Polls of New York Republicans (not an oxymoron) say that Donald Trump should be getting slightly over 50% of the vote in the April 19 primary there. That wouldn’t be surprising; after all, it’s his home state. But let’s take a closer look at those polls.
A recent one (4/3-4/5) was done by Monmouth. Since New York is a closed primary state, it sampled registered Republicans only. However, it has what I would consider a small sample for a poll: 302. What’s more, although it’s weighted for age and gender it’s random in terms of geography, and I very much doubt that New York’s Republicans are evenly distributed throughout the state. The margin of error is 5.6, which is somewhat high but certainly not astronomical, but for subgroups it’s higher (for example, for males it’s 8.1 and for “very conservative” voters it’s 11). That has to do with the fact that once you get into subgroups the already-small n becomes much smaller.
For example (and I find this interesting in terms of the makeup of the Republicans of New York state), of the 302 subjects, only 80 described themselves as “very conservative,” 98 as “somewhat conservative,” and a whopping 117 as liberal or moderate. Liberal Republicans? There’s a long tradition of that sort in New York, actually; think of the term “Rockefeller Republicans.”
Then we have the curious fact that 9% of the registered Republicans in that poll say they will vote for Hillary Clinton against Trump, and if Cruz were the nominee 10% would vote for Hillary Clinton. Again, who are these people? Kasich does best of all with them against Hillary; I guess those liberals and moderates among them figure he meets their needs as well or better than Hillary does.
In addition, the repondents’ votes are not set in stone. Although 2% have already voted, and 38% have decided for sure, 34% have only a strong preference, 12% a slight preference and a full 14% are undecided. That’s quite a bit of wiggle room; if you add those last two categories you get 26%.
Another recent poll is this one, taken by CBS News just a few days earlier (3/29-4/1), with somewhat similar results. In it, 20% of Republican respondents (the number of Republicans was 545, significantly more than in the other poll) seemed willing to change their minds.
Another interesting question is how often these respondents have voted for the GOP candidates in the past. 34% said “almost always” and 40% said “usually,” which leaves 19% split about evenly and the remaining 7% usually or always have voted for Democrats. And only 81% of respondents in the poll said they probably or definitely will vote in the primaries.
In that CBS poll, however, the subjects were registered voters and there seems to have been no attempt to make sure that the respondents who answered about the Republican primaries actually are registered Republicans who can vote in the primary (New York’s is closed)—or it there was such an effort, the poll results don’t mention it. They also don’t mention whether the sample was weighted and if so in what way. In general, although there are a lot more questions in that particular poll, the pollsters offer a lot less information about its methodology.
One of the most recent NY polls of all was by Emerson (4/6-4/7). Unfortunately, I can’t access the full results (they say it can be found here, but when I try to download the file I get something that looks like impenetrable gibberish) so it’s hard to tell as much about it. But it involves “321 likely [GOP] primary voters, with a margin of error of +/5.4%,” which again seems like a very small n. And I have no idea how they define “likely voters” (are they, for example, registered Republicans? Because if they’re not they can’t vote in the GOP primary in NY). They were weighted for gender, but not for anything else, which makes me wonder once again how the geographical distribution goes, which is important because the majority of NY’s delegates are awarded by winning districts. The rules on that last point are so complex that it’s hard to summarize them (the Emerson link doesn’t do a very good job, either), so if you’re really interested I suggest you go here, but don’t say I didn’t warn you.
Trump’s Emerson score is good (56 to Cruz’s 22 and Kasich’s 17). But interestingly enough, it represents a drop from his score from just three weeks earlier in a poll by the same group. In an Emerson poll taken March 14-16, Trump had received a hefty 61% of the vote, with Kasich only at 1 and Cruz 12. You can see what’s happening here, and the difference does not reflect the previous presence of Rubio, who had only polled at 4% in that mid-March poll before dropping out.
So the real question—aside from whether any of these polls are correct, which is always the question with polls—is whether the current trend of Trump falling and Cruz/Kasich rising will continue in the next week or so, and if so how far it will go. The margin of error in that recent Emerson poll is already 5+ percent, which means that Trump is within the margin of error for achieveing the 50% level he would need in New York to trigger certain winner-take-all provisions for the state delegates (not the district delegates). Again, it’s so complex I suggest you go to that website if you’re interested in the details.
I am nearly positive that Trump will win the NY GOP primary; all indicates suggest that very strongly. But by how much, and how many delegates will it mean he actually receives?
Plus, the delegates are chosen by the state GOP—even the Trump ones—which could create another problem for Trump, because they may not be particularly loyal to him. That, of course, would probably anger his supporters very much, and understandably so. But as many people have pointed out, if Trump the smart dealmaker didn’t know this and prepare for it, maybe he’s not such a smart dealmaker after all.
By the way, that most recent Emerson poll indicates that in a general election, all of the GOP candidates would be toast and the Democrat would win the state.
One other curiosity I noticed in the same poll:
Among New York’s 27 Congressional districts, each worth 3 delegates, [Trump] appears to dominate in five (the First, Second, Third, Fourth and Eleventh) and stands a good chance of passing the crucial 50% threshold in 11 other districts which would give him a winner-take-all victory [that is, in terms of the three delegates in each of the districts where he might get 50% of the vote]. In contrast, Cruz is competitive in only seven Congressional districts.
But recall that there were only 321 likely GOP voters in that entire poll. You’ll note that there are 27 districts in NY. That means that there was an average of about 12 poll respondents per district. How could any valid conclusions possibly be drawn on a such a basis?
A poll with almost double the number of respondents was the most recent one by Fox, taken March 4-7, and featuring 602 likely Republican primary voters. That’s almost double. Trump’s score was 54, Kasich’s 22, and Cruz’s 15, so this one was the lowest of them all for Cruz. 23% of respondents say they could change their minds, however.
Here’s the description of how the respondents were chosen:
…[They] were randomly selected from a statewide voter file of registered New York voters using a probability proportionate to size method. This simply means phone numbers are proportionally representative to the number of voters in all regions across the state of New York.
I don’t know quite what that means. We know that only registered Republicans are allowed to vote in the GOP primary in NY, so were only registered Republicans asked about the Republican primary? It’s not at all clear from the way that is worded. As for the geographic proportionality of the sample, districts in NY vary greatly in their composition—some districts are overwhelmingly Democratic and some are majority Republican, and New York’s delegates are selected on a district-by-district basis, so in some districts each single voter has far more selective power in a primary than in other districts with more Republicans. Was that reflected in the geographic distribution of poll respondents, or are the samples merely assumed to be proportionate because the list is of all registered voters in a region and they were randomly selected within each region? It’s very confusing, I must say, but it’s important to know the answer in order to understand how valid the poll is.
As I already indicated, my guess is that Trump’s support in New York is very very strong. But since the majority of New York’s delegates are not selected on a statewide basis (only 14 out of 95 are), it’s hard to know what to make of any of these polls, because they don’t seem to take the importance of districts into account.
You may ask why on earth I’m paying so much attention to them, then. Good question. Well, they’re the only evidence we have. Plus, when I read headlines about the results of polls, it usually piques my curiosity. How was the poll done? Is it really valid? It’s a weird little peccadillo I have, but I happen to be interested in polls.