Bad Science



Patrick is a mid-40 year old geek with an undergraduate degree in mathematics and a master's degree in Information Systems. Nothing he says here has anything to do with the official position of his employer or any other institution.

Related Post Roulette

48 Responses

  1. Avatar Burt Likko says:

    Three things jump out at me — other than the unexplained (and easy to miss!) bid of statisticplay you point out, @patrick:

    1. Deaths by auto accident have had four dramatic dips: 1973 (mandatory seat belts?), 1981 (ABS?), 1988-1992 (mandatory air bags), and 2007 (mandatory crumple zone construction?). I’m interested if my guesses about why these safety improvements happened are right.

    2. One-third of people are living to age 85 or older, a two-and-a-half-fold increase since the statistics begin in 1968. That’s pretty remarkable when you think about it.

    3. Given that lifespans are increasing so, it’s remarkable that spending on nursing homes, hospice, etc. for those of advanced age has declined as an overall piece of the healthcare spending pie. That must mean that other kinds of medical care are growing in cost much, much faster. What are these dollars being spent on?Report

    • Avatar Patrick says:

      (1) That sounds like pretty good guesses to me. 2007 might also be an inflection point for young driver population; I’m given to understand that the current generation has a much lower percentage acquiring their DL’s in their mid-teens.

      (2) Yeah, that bodes well for my plan to make it to 2112.

      (3) My understanding is a big chunk of the increase in cost is spent in the last 30-90 days of life, usually prolonging terminal patients. I’m very possibly wrong, though.Report

      • Avatar Burt Likko says:

        Let’s assume @patrick is right about (3); I’ve got similar notions floating about atop my memories too. One wonders how this category is not necessarily self-defeating: you identify the date of the patient’s death, count backwards X number of days, and add up the cost of the healthcare incurred in that time. It seems that almost definitionally, you’re going to be getting a much higher number than you get earlier in the patient’s life.

        Significant medical events seem like they’d tend to be expensive as an inherent matter. If a person spends, let’s say, a million dollars over the course of a lifetime on medical care, isn’t it fair to assume that most of that comes in one or two big chunks? A post-trauma surgery and possible recovery therefrom, or the manifestation of a grave illness, just naturally consume huge amounts of money. So of course we’re going to have a big ol’ number looking at immediately-before-death medical care, precisely because it’s looking at major medical events that both triggers risk of death and large amounts of medical spending.

        Point is, if we’re looking at terminal-care costs, I think we’re not getting a good picture unless we also factor in morbidity rates for big blocks of treatment too. Only that’s, like, complicated.Report

      • Avatar Patrick says:

        All that’s true, B-man.Report

      • Avatar Tod Kelly says:

        As far as I know, there are several things that coincided around 2006-8 that dropped fatalities. One has to do with an explosion of black boxes in corporate fleets, which is actually a big pool of vehicular accidents victims. (Vehicular deaths is the runaway perennial leading cause of workplace deaths.)

        The other is that for various economic reasons, people were starting to drive a lot less, both business and personal.

        I have never heard the crumple-zone reason given, Burt, but it certainly makes a hell of a lot of sense. I would be surprised if that isn’t a significant factor.Report

    • Avatar Michael Cain says:

      (1) ABS still isn’t mandatory in the US, IIRC. Consumer preference is driving the adoption, much as it has driven adoption of automatic transmissions. How do the sharp declines line up with recessions and oil-price spikes, both of which have dramatic short-term effects on driving habits?

      (3) We’re up to almost half of Medicaid spending going to long-term care, primarily for the elderly. In most states, the nursing home industry isn’t viable without the Medicaid payments coming in. States have worked like hell to hold down payments to the nursing homes.Report

    • Avatar Snarky McSnarkSnark says:

      I would like to hear the libertarian response to #1…Report

      • Avatar J@m3z Aitch says:

        I don’t know the details on all of those, but I do know seat belt use became common before being mandated. Is it really a surprise, given that there’s no significant externality there? If you get chucked through the windshield, you suffer the full consequence–it’s not someone else’s face that gets smashed or whose neck gets broken.

