## Section 3–4: Empirical Strategy, Data, and Main Results

## Source details

**Canonical URL:** [Section 3–4: Empirical Strategy, Data, and Main Results](https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025100-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2025/english/wpiea2025100-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2025/english/wpiea2025100-print-pdf.pdf.json)

---

### Data
- Source: Mexico’s ENOE (nationally representative, rotating household panel) covering 2005–2019.
- Panel structure: individuals tracked for five consecutive quarters; estimation uses individual and time fixed effects and ENOE probability weights for national representation.
- Primary sample size: 16,530,279 individual-quarter observations spanning 2005–2019.
- Homicide data: municipal-level statistics published by INEGI; quarterly homicide rates per 100,000 inhabitants computed using population estimates from CONAPO.
- Additional geographic indicator: presence of drug-trafficking organizations (DTOs) from Sobrino et al. (2019) to identify regions with elevated violence.
- Informal employment definition: unprotected work in agricultural activities, unprotected paid domestic work, and subordinate workers who, although they work for formal economic units, work without social security.
- Separation data: from the first quarter of each year module on ever-separated and reasons for separation (demand vs supply); sector coding follows NAICS.
- Controls: household characteristics (marital status, number of children), receipt of government transfers; municipality-level pre-study 2005 sectoral employment shares (agriculture, manufactures, construction, services) interacted with quarter dummies.
- Gender analysis: analyses conducted for full sample and separately for men and women.

### Empirical Strategy
- Main estimation equation:
  y_im,t = γ_i + δ_t + β_1 HomicideRate_m,t + β_2 HomicideRate_m,t−1 + Γ′ X_m,t + I′ Z_i,t + ε_im,t
  - y_im,t: outcome variable (labor force participation, employment, job separations).
  - γ_i and δ_t: individual and quarter fixed effects.
  - HomicideRate_m,t and HomicideRate_m,t−1: natural logarithms of contemporaneous and lagged municipal homicide rates.
  - X_i,t: individual characteristics (marital status, number of children, whether receives government transfers).
  - Z_m,t: municipality controls, including 2005 sectoral employment shares interacted with quarter dummies.
- Estimation: whole sample and male/female subsamples; standard errors clustered at the individual level.
- Outcomes scaled: outcome dummies multiplied by 100 so coefficients interpreted as percentage points.

### Main Findings — Impact of Homicides on Labor Force Participation (LFP), Employment, and Informality
- Aggregate employment: largely unaffected by contemporaneous or lagged homicide rate increases; gender-specific effects are significant.
- Key coefficients (coefficients multiplied by 100; standard errors in parentheses):
  - Panel A. Employment:
    - HomicideRate_t: Total 0.008 (0.015); Men -0.009 (0.020); Women 0.021 (0.022).
    - HomicideRate_t−1: Total 0.007 (0.015); Men -0.035* (0.020); Women 0.045** (0.022).
    - R^2: Total 0.744; Men 0.706; Women 0.714.
    - Mean of dependent variable: Total 57.37; Men 75.34; Women 41.2.
  - Panel B. Labor Force Participation:
    - HomicideRate_t: Total 0.007 (0.014); Men -0.004 (0.018); Women 0.015 (0.022).
    - HomicideRate_t−1: Total 0.005 (0.014); Men -0.037** (0.018); Women 0.043** (0.022).
    - R^2: Total 0.759; Men 0.737; Women 0.714.
    - Mean of dependent variable: Total 59.87; Men 78.64; Women 43.1.
  - Panel C. Informality:
    - HomicideRate_t: Total 0.014 (0.017); Men 0.000 (0.025); Women 0.025 (0.022).
    - HomicideRate_t−1: Total 0.017 (0.017); Men -0.023 (0.025); Women 0.053** (0.022).
    - R^2: Total 0.656; Men 0.672; Women 0.604.
    - Mean of dependent variable: Total 33.24; Men 46.32; Women 24.1.
  - Observations: Total 16,530,279; Men 7,797,059; Women 8,733,220.
  - Notes: ** * p<0.1, ** p<0.05, *** p<0.01. Standard errors clustered at the individual level.
- Magnitude interpretation for Mexico 2005–2019:
  - Homicide rates increased by 247 percent during the study period.
  - This rise corresponds to a nationwide decline in LFP of approximately 0.22 percentage points for men and a 0.25 percentage point increase for women.
  - The rise in female participation and employment is driven by a movement of women into informal labor.
- Mechanism interpretation:
  - Evidence supports an “income-compensation mechanism” where one gender increases labor force participation to offset the decline of the other.
  - Lagged effects suggest gradual household adaptation ("wait-and-see" approach).
  - Shifts mainly reflect composition changes (men leaving, women entering informality) rather than large changes in aggregate employment.

