Russ Roberts

Taylor on Fiscal and Monetary Policy

EconTalk Episode with John Taylor
Hosted by Russ Roberts
PRINT
Banerjee on Poverty and Poor E... Hennessey on the Debt Ceiling ...

John Taylor of Stanford University talks with EconTalk host Russ Roberts about the state of the economy and the prospects for recovery. Taylor argues that the design of the fiscal stimulus was ineffective and monetary policy, so-called quantitative easing, has also failed to improve matters. He argues for a return to fiscal, monetary, and regulatory normalcy as the best hope for economic improvement. The conversation concludes with a discussion of the impact of the current crisis on economics education.

Size: 27.4 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: July 12, 2011.] Start off by talking about what's going on in the economy right now. We're doing this in July of 2011; the recent jobs report for June of 2011, with the revisions for April and May were very disappointment. unemployment has been over 9% for roughly two years. There were a couple of months it dipped under 9%. Things don't look very good. What's your assessment? I think has really been a very disappointing recovery from the beginning. It's now just about two years old. Compared to the recovery from the most recent deep recession in 1981-1982, it's incredibly weak, almost non-existent. For example, coming out of the 1981-82 recession, that recovery's first two years had 7% growth, and so far this is 2.8%; and it looks like the second quarter, when we get the number, will be even below 2%. We don't know yet, but it's very disappointing. I think to analyze this you have to think about the whole picture. I think that's why I keep coming back to the policy. What's different now is the policy. We are doing many different things than in the last recovery, and I tend to focus on that. There's differences of opinion. Some people talk about the balance sheet out of line. Some people talk about the 0-interest bound. But I think that basically there is an erratic set of policies across the board. Let's talk about the balance sheet for a minute, because the standard defense of the policies has been: "It's worse than we thought." That's to me not a very useful analysis. But it is possible there were things holding the economy back that we didn't fully appreciate. One I worry about and don't really know how to think about, and I'd like your thoughts, is the housing mess that we are still in. We have a large number of houses that are either under water and potentially going to have to go into foreclosure, or simply in limbo. The foreclosure process isn't taking its normal path or is taking a long time; that is potentially deterring both the financial sector and the household sector from acting in a positive and risk-taking way. What do we know about that? I think what we've learned is a lot of the interventions to try to fix this, working for the foreclosures, first-time home buyers, have really not done the trick. The lesson to me is you really have to let market forces move, let the prices adjust, and then we'll have recovery of the housing market. It is a difficult problem, especially for lots of people, and there's the strategic default issue, which is very hard to deal with. What do you mean by that? And in addition, there's the problems of the banks and what their balance sheets are like. A strategic default is the idea where people see their mortgage is more than the value of the house and they say, well, I'm not paying; I'm getting out; I'm moving. That can happen. I think though that if you think about this slow recovery, that cannot possibly be the whole answer. While housing is a part of the economy, remember it's not the only part, and business investment in general is low. Business fixed investment as a share of GDP is quite low. So, we basically have lots of opportunities for growth as we come out of all recessions, no matter what they are caused by. And housing, by the way, has been a big factor in recessions in the past. So, I don't think that's the full answer. I think it's a drag on the economy, but there is always some kind of drag on the economy.
5:01Let's look at the two major areas where we've attempted to get things started getting better. Fiscal and monetary policy, the two traditional tools that macroeconomics counsels. You've recently done very interesting analysis with John Cogan on the stimulus package, so tell us what you found and what we learned. We think it's very important to estimate the effect of a stimulus package by looking at what actually happens. Not just simulate another model again, because the models differ. To simulate one model that said the package would work in advance, that same model will say it worked after the fact. A model that wouldn't say it worked in advance will say it didn't work after the fact. So you've got to go beyond the models, and that's what John Cogan and I have tried to do. We've looked very carefully at the data. We've looked at the money sent to the households--that seemed to have been saved rather than jump-starting consumption. That was about a third of the $870 billion stimulus? Yes, about a third. We looked at the money sent to the states, the grants that were meant to jump-start infrastructure spending, other government purchases. It didn't. Infrastructure spending didn't increase. Purchases by the states didn't increase. What we've been able to determine by following the money is the states reduced the amount of borrowing they have. They basically just saved, just like the households. So, when you look carefully, it's basically what you'd expect from basic economic theory. In fact, it's what we learned from studying the stimulus packages in the late 1970s, where grants were sent to the states. So I think the data are quite convincing; and as people look at it more and more they will realize that this package didn't really do what it was supposed to do, and the way it was supposed to do it. Now there was some Federal spending directly, correct, in the $800+ billion? Yes. When I see the road signs, some work and other projects were done, presumably with some of the stimulus money. Well, at the Federal level, Federal purchases of goods and services, it's almost an immaterial amount of the total. It's like less than 0.1% of GDP. So at the Federal level there wasn't really this. And by the way, when you see the signs on the roads--those could quite well be things that would happen anyway. That's the notion of shovel-ready. And all you are thinking about is they may have been financed in a different way. Rather than bond issuance, the municipality, local government, or even the state may have simply just used the grant money for that. No difference in the amount spent. So, you see--let me try to summarize what I've learned from your paper and see if it's the right lesson. It's an interesting critique of the Keynesian spending model. It basically says, ironically--in the way we carried it out in this particular time--is something like pushing on a string: that the planned expansion of economic activity literally didn't take place because it was either saved by households or offset by changes in spending by the states. Is that correct? Yes. And that is really not a surprise. This is really one of the major critiques of the Keynesian stimulus package from the beginning: you can't really change on a dime what government is doing. That's why, in fact, this program was designed not to have so much Federal spending but to have more purchases of the states. People realized that. Ironically, they didn't realize that if you just send money to the states, a bunch of money, they are not instantly going to go change their spending patterns. In fact, I think our data show they did it in a perverse way. They actually took some of this money and increased their own transfer payments. Say, for example, Medicaid; and reduced their infrastructure purchasing. They basically--the way that the American Recovery and Reinvestment Act of 2009 (ARRA), the stimulus package, was stated in the law, encouraged the states to actually reduce their purchases and increase their transfer payments, which again is counterproductive. So, that's one side. One side of the set of tools. The proponents of stimulus have argued it simply wasn't big enough--that's one of their arguments. And they will concede, I think, that it was poorly planned, that it wasn't targeted. Did they accept? Some of the proponents have accepted that the states simply borrowed less and that there was no real net stimulus. Can you summarize what kind of reaction you've gotten from the paper. Well, the initial reaction was quite negative. Hostile. But I think as people look at it, it's kind of hard to refute. As people look at the numbers, and there still will be attempts to do that; but I think I feel quite pleased that this message seems be getting out, as you say; and people seem to agree. I don't think there's evidence for the view that it just should have been bigger. That's certainly not how I read this. In fact, with these data we build a little model, and simulate the counterfactual of a larger stimulus. In other words, designed, in the same way--which I think is a practical fact of life--a much bigger, and the effects are the same, because people just save it; the states just put it in their coffers; and you don't get more stimulus. So I think it's a misinterpretation of what happened to say it just should have been bigger. It was large already. We increased our debt. It wouldn't have been more stimulative. Had it been larger, it might have been more counterproductive.
11:01Let's talk about the monetary side. So, we've just finished this slightly bizarre episode called Quantitative Easing 2 (QE2). If you go back to some EconTalk episodes we did--I remember the one with Allan Meltzer, where he basically said, and I think you'd agree, that monetary policy isn't hampered by the zero-interest rate lower bound because the Fed can always inject money directly into the economy by buying assets of various kinds. Essentially printing money and putting money into the hands of either investors or consumers, whoever gets that money. Two things struck me about this episode of QE2. First of all, it was quite large--$600 billion. Approximately the magnitude of the stimulus package; and it came on the heels of some very aggressive monetary policy before that. It's strikingly ineffective. Why? And how do you reconcile that ineffectiveness with the confidence with which someone like Ben Bernanke, but even Allan Meltzer and others who are proponents of the efficacy of monetary policy? It seems to have done nothing. Is that a correct analysis or do we not really know yet? I've looked at the numbers pretty carefully, more of QE1, which was similar purchases of smaller amounts of Treasuries, but much smaller in the way of mortgage-backed securities. $1.25 trillion of mortgage-backed securities were purchased and $300 billion of Treasuries. I looked carefully at the mortgage-backed securities' purchases and can find very small if any effect and they are very uncertain. I think basically both QE1 and QE2 were disappointing from that perspective. Now, with respect to the premise here that monetary policy is still effective--I think it is, but I would put it this way: When the interest rates hits the zero bound, then you should focus more on keeping money growth from falling. And that's what the kind of strategies that Milton Friedman would have advocated. He found, and others found in the Great Depression, that they let money growth decline. The right strategy would have been if the interest rate hits zero is just to make money growth didn't decline rather than have these gigantic injections of what we call high-powered money or bank reserves into the economy hoping to drive mortgage rates down, hoping to drive medium Treasury rates down, because we have a lot of theories out there that say those won't affect longer term rates by a very large amount. So, help me out here. The actual activities that the Fed did, which, getting the