…and a Rejoinder to Posner.
The CEA Analysis of ARRA’s Impact
Yesterday, the Council of Economic Advisers released the first of its mandated reports on the impact of the ARRA on economic activity. Based upon a variety of approaches (VAR, multiplier based), it concludes:
“…our multiplier analysis and estimates from a wide range of private and public sector forecasters confirm the estimates from the statistical projection analysis. There is broad agreement that the ARRA has added between 2 and 3 percentage points to baseline real GDP growth in the second quarter of 2009 and around 3 percentage points in the third quarter.
There is also broad agreement that it has likely added between 600,000 and 1.1 million to employment (again, relative to what would have happened without stimulus) as of the third quarter.”
The CEA actually conducted a series of analyses. The first is an informal approach, examining the contributions of components of GDP to overall growth over time, in an accounting exercise, and then correlating changes in component trajectories that might have reasonably been expected to impacted by the stimulus bill (e.g., state and local government spending on goods and services). The second is a bivariate (in levels) VAR in GDP and employment, and attributing the difference between out-of-sample forecast values for 2009Q2 and actual to the stimulus package, as well as financial interventions (both positive) and an unprecedented credit crunch (negative). The third is a multiplier analysis, that applies multipliers to various injections into the economy. (There is also a cross-country analysis, which yields suggestive results, but I will not discuss at length.)
The second (VAR or “projection”) approach yields Figure 8 from the report.
Figure 8 from CEA, The Economic Impact of the American Recovery and Reinvestment Act of 2009: First Quarterly Report, September 10, 2009. Estimates from “VAR approach”.
The difference between the projected and actual is given by the darker color in the bars in 2009Q2 and 2009Q3. For 2009Q2, that difference is 2.3 percentage points, at annualized rates.
In the third approach (the multiplier approach), the CEA estimates a 3.1 percentage points (at annualized rates) increment to growth in 2009Q2, and 3.6 ppts in 2009Q3. How does this compare to other estimates? This is addressed in the document’s Table 7.
Table 7 from CEA, The Economic Impact of the American Recovery and Reinvestment Act of 2009: First Quarterly Report, September 10, 2009. “CEA Model Approach” is a multiplier approach.
I think it’s interesting to see the remarkable coherence in estimates (aside from the guesstimates of the NABE survey).
Actual Data, and Posner’s Assertions of Negligible Impacts
What is particularly of interest is the updated information regarding what funds have been disbursed. There are two ways to report the data — the first by administrative category, and the second by function category. The relevant information (cumulative, not annualized) is reported in Tables 1 and 2.
Table 1 from CEA, The Economic Impact of the American Recovery and Reinvestment Act of 2009: First Quarterly Report, September 10, 2009.
Table 2 from CEA, The Economic Impact of the American Recovery and Reinvestment Act of 2009: First Quarterly Report, September 10, 2009.
From the report:
We divide the total dollars of stimulus expended to date into six categories: individual tax cuts and similar payments; the tax cut associated with the adjustment of the Alternative Minimum Tax (AMT); business tax incentives; state fiscal relief; aid to those most directly hurt by the recession; and direct government investment spending. The first three are tax changes of some kind and were established at passage to be roughly one-third of the total package; the second two represent emergency measures and were again estimated to be roughly one-third of the total; the last encompasses a range of direct spending and covers the remaining one-third of the total. At passage, it was anticipated that the tax changes and emergency measures would occur more quickly and direct government spending would be a larger fraction of later expenditures.
We divide the outlays and tax reduction data into these functional categories as follows. Individual tax cuts include the Making Work Pay tax credit, the child tax credit, and a number of smaller individual tax reductions. We also include direct payments (from Recovery.gov) that were made in lieu of a tax cut to certain groups. These include payments of $250 distributed to individuals who receive Social Security and Supplemental Security Income, Railroad Retirement benefits, or veterans’ benefits. The business tax incentives and AMT relief are calculated directly by the IRS as part of their simulation process.We define state fiscal relief to include just the two main programs in this category: a substantial increase in the Federal government’s matching percentage for Medicaid spending (FMAP), and formula grants to state governments for education through the State Fiscal Stabilization Fund. Aid to those directly impacted by the recession includes the increase and extension of unemployment benefits, increased funds for nutritional assistance, and increases in the Temporary Aid to Needy Families (TANF) program. It also includes the government’s substantial subsidy of continuing health insurance benefits (COBRA), which is technically treated as a tax cut.
Government investment outlays include everything else. The obvious components are spending on infrastructure, health information technology, research on renewable energy, and other forms of direct spending excluding transfers. Also included here are tax credits for particular types of private spending, such as weatherization or research and experimentation, since these credits are functionally similar to the direct government spending.
These points are of particular salience in the context of Mr. Posner’s post yesterday, in which he states:
No one seems to know how much of the $40 billion was actually received by taxpayers, as opposed to reducing their future tax payments; and of the amount actually received by them, no one seems to know how much they spent rather than saved.
