Modern Myths of Keynesian Economics

0:00This morning I'd like to speak about two Keynesian myths. A myth, by the way, is something that does not exist. One is the Phillips Curve and the second is the multiplier. Phillips Curve, of course, is the alleged trade-off between inflation and unemployment and the corollary proposition that the government must tailor physical and monetary policy to this trade-off. It's entirely based upon empirical evidence. A graduate school professor of mine used to enjoy calling the Phillips Curve fact in search of a theory. I prefer Professor Rothbard's characterization. In the free market reader recently, he calls it perhaps the greatest, single most absurd error in modern economics.

0:49There are three reasons, categories of fallacies within the Phillips Curve that I'd like to talk about. One, very briefly, is epistemology, and the second is just as a theory, and lastly, as an empirical observation. Epistemologically, of course, facts cannot give us economic knowledge. Facts do not search theory quite the opposite. Mises has told us that theory is necessary to give meaning to facts. This is true for two reasons. One is the fundamental axiom of economics that human action exists. It is from that foundation, which by the way is irrefutable because any attempt to refute it is in fact action.

1:38It is from that beginning point that a process of deduction, the praxeological procedure, can give us economic knowledge. We can deduce true implication from the true premise. These conclusions that we derive furthermore are universally true. They cannot be refuted by any kind of evidence. In an empirical fashion, it would be similar to trying to attempt to refute the proposition that a ball cannot be both read and non-read all over simultaneously by testing that out in different locations, seeing if that were true in Chicago, whether it were true in 1950 as well as 1989.

2:25The second reason in epistemological sense that facts cannot give us economic theory is that the facts that are created by human action are what Mises called complex. They are not statistical facts. The data that forms inflation figures and unemployment numbers is formed in a complex non-statistical fashion, and I'll talk about this in more detail later, but it's unsuitable information for empirical work. In fact, even the positivists recognize the truth in this epistemological procedure of praxeology.

3:13They teach in their classes daily laws of economics and economic conclusions that are falsified. The law of demand, the law of supply, these things are mental constructs. They all understand that and they all teach those things And they all say that those are economic principles, but they are not falsifiable, they are not empirically verified. To turn to the theory of the Phillips Curve, all authors who have attempted to provide theory to explain the fact of the Phillips Curve have presented it as a theory of the labor market. I don't intend to repeat Professor Hoppe's praxeological explanation of how the labor market operates, but I would like to state just two things that must necessarily be incorporated into any theory of either the labor market or any market for that matter.

4:15And these are principal to the rest of the discussion about theory. One is that theoretically all propositions must be linked in some fashion to the existence of human action. That I think is a foundation of the criticism, the Austrian criticism of all non-Austrian schools of thought that if you cannot demonstrate how these principles are connected or consistent The second point, theoretically, is that when any proposition is made about market phenomena, it needs to be kept in mind that the market is just a shorthand term economists use for voluntary exchange, and voluntary exchange, on the other hand, is just a form of financial exchange.

5:13The form of action that people engage in in order to accomplish their end, to say that in a more traditional language, market phenomena is always the phenomena of cooperation and coordination. There is a prerequisite that assumption that when you go into the market you are attempting to engage in cooperation and coordination with other people. Hence, all problems that can be called market problems are necessarily problems of coordinating or cooperation. They cannot be characterized as problems of aggregate things. To continue then into the theory, let's start with a quote from Phillips' original article on the Phillips curve.

6:00He says this, when the demand for a commodity or service is high relative to the supply of it, we expect the price to rise. The rate of rise being greater, the greater the excess demand. A couple of points about that statement. First of all, this is somewhat incorrect as a statement of equilibrium. When demand is high relative to supply, it makes the price high. That, in fact, is the indication that prices are a reflection of subjective value, as Mises told us. and Socialism. High demand relative to supply has nothing to do with changes in prices. Changes in prices in the market come about through entrepreneurial arbitrage. They come about through entrepreneurs buying resources at lower prices and combining them in the market and selling the product at higher prices. In addition to that, the rate of change of price in a market is completely unknown theoretically. This is purely an empirical matter. Some prices can change very rapidly like stock prices as people discovered to their dismay. Other prices change more slowly. There's nothing to be

7:15said theoretically without just an ad hoc assumption as to the rate of with Changes in Prices. And yet, Phillips designed the Phillips Curve to explain the rate of change of price. In particular, the rate of change of all wages, the collection of wages. Brings me to my next point. What is the significance of that? If we build a theory to explain the rate of change of a collection of wages, what significance does this have? Why do we want to know what determines the rate of change of a collection of prices? Notice that has no significance in determining the coordination on the market.

