It’s the first anniversary of the American Recovery and Re-Investment Act of 2009 (ARRA). Politicians, economists, and the public are either celebrating or bemoaning the event, while the media is delighted to report anyone’s viewpoint. Indeed, there is no precedent that we can recall for the attention and controversy surrounding year-old fiscal policy. This is no doubt due to the extraordinary efforts of the Obama administration to track the stimulus money while promoting the benefits of the policy even as many Republicans argued the opposite.1 In any event, we’re being asked frequently to comment on ARRA. This a good time to review our estimates of the macroeconomic effects of the stimulus package and sift through the rhetoric of the past year — some of which was pointed criticism of our own analysis.
Tracking the Progress
Sorting through the various estimates of how much money has been appropriated, obligated, and spent can be confusing. From a modeler’s perspective, however, the most useful figures are those reported by the Bureau of Economic Analysis (BEA), which publishes estimates of the impact of ARRA on federal receipts and expenditures in the National Income and Product Accounts. These estimates are important because they purport to represent the direct impact of the legislation on current incomes and production. They also correspond directly to the fiscal assumptions fed into our econometric model to arrive at our estimates of the overall (e.g., including multiplier effects) effects of ARRA. Figure 1 (on page 2) shows the BEA’s estimates, both quarterly (at annualized rates) and cumulatively through the end of last year.2
By the end of 2009 the cumulative tax cut was roughly $85 billion and the cumulative spending increase was roughly $140 billion, for a total of $224 billion, or less than one-third of the total amount appropriated. This compares to $169 billion assumed through 2009 by MA when the legislation was enacted. However, one should not conclude from this that the money is being spent faster than we anticipated. We assumed that the expensive fix for the alternative minimum tax and part of the temporary extension of unemployment benefits would have occurred anyway. Accordingly, these initiatives were treated as part of our baseline forecast rather than part of the fiscal initiative. Allowing for this adjustment, our earlier assumption for the stimulus total to date is similar to the BEA’s reckoning. The composition is somewhat different than we anticipated, with more of the spending on grants-in-aid to states and less on direct federal purchases. However, the macroeconomic effects of this compositional shift probably are small since the states likely used the grants to increase direct spending at the state level.
In February of 2009 we published a MACRO FOCUS entitled “Fiscal Stimulus to the Rescue — Final Answer” in which we estimated, relative to a “pre-stimulus baseline,” the macroeconomic effects of the legislation.3 Those results are shown in Charts 1 and 2 on page 3.
As just argued, the timing and the composition of the stimulus have been close enough to our previous assumptions that we remain comfortable with our earlier findings. Hence, we still estimate that the peak effect of the stimulus on employment will reach about 2.5 million by the end of this year before then fading gradually. The effect on real GDP growth (measured 4th-quarter-to-4th-quarter) will peak at around 1.5 percentage points early in 2010 before turning negative in late 2011 as the stimulus unwinds. These results fall towards the lower end of the range of estimates reported by the Congressional Budget Office.
Measure or Estimate the Jobs Effect?
What else can we say about the analysis and rhetoric of the past year? For starters, parties receiving stimulus funds are asked to file an estimate of the number of jobs saved or created with the money. When these estimates were tallied and audited by the government and press, it was clear the figures were problematic. There were just too many examples of nonsensical results. Perhaps, as some argued, the questions were not well worded. Just as likely, however, is that respondents didn’t really know the answer to the question. Furthermore, even if such estimates are accurate, they reflect only the direct effect on jobs, not the indirect or “multiplier” effects. The point is that it simply is not possible to measure directly (that is, by survey) the total effect on jobs. One needs to estimate the effect with an economic model that can be used to compare a pre-stimulus baseline with a simulation that includes all the effects of the stimulus. This is the only methodology that really makes sense to us.
Demagoguery: Baseline versus Incremental Effect
Frequently, partisan commentators — and even some economists — exclaim that the stimulus has failed because the unemployment rate now exceeds the peak shown in projections prepared before ARRA was implemented.4 This argument, which clearly — and perhaps intentionally — confuses the pre-stimulus baseline with the incremental effects of the stimulus, would be laughable if it was not taken so seriously in some quarters. For the record, last spring, as the financial crisis that engulfed the economy worsened unexpectedly — but before the stimulus could possibly have had any real effect on the economy — the unemployment rate already had moved above the Administration’s (and many others’) last pre-stimulus projection. So, this is simple: the baseline forecasts were optimistic, but unemployment would be even higher now without the benefit of the stimulus package.
The last year has seen plenty of criticism of the methodology that underlies not only our estimates but also those prepared elsewhere in the private sector, by the CBO, and by the White House. The critiques fall into two broad categories.
Permanent versus Temporary
In a Wall Street Journal editorial, John Taylor et. al. pointed to the one-time payment of $250 to Social Security beneficiaries last spring as an example of a transitory payment with little or no stimulus effect, and strongly implied that macro modelers such as MA were unaware of the “modern” life-cycle (or permanent income) theory of household behavior.5 However, all the principals of MA are intellectual descendents of Franco Modigliani, and we are fully aware of the life-cycle model. It is strongly reflected in both our econometric model and in the judgments we apply to that system. Case in point: we explicitly assumed the marginal propensity to consume out of that one-time transfer payment was essentially zero.
More generally, in preparing our analysis we made an initial judgment about the perceived permanency of the major elements of the stimulus package. For example, the most important tax cut in ARRA was the “Making Work Pay” (MWP) tax credit, which was legislated as temporary. However, we assumed it was perceived as permanent and so would have effects normally associated with a permanent tax cut. Our decision was based on President Obama’s campaign pledges, and today our decision still seems justified given that the Administration’s proposed budget for fiscal year 2011 seeks to extend the MWP credit beyond the end of 2010.
