One of these guys has got to be wrong.

Here's George Bush addressing the Republican National Convention:"We've created 17 million new jobs the past five years . . . and they're good jobs. The majority of them created in the past six years paid an average - average - of more than $22,000 a year."

Here's Michael Dukakis addressing the AFL-CIO board:

"Now between 1979 and 1985 . . . the average weekly wage of the jobs we lost in this country was $440. What replaced them? On average, jobs paying one-third less."

Which candidate is right? Neither. An investigation of the reasoning behind their assertions shows that both men are playing fast and loose with statistics. In fact, precise figures don't even exist on the pay of individual workers who gained and lost jobs.

So it goes in the great American jobs debate, where Republicans try to make it sound as if most of the people who obtained employment during the Reagan era found lucrative positions in the high-tech industry, and Democrats suggest that the more typical cases involved skilled factory workers forced to accept work in fast-food joints.

The truth, of course, lies in between. There is some evidence that the proportion of jobs paying low wages has risen, although the data are not nearly sufficient to justify a picture of the United States' turning into a nation of hamburger-flippers. There also is evidence that better-paying jobs are on the increase.

But whatever the truth, confusion surrounding the issue appears certain to grow in the weeks ahead; the debate, which had been simmering for months, has recently begun heating up in earnest. The presidential election campaign is focusing fresh attention on the issue, and economists on both sides are presenting a host of new studies to bolster their arguments.

The political and economic stakes in the debate are substantial. There is little disagreement that the American economy has been generating jobs at an impressive clip, but much is hanging on the question of whether the quality of the new jobs indicates that the country is headed on the right track.

If the public is persuaded that new jobs are concentrated primarily in minimum-wage, dead-end occupations, then not only will the Democrats be able to achieve an important public-relations victory, but policies favored by left-wing economists - such as national industrial planning aimed at preserving high-paying union jobs - will gain momentum.

If, on the other hand, the public becomes convinced that many of the new jobs offer attractive opportunities, then not only will Republicans win the political advantage, but so will the free-market-oriented policies favored by conservative economists.

As a result, some economists worry that participants in the debate are engaging in misleading propaganda in order to advance their causes.

"You can pick numbers that will prove anything in this area," said Gary Burtless, a senior fellow at the Brookings Institution. "A lot depends, for example, on whether you pick a recession year, or an expansion year, to start your calculations. There's the old expression, `Figures don't lie, but liars sure can figure.' I don't want to say that people here are liars, but I think different people are trying to make different partisan points."

Consider, for example, Bush's statement at the GOP convention that the majority of new jobs pay an average of $22,000 a year.

Behind the vice president's claim was a set of figures from the Bureau of Labor Statistics showing which occupations have grown during the 69 months since the end of the 1981-82 recession.

The BLS figures show that 16.8 million more jobs were created than were lost during the period. More than half of these were in occupations - such as managerial-professional, and craft and repair - where the average pay is $22,000 or more.

But these statistics don't really tell what has happened to the people who got those new jobs during the period. The fact that growth occurred in a high-paying occupation doesn't mean that all the people who found new jobs in that occupation received the average salary or higher. Indeed, common sense suggests that most new-job recipients were paid considerably less than the average for their occupation.

"Saying what the median annual income of an occupation is is not the same as saying what a person entering into it makes," observed one BLS official who asked not to be identified.

In fact, this official added, because statistics on the individuals who gained and lost jobs don't exist, all economists can do is look at "net" numbers, such as

he figure of 16.8 million new jobs - derived by looking at the differences shown in a series of statistical snapshots of the work force over time.

Those figures conceal the vast amount of job loss and job creation going on in the economy. They don't show, for instance, how many steelworkers became hamburger-flippers.

The net figures are useful for comparing this year's work force with previous years' work forces. The figures show that more jobs were created than were lost.

They can show that the people working today hold more jobs in certain occupations than before, and they can show that workers hold more jobs at certain pay levels than before. But nobody can be sure what such data reveal about actual job-getters and job losers.

Dukakis, for his part, is guilty of the same sort of statistical fallacy that Bush engaged in - although the Massachusetts governor drew the opposite conclusion.

Dukakis's statement that newly created jobs pay only two-thirds the wages of lost jobs was based on figures compiled from BLS data by Lawrence Mishel, director of research at the Economic Policy Institute, a labor-backed think tank.

These figures show that in the years since 1979, industries with relatively high average wages shrank and industries with relatively low average wages expanded.

Again, the figures, while informative, don't prove anything as precise as what Dukakis said. For example, the fact that jobs were lost in the relatively high-paying manufacturing sector doesn't mean that all of those jobs paid well.

As Janet L. Norwood, commissioner of labor statistics, has pointed out, some of those jobs were in industries such as leather, textiles and apparel, which tend to pay poorly.

By the same token, the fact that jobs grew in the service sector doesn't mean that all of those jobs were bad ones; many service industries, such as health services, tend to pay well.

Where, then, does the truth lie? A look at the data shows there is some evidence that bolsters the Democratic argument, even if it doesn't prove Dukakis's arithmetic.

At least by some measures, the proportion of jobs providing relatively low wages appears to have grown, notably among workers who work year-round and full-time.

Meanwhile, wages for the entire work force have failed to rise above inflation during the 1980s; by some measures, real wages have declined. Compensation per hour, adjusted for inflation - a good yardstick for measuring pay, because it includes fringe benefits - has only recently returned to the 1978 level.

And industries such as steel and autos, which pay well almost uniformly, have been among the biggest job losers, while industries such as retailing and restaurants - which almost uniformly offer low pay - have been among the biggest gainers, BLS data show.

But it is easy to exaggerate the conclusions that can be drawn from these facts. Even if an increasing proportion of workers has been receiving low wages, that doesn't necessarily mean most new workers are getting "bad" jobs, economists say. Many low-paid workers during the periods studied were presumably baby boomers who may have been starting promising careers.

Thus the Republicans can legitimately be optimistic over the quality of new jobs: Occupations such as managers and mechanics are growing, along with industries such as health services (1.4 million additional jobs during the current expansion) and business services (2.2 million additional jobs). The latter category includes computer and data processing, legal services and advertising.