Tuesday, November 12, 2013

The Predictive Power of (Very) Early Presidential Primary Polling Part I - Historically Unreliable

Colin Powell was grabbing headlines as a potential candidate for President in 1996 as early as 1993. Though he never actually led Dole in primary polling until just before he announced he would not be a candidate in November of 1995.

Though the 2016 presidential primaries won't officially get under way for another two years, the shadow campaign is upon us, as indicated by the flourish of recent polls and articles on the subject.  And the tea leaves for both parties couldn't be more different; for Democrats, the race appears to be Hillary's to take - if she wants it, while the Republican race is anyone's guess.

Hillary Clinton, should she decide to run, looks poised to wrap up the Democratic nomination in one fell swoop, as early polling has shown her very strong in the Democratic primary - stronger, in fact, than any candidate in a contested primary dating back to 1976.

Republican primary polling has been a bit more topsy-turvy to date, featuring no less than five different leaders over the last year (making it awfully reminiscent of the 2012 primary process). Those leaders have been Ted Cruz, Chris Christie, Paul Ryan, Rand Paul, and Marco Rubio:

In spite of the deluge of early primary surveys since last year's election, there are many who find the early speculation premature at best, or entirely futile at worst. “But early primary polling isn’t predictive of actual primary results!” is a frequent refrain for the many that abhor such early conjecture about an election that’s still three years away.

So in an attempt to settle the issue, I decided to take a look at the predictive power of Democratic and Republican Presidential primary polling from 1976-2012. And there were more than a few surprises.

Note, Nate Silver did something similar a few years ago, though if you click on the link, you'll see his focus was on polling one-year out from the start of actual primary contests. This piece will examine the predictive power of those polls two and three-years out from the start of the contests (where such polling data is available).

First off, early primary polling is RARELY predictive of actual primary results. Though there have been exceptions.

We'll begin this analysis by looking at those primaries in which the first year-to-two-years of primary polling following the previous presidential election wound up being predictive of the final result. Of the 15 total contested Democratic or Republican presidential  primaries since 1976, only 3 can arguably claim to have foretold the actual result through very early polling.


Of the three "predictive primaries" noted above, unquestionably the most foreshadowing of the actual result was the 2000 Democratic Primary contest between Al Gore and Bill Bradley. Consider the table below:

The two-term Vice President was more fortunate than Joe Biden is today, in that it was he who was considered the Democratic standard-bearer as Bill Clinton left D.C., and early 2000 primary polling showed it. Of the eight Democratic primary polls taken in the year following the 1996 presidential election, Gore led in all of them, obtaining no less than 28% of the vote, no more than 56%, and averaging 44%. That average jumped nearly 10 points by 1999, and peaked at 63% as the primary contests waged on.

Of the 122 surveys taken of the 2000 Democratic primary electorate between 1996-2000, Gore led in each one. Bill Bradley's best showing during the state contests came with a 50-46% loss in New Hampshire. By 'super Tuesday' on March 9th, Bradley withdrew. Al Gore ultimately won 75% of Democratic primary voters, vs. Bill Bradley's 21%.

Graph of 2000 Democratic Primary Polling, 1996-2000.


The political forecasters that relied on early '88 GOP primary polling in framing their presidential predictions wound up looking awfully prophetic that year. Like Al Gore 10 years later, two-term VP George H.W. Bush faced minimal interparty opposition in his quest for the presidency. Against a field of lesser-knowns like former Delaware Governor Pete du Pont, New York Congressman Jack Kemp, and former Secretary of State Alexander Haig, George Bush led consistently.

Of the 12 surveys taken in the two years following the 1984 Reagan landslide, George Bush led in all but one. His performance among GOP voters never dipped below 26%, never rose above 47%, and averaged 39%. His closest competitors (Bob Dole and fmr. TN Senator Howard Baker) both averaged 18% in the early days of the '88 campaign. And while 1988 primary polling was far less numerous than in 2000, Bush was still roughly as strong as his 2000 Democratic counterpart. Of the 41 total GOP primary surveys taken between 1984 and 1988, Bush led in 38, while then-Senate Minority Leader Bob Dole led in just three. Though it's worth pointing out that Bush never rose to 50% or higher in any GOP primary poll until AFTER Super Tuesday in March, 1988.

Graph of 1988 Republican Primary Polling, 1984-1988.

In the end, Bush carried the 1988 Republican Primary with 68% of the vote, to Bob Dole's 19% and televangelist Pat Robertson's 9%. Dole's best poll result against Bush came in a December 1986 Gallup survey showing him ahead of the Vice President in a 2-way race 42-36%*, as the Iran Contra Scandal took their toll on the Reagan Administration's approval ratings.  Pat Robertson never attracted more than 14% in any poll.

Fun fact: Ex-President Gerald Ford, who would have been 74 years old at the time (younger than Reagan), attracted 14% and 7% respectively in February, 1986 ABC/WaPo and LA Times polls.


The last bit of early primary polling one could argue was truly representative of final results was the 1996 GOP contest, which pitted the 1988 'runner-up' against a slew of participants ranging from a perennial candidate to a moderate ex-military General. Senator Bob Dole eventually became the obvious, strong frontrunner of the race, though his polling advantage in early surveys was less significant than Al Gore's in 2000, or even George Bush's in 1988.

Dole averaged nearly 30% of GOP primary voters in early 1996 polling, to Colin Powell's 23%, Kemp's 16%, and Quayle's 14%. Dole's comparatively small-but-discernible edge became even more pronounced as the primary season wore on. In 1995, he averaged 42% across 53 surveys, while in 1996 he averaged 43% across 22 surveys. Out of 84 total polls taken between 1992 and 1996, Dole led all but four (each one he lost was to Colin Powell in the Fall of 1995). He wrapped up the primary early, despite a narrow win for Pat Buchanan in New Hampshire, with 59% of the vote, vs. Buchanan's 21%, vs. Steve Forbes 11%.

Graph of 1996 Republican Primary Polling, 1992-1996.

Side note: get a glimpse of Dick Cheney's polling performance in the first chart under 1996 primary heading. Yep, that was our yet-to-be-former-Vice President, polling consistently in the double digits in early 1996 presidential primary. Of course, Cheney never took the plunge. But given his early strong showing, it's fun to wonder what if...Oh, and Republicans apparently seriously considered nominating Dan Quayle as their candidate for President, though for only a brief period in late 1994, early 1995. He averaged 18% in primary polling during this time period.

______                            ______                            ______

Obviously, if your name is Hillary Clinton, you're hoping 2016 primary polling is mimicking the early numbers from the 1988 and 1996 Republican, and 2000 Democratic primaries. That's because all 3 featured moderate to strong frontrunners who maintained that status throughout the entire primary, and went on to win BIG in the end. Unfortunately for Hillary, she knows all too well that big, early primary leads can evaporate over a matter of just days.

And if you're one of the many Republicans who has led in a GOP primary poll so far this year, the early numbers have given you very little to go on in the way of predictive value. That will change in my follow-up piece tomorrow, which examines the rest of early primary polling from 1976 to present day and finds that more often than not, the very early primary numbers tell us little about who will actually wind up as their party's presidential nominee.

*Gallup/Newsweek Poll, Dec, 1986. Retrieved Nov-1-2013 from the iPOLL Databank, The Roper Center for Public Opinion Research, University of Connecticut. http://www.ropercenter.uconn.edu.libproxy.uncg.edu/data_access/ipoll/ipoll.html

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.