The labor market effects of Venezuelan migration to Colombia: reconciling conflicting results†

Notes: Outcomes are residualized and multiplied by 100. Observations are weighted by city–year population. Column 3 presents a Hausman test for endogenous regressors with robust standard errors. Cluster-robust standard errors are in parentheses.2SLS, two-stage least squares; FE, fixed effect; K-P Wald stat., Kleibergen-Paap Wald statistic; OLS, ordinary least squares.

* p < 0.10,

** p < 0.05,

*** p < 0.01.

Notes: Outcomes are residualized and multiplied by 100. All models include Year FE and City FE. Observations are weighted by city–year–group population. Cluster-robust standard errors are in parentheses. FE, fixed effect; K-P Wald stat., Kleibergen-Paap Wald statistic; LFP, labor force participation.

* p < 0.10,

** p < 0.05,

*** p < 0.01.

Notes: Outcomes are residualized and multiplied by 100. All models include Year FE and City FE. See text for description of robustness checks. Observations are weighted by city–year–group population. Cluster-robust standard errors are in parentheses. FE, fixed effect; K-P Wald stat., Kleibergen-Paap Wald statistic; LFP, labor force participation.

* p < 0.10,

** p < 0.05,

*** p < 0.01.

Notes: All models are 2SLS linear probability models for probability of employment in a group (multiplied by 100), not conditional on working, with Year FE and City FE. Occupations are ranked according to mean education of natives pre-2015 and grouped into quintiles. Workers are underemployed if they say they would like to change jobs to improve use of skills or training. Observations are weighted by city–year–group population. Cluster-robust standard errors are in parentheses.2SLS, two-stage least squares; FE, fixed effect; K-P Wald stat., Kleibergen-Paap Wald statistic.

* p < 0.10,

** p < 0.05,

*** p < 0.01.

Notes: Models are 2SLS linear probability models for probability of employment in a group (multiplied by 100), not conditional on working, with Year FE and City FE. Wage-sector workers are split by formality, defined according to compliance with mandatory health and pension schemes. Self-employed workers are split into own account (with no employees, predominantly informal) and employer. Observations are weighted by city–year–group population. Cluster-robust standard errors are in parentheses. 2SLS, two-stage least squares; FE, fixed effect; K-P Wald stat., Kleibergen-Paap Wald statistic.

* p < 0.10,

** p < 0.05,

*** p < 0.01.

Notes: Sample includes the years 2014 and 2019. In- and outmigration rates are taken as shares of the 5-year lag population and multiplied by 100. All models include Year FE and City FE. Observations are weighted by city–year–group population. Cluster-robust standard errors are in parentheses.2SLS, two-stage least squares; FE, fixed effect; K-P Wald stat., Kleibergen-Paap Wald statistic; OLS, ordinary least squares.

* p < 0.10,

** p < 0.05,

*** p < 0.01.

Notes: Both the endogenous variable and the instrument are interacted with the interaction variable indicated in each row. All interaction variables are centered around zero and divided by the standard error. 2014 metro-level economic indicators are measured using the GEIH. Doing Business measures at the department level were taken from the World Bank 2017 Doing Business report. The values of 2014 per capita GDP at the department level downloaded from DANE. All models include Year FE and City FE. Observations are weighted by city–year population. Cluster-robust standard errors are in parentheses. DANE, National Administrative Department of Statistics; FE, fixed effect; GDP, gross domestic product; GEIH, Colombian National Integrated Household Survey; K-P Wald stat., Kleibergen-Paap Wald statistic; WB DB, World Bank Doing Business.

* p < 0.10,

** p < 0.05,

*** p < 0.01.

[i] Notes: Means are presented, SDs are in parentheses; data are restricted to urban residents of age 15–64 years. Migrant is defined as anyone in 2019 who was living in Venezuela 5 years ago. Population weights are applied. GEIH, Colombian National Integrated Household Survey; SD, standard deviation.

[ii] Source: GEIH (2019).

Notes: M_c,2005 is the 2005 Venezuelan share of the metro area, which ranges from 0% to 0.9%, and has been normalized to a mean of zero and SD of one. Observations are weighted by 2014 population. Robust standard errors are in parentheses. LPP, Labor force participation; SD, standard deviation.

