Voters for Obama

Abstract

The percentage-point difference between voting for Obama over McCain for those people who only voted for the president is +47.6. But when multiplied by 0.02424, the proportion of the electorate voting only for the president, their contribution to Obama’s victory margin shrinks to +1.150. The percentage-point difference between voting for Obama over McCain for those people who voted for all positions on the ballot is +5.56, a much smaller difference. But when this difference is multiplied by 0.88350, the proportion of the electorate voting for candidates on all positions of the ballot, their contribution is +4.912. It is more than four times larger than the contribution of Obama-only voters. To highlight such differences, this chapter develops a new procedure rooted in decision theory and applies it to answers this question: What kinds of practical voters facilitated Obama’s victory over John McCain? This chapter studies discontent with the incumbent president’s policies, variables of the voting process, and the stability of the vote. It finds that people who voted for Obama compared with those voting for McCain did not approve of Bush’s policies; made up their minds early; and voted early through absentee or mail ballots. Voters who previously were less politically active—African Americans, young people, previous nonvoters, the “all others” category (i.e., missing responses, refused, other), and people not sorted by partisanship—boosted Obama’s margin of victory. The party of the incumbent congressperson, a context variable, had less effect. Democratic voters in 2008 were more diverse than Republican voters; Obama captured the vote of people across the political spectrum whereas voters for McCain were more homogeneous.

State Sen. Barack Obama , a Chicago Democrat, says Daley profited from the economy and “this aversion to ideology and an emphasis on management that plays well to an executive.” The end of the machine has immensely strengthened the power of the mayor. The decline of patronage and political organizations that delivered voters to the polls has reduced the power of local ward leaders to challenge mayoral authority. “It’s harder for folks to build their own independent organizations.” Money and advertising rule now. That leads to a new urban politics, built on what Obama calls “pinstripe patronage .” It includes not only city contracts but also work that has parceled out to law firms and the fees that go to the brokerage houses that float city bond issues. “They do well, and you get a $5 million to $10 million war chest.”

—E.J. Dionne, Jr. (1999)

Unlike Ronald Reagan , who could keep demonizing government as he attempted to minimize the political costs of the sharp recession heading into the 1982 midterm elections, Obama found it difficult to voice any full-throttled attack on the bankers and corporate chieftains he and his advisors depended upon to spark the economy as soon as possible.

—Theda Skocpol and Lawrence Jacobs (2011, 32)

Notes

• 1. These figures were accessed 26 December 2011 from the web site of the Federal Election Commission , http://www.fec.gov/finance/disclosure/srssea.shtml, Summary Reports, 2007–2008 Election Cycle, as of December 31, 2008:

Campaign	Total net receipts	Net disbursed	Cash	Debt
Obama	$778,642,962	$760,370,195	$18,272,367	$434,954
McCain	$383,913,834	$358,008,447	$26,377,840	$1,603,974

• 2. The landmark legislation of the first two years of the Obama administration focused on the reform of healthcare insurance, making educational loans available, and regulating powerful economic interests (Skocpol and Jacobs 2011, 44). These changes are consistent with Obama’s campaign promises and with the dimensions of the human development index. Hopefully, their implementations will improve the standing of the U.S. on this index.

• 3. These data are from Table 2, Human Development Index Trends, 1980–2011, UNDP, Human Development Report 2010, 148–150, and are based on the 2010 revision of this index.

• 4. The American Human Development Report 2008–2009 reports a more severe drop, from second in 1980, 1985, and 1990; to sixth in 1995; to ninth in 2000; and to twelfth in 2005 (Fig. 1.2, 13). This trend is based on the earlier measure of human development. The authors of this American report (Burd-Sharps et al. 2008, 13) did not report that in 1975 the U.S. was tied for sixth, the actual trend is thus more jagged. For the original data see Table 2, UNDP, Human Development Index Trends, Human Development Report 2007/2008, 234.

• 5. These data are from Table 3, Inequality-Adjusted Human Development Index , Human Development Report 2010, 152–155.

• 6. Stepan and Linz (2011, Table 3, 844, 847) compare inequality in the U.S. with 21 other countries that compose their “comparison set” of other long-standing democracies in advanced economies. By only studying these similar countries, the variation across all of the countries is missed. By the judicious use of indicator variables, the differences between the U.S. and other long-standing democracies can be quantified readily.

• 7. See Stepan and Linz (2011, Table 3, 847) for the data.

• 8. Table 3, Inequality-Adjusted Human Development Index, Human Development Report 2010, 152–155.

• 9. Stepan and Linz (2011, Fig. 3, 847); Smith (2011, Table 13.1, 385).

• 10. Smith (2011, 386–387) discusses the poor showing of the United States on these indicators of the quality of healthcare .

