Exploring Changes Over Time in the Targeting of a Cash Transfer Program in Ecuador

Matthew S. Winters University of Illinois at Urbana-Champaign [email protected]

April 2010

ABSTRACT: The Bono de Desarrollo Humano (BDH) program in Ecaudor is a conditional cash transfer program in existence since 2003 and targeted at poor households throughout the country. The original recipient list for the program was based on a 2002 survey of 2 million households. After the initial creation of the beneficiary list, households joined the program in a piecemeal fashion when they petitioned the government, and very few households were taken off the recipient roster. In the 2006 election campaign, the ultimate victor, Rafael Correa, promised to double the size of the program‟s transfers. I examine the extent to which increases in BDH coverage at the canton level during the first year of the Correa administration are associated with the voting patterns of those cantons. I find evidence that pro-Correa cantons received disproportionately large increases in the number of cash transfers. In addition, when a new household survey was used to expand the recipient lists in 2009, pro-Correa cantons received disproportionately small increases in the number of cash transfers. I suggest three possible explanations for these patterns in the data.

A previous version of this paper was presented at the 2009 Annual Meeting of the Midwest Political Science Association. Thanks to Sebastian Karcher for useful comments, to Ben Shepherd and Rebecca Weitz-Shapiro for useful conversations and to Amanda Cronkhite for useful research assistance.

A number of recent studies have demonstrated that social welfare programs are susceptible to electoral manipulation for the benefit of the governing party (Molinar and Weldon 1994; Bruhn 1996; Shady 2000; Rocha Menocal 2001; Takahashi 2005; Weitz-Shapiro 2006; Diaz-Cayeros, Estévez and Magaloni 2007; Penfold-Becerra 2007). These papers generally have found that political incumbents increase social welfare transfers to core supporters, although there also is evidence that incumbents will try to target swing voters (Schady 2000) or even punish opposition voters (Rocha Menocal 2001). Other studies have demonstrated the effectiveness of these redistributive strategies: targeted social welfare benefits clearly increase incumbent vote share (Chen 2008a, 2008b; Manacorda, Miguel and Vigorito 2009). In this paper, I examine the Bono de Desarrollo Human (BDH) program in Ecuador for signs of manipulation during President Rafael Correa‟s first year in office.

Given that the

program is the largest social welfare program in Ecuador and that Correa emphasized the expansion of benefits during his electoral campaign, it seems likely that the BDH program would be a key means by which the government could reward loyalists, maintain support and possibly increase support. Indeed I find that the level of electoral support for Correa in a canton is a significant predictor of the increase in program transfers from 2006 to 2007, even after controlling for the level of poverty. In addition, by looking at changes in the recipient roster from 2008 to 2010 – after Ecuador had conducted a new eligibility survey – I show that pro-Correa cantons had disproportionately small increases in this latter time period, since their rolls already had swelled in the earlier period. Although these associations might suggest a supply-side intervention by the Correa administration, I also offer two demand-side possibilities through which pro-Correa cantons could come to also have first larger and then smaller increases in BDH recipients as compared to non-Correa cantons. I also investigate whether first or secondround election results are the stronger predictor and whether the association is stronger in swing or core districts.

1

The Bono de Desarrollo Humano Program Ecaudor endured a serious economic crisis in 1998-99.1 In the wake of the international financial crisis of 1998-99 and devastation caused by the El Niño weather phenomenon, Ecuador sustained a GDP growth rate of 0.4 percent in 1998 and then -7.3 percent in 1999. The proportion of households living below the poverty line swelled from 34 percent in 1995 to 56 percent in 1999. In September 1998, the government introduced the Bono Solidario program throughout the country. The goal was to replace fuel subsidies with a direct cash transfer to the poor.

Impoverished households with children under 18 and those over 65 years old were

eligible to receive an approximately $15 per month transfer. government also made disabled individuals eligible.

Beginning in April 1999, the

Citizens could enroll in the program

through local churches and then received the transfers at local banks. A 1999 household survey revealed that Bono Solidario transfers made up an average of 10 percent of annual income for households in the poorest quintile of the income distribution in Ecuador (Velásquez Pinto 2003: 7). That year, the program was reaching 44 percent of Ecuadorean households. The two main problems with Bono Solidario – as with other social welfare programs -were leakage (i.e. undeserving households receiving the benefit) and undercoverage (i.e. deserving households not receiving the benefit). In particular, the project was alleged to have an urban bias (Velásquez Pinto 2003: 8). Therefore, in 2002, Ecuador fielded the Sistema de Identificación y Elección de Beneficiarios (SELBEN) survey to identify poor households deserving of cash transfers. The initial SELBEN survey covered 90 percent of rural households in Ecuador and 90 percent of households in the most impoverished urban areas (Schady and Araujo 2006: 4). From this survey, the government created a 28-element poverty index, and households with scores in the lowest two quintiles of the SELBEN index became eligible for the new Bono de Desarrollo Humano program.

1

However, because of budget constraints, the

Background information in this paragraph comes from Velásquez Pinto (2003).

2

government could not immediately provide transfers to all households, and the program has expanded over time with new recipients being added to the BDH roster. Until August 2009, updating of the BDH roster was piecemeal at the request of individuals who believed that they qualified but were not yet receiving the transfer.2 In June 2008, the government began collecting a new SELBEN survey, and the BDH roster was updated based on this survey beginning in August 2009. In this paper, I make use of these two different time periods of program expansion – one piecemeal and one systematic – to provide evidence of an association between electoral support for the Correa government and BDH program expansion. Although the goals of BDH – as evidenced by giving the transfer to mothers with children at home – included increasing school attendance, decreasing child workforce participation and increasing child visits to health care centers, until May 2008, there was no monitoring of beneficiaries with regard to these outcomes.3

Beginning in June 2008, the Social Welfare

Ministry began to warn people that they could lose their eligibility for the transfer if their children were not in school, and the ministry began some selective monitoring, as well. Even before the beginning of monitoring, program evaluations revealed that BDH has increased school enrollment and decreased child labor force participation at levels comparable to other (truly conditional) cash transfer programs (Schady and Araujo 2006; Litschig and Winters 2008; Oosterbeek, Ponce and Schady 2008).

