Tackling Drug-Lords in a Nascent Market: Raids and Drug Crime in Uruguay

Abstract

This article analyzes the effects of police raids for different types of crime in the most conflictive neighborhoods of Montevideo, Uruguay. Interrupted time-series and intervention models are estimated using different specifications of geographical area where the crackdowns occurred and also different control strategies to produce robust results. The effect of crackdowns on crime reporting is mixed; evidence suggesting crackdowns may produce short- and long-term effects on crime depending on their ability to affect gangs’ competition for the territory and the market. It appears that the effects of raids are sensitive to the context of the criminal situation. Crackdowns are not consistently effective in influencing crime. Evidence shows it is hard to reach levels of critical enforcement through 1-day crackdowns and that crackdowns’ ability to alter drug-market conditions would depend not only on the ability to extract drug dealers from the territory but also in preventing a rapid return.

Keywords

crime crackdown Uruguay drug police

Introduction

On August 1, 2019, Hamburg port authorities found 4.5 tons of cocaine hidden in a shipment of soybeans coming from the port of Montevideo. The news was shocking on the two sides of the Atlantic. This cocaine seizure set a record for Hamburg port authorities in terms of the amount of drug, even in the context of a steep upward trend in which 2017 seizures around Europe increased to 140.4 tons from a previous 70.9 tons in 2016.¹ This seizure accounted for the biggest drug shipment ever recorded in Montevideo. The long-lasting war on drugs in the north of Latin America has had the consequence of moving drug routes for Europe and the rest of the world to the south.

In this context, Uruguay has seen a steep increase in drug-related crimes and a continuous growth of drug dealing operations, which has proven to be particularly pervasive to poor slums at the outskirts of the city of Montevideo. The city is going through one of the worst periods of violence and crime in recent decades. In 2018, homicide and robbery figures were the highest on record. These figures also show worrisome signs of spatial inequality. While the country average homicide rate for 2018 is 11.8 per 100,000 inhabitants—similar to Costa Rica or the Dominican Republic—Montevideo’s homicide rate of 16.1 is closer to figures of the Democratic Republic of Congo or Lesotho.²

In Montevideo, violence and crime incidence rates change dramatically between neighborhoods that are just a few kilometers away.³ Violence is concentrated in the most disadvantaged social contexts, exacerbating living conditions and generating a vicious cycle of violence and marginality that conspires against any attempt at social development or even social order (see Averdijk et al., 2016; Jacottet, 2018). A 30% of all homicides committed in Uruguay in 2018 were concentrated in four neighborhoods in Montevideo, home to 9% of the country’s total population. Most of these homicides occur in a few poor neighborhoods in the north area of the city, among which the neighborhood of Casavalle ranks at the top, with a homicide rate that exceeds 70 homicides per 100,000 inhabitants.

The relation between local distributors and international smugglers, as between distributors and crime, has been widely documented (see Reuter, 1986, for a classic study). The drug market in Montevideo began a worrisome trend during the 1990s with the arrival of coca paste, which rapidly penetrated all social sectors. With it, a dealer market began an expansive stage due to increased consumption. However, with the economic effects of the commodity boom, coca paste began to be replaced in the upper social strata, which shrunk the market and increased violence between drug dealers due to market competition.

The governmental response to this steep increase in crimes has been varied both in terms of the level of geographical focus and the diversity of approaches, following common international practices. Three strategies are worth mentioning. First, the government implemented a strategy of Crime Prevention Through Environmental Design (CPTED) beginning in 2012, which involves a high geographical focus and wide array of approaches. For example, surveillance cameras were installed at different points in Montevideo, but mostly with a focus on the city downtown, characterized by frenetic business activity during office hours but lonely otherwise and with hot spots of middle- and lower-middle-class residential areas.⁴ By 2017, the government claimed (although no evaluation has been made public) this system had reduced thefts by 80% and robberies by 73%.⁵ Second, in the same vein, the government initiated a Program of High Operative Dedication (PADO, by its Spanish acronym), which consists of focused police patrolling in different neighborhoods.

Third, the policy we focus on in this article is that the government began another strategy of hot-spot policing in 2017: localized raids. These were conducted in areas with particularly high levels of drug-related crime and violence as well as a presence of organized micro-drug-trafficking gangs. This strategy is combined with PADO afterward in the most complex neighborhoods.

This article uses an interrupted time-series design to evaluate the effects of such raids on crime over time. We aim to learn about the effects of crackdowns on robberies and violent crimes (threats, damage, injuries, and homicides), the two types of crime both the literature and local police authorities relate to the drug-trafficking problem. We also consider the special case of Casavalle, unique in that rival gangs are known to be battling for the territory, crime is especially high, and eight major raids have been conducted in this neighborhood. We use the method of intervention analysis for modeling the effect of raids on crime in Casavalle.

Theory and Cases

Specialized literature on the effect of hot-spot drug policing can be divided into two broad groups: the literature modeling the theoretical relation between crackdowns and crime, and the literature based on case studies, whatever identification strategy is used. To understand the effect of crackdowns on crime, it is important to understand their effects on drug dealing because the literature repeatedly suggests that the trafficking of illicit drugs plays an important role in the generation of crimes, especially violent crimes with firearms (Goldstein et al., 1989; Levitt & Venkatesh, 2000; Ousey & Lee, 2004).

Caulkins (1993) developed a model to understand the effect of market size on drug dealing, and therefore on the amount of effort needed to collapse a market (see Baveja et al., 1993; Kleiman, 1988). Baveja et al. (1997) modeled this in terms of time and number of dealers. Baveja et al. (2000) found, in modeling crackdown strategies with limited resources, that crackdowns normally lead to significant results within a matter of a week. The authors also proposed that, if crackdowns are not successful by that time, prospects for success decay greatly afterward. Sherman (1990), however, argued for the potential benefit of repeated short-term crackdowns in analyzing their effect on crime. This is particularly important because it seems to be the modus operandi for the intervention in one of Montevideo’s neighborhoods, Casavalle, where there are contentious relations between gangs.

A second important question is about the type of market the police are confronting (Baveja et al., 1997; Caulkins, 1992). Real drug markets fall between two poles: a seller’s market where demand for illicit drugs is abundant and a buyer’s market where demand is reduced, and dealers fight for market share. The latter is mostly the situation in north Montevideo, and particularly Casavalle, after the shrinkage of the coca paste first, then the legalization of marijuana in 2013.

Another question references whether or not a crackdown produces lasting results. Caulkins’s (1993, see Equation 9) model suggests the ability of dealers to jump-start a market depends on their ability to coordinate. Narco-gangs can provide this level of coordination, even when not securing for themselves big shares of drug dealing on the market. To this respect, the literature agrees that maintaining good police–community relations is key for residents being more likely to report promptly a resurgence in drug dealing (Caulkins, 1993; Maher & Dixon, 2018). In general, long-term crackdown effects are dependent, on a theoretical level, on provisions taken to prevent a market from returning (Baveja et al., 1993).

