Local Unemployment and Voting for President: Uncovering Causal Mechanisms

http://​www.​presidency.​ucsb.​edu/​ws/​index.​php?​pid=​45010.

http://​www.​presidency.​ucsb.​edu/​ws/​index.​php?​pid=​86815.

Indeed, this theoretical view of voter accountability underpins foundational views of presidents as seekers of the national good against local passions of members of congress (Kriner and Reeves 2015).

See Cho and Gimpel (2009) for a notable exception with respect to presidential voting and Rogers (2014) with respect to MSA-level unemployment and approval of President Obama.

The definition of what is local is, of course, subjective. Our focus is county-level though others have found evidence of state economic context and presidential voting (Abrams and Butkiewicz 1995; Ansolabehere et al. 2014; Ebeid and Rodden 2006).

Though we focus here on unemployment, others have found evidence that voters hold presidents accountable for factors such as local income growth, local federal spending, local loan defaults, changes in wages, or local relief spending in response to natural disasters (Gasper and Reeves 2011; Healy and Lenz 2017; Kriner and Reeves 2012).

Though see Books and Prysby (1999) and Rogers (2014).

Though we do not consider it here, others have examined the determinants of perceptions of the local economy itself. See Rogers (2016, 2014).

For an overview of this perspective see Bélanger and Nadeau (2014), Bélanger and Nadeau (2015), Martinsson (2009), Wright (2012), and Petrocik (1996).

Previous research on presidential approval found a similar pattern. Swank (1993) shows that the approval rate for Democratic presidents benefits from rising unemployment whilst their Republican counterparts are hurt by rising unemployment. Carlsen (2000) presents the same evidence. See also Powell and Whitten (1993) and Martinsson (2009) that discuss issue-ownership effects associated with economic voting in the cross-national and Swedish contexts, respectively.

Though related to issue ownership, an alternative explanation is that voters engage in policy oriented-economic voting (Kiewiet 1981), where voters have preferences for particular policies and vote for the party the believe to be more likely to enact that policy. The two are related since a party’s reputation is likely influenced by the policies they propose.

We posit the health of the local economy as a single indicator for voters, an assumption which we justify later in the paper by examining other local economic factors such as gas prices and median household income.

It would be ideal to specify the issue-ownership voting mechanism as a causal pathway via individual beliefs about which party is better at dealing with local economic turmoil. However, testing such a mechanism is not possible due to limited data on individual beliefs about distinct partisan reputations available in nationally representative surveys.

More specifically, rising local unemployment (\(\uparrow\)) makes subjective evaluations negative (\(\downarrow\)), and negative evaluations (\(\downarrow\)) decrease support for the incumbent (\(\downarrow\)). Thus, the overall effect is predicted to be negative.

One might argue that the predicted effect should differ for the Democratic and Republican parties. Two possible scenarios come to our mind. First is that the effect of local unemployment on subjective evaluations of the national economy may be conditioned by the incumbent’s party. Though subjective economic evaluations are largely driven by individual party attachments (Enns et al. 2012; Evans and Pickup 2010; Gerber and Huber 2010), we are less concerned about this scenario because our prediction considers average effects rather than variability in individual voters. Furthermore, we are interested in how much voters respond to changes in local unemployment, once their partisan preferences are accounted for. We expect that this residual effect is not conditioned by the party in office. Indeed, this expectation is supported by our analysis result showing that the effect of rising local unemployment on sociotropic evaluations is negative and its size is more or less consistent across the 2008-2016 elections. The second possible scenario is that the effect of sociotropic evaluations on vote choice may be conditioned by the incumbent’s party (Kinder and Kiewiet 1981). However, negative (positive) evaluations are found to be associated with punishing (rewarding) incumbents in previous research, and these relationships appear to not be conditioned by the incumbent’s party (Godbout and Bélanger 2007; Gomez and Wilson 2001). While there is evidence for conditioning effects of an open-seat race (Godbout and Bélanger 2007), such conditioning effects do not make the sign of the effect of sociotropic evaluations vary. These previous findings reduce the concern that the second scenario could occur.