        That said, mandating them can increase usage, hence increase lives saved. However because safety increases costs, it can cause older cars to be kept on the road longer, and can price poorer people out of the safety market. So an important consideration is how much safety is mandated, how much price increase it creates. That is, poorer people would like to buy some increased safety, but can’t afford all the available safety features. By mandating all available safety features, we can create a situation where you can’t buy an increment of increased safety, so people who can’t afford all the mandated safety buy older used cars that are less safe than they would prefer to buy.

        As an aside, I doubt ASB is a cause of a dip. From what I read some years back. ASB seems to create a moral hazard. People rely on the ASB, so they drive closer to the cars in front of them, and–as of when I saw this–ASB had not produced a decrease in rear-end collisions. I’d love to hear that had changed, but I’m dubious. Moral hazard is too common a human behavior to be very optimistic.Report

      • Avatar Snarky McSnarksnark says:


      • Avatar ktward says:


        Your first link is to a 13yo study. So I stopped there.

        … because safety increases costs, it can cause older cars to be kept on the road longer, and can price poorer people out of the safety market.

        Are libertarians really still debating auto safety regs? Really? It’s 2014. When was the last model year that cars were legally cranked out sans seatbelts? I admit I’ve no idea what that answer is, but I have a really hard time imagining that any of those cars are owned by anyone not wealthy enough to afford a vintage car. Crikey, I see 20+yo crap Toyotas and Hondas on the road every day and they all have shoulderbelts. Eventually, even the poorest among us benefit from safety regs.Report

      • Avatar dhex says:

        “Eventually, even the poorest among us benefit from safety regs.”

        the “eventually” – the speed of trickle-down safety – would be his point, i think.Report

      • Avatar Jim Heffman says:

        “Are libertarians really still debating auto safety regs?”


      • Avatar Mad Rocket Scientist says:

        Are libertarians really still debating auto safety regs?

        Case in point, rear view cameras. The mandate just got passed down that all cars will have them in the next few years.

        Safety advocates have been pushing for these for quite a few years, but there has been stiff resistance. Now the resistance is fading. Why?

        The cost of the factory installing a backup camera system in a new car is ~$45 for the camera & installation. The big cost was never the camera, but the display. Nowadays, all new cars come with in-dash flatscreens that cost less than $100, so tacking on the camera is a pittance.Report

      • Avatar Mad Rocket Scientist says:

        To expand, mandates that occur after the technology is mature and the cost has become trivial don’t cause problems. Mandates that happen before this point tend to cause a price spike that takes a while to dissipate.

        When evaluating a safety mandate, the cost versus gain has to be worth it. Backup cameras will probably save a handful of lives every year. If the mandate puts $500+ on the price of every new car sold in the US, the cost would be almost $8 B (15.6 M sales * $500 = $7.8 B). At $50, the financial impact to hopefully save a few lives is a lot less.Report

      • Avatar J@m3z Aitch says:


        I just selected a handful of studies from respected sourced for others to peruse. I didn’t choose ones that matched my viewpoint. In fact I think that first one may work against my argument. And if we’re talking about seatbelt regs, which came in well before 13 years ago, then objecting to a 13 year old study is hard to defend.

        One of those you chose not to look at is an NIH study. Obviously a libertarian plot to present dated and anti-government research.

        And, yes, there’s still debate to be had on auto safety regulations. Not just libertarian debate, but economic cost-benefit debate, regulatory policy analysis debate, whether the next step forward in safety ought to be mandated and if so how quickly. To pretend there’s no more room for reasoned debate is to demonstrate superficial knowledge of the topic.

        If you want to phrase this all in ideological terms, that’s all on you. But there is academic literature on these things.Report

      • Avatar J@m3z Aitch says:

        To follow up on Mad Rocket Scientist, if we peg the value of a human life at $5 million, then if the backup camera mandate cost–using his hypothetical–$8 billion a year, they’d need to save 1600 lives a year for the regulation to make sense.

        In fact they’d have to save X more lives, because the mandate will orice some people out of the new car market, so they’ll continue driving a less safe older car. X of those people will have deadly accidents because they are driving an older, less safe car, instead of a newer car.

