Thursday, October 4, 2012

Dynamic Tax Scoring

The Tax Foundation study, "Simulating the Effects of Romney's Tax Plan" is worth reading and thinking about, especially in contrast to the standard static analysis that I complained about at the CBO.

Gov. Romney has proposed, at heart, a reduction in marginal rates, together with tightening of deductions. He hopes to make the latter large enough so that the program is revenue neutral, or at least deficit neutral when some spending cuts are included, and as close to neutral across the income distribution as possible.

Unlike a Keynesian plan, whose purpose is to transfer wealth to the hands of people (voters) likely to "consume" it, or a redistributionist plan, whose purpose is to transfer wealth from one category to another of people, the point of a revenue-neutral, income-neutral tax reform is to permanently and predictably lower marginal rates, giving rise to incentives to work, save, invest, and increase economic growth over the long run.

What possible sense does it make, then, to evaluate such a plan by assuming off the bat that it has no effect at all on output, employment, investment and so forth? Yet that is precisely what the standard "static" scoring does!  We build a rocket ship to go to the moon, and we evaluate its cost effectiveness by assuming that it never leaves the launch pad?

There are all sorts of things to appluad in this approach.
we have simulated the effects using a model built on a standard neo-classical growth model
Hooray! The "standard neoclassical growth model" is exactly the right building block for this sort of exercise, one that has pretty much taken over all academic tax analysis for the last 20 - 30 years, but has been virtually absent in Washington, still using Keynesian macro models from the 1960s. What does it mean? In general, we model households making decisions between work and leisure, consumption and savings; we model businesses making investment, hiring, and output decisions to maximize profits, and find the equilibrium.

The general consensus, even from (sensible) Keynesians, is that this is the right sort of model to use for long run -- several years -- analysis. It's the benchmark model in which margins matter, in which lowering tax rates, while getting rid of deductions so you pay the same taxes, can possibly have an effect. The Keynesian models, which I criticized here also pay no attention to margins, and so assume away the effects that a reform focused on lowering marginal rates is trying to achieve.
This model produces a simulation of what the policy change would do to the economy, incomes, and tax revenues after all economic adjustments are given time to work, which is roughly 5 to 10 years. It does not show the annual progression, year by year, from the starting point to the final outcome, but most of the effects occur within 5 years.
More hooray. Nobody knows the exact adjustment path. The point of tax policy is to get things right for the long run, not to try to manage the year to year. So don't even try to produce numbers that we all know are meaningless.

What are the effects?
The Romney plan would raise actual and potential GDP by about 7.4 percent over a five to ten year adjustment period.
Here I actually think the model is being conservative. It seems the model is removing labor and capital distortions, but assumes no effect of tax rates on growth;  the rate of technical change is given. I suspect that lowering marginal tax rates also makes people work harder at inventions. If that's right, then there is a "growth effect" not just a "level effect." Yes, we don't know much about how large it is. But I submit that we know the sign!
...relative to the static revenue loss, the biggest bang for the buck comes from the capital gains and dividend relief, followed by the corporate rate reduction and the elimination of the estate tax
Here we see the standard conclusion from the optimal-tax literature, that taxing investment is particularly distorting.

I don't think the model has an evasion margin (it should) -- that as you reduce rates people find it less worthwhile to get their lawyers and lobbyists to find ways around them. If it did, that would increase the argument against the estate tax. There is nothing like confiscating half your wealth every generation to get you to go visit the tax lawyer.

Of course, the question of how large the dynamic effect is goes right to the heart of whether the plan is, in fact "revenue neutral," and just how many "base broadeners" are required. This analysis does not say the tax cuts pay for themselves,
The Romney tax plan would recover nearly 60 percent of the static projected revenue cost due to economic growth, higher wages and employment, and higher tax collections on the higher incomes. To keep the reform revenue neutral, the government would only need base-broadeners equal to about 40 percent of the static cost
In my view, we should get rid of all deductions period, for simplicity and to allow even more rate reductions. But this has been important in the political debate.

Models are built on assumptions, and here too. How much do people work more, save more, invest more when they face a 20% reduction in tax rates? That's an important question, and if you want to quibble with the answer, where you should look. I got as far as the description here of the underlying model, starting on p. 19.

The model assumes a rather low labor supply elasticity of 0.3. That's important and conservative -- the model is not assuming that, being allowed to keep more after-tax wages, people go out and work incredible numbers of more hours, or lots of people join the labor force. I suspect the real number is higher, especially on the latter margin.

On capital, the model starts with a central and very powerful observation
The long run real after-tax rate of return to physical capital is virtually constant over time, implying that the supply (quantity) of capital is very responsive to changes in the rate of return.
This is what's great about neoclassical growth models. A few basic facts have powerful implications. In this case, that small changes in rates of return will bring in lots of capital, from abroad if not from savers. If you want to assume otherwise, you have some hard facts staring at you.

I was not able to determine whether the model included payroll, state, local, and sales taxes. The overall tax wedge counts to distortions, and including those in the baseline will substantially raise the effects of lowering Federal taxes.

I haven't looked at the calculations, and I'm not vouching for them. But this is certainly the right way to ask the question! And if you want to disagree, we have a disciplined way to disagree. Disagree with the labor supply elasticity, the substitutability between capital and labor, and so on, if you will, but we know what we're debating.

Reading these analyses I wish they would do a Romney vs. Obama plan comparison, using the same methodology. For example, I was struck that the original Tax Policy analysis of Romney's plan concluded that, since in their static analysis there weren't enough base broadeners, that Romney must have a secret plan to raise middle-income taxes. OK, but Obama's budget numbers don't even pretend to reduce deficits. So what sense does it make to say, Romney has a secret plan to raise taxes because we forecast a  deficit, but Obama's plan has... a deficit? If we're going to hold plans to a deficit path and make up taxes to do it, shouldn't we see how both plans stack up on the same deficit path?

But, dear reader, remember to take all of this with a grain of salt. Campaign plans have very little to do with what Presidents propose, and what Congress actually votes. President Obama's plan last time was to cut the deficit in half, and you saw how that turned out. And rightly so. Events change the best plans. These are best read as general indicators of broad-brush themes, not for whether the tax credit for windmill powered cars will really truly be $5,000 or $7,000.