8:01When individuals go into voluntary exchange and begin to attempt to coordinate their activities with other people, prices emerge as an outcome of that. Prices are not formed outside of this process that people engage in in an attempt to coordinate with other people. Once those prices exist as data, then they can be combined mathematically into any kind of construct. You can add them together, you can subtract them, you can divide them into each other, etc. Some of those constructs are meaningful economically and some are not. One that certainly is meaningful is economic profit. Certainly meaningful, in other words, to multiply prices times quantities of output and resources to subtract the two to make Bank Economic Calculation.

8:49On the other hand, the price indexes that are used in Phillips Curve presentations are economically arbitrary. There's no more reason to, for example, to get a collection of wages than there is a collection of the prices of capital goods. If we are to say, if we want, in other words, to explain the rate of change and the price of wages, why not of capital goods? Why don't we construct another theory which tells us or explains for us how the prices of land change? Or why do we take all labor together? Why don't we take some subset and explain why or how there's a certain rate of change in the wages of manual labor or management or something else? The reason we don't do that is because those things have and I would argue that the same thing is true about the Phillips curve itself, at least as presented by Phillips.

9:51Lipsy, who wrote an article in 1960 following Phillips, tried to integrate in Phillips' work more into the mainstream, and he provided this as an explanation of why it was important to explain this phenomena that he called the Phillips curve. He says, this relationship between wage inflation and the unemployment rate is an extremely simple one and it holds considerable promise for empirical testing. So we are presumably under this thesis to ignore all theory, especially praxeological theory, and substitute only those things we can test. And the simpler the better, easier to test.

10:41Lipsy then, as I mentioned, attempted to reconcile, a little bit more directly, Phillips' original work and he said this about the theory of the Phillips curve. He said, we now introduce a dynamic hypothesis that the rate at which the wage changes is related to the excess demand. And specifically, the greater is the proportionate disequilibrium, the more rapidly will wages be changing. For this relation to be observed, it is necessary only that there be an unchanging adjustment mechanism in the market, i.e. that a given excess demand should cause a given rate of change of price. I'd like to go through a couple of point highlights about that. One is that the Phillips-Lipsy original work was clearly a disequilibrium phenomena, unlike The Keynesian work, or at least some interpretations of Keynes, who viewed this as an equilibrium theory.

11:41In addition to that, Lipsy here is making the assumption that, an ad hoc assumption, that there is an unchanging adjustment mechanism in the market. And then finally, Lipsy makes another assumption that at equilibrium, there is a certain level of what he calls frictional unemployment. This is the diagram he uses to construct the theoretical presentation of the Phillips curve, where the labor market is in the upper diagram and the Phillips relationship in the lower diagram. And essentially, again, what he's saying is that at point A, there is a certain level of unemployment given on the diagram at point A in the bottom panel.

12:28And that amount is what he calls frictional. That's the amount of unemployment that is unrelated to the condition of either excess supply or excess demand in the labor market itself. Now, a couple of points about this in terms of criticism of this particular presentation of Lipsy. One is that, again, it's simply ad hoc to say that there's an unchanging adjustment mechanism in the market. It's simply incorrect that whatever happens in the market happens according to human action. Individuals bring about these adjustments, and there must be some explanation of how individuals do this. And again, the explanation is entrepreneurial arbitrage based upon economic calculation. That there is an adjustment mechanism, so to speak, in the market, And there is also the underlying desire to be adjusted, that is for coordination to come about on the market.

13:21Another thing, and Professor Hoppe has pointed this out already, and this is here simply a theoretical point. The theory of equilibrium is one of what I call perfect coordination. There is no such thing as either excess demand or excess supply at equilibrium. This is the way we teach equilibrium analysis as economists, of course. We say, at wages higher than the equilibrium wage, like W0, there's excess supply. There's discoordination in the market. At wages down here, like W1, there's excess demand. But at the equilibrium wage, there's none of these things. There's perfect coordination. Now, Lipsy, again, what he's saying is that we need to, I guess, redefine equilibrium So that at point A, equilibrium on the diagram, we actually have dis-coordination, we have unemployment.

