A Typology of Forbearance

Enforcement (and its absence) has distributive consequences. More specifically, forbearance can be divided into subtypes based on its distributive effects and mode of distribution. I define the distributive incidence as in the economic analysis of taxes and transfers. Economically 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.

Just as with tax or social welfare programs, how to calculate distributive benefits can be tricky. Some laws are hard to categorize because the physical incidence (who violates the law and benefits directly from forbearance) differs from the economic incidence (who benefits indirectly once market prices have adjusted). I focus on the direct effects on those who violate the law because they tend to be most salient for enforcement politics and transparent to calculate. As Wilson (Reference Wilson1980) stresses, enforcement imposes concentrated costs on clearly identifiable individuals. It follows that politicians are sensitive to whom they are seen as sanctioning (or sparing), even when enforcement is intended to prevent other groups from absorbing ancillary benefits. This dynamic contrasts with the legislative arena, where politicians debate the overall desirability of regulation and thus consider the indirect benefits.Footnote ⁵ While including the indirect effects can reveal additional interest groups with a stake in enforcement, there also tends to be little agreement about which effects to calculate or how to judge their magnitudes, as I discuss in the Appendix.

There are two ways that a government impacts progressivity through its enforcement decisions. First, choices about which laws to enforce have distributive consequences. In this case, progressivity depends on how the marginal utility of an offense changes with income. Forbearance must be directed toward laws that the poor tend to violate to distribute resources progressively. The criminalization of homelessness clarifies the point. Although facially neutral, vagrancy laws have a discriminatory impact on the poor, given that few wealthy individuals sleep on the streets when they have a home at their disposal. Enforcement of laws that the wealthy tend to violate (white-collar crimes), 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, meaning the unauthorized occupation of vacant land or buildings. If a government does nothing to prevent squatting, then the wealthy can take valuable pieces of land to build large houses and absorb the greatest absolute benefit from government leniency. However, the poor may receive the maximum relative benefit given that they spend a greater share of their income on housing 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 upper-class squatters and ignores houses in poor neighborhoods, as I demonstrate in Bogotá, the result is solidly progressive. If only the very wealthy are allowed to violate land laws, as Sun (Reference Sun2015) finds in China, it is regressive. The incidence depends not on the type of law, but rather on how governments use their discretion in enforcement.Footnote ⁶

The second dimension distinguishes between modes of distribution. Forbearance can be provided in a contingent manner to individuals or groups as a reward for other political behaviors, such as campaign contributions, bribes, votes, or turnout. But, forbearance also can be extended in a transparent and unbiased way. I distinguish this mode as noncontingent in that forbearance is granted to individuals or groups regardless of whether they supported or opposed a politician.Footnote ⁷

These dimensions combine to generate four subtypes of forbearance shown in Figure 2. The left side corresponds to familiar regressive subtypes. When legal exemptions are provided to specific wealthy individuals, forbearance constitutes a standard form of corruption or cronyism, as shown in the upper left. It is the flipside of targeted enforcement against political enemies. The quote ascribed to Brazilian President Getúlio Vargas best captures the sentiment: “For my friends, everything; for my enemies, the law.” The lower-left corner corresponds to circumstances where forbearance extends to the rich as a class. The distinction from corruption is that the rich or powerful gain a general impunity that need not be negotiated by specific individuals. Gordon and Hafer (Reference Gordon, Hafer, Carpenter and Moss2014) show that some American politicians reduce enforcement of coal mining regulations industrywide to attract probusiness voters. In highly unequal contexts, such as Latin America, legal immunity for the wealthy can extend to most areas of law (Méndez, O'Donnell, and Pinheiro Reference Méndez, O'Donnell and Pinheiro1999). I refer to such enforcement exemptions for powerful groups as plutocratic forbearance.

FIGURE 2.