In response to my assertion that he had accused Dr. Romer of making up statistics:
I never accused her of lying about the figure…
This is actually a “revised and extended” version of the argument posed in Mr. Posner’s first posting (August 18) on Romer’s speech. Reviewing that statement is illuminating:
Romer argues in her talk that by the end of the second quarter of this year, $100 billion of stimulus money had been spent. That is a suspiciously round number, and it is unclear how it was arrived at; but let us assume it is accurate. She then argues that this small expenditure–about two-thirds of one percent of the Gross Domestic Product–is responsible for the fact that the decline in GDP fell (on an annualized basis) from 6.2 percent in the first quarter of the year to 1 percent in the second quarter (though the latter figure is likely to be readjusted upwards).
This assertion is groundless. No one has the faintest idea what effect the stimulus has had. [emphasis added – mdc]
I’ll ignore the fact there’s a footnote 4 in the Romer speech that tells one how the number was arrived at; further I’ll let readers decide what the reference to the “suspiciously round number” is meant to insinuate and focus on the math mistake. Let me reiterate, the emphasized text constitutes a math mistake. $100 billion, divided by quarterly GDP of $ 3535.8 billion is 2.8 percent.
Now, Posner responds on August 21:
Critics have been particularly unsparing of my having expressed the so-called $100 billlion in stimulus disbursements as a percentage of annual GDP. I think it’s a fair criticism–and so it is amusing to note the identical procedure in the second sentence of Romer’s speech, where she states that the $787 billion is “roughly 5 percent of GDP.” It is roughly 5 percent of this year’s GDP. But it is to be spent over at least two years.
A couple of observations First, the $787 billion is to be spent over five years (this was, after all, the chief complaint of stimulus critics), although most is to be spent within 18 months of the passage of ARRA. One can see a graph of projected spending (as estimated by CBO) here. However, on Mr. Posner’s source of amusement, when Dr. Romer indicated the share of GDP, she was placing the total stimulus package in context, and not conjoining that calculation to a statement that the entire $787 billion would have an impact within a given year. That sort of calculation only Mr. Posner has made.
Let’s return to Posner’s assertion that we don’t know anything about the $40 billion on the tax side. Table 1 provides the less “round number” of $43.5 billion cumulative in tax reductions through end-June. I agree that it’s hard to say whether any of those tax reductions were “spent” by the recipient households and firms. That’s why the functional breakdown in Table 2 is very helpful. We can then consider what is likely to have been spent, on both the tax and expenditures sides.
First, let’s take the conservative approach and assume absolutely none of the individual tax refunds, the AMT relief, and the business investment incentives are spent. Second, we have government investment outlays at $5.9 billion; well, that’s G in national income accounting-speak, so GDP has to go up there. Third is aid to directly impacted individuals (unemployment insurance, TANF, food stamps) at $14.4 billion. I suspect that these components have a pretty high marginal propensity to consume. Finally, there is state fiscal relief, which is mostly matching for Medicaid, and for education, at $28.2. Here one could argue that the spending is less than dollar for dollar of transfers — although with a majority of states emplacing furloughs on state workers, I bet this is pretty high. In addition, the turnaround in state and local spending in 2009Q2 is suggestive that the spend-out of these funds is pretty high.
Let’s set MPC to zero for all functional tax components, and set it for unity for government investment and for UI/TANF/food stamps, and 0.5 for state fiscal relief. Write this out in plain mathematical terms:
(29.3 × 0) + (7.6 × 0) + (14.4 × 0) + (5.9 × 1) + (14.4 × 1) + (28.2 × 0.5) = 34.4
Convert to Ch.2005$ by dividing by 1.10: 34.4/1.10 = 31.27
Divide by 2009Q2 GDP in Ch.2005$: 31.27/ 3232 = 0.0097
Annualize the quarterly increment: (1.0097)4 = 1.039
Convert to percent: (1.039 – 1) × 100% = 3.9%
Let me reiterate, in these calculations I’ve assume zero consumer and investment spending out of the tax-related provisions, and yet one obtains a 3.9 ppt increment (annualized q/q) growth in 2009Q2.
Now, in his latest post, Mr. Posner has made a new, bewildering statement:
“A neglected point is brought out in an article by Michael Cooper in the September 5 New York Times. The article points out that federal stimulus spending can be nullified by state cutbacks. For example, a federal grant of stimulus money for mass transit has been nullified by reductions in state expenditures on mass transit. The question then becomes what was the consequence of those reductions? Maybe they enabled a state to rescind a tax increase, in which event state taxpayers would have more money in their pockets.”
The nullification point is an odd assertion. The state cutbacks presumably would have occurred in the absence of the transfers from the Federal government, so in the absence of the stimulus package, state and local spending would have been even lower. In other words, Mr. Posner is asserting a state and local government spending reaction function that implies each dollar of spending transferred from the Federal government induces and equal and offsetting spending reduction. I’m open to this possibility, but I’d appreciate seeing some data (i.e., not anecdotes) to that effect, rather than mere conjecture.
In addition, Mr. Posner sets forth an alternative scenario, where tax increases are rescinded as a consequence of transfers to the states. On this count, CEA has investigated this hypothesis. Figure 7 from The Effects of State Fiscal Relief is illuminating:
Figure 7 from CEA, The Effects of State Fiscal Relief, September 10, 2009.
