In his column today Dr. Paul Krugman argues that the deficit impact of a large ($1 trillion) stimulus would be mitigated by the effects of higher GDP growth:

Consider the long-run budget implications for the United States of spending $1 trillion on stimulus at a time when the economy is suffering from severe unemployment.

That sounds like a lot of money. But the US Treasury can currently issue long-term inflation-protected securities at an interest rate of 1.75%. So the long-term cost of servicing an extra trillion dollars of borrowing is $17.5 billion, or around 0.13 percent of GDP.

And bear in mind that additional stimulus would lead to at least a somewhat stronger economy, and hence higher revenues. Almost surely, the true budget cost of $1 trillion in stimulus would be less than one-tenth of one percent of GDP … not much cost to pay for generating jobs when they’re badly needed and avoiding disastrous cuts in government services.

Dr. Krugman focuses only on the long-term debt service costs of a large new stimulus. This assumes we keep the full trillion dollar cost of his hypothetical stimulus as public debt forever, and he argues the “true budget cost” is just the added burden of interest payments. I think looking at only the interest costs is an incomplete way to measure the true budget cost of a policy change, but today I want to focus on the bolded sentences and walk through Dr. Krugman’s logic.

Let’s grab an envelope and write on the back. Please note that this is not my “estimate for the Krugman stimulus.” I’m simply constructing a numeric example to demonstrate a concept. Except for the trillion dollar starting point, these are my numbers, not Dr. Krugman’s:

  • Enact a stimulus law that increases the deficit by $1 trillion.
  • Assume the stimulus law works and increases GDP growth. Even if you think fiscal stimulus is largely ineffective, please play along. Let’s be super generous and assume it’s all immediately effective and there’s no waste, so it adds four percentage points to GDP this year. I don’t think this is possible and I’m just picking four as a nice round number for this example.
  • We expect to have a $15-ish trillion economy this year. ($14.7 T according to CBO.) That’s $15,000 B.
  • +4% X $15,000 B = +$600 B higher GDP from our perfect magic $1 Trillion stimulus.
  • Government takes on average about 18% of GDP in taxes. It’s lower during recessions but higher over the next decade under the President’s policy. Round up to 20%.
  • 20% X $600 B = $120 B in higher revenues resulting from the higher GDP growth resulting from the perfect magic stimulus.
  • $1,000 B gross deficit impact – $120 B higher revenues = $880 B net deficit impact of this hypothetical stimulus law, 12% less than the original gross score.

As a technical matter you can break this process up into two steps. The first step is where you guess how much your policy will increase GDP. The technical experts call this dynamic analysis. In the second step of dynamic scoring you translate the higher GDP estimate into an estimate of increased government revenues and therefore a smaller budget deficit.

Note the policy still increases the deficit, in this case by almost $900 B. This perfect magic spending stimulus does not “pay for itself” by a long shot. As Dr. Krugman points out, the dynamic aspects of it merely reduce the gross deficit increase: “additional stimulus would lead to a somewhat stronger economy, and hence higher revenues.”

You also need to think about the effects on interest rates, so a key assumption in the calculation is how the Fed will react. If you think the Fed will raise interest rates in reaction to your big fiscal policy change, then GDP growth will increase less rapidly and government debt service costs will increase, wiping out some of the dynamically scored benefit. The Fed’s reaction function probably depends on the strength of the economy – in a weaker economy the Fed won’t raise interest rates as much and so you get a bigger dynamic effect. If the economy were humming along at 5% unemployment rather than almost 10%, the Fed would worry about instantly adding four percent to GDP and would be more likely to raise interest rates, causing the dynamic benefit in the above example would be smaller.

As a practical matter you’d only want to dynamically score fiscal policy changes that are big enough to move the GDP needle in a measurable way. It would be silly to repeat the above back-of-the-envelope calculation for a $1 B law, because that’s way too small to measurably effect a $15 trillion economy. There’s some large de minimis threshold you would need to exceed for your estimate not to be silly.

Also, the policies within the law affect your estimate. The +4 percentage points I assumed depends on how much you think the policy will increase GDP, and how quickly. Economists love to debate these questions about the multiplier of certain types of spending increases vs. other types of tax cuts. My favorite paper is by Greg Mankiw and Matt Weinzierl: Dynamic Scoring: A Back-of-the-Envelope Guide. They write:

The feedback is surprisingly large: for standard parameter values, half of a capital tax cut is self-financing.

The (rounded and oversimplified) 20% number I used depends on what components of GDP will be increased, since different parts of the economy are taxed at different rates.

Dynamic scoring was originally advocated for tax cuts by conservatives who correctly argued that “static scoring”overestimated the deficit impact of large, broad-based tax cuts and incorrectly argued that the dynamic effect was so large that tax cuts would “pay for themselves.”

