Statistics

Update: That Silly Asylum Applications & Temperature Fluctuations Paper

Update at bottom. I was asked by Axel Bojanowski, Science Editor at Der Spiegel to examine the Science paper “Asylum applications respond to temperature fluctuations” by Anouch Missirian and Wolfram Schlenker.

Here is what I said (regular readers will recall that ever since the Willie Soon affair, I fully and publicly document all my replies to reporters, who cannot in any way be trusted):

It is the dumbest, most idiotic use of statistics I have seen in over a decade. (And I have seen a lot.)

You have to possess a near-miraculous view of statistics, and a complete ignorance of politics, to have suggested or believe in this model. How come asylum applications aren’t swelling in chilly Chili [sic!]? They had, say, 6,000 years of history to draw upon yet they chose only the last fifteen? And they believe the hideously complex political relationship between Europe and (largely) the Levant can be described by quantifying each country’s “corruption index”? I despair.

It’s difficult, really, to blame the authors, God bless them, for it appears they were responding in the way they were taught. Statistical models cannot discern cause, though many believe that they can. I don’t have the space here to explain this, but I wrote a book on the subject, linked here:

I will write an article responding in detail to this paper, and post it to (at least) my website:

Books

Best,

William M. Briggs

Bojanowski later asked about my dumb typo and pressed me for further details. I replied:

Yes, an idiotic typo on my part (typos will be the death of me).

The whole study is so silly. Do we really suppose it’s because of larger *increases* in temperature that Kuwait had more applications than Peru?

Physically the study makes no sense, either. A handful of minor temperature changes is not what is driving asylum applications, as even a cursory reading of the daily headlines shows. To claim one has found a “non-linear” *causative* relationship in so short of record is absurd. These fellows ought to crack open a history book or two (written before this century).

I don’t have time for more in-depth commentary before the embargo is lifted. But see my site.

Briggs

My interest in dissecting these (increasingly ridiculous) studies wanes. They are endless. And they will remain that way as long as governments, with their limitless ability to print money, fund them. Meanwhile, I’m just one guy, with no support. Every time I critique one of these sillinesses (you heard me: sillinesses) it causes me nothing but grief.

Nevertheless…

Headline: French Mayors Panic As Migrants Overwhelm Cities, Beg Macron For Help.

Mayors from seven major French cities overwhelmed by the flow of migrants, have written a joint letter to Paris published in LeMonde on Saturday, begging the government to step in and help.

According to the letter, the cities of Lille, Bordeaux, Strasbourg, Grenoble, Rennes, Toulousa and Nantes are taking in “several thousand” refugees per month, which the mayors say is causing a social emergency as they are “backed up against a wall” and “completely saturated” by a seemingly endless flood of asylum seekers.

The article mentions Afghanistan and Syria.

Headline: Commission: EU Too White, Mass Third World Migration Must Be ‘New Norm’

Brussels has said that Europeans must accept mass migration from the third world as the “new norm”, warning that neither walls nor policies will allow any part of the EU to remain “homogenous and migration-free”.

“It’s time to face the truth. We cannot and will never be able to stop migration,” writes EU Commissioner for Migration Dimitris Avramopoulos, in a piece for POLITICO, published Monday entitled, “Europe’s Migrants Are Here to Stay”.

Now quoting the “science” in the paper:

Here we examine how, in the recent past (2000–2014), weather variations in 103 source countries translated into asylum applications to the European Union, which averaged 351,000 per year in our sample. We find that temperatures that deviated from the moderate optimum (~20oC) increased asylum applications in a nonlinear fashion, which implies an accelerated increase under continued future warming. Holding everything else constant…

So in 15 years of data, they hope to discover a non-linear response in asylum applications caused by tenth-of-a-degree changes in temperature, where they can “hold everything else”, like the politics, “constant.”

My favorite three lines from the paper is this wild insinuation:

Previous studies had found a relationship between weather variations and migration (4, 5, 6), but ours is the first to focus on distress-driven migration (as measured by asylum applications) on a global scale. Two centuries ago, the “year without a summer” (1816), following the volcanic eruption of Mount Tambora in Indonesia, saw massive crop failures throughout the Northern Hemisphere, caused by the aerosol-obscured atmosphere and unseasonal climate. It triggered sizeable migrations as peasants
deserted their fruitless farms.

Climate change caused major grief before, therefore minor fluctuations in temperature, which by hint are given to be unique to our times, cause migration.

The authors don’t say economics doesn’t play. “Migration’s response to income or
wealth corresponds in an inverted U shape”. No word on what kind of shape it is that corresponds to the EU’s progressive policies. Call it the Merkel Rocket Slide.

There appears in this paper to be no recognition that politics inside the EU plays any role. Instead, it’s to the regressions and wee p-values!

