The raids on homes around the country were billed as carefully planned hunts for dangerous immigrant fugitives, and given catchy names like Operation Return to Sender.
And they garnered bigger increases in money and staff from Congress than any other program run by Immigration and Customs Enforcement, even as complaints grew that teams of armed agents were entering homes indiscriminately.
But in fact, beginning in 2006, the program was no longer what was being advertised. Federal immigration officials had repeatedly told Congress that among more than half a million immigrants with outstanding deportation orders, they would concentrate on rounding up the most threatening - criminals and terrorism suspects.
Instead, newly available documents show, the agency changed the rules, and the program increasingly went after easier targets. A vast majority of those arrested had no criminal record, and many had no deportation orders against them, either.
Internal directives by immigration officials in 2006 raised arrest quotas for each team in the National Fugitive Operations Program, eliminated a requirement that 75 percent of those arrested be criminals, and then allowed the teams to include nonfugitives in their count.
In the next year, fugitives with criminal records dropped to 9 percent of those arrested, and nonfugitives picked up by chance - without a deportation order - rose to 40 percent. Many were sent to detention centers far from their homes, and deported.
The impact of the internal directives, obtained by a professor and students at Benjamin N. Cardozo School of Law through a Freedom of Information lawsuit and shared with The New York Times, shows the power of administrative memos to significantly alter immigration enforcement policy without any legislative change.
The memos also help explain the pattern of arrests documented in a report, criticizing the fugitive operations program, to be released on Wednesday by the Migration Policy Institute, a nonpartisan research organization in Washington.
Analyzing more than five years of arrest data supplied to the institute last year by Julie Myers, who was then chief of Immigration and Customs Enforcement, the report found that over all, as the program spent a total of $625 million, nearly three-quarters of the 96,000 people it apprehended had no criminal convictions.
Without consulting Congress, the report concluded, the program shifted to picking up "the easiest targets, not the most dangerous fugitives."
It noted, however, that the most recent figures available indicate an increase in arrests of those with a criminal background last year, though it was unclear whether that resulted from a policy change.
The increased public attention comes as the new secretary of Homeland Security, Janet Napolitano, has ordered a review of the fugitive teams operation, which was set up in 2002 to find and deport noncitizens with outstanding orders of deportation, then rapidly expanded after 2003 with the mission of focusing on the most dangerous criminals.
Peter L. Markowitz, who teaches immigration law at Cardozo and directs its immigration legal clinic, said the memos obtained in its lawsuit reflected the Bush administration's effort to appear tough on immigration enforcement during the unsuccessful push to pass comprehensive immigration legislation in 2006, and amid rising anger over illegal immigration.
"It looks like what happened here is that the law enforcement strategy was hijacked by the political agenda of the administration," he said.
Kelly A. Nantel, a spokeswoman for the immigration agency, defended the program. "For the first time in history, we continue to reduce the number of immigration fugitive cases," she said, noting that the number of noncitizens at large with outstanding deportation orders, which peaked at 634,000 in the 2007 fiscal year, is now down to about 554,000. "These results speak for themselves and they are consistent with Congress's mandate: locate and remove immigration absconders."
Ms. Nantel said the number of fugitives with criminal backgrounds arrested in the 2008 fiscal year rose to 5,652, or 16 percent of 34,000 arrests, and nonfugitives fell to 8,062, or 23 percent.
Many Americans have welcomed roundups of what the agency calls "ordinary status violators" - noncitizens who have no outstanding order of deportation, but are suspected of being in the country unlawfully, either because they overstayed a visa or entered without one.
But Michael Wishnie, one of the authors of the report, who teaches law at Yale, said that random arrests of low-level violators in residential raids not only raised a new set of legal and humanitarian issues, including allegations of entering private homes without warrants or consent and separating children from their caretakers, but was "dramatically different from how ICE has sold this program to Congress."
"If we just want to arrest undocumented people," he said, "we can do it much more cheaply."
Congressional financing for the fugitive operations program rose to $218 million in the 2008 fiscal year, from $9 million in 2003, as the number of seven-member teams multiplied to 104 from 8.
In Congressional briefings and public statements since 2003, agency officials have repeatedly said that given the vast number of immigrants with outstanding deportation orders, the program will focus its resources on the roughly 20 percent with a criminal background.
An Immigration and Customs Enforcement memo dated Jan. 22, 2004, underscored that commitment: "Effective immediately, no less than 75 percent of all fugitive operations targets will be those classified as criminal aliens" - noncitizens with a criminal record as well as an order of deportation. It added that "collateral apprehensions" - immigration violators encountered by chance during an operation - would not be counted in that percentage.
But on Jan. 31, 2006, a new memo changed the rules. The directive, from John P. Torres, acting director of the agency, raised each team's "target goal" to 1,000 a year, from 125.
And it removed the requirement that at least 75 percent of those sought out for arrest be criminals. Instead, it told the teams to prioritize cases according to the threat posed by the fugitive, with noncriminals in the lowest of five categories. And it repeated that "collateral apprehensions will not count" toward the 1,000 arrest quota.
But that standard, too, was dropped nine months later. A new memo from Mr. Torres said "nonfugitive arrests may now be included" to reach the required 1,000 arrests. On average, however, it said at least half of those arrested by each team should be fugitives. It also promised to "ensure the maximum availability of detention space for fugitive arrest operations."
One result was an increase in noncriminals held in immigration detention. Another, the Migration Policy Institute report concluded, was that the percentage of criminal fugitives arrested plummeted, to 9 percent in the year that ended Sept. 30, 2007, from 39 percent in the 2004 fiscal year.
That same year, 15,646, or 51 percent of those arrested, had an outstanding deportation order, but no criminal record, and 12,084, or 40 percent, were termed "ordinary status violators" who did not fit any of the program's priority categories.
The report said the program relied on a database riddled with errors, and that many deportation orders were issued without the subject in court, sometimes because of faulty addresses.
The looser rules were reflected in sweeps like one conducted in New Haven in June 2007. During the raid, lawyers at Yale's immigration law center said, agents who found no one home at an address specified in a deportation order simply knocked on other doors until one opened, pushed their way in, and arrested residents who acknowledged that they lacked legal status.
Of the 32 arrested and scattered to jails around New England, only 5 had outstanding deportation orders, and only 1 or 2 had criminal records.