Many of the longtime readers of dailykos probably remember my "Empirical Cattle Call" project, which I updated semi-daily for the final five months of the invisible primary leading up until Iowa. If any of you were following it closely, you will also probably remember that my final projections were, um,
slightly off base.
For the past two months, I have mostly been away from the blogosphere that I so grew to love during 2003. During that time I have been working at a number of organizing projects for the Illinois Federation of teachers, and the travel, combined with the work and my general post-primary funk have significantly reduced my blogging time. Still, during those two months, once and a while I wondered, "why was I so damn wrong?" Well, now that I am back in Philadelphia for a glorious free weekend, I decided to spend Friday morning re-calculating the final Empirical Cattle Call numbers based on information I have learned over the past two months. Here are the most important things I should have known and recognized about the contemporary primary season when constructing the Empirical Cattle Call:
- Momentum, a primary season dynamic significantly on the wane during the late-1980's and throughout the 1990's, was back in full force in 2004. This was largely due to three factors: the rapidly declining influence of paid political advertisements, the massive decrease in campaign coverage on the part of televised news, and the still declining tendency of Americans to join groups and organizations (I think my newly reborn commitment to joining, researching and discussing politics blinded me to these trends). Taken together, these three factors have significantly restricted the amount of campaign related information within the electorate prior to elections, and thereby created a far more volatile invisible primary season. Any lead for a frontrunner is far more tenuous than in the past, as people claiming they will vote for candidate x are doing so based on far less information than in the past (and thus are more likely to switch as new information becomes available).
- Momentum can actually be measured (or at least estimated). Although I wish I had known when I was constructing the Empirical Cattle Call, over the past thirty years, the average national poll bounce for the Iowa winner is 14 points, and the average national poll bounce for the NH winner is 9 points (this year, Kerry received around a 20 point national bounce for Iowa, and a 12 point bounce for NH). It would not have been difficult to incorporate this bounce into the Empirical Cattle Call. All I needed to do was give whoever was leading Iowa at any given time 14 points in every other poll in the ECC, and every other candidate 14% less. After that, I could do the same, but with 9%, for New Hampshire.
- All press is clearly not good press. However, since measuring the "tone" of every single news story on a candidate is not quantifiable, it would have been against the founding principle of the ECC to even attempt such a measurement. This is not even to mention how such a measurement would have required more hours than there are in a lifetime. Thus, I should never have included media coverage as a category in the Empirical Cattle Call.
- Because momentum was indeed an important factor, I should never have included any polls after the seven February 3rd states. Before Iowa, none of the major candidates were seriously campaigning in any post-February 3rd state. Combine this with the inevitable post-2/3 dropouts, and I should have considered anything after 2/3 a blank slate.
- I should never have used Cash on Hand. I should have used total money raised, just like all the academic models that served as the basis for my model used. For starters, no one actually knows what the Cash on Hand is at any given time, and for seconds the total fundraising picture offers a better glimpse into future fundraising possibilities.
Had I realized all of these points, my final projections would have looked like this:
Kerry: 100.0
Dean: 73.9
Clark: 57.0
Gephardt: 26.8
Lieberman: 25.1
Edwards; 23.4
Sharpton: 10.4
Kucinich: 5.7
Iowa: Kerry, Dean, Edwards
New Hampshire: Kerry, Dean, Clark
Arizona: Clark, Kerry, Dean
Delaware: Kerry, Dean, Clark
New Mexico: Kerry, Dean, Clark
Oklahoma: Clark, Dean, Kerry
South Carolina: Kerry, Clark, Dean
Missouri & North Dakota: No information
Super Tuesday viable: Kerry, Dean and Clark
Nominee: Kerry
Of course, this would still have been wrong, as it would have vastly underestimated Edwards. On the morning of the caucuses, however, I don't think there was any reliable way to imagine that Edwards would surge to the extent he eventually did. Most polling information that morning had him a close third behind Dean, and still nowhere to be found outside Iowa and the Carolinas. Had pre-caucus Iowa polling been accurate, and had there been no big "scandal" moment for Dean that night, February would probably have been a very different month. My best guess on Edwards's surge is that, in the end, the invisible primary was almost entirely about Dean vs. non-Dean. Because Edwards beat Dean in Iowa, Edwards's national bounce was roughly the same as if he had won the state.
Anyway, political prediction is a tough game, and I bet wrong. Had I predicted a Kerry nomination the morning of the Iowa caucus, and then had Dean won the state, I would have had an even bigger plate of crow to eat than the one I'm still trying to finish. Even beyond any changes to my formula, perhaps the best change I could have made the Empirical Cattle Call would have been to be more humble, and have never believed my hobby was anything more than amateurish.
Coming soon:
Behind the polls: The Empirical Cattle Call, General Election Edition!