A Typology of Forbearance

The typology also yields two progressive types of forbearance. The upper right, which corresponds to a particularistic distribution of forbearance to favor the poor, is a clientelistic subtype. The poor's primary political assets are votes, not money. They can offer their political support in exchange for favorable legal treatment, and run the risk of enforcement if they switch loyalties. Classic accounts of urban machine politics, such as Chubb's (Reference Chubb1982) study of how the Italian Christian Democrats won support in Palermo, recognize that politicians manipulate enforcement to reward and punish voters. As Scott (Reference Scott1969, 1142) notes, “A large portion of individual demands, and even group demands, in developing nations reach the political system, not before laws are passed, but rather at the enforcement stage.” Last, forbearance can be extended to the poor in a generalized manner. Politicians may use notions of economic deservingness in their enforcement choices and hope to be rewarded at the polls, but they do not demand political behaviors in exchange. Zambian presidential candidate Michael Sata, for instance, pledged to stop enforcement against all street vendors and squatters to appeal to the urban poor in 2006. As Sata justified his position, “If you fail to provide for [the poor], don't bring punitive measures against them” (Resnick Reference Resnick2013, 78).Footnote ⁸ Forbearance functions as an informal welfare policy when directed at the poor as a class. I focus on this welfarist subtype below.

There are limitations to this distributive typology, especially those that come from the moral values of a society. It may make little sense to think in a purely economic way about, say, whether to kill bankers or sell organs, because these are construed as moral prohibitions whose relevance is independent of the economic position of the individuals involved. In contrast, many property laws and economic regulations primarily raise distributive concerns. Our moral judgments turn on the needs and alternatives available to those who violate the law. This division follows Rawls (Reference Rawls1999 [1971]), who separates some values, such as freedom from violence, as inviolable and therefore not subject to principles of distributive justice. Although the concept of forbearance applies to moral and symbolic regulations, this typology works best for laws not subject to such deontological constraints. Even in this distributive realm, citizens and politicians hold divergent views on whether laws represent binding goals to be defended over recalcitrant social conditions, analogous to moral prohibitions, or whether laws need to take into account the circumstances of those who violate them, closer to social policy. As I turn to next, many politicians take the latter view when they lack access to other distributive instruments.

Cross-Equilibrium Correlations

A necessary, but not sufficient, condition for forbearance is that government actions, rather than offense levels, drive enforcement variation. As in work on law and economics, I model enforcement as an equilibrium outcome determined by the interaction between offenses and sanctions. The goal is to characterize the related, but distinct, phenomenon of variation in sanctions due to government choices and due to the level of offenses, just as a supply and demand framework separates changes in production inputs from changes in consumer preferences in economic markets. This method on its own cannot establish why the political supply of enforcement varies, but it can rule out differences in societal compliance as the driving factor.

All studies of enforcement struggle with an issue of reverse causality between offenses and sanctions. This endogeneous relationship has led some scholars to interchange the concepts of compliance and enforcement, as well as their measures. For example, Levitsky and Murillo (Reference Levitsky and Murillo2009, 117, 129) define enforcement as “the degree to which parchment rules are complied with in practice,” and advocate operationalizing enforcement through measures of offenses or sanctions. But this approach to measurement provides limited leverage to understand whether citizens voluntarily comply with the law, reducing the need for sanctions, or whether politicians cannot or will not enforce the law.

Instead, sanctions and offenses are better thought of as functions of each other. On the one hand, the most basic assumption of crime control is that fewer individuals will violate the law as governments invest more in punishing offenses. We therefore can think of a demand curve (Compliance), which represents the level of offenses in a society as a function of government sanctions. This curve slopes downward because offenses decline with sanctions. On the other hand, the equilibrium is also determined by a government supply curve (Enforcement), which traces out the level of sanctions as a positive function of the level of offenses.Footnote ¹⁰ Even governments that are not keen to enforce will do more sanctions as the offense level increases because the negative externalities multiply, creating a rationale for additional sanctions, and more offenses are caught for a given resource investment.Footnote ¹¹ The slopes of the enforcement curves vary because a government that engages in forbearance responds less to changes in offenses than one eager to enforce.

Figure 3 visualizes the two ways in which limited sanctions can be observed, through differences in the level of offenses or in government behavior. Think about two political units, a and b, which are alike in most ways. Citizens in a may have strong social norms against a given legal violation so the entire compliance curve is shifted inward relative to district b, as in the left panel of Figure 3. So, for any level of sanctions that government a does, it faces fewer offenses. Another possibility is that a has a similar culture of compliance as b, but government a cannot or will not enforce. The entire supply curve is shifted outwards and flattened, as in the right panel of Figure 3. For any level of legal violations, district a does fewer sanctions than b. A comparison of sanctions alone gives us limited leverage to distinguish these two scenarios. Similarly, data only on offenses do not allow us to draw inferences about sanctions.

