In Defense of Allan Meltzer

Allan Meltzer is one of the greatest monetary economists to have ever lived (although credit is also due to his long-time collaborator Karl Brunner, who Meltzer would be the first to acknowledge). This is not hyperbole. Meltzer’s work emphasizes the role of imperfect information, uncertainty, and transactions costs in developing an understanding of the role of money in exchange (AER, 1971 with Karl Brunner), the business cycle (with Alex Cukierman and Karl Brunner, JME 1983), evaluating alternate monetary regimes (1986 JME), and the role of monetary and fiscal policy (to be concise, on monetary policy, with Cukierman, Econometrica 1986; Economic Inquiry, 1987; with Mascaro, JME, 1983; with Brunner Carnegie-Rochester Series, 1983; on fiscal policy, with Scott Richard, JPE, 1981; with Brunner, 1993). This is not to mention Meltzer’s work on the history of the Federal Reserve and the monetary transmission mechanism.

For all the talk about economics needing to find a new vision, the work of Allan Meltzer (and his co-authors) represents a strong foundation on which to build.

What’s more, Meltzer has long embraced Keynes’s vision of uncertainty. More importantly, and unlike most others who have done so, he has recognized that uncertainty plagues not only individuals and firms, but also government policy makers.

With that being said, it was particularly disheartening to see Brad DeLong address Meltzer as follows:

Exercise some moral responsibility, Allan.

Shameless partisan hack.

Why would DeLong write such a thing? It seems that he doesn’t much like Meltzer’s critique of quantifying jobs “saved or created” by the stimulus package. Specifically, in a memo to John Boehner, Meltzer writes:

There is no greater recognition of the failure of the stimulus program to create jobs than the efforts to mislead the public into believing the program had saved thousands, or millions, of jobs.

One can search economic textbooks forever without finding a concept called “jobs saved.” It doesn’t exist for good reason: how can anyone know that his or her job has been saved?

The Administration can make up any number it pleases. The number has no meaning.

DeLong argues that this is ridiculous and appeals to authority by citing Milton Friedman arguing that the Depression was worsened by the Fed’s failure to prevent the money stock from declining:

If the concept of “jobs saved” does not exist, how come Milton Friedman says that an extra $1 billion of open market operations in late 1931 would have stopped the Great Depression in its tracks.

He continues:

You can critique models. You can critique parameters. You can critique parameters. You can critique how the calculations are done, but you cannot deny their existence, for the kind of counterfactualcalculations that Milton Friedman does are, of course, the steady diet of what economists and other policy analysts do every day.

Of course, if we expand the earlier quote of Allan Meltzer, we would find that this is precisely what Meltzer is doing in his memo. Here is the full quote:

There is no greater recognition of the failure of the stimulus program to create jobs than the efforts to mislead the public into believing the program had saved thousands, or millions, of jobs.

One can search economic textbooks forever without finding a concept called “jobs saved.” It doesn’t exist for good reason: how can anyone know that his or her job has been saved?

The Administration can make up any number it pleases. The number has no meaning. The Council of Economic Advisers gets a number for jobs saved using the same model that Dr. Christina Romer and Jared Bernstein used when they forecast that the $787 stimulus program would keep the worst unemployment rate in this recession at about eight percent. But as we all know, since that bill became law, our economy has shed some three million jobs and the unemployment rate is nearing double digits.

In other words, Meltzer believes that the model that was used to assess the benefits of the stimulus package is significantly flawed and an impractical guide to assess job creation. In addition, his earlier point is that there is no meaningful basis on which to calculate jobs “saved or created”.

Greg Mankiw has made much the same point:

That is, I do not object to claims such as,

A: “Based on our models of the economy, we believe there would be X million fewer jobs today without the stimulus.”

But it is absurd to suggest that you can say,

B: “We have measured how many jobs the stimulus has saved or created, and the number is X.”

Economists are capable of making statements such as A, but it is beyond our ken to make statements such as B. Statement B is,of course, much stronger than statement A, as it purports to be based on data rather than on models. Unfortunately, we are hearing statements like B much too often from administration officials.

Going beyond Mankiw’s description above, statements exemplified by A are dependent on the model that is used. Several questions need to be addressed. How useful is the model? What are the assumptions? How does it compare with other models in the literature?

