Why Bahar and Hausmann Tell Us Nothing About Venezuelan Migration Flows to the United States
Abstract
Bahar and Hausmann (2025a) claim to find evidence against the hypothesis that oil sanctions on Venezuela lead to increased migration flows to the United States. We show that their findings derive from applying a nonstandard, misspecified Engle-Granger test to first differences. This specification is incorrect because cointegration tests are designed to evaluate relationships between the levels of variables, not their first differences. Since the residuals from regressions of I(0) variables will, under general conditions, be stationary, testing for cointegration between first differences of I(1) variables virtually ensures a spurious finding of cointegration. Using Monte Carlo simulations, we show that the misspecified Bahar-Hausmann test on first differences exhibits a false positive rate of 100 percent. Once the Engle-Granger test is applied correctly to the logarithms of levels, the evidence of cointegration vanishes. The Bahar-Hausmann regressions therefore provide no valid basis for inference about any underlying relationship between migration and Venezuelan oil revenues.
Summary
This paper critiques a study by Bahar and Hausmann (2025a) which claimed to find evidence against the hypothesis that oil sanctions on Venezuela lead to increased migration flows to the United States. Rodríguez and Bravo argue that Bahar and Hausmann's conclusion is based on a flawed application of the Engle-Granger cointegration test. Specifically, Bahar and Hausmann incorrectly applied the test to first differences of the variables (Venezuelan oil income and US border encounters of Venezuelan nationals) instead of the levels, which is the standard and appropriate procedure for I(1) variables. The authors demonstrate, both theoretically and through Monte Carlo simulations, that this misspecification virtually guarantees a spurious finding of cointegration. Their theoretical proof shows that the probability of falsely rejecting the null hypothesis of no cointegration approaches 1 as the sample size increases. Their Monte Carlo simulations, using a sample size comparable to Bahar and Hausmann's, show a false positive rate of 100%. When the Engle-Granger test is correctly applied to the logarithms of the levels of the variables, the evidence of cointegration disappears. The paper concludes that Bahar and Hausmann's regressions provide no valid basis for inference about the relationship between migration and Venezuelan oil revenues. This matters to the field because it highlights the importance of proper econometric methodology and demonstrates how a flawed application of a standard test can lead to incorrect conclusions, particularly in the context of policy-relevant research.
Key Insights
- •Bahar and Hausmann incorrectly applied the Engle-Granger cointegration test to first differences of I(1) variables instead of their levels.
- •The authors theoretically prove that applying the Engle-Granger test to first differences of non-cointegrated I(1) variables leads to a false positive rate approaching 100% as the sample size increases.
- •Monte Carlo simulations with N=48 (similar to Bahar & Hausmann) confirm a 100% false positive rate when the misspecified Engle-Granger test is used.
- •When the Engle-Granger test is correctly applied to the logarithms of the levels of Venezuelan oil income and US border encounters, no evidence of cointegration is found.
- •The authors also highlight the inconsistency of testing for cointegration in untransformed variables and then estimating empirical specifications premised on cointegration in the logarithms of those variables.
- •Bahar and Hausmann's own replication code reveals that the test statistic is -7.923 and not -7.903 as stated in the paper.
- •Diagnostic Dickey-Fuller and augmented Dickey-Fuller tests for the levels of the variables show inconclusive evidence of nonstationarity.
Practical Implications
- •This research serves as a cautionary tale for applied econometricians, emphasizing the importance of carefully considering the assumptions and proper application of statistical tests, especially cointegration tests.
- •Researchers studying migration patterns and the impact of economic sanctions should be aware of the potential pitfalls of using flawed econometric techniques.
- •Practitioners analyzing time series data should carefully examine the properties of their data (e.g., stationarity, order of integration) and choose appropriate econometric methods.
- •The paper opens up avenues for future research to explore the true relationship between Venezuelan oil revenues, sanctions, and migration flows, using more rigorous and appropriate econometric techniques.
- •The critique encourages more careful scrutiny and replication of published research, particularly in politically sensitive areas.