The Politics of Informal Redistribution in Latin America

Voting for Forbearance: The Politics of Informal Redistribution in Latin America Conventional wisdom is that poor voters in Latin America have amorphous electoral interests that lead them to sell their votes, support anti-establishment politicians, or swing between parties. This paper argues that the urban informal sector poor have a coherent material interest in forbearance toward property laws that they tend to violate. Because informal welfare benefits are provided through inaction, politicians can send more credible electoral cues of their affinity with poor voters through forbearance than through traditional social policy promises. To demonstrate this, I develop an electoral model in which voters use enforcement as a way to learn about which politicians are likely to represent them in office, and politicians in turn avoid enforcement in poor districts due to the electoral consequences. I then test these implications using a survey experiment in Bogotá, Colombia and a database of campaign platforms from Lima, Peru. This paper suggests that forbearance functions as an informal mode of redistribution, and that it structures electoral behavior more clearly than traditional tax-based redistribution in Latin America. 1 The Law, in its majestic equality, forbids, the rich, as well as the poor, to sleep under the bridges, to beg in the streets, and to steal bread. —Anatole France, The Red Lily (1894) The urban poor have been placed at the center of Latin America’s weak system of political representation. Changes associated with the urban poor’s employment in the informal sector have been credited with the failure of class-based political parties (Roberts 1998, 2002), the breakdown of traditional party systems (Cameron 1994; Morgan 2011), and the lethargic adoption of programmatic social policies (Haggard, Kaufman, and Long 2013; Weyland 1996). Scholars grapple with what Robert Kaufman (2009) calls the “inconvenient facts” that Latin America’s poor do not demand more redistribution or vote for leftist politicians who advocate it. Instead, the sense is that informal sector voters have amorphous interests that lead them to swing erratically among parties, sell their votes, or support politicians based on populist appeals. This paper pushes back against these dim portrayals of the urban poor’s electoral behavior.1 It introduces the concept of forbearance, or the intentional non-enforcement of law, and argues that poor voters have a coherent material interest in forbearance toward laws that they tend to violate, such as those against squatting, street vending, and unlicensed taxi driving. Poor voters understand forbearance as an informal welfare policy for those willing to assume the legal risks. They support candidates who advocate it. Nonetheless, forbearance tends to be a second-best option to formal versions of the same goods, such as a house or a job. Support for forbearance comes from the fact that it is a credible signal of a politician’s distributive commitments in contexts where social welfare policies often fail to come to fruition or favor formal sector workers. These heuristics thus are likely to matter most in countries with welfare states that do little to aid the poor. In such contexts, 1 Due to a high correlation between urban poverty and informality, I refer interchangeably to the urban poor and the informal sector poor. Three-quarters of the bottom quintile of the income distribution work in the informal sector in Latin America (Perry 2007). 2 politicians use forbearance as a way to provide material benefits to constituents and indirectly signal their affinities when they depend on the poor’s backing to win office. Drawing on an original survey and campaign data, this paper tests this theory in two ways. First, the electoral effects of policies, including those stemming from enforcement (or its absence), are typically difficult to infer because they are associated with other unobservable candidate characteristics. I therefore ran a survey experiment in Bogotá, Colombia to isolate whether forbearance serves as a heuristic for the distributive commitments of candidates and motivates electoral behavior. I find that forbearance indeed shapes candidate assessments and vote intention in the ways my model predicts. While 60 percent of poor voters believe that a candidate who promises not to enforce the law serves their interests, only 25 percent see a candidate who plans to enforce as serving their interests. Second, the drawback of survey experiments is that they abstract from the natural world; rarely do voters receive isolated bits of policy information on which to base their decisions. To better understand the salience of forbearance as an electoral issue, I turn to observational data on more than 300 local campaign platforms in Lima, Peru. I show that forbearance platforms are ubiquitous in poor electoral districts, where my micro-level findings suggest that they are most likely to gain favor. Combined, these results suggest that politicians face variable incentives to enforce property laws, and may hold back due to the perceived electoral repercussions. This electoral theory contrasts with dominant alternatives rooted in institutional capacity constraints, sociolegal culture, or special interest group politics alone. Attention to enforcement contributes to our understanding of the distributive issues at stake in politics, especially for poor voters historically marginalized from the formal welfare state. More broadly, in introducing the concept of forbearance, this paper highlights the importance of thinking of enforcement as a variable that can be manipulated by politicians for electoral ends. As Levitsky and Murillo (2009) emphasize, different levels of enforcement can alter our theories about 3 institutional design, effects, and change. But we still know little about the sources and variability of enforcement outcomes apart from the institutional constraints that mean that politicians cannot enforce their own laws. Thinking in terms of forbearance opens rich avenues for future research on the electoral determinants of enforcement and the interaction with other social policies. The Concept This paper introduces a new concept, forbearance, to distinguish deliberate choices not to enforce state laws from garden-variety state weakness. In common parlance, forbearance describes a broad set of situations. It can imply a decision not to enforce a contract, debt, law or regulation. Here, I define forbearance as intentional and revocable government leniency toward violations of the law. There are three components to this definition—capacity, intention, and revocability—that I walk through and use to distinguish related concepts in Figure 1. First, institutional capacity distinguishes forbearance from forms of weak enforcement in which actors cannot enforce the law. In states with scant personnel or budgets, the enforcement outcome is uniform. Forbearance only can occur when states plausibly muster the infrastructure to sanction offences.2 Second, and most critically, forbearance occurs when a political actor chooses not to enforce a law. In cases of forbearance, the capacity to enforce exists, but the intent to do so is absent. Enforcement preferences can diverge between the actors who design a rule and those who implement it. A contemporary example helps clarify the point. In 2012, President Barack Obama suspended the deportation of young illegal immigrants. Although immigration authorities have the budget and agents to deport youths, and the law unambiguously orders authorities to do so, the 2 The empirical identification of a minimum level of state capacity is delicate because governments can withdraw state resources as a strategy not to enforce, or work with private actors to bolster internal capacity (e.g. Amengual 2013). States with equal capabilities also can assume different regulatory “loads” (Centeno and Portes 2006) and struggle to perform routine operations while still engaging in ‘crackdowns’ or exceptional tasks (Eeckhout, Persico, and Todd 2010; Taylor 2011). I tackle this issue of how to separate capacity constraints in separate work (Holland 2014). 4 President has decided not to follow the letter of the law. The choice evidences a political preference to alter the law’s impact at the implementation stage. A third core definitional element is that forbearance is revocable. The state reserves the right to enforce the law and offenders believe that the rules can carry a credible sanction. This insecurity differentiates forbearance from cases of amnesty, pardon or legalization. There, the state formally surrenders its enforcement rights. Revocable decisions, on the other hand, are open to revision and are more likely to be negotiated informally. A revocable decision can be formal; for instance, in 2012, Obama issued an executive order not to deport undocumented immigrants who were brought to the United States as children. The next president could reverse this written order. More often, revocable enforcement decisions are informal, as with verbal directives not to deploy the police against rioters in India (Wilkinson 2006). As long as formal decisions are provided on a temporary basis and leave those affected in a state of legal insecurity, I consider them to be revocable. The ability to go back on an enforcement decision also separates forbearance from alternative concepts of “benign neglect” or “standoffish” state behavior (e.g. Dorman 2007; Slater and Kim 2013), where states intentionally eschew administration of an area. Prolonged neglect means that states cannot easily reverse course and enforce at whim. In cases of forbearance, politicians gain leverage through the maintenance of information and procedures to change their enforcement decisions. 