Bono de Desarrollo Humano and Ecuadorean Politics Former Economy and Finance Minister Rafael Correa ran for president of Ecuador in 2006 as head of the new Alianza PAIS party. His campaign emphasized national sovereignty 2

Personal communication with members of the BDH evaluation team at the National Ministry for Planning and Development (Secretaría Nacional de Planificación y Desarrollo), June 2008. 3 Oosterbeek, Ponce and Schady (2008), however, say that “a substantial part of the potential beneficiaries” nonetheless viewed the transfers as conditional rather than unconditional (2). There was a publicity campaign with regard to school and health care center attendance that accompanied the rollout of BDH.

3

(i.e. from relationships with the international financial institutions and the United States) and increased social welfare benefits for the poor. With regard to the latter, Correa specifically promised to double the size of the transfers in the Bono de Desarrolo Humano program. In addition, Correa promised to revise Ecuador‟s constitution to increase political stability in the country. The incumbent president, Alfredo Palacio, had come into office from the position of vice president following the legislature‟s removal in April 2005 of President Lucio Gutiérrez. Palacio was not affiliated with any political party, although he is rumored to have cultivated ties to several during his time in office. His administration was seen as a caretaker administration until the 2006 elections, and he did not participate in the race. In October 2006, Correa finished second in the first-round election behind banana tycoon Álvaro Noboa. In the second-round runoff the following month, Correa captured 57 percent of the vote against Noboa, and he took office on 15 January 2007.

True to his

campaign promise, the government doubled the size of the BDH benefit from $15 to $30 per recipient in February. In addition, the overall size of the program expanded at an increasingly rapid rate: the number of transfers increased by 7.4 percent in 2007, whereas growth had been 4.0 percent in 2006 and 6.6 percent in 2005 (see figure 1). Previous research on the manipulation of social welfare programs for electoral gain has focused on the year before the election, on the assumption that this is the time period during which an incumbent regime has the most incentive to use government spending to either motivate its core supporters to vote or else to try and win over swing voters. In this study, I look at the year after the election and explore the association between voting in the preceding election and the expansion of the social welfare program in the first year of the new government. Insofar as the government may have directed the expansion of the BDH in particular areas, there are two reasons why it makes sense to look at the post-election period in Ecuador 4

in 2007. First, the fact that BDH was such a major part of Correa‟s campaign suggests that his administration would want to increase the program among supporters as a means of making good on campaign promises. Second, even though Correa had won the election in November 2006, on his inauguration day he already was announcing a new vote – for a referendum on whether or not there should be a new constitution, which, if approved, then would be followed by a vote for members of the constituent assembly that would draft the document. The referendum vote was scheduled for March and later moved to April 15.4 Given a desire to build support in advance of these two votes, we might expect the Correa administration to roll out increases in the BDH program in such a way as to sure up core supporters and possibly convince swing voters to side with the government in the referendum and when electing members of the constituent assembly. These two ideas lead to my first hypothesis: H1: Areas that voted for Correa in the 2006 elections saw disproportionately large gains in BDH transfers in 2007. At the same time that Correa was expanding the BDH program, there was recognition within the government that the piecemeal approach to adding new recipients to the BDH roster was incomplete and sub-optimal for poverty alleviation purposes. Therefore, a new SELBEN survey was designed in order to systematically update the recipient roster. The survey went into the field in the summer of 2008, and the BDH rosters were updated beginning in August 2009. This update of the rosters involved not only the addition of eligible recipients to the program but also the deletion of ineligible recipients. Because the updating of the rosters in 2009 was based on this new survey, additions to the BDH roster in 2009 and 2010 should not be directly correlated with political support for the Correa government. And insofar as pro-Correa areas already had seen an increase in the number of transfers that they received during the earlier time period, these areas should have 4

Ecuador voted overwhelmingly in favor of drafting a new constitution, and elections for the constituent assembly occurred on September 30. Voters approved a new constitution one year later, and Correa was reelected under its provisions in April 2009.

5

been less likely to see as large an increase (per capita) as other areas when the rosters were augmented based on the new SELBEN. Therefore, my second hypothesis is H2: Areas that voted for Correa in the 2006 election saw disproportionately small gains in BDH transfers beginning in August 2009. Although I have motivated these two hypotheses in terms of a supply-side story in which the Correa government makes use of the BDH program for political ends, following the presentation of the results, I will suggest some alternative demand-side hypotheses that would produce similar patterns in the data.

Data The outcome variables of interest are, first, the change from 2006 to 2007 and, second, the change from 2008 to 2010 in the number of BDH recipients5 in each canton in Ecuador.6 With the first variable, I look for the relative size of the piecemeal increases during Correa‟s first year in office, and with the second variable, I examine the relative size of increases following the survey-based updating of the BDH recipient list that began in August 2009. I operationalize these outcome variables in several ways. I look at the year-to-year change in transfers per capita.7 And I also look at the total number of recipients, using a lagged value of the variable as an explanatory variable. (In addition, I look at the year-to-year percent change in total transfers to each canton, but I reserve these results for the appendix.)

5

Some previous studies of political targeting in social welfare programs have looked at expenditures per capita. Although there is variation in the size of the transfer across the three categories of BDH recipients – the elderly and disabled receive transfers that are smaller than those received by mothers – for the moment, I look only at the number of transfers. 6 Ecuador’s 24 provinces contain a total of 223 cantons, and there are an additional three cantons that are not part of any province. These cantons are further divided into parishes. Although I have BDH recipient numbers at the parish level, I only have voting returns at the canton level. Therefore, the unit-of-analysis is the canton. 7 Although cash transfers are often thought of as going to households rather than to individuals, households in Ecuador sometimes include multiple BDH recipients (e.g. a mother of school-aged children and an elderly person). Therefore, a per capita measure is more appropriate than a per household measure.

