_Unequal Democracy_ is one of the best books on American politics in the past generation. It has few peers in its engagement with important, topical questions based on careful empirical analysis reported in accessible prose.
Among the book’s many interesting findings is this:
In the fourteen presidential elections from 1952 to 2004, a majority of white voters in the lower third of the income distribution chose the Democratic candidate over the Republican in only four.[1] Why is that? Bartels offers a fascinating answer. I’m going to summarize his argument and findings and attempt to extend them a bit.
_Unequal Democracy’s explanation_
For most Americans, and particularly for those in the lower part of the income distribution, incomes have tended to grow much more rapidly under Democratic presidents than under Republican ones. From 1948 to 2005, family income at the twentieth percentile of the income distribution (P20) grew at an average rate of 2.4% per year under Democratic presidents, compared to just 0.3% per year under Republicans.[2] Cumulated over time, that’s a huge difference.
Despite much hand-wringing by progressive commentators, there has been no decline in the propensity of low-income whites to vote Democratic in presidential elections.[3] But as the chart above shows, that’s partly because the share voting Democratic was pretty low even in the 1950s, 1960s, and 1970s. What’s interesting, and surprising, isn’t so much that it hasn’t decreased; it’s that it was never particularly high.
How can that be, given that the Democrats have long claimed to be the party of the people and that the evidence, at least with respect to income growth, supports their claim? The reason, according to Bartels, is not race or crime or feminism or abortion or foreign policy. It isn’t roast beef, apple pie, and postmaterialist values. It’s not religion. It’s not a leftward lurch or a right turn by the Democrats. It isn’t a belief that there’s no difference between the two parties. It isn’t “false consciousness.” The reason is that voters are myopic, and Democratic presidents have been less likely than Republican ones to produce healthy income growth in election years.[4]
_Extending the theory and the empirical assessment_
Does the sharp contrast in overall income growth under Democratic versus Republican presidents have any effect on the partisan preferences of low-income whites? If it’s had little impact on presidential voting because of myopia, perhaps it’s more likely to affect an indicator of deeper, underlying party preference, such as party identification. The data in the following chart are supportive. Lower-income-third whites were more likely to vote Republican than Democratic in presidential elections even in the 1950s and 1960s (chart 1 above), but during those decades this group was much more likely to identify as Democratic than as Republican. The Democrats’ advantage has shrunk (I’ll come back to this), but it has persisted.
Though Bartels doesn’t put it quite this boldly, these findings suggest a simple but powerful explanation of party identification and voting among low-income whites: presidents’ success at generating income growth influences party identification and presidential voting, and income growth in election years exerts an even stronger impact on voting.
The partisan gap in income growth has varied over time and across regions. This can be used to get additional insight into its impact.
The magnitude of the partisan difference in income growth (overall, not just election years) has shifted over time. As the following chart shows, for families at the twentieth percentile the gap between Democratic and Republican presidents was much larger in the 1950s, 1960s, and 1970s than in recent decades.[5] Given this, we should observe a decrease in the gap in party identification among the lower income third. As chart 2 above shows, we do. And the timing fits: both occurred after the 1970s.
There also has been significant regional variation in the partisan gap in income growth, as is evident in the next chart. In the period from 1963 (the first year state income data are available) to 1979, the partisan gap in P20 income growth was much larger in the south than in the rest of the country.[6] Moreover, the decrease in the partisan gap between 1963-79 and 1979-2004 was greater in the south than in the nonsouth (though it shrunk in both regions).
This suggests two hypotheses about regional differences and over-time shifts in party identification among low-income whites: First, through the 1970s Democratic identification should have been stronger in the south than elsewhere. Second, after the 1970s we should observe a larger decline in Democratic identification in the south than in the rest of the country. The following chart shows that both of these hypotheses are supported.
Now let me return to voting in presidential elections. Low-income whites have been a bit more likely to vote Democratic since the early 1990s than they were in earlier decades (chart 1 above). The difference isn’t huge, but it appears to be real. Does the theory have an explanation for this?
Yes. On the one hand, this development is surprising given that the Democrats’ advantage in overall income growth has lessened (chart 3). But the theory suggests that the partisan gap in election-year income growth matters more, and here the recent trend has been favorable for Democratic candidates. As the next chart indicates, in 1992 and 2004 Republican presidents produced relatively anemic income growth.[7] Democratic presidents did no worse in 1996 and 2004. Given the continued (albeit narrow) Democratic advantages in low-income whites’ party identification and in overall income growth (charts 2 and 4), this new parity in election-year income growth ought to have helped boost Democratic voting among low-income whites.
_Alternative explanations_
There are other explanations of low-income whites’ party identification and presidential vote choice. How well do they square with the data?
