“2010: What happened?” in light of 2018

Back in November 2010 I wrote an article that I still like, attempting to answer the question: “How could the voters have swung so much in two years? And, why didn’t Obama give Americans a better sense of his long-term economic plan in 2009, back when he still had a political mandate?”

My focus was on the economic slump at the time: how it happened, what were the Obama team’s strategies for boosting the economy, and in particular why they Democrats didn’t do more to prime the pump in 2009-2010, when they controlled the presidency and both houses of congress and had every motivation to get the economy moving again.

As I wrote elsewhere, I suspect that, back when Obama was elected in 2008 in the midst of an economic crisis, lots of people thought it was 1932 all over again, but it was really 1930:

Obama’s decisive victory echoed Roosevelt’s in 1932. But history doesn’t really repeat itself. . . With his latest plan of a spending freeze, Obama is being labeled by many liberals as the second coming of Herbert Hoover—another well-meaning technocrat who can’t put together a political coalition to do anything to stop the slide. Conservatives, too, may have switched from thinking of Obama as a scary realigning Roosevelt to viewing him as a Hoover from their own perspective—as a well-meaning fellow who took a stock market crash and made it worse through a series of ill-timed government interventions.

My take on all this in 2010 was that, when they came into office, the Obama team was expecting a recovery in any case (as in this notorious graph) and, if anything, were concerned about reheating the economy too quickly.

My perspective on this is a mix of liberal and conservative perspectives: liberal, or Keynesian, in that I’m accepting the idea that government spending can stimulate the economy and do useful things; conservative in that I’m accepting the idea that there’s some underlying business cycle or reality that governments will find it difficult to avoid. “I was astonished to see the recession in Baghdad, for I had an appointment with him tonight in Samarra.”

I have no deep understanding of macroeconomics, though, so you can think of my musings here as representing a political perspective on economic policy—a perspective that is relevant, given that I’m talking about the actions of politicians.

In any case, a big story of the 2010 election was a feeling that Obama and the Democrats were floundering on the economy, which added some force to the expected “party balancing” in which the out-party gains in congress in the off-year election.

That was then, this is now

Now on to 2018, where the big story is, and has been, the expected swing toward the Democrats (party balancing plus the unpopularity of the president), but where the second biggest story is that, yes, Trump and his party are unpopular, but not as unpopular as he was a couple months ago. And a big part of that story is the booming economy, and a big part of that story is the large and increasing budget deficit, which defies Keynesian and traditional conservative prescriptions (you’re supposed to run a surplus, not a deficit, in boom times).

From that perspective, I wonder if the Republicans’ current pro-cyclical fiscal policy, so different from traditional conservative recommendations, is consistent with a larger pattern in the last two years in which the Republican leadership feels that it’s living on borrowed time. The Democrats received more votes in the last presidential election and are expected to outpoll the Republicans in the upcoming congressional elections too, so they may well feel more pressure to get better economic performance now, both to keep themselves in power by keeping the balls in the air as long as possible, and because if they’re gonna lose power, they want to grab what they can when they can still do it.

In contrast the Democratic leadership in 2008 expected to be in charge for a long time, so (a) they were in no hurry to implement policies that they could do at their leisure, and (b) they just didn’t want to screw things up and lose their permanent majority.

Different perspectives and expectations lead to different strategies.

The post “2010: What happened?” in light of 2018 appeared first on Statistical Modeling, Causal Inference, and Social Science.

MRP (or RPP) with non-census variables

It seems to be Mister P week here on the blog . . .

A question came in, someone was doing MRP on a political survey and wanted to adjust for political ideology, which is a variable that they can’t get poststratification data for.

Here’s what I recommended:

If a survey selects on a non-census variable such as political ideology, or if you simply wish to adjust for it because of potential nonresponse bias, my recommendation is to do MRP on all these variables.

It goes like this: suppose y is your outcome of interest, X are the census variables, and z is the additional variable, in this example it is ideology. The idea is to do MRP by fitting a multilevel regression model on y given (X, z), then poststratify based on the distribution of (X, z) in the population. The challenge is that you don’t have (X, z) in the population; you only have X. So what you do to create the poststratification distribution of (X, z) is: first, take the poststratification distribution of X (known from the census); second, estimate the population distribution of z given X (most simply by fitting a multilevel regression of z given X from your survey data, but you can also use auxiliary information if available).

Yu-Sung and I did this a few years ago in our analysis of public opinion for school vouchers, where one of our key poststratification variables was religion, which we really needed to include for our analysis but which is not on the census. To poststratify, we first modeled religion given demographics—we had several religious categories, and I think we fit a series of logistic regressions. We used these estimated conditional distributions to fill out the poststrat table and then went from there. We never wrote this up as a general method, though.

The post MRP (or RPP) with non-census variables appeared first on Statistical Modeling, Causal Inference, and Social Science.

