Thursday, October 27, 2011

Against Actuariocracy

I really love this point that Matt Yglesias makes:
Instead of trying harder to improve our long-run budget forecasts, what we ought to be doing is discounting long-run budget forecasts much more in our evaluation of what belongs on the policy agenda. People in Washington treat it as obvious that a CBO projection of Social Security outlays in the year 2065 is a very serious subject to talk about, but that the legality of banning genetically engineered super-sprinters from Olympic track & field events in the year 2065 is frivolous. But I’m much, much, much more certain that we’ll have genetically engineered sprinters in the future than I am about the future productivity growth rate.
I can't really speak to the pressing issue of future Olympic sprinters, but...yes!

Look, it makes sense to run long-term projections for things like Medicare and Social Security. But -- and Yglesias has made this point before -- the margin of error for those projections is going to be just huge. Maybe health care costs will crash in the future, with Dr. McCoy type treatments available at low cost, yielding long productive lives followed by a quick, quiet death from "old age"...or maybe most of us will live well past 120 but spend our last 50-75 years or more in terrible physical shape, unable to work and draining all the resources that the under-70s can produce. Or maybe we'll be pretty much like we are now, only more so. Who knows? CBO sure doesn't. Yes, it makes sense to project and pay some attention to the "like now, only more so" possibility, but if either of the others happens the policies we enact now based on that possibility won't make any sense at all in retrospect.

(That includes if the dystopian answer is our future -- if that's what's coming, it's not going to make any practical difference that our actuaries in 2011 said everything would be great in 2080. And of course if the utopian future is on its way, we're just punishing ourselves now for no good reason).

So, no, I'm not going to say that the answer is to ignore the future entirely, but for public policy planning purposes as far as I'm concerned it's real important to pay attention to the present, somewhat important to keep track of where we're heading for the next decade or so, and worth keeping an eye on anything beyond that. If you think that's shortchanging the future, then the answer isn't to pay more attention to the actuaries; the answer is to pay more attention to long-term research and development (however you believe that's best accomplished).

By the way: how much did a good medical tricorder cost in Kirk's time, and how widely available were they anyway? We're totally ahead of pace on that one, right?


  1. If there ever are tricorders, Apple will claim exclusive design patents and sue everyone else to protect them. So the price will be high for that reason, at least.

  2. Jonathan, it seems to me you're missing Yglesias' point a bit. Yglesias is arguing that error is inevitable in forecasts because we can't easily model predictable Black Swan-type events (e.g. grinding recessions), which is somewhat different from your position that 'who the heck knows what will happen more than one decade out, might as well not worry about it until its closer'. All due respect, I don't think Yglesias is advancing that argument.

    Around this household, you often find yourself watching Annie with little girls. In one memorable scene, Daddy Warbucks excoriates FDR over the New Deal, which Warbucks describes as "badly planned and administered. You don't think through the implications of your programs, Franklin!" Warbucks thunders. The movie doesn't elaborate, but one can easily imagine Warbucks pointing out that the low burden of SS, given the short life expectancy of potential recipients, was unsustainable given the almost-certain lengthening of life spans as the program took hold.

    And while it was a bit early for science fiction, perhaps FDR pushed back with a science fictiony retort like from the open to this thread, that Warbucks wouldn't know that people in 1955 wouldn't die miraculously at age 65 or be whisked away to Klingon or something like that. And even if they weren't, how the hell would Warbucks know exactly what life expectancy would be in 1980?

    Answer is, of course, that Warbucks wouldn't. But he'd be on much safer ground arguing that life expectancy would be "something much higher than the 1930s, given, among other things, the beneficial impact of SS", than going for the science-fiction type explanation.

    So if you know a pundit that is saying, with certainty, that the unsustainable debt load in 2040 will be exactly $57 T, or whatever, that's a blowhard - but making a strawman of such a pundit, and following with "Hey people are as likely to voluntarily enter the StarTrek-brand cryogenics program at age 80 as just keep living on", well, that's not an especially confidence-building iteration of futurism either.

  3. Yglesias used a very unfortunate example to make a legitimate point. We can actually predict with the outlays for Social Security in 2065 for a given set of policy choices with more than enough accuracy to evaluate the impacts of various policy changes. The reason is very straightforward if you spend about 30 seconds thinking about it. The only uncertainty is how many people will qualify for benefits. We have really good tools to make projections like that because there is enormous inertia behind the trends. To qualify for Social Security in 2065, you have to be at least 10 years old as of the 2010 census and not die before 2065. For an initial estimate assume no immigration, just look the data up in the various actuarial tables, do the benefits math, and you know how much you are going to spend. Spend a half-hour running various immigration scenarios and you will quickly figure out that it doesn't matter that much. We'll have plenty of time to react if there are substantial changes in immigration.