5 Institutional Capacity? Divergent Preferences? no Revocable? no Weak Enforcement Effective Enforcement no yes Legalization Forbearance FIGURE 1. Defining Forbearance Forbearance can be divided into subtypes based on its distributive effects.3 I define the progressivity of forbearance as in the economic analysis of the effects of taxes and transfers. Progressive forbearance occurs when lower tranches of the income distribution receive greater absolute or relative benefits through the tolerance of law breaking than upper tranches. Regressive forbearance implies that upper tranches capture greater benefits. A benefit is defined as the net value of the offense to most individuals.4 Some offenses, like copyright infringement, generate important economic gains and limited moral disutility. Other crimes, like allowing individuals to kill bankers or sell their organs, generate risks and ethical quandaries that dissipate the income gains for most individuals. Progressivity is clearest when property and regulatory violations are at stake, rather than “hard” crimes. I focus on cases of progressive subtypes because few expect the poor to influence enforcement in their favor, particularly if the written law represents the interests of more 3 Some types of forbearance do not have clear distributive consequences, such as choices about the enforcement of moral and religious regulations (laws against sodomy, abortion, and so on), but they are not the focus of this paper. 4 While I focus on the benefits derived by the person violating the law, there can be ancillary distributive effects that benefit other class groups. For example, in the cases of domestic service or illegal immigration, the benefits to the worker who violates the law may be trumped by the economic benefits to employers who gain a cheaper workforce. In classifying types of forbearance, I follow the conventions in the analysis of social policies and consider the policy incidence. But in understanding the politics of forbearance, I do allow other class groups to benefit and separately consider the social welfare effects of the policy as a whole. 6 powerful groups. It comes as little surprise that the rich can deploy power, money, and lawyers to distort the law in their favor. There are two ways that a government impacts progressivity through its enforcement decisions. First, choices about which laws and regulations to enforce have distributive consequences. In this case, progressivity is measured by how the marginal utility of an offense changes with income. Forbearance must be directed toward laws that the poor tend to violate to progressively distribute resources. The criminalization of homelessness clarifies the point. Although facially neutral, vagrancy laws have a discriminatory impact on the poor, given that few wealthy individuals choose to sleep under bridges when they have a home at their disposal. Enforcement of laws that the wealthy tend to violate (“white collar” crime), like those against pure cocaine use, foreign tax shelters or insider trading, has the opposite distributive consequences. Other laws can benefit all income groups if ignored. In these cases, progressivity depends on which groups are targeted for enforcement. Consider the case of enforcement against squatting. If a government does nothing to prevent electricity theft, then industrial users of electricity, who tend to be wealthy, absorb the greatest absolute benefit. However, the poor may receive the maximum relative benefit given that they spend a greater share of their income on basic services and they are potentially judgment proof in the case of prosecution. Weak enforcement thus has mixed distributive effects: it is likely regressive in absolute terms, and progressive in relative terms. However, if the government prosecutes industrial electricity users and ignores illegal connections in poor neighborhoods, the result is solidly progressive. The incidence depends not on the type of law, but rather on how governments use their discretion in enforcement of a law. Min and Golden (2014), for instance, find that electricity loss spiked prior to elections and concentrated among wealthy landowners in Uttar Pradesh. The fact that forbearance encompasses cases of uneven 7 enforcement of a law differentiates it from related concepts like selective policy implementation (e.g. O’Brien and Li 1999).5 The mode through which forbearance is distributed also can differ. Clientelism is often defined by individual-level quid pro quo protocol. Consistent with this definition, forbearance can be provided in specific individual cases as a reward contingent on political loyalty (Chubb 1982). The quote ascribed to Brazilian President Gétulio Vargas captures the sentiment: “For my friends, everything; for my enemies, the law.” When forbearance implicates groups of citizens, it becomes a club good. Club goods can be exchanged for a group’s turnout or votes in which case the relationship still may be clientelistic (e.g. Abente Brun and Diamond 2014; Kitschelt and Wilkinson 2007). Direct political exchange, cemented through personal or group reciprocity, can drive forbearance in some cases. Crucially, however, forbearance is not a subtype of clientelism. Much like a programmatic policy, forbearance can be provided in a transparent way and independent of political loyalty. Minimalist definitions of programmatic policy center on the idea that disbursements are 1) made based on transparent criteria, regardless of whether a particular individual supported or opposed the politician or party that designed the policy, and 2) that the criteria of distribution are formalized (Kitschelt 2000: 850; Stokes et al. 2013: 7). To continue with the immigration example, Obama instructed prosecutors to exercise their discretion to exempt all young immigrants, not to reward select followers in the Democratic Party. Politicians often are open in their beliefs that certain laws should not be enforced when offenders are poor or vulnerable. While this position can be motivated by an instrumental desire to secure votes, it is a generalized position that applies to an entire class of individuals. Indeed, as I argue below, one of the main drivers of progressive 5 Forbearance is distinct from selective implementation in two additional ways: 1) it can occur even when politicians have the resources to enforce, whereas selective implementation tends to be conceptualized as a choice about which policies to prioritize given limited resources, and 2) the actors making enforcement decisions are conceptualized as politicians, rather than bureaucrats. 8 forbearance is the incentive to signal broad affinities with the poor. Despite public positions, forbearance often is communicated outside of formal political channels and thus falls short of standard definitions of programmatic policy. Forbearance is an inefficient form of targeted redistribution that results in material and abstract costs. Law breaking imposes public harms such as the deterioration or appropriation of public goods, insecure property rights, unfair competition, and so on. Widespread noncompliance can jeopardize a state’s legitimacy and indicate an ineffective state. However, forbearance also has some advantages in contexts with weak welfare states. Most notably, it can raise the level of social good provision. Because forbearance can be provided immediately by withholding state capacity, it can result in immediate, and thus more credible, distributive improvements compared to the expansion of state welfare programs. Latin American welfare states historically have done little to aid the informal sector poor (DeFerranti, Perry, and Ferreira 2004; Díaz-Cayeros and Magaloni 2009; Haggard and Kaufman 2008). Despite recent reforms to extend coverage and improve targeting, progress has been uneven across space and policy sector (Garay 2015; Pribble 2013). I now consider why progressive forms of forbearance arise and persist, despite the fact that they impose costs on both the state and society. An Electoral Theory of Forbearance The idea that enforcement has distributive consequences is not new. But, the conventional wisdom is that those who benefit from breaking the law pay bribes, lobby or use their political connections to discourage enforcement. Much like other forms of corruption, negotiations around enforcement occur under the table. And, to the extent that politicians are involved, it is to extract resources or support by exploiting the insecurity of those who violate laws (Boone 2009; Chubb 1982; Cross 1998; Onoma 2010; De Soto and Ghersi 1989; Tripp 1997). In contrast, my electoral theory departs from the premise that some forms of forbearance garner substantial popular support. 