# Fiscal stimulus

My colleague UCSD Professor Valerie Ramey has an interesting new paper looking at the effects of higher government spending on GDP.

Ramey (2012) approaches the question from a forecasting perspective. Suppose a certain event (examples of which are detailed below) causes you to revise your forecast of how high government spending is going to be over the next few years. How would this news cause you to change your forecast of how high private GDP (that is, all the components of GDP other than government spending) is going to be? If your prediction of private GDP goes up, that is evidence consistent with a fiscal multiplier greater than one– added government spending not only contributes directly to GDP from the accounting identity, but also helps boost private spending as well. If private GDP goes down, that suggests a multiplier less than one.

The forecasting models she looks at are vector autoregressions, in which one tries to predict a set of variables such as the log of real government spending per capita, log of real private GDP per capita, the marginal tax rate, and the 3-month T-bill rate based on what all of those 4 variables have been doing over the last year. In her simplest exercise, Valerie looked at how the forecasts of each of the 4 variables would change if real spending this quarter comes in higher than you would have expected according to the model. The top panels in the figure below are based on forecasting relations using data all the way back to 1939. The graph in the left panel shows how the model’s k-quarter-ahead forecast of real government spending would change in response to news of higher government spending at time 0, plotted as a function of k, how far into the future you’re looking. (It will help economist readers, but perhaps not anyone else, if I were to describe this as an impulse-response function based on a Cholesky factorization with government spending ordered first as in Blanchard and Perotti (2002)). Given the positive serial correlation in government spending, if you learn spending is about 0.4% higher this quarter, you’d expect further spending increases over the next year, with the graph normalized such that the news causes you to expect 1% higher real spending per person 4 quarters following the original information.

The right top panel shows how the model’s forecast of future real GDP per person excluding government spending would change in response to the news. The model predicts that private spending will be 0.7% lower after a year. This negative effect is statistically significant.

##### Revision in forecast k quarters into the future following an unanticipated increase in real government spending per person at time 0. Left panels: change in forecast (in percentage points) of real government spending per person. Right panels: change in forecast (in percentage points) of real GDP less real government spending per person. Estimates based on sample period indicated at top, 95% confidence regions in gray. Source: Ramey (2012).

In part this inference is based on what happened during World War II. Between 1941 and 1944, real government spending increased by \$75 billion (in 1937 dollars), but real GDP only rose by \$60 billion. That’s consistent with the patterns above, in which private spending falls in response to higher government spending. One might argue that there were other special factors such as rationing that reduced private GDP at the time. The second row in the figure above leaves out the World War II data, and just bases the inference on what we saw over 1947-2008. The effects are similar to those found using the full sample. The last panel of the above figure leaves out both World War II and Korea. With less data, the inferences are less reliable, but the overall estimates remain quite similar to those for the full sample.

Valerie’s paper explores a number of other possible ways one could form the forecasting question. One concern is whether government spending rose in response to some other events that might have had direct effects on the economy, in which case the revision in forecasts might represent the consequences of those events instead of an effect of government spending itself. The next graph uses an alternative idea proposed by Perlotti (2011) of specifying the news variable in terms of defense spending alone rather than overall government spending (this is now a 5-variable VAR with defense spending ordered first). Results are essentially the same as for the first exercise.

##### Revision in forecast k quarters into the future following an unanticipated increase in defense spending at time 0. Left panels: change in forecast (in percentage points) of real government spending per person. Right panels: change in forecast (in percentage points) of real GDP less real government spending per person. Estimates based on sample period indicated at top, 95% confidence regions in gray. Source: Ramey (2012).

In yet another approach, Valerie used a news series constructed in Ramey (2011) that is based on reading of Business Week and other historical sources to construct a series of changes in the expected present
discounted value of government spending caused by military events. Although the effects on private GDP are not measured as precisely using this indicator, the overall inference confirms the view that higher government spending raises GDP by less than the spending itself.

##### Revision in forecast k quarters into the future following an unanticipated increase in future defense spending as measured by Ramey (2011). Left panels: change in forecast (in percentage points) of real government spending per person. Right panels: change in forecast (in percentage points) of real GDP less real government spending per person. Estimates based on sample period indicated at top, 95% confidence regions in gray. Source: Ramey (2012).

Ramey (2012) concludes:

Using a variety of
identification methods and samples, I find that in most cases private spending falls
significantly in response to an increase in government spending. These results
imply that the average GDP multiplier lies below unity.