In addition, our model uses modern econometric techniques of co-integration and error-correction to differentiate households’ responses to transitory as opposed to permanent events. So, in yet another example, we assumed that the extension of unemployment benefits legislated under ARRA was temporary. Over the assumed duration of the provision, the propensity to consume such benefits is only about one half the eventual response of spending to the increase in after-tax wages associated with the permanent MWP credit. This still seems quite sensible to us.
Finally, there is the estimated response by state governments to the receipt of grant monies. Almost all states have a balanced-budget requirement, so stimulus grants likely have prevented belt-tightening at the state level. Our estimates of this effect are based on fiscal “reaction functions” that relate components of states’ spending and taxes to their recent budget deficits. We do not have to worry about whether the impact of grants on states’ deficits is temporary or permanent because the states’ balanced budget-requirements don’t make that distinction. If states’ deficits widen, their budgets must be tightened, and historically that has been the case. Hence, we can be fairly confident that the empirical regularities reflected in our reaction functions are applicable in the current environment.
Taylor has also argued that fiscal multipliers in large-scale econometric models are too large, in part because they do not incorporate the standard monetary response (e.g., to tighten) in the face of a stimulus. As evidence, he points to small multipliers often found in reduced-form models.
In our model, however, there is not a simple, single multiplier; the system is far more complicated than that. Different kinds of stimulus — personal taxes, business taxes, direct spending, subsidies, transfer payments — all have different direct and indirect effects on the economy. Furthermore, those effects depend upon whether we view an initiative as temporary or permanent, whether we are looking at the short run or the long run, and whether we assume an offsetting monetary policy. Each situation, and the associated multiplier, is different.
For example, consider the temporary “bonus expensing” of tangible property included in ARRA. This could have a very small pull-forward effect on fixed investment — we did allow for a modest effect here — but bonus expensing is mostly a lump-sum transfer to shareholders. There is associated with this transfer a normal wealth effect on consumption, but the effect is very small. Similarly, the other business tax cuts in ARRA have near-zero impacts.
Now it must also be mentioned that in our model a permanent increase in government spending temporarily boosts GDP. The long-run impact is to reduce GDP; that is, the long-run multiplier is, as neoclassical theory suggests, negative, thanks to the “crowding out” of productive investments in the private sector. The vehicle for crowding out is a rise in real interest rates that can be either advanced or delayed — but not prevented — by the Fed. This means that by varying our monetary assumptions, we can use our model to produce practically any short-run fiscal multiplier we like, including those of the magnitude championed by Taylor. For example, if we assume that the Fed responds to ARRA by raising interest rates according to a Taylor rule, the resulting short-run multiplier in our model is far smaller than if we assume real interest rates are fixed.
This last point is particularly relevant in today’s economy. Currently the nominal federal funds rate is pinned at nearly zero, and our favorite monetary rule suggests that it should be negative. Hence, when we prepared our analysis of ARRA, we explicitly assumed the Fed would not raise nominal interest rates in response to the stimulus because it wanted to encourage the largest possible impact on the economy. Furthermore, since the stimulus reduced slack in the economy and so put upward pressure on inflation (or prevented further disinflation), the real interest rate fell in our simulations. This is opposite the “normal” response of real rates in most reduced form models with an implicit (or explicit) monetary rule, and works to magnify the multiplier effects — as intended by the Fed. Here is a case in which a detailed, general structural model that can capture today’s special circumstances is better for understanding the likely response of the economy to the stimulus than an a-theoretical, reduced-form model based on historical averages.
Was it Right, Was it Enough, Do We need More?
Certainly everyone can agree that ARRA was messy and inefficient. With extra time and deliberation a more effective stimulus plan surely could have been devised! Everyone can also agree that it was not a silver bullet; after all, the unemployment rate breached 10%. Yet our feeling remains that ARRA is having a measurable and sizable, if temporary, impact on economic growth and employment, and that by reducing slack in the labor market, it provided insurance against the risk of deflation that was palpable when the legislated was enacted. Indeed, part of ARRA’s value was its signal that the federal government did not intend to let economy slide over a precipice into another Great Depression, something that seemed quite possible at the time.
For those who argue that we couldn’t afford ARRA, we offer these counterpoints: (a) our analysis showed that the legislation “dynamically” paid for more than a third of itself by restoring the economy to full employment sooner; (b) because it is temporary, ARRA has essentially no lasting impact on the unsustainable nature of the country’s current fiscal stance; (c) our calculations suggested that the benefit (e.g., the present discounted value of temporarily higher GDP) exceeded the budgetary cost of the legislation.
Of course the package could have been larger and, in that case, both the effects and the messiness of the legislation would have been larger as well. Perhaps that would have been desirable. Yet the size of the package ultimately was limited the political and budgetary realities, and those same considerations today suggest that a major follow-on is not in the works unless the nascent recovery, which looks promising, suddenly stalls.
The Fiscal Uncertainties Just Ahead
As Chart 2 makes clear, the stimulus associated with ARRA will turn to drag in 2011 as the policies unwind…and that assumes the MWP credit in fact becomes permanent. The shift towards fiscal drag could be significantly exacerbated if the Bush tax cuts are allowed to sunset as currently legislated. If the economy is booming by then, such a sharp reversal of fiscal policy might be desirable not only as short-run stabilization policy but also as a partial solution to the nation’s long-term fiscal predicament. However, if the recovery remains fragile by then, such a sharp fiscal contraction would almost certainly seem undesirable, and it will seem that ARRA was a “bridge” not long enough to reach a sustainable recovery.