From the set of studies with a clear-cut causal identification mechanism, evidence is mostly from different cities and towns of the United States. Across extensive scholarship on hot-spot policing, in particular focalized raids, evidence is mixed about the existence of long-term effects, short-term effects, or non-effects of crackdowns on crime (see Braga & Bond, 2008; Caulkins et al., 1993; Cohen et al., 2003; Kim et al., 2019; Lawton et al., 2005; Mazerolle et al., 2006; Moeller & Hesse, 2013; Novak et al., 1999; Nunn et al., 2006; Phillips et al., 2016; Shepard & Blackley, 2005; Sherman & Weisburd, 1995; Smith, 2001; Werb et al., 2011; among others).

Among those studies identifying at least short-term or even long-term effects of crackdowns, evidence is mixed about possible displacement effects of crackdowns on crime, increases of violence in the short term as a consequence of crackdowns, or even (on the positive side) diffusion or at least deterrence effects. Among the main issues identified by these studies as potential intervening factors in moderating or even washing out the desired effect of a crackdown are the other risk factors present in the area, such as organized gangs, the inability of the police to sustain the intervention over a level of critical enforcement, or even the inability to modify market conditions to decrease the probability of the market’s returning after the crackdown. The success of a crackdown maybe even related to how the detainees navigate the judicial system. In other words, crackdowns by themselves, like almost every other policing strategy, are unlikely to reduce crime when isolated from other coordinated interventions, particularly in complex environments (see Mazerolle et al., 2006).

There is also a growing stream of literature on Latin America, with a strong focus on Mexico and Central America, analyzing the problem of drug-related crime and policing strategies (see Cruz, 2010). Several scholars addressed issues of police militarization and open confrontation in contexts of weak state capacity and control over the territory (see Calderón et al., 2015; Duran-Martínez, 2018; Flores-Macías, 2018). Initial evidence from these works suggests police militarization augments violence levels, particularly kidnapping and homicides. We should note that the size and structure of gangs in Uruguay are not comparable with their Central American correlates. Furthermore, the raids we investigate are not designed as an instrument of open war on drugs as has been seen in other countries. Nevertheless, the evidence from Mexico and Central America is telling for short-term expected effects of direct intervention.

Based on the extant knowledge of the effects of crackdowns on crime, our goal is to learn about post-crackdown effects in Montevideo’s most conflicted neighborhoods for violent crimes and robberies. Our hypotheses are the following:

Hypothesis 1: We expect the analyzed crackdowns to be associated with a short-term change in levels of crime.

Hypothesis 2: We expect the effect of crackdowns to be higher for violent crimes (threats, damage, injuries, and homicides) than for robberies.

Our expectations about the effects of crackdowns in Montevideo on crime reporting are based on the comparison between evidence-based theory and the particular characteristics of these operations in the territory.

The analysis assumes that levels of reported crime follow the same trends as levels of actual criminal activity. However, we have to acknowledge that it is also possible for crackdowns to have effects on citizens’ willingness to report, which would add measurement error that would be a confounder to our causal claim. This potential problem, of course, affects all studies that examined aggregate crime statistics in response to some intervention (e.g., a gun control law’s effect on armed robberies in Boston, as studied by Berk et al., 1981; Deutsch & Alt, 1977). Unfortunately, we cannot differentiate the treatment’s effect on willingness to report from its effect on the actual level of criminal activity. Nevertheless, we do expect that any effect on the willingness to report to be short term; therefore, we assume any long-term trend could safely be attributed to actual levels of crime.

We analyze raids done by the Montevideo police between January 2017 and April 2019 as part of a renewed hot-spot policing strategy adopted in response to a sharp increase in crime and particular drug-related crime. The analyzed crackdowns were carried out in close coordination with other state agencies: Energy Company (UTE), Telecommunications Company (ANTEL), Montevideo city government, Ministry of Housing and Territorial Ordering (MVOTMA), and other state agencies. The main declared goal of these operations was the re-establishment of law and order by a reinvigoration of the state presence in areas where micro-trafficking gangs had gained power and territorial control, advancing a culture of illegality. The government understands the importance of avoiding violence escalation by entering an open drug war as well as the long-term character of the policing strategy oriented to curve down the drug market as well as crime. However, by 2017, some of the gangs began to expel people from their homes through threats and extortion, improving the control of the territory. The police also perceived an increase in the use of dissuasion mechanisms to prevent public services to enter various areas of the neighborhoods, denying access to ambulances and public transportation.⁶ This was seen by the government as a qualitative change in the extant equilibrium of forces with the gangs, which triggered a response based on the combination of crackdowns and PADO policing right afterward, with some small urban interventions such as street openings, house demolitions, and improved street illumination. Figure 1 shows the trend in terms of total crime in Montevideo, whereas Figure 2 illustrates the differential prevalence of crimes by neighborhood in Montevideo, as well as the areas where the crackdowns took place.

Figure 1.

Number of crimes by week, Montevideo.

Figure 2.

Total crimes by neighborhood and areas with raids.

As these highly coordinated raids were an innovation in terms of policing strategy in Montevideo, there was no previous experience of this type of hot-spot policing in the government or the police. Most raids lasted by only 1 day, not being repeated afterward, at least in the short term. There are two exceptions to this pattern: In Cerro Norte, the police made two interventions in a week, during Weeks 172 and 173. In Casavalle, as already explained, eight raids occurred between Weeks 156 and 221. Table 1 shows the crackdowns in order of occurrence, detailing the number of police officers directly involved (if known), the number of days each operation lasted, the number of detainees, the number of properties entered, and finally a dummy variable for when drug seizures occurred.

Table 1.

Montevideo Crackdowns.

Neighborhood	Date	Police officers	Days	Detainees	Raids (houses)	Drug seizures
Casavalle	December 20, 2017	600	1 (first one)	35	64	1
Cerro Norte	April 10, 2018	—	2 (within a week)	26	88	2
Malvín Norte	June 13, 2018	376	1	13	29	1
Jardines del Hipódromo	October 4, 2018	—	1	8	12	1
Villa Española	October 16, 2018	350	1	16	54	1
Peñarol	November 15, 2018	—	1	16	18	0
Tres Ombués	December 8, 2018	—	1	16	27	1

Source. Elaborated with data provided by the Ministry of the Interior and Official website of Ministry of the Interior.

As shown in Table 1, most crackdowns were effective in seizing drugs. The average number of search warrants per crackdown is 32 properties, and the average number of detainees is 16 people per crackdown. For those cases where the number of police officers was reported on official documents, the average is 442 per raid.