FIGURE 3.

Compliance-Demand (left) and Enforcement-Supply (right) Shifts

The key insight, which is often overlooked, is that information about the distribution of both offenses and sanctions can distinguish different causes of limited enforcement. Classic theory of identification of supply and demand curves in economics, such as Working (Reference Working1927), recognizes that supply-side shifts present different patterns of prices and quantities than those driven by consumer-side dynamics. In the same way, suppose the variation across units, such as political districts, is driven primarily by the demand (compliance) curve rather than the supply (enforcement) curve.Footnote ¹² In this case, we see each unit has its own compliance curve, represented by the shifted dotted lines in the left panel of Figure 3. When we look across units (each in their own equilibrium marked by a point), the units trace out an upward sloping relationship. This positive correlation corresponds to a “rational” enforcement model in the sense that governments engage in more sanctions when there are more offenses. As we will see below in Santiago, hard crime follows this pattern across districts. If, on the other hand, the enforcement choices of governments fluctuate more than compliance (corresponding formally to more variation in the supply curves across units), a different pattern emerges. A line drawn to fit a range of political units will slope downward, as shown in the right panel. A “political” enforcement model predicts that governments exert less effort against offenses, which in turn encourages more legal violations. As we will see in the case of street vending, mayors face different electoral pressures that lead some to oppose enforcement. Lax enforcement then encourages more residents to work as street vendors. Put another way, a positive correlation suggests that political units follow the same enforcement policy, with sanctions allocated in proportion to offenses (compliance-side variation), while a negative correlation indicates that units follow different enforcement policies (enforcement-side variation).

The fact that enforcement variation stems from the government supply of sanctions does not isolate whether politicians are unable or unwilling to enforce the law. One strategy is to control for observable differences in resources. Another, as we will see in Santiago, is to select political units that share enforcement institutions. But, given possible unobserved institutional variation, this equilibrium strategy is most effective when there are exogenous shifters in the motivations for enforcement, such as those used in work on distributive politics like electoral cycles across time or reelection incentives across districts. I now suggest how changes in enforcement patterns in response to such shifters can distinguish the reasons for government-side variation.

Resource and Electoral Elasticity

The main hurdle to show forbearance is evidence of political intent. Simply asking politicians why enforcement lags raises concerns about “cheap talk.” Politicians may blame a lack of resources when they use forbearance to extract votes or resources, or they may claim benevolent welfare motivations to mask institutional weakness. Due to this risk, I consider a strategy to infer intent from the elasticity of enforcement. Elasticity refers to how responsive one variable, in this case enforcement, is to a change in another, like budgets, enforcement costs, or competing social concerns. Building off the equilibrium framework, I examine which factors lead to small or large shifts in the enforcement supply. A theory of forbearance implies that resources produce minimal changes in enforcement, while electoral factors result in large disturbances.

To walk through this reasoning, start with a budget constraint, representing government resources that can be spent on a set of goods and services. A politician allocates resources between enforcement and all other social goods, as shown by the budget constraint B. An exogenous change may reduce the cost of enforcement. For example, a technological advance like satellite imagery has decreased the costs to detect and sanction illegal logging activity. Such a reduction moves the budget constraint to B′. If a politician wants to enforce the law, but previously was constrained by funds, the elasticity of response will be large. This responsiveness to changes in costs is represented visually by a steep indifference curve in which the politician is willing to trade substantial social goods for improvements in enforcement. In this case, the fall in the price of enforcement leads to an increase in the amount of enforcement, as shown by the move from a to a′ in the left panel of Figure 4. This scenario corresponds to theories of selective policy implementation where the limiting factor is resources to provide a range of social goods and enforcement.

FIGURE 4.