numbers right--$1.4 trillion was Quantitative Easing 1 (QE1), which was mostly mortgage-backed securities. Those were assets that were either on the books of Fannie and Freddie, so-called government-sponsored enterprises, or investment banks. Correct? Yes. Mortgage-backed securities backed by Fannie and Freddie. So, that was most of the amount? $1.25. That pushed money where? Onto the balance sheets of whom? So, the way the Fed has to have money to buy all these securities--they get the money by just crediting the banks with deposits at the Fed. That's the notion that you are creating money. You are creating reserves at the Fed; that's a measure of creating money. And that's what you called high-powered money. Yes. Lots of names for it--monetary base, central bank money--it simply means the deposits the banks hold at the central bank, the Fed. Currency is part of it, too, but the main thing is the change in deposits held by banks at the Fed. So, the Fed gets its money by simply taking those deposits, putting it on the books of its banks. So, the idea would be then, as a mechanism for economic activity, that the banks would go out and use those reserves. They wouldn't want to keep money just sitting around on their balance sheet with the Fed--they are going to go do something with it. Is the case that they haven't? They haven't. We saw that right away. When the balance sheet of the Fed--the reserves that the banks hold at the Fed--when they exploded in the panic period of late 2008 and then continued with these quantitative easings, we saw the banks just took it on. And just basically the Fed purchased more securities; banks would just hold these. They didn't go out and lend them separately. So, they are paying some interest on these reserves. It's a very small amount, less than the banks could get by doing other things. So, the banks didn't lend it out. Of course, if they'd lent it out you would have had an explosion of the money supply very quickly. By not lending them out, the Fed has actually been given another discretionary tool to buy all these securities. So, I remember in the early days of the crisis, Hank Paulson was Secretary of the Treasury at the time, who had initiated the Toxic Asset Relief Program (TARP)--Reserve Program? Can't remember now; so long ago; back in the early aughts. The government had these banks whose balance sheets they worried about and they injected--it was separate now, not the Fed? Or is it part of the Fed? The TARP was not the Fed. They also put a lot of resources into these distressed banks, and they got very upset that they just sat on them. They basically yelled at them. I don't know whether this was theater or something else, but I found it very strange that the Secretary of the Treasury was browbeating the banks: We only gave you this. It's like giving your teenager and you expect him to go buy organic vegetables and instead he goes and buys beer. I gave it to you for organic vegetables. Why, in the face of that behavior, which was clearly a response either to unease or a lack of opportunity, why would we do it again? What do you think Bernanke had in mind with QE2 when he had these banks already well about their minimum of reserves? What was he hoping to accomplish? What do you think his public and private story was there? He was very clear. He wanted to reduce longer-term interest rates by buying longer-term securities, which would drive their price up and therefore their yield down. That was clearly the purpose. Similarly the purpose with purchasing the mortgages, to reduce mortgage interest rates. The connection with your question about the banks holding the reserves--effectively the fact that the banks took on all these reserves allowed the Fed to do all these purchases, and tried to use those funds to drive interest rates down. And what didn't work, at least as I can tell, was that it didn't drive those rates down, and those markets are gigantic. And they depend on expectations of the future. Of course you can, by big injection, move them temporarily. But by and large the Fed didn't move those very much. And I think that's what we would expect from basic theory.
18:59So, one story would be they were disappointed with the outcome of the policy. A more cynical story would be that the Fed has political allies who are the financial community; they had bought a bunch of bad stuff; and this was a way to get it off their books and make them healthy without making it look like welfare. Is it correct to characterize the Federal Reserve in the last few years as being a sugar daddy to the financial sector, where the risks of those assets are going to be borne by the taxpayer rather than those original institutions? Well, there's a couple of things there. I think piecing the panic with the rescue of Bear Stearns's creditors and AIG's creditors is obviously a situation where the Fed took on risk. Even though it's intentions were placed otherwise as far as we know, did it help directly those creditors? Absolutely. With respect to these quantitative easings, which, I'll refer mainly to these large-scale asset purchases of Treasuries and mortgage-backed securities, Treasuries are highly safe securities. So, you are not doing anybody a favor--those aren't toxic assets. So, the motivation of the Fed there was not to take the toxic assets off of the banks' balance sheets. It couldn't be. But to basically reduce those rates. Now to follow up on your other question about helping the banks, the zero-interest rate or near-zero on Federal Funds obviously is a way--it increases banks' profits, because they can lend at higher rates than that. So, the margin between the Fed's rate, which they are able to control now, and the lending rates of the banks is a profit margin, and so that does help the banks. To what extent that's motivating the Fed, I don't know. At least an indirect effect. What do we know about the volume of activity? We talk a lot about prices, which are the interest rate. We say that banks have all these reserves on their books. It reminds me a little bit of this claim that consumers aren't spending. Well, they are spending--they are spending an enormous amount. They are not spending as much as they did before, but if you go into a shopping mall, people are spending their money. They are not saving 80%. They might be saving 5% or 8%, close to zero, and some say it might be negative. So when we say that the banks are "sitting on the money"--do we have any evidence that they are really literally sitting on the money? That it's hard to get a loan. If I want to start a small venture, can I get a loan? Or is simply I don't want to ask for it because I'm worried about the future, if the volume is low? Do we know anything about those volumes? The volumes are gigantic. We can see them. We get the data from what the banks are holding at the Fed--$1.5 trillion. The holdings are gigantic. But are they doing anything with it? You're question is really why they are holding it. No, I'm asking a different question. They are doing something. The banks aren't just sitting around twiddling their thumbs and saying I hope it gets better some day. They are lending some of that money, right? Absolutely. There are bank loans; bank loans are out there. But the amount of money they have to lend from these reserves is much larger; they are holding a big excess. That's there in the data. Why are they holding large excess reserves? Are they purposely holding back because they don't want it? Well, no, they want to make money like everyone else; and it seems to me the demand in the economy is low compared to that gigantic amount of reserves. Clearly they are cautious; some may be more cautious than others. Either supply or demand--and I think it's mainly demand--is low. But is actual loan activity a fraction of what it was during good times? In other words, I understand they are not lending out the whole thing. But are they lending out a sufficient chunk? Well, they are lending out an amount that corresponds to the weak economy. The economy is weaker; business investment is weaker. Another way to think about this demand side is the corporations are sitting on a lot of cash. Their own balance sheets have a lot of short-term liquid assets that they could invest and build things with, so that's another reason they don't even need all this bank lending. They are holding back for other reasons. Presumably unease about the future. Uncertainty. The things we keep mentioning--uncertainty about the regulations of health care and the financial services area, are taxes going to increase, what are we going to do about the budget deficit. When we talk about this to businesses, they focus on this uncertainty a lot. You can't prove it. Speculation, obviously.
24:15One view of where we are right now is that the traditional tools of fiscal and monetary policy either have failed or been implemented poorly. What do we do now? What would you do now if you could do what you wanted? When you say the traditional tools have failed or are implemented poorly, I would say what we learned is lessons from a couple of decades of pretty good times, the 1980s and 1990s--a follow a steady-as-you-do policy. With respect to the Fed, don't try to reduce rates too much. With respect to fiscal policy, don't try these short-term stimuluses. We knew they didn't work in the past; they caused, probably, they caused uncertainty and made things worse. And now, in the Fed's case, this monetary overhang, which we've been discussing--this big increase in reserves; and of course we have this gigantic debt at the fiscal level. So, I would say, the thing to do that would be most powerful for the recovery is simply to get back to sound fiscal policy, sound monetary policy. And I think that would be positive for the economy. It's not like we don't have any tools. We do have a tool: Get back to sound fiscal and monetary policy. What would that involve? On the fiscal side it would mean a good plan to reduce the deficit over time from the very high levels it is now. I think the way to do that, most constructive for economic growth, is to reduce spending, not to raise tax rates. There are some plans out there. The political system can't get there quite well, but there's progress. I think in the last year things looked quite dismal that we were even going to address this; now, with election in November, President Obama's first budget was basically withdrawn and he's got new proposals to reduce spending. So, we're making progress on the fiscal side. On the monetary side, I think there's also a realization that these interventions have not worked very well; and that's what you're saying here. They seem to be pulling back from that. So, in their case, they should lay out an exit strategy that shows how they are going to get that balance sheet back down in a reasonably predictable way that doesn't cause too much uncertainty. And then, when the time comes, start to raise rates a little bit. And eventually try to prevent the uncertainty about inflation which is out there, and if you like, another contraction that people worry about. So, today is July 12; we are 21 days--3 weeks--away from a supposed crisis August 2nd, when if nothing is done it's alleged the Federal Government will be unable to meet its obligations without borrowing. The debt limit will have been reached; and we won't be legally allowed to borrow; and we face the risk of default, at least in some dimension. How do you read that analysis and what do you think ought to be done? I think they ought to make an agreement to reduce spending, at least by a good chunk, two and a half trillion is what I talk about over 10 years; and simultaneously increase the debt limit by that amount. That establishes the principle that you are not going to just increase the debt without dealing with the spending problem; I think that's a principle we should continue with and I'm hoping that's where they get to. I think it's possible to do that right now. One thing I always find troubling is this 10-year thing. So, $2.5 trillion sounds like a lot of money; over 10 years, it's