Fertility and Parental Labor-Force Participation: New Evidence from a Developing Country in the Balkans

Get Complete Project Material File(s) Now! »

Women’s Justice Centers Program

The 1994 Inter-American “Belem do Pará » Convention on “Prevention, Punishment, and Eradication of Violence against Women” significantly expanded Latin America’s definition of domestic and sexual violence. As a consequence, many countries in the region modified or enacted new legislation incorporating those issues into their political agenda. In particular, Peru altered its Police and Justice System’s jurisdiction to encompass domestic and sexual violence complaints. This new legal framework paired with the government’s awareness of the country’s high levels of domestic violence led in 1999 to the creation of the women justice centers (WJCs) –“Centros de Emergencia para Mujeres”– by the Peruvian Ministry for Women and Vulnerable Populations (MIMP) as part of the National Program against Sexual and Family Violence.
The women’s justice centers (WJC) are free of charge public centers that aim to strengthen the justice system’s capacity to detect, process and assist victims of domestic and sexual violence from an inter-disciplinary approach that includes legal, social and psychological dimensions. Basically, incoming victims receive a service designed to integrate all steps of the complaint process (e.g. police station, attorney’s office and medical doctor) in a single office in order to reduce as much as possible the time dedicated to issue the complaint and to follow the legal procedure in the corresponding court of justice. Hence, WJCs are frequently located within a short distance from partner establishments such as police stations, prosecutors’ offices and In addition to assisting incoming victims, WJC center’s aim is also to undertake a local violence prevention program. The prevention component intends to identify, control and reduce the risk factors. In this regard, the WJC centers have put in practice courses for training justice promoters –“facilitadoras en accion » and “promotores juveniles”–, which are volunteer women that advocate and execute campaigns, talks, workshops and seminaries to raise awareness about the problem of domestic violence in their region. Lastly, WJCs keep a record of cases that allow for monitoring and evaluation of the persistence of domestic and sexual violence (MIMDES, 2007).
The first women’s justice center opened in the District of Lima in 1999. During 1999-2014, the number of centers has grown from 13 in the first year to 226 by the end of 2014, covering 100% of the 24 regions of Peru and 96% of the provinces (188 of 196 provinces). Figure 1.1 shows the distribution and growth of the opening of the WJC centers over time. Whereas WJCs centers opened gradually throughout the first years of implementation, the program expanded exponentially after 2006. Up to that year, the average opening rate was about 6 WJCs/year; from 2006 to 2014 it augmented to 22 WJCs/year. Such escalation was provoked by a 2006 decentralization decree that granted local governments the right to open their own WJCs at the district level.
From a geographical coverage point of view, by 2014 most of the WJCs are concentrated in Metropolitan Lima and Lima Provinces (31 WJCs); in the Callao region there are 4 WJCs; the rest of the coastal region have 46 WJCs; in the sierra region there are 117 WJCs and in the jungle region there are 28 WJCs (Figure 1.2). Given the before-mentioned strong ties to local justice and health institutions, WJCs are highly located within urban areas.
According to MIMP’s statistics, the number of domestic violence cases registered in the WJC centers has increased substantially: from 29,759 in 2002 to more than 60,000 in 2016 (See Figure 1.3). Whereas 40% of reported cases are from women between 25 and 45 years old, children and teenagers (0-17 years old) constitute the second largest group – 30%. Additionally, a 2006-2008 survey administered by MIMP on 51 WJCs revealed that for the majority of the 10 The service provided in these centers is staffed by representatives of various government institutions such as police officers, prosecutors, counsellors, psychologists and public welfare agents whose objective is to help the victims of domestic abuse (MIMDES, 2007).
Albeit these evaluations, the program’s effectiveness remains unclear. MIMP’s statistics lack a rigorous study of the causality of factors and its mechanisms. Although globally, WJCs seem to have a positive effect on curbing the incidence of domestic violence, it remains uncertain if such effect is due to the centers’ ability to act as catalysts for women empowerment –which indirectly could enhance women’s capacity to care for their children through improved social protection and access to justice. Additionally, given our detailed micro-level of analysis, our results could become instrumental to understand the effectiveness of similar programs implemented in other Latin American countries.

