Really bad predictions

Hard scientist Clifford Stoll writing about the internet in Newsweek in 1995.

Not only economists make bad predictions!!

Via the Agitator. Feel free to insert your own snide remark about Paul Ehrlich.

"Confirmation bias is everywhere"

Reason's Ron Bailey summarizes provocative research on the role of scientific literacy and cultural commitments in views on global warming / climate change.

Thought question: does the confirmation bias study itself confirm my biases? That would, I suppose, be meta confirmation bias.

Underwater shipping containers

A report on scientists studying a shipping container that landed at the bottom of the ocean.

I was surprised by the number of containers lost in the sea each year.

Via MR

Attorney general out of control

Having state attorney generals use rules designed to prevent fraud to conduct amateur science reviews with real penalties seems like a bad idea to me.

Is there no actual crime in Virginia for the attorney general to attend to?

Truth in labeling would be increased if the name of the attorney general position were changed to "media whore who wants to campaign for governor or senator while collecting a taxpayer funded salary".

Full disclosure: I have a graduate school friend who is a state attorney general (in a different state). My sense is that he behaves better than most.

Economists versus climate scientists

This piece compares the reactions - calm, measured economists versus freaked-out climate scientists - between two sets of scholars studying climate change.

The piece argues that climate scientists and economists differ in their beliefs about the adaptability of social and economic institutions to climate change. I think that is correct.

I think there are also likely differences in the process of selection into economics and climate science that feed into these differences in attitudes. A cynic might also argue that the two groups face different elasticities of research funding with respect to predictions of doom.

There is room for some deeper sociology of science in this area.

Oh, and we should have a carbon tax.

Via this Matt Yglesias post.

The science of hugs

A snippet on research on the duration of hugs, using video from the Olympics.

The thing that struck me about this was the mismatch between the theory and the test statistic. If it really is true that we all experience life in three second bursts, then all hugs should last some multiple of three hugs. The research as reported instead looks at the average. But the average could be three if half of all hugs lasted one second and half of all hugs last five seconds, in which case the theory would hold for no one.

Hat tip: Charlie Brown

Evidence-based medicine: flu vaccine

This Atlantic piece, now a few months old, shows the difficulty in really, really using the evidence to drive public health policy. Many of these same phenomena carry over into the world of labor market policy, such as the existence of dozens of studies, only a handful of which should receive non-zero weight in any serious synthesis of the evidence.

Biased research on smoking ...

Here is an example of why moving all research from the drug companies to government funded labs and university researchers will not solve the problem of biased research.

Though one should not omit the possibility that the researchers hired by Health Canada to do the research simply were not very good or did not include on their team someone familiar with survey design. An easy way to deflate nearly any applied economist doing research using survey data is to start asking about the details of the data collection, either in terms of the sampling or in terms of the survey structure and question wording.

New measurement error paper

A new paper on measurement error in survey measures of earnings, specifically in the U.S. Current Population Survey data used to calculate the unemployment rate, from Chris Bollinger and Barry Hirsch.

My general feeling is that measurement error is ubiquitous, sometimes substantively important, and grossly understudied by economists and other social scientists.

This is meat and potatoes research that has potentially large spillovers across subject areas and disciplines. Research funding agencies would do well to fund it.

Cheap Talk on self-abuse

Cheap Talk mounts an evolutionary defense of wanking. I had never thought of comparing it to tickling, which is the key point in the argument. Clever!

Seems appropriate to link to this as well.

Measuring production in science

My friend and co-author Julia Lane at NSF offers some thoughts on measuring the output of scientists, broadly conceived.

The problems here are many and deep. The two I think are toughest are dealing with disciplinary differences in the nature of scientific output and dealing with the perverse incentives created by unsophisticated performance metrics. In the latter case, examples are easy to find in the private sector, in universities and in government of poorly designed performance measures consuming resources and diverting activity away from production toward strategic manipulation.

But, it is surely worth trying to do better.

Scientists and engineers

Both clever and funny ...

Hat tip: Laura Kawano

Science news cycle

This captures it perfectly!

It is a mixture of universities wanting publicity, news outlets wanting to generate excitement and fear, and ignorance of how science works on the part of journalists and the public. The only thing the comic leaves out is the wild over-representation of counter-intuitive findings in the media, because of course they are news, and confirmation of existing views is not.

Hat tip: the agitator

The AP gets a C for their coverage of a study on prenatal pollution exposure

The Associated Press reports on a study of the effects of localized pre-natal pollution on child IQ:

The results are in a study of 249 children of New York City women who wore backpack air monitors for 48 hours during the last few months of pregnancy. They lived in mostly low-income neighborhoods in northern Manhattan and the South Bronx. They had varying levels of exposure to typical kinds of urban air pollution, mostly from car, bus and truck exhaust.

