Deinonychus antirrhopus

New Economy Redux?

This is an interesting article on Social Security over at the American Enterprise Institute. The thrust of the article is sure, there is a demographic problem for Social Security (SS) and yes, it is quite likely the economy wont save us (i.e., economic growth wont be large enough to offset the increased expenditures in SS), but there is every reason to believe that science will save us.

The economic implications of this growth in computing power are staggering. We are used to thinking of the capital stock as growing slowly. As computers take up an increasing share of the capital stock, and as they continue to improve by a factor of two every two years or so, the capital stock will begin to grow at astonishing rates. Labor productivity, which we are used to thinking of as increasing at a rate of 1 to 2 percent per year, could increase by 4 percent or more in the next decade, and we could see double-digit increases in the years that follow. Consequently, Kurzweil foresees a future 25 years from now in which “there is almost no human employment in production, agriculture, and transportation.”

While all of this might be true, I don't think this should be the bed rock of public policy. It is sort of a "hope that science will save us" view. This strikes me as policy that will ramp down expenditures on freeways and roads, because in the future we'll have flying cars.

Look at all the goodies we have now. Pretty impressive computers. Thirty years ago, many people might think you're crazy if you told them the future would have computers on desktops that could crunch through gigs of data and be point-and-click simple. We have cell phones, and ways of storing vast amounts of data in amazingly compact ways (a DVD can hold gigs of data).

Suppose we could construct a technology index and set say, 1968=100. What would the current value of the index be? 200? 300? 600? Has human productivity increased by that much? No. What happened? Well, all these advances in technology have allowed people to have more time to engage in non-productive activities.

For example, you boss says, "Prepare a report on costs for district 5 for the last six months." Now instead of having a few sheets of white paper with some typed up tables you have used Excel to put together a slick little report. You have borders, shading, bold, italics, color, and maybe even some cute little pictures. Now you have spent a considerable amount of time putting all that together, but it does not provide anymore information than the old report that was done on a typewriter.

Lets not get into how many man-hours of work have been lost playing mine sweepr, soletaire and other games. Factor in the net and browsing and it becomes pretty clear that people aren't doing that much more work even though technology has increased tremendously.

The economics of this is pretty simple. People make a labor-leisure decision. How much am I going to work and how much non-work am going to engage in. If they can engage in the same amount of work for less time, it makes the person better off. The person is not going to do more work without additional compensation.

It strikes me that these kinds of "exponential reasoning" while very seductive is actually potentially very misleading. I could very well be wrong, but I don't think that science and additional computing power are going to bail us out of this problem.

Link via Arnold Kling

Update: Arnold Kling responds to the comments on this topic here.

His response about the nonlinear nature of these sorts of things does have some validity to it. Here is a simple example.

The blue line is a simple exponential function, and the red line is a linear approximation. One might believe that an exponential phenomenon is linear early on. This could be the case.

I'm still not convinced that this is going to be the near utopian boom that the initial article is claiming. Don't get me wrong, I hope it is, but I'm doubtful.

Also, we are not arguing fine points. Arnold Kling and I do agree that basing public policy on such outcomes as if they were likely or some sort of baseline is probably a bad idea.

Update II: Arnold Kling, in comments, notes that my graph underestimates the potential for error. This is true. If we extended both series out say another 20 or more data points we'd see that the linear approaximation is way off.

The point of my graph was to show how easily one could mistake a nonlinear process as an linear one when you don't have sufficient data. The example above has 30 data points, which isn't great, but isn't totally horrid either.

Its an interesting issue. I'm hoping to look into it a bit more.

Posted by Steve at October 30, 2003 01:20 PM