        Now for some real nimbets. According to this report, they may save 70 lives a year. At $5 million/life that’s a $350 million savings. At 15.6 million car sales a year (using MRS’s numbers), the cameras need to cost about $22 1/2 each. That’s very rough, and since most cars will be on the road more than a year (although the real factor to consider is how long the cameras are effective, and which is lower, the average camera life or the average car life), they can actually cost more and a mandate still be cost effective. Using MRS’s calculations again, the total cost appears to be ~$140. Now there’ll be some discounting I think, for the additional years, but I’m not sure what that’d be, so the final calculation is beyond what I can do here. But done correctly it will either demonstrate cost-effectiveness or not.

        Of course if we accept a much lower value of human life, say $129,000, then the cameras are going to need to cost about $1.70 to be a cost effective mandate.

        Of course if they’re not mandated, but just an option, any consumer can estimate their own expected net value of buying one. Insurers might get in on the act, too, giving discounts for having them.

        It’s true that backup cameras will not become universal, or will do so more slowly, without a mandate. But that is a bad argument for the mandate, because it relies on absolutes, rather than a cost-benefit analysis.

        So, yes, these issues are still debated. It would be irresponsible not to debate them.Report

      • Avatar Snarky McSnarkSnark says:


        To expand, mandates that occur after the technology is mature and the cost has become trivial don’t cause problems. Mandates that happen before this point tend to cause a price spike that takes a while to dissipate.

        This is quite chicken-and-egg. I grew up in Los Angeles, where, on summer days, visibility was often no more than four or five miles because of smog. California had its own “clean air” standards, then, and because of its large population was more-or-less able to mandate pollution controls to car manufacturers.

        Every cycle, California would mandate an emission level (or specific technology) for cars to be sold in California several years hence. And every time, the car manufacturers would asset that it was “impossible” to meet the mandate, and that it would add many thousands of dollars to the price of the car. And every time (without exception, to my knowledge), they would manage to conform to the mandate, at a net cost-per-car that was a mere fraction of their panicked early estimate.

        The price of these technologies (and most technologies) goes down when they begin to be manufactured in scale. And, pretty soon, they simply begin to be included universally, because that turns out to be more economic than contingent design and assembly.

        So I would assert that without the mandates, no cars would have air bags, catalytic converters, or even brake redundancy.Report

      • Avatar Will Truman says:

        Snarky, I think you’re right about the mandating something having the effect of lowering price, but automakers were already ramping up on airbags (again) by the time the legislation passed. It was already the case that they were likely to become common, if not universal.Report

      • Avatar Mad Rocket Scientist says:

        As Will said, in most cases, the technology already exists & is already offered as standard on high end cars & optional on others. Catalytic converters are one of the few times inclusion resulted from regulations (instead of the other way around). Granted, cats were already around prior to the 1970s (developed in the 50s because an engineer was worried about LA smog & wondered if something could be done…). Part of the reason they were not widely used on cars was because of leaded gas (it destroyed the cats), and because they were very expensive with rare earth metals needed for the reaction (even today, a new cat for my Tribeca is ~$500). So not only were cats expensive, regulations had to change fuel formulation which caused engine re-designs.

        So I suspect that yes, the cost of cars post cat mandate was much higher than pre. Of course, the best way to know that would be to look at historical price data from before & after the mandate & see. I tried a quick search to find such a resource & came up with this:

        It also includes links to the People’s History, with some specific pricing data.

        In short, quick analysis of price data from the 1970’s is problematic because inflation in the 70’s was pretty severe, so one has to adjust for that as well, but the cost of cars did dramatically increase in the 70s, much faster than wages rose.Report

      • Avatar Kim says:

        Estimated life of cars on the road now is a lot longer than it was in the 1970’s, so I assume that regulations ought to be getting more scrutiny now.

        Still, the more we automate cars (that sonar idea sounds awesome), the closer we get to fully automated cars, which will save TONS of lives.Report

    • Avatar The Priests says:

      @patrick that bodes well for my plan to make it to 2112.

      We’ll be waiting for you in our nice contented world, my son.

      We’ve taken care of everything.Report

  2. Avatar trizzlor says:

    Nice catch. But the slides describing the impact of the AIDS epidemic hit really hard, it’s rare to see such substantial trend shifts. I would call it malpractice the way our government treated that outbreak.