VP Cheney said this once and caught years of grief for it. He was incorrect – no conceivable tax cut from our current position can fully pay for itself. Using the logic described above, a broad-based reduction in rates on labor or capital income could, however, partially pay for itself.

In the world of DC fiscal policy combat, you will rarely hear a conservative admit that tax cuts do not fully pay for themselves, just as you will rarely hear a liberal acknowledge that large broad-based tax rate cuts do increase GDP growth and partially offset the gross deficit effect.

CBO began limited use of dynamic scoring (which they refer to as “relaxing the assumption of fixed nominal GDP”) in 1995 at the urging of Congressional Republicans.

While Dr. Krugman has long acknowledged the logic of dynamic scoring, it appears he supports it only for spending increases, and only when Democrats are in charge (seriously).

Here is Dr. Krugman in February 1995 (emphasis added by me):

But if you want to know whether we are really on the way to becoming a Latin American look-alike economy, the key issue to watch is a seemingly arcane one: the idea of “dynamic scoring” for tax cuts.

The basic idea of dynamic scoring is reasonable: People will alter their behavior when you change tax rates, so if you want to figure out how much revenue is gained or lost from these changes, you should take that altered behavior into account. For example, if you cut the federal gasoline tax, people might buy less-fuel-efficient cars, raising gas consumption, so the feds could conceivably end up with more rather than less revenue.

The problem is that economics is not (to say the least) an exact science, so that attempts to predict the effects of tax changes on behavior are both uncertain and controversial. Thus, the only kind of person you want to trust with dynamic scoring is someone who not only knows his stuff but will consistently bend over backwards to avoid reaching comfortable conclusions simply because they are politically convenient.

Do the Republican leaders and their economic advisers who are calling for dynamic scoring meet this test? Does the Republican majority believe that a cut in the capital gains tax will actually reduce the deficit because it has made an objective study of the statistical and economic issues involved, and has reached the conclusion in spite of a determination not to engage in wishful thinking?

So how should you approach the idea of dynamic scoring? With great caution. If the Republicans show a lot of enthusiasm for the idea, or if they choose a director for the Congressional Budget Office who loves the notion, then you might want to think about investing your money someplace where bitter experience has taught politicians and the public the virtues of responsibility–someplace like, say, Argentina.

Here he is in January 2003:


blockquote>Will this alcoholic

[President Bush] eventually go back on the wagon? Not for a while; he has too many enablers. The Congressional Budget Office will soon start using “dynamic scoring” to assess proposed tax cuts – that is, it will build in the supply-side assumption that tax cuts raise the economy’s growth rate, and therefore generate indirect revenue gains that offset the direct revenue losses. In the past, budget officials have opposed this practice, because it’s so easy to slide from objective analysis into wishful thinking. With Republicans controlling both the White House and Congress, does anyone doubt that future C.B.O. analyses will take a very favorable view of big tax cuts for rich people?

Now that Democrats are in charge, today Dr. Krugman titles his column “The Bad Logic of Fiscal Austerity” and argues for dynamic scoring of a trillion dollar stimulus. Given CBO’s projection of budget deficits exceeding four percent of GDP for the next ten years, it makes me wonder if Dr. Krugman thinks this quote from his 2003 column now applies:

It’s O.K. to run a deficit during a recession, as long as the deficit is clearly temporary. But both the numbers and the administration’s search for excuses tell us that there’s nothing temporary about the red ink. On the contrary, we’ll probably be on a deficit bender until the baby boomers retire – and then it will get much worse.

My view

  • Dynamic scoring is conceptually valid.
  • It should be applied to both the tax and spending sides of the ledger.
  • When doing dynamic scoring you need to think hard about (1) the effect of the proposed policies on short-term economic growth, (2) how the Fed will react, and (3) whether the specific policies increase the capacity of the economy and the potential for higher long-term economic growth. You’ll probably have different parameters based on both the specific policy change and the short-term status of the economy.
  • It’s hard to estimate the parameters, so dynamic scoring should be applied in very limited cases: only to very large, broad-based policies that will undoubtedly affect the level of GDP. I would probably set a threshold of a gross deficit effect of at least half a percent of GDP per year. That’s $75-ish B per year, or $750 B over ten years. Really big.
  • While tax cuts do not fully pay for themselves, large broad-based cuts in tax rates on capital or labor income can partially pay for themselves.
  • The deficit effect of a proposed policy should be one important factor in decision-making. I don’t need a tax cut to pay for itself to think it’s good policy if there are other reasons to do it. Others might argue the same for particular spending increases.

(photo credit: Wikipedia)