A new federal law is letting thousands of federal convicts out of prison early, and no one is tracking who these felons are or what they do as they return to the streets. A belated April Fool's joke? No, it's true, and it's no joke.

It's all because of a radical policy change making federal inmates convicted of crack charges eligible for reduced prison sentences. Congress allowed the new law to go into effect last month, including a provision that made the reduced guidelines retroactive.

Since 1986, those convicted of crack-related offenses have received substantially more jail time than those convicted of powder cocaine offenses. Until last year, the "disparity" was 100 to 1: Five grams of crack triggered a five-year minimum sentence, versus 50 grams of cocaine.

The disparity shrunk significantly last year, after the U.S. Sentencing Commission revised its sentencing guidelines to increase the amount of crack needed to trigger minimum sentences.

Reducing the disparity wasn't especially controversial. A broad consensus had been building for years that the penalties for crack, compared to those for cocaine, were excessive and unjust.

But what was controversial was when the Commission took the unusual step of making the new, lower sentences available to prisoners who had already been convicted and sentenced. The change went into effect on March 3. Since then, 3,107 convicts have received shorter sentences.

As any fan of "Law & Order" knows, prosecutors exercise great discretion in developing cases. Every prosecutor knows which crimes are easy to prove and which require more resources. That analysis factors into the course of most prosecutions, from the investigation and indictment to bargaining with the defendant and sentencing.

An example: Proving distribution or possession of crack is relatively easy, so when it won't affect (increase) the overall sentence, a prosecutor may pursue the crack-related charge but not weapons possession or even assault charges. After all, proving the extra charges may take more resources but won't make the public any safer by putting the perpetrator away for a longer period of time.

The Sentencing Commission's retroactivity policy will produce unfair results. For example, a prisoner who received a 10-year sentence as part of a plea agreement for a crack offense - but who wasn't charged with a related weapons offense - may get a few years knocked off of his sentence under the new guidelines. Meanwhile, a prisoner with a 10-year sentence for the same weapons offense alone gets nothing.

Furthermore, as Attorney General Michael Mukasey warned Congress earlier this year, some career and violent offenders will get out of jail early. Newspapers such as The New York Times roundly criticized Mukasey's warning, accusing him of trying to scare the public. But there is good reason to believe that violent and repeat felons are being released early.

The Sentencing Commission says its policy changes don't apply to violent and career criminals. But the federal public defender service disagrees, arguing that these criminals should be treated exactly the same as first-time, nonviolent offenders under the new rules. The service has already issued a thoughtful memo to its lawyers explaining ways to exploit possible loopholes in the new guidelines. Their arguments are sound, and no doubt are influencing federal judges who decide who gets out early.

So, who is getting released early? Are federal judges following the Sentencing Commissions' policy guidelines? Are judges releasing armed career criminals because the law, as written, and Supreme Court precedent demand that result? How many of those just released have committed new crimes? No one is keeping track.

If crime rates rise, especially in inner-city neighborhoods, policymakers will need to know if retroactivity is to blame. Without good statistics, there's just no way to answer that question.

And it's inevitable that the issue of retroactivity will come up again. Without the facts, policymakers won't have any firm basis to decide whether retroactivity makes sense or puts the public at risk.

This is a job for the Department of Justice, which needs to make collecting and analyzing this data a priority, and official policy. Congress should ask the department for periodic reports, and those facts should be made public.

Reducing the prison terms of individuals already convicted and sentenced was an unusual move, one with uncertain consequences. Did Congress make a wise policy choice, or gamble with our safety? There is no way to know unless our government keeps track of the facts.

Charles D. Stimson, a former prosecutor and defense attorney, is Senior Legal Fellow at The Heritage Foundation (www.heritage.org), where Andrew M. Grossman is a Senior Legal Policy Analyst.