The article looks at how smoking affects birth outcomes. Instead of using instrumental variables, the researchers used panel data from mothers with multiple births. By controlling for individual differences, they found different results compared to previous methods. They also considered potential issues with the data.