With the midterm elections only two weeks away and the Democrats in jeopardy, the prevailing wisdom is that the election will be a referendum on the Obama administration's stewardship of the economy. A large fraction of 2008 Obama voters now cite the economy and jobs as the key reason they will vote Republican this year, according to an October 17 AP poll. "The president must zero in on the economy if he wants to help himself and his party," writes Eleanor Clift. The basic notion here, promulgated by pundits and political analysts, is that the current political environment turns on the vagaries of the economy. This amounts to a cyclical theory of American politics. And, in fact, several decades ago, the political scientist Douglas Hibbs advanced his seminal theory of the "political business cycle," which argues that economic movements have a sizable effect on American elections.
But another line of thinking suggests that American politics turns on deeper structural changes in the economy and society. In the influential Red State, Blue State, Rich State, Poor State, Columbia University's Andrew Gelman
and his colleagues uncovered a paradox that both confirms and defies
the conventional wisdom about American elections. While rich voters
trend Republican, rich states trend Democratic, he found. The
opposite holds as well. Though poor and minority voters overwhelmingly
pull the lever for Democrats, poor states consistently end up in the
A second version of the structural approach comes
from John Judis and Ruy Teixeira, who argue in The Emerging Democratic Majority that
the rise of the post-industrial economy has tilted the playing field
toward Democrats who gain an advantage in wealthier urban "ideopolises"
while holding onto the votes of the poor and minorities.
A third perspective comes from Ronald Inglehart of the University of Michigan, whose detailed World Values Surveys identifies a shift in political culture from the more traditional, religious, and materialist orientations of the industrial age to post-materialist values of self-expression, openness to diversity, secularism, and broad public goods like concern for the environment.
~ ~ ~
This juxtaposition thus mirrors the debate over the economy: Will shorter-term cyclical factors determine the outcomes of the mid-terms or are deeper structural factors at play?
With the help of my colleague Charlotta Mellander,
I decided to take an empirical look at this question. On the one hand,
we considered a series of key cyclical variables such as the
unemployment rate and its change since the economic crisis began, and
also housing prices and their change since the bubble burst. And, on the
other hand, we considered key structural factors, such as income a la
Gelman, post-industrialism a la Judis and Teixeira (measuring the
prevalence of creative class jobs versus working class jobs), and
post-materialist political values a la Inglehart, including the
prevalence of religion and openness to both immigrants and gays and
We confined our analysis to the state level, using pooled
polling data for both Senate and governor races across the country, which we drew from Nate Silver's FiveThirtyEight election forecasts at The New York Times.
We conducted a basic correlation analysis and compared the results for
the current midterm to those for Obama and McCain in the 2008
presidential race. (The graph below summarizes the key findings.)
This kind of analysis can only point to associations between factors and does not identify any causal pattern, and of course other factors may come into play. Polling data covers a much smaller number of observations than election returns and suffers from other problems. For these reasons, we caution against drawing overly broad conclusions from this exercise. Still, the patterns it points to are quite interesting.
Despite all the "it's the economy, stupid" hubbub among the chattering classes, our analysis finds little empirical support for the cyclical view. There was no statistical association at all between the share of voters leaning Democrat or Republican for either Senate and governor races and our key cyclical factors -- the unemployment rate, the change in the unemployment rate, housing values, or change in housing values. This is not to say that these factors do not matter at the margin, as polling data clearly tell us that many individuals are shifting their 2008 Democratic vote to a Republican vote in these midterms.
But it's structural factors on which this election is much more likely to turn. We find significant statistical associations between most of the structural variables in our analysis and the share of voters leaning Democrat or Republican in both Senate and governor races, as detailed below.
Income: Higher income states went for Obama in 2008
while lower income states went for McCain. The trend continues, even in
light of the ongoing economic malaise. Income is positively associated
with Democratic share for Senate (.4) and governor (.36) races. And it
is negatively associated with Republican share for Senate (-.54) and
governor races (-.38). These associations have weakened more on the
Democratic side (.52 for Obama) than for the Republicans (.-51 for
Class: Class played a role in the 2008 presidential election and it continues to do so in the midterms. Creative class states went for Obama in 2008 and working class states went for McCain, and this holds up for the midterms as well. The creative class is positively associated with Democratic share in both Senate (.34) and governor (.36) races, and negatively associated with Republican share in each (-.38 for Senate and -.52 for governor). These associations have again weakened more for the Democrats (.52 for Obama) than for the GOP (-.46 for McCain) in 2008.