### Labor Demand vs Supply Drivers (Table 2)
- Approach: classify ENOE non-participation reasons into demand-driven and supply-driven; reestimate main equation with outcome dummies for each motive (multiplied by 100).
- Selected coefficients and means:
  - Panel A. Labor Demand Reasons — "No jobs or other market reasons":
    - HomicideRate_t: Total -0.00048 (0.0047); Men 0.0014 (0.003); Women -0.0102 (0.008).
    - HomicideRate_t−1: Total 0.0036 (0.0047); Men 0.0082** (0.003); Women 0.000 (0.008).
    - R^2: Total 0.296; Men 0.251; Women 0.301.
    - Mean of dependent variable: Total 0.52; Men 0.15; Women 0.86.
  - Panel B. Other Personal Reasons (Supply):
    - HomicideRate_t: Total -0.0019 (0.0034); Men -0.0006 (0.0040); Women -0.0030 (0.0053).
    - HomicideRate_t−1: Total -0.0056* (0.0034); Men 0.0015 (0.0039); Women -0.0121** (0.0055).
    - R^2: Total 0.286; Men 0.289; Women 0.284.
    - Mean: Total 0.56; Men 0.36; Women 0.73.
  - Panel C. Caretaking Reasons:
    - HomicideRate_t: Total -0.0112 (0.0077); Men 0.00467 (0.0046); Women -0.0254* (0.01401).
    - HomicideRate_t−1: Total -0.0110 (0.0077); Men -0.0002 (0.0047); Women -0.0206 (0.0139).
    - R^2: Total 0.369; Men 0.292; Women 0.364.
    - Mean: Total 2.58; Men 0.42; Women 4.51.
  - Panel D. Public insecurity:
    - HomicideRate_t: Total 0.0004 (0.0003); Men 0.00005 (0.0002); Women 0.0007 (0.0006).
    - HomicideRate_t−1: Total -0.0002 (0.0004); Men -0.0002 (0.0003); Women -0.0001 (0.0007).
    - R^2: Total 0.265; Men 0.253; Women 0.268.
    - Mean: Total 0.00007; Men 0.00003; Women 0.00011.
  - Panel E. Age or Appearance:
    - HomicideRate_t: Total -0.0097** (0.0039); Men -0.0056 (0.0045); Women -0.0134** (0.0062).
    - HomicideRate_t−1: Total -0.0009 (0.0039); Men -0.0003 (0.0046); Women -0.0015 (0.0062).
    - R^2: Total 0.286; Men 0.290; Women 0.283.
    - Mean: Total 0.67; Men 0.48; Women 0.85.
  - Panel F. Education or Experience:
    - HomicideRate_t: Total 0.0007 (0.0016); Men -0.0001 (0.0017); Women 0.00134 (0.0025).
    - HomicideRate_t−1: Total 0.0012 (0.0015); Men 0.0020 (0.0016); Women 0.0005 (0.0025).
    - R^2: Total 0.274; Men 0.287; Women 0.267.
    - Mean: Total 0.11; Men 0.07; Women 0.14.
- Interpretation:
  - Demand-side: Significant increase in men citing demand-side constraints (HomicideRate_t−1 coefficient 0.0082** for men) — suggests declining male LFP is partially driven by reduced labor demand.
  - Supply-side: For women, higher lagged homicide rates associated with reductions in supply-side non-participation (Other Personal Reasons HomicideRate_t−1 = -0.0121** for women), indicating women enter the labor market to stabilize household income.
  - Public insecurity and caregiving are not reported as significant reasons for withdrawal in response to higher homicide rates; absence of self-reported insecurity effects suggests crime operates primarily as a negative labor demand shock for men.

### Heterogeneity: DTO Areas vs Non-DTO Areas (Table 3)
- Definition: DTO areas = municipalities where cartel presence was higher than the historical median weighted by employment (cartel presence from Sobrino et al., 2019).
- LFP responses by area type (coefficients and means):
  - Panel A. DTO Areas:
    - HomicideRate_t: Total 0.024 (0.033); Men -0.035 (0.042); Women 0.077 (0.050).
    - HomicideRate_t−1: Total 0.0075 (0.033); Men -0.102** (0.042); Women 0.107** (0.049).
    - R^2: Total 0.771; Men 0.761; Women 0.734.
    - Mean of dependent variable: Total 60.97; Men 77.34; Women 46.1.
  - Panel B. Non-DTO Areas:
    - HomicideRate_t: Total -0.005 (0.016); Men 0.005 (0.021); Women -0.015 (0.025).
    - HomicideRate_t−1: Total 0.0045 (0.016); Men -0.021 (0.021); Women 0.027 (0.025).
    - R^2: Total 0.755; Men 0.719; Women 0.701.
    - Mean: Total 58.77; Men 79.84; Women 40.0.
  - Observations: Total 16,530,279; Men 7,797,059; Women 8,733,220.
- Quantified impact in DTO areas (2005–2019):
  - Increase in the homicide rate corresponds to a decline in LFP of approximately 0.58 percentage points for men and an increase of 0.61 percentage points for women.
  - The 0.61 percentage point increase for women accounts for 26.7 percent of the total rise in female LFP in DTO areas over the same period.
  - The 0.58 percentage point decline for men explains 17.5 percent of the total reduction in male LFP in DTO areas.
- Interpretation:
  - Significant gender asymmetries concentrated in cartel-dominated (DTO) regions.
  - No significant LFP responses for either gender in non-DTO areas.
  - Employers in DTO areas may strategically reduce demand for male workers to mitigate perceived risks of employing males affiliated with DTOs or exposed to victimization.

