Russ Roberts

Garett Jones on Stimulus

EconTalk Episode with Garett Jones
Hosted by Russ Roberts
PRINT
Frank on Competition, Governme... Rosenberg on the Nature of Eco...

Garett Jones of George Mason University talks with EconTalk host Russ Roberts about the workers who were hired with money from the 2009 American Recovery and Re-investment Act--the stimulus package. Jones (with co-author Daniel Rothschild) recently completed two studies based on surveys and interviews with firms who received stimulus funds and workers who work at those firms. They found that 42% of workers hired had been unemployed. The remainder came from other jobs or from outside the labor force such as retirement or school. Is 42% a big number or a small number? Jones argues it is small and defends his conclusion. The conversation also includes a discussion of the labor market generally and why the stimulus spending may not have been effective.

Size: 28.7 MB
Right-click or Option-click, and select "Save Link/Target As MP3.

Readings and Links related to this podcast

Podcast Readings
HIDE READINGS
About this week's guest: About ideas and people mentioned in this podcast:

Highlights

Time
Podcast Highlights
HIDE HIGHLIGHTS
0:36Intro. [Recording date: September 14, 2011.] Our topic for today is the economics of stimulus. The Mercatus Center here at George Mason recently published your study with Daniel Rothschild, "Did Stimulus Dollars Hire the Unemployed?" And we're going to talk about that study, but I want to talk first about Keynesian stimulus more generally. How is it supposed to work? The Keynesian story is that in the short run economies sometimes get tied up in knots where there are a lot of workers, a lot of firms they just aren't going to be very useful or very productive for a while. There are unemployed workers, unemployed firms. In Keynesian story is that while the market is figuring out what to do, while the market is healing itself, while prices are adjusting and wages are adjusting--or not there are some skeptical summer not skeptical--the traditional new Keynesian story which I thought was the mainstream view for decades was that in the short run the government might be able to go in and use those workers while the market is figuring out what it's going to do 2 or 3 years down the road. So in a sense this is quite similar to the old debate notice the socialist calculation debate. So back in the early 20th century Mises and Hayek were engaged in a debate socialist-leaning economists. The Socialists said that markets obviously don't work very well. We see these crises, these recessions happening every once in a while. It seems like there is a lot of unemployment. Maybe conscious planning can do better. Maybe government planning can do better. The Austrians led by Mises and Hayek argued that no knowledge actually wasn't there, the government would not be able to acquire enough knowledge about how to use workers wisely, about how to meet consumer demands wisely. What the trade-offs were between different consumer goods, different production techniques. Without prices they couldn't do that and they couldn't calculate any sort of artificial prices. There was hope for a while that with sufficient computing power and time--time might be 70 years--so that would rule out activist government policy on that ground. Some Mises and Hayek argued that the knowledge just wasn't there to be able to run anything like a modern economy without the price system. Just to comment on that: the knowledge might be there or but it couldn't be gathered to one particular decision maker or decision-making body. Knowledge is dispersed among millions of individuals, which made it impossible to be pulled together the way price system does. The Keynesian story about short run stimulus basically concedes the point that in the long run markets do a better job, but in the short run--in a few months after a financial crisis for instance when a million or so workers get laid off quite quickly--maybe the government can go in and use those workers for while on some projects, to build some schools, to build some roads, to ransom eldercare centers, during that time when the market is sort of figuring itself out. So while the market is healing itself the government could come and use these workers without too much cost to society. Because there is slack resources, they're sitting around unemployed, literally in the case of workers; but there are machines that are "unemployed"--they are not producing--so there's a chance to get things going again. I always think of it as a chance for that to get things going again. It's interesting. You called it, during this interim period while the market figures things out. But I always think of Keynesians arguing that by giving these workers money, and unemployment insurance by the way isn't invoked as a job creator on these grounds, they'll take the money, they'll spend it, is multiplier effect will kick in, and that will cause the broken markets to heal themselves. That's to use one more metaphor. You are right that a lot of people use that, especially, I say, non-economists use that argument more often, this jumpstarting mechanism that the government can jumpstart a recession. But serious economists, even the folks at the Congressional Budget Office [CBO], when they put out their models for how the stimulus is supposed to affect the economy, their assumption is that without the stimulus the economy would fix itself over the course of a few years but the government can help sort of closed gap in the meantime. So there's this middle period of weakness where government action could in theory coming in and grow the economy. For the short run. But the way you are describing it, the reason to do that is twofold. One is compassion for the workers who in the meanwhile have nothing to do. And second is we may as well get some productive use out of them in the form of bridge building, etc. Yes, so they might claim that it's truly a free lunch; at least it might be a low-cost lunch and that's the claim of Keynesianism, it's that grab some unemployed workers, grab some unused machines, let's use them to build something useful, even if it's not perfect, it's better than if we just sat at home watching TV. But Keynes himself--modern defenders of Keynes, and I'm thinking of Joseph Stiglitz, because I heard you say this in Congressional testimony--they believe that digging holes and filling them back in will help the economy. So they are invoking some kind of jumpstart multiplier, not just: well let's get something out of these folks meanwhile. That's true. And the CBO does use multipliers as well, right? They use multipliers that are bigger than one for a lot of activity. So, yes, Keynesians think there will be these additional spillover effects because if I get some extra money I'll go down to the store and buy extra stuff. And then the store will order some more and then the factory or producers I will hire some more workers, right? Yes. And Samuelson, one of Keynes's most important disciples really mathematized some of Keynes's most important ideas, pointed out that this would happen to firms, too. Firms have a tough time getting money in a recession, so if the government is giving them a lot of business, and use it to buy more machines, more equipment, and build up their capital stock. So Keynesian theory, at least the sophisticated Keynesian theory up the 1950s and 1960s, was not really this solely consumer oriented theory that a lot of us think of it as. They kept an attention to what was going on inside of businesses. Of course, in today's world businesses are sitting on a lot of cash. Indeed. And so, access to capital is not holding them back. Correct? At least not the biggest firms, no. But that's what we expect to happen in recessions, that the biggest firms have this big stock, this big war chest, that they sit on, and it's the small and medium-sized firms that really have a tough time getting access to capital because banks aren't sure who to trust. Because they are nervous and their own balance sheets might be in trouble which is particularly true right now.
7:37So, in February 2009, the President signed the American Recovery and Reinvestment Act [ARRA]. But we are going to call it the stimulus for shorthand, which I think is something of an optimistic description of it. But it is what the defenders of it called it until it didn't seem to have worked as well as they hoped it would. Now the President doesn't use the "S" word, but when he passed it, it was called stimulus. Just to avoid having to say ARRA on the air, which is hard to understand, we'll call it the stimulus. That's fair. It's unfortunate that these words get politicized, but that's part of the name of the game. So, he passes this law, passes this legislation--excuse me. A set of predictions were made at the time about the impact on output, unemployment, employment. A famous study was done suggesting that if it weren't passed unemployment would go over I think 8%. Of course, it was passed. Unemployment went over 10% and the defenders said it would have gone even higher if it hadn't been passed. That's one argument. The other is it made it worse. How would we possibly adjudicate those two viewpoints? We have smart economists on both sides. What your study will talk about in a minute is not sophisticated econometrics. Which others may think of as a plus, but which I think limits its application--there's been a lot of sophisticated econometrics done on both sides of the fiscal stimulus jobs debate. What we learned? Have we learned anything other than that economists don't agree on this? Well, this is a tough one. I'm a fan of using econometrics judiciously and cautiously, especially if you can find anything that looks like a natural experiment. So, wars have been a tragic natural experiment for the United States. And sometimes they happen at fairly random times--the Korean War, the Vietnam War, the wars after 9/11. So economists have been able to use these, especially Valerie Ramey, one of my mentors at University at California, San Diego. She used these wars as a natural experiment to see what happened to the economy when the government just goes and buys a lot of stuff really quickly. Especially when it's a large amount of stuff. So, the standard econometric analysis, which tries to simulate a natural experiment, right? It tries to look at changes in government spending to see if you try to tease out the impact on employment "holding everything else constant," which can't be done, you hope that you have isolated and chosen in your database things that are enough isolate the impact of the government spending itself. But the idea of an actual experiment is it's an exogenous--meaning outside the model itself, large ideally--impact. So, what did she find? Her big innovation is that she realized that what matters for business people is not when the government spending actually happens. What matters is when business people realize the spending is going to happen, when they expect it to happen. Why is that important? Because businesses will start ramping up in advance of the actual spending. So you don't want to look when it actually happens because then you might find no effect, falsely. Yes. In our stimulus interviews one of our government contractors told us: to get one of the