[UPDATE: rough interview transcripts now included, down at the end.]
This saying may not be catchy, but it’s true: “she who lives by the demographer dies by the demographer.” So: meet the demographer, Peter Morrison.
Anaheim Council member Kris Murray now firmly dominates the City Council’s 3 or 4 person majority — in fact, at this point, let’s just call it the “Murrjority.” This past week, in brushing off the ACLU’s and local Latino plaintiffs’s California Voting Rights Act suit against Anaheim, she placed pretty much every one of the Murrjority’s chips on the “expert report” of demographer Dr. Peter Morrison, pictured above. Morrison, until recently of the RAND Corporation, is the demographer who was retained by Anaheim to study its demographics and is presumably intended to be an expert witness in proceedings before Judge Franz Miller. That may include Tuesday’s hearing and potentially a trial beyond that.
In her extensive statements from the dais justifying rejection of the Citizens Advisory Committee to place districts with “voter residency” requirements (or, as they are known by law, just “districts”) on the ballot, the chilly cucumber of a councilwoman continually cited the Demographer’s Report as the decisive reason for her vote. (I’m capitalizing this term because, until this episode in the city’s history is finished, it should pretty much be a proper noun.) Murray expressed some other reasons as well, but they were either trivial or factually flawed; as such, I will get to them only if I have extra time and want to have some fun.
The Murrjority (which is not guided by Murray alone; no doubt others have left their Pringleprints all over the plans) clearly considers Demographer’s Report its best hope to get through the court process without being forced to create actual districts — ones in which a person must reside to qualify to vote in a district election — in order for the city to comply with the California Voting Rights Act. (Anaheim last Tuesday create “candidate residency” districts, which I’ve been calling the “Santa Ana plan”: elected representatives chosen primarily by the rest of the city. These are not to be confused with voter residency districts — which I stress mostly because Murray repeatedly did just that — and tried to convince everyone else to do the same.)
As will surprise no one who knows my writing, this story has turned out to be really long. So, I’ve divided it up. Today (which, being in the middle of a holiday weekend, is a likely low readership day) I’m just going to set down some preliminaries, as well as some interviews I did with Morrison for those who may have been waiting for them. Tomorrow, we’ll go over the contents of the Demographer’s Report. Monday (and maybe Tuesday as well), we’ll review how Kris Murray has intensified the damage by lashing the city’s fate to this report. Without the Demographer’s Report, it’s unlikely that Anaheim’s leaders could work up the brazen will to spit on the California Voting Rights Act. So, when the Demographer’s Report is found lacking, Kris Murray will have made it much easier the actions of the City Council get overturned in court. (Thanks!)
For today and tomorrow, though, let’s just try to understand what the Demographer’s Report says.
1. The Murrjority’s Secret Weapon
The Demographer’s Report was intended to prove why Anaheim doesn’t need to do any better of a job of representing Latinos than it already does. (The shootings and protests and riots? Details!) Tomorrow, we’ll go through it slide-by-slide; I’ll try to be cheeky and entertaining enough here to make you willing to sit through the lecture, even against your better judgment. (No small task, that.)
For now, let’s just note that the Demographer’s Report served an even more important political purpose than merely justifying Anaheim’s inaction. Because the awesome and powerful information contained within it had not been made available to the Citizens Advisory Commission, it gave the Murrjority all of the permission it needed to ignore that Commission’s advice. (To paraphrase Murray: “I’m sure that they’re good people, but they didn’t have the benefit of reading the Demographer’s Report like we did!”)
Unfortunately for the Murrjority, the Demographer’s Report fails gloriously as a means of justifying Anaheim’s position. I don’t ask that you take my word on it alone, Dear Reader. I invite you to accompany both me and Kris Murray, whom you can imagine is reading this article at the same moment as you to try to find out why I have been calling her a moron since Wednesday, on an exploration of Social Science gone wrong. (In this scenario, Kris Murray was violently shaking her head “no!” at the moment she read the first sentence of this paragraph. Patience, Kris. The explanation is coming. Relax. Have a drink. Have five.)
The Demographer’s Report is forceful, but like much social science it is only as good as its assumptions — which in this case are woefully inappropriate. You see, when applying demography to electoral politics, if you want to be sure about what you’re saying it’s best to double-check your assumptions with, oh, someone like a statistically trained social science Ph.D. who taught university-level research methods as a Political Scientist focusing on, among other things, the American electoral process.