Elastic (left) and Inelastic (right) Responses to Changes in Enforcement Costs

Conversely, a politician who does not want to enforce has an almost horizontal indifference curve. Additional enforcement requires a substantial sacrifice in other social goods, as shown in the right panel of Figure 4. The same change in resources from B to B′ produces little increase in the level of enforcement (here a and a′ are equivalent). Forbearance thus can be distinguished by a relatively inelastic response to changes in the cost or budget available for enforcement. This point reinforces the empirical meaning of intent: a politician is unwilling to increase enforcement no matter how cheaply it can be done, or the extent to which other social problems command attention. It also makes clear that selective policy implementation and forbearance fall on a continuum: the less willing politicians are to dedicate resources to enforcement as they become available, the more their behavior looks like forbearance.

While forbearance implies inelastic responses to resource changes, it involves a high elasticity to changes in the political costs of enforcement. Forbearance can be distinguished by many of the same electoral patterns predicted for other distributive goods, such as political enforcement cycles. Observations of electoral patterns suggest institutional slack to change enforcement along with shifting political conditions, rather than binding resource constraints.

Elasticity measurements suggest the causes of enforcement variation. Still, finding valid instruments where enforcement resources or electoral incentives shift without other major changes in conditions can be tricky even with the best data, let alone often-patchy data available in developing countries. The elasticity approach also paints a stark, and sometimes unrealistic, dichotomy between institutional weakness and political incentives. The final empirical strategy takes another approach to the endogeneity problem suited to data-poor environments in which capacity constraints and political incentives intermingle.

Enforcement Process Tracing

Enforcement process tracing tackles the endogeneity between offenses and sanctions by looking at government behavior in response to a given set of legal violations. Enforcement entails a sequence of behaviors that runs from the detection of an offense to administrative action to the imposition of a sanction. Much like using a lagged variable, isolating reactions to a set of cases lets us set aside the issue of how past enforcement affects the offense level. The drawback is that we need strong assumptions about the enforcement process to identify forbearance.

Empirical studies of forbearance must have some notion of enforcement under institutional constraints as a counterfactual. Perfect enforcement is impossible. In any country, rich or poor, we expect some natural “loss” to occur at each step of an enforcement process. The challenge is that we do not know how hard each part of the process is, or more technically speaking, the functional form of the loss. A reasonable null hypothesis is one of constant attrition, or a log-linear function in which the same proportion of cases fails at each stage of the enforcement process. For count data, constant proportional differences are a more plausible and typical assumption than absolute differences. A government that arrests 1,000 people for 100,000 offenses is unlikely to convict 1,000 of 2,000 people charged with a crime. With a constant hazard rate, the same percentage would hold for both. A government that makes 1,000 of 100,000 arrests also prosecutes 20 of 2,000 people charged. Constant attrition seems a sensible baseline assumption, but a more precise approach might calibrate the loss function using data on other offenses or governments.

Compared to an alternative theory of institutional weakness, forbearance implies that enforcement procedures stop at political bottlenecks. Politicians should make decisions to halt enforcement, even when bureaucrats and police perform their jobs. A sharp deviation from a log-linear trend occurs at the points at which politicians influence the enforcement process. The empirical challenge is that bottlenecks also can occur because some procedural steps are harder than others. Additional causal process observations, such as evidence that politicians engage in direct actions to bring about the bottleneck, larger drops for politically sensitive cases, or information about the relative difficulty of each step, can strengthen a claim of forbearance.

In sum, none of these empirical approaches is a silver bullet. But they all offer falsifiable predictions that distinguish forbearance from often-confused theories of limited enforcement, such as societal compliance and selective policy implementation. To confirm all three observations, preferably through a combination of qualitative and quantitative data, in which various types of evidence corroborate one another, amounts to a gold standard in a challenging empirical context. I now apply these strategies as a way to make their use and mixture more concrete.

Street Vending in Santiago, Chile

On standard measures of state capacity, such as bureaucratic quality and tax collection, Chile excels (Bergman Reference Bergman2009, 4–5). High levels of institutional capacity make it unlikely to observe weak enforcement under capacity-based theories. Yet, an estimated 53,000 unlicensed street vendors work selling everything from feather dusters to used clothing in Santiago (Appendix). Why do these public violations of the law persist? I suggest that politicians rely on forbearance in low-income districts as a way to win votes and signal their commitments to the poor.

