BAD READING

This month I will start with a book review, R. Formaini's Myth of Scientific Public Policy. It is, in my opinion, a bad book. The author is a vice president of the Cato Institute, he abundantly quotes Austrian economics, Mises, Hayek, Rand, Friedman and other friends of free enterprise and liberty, and comes highly recom-mended by M. Rothbard.

Is that bad?

Yes, very bad when the author delves into a subject he does not understand, because that is bad publicity for good names.

The author essentially says that basing public policy decisions on scientific methods such as risk assessment is a myth, not because (as I say) such methods merely act as window dressing for policies that are in fact based on optimizing media approval un hopes of programming the electorate, but because he thinks risk assessment and the rest of probability theory is a mathematical exercise that can neither be proven correct nor falsified and has no bearing on reality.

Now that is a blunder so absurd that it would not be worth noting were it not for the names the author uses as an umbrella (including the name of Ludwig von Mises' brother Richard, who was a mathematician and probabilist, but who is quoted here in utterly misunderstood connections).

If decisions in real life cannot be based on probability theory, what made the Las Vegas game halls change the dealers' rules of playing poker, after a professor of probability theory kept winning by basing his decisions on game theory?

Not only can such decisions be taken, but they can be automated. When a computer fishes a signal out of random noise, it has nothing but probabilistic criteria to go by, and it is, of course, quite often wrong, mistaking noise for a signal or missing the signal in the noise. Yet over many trials each second it brings forth a very reliable signal from a very unreliable channel.

Nor is this technique a modern byproduct of the electronic age. The moon should cause tides of the atmosphere just as it causes tides of the seas, i.e., a waxing and waning of air pressure; but be-cause of the relatively small mass of the atmosphere and the vastly greater changes in pressure caused by the "weather," it cannot be measured directly. Long before the advent of computers, the effect was demonstrated by just such signal detection methods, with the signal (lunar variation of pressure) 60 decibels below the noise (other variations)--i.e., the noise one million times the power of the signal.

But to get closer to public policy, how do you optimize the main-tenance policy of equipment, structures, roads, weapons? Is it bet-ter to repair only after failure (yes, sometimes), and if not, at what intervals should one preventively replace which parts?

Suppose you have a limited number of bombers to search vast areas of sea for submarines; what is the optimum schedule and route of the patrols? This problem was solved by the British in WWII and started a branch of probability theory known as Opera-tions Research.