On the estimation of the effect of stressors on birth outcomes

•The estimated effect of tornadoes on birth outcomes changed over time.•Attrition before birth can lead to better outcomes among the surviving infants.•The estimated effects of stressors on birthweight and prematurity have to be interpreted in the context of changes in mortality in-utero.

Using tornadoes as a source of random variation in exposure to stressors, this papers shows that failure to account for selection through attrition before birth, can lead to significant errors of interpretation of the results of commonly run regressions.