Tuesday, June 12, 2012

Race and the US Presidential race

Conservative philosopher-blogger Keith Burgess-Jackson had already predicted that, if Obama loses the November 2012 election, the Left will say it was because of race. Now some social science has popped up to support precisely such a claim. A study (pdf) has used racially-charged Google searches to estimate how much less Obama's share of the popular vote was in November  2008 than one would expect based on the performance of other Democratic Party candidates in that election. (A newspaper blog post by the author summarising the research is here.) As predicted, Andrew Sullivan is already citing the data as explaining why Obama's chances for re-election is poor.

There is something a little odd about the result in the study, given that the only other postwar American Democratic Presidential candidates to get a majority of the popular vote were Carter in 1976 (50.08%) and LBJ in 1964 (61.05%). So Obama's 52.87% is the second-highest Democratic Presidential result in 16 Presidential elections. Harvard economics doctoral candidate Seth Stephens-Davidowitz's study implies that Obama "should" have got 56-58% of the vote, which would have made been an extraordinary performance. (The study estimates a 4-6 %point loss from racial animus and a 1%point gain from being black, netting out to a 3-5 %point loss in votes.)

How extraordinary? Well, the big winners in US Presidential elections are typically incumbent Presidents. George Bush senior 1988, the incumbent VP for a popular President, did better than Obama (53.37%). But the big postwar winners were: Ronald Reagan 1984 (58.77%); Richard Nixon 1972 (60.67%); LBJ 1964 (61.05%); Eisenhower 1956 (57.37%); and, the only non-incumbent big winner, Eisenhower 1952 (55.18%).  So, Seth Stephens-Davidowitz's regression on the results from a single election suggest that Obama should have got the best non-incumbent Party candidate result in the postwar era, even beating out war-hero Eisenhower, who won big after the Democrats had been in office for 5 consecutive Presidential terms (1932-1952) and (like the Republicans in 2008) had the burden of a very unpopular incumbent in the midst of a much bigger war.

Thus, just eye-balling the historical data, we have reasons to be sceptical about the result of Stephens-Davidowitz's study.

There is also some social science data which suggests the racial factor may not have been nearly so large. Estimating expected Presidential vote (actually, share of two-Party vote) purely on the basis of per capita GDP growth using data from every post-war Presidential election from 1952 onwards suggests that Obama slightly under-performed, but well within historical patterns. (Al Gore under-performed much worse in 2000, for example.) If, however, per capita GDP growth is weighted by the "military casualties effect", then Obama's 2008 performance was almost bang on what political scientist Douglas Hibbs's "Bread and Peace" model predicts. (Al Gore remains the stand out under-performer.)

So, we have a claim, based on regressions for a single election, that Obama "should" have put in the best performance by an non-incumbent candidate in the postwar period. (The "Bread and Peace" model only does elections from 1952 onwards, due to data limitations, but Truman, the incumbent, won in 1948.) Or, we have data, based on 15 elections, suggesting that Obama did about what you would expect (in achieving the second-highest share of the popular vote in the postwar period both by a Democratic Presidential candidate and by a Presidential candidate not representing the incumbent Party).

Stephens-Davidowitz's study of racial animus and the 2008 election reads well and appears to be analytically sound. Though using unemployment as an indicator of economic conditions is not very persuasive, the evidence is unemployment has much less impact on general voting patterns than one might expect.

A notable result in the study is that its measure of racial animus is not correlated significantly with Democratic candidate John Kerry's share of the 2004 vote which, as author Stephens-Davidowitz points out, is evidence against the view that racial animus is predominantly a Republican phenomenon.

One of the criticisms of my essay on post-modern conservatism was that I did not deal with the race issue. Race struck me as fraught, not particularly relevant to the phenomenon I was examining and an example of a general tendency for conservatives to be agin groups making a bid for equal treatment. Also, despite claims to the contrary, racial animus is not something that sits on just one side of the ideological spectrum.

That African-Americans overwhelmingly vote Democrat means that African-Americans effectively have a monopoly political provider--the Democratic Party. It is reasonable to ask how being effectively electoral prisoners of a single political Party has worked out for them. As a corollary of the above, black concerns do not have much of a voice within the Republican Party; both through lack of activists and through not being a targeted voting group. Which has not stopped folk such as Clarence ThomasCondoleeza Rice and Colin Powell being popular among Republican voters. But it does mean, as one commentator has pointed out, that;
The racial diversity among Democrats and the lack of it among Republicans means that the two bases bring differing sets of concerns to the national debate.
Which is very much not the same as saying that racial animus is somehow a specifically Republican problem.

Trying to get hard data on racial animus and its effects is difficult, which makes having an informed debate about it difficult. (The principle of equal protection of the law is much simpler.) Stephens-Davidowitz should be congratulated for his innovative use of Google data; as he says, a data-rich (and timely) analytical tool. Still, his results do seem to be stronger than historical patterns, and other social science research, suggests. Which brings back to the perennial problem about social science research; it is difficult. Both because of the complexities involved and because it is so often so ideologically charged. (As indicated by the principle that if a white voter votes for a white candidate because the candidate is white, that is racism; but if a black voter votes for a black candidate because they are black, that is not.)

If, in November 2012, Obama gets around about what Douglas Hibbs's "Bread and Peace" equation suggests, then Stephens-Davidowitz's study will not be vindicated. If, however, Obama gets significantly less, then Stephens-Davidowitz may well be on to something.

Which is a different thing than saying if Obama loses Stephens-Davidowitz's study will be vindicated. But the difference is likely to be lost in the shouting.

[Cross-posted at Skepticlawyer and at Critical Thinking Applied.]


  1. A correction: George H. W. Bush (the father), was elected president in 1988, while he was Reagan's vice president. In 1992, Bush was defeated by Bill Clinton (remember him?).

    1. Oops, will correct. I wrote some of this late at night ...