Figure 7 compares the size of total funds from these programs that had been transferred to each state through July 3 with the state’s maximum reported fiscal 2009 budget gap.22 While these transfers were sizeable in magnitude, in 35 states they are smaller than the reported budget gap. This means that states could have used the funds to increase their rainy day funds only if they reduced expenditures or increased taxes in the same fiscal year. Given that states underwent deep expenditure cuts in 2009 (42 states cutting their fiscal year 2009 budgets by $31.6 billion through June 2009), it seems unlikely that this would have occurred.
In conclusion, it appears to me that we have several pieces of information, from a variety of sources, that suggest a positive impact on 2009Q2 q/q growth stemming from the ARRA. In addition, if one were to do a meta-analysis of estimated impacts, it would seem to me likely that the mean impact would be in range estimated by CEA.
Let end on a note of agreement. Mr. Posner wrote yesterday:
My lefty critics don’t believe me when I say I support the stimulus.
Since Mr. Posner is referring to me (among others), let me say I understand that Mr. Posner is a Keynesian and supports the stimulus. My disagreement is with his nihilistic proposition that we can’t know anything about whether there was an impact of the ARRA in 2009Q2. (Personally, I don’t count myself a “lefty”, but that’s another debate.)
Oh for god’s sake, can’t you just start a “Defending Romer with curious zeal: Come see me rail endlessly against Posner” blog? Half of this post is interesting and informative, the other half is completely unnecessary, and on the whole this entire Posner ‘fight’ appears childish if only for the fact that it keeps going, and going, and going here. I am sincerely lost as to why you are expending so much effort – still – on this, and I wince at the slow but accelerating dive of so many of these blogs toward little more than petty cross-site squabbling.
At least there are always the guest (or other rarely-posting) contributors.
“There is also broad agreement that it has likely added between 600,000 and 1.1 million to employment (again, relative to what would have happened without stimulus) as of the third quarter.”
This staement doesn’t appear to carry any water. The same groups initial projections on UE w/wo ARRA.
http://m.factcheck.org/Images/image/2009/Articles/6_16_2009_Making_Sense_Stimulus_Spending/Romer-Bernstein_Chart.jpg
With all due respect, the real news is that the monthly run rate of stimulus (outlays and tax reductions) peaks in August.
What do you call declining stimulus after August? A policy that is contractionary on a sequential basis.
Of course, stimulus has to decline sometime. However, I don’t think economists pay enough attention to the de-stimulative effect of letting stimulus simply wear off. That’s because they believe in inventory correction cycles, and in those cycles, the economy has recovered by the time the stimulus becomes contractionary. Nothing in this blog’s posts leads me to believe that we are at that self-sustaining recovery point.
So instead of writing about Posner, why not write about:
1) the impact of a declining fiscal stimulus;
and
2) the impact of a non-existent monetary stimulus (as the entire expansion of the Fed’s balance sheet went back into excess reserves). Further, M2 is now in outright decline on a monthly sequential basis.
I look at deflatio in the private sector, as in:
http://www.inflationdata.com/inflation/inflation_rate/historicalinflation.aspx
I see that since the stimulus began, the private sector has deflated by 2%. Does the regression hold up? If oil is the constraint, then yes, because the dollar deflator is still in lock step with oil prices. So the simpler regression is see if the stimulus is decorrelating the economy from oil volatility. The answer is no.
You seem to have picked a red herring (Judge Posner) to argue with in this piece. Few knowledgeable people look to him for economic assessments or advice, why should you waste your time in critiquing his silly arguments? Moreover, why should you accept the White House’s self-report card at face value?
Instead, take a look at the bigger issues on the impact of the ARRA. Some are presented above, both in looking backward at the last quarter and forward to impacts down the road.
To me, the bigger issues of the ARRA have to do with its impact (or not) on employment and unemployment. I am one of those who believes that unless the country is at near full employment, any economic growth is tenuous at best and potentially socially distabilizing (with new “wealth” increasingly concentrated).
So what is the picture today? Already our unemployment rate is at 9.7% and rising while the CBO’s assessment of the ARRA indicated a baseline (non-ARRA)4Q2009 unemployment rate at 9.0% with ARRA reducing that rate to 7.7-8.5%. And please note that these projections used the much-maligned U3 estimate of unemployment. So much for ARRA’s positive impact on unemployment so far.
On the other side of the coin, employment has plunged to 139.6 million according to the latest BLS report whereas the CBO projected the 4Q2009 employment level to be between 142.4-143.9 with ARRA & 141.6 million without. Another strike against the impact of the ARRA.
And, of course, the WH projections were rosier than that. SO, JUST WHERE ARE THOSE ONE MILLION JOBS THAT ARRA HAS “SAVED?”
In the words of the immortal Burger King commercial: Show me the beef!
BKarn: Thank you for your suggestions. If Posner’s nonquantitative assessments were not so representative of lots of commentary on the web, I would cease and desist. But until that occurs, I’m going to keep on boring you with math.
Dave Pearson: I don’t have a monthly time series on outlays (as opposed obligations). If you can provide it, then I’ll be quite interested in seeing it. Your assertion regarding the time profile of stimulus is surprising given the known annual time profile (by FY) of expected stimulus as estimated by CBO (see graph here — peaks in FY10) and CBO’s assessment of the impact on the output gap.
Terry: The Becker-Posner website today ranks 4868 according to the Technorati; it is ranked 33rd by visits on Gongol’s ranking of business/economics websites.
