Michael Barone wrote a piece this week where he suggested that Clinton could still win the popular vote. I took this as a challenge: how would I predict the popular vote contest, and what margin would Clinton require in order to take the popular vote lead.
Before I begin my analysis, I want to make clear that I think this is the wrong standard to apply. In my review of past close primary fights (1976 GOP, 1980 and 1984 Democratic) the standard has always been pledged delegates. This makes sense: if popular vote were the standard no state would ever choose to elect delegates by caucus. In fact, applying the popular vote standard is tantamount to disenfranchising caucus states, some of which are true swing states (Iowa, Minnesota, Colorado).
Having said this, let's start by trying to predict the popular vote outcome. The hardest part is predicting turnout. Barone used the Kerry turnout in 2004. However, if you bother to examine the turnout to date, you find this reasoning fundamentally flawed. To try and predict the turnout in the remaining 10 contests, I began by examining the turnout in the primary contests to date. This allowed me to arrive at a turnout prediction for each of the remaining states. I then used an average of recent polling (data from realclearpolitics.com and pollster.com) to predict the results.
As this table shows, Clinton would need a 14 point average margin in the remaining states to overtake Obama (9% if you include Florida).
How did I get these numbers? This table shows predicted turnout, and the results (based on the most recent polling data). Note that I could not find recent polling in Oregon. More on the flip..
One of things this analysis shows is the relationship between North Carolina and Pennsylvania. Based on current polling Clinton only nets about 8,000 votes between Pennsylvania and North Carolina. Moreover, there are solid reasons to suggest that my predictions are at the high end for Pennsylvania and the low end for North Carolina. As Chris noted today, Pennsylvania was an extremely hard fought state in 2004. Conversersly, North Carolina was conceeded early in the process. As a result, one would expect turnout in North Carolina to be much closer to Kerry's '04 number than turnout in Pennsylvania.
To estimate turnout, I compared the Kerry 2004 turnout versus the primary turnout in 2008. Here is what I found:
Obviously the relationship between the Kerry turnout and the 2008 primary after the Super Tuesday primaries is the more relevant since
the GOP contest was essentially resolved on Super Tuesday. Perhaps the best states to use for predicting the Pennsylvania turnout are Ohio and Wisconsin. Both of these contests were strongly contested in 2004, and both were the focus of significant efforts by Clinton and Obama. Total turnout in the Ohio primary was 81% of Kerry's vote in 2004 and turnout in the Wisconsin primary was 74% of Kerry's total. However, unlike Ohio and Wisconsin, Pennsylvania is a closed primary.
My state by predictions are based on assuming that Pennsylvania turnout is 75% of Kerry's, and for the rest of the primaries I assumed turnout of 80% for open primaries and 70% for the remaining closed primaries.