9 Because forbearance can serve socially acceptable ends of advancing material equality, politicians sometimes openly and deliberately eschew enforcement to win votes. In the context of economically progressive offenses, I expect poor voters to support less enforcement than the nonpoor. Some poor individuals benefit directly if they engage in an offense. Others gain an indirect benefit of insurance against an income shock, or consumption benefits from cheaper goods. In contrast, the middle class suffers negative externalities from forbearance. They can experience material losses, such as reduced access to public goods like sidewalks and conservation land that the poor appropriate for private use, or abstract harms, such as the state’s weaker reputation for property rights protections. While the poor suffer many of these same harms, they are offset by concrete benefits. This materialist view of enforcement preferences makes distinct empirical predictions from two dominant alternatives. On the one hand, a common view is that legal violations persist because special interest groups who violate the law, such as organized street vendors or squatters, impose their preferences on a majority that prefers consistent enforcement (Cross 1998; Donovan 2002; Roever 2005). On the other hand, sociocultural theories of law propose that a cultural disrespect for formal rules accounts for widespread law breaking in the region and hinders the state’s ability to enforce the law. Cultural idioms (“The law is for the peasants”; “Make the law, make the way around it”; “I obey, but do not comply”) are taken as evidence of widespread cultural norms that reject enforcement and the vestiges of Latin America’s clientelistic traditions (DaMatta 1991; García Villegas 2009; Mockus, Murrain, and Villa 2012; Mockus 2002). According to this view, individuals across the income spectrum take a dim view of law and do not discriminate based on the distributive effects of law. I agree that legal norms are sometimes viewed as flexible and jettisoned by large segments of the population in Latin America. But, in contrast to special interest group theories, my argument is that forbearance against laws with progressive consequences attracts substantial support 10 among poor voters. Legal violations without progressive consequences are condemned across the income spectrum in ways that are ignored by sociocultural theories. Enforcement preferences thus are tightly bound to an individual’s material interests, much as redistributive preferences vary with an individual’s income and the progressivity of social policy: Hypothesis 1. The poor support less enforcement against progressive legal violations than the nonpoor; both poor and nonpoor voters reject regressive legal violations. As special interest group theories emphasize, those who violate the law rarely are the median voter. It is not clearly in the material interests of most poor voters to support forbearance. Nonetheless, I still expect poor voters to support forbearance because of their instrumental expectations. Politicians use forbearance to provide direct material benefits to the poor, but also to signal their broader willingness to resolve the poor’s needs. Local politicians that I interviewed stressed that they would be perceived as callous to the social circumstances of the poor if they enforced the law. Blocking enforcement, in contrast, demonstrated an understanding of the economic circumstances facing the poor and a willingness to tailor the law to resolve them. For example, to give a flavor for the way that forbearance signals an affinity for the poor, consider the logic of a councilor in a poor district in Lima, Peru: “Democracy creates more demands on the local governments because we are closer to the people and absorb the social problems. My support comes from the poor, so when there is a land invasion, I visit to show my sympathy. If you bring in the police, then it’s clear to everyone that the mayor does not really fight for the social needs of people and isn’t really looking for ways to help.”6 Empirically, then, poor voters should be more likely to perceive politicians who forbear as more broadly representative of their interests than those who enforce. In other words, forbearance serves as an affinity signal that demonstrates a politician’s distributive commitments. This heuristic can translate into electoral outcomes. Voters must decide on candidates based on observable 6 Author interview with Erasmo Segundo Cardenas Obregon, Local Councilor, Municiaplity of Ate, Lima, Peru, November 23, 2011. 11 characteristics, but they often lack informational cues like those provided by party labels in advanced democracies or ethnicity in some developing democracies (e.g. Bates 1982; Chandra 2004; Posner 2005). Enforcement can send clear signals about a politician, given that it requires the use of coercive force, and for incumbents who control police actions. Forbearance is a subtler cue, but it can be conveyed through public statements and alliances with groups in violation of the law, such as street vendors and squatters. It follows: Hypothesis 2. Poor voters perceive politicians who forbear (enforce) as more (less) broadly representative of their interests, and thus are more likely to vote for candidate who propose forbearance. A natural counterpoint to this argument is that the poor desire formal versions of social goods, rather than run legal risks and position themselves as criminals to secure basic goods. So, why would citizens hold politicians accountable for forbearance, rather than the welfare outcomes? The lynchpin is that forbearance motivates the poor’s electoral behavior because it is a credible and targeted welfare policy, not because it is necessarily the preferred one. Talk of social inclusion and welfare state programs is cheap. But, informal sector voters do not necessarily expect to receive the benefits from these policies. Khemani (2007: 59), for instance, has described a vicious cycle of “low performance and low expectations” for programmatic social policies (also see, Soifer 2013). In Latin America, low expectations of welfare programs come from a history of social programs that have been “truncated” to formal sector workers. Many types of social expenditures continue to have low coverage rates or do little to redistribute income to the poor (Levy and Schady 2013). This logic also is important in the case of politicians at the subnational level, who rarely can change the design of welfare programs and maintain limited budgets. Instead, many of their powers come at the enforcement stage. In contexts where formal social welfare programs often have failed to reach the poor, I expect: 12 Hypothesis 3: Enforcement platforms more clearly communicate a politician’s distributive commitments than promises of social expenditures that achieve functionally similar ends.7 Class-based differences in preferences and affinity signals suggest that politicians will face different incentives to enforce the law, even within the same country or city, depending on the demographics of the district where they run for office. Due to residential segregation, some electoral districts overwhelmingly concentrate poor voters. Others group together nonpoor residents. Politicians in majority-poor districts can reject enforcement in their campaign promises to woo voters. Enforcement carries heavy electoral costs. In nonpoor districts, politicians should pledge to enforce or avoid the issue altogether. Hypothesis 4: Politicians run on platforms that favor forbearance in districts with higher concentrations of poor voters, and embrace enforcement in those with lower concentrations. I test these hypotheses through two types of data. First, few existing surveys probe legal attitudes beyond general perceptions of the “rule of law.” I therefore fielded an original face-to-face survey of 900 voters in the capital city of Bogotá, Colombia to examine individual-level variation in support for enforcement against different types of legal violations.8 An innovative aspect of this analysis is that I rely on a battery of questions about both offenses and sanctions. I also include an embedded survey experiment to understand how voters perceive candidates with varying enforcement and social policy promises. Second, I look at variation in promises to enforce across electoral districts within Lima, Peru. These components purposely were conducted in two different capital cities to exploit differences in their electoral structures: Bogotá is a politically centralized city, meaning that residents elect a single mayor for the entire city. The survey experiment therefore invokes a hypothetical candidate reacting to the same set of citywide conditions. Lima is a politically 7 A corollary to this hypothesis, although beyond the scope of this paper, is that forbearance will be a less effective cue of a politician’s distributive commitments where social policies become more generous and targeted at the poor, or local politicians have greater powers over subnational social policies. 8 Details of the survey methodology are included in the Appendix. 