Russ Roberts

Ramey on Stimulus and Multipliers

EconTalk Episode with Valerie Ramey
Hosted by Russ Roberts
PRINT
Wapshott on Keynes and Hayek... Avent on Cities, Urban Regulat...

Valerie Ramey of the University of California, San Diego talks with EconTalk host Russ Roberts about the effect of government spending on output and employment. Ramey's own work exploits the exogenous nature of wartime spending. She finds a multiplier between .8 and 1.2. (A multiplier of 1 means that GDP goes up by the amount of spending--there is neither stimulus nor crowding out.) She also discusses a survey looking at a wide range of estimates by others and finds that the estimates range from .5 to 2.0. Along the way, she discusses the effects of taxes as well. The conversation concludes with a discussion of the imprecision of multiplier estimates and the contributions of recent Nobel Laureates Thomas Sargent and Christopher Sims.

Size: 31.7 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: October 11, 2011.] Effect of government spending on output and employment, the effectiveness of what has come be called stimulus. You've done a lot of important work in this area, including a very important and useful survey of past work done by yourself and many others. And you lay out the range of effects that people have found. We'll put a link up to that paper as well as to your other work on the impact of spending and military spending in particular on the economy. Let's start by talking in very general terms about a term I think people hear all the time, certainly in economics literature, but it even comes into the everyday language of the press occasionally, which is the multiplier. What do people mean by the multiplier when they use that term and what are the relevant magnitudes that interest us in policy? The multiplier is a concept that comes from the good, old-fashioned traditional Keynesian model that many people may have seen in their first course in macroeconomics. And it's simply asking the following: If the government increases its spending by $1, how much will overall Gross Domestic Product (GDP) or output increase? So, a multiplier of 1 means that if the government increases its spending by $1, GDP will increase by $1. If we have a multiplier of 1.5, it means that an increase in government spending of $1 will increase GDP by $1.50. Which is glorious. Now, since GDP consists of the sum of government spending, consumer purchases, and investment, a multiplier of 1 is basically saying that the government increase in $1 is not leading to more consumption or investment or net exports, which is yet a fourth component. It says it's just going straight through to GDP and not really stimulating the private economy. So, people are really looking for multipliers greater than 1 in order for short-run stimulus policy to have a good effect. And conversely, a multiplier of less than 1 indicates that there's been some destructiveness of that increase on the private sector. Exactly, correct. And that's what we've traditionally called crowding out. So, if the government increases its spending by $1 and we see that GDP increases by $0.60, that suggests that that government spending has crowded out some kind of private spending, such as investment or consumption or net exports. Now, one of the confusions in thinking about this carefully, of course, is that measuring that and drawing policy conclusions from that magnitude is going to depend on how the government spending is financed, and if the government spending is debt-financed there are presumably future taxes that are going to have to be raised that have their own independent effects, I assume. So, when we are talking about a multiplier of less than one, equal to one, or greater than one, we are usually ignoring any future impact from tax increases unless they are anticipated. Is that a fair summary? Yes. So, there are several notions we have to think about in terms of taxes. When I talk about this typically I try to focus on the question at hand right now, which is: What is the impact of an increase in government spending that's financed with deficits now, but where people know that taxes are going to be raised in the future? If taxes were the sort of thing like head taxes, where if you are alive you pay $100 sort of tax, it's not clear that there would be much difference whether it's deficit-financed or tax-financed now. But we don't tax people that way. And the evidence suggests that the way we do tax people, which is basically taking a percentage of their income, or taking a percentage of their profits, has a much more negative effect on the economy. The leading estimates are the ones by Romer and Romer, who find tax multipliers of between -2.5 and -3. Meaning that a $1 increase in taxes leads to a $2.50-$3.00 decrease in GDP. Big impacts, distortionary effects, of taxes. Which suggests that government spending better be doing something really productive if it's going to be financed. That's right. So, presumably when the economy faces 9% unemployment, people might be willing to get that short run benefit and pay that long run cost. But that short run benefit, as you say, had better be pretty big to make it worth it to pay that long run cost.
5:53Let me ask you a question about that effect of taxes. I'm someone who generally would like smaller government and am not a big fan of the stimulus, so I have that bias, so I tend to applaud--and people like me tend to applaud--these numbers that find large negative impacts of taxation. Why do they find those? So, let me challenge my priors here. Given that labor supply, certainly for men, less so for women, is relatively inelastic--that is, not so responsive to changes in net income that you get from an extra hour of work--how do they generate, what's going on, what's the underlying economics for why taxes are having such large negative effects? Just to make it clear, the argument is that when you take away a portion of what people earn, they are going to respond to that by not working as much; and the economy will get smaller and that's going to offset the gains that the Keynesian multiplier promised. Given that people don't seem to respond very much in the microeconomic literature to changes in tax rates, why do Romer and Romer, Christina Romer and David Romer, find such big effects? They haven't looked at it, but here's the following explanation based on what we know from various empirical studies in the literature. So, you are absolutely right: if you look at a typical full-time employed male, we know that they do not shift their hours much in response to changes in tax rates. You are absolutely right on that. However, now more studies have looked at what we call the extensive margin, which is not just changes in hours while you are employed, but rather your movements in and out of employment; and in particular at both ends of the life cycle. So, for example, people in Europe, on average work fewer hours over their lifetime than people in the United States. And they are taxed a higher rate. Ed Prescott first made this point: why do Americans work so much more than Europeans? And he argued it's because of the difference in the tax rate. And when other people, such as Richard Rogerson looked at this more closely, what they are finding is even though the average hours aren't that different for prime-age individuals, say people ages 30-50 something, you see much lower hours worked in Europe for those who are in their 50s and 60s in the United States, and also for the younger people. So, what we are seeing is movements in and out of the labor force in a way that seems to be a response to different tax rates. Of course, one of the challenges that is lost in the microeconomic literature typically is a variation of the point you are making about Prescott and Rogerson's work, which is in many ways it is lifetime hours that matter, not this particular quarter when I survey you. And high-income, high-productivity folks--not the same thing--but people who are both tend to work very hard while they are working and retire earlier than people used to do in the past. So, it looks like they are not responding to incentives, but they are; you just have to look over the whole lifetime. Right. Exactly. And it's not just taxation of labor income. When Romer and Romer looked at tax increases, they didn't distinguish what kinds they were. So, some of them could have been taxes on firms, variations in investment tax credits. So, these tax effects could be working through job creation, and when you raise taxes, there are fewer people who decide to start businesses; profits are lower so they want to hire fewer workers. So, overall the growth in job creation might be made lower by the increase in tax rates. And that could also explain why the tax multiplier is so negative.
9:56Let's talk for a minute about what others have found. I want to look at the general empirical challenge of trying to assess the impact of government spending on the economy. Every once in a while on this program I make the joke, which I have to repeat, which is: How do you know economists have a sense of humor? And the answer is: They use decimal points. So, obviously you have to use a decimal point when you are under 1; but when you are over 1--for example, I think the Administration's estimates of the impact of the stimulus, which were I think written by Jared Bernstein and Christina Romer, the same one of the Romers we've been talking about, they used a multiplier I think of 1.52. Not 1.5 or 1.57 but 1.52. Some of the estimates I saw. That's a very precise estimate I saw of something I think we are about to learn is something that is hard to measure precisely. Talk about the general challenge. What you are trying to do in these exercises when you assess the effect of government spending is to "hold everything else constant," or control statistically and assess the impact of government spending. Why is that so hard to do with any precision? The main reason that it's difficult to do is that economists are not allowed to run experiments on the economy. So, think of the example--I was at the International Monetary Fund (IMF) a couple of weeks ago--and I said: Ideally the IMF would go out and take a bunch of countries that were similar, and put in some control variables, and then randomly assign countries to those who would have a 2-year increase in government spending that was known to be temporary, deficit-financed and taxes would go up later to finance it; and the control group of countries wouldn't have anything like that happen. Then 2 years later the good statisticians from the IMF would go in and run some very simple statistical analysis to see what was the difference in the growth rate in GDP in the countries that were the treatment group, where government spending was just randomly increased, compared to those countries where it didn't. Unfortunately--well, probably fortunately--the IMF is not allowed to run that experiment. So, that's why economists have to use econometrics. And in fact the Nobel Prize winners this year, Chris Sims and Tom Sargent, were instrumental in improving the way that macroeconomists try to reproduce controlled experiments in a very uncontrolled environment. So, there are several issues. Suppose we say: What happens to the economy when the government has stimulus packages? What we really want to do is compare it to the counterfactual: What would have happened had they not had a stimulus package? The problem is: it's very hard to figure out the counterfactual. So, there's the infamous diagram by Christina Romer and Jared Bernstein that says: If we don't have a stimulus package, the unemployment rate will go up to 8%. 