Determinants of the Outcomes of U.S. Senate Elections

Though House election outcomes have received much attention in recent years, comparable attention has not been paid to the Senate. We test a model of Senate seat change that includes changes in per capita real disposable income, presidential popularity, and party control of seats at stake in an election. The data consist of all Senate elections since 1920. A partial model explains 38% of the adjusted variation, and the full model almost 50%. The full model is then used to predict results for the 1986 elections over various levels of presidential popularity and changes in real disposable income.