14:17He calls it again, frictional. But the equilibrium construct is not designed that way. Equilibrium means perfect coordination. I think Professor Hoppe has discussed that in more detail. Notice if we make the assumption then that we're going to stick with that definition of equilibrium, that at point A, we have perfect coordination, then the actual Phillips curve would look like this. We'd have to extend the horizontal axis and write negative unemployment on this axis to the left of zero. And the Phillips curve would actually run down through this point, where this would be point A. This would be an area of excess demand on the market, this an area of excess supply. And again, that's not exactly what Phillips had in mind.

15:05In other words, that is not the traditional Phillips trade-off. there is no trade-off there but let's just continue to go along with Phillips definition and say okay frictional unemployment does exist at point A then what can we say about theoretically within the framework of equilibrium as we understand it and teach it as economists about the Phillips relationship notice again we'll grant Phillips point that at point A we have some positive amount of unemployment we'll call frictional now as the wage When the rate increases above that level, these individuals that are brought into the market at higher wages, like W-0, that are induced into the market, are all unemployed. The level of employment is given along by these points on the demand curve. That's the actual amount of employment in the market.

16:01So, one for one, as individuals enter this market at higher wages to try to find jobs, they are unemployed. infers that the unemployment rate is increasing from the frictional level and so we do have this section of the Phillips curve that Phillips draws downward to the right because at this point the wage, in order to adjust the equilibrium would have to fall, we have a negative wage inflation and at the same time we have greater rates of unemployment because each additional individual brought into the market is unemployed so the unemployment rate must go up. In other words In other words, the labor force is increasing one for one with the number of unemployed. But what happens when we move back down the supply relationship?

16:48When we reach point A again, if we're moving back down this way, when we reach point A, we have a frictional level of unemployment. Now as we move to wages below that, what must necessarily be true about the individuals who are now dropping out of the labor force? If we start with a frictional amount of unemployment at point A, these individuals, as the wage falls and we move to these points on the supply curve, what can we say about the individuals who were employed at point A as opposed to the number of individuals employed back at the wage W-1? Well, notice those individuals, unless we are willing to make some other assumption about frictional unemployment, those individuals must have been employed at point A. If there's a given amount of frictional unemployment and no unemployment, no other type of unemployment, then as people drop out of the labor force as the wage goes down, they're taking leisure now instead of work.

17:44Those individuals were employed. And so the amount of frictional unemployment is the same while the labor force is shrinking. Notice that would make the Phillips curve do this, which again is not the traditional trade-off, right? In other words, inflation and unemployment would increase together under that scenario. And Phillips provides no mechanism by which the frictional amount of unemployment decreases as the extent of excess demand grows, that is, as we move to lower wages on the diagram. One more point about this theoretically, even if we're willing to grant to the Phillips-Lipsy argument that frictional unemployment can be reduced as we move to greater and greater levels of excess demand.

18:34It still doesn't hold that wage inflation and unemployment data can be generated by this representation of the labor market. The reason is because the rate of wage inflation is what we might call a flow variable. It occurs, it is only measured over some period of time, whereas the unemployment rate is a stock variable measured at a point in time. And notice on the diagram, we cannot create both wage inflation and a given level of unemployment simultaneously. If we start, for example, at W1 with some amount of excess demand given on the diagram, let's say by the difference between C and D, That will give us a level of unemployment and a level of the wage rate.

19:26Now, if we assume, in order to generate, we must assume that the wage is going to change, it's going to increase, so the wage is going to increase back to the equilibrium level, then that change in the wage will generate wage inflation. But notice, it also generates a change in the rate of unemployment. As we move from a wage of W1 at point C and D, back to point A, we have less unemployment. So the only thing the diagram, the only thing the theory can generate are either levels of the wage and levels of unemployment, or changes in the wage and changes in unemployment. The Phillips Curve, on the other hand, tries to correlate changes in the wage rate with levels of unemployment. The Theory simply can't do that. It's incapable.

20:16Now, as an alternative, let's try the Keynesian presentation, and with some misgivings about this, especially after Professor Garrison's discussion, because it's somewhat unclear as to what exactly Keynes meant in this area of the labor market. I like the quote, I believe, by James Tobin out of Seymour Harris's book, The New Economics, where he says it's impossible to find a representation of Keynes' work that is both comprehensible and true to the original. Anyway, it seems that Keynes had something like this in mind.

21:01First of all, he seems to imply that this is an equilibrium theory, that his theory of employment is an equilibrium one. He says, the volume of employment is given by the point of intersection between the aggregate demand function and the aggregate supply function. Traditionally, it's what we call equilibrium. In addition, employment depends upon aggregate demand in the Keynesian view. Again, quoting from Keynes, he says the object of the employment function is to relate the amount of effective demand with the amount of employment. And finally, again, with some misgivings about this, Keynes seems to imply, and most of his followers embodied this idea, that money wages do not readily adjust downward.