Features of Santiago's institutional design make it an ideal case to identify forbearance through an equilibrium analysis. On the one hand, Santiago is divided into dozens of districts (comunas), each of which elects a local mayor. Mayors therefore can select enforcement policies consistent with their electorate's preferences. On the other hand, a single national police force (Carabineros) controls enforcement of legal violations, including street vending. If most variation comes from mayors tailoring enforcement policy to their constituents’ needs, then I expect to observe a negative correlation between offenses and sanctions across districts. At least formally, there are no differences in institutions across space to generate this supply-side variation. This prediction contrasts with a null hypothesis that most variation comes from compliance rates, such as how profitable it is to work as a street vendor. Demand-side variation would lead to a positive correlation between offenses and sanctions.

Second, beyond a basic correlation, district demographics permit an elasticity analysis. My fieldwork indicated that the electoral costs of enforcement against street vending vary with district composition. On material grounds, the poor tend to favor less enforcement than the nonpoor because they are more likely to work as street vendors and purchase goods from vendors. While all districts run employment programs of varying quality (Pribble Reference Pribble2015), mayors stressed that they cannot find sufficient positions for job seekers and receive no national funding. Mayors tolerated unlicensed vendors as a way to assist constituents. As one mayor expressed, “The level of education in [this district] is fifth-grade so it's hard to find other stable jobs. Maybe I can find someone a job for a month, but it ends and they are back to the streets so you have to let them stay to work.”Footnote ¹³ Beyond direct employment benefits, mayors use forbearance to signal their affinity with poor voters. Politicians in low-income districts feared being seen as “anti-poor”Footnote ¹⁴ and sacrificing their “reputation for understanding social realities”Footnote ¹⁵ if they called in the police. Those in middle-income districts, in contrast, stressed that residents do not need to work as vendors and “applaud control.”Footnote ¹⁶ To draw more general conclusions, I surveyed bureaucrats in 34 districts and asked whether they believed that mayors lost or gained political support when they removed street vendors. The perceived electoral costs diverged: only bureaucrats in poor districts believed that mayors lost support (Appendix). As such, I expect enforcement to decrease with district poverty.

In most contexts, a prediction that low-income districts enforce less also is consistent with capacity-based explanations. While a common police force controls enforcement in Santiago, there still may be informal institutional differences if the police are less attentive to low-income areas of the city. As a type of placebo test for this competing explanation, I look at enforcement patterns for a valence issue that all citizens want to control, hard crime. I expect the level of offenses, not political choices, to be the main determinant of enforcement in Santiago.Footnote ¹⁷ The relationship between sanctions and offenses should be positive if the police are equally responsive across districts, and there should be no differences by district type. If the police ignore poor districts, then a negative relationship between offenses and sanctions will emerge, and poverty also will correlate with enforcement against hard crime.

Figure 5 presents the results of a comparison of enforcement patterns across district types and offenses. I focus on the patterns in the raw data here to convey the strategy to identify forbearance, and use regression analysis to show that these results are robust to a number of controls and replicate across cities in separate work (Holland Reference Holland2015). The left panel plots the number of unlicensed vendors (offenses) and police operations (sanctions) in each district in Santiago. Consistent with my electoral interpretation, low-income districts, marked by filled circles, both have more street vendors and do less enforcement. Middle-income districts, represented by open triangles, generally have fewer street vendors and do more police operations. The correlation between vendors and police operations thus is negative across districts, suggesting that most variation indeed comes from differences in the government supply of enforcement. To see this more clearly, I estimate a simultaneous equations model that quantifies the theoretical intuition presented in Figure 3 (see Appendix for details). In an average low-income district, police operations barely increase with the number of vendors. As the left panel of Figure 5 shows, the estimated supply curve is nearly flat. In contrast, in a typical middle-income district police operations are predicted to increase substantially with a change in the number of vendors. The fact that the supply curves account for much of the variation can be seen by the fact that these structural supply curves fit quite closely through the cross-sectional relationship between vendors and operations in the two types of districts. Demand curves are suppressed for ease of presentation and account for the residual variation along the supply curve. However, overall the supply curves fluctuate more and thus account for the negative correlation between vendors and police operations.

FIGURE 5.