    There are some systemic shocks (pandemics, nuclear war, etc.) that could invalidate your numbers, but rest assured that if any of them occur, no one is going to be worrying about the SS projections.

    Genetically engineered sprinters, on the other hand, are not going to be a problem in 2065. Sprinters don't reach their full potential until they are past 20 years old. That means the genetic engineering would have to be perfected by 2045. We don't have any idea what combination of genes would need to be tinkered with. And I don't see anybody down in Jamaica trying to figure it out.

    Of course, the correct example to have used was Medicare because we have no clue what will happen with health care costs (except that they will go up by some unfathomable amount).

  4. Of course, the question isn't just what life expectancy will be later in this century (maybe infinite, if they figure out how to stop DNA from degenerating), but also what the inputs into the system will be. What will be the means of production, and how much relative wealth will they generate? I would say it's safer to predict, today, that they'll be very different in 50 years, than it would have been to make a similar prediction in the 1930s. From then 'til now, we've had an economy that depends on (a) burning stuff, namely limited supplies of fuel, with masses of wasteful and dangerous by-products, and (b) acquiring food from actual plants and animals, more efficiently but not, in principle, differently than people have been doing for thousands of years. It's pretty clear that both of these are likely to change. We know, in princple, how to harvest energy from the sun, wind, ocean waves, and nuclear fusion (much more powerful than fission and without the radioactive waste). The problems of scaling these processes up and making them cost-effective are essentially technical -- not simple, but not likely to go unsolved. (See this, for instance: As to food, again it's pretty clear that having to grow plants and raise farm animals, although it will presumably still be done, will no longer be essential because food can be synthesized. There have even been successful experiments, which however I won't describe further, in creating edible food by extracting proteins from sewage, thus killing two birds with one stone.

    The point is, we know scientifically that all these things can be done, it's just a matter of solving some problems in engineering. If the leap from the 1930s to the 1980s included things like electronic computing, transistor and silicon technology, televisual broadcasting, space travel, nuclear fission, etc. -- all either unknown or at very preliminary stages in the '30s -- then I have trouble believing that the next 50 years won't see major, "game-changing" breakthroughs in areas where we're already pretty confident we know what can be done.

  5. Yglesias specifically mentioned Social Security outlays. Not revenues, not trust fund balances, but outlays. That's spending on a program that is driven almost entirely on actuarial projections. For that projection to be wrong in any material way would imply changes so vast that it won't matter that the projection is wrong.

    I completely agree with Yglesias's point. I just wish he had used as an example something like Medicare costs which we aren't very good at projecting even 5 to10 years in the future.

  6. Jeff, you make several good points. Two quick thoughts in reply:

    1) When we think of discontinuous innovation that 'sticks', there's more to the story than simple technical innovation. It seems to me that you also need a) a simple chain of command (i.e. one entity mostly/entirely responsible for effecting the change) and b) a non-dramatic habitual leap for the user.

    Perhaps this can be thought of via those ubiquitous Tom Selleck ads for AT&T in the early 90s, forecasting all sorts of great, unimaginable things that would be brought to you by AT&T. Some are so second-nature that its hard to remember they didn't exist ("Did you ever imagine you would use your phone as a camera? You will") and others still seem 100 years in the future ("Did you ever imagine you would fly through the grocery checkout without stopping at all? You will").

    What's the difference? The camera phone wasn't that discontinuous, and most of the innovation came through a single institution. By contrast, the 'fly-through-the-checkout-line' is a very big change, requiring participation from many entities.

    So food from sewage? Conceivable, but in action, much more like the 'fly-through-the-checkout-line' than the camera phone.

    2) Even when the market encourages complex discontinuous innovation, it doesn't necessarily create revolutionary GDP growth. Consider the electric car. Been around in prototype form for years, finally catching on because we're sliding down the backside of peak oil and gas prices are soaring accordingly.

    Is the electric car a GDP game-changer? Well, maybe, and I'm not that well-read, but my impression is that old-world entities (e.g. Exxon) have almost as much to lose - and thus the GDP loses - as efficiency gains.

    In general, it occurs to me that there are two or so great, discontinuous GDP growth periods since WWII. The first, from the late 40s to early 60s, was driven by improvements in infrastructure and related economic efficiencies. The second, in the 80s and early 90s, came from efficiency gains from technology and outsourcing. For the most part the union jobs lost from outsourcing were replaced (then), so these changes were not "zero sum game", and the economy grew.

    Is the food from sewage also a rising tide that lifts all boats? Or are there established aspects of the old economy that will be hurt by such change, perhaps almost as much as the helps that flow through from efficiency?