Research Design and Causal Identification

Data

Crime and crackdown information was provided by the Center for Criminal Data (UAC—Unidad de Análisis Criminal) and the Montevideo Police Department, both under the authorization of the Ministry of the Interior (MI). Because the temporal unit of analysis is the weekly count of events, each variable is presented as the number of incidents per week. The time frame for the study includes 156 weeks before the first raid and between 20 and 65 weeks after each raid occurred, resulting in a total of 221 week-observations.⁷

Dependent Variables and Synthetic Control

The dependent variables are the logarithms of (a) reports on robberies and (b) violent crimes (sum of threats, damage, injuries, and homicides) per 1,000 people. Each model (see Equation 2 in the “Statistical Methods” section) uses one of these dependent variables. For each crackdown (except for Casavalle), we model the two outcomes at two levels and with two strategies for comparison. The first level of comparison is the neighborhood, while the second level of comparison is the exact area in which the raid was performed, built from census segments (block groups). We use the two strategies as we recognize using the neighborhood as the unit of analysis allows for a potential misrepresentation of the substantive geographical area in real terms. For simplicity, however, we focus on the neighborhood-level analysis in this article, with the segment-level analysis in the appendix.⁸ For Casavalle, we only model at the neighborhood level because multiple places inside the neighborhood have been intervened in continuously.

For each analyzed crackdown, we built a synthetic control neighborhood following Abadie et al.’s (2010) proposed methodology, as shown in Equation 1. As explained by Abadie et al. (2010), a synthetic control is a weighted combination of control neighborhoods chosen to approximate the treated neighborhood in terms of the crime outcome predictors.

\sum_{m = 1}^{k} v_{m} {(X_{1 m} - X_{0 m} W)}^{2},

(1)

where following Linden (2018), X_1m is the value of the mth preintervention covariate for the treated neighborhood and X_0m is a 1 × J vector containing the values of the mth preintervention covariates for the non-treated neighborhoods in the sample. The neighborhood donor pool is composed of each neighborhood that did not has had an intervention by the time the treated neighborhood receives the intervention. Next, v_m is a weight that reflects the relative importance that is assigned to the mth preintervention covariate. When measuring the discrepancy between X₁ and X₀W for each of the three predictors of crime that we use (prevalence of poverty, percentage of young males aged between 15 and 29 years old, and average years of education), we want the predictors with large predictive power on crime for the treated neighborhood to be assigned large v_m weights. All predictors are drawn from Household Survey data for years 2015 to 2017 (Instituto Nacional de Estadística [INE], n.d.). For a seminal example of the analysis of the relation between drug selling and poverty, see Reuter et al. (1990).

In constructing each control, we modeled the dependent variable for every week between Week 1 and the week before the intervention. We only included neighborhoods without a crackdown to participate in the analysis. Therefore, the evolution of the outcome for the resulting control group is an estimate of the counterfactual of what would have been observed for the affected unit in the absence of a crackdown. The combination of interrupted time series and synthetic controls helps improve causal inference in policy evaluation (Linden, 2018).

Descriptive Statistics

Figure 3 shows time-series line graphs of the changes in the levels of crime reports for our cases from January 2015 to March 2019. Each vertical black line represents an intervention week in a certain place.

Figure 3.

Montevideo weekly assault report and crackdowns (vertical line), by neighborhood.

Examining Figure 3, a few facts become apparent. First, the sharp increase in crime reporting shown in Figure 1 does not represent a homogeneous change in the level of crime reporting in the treated problematic cases. Rather, we see different patterns across them. The first crackdown occurs immediately after the sharp increase shown in Figure 1 (during Week 154), at Week 156 in Casavalle. Second, the vertical lines in Figure 3 represent the crackdowns. The neighborhood standing out by being treated on several occasions is Casavalle, the most problematic in terms of organized drug gangs where two of them—the Chingas and the Camala—are engaged in a continuous fight over the territory.

Statistical Methods

The study is based on interrupted time-series analyses to examine whether police crackdowns reduce crime reporting. There is a two-step procedure: determining an Autoregressive Integrated Moving Average (ARIMA) model to evaluate stationarity and need to account for autocorrelation and then building models with treatment and control groups.

We first tested for unit roots using Dickey–Fuller Generalized Least Squares (DF-GLS). As shown in Table 2, the data by neighborhood are stationary for the dependent variables. Even with differencing, these outcomes required an additional autoregressive process, meaning a full ARIMA model was necessary, and accounting for autocorrelation is essential in model estimation.

Table 2.

Augmented Dickey–Fuller Test for Unit Root.

Neighborhood	Observations	Assaults		Violent crimes
Neighborhood	Observations	Test statistic (three lags)	MacKinnon approximatep value for Z(t)	Test statistic (three lags)	MacKinnon approximatep value for Z(t)
Cerro Norte	216	−4.726	.0006	−5.430	.0000
Malvín Norte	216	−5.289	.0001	−7.172	.0000
Jardines	216	−6.395	.0000	−7.561	.0000
Villa Española	216	−5.881	.0000	−6.896	.0000
Peñarol	216	−5.357	.0000	−7.595	.0000
Tres Ombúes	216	−5.798	.0000	−7.126	.0000

The final equation for each model is as follows:

{Crime}_{t} = β_{0} + β_{1} T_{t} + β_{2} X_{t} + β_{3} X_{t} T_{t} + β_{4} Z + β_{5} Z T_{t} + β_{6} X T_{t} + β_{7} Z X_{t} T_{t} + ε_{i},

(2)

where Z is a dummy variable to denote the cohort assignment (treatment or control), T is a time trend, and X is a dummy variable to indicate whether it is a post-intervention period. Therefore, ZT_t, ZX_t, and ZX_tT_t are all interaction terms among previously described variables (see Linden, 2015).

From Equation 2, β₀ to β₃ represent the control group whereas β₄ and β₅ represent the pre-treatment differences in level and slope between the treatment and control groups. Therefore, the non-significance of the β₄ and β₅ parameters suggests ideal conditions for the comparison (Linden & Adams, 2011). β₆ represents the difference between treatment and control groups in the level of the outcome right after the crackdown. β₇ indicates the difference in slopes right after the intervention (crackdown). Finally, post-intervention trends and differences between trends are computed for the treated unit and the control one. By using synthetic controls at the neighborhood level, we assume confounding omitted variables affect both treatment and control groups similarly. For evidence that this is correct, in all models, β₄ and β₅ ideally will be close to zero.

