The Ground Game in the 2012 Presidential Election

Abstract

Relatively little research has examined the effects of campaign-led field activity in a competitive election. In this article, we leverage a unique data set containing the location of every Barack Obama and Mitt Romney field office and county-level data on the presidential vote to understand how communication with voters in the field may have affected the outcome of the 2012 presidential election. We find that the presence of Obama field offices was associated with greater Obama vote share at the county level, although we cannot detect a similar relationship for Romney field offices. We conduct additional robustness tests to address the potential limitations of these observational data. Ultimately, we conclude that even if Obama’s field organization out-performed Romney’s, the aggregate impact of Obama’s field organization was not large enough to determine the outcome of the 2012 presidential election.

Keywords:

• campaign
• election
• field organization
• ground game

Notes

1. Political science field experiments on turnout have helped isolate the effects of electioneering during real campaigns, but few designs have deployed treatments intense enough that opposing campaigns might detect the treatment and counter it. The design of such an experiment might also require opposing campaigns to detect the treatment in non-competitive areas and make the decision, for example, to ignore these areas (as is common when campaigns have scarce resources to deploy). Even when field experiments have delivered treatments that could be observed by opposing campaigns (Gerber et al., 2011), they have done so in elections without challengers, perhaps due to ethical concerns about influencing the outcome of the election.

2. For example, upon visiting a Romney field office in Cuyahoga County, Ohio, we observed campaign signs for many other Ohio races, especially the Senate race between Josh Mandel and Sherrod Brown.

3. Bird also said the campaign stratified battleground states into those believed to be crucial—these were Colorado, Iowa, Nevada, New Hampshire, Ohio, and Virginia—and those believed to be secondary (such as Florida and North Carolina). The former group received more resources—“they got whatever they wanted” and were “saturated,” said Bird.

4. This was again consistent with what we observed at the Cuyahoga and Medina County Boards of Election in October 2012. On the night before in-person early voting began in Ohio, Obama field organizers in Cleveland set up tents and portable bathrooms along the sidewalks leading up to the polling location. There was also entertainment from professional disc jockeys. This was to prepare for people who would get in line to vote and wait all night. In Medina County, a much more Republican-leaning county relative to Cuyahoga, we observed no such mobilization tactics by either party.

5. This analysis is available in the supplementary materials for their article: http://sites.sas.upenn.edu/mleven/files/gg_appendix_final.pdf

6. We coded the following states as battleground states: Colorado, Florida, Iowa, Nevada, New Hampshire, North Carolina, Ohio, Virginia, and Wisconsin.

7. We rely on county-level data because this is the lowest level of aggregation for which election returns are currently available across the states.

8. All variables were obtained from the U.S. Census Bureau (http://quickfacts.census.gov/qfd/download_data.html), with the exception of unemployment data, obtained from the Bureau of Labor Statistics (http://www.bls.gov/lau/laucntycur14.zip).

9. For example, we estimated the effect of having zero, one, two, or three or more Obama offices or Romney offices, accounting for the other variables in Table 2. For the Obama office variables, the coefficients and standard errors were 1 office (b = .30; SE = .16), 2 offices (b = .84, SE = .32), and 3 or more offices (b = .43, SE = .38). For the Romney office variables, the comparable statistics were 1 office (b = –.14; SE = .18), 2 offices (b = –.32, SE = .38), and 3 or more offices (b = .09, SE = .23). Thus, for both Obama and Romney, the apparent effect of three or more offices is smaller than that of two offices. Estimating the separate effects of one or two or more field offices suggests a fairly linear pattern. Having one Obama office in a county is associated with a .29 gain in vote share (SE = .16); having two or more is associated with a .72 gain in vote share (SE = .34). Having one Romney office in a county is associated with a –.16 drop in Obama vote share (SE = .18); having two or more is associated with a –.32 drop in Obama vote share (SE = .30).

10. One question that arises is whether these associations are confined to battleground states, given that field office activity in non-battleground states may have been directed at battleground states (e.g., by having supporters in, say, Idaho making phone calls to Nevada voters). We interacted the Obama and Romney field office measures in Model 2 with dummy variables for battleground state status (with CO, FL, IA, NV, NH, NC, OH, VA, and WI coded as battleground states), and included these interaction terms in the model. Neither interaction was statistically significant.

11. If we use the difference in the raw number of field offices, the coefficient is .088 (SE = .056). In an auxiliary analysis, Darr and Levendusky (2014) also find a statistically significant effect of the difference in the 2012 field offices, using a somewhat different specification than we use here.

12. The question, of course, is whether Obama might have placed field offices in places that were already trending Democratic. If so, and if the control variables in the analysis do not fully capture the characteristics that may have affected the placement of field offices, then our estimate is biased. One piece of evidence that supports our finding comes from Darr and Levendusky (2014), who find that changes in field office placement from 2004 to 2008 were not associated with House vote share in 2006, suggesting that in 2008 Obama was not placing field offices in districts that were trending Democratic.

13. This is very similar to Darr and Levendusky’s estimate that the field offices brought Obama an additional 275,000 votes nationally in 2008.

14. An anonymous reviewer speculated that Republican voters might simply be less susceptible to the effects of field organization than are Democratic voters. This is plausible, inasmuch as “marginal” voters—those influenced by exogenous influences on turnout—are disproportionately Democratic (Fowler, 2015).