### Mechanisms and Economic Implications
- Labor reallocation channel:
  - Rising homicides induce men to leave the labor force and women to enter (often into informal employment), producing compositional changes rather than large aggregate employment shifts.
  - Movement into informality implies low-cost-of-entry jobs with no formal contracts or benefits, associated with low productivity.
- Productivity and welfare:
  - Labor market misallocation is a critical channel through which crime undermines labor productivity.
  - The gendered pattern may reflect employer perceptions and affiliations with DTOs: perpetrators and victims of crime are overwhelmingly men, potentially making firms view male hires as higher risk.
- Policy-relevant observation:
  - Effects are localized to areas with DTO presence, underscoring the role of local institutional and economic conditions in shaping the labor market consequences of violence.

### Impact of Homicides on Job Separations (Section 4.2)
- Overview:
  - Investigates how rising local homicide rates affect job separations, distinguishing supply- versus demand-driven separations and gender-specific responses.
  - Uses an annual survey module (available in Q1 of each calendar year) with individual fixed effects and clustering at the municipality level.
  - Reestimates Equation 1 with individual fixed effects substituted for Mexican state-level fixed effects.
- Main empirical findings (Table 4):
  - Panel A — All separations:
    - HomicideRate_t: Total = 0.185** (0.072); Men = 0.164 (0.102); Women = 0.207*** (0.056).
    - HomicideRate_t−1: Total = 0.242*** (0.072); Men = 0.172* (0.099); Women = 0.279*** (0.057).
    - R^2: Total = 0.384; Men = 0.383; Women = 0.354.
    - Mean of dependent variable: Total = 67.2; Men = 55.4; Women = 76.3.
    - Interpretation: The observed increase in the homicide rate during 2005–2019 corresponded to an increase in job separation by about 1.4 percentage points overall; women experienced a larger increase (1.6 percentage points) versus men (1.0 percentage point).
  - Panel B — Separation for supply reasons:
    - HomicideRate_t: Total = 0.195*** (0.068); Men = 0.195** (0.088); Women = 0.179*** (0.067).
    - HomicideRate_t−1: Total = 0.229*** (0.067); Men = 0.233*** (0.087); Women = 0.226*** (0.065).
    - R^2: Total = 0.195; Men = 0.227; Women = 0.141.
    - Mean of dependent variable: Total = 50.9; Men = 36.0; Women = 61.4.
    - Interpretation: Supply-side factors dominate; increases in homicide rates raise separations driven by supply-side motives (safety concerns, worsening workplace conditions, insufficient compensation) and effects are consistent across genders.
  - Panel C — Separation for demand reasons:
    - HomicideRate_t: Total = -0.036 (0.027); Men = -0.031 (0.033); Women = -0.036 (0.032).
    - HomicideRate_t−1: Total = -0.060** (0.025); Men = -0.073** (0.031); Women = -0.061** (0.031).
    - R^2: Total = 0.095; Men = 0.174; Women = 0.064.
    - Mean of dependent variable: Total = 6.9; Men = 7.5; Women = 6.5.
    - Interpretation: Separations for demand-side reasons (layoffs, employer terminations) decline following increases in homicide rates.
  - Observations (Table 4): Total sample = 6,322,760; Men = 2,759,926; Women = 3,562,834.
  - Notes: Significance indicators: *** p<0.01, ** p<0.05, * p<0.1. Standard errors clustered at the municipality level.
- Labor margin distinction:
  - Job separations and labor force participation capture different margins: separations reflect transitions within employment (job-to-job moves, temporary exits into unemployment), while labor force participation measures entry into or exit from the labor market.