We find higher order serial correlation in the error terms of our model.⁹ To deal with this, one approach that we could have taken in response is to estimate the model with ordinary least squares (OLS), which produces inefficient coefficient estimates in the presence of serial correlation but would allow us to compute Newey–West Standard Errors, which are unbiased measures of uncertainty in the presence of higher order serial correlation. The approach we instead took was to estimate the model using the Prais–Winsten algorithm for Feasible Generalized Least Squares (FGLS). While Prais–Winsten only allows for first-order serial correlation, the coefficient estimates themselves will be computed without making an inaccurate assumption of independent errors. In addition, if we estimate the model with OLS, we jeopardize comparison conditions with significant values β₄ or β₅ for several models. When we instead estimate the models using Prais–Winsten FGLS, not only do the coefficients gain efficiency in the presence of serial autocorrelation, but the efficient coefficients jeopardize ideal comparison conditions with a significant β₄ or β₅ in fewer cases than the OLS model does. Given both the theoretical properties of the estimator and the added power for our causal inference goals, we estimate models with Prais–Winsten FGLS. Results are shown in Table 3. Alternative specifications, as well as OLS models with Newey–West Standard Errors, are presented in the appendix.¹⁰

Table 3.

Robberies and Violent Crimes by Neighborhood, 30 Weeks After and Before (Synthetic Control).

Variables	Cerro Norte		Malvin Norte		Jardines		Villa Española		Peñarol		Tres Ombues
Variables	Robberies	Violent crimes	Robberies	Violent crimes	Robberies	Violent crimes	Robberies	Violent crimes	Robberies	Violent crimes	Robberies	Violent crimes
T	−0.006	−0.003	0.013	−0.006	−0.002	−0.013	0.004	0.004	0.002	0.01	0.004	0.008
	(0.007)	(0.007)	(0.01)	(0.019)	(0.01)	(0.008)	(0.007)	(0.01)	(0.005)	(0.008)	(0.01)	(0.012)
Z	0.188	−0.312**	0.402*	−0.16	−0.719***	−0.067	−0.237	−0.058	0.563***	−0.1	−0.393*	−0.146
	(0.173)	(0.155)	(0.235)	(0.439)	(0.227)	(0.178)	(0.168)	(0.226)	(0.116)	(0.184)	(0.228)	(0.278)
ZT	0.009	0.014	0.01	−0.001	0.028**	0.011	0.001	−0.01	−0.025***	−0.013	−0.012	−0.003
	(0.011)	(0.01)	(0.014)	(0.026)	(0.014)	(0.011)	(0.01)	(0.014)	(0.007)	(0.011)	(0.014)	(0.017)
X	0.349*	−0.286*	−0.251	0.08	−0.032	0.31	−0.124	0.082	−0.044	−0.04	0.022	−0.299
	(0.176)	(0.159)	(0.239)	(0.404)	(0.243)	(0.194)	(0.188)	(0.253)	(0.136)	(0.217)	(0.277)	(0.344)
XT	0.015	0.01	−0.011	0.007	0.033*	0.006	0.007	−0.01	0.031**	−0.011	0.001	0.01
	(0.01)	(0.009)	(0.014)	(0.027)	(0.017)	(0.013)	(0.014)	(0.019)	(0.012)	(0.02)	(0.031)	(0.038)
ZX	−0.421*	−0.063	−0.405	0.237	0.124	−0.235	0.092	0.078	0.284	0.204	0.308	−0.468
	(0.25)	(0.224)	(0.339)	(0.572)	(0.344)	(0.275)	(0.265)	(0.357)	(0.193)	(0.307)	(0.391)	(0.487)
ZXT	−0.019	−0.02	0.027	−0.005	−0.088***	−0.009	0.014	0.033	−0.023	0	0.03	0.066
	(0.015)	(0.013)	(0.02)	(0.038)	(0.024)	(0.019)	(0.02)	(0.026)	(0.017)	(0.028)	(0.044)	(0.054)
Constant	−1.260***	−0.993***	−1.769***	−1.362***	−1.057***	−1.377***	−1.295***	−1.704***	−1.080***	−1.526***	−0.947***	−1.227***
	(0.122)	(0.11)	(0.166)	(0.311)	(0.16)	(0.126)	(0.119)	(0.16)	(0.082)	(0.13)	(0.161)	(0.196)
Post-trend control	0.0083	0.0072	0.0024	0.0015	0.0310**	−0.0066	0.0111	−0.0063	0.0329***	−0.0007	0.005	0,0177
	(0.0071)	(0.0064)	(0.0097)	(0.0177)	(0.0139)	(0.011)	(0.0119)	(0.016)	(0.0112)	(0.0179)	(0.029)	(0.0360)
Post-trend treatment	−0.0023	0.0016	0.0397***	−0.0051	−0.0293**	−0.0052	0.0255**	0.0163	−0.0145	−0.0138	0.0235	0.0810**
	(0.0071)	(0.0064)	(0.0097)	(0.0177)	(0.0139)	(0.011)	(0.0119)	(0.016)	(0.0112)	(0.0179)	(0.029)	(0.0360)
Post-trend difference	−0.0106	−0.0057	0.0373***	−0.0065	−0.0603***	0.0015	0.0144	0.0226	−0.0474***	−0.0132	0.0184	0.0633
	(0.0101)	(0.009)	(0.0137)	(0.025)	(0.0196)	(0.0155)	(0.0168)	(0.0226)	(0.0159)	(0.0253)	(0.041)	(0.0509)
Weeks before	30	30	30	30	30	30	30	30	30	30	30	30
Weeks after	30	30	30	30	23	23	21	21	17	17	14	14
Observations	118	118	118	118	104	104	100	100	92	92	86	86
R ²	.17	.135	.463	.032	.301	.051	.169	.08	.397	.206	.302	.12

Note. Standard errors in parentheses. T, Z, and ZT (B₁, B₂, and B₃ represent the control group), X (B₄: pre-T diff level T vs. C); XT (B₅: Pre-T diff slope T vs. C); ZX (B₆: diff-in-diff levels T vs. C); ZXT (B₇: diff-in-diff slopes T vs. C).

p < .1. **p < .05. ***p < .01.

The neighborhood of Casavalle is a special case. The level of crime there is very high, the land area of the neighborhood is very broad, there are two gangs at war with each other, and the police intervened 8 times there during our time frame. Given the wholly unique features of Casavalle, there is no reasonable comparison group for this neighborhood with which to do difference-in-differences analysis. For this reason, we turn instead to a univariate time-series intervention analysis. That is, we find the Box–Jenkins noise model that describes the correlation patterns of crime (ARIMA model) and measures the dynamic impact that the eight interventions appear to have on the crime series in this neighborhood. Ideally, we would have been able to assess how changes in Casavalle contrast from similar neighborhoods. However, given that the neighborhood received eight separate treatments and has no approximate counterpart, a single series analysis is the only viable option.

