Resumen

The incidence of twins has been used to identify the impact of changes in fertility on measures of investment  in children born prior to the twins, and the emerging consensus in this literature is that there is no evidence of a quantity-quality trade-off. We argue that the standard approach is flawed. Even if twin conception is random, bringing twins to term is a function of maternal health which is difficult to fully observe and which tends to be correlated with child quality, rendering the instrument invalid. The neglect of this fact in the existing literature will tend to lead to under-estimation of the quantity-quality (Q–Q) trade-off and so could contribute to explaining the negative results in the literature. Our contention that women who produce twin births are positively selected is demonstrated using data from richer and poorer countries. Using large samples of microdata from developing countries and from the USA which include indicators of maternal characteristics including health, we show that a significant trade-off emerges upon correcting for these biases. We show that this result is likely to be only a lower bound of the true Q–Q trade-off and discuss how to estimate the size of these bounds. These results have important implications for twin studies in all contexts examined here.