not a lot of money. It's $250 billion, in a budget of $3.6 trillion, it's a cut, but not a very dramatic cut. It's nothing close to what would put us back to the level of Federal activity of, say, even 5 years ago, 4 years ago, before this enormous expansion. It's strange what the baseline is. Why do you pick such a small amount? You are quite right to question the trillions and the 10-year, so I like to look at the actual path. Look at graphs; try to have graphs that show what these trillions mean. For example, if you took $6 trillion off the current projections, that really would bring Federal outlays as a share of GDP back to where they were before the mess began, around 19%. As a percentage of GDP. I think that's a good goal; can balance the budget without raising taxes. So, $6 trillion, if you go there at a reasonable pace, will do that. I think it's very important to think about it in those terms. I'm talking about $2.5 trillion, by the way, so that's not all the way to $6 trillion. I would say if we could get $2.5 trillion now and the remaining $3.5 trillion, I think that could be part of the discussion over the next year. We have a Presidential election, current President says he wants to raise taxes; his opponents on the Republican side say they don't; have a debate about that $3.5 trillion. Get $2.5 out of the way; that's a good game changer, big difference from what it was a year ago. I think that would be a sensible way to proceed. One of the things I find so dysfunctional, though, about the political system is how different it is from what we might call sensible family budgeting. If my family, if I lost a source of income, found myself indulging a taste for fast cars--which fortunately I don't have--maybe to please my teenage sons I bought a car I couldn't afford; and I realize I'm living beyond my means. Eventually somebody's going to say: I don't think I'm going to keep financing that. I'd look around and I'd say: Let's look at the stuff that isn't so important. Another strategy would be: Let's cut 15% across the board--eat out 15% less, spend 15% less on vacation, on clothing. Obviously, there are some things it's hard to spend 15% less on. The government--I have no idea what the process is, other than it's not what economists would like it to be. We'd say things like: Let's look at the stuff the private sector can maybe do as effectively; let's privatize some things that aren't really being done well. Do you have any idea what they are actually doing? I don't know what they are doing, when they talk about these cuts. The more worrisome part is: Do they know what they are doing? That's a serious question. This is a very complex budget; there's a lot of devil in the details. You go down and you look at what they a given agency's doing. One way to think about this and make some common sense judgments is: Think about share of GDP, the whole thing. Now over 24% of GDP. In 2007, it was 19.5% of GDP. Why can't we get spending back? What's the big harm, what's the pain involved with getting government agencies to spend at the levels they were spending at in 2007. And even with inflation--there's still growth of spending. I think the bottom line people should be asking each agency, each account: Why can't you get back to where you were in 2007? Wouldn't that be sensible? We could look at your budget and say what were we doing 3-4 years ago? When we were okay. We'll get back there.
32:35So, one answer would be: Well, we're bombing Libya now; and that's an extra commitment we made. We might decide we can't afford it. But that would be one example of something we'd know we didn't do in 2007. What I find striking about the political debate is I have not seen a careful analysis of which agencies are spending more relative to 2007. We understand that when unemployment goes up, unemployment insurance is higher, food stamps are higher. But those are relatively small amounts of money. Where is the other 5 percentage points of GDP, which is huge? It's almost a trillion bucks. There's a lot of incentive in agencies and the people getting these funds not to go through that calculation because they like the fact that they got extra spending in the last 2-3 years. They think it's good. Anyone would defend what they have. So I think there's a real incentive not to say: Well, we can get along with what we had 3 years ago. People like the extra. Places like this university where I teach, Stanford, receives Federal funding. And they got more Federal funding in the last few years. A lot of universities did. And so they are going to be defending that, as anyone would. You'd think they could just look. It's a good economic activity. If somebody out there is listening. One of the things you see as so important--people quickly tune out about the numbers. And charts, even. Their eyes glaze over. So, I think it's a real challenge to get people to focus more on these numbers, these charts. Simple stories like you emphasized here--the facts would be so revealing. It doesn't help resolve all the differences of the opinion, but I think we could go so much further just by explaining here's what this agency was doing, what that agency was doing. To take an example from the Fed, by the way, one of my pet peeves: You mention the bailout of the creditors of Bear Stearns, which I think is the right way to talk about it. It wasn't a bailout of Bear Stearns; it was a bailout of their creditors; as listeners know I care deeply about that distinction. To get JPMorgan Chase to take on the obligations of those creditors, the government acquired--guaranteed, or both--about $30 billion of so-called toxic assets of Bear Stearns. I remember that weekend very vividly, March 2008 weekend; as has been the case in every one of these expansions of government discretion, we were assured that: This may turn out to be a money-maker, because when the Fed acquires these, we don't know what they are worth now, but the market is thin; these aren't so liquid; when things get better. I find it remarkable that I--speaking for myself now, a relatively informed person in this area, in the top 1%; not the most informed person--I have no idea whether that turned out to be a good idea or not. I have no idea what the state of those assets is. I suspect it would be extremely difficult to discover what has happened to those. They got rolled into lots of stuff. The lack of transparency--certainly of the Fed, which is famous for it--but even in this case where you point out, the state of the budget is so complex. A lot of the time we are talking about future activities where the rules aren't clear, what the baseline is. It would be a very useful thing just to get a fact or two about activity. It would be a good project for somebody out there listening. I'd encourage it. The thing about the Fed and the banks, which I guess goes without saying, is the banks are the most heavily regulated financial institutions we have. No they're not! We have deregulation in America! Banks do whatever they want. Didn't you know that? That's why we're in the mess we are in: free market philosophy. Sarcasm. What this all points out is this concept of regulatory capture: you have the banks being regulated by groups that seem to be winking about certain things they are doing, so it's not the fact that there isn't sufficient regulation. It's that they didn't seem to be doing their job in doing the regulations. We need to find out more about that; investigative reporting is needed about that.
37:45Do you want to say anything else about the debt ceiling? You suggested what you'd like to see happen--in the short run a serious attempt to slow the growth of spending or maybe take it down a little bit, followed by some additional activity if the political will was there. That may not happen. Are we at risk of heading in the direction of Greece? Again, it's July 12--Italy was in the headlines this morning. They face a different problem that we may face in the longer run; but that's not our short-run problem, correct? Well, the short run turns into the long run really fast in this business. I think the lessons from Greece and Italy are very relevant to us. First of all, they are part of the same story that what is causing the mess here is some government policies--spending more than you have, deficits; or this monetary policy. It's just another example: how could you blame the markets on what's happening in Greece or Italy? That's basically a policy question. I think the lesson for us is very clear and may have been one of the reasons we see this shift in the United States that people don't want to go down that road. I think the lesson is there. The way the financial markets work is people can turn against you pretty quickly. People get a perception that we haven't dealt with this problem or we don't have the political will and it could be very damaging. So, if that's the case, if there is a serious risk here--and I ask this question because I am personally, historically, a deficit dove. I've often said deficits aren't that important. I learned that from Milton Friedman; it's spending that matters, it's not how you finance it. A deficit is simply a mix of taxes today and tomorrow rather than just taxes today. The real drag on the economy comes from taking real resources out of the private sector and putting it in the public sector. To the extent it's used well, it's okay; but if it's not, it's a harmful thing. Yet, when I see what's going on in Greece, I get a little apprehensive that maybe we're on the verge of not just borrowing a little more than we did before, but potentially getting into a situation where, as you've suggested, where suddenly people say: We don't see that as a safe asset any more. And yet, we see nothing in the market place that suggests that's the case. Treasury rates remain remarkably low, suggesting that world investors--meaning both American and foreign--don't seem to think there's much of a risk of a default. How do you read those numbers? We have a weak economy; that's why it's come down recently. I think there's a sense in which people feel we will figure this out, long term; we will get our act together. America has a history of getting its act together eventually, and they are making a bet on that. So, that's why I say things can move against you quickly. If you look at emerging market countries, when they got into trouble, there may be a rescue. Maybe Argentina gets a rescue from the International Monetary Fund (IMF) and interest rates come way down. Confidence is back. But everyone knows that 2-3 years later there's going to be a problem again, and they go ratcheting back up again. And they move against you quickly. I also think that's part of the answer to your question about deficits mattering. At some point the debt gets too high, and especially when a substantial fraction is held abroad, as ours--half of our debt is held abroad. That increases the chance that things can move against you rapidly. I don't think that's a reason to be complacent. So often in these cases things look good and the rates are low; and we saw that actually going into the crisis recently. You had risk premiums seemed very low; people didn't seem to be paying attention to the unsustainability of the housing boom; and it goes the other direction quickly. I recently heard Sam Peltzman say: Things go well until they don't any more. That turning point--events often mask what's going on.
