He throws a couple oft-neglected factors into the mix to create an expected approval rating, which includes the state’s size and partisan leanings, as well as other, more standard metrics.
Who’s the most popular governor in the land? A fairly well-informed observer might guess Jon Huntsman, Jr. of Utah (who won’t be a governor for much longer), Charlie Crist of Florida or Sarah Palin of Alaska. Those would all be pretty good guesses, except for Palin, who was once extremely popular in Alaska, but no longer is. The answer, however, appears to be governor Dave Freudenthal of Wyoming, who when his approval was last tested back in August, was viewed favorably by some 81 percent of voters.
What makes this more remarkable is that Freudenthal is a Democrat — something which a only 26 percent of Wyoming voters are (52 percent are Republicans). On the other hand, Freudenthal has one big advantage: Wyoming is a small state, and as our own Andrew Gelman has noted, it seems to be much easier to maintain higher approval scores when you have fewer constituents to please.
What I’ve set out to do then, is to develop an index of gubernatorial power rankings that takes a governor’s approval scores and adjusts it for these two things: the size of a state and its partisan makeup. Let me explain in a bit more detail.
Approval Ratings. Approval ratings are taken from all available public polling data collected in the last three months (since March 1st). We take a governor’s positive rating less his negative rating (for example, 52 percent approve and 44 percent disapprove is a +8) and then simply average the numbers across all the polls. A maximum of one poll — the most recent — is used from each survey firm. In other words, this is basically the Real Clear Politics approach. Where no polls have been conducted on a governor within the past three months, we take the most recent available one; such instances are denoted in the chart below.
Obviously this is not a perfect metric. Approval ratings tend to be associated with fairly large house effects at different polling firms; a more sophisticated version of this analysis would correct for that. But, this is what we’re going to go with for now.
Adjustments for Size of State and Party ID. The next step is to create an expected approval rating based on two factors: the size of a state and its partisan leanings. The adjustment for the size of the state is based on a regression analysis of the unadjusted approval scores against the square root of a state’s population. Small-state governors are expected to have above-average approval ratings, and large-state governors below average ones.
The other adjustment is for party ID. Party affiliation isn’t quite as important for governors as it is for other types of office-holders, as they aren’t necessarily beholden to the whims of their national party and have somewhat more flexibility to define themselves. That’s why Freudenthal is the Governor of Wyoming, and Republican Jim Douglas is the chief in Vermont. Nevertheless, party has some influence. Based on previous (unpublished) work I’ve done on the 2006 election, each additional point of advantage a gubernatorial candidate has in terms of his partisan ID advantage translates into about half a point at the ballot box on Election Day. That is, all else being equal, we’d expect the Democratic gubernatorial candidate in Michigan, where Democrats had a 12-point party ID edge on the 2008 exit poll, to win by about 6 points.
We take this rule of thumb and apply it to a governor’s expected approval rating. For every point’s worth of advantage a governor’s party has in his state (as according to 2008 exit polls), we add half a point to his expected approval rating. Obviously, in cases where the governor’s party is at a disadvantage, this lowers his expected approval rating.
The Results. The final step is to compare a governor’s actual approval average against his expected one; this is what we refer to as our power rating. For example, Hunstman’s approval rating in a recent Dan Jones poll was a remarkable +61 in Utah, but because he hails from a state which is both very red and relatively small, we’d expect him to score a +18. His power rating, then is, 61 less 18, or +43 — still very good.
Power ratings for all 50 governors are listed below. (And yes, I know that Kathleen Sebelius is no longer the governor of Kansas, but nobody has tested Mark Parkinson yet).