government contracts, you already had to be at the trough waiting. So businesses have to plan in advance for something like doubling their workforce or expanding business capacity. So what she did is she actually looked at Business Week magazine articles, to see when the readers, the journalists of Business Week actually foresaw when the government was going to be ramping up military spending. And she noticed that that happened quite a bit before the actual spending increased. And so she used the Business Week dates as a measure of when the market became aware of extra government spending. And that was to make sure that she didn't underestimate the impact. Because if you ramp up early and then the spending happens--or you even might start hiring early--and then the check comes to pay for it, it will look like the check had no effect because you had already responded to this stimulus. Yes. So, anticipation effects, expectation effects, are a big part of doing serious economics. Paying attention to the fact that people and consumers look to the future when making their decisions today. And what she found was a multiplier that was between 0 and 1. So, in other words, yes, the economy grew after a defense ramp-up, but the private sector shrank. Offsetting some of the stimulus. So, the optimistic Keynesian view is that the government buys a billion dollars' worth of stuff and then consumers go buy an extra half billion dollars' of stuff. Then people get that money--or you've gone through all the cycles, and it adds up to another half billion? It could. Perhaps businesses buy some extra stuff to build up their business capacity. But what she found was instead when the government grew, the private sector shrank. And that's an important benchmark. I'd say a number of the recent studies that have used more sophisticated econometrics have found numbers about in that range, where yes, the government does grow the overall economy, but at the expense of shrinking the private sector slightly. So, you end up with more tanks but fewer refrigerators, because the government, say, bid up the price of steel. The example I always use is there are a lot of unemployed workers before WWII, or before any way--there's always some unemployment. So, the government comes in, puts them to work through conscription--until recently they drafted them, forced them into the army. So unemployment goes down. It's easy to reduce unemployment if you can force people to join the army. Then the question is: What do you do with them? Well, if you put them in tanks, people say that's good because that increases, puts money in the pockets of the tank-makers. Well, it does, but it also drives up the demand for steel. And if steel is not unemployed, which it often is not, it pushes up the price of steel, which discourages private users of steel. So, getting more tanks in return for fewer refrigerators--it's a bigger measured economy, but it's depressing. It's not good. It's a genuine tradeoff. It becomes a cruel tradeoff at this point; it's not a free lunch. And it's something people should think seriously about before diving into something like that.
14:35What was the reaction to her work? And when did it come out? Early versions of it she was presenting back in the late 1990s. And it was eventually published. I'd say it's been heavily cited. And these dates, now known as they Ramey-Shapiro dates, get used in a lot of work. I wish her work had gotten more mainstream attention, because unfortunately we spend a lot of our time talking about WWII. And free market economists make the point that consumer spending shrank during WWII. And Keynesians make the point that there was rationing on consumption during WWII. Consumers weren't allowed to buy stuff. So, WWII was an interesting experiment, our biggest one. But there were a lot of government rules there. But, the reason it was rationed was there wasn't enough steel to go around to build refrigerators, so the government decided that rather than let the price of refrigerators go really high, they would ration them. So, I've never really understood that argument as a defense of Keynesianism during WWII. I think the more important point is that America's intervention in WWII was widely anticipated. So, the date you would use is very difficult for figuring out the impact of assessing WWII. 9/11 came out of the blue, so no one was thinking: Soon the United States is going to be attacking Afghanistan and Iraq. So, I think it's probably a little cleaner. And similarly with the Korean War--quite a surprise. Relative to WWII. At least the date of it. The nice thing about the Ramey dates is at least it gets away from having to talk about this one case study and we can look at a number of events that have happened to a massive economy when government spending ramped up dramatically. Now, Robert Barro has done related work, found similar effects--he's focused on war as an exogenous, unexpected increase in government spending. But his work, or Valerie Ramey's, hasn't convinced the Keynesians. They haven't said: I guess I was wrong. They must have a counter. Well, one possibility is to say: So what if it's smaller than 1. At least it's bigger than 0. The normal view of economists is that at least in the long run, government spending crowds out mostly investment spending. So, it shrinks savings; it shrinks the capital stock in the very long run. So, market-oriented economists will tend to believe that the multiplier is actually negative. A lot of this debate goes on about whether the multiplier is bigger than 1 or less than 1. On some level the most important question is: Is it positive or negative? Once we get into the 0-1, we get into debating how much we like government stuff versus non-government stuff. If you are fighting a war and you are worried about your survival, you are thrilled to substitute tanks for refrigerators. We are not arguing here whether defense spending is good or bad. Presumably there are times when it is crucial; there are times when it's unnecessary. We are just talking about the economic impact. Whenever I write about this, I always get some commenter who says: Are you saying we shouldn't have fought the Nazis? That's not what we are talking about here. We are only asking--let's assume we had to fight the Nazis. What was the impact of that on the economy? Not on the world or on the future of the United States. We are just asking about the private economy. To me, the Keynesian story is it's supposed to stimulate the private economy because there are all these resources lying around doing nothing. It didn't. It shrunk the private economy. You can still argue: that's because it had to be so big that it eventually crowded things out. So, it's not a refutation of Keynesian. For me, it's merely saying that you can't use this example, which Keynesians always use, that GDP grew during the 1940s. And my answer is: So what? Yes, you can make GDP grow, buy a lot of tanks and bombs and airplanes, and it was great for those industries, in those areas. The question is: Did the Keynesian multiplier for the private sector, which is again one of these folklore stories that you hear, did that kick in? The answer is: evidently not. Although there is some dispute, I'm sure about consumption spending in WWII. But I think the people who lived through it were pretty overwhelmingly not happy with their consumption experience. And it appears from Ramey and Shapiro's work that the same thing happened in the wars that we've had since WWII: extra government spending is associated with shrinking private sector spending.
19:12So, that's some evidence. You said positive, a fraction, something between 0 and 1 would be better than nothing. I'm sure they have quibbles about how the econometrics was done. Has anyone tried to re-do her work with a different set of assumptions? I have to say, first of all, I'm not an expert on that line of work. But I've poked around quite a bit, and I've not seen anybody pushing for numbers bigger than 1. In the wartime spending. Yes. But there are lots of studies that find bigger than one not using this natural experiment approach, just taking all the data, not just looking at wars. Yes. The typical results that the CBO uses, when they generate their big multipliers come from more routine econometric studies, where you just look and see when people get extra big checks from the government, how does that affect their consumer spending; and let's just sort of assume that that spending wouldn't have happened otherwise. So, there are a number of these multipliers floating around, but I have to say that in the last 5-10 years, all of the multipliers that I'm familiar with find something that I really wouldn't have expected before, which is that the multipliers for tax cuts are bigger than the multipliers for government spending. That's kind of weird. You'd think it would be symmetric. That could be a data issue or a measurement issue. The people who find that--do they have a story? The most famous one of these is Romer and Romer's study. Christina Romer, recently of the Council of Economic Advisers, now back at Berkeley, and her husband David Romer. Theirs doesn't look at government spending itself. But they found a tax multiplier of 3. So, they found that a dollar of government tax cuts, in the short run, increases GDP by $3. Surprisingly large number. But other studies using traditional econometrics have found something similar. So, Blanchard and Perotti have a widely cited paper, Olivier Blanchard, MIT economist. They found that government spending multipliers are smaller than the tax multiplier. There's a U. of Chicago economist whose name escapes me who has a similar result. On the flip side, when it comes to times when we want to shrink government, or shrink government deficits, Alesina has a study that has been widely attacked that showed that it's possible that expansionary fiscal contraction--that's the jargon they use--it's possible that when government spending shrinks as part of a deficit reduction plan that the private sector actually grows. Keynes's animal spirits become awakened. A lot of people attacked that idea. The OECD [should be the International Monetary Fund, the IMF] tried to attack it, but even their attack, they themselves found that when governments try to fix their long-run budget deficits by cutting government spending, there is only a small negative impact on the economy. When they try to cure their government budget deficits by tax increases, there is a bigger negative effect on the economy. So, if you are trying to close your long-run budget gap, it's better to do it through government spending cuts than through tax increases. And again, this is the flip side of the multiplier debate, but it's telling us the same story these other studies have found: that changes in government spending have small influences on the economy and tax changes have big influences on the economy. I think the paper Christina Romer wrote with Jared Bernstein, Vice President Biden's economic adviser, paper that predicted the 8% unemployment if the stimulus were not passed--I think they used a very high multiplier. I think it was 1.52. I always have to tell this story of how do we know economists have a sense of humor, macroeconomists--they use decimal points. They didn't use 1.5. They used 1.52. I'm not sure that's what that actual study used, but I'm sure that actual number was in the air when we were talking about the stimulus. And some Keynesians would push for numbers even larger than 1.52. So, there is a growing literature that's skeptical either about spending relative to tax cuts or literally optimistic about spending cuts being stimulative. There's a little bit of a natural experiment at least talked about in England. England has embraced austerity--meaning reducing government spending. I don't know if they really did. I haven't been following that, but that's a possible source of information. Sadly, the Western world is going to be running this experiment a lot of times in the next few years. So, we'll have a lot more data in the next 5 years. Because a lot of governments don't have enough money to cover their promises. They made a lot of promises, especially to the elderly; and those are tough to keep. Demographically.