### Sectoral decomposition of separations (Table 5)
- Definition:
  - "Origin" sector = sector of the firm a worker separated from.
  - "Destination" sector = sector where workers who separated are currently employed.
- Selected sectoral patterns and exact coefficients:
  - Panel A — Construction (Mean: Origin Total = 0.26; Destination Total = 0.48; Origin Men = 0.53; Destination Men = 0.99; Origin Women = 0.02; Destination Women = 0.03)
    - HomicideRate_t: Total Origin = -0.0101 (0.0072); Total Destination = -0.0106 (0.0101).
    - HomicideRate_t: Men Origin = -0.0191 (0.0159); Men Destination = -0.0200 (0.0224).
    - HomicideRate_t: Women Origin = -0.0005 (0.0013); Women Destination = 0.0033** (0.0016).
    - HomicideRate_t−1: Total Origin = -0.0124* (0.0069); Total Destination = -0.0022 (0.0106).
  - Panel B — Manufacturing (Mean: Total Origin = 0.40; Total Destination = 0.90; Men Origin = 0.45; Men Destination = 1.34; Women Origin = 0.36; Women Destination = 0.50)
    - HomicideRate_t: Total Origin = -0.010 (0.009); Total Destination = 0.003 (0.017).
    - HomicideRate_t−1: Total Origin = 0.001 (0.008); Total Destination = 0.026 (0.018).
    - HomicideRate_t−1: Women Destination = 0.029** (0.014).
  - Panel C — Trade (Mean: Total Origin = 0.47; Total Destination = 1.02; Men Origin = 0.42; Men Destination = 1.31; Women Origin = 0.51; Women Destination = 0.75)
    - HomicideRate_t: Total Origin = 0.021*** (0.008); Total Destination = 0.067*** (0.019).
    - HomicideRate_t: Men Origin = 0.028*** (0.010); Men Destination = 0.111*** (0.028).
    - HomicideRate_t: Women Origin = 0.016 (0.010); Women Destination = 0.035** (0.016).
    - HomicideRate_t−1: Total Origin = 0.023*** (0.008); Total Destination = 0.050*** (0.018).
    - HomicideRate_t−1: Men Origin = 0.039*** (0.010); Men Destination = 0.064** (0.027).
    - HomicideRate_t−1: Women Origin = 0.010 (0.011); Women Destination = 0.041*** (0.015).
  - Panel D — Services (Mean: Total Origin = 0.85; Total Destination = 2.03; Men Origin = 0.78; Men Destination = 2.71; Women Origin = 0.91; Women Destination = 1.41)
    - HomicideRate_t: Total Origin = 0.017 (0.011); Total Destination = 0.020 (0.030).
    - HomicideRate_t: Men Origin = 0.032** (0.014); Men Destination = 0.037 (0.043).
    - HomicideRate_t: Women Origin = 0.005 (0.014); Women Destination = 0.014 (0.024).
    - HomicideRate_t−1: Total Origin = 0.035*** (0.011); Total Destination = 0.086*** (0.030).
    - HomicideRate_t−1: Men Origin = 0.019 (0.014); Men Destination = 0.113*** (0.042).
    - HomicideRate_t−1: Women Origin = 0.048*** (0.014); Women Destination = 0.067*** (0.025).
  - Panel F — Agriculture (Mean: Total Origin = 0.23; Total Destination = 0.51; Men Origin = 0.38; Men Destination = 0.98; Women Origin = 0.10; Women Destination = 0.09)
    - HomicideRate_t: Total Origin = -0.028** (0.011); Total Destination = -0.002 (0.021).
    - HomicideRate_t: Men Origin = -0.053*** (0.020); Men Destination = -0.033 (0.041).
    - HomicideRate_t: Women Origin = -0.005 (0.009); Women Destination = 0.023** (0.010).
    - HomicideRate_t−1: Total Origin = 0.007 (0.011); Total Destination = 0.004 (0.020).
    - HomicideRate_t−1: Women Origin = 0.015* (0.009); Women Destination = 0.005 (0.009).
  - Observations (Table 5): Total sample = 6,322,760; Men = 2,759,926; Women = 3,562,834.
  - Notes: Significance indicators: *** p<0.01, ** p<0.05, * p<0.1. Standard errors clustered at the municipality level.
- Sectoral interpretation:
  - Rising homicide rates increase turnover into trade and services.
  - Construction shows minimal sensitivity to violence-driven separations.
  - For men: elevated homicide rates associated with higher likelihood of leaving trade jobs and taking new trade jobs—indicative of labor reallocation within the trade industry.
  - For women: separations concentrate in service-oriented transitions; homicide rate increases both the likelihood of separating from service jobs and taking new service jobs—suggesting women move toward service sector jobs that may offer higher perceived safety and stability.

### Mechanisms and implications (job separations)
- Supply versus demand channel:
  - Evidence indicates supply-side drivers dominate: workers, especially women, adjust labor supply to mitigate exposure to violence (safety concerns, workplace conditions, compensation).
  - Demand-side separations (layoffs, employer terminations) decline following higher homicide rates.
- Labor reallocation and misallocation:
  - Homicides prompt reallocation of labor across firms and sectors driven by worker preferences for non-pecuniary job attributes rather than skill alignment.
  - Such preference-driven shifts increase mismatches between workers’ skills and job assignments, potentially reducing overall labor productivity.
  - The results provide a micro-foundation for the documented negative effects of violence on labor productivity.

### Summary conclusions (chapter statements)
- The rise in violence leads to a significant increase in job separations, particularly among women, who are more likely to leave jobs due to worsening work conditions and increased risks.
- Women’s increased labor force participation can act as a compensatory mechanism to offset declines in male labor demand, but the reallocation toward more informal or safety-perceived sectors suggests negative effects on aggregate labor productivity.
- Violence exacerbates inefficiencies in labor markets via misallocation, reinforcing its role as a barrier to productivity growth.