## 50 thoughts on “Fiscal stimulus”

1. CoRev

“These results imply that the average GDP multiplier lies below unity.” OMG, how many heads are exploding?

2. 2slugbaits

JDH Thanks.
I’m not quite sure what the takeaway is supposed to be. I don’t think anyone has seriously claimed that increasing fiscal spending always and everywhere increases private employment. Clearly, way most of the time government spending crowds out private sector spending. And the Fed is pretty effective at making sure that happens. That’s one reason Krugman et al. were so upset about Bush’s increased military spending when the economy was at something like full-employment and the fact that this spending was deficit financed. The interesting economic and econometric question is whether fiscal policy increases private sector employment when the Fed is at the zero lower bound and cannot effectively sterilize the effect of increased fiscal spending. That’s what today’s argument is all about and I worry that some folks (read Sean Hannity and Rush Limbaugh) will misapply the paper’s conclusions.
Some specific comments regarding the paper:
(1) Ramey uses four lags in her VARs; however, it appears that all of the lags are against levels data. So if the variables are I(1) stationary, then she is really only using 3 lags, which could be a problem given that she uses quarterly data.
(2) Rather than just saying that she used 4 lags because that’s standard, I would have preferred to see some information criteria to support that statement. Not saying something like an AIC or BIC wouldn’t have recommended 4 lags, only that all econometric software provides that information, so why not report it.
(3) It was unclear to me if she was using one large VAR with 4 or 5 variables, or if she was using 4 or 5 simple bivariate VARs. If the former, then we would want to know the ordering and the var-covar matrix since this was done using a Cholesky decomposition.
(4) Part of government spending would include interest payments to the private sector or to foreign entities. Presumaly those would lag government spending if spending were deficit financed. That’s extra income for someone, so you would want to consider the MPCs and tax elasticities of those who receive most of the interest payments. I’m not sure that using average tax rates really corrects for this.
(5) She defines private sector spending as Y-G. Well, okay…but it’s not like the X – M = NX component of Y is completely independent of how G finances spending.
(6) Was there any interesting information in the variance decompositions? They were not shown in the paper.

3. Anonymous

I find that in most cases private spending falls significantly in response to an increase in government spending.
I don’t think many Keynesians would dispute that during normal times, but the exception would be during a depression when aggregate demand is severely depressed, capital utilization is low and unemployment is very high.

4. arthur

But we are not interested in the *average* GDP multiplier over the business cycle. We are interested in the multiplier in a depression.
Isn’t an average GDP multiplier below unity perfectly consistent with a near-zero multiplier during a boom and an above unity (or even a 2+) multiplier during a depression?

5. arthur

cont from prev comment …
Not to forget the ZLB.
I must say I share 2slugbaits’s puzzlement. The paper seems to be asking a largely irrelevant question, and getting an answer that is likely to be misinterpreted, quite possibly deliberately.

6. Rick Stryker

A number of commenters have questioned the relevance of Ramey’s article given the zero lower bound. But Ramey does explicitly test the multiplier during 1939-49 when interest rates where near zero and concludes that there in no evidence for new Keynesian arguments for stimulus spending. She did that by comparing the multiplier implied over that period to the multiplier implied by the same VAR over 1939 – 2008. The multiplier over 1939-1949 was actually lower than over the longer sample. Commenters need to acknowledge that. No doubt these same commenters will dismiss Ramey’s results by saying these WWII years are not representative since the U.S. economy had shifted to war production and was command and control. That’s how Krugman dismissed Barro’s results. But rather than just dismiss results without argument, Krugman and his acolytes need to provide evidence of their own, something they’ve failed to do so far.

7. 2slugbaits

Rich Stryker Wouldn’t you expect 1939-1949 to have a lower multiplier? During most of that period private spending was rationed.
And it sounds like you misunderstand the zero lower bound. The zero lower bound does not just mean the short term rate is very low; it means the short term rate is very low and the Wicksell natural rate is negative. In terms of old school Keynesian models it means the LM curve is flat. No one would argue that the LM curve in the 1940s was flat. And no one would argue that the Wicksellian natural rate was negative.
Ramey’s VAR did not have any threshold parameter, so presumably a reduction in government spending would show up as an increase in private spending. And in normal times that is exactly what we woud expect. But how’s that austerity thing working out in Spain? Or Italy? Or Portugal? Or Britain? Or Ireland? Ramey’s conclusion pretty much matches that of just about any New Keynesian model. The relevant question today is whether a New Keynesian model with all the nice DSGE flourishes makes sense in the current environment. I didn’t see anything in Ramey’s paper that would support an Austerian view of fiscal policy in today’s economic environment. Remember, the last observation is 2008:Q4, which means that after adjusting for VAR lags the model effectively missed the Great Recession
One other thing about the graphs in Ramey’s paper. It’s hard to tell one way or another, but visually it’s not obvious that all of the Impulse/Response charts actually converge to some long run stationary position. To be sure, most do, but there are a few that look like they might just keep growing.