21:46If you read Keynes, he seems to imply that in two ways. is one that workers simply weren't willing to take money wage cuts. He says, while workers will usually resist a reduction of money wages, it is not their practice to withdraw their labor whenever there is a rise in the price of wage goods. He also thought that workers may be unable to lower the money wage. He says, again, except in a socialized community where wage policy is settled by decree, There is no means of securing uniform wage reduction for every class of labor. If we follow those particular prescriptions then for the labor market, again this is not the only way to interpret Keynes's view of the labor market, it would look something like this.

22:34This is a diagram lifted from Leontief's presentation in Seymour Harris's The New Economics of Keynes's labor market. The inelastic portion of the supply of labor represents Keynes' assumption that workers will not accept a reduction in their money wage, that they would rather accept less employment represented by point A, refusing to accept the lower money wage. Then on the diagram, Leontief at least claims that point F is the point of full employment. Any aggregate demand then that would leave demand for labor short of full employment would result in a level of involuntary unemployment as the difference between point A and point F.

23:26Again, without going into this theory in any kind of detail, Professor Hoppe has talked about some of these ideas already. I would just like to present some of the criticisms in terms of the Phillips curve, what this has to do with the Phillips curve. First of all, an equilibrium theory, unlike the Phillips ellipse, disequilibrium theory, cannot generate, without additional assumptions, the necessary data of the Phillips curve. It cannot generate changes in the wage rate and unemployment rates. All equilibrium theory can do is generate levels of those things. In addition, if we add the assumption about changing aggregate demand affecting the labor market in the way Keynes envisions it, then the Phillips curve, if we were to draw the Phillips curve according to this diagram, let's say we start at point A with a certain level of demand for labor based on aggregate demand and aggregate demand increases and hence labor demand expands and employment increases, lowering the level of unemployment.

24:37Notice, nothing happens to the wage. The money wage stays the same. So we don't get no trade-off here at all. In fact, all we would get if we were to draw it would be the horizontal axis of the diagram. That is to say, this would be a wage inflation of zero, would be consistent with various levels of unemployment as we move along the inelastic section of labor supply. Then when we reach point F, if demand were to be increased beyond point F by an increase in aggregate demand. It's somewhat uncertain as to what happens in this case, but if by full employment, Keynes means nobody else can be employed, and again, it's not clear that that's what he meant, then the rest of the Phillips diagram would just be a vertical line with wage inflation on the vertical axis because nothing would happen to the unemployment rate as the wage was driven up past point F.

25:30Again, there may be some other interpretations of the Keynes. But again, the diagram provided by Leontief cannot give us the Phillips relationship. And of course, in a more general way, the Phillips trade-off was a great setback for the Keynesians. They didn't want a trade-off. The Keynes and Hansen, some of the other earlier Keynesians believed that you could just increase aggregate demand right up to the point of full employment with no inflation. There was no inflation trade-off. If you went beyond that point, King said, prices would rise without limit. That is, it was like a light switch. There was no trade-off. It was just either all unemployment or all inflation.

26:18Now one other point to make, and I don't want to dwell on this for any length of time because it doesn't have much to do with the Keynesian myth but neoclassical economists and others that have been talked about have attempted also to provide theories to explain the Phillips curve and as I'll go into in a minute that that effort has all been wasted because there is no Phillips curve but anyway the the neoclassical work of if I may call it that of Phelps, Friedman, etc. is I believe I believe an attempt to blend Cain's work with the original disequilibrium view of Phillips and Lipsy. It's an attempt to integrate aggregate demand, not into an equilibrium construct, but a disequilibrium one, where there are some additional assumptions provided by Phelps as to what happens in the disequilibrium state.

27:18He says this, the frictions in an advanced economy create a linkage between the path of the unemployment rate over time and the path of the inflation rate generated by the course of aggregate demand. These frictions then in turn stem from faulty information, bad information, and are modeled in this work by search theories of the labor market and certain assumptions about how individuals adapt to this uncertainty that they face in the market. Just a couple of points about this, again, without going into detail. The adaptive expectation assumption is completely arbitrary. Just like all other ad hoc assumptions about labor market adjustments or adjustments in any other market, adjustments come about through human action.