    The visualization is terrific. I happened to be listening to this while watching and it felt like I was scrolling through a slick movie opening. It’s interesting that the structure is more like that of a PowerPoint presentation than a news article. The “data journalists” at Vox seem to be going in the same direction as well with what they’re calling “cards” instead of slides. I wonder who decided this was a good way to present things, given that most people I know and work with can’t stand the thought of PowerPoint.Report

    • Avatar Michael Cain says:

      I’m probably as vocal as anyone in my dislike of PowerPoint, but that’s mostly because so many people are so bad at slide design. This visualization works well in the up-close environment under one viewer’s control in which they’re intended to be viewed. Put it up in the usual sort of conference-room environment and it probably sucks. For example, the font size is almost certainly too small to be legible, stuff would be appearing and disappearing at the speaker’s speed rather than the viewer’s, etc.

      One of the things too many PowerPoint users do that just drives me batty is to have the font size change from slide to slide depending on how much text there is. First it’s 16-point; then it’s 12-point; then it’s 32-point; for God’s sake, man, pick a size that’s comfortable to read and stay with it!Report

      • Avatar trizzlor says:

        Good point, but much of this information could have been conveyed to me through standard narrative text rather than in figures. I wonder where the decision of putting the data front-and-center comes from. Part of me thinks that people are just less likely to sit down and read something on the internet. Another part of me wonders if this “just the facts” style is meant to shield the writer from any claims of bias, when – as Patrick pointed out – data presentation can be just as biased as an editorial.Report

      • Avatar Patrick says:

        One of the things too many PowerPoint users do that just drives me batty is to have the font size change from slide to slide depending on how much text there is.

        That’s the default.

        It’s a surprisingly huge pain in the butt to change that.Report

      • Avatar trizzlor says:

        Actually, I may be contradicting myself. Vox puts their information into “cards” ( but each card is still a fairly typical narrative. I don’t really know why they do it, or why they’re called cards, or what experience they’re trying to get at. Perhaps it’s taking advantage our lizard-brain love of lists and clicking “right” on the keyboard. But I do seem to read them to the end a lot more than the articles on Wonkbook.Report

      • Avatar Tod Kelly says:

        Not that anyone cares, but about seven years ago I made two adjustment to my presentations that have really had a huge impact on their popularity.

        The first was switching from Power Point to Keynote.

        The second was to cut almost all of the written words out of them, and use images instead.Report

      • Avatar Jim Heffman says:

        It’s important to keep the application in mind when creating a presentation.

        If there will be someone present to say words at the audience, then you don’t need to put words on the slide. Say, a keynote speech or a fundraising pitch.

        If there will *not* be someone present, then all the words *do* need to be there. This is for course material or a writeup of scientific results, where the slide package is intended to be the actual deliverable content.Report

  3. Avatar zic says:

    I frequently see this kind of error in economics (even by economists,) comparing household incomes instead of hours worked; particularly a problem when we look at incomes over time, and than discussing it as if it equals wages.

    I think there are similar errors in looking at employment rates over time because we stopped counting those who’d given up looking for work in the Clinton admin.Report

    • Avatar Patrick says:

      Context: my undergraduate degree was in mathematics, I’m a passable but lapsed statistician, and the last several years I’ve been working part time, very slowly, on a degree in information science that has a double minor in business and politics and economics.

      I’ve taken four classes on research methodologies, which is probably three more than most people who do economics or policy analysis.

      Still, it’s hammered freakin’ home that you don’t make these sorts of errors, by most competent faculty. You get lambasted for this stuff. An error like this in a presentation like this one won’t get you killed, it’s not central to the thesis of the presentation. But you’re still going to get points taken from Gryffindor.

      So given the frequency with which I see these errors in public policy discussions, I’m forced to come to one of a limited number of conclusions:

      (a) Most schools who teach higher level economics or policy analysis do a terrible job of explaining research methodologies;

      (b) Lots of people in policy analysis didn’t get degrees in the field; or

      (c) They know exactly what they’re doing.

      Given that I see a much higher propensity for these sorts of errors from public policy think tanks with an ideological bent, than I do from, oh… say… folks who actually do stuff like work for 20 years in public service doing data analysis, it’s pretty much become one of my giant red flags for, “Hey, these people are lying sacks of shit”.