8 point something. And then, with the stimulus package, it will only go up to 7%, 7 point something. It will decline more quickly. Exactly. Now, in fact a huge stimulus package was passed, and the unemployment rate went up almost to 10%. It went a little over 10%. That's right. So, the question is: one possibility would have been that had we not passed the package, the unemployment rate would have been 12%. Or, it could be that the stimulus package didn't do anything and the unemployment rate is exactly what it would have been had it not been passed. Or it could be that the stimulus package actually made it worse. It's so hard to figure out what the counterfactual is. And that's the statistical challenge that we face as economists. And that's why people are giving so many answers. It's not that they are doing wrong techniques. It is that it is very hard to get a handle on this, and there is no one way to do it. There's also a simultaneity problem, which is that: when a government decides to increase spending, how it does it, and what it does alongside that in the monetary area and in tax policy and in regulations and everything else. All going on at the same time and you can't always control for that. First of all, you can't always control for all the things that have changed. That's one of the general problems when you have a multivariate world, a world where more than one thing changes at the same time. But worse than that, you've got causation running in both directions. You could have the economy doing really poorly, encouraging stimulus, leading you to a false conclusion that stimulus causes the bad economy. Or vice versa: the economy would have recovered anyway and then because correlation is not causation you could falsely conclude that. To put it in layperson's terms, you have a big, messy set of complexity and you don't have data on everything, and you've got to do the best you can. 1535
15:01Before we go to your particular work, two special challenges in these empirical measures: one of them is expectations, which are never measured directly but which of course matter. People will look ahead. And the other is timing, which is related to expectations. In your work you spend a lot of time worrying about when did those expectations really have a chance to form. Let's talk about expectations generally for a minute. There was a big debate in the 1980s in the economics literature which is of course now coming back to life about when future tax increases would induce behavior now. When we talked about it before when we talked about the Romer coefficients, the negative tax multiplier, you were just looking at the fact that when a tax increase comes, people respond. But in fact, a lot of people argue that they respond now, even though the tax increases are coming later; and the same would be true for changes in government spending. Why are those expectations relevant and how can you possibly measure those effects accurately? Let me give two examples where they are relevant and then talk about trying to measure them. Let's talk about taxes. Suppose you were told and you believe that taxes would be unchanged this year--let's take sales taxes because that's really clear--that sales taxes are going to stay where they are this year but that next year they are going to go up to 15%. If you are somebody who is thinking of buying a major durable good such as an automobile next year, what would you do. Try to move it forward. So, in fact, thinking that taxes are going to be higher next year can actually spur the economy this year. Now, consumers will pay attention to this on the margin. People will often change their purchases by several months. But for the most part, consumers don't spend as much time on this. Businesses really pay attention. Some of the work by Christopher House and Matthew Shapiro shows that the base end of some of the Bush tax cuts, the particular parts on business, actually led the recovery from the 2001 recession to be slower than it otherwise would have because people were waiting for those tax cuts that they knew were going to be phased in in the future. That's fascinating. I wonder how robust those findings are. It's certainly a real effect; the question is always how big is it. Yes. The second thing is with respect to government spending; and that's where I've been doing work. What I found is that when there are major increases in government spending that are anticipated, that everybody's paying attention to, and the major increases I'm thinking about are the period that led up to WWII; what happened when North Korea baited South Korea in 1950; the Vietnam War getting off; and part of the Reagan buildup; as well as 9/11. That, even though government spending cannot increase much in the short run, people know it's going to increase later, and firms start gearing up. Consumers start responding, is what I see in the data. So, even though you don't see the government spending increase right now, it's that anticipation of the future increase that leads to a change in behavior now. And it affects how you measure the impact, because if change has already taken place and you start to measure from that baseline, you are going to get a very different estimate than if you start at the right time. Exactly. So, the standard way of doing this takes the change in government spending itself to be the news about the future; and therefore they are using the wrong baseline because they are looking at it from that point in time. What I did in my research was to go back to Business Week for the most part, but also The New York Times, and look at where Business Week predicted a major change in the path of government spending; and then use the variable at that point; and found that the economy started responding right away. Which is really incredibly clever and interesting and could actually be true.
19:18Now we're going to segue into your work. One of the reasons you looked into military spending--and you are not alone in doing this; a number of researchers have tried to just look at military spending because that has the best claim to being exogenous. Meaning not complicated by this possible two-directional interaction between the government response and the state of the economy. It's independent. It's closer to a natural experiment. Correct? Exactly. Particularly the big build-ups. So, I try to focus on ones that were driven by external political and military events. During normal times what I found from reading Business Week every week from 1939 to the present--almost as impressive as Allan Meltzer's reading of all the minutes of the Federal Reserve since 1913, not quite as impressive, though. But still very impressive. I would rather read Business Week--a little bit more exciting. But during normal years, how much they cut the defense budget is related to how much of a deficit there is; so we want to make sure to take those parts out. But the really big increases that I focused on, or, say, the fall of the Berlin Wall leading to the end of the Cold War, are ones that are not related to how the economy is doing. And that's exactly why so many of us like to use military spending. Tell us what you find when you do that and look at the impact of maybe the closest thing you can find to exogenous spending, an outside threat that's not related to the state of the economy or the deficit, that everyone agrees we've got to spend more money; that money kicks in at date x, but you argue that before that people start responding to it because they realize it's almost inevitable or at least has a high probability. And what do you find the impact of those government spending episodes is. I find estimates of the multiplier--a $1 increase in government spending, what does that do to GDP--that range from 0.8 to 1.2. So, 1 is right in the middle there. It depends on the sample and small little ways of calculating the multiplier. The standard errors are pretty big. Which means? Which means that the amount of uncertainty about the statistical estimate is quite large. It's imprecise. It could be a lot lower than 0.8; it could be a lot bigger than 1.2. Exactly. Don't you have some that are 0.6, also? Robert Barro has 0.6. I might when I just did it over WWII, and we can talk about that when we talk about the lower bounds. I find that in the short run, consumption actually falls a little bit. For the most part it's statistically significant, but consumption falls and then recovers somewhat as government spending starts going back down. That's completely consistent with the mainstream neo-classical model, but it's completely inconsistent with the traditional Keynesian model. So, it's a challenge to the Keynesian model. I'd add that it's consistent with the classical model; you don't have to go neo-classical. Right. The classical argument would say: the government spends more; there's less resources available for other things, and so by definition people are going to have to consume less. If you have more tanks, there's less butter. Exactly. Whereas the Keynesian argument is: more tanks, tank-makers have more money, they spend it, that stimulates the butter industry, they eat butter. It requires some general slackness, not just specific slackness. So, on investment, I tend to find that investment actually goes up at first and then falls. And that's also consistent with the neo-classical model. Private investment, you are talking about. Yes. Because initially firms know there is going to be an increase in government spending and they actually want to increase the number of factories they have. Now, part of that consumer spending gets crowded out, but at the very beginning they do increase; but then it goes down as it becomes more and more crowded out. That's because there's more steel going into tanks and less available to build a factory, etc. The price goes up. Yes. Price effects; labor gets more expensive; if it's not totally unemployed there's all kinds of effects that are going to discourage private consumption and investment. Also I find that hours worked does increase; employment increases. But, real wages measured at the firm level--so it's the wage of the worker divided by the price of your output--those actually go down as well.