28:04They come about through entrepreneurial arbitrage based upon people's calculation of profit. doesn't further our knowledge about how the real world works to simply assume that it works a particular way and especially a mechanistic way. In addition to that, it doesn't advance our knowledge very much either for the neoclassical economist to replace the homo economicus of classical economics with the homo ignoramus of neoclassical economics. Ludwig von Mises told us a long time ago that the correct view is homo agena, that man is the acting animal, that's the correct representation. Also, it has always seemed somewhat paradoxical to me that the Keynesian view of this Phillips curve literature has aggregate demand determining optimal employment.

29:02Aggregate demand for products determines full employment or optimal employment, whereas the Phelps-Freedman literature, which focuses on adjustments in the labor market, is supposed to tell us about optimal inflation. It would seem it would be the other way around, that aggregate demand would be more closely linked with inflation, and adjustments in the labor market more closely linked with employment and unemployment. Final point theoretically, and not to go into any detail, but the... Whatever link there is theoretically between inflation, price inflation, and the unemployment rate can be completely explained or is at least completely consistent with the Misesian theory of the business cycle. There isn't any need to go beyond that. Mises has already demonstrated that artificial expansion of credit will in fact result in a boom followed by a bust, as Professor Hoppe has already described to us.

29:58Whether or not that will be associated with higher rates of price inflation and lower rates of unemployment depends upon a multitude of other empirical factors that cannot be predicted in advance. Sometimes that will happen, sometimes it will not. There again is no way to know that in advance of the complex of human action that occurs that generates these prices and these unemployment figures. Secondly, I'd like to discuss in terms of the Phillips Curve, the data of the Phillips Curve itself. And then we'll talk about the actual empirical implementation or use of this data. First of all, in terms of problems with this data, and this I think is somewhat of an underappreciated problem, this data is produced by government bureaucrats and that ought to take us back a little bit.

30:56We know that government production processes are inefficient. Why do we have this great faith that the numbers they generate, unemployment, inflation, etc., are written in stone? Where does this faith come from? It would be certainly more than ironic if sometime in the future it was revealed that all these numbers were just made up, just invented. I mean economists Students have spent their whole careers trying to explain the Phillips Curve, and the Phillips Curve may be nothing more than just a bunch of bureaucrats in a back room generating random numbers. I remember again in graduate school, it sounds a little far-fetched, but again in graduate school, Opel and State brought in a speaker who used to be sort of a middle-level bureaucrat in the Bureau of Labor Statistics, and he used to be involved in the calculation of Unemployment Rate, and he told us, presumably never thinking anyone would ever repeat this, that he was wrong, that sometimes when they go through the calculation, they get the surveys

32:10back and they crunch out the numbers, that the numbers would come out wild. You get an unemployment rate of 25%, you know, when it had been seven or six. And he said when those things were sent on to the top muck mucks, the unemployment rate would be published at 7.5. And they used to have a joke apparently about this in his circles. They called this invoking the omega factor. Another problem, it seems to me, with allowing and Government Bureaucrats to produce this information for us is that this is politically sensitive information. This type of information makes and breaks political campaigns and careers and it's certainly not beyond the realm of possibility that it would be deliberately manipulated for political gain. And again, Reagan's, the Reagan administration's manipulation of the CPI is a pretty good example of that. Economists are fond of talking about the political The mathematical business cycle where government officials manipulate the levers in order to create good numbers, but it's much simpler just to manipulate the numbers themselves.

33:22Second data problem is that this data is aggregated, and you've heard much about the theoretical difficulties that aggregation creates already, And I won't go into that in detail, but just to point out again that when you aggregate prices, you wipe out the information that's necessary to understand what is going on in a relative sense as people attempt to coordinate themselves in the marketplace. Also, part of this aggregation process is aggregating unemployment figures into one number across all industries.

34:09And again, if you disaggregate that, you find that unemployment varies widely across different industries, across different geographic locations, etc., which poses something of a problem for the Phillips curve doctrine, especially the Keynesian view of it. And aggregate demand or lack of it is what causes unemployment. Why do we get these differences? Why relative differences? Why more unemployment for young teenagers? Why more unemployment in auto industry in certain times, et cetera? That begs for an answer that can be provided only by observing and analyzing these relative differences. In addition to that, though, aggregation creates an econometric problem, creates a problem In terms of estimating, as economists attempt to generate equations for the Phillips curve, briefly without going into trying to reproduce this in detail, when the least squares procedure is used to estimate a regression line, which is what is done with the Phillips curve, the The information that is used to generate the line is variation.

