Some Problems in Studying the Effects of Resource Allocation in Congressional Elections

Abstract

After analyzing (1) data aggregated to the congressional district level, and (2) individual-level data from the 1978 CPS election survey, Johannes and McAdams conclude that congressional casework has no electoral impact. The following commentary explains such null findings as the product of oversimplistic expectations and methodological weaknesses. Specifically, the Johannes and McAdams aggregate data analysis is misspecified on two counts. First, it attempts to reduce to a single regression equation a temporal sequence in which casework activity and electoral outcomes are mutually intertwined. Second, even were such a drastic reduction possible, the single equation employed would be poorly specified because of inattention to the differential productivity of cases, constituents, and representatives. The individual-level analysis also is rife with statistical problems chief among which is multicollinearity aggravated by small numbers of cases. Analyses that do what is possible to minimize such problems reveal a statistically and substantively significant impact of casework on electoral outcomes. Further analyses that go beyond the Johannes and McAdams limitation of casework effects to the individuals directly helped suggest that the electoral effects of an incumbent's reputation for service may approach the effects of party identification.