This conceptualization is from Becher and Donnelly (2013, p. 971). We acknowledge that the Becher and Donnelly study provided us with an insight for applying the potential outcomes framework to our analysis.

The 2008 dataset contains 2335 counties, and the 2012 dataset contains 2497 counties. The 2016 dataset includes 2473 counties.

http://​www.​bls.​gov/​lau/​.

As discussed below, we check for the robustness of the estimated effect of the annual change in unemployment.

For further details about the required assumptions, refer to the sequential ignorability assumption proposed by Imai et al. (2010b).

The U.S. has a long history of systemic residential segregation by race and income (Gordon 2014; Massey and Denton 1987; Williams and Collins 2001). Furthermore, political factors such as partisanship and ideology could confound the causality between economic evaluations and vote choice (Anderson et al. 2004; Evans and Andersen 2006; Evans and Pickup 2010; Gerber and Huber 2010; Wlezien et al. 1997).

Detailed descriptions of the pretreatment covariates and the sensitivity analysis results are in the online appendix (See A2 and A5).

See A3. Also, see the general estimation algorithms for the mediation analysis that can be found in Imai et al. (2010a, Appendix D). Furthermore, see Becher and Donnelly (2013, pp. 971–974) that applies the causal mediation analysis based on the potential outcomes framework to the study of economic voting.

See online appendix A4 that includes six tables presenting the regression estimates from the models of retrospective evaluations and the models of vote choice. Tables 1–3 in the appendix demonstrate that an increase in county-level unemployment leads to negative retrospective evaluations of the national economy in the 2008, 2012, and 2016 elections. This finding is robust to several alternative model specifications. Tables 4–6 show that as people see the economy more negatively (positively), they tend to become less (more) likely to vote for the incumbent party. This finding is robust to several alternative model specifications. From these regression estimates, we may make a preliminary inference that rising local unemployment forms negative individual retrospective evaluations of the national economy which in turn decrease support for the incumbent party. In the main text, beyond this preliminary inference, we present the estimates of AME, ADE, and ATE as illustrated by Fig. 3. By doing so, we provide specific effect sizes. Furthermore, we make an inference for the direct causal pathway that the preliminary inference does not address.

We conduct a sensitivity analysis to assess how much the estimated AME would be influenced by a possible violation of the ignorability assumptions. See Fig. 1 in the online appendix A5. To summarize, our sensitivity analysis demonstrates that the estimated AME is robust to a fairly large degree of violation of the assumptions.

If Obama’s share goes down by 0.36%, then Romney’s share goes up by that amount resulting in a swing of 0.72%. As the AME is calculated by taking the average over individual respondents’ predicted vote choice, the AME indicates the difference between expected vote shares of the incumbent party (Becher and Donnelly 2013, p. 973). Therefore, a 1% point increase in unemployment decreases individual respondents’ predicted voting for Obama, which translates into decreasing the expected vote share of Obama.

Previous research has developed a multidimensional economic voting theory. For example, see Lewis-Beck and Nadeau (2011) and Bélanger and Nadeau (2014).

We only considered the 2008 and 2012 cases for which all the local economic measures were available to us.

In 2012, we compare full-time workers (N = 13,118) and unemployed and part-time workers (N = 6532) from the 2012 analysis. The analysis considers only economically active respondents, excluding retired people, students, people with disabilities, and homemakers. The magnitude of the estimated ADE reveals that a 1% point increase in the county-level unemployment rate increases these people’s likelihood of voting for Obama by about 2.5%, holding the effect driven by national economic evaluations constant. This translates into about 5% vote swing from Romney to Obama. The estimated ADE associated with full-time workers is practically zero. The difference in the estimated ADE between these two different types of employment status is statistically reliable at the 95% level. While we find that part-time and unemployed voters’ likelihood of supporting the Republican incumbent is more negative than that for full-time employed voters in 2008, the difference does not reach statistical significance. In the 2016 analysis, the difference between the two different types of employment status is not statistically distinguishable from zero.