42:26You mentioned the IMF. The head of the IMF recently got into a contretemps--French word meaning against time, literally; a dustup, disturbance--he lost his job; tabloid set of events. A replacement is on its way, Christine LaGarde. Does the IMF play any useful role in the economy? Is it a net positive? What if they just said: This is awkward, bad publicity; let's just shut it down. What would be different? It was suggested around the period of the Russian financial crisis that they said we could do that. The late 1990s. I went to work in Washington with the responsibility of overseeing the IMF and the Treasury, and what seemed to me based on that experience is the most important, practical thing you can do is to really put some rules on the behavior of the IMF so it doesn't willy-nilly bail out countries. We should say countries' creditors, because that's what they bail out. There were some reforms at the IMF in 2003 with respect to emerging markets, which really reduced the amount of bailing out of emerging markets. And lo and behold, emerging markets started to behave a lot better. That constraint, I think of it as really rules rather than discretion again made the system a lot better. Unfortunately, the same thing has now emerged in Europe, so they are kind of back in this bailout mode; and we need to figure a way to get out of that again. I think that's what we are going to learn from this crisis, and hopefully that will be the end and the IMF will be confined to communicate about what your policies are, a way to emphasize the importance of free movement of capital, which they had been doing for a while--they've gotten off that track, too. Of course, the political forces are very similar there as in the United States; they just have different titles and different foreign names. As far as I understand it, the rescue of Greece--you mention rescuing the creditors--is mainly a rescue of French banks. And so the political influence of those banks is going to often override good economic policy. Absolutely. That's the same kind of thing that would go on with the IMF in emerging markets. You'd have financial institutions or banks or whatever holding debt of these countries and be very concerned about default of these countries. So they would basically be quite favorable to the interventions. I saw that myself in practice for a number of countries. To say no in those cases--you are going to get a barrage of complaints and irritation from the financial sector. But, sometimes that's the right thing to do. This whole issue of rules versus discretion, which I know you are interested in--the political barrier to that, a move to less discretion and more rules--it stifles the ability to take care of supplicants, a nice word for special interests or cronies, if you want to get negative about it. I'm increasingly pessimistic about our ability to fix that, one reason being there is a shockingly large and shockingly small amount of outrage about what's going on now, from both the left and the right. How can that be at the same time? There's more than I ever noticed in the following sense: more people are disturbed today than I've ever seen before about the discretionary activities of the Treasury and the Fed over the last three years. Certainly there has never been a time in my lifetime when there was more cynicism and dislike of the Federal Reserve, for example. And yet, it's relatively small. Mainstream economists for the most part assume we'll just have to get better strategies, we'll just tweak the regulations. So, while there are more people than ever advocating a gold standard--which is shocking compared to ten years ago, it's still a relatively small number and there are no other reasonable reforms on the political docket that I can see that are legit that are going to gather more support. What are your thoughts on the prospect for any kind of serious change in any of these areas? Well, in the Fed I think there are some prospects. It's not a gold standard proposal, but there is for example this dual mandate that the Fed has, where they are supposed to both look at what's happened to inflation and maximize employment--that creates the excuse at least for lots of discretion. It changes their role. For the good period of Fed performance in the 1980s and 1990s, until relatively recently, both Volker and Greenspan deemphasized that tool and focused on inflation. So, I think the first step is to fix that dual mandate. And the second thing, in 2000, the reporting requirements for the Federal Reserve about the money supply, money growth, were removed and I think we should put them back. What I think we should do is put them back in in a different form; I think I have ways to suggest that that be done. Things worked pretty well. It wasn't ideal, but it was better than we'd seen in history at that point. I'd like to see it possible to go back to that in a modern world we live in and achieve many of the goals that people advocating a gold standard would like to achieve. Do you know why those reporting requirements were changed? What was the public statement? They were removed in what was actually a part of a housing bill in 2000 because the monetary aggregates became hard to measure or interpret. They became less reliable. So, what was happening was the Fed, required to report about them, would constantly be saying why it can't forecast it or project it very well and why it deviated from it. The truth is the requirement did become less meaningful because it was being violated all the time and the Fed was spending all its time explaining that. So, it was removed and no one complained too much at the time. But I think we now see afterwards that this was a period when it moved in much more discretionary ways and I think to have preserved, maybe not those requirements, but a modified version of those requirements--I tend to think they could do that in terms of their strategy for setting the interest rates, for example. I think that would be a good reform. And of course on the fiscal side you have impulses for reforms, too. House budget amendment to simply budget rules as a share of GDP. I think on the fiscal side you just have to change the spending, get the spending down; that would be a big improvement. I think there are things to do, not a perfect world but we could improve things a lot.
50:43Macroeconomics as a field: You've recently been characterized as an anti-Keynesian. Would you say that's an accurate description and if so, is there a more pro- rather than just anti- description for yourself? I'm very critical and have been critical for a long time about the use of discretionary policy to try to mitigate the business cycle; and instead have stressed the importance of getting the overall stance of fiscal policy right; to have a tax system which is efficient, to rely on automatic stabilizers. That's a far better solution, so I think in that sense I would be anti-Keynesian. Based on experience, based on the models we build, you can't do this fine tuning. I think some people think anti-Keynesian means I believe there are no rigidities in the economy. That's not true; I do think there are rigidities in the economy. That's not a measure of it. I think the main thing here is: is this a practical prescription for improving the economy; and I think it's not. You and I talked before this recording as some people characterize your anti-Keynesian stance is that you believe all government spending is useless. That's not correct, is it? No, absolutely not. There is an important role for government, there are public goods, externalities. We have a long list of cost-benefit types of things to go through and there is also a role for government in maintaining the rule of law and enforcing that. So, I think it's very important to not be characterized as anti-government spending by the word anti-Keynesian. So you are mainly referring to the discretionary fine tuning? Yes, and what goes along with that, of course, is a tendency to expand spending because it kind of becomes an excuse. I think we've seen that recently; we had a big increase in spending, basically taking our eyes off the ball because we--our government--thought that this would be stimulative to the economy. So, that's another sense in which that can lead you astray. It certainly--Keynes famously said that if you pay people to dig ditches and fill them back in, you'll still have a stimulative effect. I heard Joe Stiglitz in front of Congress agree with that. He said it's not the best thing to do; it would be better to do something more productive with the money. But, I do agree with you that when you get to a point where you believe that's true, you certainly open the floodgates to a political pressure to spend money on things that are beneficial to certain groups. That are worthless. But the argument there is it's a free lunch because it increases aggregate demand. It'll be interesting to see. On the free lunch, at best you could have a really short term little blip or something, and then that's going to dissipate and then you have the costs in higher taxes that you've got to pay for the debt. One of the things that people have to think about when they take their position on Keynesian or anti-Keynesian is a quantitative thing. We're talking about numbers. So many of these things at best are like short-term little gooses to the economy. And then they leave you worse off very quickly than you were before. I don't think that is a positive. It's a negative. It's like the Cash for Clunkers--people did spend more for cars in that period. They moved their purchases forward by a few months. It did no good to the recovery. And so much of this is of that nature. I think the more people understand that--but it is a matter of understanding the timing, the numbers, the size, what's temporary and what's permanent. Kind of hard.
55:19Do you think this episode, which of course isn't over--incredible time to be alive as an economist because there is still a lot of uncertainty about how it's going to turn out; we may be facing a double dip--do you think that there's going to be much change in textbook economics? How we think about these things? We've talked about this before. Thought you said--looked really carefully at past stimulus programs, fiscal policy expansions. Pretty much unanimous conclusion that they are not very effective. In the midst of a panic and political pressure, that knowledge, if indeed it was right, was ignored and now we have a lot of smart people arguing it could have worked, it should have worked, it did work, even. People like you and I are saying it failed and we should continue learning the lessons we've already learned; but people have forgotten them. I think it's going to play out in the textbooks. Those who write textbooks who think it worked will write that and those who write textbooks who think it didn't will emphasize the latter. Since textbooks have to appeal to everybody, there will be a tendency to give the other side. But to me, the textbooks should reflect these things didn't work in practice. They should contain them in there. The textbooks that have these Keynesian crosses in there, you've got to have huge warning signs under them that this is really not how we've found this to work in practice. That's what I'd like to do. I gave a talk to the American Economic Association a few weeks ago where I kind of outlined my thoughts on how it should affect the teaching, and I also emphasized that we now have just a host of interesting examples to teach students. Say, like this Cash for Clunkers, beautiful example--they did move their purchases forward, incentives were changed; you also have information about what happened to the money when people were give it--they held it. That's just what the permanent income model shows you. Beautiful illustration. In terms of teaching economics it's a huge opportunity to show that our models work pretty well. In those examples, what I like about them--and long time listeners will know my feelings about empirical work and data analysis--those are simple. There's nothing complicated. You are looking at the data. There is always some question about interpretation, measurement; it's not perfect. But those stories you just mentioned are stories a freshman economics class could easily absorb. And since I teach Principles, I've been using them already. Very convincing. I would say the students coming into Economics 1 or Economics 101 definitely have a skepticism of some of these Keynesian things. Now there's more reason to be skeptical. They can see what happened, simple illustrations. I think that will affect things. Whether 30 years from now there's another big crisis and we do it again, I hope not. I get discouraged some times. I hear you.

