uly was a poor month for job creation in the United States.

July was an excellent month for job creation in the United States.

That tale of two employment reports is true, and it continues a trend that has persisted for two and a half years. The discrepancies have made it possible for Republicans to herald a job recovery and for Democrats to deny one exists.

Both sets of statistics were issued by the government's Bureau of Labor Statistics, but they come from very different surveys. One, the establishment survey, which questions 160,000 employers, paints the bleak picture. The other, the household survey, which questions 60,000 people about whether they or other family members are working, paints the better picture.

Which is right? Because of its smaller sample size, the household survey is always more volatile, and month-to-month changes can be deceptive for that reason. So economists normally pay more attention to the establishment survey. But the fact that they differ so drastically may mean that reality lies somewhere in between.

Over all, the household survey now shows that employment has risen by 1.9 million jobs, or 1.4 percent, since President Bush took office, while the establishment survey shows employment is down by 1.1 million jobs, or 0.8 percent.

The establishment survey concluded that July was a poor month, with a seasonally adjusted job growth of just 32,000 jobs, far below what economists were expecting.

On a nonseasonally adjusted basis, the performance was even worse. It showed that there were 1.2 million fewer people working in the United States in July than in June. The adjustment in part reflects seasonal workers who are not paid for the summer, like some school employees.

The household survey, on a seasonally adjusted basis, showed a gain of 629,000 jobs in July. Before seasonal adjustment, the gain was an even larger 839,000 jobs. That may partly reflect the fact there are more agricultural jobs in the summer - which are included in the household survey but not in the establishment one - or that some workers who have the summer off would normally say they had jobs although they had not worked in July.

The household survey is used to calculate the unemployment rate, which fell to 5.5 percent in July, the lowest figure since October 2001.

The two surveys are not intended to produce the same results. The household survey includes the self-employed as well as agricultural workers, who would not be counted in the establishment survey, and the establishment survey counts each job so that a worker with two jobs could be counted twice.

But over time the two surveys have been roughly similar, although with an interesting political difference. Republican administrations tend to produce better household numbers than establishment ones - perhaps reflecting a better environment for the self-employed. Democratic administrations tend to show better establishment figures.

Early in the Bush administration, which took office just as the economy was sliding into a recession after a period of prolonged growth that saw unemployment sink to historically low levels, both surveys showed a shrinking work force. But they began to diverge in early 2002, with the household survey finding steady job gains, while the establishment survey continued to show a shrinking number of people with jobs.

By the time the establishment survey hit bottom, in August 2003, it indicated that 705,000 jobs had vanished since January 2002, bringing the total job loss in the Bush administration to 2.6 million. But the household survey found that nearly 2 million jobs had been added since January 2002, almost reversing the job losses earlier in the Bush administration.

Since then, the two surveys have fluctuated, but the establishment survey has continued to paint a bleaker picture than the household one.

Going back to Harry Truman, Mr. Bush - by either measure - has presided over an economy that has produced the poorest job creation record of any occupant of the White House to this point in a presidential term. The second worst was turned in by President Bush's father, who failed to win a second term in 1992.

But good job records also do not guarantee electoral success. The two best terms by that measure ended in 1968, when Lyndon Johnson was president, and in 1980, when Jimmy Carter was in the White House. In each year the Democrats were thrown out of the White House by dissatisfied voters.

A Democratic partisan would note that six of the seven best presidential terms, as measured by the establishment survey through the first three and a half years, have been under Democratic presidents, with the only Republican to make that list being Ronald Reagan in his second term, when jobs grew by 9.5 percent. The seven worst by that measure were all Republican administrations.

On average, the establishment survey has shown a 10.1 percent rise under Democrats and a 4 percent gain under Republicans. The household survey has shown a less marked preference for Democrats, whose edge is 6.7 percent to 4.8 percent.