Friday, September 26, 2008

Who You Callin' Underpolled?

Just last month FHQ took a look at the relationship between the competitiveness of a state -- as measured by our weighted average -- and the frequency of polling in that state. The expectation is that the more hotly contested a race is in a particular state, the more apt we are to see a higher number of polls. The regression of this relationship (with the state's number of electoral votes thrown in as a control) explains just under 60% of the variation that we see in the number of polls from state to state. More than that, though, it allows us to predict where a state should be in terms of the number of polls given both its competitiveness and number of electoral votes.

[Click Graph to Enlarge]

In other words, that provides an indication of whether a state has been overpolled -- a concept I don't personally believe in -- and states that are underpolled. Ah, now there is something that is of interest. But there has been an awful lot of polling conducted between last month prior to the conventions and now, at the outset of debate season. The same cast of characters is still there though. Among the toss up states, Florida, Ohio and Pennsylvania are getting polled more than any other state. And that has a lot to do with the head start each had. The advantage each had last month is still there and even augmented now. However, close states like Nevada and Indiana continue to be underpolled in light of how tight they are. New Hampshire, New Mexico and Missouri are also just under where we would expect them to be given their levels of competitiveness.

There's been a lot of talk about why it is that Nevada is underpolled. It has been said that the Silver state is notoriously hard to poll. But why? Well, fortunately FHQ has someone on the inside to help us all understand the polling situation in Nevada. The other day I spoke with David Damore from the University of Nevada-Las Vegas. Dave has stopped by here on occasion to comment on topics relating to Nevada, especially during primary season. So I put the question to him and here is what he had to say on the matter:
"There are a couple of things going on here. First, there is not an entity akin to say the California Field Poll in the state that consistently polls. Thus, what we get is a hodgepodge of polls done by the two biggest papers in the state (Las Vegas Review Journal and the Reno Gazette Journal) and whatever national firms take an interest (which tends to be pretty sporadic). I do not know much about the RGJ polling process, but from what I can glean from the RJ methods is that they are not good. They typically use small samples, which yield large margins of errors and they over sample rural Nevada intentionally because the rurals tend to have higher voter turnout, but of course vote 4 to 1 in favor of the GOP, so their polls always have about a four or five percent pro-GOP bias. For instance, a couple of days before the 2004 election the RJ had a poll with Bush up around five points. At the time I was on a radio program with the Kerry guy in NV and their polls had the state dead even.

"The second big issue is finding the voters. In particular, Las Vegas is a very transient place and pretty much anyone under 30 is cell phone only. This latter group is not included in any of the sampling frames and given that they lean overwhelmingly Democratic, they are missed. This is my guess as to why the latest NV polls are favoring McCain; a dynamic that is at odds with what is happening on the ground here."
Now, what is Indiana's excuse?


Recent Posts:
Nothing to see here, folks.

The Electoral College Map (9/26/08)

The Electoral College Map (9/25/08)

8 comments:

  1. According to that regression line, Hawaii should have a negative number of polls?

    ReplyDelete
  2. Also, Indiana's excuse is probably its history and the fact that it's favored McCain for most of the election cycle.

    ReplyDelete
  3. Right. Talk about overpolled.

    I think Utah may actually be in the same boat. I left the office in a hurry to catch a ride home this afternoon and didn't save the data (wonderful!), or I'd double check that for you. I'll need to look into constraining the predictions at zero if I can.

    ReplyDelete
  4. There is another important variable here: the number of electoral votes a state has. Ohio, Pennsylvania, and Florida are heavily polled by the national firms because they're close and they're big prizes. Further to the right, California and New York are also above the line. Nevada still has no excuse on that score being polled less than New Mexico or New hampshire, though.

    ReplyDelete
  5. Agreed on Indiana. It still should be surveyed more though.

    ReplyDelete
  6. EVs are in the "full" model, Scott, but not in the graph. Once I get the data set back to where it was I'll post a follow up. I can put two graphs on top of the other, so I'll plot the predicted line for the model with EVs and put it over top of this one to show the difference.

    ReplyDelete
  7. One last question on this post: do these totals include Zogby polls?

    ReplyDelete
  8. Jack,
    It does. I was already planning a follow up to this post this week. Those Zogby polls will be dropped from consideration there.

    ReplyDelete