2. A Glossary of Social Scientific Terms (note: this section is for reference purposes; unless you are a trained social scientists, do not read the section in color straight through, or you may die)
It so happens that I am a statistically trained social science Ph.D. who taught university-level research methods as a Political Scientist focusing on, among other things, the American electoral process. (Happy to be of service!) I can best serve you for now by explaining some basic concepts before we apply them.
In my first draft of this story — yes, for most of this I am doing multiple drafts, which tells you that this stuff is important! — I included explanations of social scientific concepts in various slides. This, I came to realize, would make them unreadable for many people. So, instead, I’m putting those explanations up here and I’ll just refer to them. If you don’t understand what a term like “operationalization” or “outlier” means, you can refer back up to here. Best of all, I’m going to color code them to make them festive to read and easy to find! You’ll come back to this tomorrow and Monday — maybe every day after that. I’m going to start with just a few terms, but I’ll be adding to this list.
Operationalization: Social scientists like to say (mostly because it’s true) that social science is more difficult than natural science. “Hard” empirical science in general (with some data-driven exceptions such as astronomy and archaeology) largely involves offering a theory, deriving a hypothesis, operationalizing the terms of that hypothesis — meaning figuring out what you’ll actually measure to see if the concept you’re interested in is changing, performing the measurements, and then analyzing the results of the measured variables using statistics. The critical step with which most readers are probably unfamiliar here is “operationalization” — and it’s the rock on which plenty of social science founders.
This is less true (it’s still often true, but less so) in the physical and life sciences. In natural science, if you’re interested in something as commonplace as local surface temperature or as mysterious as the firing rate of single neurons or the decay traces of the tiniest fundamental particles, your theory and hypothesis and measurement and statistics are important, but operationalization is usually relatively straightforward.
For example: you find the brain cells that you’re interested in. You hypothesize that they will fire more or less quickly after the ingestion of a certain drug. You measure changes in the electronic charge with a very fine electrode, and then you analyze it. I skipped over the step of operationalization there, because it’s so straightforward. It was this: we will say that a neuron has fired when we see a tiny little blip in the electrode. Everything else is amazingly complicated, but the identification of “electrical surge” with “cell firing” is straightforward.
In social science, it’s usually not straightforward at all. That’s what makes social science so difficult to do well. And this is a good example of that. Dr. Morrison will ask us to read something of major significance into the fact that John Leos, a Latino, finished third to Jordan Brandman and Lucille Kring in 2012. He’ll ask us to conclude that the number of votes received by Leos is going to be a good operationalization of something like “the power of Latino voters.”
In so doing, he’ll be ignoring at least two major factors:
(1) that Leos, unlike the vast majority of Anaheim Latinos, is a Republican
(2) that even if Latinos all share the same preferences and that they override partisan affiliation, his proportion of the vote will depend among other things on how many other Latinos are splitting up the Latino vote with him.
Now the notion that Latinos will only vote for Latinos is itself a very questionable operationalization of the concept of Latino self-determination. On the merits, many Latinos might have argued that the candidate most sympathetic to their needs was Duane Roberts — an Anglo from the Green Party with almost no campaign money — who finished last. The candidate whom I’m told (without having analyzed it myself) did best in Latino communities was Jordan Brandman — who not only had gobs of money behind him but was also the only candidate endorsed by the Democratic Party.
Looking at 2012 tells you something about an election where no only one “serious” Democrat ran (with the support of both Democrats and the Resort Area machine and cashbox) and in which the only “serious” Latinos running (that’s plural if you count Steven Albert Chavez Lodge as one) were Republican — the opposite of most voters in the city. Can one extrapolate from such results? Not likely. Then toss in the fact that of the nine Council candidates, five (including Lodge) featured Latino surnames, in addition to Roberts having that community as a base and Brandman having that Democratic Party endorsement — meaning that seven of the nine had either an ethnic, partisan, or ideological claim on Latino voters.
Splitting votes seven ways leads to smaller numbers than splitting them two or three ways. In fact, if we add the totals of all of the candidates who didn’t display a Spanish surname on the ballot — Brandman, Kring, and Linder (I’m leaving Roberts out because his Latino base was probably higher than the votes he got for being “non-Latino”) we get 65,252, almost the same as the 64,213 obtained by the quintet of Leos, Rivera, Chavez Lodge, Chuchua, and Gaona.
