A serious case for dynamic scoring

Innovation policy has a budget problem. Here’s how Congress can fix it.

Doug Elmendorf is Dean and Don K. Price Professor of Public Policy at Harvard Kennedy School, and former Director of the U.S. Congressional Budget Office. Heidi Williams is Professor of Economics at Dartmouth College and Director of Science Policy at the Institute for Progress, a non-partisan think tank focused on innovation policy.

Everyone from Mitt Romney to Barack Obama to Hilary Clinton to Marco Rubio to Donald Trump has voiced support for the idea that we should make green cards more easily available to foreign-born STEM graduates. If there is so much genuine bipartisan support for this idea, why has it never come to fruition?

Data — and intuition — suggest that admitting more high-skilled immigrants to the U.S. would generate substantial benefits to the U.S. economy. Large numbers of skilled immigrants are eager to move to the U.S., and when they move here, their research is highly productive, they generate valuable innovations at disproportionately high rates, and they start new firms that lead them to act more as “job creators” than “job takers” for U.S. natives. Yet among people who write legislation, it’s an open secret that proposals to increase the number of high-skilled immigrants are currently tabulated by the U.S. government as being a net cost to the U.S. economy rather than a net benefit. For example, when the Congressional Budget Office (CBO) and the staff of the Joint Committee on Taxation (JCT) scored Section 80303 of the America COMPETES Act — which proposed increasing the availability of green cards for STEM masters and PhDs — they tabulated that it would cost the federal government $3.1 billion over 10 years.

Why would CBO and JCT be reporting back a score that tells lawmakers and the public an answer that is at odds with the evidence?

CBO and JCT didn’t make a mistake in this estimate; they simply followed procedures specified by Congress, but those procedures did not serve Congress well in this case. While understanding the causes of this disconnect requires diving into some details, the good news is that there is light at the end of the tunnel. If Congress wants scores of proposals that increase high-skilled immigration to reflect their true budgetary value, Congress can direct CBO and JCT to invest in fixing this problem. 

Conventional vs. dynamic scoring

Taking a step back, whenever Congress wants to know how a policy proposal would affect the federal budget, they ask for a “score” of the proposal from CBO and JCT: CBO provides spending estimates, and JCT provides revenue estimates. Conventional scores account for mechanical effects — for example, subsidizing flu vaccine prices increases government spending — as well as many kinds of behavioral responses —for example, lower flu vaccine prices would probably encourage more people to get vaccinated. 

What conventional scores do not take into account is how a policy could affect employment, growth, and productivity in the macroeconomy; such effects are included in so-called dynamic scores, but not in conventional scores. This distinction matters because conventional and dynamic scores can be quite different. Conservative lawmakers have argued, for example, that tax cuts could increase economic activity and spur economic growth — effects that are not included in conventional scores and that could make tax cuts less costly for the government or even save the government money. Public debate over dynamic scoring largely focused on this tax example, and dynamic scoring became branded as a partisan, conservative idea.

The problem is that dynamic scoring shouldn't be partisan. In many cases, conventional and dynamic scores will be similar, but in cases where they differ, good dynamic scores are more accurate and therefore more informative for lawmakers. Because dynamic scoring is more time-consuming and more complicated, it isn’t the right tool for all situations. One of us (Elmendorf) has argued that dynamic scoring should be applied to “major” legislation that would affect the macroeconomy. But a complementary approach is to anticipate the contexts in which we expect conventional and dynamic scores to differ. Emblematic are policies that would spur research and innovation — policies that often appear to increase deficits under conventional scoring despite strong evidence that they could increase economic growth. When we apply conventional scoring to such policies, we’re telling lawmakers and the public that these policies are not as good as we know they really are.

In the case of high-skilled immigration, conventional scores do not even capture the mechanical increase in U.S. population growth that arises when additional immigrants come to the U.S. The Section 80303 proposal, for example, would have brought (or let stay, post-graduation) more STEM masters and PhDs to the U.S., but conventional scoring doesn’t consider the additional tax revenue that would be collected from these relatively high-wage workers, much less the productivity and growth effects that have been documented from admitting more high-skilled immigrants to the U.S.

Because congressional proposals to increase high-skilled immigration like Section 80303 are currently scored as a fiscal negative, since the costs of the government benefits that would be received by green card recipients are tabulated, such proposals face a more daunting legislative gauntlet. When bargaining over a multifaceted piece of legislation, the components that improve the budget tend to be prioritized, and the parts that cost money are often left on the cutting-room floor.

Scoring can be improved

And so we are left with problems like Taiwan Semiconductor Manufacturing Company’s recent announcement that it can’t find enough skilled workers to get its new Arizona chip plants ready in time, delaying mass production to 2025. Against that background, we were pleased to see a recent request from the CBO for comments on the relationship between immigration and productivity — a request that suggests the agency is looking into revising its scoring approach to better account for the productivity and growth effects of immigrants. 

However, at the end of the day, CBO and JCT do what Congress asks of them. If Congress wants scores of high-skilled immigration proposals to reflect their true budgetary value, Congress must direct CBO and JCT to invest in fixing this problem. Requests for dynamic scoring can be made by the Chairman or Ranking Member of either the House Budget Committee or the Senate Budget Committee. For immigration in particular, there is also an intermediate adjustment that falls between conventional and dynamic scoring. Namely, for bills that would materially increase the size of the U.S. labor force, CBO and JCT have — in the past, on request — incorporated their projections of the direct effects of the bill on the U.S. population, employment, and taxable compensation; the CBO cost estimate for H.R. 2131 (SKILLS Visa Act) is one example. Any immigration-related scoring submission by any Chair or Ranking Member of a committee with jurisdiction could request “partially dynamic” scoring which accounts for population, employment, and taxable compensation.

While the immigration example is particularly stark, similar issues arise for other innovation policies. Consider federal funding for research and development (R&D). Such investments have a direct cost to the federal budget: funds are spent, and benefits such as securing national defense or maintaining public health are not in CBO’s mandate to measure. But what if federal R&D investments increase macroeconomic output through innovation and economic growth, or through inducing positive effects on private-sector productivity?

For example, recent research by economists Shawn Kantor and Alexander Whalley finds substantial positive effects on local and national output from U.S. federal R&D during the space race, even within the 10 years following the end of the space race (which is relevant timing given the standard 10-year budget scoring window). Although we agree with CBO that more evidence on the impact of federal R&D would be valuable, existing evidence like Kantor and Whalley’s shows that conventional scores of federal R&D investments will look worse than more accurate dynamic scores. Moreover, dynamic scoring can evolve as additional economic research emerges.

For policies like increasing high-skilled immigration that induce economically meaningful changes in employment and productivity, conventional scores misrepresent the true effects on the federal budget implied by existing evidence. Rather than seeing dynamic scoring as a conservative agenda, those invested in spurring economic growth should pay attention to cases where dynamic scoring would help align the incentives of Congress with the policies that are best for society.