2018 NYU Tax Policy Colloquium: Greg Leiserson on dynamic scoring, part 2

Greg Leiserson’s colloquium paper on dynamic scoring powerfully argues that dynamic scoring, as currently practiced, is a flawed and asymmetric approach that leans unduly in favoring tax cuts – almost, but not quite 100%, the bigger the better – without properly addressing the relevant tradeoffs, such as the fact that there will be losers in the future from reductions in net revenue.

But let’s start here with why one might want to do dynamic scoring. By saying that “one” might want to do it, I am abstracting for a moment from the political process, in which dynamic scoring is a weaponized tool pushed by those on the political right to advance their own subjective policy preferences, to ask only why it might be of interest to good faith players who wanted more and better information, even absent that entire side of things.

To set things straight from the start, conventional revenue estimates, no less than “dynamic” ones, are completely dynamic in a microeconomic sense. They build in, for example, the point that, if one doubled tax preferences for solar panels on people’s roofs, there would likely be a supply side response (more $$ spent by more people on such panels). What conventional revenue estimates treat as fixed is the macroeconomic side of things – aggregate labor supply, savings, available capital, etc.

Now, those things are not in fact fixed. They can change in response to a tax change, at least if it’s big enough and has some direct impact on them. So in principle one ought to take those things into account.

Suppose, for example, that the 2017 tax act actually would have paid in full for itself. That would have been worth knowing. And even if it, say, one-third pays for itself, that, too, is worth knowing.

Suppose further that members of Congress ask the staff: At what level do you project 2020 GDP, (a) without vs. (b) with a particular tax bill’s being enacted? It’s a reasonable question to ask, and one that merits a fully informed answer.  But this brings us to the various “buts” regarding dynamic scoring, especially as practiced.

An iniital point is sharp dissensus among macroeconomic models. But that of course doesn’t mean we should have no such estimates, it just means they should be handled with care and that there should probably be multiple forecasts presented. (But JCT was instead told to provide a single operative estimate this time around.)

A second, more telling point, is that estimators are being asked to forecast incompletely specified policies. A budget forecast should be forward-looking for two reasons. One is that we should care about the future. The second is that actual economic actors may be forward-looking, so what they do may reflect their expectations about the future.

When there’s a preexisting fiscal gap, and a giant tax cut would make it even worse (even building in today’s actors’ expectations), the actual full set of policies, including the consequences of the set of pay-fors, remains unspecified. Now, this does not entirely defeat one’s making estimates about what will happen over the next ten years if the announced policies remain in force. But it creates a set of discontinuities – inside vs. outside the budget window, and announced vs. as yet unannounced policies – that can be misleading to policymakers and the public even if the 10-year forecast is accurate.

If there’s one fundamental political economy problem in budget policymaking, other than power imbalances and disregard for the interests of the less powerful, it’s short-termism. Politicians and voters want good stuff now without due regard for the nation’s future. (This is the subject, for example, of James Buchanan and Robert Wagner’s famous book, Democracy in Deficit.)

Dynamic scoring, when added to a budget window, accentuates short-termism. (One could reasonably argue, however, that the underlying problem is more about the budget window than dynamic scoring as such.) It gets back to the question that, as I noted in my prior post, policymakers might reasonably want to ask their staff: What would a given tax bill do, say, to GDP in 2 or 3 years? Revenue-losing tax bills such as the 2017 act have a tendency to increase short-term GDP growth in exchange for lowering it in the long run, due to fiscal crowdout. So while one ought to supply the information, one makes the political process worse if the result is to increase focus on the short term at the expense of the long term.

To put it another way, budget rules aren’t just about the quantum of information, they’re also about relative emphasis (as well as imposing specific constraints in particular cases, although the dynamic score was not used that way in 2017). Dynamic scoring as used – not necessarily, as an inherent conceptual thing – accentuates short-termism, when arguably the single most important thing that budget rules and scoring methodologies should do, apart from just supplying information, is to counter short-termism.

What happens if one puts dynamic scoring at the center of the process, and cares only about net revenue cost and GDP growth within the 10-year window? It’s possible that this would lead to one’s deriving the result that GDP would be the highest if federal revenue went all the way to zero – i.e., no taxes whatsoever.

To put it more precisely, the only way that wouldn’t be true is if crowd-out within the ten year period from having no tax revenues whatsoever, hence skyrocketing public debt and annual interest charges, sufficiently suppressed growth within the 10-year period by reason of crowd-out. Now this might happen in the forecast, even if estimators didn’t decide that they needed to consider modeling a full-blown fiscal crisis inside the budget window.

But there’s still a sense in which the wrong question (at least as a matter of emphasis) is being asked. If one centrally relies on dynamic scoring PLUS exclusive focus on the 10-year budget window, the optimal level of taxation appears to be that which would be optimal if (a) one had to pay interest charges and deal with crowd-out within the 10-year period, but also (b) at the end of 10 years, one got to cancel the public debt run-up – settle it for zero -without this having been anticipated or having adverse effects afterwards.

That is obviously not a sensible way to approach actual budget policy. As the Leiserson paper convincingly shows, dynamic scoring – again, conditioned on how it is being used, not necessarily in the abstract as one tool among dozens – creates bias in favor of ever larger tax cuts without offering any clue as to the fact that diminishing government revenues adversely affects people who therefore either get less from the federal government, now or in the future, or else end up paying more taxes later due to new enactments.

One last perspective that I want to offer with regard to dynamic scoring is its having been used as a “weaponized” tool in the federal budget and tax policy wars, and its not being the only possible weaponized tool, in support of the only existing interests, that one could imagine. But I will leave that point for a separate post that is to follow shortly.