13 decentralized city, meaning that residents elect a mayor for their sub-city district. Candidate platforms therefore should vary depending on local voter demographics. Both cities are situated in countries with weak welfare states where I expect promises of forbearance against street vendors and squatters to be salient electoral issues. These two sets of analyses produce highly consistent results, strengthening my conclusions. I concentrate here on understanding preferences and political messages about enforcement. The goal is to underscore how politicians at times face incentives not to enforce the law, separate from the budgetary and institutional resources at their disposal. Of course, this theory also has implications for actual enforcement actions. I show in separate research (Holland 2014) that my model explains differences in how much and where enforcement occurs. Class and Enforcement Preferences The most basic assumption of redistributive politics is that the poor prefer social welfare spending and the rich resist it. It follows that income predicts support for redistribution (Meltzer and Richard 1981; Romer 1975). If forbearance serves as an informal mode of redistribution, then enforcement preferences should polarize along income lines in a similar manner. Measuring enforcement preferences is tricky because individuals can hold separate attitudes toward legal violations (compliance) and state sanctions (enforcement). Both attitudes likely affect the extent to which parchment laws shape behavior. Beliefs about compliance can be governed by political and social norms that make certain forms of legal violations socially acceptable. For example, taking the iconic example from Les Misérables, the theft of a loaf of bread by a starving man like Jean Valjean may be an accepted action in the eyes of most citizens. Others may disagree with the theft, but hold separate beliefs about the sanctions that states use to control offenses. For example, even though many individuals disapprove of stealing bread, still many more may find imprisonment to be abhorrent. Whether questions are framed in terms of support for law breaking 14 or law enforcement likely affects the level of popular support, but not necessarily the individual-level determinants of legal attitudes. To operationalize preferences, I use two measures of attitudes toward progressive legal violations: approval of unlicensed street vending (Vending) and approval of squatting (Squatting). In addition, I consider a pair of measures of enforcement attitudes based on the most common type of sanctions for these offenses, decommission of vending merchandise (Decommissions) and the eviction of squatters (Evictions). I also ask questions about two legal violations done by the nonpoor, using business tax loopholes (Tax Loopholes) and paying less than registered on industrial electricity meters (Electricity). These offenses were selected because they are common administrative infractions that vary in their distributive effects. Each question was measured on a 4-point scale running from “strongly disagree” to “strongly agree.” All variables were rescaled from 0 to 1 for ease of interpretation. My key independent variable is an individual’s socioeconomic status. Colombia divides the population into six class strata that are established through a detailed household survey that accounts for current income levels, as well as the precariousness of that income level (due to education levels, type of work, household size, dwelling quality, and so on). Individuals are highly knowledgeable of their household’s stratum, and it offers a more precise measurement of material interests that selfreported income due to issues of nonresponse and volatile income flows.9 I therefore use socioeconomic stratifications to capture material interests (Class). I also control for political identification on a five-point scale that runs from the political Left to Right (Right). Greater attachment to law and order, and weaker support for redistribute policies, often are taken as core tenets of conservative ideology. Individuals who identify with the political Right may be less 9 The correlation between the strata identified on government documents and the self-identified household strata is 0.97. The results are robust to using self-reported household income as a covariate. 15 supportive of forbearance, although I expect these effects to be weak given that many respondents struggle to answer political ideology questions or converge on centrist answers. In addition, I expect that individuals who have worked as a street vendor (Vendor Past) or have occupied land illegally (Squatter Past) are more supportive of forbearance, given that they realize the distributive benefits at stake. They also may be more likely to understand forbearance as an insurance mechanism in the case that they need to violate the law again in the future. I also include a set of control variables that plausibly affect preferences: age, employment status, and gender. Figure 2 presents a plot of the estimated coefficients and 95% confidence intervals from an ordinary least squares regression using the individual survey items. Each coefficient corresponds to a move from the lowest to highest value (class, political ideology, and past legal violation), controlling for basic sociodemographic characteristics (not shown). A single ordinal category increase in approval for legal violations, say from disagreement to agreement, corresponds to a value of 0.25. Looking at the class variable for progressive legal violations, the fact that none of the confidence intervals cross the vertical line at zero underscores that they are all significant in the direction expected. The sign on class for law breaking is negative—the wealthy are less likely to approve of squatting and street vending—and the coefficients for enforcement are positive—the wealthy are more likely to favor decommissions and evictions. These effects are substantively as well as statistically important: nonpoor respondents are roughly an ordinal category less supportive of legal violations and more supportive of enforcement than poor respondents, even controlling for direct experiences violating the law. The fact that class predicts attitudes for both street vending and squatting also is suggestive that indirect material benefits, such as the availability of low-priced goods from street vendors, do not drive the results. While street vendors produce auxiliary material benefits for poor consumers, squatting often imposes costs that come from the use of public services. The effect of class is both larger and more consistent than ideology and direct material 16 benefits from legal violations. Political ideology does not seem to shape responses, while past violations have the expected sign (although falling short of statistical significance for attitudes toward evictions). Reassuringly, poor respondents are no less tolerant, and equally willing to punish, law breaking by the wealthy. It may seem surprising that the rich are not more supportive than the poor of the use of tax loopholes and industrial electricity theft, given that the benefits accrue to the wealthy. In part, the weak coefficient on class comes from the near universal condemnation of these legal violations: just 2 percent of survey respondents approve of the use of tax loopholes and electricity theft. Social desirability bias may dampen the rich’s support even among the subset of respondents who receive distributive benefits. But even if the low approval rates stem from an unwillingness to answer survey questions truthfully, they strengthen my argument that regressive legal violations cannot be justified publicly. In contrast, legal violations by the poor have publicly acceptable distributive effects that divide the public. Class Right Past Violation -0.4 -0.3 -0.2 Vending Squatting Decommission Eviction Electricity Taxes -0.1 0.0 0.1 0.2 0.3 0.4 Mean and 95% Confidence Interval by Survey Question FIGURE 2. Correlates of Enforcement Preferences 17 The measured level of popular support for forbearance against legal violations by the poor is dependent on the question framing. The fraction of voters that support legal violations (i.e. approves or strongly approves of squatting or street vending) is relatively low: 30 percent approves of unlicensed street vending and 13 percent of squatting. Popular backing for forbearance comes in reactions toward sanctions: 86 percent oppose decommission of merchandise from street vendors, and 63 percent oppose the eviction of squatters.10 This gap helps make sense of why politicians engage in forbearance, even when they have the formal powers to change the law. Most respondents want others to comply with the law, but they do not support the most common sanctions deployed to make property laws effective. Affinity Signals and Mayoral Candidates Even if citizens understand the distributive functions of forbearance, it does not necessarily follow that they hold politicians responsible or change their voting behavior in response to enforcement choices. I now explore my second and third hypotheses: that politicians who propose forbearance are identified as sympathetic to the poor’s broader distributive interests, and that, given the more credible nature of the benefits, forbearance provides a clearer cue than traditional social policy platforms. The Bogotá survey embedded a between-subjects experiment that allows me to analyze the microlevel implications of my electoral model: enforcement information signals a politician’s distributive commitments, and does so more effectively than formal social policy platforms. Each survey respondent heard two short statements about a hypothetical candidate considering a run for city mayor. The combination and order in which they heard the candidates varied at random. All of 10 A second issue in the measurement of enforcement preferences is whether they are framed in terms of vignettes about individuals who violate the law or face sanctions, or in terms of abstract principles. As expected, abstract principles attract lower levels of support. For instance, I asked a question about whether an individual supported the eviction of a single family that had occupied land illegally and one that asked generally about support for the eviction of squatters. The level of opposed to the eviction dropped from 63 to 49 percent. 