8. jonathan

It’s an interesting approach, but I’m not sure it says what people hope it says.
As I read the paper, the question in my head was “is this really what people mean when they speak in general terms about deficit or even tax-financed spending?” It isn’t how I think of it. Spending comes up in two general cases: normal times and bad times. These are very different things. In normal times, the cost of government is part of the cost of maintaining society; it does things – like education, like welfare spending, like aid to sick children – which are part of what makes us a country. While we could imagine removing those things – just eliminate Medicare – that would change the country and all the underlying assumptions in the model. If the belief is that course would generate greater prosperity, I suggest looking at the rest of the world: we could be like Columbia or Nigeria, where there is wealth and squalor, or maybe like Venezuela where control of the economy (and society) by a group of a few hundred families led to Chavez. The assumption is we’d still look exactly like us except richer, but where are the examples of this anywhere in the world? They don’t exist.
In the normal case, if government spending actually increased total GDP beyond 1 to 1, we could spend and tax our way to nirvana, always making more, sort of the economic equivalent of a perpetual motion machine. The results reported make absolute intuitive sense. But does that matter? If the multiplier is negative, then we see the cost of government can be estimated at this level. Does anyone believe there is no cost?
Multipliers usually come up in the case of recession or crisis. I didn’t see any discussion of what private spending or total GDP would have been without government spending increases. If there is a multiplier effect, it would be in the gap between what would have happened and what did. That is really hard to do.

9. ppcm

This paper carries a substantial content that is driving peripheral thoughts about well accepted theories and induced empirical data.
May be Ricardo (Keynes well known critic) will rejoice through the content and findings, where tax material burden and anticipation for further taxes may occur when government spending are lifted. Hayek will be prompt to find few government misallocation of expenditures. At risk of over simplifying, there are variables such as trends anticipations and there are data stocks. Is it possible to increase the number of variables within the paper to the extent of debts burden (private and public) and their impacts on growth, their impacts on the rate of return of governmental expenditures and private investments. Which factors carry the most weighing and when?
Few papers may assist in the research.
“This time is different” is dealing with the debt stocks and repayment hurdles when “The real effects of debts” (BIS publication) is dealing with the debts dynamics private and public sectors and their marginal decreasing contribution to GDP growth.
In economist language, what Granger causes, marginal decreasing returns,where are the pivot points where are the switching points private expenditures versus government. As always increasing debts, should in theory lift up interest rates as risks preventive measures. If they do not, where is the turning point of decreasing marginal return?
It would be educative to understand how is the +beta3Newst quantified within the Barro and Redlick formula

10. bakho

No one seems to be focusing on this conclusion:
“the results suggest that direct hiring of workers by the government may be more effective than relying on multiplier effects of government purchases.”
Government employment (if states and locals are included) has declined since 2009.
Government stimulus is really “What stimulus” because the money spent by the Federal Government is largely cancelled by cutbacks at the State and Local levels.
Government spending is “leaky” in a global economy. Buying stuff may put people to work in China but not the US. Hiring Americans to work on American infrastructure, education and health ensures that all spending goes to unemployment.
Our biggest economic problem today is not GDP. It is rising at a reasonable rate. The GDP gap is not closing, meaning the GDP could rise at a faster rate. The Biggest problem is unemployment and the damaging social effects of long term unemployment. These effects have HIGH negative social costs and should not be tolerated.
The paper makes the conclusion that BigG can easily reduce unemployment by simply hiring workers to do what needs to be done. So why is this NOT the focus of discussion?

11. aaron

2sb, I’m not sure anyone has said austerity here.
The implication is simply to cut spending.
None of which has be done significantly (at all?) by most those countries.
You are also forgetting the flip side of the implication. Taking more money from the private sector decreases private sector growth.
The implication isn’t austerity, it’s cut spending.

