The injustice of white-collar sentencing rules.

Even those who support tough penalties for white-collar fraud were startled in March when Jamie Olis, a former midlevel tax expert at the energy firm Dynegy, was sentenced to more than 24 years in federal prison for his role in a relatively minor financial scandal. In a world where Enron's Andrew Fastow will serve only 10 years and some of Fastow's colleagues might walk, Olis' punishment seems bizarrely harsh and unfair. One reason for this is a flawed method for calculating sentences for financial crimes, especially fraud at a public company.

Olis' saga began in April 2001, when he, his boss Gene Foster, and a colleague named Helen Sharkey used an accounting gimmick called a "special purpose entity" to temporarily increase Dynegy's operating cash flow. Pre-Enron, the rules regarding the legal use of special purpose entities were fuzzy, but, for arcane reasons, the way Foster, Olis, and Sharkey used this one was declared illegal. Dynegy eventually restated its results, and, in June 2003, federal prosecutors indicted Foster, Olis, and Sharkey for conspiracy, securities fraud, mail fraud, and three counts of wire fraud. Foster and Sharkey pleaded guilty to a single count of conspiracy and likely will be sentenced to no more than five years—the statutory maximum for a single conspiracy conviction. Olis maintained his innocence and went to trial. Last November, he was convicted on all six counts.

Olis' sentence was long for two reasons. First, unlike his colleagues, he took the risk of trying to defend himself. As is often the case, prosecutors had piled on charges, perhaps to pressure the defendants to plead. Olis' conviction on multiple counts raised the statutory maximum sentence that could be imposed. (For sentencing purposes, it was irrelevant that Olis presumably could not have committed securities fraud without also committing mail fraud and wire fraud—false financial information and accounting opinions aren't disseminated via ESP.) Second, Olis fell afoul of recently updated rules in the federal sentencing guidelines for financial crimes.

As with other federal crimes, fraud sentences are calculated using a complex formula set forth by the United States Sentencing Commission in a tome called the Guidelines Manual. The Manual calls for adjusting sentences based on the severity of the fraud, with severity determined largely by the magnitude of financial losses and the number of victims.

For a defendant with no criminal history, for example, the "base level" guideline for a fraud conviction is zero to six months in prison. Judges then have to consider approximately 20 possible adjustments to this sentence, including:

• The amount of the "loss" attributable to the fraud
• The number of victims of the fraud
• Whether the fraud involved "special skills" or "sophisticated means"
• Whether the defendant worked in the investment business or as an officer or director of a public company

Of these factors, by far the most influential is the estimated loss. Under guideline revisions adopted in 2001 and 2003 in response to public outrage about corporate scandals, if a fraud by an investment adviser or the officer or director of a public company is estimated to have caused losses of less than $5,000—the lowest category—there is no increase in the sentence. Above $5,000, the sentence increases, stepping up through 15 categories to losses of "more than $400 million." All else being equal, for example, if the loss estimate is $1 million to $2.5 million, the guideline sentence jumps to 63 to 78 months, or five to seven years. If the loss is more than $400 million, the guideline is 292 to 365 months, or 24 to 30 years. If a $400 million fraud also affected more than 250 victims, moreover, the guideline calls for a life sentence.

In theory, this sounds reasonable: Steal more money or cause more losses, and you do more time. In the case of securities fraud, however, it can produce outcomes that are so arbitrary and harsh as to be absurd.

Judges have discretion to determine estimated losses, but the guidelines advise them to base the estimates on factors that, where appropriate, include "the reduction that resulted from the offense in the value of equity securities"—in other words, the amount the crime caused a company's stock to fall. This is where the trouble starts.

First, it is pleasant to think that we know what makes stocks go up or down, but in most cases, we don't. Stock prices are the product of thousands of buy-and-sell decisions based on dozens of factors, including liquidity, interest rates, market direction, technical indicators, industry and company performance, and the relative attractiveness of other securities or asset classes. To conclude that a fraud "resulted in" such-and-such a loss by analyzing the stock price, therefore, is a delusion (although the folks who wrote the sentencing guidelines aren't the only ones who suffer from it).