*Source: wpiea2025100-print-pdf (Sections 3–4).*

### Section 3 outlines the empirical strategy employed in the analysis.  Section 4 presents the main

### Section 3–4: Empirical Strategy, Data, and Main Results

### Data
- Source: Mexico’s ENOE (nationally representative, rotating household panel) covering 2005–2019.
- Panel structure: individuals tracked for five consecutive quarters; estimation uses individual and time fixed effects and ENOE probability weights for national representation.
- Primary sample size: 16,530,279 individual-quarter observations spanning 2005–2019.
- Homicide data: municipal-level statistics published by INEGI; quarterly homicide rates per 100,000 inhabitants computed using population estimates from CONAPO.
- Additional geographic indicator: presence of drug-trafficking organizations (DTOs) from Sobrino et al. (2019) to identify regions with elevated violence.
- Informal employment definition: unprotected work in agricultural activities, unprotected paid domestic work, and subordinate workers who, although they work for formal economic units, work without social security.
- Separation data: from the first quarter of each year module on ever-separated and reasons for separation (demand vs supply); sector coding follows NAICS.
- Controls: household characteristics (marital status, number of children), receipt of government transfers; municipality-level pre-study 2005 sectoral employment shares (agriculture, manufactures, construction, services) interacted with quarter dummies.
- Gender analysis: analyses conducted for full sample and separately for men and women.

### Empirical Strategy (Section 3)
- Main estimation equation (individual and time fixed effects, controls for supply and demand):
  y_im,t = γ_i + δ_t + β_1 HomicideRate_m,t + β_2 HomicideRate_m,t−1 + Γ′ X_m,t + I′ Z_i,t + ε_im,t
  - y_im,t: outcome variable (labor force participation, employment, job separations).
  - γ_i and δ_t: individual and quarter fixed effects.
  - HomicideRate_m,t and HomicideRate_m,t−1: natural logarithms of contemporaneous and lagged municipal homicide rates.
  - X_i,t: individual characteristics (marital status, number of children, whether receives government transfers).
  - Z_m,t: municipality controls, including 2005 sectoral employment shares interacted with quarter dummies.
- Estimation: whole sample and male/female subsamples; standard errors clustered at the individual level.
- Outcomes scaled: outcome dummies multiplied by 100 so coefficients interpreted as percentage points.

### Main Findings — Impact of Homicides on Labor Force Participation (LFP), Employment, and Informality (Section 4.1)
- Aggregate employment appears largely unaffected by contemporaneous or lagged homicide rate increases, but gender-specific effects are significant.
- Key summary statistics and coefficient highlights from Table 1:
  - Panel A. Employment (coefficients multiplied by 100; standard errors in parentheses):
    - HomicideRate_t: Total 0.008 (0.015); Men -0.009 (0.020); Women 0.021 (0.022).
    - HomicideRate_t−1: Total 0.007 (0.015); Men -0.035* (0.020); Women 0.045** (0.022).
    - R^2: Total 0.744; Men 0.706; Women 0.714.
    - Mean of dependent variable: Total 57.37; Men 75.34; Women 41.2.
  - Panel B. Labor Force Participation:
    - HomicideRate_t: Total 0.007 (0.014); Men -0.004 (0.018); Women 0.015 (0.022).
    - HomicideRate_t−1: Total 0.005 (0.014); Men -0.037** (0.018); Women 0.043** (0.022).
    - R^2: Total 0.759; Men 0.737; Women 0.714.
    - Mean of dependent variable: Total 59.87; Men 78.64; Women 43.1.
  - Panel C. Informality:
    - HomicideRate_t: Total 0.014 (0.017); Men 0.000 (0.025); Women 0.025 (0.022).
    - HomicideRate_t−1: Total 0.017 (0.017); Men -0.023 (0.025); Women 0.053** (0.022).
    - R^2: Total 0.656; Men 0.672; Women 0.604.
    - Mean of dependent variable: Total 33.24; Men 46.32; Women 24.1.
  - Observations: Total 16,530,279; Men 7,797,059; Women 8,733,220.
  - Notes: ** * p<0.1, ** p<0.05, *** p<0.01. Standard errors clustered at the individual level.
- Magnitude interpretation for Mexico 2005–2019:
  - Homicide rates increased by 247 percent during the study period.
  - This rise corresponds to a nationwide decline in LFP of approximately 0.22 percentage points for men and a 0.25 percentage point increase for women.
  - The rise in female participation and employment is driven by a movement of women into informal labor.
- Mechanism interpretation:
  - Evidence supports an “income-compensation mechanism” where one gender increases labor force participation to offset the decline of the other (cited parallel with Mishra et al., 2021).
  - Lagged effects suggest gradual household adaptation ("wait-and-see" approach).
  - Shifts mainly reflect composition changes (men leaving, women entering informality) rather than large changes in aggregate employment.