12. Brian

It seems to me that when thinking about stimulus, you need to consider two regimes: when interest rates are at the zero bound and when they’re not. I don’t think anyone nowadays thinks that fiscal policy should replace monetary policy, when traditional monetary policy tools can be used. The interesting question is whether fiscal policy can play a stimulative role once all the monetary instruments have been played (ban pun). I don’t see how Ramey’s VAR’s distinguish between these two regimes, since there are no dummy variables or fancier threshold effects in her models to control for which of the two regimes is in effect at any given time.

13. aaron

bakho, “the results suggest that direct hiring of workers by the government may be more effective than relying on multiplier effects of government purchases.”
This isn’t very important since it’s found that the effect on emploment is temporary and not stimulative to the economy. To keep the employement rate from falling, the spending has to continue, and the spending decreases growth.
Unless lower unemployment is going magically shock “animal spririts” awake, it is not a good metric.
She is simply stating that if your goal is fudging employement numbers, direct government hiring is the way to go. However, doing so is still a bad idea.

14. JDH

arthur and others: Note that Ramey’s forecasting model includes past values of private GDP, and so the forecast does take into account whether the economy is weak or strong. However, the specification implies that the incremental contribution of government spending to the forecast does not depend on whether the economy is weak or strong. This is an inherent implication of any linear forecasting model. To have any other implication, one would need to build a nonlinear forecasting model.

Linear models are the first choice of most researchers, regardless of the question being asked. The reason is very simple– it takes a lot more data to detect nonlinear relations, and there are all kinds of possible forms nonlinearity could take. Moreover, to detect any relation at all, one needs variation in the explanatory variables, and the biggest variation historically came from wartime spending. I think Ramey’s research may inspire others to look for possible nonlinearities here, but it is not at all clear they would find anything. Summarizing the average effect is not a conspiracy to mislead. Instead, it is the key summary statistic that a researcher can accurately report.

One thing we can do is look for weak evidence of differences in the multiplier in 1939 compared with later– see Rick Stryker’s comments above. I think many economists do believe the situation today has much in common with that in 1939.

15. tj

Maybe this section from page 13 will help clarify the point JDH makes above.
Because both government spending and taxes are potentially affected by the state
of the economy, which also impacts private spending, we would expect these fiscal
variables and tax rates to be correlated with the error term . Thus, estimation calls
for instrumental variables. A natural instrument for the tax variable is the Romer and
Romer (2010) narrative series on exogenous tax changes. This variable calculates the
annualized change in tax liabilities due to legislation based on either deficit concerns
or long-run growth promotion. Thus, the identification assumption I am making is
that the tax legislation changes affect the economy only through changes in tax rates.
Because the Romer-Romer tax instruments are available only from 1945 to 2007, the
estimation must exclude the World War II sample. For government spending, I use lags
of my news variable as an instrument. Because the current value of news is an included
variable, my identification assumption is that while current news can independently
affect private spending, lagged values of news affect the economy only through current
changes in government spending. This assumption might be questionable if there are
additional lags in the effects. Thus, I will assess the robustness of the results to adding
lags of spending growth, government spending, and taxes to the specification. Using
the period 1948 to 2007, I explored various lags of the two instruments up to 12 lags.
I use four lags of each instrument since this number of lags maximized the Cragg and
Donald (1993) statistic.

16. Dave

Interesting. Besides the issue if the economy is in recession I think type of spending needs to be considered. Military spending is often outside of the country and undermines good will while not training for the private sector. we need teachers not troops.

17. Ricardo

Professor,
I am actually surprised. I would have expected that initially increased government spending would generate increased private sector activity in response to the artificial boom.
The graphs are curious when you consider weighting of the gold exchange standard. In an eyeball comparison there is a slight spike immediately after spending stimulus in the 1954-2008 graphs that is less evident in graphs including more gold standard years. I would expect this spike to be even more pronounced if only fiat currency periods were analyzed.
I expect that money supply may have skewed the data. I would be curious what the results would be if the data were adjusted for increases in money supply and during a period without an anchor to the currency.

18. Joseph

So does this study also imply the inverse? If the government cuts spending, does GDP go up? Austerity, here we come! Because it is working so well in Europe (and Wisconsin).

19. JDH

Joseph: The claim is that when government spending goes up, then GDP goes up, but by less than the increase in government spending.

20. Joseph

So then the inverse would be that cuts in spending cause GDP to decrease by more that the amount cut.

