I thought it would be interesting for everyone to take a look at the complete dataset on the Democratic primary for Governor from the poll that, as Ben notes below, is the subject of a front-page article in today’s Globe:
Here are a couple of things that leap out at me.
What do you see?
Please share widely!
56-16 pct in the less than 30k income group , 38-20 in the less than 60k to 30k income group, and a shocking 39-6 African American vote,all in Reilly’s favor dictates that he will pull a large number of votes from the urban areas. In both the primary and general election these numbers are key because of the huge number of votes the cities generate. If Reilly simply breaks even in the suburbs he wins.
You are Tom Reilly and you have been either DA in the state’s largest county or AG for fifteen years. You claim the support of just over one third of the Democratic electorate. You are only about a dozen points ahead of a guy that half the voters don’t know yet. That guy is going to beat you at the state convention in early June, a convention that will be the biggest media event of the campaign to that point (with the possible exception of the February flap over your running mate). That guy is also more articulate and compelling than you are — as people get to know him they tend to like him.
<
p>
You are mighty glad you have $3 million more than that guy. You’re going to need it.
<
p>
BTW, we can toss most of the subgroupings in the poll out the window — the sample sizes are way too small.
<
p>
Anyone really want to bet that Reilly will beat Patrick among African-American voters in September?
<
p>
assumptions are a dangerous thing in politics.
<
p>
And for each of yours, there’s an opposite thats equally plausible and as likely to happen. Try this, you’re Tom Reilly, your worst month is behind you and no one is going to remember it in September. Your opponent has a near 100% burn rate and has little to show for it beyond the built in support he started with. No one expected you to win the convention, but then again, the convention means, well … nothing come September or November for that matter.
<
p>
As for tossing the subgroupings, no problem here, I like the big picture of the poll.
Why do wealthier, more educated people favor Patrick?
<
p>
I bring another statistic from that same survey to everyone’s attention. 50% of the people surveyed still don’t know about Deval Patrick. For a comparison, about 45% didn’t know Mihos… so more people know a brand new, independent candidate than Patrick, who has been running for at least a year.
<
p>
Since well educated people are more politically aware anyway, it serves to note that they are more likely to know him. This could be stretching it a bit, but I think this is a positive corellation: the better you know Deval Patrick, the more likely you are to vote for him.
<
p>
This also illustrates why African American and younger voters are less likely to vote for Patrick, although I think the sample size really hurts data with this much detail. If there’s about 400-500 people who participated in this survey, about 50 of them were probably African Americans and there’s likely a similarly small number for 18-34.
<
p>
The Catholic numbers are a real problem for Deval Patrick. He’s going to need to make some inroads there – but I don’t know how he’ll do it. He can’t win the primary if he gets trounced by those kinds of numbers with Catholic voters. Ideology should have little to do with it, Reilly has came out as favoring gay marraige and he’s pro-choice. Aside from the fact that Reilly is Catholic (I think) I’m clueless as to why the difference is so great.
Note that of the 34 African-Americans identified as likely primary voters in this survey, 41% (i.e., 14 individuals) were undecided, so the actual sample size is only 20 individuals. That number is small enough that Patrick’s poor showing among the group could easily be an artifact of the sample rather than indicative of what’s really going on in the African-American community. Nonetheless, there are reasons for Patrick to worry about that part of the electorate.
<
p>
Also, I agree that Patrick’s continuing difficulty in getting non-political junkies to be aware of his existence is a problem, and that the comparison to Mihos is telling, though maybe Mihos has some residual name-recognition from the Turnpike flap.
incumbent or incumbent party in Presidential elections. I suppose Reilly could be considered an incumbent in a sense and hence the young voter preference for Reilly.I realize this is not a presidential election but I think a political scientist would support that the same political theory is in play here.
Interesting that Patrick gets a zero among people who may vote. Should he hope for a September monsoon?
<
p>
Also, geographically, he seems to be weakest in the southern part of the state. Someone who knows better can correct me, but didn’t he do quite well in the caucuses in that part of the state? It suggests something of a disconnect between regular voters and party activists.
<
p>
And I’d love to know who was named in the “other” category, especially among registered Democrats.
Is Patrick’s best region, which I find somewhat surprising, given Reilly’s roots there. The small sample sizes must have distorted some of the numbers (I have a hard time accepting both Patrick’s poor showing among African-American voters and Gabrieli’s 30% among Hispanics).
<
p>
What’s up with the category of “vote unless emergency”? One would think that the results in this category would be similar to the “definitely vote” category, but Deval does much better in the former. I think the sample sizes may be wreaking havoc with some of the answers…
<
p>
Deval’s name recognition is clearly growing, but I do worry about what will happen when Reilly starts pumping millions into TV ads, which will help casual voters determine how to vote. The relatively high Don’t Know numbers in this poll show that many voters are waiting until summer to make a decision, and that decision will be impacted by whose name they are hearing more.
I don’t pretend to speak with authority on the subject of polling samples, but it is my understanding that the magic # of a predictive sample size is around 25. Yes the margin on error will be higher, but it’s my understanding that pollsters view results of a 25 response or great to be a valid indicator.
With a sample size of 25, for any large population, the margin of error is +/-20%. That is practically useless for any analysis in a race where the two candidates are seperated by less than that.