I refer to the CEA reports because I thought they were well done. I’d welcome your insight into how a better analysis would be conducted; that would be constructive.
Menzie: I admire and appreciate all your hard work here. It is nice to have full documentation in a prominent place.
Thank you for taking the time to carefully analyse and refute Prof. Posner’s remarks. Given his scholarly contributions in economics and the high profile of his blog, I think the effort you’ve put in serves a useful (if not essential) purpose in promoting a rational fact-based policy debate.
I continue to be astonished by the dispersion of views among highly regarded economists (compare John Taylor and Paul Krugman, just to mention two.) There can be no rational policy discussion with a “he said/she said” style of economics reporting. Economics, econometrics and the system of national accounts were created to tackle this problems. Thanks for applying them.
Although the assumptions about the contributions of government stimulus are pretty clear, I haven’t found anything specific on what assumptions are made about the private debt change for these projections in this report. What are these? Does anyone know?
I have got the impression with these projection and all the “recovery”-speech again and again that there is a big elephant in the room.
rc
Menzie,
I mischaracterized Mark Zandi’s work. He claims the maximum stimulative EFFECT comes in 3q (I’ve assumed an August peak). That is because the maximum variation in sequential stimulus occurs in that month. So basically, he is looking at the second derivative shrinking, whereas I thought it was turning negative. While a shrinking second derivative is not exactly contractionary, it is possible that Zandi’s point is the relevant one: on a run rate basis, 3q is the point of maximum stimulus.
BTW, you didn’t react to my assertion that monetary policy is not stimulative, or that M2 is now declining. Surely, this is relevant to the question of how effective the comprehensive stimulus will be going forward.
@Menzie,
The Posner battle is like the old childhood game of Whack a Mole. Each time you think you’ve won, he pops up in a different place. I wrote the comment at his blog today and thought you might find it interesting. Keep up the good work.
——————————————
First Posner says:
“I simply do not believe that it is possible to attribute the improvement in the economy to the actual spending of stimulus money through the end of the second quarter, because I believe the actual spending was small and that the significance of the stimulus program is more psychological than (as yet) economic.”
But then he says:
“I attribute Christina Romer’s August 6 speech, and Vice President Biden’s speech of September on the first 200 days of the stimulus program, to the political imperative of shorting up public support for the stimulus. Oddly, the program is unpopular, and not only among Republicans.”
Seemingly oblivious to the inherent contradiction. In short, he’s saying the Recovery Act has led to the sudden improvement in the second quarter, not due to any actual spending, but due to its psychological effect, presumably in the form of increased optimism about the economy. But then he goes and points out that the Recovery Act is unpopular. How is it possible for a program to both increase economic optimism and yet generate antipathy?
The simple answer is it’s not possible. The Recovery Act is having a positive effect, but it’s not through the sheer power of optimism, it’s through the tangible effects of net new consumption, investment and government spending.
He writes:
“The stimulus program probably should have been adopted last fall rather than in February of this year, and expedited more vigorously than it has been.”
Probably, but so much happened so fast it’s hard to remember. The financial crisis erupted in September. The bad economic reports didn’t start rolling in until November. It really wasn’t apparent until then that the financial crisis had spilled over to the general economy. It was only then that talk of a fiscal stimulus arose. It wasn’t until December 1st that the NBER officially declared that we were even in a recession and it wasn’t until December 17th that the fed funds rate was dropped all the way to 0%. It was only then that it was clear we had exhausted monetary policy. We were officially in a liquidity trap.
On January 10th the incoming administration revealed a plan for a discretionary fiscal stimulus. On January 28th the House passed a bill. On February 10th the Denate passed a bill. After passing conference the bill was signed into law on february 17th. The first spending occurred the following month. Exactly two months had elapsed from the Federal Reserve dropping the fed funds rate to zero to the president signing the stimulus into law. And that despite the change of government.
He writes:
“Romer has warned in her academic writing (which incidentally I admire) that Keynesian measures for spurring recovery from an economic downturn tend not to be implemented until the downturn has reached its bottom and thus they risk overheating the economy by adding public demand when private demand is again growing. This should worry Romer, although because of the continuing credit crunch, continued stimulus expenditures may not in fact overheat the economy.”
The more important consideration is the zero lower bound for the federal funds rate. As long as that’s deemed necessary there’s no chance that a stimulus is counterproductive. The moment that is no longer the case we might consider curtailing or withdrawing the fiscal stimulus because of the crowding out effect. But given the analysis done by Glenn Rudebusch of the FRBSF there’s no chance of that for several years. The bottom line is that a discretionary fiscal stimulus is only necessary and is most effective in a liquidity trap. This is not an ordinary recession.
He writes:
“No one seems to know how much of the $40 billion was actually received by taxpayers, as opposed to reducing their future tax payments; and of the amount actually received by them, no one seems to know how much they spent rather than saved. These distinctions have eluded Professor Chinn, though I had emphasized them in my August 25 blog posting–the target of his criticisms. He seems to think that it is possible that all the $40 billion took the form of rebates, but as far as I know none of it did. A rebate would be a check to the taxpayer, as distinct from a tax credit. Bush’s spring 2008 stimulus package consisted entirely of tax rebates; the Obama stimulus does not.”
