When the federal government offered to help thousands of poor people pay their rent a few years ago, half of those eligible took the money.

When it offered to give families living in substandard housing enough money to secure better units, only about a third did so.

These findings from one of the largest social experiments ever undertaken by the government were disclosed yesterday by Donna E. Shalala, an assistant secretary of the Department of Housing and Urban Development.

Many of the findings "challenge some of our most widely held assumptions" and have important implications for federal housing policy in the future, she told the National Association of Housing and Redevelopment Officials, which is meeting in Seattle.

The experimental housing allowance, which sent rent-supplement checks averaging $78 a month directly to families, was the Nixon administration's most important housing initiative. Some top HUD officials at the time hoped housing allowances would eventually take the place of subsidized housing construction or substantial rehabilitation" or neighborhood improvement.

They do lead to some improvement in the quality of a recipient's dwelling, she said. Another HUD source said quality - measured by higher rent and improved physical conditions - rose 20 percent.

HUD officials had expected greater participation in the experiment which began in phases in 1973 and 1974 and will end next summer, and they thought that most of the families who did take part would move more quickly, she said.

Instead, after extensive government efforts to reach families whose income was from $4,000 to $4,800, "about 60 percent of eligible renters and around a third of the eligible home-owners enrolled," said Shalala, who heads the policy development and research division.

Ray Struyk, deputy assistant secretary for research, said participation lagged because many of the poor, especially the elderly, did not want their incomes or housing conditions to be checked by government bureaucrats; some did not want "welfare," and many others did not want to move out of their neighborhoods, away from family and friends.

"The results also contradict sharply the assumption that, given the opportunity, all low-income families will flock to the suburbs," Shalala said. "No such pattern occurred.

"This finding also has important inplications for our racial integration goal. It suggests that housing allowances by themselves will not increase racial integration.

"Nor did families use their housing assistance to get luxury units," she added. "Some people believed they might have because there was no ceiling on the rents they could pay. . . As a matter of actual fact, the low-income people in the program behaved quite conservatively."

In one locality, Struyk said, 40 percent of those in the program were already in adequate housing and simply used the extra money to reduce their rent burden. The average poverty family spends 45 percent of its disposable income on housing, he said.

And, 30 percent of those in the program got their unit fixed up and stayed in it. Of the remaining 30 percent, who were in substandard units, 9 percent moved to a standard dwelling, 16 percent gave up trying to find one and 5 percent are still looking, Struyk said.

The housing allowance program differs from the government's current low-income housing program under Section 8 of the 1974 housing law in that the subsidies go to the tenant. They go to the landlord or developer under Section 8, which is designed to encourage construction as well as use of existing apartments for the poor.

Shalala said HUD had expected housing allowances to induce undue rent inflation but that did not happen.

Some 23,000 families have taken part in the $180 million experiment in Pittsburgh; Phoenix; Green Bay, Wis.; South Bend, Ind.; Springfield, Mass.; Durham, N.C.; Jacksonville, Fla.; Peoria, Ill.; Tulsa, Okla.; San Bernardino, Calif.; Salem, Ore., and four rural counties in North Dakota.