Why isn't Mitt Romney doing better in his quest for the presidency? That's the question a lot of pundits, left, right and center, are asking. Given the weak economic recovery and the president's tepid approval ratings, one might expect Romney to run away with this.
The answers usually come back to those offered up Friday by Charlie Cook, the dean of Washington psephologists: Romney is a weak candidate; Romney’s advertising has been weak; Romney didn’t reach out to Latinos; Paul Ryan was a poor veep choice.
I’ve expressed agreement with many of those criticisms to varying degrees. But there are two important things to bear in mind. First, summer polling has been amplified by the 24/7 news cycle, yet it is of limited utility. The correlation between summer polls and fall outcomes is weak; one need only look at years like 1980 and 1976 to reach this conclusion.
Even so, I doubt that Romney will win by more than a few points, if he wins at all. Obama’s job approval today is at 48 percent, right at the political Mendoza Line (as Josh Kraushaar once put it) separating acceptable mediocrity from unacceptable mediocrity. With job approvals above 45 percent, it is unlikely that Obama would lose by more than a few points, if he loses at all.
More importantly -- and I think a lot of commentators miss this -- the economy isn't as bad as it was in 1980, or even in 2008. Yes, the recovery is weak, and is weak enough to endanger Obama's re-election bid. But it is still a recovery, and might be strong enough to re-elect the president.
Indeed, if you look at various econometric models of the election, the lion’s share (10 of 13) predict an Obama victory in the fall. Only two, Douglas Hibbs’s “Bread and Peace” model and Alfred Cuzan’s model, give Romney an overwhelming shot at winning.
Of course, the downside of these models is that they tend to focus on a single economic variable, and most include endogenous data such as presidential approval or even polling data. So I wanted to look at a heuristic device I developed, based on an approach I took in the winter, which tries to take a holistic look at the economy.
The technique is as follows. In late 2011, Nate Silver identified a host of economic variables that correlated with previous presidential election outcomes. I took those variables with a fairly robust relationship to presidential election outcomes (r-square of more than 0.25). Then, to eliminate double-counting (or if you prefer, multicollinearity), I went through and excluded variables that correlated strongly with other variables (r-square of more than 0.6). Basically, there’s no need to include changes in real disposable income and changes in per capita RDI, since those both tell us roughly the same thing.
That left me with the following economic variables (all taken over three quarters or nine months): the ISM Manufacturing Index; change in non-farm payrolls; change in the unemployment rate; change in real personal income; change in the employment-to-population rate; real GDP growth; real non-farm output; and changes in real commercial and industrial loans.
I then went through and ranked each presidential year for each economic variable. For example, the best ISM Manufacturing Index we’ve ever seen was in 1984 (61.7), so it was ranked first. The worst was 1980 (11), so it is ranked last
Why isn't Mitt Romney doing better in his quest for the presidency? That's the question a lot of pundits, left, right and center, are asking. Given the weak economic recovery and the president's tepid approval ratings, one might expect Romney to run away with this.
ReplyDeleteThe answers usually come back to those offered up Friday by Charlie Cook, the dean of Washington psephologists: Romney is a weak candidate; Romney’s advertising has been weak; Romney didn’t reach out to Latinos; Paul Ryan was a poor veep choice.
I’ve expressed agreement with many of those criticisms to varying degrees. But there are two important things to bear in mind. First, summer polling has been amplified by the 24/7 news cycle, yet it is of limited utility. The correlation between summer polls and fall outcomes is weak; one need only look at years like 1980 and 1976 to reach this conclusion.
Even so, I doubt that Romney will win by more than a few points, if he wins at all. Obama’s job approval today is at 48 percent, right at the political Mendoza Line (as Josh Kraushaar once put it) separating acceptable mediocrity from unacceptable mediocrity. With job approvals above 45 percent, it is unlikely that Obama would lose by more than a few points, if he loses at all.
More importantly -- and I think a lot of commentators miss this -- the economy isn't as bad as it was in 1980, or even in 2008. Yes, the recovery is weak, and is weak enough to endanger Obama's re-election bid. But it is still a recovery, and might be strong enough to re-elect the president.
Indeed, if you look at various econometric models of the election, the lion’s share (10 of 13) predict an Obama victory in the fall. Only two, Douglas Hibbs’s “Bread and Peace” model and Alfred Cuzan’s model, give Romney an overwhelming shot at winning.
Of course, the downside of these models is that they tend to focus on a single economic variable, and most include endogenous data such as presidential approval or even polling data. So I wanted to look at a heuristic device I developed, based on an approach I took in the winter, which tries to take a holistic look at the economy.
The technique is as follows. In late 2011, Nate Silver identified a host of economic variables that correlated with previous presidential election outcomes. I took those variables with a fairly robust relationship to presidential election outcomes (r-square of more than 0.25). Then, to eliminate double-counting (or if you prefer, multicollinearity), I went through and excluded variables that correlated strongly with other variables (r-square of more than 0.6). Basically, there’s no need to include changes in real disposable income and changes in per capita RDI, since those both tell us roughly the same thing.
That left me with the following economic variables (all taken over three quarters or nine months): the ISM Manufacturing Index; change in non-farm payrolls; change in the unemployment rate; change in real personal income; change in the employment-to-population rate; real GDP growth; real non-farm output; and changes in real commercial and industrial loans.
I then went through and ranked each presidential year for each economic variable. For example, the best ISM Manufacturing Index we’ve ever seen was in 1984 (61.7), so it was ranked first. The worst was 1980 (11), so it is ranked last