COMMENTS (30 to date)
Anne writes:

I have been trying to figure out a question related to the comments in this podcast re which federal agencies have increased spending vs. which have not. Historically, “black budget” spending has not been included in the published federal budget. The black budget covers the cost of the intelligence services (i.e., CIA, NSA and perhaps DIA) and is approved through various Congressional intelligence committees. In the little research I’ve done, this amount is quoted as being somewhere between $80B to in excess of $500B. My understanding is that President Obama has mandated that much (all??) of this spending be included in the published budget. This raises the basic questions of whether all federal spending is actually included in the budget, and whether the current budget proposals represent a real increase in spending or simply increased transparency (aside from stimulus spending, of course). I’d really like to hear someone discuss this in detail. Thanks for all the work in putting these podcasts together.

Mark Crankshaw writes:

The Office of Management and Budget does produce detailed summaries of federal outlays. The GPO has some handy spreadsheets available at their website (gpoaccess.gov). For example, at (http://www.gpoaccess.gov/usbudget/fy12/xls/BUDGET-2012-TAB-4-1.xls)
one may access an excel spreadsheet that details historical outlays by agency from 1962 to the present. This spreadsheet includes estimates for 2011 to 2017. From this spreadhseet one can quickly determine how much budget growth there has been by department. The dirty little secret: almost EVERY Federal department and agency has seen a substantial level of budget growth in recent years. From 2007 to 2010: Defense up 26%, Social Security up 21%, Dept of Agriculture up 53%, Department of Commerce up 104%, , HHS up 27%, Dept of Labor up 264%!, Dept of Education up 40%, the EPA up 33%, Dept of State up 73%, etc., etc.

The Federal spending binge didn't leave anyone out...

keatssycamore writes:

At about the 11 minute mark, John Taylor says the following, "We increased our debt. It wouldn't have been more stimulative. Had it been larger, it might have been more counterproductive." This comment comes right after Taylor claiming (about minute 10:30) that he & his peeps DID construct a model & run the numbers for what a much larger stimulus could/should have wrought (" In fact, with these data we built a little model, and simulated the counterfactual of a larger stimulus").

So why is it that 20 seconds after telling us about this modeling work, Taylor will only venture a "might have been counterproductive"? I mean, c'mon JT, you just said you did the work, so don't play hide the ball with us. Just tell us if your model showed a bigger stimulus was, or was not, more counterproductive.

Or I guess you could just start leaving out the "might have been more counterproductive" line.

Chambana writes:

Funny how Russ rarely challenges guest's methodology when his/her story is consistent with his view of the world. Contrast this and last week's podcasts...

FYI, here are some interesting stimulus findings from this week's The Economist.

"Food stamps also help stimulate the economy more than other forms of government spending, points out Jim Weill of Food Research and Action Centre, a charity, since their recipients are so poor that they tend to spend them immediately. When Moody’s Analytics assessed different forms of stimulus, it found that food stamps were the most effective, increasing economic activity by $1.73 for every dollar spent. Unemployment insurance came in second, at $1.62, whereas most tax cuts yielded a dollar or less."

http://www.economist.com/node/18958475

Gary Rogers writes:

I found the discussion about the efficacy of the recent stimulus interesting by what was missing from the discussion as much as by what was said. What I heard discussed was that that the money was held by states, banks and other recipients so that the expected multiplier effect was nullified by the inefficient distribution of stimulus funds. This was very enlightening but what I did not hear was the other side of the stimulus transaction. Whenever the government spends money on a stimulus, the money must come from someplace and that is rarely discussed. The only place stimulus money can come is from is the private sector so you cannot evaluate the results of a stimulus without looking at the effect of pulling that money out of the economy someplace else. So what happens when you pull a couple trillion dollars from the economy through taxation, borrowing or inflation? If quantitative easing is a valid way to inject liquidity into the economy through the purchase of securities, doesn’t financing a stimulus package through the sale of bonds have exactly the opposite effect? If the government borrows from me by selling me a bond, wouldn’t I have less available cash and forego some of my expenditures? Wouldn’t this cause the people who would otherwise have sold to me to cut their own expenses? Doesn’t this have as much of a negative multiplier effect as the stimulus has a positive multiplier? Could it be that the reason the Fed has had to go to a long term zero discount rate supplemented by quantitative easing is a direct counteraction to the effects of financing our huge government deficits? I would like to have heard Dr. Taylor’s answers to some of these questions.

