Modelling the determinants of swing ratio and bias in US House elections 1850–1980

Electoral data for the US Congress for the period 1850 to 1980 are analyzed. We distinguish two important features of the seats-votes relationship that are often confounded: swing ratio and partisan bias. The swing ratio and electoral bias are shown to be a product of the shape of the constituency partisan distribution (CPD). We confirm earlier research showing that the decline in the congressional swing ratio noted by Tufte (1973) and Calvert and Ferejohn (1984) among others can largely be attributed to a reduction in the number of competitive House seats which has been ongoing for at least 70 years and has, for the entire country, exhibited a near straight-line decline in moving average over virtually the entire period. But, following Gudgin and Taylor (1979), we also show that the swing ratio is responsive to the kurtosis of the distribution of Democratic vote share across congressional districts. Finally, we show that partisan bias is responsive to the skewness of this distribution.