• 11. Circa May 2012 after his first term in office Obama’s job performance ratings , 48% approve and 47.8% disapprove, were considerably better than the 65% disapproval for Bush in November 2008 after his second term in office. About 58.5% in May 2012 thought the country is headed in the wrong direction versus 78% in November 2008. Accessed circa 25 May 2012 from the Real Clear Politics web site.

• 12. Stuart Stevens , a lead strategist and media consultant for the Mitt Romney campaign in 2012, opined that not taking into account the difference between the percentage-point difference and the contribution to a candidate’s margin impaired their campaign (Jamieson 2013, 40–41): “Our goal was always to get up to converting at least 75% of wrong track voters and to get you [the Democrats] at least down to 85% of right track voters. We slowly crept up. On election day we were close to 75% of wrong track voters. The problem was there were fewer of them.”

• 13. Kahneman ’s (2011, 322–323) analysis of fear of bus bombings in Israel guides this procedure.

• 14. Helevik (2009) discusses the advantages and limitations of the linear probability model . Murname (2013, 395–398) applies this model to predict high-school graduation rates.

• 15. For the initial reports see Greenberg et al. (2008) and Greenberg Quinlan Rosner Research (2008).

• 16. In-depth analyses of standard surveys are provided by Kenski et al. (2010) for the 2008 Annenberg Election Survey; Lewis-Beck et al. (2008) for the 2000 and 2004 American National Election Surveys; and Ellis and Stimson (2012) for the General Social Survey. Reanalyses of these datasets most likely would lead to redundant findings.

• 17. The exit poll data were presented by CNN.com, Election Center 2008, updated, November 17, 2008; accessed for use here on 1 September 2011.

• 18. This distribution is derived from the variable coded presev3 in the computer code syntax.

• 19. To take into account the survey weights the analyses apply the following SAS survey modules : Proc SurveyFreq, Proc SurveyMeans, Proc SurveyRegression, and Proc SurveyLogistic. The analyses also apply Proc Factor, Proc Mixed, Proc Glimmix and Proc Calis, which do not explicitly provide for survey weights.

• 20. Separate analyses suggest that evaluating the direction of the country as wrong (x) is prior to evaluating Bush’s job performance as poor (t). The latter mediates the effects of the former on political response variables (y).

• 21. The next chapter disaggregates the items in the discontent index. The item about Bush’s job performance shows about the same correlations as the discontent index: Unfavorable job performance is negatively correlated with warm feelings for Bush (Spearmen r _s = −0.723) and positively correlated with voting for Obama (r _s = +0.697), p < 0.0001 for both correlations.

• 22. The recalculated percentages are (51.65/95.55) − (43.90/95.55) = 54.06 − 45.94 = 8.12.

• 23. Kenski et al. (2010, 251–264, 304–305) studied absentee and early voting in the 2008 campaign. Their Table 11.4, p. 260 clearly shows that the earlier the vote the larger the Obama edge over McCain. However, the contributions to Obama’s margin of victory of the different categories of time of voting were not reported. The table below does this; showing that the percentage-point differences (δ) strongly favor early voting but the probability adjusted contributions to Obama’s vote are about even across the three categories of time of voting:

Contributions to Obama’s victory margin from Kenski et al. (2010, Table 11.4)

Time of vote	N	Freq. prob.	Obama (%)	McCain (%)	Difference = δ (%)	Freq. prob. × δ (%)
Election day	2074	0.651996	52.6	47.4	5.2	3.4
1 Week prior	660	0.207482	57.9	42.1	15.8	3.3
2 Weeks prior	447	0.140522	62.4	37.6	24.8	3.5
Marginal prob.	3181	1	55.1	44.9	10.2	10.2

• 24. SAS’s Proc SurveyLogistic produces the odds ratios and their lower and upper confidence bounds; Proc SurveyReg produces the linear probability effects and the significance of the b coefficients.

• 25. With voting for all positions on the ballot as the base, the linear probability effect of only voting for president on voting for Obama is b = +0.224 (t = 2.71, p = 0.007).

• 26. When African Americans are the test factor, the linear probability model produces an effect of only voting for the president relative to voting for all positions on the ballot of b = +0.093 (t = 1.40, p = 0.161), which is not statistically significant.

• 27. From the linear probability model the effect of not voting in the 2006 congressional election on voting for Obama is b = +0.135 (t = 3.65, p = 0.0003).

• 28. When the respondent’s age category is controlled, the linear probability effect of not voting in the 2006 congressional election becomes b = +0.074 (t = 1.91, p = 0.057), not significant.