This was another great EconTalk. Thanks!

Frank Flynn writes:

for me two things stand out that I would like to call attention to:

First: at about 5:01 "We looked at the money sent to the states, the grants that were meant to jump-start infrastructure spending, other government purchases. It didn't."

The news in 2008 & 9 was of potential breadlines as everyone (public / private sector) killed projects. This stimulus was not promised to increase infrastructure spending - it was to prevent the total collapse of existing infrastructure projects. States saw their revenues plummet and were laying off thousands of workers.

So if States we able to continue as before then it did work, no?


Second: at about 24:15 "I think the way to do that, most constructive for economic growth, is to reduce spending, not to raise tax rates."

Taxes rates are at an historic low - they are significantly lower than in 1981-82 when "coming out of the 1981-82 recession, that recovery's first two years had 7% growth".

So there's something else going on here - the economy can thrive with somewhat higher taxes - it has before even in bad times.

I'm personally inclined to cut spending but I think we should raise the taxes we need - and I have never heard a convincing argument as to why cutting taxes is always good and raising taxes is always bad.

Bob Anderson writes:

I am disappointed that John Taylor was asked to dissect the past and predict the future, but was never held to account for the inaccurate predictions he made in previous programs. Why was that? If you want your interviews to be something other than sychophants sharing the same opinion, ask some tough questions.

Frank Howland writes:

An interesting podcast. I too think that Russ should have challenged his guest a bit more. Some comments:
--Most important, Taylor ignored the problem of medical cost increases in describing the government's rising spending. I agree that most if not all federal agencies should have their budget cut, but it is the cost of medical care which is the major problems. In setting a 19.5% of GDP limit on spending, Taylor is really asking for cutbacks in Medicare and Medicaid.
--In the 18:59 section of the podcast, Taylor said that the Treasurys the Fed is buying are safe, but he didn't address the mortgage backed securities.
--Does Taylor really support a balanced budget amendment? Russ should have probed him on this.
--The remarks about textbooks were interesting. Cash for Clunkers is a great example. I wonder whether textbooks in financial markets and institutions and money and banking will be revised. I doubt that there will be any significant changes.

Frank Howland writes:

Here's a very recent citation which supports some of what Taylor said regarding the stimulus:

Aizenman, Joshua and Pasricha, Gurnain K. (2011) "The Net Fiscal Expenditure Stimulus in the US, 2008-9: Less than What You Might Think, and Less than the Fiscal Stimuli of Most OECD Countries," The Economists' Voice: Vol. 8: Iss. 2, Article 5.

Form the Summary: Understanding how the economy reacted to fiscal stimulus in the aftermath of the deepest recession of the last fifty years is essential. Joshua Aizenman ... and Gurnain Kaur Pasricha ... show that aggregate fiscal expenditure stimulus in the United States, properly adjusted for the declining fiscal expenditure of the fifty states, was close to zero in 2009. Furthermore, the USA is ranked at the bottom third in terms of the rate of expansion of the consolidated government consumption and investment of the 28 OECD countries they studied recently.


blu writes:

re: Mark Cranshaw reference to OMB budget spreadsheet, total outlay in the past decade increased from (in millions) $1,862,846 (2001) to $3,818,819 (2011 est) or almost $2 trillion, over $1 trillion in last 5 years.

[Edited per email correspondence.--Econlib Ed.]

Zane Rockenbaugh writes:

As others have noted, there were some unchallenged straw men, but overall thought Taylor did a really good job laying out the case against stimulus and monetary fiddling. It definitely moved the needle for me on the subject even if some of the assertions were, IMO, a bridge too far.

Personally, I've a bit of a fetish for the the gold standard proponents. IMO, it's a terrible idea, but I share the desire for a mechanistic/rules based monetary policy. So I especially appreciated that portion of the talk.

Thanks guys.

John Doe writes:

Like the other commentators previously said, it's obvious when the guest and the host are more aligned ideologically. I'm not saying Russ should interview him from the perspective of Paul Krugman or Brad DeLong, but he should at least do the most basic of research by reading their blogs. He could have had a much more interesting interview by asking about these disagreements. Or maybe Russ is aware of all the debates and they are uninteresting and Taylor is clearly correct (I'm not an economist after all).

At 32:50 in the podcast, Russ says: "I have not seen a careful analysis of which agencies are spending more relative to 2007. We understand that when unemployment goes up unemployment insurance is higher, food stamps are higher, but those are relatively small amounts of money. Where are the other 5% points of GDP?"

And then John Taylor blames it on government workers defending their turf and doesn't even bring up the points that Krugman raised during the debate they had on their blogs a few months ago. You can find the two posts by Taylor and the three posts by Krugman by following the links.

Krugman says: "I pointed out that about half the rise in the ratio of spending to GDP reflected low GDP, not high spending, and that the other half was basically safety net programs..." So that is where the 5% is, the "income security" programs and the financial crisis. Taylor doesn't actually debate Krugman on this point exactly (conceding the point I assume), but on a related question. That's obviously not a detailed analysis but it's certainly relevant to the question Russ asked. But it's interesting that Taylor doesn't even bring this up.

And you could have asked Taylor about this quote from his own textbook: "An expansionary fiscal policy will have a relatively strong effect on aggregate demand if interest rates don't rise by much [when government spending increases]..." Last time I checked interest rates haven't risen dramatically (in June 09' he blamed the modest rise on long-term debt concerns) and still claims the stimulus may have been counterproductive (something he was claiming as early as June 09' before the results could even be evaluated, which Krugman pointed out during the show).

John Doe writes:

I may be pointing out the obvious here, so please forgive me if I am. But to add to my last point, it sounds like Taylor may believe the last paragraph of my previous post and essentially agree with Paul Krugman. So why isn't he writing and speaking about the need for another properly constructed stimulus?

If you go to Taylor's blog he basically says he thinks it isn't possible politically and operationally to have the right kind of stimulus. But that's an entirely different debate. But I think that point is easy to miss (or maybe it's just me). Taylor comes off as if he dismisses the idea of stimulus theoretically as well, which just doesn't make much sense if he subscribes to his own economic paradigm. As Krugman points out here.

And while Taylor does cite evidence for his view that a good stimulus isn't likely to happen in reality, I think his political ties and views should be considered. And the way he frames his view of the stimulus being a failure often comes across as if he's dismissing the theoretical basis for it as well (by not making it clear that the stimulus package wasn't a good Keynesian stimulus package, which it wasn't). But that would make sense considering his political views and his opinion that they are impossible in the real world.

On to his concern about the debt. When tax cuts are proposed the debt is secondary, or maybe there is some actual evidence that cutting taxes doesn't contribute to our debt problem. Because as far as I can tell, this and other similar assertions that Bush's economic plan "is a prudent plan that assists the lowest-earning Americans the most with tax cuts while paying down our (nation's) debt" is also operationally and politically impossible.

The above and other quotes are here.

Charles writes:

Econbrowser attempts to answer the same question for the effect of moneteray policy:

here:
http://www.econbrowser.com/archives/2011/07/evaluating_quan.html

and here:
http://www.econbrowser.com/archives/2010/10/qe2_estimates_o.html

and seems to reach the opposite conclusion: large sscale asset purchase as had about -40/60bp contribution to the negative trend in 10 year treasury yield.

Charlie writes:

Agree with some of the other comments that Russ was uncharacteristically non-chalant about accepting JT's empirical work. It was especially striking only a week after Russ spent several minutes challenging, whether Banerjee could really create confidence intervals in a randomly controlled trial (the answer as Banerjee was trying to explain was, of course you can, but Russ didn't seem to accept it).

