Wright - Contextual Models of Electoral Behavior: The Southern Wallace Vote

This paper seeks to resolve the inconsistency in studies of the black concentration hypothesis in the context of the 1968 George Wallce election.

One study finds that percentage black was correlate of the Wallace vote and suggested this demonstrates contextual effect. But other studies did not find strong support between black concetration and higher Wallace vote.

Problems in the Use of Aggregate Data
Wright blames inability of previous studies to validate or reject the black conentration hypothesis by using only aggregate data. To the extent that blacks are a proportion of all voters, previous studies inappropriately equate the vote for Wallace with the number of white voters for Wallace.

Individual level surveys have a hard time picking up contextual effects.

A Method of Contextual Analysis
Authors state need to synthesize multiple data files, contextual information must be integrated with data on individual attitudes and behavior.

Study uses county as the contextual unit.

The dependent variable is vote for or against George Walllace.

Measuring the indepedent variable - black concentration - uses black concentration in 1940 because it catches adult population in formative years of adolescence. Past contextual influence matters for later presidential vote.

Local Black Concentration and Region
What about stuides where aggregate data finds opposite correlations in black concentration and Wallace vote in the Deep and Marginal south?

We have to models to figure this out:

Model One - explaining vote for Wallace as a function of the single variable of local black concentration

Model Two - explaining vote for Wallace as a function of local black concentration AND a dummy variable for the effects of region.

Model Three - allows slopes of Wallace voting to vary within regions - the marginl south and deep south.

State Black Concentration
Model Three takes into account regional differences in Wallace vote. State political systems may be influenced by proportions of blacks at that level, just as with more local measures. A probable mechanism for state contextual influence is exposure to the levels of racism in the electoral campaigns for state office. States in the deep south have more openly racist campaigns than in the marginal south.

Model Four: Wallace support a function of local black concentration differenially related within each region, state balck concentration, and region. This model adds to the explanatory power of Model III. It adds the addition of state proportion black to Model III.

Model Five: remove the rgion variable and force a single local black concentration slope for both regions. Doing so does not significantly weaken explanatory power from Model IV.

Although the Deep South did give higher support to wallace, this can be expalined in terms of white responses to black concentration at the local and particularly state levels.

Model Six: a final model tests for spuriousness with several other possible explanatory variables and doesn't have any more explanatory power than Model Five.

Primary Groups and Issue Proximity
What are the mechanisms for linking contextual factors to voting behavior?

(1) Peer group support for Wallace (see Putnam social interaction article)

(2) Increasing the relative size of the black population results in greater racial hostility among whites, resulting in higher Wallace vote.

So context may be one factor, what about the voter's actual issue proximity to Wallace? Issue proximity and social context lead to vote choice. Based on data analysis, adding both issue proximity and the contextual factors gives a very strong correlation with Wallace vote

---Once you add both issue stance and primary group context into a model, you get a very strong correlation, suggesting an intervening effect.

Summary
The black concentration hypothesis appears to be correct. There is some lag between attitude change and changes in the environent. Race has a contextual effect at the stae and community levels. It is useful to posit hypotheses of alternative models and then empirically testing the differents in the models to reach empirical explantions. We should be synthesizing aggregate and local data.