6

As in previous studies of this type, I seek to juxtapose a need-based explanation – where places with more poverty receive more cash transfers – against a politics-based explanation – where places that are electorally desirable to the incumbent regime receive more cash transfers. My main explanatory variable of interest is a measure of Correa‟s political support within a canton. I measure this in a number of different ways: as the vote share that went to Correa in the first round, as the vote share that went to Correa in the second round, as an indicator for whether or not Correa had the plurality of votes in the first round and as an indicator for whether or not Correa had the majority of votes in the second round.8 To capture needs-based increases in program size, I include a measure of the percentage of individuals living in poverty in each canton, as calculated by the National Statistics and Census Institute from 2001 census data adjusted by 2006 household survey data.9 In addition, I control for the percentage of the canton population that is of indigenous origin using data from the 2001 census.10 One of the 28 household-level factors that determine recipient selection in the SELBEN is whether or not the household is of indigenous origin. This variable is particularly important to include in the statistical models given the fact that Correa did very well among indigenous voters.11 Insofar as this variable then correlates both with Correa‟s vote share and with the number of program recipients, its exclusion might yield omitted variable bias in the estimations. In analyses where the total number of BDH transfers is the outcome

8

In all cases, I use the number of valid votes in the denominator, removing blank votes and votes that were nullified. 9 All population numbers used in this analysis are based on these 2006 extrapolations from the 2001 census. Unfortunately, this means that the analysis is unable to take account of any disproportionate changes in population during the four-year period under study. However, this relatively short time period should alleviate concerns that changes in program enrollment were due to drastic changes in population size. 10 Around seven percent of the population considers themselves indigenous. The overwhelming majority of Ecuadoreans – 77 percent – classify themselves as mestizo, meaning of a mixed Spanish and indigenous ancestry. The census data do not allow me to distinguish the extent of indigenous identity among mestizos. The proportion of indigenous residents varies greatly across cantons from about one-tenth of one percent in several cantons in the province of Loja to over 90 percent in the cantons of Chimborazo in Guamote, Taisha in Morona Santiago and Arajuno in Pastaza. 11 He speaks Quichua, the most prevalent indigenous language, and once spent a year doing volunteer work among indigenous communities.

7

variable, I also control for population. Unfortunately, all of these variables are measured at a single point in time only, either 2001 or 2006. This means that insofar as there were changes in the canton-level poverty rates between 2006 and 2010, they are not reflected in the analysis that examines the period of systematic updating of the recipient roster. In addition, although the poverty rate and the indigenous population are meant to proxy for the likelihood of BDH eligibility criteria being met in a particular canton, the actual SELBEN measure is a formula based on 28 household characteristics, meaning that they are ultimately brute indicators for the canton-level eligibility rates. Insofar as some of the specific criteria in the SELBEN formula may cluster geographically – for instance, the index includes questions about whether the household has members living abroad and whether the household is receiving remittances – then it is possible that there may be clustering of citizens eligible for BDH in cantons that also voted for Correa. In this case, an association between Correa‟s vote share and the rate of increase in BDH participants might be spurious even after controlling for the poverty rate and the size of the indigenous population.

Results Changes in BDH Transfers During Correa’s First Year in Office First, I look at the changes in BDH transfers during Correa‟s first year in office, from 2006 to 2007. Figure 2 shows that all cantons received increasing per capita BDH transfers during this period. I expect those cantons that voted for Correa in the 2006 elections to receive a greater increase in BDH transfers than those cantons that did not support Correa. The descriptive statistics in tables 1 and 2 display exactly this trend. Across all cantons between December 2006 and December 2007, the number of BDH transfers per capita increased by an average of 0.0011. In cantons won by Correa in the first round, the increase was 0.0012, whereas in cantons not won by Correa in the first round, the increase was only 0.0010. This difference is just shy of the 95 percent confidence level (p < 0.06). Similarly, in cantons won by 8

Correa in the second round, the increase was 0.0011, whereas in cantons not won by Correa in the second round, the increase was only 0.0009. This difference is statistically significant at conventional levels. In both the first and second rounds, Correa won cantons where the number of transfers per capita in 2006 was lower than the national average, so the fact that these cantons sustained a higher than average increase in cash transfers may be something of a reversion to the mean effect whereby Correa supporters were clustered in cantons that were receiving less than their fair share of BDH transfers.

Therefore, it is important to control for the background

characteristics of these cantons, specifically their level of poverty. Turning to a multivariate framework, table 3 displays the results of OLS regressions of the canton-level absolute change in BDH transfers per capita between 2006 and 2007 on Correa‟s political support controlling for the canton‟s poverty rate.

Across four different

measures of canton-level support for Correa‟s candidacy, pro-Correa cantons experienced larger increases in per capita BDH transfers during the Correa administration‟s first year in office than did those cantons not supporting Correa. According to model (1), a one percentage point increase in the vote for Correa during the first round election corresponds to a greater increase in BDH transfers than would a one percentage point increase in the poverty rate! This is clearly a substantively important association. It is not clear whether it was better to be a Correa-supporting canton in the first or second round.12 According to the point estimates for the vote share variables, an additional percentage point of support for Correa during the first round resulted in a greater increase of BDH transfers in 2007 than did an additional percentage point of support for Correa during the runoff election. However, the coefficient estimate on an indicator for a Correa victory in a canton during the first round is smaller than the coefficient estimate on an indicator for a Correa

12

The canton-level correlation between Correa’s first and second-round vote shares is r = 0.41.

9

victory during the second round.13 I explore the relative strength of the first and second round associations in more depth in the final section of the paper. In table 4, the outcome variable is the total number of BDH transfers in 2007. In these regressions, I include as explanatory variables the number of 2006 transfers, the canton poverty rate, the indigenous population proportion and also the canton‟s population. Across all four models shown in this table, we see a very consistent estimate that cantons, controlling for their other characteristics, had, on average, 1.05 BDH transfers in 2007 for every one that they had in 2006. The fact that the poverty variable is not statistically significant suggests that BDH underwent a general broadening in the first year of the Correa administration and not necessarily a targeted broadening.

(Much of the effect of the poverty rate is likely to be

subsumed in the variable measuring the number of 2006 transfers – that is, the previous year‟s program size already incorporates the poverty characteristics of the canton.) The coefficient estimates on the political variables in table 4 are all statistically significant, suggesting that cantons voting for Correa received larger increases in BDH transfers than cantons otherwise comparable in terms of 2006 transfers, population size and poverty rate. The coefficient on Correa‟s first round vote share in column (1) of table 4 suggests that a canton that gave 10 percentage points more to Correa in the first round would receive 62 more BDH transfers than an otherwise comparable canton in 2007. The coefficient in column (2) on the Correa-victory indicator similarly suggests that the 58 cantons Correa won in the first round each had, on average, an additional 115 recipients added to their BDH rosters by virtue of being pro-Correa constituencies. Appendix Table A1 displays the results of OLS regressions of the canton-level percent change in BDH transfers between 2006 and 2007 on Correa‟s political support. These models produce similar substantive results but do not perform as well in terms of overall amount of explained variation. 13

The table makes them look indistinguishable only because of rounding.