25:25Let's move on to your study, with Daniel Rothschild. Tell us what you did and why you did it. What I like about it is, no doubt it's flawed, imperfect, doesn't answer every question--but you did something most economists don't do. You went around and asked people what they did. Which is interesting. Within just a few weeks of the stimulus bill being passed, my colleagues at the Mercatus Center decided that this was a chance to do a once-in-a-lifetime study of how the stimulus was influencing the economy. So, a number of us batted around ideas about what would be a good project to work on. And the Mercatus Center is working on a number of projects about the stimulus. One idea that I threw out is: We should go talk to people. In the last decade or so, economists have been comfortable with studies where economists go and interview people or send out surveys and ask people how they react to the economy. My favorite example of this is Truman Bewley's book, Why Wages Don't Fall During a Recession. In the early 1990s recession--he's a professor at Yale; he had written many theoretical papers about why unemployment might exist and why it might get worse in a recession, fancy game-theoretic models, heavily mathematized--started rolling around, he saw the economy weaken and he thought this might be a chance we could actually go talk to some business people, some union leaders, some hiring managers. And that's what he did. This is a good-sized book but there is a shorter version of it available on the St. Louis Federal Reserve's website, where you can distill some of his big ideas. What he found was that businesses are really concerned about fairness. They are really concerned about their reputation. So, they are terrified of giving across-the-board wage cuts during a recession, even if that would save all of their workers' jobs. So, instead of cutting everybody's wages by 5%, they decide to fire 5% of the workers, or lay them off. He heard this again and again, that business managers, business owners thought that morale would be hurt at the firm if they cut everybody's wages by 5%. So, morale is a repeated theme. But he learned this by talking to people. He learned this by actually sitting in rooms, by driving around Connecticut and getting out of his office. Which is something economists don't do that often. I have to say I'm somewhat skeptical. To me, the Bewley story indicates some of the dangers of surveys, which--it's one thing to ask people what they did. To ask them why they did what they did, I think you are treading on dangerous turf. It could be merely that they knew there was at least 5% of the labor force that was not very fit for their jobs; this was a nice time for morale relative to other times to fire them. But they tell you a fairness story because that sounds better. That's possible. Put that to the side. The point I think is interesting is that a theorist decided to get his hands dirty, so to speak, in this way, is fascinating. And he should be applauded for it. I agree. So, what did you do? The first part of our study involved interviews. We sent out teams, led by Dan Rothschild, to five different metropolitan areas in the United States. They made phone calls to stimulus recipients just using the recovery.gov list. Recovery.gov being the government's website summarizing the stimulus impact and different state impacts. They have mailing addresses, dollar amounts, names of the organizations. It's quite comprehensive, so more people should be making use of this. So, they just called up people and asked if they could drop by for 20-30 minutes for a talk. Almost all the people who said yes agreed to be tape-recorded; we gave them anonymity. And we have recordings of these interviews. I spent some time listening to them. Our teams wrote digests of all of interviews and I spent some time looking at all of their notes. And Rothschild and I spent some time talking about all of this, about what the big lessons were that they saw. How many firms did they talk to? In the interview-based study, which is separate from our written survey that we mailed out, we interviewed 85 organizations. They were almost equal mix between non-profits and private businesses, with a small number of government businesses as well. Give me the flavor of what kind of businesses and nonprofits there were. Now, of course, there could be a bias in terms of who said yes. Yes, exactly. There are many places bias could come in, but that would be one of the worries. So, you don't know exactly how representative they are, but you know something about the universe of people who said yes that were interviewed. Tell us something about that. Sure. Well, a typical private-sector business would be a government contractor, a construction contractor, who might do mostly work for the Federal government; some work for state governments, some work for the private sector. So, they take a lot of different jobs; the Federal government is just one of their bosses every once in a while. So, my team would sit down with this person and ask them a few questions. How efficient was the stimulus funding? Did the money come in when they said it would? What did you have to do to get it? What were the reporting requirements like? We heard a lot of complaints about the reporting requirements, how difficult they were at the beginning; but apparently the Federal government fixed that, made it quite a bit easier, quite a bit more user-friendly after a while. I'm at the Mercatus Center also, as you are. We are of course a market-oriented, free-market oriented research center. Did you worry, and if so what did you do about the fact that many of the people who were sent out to the interviews were unsympathetic to the stimulus? Which I assume they were. When people asked us, we would always say we were from the Mercatus Center, a market-oriented research center associated with George Mason University. And, I'm guessing that a lot of the people who met with us must have done a Google check. That's what I would do if somebody were coming in to talk to me. But that's because you are an academic. Is that true? I think a lot of them probably just said: Ok, it's a university. People in the real world actually work, like all the time. We work less during the day and more at other times. So, if you are in what's called a job, your 20 minutes was already a big deal. I don't think they spent another 5 or 10 looking around, reading about the Mercatus Center. I'm just guessing. One kind of bias would be that the interviewee might try to please you or not, depending on the perceived attitudes of the Mercatus Center. I'm more interested in the attitudes of the team itself shaping the interview in a direction that might go in the direction that would be anti-stimulus. We asked open-ended questions. I think that's part of the way we neutralized that possibility. We would ask them: How many workers did you hire? How many jobs did you create or save? If you are asking them an open-ended question like how did the reporting system work, they can tell you it worked well or it worked poorly. By the way: the transcripts and actual recordings--are they available? Actually, we have to keep them anonymous, so we are not making them available now. Okay, so that's awkward. Let's put that question aside. Our surveys are all available, from our 1300 respondents. That's a different part of the study. I mention the transcripts and recordings because it would be nice to hear any tones of voices. Let's put that to the side.
33:47Who are you talking to at the company? Who is the person on the other side of the table, typically? Sometimes it's the president of the organization, the owner or co-owner. So, when we spoke with some small engineering firms, it would be the two owners of the firm would come to the interview and sit there and talk for 20 minutes. So, if you asked them, how many people did you hire as a response to the stimulus, would they say: Oh, I'd have to look that up? Would they say: Oh, about 100? Would they have that number easily? Make it up? Business people thought this through. Some of the people gave us a response that was just what they put on the government form. These were people who would have the answer right at their head: Oh, we saved this many jobs. The government's formula though just took it for granted that if the government spent a certain amount of money and if there were certain people working on that government project, those were jobs created or saved by the stimulus. That sounds like a good answer. What's wrong with that? We don't know what would have happened otherwise. That's one problem. A lot of companies told us that things had been slow, but they still had their jobs. Things had been slow, and then the stimulus came along, and gave us more work to do. So, in that setting, it's hard to imagine that the government stimulus either created or saved those jobs. Nobody really was hired. It was just that workers who were sitting at work playing Tetris or checking their email--like the academics you mentioned a minute ago--instead they would actually do a lot of work during this period of the stimulus. So, we don't know if they would have lost their jobs otherwise. They were working beforehand without a lot of work to do. And work that actually expanded, of course, we don't know where those workers came from, whether they were unemployed before. We'll come to that in the second study. Summarize what you learned or think you learned from the interview part of this project. What surprised you? The first thing we learned was that when people talked to us about where their workers came from--because some firms were small or medium-sized firms where they really knew the workers--it was surprising how few people came from unemployment lines. These companies would tell us--I can think of one Federal agency actually that did some hiring based on the stimulus. They said: We hired 5 people based on the stimulus funds; two of them were retirees from our own agency that we brought back under a special program that let them keep their retirement benefits while earning a salary. Nice. One was transferred over from another agency. One was a part-timer that we made full time. And one person switched over from a private-sector job. I'll count somebody going from part time to full time. I'm going to call it a .5. A point 5. The rest are quite a bit more ambiguous. And I'm not cherry picking now. That's the kind of story that came up quite a lot. Now, defenders of the stimulus would point out that that one person who came from the private sector, who was employed beforehand--so of those 5, you got a 0.5 increase in employment. Net. The other 4.5 already had jobs. But you could argue that the one person who came from the private sector--that firm now has an opening. And you don't know whether that firm went out an hired somebody. Well, it's possible that that firm posted an opening as soon as that worker left. But a lot of what we've seen in this recession is businesses getting by with fewer workers. Productivity keeps increasing and as a result even a market-oriented economist would say that maybe the businesses learned to produce just as much stuff with fewer workers. A Keynesian response is that actually in recessions firms hold onto a lot workers. Keynesians call this labor hoarding. So, you are afraid to let go of your core workers during a recession even though things are really slow. And as a result, if somebody walks out the door during a recession, they are actually doing you a favor. That's one less person you have to worry about firing. But it could be the other way. It's absolutely possible that some firms will try to hire someone. As we'll see in the survey results, it looks like it might be actually kind of hard to hire someone.