      They *know* this stuff is misleading, and they do it on purpose.Report

      • Avatar zic says:

        What little I know, I learned from a book on math and statistics for reporters and the Wall St. Journal’s style guide.

        But I learned more from talking to people who actually generate numbers while reporting. Because I was pretty ignorant, I’d listen/review/study whatever numbers they had, and they try to ask questions that helped me put those numbers into context for my readers. I made it a habit to always go back — hey, I want to make this analogy, does it work? Have I got it right?

        And did they ever love me for that. I had dozens and dozens of researchers who’d let me know when they were publishing new work because they trusted me to report on it correctly and not cause a media cascade. I think that’s unusual, too; from the responses I got from sources when I’d call back to verify that what I’d written actually reflected what their data supported. I think this also applies to blogs, I’d encourage people writing here to be fearless about tracking down the people who generate numbers and ask for help giving them meaningful context. They’ll appreciate it, and are typically generous with their time.

        So it’s a compounding error — folks with ideology and a point to prove and reporters to lazy and sure of themselves to verify the conclusions they draw. (Editors, too, sometimes, writing headlines not supported by the words that follow.)Report

      • Avatar Troublesome Frog says:

        The Wall Street Journal has a style guide that deals with properly using statistics? That must be one really sad document.Report

      • Avatar zic says:

        It was written before the present owners took the helm.Report

      • Avatar Kim says:

        you should see doctors’ research. It’s notoriously bad with statistics.
        And they’re not trying to mislead…Report

      • Avatar Kim says:

        yes, it’s very very unusual to find a good science writer. you’d think that talking with the folks after you’re done with your writeup would be “ordinary business” but sadly it’s really not.

        I’m convinced this is why a lot of scientists/economists have blogs online… better to be able to say it in your own words.Report

      • Avatar Jim Heffman says:

        “So given the frequency with which I see these errors in public policy discussions, I’m forced to come to one of a limited number of conclusions…”

        There’s a fourth option: “I’m smart, and I’ve thought about this, therefore I’m right, because when smart people think about things they always get the right answer!”

        Combine that with a healthy dose of the Dunning-Krueger effect and you see how someone can say “it used to be twenty percent, now it’s ten percent, that’s only a ten percent reduction in support for the policy!”Report

      • Avatar zic says:

        @jim-heffman I’m sorry, but comparing apples and oranges and then writing a report on bananas does not one a fruit expert make.Report

      • Avatar Patrick says:


        That goes a long way to explaining bad science reporting, but typically “I’m smart therefore I’m right” doesn’t get you past peer review.Report

    • Avatar Road Scholar says:

      Some of that is undoubtedly the streetlight effect. The statistics for incomes, both individual and household, are readily available from public sources like the IRS. Hours worked is a lot harder to get at, requiring surveys and such, and historical data may be unreliable if available at all.

      I don’t fault researchers for using proxies when that’s all that’s available. But I do fault them for then turning around and pretending that they’ve proven something the data doesn’t support, just because they really, really want to prove that particular point. I can much better respect an honest “I dunno.”Report

      • Avatar Patrick says:

        The honest answer is, “Analyzing the data that we have supports the need for better research, better data. It seems more likely than not that there is a problem here, but we can’t say with a reasonable degree of accuracy or certainty. Give me more funding, you tightwads!”Report

    • Avatar Jim Heffman says:

      “I frequently see this kind of error in economics (even by economists,) comparing household incomes instead of hours worked; particularly a problem when we look at incomes over time, and than discussing it as if it equals wages.”

      Careful, you’re dangerously close to claiming that the apparent wage gap between men and women is actually a result of improper statistical analysis rather than an actually lower per-hour payment rate for women.Report

      • Avatar zic says:

        Actually, I’d argue the wage gap is mostly the result of men failing to exercise their right to engage more with their families.Report

      • Avatar Kim says:

        zic’s right. The wage gap is mostly because women have fewer years worked due to spending time raising a family (childbirth, caring for the kid the first year or two…)Report

  4. Avatar Christopher Carr says:

    Is committing suicide by lethal injection more humane than committing suicide by firing squad?Report