In government, too: you'll find illegal aliens in the darndest places.

NEARLY everyone professes to agree that the government should crack down on employers who hire illegal aliens. But it appears that the government needs, first, to crack down on itself. That's the story told in a footnote of an audit report quietly released a year ago by the inspector general of a federal agency that seemingly has nothing to do with immigration.

It indicates that the Bush administration itself has employed aliens not authorized to work in the United States. Year after year, first the Justice Department and then the Department of Homeland Security were given information that would have allowed them to pinpoint exactly who these unauthorized alien workers were and which federal agencies employed them. Yet, by the end of the period covered by the audit report (which ran through tax year 2003), the problem remained unsolved. (The inspector general has not yet analyzed the relevant data for tax years after 2003.)

Which agencies employed these aliens? The government won't say. Citing a provision in the tax code, the inspector general's office says it is prohibited from publicly identifying them, and so, for now, they stay "in the shadows."

The audit report is innocuously titled "Employers with the Most Wage Items in the Nonwork Alien File." It was published in June 2006 by the inspector general of the Social Security Administration (SSA), and is publicly available on SSA's website. The audit was designed to discover the 100 employers that had filed the largest number of W-2 forms in tax years 2001 through 2003 for employees who were using what SSA calls "nonwork" Social Security Numbers (SSNs). These are SSNs that SSA gives to aliens who are not entitled to work in the U.S. but who claim to need an SSN anyway.

Prior to 1996, SSA handed these numbers out like candy. According to a 1999 inspector general's report, an alien who had no right to work in the U.S. could nonetheless get an SSN "for a variety of reasons including tax, banking, school, insurance, driver's license, and government benefit purposes." After reforms made in 1996, an alien could still get a "nonwork" SSN in order to secure a driver's license or claim a government benefit to which he was entitled even though he was not entitled to work in the U.S. In 2003, SSA stopped giving out nonwork SSNs for driver's licenses.

But the damage was done. By 1998, there were already more than 7 million "nonwork" SSNs in circulation--a number exceeding the combined populations of Rhode Island, New Mexico, Wyoming, Montana, North Dakota, South Dakota, and Delaware. Not surprisingly, many aliens used these "nonwork" SSNs to work illegally in the U.S.

To deal with this problem, the Illegal Immigration Reform and Immigrant Responsibility Act of 1996 mandated that SSA create a list every year called the NWALIEN file. This list includes the names and addresses of every worker and every employer associated with a W-2 bearing a nonwork SSN. The law mandated that SSA provide this list annually to the Justice Department, which contained the Immigration and Naturalization Service (INS). In 2003, when the Department of Homeland Security (DHS) took over the functions of INS, SSA started sending the list to DHS.

Because some aliens originally given nonwork SSNs eventually secure work visas or green cards, and because the DHS does not notify SSA when this happens, the raw NWALIEN file produced by SSA includes names of many people who are currently authorized to work in the U.S. All DHS would need to do to create a detailed road map for worksite enforcement against one type of immigration lawbreaking, however, is double check the NWALIEN list against its own work-authorization data.

As part of its audit, for example, the SSA inspector general's office checked a random sample of 275 "nonwork" W-2s against DHS data. Thirty-seven percent of the workers named on these W-2s, it turned out, had indeed secured authorization to work in the U.S. Sixty-one percent, however, had not. For another five people in the sample (2 percent), DHS could find no records at all.

Thirty-six of the 275 randomly sampled "nonwork" W-2s were for government workers. Sixteen of these 36 government workers (44 percent) had never been authorized to work in the U.S. Also, one of the five workers for whom DHS had no records was a government employee.

In fact, government itself, the inspector general determined, was the nation's most egregious filer of W-2s with nonwork SSNs. Seventeen of the top 100 employers for filing these questionable W-2s were federal, state, or local government agencies. From 2001 to 2003, these agencies collectively filed 49,380 W-2s with nonwork SSNs, reporting $1.5 billion in tax dollars paid to employees who might not be authorized to work in the U.S.

If, in keeping with the inspector general's random sample, 44 percent of the 49,380 "nonwork" W-2s filed by government agencies were filed on behalf of aliens not authorized to work here, that would mean these agencies filed more than 21,000 W-2s for unauthorized aliens in a three-year period.

A footnote in the report says: "Of the 17 employers in government, 7 were Federal agencies, 7 were State agencies and 3 were local government entities." But the report names no names. I asked the inspector general's office what they were. "I'm afraid we cannot provide that information," the office told me, saying Section 6103 of the Internal Revenue Code prohibits it from doing so. "What, if anything, did DHS do about the government agencies that were filing W-2s with nonwork SSNs?" I asked. "The Department of Homeland Security would be the appropriate party to address that question," answered the IG's office.

I sent DHS spokesman Russ Knocke a copy of the audit report and asked him to name names for me and tell me what DHS had done "to deal with these government agencies filing W-2s with 'nonwork' SSNs." The totality of his e-mailed answer was: "A non-work SSN might not necessarily mean that the individual is unauthorized to work. The SSA IG's report itself questions the adequacy of the records. Your first question should be directed to the IRS."

The IG's audit report raised a national-security issue. "Non-citizens who work without DHS authorization," it said, "could affect homeland security because they may obtain employment in sensitive areas." Leaving aside the government agencies, 13 of the employers ranked among the top 100 filers of W-2s with nonwork SSNs were in the "technology" sector. Another 15 were universities. One transportation company filed 3,635 of these W-2s. Was it an airline? The report doesn't say.

The report does say, however, that the People's Republic of China ranked fourth and Iran sixth among the top ten countries of birth for employees using nonwork SSNs to work for the 100 employers audited. These 100 employers filed 6,611 W-2s with nonwork SSNs on behalf of PRC nationals, and 4,805 on behalf of Iranians. The report does not indicate if any of these Iranians or Chinese worked for the government.

Another SSA inspector general's report published in September 2005, however, does name a government agency and does mention PRC and Iranian nationals working for the government on nonwork SSNs. It was the Defense Department. In this report, which covered tax years 1999 through 2003, SSA auditors said "we found that 5,192 DOD employees were working under SSNs that were originally issued as nonwork SSNs. For these 5,192 employees, DOD components submitted 19,777 W-2s totaling approximately $573 million in wages." The report also said 179 of these DOD employees were born in the People's Republic of China and 139 in Iran. The auditors did not check DHS data, however, to see how many of the employees using nonwork SSNs had been authorized to work in the U.S. sometime after they had been assigned a nonwork SSN. It is thus theoretically possible--if unlikely--that all of them were legal workers.

We've learned over the years that it's too much to expect the government to enforce immigration laws. But how about abiding by them?

Mr. Jeffrey is editor-at-large of Human Events.