Patterns of Sanctions and Offenses, Street Vending (left) and Crime (right) in Santiago

A clear contrast emerges between street vending and hard crime that all citizens want to control. The right panel of Figure 5 shows that the relationship between reported crimes (offenses) and police captures (sanctions) is positive. Crucially, minimal differences exist between low- and middle-income districts. The very similar estimated supply curves suggest that most variation across districts must come from variation in the demand to commit crimes. They show that police responsiveness does not differ across these district types for a less politically controlled form of offenses. Thus, the spatial patterns are consistent with my theory that variation in enforcement against street vending stems primarily from the choices of local mayors to respond to divergent constituency pressures.

A loose version of enforcement process tracing confirms that mayors intervene in poor districts to produce these results. While the police can initiate operations, they rarely do so. As one mayor described, “The police don't obey the mayor by law, but they also don't enforce unless I ask them to. . .[because] all the fallout of operations hits the mayor.”Footnote ¹⁸ If enforcement flags in poor districts because the police are less responsive to these areas, I should observe that mayors demand police presence and wait for an institutional response. This is not the case. Rather, in 13 of 19 low-income districts, mayors never called the police. In several districts that conducted operations, I also found evidence that enforcement dropped off in election years. For instance, operations went from three per month to zero in the low-income district of La Florida prior to the 2012 election. The mayor allegedly transferred bureaucrats that attempted to enforce vending regulations to different offices and curbed their responsibilities. As one bureaucrat explained, “The mayor thinks that these are poor people who don't have jobs and he needs to help, but I was unwilling to mix up social criteria with the law so he decided to appoint his own person for the sensitive [election] period.”Footnote ¹⁹

The lack of enforcement is suggestive of the subtype and motivations for forbearance in Santiago. If mayors negotiated contingent exchanges, they would enforce selectively against particular vendors or groups that did not support them. Out of two-dozen street vendors that I interviewed, not a single one reported paying a bribe or fearing eviction if they voted against the mayor. Politicians described something closer to constituency service in which they used forbearance to expand the goods at their disposal, and further softened enforcement as elections neared to reinforce their reputation among less well-off voters. Nevertheless, few politicians proposed changing the law or legalizing vendors because they preferred the flexibility to exert control if economic or social policy conditions were to change. Additionally, some officials stressed that legal authorizations require local council approval, and business owners tended to oppose authorizations near their properties in public debate. Forbearance thus allowed politicians to boost the level of employment assistance covertly.

Taken together, these results are more consistent with a theory of forbearance than alternatives rooted in state capacity. Using a unique dataset on both vendors and control operations, I documented a telltale negative correlation between offenses and sanctions, despite common and capable institutions. I also found that enforcement mapped on to electoral pressures, which decline in poor districts for street vending but not for valence issues like hard crime. There also were clear political bottlenecks in which mayors refused to call in the police to control street vending. Electorally motivated enforcement choices, however, cannot be reduced to a contingent exchange for votes. Instead, qualitative evidence suggests that politicians used forbearance to expand the welfare goods they could provide and signal a commitment to poor voters.

Squatting in Bogotá and Lima

I now compare squatting in Bogotá and Lima to illustrate the problems with standard enforcement measures, and the identification challenges present in weak institutional environments. Evictions are the required, but infrequent, sanctions for squatting. Given their rarity, an equilibrium approach provides limited analytic leverage. Enforcement process tracing, as well as elasticity measures, are best suited to the complex administrative environment. I collected data on each step of the enforcement process over a five-year time period (2006–11). Because of the weakness in institutional procedures, there is much greater uncertainty about how cases move through the enforcement process in Lima. Nevertheless, in both cases, I find evidence that mayors engage in forbearance to win votes and expand housing options.

Enforcement process tracing begins by measuring the universe of offenses against which a government could have acted. An offense involves a single construction on land for which a household lacks building rights. In Bogotá, a plausible measure of offenses comes from the number of “provisional” water connections. Each household almost immediately applies to receive water, and the city's water company acquiesces to avoid service theft and comply with court mandates. According to the water company, there were about 23,000 new provisional connections in Bogotá. Lima does not automatically grant squatters access to water so the same measure cannot be used. However, the Peruvian household survey does ask respondents if they live in a house formed through a land invasion and if they have (or have applied for) a property title. In 2011, only squatter settlements formed prior to 2004 were eligible for property titles so most untitled invasions likely formed after that date. By this metric, there were almost 90,000 offenses in Lima.

