Tag Archives: Bayesian

BEYOND THE RUNNING TALLY

BEYOND THE RUNNING TALLY: PARTISAN BIAS IN POLITICAL PERCEPTIONS

L. M. Bartels

Political Behavior, Vol. 24, No.2 Special Issue: Parties and Partisanship, Part One (Jun., 2002)

A Summary

In a Nutshell

The Michigan framework posited that partisan loyalties are formed early in life and remain stable throughout adulthood and they serve as the determinants of more specific political attitudes. This means that party-identification on attitudes toward politics is far more important than the influence of these attitudes on party identification itself (Campbell).

More recent work has described a running tally of retrospective evaluations of party promises and performance as being the driving force behind party identification (Fiorina), and this is consistent with the view that voters maximize their expected utility based on past political experience.

An essay by Gerber and Green argue that whilst political beliefs do change, including evaluations on the performance of the economy, they change to approximately the same degree amongst those with different political allegiances. Thus “biased learning” (coloured by partisanship) appears to have little support. For these authors Bayesian learning means that people with substantially the same prior information, but different partisan affinities, will learn from new information in a similar direction and to a similar extent, and the evidence for this is that opinion shifts (mapped from survey data) are largely parallel i.e. unaffected by partisan bias.

What Bartels argues is that within the Bayesian framework parallel shifts are impossible unless partisan bias is built into the groups, as if it were not there would be a convergence of opinion. Thus empirical evidence that suggests that shifts in opinion are parallel are evidence against unbiased information processing.  Thus it is the failure to converge that needs explanation.

Bartels points to a number of survey responses to political issues and examines trends in opinion of both Democrats and Republicans. He shows that the parallel trends in opinion over management of the Gulf War, Bush’s performance and economic conditions can only be explained by a significant partisan bias, otherwise the result would have been convergence of opinion. Furthermore, whilst opinion on certain policies may present a bias accounted for by the intrinsic value differences between the two groups it is harder to argue that opinion over e.g. The Gulf War, are driven by value differences between the groups. The case for the existence of partisan bias is even stronger when opinions differ over objective facts i.e. the level of unemployment/inflation. In such circumstance differing values cannot explain stark partisan differences in opinion.

In this regard he uses as an example unemployment under the Reagan administration which fell from 7.1% to 5.5%, and inflation which fell from 13.5% to 4.1%. In response to surveys 50% of Democrats thoughts that inflation had got worse, or somewhat worse, and only 8% thought it had got better. The results were 13% and 47% respectively for Republicans. This is evidence of substantial partisan biases in perceptions of how the country fared during the Reagan years. A similar story is told regarding the Clinton administration.

Thus the evidence suggests that partisan loyalties have pervasive effects on perceptions of the political world. In some cases this produces divergence of opinion, but in a great many more cases it significantly inhibits what would otherwise be a strong tendency to convergence. Thus partisanship is not based on a running tally, but is a dynamic force in opinion formation.

Advertisements