About this Blog
About The Author
Email Me

RealClearPolitics HorseRaceBlog

By Jay Cost

« Mitt's Ham-Handed Campaign | HorseRaceBlog Home Page | On Clinton's Attacks »

Why I Can't Call Either Nomination

This nomination cycle has been one of the most interesting since the open era began in the 1970s. Only 1988 really compares to it.

This cycle has been a surprising one, too. Time and again the conventional wisdom has been abandoned. Guiliani was not supposed to maintain a national lead. He has. Romney was not supposed to face a serious challenge in Iowa. He does. Clinton was supposed to be inevitable. She isn't.

For my part, I have been relatively quiet in predicting what will happen. While I have my opinions about the future, I have kept most of them to myself. Readers of my 2004 blog might be surprised by this. That year, I made some pretty bold predictions while the mainstream press shrugged its collective shoulders and declared the race a toss-up. This year, it has been the other way around. Why's that?

The big difference is that in 2004, there were some solid, reliable indicators that pointed toward a Bush victory. They were not obvious indicators, but they were there. In this cycle's Republican and Democratic nomination contests, those kinds of metrics are lacking. And when you don't have good measuring sticks - you can't measure very well!

There are five reasons that I have remained stubbornly agnostic about this nomination contest. They apply to both the Democratic and Republican nominations. I have been hinting at them all cycle. Now that both races are clearly more competitive than most pundits once thought, I figure people might appreciate a more formal treatment of my reticence.

(1) Any estimation of what will happen must be predicated upon what has happened in the past. This goes for politics, physics, chemistry, anything. Empirical knowledge proceeds by trial and error. We observe the same process again and again - and in so doing, we acquire a data set that enables us to test theories about why the process occurs as it does.

The following axiom is applicable to all empirical knowledge, regardless of whether it is quantitative or not: as the number of observations decreases, the expected divergence between our prediction and the actual result increases, and the precision of our prediction decreases. Simply stated, the fewer the observations, the less confident we can be about our predictions.

The "open" era of presidential nominations began in 1976. This was the first year that primaries and open caucuses were widely used as something more than "beauty pageants." Instead, they became the decisive factors in delegate selection. This means that we only have a total of sixteen observations (eight presidential elections times two party nominations) to draw inferences about the 2008 cycle. This is the greatest problem for predicting the outcome of this nomination cycle. This small data set prohibits us from being precise.

(2) Further complicating this process is that this number has to be qualified by the type of contest. It is all well and good that there have been sixteen total nominations in the open era - but the 1984 Republican contest is practically no help for us when it comes to understanding 2008!

Scholars tend to classify the nomination contest in one of four ways:

(a) One candidate campaigns. These would include the Democratic contest in 1996, and the Republican contests in 1984 and 2004. They are nomination contests in which an incumbent president runs without major opposition.

(b) Two candidate campaigns. These are frequently campaigns where an incumbent president is challenged by an ideologically distinct opponent. So, here we would place the 1976 and 1992 Republican contests, and the 1980 Democratic contest. We would place the 2000 Democratic contest here, too.

(c) Multi-candidate contests without an obvious frontrunner. These have been exclusively Democratic until this year. These have included the Democratic contests of 1976, 1988, 1992, and 2004.

(d) Multi-candidate contests with an obvious frontrunner. Here, we would have the 1984 Democratic contest, as well as the Republican contests of 1980, 1988, 1996, and 2000.

These contests have vastly different dynamics - and so it is difficult to draw inferences about, say, the 1988 Democratic contest from the 1992 Republican contest. If we want to be strict about things, we would not make any cross-type comparisons. Minimally - it would be irresponsible of us to make cross-type comparisons without serious qualifications. After all, comparing a type (b) race with a type (d) race introduces at least some fuzziness into the picture.

We would probably place this year's GOP contest in category (c), and the Democratic contest into category (d). This means that we have only four previous observations to get a read on the GOP nomination contest, and five for the Democratic contest.

What is more, we might claim that there is an apples-to-orange quality about comparing Democratic contests to Republican contests. This means that, once again, we probably should make cross-party comparisons only with some real caveats put in place. Doing so means that, practically speaking, the number of observations drops even further. After all, there is only one other Democratic contest in category (d) - and there has never been a Republican multi-candidate contest without an obvious frontrunner before this year.

(3) Further complicating this is the fact that there has been a secular change in the nomination process itself. Namely, it has slowly become front-loaded. The path that Jimmy Carter took to acquire the nomination in 1976, building momentum slowly-but-surely via a string of small victories, is no longer possible. More than half of the country will have finished voting five weeks after the Iowa caucus. The implication is that our prior data points provide even less of a guidepost than point (2) allows. After all, they were the product of a process that is different than today's.

(4) Despite the plurality of contest types and the secular trend in the way contests are scheduled, we used to have some stable indicators we could utilize. Again, we had a paucity of data, so we had to be cautious about drawing inferences - but we once had at least a few stable indicators that seemed reliable guides despite all this uncertainty.

Namely, after 1976 candidates who had a lead in the final national polls of the pre-election year were always the best fundraisers. They also always won the nomination.

Until 2004, that is.

Howard Dean was leading in the national polls at the end of the 2003. He had also raised more money than any of his competitors by the end of 2003.

The fact that Dean lost despite his leads could be explained in one of three ways:

(a) Money and poll position were never decisive causal factors in winning the nomination. Instead, they usually correlated with some (still unknown) causal factor, which is the true reason a candidate wins. The relationship between money, polls, and victory is therefore at least partially spurious.

(b) Money and poll position are usually decisive causal factors - the result in 2004 was an aberration induced by the unique circumstances of that particular election. Like much of life outside Newtonian physics, the political environment is stochastic - and so general processes are often interrupted by unique, unpredictable factors.

(c) Money and poll position were once decisive causal factors, but the nature of the nomination process has changed such that they are no longer sufficient causes of victory. Other factors have come into play. Perhaps it is the frontloading - Dean never had an opportunity to recover from what was basically a small and potentially insignificant loss. Perhaps it is the fact that party elites (i.e. those who fund candidates and give them early boosts) have become more partisan than the party rank-and-file - so that a hyper-partisan candidate like Dean could develop an early lead only to see it wash away to a more mainstream Democrat like John Kerry.

Which one is correct? Your guess is as good as mine - though I tend to agree with Dean McSweeney. The process that connects money, poll position, and victory is more complicated than we once thought. Either it always was more complicated (point a), or the change in the cycle has introduced complicating factors (point c). Until we have some more observations under our belt - we cannot know for sure. This, in turn, means that we cannot be as confident as we once were about these metrics.

(5) Even if Howard Dean's fall had never occurred - we still would have trouble using these two metrics this year.

On the Republican side, there is a split between the money leader and the poll leader. Giuliani has led in the polls all year. Romney is the fundraising champ. On the Democratic side - while Clinton has a lead in the polls, she and Obama are essentially tied in terms of money. Obama has a small lead on her when we factor out the amount she contributed from her Senate reelection fund.


So - there it is. Five reasons not to be hasty in your conclusions. At this point, I see no metric to determine who will win the nomination. All I'm left with are my gut feelings - and I try not to publish them.

-Jay Cost