THE ADMINISTRATION's budget embarrassment is generating a lot of helpful suggestions that are bright, plausible and fundamentally wrong. Sen. Robert Dole, chairman of the Finance Committee, observes that present law would cut income tax rates in July, cut them again in 1983 and then index them to inflation. Why not, he proposes, cancel the 1983 cut and instead proceed immediately to index them?

Here's a better idea: why not cancel both the 1983 cut and the whole scheme for indexation? Of the two, indexation is by far the more dangerous. It will undercut any attempt to restrain inflation.

The purpose of indexation was to try to make the tax system inflation-proof. The theory was that if all the tax brackets were lifted under an automatic formula at exactly the same rate as the cost of living, people would not be bumped into higher brackets by pay raises that only kept them even with inflation. The concept had a special appeal to fiscal conservatives, who argued that as long as inflation kept increasing tax revenues that way, Congress would only spend them.

Those are all nice thoughts. It's a pity that none of them is right. For example, Congress has never regarded a lack of revenues as a fatal constraint on spending. Otherwise the country would not have had all those deficits over the years.

More important, indexation is not so simple as it looks--a lesson that you might think Congress had learned from its enormous errors and misjudgments in indexing the Social Security benefits in the 1970s. The present financial strains in the Social Security system, and the prospect that the benefits will outrun the system's income within the next year, can be wholly attributed to the mistakes in indexation. Now Congress has applied the same flawed principle on an even larger scale to the tax system.

Congress chose to tie the tax brackets to the consumer price index. Unfortunately, the CPI tends to overstate inflation. It rose 13.3 percent in 1979 compared with a rise of 8.7 percent in the GNP chain price index, an alternative measure that most statisticians consider more accurate. In 1980, the CPI rose 12.4 percent, while the chain price index rose 8.6 percent. Those are not trivial differences. As more is linked to the CPI, any attempt to correct its eccentricities will be fought by the people who benefit from them.

Suppose that another inflationary shock--another oil crisis, for example, or another bad harvest--hit an American economy with indexed federal benefits and indexed tax brackets. In response, federal benefits would automatically rise while taxes would automatically be cut. Because of the peculiarities of the CPI, both the benefit increase and the tax decrease would probably be greater than the actual inflation rate.

Benefits up, tax revenues down. Does that sound like a way to balance the budget? Does it sound like a way to end inflation?