-
Appendix E - A Structural Model Approach
- W.E. Upjohn Institute
- Chapter
- Additional Information
193 Appendix E A Structural Model Approach This appendix uses sequential equations to posit structural models for support of full employment legislation in the House. It is intuitively clear that the forces of constituency, economic factors, personal ideology, party, and electoral security all did not converge simultaneously in 1945 to influence the outcome of the full employment bill. The economic traits of the district and the constituency characteristics already had affected the type of person elected to Congress, particularly the question of whether the representative was a Democrat or Republican. These district characteristics may also have a further effect on the likelihood that the representative supported the full employment bill, an effect that is distinct from the party of the member, as the earlier analysis of those who deviated from their party’s position suggests. The most appropriate models of support for full employment are those that approach it as a sequential process. The traits of the district are used to predict the party of the candidate who won the most recent election—i.e., the current member of Congress. In turn, the member’s party identification is used to predict his or her position on full employment. Of course, this set of relationships is more complex than this simple statement implies, and it captures other variables as well, such as the 1937 unemployment rate’s influence upon the party of the person representing the district as well as upon his or her support for full employment. One way of depicting this set of relationships is presented in Figure E.1. This general model assumes that the influence of the district traits—economic and constituency—upon party moves in one direction. That is, the district traits affect party, but party does not affect the district traits. Moreover, the general model assumes that electoral security measures and biographical status of the member do not affect the party of the representative elected by the district. Finally, this model assumes that support for the full employment bill does not “feed back” on the district, electoral, or personal characteristics. In other words, support for full employment does not influence the district, electoral, or personal characteristics that were measured before the vote.1 In such models, a long-standing method used in various disciplines involves structural equations. Essentially, this approach is based upon a series of multiple regression equations that refines the information on each exogenous variable’s direct and indirect effects on the dependent variable.2 Another advantage of using structural equations is that the method enables us to address In order to view this proof accurately, the Overprint Preview Option must be checked in Acrobat Professional or Adobe Reader. Please contact your Customer Service Representative if you have questions about finding the option. Job Name: -- /347091t 194 Wasem the push for full employment as a sequential process and, thus, to place support for full employment in a context that comes a bit closer to the actual dynamics. As the structural equation model is constructed, party becomes an endogenous variable—i.e., one that is dependent on external forces. The multiple regression of party upon the districts’ constituency and economic variables reveals several interesting relationships. Wealth, as measured by the adjusted living level, is strongly and negatively associated with a Democratic representative . The percentage of the district’s population that is African American and the percentage that is foreign-born are also good predictors of electing a Democratic member of Congress. The districts’ percentage of working-class citizens and percentage of urban change are significant as well. Several interesting patterns emerge in the first structural equation model (Model A) when the indirect effects of the district traits through party are compared with these traits’ direct effects on support for full employment (Table E.1). Percentage working-class, which has one of the highest simple correlation coefficients with full employment, appears to have more of an indirect effect through party than a direct effect on the full employment scale. Most of the effect of urban change is felt through party, as are the effects of race and ethnicity. The unemployment rate, on the other hand, has some direct effect on support for full employment over and above its effect on party. Figure E.1 Basic Model of Support for Full Employment: Model A Constituency Economic Party Electoral Personal % foreign born % African American % urban change median education % working class 1937 unemployment rate adjusted living level tenure in Congress margin of victory New Deal realignment age biographical...