38:35Let's turn to that. So, what we've been talking about so far was 85 face-to-face interviews with business executives and non-profit execs about their experience as stimulus recipients. That study has been published by the Mercatus Center, and the title is "No Such Thing as Shovel Ready." And the reason that's the title is? The President actually used that term to describe what was going on with the stimulus. He said: the big thing he learned from the first year of the stimulus program is that there is no such thing as shovel-ready projects. And why is that relevant for the survey results? It means that it's hard for the government to ramp up quickly and find a lot of unused workers and unused organizations. Part of what we found in the interviews is that government and government agencies often went to the organizations they trusted. They went to the firms they trusted. People they had experience with. They didn't start up a new firm: I'm going to hire 20 new workers. No. Smart, by the way. They didn't want to waste taxpayer dollars. They didn't want boondoggles. And because they didn't want boondoggles, they went to trusted firms. And trusted firms are often already busy. It's also convenient. So, let's turn to the second part of the project, which is a mailed survey. What did you do there? There, we sent out a 2-page survey to thousands of stimulus recipients. We went off of Recovery.gov. How many were there together, by the way? How many recipients are there? That I don't have an absolute answer to. I want to say 13,000? Yes, that's got to be in the ballpark because after we did a screening, there were 8,000 left in our screen. So, at the point you are doing the survey, there are about 13,000 firms that received stimulus. You threw out 5,000 who you didn't mail to because they were very small. The very smallest ones we didn't send anything to. If you received less than $100,000 in stimulus funding, we didn't send it to you. Also, if you were part of the state, part of the Governor's office, we didn't send it to you. Why? We wanted to send surveys to people who were close to the spending. And we knew the Governor's office wasn't hiring people; they were sending it out to somebody else, and we wouldn't be able to get information on those folks. But local governments, we sent them to; county governments, we sent them to; non-profits, we sent them to. And mostly private sector firms. So, I just have to mention: non-profits don't wield a lot of shovels. I'm kind of intrigued by that. I think most of us incorrectly think the government--we saw on the highway, road-building; you'd think: That's the stimulus money. But that's a very tiny percentage of the $800 billion of ARRA, of the stimulus money went toward. Whether shovels were ready or not, went to shovels. It went to food banks. It went to people running elder care centers. Non-profits that a lot of us think were doing things we would approve of. So, how many surveys did you send out? We sent out a total of 8000 surveys. How many came back? We had 1300 responses, but I have to say that those 8000 surveys, each of them included three pieces of paper. One piece of paper was for the business itself to answer; and two pieces of paper were for workers to answer. They said: "and these two workers of your choice." So, we are getting a look inside these firms. We don't know who is responding, because those are all kept anonymous. There is no identifying information. But it is getting us a chance to look inside of these firms. So, when I say "how many responded," of course sometimes you only got one piece of paper back, I assume. And as soon as you got all three...? We gave people separate envelopes, so that gave people further anonymity. So, if a worker didn't want to send it in, he didn't have to send it in. So, how many pieces of paper came back? 1300 pieces of paper. So, in a way that's extraordinarily good. 5% is phenomenal for a mail survey of some kinds. But in this, you could argue that you got what you got. There were 3000 pieces of paper; 8000 places got sent to; that's 24,000; you got over 1000 back. Not bad. The point I'd like to make is: We shouldn't have had to be the ones to do this. Every single question we put on these surveys were ones the Federal government could have easily put on their quarterly stimulus reports. They could have asked every stimulus recipient: Of the workers you hired, how many came from the unemployment lines?
43:33So, let's take that question, which I think is the most interesting one of the things you report. First question: Of the 1300 or so, over a thousand or so responses, how many of those were employer responses? It's right around a third. So, let's call it 400. You had 400 companies write back to tell you that they are stimulus recipients. One of the questions was: How many of the workers did you hire and how many of them were unemployed before you hired them? How would they know? Actually we didn't ask them that question. We asked the workers that question. We asked the workers: When were you hired at this firm? And then we asked them, if they were hired after 2009, we asked them a series of questions. And one of the questions we asked them was: What were you doing before you worked at this firm? We asked them: Did you work at another job immediately beforehand? Were you unemployed? Retired? In school? Out of the labor force? Thief--what we call the uncovered sector in some sectors, not covered by Social Security, called the uncovered sector. But you also asked the employer. We asked the employer what the hiring side was like. We did ask them how big was the stimulus compared to your annual revenue? How many workers did you hire? And, what was the size of your labor force? We actually broke it out the way the government does. We asked how many workers did you keep from firing? How many did you hire? And, how many did you bring back? Of course, keep from firing is a very subjective term. But we put some weight on what people said about their own firms. We also asked them if it was harder to find good people now compared to before the recession. So, again, you are asking them to make a judgment call; but these are the people who are making these decisions. So, we asked them: Is it harder or easier to find good workers now or before the recession? In theory, it should be easier. Much easier. There is a pool of unemployed labor. One more point about methodology. Because it was anonymous you were not able to compare employer responses to employee responses. No. We could not. One could make a very crude comparison if we looked at people in the same states. That's about it. So, what did you find? We found that only 42% of the workers came from unemployment. 47% said that before they got this job, they were working somewhere else. And where the rest came from was from school and out of labor force. So, small numbers. Why do you say only 42%? Could say 42% of the jobs created by the stimulus program for recipients--and you are always going to add the qualifier in the recipients of the survey who responded. You don't know what this captures; this is a very crude estimate. But it's interesting. It tells you something. It would be great if the government replicated our results and found that the numbers were the same or different. Replicate your survey. Yes. So, why do you say only 42%? Those are new jobs! That's great. Well, we can't tell if they are new jobs. The problem is that we don't know what would have happened otherwise. But here's what we do know. We do know that in recessions, in normal recessions in the United States, more people get hired from unemployment than usual. On average, in the United States, when firms are hiring, about half the people they hire come from unemployment lines and about half are job-switchers, coming over from other firms. So even in a healthy environment, even in a healthy job market, it's very common for people to be hired from unemployment. There are always jobs breaking up. There are people moving from one town to another. It's always possible to hire unemployed people even when the labor market is quite healthy. We would think that in a recession it would be much easier than usual to hire workers from unemployment lines. If the stimulus was actually targeted at areas of weakness in the economy, then we should be able to do much better than average. They actually did a little worse than average. Which is either a statement about stimulus or a statement about this particular recession. We don't know. And it's possible it's a statement about this particular stimulus. A lot of us noted early on, when the stimulus bill first came out, that a lot of it seemed to be targeted at highly skilled sectors. You've heard a lot about these green jobs, for instance. Going to engineers, for instance. I was always taken aback by how much went to, say, universities who do health research. I like health research. We do a lot of it already. Doing a little bit more isn't the worst thing the government can do with my money, but it doesn't put a lot of unemployed Nevada carpenters back to work. If Washington University in St. Louis gets to increase its spending on Parkinson's research. Top medical researchers do not have very high unemployment rates. Like zero. Very close to 0. It's a good point where the unemployment in this recession, as in other recessions, is very, very different for different educational groups. So, if you have been to college, and even more so if you have an advanced degree--if you have an advanced degree, your unemployment rate is very low. I think college rate is about 5%, which is "low", which is for non-college, if you haven't finished high school, if you have just finished high school, it's over 10% or maybe over 15% for some education groups. In our survey, half of the respondents had at least a bachelor's degree (BA), and a fourth had a graduate degree. So, stimulus money appeared to go--as casual evidence suggested beforehand--to hire a lot of very skilled people. Federal government tends to hire skilled people. That could also just because the owner of the firm just gives it to my buddies, the owner just filled it out, could be fine. Down to the guy working the backhoe. We don't know. We don't know who within the firm decided to respond.
