rational but there are information asymmetries between voters and incumbents. The empirical evidence from Less Developed Countries (LDCs) tends to support the predictions made in the political business cycle literature. In a study covering seventeen Latin American countries over 35 years, Ames (1987) shows significant increases in total public sector expenditures in electoral years. More recently, Schuknecht (1996) finds further evidence for political business cycles in a study which covers 35 countries in the 1970 to 1992 period. Rodrik (1991, cited in Waterbury 1993, p. 201) and Buffie and Sangines Krause (1989, pp. 144-45) also observe pre-electoral expansions in Turkey and Mexico, respectively. One partial exception to this body of evidence is Remmer (1993), who uses recent data from six Latin American countries to argue that economic conditions generally deteriorated in the period leading up to an election and improved thereafter. This suggests that politicians may try, but are often unable to manipulate macro-economic variables successfully, an argument which has also been made for Western Europe (Lewis-Beck 1988). A different issue, which has not been resolved in the empirical literature, is whether voters in LDCs actually reward politicians for their pre-electoral spending sprees (Dornbusch and Edwards 1991, Stokes 1996a). Political manipulation of expenditures is likely to have different implications when it is concerned with a specific program like FONCODES rather than the aggregate macroeconomy. The stakes are obviously lower: it may be easier and less risky to try to accurately time the disbursement of funds from a specific program like FONCODES, but the electoral returns from such spending sprees are also likely to be smaller. Still, political manipulation of specific social programs like FONCODES may be one of the 6  few remaining ways a government committed to an agenda of low inflation, fiscal responsibility, and a stable exchange rate can attempt to influence the vote in the months leading up to an election. Moncada presents a simple time line of quarterly FONCODES expenditures and elections between 1991 and 1995. Interpretation of Moncada's graph is not easy, however, because elections do not always fall conveniently at the end of a quarter. Nor does Moncada formally test the relationship between elections and expenditures--perhaps because of the small number of quarters (16 quarters) and the large number of national elections (3 elections) during the period. Figure 1 graphs monthly data for the amounts approved for community-based and special projects between November 1991 and May 1995. The chart shows that expenditures are highly irregular, especially (but not only) for the special projects. There are at least four distinct spikes in total expenditures. Three of these spikes coincide with the three national elections held during the period-the November 22, 1992 election to the constituent assembly (month=13), the October 31, 1993 referendum (month=24), and the April 9, 1995 presidential election (month=42). Figure 1 thus provides some prima facie evidence that FONCODES expenditures were affected by elections. Figure 1: FONCODES expenditures, by month Apit 91995 40 Noenbr192Ocrtober 1993 2 0 AX -~community November~~~~Mnt 1992emer191 35- 30 I I~~~~~~~~~Mnh(-Nvme 91 a~~~~~~~~~~~~~~~~~~~~~~  Estimates of election effects If there is a political expenditure cycle, FONCODES expenditures at a particular point should be a function of the electoral calendar: (1) Expt =f(Elecb zd where Expt stands for FONCODES expenditures in time t, Elect is the presence (or absence) of an election at the end of time period t, and z, is a vector of exogenous variables which affect expenditures. The timing of elections is taken to be determined outside the model. Insofar as it is easier to manipulate the timing of FONCODES expenditures than the timing of elections, this is a reasonable assumption. Table 1 reports results of OLS regressions of total expenditures and expenditures for comnunity- based and special projects on electoral dummies, a linear time trend, and quarterly seasonal dummies. Columns (i), (iii) and (v) report results from the regressions of expenditures on a single electoral dummy variable which takes on the value one for the three months immediately prior to an election, and zero otherwise.6 Columns (ii), (iv) and (vi) include three separate electoral dummies to allow the effect of elections on expenditures to vary by election. Table 1: The impact of elections on the timing of FONCODES expenditures Community-based projects Special projects Total expenditures (i) (ii (iii) (iv) (v) (vi) ELECTION 7.05** - 5.07** - 12.72*** - (2.03) (2.39) (2.78) Election 1 - 6.08 - 0.82 - 3.73 (1.14) (0.36) (0.57) Election 2 - 4.51 - 14.74*** - 21.84** (0.69) (3.53) (2.58) Election 3 - 10.49 - 11.72*** - 17.93** (1.61) (2.73) (2.11) Monthly linear trend 0.04 0.03 0.04 -0.00 0.09 0.01 (0.30) (0.15) (0.85) (0.09) (0.60) (0.08) Seasonal dummies Yes Yes Yes Yes Yes Yes Durbin-Watson statistic (original) 1.20 1.22 2.45 2.43 1.49 1.42 Durbin-Watson statistic (transformed) 2.13 2.10 2.00 2.05 2.03 2.06 R-squared 0.167 0.173 0.278 0.370 0.274 0.328 Number of observations 42 42 42 42 42 42 * indicates significance at the 10% level; **indicates significance at the 5% level; *** indicates significance at the 1% level. Note: Absolute t-statistics are reported in brackets, and are corrected for serial correlation using the Cochrane-Orcutt procedure. A constant was calculated but is not reported. 6Because of the two-week time lag between approval of a project and disbursement of funds, the relevant month for the parameter Elect in equation (1) is taken to be that in which the project was approved if the election took place in the second half of the month, but the month preceding it if the election took place in the first half of the month. 