To Win In Iowa Or New Hampshire, It May Be Better To Poll Worse Nationally

Ted Cruz is doing considerably better in polls of Iowa than he is in national polls. Chris Christie is doing better in New Hampshire than he is with all Americans. Donald Trump is doing worse in both Iowa and New Hampshire than he is in the country at large.

Just as during the 2012 general election, state polls and national polls disagree. And our advice this year is the same as it was then: Trust the state polls. In fact, there’s evidence the national polls may be a negative indicator once you control for the state-level survey results. If you’re a candidate who wants to win one of the first two contests, you’d rather have good state polls and bad national polls than good state and good national polls.

How can that be? I’ll get to the math in a moment, but let me give you the implications first. According to my analysis, if Jane Doe is polling at 30 percent in Iowa and 10 percent nationally, she would be expected to win 34 percent when the final Iowa results are tallied.1 Meanwhile, if Jane Doe is polling at 30 percent in Iowa and 30 percent nationally, she would be expected to get only 29 percent of the Iowa vote.

That dynamic is evident in real-world results too. Take the 2008 primary campaigns: Mike Huckabee averaged 22 percent in Iowa in the second to last month before the caucus but 8 percent nationally. He went on to earn 34 percent of the Iowa vote. Rudy Giuliani was polling at 13 percent in Iowa and 28 percent nationally, and finished with just 4 percent of the Iowa vote. On the Democratic side that year, Barack Obama was polling at 25.5 percent in Iowa and 22.5 percent nationally, and ended up with 35 percent of the Iowa vote.

Of course, this result could just be statistical noise. That is, the result could just be happening by chance. We’ve only had 12 primary campaigns without an incumbent president running since 1980 (the sample I used) — that’s not a ton of data.

Still, there are some common-sense ways to explain why it may be better to do worse in national polls. Before I get to those, though, let me walk you though my analysis. (The next two paragraphs get a teensy bit technical, so skip ahead if you’re not interested in the methodological details.) I calculated polling averages for all primary candidates since 1980 who were present in at least one Iowa, New Hampshire and national poll from 35 days to 65 days before the Iowa caucuses. (In other words, a monthly average at about the point in the campaign where we are now.)2 I then compared each candidate’s polling averages to the results in Iowa and New Hampshire. I did so in a simple linear regression that untangled the relationship between the state polls, national polls and the results in Iowa and New Hampshire.

In the Stata output above, the most important thing to look at is the “Coef.” column. It’s positive and about 1.1 for the Iowa and New Hampshire polls3, meaning candidates finish 1.1 percentage points higher in the final statewide results for every percentage point they do better in a given state’s polls. For the national polls, the effect is negative and around 0.3, meaning candidates do 0.3 percentage points worse in the final statewide results for every percentage point they do better in the national polls. You’ll note that the “Root MSE” (the root mean squared error) is 8.6 percentage points, which means that there is a wide margin of error.

So what might be going on? I can’t be sure, but the candidates who underperform in early states are often those with the highest name recognition — like Giuliani or longtime Republican frontrunner George W. Bush in 2000. (Bush was upset by John McCain in New Hampshire.) It could be that a disproportionate share of these candidates’ support in the early states is due to high name recognition, which the national polls pick up on, and not because they line up well with the state’s voters. As the voting gets closer and more voters tune in, name recognition tends to even out, and voters may decide there are better options out there.

Relatedly, the phenomenon I found could be due to news coverage. Sometimes celebrity candidates receive a lot of national media coverage that helps drive their support in the national polls higher than it would be based on how well they’re actually liked. There’s no guarantee you’ll have the media spotlight forever (though maybe Trump is the exception).

It could simply be that early state voters — who are paying more attention and getting more candidate exposure than national voters — are seeing something in the candidates that people outside the state are not. Maybe a candidate has a flaw or strength that becomes apparent only upon close inspection. And while voters in these early states are paying more attention at this point than national voters, even in Iowa and New Hampshire, many residents don’t really tune in until the final couple weeks or days of the campaign, which is when the preferences of early deciders funnel down to the rest of the state’s voters.

It could also be about whether a candidate is a good fit for Iowa or New Hampshire ideologically and/or temperamentally. That’s something that can be difficult to measure, a mix of perception and reality. Better early state numbers compared to national numbers may indicate a candidate is better lined up with a state’s electorate. Hence, his or her standing in a state will only improve as voters get to know them more. That’s especially the case for candidates who might not be as well-known nationally, such as Huckabee in 2008 or McCain in 2000.

Alternatively, it could be about campaigns’ efforts to persuade and turn out voters. While a lot of people think of the ground game as some mystic force that polls can’t possibly pick up, good pollsters screen for likely voters or a likely electorate. That screen is partly based on voting history, but pollsters also ask respondents if they plan on voting. A gap between the state and national polls may indicate that a certain candidate is either doing a good or bad job of motivating his or her supporters to vote in a particular state. The benefits (or harm) of a good (or bad) ground game may accrue as the campaign goes on.

Candidates with better national numbers may also be more inclined to focus on more than one state. The Trump campaign has been campaigning in Iowa, New Hampshire, the South and elsewhere. That lack of a singular focus may be a relative disadvantage when Iowa and New Hampshire cast their ballots. Meanwhile, Christie has all but moved to New Hampshire and Cruz has courted Iowans with gusto.

This phenomenon may not hold in 2016. Cruz could collapse in Iowa. Christie could fall apart in New Hampshire. And maybe Trump will outperform his polling in both contests. But the gap between their current state and national polls suggests the opposite is more likely.

## Footnotes

1. Remember, you’d expect them to do better than they are polling now even if the race stayed static because there are undecided voters in the polls.
2. For the 1980 campaign, I had to go back to 110 days before the Iowa caucuses to get a New Hampshire poll. The inclusion of 1980 New Hampshire polling doesn’t alter the analysis.
3. Running the regression separately for Iowa and New Hampshire results reveals basically the same thing.

Harry Enten is a senior political writer and analyst for FiveThirtyEight.