10

Consistently across multiple measures of the change in BDH recipients and multiple measures of support for Rafael Correa in 2006, the first hypothesis is supported: those cantons that more strongly supported Correa received disproportionately large increases in BDH transfers during his first year in office.

Changes in BDH Transfers Following the New SELBEN Survey Now I look at the changes in BDH rosters following the new SELBEN survey. Whereas new additions in the 2006-2007 period were based on piecemeal updating of the list, beginning in August 2009, the list was systematically updated nationwide using new household survey data. This updating involved not only the adding of new recipients to the BDH rolls but also the deletion of ineligible recipients. As figure 3 shows, in 10 cantons, this led to an overall decline in the number of transfers.14 Given the disproportionately large increases in pro-Correa cantons in the 2006-2007 period, I expect that there will be disproportionately small increases in proCorrea cantons in the 2008-2010 period as a result of the BDH rolls in these cantons already having been augmented in the earlier time period. Tables 5 and 6 provide a priori evidence of this. Whereas tables 1 and 2 show that the increases in BDH enrollment from 2006 to 2007 were greater than in pro-Correa cantons than in non-Correa cantons, tables 5 and 6 show that the increases in BDH enrollment from 2008 to 2010 were greater in non-Correa cantons than in pro-Correa cantons, suggesting that under the more systematic reassessment of the beneficiary lists that began in August 2009, the proCorrea districts that had added a disproportionate number of beneficiaries during the early part of Correa‟s presidential terms suffered from a reversion to the mean effect later on. Using Correa‟s performance in either the first or second round of the 2006 elections to proxy for his

14

Since, as noted above, the same population number from 2006 is always used as the divisor, it is likely that additional cantons experienced an overall decline on a per capita basis.

11

support in a given canton, the growth of BDH recipient lists in pro-Correa cantons is statistically significantly less than in non-Correa cantons at conventional levels. Tables 7 and 8 show that the basic relationship in the data holds up to the inclusion of controls for the poverty rate and the size of the indigenous population. In table 7, all four indicators of political support for Correa are negatively signed, although only the first-round vote share is statistically significant. (The pattern in appendix table A2 is similar.) The coefficient estimate in column (1) again suggests that a one percentage point change in Correa‟s performance in the 2006 elections was approximately equivalent in magnitude – although now in the opposite direction – to a one percentage point change in the poverty rate. In table 8, all four measures are both negatively signed and statistically significant. Cantons where Correa met with electoral success in 2006 – those cantons that were likely to receive disproportionately large increases in the number of BDH transfers from 2006 to 2007 – see disproportionately small increases (or in some cases decreases) in the number of BDH transfers from 2008 to 2010. Therefore, looking at the changes in the BDH recipient lists over the period 2006-2007 and then 2008-2010 produces complementary evidence that those areas that supported Correa‟s election in 2006 saw disproportionately large increases in the first period followed by disproportionately small increases (or even decreases) in the second period. The two hypotheses offered again have found support in the data. I motivated these hypotheses with a supply-side story about government behavior. In the next section, I explain this mechanism in more depth and also offer two alternative demand-side mechanisms that might lead to the same patterns in the data.

Possible Mechanisms for the Changing Returns to Voting for Correa The existing political science literature on the use of social welfare programs for electoral gain focuses on supply-side explanations: incumbent governments are assumed to increase 12

social welfare transfers to particular districts for electoral rewards. Often, the specific fashion in which this happens is not described, but the implicit assumption seems to be that government appointees in the relevant ministry have significant discretion for selecting beneficiary areas and therefore do so in a fashion that leads to disproportionate selection of politically-important areas. This possibly was the case in Ecuador, as well – Correa appointees in the Ministry of Social and Economic Inclusion may have sent orders to the local offices responsible for registering people for BDH, telling them to go out and actively find people to add to the BDH roster. However, I have been unable to find anecdotal evidence of such supply-side pressures in the Ecuadorean press or through interviews with key contacts. In addition, such a story would seem to contrast with the increasing professionalization of the program during Correa‟s presidency, such as the move to actually make the cash transfer conditional on school attendance and health center visits and the initiation of the new SELBEN survey to be used to update the recipient roster.15 Therefore, it is worthwhile to consider two demand-side stories related to Correa‟s election campaign.

As described above, the BDH program formed a major component of

Correa‟s campaign – he lauded the success of the program so far and promised to double the size of the transfer if elected.

Insofar as Correa performed well in districts where he

campaigned strongly, we might expect that the important role of BDH in his campaign messages had observable behavioral impacts in these districts. One demand-side story is a knowledge story. Based on Correa‟s campaigning, people in the districts where he campaigned became more aware of the program and subsequently applied for it. Although the program already was widely known throughout Ecuador – such that we might be skeptical of thinking that Correa‟s campaigning brought new attention to it – Correa‟s campaign promise to double the size of the transfer may have stimulated interest in

15

Although it is worth noting that the updating of the roster based on the new SELBEN survey did not occur until after Correa had won reelection in April 2009.