24:40Don't you find that most of the employment increase is government employment, not private? Yes. So, that's my more recent paper, the working paper, where I changed the question a little bit because it became obvious we were going through this recession and I don't even think we have now what we've called a recovery--it's just sort of staying near the bottom. That, in a way, we don't care so much about the multiplier on GDP. Certainly that matters, but what really matters is: Can government spending create jobs? That is the central question. What was happening at the beginning of the recession was employment was going down. GDP wasn't going down so much because productivity was increasing. An increase in productivity is good, but we'd like the government stimulus to create jobs. So, I started looking back through the old data and figured out that almost all of the increase in employment that I was finding from government spending was an increase in government employment. So, for example, during WWII and during the Korean War, we had the draft. And when I look at the number who were in the military, a big part of the increase in employment is just an increase in the number of people in the military. Now, this is very consistent with the work of Robert Higgs, the economic historian, who argues that the standard story, which we've talked about a number of times on this program--the New Deal didn't cure the Great Depression because it wasn't big enough, it wasn't a real test of Keynesian stimulus and then Roosevelt messed up in 1938, raised taxes, and the money supply contracted, so we had a double dip; but in 1939, 1941, as the U.S. government started cranking up the war machine, that finally got us out of the Depression. And what Higgs argued is that private consumption went down--and I think it's very important to emphasize that the endless claim that war is good for the economy is a very strange claim given that wartime England, Germany, and the United States had to be some of the most unpleasant times to be a consumer and to have normal well-being. Everyone understands it's a time of austerity, not prosperity. In fact, you could argue there are other things going on. That's fine. But it's hard to understand the argument. It's clearly good for tank-makers. And what Higgs argues is that unemployment went up because people went into the army--that's the point you just made--and that sure, GDP went up because we are artificially counting in GDP a tank but the real thing we care about is the butter side, the private consumption of the goods and services that we get pleasure, satisfaction, and utility from. But in Higgs's work--and in this I want to challenge his conclusion and yours--one of the challenges for him, and I think for your work is its extremely difficult to measure consumption in wartime. Because first of all we have price controls. If you look at the raw data, I think Higgs shows actually consumption goes up. But that's because, Higgs argues, because we mis-measure prices. There's a lot of stuff that is artificially cheap; there's stuff that wasn't even available, you couldn't even buy it because of rationing; and so you can't take the standard government data series and time series on their face. You've got to look deeper. How do you deal with that in your work? You are looking across dozens of data on consumption and government spending, etc. How do you deal with the fact that the prices, which are a huge challenge, are probably not accurate? I did the best that I could. That's an honest answer; a rare one. We only have quarterly data from 1947 on. And because in my initial paper it's all in the timing, I really needed to have quarterly data going back to 1939. Because that's our best natural experiment--WWII. So, I found in a government publication data on nominal--that's in dollar terms--GDP and all the components: consumption, services consumption, non-durable, durables, those sorts of things. Then what I did was take consumer price indexes and producer price indexes and spent a lot of time trying to figure out what kinds to use, and then used those to try to turn the dollar-denominated variables into "real" quantities. Corrected for price changes, purchasing power then. Exactly. I was doing this independently and just happened to get an email from Robert Gordon at Northwestern U., who didn't realize I was doing this as well. He was working with one of his students, where they were going back to 1919 using interpolation to try to create data from 1919 through the 1940s. Well, we heard that each other had done it; we'd done it completely different ways. We shared the data and started plotting them against each other; we were astounded how close the two data series were. So, we felt it's nice that we used different methods; and the fact that we were getting similar results made us both feel really good. It's some benchmark at least. But to be clear neither of us was dealing with the fact that there was indeed rationing. There was very severe rationing on durable goods such as automobiles--basically the automobile manufacturers stopped making automobiles and started making tanks. So at any price, consumers could not buy a car. There were also tires and rubber and all the standard things one hears. Sugar. The rationing started going in in spring 1942. I'm still seeing somewhat slower growth in consumption even before that. Gives you a little bit of justification. I think the rationing, some of it, dampened it somewhat. But I think we would have seen the dampening. And even when I omit WWII and just have Korea and on, I'm finding the decrease in consumption as well.
31:00So that raises an interesting question. One of the more fascinating natural experiments is just starting to come out in the blogosphere and some essays and journals is the post-WWII period. Not WWII, not Korea, but 1945-1946--an incredible natural experiment. Government spending falls by 60-70%. The Keynesians of the day, including Paul Samuelson, who becomes one of Keynes's great champions, predict a horrible contraction. Samuelson predicts the worst depression we will have ever seen if we let government spending fall in the aftermath of the war without planning for that. Of course, that planning does not take place. Government spending just plummets. Millions of soldiers--some go to school through the GI Bill, but millions of soldiers and production workers from military spending are suddenly looking for work. And incredibly, job creation takes off like crazy. So, in that time period, that's in your data, correct? And by the way, the standard Keynesian response, ex post, is: Well there was a lot of pent-up demand. That was easy to say ex post; ex ante they didn't seem to notice that as the possible response of the economy. But more importantly, the reason there was pent-up demand is because you couldn't buy stuff--not because of rationing, but because the steel wasn't available to make the car. That's why there was rationing. Rationing wasn't some exogenous thing. Do you have anything to say about that contractionary period in government spending? I think that period is very interesting; in fact I discuss it in some of my discussions of other people's government spending papers. To give you some of the numbers: in mid-1945 the military employment was 12 million; by mid-1946 military employment was only 3 million. The civilian labor force in 1945 was 55 million; by mid-1946 it was 60 million. So, the total labor force shrunk by 6%. Part of that was that women had flowed in to help with the war effort and afterwards they went back home. Some people went to school. But you still had a surge in the civilian labor force by 5 million, which was 9%. And when you look at what happened, the unemployment rate went from about 1% to a peak of 5%. And the question is: How did the U.S. private sector absorb so many workers so quickly? The most recent Journal of Economic History has a paper on this where they said it was export demand because much of Europe's capital stock was destroyed. The United States had a big surge in its export markets and that that was one of its prime movers. This person went through input-output tables; it seems to have been a really nicely-done paper; I've only had a chance to skim it. Found that there was basically a big surge in demand, and that led to the extra job creation. The problem with those kind of explanations for me is always that ex post we can look back and find it. Right. If it hadn't been export-driven, maybe it would have been driven by another part of the economy. My answer, which is not very helpful but still could be right, is: There are times when the economy works well and there are times when it doesn't. When it works well, it's easy to find a job; when it doesn't, it's really hard to find a job; but we don't really understand those differences. I always use the tech bubble or whatever you want to call it--the tech boom and bust. In 1999 or 2000 if you were a skilled person in some aspect of the web you made a lot of money and people were desperate for your services. 2002 that wasn't so true any more, for a whole bunch of reasons; but it still wasn't that hard to find a job. And before the recession hit in 2007 people quit their job with not a lot of fear because they thought they'd find another one. Or if they lost their job because their firm went bust they knew they could find another one if they had skills. That's not so true right now; and I don't think we have a very good understanding of why that's the case.