8  The results in Table 1 confirm the general picture in Figure 1. The electoral parameter in columns (i), (iii) and (v) is positive and significart at the 5% level or better in all three specifications. This suggests that expenditures on both comrnunity-based projects and special projects were significantly higher than expected in the three months leading up to elections. When the three elections enter into the regression separately, the influence of elections on the timing of expenditures is most apparent for the special projects, especially for the October 1993 referendum and the April 1995 presidential election.7 The empirical analysis above suggests that Fujimori boosted FONCODES expenditures when it mattered most. But there are obviously other factors at work. Consider, for example, the expenditure spike in September 1994 (month=35), the second-largest in 1he data set. In August 1994, Manuel Vara Ochoa replaced Manuel Estela as the executive director of FONCODES. Estela was a technocrat who was highly regarded at the World Bank and the Inter-American Development Bank. Prior to being appointed executive director of FONCODES, he had been the head of SUNAT, the govemrnment agency charged with the collection of taxes in Peru. Estela brought to FONCODES his advisors from SUNAT, and a working method which emphasized a rigorous screening of project proposals. This was not, however, to Fujimori's liking, and the president quickly became dissatisfied with Estela's unwillingness to meet "disbursement targets". Estela's replacement, Manuel Vara Ochoa, was a friend of President Fujimori's from their days together as professors at the Agrarian University in Lima, and was understood to be more sensitive to the political importance of a speedy approval process. The massive spike in expenditures in September 1994 may be a result of the wholesale approval of projects which had not been funded during the Estela period. In conclusion, there is some evidence of a political expenditure cycle in FONCODES projects, but the electoral calendar is by no means the only determinant of changes in FONCODES expenditures between 1991 and 1995. 7 The changing political context in Peru may explain differences in the effect of elections on FONCODES expenditures. The Fujimori government had much at stake in the October 1993 referendum and the April 1995 presidential election. The referendum on the new constitution was widely seen as a referendum on three years of Fujimori government (Roberts and Arce 1998, p. 229). Moreover, the new constitution included a clause which would allow Fujimori to run for re-election--something which he could not to do under the 1980 constitution. The April 1995 election, meanwhile, would assure Fujimori five more years as president. By contrast, the November 1992 election to the constituent assembly took place at a time when the government was virtually unchallenged: the opposition was in disarray after Fujimori's "self-coup" of April 1992, and Fujimori's popularity was very high. Indeed, three of the four "historical" parties--Popular Action, APRA, and the United Left--did not even participate in the 1992 election, alleging that this would "legitinize" Fujimori's zoup. The Popular Christian Party participated, but won only 9.8% of the vote. Fujimori's Cambio 90-Nueva Mayoria party won 49.2% of the vote, and 44 of the 80 seats in the constituent assembly. The remaining 41% of the vote went to smaller parties, many of which had been formed shortly before the election. 8 Author's conversations with staff at FONCODES and at the Ministry of the Presidency. One knowledgeable staff member described the relationship between Estela and Fujimori as "strained" and related a telephone conversation in which Fujimori gave Estela a serious warning ("un toque de atencion de palacio"). 9  5. The geographic distribution of FONCODES expenditures This section discusses the influence of elections on the geographic distribution of FONCODES expenditures on community-based projects. Specifically, it tests a model in which past electoral outcomes are used to predictfuture expenditures. First, I briefly review the literature on the topic, and describe some features peculiar to the Peruvian setting. I next discuss the econometric specification, and present results. A basic theory of the determinants of voting behavior might consider the affinity of voters for different candidates (an "ideological" component) and the changes in welfare that voters can expect to see after an election (an "economic" component) (Lindbeek and Weibull, Markus 1988, Peltzman 1990, Dixit and Londregan, Landon and Ryan 1997). If voters face a choice between an incumbent and his challenger(s), they might use the performance of the incumbent while in office as a reasonable guide to the policies they could expect to see enacted if the incumbent were re-elected.9 The probability that voter i will vote for the incumbent at time t, Pit, might then be described in the following terms: (2) Pit= g(v1, YirYit-l, Wirwit-d where v, is a time-invariant, non-economic parameter which affects the vote. This parameter could include voter i's perception of the incumbent's character, or the ideological affinity between voter i and the incumbent. I assume that the parameter v, ranges from -oo to x, representing strong dislike and like for the incumbent, respectively; Y1t -Yi,, is a parameter that reflects aggregate changes in the economy that can plausibly be attributed to the incumbent; and W1,-Wit_1 is a parameter that measures the change in net transfers (transfers-taxes) for voter i during the incumbent's previous period in office. '