49:46To come back to your point about roughly half of new jobs in healthy times going to unemployed workers, and here it was 42%, surprisingly, should have been above the usual. How do we know that about healthy times? Where do those data come from? This is all a lot of work that's grown out of Mortensen and Pissarides, who won the Nobel Prize, and the work of John Haltiwanger, U. of Maryland. So, John Haltiwanger was the first person, with his co-authors, who started collecting data on labor market dynamics, what we now call job churn. They are the ones who created the new stylized fact that we didn't even know of which is that on average in 20 years ago in the United States, there are about 4 million jobs created and destroyed every month. That's an amazing number, an amazing amount. Either by firms that are shrinking or going out of business. And then, in healthy times, slightly more than 4 million jobs are created every month because firms are either expanding or getting started. In healthy times it's even below 4 million. It's amazing that the amount of churn goes on in this economy. It's through the work of Haltiwanger and his co-authors, getting the Federal government, the state governments, and governments in other countries to start keeping track of this. They are the reason we now think of labor markets as being so dynamic. So, in America, the Bureau of Labor Statistics [BLS], I think it's called the Job Openings and Labor Turnover Survey [JOLTs], which is relatively recent response to this work. Fantastic data. So, let's go back to your results. One of the things you found: 42% of the people who were hired, were hired from the unemployed. So 58% came from either other jobs or out of the labor force. Some of those out of the labor force presumable could have been discouraged workers, though. They could have been. But they were a tiny fraction. I think a lot of people who according to the government definitions were out of the labor force, they thought of themselves as unemployed. They used the common-sense definition of unemployed. Interesting. Good point. Now, Matthew Yglesias, on his web page, suggested that actually 42% was a big number. He said: when unemployment is only 10% of the economy, anything over 10% is gravy. Part of it is the point I made before, which is that when somebody leaves a firm, when somebody switches jobs, we're not sure; we don't have any reason to believe that person is going to get replaced any time soon. Our data gives some evidence that there really was across the board labor market tightness. So, we have the fact that there was this high rate of job poaching or job switching. We also know from the survey that businesses said it was at least as hard or harder to find good people before the recession. Slim majority. About 15% said it was harder to find good people. About 37% said it was no easier--just as easy or just as hard. How many said it was easier? A little less than half--something like 47%. So, that was the most common response, it was easier to find people. That was the modal response. So you are saying you are surprised because it was a relatively low number, even though it was the most common response. Yes. When most people find it as hard or harder, that truncation is what I didn't expect. If the stimulus was well-targeted, then the data would be telling us overwhelmingly that it's easier to find good people. Once we got the money, we just went out and scooped them up. So, my thought, on the Yglesias point, which I think is an important point to make about how you think about unemployment and aggregate demand, Keynesian demand, is that workers aren't homogeneous. They are not all the same. If you think of them as all the same, then for every 90 workers who are working, there's 10 marbles who aren't in somebody's pouch. The reason they are not in somebody's pouch is they are off in a corner somewhere; imperfect information or people haven't matched. But it could be that those workers don't have the skills people want to use. What you are finding is that what skills you have matter a lot. It's not just being a body. Labor is not homogeneous. Yes. This is a point that Steven Horowitz, the economist, did a great job responding to in a blog post. You can't just assume that workers are fungible, that job skills are fungible. Search takes a long time. Any of us folks know just how hard it is to find a job. There's a reason why labor economists often talk about the labor market as being like the dating market. Search takes a long time, and finding a good fit is more the exception than the rule. And Arnold Kling at EconLog has written very eloquently on what he calls Patterns of Sustainable Specialization in Trade--if you are an unemployed carpenter, you might have to find a whole new area to be productive in and carpentry might not be a viable alternative for the next few years. People who have defended the stimulus say to me: Well, you go to that construction firm, you see them on the highway; construction workers can go work on that highway job. But my suspicion, and I think it's somewhat consistent with your finding, is I'm not so sure a person who can lay drywall can put asphalt down on a major American highway. Their resume is going to go very low in the pile. This is a point Gary Becker made early on, when the stimulus was being talked about. Sectors that economists think are workers should be able to switch across--it doesn't really work that way on the ground. It's like somebody telling us: You are an economics department; why don't you just hire sociologists? You are both social scientists. Right. The problem with that criticism, though, and this is where I find the mystery, is that when things are good, and 4 million jobs are getting destroyed every month and another 4.2 are getting created, somehow people just move in and out of jobs. You quit your job, you don't worry about finding a job, in good times; the tech bubble bursts, there are all these unemployed web designers because now web design doesn't pay much. It used to be a cool, novel skill; now it's worthless because there are so many templates that a 9-year-old can create a web page. But those people, most of them, find work elsewhere, or they tooled up. So, why is it--you are saying it's so hard to find these matches. In healthy times it's not so hard. I think you are right. That's a very good question. Let me talk a little bit about this idea of the government hiring people from unemployment lines. If you just believe, if you just assume, that Keynesianism works, then 42% being hired from unemployed might sound like good news to you. But we should remember that in a healthy economy, when the job market is tight or hot, the government could go in and hire workers quite easily. A massive government spending program can always scoop up workers. If you pay enough. So, we saw people coming out of unemployment at about the rates you would see in a healthy economy. And on the question of paying enough, we actually asked workers--the job switchers, whether they were unemployed or job switchers--how are your pay and benefits here compared to your previous job. On the pay issue, there was a slight tendency for people to say the pay was a little better than at their last job. Definitely true for the job switchers, for the people who were poached. People who were unemployed were taking pay cuts on average. But, on benefits, across the board, people were twice as likely to say they had better benefits now than at their old job. So, about 40% said their benefits were better; about 40% said the same; and 20% said worse. So, the stylized fact that people put a lot of attention on benefits seems to be true. And, it looks as if when the government and government contractors wanted to get good workers under the stimulus, what they did was give about the same pay package but much better benefits. And attracted some people in from other firms.
58:32That was part of my other criticism of the stimulus logic, which was very similar to the steel story I told earlier, which was: increase the demand for medical researchers; not only are you not going to put a lot of Nevada construction workers back to work; you are going to raise the wages of medical workers--or the compensation--and make it more costly for other private firms. So, if anything, when people say maybe other jobs are created elsewhere--it's probably were not created. Because they got more expensive, those skills. That's part of the reason we title this "The Supply Side of the Recovery Act." If there are more workers working for the government, then it may be the case there are just fewer workers available for the private sector. This is supply-side crowding out. A lot of people who criticize Keynesian spending talk about how there might be some financial crowding out. Like, if the government is borrowing a lot more money, that's less money that can be borrowed by the private sector. I'd like to make the point that what we've found evidence of here is that when workers are pulled toward the government sector--especially when the best workers are pulled over to the government sector--those are workers that aren't available to the private sector. Now, I'll be interested in talking to people on the other side of this issue, people who defend Keynesian stimulus, and I'm always eager to find them. I'm sure they'll have some critical things to say about your survey. As we said earlier, it's an imperfect survey. Many things limit its applicability to make universal statements about the stimulus package. But it's a really interesting thing that you did, so I salute you for what you did. Thanks; and I hope that the Federal government takes some lessons from us and asks these kinds of questions of stimulus recipients in the future. Now, what kind of reaction have you received? I mentioned Matt Yglesias, who was critical on the web. Other bloggers have been critical of it. Have you received any feedback from academics or any hate mail? Anything you want to share with the audience? I've received a couple of nice emails from government contractors who have given me some insight. It's almost as if I have my 86th and 87th interviews now. People have been very insightful in telling me how job markets work, and how the government contracting system worked when the government is trying to ramp up on short notice. Some other academics have had some nice things to say and I'm glad it's gotten some attention in the mainstream media. Are you planning on doing anything similar down the road? We are probably going to work on a full-blown academic paper that probably combines both papers together to sort of do fancier econometrics that economists care about and compares a few more hypotheses we were able to test. We asked people, for instance, how much they cut back their spending when they were unemployed, and how much they borrowed money or saved, so we were able to test some of Milton Friedman's ideas about the permanent income hypothesis. And a number of other questions.

