Whenever I hear the phrase "studies prove" this or that, it makes me think back to the beginning of my career as an economist at the Labor Department in Washington.

Secretary of Labor Arthur Goldberg was scheduled to appear before Congress to argue in favor of some policy that the Labor Department wanted enacted into law. Down at the bottom of the chain of command, I was given four sets of census data that had not yet been published and was told to analyze these data for a report to go to the Secretary of Labor.

Two of these sets of data seemed to support the Labor Department's position but the other two went counter to it. When I wrote up a paper explaining why this was so and concluded that the statistics overall were inconclusive, there was much dismay among those in the hierarchy between me and the Secretary.

They were also puzzled as to why anyone would write up such a paper, knowing what the Department's position was on the issues. They took my paper, edited and rewrote it before passing it up the chain of command.

Secretary Goldberg then made his usual confident presentation of the rewritten study to Congress, probably unaware of the contradictory data that had been left out.

It was a valuable experience so early in my career to learn that what "studies prove" is often whatever those who did the studies wanted to prove. Labor Department studies "prove" whatever serves the interest of the Labor Department, just as Agriculture Department studies "prove" whatever serves the Department of Agriculture's interests.

It is the same story on the other side of the Atlantic, where a new book about Britain's criminal justice system exposes the fraudulent methods used to generate statistics about the "success" of various programs of alternatives to imprisonment. The book is titled "A Land Fit for Criminals" by David Fraser.

The numbers may be accurate but the definition of "success" makes them meaningless. When a criminal is put on probation and the probation is not revoked for a violation, that is "success."

Unfortunately, the British criminal justice system does not automatically revoke probation when a criminal commits a new crime.

A criminal on two years' probation can commit a crime after six months, be convicted and sentenced -- and, after serving his sentence, go back to completing the remaining 18 months of his probation, producing statistical "success" for the probation program. That is the whole point of the "study."

On either side of the Atlantic, it is a terminal case of naivete to put statistical studies under the control of the same government agencies whose policies are being studied.

Nor will it do any good to let those agencies farm out these studies to "independent" researchers in academia or think tanks because they will obviously farm them out to people whose track record virtually guarantees that they will reach the conclusions that the agency wants.

Climate expert Richard S. Lindzen of M.I.T. has indicated that the vast amount of government research money available for studies of "global warming" can discourage skeptics from being vocal about their skepticism.

This is not peculiar to studies of "global warming." Many people who complain about the corrupting influence of money never seem to apply that to government money.

If high government officials were serious about wanting to know the facts, they could set up an independent statistical agency, along the lines of the General Accounting Office, to do studies of the effects of the policies of the operating agencies.

That would mean that the fox would no longer be in charge of the hen house, whether the fox was the Labor Department, the Commerce Department, or any of the other departments and agencies.

It would also mean that various bright ideas originating in Congress or the White House would now be exposed to the risk of being shown to be costly failures or even counterproductive. Whole careers could be ruined among both elected officials and bureaucrats.

Don't hold your breath waiting for it to happen. But do keep that in mind when someone says "studies prove . . . "