13

receiving the transfer and given people a new incentive to apply for it. So perhaps in the areas where Correa campaigned, more people were aware of the increase in the program benefit and therefore more interested in applying for the program. Another demand-side story is an entitlement story. In this scenario, Correa voters – knowing about Correa‟s promise to expand and increase BDH – more actively sought out entry into the BDH program, feeling entitled to benefit from the program because they had cast their vote for president for Correa. This explanation does not even necessarily rely on Correa having campaigned more actively in the cantons where he won. Instead it could simply be that Correa voters, knowing generally that their candidate had promised expansion of this program, more actively investigated being added to the program, leading to the disproportionate increases in BDH recipients in pro-Correa districts. Any of these three explanations could plausibly explain why there was a larger-thanaverage increase in BDH transfers in pro-Correa districts during his first year in office followed by a smaller-than-average increase in BDH transfers in pro-Correa districts during the period following the use of the new SELBEN survey results to refine the recipient lists. Unfortunately, the current data does not allow me to diagnose which explanation is the most accurate. I have no information on voter attitudes or on what motivated individuals to apply for BDH – whether they did so because they were sought out by officials from the new Correa administration or because they were motivated to do so by having heard about the program during the campaign. Also, for any of the three cases, it is worth highlighting that they suggest the relatively capable functioning of the Ecuadorean bureaucracy. The fact that the pro-Correa districts, on average, see less of an increase than other places after the completion of the new SELBEN survey implies that the bureaucracy was implementing the changes due to the new survey in an impartial and technocratic manner. Because potentially eligible recipients already had been added to the roster during Correa‟s first year in office in pro-Correa districts, the roster

14

adjustments beginning in August 2009 tended to emphasize the addition of recipients in other areas.

Additional Analyses In this section of the paper, I examine some of the other patterns in the data from the 2006 to 2007 increases. Specifically, I assume the supply-side story is true – that the Correa administration rewarded supporting areas with additional BDH transfers during its first year in office – and look to see whether first-round or second-round supporters were likely to receive larger increases and then whether core districts or swing districts were likely to receive larger increases. First Round versus Second Round Effects In tables 3 and 4 (and appendix table A1), the coefficient estimate on first-round vote share is always slightly larger than that on second-round vote share, whereas the coefficient estimate on the second-round victory indicator is always slightly larger than that on the firstround victory indicator.

Therefore, there is mixed evidence with regard to whether the

association between voting results and the change in BDH transfers is stronger based on first or second round results. In appendix table A3, I include both the first and second round voting variables in the same estimating equation. The variables measuring support for Correa in the runoff election generally perform better than those measuring first-round support. When both variables are measured as either vote shares or indicators, the second-round variable is larger and estimated with more certainty. In mixing vote shares and victory indicators, the variables from the runoff elections still perform better. Therefore, insofar as the supply-side story is correct, it seems that being a Correa loyalist in the second election – where only Correa and Noboa were candidates – was more likely to lead to increased BDH transfers in 2007.

Core versus Swing Voters 15

There is a long-running debate in political science with regard to whether incumbent parties target redistributive goods at their loyal core voters, who they expect to vote for them, or rather at swing voters, who they are hoping to convince to vote for them.16 In order to examine what type of cantons were preferred by the Correa administration in their distribution of BDH transfers, I divide cantons into core pro-Correa cantons, swing cantons and anti-Correa cantons. I define core cantons as those in which Correa‟s vote share was in the 75th percentile or above for his vote share across all cantons.17 (I define variables for both the first and second rounds of the election.) I use two possible definitions of swing cantons. Taking a broad view, I first define swing cantons as those where the Correa vote share fell between the 25th and 75th percentiles. Then I take a more conservative view and define swing cantons as those where the Correa vote share fell between its median and the 75th percentile value.18 In appendix table A4, the models in columns (1) and (2) look at the difference between core and swing districts in the first round election only; model (1) uses the more liberal definition of swing districts, and model (2) uses the more conservative definition. In both cases, the coefficient point estimate on core districts is larger – although it is quite similar to that for swing districts in model (2) – but the swing district variables achieve slightly greater precision. In models (3) and (4), this pattern is repeated when looking only at the second round elections: the magnitude on the core canton indicator is larger but estimated with less precision. This is weak evidence supporting the claim that increased BDH transfers in the first year of the Correa administration largely were aimed at core cantons.

16

Most theoretical models suggest that targeting swing voters is cheaper: given their looser ideological attachments, their votes can be won over with less money (Dixit and Londregan 1996, 1998; Persson and Tabellini 2002). Looking retrospectively, Susan Stokes argues that “voters who are predisposed in favor of *a party+ on partisan or programmatic grounds cannot credibly threaten to punish their favored party if it withholds rewards. Therefore the party should not waste rewards on them” (2005: 317). The Cox and McCubbins (1986) model of redistributive politics, however, suggests that parties should target core voters. 17 th In the first round, Correa’s 75 percentile vote share was 27 percent. In the second round, it was 69 percent. 18 th The 25 percentile value in the first round was 11 percent, and the median was 20 percent. In the second th round, the 25 percentile was 51 percent, and the median was 63 percent.

16

Adding both the first and second-round variables together in model (5), the core district variable is larger during the runoff election but not during the first-round election. The swing district variables are still estimated with more precision.19 When the more conservative secondround swing district indicator is used in model (6), all of the voting variables lose statistical significance, although the point estimates for the two first-round variables remain relatively stable as compared with the estimates in model (5) and both are greater in magnitude than the second-round indicators. Therefore, it is difficult to make conclusive claims about whether core or swing districts were more likely to receive increased BDH transfers.

Summary and Conclusions Recent scholarship has documented that governments – particularly in Latin America – are using cash transfer programs as a means of conducting electoral politics. In particular, this research has shown that governments provide redistributive transfers to core supporters in expectation of their votes in the next election. I investigate whether or not there is evidence of such manipulation in Ecuador‟s Bono de Desarrollo Humano program in the period immediately following Rafael Correa‟s election. In addition, I look to see what the patterns in the data are several years later when the government updated the BDH roster through a new household survey aimed at finding eligible recipients. During his campaign, Correa highlighted BDH as Ecuador‟s signature social welfare program and promised to double benefits to recipients. Following his election, his government immediately began politicking for the approval of a referendum to draft a new constitution and then for seats in the constitutional assembly. Therefore, increasing redistribution to supporters

19

The collinearity between the two indicators is not overwhelming. Of the 54 cantons considered first-round core districts, only 22 of them are also considered second-round core districts, whereas 33 non-core districts from the first round are classified as core districts in the second round. The hypothesis test for the joint significance of the two core canton indicators falls shy of standard statistical significance levels.

