The Commerce Department and the FBI are currently investigating several dozen government employees to try to find out who leaked GNP statistics before their public release. The investigation, described by Laurie McGinley in The Wall Street Journal, has not come up yet with a culprit. But it does teach three useful lessons -- lessons from which, unfortunately, the Reagan administration is not likely to learn much.

The first is that lie detectors don't work well. Commerce seems to know this: it allowed tests only at the prompting of the FBI and has assured employees that they won't be penalized for refusing to take them. Commerce is right to try to track down the leak of its estimate for third-quarter economic growth, which was released Sept. 20; that informatin is worth a great deal to various traders and speculators, particularly now that there is futures trading in economic indicators. Unfortunately, the lie detector tests aren't likely to provide reliable information. The tests measure physical responses to stress; some liars can pass them, and some truth-tellers fail. If Commerce conducts its investigation as intelligently as it says it's doing, lie detectors are not going to be much help.

Which leads to lesson two: if you want to prevent leaks in the future, there are plenty of methods better than lie detectors. Sixty people at Commerce knew the Sept. 20 number in advance; for the corresponding release scheduled for Oct. 18 that number was reduced to 10. The interval between the calculation of the number and its release was reduced from 48 to 24 hours. Advance word on the number was sent by messenger to the Council of Economic Advisers rather than telephoned. There was no leak in October. A government that classifies too much material and gives access to rightly classified material to too many people is going to have too many leaks. A government that takes easy, sensible precautions will have far fewer.

The third lesson is that government statistics are valuable. Anyone who solicited the Sept. 20 leak (if anyone did) certainly thought so. One of the most useful things the government does is to compile reliable numbers that make it possible for people to make some sense of a large and complicated world. Over many years the federal government has built statistical agencies of great expertise and integrity. Sadly, Reagan administration budget-cutters, seemingly out of a dogmatic belief that government can do nothing well and shouldn't try, have gone after the statistical agencies and have done great damage, though the amounts of money that can be saved are exceedingly small and the amount of damage to society exceedingly large. The Commerce Department deserves credit for maintaining the reliability of the GNP figures and for investigating possible leaks. It's too bad it is using the mumbo-jumbo of lie detectors to perform this worthwhile task.