### Labor Demand vs Supply Drivers (Table 2)
- Approach: classify ENOE non-participation reasons into demand-driven and supply-driven; reestimate main equation with outcome dummies for each motive (multiplied by 100).
- Key findings (selected coefficients and means from Table 2):
  - Panel A. Labor Demand Reasons — "No jobs or other market reasons":
    - HomicideRate_t: Total -0.00048 (0.0047); Men 0.0014 (0.003); Women -0.0102 (0.008).
    - HomicideRate_t−1: Total 0.0036 (0.0047); Men 0.0082** (0.003); Women 0.000 (0.008).
    - R^2: Total 0.296; Men 0.251; Women 0.301.
    - Mean of dependent variable: Total 0.52; Men 0.15; Women 0.86.
  - Panel B. Other Personal Reasons (Supply):
    - HomicideRate_t: Total -0.0019 (0.0034); Men -0.0006 (0.0040); Women -0.0030 (0.0053).
    - HomicideRate_t−1: Total -0.0056* (0.0034); Men 0.0015 (0.0039); Women -0.0121** (0.0055).
    - R^2: Total 0.286; Men 0.289; Women 0.284.
    - Mean: Total 0.56; Men 0.36; Women 0.73.
  - Panel C. Caretaking Reasons:
    - HomicideRate_t: Total -0.0112 (0.0077); Men 0.00467 (0.0046); Women -0.0254* (0.01401).
    - HomicideRate_t−1: Total -0.0110 (0.0077); Men -0.0002 (0.0047); Women -0.0206 (0.0139).
    - R^2: Total 0.369; Men 0.292; Women 0.364.
    - Mean: Total 2.58; Men 0.42; Women 4.51.
  - Panel D. Public insecurity:
    - HomicideRate_t: Total 0.0004 (0.0003); Men 0.00005 (0.0002); Women 0.0007 (0.0006).
    - HomicideRate_t−1: Total -0.0002 (0.0004); Men -0.0002 (0.0003); Women -0.0001 (0.0007).
    - R^2: Total 0.265; Men 0.253; Women 0.268.
    - Mean: Total 0.00007; Men 0.00003; Women 0.00011.
  - Panel E. Age or Appearance:
    - HomicideRate_t: Total -0.0097** (0.0039); Men -0.0056 (0.0045); Women -0.0134** (0.0062).
    - HomicideRate_t−1: Total -0.0009 (0.0039); Men -0.0003 (0.0046); Women -0.0015 (0.0062).
    - R^2: Total 0.286; Men 0.290; Women 0.283.
    - Mean: Total 0.67; Men 0.48; Women 0.85.
  - Panel F. Education or Experience:
    - HomicideRate_t: Total 0.0007 (0.0016); Men -0.0001 (0.0017); Women 0.00134 (0.0025).
    - HomicideRate_t−1: Total 0.0012 (0.0015); Men 0.0020 (0.0016); Women 0.0005 (0.0025).
    - R^2: Total 0.274; Men 0.287; Women 0.267.
    - Mean: Total 0.11; Men 0.07; Women 0.14.
- Interpretation:
  - Demand-side: Significant increase in men citing demand-side constraints (HomicideRate_t−1 coefficient 0.0082** for men) — suggests declining male LFP is partially driven by reduced labor demand.
  - Supply-side: For women, higher lagged homicide rates associated with reductions in supply-side non-participation (e.g., Other Personal Reasons HomicideRate_t−1 = -0.0121** for women), indicating women enter the labor market to stabilize household income.
  - Public insecurity and caregiving are not reported as significant reasons for withdrawal in response to higher homicide rates; absence of self-reported insecurity effects suggests crime operates primarily as a negative labor demand shock for men.

### Heterogeneity: DTO Areas vs Non-DTO Areas (Table 3)
- Definition: DTO areas = municipalities where cartel presence was higher than the historical median weighted by employment (cartel presence from Sobrino et al., 2019).
- Table 3 highlights LFP responses by area type:
  - Panel A. DTO Areas (coefficients and means):
    - HomicideRate_t: Total 0.024 (0.033); Men -0.035 (0.042); Women 0.077 (0.050).
    - HomicideRate_t−1: Total 0.0075 (0.033); Men -0.102** (0.042); Women 0.107** (0.049).
    - R^2: Total 0.771; Men 0.761; Women 0.734.
    - Mean of dependent variable: Total 60.97; Men 77.34; Women 46.1.
  - Panel B. Non-DTO Areas:
    - HomicideRate_t: Total -0.005 (0.016); Men 0.005 (0.021); Women -0.015 (0.025).
    - HomicideRate_t−1: Total 0.0045 (0.016); Men -0.021 (0.021); Women 0.027 (0.025).
    - R^2: Total 0.755; Men 0.719; Women 0.701.
    - Mean: Total 58.77; Men 79.84; Women 40.0.
  - Observations: Total 16,530,279; Men 7,797,059; Women 8,733,220.
- Quantified impact in DTO areas (2005–2019 period):
  - Increase in the homicide rate corresponds to a decline in LFP of approximately 0.58 percentage points for men and an increase of 0.61 percentage points for women.
  - The 0.61 percentage point increase for women accounts for 26.7 percent of the total rise in female LFP in DTO areas over the same period.
  - The 0.58 percentage point decline for men explains 17.5 percent of the total reduction in male LFP in DTO areas.
- Interpretation:
  - Significant gender asymmetries concentrated in cartel-dominated (DTO) regions.
  - No significant LFP responses for either gender in non-DTO areas.
  - Employers in DTO areas may strategically reduce demand for male workers to mitigate perceived risks of employing males affiliated with DTOs or exposed to victimization.