According to the Joint Committee on Taxation there are approximately $121 billion in tax expenditures contained in the FY 2009 portion of the Recovery Act. Not all of these tax expenditures will go out in the form of a rebate or decreased withholding. In fact I estimate that only about 60% of it will. Recovery.gov has recently been releasing those figures. You can find them here:
http://www.recovery.gov/?q=content/estimated-625b-tax-relief-recovery-act
So what are the tax expenditures that are going out immediately week after week?
1) Make Work Pay. As of the end of the second quarter $15 billion had gone out as reduced payroll withholding. If Mr. Posner is not aware of this he hasn’t looked at his paycheck stub recently.
2) Other Individual Credits
a) Expanded Tax Break for 2009 First-time Buyers.
This rebate is available immediately as a refundable rebate to taxpayers if you file an amendment to your 2008 Form 1040.
b) Sales Tax Deduction for Vehicle Purchases.
This is available immediately through reduced withholding.
c) Exclude up to $2,400 of unemployment insurance benefits from gross income.
This is available immediately through reduced withholding.
Altogether $8.5 billion in additional individual credits had gone out as a rebate or through decreased withholding by the end of the second quarter.
3) Energy Incentives.
As of the end of the second quarter about $400 million had gone out in the form of tax rebates.
4) Tax Incentives for Business.
The bonus depreciation is available immediately as a rebate and the
carryback of net operating losses is available through decreased withholding. As of the second quarter $14.5 billion had gone out.
5) Cobra.
As of the end of the second quarter $4.9 billion had gone out in the form of decreased withholding.
So in total $43.2 billion had gone out as a rebate or through decreased withholding by the end of the second quarter.
Posner writes:
“Professor Chinn says that even if none of the $40 billion in tax relief was actually spent by recipients of the relief in the second quarter, the remaining $60 billion of Romer’s “more than $100 billion” was ample to increase GDP that quarter by two percent. But that assumes that all of the $60 billion was spent, and no one seems to know how much was; Chinn assumes it all was, which is false.”
Dr. Chinn made no such assumption. His main point was that even if only a small fraction of the money that had gone out by the end of the second quarter had resulted in net new spending this would have resulted in the effect on GDP claimed by Romer and the several private forecasters. But let’s suppose that this assumption were true. That means that without last quarter’s stimulus GDP would have been approximately $3.49 trillion instead of $3.55 trillion. In other words it was approxiamtely 1.7% larger and at an annual rate that would have added approximately 7.1%.
Now the change bewteen the -6.4% decline in GDP in the first quarter and the -1.0% decline in the second quarter was only 5.4%. So if only 60% of the stimulus spending through the second quarter ended up as net new spending it would have accounted for about 130% of the change.
Posner writes:
“A neglected point is brought out in an article by Michael Cooper in the September 5 New York Times. The article points out that federal stimulus spending can be nullified by state cutbacks. For example, a federal grant of stimulus money for mass transit has been nullified by reductions in state expenditures on mass transit. The question then becomes what was the consequence of those reductions? Maybe they enabled a state to rescind a tax increase, in which event state taxpayers would have more money in their pockets. And then the questions would be: how long does this process take, and how much of the additional money do the taxpayers spend rather than save?”
I’ll let Dr. Chinn handle this. He responds:
“he nullification point is an odd assertion. The state cutbacks presumably would have occurred in the absence of the transfers from the Federal government, so in the absence of the stimulus package, state and local spending would have been even lower. In other words, Mr. Posner is asserting a state and local government spending reaction function that implies each dollar of spending transferred from the Federal government induces and equal and offsetting spending reduction. I’m open to this possibility, but I’d appreciate seeing some data (i.e., not anecdotes) to that effect, rather than mere conjecture.
In addition, Mr. Posner sets forth an alternative scenario, where tax increases are rescinded as a consequence of transfers to the states. On this count, CEA has investigated this hypothesis. Figure 7 from The Effects of State Fiscal Relief is illuminating:
http://www.whitehouse.gov/assets/documents/CEA_ARRA_Report_Final.pdf
‘Figure 7 compares the size of total funds from these programs that had been transferred to each state through July 3 with the state’s maximum reported fiscal 2009 budget gap.22 While these transfers were sizeable in magnitude, in 35 states they are smaller than the reported budget gap. This means that states could have used the funds to increase their rainy day funds only if they reduced expenditures or increased taxes in the same fiscal year. Given that states underwent deep expenditure cuts in 2009 (42 states cutting their fiscal year 2009 budgets by $31.6 billion through June 2009), it seems unlikely that this would have occurred.’
In conclusion, it appears to me that we have several pieces of information, from a variety of sources, that suggest a positive impact on 2009Q2 q/q growth stemming from the ARRA. In addition, if one were to do a meta-analysis of estimated impacts, it would seem to me likely that the mean impact would be in range estimated by CEA.”
https://econbrowser.com/archives/2009/09/the_arraas_prog.html
Now in conclusion I want to reiterate some of my own points.
After falling considerably, and progressively more deeply in each of the three quarters before the most recent one, the fall in GDP moderated substantially. After declining at an annual rate of 6.4% in the first quarter of 2009, it fell at a rate of 1% in the second quarter. The rise in GDP growth from the first quarter to the second was the largest in almost a decade, and the second largest in the past quarter century.
