The pollsters argue -- Baselice vs. Montgomery
Mike Baselice, Perry's pollster, has been criticizing the poll by Jeff Montgomery that found KBH leading Perry 60-32. Most of the sniping has been on Quorum Report.

When the poll came out, I noted that I didn't think the lead was 28 points. I guessed (and that's what much of polling is) the lead to be about 10 percent.

I'll summarize and excerpt some of the arguments in timeline form.

1. Montgomery releases press release with poll results showing Perry losing to KBH by 28%, but with Perry leading Strayhorn by a similar margin.

2. Perry campaign manager Luis Saenz says:

This data and a $1.95 will buy you a venti-sized cup of coffee at Starbucks. This is a transparent attempt by a Democrat pollster to taint the Republican primary process by cooking up flawed
numbers derived from non-voters and unlikely voters.

3. Montgomery responds:

There are two traditional, accepted methodologies for identifying party and voting propensity in political polling. One is to use a random sample and a series of screen questions; the other is to work from a voter list. Both methods have strengths and weaknesses.

The random-sampling method [that Montgomery used for this poll] requires the pollster to rely on what respondents say about their voting tendencies, but it also allows us to call both listed and unlisted numbers. The voter-list method excludes those with unlisted phone numbers, which is up to a third of the statewide population.

Both methods are widely used and accepted. In fact, Gov. Perry's own campaign pollster, Michael Baselice, uses the same random-sample method we used in this poll -- as do many other reputable national pollsters.

4. Baselice responds. The second paragraph is a summary by Harvey Kronberg.

"It is not accurate to state that Baselice & Associates, Inc. uses the same random-sample method that Montgomery & Associates used in its recent poll," Baselice said. "To determine whether any two sample methods are similar, we would have to compare the calling designs, the number of completed interviews desired in each region of the state, gender quotas, etc."

Baselice said that while his group does use random-digit dialing for statewide voter surveys of the general electorate, it recommends using a voter list of known Republican primary voter households when attempting to measure the opinions of Republican primary voters.

"However, our experience is that an RDD methodology does not produce a
representative sample of primary voters," Baselice said. "The recent
Montgomery & Associates survey attempts to measure support among Republican
voters in a primary election, however we believe the respondents should be
termed "casual" Republicans since they are not 'actual' primary voters."

Baselice said if 46 percent of the adult population in Texas voted in a
Republican primary at least some of the time (as the Montgomery & Associates
poll represents with 478 casual Republican primary voters out of 1,035
adults interviewed), then, proportionally, there should be more than seven
million Republican primary voters identified statewide on voter files.

"There are not even half that many GOP primary voters when we look at recent
primary elections combined. For comparison, there were 7.4 million total
Texas votes cast in the entire 2004 general election for president,"
Baselice said.

Baselice said he does not agree with a sampling methodology that produces
"casual" Republicans and then attempts to portray those opinions as
reflective of "actual" Republican primary voters.

5. Montgomery responds:

I appreciate Michael Baselice's thoughtful comments on our poll. I am also
confident that our numbers are accurate. I might share his concern about
polling "causal voters," but for two reasons.

First, there was no statistical difference in the trial heat response
between those who said they always vote in the GOP primary and those who say
they vote in the GOP primary only some of the time.

Second, I think Mr. Baselice may underestimate just what a high-profile race
this would be. In a district judge or low-profile state representative race,
for example, where only hard-core voters participate, I agree that voter
lists do work better. But I believe that RDD (random digit dialing) is a
better technique when a larger population becomes deeply engaged in a
contest--as in a general election, as he notes, but also in a truly historic
primary like this one, with a sitting senator giving up her seat to
challenge a governor from her own party. It would certainly be a heated and
expensive race as well. I believe turnout would be very high---especially
with a hotly contested primary to fill the empty U.S. Senate seat as well.

So even though it is a GOP primary election, it is likely to include some
general election voters, Independents, and even some Democrats. A generation
ago, it was common for non-Democrats to vote in the Democratic primary,
since that was the only place to make their vote count. As Texas becomes a
Republican stronghold, I believe the same may be beginning to happen in
reverse.

For all those reasons, I believe RDD plus screen questions was the best
methodology for testing this potential race. But the more interesting
question is this: what has the Perry campaign found in its own polling on
this race--specifically, what are the results of any trial heat question
they have tested before testing messages (which sways voter response). I'd
be very curious to hear whether they found that the Governor would begin a
race with Sen. Hutchison ahead or behind in the polls.

Where to begin?

Often people call polling both "art and science," because it is science based on many assumptions. Baselice and Montgomery are assuming different things. Baselice is assuming that the primary election will be a relatively normal primary, while Montgomery is assuming that this will be an abnormal, super-heated primary that will boost turnout. Much of their disagreement flows from that simple disagreement.

Montgomery closes by challenging Baselice to release internal Perry poll results if Baselice disagrees with Montgomery's methodology. Montgomery knows that the Perry campaign won't let Baselice release Perry's internal polls because they will show KBH in the lead. Perry doesn't want to confirm a KBH lead, even if the race is much closer than Montgomery's poll suggests.

My take: Like I've said before, my guess is that KBH leads Perry by around 10 points right now among likely primary voters. Perhaps it's more like 15%, but I'm comfortable guessing a lead of around 10%.

Is Montgomery right that this primary will see a huge turnout? Well, my view of political history is that many campaigns have been based on turnout. They've almost all lost (for a recent example, see Sanchez vs. Perry in 2002).
- posted @ 11:15 PM permalink