What does one make of this? I don’t know — and that’s actually an important answer. I think that it’s likely that the 2012 election tells us not much about Latino voting power in Anaheim — and as such it cannot be used as a cudgel to beat away the requirements of the CVRA.
Here’s another problem of operationalization: should we even care about the ethnicity of the candidates who get elected as opposed to how much power the Latino electorate has to get their choices into office? We’re asked here to look at the former outcomes — largely because they are easier to measure — but there’s a strong case to make that Voting Rights law is concerned much more with the latter. After all, African Americans would not, if asked, be likely to vote in favor of the anti-Affirmative Action crusader Clarence Thomas being on the U.S. Supreme Court.
And finally, here’s a little thing, but it may gall you. Lorri Galloway is half Iberian Spanish (as in, from Spain rather than Latin America) and half Filipino. As is sometime true of both ethnicities, she is quite fair — and of course her name on the ballot does not show Latino heritage. And yet, the Demographer’s Report can only show that Latino candidates do well in Anaheim because it counts Galloway as — not 50%, but 100% Latino. This, again, is a matter of operationalization: defining concepts in concrete terms. If we’re going to count Latino candidates, then who counts as a Latino candidate? (I can’t remember who it was, but someone in Tuesday’s meeting briefly tried to make a case for Harry Sidhu a, more or less, “close enough.”) I don’t think that a judge will buy it.
Outlier: an “outlier” is a data point — it could be a person, a place, a time period, an event — that is very unusual. Because usually we can only take a relatively small sample of data — and, if we’re trying to predict the future based on the past, we can’t same future data at all — outliers can make it harder for us to accurately describe a group and even harder to predict what will happen in the future. For example, if you were trying to describe the number of NBA players per city in OC and one of the five places you chose to review was Newport Coast, you’re probably going to get a really skewed estimate of how many NBA players are in OC overall, because Newport Coast is an outlier. In describing or predicting data, one danger of including outliers is that the resulting model will tend to be unstable, such that including or excluding just one datum could lead to widely divergent results. If the results could vary widely based on small choices like this, why should anyone — like a court of law — rely on them?
Selection Bias (example chosen for the Winships): In describing existing data or predicting future data, your choice of which data to use in making the prediction lead to poor results. If you were trying to estimate the likelihood of how many championships the Los Angeles Lakers are going to win over the next five years (2014-2018), you might, for example, look at how well they’ve done in previous five-year periods. Here’s a table:
2009-2013 — 2 championships
2004-2008 — 0 championships
1999-2003 — 3 championships
1994-1998 — 0 championships
1989-1993 — 0 championships
1984-1988 — 3 championships
1979-1983 — 2 championships
1974-1978 — 0 championships
1969-1973 — 1 championship
1964-1968 — 0 championships
1959-1963 — 0 championships
Add up the numbers all the way back to 1959 (when the Lakers moved to LA) and you get 11 championships in 11 5-year periods, or 1.0 championship per 5 year period. Go back only to 2009 and you’d predict 2.0 per 5-year period. Go back to 2004 and you’d predict 1.0; but go back to 1999 and you’d predict 1.7. Go back to 1994 and you get 1.25; go back to 1989 and you’re back to 1.0. Go back to 1984 and your average is 1.3; go to 1979 and you’re back up to 1.4. (Beyond that it starts falling off.) There’s no “right answer” to the prediction based on the data alone; you have to make a case that the Lakers of today are or are not in a similar sort of situation to the sample from which you make the prediction. Now, if you know that someone wants to find a high number and they go back to 5 years or 15 years 25 years or 40 years (all of which are higher than the periods near them), you have to wonder if they just got lucky or they were “cooking the books.”
In a sense, though, it doesn’t even matter whether such an error is intentional: if you’re looking at the Boston Celtics and you randomly choose the 10-year period from 1960-1969 to make the prediction that they’ll win 9 of the next 10 NBA championships, you don’t just say “well, I chose randomly, so I get to stand by the data!” You look at the data and admit that you ran into a huge outlier — and that as a result your prediction of the future is worthless.
