The ideological gap separating the Republican and Democratic parties in Congress has grown dramatically wider in recent decades. An analysis of the presidential vote in congressional districts over the last 60 years finds that the degree to which most districts are different from the “average” district has grown, supporting the theory that polarization stems from geographic clustering.
Over the last 40 years, the geographic distribution of the American electorate has become more clustered with respect to party voting, college attainment, median family income, and the marriage rate. This is responsible for an estimated 30 percent of the increase in polarization in the House of Representatives over time.
Large efficiency gains and substantial reduction in omitted variable bias are demonstrated in an application to sociodemographic differences in the risk of child obesity estimated from two nationally representative cohort surveys.
This research challenges the assumption that the timing of deployments and their distribution over time are serially independent, arguing that military interventions occur in temporal clusters driven by the number of interventions in the recent past.
This report challenges the assumption that the timing of deployments and their distribution over time are serially independent, arguing that military interventions occur in temporal clusters driven by the number of interventions in the recent past.
There is a bias-variance tradeoff at work in propensity score estimation; every step toward better balance usually means an increase in variance and at some point a marginal decrease in bias may not be worth the associated increase in variance.
Ownership of risky assets rises as the risk of medical expenditure decreases. Medigap or employer supplemental insurance increases risky asset holding by about 7 percentage points. Belonging to a Medicare HMO increases risky asset holding by 13 percentage points.
Examines implications for statistical power of using predicted probabilities for a dichotomous variable rather than the actual variable. Information loss from not observing actual values can be large. Direct substitution is easy and efficient.