18 the statements were loosely based on actual candidate proposals about enforcement against unlicensed street vendors from past mayoral elections in Bogotá, and subsequently verified by country experts. I focused on street vendors in the survey because they create common problems across urban space. Squatters tend to build their homes in peripheral districts of a city and therefore not all residents are familiar with the legal violations. Individuals evaluated a candidate with a plan for either forbearance or enforcement against unlicensed street vendors. As with real politicians, the forbearance vignette says that a politician opposes enforcement and will privilege workers’ rights, while the enforcement vignette involves promises of actions against vendors. The vignettes read as follows: Forbearance: As a City Councilor, the candidate criticized the abuses of previous mayors against unlicensed street vendors. If he were mayor, he would respect the rights of workers in the informal economy until the government provides other job alternatives. Enforcement: As a City Councilor, the candidate criticized the permissive attitude of previous mayors against unlicensed street vendors. If he were mayor, he would work for greater order in the streets and clear out unlicensed vendors. As a comparison point, individuals also heard one vignette about a candidate with a plan for pro-poor job creation (Pro-Redistribution) or pro-business investment (Pro-Business). My goal was to simulate a classic Left-Right division on employment issues as a comparison point for the cleavages created by enforcement platforms, rather than use a skeletal “control” vignette with no information. The platforms read as follows: Pro-Redistribution: As a City Councilor, the candidate criticized the lack of projects for vulnerable sectors of the population. If he were mayor, he would promote social inclusion, and specifically, invest in government employment programs to help unskilled workers find jobs. Pro-Business: As a City Councilor, the candidate criticized the lack of projects to generate private sector investment. If he were mayor, he would promote economic development, and specifically, cut government bureaucracy to help businesses grow and generate jobs. 19 Respondents were randomly assigned to each candidate vignette, and standard randomization checks find no evidence that random assignment was compromised.11 After hearing each candidate’s proposal, respondents assessed the candidates separately.12 They evaluated the candidates on a four-point scale running from “highly unlikely” to “highly likely” on three items: (1) How likely is it that the candidate promotes the interests of the poor? (2) How likely is it that the candidate promotes the interests of people like you? and (3) How likely is it that you would vote for this candidate for mayor? The first pair of questions was intended to tap into representation and expectations of benefits, whereas the last analyzed vote intention. Finally, respondents placed the candidate on a five-point ideological scale ranging from “Left” to “Right.” Figure 3 shows the perceived likelihood that the candidate will promote the interests of the poor, as well as the ideological placement. Given that there is no single “control” condition, I plot the raw means and 95 percent confidence intervals for each experimental condition.13 The vertical line at 0.5 indicates the point of indifference where respondents are equally likely or unlikely to identify a candidate as representing the poor’s interests and the political Left. The results strongly confirm my hypothesis that enforcement/forbearance messages send clearer anti/pro-poor cues than traditional social policy platforms. Provided with the enforcement message, 78 percent of respondents believed that the candidate is unlikely or very unlikely to favor the poor’s interests. Just 11 Table A1 in the Appendix reports the average class, education, age, gender, political ideology, and labor history in each condition. Significance tests for differences of means using standard t-tests suggest no significant differences (at 95% confidence) between conditions on these characteristics. An omnibus F-test also found no imbalance. 12 A direct comparison of candidates might better approximate the theory that voters favor candidates who forbear over those who enforce or run on formal social policy platforms. Within the context of this survey, though, matched comparisons might have created a consistency bias. If respondents were to see both candidate pairs, they may attempt to make their answers consistent by preferring the perceived left- (or right-) leaning candidates in both cases. 13 I also follow the convention of using difference-of-means tests to measure average treatment effects in the Appendix, and corroborate my results using regression analysis in which the experimental treatment condition is included as an independent variable with and without covariates, namely gender, age, education, and past history as a street vendor. 20 33 percent of respondents thought the same about the forbearance candidate. The difference between these conditions was statistically significant (the a raw scores on a 1-point scale were 0.75 and 0.42, p < 0.001), suggesting that the information on a candidate’s enforcement positions efficiently communicated a politician’s affinities. In contrast, the traditional social policy candidates did little to signal pro-poor commitments: 57 percent of respondents viewed the pro-redistribution candidate as unlikely or very unlikely to favor the poor’s interests compared to 59 percent for the pro-business platform (raw scores of 0.55 and 0.56, respectively). In sum, learning that a candidate favored forbearance most efficiently signals a politician’s affinity with poor voters, even relative to a pro-redistribution message. Politicians not only avoid an electoral cost and reputation as an “antipoor” candidate that comes from enforcement, they also can gain a reputation for representing the poor’s interests. Benefits the Poor (Likely to Unlikely), N=890 Ideological Placement (Left to Right), N=787 Pro-Redistribution Pro-Business Forbearance Enforcement 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Mean and 95% Confidence Interval by Candidate Statement FIGURE 3. Perceptions of Distributive Benefits and Ideological Placement by Candidate Statement The findings are similar, although weaker, for the ideological placement of candidates. Both the pro-redistribution and pro-forbearance candidates are perceived as representing the center-Left, 21 and there is no distinguishable difference between them. The enforcement candidate is identified as significantly farther to the political right than the other candidates. The inability to place the candidates ideologically, other than the enforcement candidate, may be explained by the declining salience of Left-Right distinction in Colombian politics. Nonresponse on political ideology questions is a major problem particularly for low-income respondents (22 percent nonresponse, compared to 5 percent for the upper-income brackets). The key point is that, despite muddled opinions about a candidate’s ideology, the idea of whether a candidate “benefits the poor”—which often is used as the defining feature of the political Left (Levitsky and Roberts 2011)—evokes clear opinions across the income spectrum. It also leads voters to identify candidates with forbearance promises as supportive of the poor. Consistent with my theoretical expectations, poor and nonpoor voters respond differently when asked whether a candidate supports “people like you” and whether they would vote for each candidate.14 Colombia divides the population into six class strata that I rescale from 0 (the lowest) to 1 (the highest).15 To see how responses differ by class group, Figure 4 reports the mean response and standard deviation by class strata (shown with local polynomial smoothing and simulated 95% confidence intervals). The upper left panel shows that poor respondents believe that the forbearance candidate is more likely to benefit them by a wide margin compared to the enforcement candidate. Among the bottom strata of the income distribution, 60 percent of the poor believe that the forbearance candidate is likely to benefit them, compared to just 25 percent who hold that belief 14 The phrase “people like you” is left deliberately vague so individuals can interpret their own class position or identity. 