Debate about genetics and school performance

Jag Bhalla points us to this article, “Differences in exam performance between pupils attending selective and non-selective schools mirror the genetic differences between them,” by Emily Smith-Woolley, Jean-Baptiste Pingault, Saskia Selzam, Kaili Rimfeld, Eva Krapohl, Sophie von Stumm, Kathryn Asbury, Philip Dale, Toby Young, Rebecca Allen, Yulia Kovas, and Robert Plomin, along with this response by Eric Turkheimer.

Smith-Wooley et al. find an association of test scores with genetic variables that are also associated with socioeconomic status, and conclude that “genetic and exam differences between school types are primarily due to the heritable characteristics involved in pupil admission.” From the other direction, Turkheimer says, “if the authors think their data support the hypothesis that socioeconomic educational differences are simply the result of pre-existing genetic differences among the students assigned to different schools, that is their right. But . . . the data they report here do nothing to actually make the case in one direction or the other.”

It’s hard for me to evaluate this debate given my lack of background in genetics (Bhalla shares some thoughts here, but I can’t really evaluate these either), but I thought I’d share it with you.

The post Debate about genetics and school performance appeared first on Statistical Modeling, Causal Inference, and Social Science.

Can we do better than using averaged measurements?

Angus Reynolds writes:

Recently a PhD student at my University came to me for some feedback on a paper he is writing about the state of research methods in the Fear Extinction field. Basically you give someone an electric shock repeatedly while they stare at neutral stimuli and then you see what happens when you start showing them the stimuli and don’t shock them anymore. Power will always be a concern here because of the ethical problems.

Most of his paper is commenting on the complete lack of constancy between and within labs in how they analyse data. Plenty of Garden of forking paths, concerns about type 1, type 2 and S and M errors.

One thing I’ve been pushing him is to talk about more is improved measurement.

Currently fear is measured in part by taking skin conductance measurements continuously and then summarising an 8 second or so window between trials into averages, which are then split into blocks and ANOVA’d.

I’ve commented that they must be losing information if they are summarising a continuous (and potentially noisy) measurement over time to 1 value. It seems to me that the variability within that 8 second window would be very important as well. So why not just model the continuous data?

Given that the field could be at least two steps away from where it needs to be (immature data, immature methods), I’ve suggested that he just start by making graphs of the complete data that he would like to be able to model one day and not to really bother with p-value style analyses.

In terms of developing the skills necessary to move forward: would you even bother trying to create models of the fear extinction process using the simplified, averaged data that most researchers use or would it be better to get people accustomed to seeing the continuous data first and then developing more complex models for that later?

My reply:

I actually don’t think it’s so horrible to average the data in this way. Yes, it should be better to model the data directly, and, yes, there has to be some information being lost by the averaging, but empirical variation is itself very variable, so it’s not like you can expect to see lots of additional information by comparing groups based on their observed variances.

I agree 100% with your suggestion of graphing the complete data. Regarding measurement, I think the key is for it to be connected to theory where possible. Also from the above description it sounds like the research is using within-person comparisons, which I generally recommend.

The post Can we do better than using averaged measurements? appeared first on Statistical Modeling, Causal Inference, and Social Science.

The Axios Turing test and the heat death of the journalistic universe

I was wasting some time on the internet and came across some Palko bait from the website Axios: “Elon Musk says Boring Company’s first tunnel to open in December,” with an awesome quote from this linked post:

Tesla CEO Elon Musk has unveiled a video of his Boring Company’s underground tunnel that will soon offer Los Angeles commuters an alternative mode of transportation in an effort to escape the notoriously clogged highways of the city.

The only thing that could top this would be a reference to Netflix. . . . Hey, wait a minute! Let’s google *axios netflix*. Here’s what we find:

The bottom line: Most view Netflix’s massive spending as reckless, but Ball argues it isn’t if you consider how it’s driving an unprecedented growth that could eventually allow Netflix to surpass Facebook in engagement and Pay-TV in penetration. At that point, they will have the leverage to increase prices, bringing them closer to profitability and making the massive spend worthwhile.


See, for example, here:

At this point, I’d imagine most of our regular readers are getting fairly burned out on the Hyperloop. I more than sympathize. In terms of technology, infrastructure, transportation, and public policy, we’ve pretty much exhausted the topic.

The one area, however, where the Hyperloop remain somewhat relevant, is as an example of certain dangerous journalistic trends, such as the tendency to ignore red flags, warnings that there’s something wrong with the story as being presented be it lies or faulty data were simply a fundamental misunderstanding. . . . One of the major causes of this dysfunction is the willingness to downplay or even completely ignore massive warning signs. . . .

and here:

What does Netflix really want? To make sense of the company’s approach toward original content, it is useful to think in terms of long-term IP value vs the hype-genic, those programs that lend themselves to promotion by being awards friendly or newsworthy. . . . If Netflix really is playing the wildly ambitious, extremely long term game that forms the basis for the company’s standard narrative and justifies incredible amounts of money investors are pouring in, then this distribution makes no sense whatsoever. If, on the other hand, the company is simply trying to keep the stock pumped up until they can find a soft landing spot, it makes all the sense in the world.