35:28Let's go to the literature which you've done a magnificent job surveying. I suspect you read the papers. Which would put you in a different group from myself. You survey a great deal of work. Your technique; you are focusing on military spending; you are using a particular style of equations and econometric estimation. Other people look at different measures of government spending; they use different statistical techniques. So, there's a much wider range than 0.8 to 1.2. What do other people find? Give us a feel for the range. One of the points I make in my paper is that the range within papers is often as big as the range across papers. Very important finding, actually. Because it often depends on the sample period. So, why does the sample period matter? Well, for example, if one includes Korea, government spending went up a lot, so it's great to include it. But in contrast to current politicians, the leaders back then felt that if one had a high debt-to-GDP ratio, which of course the United States had after coming out of WWII, that any increase in spending had to be financed by an increase in current taxes. So, the problem is: Does Korea have a smaller multiplier because the multiplier is small or because taxes were increasing? So, including or excluding Korea can make a difference. So if someone is reporting on their findings, one of their findings might include Korea and one would exclude it; and that gives you a big range within that one paper. Exactly. The other sorts of possibilities are just different specification. For example, it's a much older paper, but I hope to see more of the current papers doing this--this is by Davis, Loungani, and Mahidhara. It was never published, but Steve Davis of the U. of Chicago was one of those co-authors. They were looking at military spending across states, and they found that whether they used one employment survey that just looked at establishments versus a household survey that in some measure accounts for workers, that they got very different multipliers in terms of jobs. I'm sure you would. So the range within that paper was huge. It would depend not on issues of cherry picking or making decisions about what period the study should examine, but just what source you use for hours or employment. Fascinating. Another example is a recent paper by Auerbach and Gorodnichenko, who are both at Berkeley, where they were looking at a very interesting hypothesis that the government spending multiplier might be greater during recessions than during expansions. And that has a lot of intuitive appeal because during recessions there are a lot of unemployed people or you might get less crowding out. When they run their baseline specification, they find multipliers of near 0 during expansions and multipliers of 3 during recessions. Three point what? I can't remember. There's a decimal point. I tend to round because I don't believe the decimal points. That's very important, because they've got these placid periods when everything is going fine; but if you include when government spending is swinging around in there you are going to get a very blah multiplier; but if you only focus on the bad times you are going to get the real one, which is the one you care about, usually. Exactly. But when they make a change in the specification, and this has to do with a highly technical kind of point, but in their baseline specification they do not allow the increase in government spending to move us to a different regime, because they do this with sort of a regime-switching model. Whereas in fact you would think--the typical recession doesn't last very long. You tend to expect to get into an expansion before too long. They don't allow for that effect. They assume that one stays in recession. When they modify their model to change that effect, then suddenly their top multipliers during recessions are 1.5. So, even within that paper the estimates are over a wide range. So, one of the things that's fascinating to me, and used to depress me; now I just think it's just the way it is: We are not very good at this as economists. We should admit it. But the fact is you have people who think the multiplier is really big and some of them are scholars of multiplier studies like yourself, who are actually in the data. Some are not--they are just pundits or smart economists who are talking about this even though it's not their research area--and they say we need to spend a lot more. On the flip side you have people who similarly, like yourself, are deeply involved in the data who say: No, the stimulus is not going to be very effective. But you also have people who don't know anything about it and they agree with that. So we have this enormous range of estimates and we have this big bias of priors and ideological presumption. It's kind of awkward for a profession. You've been talking about the within-paper range. What is it across papers? What's the possible range that the multiplier might be, and what's the real, full range? I said I thought the multiplier was most likely between 0.8 and 1.5. Across all the work, not just your own. Right. Some find it as high as 3 point something--that's the one we just talked about. And some tax multipliers are even -5, so the tax multipliers have an even wider range than the government spending multipliers. And some are as low as 0.5. I think in your paper you said 0.5 to 2, because the 3 ones are for special cases like recessions. Realistically I think you could argue that belongs in the range.
42:07One thing this teaches us of course is that assumptions matter a lot, and all of these results come from different assumptions, different specifications, different fundamental techniques. But a range of 6-fold, or merely 4-fold, is not good. No. As one economist--I believe it was Roger Farmer at UCLA said when we were initially talking about the stimulus--he said: Can you imagine if the Pentagon called over to the physicists and said, What's the key physics equation behind this nuclear submarine; and they said, Uh, estimates of 0.5, maybe 2? Would they build nuclear submarines based on people giving them those kinds of answers? And one thing we haven't talked about is recent work of Alesina and others that actually finds that government contraction actually expands the economy, similar to what we were talking about with the post-WWII period. Right. I don't know if that's done in a multiplier context, but I guess that would be a less-than-zero multiplier on fiscal. That's right--it's a negative. The expansionary contractions of fiscal policy suggest a negative multiplier. I have to quote the lyrics from the second Keynes-Hayek video, Fight of the Century," where John Papola and I say: "The economy's not a course you can master in college. To think otherwise is the pretense of knowledge." And that's an allusion to Hayek's 1974 Nobel Prize Address, "The Pretense of Knowledge," where he says: The economy is just too complicated to model in a specific and precise way. What's your take on that? It is. We cannot be as precise as we would like, say, as physicists are. However, we can come up with parsimonious models that are reasonable approximations to the economy and it can give us much more insight than just shooting from the hip. What's the evidence for that latter part of your statement? In other words, I accept the fact that we know more about the economy than we did 80 years ago. When this podcast, if all goes well, gets posted and I think it will, it'll be a week after I interviewed Nicholas Wapshott, who is the author of this new Keynes Hayek biography; and one of the things you have to notice in there is that economic debates in the 1930s about macroeconomics were so frustrating because people couldn't even agree on what the terms meant--what was meant by investment. We've made some progress; certainly made some progress in what doesn't work. But aren't we still shooting from the hip when we've got a 0.5 to a 2.0 and we've got to decide whether to pass some package of increased government spending? Do we have any scientific unbiased way to assess its impact ex ante? Unbiased scientific way. I mean, you are right that I think that sometimes researchers' biases come--we do have a range. Nobody's saying it's 10. In the standard Keynesian model, they often have multipliers of 5. Very few people are suggesting that, and most people don't believe that. We think that probably the maximum is 2; and even with a multiplier of 2, it is not clear that it's worth passing a stimulus that is going to help us in the short run and yet have very negative consequences in terms of taxes later. So, I think people can make reasoned arguments based on the wide range we are giving. That's a great point. Although that level of subtlety often doesn't make it into policy. But it's a very interesting point.
46:30Let's talk about your work in particular, which is similar to what Barro and Charles Redlick find. They are using a very different econometric technique than you are. So again that's maybe some confirmation again that you are both onto something. But let's take the lower bound of your work--the 0.8--or Barro's 0.6, meaning there's a lot of crowding out. And then on top of that you have the taxes, so it doesn't look like a very good deal. Obviously a lot of people have disagreed with you. They don't like that policy conclusion; and their own work suggests a substantial multiplier, and then they'd probably pick holes in the Romer and Romer work, and other work on taxes, and say it's not that large--the negative effect of future taxes--or they'd say, that's true that there are going to be these negative effects of the taxes, but the economy will have recovered somewhat by then and we can afford those. What's your response? How confident are you in your own estimates and how confident do you think people should be in their counter to yours? In other words, somebody who challenges you and says: This military thing, it's a gimmick. What would they say if they wanted to attack you? What kind of responses have you gotten from other economists and what are the biggest criticisms they've made of your methodology? And Business Week--come on, that's a magazine! Criticism--I would say that there's something special about WWII and Korea. Those of us who look at military spending, when we look at it in the aggregate, so Robert Hall, Robert Barro, and I, all agree that there is so little variation in government spending after Korea that there's simply not enough information in the data at the aggregate level to actually be able to identify any multiplier. To tease out the precise effects. Exactly. So that's why all of us have focused on at least Korea and often on Korea and WWII. The key criticism is: We don't want to know what the multiplier is when there's a big draft; and there might be other things going on--people worried about the Soviets during Korea, about Nazi Germany during WWII. There were so many other things going on, rationing, that this only gives us the answer on average over the history of the United States back to 1939. That's not necessarily the answer for what happens if the government raises government spending for two years to get us out of a recession right now. Fair enough. That's the best critique of why you