21. JDH

Joseph: No, when spending goes down, GDP goes down, but by less than the budget cuts.

22. arthur

@JDH
Thanks for the reply and the clarifications.
“…the specification implies that the incremental contribution of government spending to the forecast does not depend on whether the economy is weak or strong. This is an inherent implication of any linear forecasting model”
Your point about the paucity of data and the problems with non-linear models is well taken, but does this not amount to assuming a large part of the conclusion that the size of the multiplier does not vary with the state of the economy? Is this assumption justified?
“I think many economists do believe the situation today has much in common with that in 1939.”
1939: Yes.
1942: No
“Traditional Keynesian multipliers assume that there are no capacity constraints to impede a fiscal-driven expansion in aggregate demand. On the contrary, we find ample evidence of capacity constraints in 1941, particularly in the second half of that year. As a result our preferred government spending multiplier is 1.80 when the time period ends in 1941:Q2 but only 0.88 when the time period ends in supply-constrained 1941:Q4. Only the 1.80 multiplier is relevant to situations like 2009-10 when capacity constraints are absent across the economy.”
The End of the Great Depression 1939-41: Policy Contributions and Fiscal Multipliers
Robert J. Gordon and Robert Krenn
NBER Working Paper No. 16380
September 2010

23. Anonymous

Looking at the graphs, I don’t see how the author can conclude “that in most cases private spending falls significantly in response to an increase in government spending.” To the contrary, private spending increases significantly, after a few quarters of delay.
What am I missing?

24. DS

As a lay person to this debate, I have a number of questions:
Is the way she calculates private sector spending a common and accepted method of doing so, or did she invent it?
She indicates that private spending includes consumption + investment + net exports. It would be interesting to know how each of these components are affected. One would expect, that government spending would have little effect on exports.
Am I to understand that in the field of economics, there are currently several competing theories about how to measure the output multipliers and that you basically get a positive or negative result based on which one you use? She mentions that SVARs tend to find rises in consumption whereas the method she typically uses, EVARs tend to find falls in consumption and then using cross sections of states finds very positive multipliers. The disagreement between the measures is very large.
And if we don’t agree on how to measure the effects, how in the world can we really say anything about any of this?
The paper actually seems to find a negative multiplier only in the long term. She admits a short-term bounce:
“In the 1939-2008 sample, private spending rises slightly on impact,
but then falls significantly below zero, troughing at around 0.5 percent of GDP.
In the 1947-2008 sample, private spending rises significantly on impact, to about 0.5
percent of GDP, but then falls below zero within a few quarters. These results are consistent
with the effects of anticipations discussed in the theoretical section of Ramey
(2009b). As that paper showed, in a simple neoclassical model, news about future
increases in government spending lead output to rise immediately, even though government
spending does not rise for several quarters. Thus, the theory predicts that
private output should jump on impact and then fall.” p.10
She posits that the initial bounce is due to other factors, but doesn’t provide any evidence that this was in fact the case.
And a short term bounce is what we expect from a stimulus particularly because the spending is normally short term itself and expected to be short term.
Finally, in talking about how government crowds out the private sector, one of the specific mechanisms she dicusses for how this occurs is that government makes fewer workers available to the private sector. But this makes the point made by others about whether her conclusions apply to the situation of a downturn very relevent since in a downturn like we’ve just experienced, there is a significant excess of available labor. (p. 6)

25. Robert Keyfitz

But, suppose the government increases public spending when it correctly anticipates that private spending is about to fall. Or, suppose higher public sector capacity utilization triggers a public investment cycle which lags the business cycle. In neither case would a negative correlation bewteen public and private spending imply anything about the size of multipliers or the productivity of government spending. Prof. Ramey will need a more fully articulated model to draw policy conclusions from her findings.