17

both would realize a campaign promise and also rally supporters in advance of new, important votes. Comparing the number of BDH transfers in 2007 to the number in 2006, there is a clear positive association between a canton having voted for Correa and having received a substantially larger increase in BDH transfers. Following the revision of the eligibility list that began in August 2009 based on the new SELBEN survey, these same cantons saw significantly smaller increases – and sometimes decreases – in the number of BDH eligible citizens within their borders. On the one hand, this evidence can be read as suggesting that the Correa administration did indeed increase BDH transfers to loyal cantons in the year 2007 and that these cantons then reverted to the mean number of transfers (conditional on poverty levels) in the later period. On the other hand, it also is possible to read this evidence as suggesting one of two demand-side stories – either Correa‟s campaigning provided people with knowledge about BDH and about his intent to double the benefit such that they were incentivized to then apply for the program or else Correa voters felt a sense of entitlement to the program and so were more likely than non-Correa voters to apply for it following his election. The current data unfortunately does not allow me to tease out which of these stories is the most plausible. Additional information about the correlation between Correa‟s campaign activities and his vote share or about individual motivations for applying to the BDH program during the period of piecemeal addition is needed. Nonetheless, the fact that the pro-Correa cantons stand out in both periods as different from the other cantons alerts us to the existence of some relationship between the Bono de Desarrollo Humano program and the electoral politics of Ecuador.

18

Figure and Tables

Figure 1. Total Number of BDH Participants. Note that the 2001-03 figures include individuals receiving transfers under the older, less exclusive Bono Solidario program.

30 20 10 0

Number of Cantons

40

50

Histogram of Changes in BDH Transfers 2006 to 2007

0

.01

.02

.03

.04

Per Capita Change in BDH Recipients 2006-2007

Figure 2. Changes in BDH Transfers from 2006 to 2007. Measured in December 2006 and December 2007.

19

30 20 10 0

Number of Cantons

40

Histogram of Changes in BDH Transfers 2008 to 2010

-.05

0

.05

.1

.15

Per Capita Change in BDH Recipients 2008-2010

Figure 3. Changes in BDH Transfers from 2008 to 2010. Measured in December 2008 and April 2010.

Whole Country

Correa Cantons (First Round) (N = 55) Non-Correa Cantons (First Round) (N = 154) Difference

BDH Participants Per Capita 2006 2007 0.125 0.136 (0.003) (0.003)

Difference 0.011 (0.0004)

0.116 (0.005)

0.128 (0.006)

0.012 (0.001)

0.128 (0.003)

0.139 (0.003)

0.010 (0.0004)

-0.012 (0.006)

-0.010 (0.006)

0.002 (0.001)

Table 1. Descriptive Statistics Comparing First Round Pro-Correa Cantons to First Round Non-Correa Cantons from 2006 to 2007. Each box contains the number of transfers per capita. Standard errors in parentheses. The Welch two sample t-statistic for the difference in the lower right hand corner is 1.95. A pro-Correa canton is one in which Correa won a plurality of votes in the first round.

20

Whole Country

Correa Cantons (Second Round) (N = 159) Non-Correa Cantons (Second Round) (N = 50) Difference

BDH Participants Per Capita 2006 2007 0.125 0.136 (0.003) (0.003)

Difference 0.011 (0.0004)

0.123 (0.003)

0.134 (0.003)

0.011 (0.0004)

0.133 (0.005)

0.142 (0.005)

0.009 (0.0005)

-0.011 (0.006)

-0.008 (0.006)

0.002 (0.0006)

Table 2. Descriptive Statistics Comparing Second Round Pro-Correa Cantons to Second Round Non-Correa Cantons from 2006 to 2007. Each box contains the number of transfers per capita. Standard errors in parentheses. The Welch two sample t-statistic for the difference in the lower right hand corner is 3.58. A pro-Correa canton is one in which Correa won a majority of votes in the run-off election.

Outcome Variable

Poverty Rate Indigenous Population Correa First Round Vote Share Correa First Round Victory Correa Second Round Vote Share Correa Second Round Victory Prob > F R2 N

(1) Per Capita Change 2006-2007 0.015*** (0.004) -0.004 (0.003) 0.016** (0.006)

(2) Per Capita Change 2006-2007 0.014*** (0.004) -0.004 (0.003)

(3) Per Capita Change 2006-2007 0.014** (0.003) -0.008** (0.004)

(4) Per Capita Change 2006-2007 0.013*** (0.004) -0.007* (0.003)

0.003** (0.001) 0.012** (0.004) 0.003** (0.001) 0.01 0.18 209

0.01 0.17 209

0.01 0.20 209

0.01 0.18 209

Table 3. Vote Share and Absolute Per Capita Change in BDH Transfers from 2006 to 2007. OLS regression results. All models include an unreported intercept. Robust standard errors clustered at the province level in parentheses. * = p < 0.10, ** = p < 0.05, *** = p < 0.01.

21

Outcome Variable BDH Transfers 2006 Poverty Rate Indigenous Population Log(Population) Correa First Round Vote Share Correa First Round Victory Correa Second Round Vote Share Correa Second Round Victory Prob > F R2 N

(1) BDH Transfers 2007 1.05*** (0.005) 68.2 (95.6) 95.0* (54.9) 103*** (20.8) 620** (229)

(2) BDH Transfers 2007 1.05*** (0.005) 23.4 (71.6) 94.5 (62.8) 100*** (19.9)

(3) BDH Transfers 2007 1.05*** (0.005) 90.7 (94.3) -121.4* (66.9) 114*** (21.3)

(4) BDH Transfers 2007 1.05*** (0.005) 69.1 (96.8) -44.1 (56.0) 112*** (23.0)

115** (42.7) 545*** (143) 156*** (55.0) 0.00 0.99 209

0.00 0.99 209

0.00 0.99 209

0.00 0.99 209

Table 4. Vote Share and 2007 BDH Transfers. OLS regression results. All models include an unreported intercept. Robust standard errors clustered at the province level in parentheses. * = p < 0.10, ** = p < 0.05, *** = p < 0.01.

Whole Country

Correa Cantons (First Round) (N = 55) Non-Correa Cantons (First Round) (N = 154) Difference

BDH Participants Per Capita 2008 2010 0.141 0.201 (0.003) (0.004)

Difference 0.059 (0.002)

0.135 (0.006)

0.183 (0.008)

0.049 (0.005)

0.144 (0.003)

0.207 (0.004)

0.063 (0.003)

-0.009 (0.006)

-0.023 (0.008)

-0.014 (0.006)

Table 5. Descriptive Statistics Comparing First Round Pro-Correa Cantons to First Round Non-Correa Cantons from 2008 to 2010. Each box contains the number of transfers per capita. Standard errors in parentheses. The Welch two sample t-statistic for the difference in the lower right hand corner is 2.45. A pro-Correa canton is one in which Correa won a plurality of votes in the first round.