Next, government officials must detect the offenses that occur. In 2003, Bogotá created an agency to monitor squatting through a combination of satellite images and teams of engineers. The agency registered almost 14,000 illegal land occupations, spotting around 60 percent of offenses. In contrast, calculating how many squatters are detected in Lima is complex because no single agency monitors them. By law, the Lima metropolitan government is supposed to report squatters to the state agency that owns the land. But the city follows land invasions through newspaper reports, and not a single inspector monitors invasions. Other levels of government do little more. In 2011, the national land agency had no monitoring procedures, or even a complete registration of its holdings. Likewise, only a handful of subcity governments (distritos) engage in proactive monitoring. Combined across institutions, I found that roughly 5,900 cases, or just 6 percent of squatters, came to the attention of authorities.

After detection, bureaucrats open an administrative case against illegal land occupations. In Bogotá, detection is the responsibility of the city government, but subcity districts (localidades) take responsibility for subsequent proceedings. District bureaucrats face obstacles unique to enforcement in a context without addresses. District officials rely on photographs and approximate locations to match constructions to the geographic coordinates provided by the city. They struggle to identify land occupations because physical appearances change rapidly. Between visits by officials, an isolated, tin shack can be expanded or enclosed by hundreds of similar ones. Once located, all legal correspondence must be hand-delivered, which means that the identification process needs to be repeated multiple times. Given these obstacles, roughly 20 percent of detected occupations result in administrative actions. Administrative procedures to identify legal claimants greatly slow enforcement in Lima too. Unlike in Bogotá, cases are filed against entire settlements, to which claimants are then added as the settlement expands and changes. The Attorney General filed 180 legal dockets involving an estimated 4,500 claimants in Lima during the period studied.

Courts must approve a sanction after an administrative process is open. In Bogotá, the court process allows those affected by enforcement to file constitutional appeals (tutelas), which can delay and stymie sanctions. Once the court authorizes an action, the case returns to the district government. About 39 percent of administrative actions make it through the judicial review to end in a demolition order. Unfortunately, determining the number of cases with eviction orders is impossible in Lima due to the fragmentation of court filing procedures.

The final step in the enforcement process—and the one most commonly measured—involves the imposition of a sanction. In both cities, politicians play a key role in this stage. In Bogotá, mayors must sign and schedule the removal of an illegal construction. But mayors largely refuse to sign or postpone evictions until cases pass their statute of limitations. Only 4 percent of court orders result in a sanction being imposed, which is the lowest rate of efficacy in the entire enforcement process. In Lima, mayors also try to stall police action, often by lobbying the state to purchase seized land. The data again are less clear, but roughly 5 percent of cases end with an eviction.

Figure 6 visually summarizes the “leakage” at each step of the enforcement process, ending in almost no sanctions. On the vertical axis, I log the number of cases so the change can be interpreted as the percent decline at each step. The straight dotted line represents the proposed counterfactual of constant attrition. What is most important is that a sharp deviation occurs from the log-linear trend when mayors have important roles in the enforcement process. In Bogotá, the bottleneck comes at the very end, when mayors decide whether to sign eviction orders. In Lima, a major drop-off occurs in the number of cases that enter the legal system in the first place, and then again when mayors coordinate evictions.

FIGURE 6.

Enforcement Process Tracing of Squatting in Bogotá (left) and Lima (right)

This exercise highlights how standard indicators of enforcement can lead to misleading conclusions. Think what happens when we measure enforcement through offenses, bureaucrats, or budgets. On all of these measures, we would infer that Bogotá engages in much more enforcement than Lima. Bogotá has a third of the number of squatters and an entire agency dedicated to monitoring. But Lima actually does more sanctions in absolute terms. Following each step leads us to the more accurate conclusion that both governments do very little enforcement, albeit with different institutional expressions.

Another empirical challenge is to determine if the observed bottlenecks represent political interventions, rather than just administratively challenging procedures. Several facts suggest the former. First, particularly in Bogotá, bureaucrats lamented that politicians blocked the enforcement process. They engaged in “Orwellian” jobs. Despite bureaucrats’ efforts to bring cases, eviction orders sit on mayors’ desks because signing them would create an “unfavorable image of the mayor with the community.”Footnote ²⁰ More generally, bureaucrats stressed the political, not technical, challenges of evictions. I surveyed 41 housing bureaucrats, asking whether resources or mayoral intervention posed a greater constraint on enforcement. In Bogotá, about two-thirds of bureaucrats saw political intervention as the primary obstacle. In Lima, bureaucrats divided. About half discussed resource constraints including the availability of cars, maps, and staff, and half complained about the lack of political support.

