Nothing touches off panic in taxpayers more than the prospect of an audit. But there are such wide disparities across the country in who gets audited that residents in New York City and its suburbs continue to be treated with kid gloves by the Internal Revenue Service, while those in California have been hit comparatively hard.

Part of the reason for the difference is that the IRS has been looking harder in California for people who do not file tax returns, but it is also because the agency relies on a decade-old study of taxpayer behavior that critics call obsolete.Newly released IRS data show that Los Angeles County taxpayers are four times as likely to be audited as those in New Jersey, even though the two areas have roughly the same population and the same average income.

The new tax information, obtained by researchers at Syracuse University, discloses another trend that is likely to add to popular disenchantment with the IRS: Back taxes and penalties are increasing for the poor and declining for the rich.

This shift is largely a result of a 1995 directive from Congress for the IRS to monitor the tax returns of the working poor more thoroughly - not only to collect unpaid taxes but also to make sure people eligible for the earned-income credit, a form of negative income tax, do not get more than they are supposed to.

This year, 174 million Americans must file an estimated 124 million individual and joint tax returns before midnight on Wednesday. About one in 150 will be audited.

The agency bases decisions on which returns to audit on several factors, including suspicions about the integrity of the preparer or information from a related return.

But it also gives huge weight to statistical departures from norms established in a 1988 study - a study that critics say is outdated.

The 1988 research was conducted by picking thousands of taxpayers at random and making them prove each entry on their returns. Using that data, the IRS was able to determine how much people in various income groups allocated to tax-deductible costs. It then devised a statistical formula and focused on taxpayers who deviated the most from the norms.

In interviews, a former IRS commissioner criticized the continued use of the old study. "The taxpayer-compliance measurement data are obsolete," said Donald C. Alexander, a tax partner at Aiken & Gump in Washington and the tax commissioner in the Nixon administration.

In 1995, the IRS planned to update the profiles by putting 93,000 individuals and 63,000 companies through similarly intense audits, but the Republicans eliminated money for the project.