Fertility treatments and the use of twin births as an instrument for fertility

Twin births are often used as an instrument for fertility in models investigating the impact of family size on labour market and child outcomes. However, a large share of twin births (24% in our sample) are the result of fertility treatments, potentially causing twin births to be endogenous and biasing estimates. Using data from the British Millennium Cohort Study we show that (a) mothers with and without fertility treatment are different, (b) twin births are still random after conditioning on fertility treatments, (c) both labour supply regressions and quantity-quality-tradeoff regressions for children’s outcomes relying on the twin birth instrument appear to be biased and (d) the bias makes it less likely to find any results.