might not want to apply the multipliers I've estimated on historical data to a particular policy that we are considering now. Good point. Of course, the challenge there--it reminds me of a paragraph I'll post on this podcast website where one of the great myths of the current crisis is that the Congressional Budget Office (CBO) found that the stimulus worked; and they estimated 1.x-3.y million jobs, the stimulus package of 2009. When you actually look at what the CBO did, is they took the multipliers they had used to predict the effect of the stimulus and they re-multiplied them times the actual levels of spending rather than what they thought they would end up being. So they basically took a forecast and re-did the forecast and said that's what happened. Which is horrifying and dishonest when it's used by other people. They are very honest about what they did. They add a paragraph which I absolutely loved, which is: Now some say we should actually look at what happened. We should have gone and seen what happened to employment and what happened in the aftermath of the 2009 stimulus package; and they concede, with remarkable honesty, and I think it's something everyone should clutch to their breast: Well, that's really hard to do because then to estimate the effect of the stimulus on employment we'd have to know what path the economy would have taken in the absence of stimulus. Which is what you said a long time ago. The counterfactual. And they basically said: Well, that's too hard. And so we didn't do that. We used the models we already had from the past. But of course the challenge there is that for that to be a reliable estimate of the stimulus, you'd have to be confident that the current situation, that the structural relationships between the different parts of the economy today, were similar to the estimates of the Keynesian multiplier in the past. This is a version of what you just described about what could make a critique of your model. But the problem is that they probably aren't; and particularly because of that forecast the CBO made of how bad it would get--the CBO made their own estimates, not just the Bernstein and Romer estimates--of what would happen with the stimulus passing. Those were grossly wrong. They confront that, too. Because they are honest. There are two interpretations of that. One is: Their model is not very good. The other is: there are some things we didn't realize that were worse than we thought. Strangely enough, they went with the second hypothesis rather than the one that the model isn't very good. But this recession, like every recession, is special. It's driven by debt, a debt-overhang, a destruction of balance sheets of homes, households, and businesses. Maybe everything goes out the window. And I think this comes back to Hayek's point that the economy may just be too complex to model with any precision. That's right.
52:50Let's talk about Tom Sargent and Chris Sims. We are taping this on October 11, 2011, the day after the Nobel Prize was awarded to both of them. What are their contributions, in particular in the area we are talking about? Why are they important for you and people like you for doing these large macroeconomic estimation challenges. Let me start with Tom Sargent's contributions. He has many, many contributions. Let me just focus on one thing. Previously, Robert Lucas, who won the Nobel Prize several years ago, argued that it did not make sense to analyze the effect of policy by simply running a regression on historical data and not taking into account what the expectations were about the permanence or temporariness of that policy were, and then just bringing that to the present. So, for example, all of the big econometric models of the 1960s-1970s would estimate consumption functions and then look at on average how taxes had affected consumption in the past; and would use those estimates to make predictions about what current policy would do. So, it was very relevant for policy. That was called the Lucas Critique. Because he said that those estimations ignored the fact that a permanent versus a temporary tax increase should have a very different and did have a very different effect on consumption and you were lumping them together, right? Exactly. So, Tom Sargent figured out ways to estimate relationships in the data that took into account that the process of taxes might differ over time, or the process of government spending might differ over time. And the way to do that was to estimate what we call deep structural parameters--the average preferences of individuals in the economy or what the production function looks like in firms--and then take those deep parameters and build back up so that we could analyze changes in policy, policies that we have seen historically. So, one of Tom Sargent's big contributions was allowing us to do that empirically, which is what we are trying to do now with the stimulus package. Chris Sims was also working on this, but came up with a nice way to sort of summarize data in something called a vector autoregression. And if you google that you'll get to the Wikipedia page, unless you already understand it. It's a hard thing. That was a way to do it, and he figured out ways to identify when you could have unanticipated increase in government spending or taxes. And then plot out on average through history what happened when there was one of these unanticipated increases in government spending, whether it was more permanent or temporary on average, and how GDP and consumption and those sorts of things, how the path of those variables changed after that policy change. So, he gave us sort of a different way of estimating these things. But they all took into account the Lucas Critique and the need for things to be exogenous. So one of the big debates in the government spending method is a paper by Olivier Blanchard, who is now the Chief Economist for the International Monetary Fund (IMF) and Roberto Perotti had used a standard Sims vector autoregression for the way they estimated government spending shocks, and what I had said was: Well, that way of doing it didn't take into account the anticipation. So I came up with my news variable, which was news, and then just augmented a Sims-style vector autoregression and then looked at the effect when news arrived. So, I'm mostly using what Christopher Sims's principle innovation was when I look at the effects of government spending. What controversy surrounds those techniques that Sims champions? There are quite a few. For example, something we've seen recently, and I've seen it in several other debates, is that sometimes if you don't estimate the vector autoregression correctly--if you don't include the proper variables or the proper lags in these--you can get results that are very misleading. For example, what some people are doing or saying is: Suppose that I know what the economy is, I write a model of the economy. Then they generate artificial data from this economy and say: What if I go and run a vector autoregression on that, given that I've got the results, given that I know what the results are? That I should be able to identify them. And sometimes, depending on how one specifies--sometimes you can get very close to the right results, but sometimes you can be very far off. And that's one of the big critiques of autoregression.
57:52Coming back to Hayek: Hayek was very critical of what he called "scientism"--the use of what looked like science either in terms of statistical techniques or mathematics to apply to things they were maybe not appropriate for, such as economics; and with his view obviously the majority of the profession does not agree. Let me ask you a question, a different version of what I was trying to get at before. Those who are listening who are not professional economists, don't know the amount of work that you put in to do the kind of estimates we are talking about. Just the reading of Business Week--we are talking about a lot of work. Judgment calls you have to make along the way. But people like you, like Tom Sargent, Robert Lucas--I've never met you but I was fortunate enough to take macroeconomics from Robert Lucas, I have tremendous respect for him. He had a deep and inspiring dedication to the truth. He was not a man of fads or the short-term or clever answer. He was deeply interested, when I was fortunate enough to be around him, in understanding how the economy worked; and I've seen Sargent a few times in action and he exudes the same kind of genuine scholarship and incredible tenacity and intellectual rigor. And I so want to believe that the kind of careful and detailed work that people like you and Lucas and Sargent do--I don't know Sims's work--and many, many others in the field are advancing our knowledge. And yet we've got this enormous range. Are you optimistic that that range might narrow down the road? Or are we really in a world--and this is related again to the point I made before--are we really in a world where every case is different? That the reason we really don't understand fiscal policy very well is just because the world's too complicated? What do you think? I am optimistic that the range will narrow. You have to understand that from the 1970s until just a few years ago, studying fiscal policy was considered a backwater in economics; and when I was doing it in the late 1990s, everybody else wanted to study monetary policy. And I felt: Why am I studying fiscal policy? Nobody is paying attention. Now we are having a resurgence in studies on this, and I'm learning a lot from reading the papers of other people who get very different multipliers. With the way scholars work, they talk to each other, they learn from each other's work. You can think of it as a mass of people with different estimates learning from each other's work. And I think we can converge to a narrow range, or at least feel more certain about the range we do have, when one number or another number applies. And my adviser, Robert Hall, would make the analogy to medicine: We've finally figured out that putting leaches on the economy is not a good idea. We still have far to go but I think that to the extent we try to use scientific principles, I think that we do make progress on how much we know. And if we compare medicine to how it was 100 years ago, there has been so much progress made. The human body is enormously complex; by some measures just as complex as the economy. And even though we can't find a perfect model to describe it, having some kinds of useful approximations that get to the essence of the key parts that will help us predict will help us make better predictions about the economy and help us to understand it better.

