Replication

A study published in PLOS ONE finding that cold showers strongly decrease work absence has implausible descriptive statistics, uses an incorrect statistical model, and has undeclared conflicts of interest.

A study published in the Journal of Population Economics finding large effects of criminalization of prostitution on rape estimates an impossible RDD specification, which effectively identifies a seasonal effect rather than a discontinuity.

A study published in the Journal of Economic Geography on the political effects of perceived immigration in Italian municipal elections constructs a shift-share exposure index that is mechanically correlated with municipality size. Controlling for electorate size renders all main results insignificant.

A study published in AER Insights finding a large effect of welfare benefits on non-EU migration to Denmark reports an ill-defined elasticity, conflates net migration with migrant residents turning 30, and disregards the origin country composition of migration flows.

A study published in the British Journal of Political Science ("Null by design") arguing that regional immigration–crime studies lack statistical power calibrates its power simulation on residual crime volatility that comes almost entirely from immigration-law offences in a few counties on the Austrian border. Without these offences, which record the act of migration rather than offending by immigrants, the design reaches 80% power at immigrant–native crime rate ratios below those observed for asylum seekers.

A study published in the Economic Journal finding that refugees assigned to more hostile German regions converge culturally toward natives faster relies on index components, chiefly risk preferences, that can respond to local hostility for reasons other than assimilation, and clusters standard errors by respondent although the treatment varies across only 38 regions. Dropping risk preferences cuts the estimate by more than half, and with clustering by region the main estimate is no longer significant.

A study published in the Journal of Political Economy finding that surname diversity causally raises innovation in US counties uses a diversity measure that rises with population growth, while patents are scaled by fixed 1900 population, and an instrument that is almost exactly the logarithm of the number of distinct surnames counted in the same census.