Econometric analysis performed by Moody’s economy.com, Macroeconomic Advisors, Goldman Sachs, J.P. Morgan Chase, IHS/Global Insight and the Council of Economic Advisors (CEA) credit the stimulus with adding 2.8%, 2.1%, 2.2%, 3.0%, 2.3% and 2.3% respectively to GDP growth at an annual rate in the second quarter. The average of these six vector autoregression analyses (VAR) is thus about 2.5%. These same private firms and insitutions are predicting that the stimulus will add 3.6%, 1.9%, 3.3%, 4.0%, 2.3% and 2.7% respectively to GDP growth in the third quarter compared to baseline or an average of 3.0%. As a result GDP will be about 1.4% or $200 billion higher at an annual rate in the second quarter thanks to the stimulus. In addition Moody’s, Macroeconomic Advisers, IHS and the CEA using VAR analysis project that the stimulus will raise employment by 1,070,000, 620,000, 690,000 and 1,040,000 respectively or an average of 856,000 in the third quarter over baseline.
As of the second quarter $56.2 billion had been spent directly and another $43.5 billion had gone out immediately as tax expenditures for a total of $99.7 billion. As of the end of August approximately $89 billion had been spent directly and $62 billion had gone out immediately as tax expenditures for a total of $151 billion. This is close to 20% of the stimulus. By the end of this quarter this figure will rise to about $180 billion. From now until the end of FY 2010 the stimulus will go out at the rate of about $100 billion. And its effect is proportional to its flow and not its stock amount.
To have the effect implied by the VAR analyses, GDP merely needed to rise by about 0.6% above baseline in the second quarter (remember we’re talking about annual rates). Since GDP in the second quarter was about $3.55 trillion that means only about $22 billion in net new consumption, investment and government spending needed to be attributable to the stimulus or only about 20% of what actually went out.
How small 20% really is is best shown by considering the details of the Recovery Act. As of the second quarter, %15 billion had gone out in the form of Make Work Pay, $28.2 nillion for other individual tax incentives, energy incentives, tax incentives for business and Cobra, $28.3 billion for state fiscal relief, $14.4 billion for unemployment insurance and food stamps and $5.9 billion for government investment outlays.
Let’s assume that only 15% of Make Work Pay was spent, and the rest was saved despite the fact it only is being paid out in paycheck increments instead of a large rebate check that’s more easily saved. Let’s assume that, somehow, the other tax relief was completely fungible and that it generated absolutely no net new spending despite the fact that a fair proportion of it was conditional on investments. Let’s assume that only 35% of the state fiscal relief was spent despite the fact that almost all states are in the midst of a severe revenue crisis. Let’s assume that only 45% of the unemployment insurance and food stamps was actually spent despite the fact the recipients are probably facing severe liquidity constraints. And let’s assume that only 65% of government investment outlays were spent that quarter, that for some mysterious reason the remainder was parked in a bank account.
With all of those assumptions you come up with a figure of about $22 billion in net new spending in the second quarter. Thus GDP was larger by about 0.62% or by 2.5% at an annual rate. Thus you would get exactly what the econometric analysis is implying.
To achieve the impact estimated by the VAR analyses in the third quarter, GDP must rise by about 1.4% relative to baseline. Thus third quarter GDP must be about $50 billion above baseline. In other words the stimulus must generate a total of about $70 in net new consumption, investment and government spending over the two quarters. This will represent less than 40% of the total amount that has gone out by that point. It is perfectly reasonable that it will have that effect.
Addendum to my previous comment:
In my hypothetical discussion of what was actually spent in the second quarter I left out the $13 billion in Economic Recovery Payments that went out as $250 checks to people receiving Social Security etc. Let’s assume that none of that was spent as well.
The last time I posted something like that it took a full three days before it was posted at the Atlantic. I wonder, was it because they were checking my claims or was it becuase they were waiting for the website traffic to fall.
So much for the ideal of a non-partisan council. But I guess it always was just an ideal.
Two or even three points to second quarter growth is just about plausible, especially if you count the Fed’s monetary stimulus and not just the ARRA.
Three points to third quarter growth is less plausible: we are talking quarter-on-quarter, and there wasn’t much if any acceleration of stimulus spending in the third quarter.
By the fourth quarter we’re already going to transitioning from the phase when stimulus overcomes risk aversion into the phase when it becomes a debilitating addiction that crowds out private investment.
Menzi- you are not correct to assert that state/local transportation would have decreased regardless of the stimulus. GAO has looked at the “substitution” effect in transportation spending in a non-recessionary context and found a clear impact. Not dollar for dollar,but 50-60 percent.
Cheno: I was surprised when I read your statement, because GAO does excellent work (many of our La Follette school grads go on to work at GAO), and your characterization seemed at odds with my understanding of what’s going on now. So I reread the report. Three observations: First, the 50-60 percent refers to the 1983-2000 period, and is retrospective, rather than current. Second, the use of 0.5 instead of unity in my calculation does not alter the end result for the impact in 2009Q2. Third, in the current situation, it is unclear whether the historical rate of substitution is relevant.