### Mechanisms and Economic Implications
- Labor reallocation channel:
  - Rising homicides induce men to leave the labor force and women to enter (often into informal employment), producing compositional changes rather than large aggregate employment shifts.
  - Movement into informality implies low-cost-of-entry jobs with no formal contracts or benefits, associated with low productivity (cited literature).
- Productivity and welfare:
  - Labor market misallocation is a critical channel through which crime undermines labor productivity.
  - The gendered pattern may reflect employer perceptions and affiliations with DTOs: perpetrators and victims of crime are overwhelmingly men, potentially making firms view male hires as higher risk.
- Policy-relevant observation:
  - Effects are localized to areas with DTO presence, underscoring the role of local institutional and economic conditions in shaping the labor market consequences of violence.

*Source: wpiea2025100-print-pdf (Sections 3–4).*

### 4.2    Impact of Homicides on Job Separations

### 4.2    Impact of Homicides on Job Separations

### Overview
- Investigates how rising local homicide rates affect job separations, distinguishing supply- versus demand-driven separations and gender-specific responses.
- Uses an annual survey module (available in Q1 of each calendar year) with individual fixed effects and clustering at the municipality level.
- Reestimates Equation 1 with individual fixed effects substituted for Mexican state-level fixed effects.

### Main empirical findings (Table 4)
- Panel A — All separations:
  - HomicideRate_t: Total = 0.185** (standard error 0.072); Men = 0.164 (0.102); Women = 0.207*** (0.056).
  - HomicideRate_t−1: Total = 0.242*** (0.072); Men = 0.172* (0.099); Women = 0.279*** (0.057).
  - R^2: Total = 0.384; Men = 0.383; Women = 0.354.
  - Mean of dependent variable: Total = 67.2; Men = 55.4; Women = 76.3.
  - Interpretation: The observed increase in the homicide rate during 2005–2019 corresponded to an increase in job separation by about 1.4 percentage points overall; women experienced a larger increase (1.6 percentage points) versus men (1.0 percentage point).
- Panel B — Separation for supply reasons:
  - HomicideRate_t: Total = 0.195*** (0.068); Men = 0.195** (0.088); Women = 0.179*** (0.067).
  - HomicideRate_t−1: Total = 0.229*** (0.067); Men = 0.233*** (0.087); Women = 0.226*** (0.065).
  - R^2: Total = 0.195; Men = 0.227; Women = 0.141.
  - Mean of dependent variable: Total = 50.9; Men = 36.0; Women = 61.4.
  - Interpretation: Supply-side factors dominate; increases in homicide rates raise separations driven by supply-side motives (safety concerns, worsening workplace conditions, insufficient compensation) and effects are consistent across genders.
- Panel C — Separation for demand reasons:
  - HomicideRate_t: Total = -0.036 (0.027); Men = -0.031 (0.033); Women = -0.036 (0.032).
  - HomicideRate_t−1: Total = -0.060** (0.025); Men = -0.073** (0.031); Women = -0.061** (0.031).
  - R^2: Total = 0.095; Men = 0.174; Women = 0.064.
  - Mean of dependent variable: Total = 6.9; Men = 7.5; Women = 6.5.
  - Interpretation: Separations for demand-side reasons (layoffs, employer terminations) decline following increases in homicide rates.

- Sample size / observations:
  - Observations (Table 4): Total sample = 6,322,760; Men = 2,759,926; Women = 3,562,834.
- Notes:
  - Significance indicators: *** p<0.01, ** p<0.05, * p<0.1.
  - Standard errors clustered at the municipality level.
- Labor margin distinction:
  - Job separations and labor force participation capture different margins: separations reflect transitions within employment (job-to-job moves, temporary exits into unemployment), while labor force participation measures entry into or exit from the labor market.