COMMENTS (23 to date)

This is an excellent podcast. It was a very enjoyable and informative discussion. One area of the podcast I found very interesting was the “multiplier” associated with tax cuts. If I understood correctly, for each dollar of tax cut there was a 3 dollar multiplier. That is impressive to say the least and it would appear that the economy would be getting a lot of bang for that buck. Russ, what do you attribute this to? Do you think that if our tax rate were only 10% that we would get the same effect or would the multiplier be a lot less? How about if our tax rate were at 75% or 80%, would a tax cut multiplier be even greater? My guess is that the lower the tax rate, the less of an effect the cut would have and the higher, the greater. But I’m basing that on my intuition. I would love some insight from you on this.

Sleeper writes:

Great podcast and THANKS for posting the Blanchard & Alesina work!

dgp writes:

Russ does a tremendous job of "challenging" assertions even when he agrees with the subject. His questioning uses a scientific technique which is sorely lacking in many journals and podcasts. Thanks for that and thanks for another interesting podcast. One other comment, Russ is usually very good at defining terms brought up in the interview, even if they seem obvious. The "Multiplier" reference was an exception to this rule. I, for one, would have been educated by a quick definition.

dgp

George writes:

Terrific discussion -- thanks to you both. The interviews conducted by Garret and the follow-up emails he later received from government contractors felt a bit like Russ' slice-of-life interviews (car salesman, hairdresser, Frito-Lay exec.) So, a suggestion for a future podcast: interview with a government contractor to see how business looks from that perspective.

Jenkins writes:

I'm confused about the claim that a multiplier is between 0 and 1 as a problem for Keynesians. If it were larger than 1, wouldn't you would have runaway growth--a single dollar spent by the government would increase the economy to an infinite size via the following process:

1) government spending increases by $1,
2) private spending increases by $1.2
3) (assuming, somewhat pessimistically, the same private/public multiplier) private spending increases by $1.2 * 1.2 = $1.44
4) private spending increases by $1.2 * 1.44 = $1.72
5) keep going until GDP is infinitely large.

In contrast, if the value is between 0 and 1 in step 2, then you get a arithmetic series that converges to a finite value. For a multiplier of one-half you get the final result 1/2 + 1/4 + 1/8 + 1/16 + ... = 1, or every dollar in increased government consumption increases final output by $2.

Matt writes:

Excellent discussion on a very interesting topic. I do think that some of Jones's more provocative conclusions are also the most flawed. Russ identified a few, so I'll just point out two.

1) It's not surprising (and it's certainly not underwhelming) that "only" 42.1% of the new hires surveyed were unemployed immediately prior to starting their new jobs, even if we assume that under normal conditions, half of new hires come from unemployment. The employee respondents were much better educated than the overall U.S. labor force. 61.9% were college graduates; 24.5% had graduate degrees. According to the Bureau of Labor Statistics, only 25.5% of Americans 25 or older held college degrees or higher in 2008. As discussed in the interview, the better educated a person is, the less likely he or she is to be unemployed. According once again to the BLS, in 2010, the overall unemployment rate in the U.S. was 8.2%. For college graduates the rate was 5.4%. For holders of master's degrees it was 4% and for people with doctorates, 1.9%. If we accept those figures as accurate, at least with respect relative unemployment by educational attainment, wouldn't we also expect that a highly educated sample of new hires like the ones surveyed would be less likely to have been unemployed before starting their new jobs than the general public? Perhaps even much less likely?

2) Jones emphasized that we "know from the survey that businesses said it was at least as hard or harder to find good people before the recession." At best, that's some glass half full analysis. We also know from the survey that only 12.2% of respondents reported that hiring was a little harder or much harder than prior to the financial crisis. In fact, 87.8% said that hiring was the same, a little easier or much easier. In sum, 47.1% reported that hiring was easier, 40.7% reported no change, and only 12.2% said hiring was harder. Doesn't that prove the opposite of Jones's claim of "across the board labor market tightness." If the labor market was experiencing no change, we would expect mostly no change responses and a smaller, but roughly equal, number of easier and harder responses at the margins representing anecdotal experiences. Instead, we see almost half of the respondents reporting easier hiring, a slightly smaller portion reporting no change, and a much smaller slice reporting tougher hiring. Clearly, the trend is toward easier hiring.

xian writes:

great stuff econtalk....definitely need more studies like these.

war spending on tanks and such is not investment. all the resources that go into producing them r destroyed and abandoned on a battlefield.

in the process, someone else's (the enemy's) fixed capital assets r also destroyed.

this is negative investment.

it's very different than building a federal highway system, air/seaport, funding medical research or maintaining a national park.

Dan H writes:

The questionable road repair currently going on outside my house notwithstanding, I don't buy the standard argument here that all government spending is by nature an inefficient use of resources, and simply digging and filling holes.

However, I am in complete agreement with xian's statement above. Government spending on military resources is a terribly inefficient use of resources. It is akin to digging holes and filling them with another country's labor force. No economist -- academic or armchair -- should be allowed to get away with the argue that a beneficial byproduct of war is economic growth.

Charlie writes:

The reporting on the empirical evidence around multipliers was terrible. Ramey's work was completely misrepresented. Ramey points to the results of her own work as showing a multiplier of .6 to 1.2, why were they reported here as 0 to 1?

The lower end estimates are in samples dominated by the Korean War. Since the Korean war was finance with tax increases, it's not a very good argument against stimulus spending.

Here's Ramey, "These government spending multipliers do not necessarily represent deficit financed increases in government spending, which is the type most likely to be used in a stimulus package. For example, the lower end of my multiplier estimates (Ramey (2011)) are from samples where the Korean War is dominant, and hence are samples in which much of the spending was financed by tax increases. Even during World War II, some of the increase in government spending was financed with taxes."

Then there is this exchange:

"Has anyone tried to re-do her work with a different set of assumptions? I have to say, first of all, I'm not an expert on that line of work. But I've poked around quite a bit, and I've not seen anybody pushing for numbers bigger than 1."

So besides Ramey herself, who includes multipliers bigger than one, her view of the literature is much different, "I will conclude that the U.S. aggregate multiplier for a temporary, deficit financed increase in government purchases (that enter separately in the utility function and have no direct effect on private sector production functions) is probably between 0.8 and 1.5. Reasonable people can argue, however, that the data do not reject 0.5 or 2."

But Jones has "poked around quite a bit." It took me about 10 minutes on google to pull this stuff up. That's ridiculous. For someone trying to gain credibility for doing serious empirical work, this is not helping his cause. I frankly am pretty disappointed because ever since Tyler Cowen put Jones on my radar screen, he has seemed to be better than this type of misrepresentation.

For the quotes I used and a more serious and honest review of the literature, an interested listener can go here . And find Ramey's paper, "Can Government Purchases Stimulate the Economy?"

Trotter writes:

Jenkins,

I may be mistaken but I believe that the multiplier is the factor by which the total increases, not the factor for each step. So a fiscal multiplier of one, means that one dollar of spending gives you one dollar of growth. I agree that if the multiplier went on ad infinitum, and it was greater than one, that would lead to infinite growth.

