So, because many Clinton supporters have been pointing to them lately, supposedly as a sign that Clinton's lead is both large and stable, I was flipping through
state polls on the Democratic primary. Here are some quick insights:
- Clinton has a northeast base. There are nine states were multiple polls have shown Clinton ahead by an average of more than double digit margins since the start of February: California, Connecticut, Florida, Massachusetts, Nevada, New Jersey, New York, Ohio and Pennsylvania. Six of these states are in the northeast, and one of them, Florida, has a large northeast émigré population. Clearly, Clinton's strength is to be found in the northeast.
- Home states There are four states with multiple polls since the beginning of February where Clinton is trailing: Illinois (to Obama), Iowa (to Edwards), New Mexico (to Richardson) and North Carolina (to Edwards). All but one, Iowa, is a homestate for one of Clinton's main competitors. It is actually quite telling that Clinton leads comfortably in Delaware and Connecticut, against Biden and Dodd respectively. While Richardson is able to hold his home turf, just as Edwards and Obama unsurprisingly hold theirs, Biden and Dodd can't even hold their homestate supporters. Keep in mind that Edwards, Obama and Richardson were only elected statewide a combined four times, while Biden and Dodd were elected at least that many times each. It is interesting to note that the state sporting the largest Clinton lead, 33%, is Arkansas. I guess they still consider her one of their own down there.
- Early States. The other area where Clinton is somewhat weaker is generally found in early states. In addition to the four states where she trails, multiple polls have shown a single-digit race in five states: Arizona, Georgia, Michigan, New Hampshire, and South Carolina. It is interesting to not that of the seven northeast states with multiple polls since February, famously early New Hampshire is the only one where Clinton does not currently hold a sizable, double-digit lead. Overall, Clinton's lead is noticeably smaller, or even non-existent in the pre-February 5th, and potentially pre-February 5th, early states of Iowa, New Hampshire, Michigan and South Carolina. Keep in mind that Arizona also has a long history of early primaries. The lone exception to the home-state / early state rule for Clinton weakness seems to be Georgia, where Obama's strength probably comes from African-Americans making up the largest percentage of the Democratic primary electorate of any state in the country.
- State polls not so out of line with national polls. Considering the eighteen states with multiple polls since the start of February, I see no real disconnect or other incompatibility between state polls and national polls. The key is to filter out those states with very few data points. Take California, for instance. In the five California polls since the beginning of March (see also Pollster.com here for one poll wikipedia left out), Clinton's average lead was 14.6%, and her median advantage was 13%. During the time period when those five polls were taken, March 2nd through April 12th, Clinton's average lead in national polls, according to my estimate, was 11.5%. That is only a 3% gap in the mean, and a 1.5% gap in the median. What difference there is can hardly be considered surprising, since California is not the home state of any Democratic candidate, is not a pre-February 5th state, and does not have a large African-American population in the primary electorate. And there is always margin of error, too. Whatever difference there is between California polls to date and national polls taken at the same time period is extremely minimal, and not demonstrative of anything in particular.
This actually leads to an instructive point. Instead of looking at individual state polls or individual national polls, it is always best to look at averages across polls. Right now, there simply are not many states where there are multiple polls to average. In the eighteen states where there is at least a rudimentary average, there is no real disconnect between national polls and state polls that cannot be chalked up to the early state / home state / high African-American concentration exceptions. Further, recent movement to Obama measured in national poll averages over the last two weeks cannot be captured in state polls, simply because there are so few statewide polls in that time period. Of course Clinton is going to do better in older polls, in non-early states, in non-home states, in states with low African-American populations, and non-Rasmussen polls (Keep in mind that Rasmussen lowers the national poll averages by a couple of points, and is not going to conduct any state primary polls in 2008). And of course she will do a little worse in most newer polls, most early state polls, most home state polls, most Rasmussen polls, and most polls with large African-American samples. The key to remember is that national polling averages are made up of universes sampled in all these kinds of polls, (except older polls). Thus, national polling averages will always produce results slightly less favorable to Clinton than will polls of February 5th states like California, Ohio and New Jersey, which tend to be the most favorable to Clinton. This doesn't mean that either national polls or state polls are flawed--it just means that it is important to take the entire polling picture into account when drawing conclusions about the campaign, rather than picking and choosing the polls you like best. Don't use isolated, individual polls as a means to develop an anecdotal, circumstantial and context-free picture that favors your candidate.
Overall, I am struck by the homogeneity in state primary polls. There simply no longer appears to be any regional blocks to speak of in the Democratic Party. Sure, there are slight differences, but we are only talking about 5-10% at most. Even among demographic groups in the party, with few exceptions Pew's recent poll showed the difference among those groups was under 10%. The entire list of "very few" exceptions includes: regular blog readers (pro-Edwards and Obama), conservative Democrats (pro-Clinton), African-Americans (anti-Edwards), Dems under thirty (pro-Obama), those who never attended college (pro-Clinton), those with incomes over $100,000 (pro-Obama), those with incomes under $30,000 (pro-Clinton), and seculars (pro-Obama, anti-Clinton). Obviously, as repeated demographic surveys of the blogosphere have shown, there is a lot of cross-over between regular blog readers, those with high incomes, and seculars.
All in all, this probably means that any candidate who is ahead by 6% or more in national polls going into February 5th will virtually sweep every state that night. There just is no longer enough regionalism and demographic skews in the Democratic Party to prevent this. In order to produce a lengthy campaign, the national picture will need to be very, very close heading into February 5th. Even if the leading candidate is ahead by only 8% nationwide on February 4th, which is the current amount I estimate Clinton to be ahead of Obama nationwide, the campaign will probably be over on February 6th.