COMMENTS (21 to date)
Hyteck9 writes:

Federal spending is not the solution. Federal enforcement of fair and just labor laws will fix the economy. Everyone is sooo busy checking their numbers to see where all the money and jobs went... that no one has bothered to think that the root problem is, in fact, the avoidance of such numbers. Companies have re-discovered the greatest financial loophole of all time. FREE LABOR. Sure, they pay you for 40 hours so its legal, but then they intimidate you to work 2nd shift, third shift.. weekends.. suddenly average Joe is working 80, 100, even 140 hours a week for fear of being fired. All those hours going unpaid, untracked, untaxed, and only benefiting the executives of the company. It prevents modivated people from getting a second job to earn extra discresionary income, destroys home life, destroys employee health, moral, etc. It also robs local economies of additional shifts of work, and in turn additional TAXABLE income. Salary exempt and "on-call" HR clauses are now being abused to the point of being criminal and no one listens to the cries for help. Pay people for their time, and the economy will bouce back.

Marc writes:

What possible scenario could lead to positive employment and growth if government deficits are reduced through spending cuts?

Doesn't the credit contraction not only lead to a reduction in employment and knock on multiplier effects?

Matt S writes:

Another great discussion. These are so valuable. It seems I hear alot of this Romer and Romer study. I can't wait to read it.

Thanks!

Fred Giertz writes:

I don't think it is appropriate to dismiss government spending (e. g.,tanks) as worthless since it does not increase private consumption. A multiplier of only 1 in Israel where a billion shekels of spending for defense increases output by a billion shekels would be viewed very favorably.

In the short run, this would appear to be a free lunch assuming the longer run multiplier is not negative.

Mark writes:

Great discussion. I was interested in the Ramey numbers after he mentioned her work in a previous podcast.

I was particularly impressed by her closing remarks comparing economics the development of modern medicine. I am a researcher in the biomedical sciences, and as I have been listening to Roberts' critiques of current economic theory I have often thought that modern economics could learn much from modern molecular/cell biology. We have certainly come a long way since bloodletting. Perhaps econ can follow the same path. Indeed, C+I+G+X-M=Y shares a striking familiarity to the theory that four humors control all health and disease.

As a biomedical scientist, I don't understand how a "macro" theory can ignore "micro" foundations. Indeed, if we have learned anything from biology this past half-century, it is that small changes at a micro level can have defining effects that would not be otherwise predictable at a macro level (think sickle-cell disease, cystic fibrosis, retinoblastoma, XSCID, etc.)

