Date of Award
Doctor of Philosophy (PhD)
This dissertation project investigates how economic conditions influence presidential support in the United States. In particular, I focus on uncovering the temporal and causal dynamics to better understand the topic.
Of the three papers of this dissertation, the first paper develops a Bayesian machine learning method and uses it to examine the temporal dynamics between macro-level income growth and presidential approval. I develop an estimation algorithm to fit multivariate time series models via the Bayesian adaptive lasso, a machine learning-based estimator that penalizes "unimportant" lagged variables and mitigates the problem of overfitting. This new methodological tool allows analysts to employ large-scale time series and a multitude of lagged terms. Consequently, it aids in discovering lagged policy effects or inertia of dynamic relationships, which have so far been difficult to theorize or test. The application of the method to quarterly data for presidential approval ratings uncovers substantial lagged effects and positive long-run effects of income growth.
The second paper conducts a causal mediation analysis to explore two causal mechanisms that shape the effect of local unemployment on presidential voting: retrospective voting and issue-ownership voting. In an individual-level mediation analysis of the 2008, 2012, and 2016 presidential elections, this paper presents evidence that both mechanisms were at work in these most recent elections. The incumbent party, Democrat or Republican, is punished when local unemployment is rising, through its influence on retrospective evaluations of the national economy. Once this mediation effect representing retrospective voting is accounted for, local unemployment bolsters support for Democratic presidential candidates and drives down support for Republican candidates, implying that the two parties' distinct reputations for the unemployment issue engender issue-ownership voting.
Finally, the third paper formulates two partisan mechanisms that might moderate the relationship between macroeconomic conditions and presidential approval: retrospective and prospective partisan mechanisms. I test this new theoretical framework against quarterly data for multiple economic indicators and presidential approval from 1964 to 2015. The effect of unemployment squares with the prospective partisan mechanism: in response to deteriorating job conditions, citizens reward Democratic presidents and punish Republican presidents with an expectation that Democratic presidents will deal better with the unemployment issue. The effects of inflation, economic growth, and income growth are best explained by the retrospective partisan mechanism: only Republican presidents are punished for deteriorating conditions in terms of these three economic indicators, suggesting that these economic issues are salient for the Republican party and citizens hold Republicans more accountable for performance on these issues.
Chair and Committee
Jeff Betsy . Gill Sinclair
Roman Garnett, Jacob M. Montgomery, Andrew J. Reeves, Steven S. Smith,
Park, Taeyong, "Temporal and Causal Dynamics between the Economy and Presidents" (2017). Arts & Sciences Electronic Theses and Dissertations. 1260.
Available for download on Wednesday, December 15, 2117
Permanent URL: https://doi.org/10.7936/K7ZG6RNN