To many veteran Congress-watchers, Sen. Russell B. Long (D-La.) is the closest the Capitol has to a legislative magician.

Those who have followed Long's machinations during a Senate debate on a tax bill have seen him make the most egregious-sounding special interest amendment seem like a well-intentioned plan to benefit the poor.

At other times, he's managed deftly to transform a giveaway laden tax bill into a relatively reasonable piece of legislation - after pulling it through the black box of a House-Senate conference committee.

And all without mirrors or even a magic wand.

But this year, Long has served notice he may try to pull off one of the neatest tricks in the annals of tax legislation - to provide voters with more tax relief than the House-passed bill, at less cost in "lost" revenues.

His secret: Change the way Congress computes the cost of a tax-cut proposal to count not only the loss of revenue that results from reducing tax liability, but also the economic feedback that a tax cut will produce.

Although the plan may seem arcane, it already has become a point of controversy among lawmakers and tax experts.

Liberals see the move as a ploy by Long to minimize the cost estimates for the widely criticized House-passed capital gains tax cut. (Long contends the capital gains tax cut would increase revenues, not lose money, if the feedback effects were counted.)

It also could have a major impact on the new congressional budget process, which sets ceilings each fiscal year on tax-reduction legislation as well as spending measures.

If Long's proposal for changing congressional revenue-estimating techniques is adopted, the capital gains portion of the tax bill would be counted as gaining a few million dollars, rather than losing $1.9 billion, as now is the case.

The result: Presto! More room in the budget for a bigger general tax cut - making everybody happy, from the more conservative, higher income taxpayers to the man in the street.

The argument Long is posing involves some technical issues:

In most tax-estimating cases, calculating the impact of a tax cut on revenues is a relatively simple matter. Tax planners just estimate how much the proposed change would reduce overall tax liability.

So the "cost" of a bill to exempt some forms of income from taxation night well be pegged at, say, a few hundred million dollars. That's the amount in revenues the Treasury would lose if the taxes now drawn from this income disappear.

But there also are other cases in which a tax change has an impact on buying habits or investment patterns, which in turn affects the amount of revenues the Treasury can expect to receive. tr for add 3.

For example, if a tax cut spurs industry to spend more on expanding its production facilities, it's apt to lead to increased business, and ultimately to bring in added revenues. This is what tax experts refer to as "feedback."

Or, revenue levels might be affected if a tax-law change prompts individuals to shift their investments or purchases from one sector of the economy to another. The same sort of impact is involved.

Admittedly. Long's contentions about including the feedback effects in tax-estimating techniques aren't without some merit.

The principle seems simple enough: If a tax change is apt to affect revenue levels, then why not count it along with other expected effects, whether they stem from reduced tax liabilities or any other factors?

Indeed, conservatives now are arguing that to ignore the economic feedback would be misleading. They point to the big Kennedy tax cut of 1964 as evidence. That bill generated so much new activity that revenues swelled.

Indeed, the whole idea of the GOP-sponsored Roth-Kemp tax cut bill is based on the assumption that the 30 percent across-the-board tax cut over three years eventually would generate enough new revenues to offset any deficit. Without that, economists say, the cuts would raise the deficit by $90 billion.

The difficulty is that tax estimators simply don't have enough expertise to compute the impact of tax changes on the over all economy with the same sort of precision that they use in figuring the effect on simple tax liability.

And in cases such as the capital gain tax proposal, opinion often is sharply divided. Estimates there range from the capital gains cut having little effect to it setting off a virtual boom.

Moreover, in some instances, government technicians already take account of some feedback effects - such as the impact of a major tax-cut proposal in the January budget.

But in that case, the feedback is computed by figuring the total impact of a big tax change with expected changes in the economic outlook and overall spending techniques.

The problem is in figuring feedback for smaller measures by themselves. It often is just impossible to compute the impact of a particular tax change by itself.

Those who favor keeping the present system - that is, continuing to measure the impact of a tax-cut proposal by its effect on tax liability - argue it may not always be adequate, but at least it is reliable and consistent.

To add to the confusion, both Treasury and Congress have been inconsistent on the question. The two sides historically have cited whichever set of figures supported the case they wanted to make politically.

For example, when Congress raised capital gains taxes a few years ago, the lawmakers claimed credit for bringing a new revenues - conveniently ignoring whatever losses may have resulted from reduced economic activity.

And when the Treasury computed the impact of the administration's proposed gas-guzzler tax, it included the feedback effects - despite officials' protests now over Long's insistence on doing it for capital gains taxes.

It still is too early to tell whether Long will be successful in persuading the Senate to count the feedback effects in estimating the cost of any tax changes.

But there is no doubt that the issue the Finance Committee chairman is raising won't go away. Many tax experts believe it is important enough to rate a serious further look next year - free from any end-of-session sleight of hand.