26. 2slugbaits

JDH I think that you and Joseph were really talking about two different things. I believe that what Joseph actually meant to say was that if an increase in government spending increases GDP, but reduces private spending, then doesn’t it follow that a decrease in government spending should increase private spending? And is this really plausible in the current environment? Does this match what we’re seeing in Europe? I don’t think so.
I don’t know why economists would think that today’s environment was a lot like 1939. According to the BEA data the private sector spending (i.e., Y- G) in 1939 right up to 1942 was growing at ~8% per year. If that’s happening today, then I guess I missed it.
As I read the paper the conclusion is that in two of the three sample periods an increase in government spending reduces private spending. You said, “In part this inference is based on what happened during World War II.” I think I would say that it was more than just “in part.” In fact, it’s almost entirely due to WWII and the effect of rationing. As Ramey notes, even during the Korea period consumers were worried about rationing during the Korean War. It’s very hard for me to understand how the pre-1954 data is at all relevant. Clearly the data shows that the more the data looks like recent history the less the conclusions from samples 1 & 2 hold. So why include pre-Korea data? I hope it’s a better reason than simply pleading that there wasn’t enough variation in the post-Korea data.
I don’t have any problems with Ramey’s conclusion regarding the third sample horizon. That is exactly what I would have expected. The sample stops at 2008:Q4, which means that there really isn’t any time period covered by the zero lower bound. Given that we would fully expect the Fed to adopt a monetary policy to sterilize fiscal stimulus, her results pass the smell test.
One curious note. While there was a spike in consumer demand for durables in 1950:Q3 with the outbreak of the Korea War, the big change in Y-G happens in 1950:Q4 due to a build up in inventories as businesses overshoot expected demand. Inventories then become noisy for several quarters. But if GDP increased because of a build-up in inventories, this is really saying that there was a mismatch between planned private spending and actual spending. Is that what Ramey really want’s to model???
Finally, the claim that news shocks might be able to predict defense spending might be true in the aggregate, but I don’t think there’s any micro foundation for this claim. So is the correlation just spurious? Defense contractors need to know more than just some fuzzy feeling that defense dollars will increase. In order to be actionable they need to know the effect on specific commodity types (e.g., aircraft, ground systems, etc.). And I’m pretty sure no one at the Pentagon knows that answer. Ask the G8 folks in the E-Ring and you’ll get one answer. Ask the G3 folks in the C-Ring and you’ll get another. And the answers won’t even be close at the commodity level even though they might be very close at the aggregate level.

27. Joseph

A couple of cause and effect possibilities:
1)Under normal conditions, whenever government spending goes up, the Fed attempts to counteract it by raising interest rates and slowing private spending.
2)During a recession, government spending on unemployment and welfare goes up as a result of private spending going down.

28. don

I wonder what your colleague (Menzie) thinks of the paper. I don’t think an econometric study can have much useful to say about fiscal policy during a liquidity trap – simply not enough degrees of freedom (unless it includes obervations across countries).

29. jonathan

I should have read the post before reading the paper. The post is much clearer that this is a forecasting exercise that applies different types of spending “shocks”. It’s a neat exercise. Having also read her 2011 paper now, I see better where she’s coming from.
A few humorous notes:
1. The paper doesn’t refer to “spending” but to temporary increases in spending that it creates and then models. It doesn’t speak to the welfare function of government. By welfare, I mean a variety of things. For example, we have a lot of prisons. (A lot more than I think makes sense.) We could get rid of them and save the money but our social imperative seems to be that the cost to society of incarceration is worth it.
2. Speaking of “cost to society”, the big “loser” is in many ways defense spending. Want to reduce private spending? Increase military spending. The military budget is now about as big as all discretionary spending and the latter includes things like Veterans Affairs that tie to defense.

30. aaron

2sb, yes, it does follow that a decrease in gov spending will be followed by an increase in private spending.
Yes, not only is it plausible, it’s highly likely.
Do you think that if Greece’s gov spent more, the private sector would grow more?

31. 2slugbaits

aaron I believe that in ordinary times central banks work overtime to ensure that fiscal stimulus is largely dampened and this can very often lead to a private multiplier less than one. My reading of Ramey’s paper is that this is what happens most of the time. The third series shows that on average government spendng has no effect on private spending over the long run. And it shouldn’t. The only exceptions were when the government rationed during WWII (and that was a function of rationing, not govt spending), and the Korean War period. But I believe Ramey misidentifies the reason…it wasn’t the Korean War spending, but Marshall Plan and Truman Doctrine spending that had the effect of reducing private spending.
Greece doesn’t have a government, so I’m not quite sure what it means to talk about government spending over there. Normally governments can tax citizens. What Greece has is an ECB credit line and an unfinished civil war waiting in the wings.

32. aaron

It would be interesting to see if the fed does tighten when spending increases in good times.
We know the fed does ease for spending in down times though.