Roberts' critique that the economy is simply too complex to be understood may be correct in part - just as we may never fully understand all that occurs inside and among cells. However, we do not need to understand everything about biology to begin diagnosing problems as they develop in the body. Sure, many of our cures could use some work, but at this point most medicinal interventions are at least better than doing nothing. The same can not necessarily be said of modern economics. By grounding their theories on basic mechanisms, perhaps economics can emerge from the dark ages. I would like to see economists move in this direction.

Shayne Cook writes:

Thank you both.

A suggestion ...
I suspect a substantial explanation of the variance (non-deterministic nature) of the multiplier lies in the examination of the "how" and "on what" Federal deficit spending is applied during recessionary periods. Or for that matter, even during growth periods. Dr. Ramey hinted and examined that for the period of WWII and Korea.

I looked at BEA data for U.S. [nominal] GDP for the period 1980-2010 (Table 1.1.5) It's curious that Government spending (in aggregate) hasn't risen during the 2008-2010 period, as a percentage of GDP. In fact, Government spending is somewhat depressed from prior averages. Of course the (G)overnment spending (G in G+I+C+NE=GDP) only measures Government purchases of new goods and services - not Government re-distribution.

The deficit-financed "stimulus" that has been applied during this economic downturn has been primarily (perhaps exclusively) applied to re-distribution, not direct Government purchases of goods and services. And that re-distribution has had the primary effect of increasing the (C)onsumption component of GDP. It is widely asserted that (C)onsumption spending is 70% of GDP. Well, it is right now, but that is several percentage points above historical average, or even trend. While increased Consumption spending (in nominal terms) has, to a very marginal degree, restored some minimal level of GDP growth, all components of (I)nvestment spending - most obviously new housing, but also for new plant/equipment - remain seriously depressed, again, in percentage-of-GDP terms relative to prior averages.

Looking at the same GDP data for the 1980-1987 (post "Stagflation" era) period, it is apparent that (G)overnment deficit spending - for new goods and services - did increase. The data also, as one would expect, indicates a substantial part of that is attributable to the increased defense spending during that period - the "Reagan Defense Buildup".

Note I'm not arguing the merits of defense spending in all circumstances. Quite the contrary, in fact. I haven't looked at data for the 1960's era, but I suspect similar artifacts regards the stimulus/multiplier effects may be seen in the Kennedy-inspired space sector build up.

Point being, the data seems to indicate that (G)overnment deficit spending has generally greater - or at least more predictable - multiplier effects when it is applied to direct purchases of goods and services rather than to re-distribution, where its GDP impacts are via the (C)onsumer component and all the related vagaries of consumer choice.

In more "Keynesian" terms, Government re-distributional deficit spending apparently has less sustainable (and predictable) Aggregate Demand shift effect than does Government direct purchase deficit spending. I do not consider it glaringly apparent from the BEA GDP data, by the way. Both the Kennedy focus on the space program (and Viet Nam) and the Reagan focus on defense spending were accompanied by significant tax cuts. It remains challenging to isolate the effects of tax cuts versus direct/re-distributional Government spending as contributory - via the "multiplier" - to GDP growth.

Peter writes:

Great podcast!

But I was a bit disappointed that no one mentioned the Sumner critique of multiplier studies (what would the Fed have done differently).

Also, I was pretty confused by some shifts between nominal GDP and the talk about real items.

Schepp writes:

Thank you for the good discussion.

I am concerned about the disregard of what money is spent on. Building a tank to be used up in war is only a capital investment in maintaining our free market. Tank capital investment can also be mis-spent and provide little protection to our markets.

A Toll road that creates a revenue stream to pay back bonds and is reasonable priced (Marginal toll equals marginal benefit) delivers value that is multiplied through the economy.

Toll Roads vs. Tanks would provide a starkly different Return on Investment (mulitplier).

As a check to my own bias Toll Roads priced inefficiently or such that they do no pay off their investors can be utter garbage. Tanks in World War II that stopped Nazi's have provided many years of relatively open markets with a substantial return on investment.

Don Tillman writes:

This is a fantastic discussion; thanks so much.

I've got a question. At around 12:50, Valerie Ramey brings up the "infamous" Romer/Bernstein chart and suggests three rough possibilities for comparing the intended vs. the actual effects of the Stimulus:

  • Case 1: The Stimulus boosted the economy and the unemployment rate would have been higher without it. This means that both the R/B curves were wildly incorrect.
  • Case 2: The Stimulus had a negligible effect on the economy, and the unemployment rate would have been about the same without it. Again, both the R/B curves were incorrect.
  • Case 3: The Stimulus had a negative effect on the economy and we'd have a significantly lower unemployment rate today without it. Of the two R/B curves, the "without" was correct but the "with" curve was wildly incorrect.

The first case might not be possible; isn't there some mechanical limit to how fast the unemployment rate can increase?

Immediately after this, the discussion moves to the topic of expectations. "People will look ahead." Expectations drive changes to the economy. Certainly; that makes sense.

But when the Stimulus was passed the expectations were very high. This was going to save the economy, this new president we have just elected with such high hopes and with the new congress all working together, these economists agree, we got the charts right here, decimal places and everything, and a huge sell with an immense propaganda campaign.

This suggests to me that there's a fourth case:

  • Case 4: The Stimulus had such an enormously negative effect on the economy that it swamped out the boost from the promise and expectation of an economic rebound and still managed to raise the unemployment rate roughly the amount that it was supposed to lower it. The means that effect of high expectations and the R/B "without" curve are both correct, and the R/B "with" curve is wildly incorrect.

Is that reasonable?

I guess I wouldn't have thought of this if the discussion hadn't gone from the Stimulus directly to the topic of expectations.

Eric Falkenstein writes:

great podcast Russ, very informative.

xian writes:

@econtalk:

i really enjoyed the discussion of the nobel economics. i think ive mentioned this before, perhaps episodes focusing on nobel or otherwise honored economic work is good fare- like a tutorial or lecture or whatever. this brief treatment was very cool...and value neutral.

hope to comment later, given time constraints...bravo!

Mort Dubois writes:

The scientific approach to economics ignores one critical factor, simply because it can't be measured. In medicine it's called the placebo effect. In economics, the term I've heard applied is "animal spirits", but might better be called confidence. I think, Russ, that you realize this because you keep tip-toeing around it, but can't quite bring yourself to say that what drives an economy is the faith of its participants, first and foremost. Short term policy interventions, like the stimulus package, cannot succeed or fail through their design. This is one of the reasons that it is impossible to isolate their effect. If everyone believed that stimulus was going to work, it would work. If not, then not. If some do and some don't, you get what we have experienced. The real multiplier is faith. Unfortunately, it will never be predictable or measurable.

DanW writes:

This was a particularly interesting and really engaging session. I'm excited to look at some of the supplemental materials you've posted. Thanks a lot to you and your guest!

