lt is robust to alternative research designs. First, we apply alternative measures of provincial public health threats and obtain consistent inferences. Second, we add a falsification test by generating a pseudo-threat. We do not find a si