Skip to main content
ABC News
Newspaper’s Endorsement Has Been Leading Indicator in New Hampshire

On Sunday, The Union Leader of Manchester, New Hampshire’s largest newspaper, endorsed Newt Gingrich for president. The endorsement represents one of the most tangible signs of parts of the conservative establishment coming around to Mr. Gingrich, who to date has received very few endorsements from Republican elected officials. It also represents a blow to Mitt Romney, who had led all polls of the state.

But does the endorsement tell us anything about how New Hampshire Republicans are likely to vote? Or is it just fodder for a slow news day?

I’ve seen numerous analyses that have questioned the importance of the endorsement on the grounds that The Union Leader has endorsed a losing candidate on many occasions (for instance, Steve Forbes in 2000 and Pete du Pont in 1987). It is probably good to be skeptical about the value of newspaper endorsements overall, especially at a time of declining circulation and increasing competition in the media space.

These analyses, however, leave something to be desired. The problem is that they do not consider how a candidate performed relative to expectations. A candidate who zoomed up to 15 percent from 5 percent in the polls after The Union Leader’s endorsement would testify to the fact that it was potentially important, even if the candidate did not win the state. Conversely, a candidate who had been dominating the state and was at 55 percent in the polls, but who won the state by only a narrow margin after The Union Leader’s endorsement, would not be a mark in its favor.

What I’ve done, therefore, is compare how the Republican candidate endorsed by The Union Leader finished in each of the past six competitive New Hampshire primaries compared with how he was polling at the time of the endorsement. By default, I use an average of all statewide polls conducted in the three weeks before the endorsement, although there were no polls within this window in the 1980 and 1996 cycles, so I use the first poll just after the endorsement instead. I do not consider races before 1980, as we have no New Hampshire polls for those years in our database.

As it happens, although only three of the six Republicans endorsed by The Union Leader during this period won their primary, all six outperformed their polling. Mr. du Pont, for instance, finished with a fairly meager 11 percent of the vote in 1988 — but this was better than his 4 percent standing in the polls at the time of the endorsement.

On average, the candidates endorsed by The Union Leader finished with 29 percent of the vote in New Hampshire — an 11-percentage-point improvement from the 18 percent they averaged in the polls when the endorsement was made.

I should note, however, that on average candidates have some tendency to improve in the actual voting from their poll standing because the polls include undecided voters, whereas everyone who actually votes will have to choose a candidate. To correct for this, as well as to check whether the results could be attributable to random noise, I ran a simple regression analysis that explains a candidate’s share of the vote in New Hampshire as a function of whether or not he was endorsed by The Union Leader and his polling average at the time. The sample consists of all Republican candidates who (i) were included in at least one New Hampshire poll at the time of the endorsement and (ii) officially ran for president at some point during this cycle.

This analysis finds that The Union Leader’s endorsement has been highly statistically significant in helping to explain the voting results. Consistent with the simpler averaging method that we used before, it pegs the endorsement as having roughly an 11-percentage-point impact.

Nevertheless, there are a couple of very important cautions as to its broader significance.

First, it does not necessarily imply causation. It is unlikely that a candidate wins as much as an additional 11 percent of the vote simply because The Union Leader endorses him. Instead, it may be more of a leading indicator for how actual New Hampshire voters will think about the candidates once they finish sorting through them. That is, it replicates in some way the thinking process that some segment of New Hampshire voters will go through, whether or not they pay any attention to The Union Leader itself. More broadly, the endorsement may serve as a proxy for various sorts of intangible qualities that may help a candidate to perform strongly in New Hampshire but that are not necessarily reflected in the early polls of the state.

Second, this finding is drawn from only six historical elections. As I often remind our readers, a regression analysis on historical data is not really the same thing as a prediction of how these factors will play out in the future. Fairly often, a relationship that is found to be highly statistically significant in past data will prove to be unreliable when applied out-of-sample.

Nevertheless, this is a pretty good sign for Mr. Gingrich. If you do take the results of the regression analysis to be tantamount to a prediction, they imply that New Hampshire could be quite close, with Mr. Romney finishing with 36 percent of the vote and Mr. Gingrich at 30 percent. What I think is fair to say is that Mr. Gingrich would at least have a shot at winning New Hampshire if he also wins Iowa, a result that could be devastating to Mr. Romney’s campaign.

Nate Silver founded and was the editor in chief of FiveThirtyEight.