Of course, Palko could be wrong in his Musk and Netflix skepticism. Palko’s been banging this drum for awhile—he’s been skeptical about these guys since way before it was fashionable—but that doesn’t make him right. Who knows? I’m no expert on tunneling or TV.

What’s really relevant for our discussion here is that an uncredentialed blogger working on zero budget can outperform a 30 million dollar media juggernaut. Even if Palko’s getting it wrong, he’s putting some thought into it, giving us a better read and more interesting content than Axios’s warmed-over press releases.

On one hand, this is an inspiring case of the little guy offering better content than the well-funded corporation. From the other direction, it’s a sad story that Axios has this formula for getting attention by running press releases, while I suspect that most of the people who read Palko only do so because they happen to be bored one day and decide to click though our blogroll—and then are so bored that they click on the site labeled Observational Epidemiology (perhaps the only blog title that’s more dry than Statistical Modeling, Causal Inference, and Social Science).

An Axios Turing test

But there’s also this which I noticed from the wikipedia article on Axios:

The company earned more than $10 million in revenue in its first seven months, primarily with native advertising that appears in between stories.

“Native advertising” . . . what’s that? I’ll look that up:

Native advertising is a type of advertising, mostly online, that matches the form and function of the platform upon which it appears. In many cases, it manifests as either an article or video, produced by an advertiser with the specific intent to promote a product, while matching the form and style which would otherwise be seen in the work of the platform’s editorial staff.

Hey! This would explain the Musk and Netflix stories. Why is Axios running press releases that could damage its reputation as a news source? Because it’s “native advertising.” Sure, it might hurt their journalistic reputation, but by the same token it could help their reputation as an ad site, as it demonstrates their willingness to run press releases as if they’re actual news stories.

By running these press releases, Axios is either demonstrating its credulity or demonstrating its willingness to print what its sponsors want to be printed. Either of these traits could be considered a plus for an ad site.

And this brings us to the Native Content Turing Test. Once your news feed contains native advertising, you have the challenge of figuring out which stories are real and which are sponsored. The better the native advertising is, the harder it will be to tell the difference, leading to a Platonic ideal in which all the ads read like news stories and all the news stories read like press releases. A sort of journalistic equivalent of the heat death of the universe, in which every story contains exactly zero bits of new information.

P.S. The last paragraph above was a joke. Apparently native advertising must be labeled in some way. According to wikipedia:

The most common practices of these are recognizable by understated labels, such as “Advertisement”, “Ad”, “Promoted”, “Sponsored”, “Featured Partner”, or “Suggested Post” in subtitles, corners, or the bottoms of ads.

I found no such identifiers in the two Axios articles above, which suggest they are not advertising but rather are just business-as-usual news articles from an organization with $30 million to burn but no time to do much more than run press releases on these particular topics.

But, again, they win: they get the clicks. And I suppose that the bland ad-like content makes the site safe for native advertising: In an environment of touched-up press releases, an actual ad doesn’t stand out in any awkward way.

The post The Axios Turing test and the heat death of the journalistic universe appeared first on Statistical Modeling, Causal Inference, and Social Science.


Taking Amtrak, going through Baltimore, reminded that Baltimore used to be the big city and Washington was the small town. My mother worked for the government in the 1970s-80s and had a friend, Pat Smith, who’d moved to DC during WW2 when they were filling up all the government agencies. Pat told my mom that, way back then, they’d go up to Baltimore to shop, because in Baltimore the stores had the newest dresses. I guess Woodies didn’t cut it.

Thurgood Marshall was from Baltimore, as I was reminded as we went by the stop for the Thurgood Marshall Airport (formerly BWI, formerly Friendship). Brett Kavanaugh was from the DC suburbs. Times have changed, what it takes to be a big city, what it takes to be a federal judge.

The post Baltimore-Washington appeared first on Statistical Modeling, Causal Inference, and Social Science.

Vote suppression in corrupt NY State

I’ll be out of town on election day so I thought I should get an absentee ballot. It’s not so easy:

OK, so I download the form and print it out. Now I have to mail it to the county board of elections. Where exactly is that? I find the little link . . . here it is:

Jeez. They don’t make it easy. Not quite vote suppression (yes, the above title was an exaggeration), but they’re not exactly encouraging us to vote, either.

To put it another way, here’s another webpage, also run by New York State:

It appears that the people who run our state government care more—a lot more—about getting its residents to blow their savings on mindless gambling, than on expressing their views at the ballot box.

It’s almost as if the people in power like the system that got them in power, and don’t want a bunch of pesky voters out there rocking the boat.

The post Vote suppression in corrupt NY State appeared first on Statistical Modeling, Causal Inference, and Social Science.