This means that the standard mechanism for substitution is unlikely to be operative to the same degree as over the 1983-2000 (essentially nonrecessionary period, excepting 1990Q2-91q1). In contrast to the 1983-2000 period, given the severe fiscal constraints facing the states, I think the amount that state governments intend to spend will not typically exceed “the amount required to meet federal matching requirements” (p.25) that allowed substitution in the past.
Tom: You state:
I don’t believe you have a strong grasp of what the CEA does. See Feldstein’s article.
While econometrics may help us understand where we are now (like knowing the prices in the stock markets), how does it help us know the future? Indeed what is the difference between finding patterns through econometrics and Technical Analysis in stock trading? If there is a pattern and it becomes widely known wouldn’t it be arbitraged away by homo economicus?
Hey Mark, I got a better version of your chart for you:
http://3.bp.blogspot.com/_pMscxxELHEg/SlEl3AH3cGI/AAAAAAAAFww/U_K7psgAl1k/s1600-h/RecoveryJune2009.jpg
source: http://www.calculatedriskblog.com/2009/07/second-stimulus-plan.html
Menzie, I’m afraid the first page of the Feldstein article, the one available to most, reinforces Tom’s point of a “of a non-partisan” council. Because it was written by a Republican appointee, I wonder if it has any significance for a Democratic appointee.
Does anyone ever stop to consider that after Nixon said “we are all Keynesians now” we had 10-12 years of malaise (to quote Jimmy Carter) while after Reagan’s “voodoo” supply side economics we had the “7 fat years” that turned into almost 20 years of prosperity. Then consider that Keynes is now ruling again in economic circles and we are facing the worst economic decline since the Great Depression. Hmmmmm!
I am amazed that the economists who engineered this disaster now point out that they added GDP even though growth is negative and they have added jobs even though we are at 9.7% unemployment. Get real! At least Christina Romer admits that her claims can’t really be proven.
DickF, do you have a link to the Romer claim?
CoRev: If you’re having trouble seeing the entire article, here’s a slightly longer excerpt. I’m using the term “non-partisan” in the sense that CBO is non-partisan, i.e., not affiliated with a party. The three members of the CEA are appointed by the President, confirmed by the Senate.
CoRev,
The Job Impact of the American Recovery and Reinvestment Plan
January 9, 2009
It should be understood that all of the estimates presented in this memo are subject to significant margins of error.
http://otrans.3cdn.net/45593e8ecbd339d074_l3m6bt1te.pdf
THE ECONOMIC IMPACT OF THE
AMERICAN RECOVERY AND REINVESTMENT ACT OF 2009
FIRST QUARTERLY REPORT
September 10, 2009
Evaluating the impact of countercyclical macroeconomic policy is inherently difficult because we do not observe what would have happened to the economy in the absence of policy.
http://www.whitehouse.gov/administration/eop/cea/Economic-Impact/
Thanks DickF. Menzie, that’s a pretty thin definition of “non-partisan”. I wish CBO was truly non-partisan, but we both know that is not absolutely true. Partially/minimally partisan or only slightly influenced may be closer to the truth.
BTW, any analysis of the efficacy of a policy that is written by the policy makers is and should always be looked at with some skepticism.
Nice work, Menzie.
It’s very important that you keep providing actual facts and analysis in a debate that too often is dominated by insults and guesses.
The discussion is enhanced by the style and content of your blog.
CoRev: I think CBO is as non-partisan (in the sense you’re using) a policy agency/organization as you get in American society today.
Erik Brynjolfsson: Thanks for the words of encouragement!
Menzie wrote:
CoRev: I think CBO is as non-partisan (in the sense you’re using) a policy agency/organization as you get in American society today.
Menzie,
I think you are right here but the problem with the CBO is that they are required to use static anlysis based on the laws as they are today. That means that they cannot build into their estimates any considerations of growth or contraction. Nor can they build in the probable tax changes such as the lapse of the Bush tax cuts. Because of this their projections often have significant variances from reality.
DickF: You might remember this post (since you commented on it repeatedly) where I excerpted CBO analysis of deficits using dynamic models. I observed at the time: “What is remarkable is how little the textbook supply side model response, with high responsiveness to tax changes, differs from the CBO’s baseline.” So I disagree with your characterization.
DickF wrote, “Does anyone ever stop to consider that after Nixon said “we are all Keynesians now” we had 10-12 years of malaise (to quote Jimmy Carter) while after Reagan’s “voodoo” supply side economics we had the “7 fat years” that turned into almost 20 years of prosperity.”
Are you saying that appropriating taxpayer money to Acorn in order that they might hire more community agitators does not have the same beneficial effect on the economy as cutting taxes? Heresy.
So if I understand correctly, if the government spent a trillion dollars in a quarter, GDP would go up, right? Well, then, why not borrow a trillion dollars every quarter to boost GDP? In no time at all, we’ll be rich!
Or are we missing something? If you put these measures before an accountant, he would have a heart attack and die. You’re presenting, if I understand correctly, single-entry accounting. There is absolutely no measure of the associated cost of such government spending. In this presentation, the money is free.
Let’s suppose you tried this in a corporate setting: “Well, let’s borrow some money, call it revenues, book the associated profits, distribute it as dividends, and our stock price will soar on the momentum!” Well, then you would be…Enron. Quite literally.
