Variable Selection for Qualitative Interactions

In this article we discuss variable selection for decision making with focus on decisions regarding when to provide treatment and which treatment to provide. Current variable selection techniques were developed for use in a supervised learning setting where the goal is prediction of the response. These techniques often downplay the importance of interaction variables that have small predictive ability but that are critical when the ultimate goal is decision making rather than prediction. We propose two new techniques designed specifically to find variables that aid in decision making. Simulation results are given along with an application of the methods on data from a randomized controlled trial for the treatment of depression.     

The effect of long-term migration dynamics on population structure in England & Wales and Scotland

We investigated the effect of migration on population dynamics in England & Wales and Scotland from the mid-nineteenth century to the present by comparing actual population size and structure with estimates based on zero net migration from a range of starting dates. In this period, Scotland had the largest net outflow among countries in Europe for which detailed information is available, whereas overall net migration in England & Wales was close to zero. In the absence of migration, population would have been over twice as large in Scotland in 2013 as the actual value, but similar to its actual value in England & Wales. Levels and pace of population ageing have been broadly similar in both countries, so the major impact of differential migration has been on population size rather than structure. We discuss these findings in relation to the debate on migration policy between political parties supporting and opposing independence in the 2014 Scottish referendum.     

Son preference in rural China: patrilineal families and socioeconomic change

This article draws on a survey conducted in six provinces in summer 2008 to investigate the determinants of son preference in rural China. The analysis confirms the conventional wisdom that son preference is embedded within patrilineal family structures and practices. We extend our analysis by exploring specific aspects of variation within patrilineal family culture. We find that the patrilineal group (clan) composition of villages and family participation in practices such as building ancestral halls and updating genealogies significantly influence son preference. Yet even though son preference is embedded within patrilineal family culture, our analysis suggests that over time the attenuation of son preference is likely. This is because determinants associated with socioeconomic change—for instance, higher levels of education, direct exposure to official policy education materials, higher income (a proxy for rural industrialization), and agricultural mechanization—all attenuate son preference. Being younger and female are also associated with weaker son preference, and both characteristics are likely to interact with education and industrialization to further dilute son preference in the longer term. Nevertheless, our findings suggest that concerted efforts are needed to ameliorate institutional discrimination against rural people in welfare provisioning and in labor markets, and to promote multiple dimensions of gender equality, including in land rights, wage rates, and education.