Second, even in the case of Lima, the incoherent detection procedures likely reflect political preferences for forbearance, not capacity constraints. Relatively simple reorganizations of the bureaucracy, such as the assignment of monitoring responsibilities to a single authority, could vastly improve enforcement capacity.Footnote ²¹ But officials, such as the director of city planning, said these reforms make little sense because, “We can't repress invasions until there are other housing options.”Footnote ²² The low rate of detection results from failures to pass information to enforcement authorities, not a lack of underlying knowledge. Squatters register their associations with the district, approach the police to verify the physical occupation of land, and ask local authorities for municipal services. Mayors use this information to report invasions to authorities when the cases involve valuable resources, such as land reserved for public goods, but they see little reason to act in most cases. This is not to say that mayors want to allow squatting. Rather, they see high electoral costs to evictions when the national government offers no housing alternative. One mayor captured a common view: “I'd love to stop invasions and put the land to more productive use, but I can't when there is no policy from the central government to offer housing to poor people. . .I'd only have a handful of votes in the next elections.”Footnote ²³

Analysis of the elasticity of enforcement further confirms that politicians intentionally oppose enforcement, and maintain latitude to act otherwise. In Bogotá, for instance, both the poor and rich have built illegal constructions in the environmental preserve that abuts the city. The typical view is that enforcement against the rich is harder, given their superior access to lawyers and political connections. Yet wealthy squatters garner little public sympathy. A “tough” test of my theory, then, is that mayors should enforce more against illegal constructions by the rich. Indeed, I find that none of the evictions conducted in Bogotá involved poor claimants, while several involved wealthy property owners. The political logic of targeting the upper class comes out in the comments of bureaucrats: “The city mayor doesn't want to disrupt things, where are you going to move all these poor people to?. . . It's different when wealthy people decide to build weekend chalets in the forest preserve. We took down 16 elegant homes last year because there is no reason that they should be there.”Footnote ²⁴ These actions are hard to reconcile with capacity-based theories, or a view that politicians only manipulate enforcement for pecuniary gains. Instead, Bogotá’s mayor, Enrique Peñalosa, perhaps best captures the paradox of states with institutional capacity to enforce their laws but not to address social demands: “The problem is that as the city becomes more sophisticated and richer, there are more controls and this is very bad because we can control informal construction but we are not offering any alternative.”Footnote ²⁵

In contrast to the sensitivity to electoral costs, enforcement behavior is unresponsive to changes in administrative costs. Consider how the Peruvian government has responded to the introduction of drones that take aerial photographs. For decades, local and national authorities have fought squatters on archeological sites with limited success. With the increasing availability of drones, the government is able to monitor encroachments. Evictions of squatters on archeological land have increased.Footnote ²⁶ This change in enforcement activity in response to a reduction in the cost of monitoring shows that the government prefers to enforce on archeological sites, but previously faced constraints. Yet, the same technology has not been used to control squatting in peripheral urban areas where the poor tend to occupy land for housing purposes. Interviews with local mayors confirmed this split approach in which “empty land” could be turned over to the poor for housing purposes, while archeological land demanded cooperation with national authorities. As one politician expressed, “How can I say that I support the poor and then go behind their back to try to get people who need housing taken off their land?”Footnote ²⁷ Hence, Peru's incomplete adoption of new technology that reduces the costs of enforcement suggests that administrative constraints are not the full story. Political calculations motivate enforcement choices.

To wrap up, this paired comparison of squatting demonstrates how to identify forbearance in weak institutional environments. Traditional measures of bureaucrats and budgets produce misleading conclusions because, on the one hand, institutional investments do not translate neatly into sanctions, and on the other, weak institutions can reflect endogenous (and mutable) political choices. Enforcement process tracing, combined with elasticity analysis, reveals that politicians intervene to block sanctions that carry electoral costs, even when other agencies do their jobs and technology improves.