Selection bias (more relevant example): If you’re studying elections in a given city to see how prone it is to elect Latinos — and you find that it elected just one fewer Latino in one election that came six elections ago as it has done in all five of the elections since then combined and more than were ever elected in all previous city elections combined — you may want to take a moment to consider whether the inclusion that election is an outlier. If it is an outlier, You may want to consider the stability of your dataset and the conclusions you offer — and, if the dataset is highly unstable as a result of it, you may want to consider whether it would make sense to remove it from your analysis, or even whether it renders your entire analysis less than useful. Remember, instead of going back six elections you could just as easily have gone back only five. If your having done so would have led to your coming up with completely different conclusions — something, by the way, that you should already know in case some blogger comes up to you after your presentation and asks you about it — then you may want to hesitate offering those unstable conclusions as a basis for making important policy decisions, let along for trying to convince a court that no, there’s really no violation of the CVRA.
That’s enough for today, except for the two videos that follow of my post-presentation interview with Dr. Morrison. You may already have a sense of the problems with the Demographer’s Report, but I promise you one thing — they will get worse.
Here’s Dr. Morrison on questions such as: “Why study the 2002-2012 period?”, “Who made that decision?”, “Why didn’t you worry about 2002 being an outlier”, and “Why operationalize as Latino voting power as ‘election of Latino candidates’ rather than ‘Latinos being able to get their favored candidates onto Council’?”
GD: Whose decision was for you to take the analysis from the period of 2002 to 2012?
PM: Thank was my choice, basically, it was the last ten years. I just picked it arbitrarily. And it was the period of time when there were interesting [?] candidacies by Latinos.
GD: OK. And no one talked to you from the Council or anyone related to it about that choice.
PM: No, no, that was my choice.
GD: Did you have any concern that 2002 was an outlier?
PM: I had heard that it was the wake of 9/11 and that some regard it as an outlier.
GD: No, an outlier in terms of the results of the elections. Two Latinos are elected.
PM: Well, I don’t know what you mean by an “outlier.” I mean, it’s a situation where two got elected, and somebody said it was because they were firemen because of 9/11.
GD: I guess, in terms of the stability of the data – would the data have been substantially different in terms of the result if you had gone only back to 2004, for example?
PM: One could take selected years and come up with, you know, different results, but I took the continuous ten-year period.
GD: I understand that you looked at the last six elections – I’m saying, would the data be substantially different if you took the last five elections?
PM: I don’t think they would be – well, it would depend on which measure you were looking at. I don’t think it would have changed the basic conclusion that Latinos seem to have a good opportunity to place second [unintelligible.]
GD: That brings me to my last question. In making a decision of how to measure the representation of the district, was there a reason that you looked at simply electing candidates – I guess by that standard South Carolina is the least racist state right now, because they’re the only one who have a Black Senator, right? – as opposed to effectuating the will of the population?
PM: The problem here is that you don’t really know whether the Hispanic candidate was in fact the preferred candidate of Hispanic voters. That’s a mystery that is only resolved with deep statistical analysis. But if you simply ask the question “are Latino candidates getting elected?” presumably by –
GD: So that’s the only question that you can answer?
PM: That’s the only one that is on the surface of the data.
GD: OK. Thank you very much!
PM: You’re welcome.
And here’s Dr. Morrison answering my questions about whether “if more seats were open, wouldn’t a faction that had previously run only two people for office (so as not to split the vote) now run three, so that the fact that a person finished third gives us little information as to whether they’d be elected if more seats were open?” — which I’ll address tomorrow.
GD: You’ve said that if a third person were elected, then John Leos would have won –
PM: Right –
GD: And that if a fourth person were elected, then Jennifer Rivera would have won –
PM: Right –
GD: How do you rule out the prospect that if there was one party that was pushing the people in there for two seats, that they would only run two candidates – there’s no reason to run a third candidate because it would split the vote – and that if there were three positions they would have run three candidates? And if there were four then they would have run four?
PM: That’s why I say that it’s a mental experiment, because I have to assume that nothing else changes. Of course, if the structure of the electoral system changes, that creates a whole set of new incentives and maybe then different people will respond in different ways. That’s why I say that all you have to go on is the history, and if you look at the history you say that this is all we can [account for?] –
GD: So there was a conclusion from the dais for example that, if we had done it this way with six seats, then John Leos would have been elected, but in fact we don’t know that to be true!
PM: That would be true if nothing else changed.
GD: But if a third candidate had run that was similar to Kring and to Brandman, that it may well be that they would have won!
PM: Yeah, but that’s purely hypothetical. You can imagine anything – [recording cuts off]