Abstract:
In a comment to Karadja & Prawitz (2019), henceforth KP, Per Pettersson-Lidbom (2020), henceforth P-L, argues that the main results in KP are severely biased. He argues that KP's results are biased due to non-classical measurement error in emigration and due to confounders related to the instrument. In this response, we show that P-L's reasoning regarding measurement error bias contradicts the results from his proposed test. More generally, P-L's results cannot exclude alternative and arguably more likely explanations. We present two straightforward tests that both indicate that measurement error does not bias KP's results. Second, we argue that KP controls for confounders in a standard way given the identication strategy. Including fixed effects at the level of the exogenous cross-sectional variation, as P-L does, severely limits the available identifying variation and decreases precision. Nevertheless, we document that KP's results are robust to non-linear frost shock controls, including fixed effects for groups of similar frost shocks. In addition, we show that our results are robust to altering regional fixed effects or dropping them altogether, in contrast to what is suggested by P-L.