Measuring Exposure to the WJC Centers

In order to be able to match the data on WJC centers with the the outcomes of interest, we construct two measures of exposure to the program: (i) WJC center within a 1km Euclidean buffer of the DHS cluster/school and (ii) WJC center in the district of the DHS cluster/school. The first measure uses the GPS coordinates of the DHS clusters/schools in order to measure a 1 kilometer Euclidean distance buffer from every DHS cluster/school location. For this method, the Euclidean buffer of 1km is first centered on each DHS cluster/school and then each DHS cluster/school is linked to a WJC center if the WJC center falls within the buffer, without consideration of district administrative borders. For instance, a DHS cluster/school located within 1km of a WJC center founded in 2008 is coded as having a WJC center within 1km of the DHS cluster/school since 2008. Figure 1.5 shows a visual representation of the Euclidean buffers for two specific regions in Peru: Lima and Tumbes.
The second measure matches the presence of a WJC center in the district, based on its date of opening and location, with the DHS cluster’s/school’s district. For instance, a DHS cluster/school in the district of Lima (150101) with a WJC center introduced in 2006 is coded as having a WJC center in the district of Lima since the year 2006.
Our preferred measure is the one that uses the Euclidean buffer since we want to estimate the impact of having a WJC center in the neighborhood of the school/household. The second measure is used as a robustness check because it might not always capture accurately the impact of the WJC centers due to the fact that districts in Peru have very different sizes. Therefore, rather than aggregating WJC center exposure in the district, we measure exposure based on how far the centers are from respective households, such that individuals residing at different points in the same district may have different levels of exposure to the WJC centers. Panel A of Tables 1.1 and 1.2 and Panel B of Table 1.3 show descriptive statistics of exposure to the WJC centers at the individual (women and children) and school level. The main reason for our choice of a 1km distance buffer instead of a larger buffer is not only because we believe that these centers have a very localized effect, but also because the measure of exposure using a 5km Euclidean buffer seems to be very similar to the one that uses presence of WJC center in the district. We present results using both measures of exposure to a WJC center principally for our main outcomes of interest.

Placement of WJC centers

A central methodological issue in our analysis is the fact that WJC centers are not placed randomly across the country. Even though our analysis will take advantage of variation over time, which will account for any fixed differences across districts and schools, it still remains important to understand what drives placement since placement decisions may not be orthogonal to other factors that could affect women’s and children’s outcomes of interest.
We address this concern in a number of ways which lead us to believe that the link between the opening of the WJC centers and the outcomes of interest is casual. First, we had several discussions with the Peruvian policymakers and WJC center managers about the location choices. Since the foundation of the first WJC center in 1999 till the end of 2005, the primary criteria they cited when deciding where to locate were population density and level of infrastructure at the regional level. In this stage, capitals and large cities were prioritized locations to open a WJC center. Starting from 2006, after the decentralization process which transferred the responsibility of the WJC centers to the local governments (districts), the Peruvian policymakers decided to open new WJC centers at the district level and they incorporated additional criteria such as proximity to police stations, district attorney offices (known as fiscalias) and health establishments.
Even though program guidelines suggested that priority should be given to poorer districts with sufficient judicial and medical infrastructures, in several occasions, political representatives had certain autonomy in deciding the order in which districts received the program. There is also anecdotical evidence from the authorities that WJC center’s placement has been primarily developed taking into account the population density but failed to take into account the rate of incidence of violence against women. This is likely due to the lack of reliable data on domestic violence or femicides for all the districts in Peru prior to the opening of the centers.
For instance, official data on femicides in Peru started to be recorded since 2009.20 Moreover, our conversations with the Peruvian policymakers suggest that educational considerations, and in particular enrollment rates or schooling performance, were never factored into program placement decisions.
Second, we are able to evaluate this endogenous placement statistically using our data. To do this we estimate at the district level: (a) the determinants of having a WJC center by the end of the sample in 2014 and (b) the determinants of adding a WJC center during 2006-2014, which is the period when the program grew substantially. We focus on several variables at the district level cited by the Peruvian policymakers such as: number of justice courts, number of district attorney offices, number of police stations and number of health establishments. We also control for district population at baseline and department fixed-effects. Moreover, in order to verify that education patterns before the program do not predict where the WJC centers are introduced, we also control for pre-program changes in primary and secondary school enrollment at the district level. Unfortunately, we are unable to perform the same test for self-reported domestic violence or femicides due to lack of pre-program data on these variables for all the districts in Peru. However, we control for baseline (self-reported) domestic violence at the district level by using the 2000 Peruvian DHS which contains a representative sample of 700 districts in Peru.