22

Whole Country

Correa Cantons (Second Round) (N = 159) Non-Correa Cantons (Second Round) (N = 50) Difference

BDH Participants Per Capita 2008 2010 0.141 0.201 (0.003) (0.004)

Difference 0.059 (0.002)

0.140 (0.003)

0.197 (0.004)

0.057 (0.003)

0.147 (0.005)

0.212 (0.007)

0.065 (0.005)

-0.007 (0.007)

-0.015 (0.009)

-0.008 (0.006)

Table 6. Descriptive Statistics Comparing Second Round Pro-Correa Cantons to Second Round Non-Correa Cantons from 2008 to 2010. Each box contains the number of transfers per capita. Standard errors in parentheses. The Welch two sample t-statistic for the difference in the lower right hand corner is 1.45. A pro-Correa canton is one in which Correa won a majority of votes in the run-off election.

Outcome Variable

Poverty Rate Indigenous Population Correa First Round Vote Share Correa First Round Victory Correa Second Round Vote Share Correa Second Round Victory Prob > F R2 N

(1) Per Capita Change 2008-2010 0.093*** (0.021) -0.052*** (0.011) -0.094** (0.043)

(2) Per Capita Change 2008-2010 0.104*** (0.021) -0.050*** (0.013)

(3) Per Capita Change 2008-2010 0.108*** (0.021) -0.047*** (0.016)

(4) Per Capita Change 2009-2010 0.108*** (0.021) -0.047*** (0.014)

-0.009 (0.006) -0.006 (0.025) -0.003 (0.007) 0.00 0.24 209

0.00 0.20 209

0.00 0.19 209

0.00 0.19 209

Table 7. Vote Share and Absolute Per Capita Change in BDH Transfers from 2008 to 2010. OLS regression results. All models include an unreported intercept. Robust standard errors clustered at the province level in parentheses. * = p < 0.10, ** = p < 0.05, *** = p < 0.01.

23

Outcome Variable BDH Transfers 2008 Poverty Rate Indigenous Population Log(Population) Correa First Round Vote Share Correa First Round Victory Correa Second Round Vote Share Correa Second Round Victory Prob > F R2 N

(1) BDH Transfers 2010 1.00*** (0.050) 2121** (942) -1804*** (516) 826*** (142) -5241*** (1617)

(2) BDH Transfers 2010 1.00*** (0.050) 2399*** (830) -1811*** (837) 837*** (120)

(3) BDH Transfers 2010 1.00*** (0.050) 2272*** (776) -393 (521) 777*** (130)

(4) BDH Transfers 2010 0.999*** (0.051) 2371** (913) -855 (520) 786*** (141)

-1087** (387) -3495** (158) -1041** (434) 0.00 0.99 209

0.00 0.99 209

0.00 0.99 209

0.00 0.99 209

Table 8. Vote Share and 2010 BDH Transfers. OLS regression results. All models include an unreported intercept. Robust standard errors clustered at the province level in parentheses. * = p < 0.10, ** = p < 0.05, *** = p < 0.01.

24

Works Cited Bruhn, Kathleen. 1996. “Social Spending and Political Support: The „Lessons‟ of the National Solidarity Programme in Mexico,” Comparative Politics 28.2: 151–77. Chen, Jowei. 2008a. “When Do Government Benefits Influence Voters' Behavior? The Effect of FEMA Disaster Awards on US Presidential Votes,” Department of Political Science, Stanford University. -----. 2008b. “Are Poor Voters Easier to Buy Off with Money? A Natural Experiment from the 2004 Florida Hurricane Season,” Department of Political Science, Stanford University. Cox, Gary W. and Mathew D. McCubbins. 1986. “Electoral Politics as a Redistributive Game,” Journal of Politics 48.2: 370–89. Diaz-Cayeros, Alberto, Federico Estévez and Beatriz Magaloni. 2007. Strategies of Vote Buying: Poverty, Democracy and Social Transfers in Mexico. Unpublished book manuscript. Dixit, Avinash and John Londregan. 1996. “The Determinants of Success of Special Interests in Redistributive Politics,” Journal of Politics 58.4 (November): 1132-55. -----. 1998. “Ideology, Tactics and Efficiency in Redistributive Politics,” Quarterly Journal of Economics 113.2: 497-529. Litschig, Stefan and Matthew S. Winters. 2008. “Unconditional Cash Transfers and School Enrollment: Non-experimental Evidence from Ecuador‟s Bono de Desarrollo Humano Program,” Columbia University IGERT Program on International Development and Globalization. Manacorda, Macro, Edward Miguel and Andrea Vigorito. 2009. “Government Transfers and Political Support,” Department of Economics, University of California, Berkeley. Molinar, Juan and Jeffrey Weldon. 1994. “Electoral Determinants and Consequences of National Solidarity,” in Wayne Cornelius, Ann Craig and Jonathan Fox, eds., Transforming State-Society Relations in Mexico: The National Solidarity Strategy. La Jolla, California: University of California at San Diego Center for U.S. Mexican Studies. Oosterbeek, Hessel, Juan Ponce and Norbert Schady. 2008. “The Impact of Cash Transfers on School Enrollment: Evidence from Ecuador,” World Bank Policy Research Working Paper 4645, June. Washington, D.C.: The World Bank. Penfold-Becerra, Michael. 2007. “Clientelism and Social Funds: Evidence from Chávez's Misiones,” Latin American Politics & Society 49.4: 63-84. Persson, Torsten and Guido Tabellini. 2002. Political Economics: Explaining Economic Policy. Cambridge, Massachusetts: M.I.T. Press. Rocha Menocal, Alina. 2001. “Do Old Habits Die Hard? A Statistical Exploration of the Politicization of Progresa, Mexico‟s Latest Federal Poverty-Alleviation Programme, Under the Zedillo Administration,” Journal of Latin American Studies 33:513-38. Schady, Norbert R. 2000. “The Political Economy of Expenditures by the Peruvian Social Fund (FONCODES), 1991-95,” American Political Science Review 94.2: 289-304. ----- and Maria Caridad Araujo. 2006. “Cash Transfers, Conditions, School Enrollment, and Child Work: Evidence from a Randomized Experiment in Ecuador,” World Bank Policy Research Working Paper 3930, June. Washington, D.C.: The World Bank. Stokes, Susan. 2005. “Perverse Accountability: A Formal Model of machine Politics with Evidence from Argentina,” American Political Science Review 99: 315-25. Takahasi, Yuriko. 2005. “Dynamic Theory of Electoral Competition and Redistribution: Evidence from Mexico,” paper presented at the Comparative Political Economy Workshop, Harvard University. Velásquez Pinto, Mario D. 2003. The Bono Solidario in Ecuador: An Exercise in Targeting. Extension of Social Security Paper No. 17. Geneva: International Labour Organization. 25