At age 5, before starting school, the children were given IQ tests. Those exposed to the most pollution before birth scored on average four to five points lower than children with less exposure.

The researchers studied pollutants that can cross the placenta and are known scientifically as polycyclic aromatic hydrocarbons. Main sources include vehicle exhaust and factory emissions. Tobacco smoke is another source, but mothers in the study were nonsmokers.

A total of 140 study children, 56%, were in the high exposure group. That means their mothers likely lived close to heavily congested streets, bus depots and other typical sources of city air pollution; the researchers are still examining data to confirm that, Perera said. The mothers were black or Dominican-American; the results likely apply to other groups, researchers said.

The researchers took into account other factors that could influence IQ, including secondhand smoke exposure, the home learning environment and air pollution exposure after birth, and still found a strong influence from prenatal exposure, Perera said.

Dr. Robert Geller, an Emory University pediatrician and toxicologist, said the study can't completely rule out that pollution exposure during early childhood might have contributed. He also noted fewer mothers in the high exposure group had graduated from high school. While that might also have contributed to the high-dose children's lower IQ scores, the study still provides compelling evidence implicating prenatal pollution exposure that should prompt additional studies, Geller said.

The good:

(1) The study is actually forthcoming in a peer reviewed journal.

(2) The article actually mentions some but not all of the conditioning variables and also quotes someone who notes a very important missing conditioning variable, though the article implicitly understates the importance of the issue.

The bad:

(1) The article does not provide a link to the study.

(2) The article does not describe how the sampling was done. Where did these women come from who are willing to wear pollution measuring equipment while already carrying around a soon-to-be-born baby?

(3) The overall tone should be more skeptical.

My take on the study:

(1) Bonus points for the cool measurement methodology.

(2) Given that the authors apparently did not control for either of the parents' IQ scores or their educational attainment, it seems quite unlikely that the selection-on-observed variables holds here. The likely direction of the bias is in the direction of the estimated impact.

(3) Bottom line: this study suggests that it might be worth it to fund a study of the same treatment with a better identification strategy. I don't think you can say much more.

Hat tip: Sue Dynarski

Reuters does social science, badly

This article from Reuters, forwarded to me by one of the students in my just-completed undergraduate econometrics course, nicely illustrates everything I complain about in that course regarding discussions of social science in the supposedly highbrow media.

First, we learn:

The survey examined sexual experiences as well as family relationships and lifestyle habits of Japanese females and males aged 16 to 49. It was carried out in September and was based on about 1,500 people.

This is all well and good but on might have a few tiny questions still remaining about the data collection. For example, was it a random sample of the Japanese population or was it a sample of, say, subscribers to a magazine? Second, what was the response rate? Was it something reasonable, like 80 percent, or something awful, like 20 percent?

Next we learn that:

The average age of first-time sex for those who said they ate breakfast every day as a middle school student was 19.4, while for those who skipped breakfast, the average age was 17.5.

Did they really just take a mean difference and report it? Perhaps there were some conditioning variables, though the text sounds like there were not. Perhaps they did some sort of highly advanced statistical procedure, like a t-test, to gain some sense of how likely the observed difference is to arise by chance in a world where this is in fact no difference in population proportions? Is there any conceivable reason why the reader should suspect that these estimates represent causal effects?

But it appears that the authors of this study had no time to waste on foolish technical trivia. There are important interpretations to be made, such as this:

"The fact that people can't eat breakfast may show something about their family environment," said Kitamura. "Before blaming individuals for having sex at an early age, it may be necessary to look into the sort of homes they are from."

There is even a hint of some policy conclusions and, of course, more things for our friends at the government to do when they aren't busy doing such a bang-up job of looking after the economy.

What we do not learn, in addition to all these details, is where to find the study so that we can read it ourselves. Would it be too much to provide a link? Isn't that the very start of using the wonderful power of the web to create new knowledge?

Absent a link, perhaps the curious reader could learn the name of the journal in which the study was published. Oh, wait, the study is not published yet. Reuters is telling the world about a study that has not even been through peer review. Sigh.

Score a big fat F for Reuters on this one and another big fat F for the Japan Family Planning Association.

Hat tip: Russell Bittmann

National differences in sexual (and survey response) behavior

The economist summarize the findings of one David Schmitt at Bradley University who studies, among other things, cross-national differences in sexual behavior. The economist piece is here and Schmitt's website with more findings is here.

The Times of London focuses on large Western countries and finds that (pun surely intended) the UK comes out on top.

I have to say, these rankings do not coincide with my casual impressions, nor with what you would expect based on the relative religiosity of various countries. In particular, I would put Canada above the US and the UK below both of them.

Passing the smell test

The economist reports on recent research involving men, women and perfume.

Is this why they start hockey games with a face-off?

According to the study described here, men (but not women) with wide faces are more aggressive.

Hat tip: Charlie Brown