Results

For the models presented in Table 3, results for the effect of crackdowns on crime reporting are mixed. As mentioned, the case of Casavalle is analyzed separately in Table 4 as it is the only case for which there were several and recurrent crackdowns beginning in Week 156 of our data and until the penultimate week for which we have data (Week 220). The results presented in Table 3 are consistent with the literature in general, in particular when taking into account that these crackdowns lasted only 1 day in every case. As single-day raids, their potential to have a lasting effect on local drug markets or even to reach a level of critical enforcement is low according to theoretical expectations (see Baveja et al., 1993, 1997, 2000, 2004; Cohen et al., 2003). The analysis of the two dependent variables (reports of robberies and violent crime) for six neighborhoods creates 12 interrupted time-series models with multiple groups.

Table 4.

Intervention Analysis of Eight Major Raids on Crimes in Casavalle.

Parameter	Assaults			Violent crimes
Parameter	Estimate	SE	p	Estimate	SE	p
December 20, 2017, Mirador I Step	0.526	0.146	.000	0.046	0.105	.659
Decay, Mirador I Pulse	−0.999	0.052	.000	0.386	0.478	.419
Effect, Mirador I Pulse	0.038	0.074	.605	−0.461	0.331	.164
March 15, 2018	−0.490	0.345	.155	0.504	0.325	.121
March 20, 2018	0.346	0.343	.313	0.115	0.325	.724
August 14, 2018	−0.158	0.342	.644	0.034	0.330	.919
August 28, 2018	−0.008	0.342	.981	0.316	0.328	.335
October 2, 2018	−0.236	0.339	.487	0.171	0.325	.600
November 20, 2018	0.226	0.345	.513	0.574	0.328	.080
March 11, 2019	0.017	0.355	.962	0.048	0.335	.886
Intercept	−1.024	0.086	.000	−1.037	0.057	.000
φ₁	0.271	0.068	.000	0.150	0.068	.026
φ₂	0.136	0.070	.052	0.178	0.068	.009
φ₃	0.056	0.071	.430	−0.025	0.069	.720
φ₄	0.092	0.070	.190	0.157	0.069	.022
φ₅	0.127	0.069	.066	0.078	0.068	.251
AIC	196.383			177.073

Note. T = 220 for all models. All interventions are pulse interventions, except the compound intervention for Mirador I. Estimated in R 3.6.1. AIC = Akaike information criterion.

Overall, we find mixed evidence, within the boundaries of the data frame we have, of crackdowns on crime. For robberies (see Table 1), there is a significant differential effect of crackdowns in crime between treatment and control neighborhoods. Crime goes down in Jardines and Peñarol, but it goes up after the raid in Malvín Norte. These are the main results of the analysis. The post-raid trend in crime significantly increases in Villa Española, albeit there is no significant differential effect versus the synthetic neighborhood. For violent crimes, there are no significant differential effects in any neighborhood. However, the post-raid trend in crime significantly increases in Tres Ombues (see Table 1). We do not find any effect for Cerro Norte under any of the proposed specifications. Hence, it appears that the effects of raids are sensitive to the context of the criminal situation.¹¹

Some limitations need to be put forward before entering into the substantive analysis. Ideal conditions for comparison—given by the pre-treatment equivalence in level and slope in the two groups, or non-significance of the β₄ and β₅ parameters—are not present in all cases. A second limitation is that we are assessing the effect of the crackdown on crime locally, not at the city level. A third limitation refers to the assessment of long-term effects in that there are ongoing operations beyond the time frame of the data. Therefore, the assessment of trends should be carefully pondered substantively in the different cases.

Findings

We first assessed the effect of crackdowns on robberies. We find a significant downward differential effect in the post-treatment trend in Jardines. During Week 197, the Jardines crackdown took place in Complejo Quevedo, a housing complex built by the government to relocate people living in slums. During this intervention, the police detained eight people, four of which ended up in prison. Three of the incarcerated were leaders of the “Figueroa” gang: led by Marcela Figueroa, her husband, and her father. The gang had built a wall to fortify the housing complex, installing a 24-hr surveillance system, including paid guards and 36 cameras around the drug-selling point. The gang emulated the same forced home-eviction system as Los Chingas did in Casavalle, with those families who remained in their houses put under continuous threat. It controlled the commonplaces of the complex, in particular the multi-purpose meeting room, using it as headquarters. The ability of the police to behead the gang, demolish the wall and surveillance system, and return the original inhabitants to their homes is consistent with the long-term downward trend that we find for robberies for the remaining 23 weeks of our data.

We also find a consistent downward differential effect in the trend of robberies in Peñarol. This is consistent with the reported evaluation of the raid made by the police. The intervention was successful in capturing and imprisoning the leader of the Algorta family gang and other relevant members. This gang was in a long-term conflict with the “Tato” gang, from a neighboring area, for the control of the territory. Between 2015 and 2018, the war between these two groups led to 29 homicides and 36 wounded people. However, the two groups greatly debilitated each other, with the Algorta gang prevailing after joining forces with the Delfino family, before the police helped the process with the capture of the latter group leaders. With the destruction of the remaining Algorta gang, the drop in robberies has been a positive outcome for Peñarol.

By contrast, we find a significant upward differential effect in the post-treatment trend in robberies in Malvín Norte. Cuyi’s gang in Malvín Norte was under internal conflict since the capturing of the Cuyi by the police on Week 138 in an isolated event. Cuyi’s wife took control of the gang, but internal rivalries appeared. The crackdown took place on Week 181, with the direct involvement of 376 policemen. The intervention included 29 home-search warrants and 13 persons were detained and incarcerated on charges of drug crimes, vehicle robbery, illegal gun trafficking, and domestic violence. However, Cuyi’s wife was not among the detainees. This anecdotal evidence goes in the direction of a short-term upward effect in crimes because of the partial success of the intervention. Although our data end on Week 221, Cuyi’s wife was finally incarcerated on Week 224 (noticias.uy, 2019). The increase in robberies during the time frame of the study is reflective of the fact that Cuyi’s gang was not fully dismantled at the time.

Turning to violent crimes, we do not find significant differential effects in post-treatments trends, but we do find a significant upward effect in crime in Tres Ombues. The case of Tres Ombues presents some particular features as it involves linkages with drug dealing in soccer stadiums. During Week 206, the crackdown involved breaking into 26 properties and detaining 16 people, although most of them were freed after investigations. Drug dealers had surveillance camera systems in the neighborhood for self-protection against other dealers. This case displays another two particular facts about drug dealers: the between-group confrontation and its relationship with drug dealing inside soccer stadiums. Guns, money, and drugs were seized, although no gang was dismantled. The crackdown involved demolition works on key houses. However, consistent with our empirical finding, 16 weeks after the raid a pick-up truck was burned with two soccer hooligans inside, who were incinerated. This crime shocked the public because of its unusual violence and its direct linkage with drug dealing in soccer. With persistent conflict among groups and gangs that have not been destroyed, violent crimes have become even more pervasive in Tres Ombues.