35:25What is actually done is a mean of unemployment would be calculated and then the line is fitted in order to minimize the variation of the actual data from this mean. Now when you aggregate data, you destroy this variation. When you sum data into one number, let's say you have 12 unemployment numbers each year here, and you sum them into one, you destroy information in an econometric sense. And it's only that type of information that can allow econometric estimation to be proper. If you aggregate data, then all of your equation estimations become biased. In fact, there really isn't anything you can say at all about the estimates with Phillips Phillips Curve data because they have been aggregated into one data point per year when in fact the government generates this data monthly.

36:20And it generates the data across different industries. And if econometrically to be correct, if economists wanted to find the actual Phillips relationship, they would have to use this disaggregated data. It's simply econometrically wrong to aggregate data, it destroys information and it makes and of course it is always improper to purposefully aggregate data, to intentionally aggregate it, which we'll see Phillips does. Third, there tends to be a misinterpretation of the data, of the Phillips trade-off. The first part of that misinterpretation is to display the data on a two-dimensional scatter diagram. Diagram. The Phillips Curve data is not generated, as we mentioned before, statistically and it cannot be represented correctly as a two-dimensional process. It at least is a three-dimensional process where time has to be taken into account in the display of the data. This is a reproduction of the Phillips Curve that I took from, Phillips Curve points, that I took from the Economic The 1988 edition of The Economic Report of the President.

37:37Now, the point here isn't to necessarily find the Phillips curve in all the points, but you should notice, by the way, that you can't readily do that. I mean, you can find all sorts of shapes in here if you look for them. But the point is here, simply, that the Phillips curve is falsified, This proposition that unemployment and inflation move negatively to one another is falsified almost half the time in this 20-year period. For example, 80 to 81, we have a positive relationship. The data, in other words, moves sequentially and you cannot display it as a two-dimensional diagram. You have to take into account the fact that you move through time from one point to another.

38:25And when you do that, sometimes the relationship is positive. Sometimes inflation and unemployment increase together, sometimes they decrease together. 82, 83 is another example, 83, 84, 84, 85, 85, 86. And you can find numerous other examples where when you move from one year to the next, the relationship is positive. And again, this is simply a misinterpretation of the data to draw a line, a Phillips curve line through here, and say that regardless of the sequence of the movement of these data points, we have a negative relationship. In actual fact, almost half the time on this diagram, as you move from one year to the next, it's a positive relationship. Notice another point about that is that you cannot predict, even if you were to run an estimation of this and come up with a negative coefficient between inflation and unemployment, you cannot predict from year to year what's going to happen.

39:19It doesn't really matter if there's a Phillips relationship in here on an average sense, in an econometric sense. What's important for government policy and what's important in a theoretical sense is, can we know with today's numbers of inflation and unemployment what tomorrow's numbers will be? Can we know if we do certain things that will move from higher inflation and lower unemployment to vice versa? Or will we, in fact, end up moving to higher inflation and higher unemployment both? And that certainly can't be precluded. In fact, that's somewhat of a normal occurrence as you look at the scatter diagram itself. One other point about misinterpretation of the data, as mentioned before in just a little bit more detail.

40:04These data points are not generated statistically. Statistically, data points are generated in a complex set of interactions between hundreds of millions of different individuals. And it's simply a misinterpretation to view this data in a unilateral, simple statistical way. It is neither, as Mises has already pointed out, it is neither repeatable, that is to say data that is brought about through history is unique. Each data point is generated in a unique sense. Statistical data, of course, must be repeatable, like flipping a coin or things of that type. It is also not random. This data is generated by motivated human beings who are attempting to accomplish something. All statistical information is generated randomly.

40:53That's a foundation of it. It's also generated, well, at least if you're going to use econometrics, this data must be generated according to a normal curve. The error must be distributed normally. And as Professor Rothbard has recently discovered, well, not to say that he only recently knew this, but recently discovered work that had demonstrated that the world is not normally distributed. It's not something that is difficult to believe, but that actual estimations have been made of the underlying statistical foundations of certain processes, and they're not normal. Then to turn to estimation to the process that Phillips and others have used to actually make an estimate of the Phillips relationship, Phillips says this in his original work, The purpose is to see whether statistical evidence supports the hypothesis that the rate of change of money wages can be explained by the level of unemployment.