READ  Allocating the Ni-Vanuatu RSE-workers

School Level Specification

Lastly, using the same identification strategy, we study the overall effect of WJC centers on education outcomes at the school level by using the following regression equation: Yst = β0 + β1W JCst + αs + λ pt + γt Xs′ + ε st (1.3).
where (Yst) is the education outcome (i.e. total number of children enrolled, standardized test scores) in school s at year t, (W JCst) is an indicator variable that takes the value of one if the school has a WJC center within 1km/in the district of the school, (αs) is a school fixed-effect, (λ pt) is a province-by-year fixed-effect, (γt Xs′ ) is a year-interacted vector of school’s initial characteristics (including initial school enrollment, presence of electricity, presence of piped water, school language (Spanish), urbanization and public school dummy) and (ε st) is a random error term. The inclusion of school fixed-effects accounts for any time-invariant characteristics at the school level. We also allow the year fixed-effects to differ by province and by measures of school’s baseline enrollment and baseline infrastructure. Since initially-different schools might be more likely to change differently, this empirical specification focuses on comparing changes in treatment and control schools with similar initial characteristics that might drive WJC center placement.
The coefficient of interest is (β1), which captures the average change in enrollment in schools that are located near the WJC centers or in districts with WJC center, to the average change in enrollment in schools that did not have a WJC center. The identification assumption is that treatment schools located in the proximity of a WJC center/in districts with WJC center would otherwise have changed similarly, on average, to those controls schools that are not exposed to the services of a WJC center. In practice, by controlling for province-by-year fixed-effects (λ pt) and by variables that drive WJC center placement, the identification assumption is that treatment schools would otherwise have changed similarly, on average, to control schools within their same province and with similar initial characteristics. Throughout this analysis, we cluster our standard errors at the school level. We also estimate this regression including district-specific time trends.
Nevertheless, we are concerned about the possibility that the results are driven by time-varying variables which might influence both the opening of the WJC centers and school enrollment. A related issue is the possibility that WJC center managers consciously decide to introduce centers where school enrollment is increasing. To address both of these issues, we use the panel nature of the school data in order to construct a placebo treatment based on the timing of the WJC centers introduction. We estimate whether future WJC centers predict current enrollment using equation 1.4 below: Yst = β0 + β1W JCst + β2W JCst+1 + β3W JCst+2 + β4W JCst+3 + αs + λ pt + γt Xs′ + ε st (1.4).
where (W JCst+1), (W JCst+2) and (W JCst+3) are indicator variables that takes the value of one if the school has a WJC center within 1km/in the district of the school starting from the year t + 1, t + 2 and t + 3. If β2 > 0, β3 > 0 and β4 > 0 are positive and significant, this would indicate that WJC centers are being introduced in areas where schooling is increasing more rapidly. While, if β2 = β3 = β4 = 0 this would indicate that WJC centers are introduced in areas in which school enrollment is growing for other reasons.24 Therefore, the coefficients β2, β3 and β4 effectively capture the effect of future openings for areas that are not covered by the WJC centers in t. Our hypothesis for the placebo regression is that total enrollment in schools that do not have a WJC center within 1km/in the district should not be affected by the fact that a WJC center may open in the future in the proximity of these schools.