The Special Case of Casavalle

Casavalle represents a special case because it is the only neighborhood in the sample in which multiple crackdowns took place, beginning on Week 156 and not ending within our period of study. In our sample, the last crackdown occurs on Week 220. This scenario resembles a strategy based on a combination of short-term crackdowns and backoffs. Casavalle is the most complex scenario for drug gangs and drug trafficking in the city. Between December 20, 2017, and March 11, 2019, eight crackdowns were carried out in Casavalle.

The first raid took place in December 2017 after 110 families were thrown out of their homes by the gang, Los Chingas, as part of a plan to take control of Casavalle and win the fight against the other criminal gang that is operative in the area, Los Camala.¹² The occupied dwellings were rented or sold out to persons related to the gang, used as a meeting place, or used as a place to deposit weapons, drugs, and stolen objects. After the first crackdown, five of the Chingas’ leaders were imprisoned, and another six were released or put under home-based surveillance, one of them being killed in a settling of scores shortly thereafter. However, the only leader of Los Chingas who was not captured managed to reorganize the gang.

Consequently, there were another six crackdowns in Casavalle during 2018, mainly linked to new reports of home-usurpations, homicides, and settling of scores between the two gangs. Many of the leaders in both gangs were arrested, but the groups reorganized with new leaders. According to some sources, they continued to be operative under the orders of jailed leaders. When considering the eight raids that took place between December 2017 and March 2019 altogether, there were 185 home-search warrants, more than 90 arrests, house demolitions, and thousands of police officers working in the area.

It is interesting to observe how this strategy of 1-day crackdowns has allowed the police to develop a varied set of strategies like demolitions to make the neighborhood more searchable and to improve the accessibility of public services such as transportation or even the PADO patrolling in stages. However, it has also made it possible for the gangs to regroup and to continue operations.

How did the eight raids that occurred within our time frame affect crime rates in Casavalle? Table 4 shows the results of our intervention analyses, where the first three numeric columns describe the results of a model of the logged total number of robberies per 1,000 persons in Casavalle by week. The final three columns show the results for the logged number of violent crimes per 1,000 residents of Casavalle. In the rows, the first three rows show the results of a compound intervention for Mirador I, the first raid conducted in Montevideo. This allows for a dynamic temporary effect followed by a persistent long-term effect. The next seven columns show the effects of pulse interventions of seven more raids. Following this is each model’s intercept, five autoregressive parameters, and finally the Akaike information criterion (AIC) as a fit index.

Focusing on the first row of Table 4, we see that there is a positive and significant long-term effect of Mirador I on robberies in Casavalle. This means that crimes tended to be more common in the portion of the series after this raid rather than before. There was not a discernible additional temporary effect, however. Furthermore, Mirador I had no discernible effect on violent crimes. The seventh raid, on November 20, 2018, appeared to raise violent crimes temporarily, but no other raid had a discernible effect on either of these series.¹³ We tried several functional forms for all of the interventions, and the story shown in Table 4 is consistently what we find: Raids in Casavalle do not show evidence of reducing crime, but on occasion, some raids correspond to a rise in crime. Given that Casavalle is a space where two gangs are at war, it makes sense that when one takes a hit from the police, the other makes a strong move to take control of the territory—hence, the fact that crime has been elevated since Mirador I, driven mainly by a rise in robberies.

Discussion and Conclusion

The study aims to analyze the effect of crackdowns on crime in Montevideo. To achieve this, we use interrupted time-series analysis and intervention models for the special case of Casavalle. Where ideal conditions for comparison are met, we find scarce effects of crackdowns in crime reporting. In five out of six neighborhoods—leaving Casavalle out—we find some significant effects for one of the two types of crime that we study. However, significant differences in post-treatment trends exist only for robberies and in three neighborhoods. But sometimes crime rises while other times it falls. It also is notable that in our separate analysis of Casavalle, crimes may have gotten worse as a response to raids. Past literature would imply that, unless a drug market is completely crushed, it will restart after a crackdown. Given the rival gangs in this neighborhood, stunning one gang may be counterproductive and an invitation for the other gang to escalate its activities. In particular, the two cases where we saw a discernible differential downward effect in crime were for robberies in Jardines and Peñarol. In each of these cases, we qualitatively observe that the gang was utterly crushed by police without a clear rival to encroach on the territory. By contrast, in the neighborhoods of Casavalle and Malvín Norte (during the time frame of study), the gangs were not fully beheaded and often had a clear rival. Hence, in the latter cases, crime was elevated for a time—though in the long run, the raids may have been essential for preventing warring gangs from becoming entrenched.

Our findings offer implications for policy development and future research. One of the most important findings is that crackdowns are not consistently effective in influencing crime reporting. This is consistent with the literature in two important aspects. First, it is hard to reach levels of critical enforcement (see Caulkins, 1992) through 1-day crackdowns. The number of police officers involved in these crackdowns vis-à-vis the results in terms of detainees, for the cases we have such data, also suggests raids have not been particularly efficient. Second, in the same vein, crackdowns’ ability to alter drug-market conditions would depend not only on the ability to extract drug dealers from the territory but also in preventing a rapid return. For the two cases in which a significant difference in a downward trend in crimes occurs, Peñarol and Jardines, qualitative evidence from police reports and press support the idea that gangs were neutralized. For the cases in which an upward trend in crimes was found, similar evidence is consistent with the idea that gangs were not fully discouraged from continuing fighting for the territory. Avoiding speedy returns, in the context of effective rule of law, will depend on effective coordinated work between the judicial system and the police for effectively finding convincing evidence of a crime. It also is important to avoid replacement by natural selection, such as new dealers with improved strategies (see Caulkins, 1992). Furthermore, the contrast in situations where raids were effective versus ineffective also suggests that future research may want to consider a key actor analysis for evaluating when and how crackdowns can be effective.

In Uruguay, the judicial system and the police patrolling strategies have been under deep reform processes since 2013, which may suggest coordination costs have risen as new actors and institutions are in place. There is strong evidence from press releases about coordination problems and public conflict between the Attorney General and the MI. However, it is reasonable to expect this capacity for coordination may improve in the long run.

