The Hidden Cost of Violent Confmict: Sorting into Local Labor - PowerPoint PPT Presentation

The Hidden Cost of Violent Confmict: Sorting into Local Labor Markets A Field Experiment in Colombia May 4, 2019 1 IHS, Vienna 2 University of Göttingen 3 DICE, University of Düsseldorf 1 / 25 Kerstin Grosch 1 Marcela Ibanez 2 Gerhard Riener 3

Motivation Confmicts have devastating consequences for development and growth. • What is the impact of confmict on labor markets? • Do individuals sort out of confmict affected areas? • Who sorts out? • Are there gender differences in sorting? • Do confmict affected areas lose the more qualifjed individuals? • How do previous experiences of confmict and violence shape sorting? • Can higher wages reduce the sorting effects? 2 / 25

Contribution Most research considers impact of internal displacement. • Displaced population higher probability unemployed or self-employed (Bozzoli, Brück, Wald, 2013; Kondylis, 2010) • Negative shock on wages and employment (Calderon and Ibañez, 2009) • Children are more likely to enter labor market (Rodriquez and Sanchez, 2012) Our contribution • First causal evidence sorting into life risk jobs using experimental approach • Asses the cost of confmict in terms of loss of talent 3 / 25

Local context: Colombia 4 / 25

Regional heterogeneity 5 / 25

Experimental Design Research assistants to work in two research projects: Santander or Putumayo. • Stage 1: Statement of interest (2206 job-seekers) • Stage 2: Randomization (797) • Stage 3: Application (383 applicants) • Stage 4: Hiring (3) 6 / 25

Experimental Design 1. Job Type • Low Risk: Garzon, Huila and San Gil, Santander (homicides 17.11 per 100.000) • High Risk: Puerto Asis, Putumayo and Vista Hermosa, Meta (homicides 63.48 per 100.000) 2. Salary for High Risk Job • Base: 1.5m COP 500€ • High: 1.8m COP 600€ 7 / 25

Regional variation 8 / 25

Results: Descriptive Statistics • The proportion of female job seekers is 56% • The average job seeker is 28-years-old. • About 10% of job seekers hold a master’s degree. • On average, participants have three years of relevant job experience. • A large share, nearly 90%, is medically insured. • Roughly 50% are vaccinated against tetanus and hepatitis B. • Good balance across treatments 9 / 25

Results: Sorting into labor markets Figure 1: Application Rates Base vs Risky low wage p=0.001; Base vs Risky high wage p = 0.057 (Fisher exact) 10 / 25

Research Question Are women less likely to apply to risky jobs? • Women are more risk averse than men (Eckel, Grossman, 2008). • This will be refmected in all aspects of their decision making. • Magnitude of gender differences is relatively small (Filippin and Crosetto, 2016) What are the consequences on sorting in labor markets? 11 / 25

Results: Gender effect 12 / 25

Research Question Is sorting associated with loss of talents? Life-risk might deter job seekers with higher opportunity cost, i.e. higher qualifjcation). Where T indicates our treatments and Z the qualifjcation/outside option of the candidate 13 / 25 Completed i = α + β 1 T i + β 2 Z i + β 3 T i × Z i + ǫ i

Results: Qualifjcation level • We use data from the World Bank STEPS survey, 2014 to estimate Mincer equation. • Two step Heckman (1979) sample selection model. • Out-of-sample prediction on job seekers in the study. • Explanatory variables: experience, area of study, and level of studies • Exclusion restriction: experience squared 1. Probability of being employed 2. Wage Equation: Gronau’s (1974) model 14 / 25

Results: Estimated Expected Earning 15 / 25

Estimation strategy: Quality and early life risk exposure • Linear probability model, dependent variable: completion of application process • Explanatory variables: Job treatment interacted with gender, predicted earnings and index of risk exposure (Homicides at city of mother when average job seeker aged 6) • Department fjxed effects • Clustered standard errors on the current city level 16 / 25

Results: Marginal effects Risk Job base salary 17 / 25

Results: Marginal effects Risk Job high salary 18 / 25

Results: Exposure to violence and application to Risky Job base salary • What are the long term consequences of crime? • Does exposure to violence affects sorting? 19 / 25

Results: Exposure to violence and application to Risky Job base salary 20 / 25

Results: Exposure to violence and application to Risky Job base salary 21 / 25

Results: Exposure to violence and application to Risky Job high salary 22 / 25

Conclusions • We fjnd that jobs with a high fatality risk discourage job seekers from applying. • A higher wage can compensate for the risk exposure and can stimulate application rates. • We do not detect any gender difference in application rates for the risky jobs. • Negative sorting out of low and very high qualifjed individuals. • Negative sorting out of low risk participants 23 / 25

Results: Qualifjcations Civil state: Partnership=1 Civil state: Single Mother=1 (0.093) -0.804*** Civil state: Separated=1 (0.218) -0.643*** (0.052) (0.210) -0.744*** Civil state: Divorced=1 (0.034) -0.643*** Civil state: Married=1 (0.090) -0.675*** Constant Civil state: Single=1 yes 0.319 (p-value) 1.003 Diff. risky jobs F-stat 1097 Obs. Experience 1.486*** yes Age yes Master yes Departments (0.369) -0.807*** (0.368) (1) -0.115*** 0.008 Vulnerability (centered) (0.021) -0.030 Predicted wage (centered) (0.042) Female 0.024 (0.048) -0.092* Risky: High salary (0.059) -0.139** Risky: Low salary (0.016) (0.057) -0.392 (0.027) Bogota (0.030) -0.007 (0.019) -0.023 0.022 (0.022) 0.015 (0.091) 0.044 24 / 25 Risky: Low salary × Female Risky: High salary × Female Risky: Low salary × Predicted wage (centered) Risky: High salary × Predicted wage (centered) Risky: Low salary × Vulnerability (centered) Risky: High salary × Vulnerability (centered)

Randomization table 0.054 (0.005) (0.005) 1105 0.716 0.005 0.011 0.008 Civil state: Single Mother (0.011) (0.012) (0.008) 1105 0.105 0.049 0.025 (0.000) Civil state: Separated (0.004) (0.005) (0.005) 1105 0.880 0.005 0.008 0.008 Civil state: Partnership (0.013) (0.013) (0.015) 1105 (0.004) (0.004) 0.068 (0.075) 369 369 367 N (0.079) (0.063) (0.067) 889 0.953 1.129 1.116 1.098 Homicides 1990: City mother (0.078) (0.071) (0.000) (0.013) Homicides 2010: Current city 0.380 0.362 0.356 0.361 1086 (0.012) 771 (0.012) Homicides 1990: City father 1.034 1.118 1.056 0.712 0.651 0.070 (1) (0.016) 27.699 27.512 Age (0.352) (0.200) (0.235) 1105 0.534 2.664 2.970 3.076 Experience (0.017) (0.016) 1105 0.799 0.752 0.117 0.100 0.104 Master N p-value Risky high wage Risky low wage Base (5) (4) (3) (2) 27.388 1105 0.084 0.431 Civil state: Divorced (0.018) (0.020) (0.019) 1105 0.370 0.859 0.824 0.853 Civil state: Married (0.006) (0.008) (0.008) 1105 0.014 (0.342) 0.995 (0.318) (0.333) Bogota 0.362 0.366 0.363 1105 0.027 (0.025) (0.025) (0.025) (0.022) Civil state: Single 0.022 25 / 25