15 Strata are established through a detailed household survey that accounts for current income levels, as well as the precariousness of that income level (due to education, type of work, household size, dwelling quality, and so on). Individuals are highly knowledgeable of their household’s stratum. Strata measures thus come far closer to capturing a broad notion of class position than straightforward income or consumption measures. Consistent with Colombian definitions, I consider Strata 1 and 2 to be poor or lower class, Strata 3 to be lower-middle class, and Strata 4, 5, and 6 to be middle and upper class. Results using income or occupational status to measure class look substantively similar. 22 about the enforcement candidate. The lower left panel shows similar in terms hypothetical electoral behavior: 57 percent of voters in the bottom quintile say they would vote for the forbearance candidate, compared to just 28 percent for the enforcement candidate. These are sizable proportions of the population, particularly for a competitive urban electoral context where margins of victory can be small. The comparison of the right and left side of Figure 4 confirms my third hypothesis that forbearance and enforcement platforms more effectively cue a politician’s distributive commitments than traditional social policy platforms. While the poor do distinguish the pro-redistribution candidate as more likely to represent their interests and win their vote than the pro-business candidate, the difference is not statistically significant (right panels). This is consistent with my model that social policy promises are less credible, and often less economically progressive, and therefore do not convince the poor that politicians will serve their interests. A surprising aspect of the survey results is that there are no significant differences in the likelihood that upper-income respondents view the candidates as serving their interests or in their vote intention. One possible explanation for this non-finding is that the nonpoor may not view street vending as an issue and thus disregard information about a candidates’ enforcement position. However, nonpoor respondents are likely to care about the structure of business and redistributive policies, and therefore should have responded to the traditional social policy prompts. Another possibility is that nonpoor respondents tend to know more about politics, which makes their preferences harder to alter through skeletal candidate information in the context of a low-stakes survey experiment. Unfortunately the survey did not include questions to test a respondent’s knowledge or sophistication about politics so it is hard to disentangle whether more engaged respondents were less swayed by the information provided. 23 Likelihood to Receive Benefits (unlikely to likely) .2 .3 .4 .5 .6 .7 .8 Likelihood to Receive Benefits (unlikely to likely) .2 .3 .4 .5 .6 .7 .8 0 .2 .4 .6 Class (low to high) 1 0 .2 Enforcement .4 .6 Class (low to high) Pro-Redistribution .8 1 Pro-Business Likelihood to Vote for Candidate (unlikely to likely) .2 .3 .4 .5 .6 .7 .8 Likelihood to Vote for Candidate (unlikely to likely) .2 .3 .4 .5 .6 .7 .8 Forbearance .8 0 .2 .4 .6 Class (low to high) Forbearance .8 1 0 Enforcement .2 .4 .6 Class (low to high) Pro-Redistribution .8 1 Pro-Business FIGURE 4. Likelihood that Candidate Will Benefit Respondent’s Group (Upper) and that Respondent Will Vote for Candidate (Lower) by Class and Statement Finally, my model suggests that individuals with stronger material interests in forbearance should be the most responsive to informal welfare appeals, but that their responses should not entirely drive the effects. The survey included a post-treatment item on whether a respondent works or has worked as street vendor. As we saw above, a past history of working as a street vendor or squatter was associated with more support for legal violations. Likewise, I expect that individuals who once derived income from street vending have a direct stake in enforcement and respond more dramatically to the candidate information. This is not a trivial fraction of the sample: among the lower and lower-middle class groups, 28 percent of respondents once have worked as street vendors.16 To test whether these respondents are more sympathetic to candidates who promote 16 Lower class includes Strata 1 and 2, and the lower-middle class is Strata 3. 1 24 forbearance—and check whether income remains important once accounting for direct benefits from forbearance—I estimate conditional average treatment effects using past history as a street vendor as a conditioning variable. A straightforward approach is to estimate causal effects separately for the different subgroups of respondents and analyze the variance. Figure 5 compares perceptions that a candidate benefits “people like them” and vote intention for lower and lower-middle class groups who have worked as vendors and those who have not. I report difference-of-proportions in the respondents that say they find it likely or highly likely that the forbearance candidate will favor their interests and that they will vote for that candidate compared to the enforcement candidate. The differences between the conditions are statistically significant for both subgroups, suggesting that a forbearance platform has a positive effect on the poor’s perception of a candidate, regardless of whether a respondent directly benefits from forbearance. The results are not consistent with a pure special interest group story in which only street vendors favor lax enforcement. Direct experience as a street vendor conditions the average treatment effect in the predicted direction. Reassuringly, previous work as a street vendor has no effect on responsiveness to traditional social policy appeals (not shown). Nonetheless, the small sample size means that these results are merely suggestive, and do not reach conventional levels of statistical significance. 25 Likely or Highly Likely to Benefit People Like Me Likely or Highly Likely to Vote for Candidate No Vendor Past, N = 422 Vendor Past, N=162 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 Difference in Proportion of Respondents from Enforcement to Forbearance Condition FIGURE 5. Average Treatment Effects, Conditioned by Dependence on Income from Street Vending In sum, the experimental results largely support the idea that enforcement and forbearance are signals of the distributive commitments of candidates with the potential to shape electoral behavior among low-income voters, and particularly those who depend on informal sources of income. The survey experiment allowed me to clear away some of the complexity of elections by presenting respondents with only one type of information, and thus identify that enforcement information can have the hypothesized effects. The main drawback, as with all experimental work, lies in the external validity of these claims. In the natural world, voters receive a wide range of information about candidates and make their choices based on a variety of policy proposals. Studies of actual campaigns clarify when and how enforcement information actually serves this role in political life. District Demographics and Campaign Platforms My electoral theory makes predictions about precisely the type of individuals who should receive campaign messages that revolve around forbearance in the first place: politicians who seek office in electoral districts where the median voter is poor should be most likely to promote forbearance. To examine my final hypothesis, and observe how forbearance operates in real world 26 elections, I gathered a dataset on campaign platforms from the 2010 district election in the Lima metropolitan area. Peru requires political parties to publish their platforms for every local race they enter.17 Peru has a “collapsed” party system (Crabtree 2010; Levitsky and Cameron 2003; Seawright 2012) so these documents reflect localized campaign issues, rather than coherent programmatic agendas.18 Most voters do not read a politician’s published campaign platform. But, voters may learn about them through motivated neighbors or groups that do pay attention to written statements. Leaders of street-vending associations often asked candidates to incorporate their policy demands into their campaign platforms. Platforms also are observable ways to understand the types of positions that candidates take and likely discuss in campaign appearances that reach a broader popular audience. Analyzing platforms also allows me to focus on a politician’s intention to enforce the law apart from their capacity to do so. Local politicians may lack the budget, police, and bureaucrats to make the law effective. But, were this the case I would still expect them to promise to control illegal activities and then confront capacity limitations. My theory of forbearance, in contrast, predicts that politicians should be open in their decisions not to use state power to enforce. Platforms were hand-coded based on the types of enforcement commitments they mentioned toward street vending and squatting.19 My key comparison is between platforms in districts with a majority of low-income residents—poor districts for short—and those with less than a majority of upper-income residents—rich districts. There are fourteen poor districts and nine rich 17 Platforms (Plan de Gobierno) are available for each district and candidate through the National Electoral Authority’s Governance Observatory (Jurado Nacional de Elecciones, Observatorio para la Gobernabilidad, http://www.infogob.com.pe). 