The effects we find are consistent with theoretical expectations given the qualitative reports available for each case. This supports the overall claims that hot-spot policing should be accompanied by other complementary strategies, and that raids alone are an important albeit insufficient strategy. In Montevideo, focused-patrolling (PADO) and CPTED strategies are two locally innovative strategies that authorities have been implementing, in a few cases in combination with raids. Moreover, a new program of architectural and urbanistic interventions has been recently announced by the authorities for Las Acacia’s neighborhood, which is expected to contribute to raising the costs of dealers and therefore moving crimes downward (see BID, 2018, for PADO evaluation and Casanova, 2019, for the urbanistic interventions).

Finally, the evaluation of hot-spot policing should not be dissociated from a deeper knowledge of the action and location of drug dealers in the territory. Uruguay, and Montevideo in particular, is increasingly the focus of action of groups coming from other Latin American countries, as Uruguay is one of the main routes for cocaine to leave for Europe and worldwide. In this sense, hot-spot patrolling’s ability to modify this higher level market should be considered, at least, insufficient. In the same vein, policing strategies should also be combined with nationally crafted frontier protection strategies. More research is needed in this nascent market that has the potential ability to seriously threaten the rule of law and the state itself.

Footnotes

Acknowledgements

For helpful assistance, we thank Peter H. Reuter, Angelica Durán-Martinez, Matthew L. Hipps, and Melissa Whatley.

Authors’ Note

Previous versions of this work were presented at the Annual Meeting of the American Political Science Association (August 2019, Washington), the Annual Meeting of the Georgia Political Science Association (November 2019, Savannah), and the Latin American PolMeth Meeting (November 2019, Montevideo).

Declaration of Conflicting Interest

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The financial support from the grant FSSC_1_2018_1_147750 from the Uruguayan National Agency for Research and Innovation (ANII).

ORCID iD

Juan A. Bogliaccini

Notes

Author Biographies

Juan A. Bogliaccini is Full Professor at the Department of Social Sciences, Universidad Católica del Uruguay. He received his Ph.D. in Political Science from the University of North Carolina, at Chapel Hill. He is Leading Editor at the Latin American Political Economy Series, Palgrave MacMillan.

Diego Pereira is an Adjunct Professor at the Department of Social Sciences, Universidad Católica del Uruguay. He received his MBA from the Universidad Católica del Uruguay.

Juan Ignacio Pereira is a PhD student at Brown University. He received his M.A. in Public Policy from the University of Chicago.

Cecilia Giambruno is a research associate at the Inter-American Development Bank. She received his M.A. in Public Policy form the Universidad Católica del Uruguay.

Ignacio Borba is a research assistant at the Department of Social Sciences, Universidad Católica del Uruguay. He is an undergrad student at the Universidad Católica del Uruguay.

References

Abadie

Diamond

Hainmueller

A. J.

(2010). Synthetic control methods for comparative case studies: Estimating the effect of California’s tobacco control program. Journal of the American Statistical Association, 105(490), 493–505. https://doi.org/10.1198/jasa.2009.ap08746

Averdijk

Van Gelder

J. L.

Eisner

Ribeaud

(2016). Violence begets violence . . . but how? A decision-making perspective on the victim-offender overlap. Criminology, 54(2), 282–306. https://doi.org/10.1111/1745-9125.12102

Baum

C. F.

Schaffer

M. E.

, 2013. ACTEST: Stata module to perform Cumby-Huizinga general test for autocorrelation in time series.

Baveja

Batta

Caulkins

J. P.

Karwan

M. H.

(1993). Modeling the response of illicit drug markets to local enforcement. Socio-Economic Planning Sciences, 27(2), 73–89. https://doi.org/10.1016/0038-0121(93)90008-7

Baveja

Caulkins

J. P.

Liu

Batta

Karwan

M. H.

(1997). When haste makes sense: Cracking down on street markets for illicit drugs. Socio-Economic Planning Sciences, 31(4), 293–306. https://doi.org/10.1016/S0038-0121(97)00002-5

Baveja

Feichtinger

Hartl

R. F.

Haunschmied

J. L.

Kort

P. M.

(2000). A resource-constrained optimal control model for a crackdown on illicit drug markets. Journal of Mathematical Analysis and Applications, 249(1), 53–79. https://doi.org/10.1006/jmaa.2000.6931

Baveja

Jamil

Kushary

(2004). A sequential model for cracking down on street markets for illicit drugs. Socio-Economic Planning Sciences, 38(1), 7–41. https://doi.org/10.1016/S0038-0121(03)00026-0

Berk

R. A.

McCleary

Hay

R. A.

(1981). Applied time series analysis for the social sciences. Contemporary Sociology, 10(6), 818–819. https://doi.org/10.2307/2067244

BID Interamerican Development Bank. (2018). ¿Cómo evitar el delito urbano? el Programa de Alta Dedicación Operativa en la nueva Policía uruguaya. In Mi.

del Interior

B. I.

de Desarrollo

(Eds.), ¿Cómo evitar el delito urbano? el Programa de Alta Dedicación Operativa en la nueva Policía uruguaya. Gráfica Mosca. https://doi.org/10.18235/0001078

10.

Braga

A. A.

Bond

B. J.

(2008). Policing crime and disorder hot spots: A randomized controlled trial. Criminology, 46(3), 577–607. https://doi.org/10.1111/j.1745-9125.2008.00124.x

11.

Calderón

Robles

Díaz-Cayeros

Magaloni

(2015). The beheading of criminal organizations and the dynamics of violence in Mexico. Journal of Conflict Resolution, 59(8), 1455–1485. https://doi.org/10.1177/0022002715587053

12.

Casanova

(2019). La mirada urbana en la prevención del delito: el caso Marconi. https://www.minterior.gub.uy/images/libros_y_revistas/MI_Barrio_Marconi.pdf

13.

Caulkins

J. P.

(1992). Thinking about displacement in drug markets: Why observing change of venue isn’t enough. Journal of Drug Issues, 22(1), 17–30. https://doi.org/10.1177/002204269202200102

14.

Caulkins

J. P.

(1993). Local drug markets’ response to focused police enforcement. Operations Research, 41(5), 816–1008. https://doi.org/10.1287/opre.41.5.848

15.

Caulkins

J. P.

Larson

R. C.

Rich

T. F.

(1993). Geography’s impact on the success of focused local drug enforcement operations. Socio-Economic Planning Sciences, 27(2), 119–130. https://doi.org/10.1016/0038-0121(93)90012-8

16.

Cohen

Gorr

Singh

(2003). Estimating intervention effects in varying risk settings: Do police raids reduce illegal drug dealing at nuisance bars? Criminology, 41(2), 257–292. https://doi.org/10.1111/j.1745-9125.2003.tb00988.x

17.

Cruz

J. M.

(2010). Central American maras: From youth street gangs to transnational protection rackets. Global Crime, 11(4), 379–398. https://doi.org/10.1080/17440572.2010.519518

18.

