For a better browsing experience, please upgrade your browser.

FiveThirtyEight

Politics

We’re more than 20 months from the first Republican presidential primary, and the pool of candidates is inchoate. Pollsters are asking voters about 10 candidates, none of whom registers higher than 20 percent support in early state or national surveys. It’s the most divided GOP field at this point in the cycle since the parties reformed the nomination process in the 1970s.

But on the Democratic side, Hillary Clinton has the field pretty much to herself, and there’s little sign that party leaders are encouraging others to run (as they did in 2008). Despite the fact that she hasn’t declared her intention to run yet, she’s polling at 60 to 70 percent — the strongest position for any non-incumbent in the modern era and 30 points higher than at this point in the 2008 cycle.

Early polls often don’t foretell the eventual margins of the primaries, something that shouldn’t be surprising. But what if Clinton in fact wins the Democratic nomination in a landslide, while the Republican nominee does so only after a long and close race? As Karen Tumulty and Robert Costa reported in The Washington Post, GOP elders believe that scenario would lead to problems for their eventual candidate. Are they right? Could a Democratic sweep and a drawn-out Republican contest offer insight into the outcome of the general election?

History suggests possibly. There’s been a strong correlation between the margin of victory in the primaries and the later margin in the general election. But correlation isn’t always causation. Let’s look at the data.

To start, I calculated the margin of victory for each party’s nominee through past primary seasons. In 2012, for example, Barack Obama won the Democratic nomination by an aggregate 87.2 percentage points based on the votes cast in each state, while Mitt Romney won the GOP’s by 31.7 points. I then compared the difference between the two parties’ nominees (55.5 points in this case) to the general election margin of victory (3.9 points for Obama).

The relationship is illustrated in the chart below. The x-axis shows the difference between the two nominees’ margin of victory in all caucuses and primaries, and the y-axis is the general election margin between the same two candidates.

Enten.Primary-vs-General

As the graph shows, general election success has usually followed primary strength. Each extra margin-of-victory point for the incumbent party in the primary season equals a little less than an extra .2 points come the general election.

If the primary votes were held today and the polling was accurate, Clinton would win the Democratic nomination by 57 points, while the Republican nominee would win by the slimmest of margins. Based on the graph, Clinton would then win the presidency by about 7 points.

Of course, the eventual primary result is likely to be different from current poll results. That means the model has a relatively wide margin of error. Still, Clinton’s lead is large enough that even if her primary margin of victory dropped by 30 points (or the Republican nominee’s rose by 30 points), she’d still be expected to win the general election.

It’s possible that President Obama’s low approval ratings could hold Clinton back. As many political scientists have found, these ratings correlate strongly to general election results.1 If voters are happy with the incumbent, they’re likely to re-elect his party to the presidency. If they are dissatisfied with the incumbent, they’ll probably kick the party out of office.

Yet even when controlling for the president’s approval and net approval ratings at the end of the primary season,2 the difference in primary margin is still a significant determinant of general election results.3

For example, if we were to hold Obama’s current net approval rating (about -9 percentage points) constant, and if the current 2016 primary polling were exactly accurate, Clinton would be slated to win the general election by 3.5 points.4

Why does the primary margin of victory seem to indicate who’s going to win the general election? Partly because presidents and their vice presidents almost always cruise to the nomination. Obama had no real challenger in 2012, for example. After all, primary voters aren’t likely to turn away an incumbent who looks like he’s going to win the general election. On the other hand, incumbents who do have trouble in the primaries — like Gerald Ford in 1976, Jimmy Carter in 1980, and George H.W. Bush in 1992 — go on to lose.

The primary margin may also provide some information about candidate quality. Many political science models (see Emory University’s Alan Abramowitz’s for one example) simply do not account for who is running. That’s partly because it can be hard to come up with tangible definitions of candidate quality. Voting results from the primaries are quite tangible, however, and they may provide some evidence that candidate quality matters.

It’s also possible that a longer nomination process can leave a candidate damaged, say by draining her funds or forcing her to take positions that do not appeal to general-election voters. Candidates often get a “bounce” in the polls once they wrap up their nominations and begin pitching their message to the general electorate. But there could be some hangover effects. For instance, the positions Mitt Romney staked out on immigration reform early in 2012 could have hurt him with swing voters in November.

One final caveat: the current nomination process is something relatively new in American politics, and this analysis is based on only 10 data points. That’s always a reason for caution.

But if the relationship holds for 2016, Clinton could win even given Obama’s low approval rating. We’ll have to keep an eye on the primary polls to see whether the Republican field remains a mess, and whether Clinton dominates the Democratic side as expected.

Footnotes

  1. One notable exception is Helmut Norpoth’s Primary Model. He has shown that the New Hampshire primary in particular is good at forecasting a vote. His model has a few key differences from mine, such as his inclusion of data from the years before all states used caucuses and primaries to decide their nominations. He also tries to strengthen his model’s explanatory value by adjusting the primary-vote variable in complex ways. My goal is to keep my model as simple as possible. ^
  2. All net approvals from 1976 to 2012 are taken as an average of May and June Gallup polls. ^
  3. Ninety-four percent of the total variation in results is explained by a model that includes both the net favorable of the incumbent president and the primary-vote variable. That 88 percent variation becomes 64 percent when I use a model that only has the primary-vote variable. The standard error for the primary-only model is 5.03 points, and for the approval model it’s 2.91 points. ^
  4. For a Republican to be favored to win, according to the model, he’d have to narrow the difference between his and Clinton’s margins of victory in the primaries to about 33 points. ^

Filed under , , ,

Add Comment

Powered by WordPress.com VIP