33. kharris

Late to the party, but just gather up a handful of points from comments that seem to be important to understanding the results…
“Too few degrees of freedom” seems quite true, given that the zero bound issue is the important one now, and for a number of private credit markets, today’s rates are lower than during WWII. Mortgage rates are at historic lows.
Since we assume that the government is competing resources away from the private sector through spending, slack labor markets and low borrowing rates mean today’s conditions are very different from most of the WWII period that was the “zero rate” period in the sample. Government can have resources now without competing them away from the private sector, so the mechanism by which government spending is thought to reduce private spending is not at work.
Hamilton notes the weakness of linear models. Marry that up with the questionable example of the earlier period of zero rates, and we have considerable reason to doubt that this study tells us much about economic behavior at the extremes. The current period is not so far outside the time of the sample, but current conditions are rather far outside of the sample, aren’t they?
“the results suggest that direct hiring of workers by the government may be more effective than relying on multiplier effects of government purchases.”
Contrary to aaron’s claim, direct hiring of government employees increases spending growth. The results of the study are that overall growth rises with government spending, though private spending falls. aaron has substituted “spending” for “private spending” which is commonly done at AEI and Heritage, but private spending is merely one category of spending. (Note that aaron crawls right on top of that AEI/Heritage view when he calls government hiring “fudging employment”. That’s the Amity Shlaes argument that employment didn’t rise in response to FDR’s policies because employment didn’t rise if you leave out all the jobs created by FDR’s policies.
When aaron goes on to say lower unemployment is not a “good metric”, he doesn’t indicate what it is meant to measure. If human welfare is the goal, then lower unemployment would seem to be a useful metric. If a particular number in the national accounts, say private spending, is the goal, then unemployment may be a useful metric, but we can’t know from what is presented here. So when aaron declares that public hiring is “still a bad idea”, he does so with no actual support for that view. When aaron sees “no reason the multiplier would be higher now” is seems to come down to seeing what he wants to see. That isn’t much of an argument.

34. acerimusdux

First, I don’t see where she says anything about VARs for the period of 1939-1949. She neither explicitly or implicitly suggests that one can derive multipliers for this period by comparing differences in estimates for 1939-2008 with 1947-2008.

What she actually says here is that the average multiplier for the entire period is likely below one, but that government spending does appear to increase employment. And with regard to that, she acknowledges: “With respect to the second question, most economists and policy makers would agree that job creation is at least as important a goal as stimulating output.”

Note that she doesn’t do a detailed case study of WWII in the first part of the paper on private spending, she includes this in the second part, specifically addressing employment. JDH above points out that “Between 1941 and 1944, real government spending increased by \$75 billion (in 1937 dollars), but real GDP only rose by \$60 billion.” But as Valerie notes, “WWII is especially interesting from a labor market point of view because the economy went from an unemployment rate of 12 percent (18 percent if one includes emergency workers as unemployed) at the start of 1939 to an unemployment rate less than 1 percent by 1944” and “as the table shows, from September 1940 to the peak in March 1945, total employment rose by 15.6 million”.

So if one believes current conditions are similar to the early 1940s, I suppose one has to ask, which do we need more right now, higher private output, or about 15 million more jobs?

I’ll add, it’s not clear to me that the model lags here would be sufficient to capture any possible long term positive effects of having those workers employed. We do know we were able to have strong growth in the postwar period while significantly cutting spending. If one looks at Figure 8, one sees that real government spending fell as sharply from 1945-1946 as it rose from 1942-1944, and by 1947 spending was back to 1941 levels.

Also worth pointing out that liquidity trap conditions were over by mid-1941, and we actually had double digit inflation by early 1942. And private employment grew rapidly as inflation rose. And both the inflation threat and the growth in private employment seem to have been stalled by a fairly modest increase in short term interest rates from 9/41 to 7/42.

35. aaron

The multiplier is less than 1. It is a bad idea.
On government employment, transferes would be as effective. Probably more effetive in improving wellbeing.
The onus is on those wishing the multiplier is greater than 1 now and even just after the crisis.
Menzie recently cited a paper that suggests it is 1-1.5 during downturns.
However, I’m highly skeptic of this being a typical downturn. No one has shown that the multiplier is higher than 1 and there is good reason to believe it is not. http://cumulativemodel.blogspot.com/2008/10/problem-with-inflation-right-now.html

36. aaron

“Given the historical experience of the US economy, our preferred estimates of the government spending multiplier are between 0 and 0.5 in expansions, and between 1 and 1.5 in recessions.”
We have been in an expansion since June 2009.
Posted by: Steven Kopits at May 22, 2012 12:44 PM

37. 2slugbaits

Appendix A says that GDP and private GDP are converted to a per capita basis including those non-civilians stationed overseas. Huh? Does this make sense? Why would Ramey compute per capita private GDP to include millions of people who were unable to consume any of that private GDP? GIs who were overseas either saved their money (a reduction in the multiplier), or spent their wages overseas (a reduction in the multiplier) or bought from the PX (which was part of G and not included in private GDP). I hope I’m reading this wrong because it doesn’t seem to make a lot of sense. It seems like this would bias the multiplier downward in a number of suspect ways.

