This article appears in the December 19, 2011 issue of National Review.

For months, pundits have been eyeballing poll numbers and spinning the story that Mitt Romney has a solid, unchanging base of support, and that the other candidates are taking turns being the "Non-Romney" of the day. This consensus view is based on a faulty premise. If eyeballing of the data were enough, God would not have given us econometricians.

To an econometrician, the consensus is testable. If it is correct, then changes in poll numbers should be highly negatively correlated among the Non-Romneys: If Newt Gingrich goes up, Rick Perry goes down. But the changes should be uncorrelated with movements in Romney's poll numbers.

The nearby chart explores these links. It is based on every poll reported by the website RealClearPolitics since the beginning of June 2011, and analyzes how the ups and downs of the different candidates correlate. To make the large amount of information digestible, the chart identifies for each candidate the relationship between changes in his poll numbers and changes in those of his competitors. A negative and statistically significant relationship is denoted by an arrow, and no significant relationship by a zero. There were no cases of a statistically significant positive correlation between two candidates.

For example, the first row of the table reveals (reading across) that Gingrich's support has come at the expense of Romney, Cain, Perry, and Bachmann, but is not related to the swings for Paul, Santorum, and Huntsman.

The table is filled with interesting information. Perhaps most interesting, the data suggest that support for Romney is related to support for Perry and Gingrich but not to support for the others. Romney is not on an island: His voters seem willing to jump to other candidates, provided that the alternative has ample political experience.

Among the Non-Romneys there does appear to be a bit of a game of musical chairs going on. Bachmann, for example, rises and falls at the expense of Perry, Cain, and Gingrich, but not of Romney. The chart also suggests that a number of candidates live in an alternative reality, far removed from the real action. Support for Ron Paul, Jon Huntsman, and Rick Santorum is uncorrelated with support for anyone else.

There are currently two frontrunners, and the correlations help clarify the likely future scenarios.
First, if Gingrich and Romney stay strong, and others drop out, then it looks as if Gingrich would gain the most support, since Romney stands to gain only from Perry, who barely has a pulse.

In the second scenario, Gingrich goes down in the polls, and his supporters spread out to many candidates. This would redound to Romney's benefit, but not necessarily decisively. Finally, if Romney goes down, most of his supporters will fan out between Gingrich and Perry. Given his large lead over Perry, that could be enough to put Gingrich over the top quickly.

Kevin A. Hassett is a senior fellow and director of economic policy studies at AEI.