Begin with the early postwar decades. The most common view is that Democrats won the hearts of low-income whites with FDR’s New Deal policies.[8] Low-income families benefit from programs that provide income security and/or redistribution, such as Social Security, unemployment insurance, and social assistance, and from laws that help to raise the earnings floor, such as the minimum wage and the right to collective bargaining. The argument is that they therefore favored the party that created and stood by such programs. This view has some difficulty accounting for the regional variation in low-income whites’ party identification during this period (chart 6).
A second view about the 1950s and (early) 1960s is that party orientations and voting were not based mainly on material self-interest, but rather on Civil War legacies. In this account, there never was a New Deal golden age in which the Democrats won the allegiance of low-income Americans by virtue of their commitment to economic opportunity, security, and fairness or via their superior performance in raising incomes. Low-income whites in the south, like most whites in the south, identified with and voted for the party that hadn’t launched the war of northern aggression. Outside the south low-income whites were more evenly split but leaned Republican.[9] An empirical problem for this hypothesis is the Democrats’ advantage in party identification outside the south (chart 6).
What about the period since the 1960s, or perhaps since the 1970s if we consider the late sixties and the seventies a transition era? The most prominent view is that low-income whites have turned away from the Democrats over social issues such as race, abortion, homosexuality, and school prayer.[10] The seeming weakness of this account, as Bartels notes, is that this group’s propensity to vote for Democratic presidential candidates has not decreased (chart 1).
A second view holds that in recent decades material issues have become more, rather than less, central to low-income whites’ partisan considerations, due to the parties’ ideological purification and their separation on economic issues. The Republican party has moved to the right and conservative southerners have switched from the Democrats to the Republicans. As a result, both parties have become less heterogeneous and the difference between them on economic issues has become much clearer than during the New Deal era.[11] A problem for this account, it would seem, is the lack of increase in Democratic identification among low-income whites outside the south (chart 6).
_Summing up_
I’ve offered an extension of Larry Bartels’ partisan-gap-in-income-growth explanation of low-income whites’ political preferences, an extension of the empirical assessment utilizing over-time and cross-region variation in the partisan gap in income growth, and some brief comments on a few prominent alternative accounts. The Bartels theory looks promising.
fn1. 1952 is the first year for which data are available, and Unequal Democracy was published prior to the 2008 election. This updates a finding by Jeffrey Stonecash in his “Class and Party in American Politics”:http://books.google.com/books?id=-_RTFNIkf4YC&dq.
fn2. Unequal Democracy, chapter 2, especially figure 2.1. These incomes are pretax, adjusted for inflation; the data are from the Current Population Survey (CPS). Bartels lags income growth one year; for instance, 1993 is counted as a Republican year and 2001 as a Democratic one.
fn3. Unequal Democracy, chapter 3, particularly figure 3.2. I’ll sidestep the question of whether low income is a good indicator of “working class”; I want to focus here on low income.
fn4. Unequal Democracy, chapter 4, especially figures 4.2 and 4.4. The logic is similar to that in “Douglas Hibbs’ “bread and peace” voting model”:http://douglas-hibbs.com/HibbsArticles/Public%20Choice%202000.pdf (pdf).
fn5. More detail “here”:http://www.u.arizona.edu/~lkenwor/challenge2010.pdf. The data I use in chart 4 are the same as those used by Bartels.
fn6. The south here is Alabama, Arkansas, Delaware, Florida, Georgia, Kentucky, Louisiana, Maryland, Mississippi, North Carolina, Oklahoma, South Carolina, Tennessee, Texas, Virginia, and West Virginia.
fn7. See also figure 4.2 in Unequal Democracy. I show income growth at both P20 and P95 in the chart because Bartels finds the latter to be more closely correlated with bottom-income-third whites’ vote choice.
fn8. Ruy Teixeira and Alan Abramowitz, “The Decline of the White Working Class and the Rise of a Mass Upper Middle Class”:http://www.brookings.edu/papers/2008/04_demographics_teixeira.aspx.
fn9. Stonecash, Class and Party in American Politics.
fn10. Thomas Edsall with Mary Edsall, “Chain Reaction”:http://books.google.com/books?id=GL80BQR6pk0C&dq; Thomas Frank, “What’s the Matter with Kansas?”:http://books.google.com/books?id=TMY2_2Zbnc8C&dq
fn11. Stonecash, Class and Party in American Politics; Nolan McCarty, Keith Poole, and Howard Rosenthal, “Polarized America”:http://books.google.com/books?id=mLqjIAAACAAJ&dq; Teixeira and Abramowitz, “The Decline of the White Working Class.”