In an eerie echo of budget director David A. Stockman's infamous line--"None of us really understands what's going on with all these numbers"--several experts in the government's war on paperwork yesterday said the measurements they use to chart their progress are imprecise at best and guesswork at worst.

But, they said, the current estimates on the millions of hours it takes businesses, state and local governments and ordinary people to fill out federal forms are the best measurements available, imperfect as they are.

Under the Paperwork Reduction Act of 1980, the number of hours spent completing federal forms was to be reduced by 15 percent in fiscal 1982, and 10 percent this fiscal year. That means counting those hours.

So yesterday the Treasury Department gathered what few experts there are on the topic for a two-day "Paperwork Burden Estimating Conference" to discuss and improve the process. Treasury, particularly its Internal Revenue Service, accounts for a large chunk of the government's paperwork burden, including the two forms considered the most burdensome, the Form 1040 income tax return and the W2, the wage and tax statement.

"You have vastly inflated estimates of burden and when you reduce it you're reducing nothing but numbers on paper," said Arnold Jones, associate director of the General Accounting Office's General Government Division. But, he added, the agencies' existing definitions of "paperwork burden" "may be . . . the best you can do in the real world."

The Treasury Department, for instance, computes the burden of one form by sending it to nine companies of different sizes, and then asking them how long it took to complete it. It then multiplies the figure it gets by the number of companies which have to file the form.

Earlier, Arnold Strasser, assistant chief of the Office of Management and Budget's Reports and Management Branch, told the bureaucrats, "Paperwork estimates are an attempt at determining an average burden. There are extreme variations around the middle. If we estimate it's going to take half an hour to do something, it may take five minutes for someone and for someone else it may take five days."

Strasser added, "As long as the methodology we use is constant, the measures of change. . . are . . . more accurate than the aggregate statistic we worked with in the first place." Or, we may not know where we've been or where we're going, but we know how fast we're moving.