Garett Jones writes:

@Charlie:

Thanks for your comments.

The midpoint of Ramey's 0.6 and 1.2 is 0.9--which is the number I had in mind for her multiplier.

I think that's the normal way to sum up multiple specifications, unless the author takes a position that some specifications should be dramatically downweighted.

And just as importantly, her results show declines, not increases, in consumer and investment purchases. One can of course argue that the confidence bounds are wide enough that perhaps those effects are zero or even positive. But the weight of her evidence is that domestic private spending shrinks after a war-driven spending shock. Unless one wants to argue for a multiplier driven by increases in net exports, then we've got a multiplier less than one.

Ramey also finds the same result--shrinking consumer and investment spending after a rise in military purchases--with her Survey of Professional Forecasters methodology, and that only looks at post-1969 data.

As you note, in her recent literature review, "Does Government Spending Stimulate Private Activity," Ramey briefly summarizes a variety of multiplier estimates, and "concluded" that the deficit-financed government spending multiplier was "probably" between 0.8 and 1.5 (N.B. The president's American Jobs Act includes tax increases, and so is likely to yield a different multiplier).

As she notes, "This range is quite wide"--and uses that as a launching point for a further discussion. That's all I meant I saw no major figure in the profession "pushing" for multiplier estimates greater than 1: I think the current debate among major figures puts weight on both sides of 1--and a lot more weight below 1 than one would think from reading undergraduate economics textbooks. The age when we started our debates with spending multipliers of 2 or more is long gone.

More importantly, Ramey notes that "For the most part, it appears that a rise in government spending does not stimulate private spending," and "For all but one specification, it appears that all of the employment increase is from an increase in government employment, not private employment."

So again, her reading of the broader literature--not just her work on military spending--is consistent with a spending multiplier and a jobs multiplier both around one.

I really think that "tax multiplier > spending multiplier, spending multiplier around 1 or less" is likely to become a default view for most settings. It'll take a while, but that's where the empirical work by top economists has pointed lately.

Perhaps we'll find, a la Sacerdote & Feyrer's recent paper, that the right kind of government spending has a bigger multiplier (setting aside productivity-enhancing investment--I'm thinking short-run here); or as in the Christiano/Eichenbaum/Rebelo model, the zero nominal bound really matters a lot. And these special cases deserve special attention. But that discussion should begin from a benchmark, and I suspect that the benchmark estimate of the government spending multiplier in coming years will be quite a bit lower than the typical economist thought during the 50's and 60's.

[typo corrected--Econlib Ed.]

Garett Jones writes:

@Matt:

Thanks for your comments.

On #1: It's possible you're right: Since the government and its contractors tend to be highly educated, we already know that it'll be hard for stimulus to target the unemployed---so maybe 42% from unemployment isn't bad, conditioned on who they targeted.

But that's just my point: The decision of who to target was the key decision. The decision to target the funding at the high-skilled was a decision (perhaps implicit) to accept a smaller multiplier.

On #2: I can't recall how many of may labor market tightness indicators I mentioned in the interview, but let me set out a few here--and I'd like to note that I'm really interested in whether the stimulus was targeted at sectors of the labor market that were relatively tight. Of course, we can only learn so much from a one-time cross-section, but with that caveat in mind:

A. As you noted, about half of respondents said it was easier to hire than before the recession. For Keynesian purposes, what matters is that a lot of firms had to hire in what they considered to be a labor market that was "normal" or worse--one sign of Brad Delong-style stimulus bottlenecks.

B. We asked workers about wage and benefit increases: the median respondent reported a slight wage increase; about 40% reported better benefits than in their old job, only about 20% reporting worse benefits than before. This looks like organizations facing an upward-sloping supply curve for workers, not the flat labor supply curve of a slack labor market.

C. We checked to see if organizations that received more funding were more likely to be the organizations that found it easier to find good workers. Under multiple specifications, the answer was no. Stimulus was not well targeted at slack sectors of the labor market within our sample.

So multiple measures found that in our survey, the average stimulus dollar went toward organizations that were facing something like "normal" labor markets. Certainly, some of the cash went for contracts where it was a snap to find good workers quickly, without facing hiring bottlenecks and rising wages and benefit packages. But I think most voters would be surprised that that wasn't the case 80% or 90% of the time.

Garett Jones writes:

@Charlie

Erratum in my second sentence: 0.9, not 0.8.

[Corrected now in original comment--Econlib Ed.]

Mort Dubois writes:

There's an awful lot of reliance being placed on an extremely small sample of an enormous and highly varied economy. If Hayekians don't believe that government can ever gather sufficient information to make good decisions about what to do with resources, why should we think that academics are likely to do any better? I run a small manufacturing company that does some work with the government. My experiences vary from the conclusions drawn here. Here are some examples:

1) Government spending allowed me to keep workers on the job whom I would otherwise have laid off. Contrary to Russ's supposition that management would rather have workers do nothing than fire them, it makes no sense to pay people to do nothing, particularly when you have no money coming in. Laying people off is no fun, going bankrupt is even worse.

2) In my corner of the manufacturing economy, I found it much easier to hire good people after the recession than before. Many of my competitors failed, and I was able to cherry pick employees who I never would have had a chance at before. It's worthwhile to note that, in my experience, the best workers spend a very short time unemployed. Good workers may spend longer, if their skill set is oriented towards a contracting industry.

3) I do agree that many jobs are extremely skill-specific. That makes perfect sense in our specialized economy. That is undoubtedly contributing to the persistent unemployment we are seeing.

This was a very thought-provoking podcast, and I thank both Russ and Garret for their thoughts. I am left, as usual, with some dismay at the disconnect between academics and actual business people, and real life economics. Russ: your occasional guests who are actually in business are invaluable in providing another point of view. Have more on the show!

Mort

engineer27 writes:

Part of the philosophy of the stimulus was to get the funds into the economy right away. Therefore, projects had to be already in the pipeline. So, it is natural to expect those projects to be in areas that were considered "high value" before the recession -- cutting edge energy and health research, for example. Thus, we should expect the Jones sample to be weighted in that direction, thus running into labor supply issues (and others). I note that this supports John Taylor's closet-Keynesian view from a few weeks ago, viz. that stimulus would work if it could be designed and implemented correctly, but that is difficult to do.

On a more general PSST note, is it possible that we have overspecialized ourselves to the point where our economy is no longer functional? My field is, broadly, Electrical Engineering, but I find that employers are seeking extremely specialized sets of skills and knowledge, and seemingly would rather have their businesses continue to limp along short-handed rather than hire someone with extensive experience in related disciplines but no product or industry-specific knowledge. That seems to me to be an example of an Unsustainable Pattern of Specialization.

Modern M. Theory writes:

In this recent podcast you made reference to the fact that cutting gov't spending is more effective then hiking taxes in a period of fiscal adjustment. There is a simple explanation for this, historically central banks ease when the public sector tries to run a surplus because of the subsequent fall in aggregate demand. When governments hike taxes central banks do not anticipate the fall in aggregate demand and actually interpret the taxes as inflation due to our inflation measurement mechanism therefore HIKE rates precisely as the economy is about to slow. All this intervention causes the appearance of higher taxes being more punitive then cutting spending when we can easily see they are the exact same thing.

W.E. Heasley writes:

Excellent episode of EconTalk!

Regarding Romer and Bernstein’s choice of carrying the multiplier of Keynesian foolishness to the degree of 1.52, the extra decimal place is likely associated with Jared “the king of spin” Bernstein the non-economist chief economist.

Bernstein is so very sure of himself, so very omni-present and omni-science, [unfortunately he is sociologist but don’t tell him that as you will burst his egotistical bubble]…. surely the 1.52 exaction is a “king-of-spin” arrogant creation.

Eric Mauro writes:

I appreciated the small unresolved discussion about worker training and skills. I actually worked for a while as a web designer before the tech bust of the early 2000s.

"The problem with that criticism, though, and this is where I find the mystery, is that when things are good, and 4 million jobs are getting destroyed every month and another 4.2 are getting created, somehow people just move in and out of jobs. You quit your job, you don't worry about finding a job, in good times"

In my experience in a tight labor market for a technically skilled job, the company will do the training, or pay for some time for a new employee with suboptimal skills to learn to do the job. It could be 3-6 months but the company figures they have to do it to get an employee fast enough, they can't wait for a kid to get out of college with an optimal skill -- in fact the way college works it's almost guaranteed that the skill training you get there will be outdated.