* p < 0.10,

** p < 0.05,

*** p < 0.01.

Notes: The outcome, namely, log hourly wage, is residualized and multiplied by 100. All models include Year FE and Unit FE. Panel A includes Colombian-born migrants in the migrant share, while Panel B excludes them. Columns 2–4 use different share components of the shift-share IV, based respectively on the complete 2005 census (as in the main analysis), the 10% subsample of the 1993 census, and the inverse minimum driving distance from the metro area or department capital to the closest Venezuelan border crossing. Geographic units include the 79 CZ-defined metro areas (as in the main analysis), the 23 official metro areas of which the GEIH is representative, and the 24 departments of Colombia. Observations are weighted by unit–year population. Cluster-robust standard errors are in parentheses. CZ, commuting zone; FE, fixed effect; GEIH, Colombian National Integrated Household Survey; IV, instrumental variable; K-P Wald stat., Kleibergen-Paap Wald statistic; OLS, ordinary least squares.

* p < 0.10,

** p < 0.05,

*** p < 0.01.

Notes: See notes to Table A3. This model is a regression of the change in residual wages from 2014 to 2018 on the 2018 foreigner share, calculated using the GEIH or Census as a share of the 2018 population. The instrument is created analogously (interacted with the national migration rate from the GEIH or census). In no case does the K-P Wald statistic fall below 10. CZ, commuting zone; FE, fixed effect; GEIH, Colombian National Integrated Household Survey; IV, instrumental variable; K-P Wald stat., Kleibergen-Paap Wald statistic; OLS, ordinary least squares.

* p < 0.10,

** p < 0.05,

*** p < 0.01.

Notes: See notes to Table A3. This is the original two-way fixed-effects model, replacing the 5-year migrant share with the 1-year migrant share (the share of the population that arrived from Venezuela in the past 12 months) and varying the end of the period of analysis. In no case does the K-P Wald stat. fall below 10.

CZ, commuting zone; FE, fixed effect; GEIH, Colombian National Integrated Household Survey; IV, instrumental variable; K-P Wald stat., Kleibergen-Paap Wald statistic; OLS, ordinary least squares.

* p < 0.10,

** p < 0.05,

*** p < 0.01.

Notes: Outcomes are residualized and multiplied by 100. All models include Year FE and City FE. Column 4 regresses the change in outcomes from 2014 to 2019 on the change in the migrant share. Column 5 replaces the migrant share with the number of migrants (divided by 100) and controls for baseline population interacted with year FEs. Column 6 assigns natives to their 5-year lagged metro area of residence. Observations are weighted by city–year–group population. Cluster-robust standard errors are in parentheses. 2SLS, two-stage least squares; FE, fixed effect; K-P Wald stat., Kleibergen-Paap Wald statistic; LFP, labor force participation.

* p < 0.10,

** p < 0.05,

*** p < 0.01.

Notes: Linear and quadratic migrant shares are instrumented with the linear and quadratic forms of the IV. Wages are residualized and multiplied by 100. All models include Year FE and City FE. Observations are weighted by city–year–group population. Cluster-robust standard errors are in parentheses. 2SLS, two-stage least squares; FE, fixed effect; K-P Wald stat., Kleibergen-Paap Wald statistic.

* p < 0.10,

** p < 0.05,

*** p < 0.01.

Notes: Outcomes are multiplied by 100. All models are 2SLS linear probability models for probability of employment in a group, not conditional on working, with Year FE and City FE. Observations are weighted by city–year–group population. See paper for description of robustness checks. Cluster-robust standard errors are in parentheses. 2SLS, two-stage least squares; FE, fixed effect; Occup, occupational.

* p < 0.10,

** p < 0.05,

*** p < 0.01.

Notes: Sample includes the years 2014 and 2019. Migration rates are taken as a share of the 5-year lag population. All models include Year FE and City FE. Observations are weighted by city–year–group population. See paper for description of robustness checks. Cluster-robust standard errors are in parentheses. 2SLS, two-stage least squares; FE, fixed effect.

* p < 0.10,

** p < 0.05,

*** p < 0.01.

Notes: Each row presents the interaction coefficient from a separate regression using the indicated interaction variable.

† Indicates that the K-P Wald stat. is >16.

See notes to Table 7. Cluster-robust standard errors are in parentheses. 2SLS, two-stage least squares; GDP, gross domestic product; K-P Wald stat., Kleibergen-Paap Wald statistic; WB DB, World Bank Doing Business.

* p < 0.10,

** p < 0.05,

*** p < 0.01.

Notes: All outcomes are multiplied by 100. All models include Year FE and City FE. Observations are weighted by city–year–group population. Excluding border locations drops cities within 100 km driving distance of the Venezuelan border. Cluster-robust standard errors are in parentheses. 2SLS, two-stage least squares; FE, fixed effect.

* p < 0.10,

** p < 0.05,

*** p < 0.01.