Enforcement Process Tracing: Forbearance and Dilution in Urban Colombia and Turkey

Abstract

Cities are complex regulatory environments. Attempts to regulate urban behavior create opportunities for politicians to manipulate enforcement to win votes and reward supporters. While some politicians choose not to enforce regulations, or forbearance, others undercut their intent, or dilution. Empirical research on enforcement has lagged behind due to the identification challenges in distinguishing weak state capacity from political manipulations. We develop a structured approach to process tracing that follows enforcement decisions sequentially across bureaucracies and specifies statistical distributions as counterfactuals to identify the causes of limited enforcement. We illustrate these strategies through original data on enforcement against squatters in urban Colombia and the provision of building permits in urban Turkey. Enforcement process tracing helps to document a form of distributive politics that is common to cities in the developing world.

Appendix

The Difference in Difference Model

Conventionally, a difference-in-difference (DiD) strategy compares the pre- and post-treatment differences in the outcome of a treatment and a control group. In our model, municipalities in election years (in months) correspond to the treatment group, and the same municipalities in non-election years (for the same months) correspond to the control group. In estimating the effect of elections on construction permits, we define month right before the election as the “treatment” period and the month preceding the treatment period as the “pre-treatment” period. Local elections occur on the last Sunday in March so we define March as an election period and February as a non-election period. We then analyze how permitting behavior changes in this two-month period preceding the election, compared to the same period in non-election years.

Table 2 presents the regression results. Our dependent variable is the monthly number of construction permits issued in a given municipality. The first column shows the results for all urban municipalities. An additional four construction permits are issued in the period prior to an election (March) compared to the same period in non-election years. Given that the average number of permits issued in the same month in non-election years is four, this result suggests that permitting activity doubles in the month leading up to an election.

Table 2 Regression results

Model Specification

$$ {Y}_{ist}=\alpha +\upgamma {\mathrm{TreatmentGroup}}_{\mathrm{s}}+\delta {\mathrm{TimeDummy}}_{\mathrm{t}}+\lambda {\mathrm{MunicipalityDummy}}_{\mathrm{i}}+\beta {D}_{\mathrm{i}}{D}_{\mathrm{s}\mathrm{t}}+{\in}_{\mathrm{i}\mathrm{st}} $$

Where DiD_st = TreatmentGroup_s × Time_t is a dummy variable which equals one for treatment units in the treatment time (March) and is zero otherwise. TimeDummy_t denotes the time dummy that we use for each February and March since we pool multiple time periods in our model.

TreatmentGroup_t equals one for treatment units (those in election years) and is zero otherwise (those from the previous year). Finally, MunicipalityDummy_i, the municipal-level fixed effects, controls for the time invariant municipality specific characteristics.

Table 3 Summary Statistics: the Monthly Number of Residential Construction Permits Issued in Non-Election Years, by Type of Municipality

Table 4 Main results including rural municipalities

Table 5 Results with Party and Core/Swing Interaction Variables

Table 6 Results of the Poisson Regression Model