D. F. Linton writes:

Another great podcast. Thanks, Russ.

BTW: This short Scientific American article Why economic models are always wrong, bears directly on the point Ms. Ramey made about using statistical techniques to fit models to known perfect data sets. The same issues probably arise in climate modeling as well. Ms. Ramey's final analogy in the progress of medical knowledge as hopeful for the future prospects for accurate econometric estimation ignores the vital difference that the economy is composed of individually willful actors with changing and unknowable motivations where as the biological processes of our bodies are comprised at the bottom by knowable chemical and electrical processes. What we can not now do and are unlikely ever to do is to bridge the methodological dualism gap.

J. Grover writes:

Thanks for a great interview with Valerie Ramey. There seemed to be an obvious question which you touched on but did not explore to the depth I would have liked. It was this: The 2009 stimulus package of roughly a trillion dollars was justified by the Obama administration’s economists using a Keynesian multiplier of 1.52. They therefore expected private sector GDP to grow by approximately 500 billion dollars. Overall growth in GDP would be about 1.5 trillion dollars. Assuming a growth in GDP of 200K creates one job, they expected about 7.5 million jobs to be “created or saved” by the package. If you further assume that the government will tax about 20% of that growth, or 300 billion dollars, the stimulus package cost “only” about 700 billion. You mentioned the work by Christina Romer, who was head of Obama’s council of economic advisors at the time, and her husband, David, looking at the effect of tax increases on GDP, who found that for every dollar taxes are increased GDP falls by 3.5 dollars. Turning that around, a dollar of tax cuts would increase GDP by 3.5 dollars. So, consider this alternative to the 2009 stimulus: Reduce taxes by one trillion dollars, GDP goes up by 3.5 trillion dollars, theoretically 17.5 million jobs are “created or saved”, tax revenues jump by 700 billion, making the net cost of the package “only” 300 billion dollars. As far as the country is concerned, the same trillion dollars would have had to be financed but the cost would have been far less and the benefits would have been far greater. A significant additional plus with this approach is that the government would not have grown to accommodate expending all that money.
So why did Christina Romer recommend the Keynesian approach when her own research clearly indicated a much better alternative existed?

[The spelling of Romer's name has been corrected.--Econlib Ed.]

emerich writes:

Excellent podcast. Ramey comes across as refreshingly modest and undoctrinaire. The one question I still had at the end is, Why don't we discuss short/medium/long-term in connection with the multiplier the way we do with demand and supply elasticities? Shouldn't we be talking about short-run vs. long-run multipliers? Isn't the real question, What's the present discounted value of the impact on GDP of $1 in government spending today?

Roger McKinney writes:

Very interesting interview! I recently read Estey’s old book (1950) “Business Cycles” in which he analyzed cycles from 1790 to 1949. Most of the cycles took place long before state intervention since they were pre-1929 and clearly the economy has a natural ability to recover from credit-induced depressions. Price changes do the main work of adjustment. Do any modern modelers take prices into account?

Also, wouldn’t it be reasonable to use pred-1929 recoveries as a baseline for determining whether stimuli work or not? For example, Estey estimates that economies pre-29 recovered on average after 22 months of depression and expansions lasted 26 months. He found that contractions caused output to fall 14 % from trend and expansions went 13% above trend. Stimuli would need to cause a shorter depression and growth in gdp greater than 13% above trend in order to be considered effective.

boileaup writes:

I appreciated the analytical findings in this podcast but I am disappointed in the clear bias of the interviewer. Although the podcast purports to present academic findings to explain economic theory in layman's terms, the assumption that a new listener might make is that the statements made during the podcast and, indeed the guests, would be unbiased because of their academic credentials. In fact, because of the bias of the interviewer, the layman-focused explanations of the findings are biased. Indeed, the guest selection is likely also biased towards the interviewer's world view.

In my view economics is an observational science rather than an emperical science and therefore interpretation of the results and the statement of causal links in the research can be highly influenced by the interpretation of economists. Therefore I believe that avoidance of personal bias is even more important in this academic sphere than in most other emperical sciences.

I believe that a program such as this should be even more careful to avoid bias given the importance that listeners might place on the content of the program because of its apparent academic nature and focus. If the program continues to present itself as an academic assessment of economics and not simply as pure entertainment, then I would recommend that a shift towards unbiased interviewing techniques and guest selection be put in place.

With kind regards

Pierre Boileau

xian writes:

criticisms/arguments that hinge on decimal places are often simple straw men.

EPZEN writes:

One of the things that I find frustrating is comparing periods that are not equivalent. Government spending will crowd out private investment when the economy is at full employment. However, we are currently not at full employment and monetary policy has pretty much run out of tools or else what tools they do have they are not willing to make much use of (pushing long term interest rates even lower). We are in a liquidity trap which we were not in during the Korean War.

To see the effects of government spending in a liquidity trap one only has to look at the current effects of austerity programs in Britain, Ireland, Greece etc. All of these countries which have practiced austerity have seen their unemployment rates increase.

In regards to the U.S. stimulus it is of no surprise that it had no effect. The stimulus was matched by an equal decrease in state and municipal spending and therefore what a surprise that unemployment stayed at the same level.

As far as the WWII era boom the consumer price index doubled between 1940 and 1949 which in essence cut the country's debt (private and public) in half which is what was necessary to get out of the liquidity trap. When monetary policy cannot be loosened the only way to get the economy moving is through government spending. Yes government has to borrow the money but part of the problem is that companies and wealthy individuals are hanging on to their money so one way to move that money into the economy is for the government to borrow it from the people that are holding it and then to spend it (either directly or through tax breaks for lower to middle income people.....they will spend the tax break, the wealthy will not....and NO this is not class warfare).

I would be a classical example of what is happening in the economy. I would be considered wealthy (i.e. top 1%) and have been reluctant to move my money out of a cash position. In fact, I became so dismayed watching the behaviour of the American Congress in the debt ceiling debate and of the European Central Bank/Governments that I am almost entirely in a cash position. This from an investor who had traditionally followed a relatively aggressive investment philosophy. I am in essence, a micro reflection of the liquidity trap.

One final point that I would make is that many economists compare time periods that are not equivalent. We are currently in the 1930's, not the 50's, not the 70's nor the stagflation of the 80's. Unfortunately, this makes the science of economics difficult when you in essence have only two data points.

J Trott writes:

I found Ramey's explanation of the tax affect on labor and GDP to be interesting but ultimately unconvincing. The idea that the correlation between lifetime work hours and tax rates means that tax rates determine work hours may be part of the affect but I think there is at least one other. I think that it is probable that people will do different work if tax rates change or they expect them to change even if they work the exact same number of hours. The intuitive story is that if one has a low risk option of government employment or private employment versus a high risk option of entrepreneurship, the decision between those two is influenced by one's expected reward. In a high tax world, the upside for entrepreneurship is lower while the relative safety of employment stays constant. Someone could do some econometrics on this idea.

Comments for this podcast episode have been closed
Return to top