18 The only party that consistently uses the same platforms in each district is the Christian People’s Party (Partido Popular Cristiano, PPC). PPC is one of the few traditional political parties that remain in existence. Even with the PPC, however, the platforms are tweaked in each district. 19 Automated text analysis is not appropriate given the relatively small size of this dataset. A team of research assistances hand-coded the platforms, and country experts subsequently verified the classifications. 27 districts in the sample, resulting in 306 platforms in total.20 To get a sense for the issue salience, I first identified whether each platform mentions informal activities. In poor districts, 57 and 80 percent of platforms discussed unlicensed street vendors and informal settlements, respectively. As expected, the frequency drops in rich districts: 30 percent of platforms talk about street vending and 55 percent discuss issues related to informal housing.21 Given that many platforms are thin on content and many enforcement positions are communicated informally, discussion in platforms probably constitutes a lower bound on the importance of these issues in local races. Street vending and squatting thus appear to be salient issues. How does the type of enforcement position vary by district? I first identified platforms with clear enforcement promises. These include platforms that mention 1) specific police actions (decommission of merchandise from vendors, relocation or removals of vendors, evictions of squatters, and so on), 2) commitments to improve enforcement capabilities (additional police or monitoring activities), or 3) promises to apply existing laws with respect to squatting and street vending. Typical examples in the last category include statements to “exercise urban control as the 20 Peru divides households into five socioeconomic groups from “A” (the wealthiest) to “E” (the poorest). I define a poor district as one in which the majority of residents are from Classes D and E. Rich districts have at least a quarter of residents from Classes A and B, and a majority from Classes C and above. I did not code thirteen mixed-income or lower-middle class districts, where the empirical predictions are less clear. 21 Notably, the character is discussion of squatting changes in rich district: many platforms discuss problems of buildings without licenses or property titles (or improper ones), or neighborhoods that began as informal settlements and thus have problems with respect to services or quality of construction. The construction of new informal settlements is a very rare event. For the time being, the “mentions squatting” category includes any discussion of regulatory issues related to informal construction (including the extension of services). Likewise with street vending, some district platforms discuss how the control of street vending has pushed economic activity into residential areas or household shops, and now the zoning is inappropriate. 28 law allows and mandates,” or “implement a regulatory plan that will allow for greater order in street vending.”22 Second, I analyzed whether platforms included statements in favor of forbearance. These fell into three broad categories: 1) calls to refrain from police action, 2) language that prioritizes the social needs of those who violate regulations, and 3) appeals that candidates will promote order by “crowding out” legal violations through formal solutions. A typical example that combines the first two types of language reads: “We will avoid abuses against street vendors in the district, improving their location and giving them the necessary support, because they are people who work day in and out, because they don’t have a steady salary like others have, and because they also need to be given the assistance appropriate to their situation and not repressed through evictions.”23 Other candidates stop short of saying that they will stop enforcement, but reiterate their comprehension of the reasons that the poor need to work as street vendors or build their houses in informal settlements. For instance, platforms that foreground social needs note that, “Residents should be allowed to undertake commercial activities that allow them to satisfy their basic consumption bundle,”24 or that “[squatters] should not be viewed as exiles but rather given even more assistance.”25 Lastly, forbearance can be conveyed through implicit appeals that candidates will “crowd out” legal violations through formal solutions. An illustration of this dynamic is a platform that decries that street vending has “grown in a disorderly fashion,” but proposes “the 22 Alianza para el Progreso, Plan de Gobierno Municipal, District of Villa María del Triunfo; Cambio Radical, Plan de Gobierno Municipal, District of Ate, Lima Peru. I exclude a third category of statements that decry violations of the law without offering concrete solutions. For instance, 36 percent of platforms in poor districts incorporate discussions of how “disorderly” urban development and failures to control squatter settlement growth have created serious social problems. Yet, frustration with the problem does not necessarily entail a willingness to take coercive actions to prevent it. 23 Agrupación Si Se Cumple, Plan de Gobierno Municipal, District of Ate, Lima, Peru, 2010. 24 Alianza para el Progreso, Plan de Gobierno Municipal, District of Ate, Lima, Peru, 2010. 25 Partido Aprista Peruano, Plan de Gobierno Municipal, District of Villa María Del Triunfo, Lima, Peru, 2010. 29 way to address this growth is to attract formal businesses to give work to street vendors.”26 Implicit is the idea that enforcement is not the way to confront the problem. In the case of squatting, another common platform proposal is to provide local legal recognition or work toward property titles for those who have occupied land illegally. In particular, mayors issue “urban authorizations” (prehabilitación urbana) in which squatters become eligible to receive a bundle of urban services, such as water and garbage collection, but do not gain property title to their land. I code these statements as forbearance promises because they offer those who take land legal security, and may encourage continued land takings. Many politicians consider promises of legal recognition to be akin to forbearance because of the incentive problems created. The quicker and easier it is to obtain property title (or recognition for public service purposes), the more attractive it is to take land and wait for the transfer of full rights.27 Of the 28 local politicians that I interviewed in Lima, every single one believed that presidential initiatives to offer property titles to squatters—which are often buttressed by local mayoral decisions to offer urban authorizations—have increased the pace of land invasions and made it harder to prevent their spread. The typical reasoning of mayors was that land invasions were acceptable because the state itself recognized them, or as one mayor put it, “If the president wanted me to stop land invasions, then he would stop legalizing them every three years.”28 As expected, the picture that emerges is that forbearance platforms are more common in poor districts. While 30 percent of platforms in poor districts mention forbearance toward street vendors, 11 percent do in nonpoor districts. The contrast is even clearer for squatters. Almost 40 percent of platforms mention forbearance and legal recognition for squatters, while 13 percent do in 26 Acción Poplar, Plan de Gobierno Municipal, District of Ate, Lima Peru. Recent empirical research, for example, finds that municipal titling programs in Brazil are associated with more land taking in surrounding municipalities as residents anticipate forbearance and future titling (Albertus, Brambor, and Ceneviva 2014). 28 Author interview with John Barrera, Mayor, District of Ancón, Lima, Peru, December 1, 2011. 27 30 nonpoor districts. The contrast in terms of enforcement platforms is less clear because they are so rare. Surprisingly, even in nonpoor districts, only a handful of politicians made enforcement promises. Across districts, less than 10 percent of politicians mentioned specific enforcement actions, and no distinct patterns can be discerned between poor and nonpoor districts. Figure 6 plots these differences across district type and legal offense. 0.45 Fraction of Platforms 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 Forbearance Vending Forbearance Squatting poor Enforcement Vending Enforcement Squatting nonpoor FIGURE 6. Campaign Platforms by District Type and Promise, Lima 2010. These observational data demonstrate that enforcement and forbearance platforms are central issues in local political campaigns. Particularly in poor districts where voters are most likely to benefit from and perceive candidates that promise forbearance as serving their interests, I find that candidates avoid enforcement promises and use forbearance cues. A logical question is whether politicians with forbearance appeals receive higher vote shares on average. This test, however, runs into serious endogeneity problems because politicians may propose to enforce when they are secure in office and most likely to win reelection. For this reason, the survey experiment provided greater leverage to isolate the effects of enforcement platforms on electoral outcomes. The fact that forbearance communicates pro-poor affinities and makes the poor more likely to support a candidate in an experimental context helps explain why politicians propose forbearance on the campaign trail. 