41:59And a couple of points about that, and this is a widespread misconception about econometrics. Econometrics, first of all, can't explain anything. What econometrics does is show how variation in one variable is correlated with variation in another. It provides only correlation, not explanation. Explanation must be provided by economic theory. For example, there may be a very high correlation between the number of dishes washed in a restaurant per day and the total receipts that are received in that restaurant per day, but that is no theory which says that more dirty dishes causes more receipts to come into the restaurant. Another point about, implicit in that statement, econometrics only tells us about variation.

42:53It tells us nothing about whether levels of variables move together or whatever. It tells us only when one variable moves, does another one move with it. It cannot tell us anything about if wage inflation is high, is unemployment at a certain level. In addition, Phillips goes on to say this, that there's a clear tendency for the rate of change of money wages to be high when unemployment is low and to be low or negative when unemployment is high. This is a reproduction of Phillips scatter diagram from which he made that statement. And I would personally take contention with that statement. I don't think that diagram shows anything of a sort.

43:38of a sort, I don't think you can look at those points and say, without any debate, that it shows that the money wage rate is high when the unemployment rate is low. In fact, if you eliminate a couple of outlier points up here at the top of the diagram, almost all the points fall in a band, almost uniformly, almost at random through that band. There's really nothing that can be said before running a test as to whether or not high rates of wage inflation correlate with low rates of unemployment. In fact, I think it's simply a misstatement. If you look at any particular level of unemployment, you'll notice that it's consistent with various levels of wage inflation.

44:26Two percent unemployment, for example, is we find a wage inflation that's negative, about a negative half percent. We find 2%, 3%, 4%, even 7% is not completely inconsistent as a wage inflation number with 2% unemployment. And the same is true for most of the bulk of the data. Even if we move out to 6% unemployment, we find wage inflation correlated with it that is widely varied. Negative rates of inflation, positive rates of inflation. And so there again isn't anything that can be said unilaterally about a Phillips relationship from that particular scatter diagram. In addition to that, the scatter diagram is actually much worse in terms of its interpretation than it appears here because Phillips has already arbitrarily separated his data.

45:25This is the Phillips diagram for the period 1861 to 1913, but Phillips also has data from 1913 and 1957 that he doesn't put on this diagram. And the reason he doesn't do it is because most of the data points are out here. And there are a few in here and a bunch of them back here. And if you put it all together, it doesn't really look like much at all. So there's been, in my opinion, some arbitrary separation of the data. In addition to that, Phillips goes through some manipulation econometrically to get his results. One thing he does, and again this is econometrically incorrect, he assumes a particular form of the functional relationship.

46:14He assumes that the Phillips curve is log-linear. And if you know anything about log-linear, that particular form is quite inappropriate for this data because you cannot take the log of negative numbers. And so you have to eliminate all of the negative numbers in his database. Then, so he obviously should not assume a log-linear form. It's inappropriate. In addition to that, Lipsy showed in his article that log-linear is statistically inferior. He has a different form of the function which actually correlates this data much better than ellipses. Second, and if you're wondering about the crosses, Phillips does not estimate the Phillips curve with the data points.

47:01He averages these data points into these crosses so that, for example, between the unemployment rate of 2% and 3%, he averages all of the wage inflation, he comes up with this. Then he does the same thing between 3% and 4%. He averages all the wage inflation. He comes up with this cross, etc. for the rest of them. Then he fits his Phillips curve to the four crosses. He just fits the curve in between those four crosses. The other two crosses provide where the curve passes for one of the other coefficients of his equation. Not to go into any detail about his equation, but he estimates two of the coefficients with four data points. And then he estimates the other one with the remaining two outlier crosses.

47:48And again, this is chicanery. You just don't do this econometrically. You don't aggregate data on purpose to provide for a particular relationship that you're looking for. That's simply incorrect. Lipsy then cleaned up some of this mess and actually did a regression estimation for all of this data, all of the 1861 through 1913 data with an equation which allowed him to use the negative points, etc. He didn't aggregate the data as Phillips did. Now, if we were just going to sit back before this was done and we wanted to have a kind of metric criteria to determine the correlation coefficient between changes in wage inflation and unemployment, I mean, how high would it have to be for us to accept a relationship that we're going to make the centerpiece of macroeconomics and the guiding post of government policy?

48:56Would you want it to be 0.99? I mean, at least 0.95, I would think. Most econometric studies would be laughed out of the room at less than 0.95. Notice this is especially true for time series data. There tends generally to be a high trend built into time series data that gives extremely high r-squares for time series data. When I did my doctoral work, it was in econometrics, and I ran several regressions. They still haunt me at night. And the R-squares were extremely high, .99, .98. If I get an R-square of .96, I'd lay awake at night, you know, worrying about whether it was going to be accepted or not.