38. 2slugbaits

Hopefully one of the academic economists out there who is fluent in NIPA accounting can help out here. In the Ramey paper she briefly mentions exports of defense equipment. Does anyone know if defense exports are counted under NX or are they counted under G? Keep in mind that under the Arms Export Control Act of 1976 private US defense companies are not allowed to make a direct sale to a foreign government for most of the expensive stuff that countries like to buy. Technically ownership must pass from the defense contractor to the US govt and then the US govt formally sells it to the foreign military sales (FMS) customer. For big ticket items the FMS country first budgets for the weapon system and then negotiates an agreement with the US govt. The FMS customer then deposits money in a special DFAS account in Denver. DOD then asks Congress for appropriation authority to buy the stuff from a US defense contractor. After production the US then buys it from the defense contractor and immediately sells it to the FMS customer turning the DFAS money over to the Treasury. Since it technically counts as a “G” purchase does that mean it is not counted as “NX” in NIPA? It would matter in terms of Ramey’s paper because if NIPA counts it as “G” then this would understate the multiplier. BTW, if the State Dept handles the “sale” rather than DOD, then it is definitely counted as “G” in NIPA. That much I know. I’m just not sure what happens if it passes through DOD.

39. aaron

2sb, yes it makes sense. The vast majority if income is spent back home (mortgage, health, family, savings, amazon.com). And cost of living is very low while deployed (and covered).

40. Walter Sobchak

Dave: “Military spending is often outside of the country and undermines good will while not training for the private sector. we need teachers not troops.”
This is the sort of blather that gives Liberalism a bad name. Here are your assignments:
How much does an enemy attack cost, in pure economic terms. Please quantify the impact of an enemy attack that destroys the central business district of a major city and kills 3,000 people. How can such attacks be prevented. You may not assume any change in human nature. Quoting John Lennon lyrics will reduce your grade to zero.
Unionized urban school districts in the United States do not teach their students basic skills. Most of their students drop out or just go through the motions. The unions have enormous political power and will not allow the system to change. Any additional money going to the system will go only to the salaries and benefits of union members. How many teachers do we need in the unionized education system? How many teachers do we need to add to the system?

41. bakho

@aaron
So what if BigG hiring is Temporary? We have temporary unemployment that is way too high and will remain too high for YEARS. Temporary hiring IS a solution to the #1 economic problem, High Unemployment. Temporary employment can also create demand that leads to higher private sector employment. The effect is Temporary only because it ends once the economy recovers and crowding out occurs. Ramey states clearly that BigG CAN reduce unemployment by direct hiring. This IS important.
It is cruel and heartless to stand by and do nothing to help the unemployed. It is BAD long term policy that damages the longer term productivity of our economy by FAILURE to adequately invest in our workforce. High long term unemployment does great social damage.
Aaron, you have no argument here. You only have a “Kick the poor in the teeth while they are down” strategy” that is heartless, based on preferences of the Greedy rich who get their kicks by bullying the rest of us. These social parasites are arguing against reducing unemployment because they enjoy making others suffer unnecessarily.
Unemployment is the #1 problem with our current economy and our sadistic, heartless, self-absorbed elites are too out of touch to care or unconscionably and knowingly cruel.

42. Kevin

But this methodology doesn’t say anything about causality, right? It could be that the government spends more because it anticipates private spending to decrease.

43. bdbd

OMB OIRA guidance for estimating the benefits of federal investment or regulatory action pretty explicitly disallows reliance on employment or other “multipliers” to justify federal actions in normal times, because the normal assumption should be full employment. OMB Circulars A-4 and A-94 (section 6b3: “Multiplier Effects. Generally, analyses should treat resources as if they were likely to be fully employed. Employment or output multipliers that purport to measure the secondary effects of government expenditures on employment and output should not be included in measured social benefits or costs.”). Of course, in recession the situation is likely to differ, but stimulus actions would be looked at differently (and elsewhere) within the govt.