Impact of WJC Centers on Gender-Based Violence

We begin by investigation the impact of WJC center’s introduction on the incidence of gender-based violence against women. From estimating equation 1.1 for the sample of women, Table 1.5 presents the results of regressing the likelihood of experiencing domestic violence (by the intimate partner) in the last 12 months against the presence of a WJC center within 1km/in the district after controlling for several covariates, district fixed-effects, district-specific time trends and province-by-year fixed effects.
Panel A of Table 1.5 shows our domestic violence estimates when exposure to the program is measured through the presence of a WJC center within a 1km Euclidean buffer. Column 1 presents our results using the entire sample of women.25 Introducing a WJC center within 1km of the women’s residence decreases domestic violence by 2.2 percentage points, which represents a 5.6% decrease in domestic violence. Column 2 shows this regression after including district-specific trends to address the concern that districts that have a WJC center are trending differently than those that do not. This coefficient is slightly smaller (1.8 percentage points) but still significant. Our preferred specification is shown in Columns 3, in which we limit the sample to just urban clusters, which means that control areas are most comparable to those which are affected by the introduction of a WJC center. Even though this reduces the sample significantly, the coefficient is a bit higher in magnitude to the overall sample (2.9 percentage points) and highly significant. Lastly, Column 4 limits further to areas that ever have a WJC.

Table of contents :

1 Access to Justice, Gender Violence and Children: Evidence from Women’s Justice Centers in Peru 
1.1 Introduction
1.2 Background
1.2.1 Gender-based violence in Peru
1.2.2 Women’s Justice Centers Program
1.3 The Data
1.3.1 Individual and Household Level Data
1.3.2 School Level Data
1.3.3 District Level Data
1.3.4 Measuring Exposure to the WJC Centers
1.4 Empirical Strategy
1.4.1 Placement of WJC centers
1.4.2 Individual Level Specification
1.4.3 District Level Specification
1.4.4 School Level Specification
1.5 Results
1.5.1 Impact of WJC Centers on Gender-Based Violence
1.5.2 Impact of WJC Centers on Children’s School Attendance
1.5.3 Impact of WJC Centers on School Enrollment
1.6 Discussion: Mechanisms
1.7 Robustness Checks
1.7.1 Assessing the Internal Validity of the Research Design
1.7.2 Accounting for the dynamic impact of WJC centers
1.8 Conclusion
2 Fertility and Parental Labor-Force Participation: New Evidence from a Developing Country in the Balkans 
2.1 Introduction
2.2 Background
2.2.1 The Evolution of Total Fertility Rate in Albania
2.2.2 Labor-Force Participation in Albania
2.3 The Data
2.3.1 Descriptive Statistics
2.4 Empirical Strategy: Instrumental Variables
2.4.1 Son Preference in Albania and Siblings Sex-Composition
2.4.2 The Relevance Condition
2.4.3 The Exclusion Restriction
2.4.4 The Econometric Framework
2.5 Results
2.5.1 First-Stage
2.5.2 Main Results: OLS and Second-Stage
2.6 Discussion
2.6.1 Two plausible mecanisms
2.6.2 Heterogeneity Analysis
2.7 Conclusion
2.A Data sources and data construction
2.B LATE without monotonicity
2.C Theoretical Framework
2.C.1 Maximization Problem
2.D Supplementary Tables
3 Returning Home After Conflict Displacement: Labor Supply and Schooling Outcomes Among Kosovar Households 
3.1 Introduction
3.2 Background
3.2.1 Kosovo War (1998-1999)
3.2.2 Forced Displacement from Kosovo
3.3 The Data
3.3.1 Measuring Conflict Displacement
3.3.2 Descriptive Statistics
3.4 Empirical Strategy
3.4.1 Identification: Instrumental Variables Approach
3.4.2 First-Stage Estimation
3.4.3 Isolating Plausibly Exogenous Variation
3.4.4 Robustness Checks
3.5 Results
3.5.1 Conflict Displacement and Labor Market Outcomes
3.5.2 Conflict Displacement and School Enrollment Outcomes
3.6 Mechanisms
3.6.1 Channels on Labor Outcomes
3.6.2 Channels on Education Outcomes
3.7 Conclusion
3.A Suplementary Figures and Tables
3.B Alternative Identification Strategy
3.B.1 Identification: Difference-in-Difference (DID) Strategy
3.B.2 Endogeneity in Municipality Displacement
3.B.3 Displacement and Schooling Completion Outcomes


Related Posts