49:42Well, anyway, when Lipsy fits this, he gets an R-square of .64. And remember, that's to the good data. He actually fits two different curves. The R-square he gets for the data that looks fairly Phillips-oriented is 0.64, but when he does this again for the rest of the data, he gets an R-square of 0.28. Now, personally I wouldn't bet my own career, let alone the whole profession, on R-squares of 0.64 and 0.28. That's just not very good, especially for time series data. On the other hand, another article by Samuelson and Solow claimed that this was actually quite remarkable, using their own words, they called these results remarkable.

50:36They said that for the 1861 to 1913 period, there was a fairly close relationship between inflation and unemployment and for the 1919 to 1957, that the curve fit about as well. And they were so inspired by this, they did their own scatter diagram for U.S. data. It looks like this. And simply without running any regressions or any other kind of sophisticated process, they just pronounced the existence of the Phillips Curve, which, in my opinion, takes great faith. In fact, Samuelson and Solo, from that article, even said, quote, that they're points all over the place. So how did they get the Phillips curve trade-off? They did it the same way that Phillips and Lipsy did, they arbitrarily separated the data, that's what the blue points are, the blue points are the data for the 1940s and even that data doesn't look real promising for a Phillips curve relationship, then again that's econometrically incorrect, you should just take all the data, lump it together, run your regression and let the chips fall where they may.

51:48One other point of interest about Samuelson and Solo is that this is the beginning of the shifting of the Phillips Curve. This article is where Samuelson and Solo claim that the Phillips Curve has simply just shifted. It was back here in the 40s and then it shifted here and then out and out. And not too surprisingly, that did not set them aback as to whether or not the Phillips Curve really existed. that simply inspired them to call for more government intervention to keep the Phillips Curve from shifting. So we have to have a set of government intervention to keep it from shifting, and then physical monetary policy to move along it. But econometrically, what a shifting curve means is what econometricians call misspecification.

52:36In more simple terms, it means your equation is wrong. It's just wrong. That's what shifting means. And it seems certainly that it would be an act of prudence to wait until we had the right equation before we ran around pulling levers of fiscal and monetary policy to exploit an equation that doesn't exist or we don't know whether it will shift or not. One other point briefly about the Phillips curve estimation is that neoclassical economists, again I use that term somewhat loosely, have also attempted to explain this shifting, explain the econometric mis-specification of the Phillips relationship. And they have used, by and large, you'll find this in the textbooks, they've used expected rates of inflation as a potential variable to put into the Phillips relationship to explain the shifting.

53:32And of course, when the expectation of inflation changes, then the Phillips relationship shifts, and in the long run, we get the long-run Phillips curve, the vertical Phillips curve. Now, econometrically, there's some problems with that. By far, the most important is you can't quantify expectations. And if you can't get data, you can't put it into the regression. And it doesn't do, of course, to use past values as an expression of expectations. Lipsy even showed his 1960 article that past values don't tell us much of anything. They're not correlated at all, really, with future values. And hence again, we're sort of in an econometric block here, where we simply, until other variables that can be quantified or identified and put into the equation, simply don't know what to do, we don't have anything.

54:24The Phillips curve simply does not exist. And of course, if you were to find all the variables and put it in there, you would probably have all the variables. I mean, it would be everything by the time you got finished. Let me just quote one more thing. This is from the economic report of the President, from the anonymous scribes who wrote that. In their viewpoint of the Phillips Curve, they said this, that recent data provide little evidence of the trade-off between inflation and unemployment. The recent experience in other countries also appears to contradict the notion of a stable trade-off between inflation and unemployment. So perhaps, even in the mainstream work, the tide is turning on the Phillips curve, but it certainly has never existed, let alone simply disappeared in the 70s and 80s.

55:16So just in summary about the Phillips curve, it doesn't exist. Second, it can't be derived from the theory that is provided to explain it. The third, the data that's used to generate estimates of the Phillips curve is highly favorable, it's aggregated and hence highly favorable to making the case of the Phillips curve and that data is inappropriate, should use raw data, not aggregated data. And finally, even with favorable data, estimation does not support the existence of the Phillips curve. R-squares of 0.64 and 0.28 are not too large. Now if I may turn to the multiplier and make some, I will not turn to the multiplier, I'll quit, I'll quit there.

56:06Thank you very much.