31 Conclusion Much scholarship is pessimistic about the ability of the informal sector poor to understand and vote for their interests. This paper highlights that drawing such a conclusion assumes that the informal sector poor vote based on formal social policies. If instead we take a broader conception of the material interests of the poor, we arrive at different conclusions about the logic of elections, particularly at the subnational level. This paper does just that. It introduces the concept of forbearance and theorizes how it can provide important informal welfare benefits to low-income groups and cue a politician’s broader distributive commitments. As such, politicians tailor enforcement appeals to the electoral weight of the poor in the district where they seek office. I tested the electoral theory of forbearance using two data sources with unique benefits and limitations. Standard survey questions revealed surprisingly high levels of popular support for forbearance against progressive legal violations, particularly among poor voters. By manipulating respondents’ exposure to information, the survey experiment provided a clear test of my hypothesis that forbearance serves as a heuristic that poor voters use to identify politicians that favor their interests. At least in the artificial context of a survey experiment, poor voters say they are more likely to vote for candidates who promise forbearance and against those who promise enforcement. Classic social policy promises, in contrast, fail to convey substantial information about candidates’ affinities. Turning to a dataset on local platforms in Lima, I showed that forbearance forms an important part of electoral campaigns, particularly in poor districts where voters are sympathetic to its use. Politicians can be open in their rejection of enforcement and support for legalization of past violations. These data help establish the importance of forbearance as a concept that is distinct from weak enforcement: politicians’ incentives to enforce the law vary across electoral districts, and the types of constituencies that they seek to court. Hence, the contribution of this paper is to offer 32 a fresh perspective on the types of issues that motivate the informal sector poor’s electoral behavior and structure urban elections. The use of progressive forbearance to signal affinity for the poor raises questions about its relationship to populism. While the concept has been stretched and disputed, a minimalist definition of populism is the appeal to mass constituencies by personalistic leaders who challenge established elites and institutions (Roberts 2006: 127). On the one hand, forbearance is an appealing distributive tactic for populists. Core to the definition of forbearance is the idea that it is a distributive tool that circumvents established institutions and welfare bureaucracies. It can form part of an anti-state position in which political outsiders offer immediate benefits to the poor without reliance on state institutions. Forbearance also plays into an anti-elite discourse, and thus has been a common distributive and symbolic tactic for populists from the political Left and Right like Hugo Chávez and Alberto Fujimori, respectively. Politicians can promise to defend the poor regardless of what some law written by out-of-touch legislators might say. Chávez, for example, has attacked mayors who attempt to enforce against street vendors as “enemies of the informal economy”29 and Fujimori blocked the national police from participating in enforcement operations due to the social costs.30 Outside the region, Resnick (2012) documents how Zambian President Michael Sata appealed to urban squatters and street vendors by forgoing enforcement. However, while populists often embrace forbearance, it is incorrect to view forbearance as a form of populism in all but the loosest sense, such as a direct appeal to the poor or a tactic to 29 For example, see “Jorge Rodríguez se convirtió en el enemigo de la economía informal” Noticias 24 8 Nov 2012. 30 In May 1997, Lima city mayor Alberto Andrade had organized an operation to remove street vendors from the central market. The city’s police were to receive protection in the removal of vendors’ equipment from the National Police. But, allegedly, the National Police received an order to abandon the operation at the last minute because “the conditions weren’t right.” Andrade said that the operation failed due to “a political hand that harms the capital, that harms the city of Lima” and accused Fujimori of supporting the vendors. See, “Batalle campal entre polícias y comerciantes,” ATV, May 14, 1997, https://www.youtube.com/watch?v=8nrg4Ncg4iw; “Aplazan desalojo de ambulantes del Mercado Central,” El Comercio 1 May 1997. 33 provide immediate benefits by neglecting long-term costs. Such a thin definition provides little analytic leverage to understand the range of politicians and circumstances that favor forbearance. The plethora of local politicians that promised weak action against squatters and street vendors in Lima were hardly united in a particular anti-elitist agenda. Forbearance may be an attractive tool for populists given its anti-state nature and immediate provision, but its pro-poor benefits appeal to a broader variety of politicians confronted with welfare state weakness and seeking election among low-income constituents. Future studies of electoral politics could fruitfully look at the affinity signals that voters use in the absence of party labels or more generally in light of the declining salience of Left-Right ideological cleavages. While Left and Right have lost salience for many voters, and the poor in particular, the idea of politicians who “help the poor” or represent “people like me” still holds weight. I showed that almost every poor respondent could respond to the question of whether a politician would help the poor, but substantially fewer were able to position the candidate on a LeftRight political spectrum. Understanding what policies communicate group representation, how they differ across political systems and levels of politics, and when they are invoked rather than valence issues could help make sense of how claims of political representation are made in context with weak political parties. Affinity signals fall short of the programmatic ideological content provided by stable parties but that they do convey group representation and tangible material benefits in ways that scholarship on populism often overlooks. While I argued that forbearance is a powerful signal of a politician’s distributive commitments, I see it as a starting point for future theorizing about a broader set of cues that can communicate political representation. Lastly, a central contribution of this paper is to introduce the concept of forbearance and provide an empirical illustration of its operation. The conceptual framework could be extended to other laws with different distributive consequences and contexts in future comparative work. 34 Already, a number of studies have found evidence of electoral cycles of enforcement in cases when legal violations by the rich are at stake: electricity theft to irrigate the fields of wealthy Indian farmers peaks in election years (Golden and Min 2013); arson to clear land for purchase by the wealthy spikes prior to Greek elections (Skouras and Christodoulakis 2011); labor regulations go unenforced in election years in Argentina (Ronconi 2010); and environmental regulations to control pollution are enforced less strictly in election years and in cases where the mayor and governor are from the same party in Brazil (Ferraz 2007). The fact that enforcement follows political cycles in a variety of “weak” institutional contexts suggests scope for choice in how laws are enforced, or forbearance by political actors. However, scholars know little about the circumstances under which law enforcement is manipulated to serve the interests of the poor. Progressive forms of forbearance hold out a relatively unique potential to communicate electoral representation. While falling short on standard definitions of programmatic policy, forbearance holds a similar potential to structure political competition because it can be openly defended and grounded in beliefs about how to help the poor. Regressive violations, in contrast, are subject to near universal condemnation, and forbearance likely persists in these contexts precisely because voters cannot observe its use. Campaign platforms revealed statements about forbearance toward the poor as necessary to improve the short-term employment options for the poor until the state can guarantee their welfare. Other politicians countered with a view that enforcement served the public’s long-term interests. None of these proposals involved changes to the formal legal structure. But these were substantive philosophies about core social policy issues. By neglecting enforcement as a potential dimension of political competition, scholars miss out on a central issue that can differentiate candidates and structure urban elections. 35