• 29. Political scientists define the relationship between political partisanship (assumed to be a prior variable) and political ideology (assumed to be a consequence) by saying that ideology is sorted by partisanship—Democrats tend to be liberals and Republicans tend to conservatives, but some Democrats and Republicans are moderates. At the extreme, there is polarization: the moderates have moved either to the Democrats or to the Republicans taking on their ideologies, which may have become more radical; the political middle has disappeared. For a very clear explication of the differences between sorting and mass polarization see Levendusky (2009, 4–8). This chapter generalizes slightly the notion or sorting by studying how the indicators of partisanship (prior variables) sort the vote (a consequence); the indicators of partisanship may be previous presidential voting choices, previous congressional voting choices, and so forth. See Fiorina and Abrams on sorting (2009, 49–74) and the contrasting view of Abramowitz (2013, 1–17) on polarization. Rather than using “polarization” which has different interpretations, the subsequent chapters will refer to differences between extreme categories as “polarities .”

• 30. In linear probability models, voting Democratic in the 2006 congressional election produces a pro-Obama effect of b = 0.693 (t = 31) and “all others” produces a pro-Obama effect of b = 0.476 (t = 15.5, p < 0.0001). The latter effect holds with simultaneous controls for discontent, age category, minority status, and gender, the b = 0.228 (p < 0.0001).

• 31. Using the Republican congressional voters in 2008 as the reference category in a linear probability model, the effect of voting Democratic in the 2008 congressional election on Obama vote is b = 0.738 (t = 37.4, p < 0.0001); the smaller effect of “all other” on voting for Obama is b = 0.579 (t = 9.7, p < 0.0001). Simultaneous controls for minority ethnicity, age category, gender and discontent do not change the significance of these relationships.

• 32. Here, the summation is the equivalent of a difference-in-differences (DID) effect: When there is a Democratic incumbent then Obama’s advantage is +12.261 percentage points. When there is a Republican incumbent then McCain’s advantage is +4.504 percentage points. The DID = +12.261 − +4.504 = 7.757 percentage points; Obama’s margin of victory.

• 33. A linear probability model also confirms the effect of a Democratic congressional incumbent on vote for Obama: the b = 0.178 (t = 6.48, p < 0.0001).

• 34. A latent class model of the issues of the 1992 election classified the voters along a Left-Center-Right continuum and found that those on the Right were more ideologically consistent than those on the Left. About 69.5% of those on the Right were conservative ideologically, whereas 43.7% of those on the Left were liberals and another 32.6% on the Left were centrists. For details see Smith (2008, 178–182) and Smith ([2003] 2004).

• 35. In the linear probability model the effect on voting for Obama in 2008 of the “all others” category for the 2004 presidential election is b = +0.24 (t = 5.5, p < 0.0001). Simultaneous controls for discontent, political party affiliation, ideology, younger people, and African Americans reduce this effect to b = 0.065, but it retains its statistical significance (t = 2.09, p = 0.037).

• 36. The less ideological political center boosted Bill Clinton ’s margin of victory over George H.W. Bush in 1992. Smith (2008, 178–182) and Smith ([2003] 2004) show that among the Left 55.8% identified as Democratic and among the Right 56.3% identified as Republican, about the same percentages. Among the Center, 37.5% identified as Democratic, 29.7% identified as Independent, and 30.5% as Republican.

• 37. In a linear probability model, the “all others” category for 2004 relative to the Bush Republican voters in 2004 boosts the Obama vote in 2008 by b = 0.63 (t = 15.61, p < 0.0001). The Kerry voters relative to the Republican voters in 2004 boosts the Obama vote in 2008 by b = 0.795 (t = 44.6, p < 0.0001). These effects remain statistically significant when discontent, party affiliation, ideology, younger age, and African American are simultaneously controlled. But the difference in Obama vote between the Kerry voters and the “all others” voters becomes minimal and the effects shrink: Now the statistically controlled effects for the “all others” category for 2004 relative to the Bush Republican vote in 2004 boosts the Obama vote by b = 0.176 (t = 3.83, p = 0.0001); and the Kerry voters in 2004 are still Obama voters in 2008, but by a much smaller amount than earlier, the b = 0.181 (t = 4.45, p < 0.0001) versus b = 0.795 (t = 44.6, p < 0.0001).

• 38. This very brief synopsis of aspects of the 2012 election is consistent with Alexander and Jaworsky ’s (2014) cultural account. Implicitly, their theoretical model is one of stimulus and response with the stimuli being the perceptions of candidates made available to the audience of ordinary voters by television, speeches, advertisements, and interpersonal influence.

• 39. This discussion draws upon Drew (2012), Toobin (2013), and Schlozman et al. (2012). The latter state (2012, 537): “Of the 2018 legislators across nine states that considered voter ID bills between 2005 and 2007, 95.3% of Republicans and 2.2% of the Democrats voted in favor.” They also suggest (2012, 563–564) that voter ID laws do not prevent fraudulent voting, which is very rare in any case, but may depress voter turnout. The recent Supreme Court (June 25, 2013) decision weakens the Voting Rights Act by no longer requiring preclearance with federal agencies for changes in voting laws by nine states mostly in the South.

• 40. I thank Ruth Leeds Love and anonymous reviewers for their comments that focused my revisions of this present chapter.