Russ Roberts writes:

Charlie,

Taylor's empirical work may or may not be iron-clad but it is different than much of the empirical work I typically criticize. Look at his work on cash-for-clunkers or the Bush tax cut stimulus. He looks at what happened to automobile purchases in the one case, or spending vs. savings in the other. He is not trying to tease out the effects of some variable while holding a bunch of other stuff constant--a situation where different assumptions will change the magnitude and direction of what you're trying to measure. Of course even in Taylor's work there is a question of interpretation. It is possible that other events caused automobile purchases to slump after cash-for-clunkers expired. But it's a relatively transparent exercise.

On the Banerjee, I stand by my skepticism. Go to the comments and you can read the follow-up where I try to make my concerns clearer. You can decide for yourself whether I'm right or not. But I didn't have an ax to grind against Banerjee, I was just trying to figure out the methodology. I think the statistics of those techniques is more complicated than he conceded--but again, decide for yourself.

Charlie writes:

Russ,

Every stimulus advocate says that state and local government aid didn't increase state spending, it just stopped states from drastically cutting spending. John Taylor says, look states didn't increase spending, therefore spending didn't work. It adds nothing to the debate. The debate is still over the counterfactual, what would have happened in absence of the stimulus package?

To take that simple fact gathering and pretend it proves that once again Federal spending didn't work is just plain wrong, and yet, Taylor said it several times. There's a reason macroeconomists do all those complicated things, the simple things don't work either. If you had applauded his transparency, but pointed out he didn't really learn anything, it'd be fine. It's just hard to imagine you applauding the transparency and letting the conclusion go unchallenged of someone on the other side.

I also just read the comments on the other site. It appears Leamer and Banerjee are both right, but you are still wrong. Here's the relevant quote from Leamer, "These controls are needed to improve the accuracy of the estimate of the treatment effect and also to determine clearly the range of circumstances over which the estimate applies." It's about increased accurracy not reducing bias.

To use your rats example, if you know a lot about certain genetic differences rats have that make them susceptible to this or that, you can do a lot of cool stuff with sampling to improve your estimator. But even if you know nothing about the rats, the experiment is still valid. The randomization does the trick for you. Suppose you want to test a rat poison, you have 100 rats and you know a specific 50 are likely immune to the poison because of genetics. You will divide the rats into their genetic groups and then into a placebo and control for each group. You'll have no variation. The 25 nonimmune rats that received the poison will die, both groups with placebo wil live as will the immune group. But say you know nothing about their genetics. You divide them into two groups, poison and placebo. Now you have variation. Some groups in the poison group won't die. You can measure that variation. Both your measure of the variation and your measure of the effect of the treatment are estimators. That is, they are one outcome of a distribution and not necessary the truth. But they are unbiased, meaning "on average" or more precisely in expectation they are true. To look at our experiment, on average will have 25 rats die that receive the rat poison. But sometimes we'll get 50 and sometimes we'll get 0.

This is why we are able to run useful experiments on things we don't know that much about. We could show that washing your hands between patients saves lives, even without knowing anything about the germ theory of disease. Now a days, we'd like to control for all the different types of exposure to germs that doctors may have, and we'll get better estimates after controls, but it isn't necessary to have a valid, unbiased experiment.

xian writes:

listening to j taylor is almost always enlightening. his blog is also a must read on these issues and makes the strongest cases for interpretations of events/data. on the whole, i learned a lot from this discussion.

it's amazing how on some topics (usually revolving around the experience of others or some specialized knowledge/skill) econtalk is the most engaging conversation i know of; a real treasure.

yet, if the topic is too easily framed in ideology, then econtalk becomes a one-handed economist.

and so the result here was a lot of "call and response". no real digging, just giddy "yeah bro, me too bro".

a recurring theme on this podcast is confirmation bias and skepticism, but it gets applied selectively. this is a very interesting behavior.

disappointingly, it looks like a case of epistemological closure, a complete absense of conflicting ideas.

i wonder if the lack of challenging an ideological comrade is bc econtalk doesnt know the arguments against? or maybe doesnt understand them enough to make the case? im sure econtalk could if it made an honest effort.

if this were science, then competition would force the practioners to know the counter arguments.

very sad economics cant bring itself to take that humble step- not even on econtalk.

still luv u!


xian writes:

woops...i meant epistemic closure.

Russ Roberts writes:

Charlie,

To clarify one issue--I was not on the "other side" of Banerjee. I was asking him a simple methodological question that was unexplained in his book--how are treatment vs. control results evaluated in experiments in poor countries. I assumed, incorrectly, that researchers checked to see if random samples were found to be different from each other on key variables that might affect the treatment effect.

I don't understand your rat example. And I'm a little confused by the hand washing example, too. In these experiments we aren't just trying to get an unbiased estimate of the effect. We're not sure there's any effect at all. To assess whether an effect is significant, you need to decide what sample size to use. I'm arguing that if randomization is not enough because the treatment and control groups are quite different, then it isn't clear which sample size you should use to measure significance and therefore it's hard to make judgments about effectiveness vs. none at all.

So to take your hand washing example. Suppose I want to test a NEW theory of germs, that wearing white shoes is good for reducing infections. I'm not just trying to measure the magnitude of the effect. In fact, as it turns out, wearing white shoes doesn't help prevent infections but I don't know that. That's what's to be determined.

Suppose I find that the patients of doctors wearing white shoes in my experiment had lower infection rates. Shouldn't I check to see if the doctors who wore the white shoes were more or less likely than the control group to wash their hands? Shouldn't I care about the prevalence of handwashing in the general population of doctors vs. my sample? And in worrying about whether the infection rate is purely random, that is, the confidence interval around my post-estimate, what is the correct sample size if almost all of the doctors in my sample wash their hands but most doctors outside my sample don't?

Charlie writes:

"To assess whether an effect is significant, you need to decide what sample size to use."

No you don't. Actually, if it's a truly new experiment, you'll have no idea what sample size to use, because you'll have no idea of the variation that you'll experiment. After you run the experiment you'll have an estimate of variation (remember it is a random variable), you can then use that to see if your result was significant. [If you thought you had info on the variation from earlier studies, you have some options, updating Bayesian priors is one example, but this is not required.]

"I'm arguing that if randomization is not enough because the treatment and control groups are quite different,"

Sometimes it seems like old school economists didn't grapple very much with expectation. In expectation (or on average in lay terms) the treatment group and the control group are exactly the same. That is the virtue of a random sample. Yes, there is a risk that you will draw a bad sample and get to the wrong answer. That is what my rats example was about. Random sampling doesn't prevent type I or type II errors (confirming a false hypothesis or rejecting a true one). But there is no bias introduced in the selection of treatment and control groups, since that is selected at random. In expectation, they are exactly the same.

"then it isn't clear which sample size you should use to measure significance and therefore it's hard to make judgments about effectiveness vs. none at all."

I think I covered this. To know what sample size to use, you'd have to know the variance within the population (which in real life you never do). Then you can decide how narrow you want the confidence interval. If you want to reject a hypothesis in favor of a null, you actually have to already know the size of the effect to choose the right sample. (of course, if you know the variation and the effect, why run the experiment?)

"Suppose I find that the patients of doctors wearing white shoes in my experiment had lower infection rates."

It's possible to infer from your example that you aren't clear on what a treatment and control group are. To run this experiment, you have to take n number of doctors and randomly assign them to wear white shoes or nonwhite shoes. You can't just split the doctors into two groups based on the shoes they're wearing. If you did that, it could be that wearing white shoes is correlated with some other latent variable (like washing your hands).

In the course of running this experiment, you will also get an estimate of the sample variation. You will use that sample variation to test whether your effect is significant. I think most of the time you would find white shoes had no effect on patient outcome. I think sometimes you would get a false positive. But suppose that you did find significant results, and suppose you even replicated the experiment again with an even larger sample. I would believe you were an honest researcher and take your results seriously. I would try to generate a hypothesis explaining your results, and try to test it.

Here is my working hypothesis: White shoes make (some) doctors self-identify with cleanliness more, so they wash their hands more.

"Shouldn't I check to see if the doctors who wore the white shoes were more or less likely than the control group to wash their hands?"