The approach you presents suggests you are indeed an egalitarian. Egalitarianism is built on the decreasing marginal utility of wealth and income (an incredibly powerful tool, by the way, for understanding social progress). An egalitarian will say, “Well, if we have a rich guy and a poor guy in a room, and we have decreasing marginal utility of wealth (assume no income), then if both guys have exactly the same wealth then social well-being will be maximized.” Mathematically speaking, this has to be true.
However, if this analysis is run dynamically over time, you will gut your savings and investment, because, by definition, the poorer guy will have a higher propensity to consume, and therefore the savings and investment in the system decrease, and of course, the richer guy looses the incentive to produce or flees the system. (Hence the Berlin Wall.)
Now, you have presented a static analysis here. You highlight that govt spending increases the growth rate, or more precisely, the level of GDP for the period of the spending. That’s a good thing. End of story.
If you were a liberal, your analysis would not stop there. Your analysis would consider returns over time based on i) multipliers + ii) incremental marginal utility of income for the down-and-out* divided by iii) the cost of capital. So, you would attempt to show that the capital borrowed for the stimulus would have a higher social return than if retained in private hands, and you would marshall some sort of multiplier argument and measurement of the need to help displaced persons. The measure of return would be something like the Recessionary Return on Capital, or some such thing.
But as you have not done that, the evidence suggests that you are indeed an egalitarian. The fact that you teach at Wisconsin and perpetually bash ex-President Bush has nothing to do with it.
* technically, this would be a socially conservative bit of analysis
Thanks for the illuminating post. I also think the CEA work is well-done. HOWEVER, the apparently universal obsession with the date at which the Feds disburse funds deserves comment.
State and local governments can anticipate, borrow, and lend. If they expect funds previously unexpected, they can adjust in-go and out-go prior to the receipt of funds, or well after. What’s really required is comparing observed outlays to a counterfactual.
The comparison of funds transferred to budget gaps is worth doing, but it is apples to oranges in one respect. The gaps (which by the way are just state governments, do not include locals, so are understated) are prospective and for an entire fiscal year. The funds transferred thus far are less than the total promised.
The rainy-day funds are disproportionately held by just two states — Texas and Alaska. This increases the likelihood of an impact from the program.
The good news is that most of the funds have an impact because of the bad news that their level is inadequate — they in all likelihood fail to exceed the extent of fiscal retrenchment which was in process before the launch of the program.
(Keep those LaFollette grads coming. We love ’em.)
Posted by: Mark A. Sadowski:
“Seemingly oblivious to the inherent contradiction. In short, he’s saying the Recovery Act has led to the sudden improvement in the second quarter, not due to any actual spending, but due to its psychological effect, presumably in the form of increased optimism about the economy. But then he goes and points out that the Recovery Act is unpopular. How is it possible for a program to both increase economic optimism and yet generate antipathy?”
I’m inclined to think Posner is confusing the unpopularity of some of the President’s other spending and economic proposals with that of the ARRA. From public opinion polls, I see no change in support for the stimulus plan. I do see this though:
link
“And please tell me if you approve, disapprove or neither approve nor disapprove of the way Barack Obama is handling each of the following issues:
The economy:
9/3-8/09
Approve 44%
Disapprove 52%
4/16-20/09
Approve 58%
Disapprove 35%
Federal budget deficit:
9/3-8/09
Approve 33%
Disapprove 56%
4/16-20/09
Approve 49%
Disapprove 41%
The above is a Roper poll, but a recent ABC News/Washington Post poll 9/10-9/12 produces a similar result.
As for consumer confidence:
2009-02-01 56.3
2009-03-01 57.3
2009-04-01 65.1
2009-05-01 68.7
2009-06-01 70.8
2009-07-01 66.0
2009-08-01 65.7
2009-09-01 70.4
The U. of Michigan survey suggests confidence did surge at the very beginning of Q2. It’s hard to say though whether that occurred before the actual spending kicked in in a big way.
Interesting though that there hasn’t been much improvement in confidence with the improvement in other data since May.
@acerimusdux,
I don’t think Posner bothered to check the actual polling data any more than he bothered to check if any of his other claims were plausible. If he did that then he would be unable to raise doubt in the guilt free manner as he so frequently does. It’s an old lawyer’s confidence trick.
But I think it’s pretty evident he is contradicting himself, and he really doesn’t care if he is. His main goal in that article was to raise doubt about the tangible (nonpsychological) benefits of stimulus while impuning Romer’s (and to a lesser degree Biden’s) ethics. While this sort of logical incoherence would be frowned on in a purely academic setting it is seemingly de riguer in the legal world.
Truthfully, I haven’t been following the stimulus polling closely. The most recent poll I’ve seen on the stimulus is from ABC. 33% think it is helping the national economy vs 19% who think it is hurting it. That evidently leaves 48% with “no opinion or I don’t know.” So, based on this single poll, it’s hard to say what the general public really thinks about it.
http://www.pollster.com/blogs/us_economy_abcpost_91012.php
And I think you’re right, Posner is probably conflating all of the polling on the government’s various recovery policies. But I also suspect that the general public is having trouble distinguishing the various recovery policies themselves so I don’t think it’s even clear what the polls specifically about ARRA really mean.
However, it is very interesting to see the rise in consumer confidence just as the economy’s death spiral paused and just as (coincidentally?) ARRA kicked in.