We budget-makers, from David Stockman on down, are hoist with our own petard. Because of accounting conventions developed somewhere back in the mists of time, we have made everybody think we are cutting the non-defense part of the budget much more than is the case.

Looking at the raw numbers, David Broder, for example, has suggested that President Reagan could be accused of "abandonment of national responsibilities," and he cites a reduction of "non-defense and non-entitlement net spending by 25 percent in one year." James Schlesinger thinks we are underestimating the real size of the budget, using as a major piece of evidence the "egregious" example of Social Security.

Separate problems are involved, but this kind of thing is endemic in analyses of the budget. Perhaps it is our fault. Let's look at what might be called the Broder problem first.

When the Jones family takes a trip to the West and pays fees to enter the national parks, those fees are "offset" against the outlays of the National Park Service, making the Park Service budget seem lower than it really is. When the Forest Service sells timber, the same thing happens--gross outlays of the Forest Service are reduced by the proceeds of the sale. This kind of thing happens all over the budget--to the tune of about $20 billion in the 1983 budget.

But the matter is much bigger than that. When the government auctions offshore oil leases, the multibillion-dollar proceeds from the bids are not counted as receipts, showing up in the tax side of the budget. They are counted as "undistributed offsetting receipts" and simply subtracted from the outlay total in the budget. Prospective sales of surplus government lands will be counted the same way, as the budget now stands. These items add another $19 billion.

There are other things of the same type, such as interest and principal payments on past loans, that artificially depress the gross outlay total in the budget. Altogether, these ''offsets" come to more than $40 billion in fiscal year 1983. And the striking thing is that nearly all of this accounting subtraction, in analyses of the budget and in our own tables, counts against the non-defense, non-entitlement, non-interest part of the budget--the domestic "discretionary" part that covers everything from the FBI to education grants to the states.

When an adjustment is made for the understatement of actual outlays, the conclusions about the budget become quite different, as shown below for 1983:

Total non-defense $536.6 billion

Less entitlements 341.5

Leaving 195.1

Less net interest 96.4

Equals "discretionary" budget 98.7

Add back accounting offsets 41.2

Equals true discretionary outlays 139.9

Once the actual size of the discretionary part of the budget is perceived, the magnitude of the reductions takes on a different dimension. No one is denying that cuts are being proposed, severe ones in some cases. But the cuts are reductions from gross outlays. In the aggregate they must be measured against gross outlays to get a proper sense of their magnitude and to make a judgment on whether federal responsibilities are being "abandoned." As it happens, the estimated accounting offsets are much larger in 1983 than earlier, which further distorts the picture.

Consider the difference between 1981 and 1983, which covers the Reagan administration reductions. For the discretionary part of the budget the first glance would indicate the following:

1981 outlays $136.7 billion

1983 outlays 98.7 billion

Reduction 28 percent

But the actual picture, using true discretionary outlays, is this:

1981 outlays $163.5 billion

1983 outlays 139.9

Reduction 14 percent

The anomalous results of our accounting system can be illustrated by a single example. Suppose the government stopped all offshore oil leasing in 1983 and left the entire discretionary part of the budget exactly the same. The "cut" would then appear as only 18 percent instead of 28 percent. Of course the actual cuts would be unchanged.

In any case, the pain and the outcries are stemming from a reduction in the discretionary part of the budget that is less than 15 percent over two years and leaves the large sum of about $140 billion.

The Schlesinger problem is different. Here the culprit is the way we treat "proposed legislation," and the issue arises with Social Security.

As everyone knows and acknowledges, the main Social Security trust fund, the one that pays old age and survivors' benefits, is technically bankrupt. The solution to the problem, for the short run at least, is to have this fund borrow from the disability and Medicare trust funds so that the checks can go out on time and in the full amount due.

The administration and Congress agreed on that solution last year. But the legislation permitting the interfund borrowing expires at the end of 1982. There is no doubt that it, or an alternative means of keeping the checks flowing, will be passed this year, but it still has to be identified in the budget as proposed legislation.

The result is that more than $14 billion of Social Security outlays are listed under proposed legislation. Schlesinger and others, looking only at present law, concluded that we were drastically underestimating Social Security outlays in 1983, showing an increase of only $4 billion. Schlesinger used this as evidence to demonstrate that the budget as a whole greatly underestimates what spending will actually be. Actual spending may, indeed, turn out higher than we estimated, but not because of Social Security. In fact, the full Social Security outlays are counted in the budget totals.

That is enough accounting arcana for one day. Perhaps we should put a warning label on the budget:

CAUTION: These figures do not always mean what they seem to mean.