On Polling and Primaries

My friend Luther wrote a post “On Michigan” at InfiniteFreeTime last week. I agree with all of it, and it set me thinking. Luther and I have compared notes on this election a bit.

I’ve learned some things from him, and he’s proven himself smarter than me about several of the electoral dynamics. If I’ve been making sense to you with these political posts, you should be reading Infinite Free Time. Luther dosen’t always write about politics but when he does, he makes sense, and he’s far more entertaining than me. I’m gonna try and elaborate today on some of what he said about Michigan.

MichiganFlag

On the Polling

(If you aren’t interested in polling, just scroll down to the next subhead. I’ll tell you how I think this Democratic primary season ends.)

Bernie Sanders’ win in Michigan was an upset. No one, including the candidates, saw it coming. Here’s a post and a discussion at fivethirtyeight.com that might shed some light. Tl;dr version of Carl Bialik’s explanation, which is the first comment on the thread: Polling organizations stopped contacting Michigan voters on Sunday, and that’s too early to stop polling in an election this fluid, so the polls missed too many late deciders.

Michigan could be a fluke. Or it could mean Sanders will be competitive in Ohio, Missouri, and (maybe) Illinois tomorrow. The reason I put IL in the “maybe” column is because Clinton has deep connections and a strong organization in that state. She was as much as 37 points ahead of Sanders there in some polls last time I looked, but given that the polling has been inaccurate in two Midwestern states now, I’m willing to consider there might be something off with the way these surveys are being conducted.

All this caused me to do some digging, because I want to know how trustworthy the polling is going forward. After doing my reading, I’ve come to the conclusion that the predictive survey data in this election might be less reliable than it’s been in 30 or 40 years. Here’s Rutgers professor Cliff Zukin explaining a few of the problems which might apply in the New York Times. I’ve also seen articles from Nate Silver at 538 and a former provost of Georgetown University which concur. I’ll break it down for you.

flagvote

  1. For a survey to have any value for predicting the behavior of voters, you need a fairly large sample, and you need it to be representative of the general population.
  2. Historically, the way this has been done is by taking huge lists of land line telephone numbers and building random samples for telephone polls. The participation rate is so low for these surveys, it’s necessary to make tens of thousands of phone calls to find even 1,000 qualified participants. Sometimes numbers have been disconnected, sometimes people just hang up, and sometimes kids answer the phone.
  3. We’ve reached the point where so many people have dumped their land lines in favor of cellphones, it’s no longer possible to use probability sampling to build a representative sample without adding more data, so the way to do it these days is to also build a sample of cellphone numbers, survey those separately, and then aggregate the landline and cellphone data.
  4. The FCC restricts the use of autodialers for cellphone surveys, so interviewers have to actually dial the numbers themselves and in some instances, polling organizations need to compensate respondents for their wireless usage. Which doesn’t seem like a big deal until you think of it from the perspective of the paid-hourly interviewer with a list of 2500 numbers they have to dial manually, or the cost of paying $5K for every individual cellphone sample of 500 likely voters for an entire election cycle.
  5. Since only the biggest and best-funded organizations can afford big, reliable phone samples that actually represent the population at this point, and the internet is so well-developed, a lot of small organizations and newcomers are using internet polls. But the demographics of internet use are skewed. So skewed, in fact, the professionals have yet to figure out how to build a representative sample using only internet data.

The result: Lots more noise in the polls than we’re accustomed to and less reliable predictions. I’ve read so much about this in the last week, I just don’t trust the polls in this election any more. So I’ll tell you where I think this Democratic primary season is going based solely on my understanding of the demographics of various states and the trends so far.

On The Primary

Clinton has a 200-delgate lead among pledged delegates, which are won based on primary outcomes. She has overwhelming support from the superdelegates who have publicly weighed in. Superdelegates can change their minds, but they’ll only do that if Sanders closes the gap and convinces them he’s the choice of the people.

clinton-sanders

Even then, he won’t get all, or even most of them. Superdelegates are hardcore party loyalists — mostly former long-serving elected officials who REALLY understand how these politics work — and they have questions about Bernie’s viability on account of he’s an unapologetic socialist (not a bad thing, in my mind, but a huge vulnerability for a general election contender) and is also untested in a national election (kind of a big deal to me, given the stakes this time around).

There is no such thing as a winner-take-all Democratic primary in the U.S. Pledged delegates are apportioned based on popular vote percentages. If you want to win all or most of the delegates in a state, you need a landslide victory. A 51/49 win results in an almost even delegate split.

This is one of the few things we’re doing right with our elections. If we weren’t staring down the barrel of a long series of winner-take-all Republican primaries in which GOP voters have to choose between a racist con man and a certifiably insane Tea Party guy, Trump wouldn’t be as strong a candidate as he is.

The Democratic Party’s primary system would be perfect if they’d dump the superdelegate idea and go to an all-primary system. (Caucuses make the process too opaque and introduce an unacceptable level of moral hazard into the nomination process, IMHO.)

So here’s what I see happening next. Tomorrow, five big, significant states will vote: Ohio, Florida, Illinois, North Carolina, and Missouri.

These states have a lot of delegates. Bernie’s performance in Michigan makes me think he might compete in the Midwest states and at least win half the delegates in Ohio. But I expect North Carolina to go big for Clinton and I expect her to at least break even in Illinois. Missouri is an unknown quantity to me, especially given the problems we’ve seen with polling in the Midwest.

I was loathe to even talk about Florida when I started this. Having lived close to Florida and been there many, many times in my life, I’ll tell you this about it. Once you get below the panhandle, it’s different from either the Deep South or the Mid-Atlantic. Central and South Florida are a region unto themselves. I think Clinton was the favorite in Florida before last week’s debate even started, and I think Bernie’s answers on the socialism questions, as much as I appreciated his effort, probably sunk him there.

floridaflag

After North Carolina and Florida vote tomorrow, the South is done until November. Clinton’s advantage among black southerners has accomplished all it can accomplish, but she’s blown Bernie out down here because he did not compete. I understand that decision — I called the southern primaries for Clinton, and decisively, months ago. But now he has to not only beat her, but beat her by wide margins in big states, to have a chance of catching up.

We’re running out of pledged delegates, and the best he’s done so far in big, diverse states is beat her by a percentage point in Michigan. Even if Michigan means something and he competes in the other states, his margin of victory there makes me think he’s just about finished.

That said, he’s done our country an important service. He’s forced Hillary to the left on a few issues, he’s energized a lot of young people, and he’s taken the first baby steps toward making it cool to be a real lefty again. So god bless him. I hope the Democratic Party has the political sense to keep him campaigning and offer him an executive branch job once Hillary clinches the nomination.

I had some things to add here about protest voting and why we shouldn’t do it this time around, but I can’t believe anyone wants to read 3,000 words from me all at one go, so I’ll save that part for another day.

Advertisements

Talk to Me!

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s