Weitz-Shapiro, Rebecca. 2006. “Partisanship and Protest: The Politics of Workfare Distribution in Argentina,” Latin American Research Review 41.3: 122-47.

26

APPENDIX

Outcome Variable Poverty Rate Indigenous Population Correa First Round Vote Share Correa First Round Victory Correa Second Round Vote Share Correa Second Round Victory Prob > F R2 N

(1) Percent Change 2006-2007 0.035 (0.026) -0.015 (0.019) 0.114* (0.064)

(2) Percent Change 2006-2007 0.029 (0.022) -0.015 (0.020)

(3) Percent Change 2006-2007 0.031 (0.018) -0.057** (0.020)

(4) Percent Change 2006-2007 0.026 (0.019) -0.039* (0.019)

0.022** (0.010) 0.111*** (0.030) 0.029** (0.010) 0.24 0.08 209

0.16 0.08 209

0.01 0.15 209

0.07 0.11 209

Table A1. Vote Share and Percent Change in BDH Transfers from 2006 to 2007. OLS regression results. All models include an unreported intercept. Robust standard errors clustered at the province level in parentheses. * = p < 0.10, ** = p < 0.05, *** = p < 0.01.

27

Outcome Variable Poverty Rate Indigenous Population Correa First Round Vote Share Correa First Round Victory Correa Second Round Vote Share Correa Second Round Victory Prob > F R2 N

(1) Percent Change 2008-2010 0.321 (0.301) -0.324** (0.138) -1.01* (0.552)

(2) Percent Change 2008-2010 0.435 (0.292) -0.310** (0.146)

(3) Percent Change 2008-2010 0.491 (0.290) -0.304* (0.164)

(4) Percent Change 2008-2010 0.481 (0.287) -0.279* (0.154)

-0.095 (0.087) 0.017 (0.239) -0.025** (0.082) 0.01 0.08 209

0.03 0.05 209

0.15 0.04 209

0.14 0.04 209

Table A2. Vote Share and Percent Change in BDH Transfers from 2008 to 2010. OLS regression results. All models include an unreported intercept. Robust standard errors clustered at the province level in parentheses. * = p < 0.10, ** = p < 0.05, *** = p < 0.01.

28

Outcome Variable

Poverty Rate Indigenous Population Correa First Round Vote Share Correa First Round Victory Correa Second Round Vote Share Correa Second Round Victory Prob > F for H0: First + Second Round = 0 Prob > F R2 N

(1) Per Capita Change 2006-2007 0.015*** (0.004) -0.007** (0.003) 0.008 (0.005)

(2) Per Capita Change 2006-2007 0.014*** (0.004) -0.006* (0.003)

(3) Per Capita Change 2006-2007 0.015*** (0.004) -0.006* (0.003) 0.011** (0.005)

0.002* (0.001)

(4) Per Capita Change 2006-2007 0.014*** (0.004) -0.007** (0.003)

0.001 (0.001) 0.010*** (0.003)

0.009** (0.004) 0.002** (0.0009)

0.002** (0.0008)

0.03

0.03

0.03

0.03

0.01 0.21 209

0.02 0.20 209

0.02 0.20 209

0.01 0.21 209

Table A3. Comparing the Effects of First and Second Round Voting on Absolute Per Capita Change in BDH Transfers from 2006 to 2007. OLS regression results. All models include an unreported intercept. Robust standard errors clustered at the province level in parentheses. * = p < 0.10, ** = p < 0.05, *** = p < 0.01.

29

Outcome Variable

Poverty Rate Indigenous Population Correa First Round Core Canton Correa First Round Swing (1) Canton Correa First Round Swing (2) Canton Correa Second Round Core Canton Correa Second Round Swing (1) Canton Correa Second Round Swing (2) Canton Prob > F for H0: 1st Round Core + 2nd Round Core = 0 Prob > F for H0: 1st Round Swing + 2nd Round Swing = 0 Prob > F R2 N

(1) Per Capita Change 2006-2007 0.013*** (0.004) -0.004 (0.003) 0.0029* (0.0014) 0.0022** (0.0010)

(2) Per Capita Change 2006-2007 0.014*** (0.004) -0.004 (0.003) 0.0021 (0.0012)

(3) Per Capita Change 2006-2007 0.014*** (0.004) -0.008* (0.004)

(4) Per Capita Change 2006-2007 0.013*** (0.004) -0.007* (0.004)

(5) Per Capita Change 2006-2007 0.014*** (0.004) -0.007* (.004) 0.0013 (0.0011) 0.0016** (0.0007)

0.0020* (0.0011) 0.0043** (0.0019) 0.0029*** (0.0010)

0.0025 (0.0015)

0.0039* (0.0019) 0.0027** (0.0010)

0.0013 (0.0009)

0.04 0.16 209

0.02 0.15 209

0.02 0.19 209

0.03 0.15 209

(6) Per Capita Change 2006-2007 0.014*** (0.004) -0.006 (0.003) 0.0013 (0.0012)

0.0016 (0.0009) 0.0018 (0.0015)

0.0008 (0.0007)

0.06

0.21

0.03

0.16

0.05 0.21 209

0.03 0.16 209

Table A4. Core versus Swing Districts and Absolute Per Capita Change in BDH Transfers. OLS regression results. All models include an unreported intercept. Robust standard errors clustered at the province level in parentheses. * = p < 0.10, ** = p < 0.05, *** = p < 0.01.