Our baseline regression links annual asylum applications from each source country outside the OECD to any EU member state. We use a panel analysis with source-country and year fixed effects, which is equivalent to a joint demeaning of all variables and accounting for common annual shocks. In other words, we link anomalies in log applications to weather anomalies once common annual shocks are absorbed (e.g., the global financial crisis in 2008).

Regression, log applications, weather anomalies, common annual shocks. Whoever taught economists how to use statistical software has a lot of explaining to do. (For why statistics can’t discover cause, and all the other failings of statistical methods, see the links at this page.)

It’s not only global warming causing Iraqis to want to move to France and Germany, folks. Global cooling, too.

We find a statistically significant relationship between fluctuations in asylum applications and weather anomalies: Applications are lowest for average temperatures around 20oC and increase if the weather is too cold or too hot.

I like a 25oC day myself, and find that extra 5o makes all the differences. Anyway (the reader can guess I am straining hard to take this seriously), their Figure 1 (that blob above) shows the regression “coefficients on temperature”.

China, we discover, because of its lovely weather, has fewer asylum applications than frigid Mongolia. The weather accounts for the difference, mind. Toasty Iraq has bunches compared to citizens of the Great White North, who mysteriously do not appear on the figure. Golly.

Yes, temperature. That’s what accounts for asylum differences. Wee p-values confirmed it! “We find a highly significant relationship (P < 0.01 for joint significance)…”

Finally the predictions of increasing migration come, after “holding everything else constant”, from climate models predicting ever-warming temperatures. This approach is not wrong per se. You feed one model into another. Fine. But then you must carry the uncertainty from the first model through the second. Otherwise, you wildly exaggerate the certainty of your predictions.

This never happens, or not that I have seen, in climatology.

People are also far too excited to see predictions which confirm their worst “fears”, so they never bother to check the predictions against reality. One claim is that “aggressive behavior, which increases with temperatures”. Does it indeed. This is why the Singaporean military is so feared. And what countries, hot or cold, have launched the bigger wars this past century?

I’ll let the authors have the last word. But I want to tell you I wept when I read the following. “Both colder and hotter temperatures increase migration flows.”

Update The Times (UK) ran an uncritical piece on the paper here. And Der Spiegel quoted me here.

Wissenschaftler klingen anders: “Die dümmste, idiotischste Anwendung von Statistik, die ich seit Langem gesehen habe”, wundert sich der Statistiker William Briggs von der Eliteuniversität Cornell in den USA…

“Die ‘Science’-Autoren nehmen die Literatur zum Thema überhaupt nicht zur Kenntnis”, ärgert sich Fröhlich. Sie ignorierten gar 6000 Jahre Menschheitsgeschichte, bezögen sich stattdessen auf 15 Jahre in nur einer Weltregion, sagt Briggs.

“Warum”, fragt Christiane Fröhlich, “berufen sich die Autoren auf Zahlen aus der EU, wo doch ein Großteil der internationalen Migration woanders stattfindet?”

Wie wichtig Daten anderer Regionen wären, macht Briggs anhand einer provokanten Frage deutlich: “Warum nehmen denn nicht im kühlen Chile die Asylanträge rasant zu nach Hitzewellen in den warmen Nachbarländern?”

Not a bad rendering of what I said. Bojanowski did not ask and I did not say I was from Cornell, though, and he must have assumed that old affiliation using outdated information. I am a nobody from nowhere.

Update More press. New Study Linking Global Warming To Refugees Is ‘Most Idiotic Use Of Statistics’ This Expert Has Seen In A Long Time

Categories: Statistics

7 replies »

  1. I’m confused. They fitted two models both of which went round the outside of the data and not through it. How does that work then? What do you conclude if all your residuals are positive?

  2. So, you’ve now become (in my eyes indispendable, vital) part of the final click-bait fights (world will remember you beeing first, I think); I’ve just seen that “focus.de” bought that article from “Spiegel.de”, their title “Forscher zerlegen Studie, die neue Flüchtlingswelle prognostiziert” under “Hochspekulativ und dumm”. Both are dying weekly magazines, but dying online at speed of light (in run-up phase):
    – alexa.com/siteinfo/focus.de
    – alexa.com/siteinfo/spiegel.de
    I think that dying online presstitutes won’t die long, there are already way too many independent blogs (& twitterati, etc) which are not nurtured by EUdSSR and their dictatocracy member states.

  3. Well done Der Spiegel, very professional.
    They have to be complemented for taking the step of bringing the study to an independent statistician (William Briggs)
    … instead of just cut and pasting it …like most media seem do.
    Be a skeptic : ‘Claims are for testing, not just accepting’

  4. ?. I read the headlines. Then your critique. Now i feel compelled to read the study to understand what mechanism they propose. Alas it’s behind a paywall. On other time maybe. But on principle, since leaving the university, I treat non-open access papers as nonexistent.

Leave a Reply

Your email address will not be published. Required fields are marked *