Cumby

R. E.

Huizinga

(1992). Testing the autocorrelation structure of disturbances in ordinary least squares and instrumental variables regressions. Econometrica, 60(1), 185–195. https://doi.org/10.2307/2951684

19.

Deutsch

S. J.

Alt

F. B.

(1977). The effect of Massachusetts’ gun control law on gun-related crimes in the city of Boston. Evaluation Quarterly, 1(4), 543–568. https://doi.org/10.1177/0193841X7700100402

20.

Duran-Martínez

(2018). The politics of drug violence. Oxford University Press.

21.

Flores-Macías

G. A.

(2018). The consequences of militarizing anti-drug efforts for state capacity in Latin America: Evidence from Mexico. Comparative Politics, 51(1), 1–20. https://doi.org/10.5129/001041518824414647

22.

Goldstein

P. J.

Brownstein

H. H.

Ryna

P. J.

Bellucci

P. A.

(1989). Crack and homicide in New York City, 1988 a conceptually based event analysis. Contemporary Drug Problems, 16(4), 651–673.

23.

Instituto Nacional de Estadística. (n.d.). Encuesta continua de hogares. http://www.ine.gub.uy/encuesta-continua-de-hogares3

24.

Jacottet

(2018). Configuraciones territoriales de las violencias y el delito en la ciudad de Montevideo. In ¿Cómo evitar el delito urbano? el Programa de Alta Dedicación Operativa en la nueva Policía uruguaya Ministerio del Interior. Montevideo. (pp. 73–95).

25.

Kim

D. Y.

Phillips

S. W.

Wheeler

A. P.

(2019). Using “symbolic” SWAT raids as a crime reduction strategy: Are their effects “instrumental” in nature? Criminal Justice Policy Review, 30(2), 176–200. https://doi.org/10.1177/0887403416664567

26.

Kleiman

. (1988). Crackdowns: The effects of intensive enforcement on retail heroin dealing. Program in Criminal Justice Policy and Management, John F. Kennedy School of Government, Harvard University.

27.

La Diaria. (2018, January 13). Casavalle: tres meses bajo toque de queda. https://findesemana.ladiaria.com.uy/articulo/2018/1/casavalle-tres-meses-bajo-toque-de-queda/

28.

Lawton

B. A.

Taylor

R. B.

Luongo

A. J.

(2005). Police officers on drug corners in Philadelphia, drug crime, and violent crime: Intended, diffusion, and displacement impacts. Justice Quarterly, 22(4), 427–451. https://doi.org/10.1080/07418820500364619

29.

Levitt

S. D.

Venkatesh

S. A.

(2000). An economic analysis of a drug-selling gang’s finances. Quarterly Journal of Economics, 115(3), 755–789. https://doi.org/10.1162/003355300554908

30.

Linden

(2015). Conducting interrupted time-series analysis for single- and multiple-group comparisons. Stata Journal, 15(2), 480–500.

31.

Linden

(2018). Combining synthetic controls and interrupted time series analysis to improve causal inference in program evaluation. Journal of Evaluation in Clinical Practice, 24(2), 447–453. https://doi.org/10.1111/jep.12882

32.

Linden

Adams

J. L.

(2011). Applying a propensity score-based weighting model to interrupted time series data: Improving causal inference in programme evaluation. Journal of Evaluation in Clinical Practice, 17(6), 1231–1238. https://doi.org/10.1111/j.1365-2753.2010.01504.x

33.

Maher

Dixon

(2018). The cost of crackdowns: Policing Cabramatta’s heroin market. Current Issues in Criminal Justice, 13(1), 5–22. https://doi.org/10.1080/10345329.2001.12036213

34.

Mazerolle

Soole

D. W.

Rombouts

(2006). Street-level drug law enforcement: A meta-analytical review. Journal of Experimental Criminology, 2, 409–435. https://doi.org/10.1007/s11292-006-9017-6

35.

Moeller

Hesse

(2013). Drug market disruption and systemic violence: Cannabis markets in Copenhagen. European Journal of Criminology, 10(2), 206–221. https://doi.org/10.1177/1477370812467568

36.

noticias.uy. (2019). Cayó la mujer del Cuyi, el líder del marcotráfico en Malvín norte. http://www.noticias.uy/index.html

37.

Novak

K. J.

Hartman

J. L.

Holsinger

A. M.

Turner

M. G.

(1999). The effects of aggressive policing of disorder on serious crime. Policing, 22(2), 171–194. https://doi.org/10.1108/13639519910271229

38.

Nunn

Quinet

Rowe

Christ

(2006). Interdiction day: Covert surveillance operations, drugs, and serious crime in an inner-city neighborhood. Police Quarterly, 9(1), 73–99. https://doi.org/10.1177/1098611105274467

39.

Ousey

G. C.

Lee

M. R.

(2004). Investigating the connections between race, illicit drug markets, and lethal violence, 1984–1997. Journal of Research in Crime and Delinquency, 41(4), 352–383. https://doi.org/10.1177/0022427803262059

40.

Reuter

MacCoun

R. J.

Patrick

Allan

Simon

(1990). Money from crime a study of the economics of drug dealing in Washington, D.C. CA: Rand, Santa Monica.

41.

Phillips

S. W.

Wheeler

Kim

D.-Y.

(2016). The effect of police paramilitary unit raids on crime at micro-places in Buffalo, New York. International Journal of Police Science & Management, 18(3), 206–219. https://doi.org/10.1177/1461355716660568

42.

Reuter

(1986). Risks and prices: An economic analysis of drug enforcement. Crime and Justice, 7(1986), 289–340. https://doi.org/10.1086/449116

43.

Shepard

E. M.

Blackley

P. R.

(2005). Drug enforcement and crime: Recent evidence from New York state. Social Science Quarterly, 86(2), 323–342. https://doi.org/10.1111/j.0038-4941.2005.00305.x

44.

Sherman

L. W.

(1990). Police crackdowns: Initial and residual deterrence. Crime and Justice, 12(1990), 1–48.

45.

Sherman

L. W.

Weisburd

(1995). General deterrent effects of police patrol in crime “hot spots”: A randomized, controlled trial. Justice Quarterly, 12(4), 625–648. https://doi.org/10.1080/07418829500096221

46.

Smith

M. R.

(2001). Police-led crackdowns and cleanups: An evaluation of a crime control initiative in Richmond, Virginia. Crime and Delinquency, 47(1), 60–83. https://doi.org/10.1177/0011128701047001003

47.

Werb

Rowell

Guyatt

Kerr

Montaner

Wood

(2011). Effect of drug law enforcement on drug market violence: A systematic review. International Journal of Drug Policy, 22(2), 87–94. https://doi.org/10.1016/j.drugpo.2011.02.002