What if you did this!!! What if after you ran your first experiment, you looked at your results and said, "I wonder if I got a bad sample. I wonder if I just stuck white shoes on doctors that happened to wash there hands more. I'll add a control variable for washing ones hands. Oh look, now there is no effect. Wearing white shoes doesn't make patients healthier."

You (potentially) just invalidated your experiment. By controlling for washing ones hands, you are assuming that washing ones hands is uncorrelated with wearing white shoes. If your assumption is wrong and my hypothesis is right, wearing white shoes improves patient health precisely BECAUSE it causes doctors to wash their hands more. The treatment works! (To reiterate, I doubt this treatment would really work).

This concept is generally true. If you KNOW things about your data, there are lots of cool things you can do. For instance, you can sample more efficiently than simple random sampling by using that knowledge up front. You can also use the data ex post in controls, which you can use to lower the unexplained variation (increase the power of your test). But you have to be right in your assumptions, if not it's garbage in, garbage out.

"Shouldn't I care about the prevalence of handwashing in the general population of doctors vs. my sample?"

I think this was covered. On average you won't get a bad sample. Rejoice! If you know TRUTH, you can use it both before and after, but it is not required.

"what is the correct sample size if almost all of the doctors in my sample wash their hands but most doctors outside my sample don't?"

I am very glad you ended this way, because it gives me a chance to reiterate my initial point. You must think in expectation. In expectation, your random sample will representative of the population. Yes, you can make errors by drawing bad samples and getting bad estimates, but on average you won't.

If you know things that are true, you can use it to your advantage. If you are wrong, you'll mess everything up. If, for instance, you KNOW that washing your hands effects the estimate of variance and is uncorrelated with being in the experiment or by the treatment, you can use that to shrink the confidence interval. [It is very strange to say, what size sample should I choose, if doctors in my sample... You have already chosen a sample size, if you have a sample. You could for instance do a stratified sample presorting doctors by how much they've washed their hands (again with lots of qualifiers) and then use that to choose your sample.]

Russ,

I appreciate that you took the time to address my comments. I meant them constructively, and have tried to address the points you made as clearly as possible (albeit verbosely). I'd enjoy continuing this discussion, if you have the time to respond.

Even when I'm giving you hard time. I'm always a very faithful listener of econtalk.

Regards,
Charlie

xian writes:

t-test. that's pretty much it.

Charlie writes:

P.S.

"I don't understand your rat example."

The rat example was just to address what you said in the podcast about confounders, underlying variables that affect the experiment. My rat example is 100% determined by confounders, but both experiments work. In the first, we find that the rat poison has a 100% mortality in one type of rat and a 0% mortality in the other. In the second, on average our estimator will show the poison kills 50% of rats. Both experiments are unbiased (our average answer will be the truth). The confounders don't matter in expectation. I'm not assuming I know the poison works. As I said, in the second experiment I will sometimes accept a false hypothesis or reject a true one. That's because as the researcher, I don't understand the genetic variation. But the brilliance is, I don't have to. [The first example has no variation by construction, but if we added some variation, say in reality, the poison kills 80% in one population and 20% in the other, we could get type I and type II errors.]

"And I'm a little confused by the hand washing example, too. In these experiments we aren't just trying to get an unbiased estimate of the effect. We're not sure there's any effect at all. To assess whether an effect is significant, you need to decide what sample size to use."

This thinking is backwards. You pick the sample size before you estimate the variation. For any sample size you pick, it will affect your chances of getting a small enough confidence interval. If you pick a million you will be very confident your estimator is close to the truth, if you pick 20 you will not be so sure your estimator is close to the truth. If you know the population variance in advance, then you can pick the size of your confidence interval. This is really uncommon in experiments, because the variation is based on a response to the treatment.

Russ Roberts writes:

Charlie,

I messed up when I wrote about picking the sample size. I meant when evaluating the significance of the finding, what n should you use--is the sample size the right n or is it something else because the sample is not representative.

One last (?) thought. In the white shoes example, suppose you don't have data on hand washing. You run the experiment and you find an that the doctors who were randomly assigned to wearing white shoes had lower rates of infection and that the result is statistically significant. You later discover that the people who were assigned the white shoes happened to be disproportionately hand washers relative to the people who had the dark shoes. Wouldn't that change your assessment of the significance? Wouldn't it change how you assess significance?

Charlie writes:

Russ,

Excellent and very clear question. I will try to make my answer as clear by explicitly spelling out the assumptions that must be true.

1. The data on hand washing must be uncorrelated with the experiment.

2. The correlation between hand washing and infection must be known.

One is likely to hold, if for instance the data was collected, before the experiment. My example was designed to break this assumption by getting the hand washing data during the experiment.

Two is particularly onerous. The correlation not only must be very well pinned down, but it cannot even be assumed to be linear. If washing your hands twice a day is not twice as good as once a day and so on, you'd have to know all the nonlinear effects.

I think in many instances researchers would argue that two holds "well enough." But I think this is why Banerjee said, "some purists would say you've corrupted the experiment." In the real world, one could always argue we don't know well enough.

This is what I meant when I said, "If you KNOW things about your data, there are lots of cool things you can do...You can also use the data ex post in controls, which you can use to lower the unexplained variation (increase the power of your test)." I put KNOW in capital letters to emphasize if you are wrong in your assumption, you've corrupted the experiment.

And to emphasize again, none of this is required, though it may be helpful when your knowledge is good.

Does that make sense? I really do not find any disagreement on this point with what Leamer said and Banerjee said.

muirgeo writes:

I sure didn't get a feeling that John Taylor had a reassuring way out of this mess. Cutting spending and maybe increasing interest rates....really? That's gonna fix things.

It's amazing that you can talk about this and realize we have record corporate profits but one of the dreariest outlooks in a very long time and not talk about that bizarre scenario. Has that ever happened in the past?

Something is very different this time and blaming it on spending, the health care bill or uncertainty and thinking firing more government workers is some how going to fix this.... wow really?


And at one point he suggested we needed to get the economy back to how it was in 2007 ?? Wasn't that a bubble based on deficit spending both public and private?

I'm sorry but this just all blows me away. There is clearly a problem with an ideological blind spot here. Trying to make this story fit a preformed theme ends up being very leaky even to the eyes of a layman.


My un-titled but objective laymans view says this is all about a lack of demand from loses in wages related to our trade policies. Until those things are addressed I'd bet cutting spending will only make things worse. The professors seems to be treating a symptom not a cause.


Seth writes:

"And at one point he suggested we needed to get the economy back to how it was in 2007 ??" - muirgeo

I believe the comment was to reduce the spending of government agencies from levels that have increased substantially since 2007.

"Wasn't that a bubble based on deficit spending both public and private?" -muirgeo

And hasn't public deficit spending expanded since then?

John Doe writes:

I doubt cutting unemployment insurance would help the economy.

"And hasn't public deficit spending expanded since then?"

Much of it due to financial crisis and the collapse of the housing bubble. A reasonable reading of the evidence leads to the conclusion that decreasing the public deficits would lead to less growth and more unemployment, which of course would lead to less tax revenue. If growth stays where it is, or starts shrinking again, the public deficit will surely rise. I'm not sure how eliminating them will help the economy, but I guess people have different priorities.

P. Scott Waterhouse writes:

Russ,

This was a terrific and informative podcast. Where is the citation or link to the study with John Cogan that is mentioned at about 5:01?

I believe as well that a larger stimulus would have been a larger mistake. Taylor certainly makes a convincing case. Still many seemingly knowledgable experts continue to hold the view that more was needed or new stimulus is needed. Is there a rational pro stimulus argument that I am missing?

The questions surrounding government getting a handle on it's excessive spending is treated exceptionally well (it gets down to incentives in the structure of government) in a brief piece by James L. Payne titled: Why Congress Can't Kick the Spending Habit. Link below. This was presented in 1991. History continues to repeat and progress is indeed slow. Professor Payne has an excellent new book on the subject as well: Six Political Illusions A Primer on Government for Idealists Fed Up with History Repeating Itself

P. Scott Waterhouse writes:

Why Congress Can't Kick the Spending Habit a speech by James L. Payne to Hillsdale College in May 1991.

Comments for this podcast episode have been closed
Return to top