### Sectoral decomposition of separations (Table 5)
- Definition:
  - "Origin" sector = sector of the firm a worker separated from.
  - "Destination" sector = sector where workers who separated are currently employed.
- Key patterns by sector and gender (selected coefficients reported exactly as in Table 5):
  - Panel A — Construction (Mean of dependent variable: Origin Total = 0.26; Destination Total = 0.48; Origin Men = 0.53; Destination Men = 0.99; Origin Women = 0.02; Destination Women = 0.03)
    - HomicideRate_t: Total Origin = -0.0101 (0.0072); Total Destination = -0.0106 (0.0101).
    - HomicideRate_t: Men Origin = -0.0191 (0.0159); Men Destination = -0.0200 (0.0224).
    - HomicideRate_t: Women Origin = -0.0005 (0.0013); Women Destination = 0.0033** (0.0016).
    - HomicideRate_t−1: Total Origin = -0.0124* (0.0069); Total Destination = -0.0022 (0.0106).
  - Panel B — Manufacturing (Mean of dependent variable: Total Origin = 0.40; Total Destination = 0.90; Men Origin = 0.45; Men Destination = 1.34; Women Origin = 0.36; Women Destination = 0.50)
    - HomicideRate_t: Total Origin = -0.010 (0.009); Total Destination = 0.003 (0.017).
    - HomicideRate_t−1: Total Origin = 0.001 (0.008); Total Destination = 0.026 (0.018).
    - HomicideRate_t−1: Women Destination = 0.029** (0.014).
  - Panel C — Trade (Mean of dependent variable: Total Origin = 0.47; Total Destination = 1.02; Men Origin = 0.42; Men Destination = 1.31; Women Origin = 0.51; Women Destination = 0.75)
    - HomicideRate_t: Total Origin = 0.021*** (0.008); Total Destination = 0.067*** (0.019).
    - HomicideRate_t: Men Origin = 0.028*** (0.010); Men Destination = 0.111*** (0.028).
    - HomicideRate_t: Women Origin = 0.016 (0.010); Women Destination = 0.035** (0.016).
    - HomicideRate_t−1: Total Origin = 0.023*** (0.008); Total Destination = 0.050*** (0.018).
    - HomicideRate_t−1: Men Origin = 0.039*** (0.010); Men Destination = 0.064** (0.027).
    - HomicideRate_t−1: Women Origin = 0.010 (0.011); Women Destination = 0.041*** (0.015).
  - Panel D — Services (Mean of dependent variable: Total Origin = 0.85; Total Destination = 2.03; Men Origin = 0.78; Men Destination = 2.71; Women Origin = 0.91; Women Destination = 1.41)
    - HomicideRate_t: Total Origin = 0.017 (0.011); Total Destination = 0.020 (0.030).
    - HomicideRate_t: Men Origin = 0.032** (0.014); Men Destination = 0.037 (0.043).
    - HomicideRate_t: Women Origin = 0.005 (0.014); Women Destination = 0.014 (0.024).
    - HomicideRate_t−1: Total Origin = 0.035*** (0.011); Total Destination = 0.086*** (0.030).
    - HomicideRate_t−1: Men Origin = 0.019 (0.014); Men Destination = 0.113*** (0.042).
    - HomicideRate_t−1: Women Origin = 0.048*** (0.014); Women Destination = 0.067*** (0.025).
  - Panel F — Agriculture (Mean of dependent variable: Total Origin = 0.23; Total Destination = 0.51; Men Origin = 0.38; Men Destination = 0.98; Women Origin = 0.10; Women Destination = 0.09)
    - HomicideRate_t: Total Origin = -0.028** (0.011); Total Destination = -0.002 (0.021).
    - HomicideRate_t: Men Origin = -0.053*** (0.020); Men Destination = -0.033 (0.041).
    - HomicideRate_t: Women Origin = -0.005 (0.009); Women Destination = 0.023** (0.010).
    - HomicideRate_t−1: Total Origin = 0.007 (0.011); Total Destination = 0.004 (0.020).
    - HomicideRate_t−1: Women Origin = 0.015* (0.009); Women Destination = 0.005 (0.009).

- Sectoral interpretation:
  - Rising homicide rates increase turnover into trade and services (Panels C and D).
  - Construction shows minimal sensitivity to violence-driven separations.
  - For men: elevated homicide rates associated with higher likelihood of leaving trade jobs and taking new trade jobs—indicative of labor reallocation within the trade industry.
  - For women: separations concentrate in service-oriented transitions; homicide rate increases both the likelihood of separating from service jobs and taking new service jobs—suggesting women move toward service sector jobs that may offer higher perceived safety and stability.
- Observations (Table 5): Total sample = 6,322,760; Men = 2,759,926; Women = 3,562,834.
- Notes:
  - Significance indicators: *** p<0.01, ** p<0.05, * p<0.1.
  - Standard errors clustered at the municipality level.

### Mechanisms and implications
- Supply versus demand channel:
  - Evidence indicates supply-side drivers dominate: workers, especially women, adjust labor supply to mitigate exposure to violence (safety concerns, workplace conditions, compensation).
  - Demand-side separations (layoffs, employer terminations) decline following higher homicide rates.
- Labor reallocation and misallocation:
  - Homicides prompt reallocation of labor across firms and sectors driven by worker preferences for non-pecuniary job attributes rather than skill alignment.
  - Such preference-driven shifts increase mismatches between workers’ skills and job assignments, potentially reducing overall labor productivity.
  - The results provide a micro-foundation for the documented negative effects of violence on labor productivity (reference to related literature framework preserved from source).

### Summary conclusions (as stated in the chapter)
- The rise in violence leads to a significant increase in job separations, particularly among women, who are more likely to leave jobs due to worsening work conditions and increased risks.
- Women’s increased labor force participation can act as a compensatory mechanism to offset declines in male labor demand, but the reallocation toward more informal or safety-perceived sectors suggests negative effects on aggregate labor productivity.
- Violence exacerbates inefficiencies in labor markets via misallocation, reinforcing its role as a barrier to productivity growth.

*Source: IMF Working Paper No. WP/2025/100 — Section 4.2 Impact of Homicides on Job Separations.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025100-print-pdf.pdf_