In a weak economy maybe the companies count on the employees to go out and pay for the training themselves. The problem is that the training is never as directed by the market as it would be if you were learning right in the company.

Charlie writes:

Garett,

I appreciate you taking the time to respond.

"That's all I meant I saw no major figure in the profession "pushing" for multiplier estimates greater than 1: I think the current debate among major figures puts weight on both sides of 1--and a lot more weight below 1 than one would think from reading undergraduate economics textbooks. The age when we started our debates with spending multipliers of 2 or more is long gone."

This statement and your statement for a multiplier "benchmark" is very problematic. The problem with this line of thinking is that modern Keynesians whether "Old New Keynesians" or "New Old Keynesians" think that fiscal spending only has an effect when the economy is not at full employment. Maybe there were some people in the 50s and 60s that thought otherwise, but it's at least arguable that Keynes never thought that way.

It points out an obvious problem with these "benchmark" natural experiments. If the economy is humming along just fine, then there is no reason to expect fiscal policy to raise private spending. Thus, in this natural experiment you would be averaging times with a high multiplier with times with a low multiplier even with undergraduate text book theory.

Auerbach and Gorodnichenko (2011) control for expansion vs. recession using the Ramey news dates and find multipliers around 2.25 in recession and .2 in expansion. I think that is quite in line with moderate textbook macro as taught for a long time. It's been awhile, since I took undergraduate macro, but I imagine this is consistent with Mankiw's book, for instance.

There is also Fisher-Peters that find multipliers of 1.8 using shocks to defense contractor earnings excess returns. Their critique is that if you have $1 million more defense spending, not all of that was necessarily unanticipated. Maybe a big increase in spending was $600k of that and the other was already expected. Thus, under there argument the Ramey numbers are biased downward. They try to control for that using excess returns. An excess return would be a better instrument, because it controls for the unanticipated amount of spending.

In this light, I just don't see how you can report there is no one calling for multipliers larger than 1. Maybe I should excuse you, since these papers aren't in journals yet, but the Ramey article was just published a few months ago and the referee process takes time. These critiques of war time gov't spending as an instrument have been around for a long time, and the counter-argument of the Keynesian side were poorly represented.


xian writes:

on significant digits:

if u carry out a calculation using numerical data, then there r rules for how precise the calculated result should b (ie how many digits the answer should have).

just bc someone reports a figure of "1.52" doesnt necessarily mean they think the "1", "5" or "2" digits in their answer is cast in stone.

it just means that the data they used to calculate an answer had digits justifying reporting three significant digits.

it's a convention of numerical calculation.

it's obvious the haters of such numerical answers have never rigorously carried them out themselves, or have forgotten elementary convention.

mind u, this does not mean the calculation/answer is valid, it just means the reported answer is being held to convention.

more here:
http://en.wikipedia.org/wiki/Significant_figures

i have not checked the calculation myself, but would give the benefit of the doubt that a reported answer of "x.yz" doesnt mean anything other than convention.

it doesnt have to b nefarious over precision.

PrometheeFeu writes:

@Garett Jones:

I found very interesting the idea of your mentor to try to figure out when businesses started anticipating government spending. The methodology leaves me a bit queasy, but it is clever and is a good stone to add to the edifice. However, the estimation of the multiplier seems a bit strange to me. Wouldn't the multiplier be expected to be higher during recessions than during normal times? If AD is down and there are lots of unemployed resources, then you would expect the government to take fewer resources from the private sector and therefore the private sector would shrink less, right? That's obviously an empirical question and I wonder if someone has looked at that using your mentor's dataset. I understand your work contributes to bridging the gap, showing that the government still uses up plenty of employed resources during recessions, but still, I'm not sure we can close the book on that one.

Mark Crankshaw writes:

Russ,

Thanks again for the mentioning of JOLTS data! I've been working on that survey at the BLS (as a mathematical statistician) since 1999 and it's nice to know someone appreciates our data. Cheers!

Garett Jones writes:

@PrometheeFeu

It's never time to close the book! Especially in a nonexperimental field, there are always good questions to ask. And of course, one certainly shouldn't close the book after a one-hour interview....

The Auerbach and Gorodnichenko (2011) paper that Charlie mentions above looks at the question you pose--whether multipliers are bigger in recessions. They also mention a couple of other papers that use similar "good times v. bad times" methodologies.

Indeed, it looks like the spending multiplier is bigger in recessions---but based on their definition of "expansion" vs. "contraction," the U.S. has likely been in expansion for about two years now, so taking their model extremely literally (likely a mistake), we'd expect a near-zero spending multiplier for any spending that started since January 2010.

Of course, by the "looking out the window" test, there's still plenty of slack in the labor market, but as I noted in my papers with Rothschild, it looks like the stimulus wasn't well-targeted at sectors with slack: Those green energy jobs, for instance, were targeted at highly educated workers. And in still-unpublished results we found that the well-educated were especially likely to be poached rather than hired from unemployment.

The better benefits for stimulus workers, the relative difficulty organizations had in finding good people, the higher-than-average rate of poaching---all point in the direction of Delongian stimulus bottlenecks, all point toward relatively poor targeting.

So multipliers are probably bigger when stimulus is targeted at slack sectors of the labor market, just the way that Larry Summers would have recommended. Indeed, the American Jobs Act appears to be at least somewhat better targeted at labor market slack (school remodeling, not solar energy research)---so to the extent we applaud the Obama Administration for doing a modestly better job designing the AJA, we're critiquing ARRA.

Another important question is whether the multiplier has changed over time---one feature of the last three recessions is the slow recovery of the labor market, with or without stimulus spending. If we can figure out why labor markets have been slower to recover from the last three recessions, we'll probably get some insight into whether multipliers are bigger or smaller than in the 50's and 60's.

I suspect that most well-paid workers are more specialized--harder to substitute, harder to replace--compared to the days of Mad Men. I think that's part of the reason labor markets are slower to recover after a recession, and it probably means that workers are less willing to switch careers for one good temporary government job, and firms are a bit pickier about resumes (This is related to Sargent and Haltiwanger's separate literatures on "economic turbulence"). So both demand- and supply-side pressures push in the direction of more bottlenecks and a smaller multiplier.

There's a small recent literature on whether spending multipliers have changed over time, and they modestly support my views: All three of these papers find that the spending multiplier has fallen at least slightly in recent decades:

The first two use Blanchard-Perotti VARs for the US:

1. Pereira and Lopes 2010:
http://cemapre.iseg.utl.pt/archive/preprints/424.pdf
"Our evidence suggests that fiscal policy has lost some capacity to stimulate output but that this trend is more pronounced for taxes net of transfers than for government expenditure...."

2. Hauzenberger's dissertation:
http://cadmus.eui.eu/bitstream/handle/1814/17875/2011_hauzenberger.pdf?sequence=2

"The main result accords well with the conventional wisdom of a declining effectiveness of fiscal policy..." The paper has a result that is difficult to believe (the long-run multiplier is bigger than the short-run multiplier) but in a small literature I mention it.

In the Euro area, similar findings, a bit stronger. Checking other rich regions of the world is helpful, especially when one has little data:

3. Kirchner, Cimadomo, Hauptmeier; an ECB study from 2010:
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1626267

"The results show that the short-run effectiveness of government spending in stabilizing real GDP and private consumption has increased until the end-1980s but it has decreased thereafter. Moreover, government spending multipliers at longer horizons have declined substantially over the sample period."

Part of the reason for the declining multipliers is probably better, more Taylor-rule-like monetary policy: The Fed tightens when it sees fiscal stimulus coming. But structural change in labor markets could be part of the story as well--both stories deserve attention.

In our current political environment, there's probably a tradeoff between fiscal and monetary policy: The House GOP will only allow so much "stimulus" regardless of its source. So even at the zero nominal bound, more fiscal policy action probably means less-aggressive nontraditional monetary policy--the Taylor rule's tradeoff is implicitly still with us as a political reality.

Comments for this podcast episode have been closed
Return to top