    Its a moot point, of course, and perhaps a pretty interesting debate, too. FWIW, I wouldn't put too much stock in zero-sum game technological innovations as GDP-growth breakthroughs.

  7. CSH, that may all be right -- you seem to know the history better than I do. But it leaves me with a few questions:

    > Didn't computerization involve a lot of entities? And wasn't at least some of what it brought along discontinuous from the standpoint of the users? Nonetheless, it still happened pretty fast.

    > Why would something like food from sewage be a problem for older entities? I would think that what would happen is that Archer-Daniels-Midland and the like would just buy up the technology(ies) and start selling the new product in place of the traditional, farm- and slaughterhouse-based products they're selling now. (And come to think of it, the sewage food might be an improvement.)

    > So what if existing entities rise and fall in line with the new developments? Isn't that what we've seen anyway over the last thirty years in, say, the media industry? Time (print) and Warner Brothers (movies/TV) merge into Time Warner, which acquires CNN, DC Comics and whatnot, then it gets bought up by AOL, then AOL flops and it spins off AOL, and meanwhile it acquires but later sells Google stock, etc. etc. From the peanut gallery, it looks like big changes -- major, long-time institutions like newspapers and magazines giving way to "new media" -- but on Wall Street, it's just an endless poker game where the deck is continually being reshuffled. Some people do well off it, some people less well, but whatever happens to GDP happens regardless.

    > Anyway, don't even some non-discontinuous innovations transform life and the economy? I don't know how economic historians rank it, but it seems to me that the internal combustion engine must have had a big impact on GDP. An economy that is making and selling cars and trucks is not only trading in those items, but is also moving other products much faster, making labor more mobile, etc., than a horse-and-buggy economy. This is bad for the horse-dealers and buggy-makers, but so what: They give way to (or become) Ford and GM. The basic user experience, though, isn't different from riding in a horse and buggy. The Tom Selleck of 1900 would be promising buggy-riders a future in which you can "fly" from city to city in your own vehicle -- and this turns out to be right. (More right than promised, in fact, because another technology is on the horizon that allows you to literally fly, but again, in an experience that involves sitting on a seat next to strangers while you're all hauled somewhere together -- essentially the same as in an old stagecoach.)

    In short, it seems to me we could have big breakthroughs that nonetheless worked their way into the existing economy pretty seamlessly. Let's say nuclear fusion is made cost-effective and comes online. This would massively increase the economy's upside, but basically you'd still be getting the end products the same way you access fossil-fuel power now: by plugging appliances into sockets (including your car, at that point). And it would all be handled by some existing companies and some new ones, presumably employing the people who used to or otherwise would be working for coal or oil companies. What's not to like?

  8. Off topic, can you change your blog settings somehow so that your comments are nofollow? When I searched for occupy wall street I get all results from "what mattered this week" replies (I was looking for that thing about the left getting a nice peaceful form of expression).

  9. The Federation's economy was sort of a communist utopia so the cost of tricorders was zero. I don't know if it's ever been addressed in the TV shows or movies whether your average Joe had access to medical tricorders. They certainly aren't presented as something that are too complicated for a typical person to operate. You would think everybody could have a medical tricorder and an emergency medical hologram in the home to take care of most medical needs. But maybe the AMA was able to use their lobbying power to require a medical license to purchase a tricorder to preserve doctors' jobs and income.

  10. Jeff,

    Thanks for your comments. What I was trying to get to was a distinction between discontinuous innovations that "replace" other critical elements of the economy from innovations that allow all participants to function more productively. So a better highway system improves the efficiency and cost structure of all market participants; it doesn't "cannibalize" any existing GDP - and thus helps facilitate disproportionate growth.

    Not many innovations are victim-free; ask a former employee of the fine Smith-Corona company if generally helpful tech innovation had no victims. But some innovation is more cannibalistic of existing GDP than others; Smith-Corona excepted, computing technology benefits generally had relatively few GDP offsets - electric cars possibly not so much. But I'm not the expert.

    Just an aside: its not likely that ADM will be doing the sewage food until its too late; put yourself in the shoes of the Senior Vice President of North America Foods at ADM, where your responsibility is for P&L growth on existing businesses. Will you go for the sewage food if it cannibalizes your current P&L and compromises your results, and bonus? Probably not until the point where your base business has been so badly compromised that you can resist it no longer.

    That being said, I think I was wrong about the sewage food last night. The aforementioned VP at ADM notwithstanding, sewage food is actually an example of discontinuous innovation that might indeed benefit almost everyone and help facilitate disproportionate GDP growth.

    (But its a bit gross to think about).


Note: Only a member of this blog may post a comment.

Who links to my website?