CHOOSING ELITES. By Robert Klitgaard. Basic Books. 267 pp., $19.95.

THIS BOOK is in significant ways an instructive primer in the hopes and disappointments of "public policy science," of which Robert Klitgaard is a professor at Harvard.

Typically, studies in this fledgling discipline promise that "policy choices" may somehow be so sterilized and systematized, indeed scientized, as to cleanse them of all taint of ethical or social predisposition. But in this respect and others, public policy studies frequently promise far more than they can deliver. And Choosing Elites is an example of that.

The book is a study, an able and candid one, of the way in which "elite" educational institutions -- that is, highly selective ones -- go about choosing and admitting their students. The admissions dilemma at such institutions is what is called, in the jargon of statistics, "selection at the right tail" -- selection from among a superabundance of candidates whose formal qualifications put them as a group way out at the extreme right end ("tail") of the bell-shaped curve showing the standard distribution of scholastic abilities.

Qualifications in such a pool of applicants are often so closely matched, or so much a matter of apples versus oranges, that it becomes a tricky task to say why one should be chosen, another rejected. Klitgaard draws heavily on the accumulated admissions experience of the various branches of Harvard. But despite (or perhaps because of) all the handwringing over selectivity, there is a wild variation from, say, law school to medical school to undergraduate college in the degree to which university subdivisions can articulate exactly what they seek in students, or what they do systematically to find it.

The firm conclusios reached in Choosing Elites are so few, or tentative, or unsurprising, as to tempt a good public policy scientist to throw up his hands and turn in his card.

There are indeed reliable techniques, mostly old and familiar, for predicting strictly academic achievement -- test scores and previous academic performance. Otherwise, it is difficult, nay impossible, to predict with accuracy the intangible social benefit ("social value added," to use Klitgaard's term) of any given attempt at social engineering in admissions policies. There is abundant evidence that diluting your meritocracy of "wonks" with good musicians, good mixers, youth of precocious social conscience, et al., tends to prevent boring and listless student bodies. But variety usually prevents dullness, whatever the setting. Nothing startling there.

The most interesting, but also the gloomiest, subject Kliting Among Groups," a delicate term for affirmative action admissions. It is, of course, an ethical and political minefield. Blacks, and to some degree Hispanics, perform far less well as groups on the usual standard cognitive tests, such as the SAT. Yet despite the clamor of recent years to the effect that standardized tests are "culturally biased," and hence put underprivileged minorities at a disadvantage, all the information Klitgaard has amassed tends to show that test scores "overpredict" the future academic performance of blacks, and substantially so.

It is a gloomy but inescapable inference from this that to gauge probable black performance in an "elite" college or university, an admissions office should in the typical case adjust SAT scores downward. No one quite knows what the implications of this are, but it is one of the many problems of selectivity which Klitgaard candidly admits -- candor on a subject so often suffused with intellectual sham and hypocrisy being a refreshing strength of this book. Most selective institutions are already practicing affirmative action, substantially relaxing their standards to recruit black and other minorities. But guarding its beneficiaries against discouragement continues to be a problem.

THE BASICALLY modest and skeptical findings of Choosing Elites suggest that in this, as in so many other fields, we are still trying to reinvent wheels which were discarded amid the turbulence of the '60s and early '70s. In that time of angst and self-consciousness over social privilege, some good and useful programs and principles were thrown out with the bad. Admissions offices at "elite" colleges and universities seem, even now, to be spending a great deal of time searching for magical formulas for recruiting the left-out and left-behind and assuring their success -- a laudable goal, to be sure, but one that has yet to foster any reliable body of predictive techniques.

The only consistently reliable "predictor" of academic success that admissions officers have to work with (apart from prior performance) is -- surprise! -- measurable cognitive ability, which is to say scholastic ability.

As Klitgaard is at pains to say, this is no excuse for abandoning the search for ways to make college selections (within sane limits) more democratic, more inclusive of human, ethnic and regional variety, and socially productive in the long run. Unfortunately, the central message of Choosing Elites is that we still know almost nothing about how to reach goals that are extraneous to the process of learning itself. Schools being schools, if you want to predict who will do well keep your eye on brainpower and intellectual self-discipline. Beyond that, the effort to find "scientific" predictors of social benefit seems to have been largely in vain. The mountain has labored mightily -- its wheezings and heavings are heavily documented in Klitgaard's footnotes and appendices. But the most it can show, so far anyway, is a pitifully underweight mouse.