These questions have been addressed in the paper by John Cogan, Tobias Cwik, John Taylor, and Volker Wieland entitled, “New Keynesian Versus Old Keynesian Government Spending Multipliers”. The model used by Bernstein and Romer is an Old Keynesian model. These authors compare and contrast the Bernstein-Romer model with the Smets-Wouters model that exemplifies the current consensus in macroeconomic thought. Here is their conclusion:

We find that the government spending multipliers from permanent increases in federal government purchases are much less in new Keynesian models than in old Keynesian models. The differences are even larger when one estimates the impacts of the actual path of government purchases in fiscal packages, such as the one enacted in February 2009 in the United States or similar ones discussed in other countries. The multipliers are less than one as consumption and investment are crowded out. The impact in the first year is very small. And as the government purchases decline in the later years of the simulation, the multipliers turn negative.

The estimates reported here of the impact of such packages are in stark contrast to those reported in the paper by Christina Romer and Jared Bernstein. They report impacts on GDP for a broad fiscal package that are six times larger than those implied by government spending multipliers in a typical new Keynesian model and our calculations based on generous assumptions of the impacts of tax rebates and transfers on GDP. They also report job estimates that are six times larger than these alternative models, and the impacts on private sector jobs are likely to be at variance with the alternative models by an even larger amount. At the least, our findings raise serious doubts about the robustness of the models and the approach currently used for practical fiscal policy evaluation.

Allan Meltzer clearly thinks that the model used by Bernstein and Romer is flawed. The work of Cogan, Cwik, Taylor, and Wieland suggests that there is reason to believe that Meltzer is correct to doubt the model.

In addition, Meltzer’s critique of the success of the stimulus package also raises important questions about attributing the recovery to the stimulus package. Specifically, he cites the role of “Cash for Clunkers” and new homebuyer tax credits in contributing to third quarter growth, neither of which was in the stimulus package.

Allan Meltzer is one of the greatest monetary economists to have ever lived. He deserves much better treatment than to be called a “shameless partisan hack”.

UPDATE: Mario Rizzo writes:

As we have been saying here, the claims that the fiscal stimulus has saved or created X number of jobs is not a simple empirical question. It must be an inference from a model that tells us what would have happened in the absence of that stimulus. Collecting reports from various firms or local governments about their job situations will not do. At best these individual reports are based on pop-theories on the part of the reporters about what would have happened.

Assessing the Stimulus

The current administration has unveiled an entirely new metric for measuring the success of stimulus spending. Rather than claim credit for “creating” jobs, they have focused on jobs that were “created or saved” by the stimulus. This, of course, is a preposterous notion. How do we know that a job was saved? The idea of transparent reporting from the government is welcome, but transparent reporting is only part of the problem. What precisely is the definition of a “saved” job? This might seem a bit facetious, but bear with me.

Suppose that a municipality receives money to pave a road. They hire a private firm to do the job. The firm was planning on laying off (we’ll say) 10 workers. However, given the new job, the firm keeps those 10 men on payroll. This seems pretty straightforward. It’s not. These 10 workers might be kept on the payroll until the completion of this job and let go thereafter. Does this still count as a job saved? How long does the person have to remain employed for it to be considered a job “saved”? Near as I can tell, this doesn’t factor in to the decision-making.

Consider another example. Suppose that a state or municipality announces that they are going to lay off teachers or police officers. If stimulus funds keep these individuals employed, this is considered a job that was saved. However, how do we know that state and local governments weren’t, at the very least, exaggerating the number of individuals that were going to lose their jobs in a ploy for more stimulus money?

Of course, all of this ignores the financing. The government doesn’t have money, it must borrow and tax in order to spend money. Thus, any metric of job creation measures gross job creation, but what we are really concerned with is net job creation.

With that being said, the number that has been released regarding the “saved or created” jobs was estimated to be between 640,329 and 1 million jobs. That means that the stimulus has cost between $160,000 and $250,000 per job. (Jared Bernstein calls that “calculator abuse.”)

White House officials have been quick to mention that these numbers do not include jobs that were saved or created through the temporary tax cuts. As I have mentioned numerous times on the blog, temporary tax cuts don’t work. John Taylor has documented this fact for the last two rebate checks. Thus, it would seem that including the cost of these tax cuts would actually inflate the cost per job.

Ultimately, I am not entirely sure what we are to get from the “jobs saved or created” metric. There doesn’t seem to be any true objective way to quantify such a thing. Regardless, based on the data on jobs and growth up to this point, one can hardly conclude that the stimulus has been successful.

UPDATE: John Taylor breaks down the GDP numbers and concludes that the “stimulus did not fuel GDP growth.”

Casey Mulligan writes that he is “still waiting for mistakes that underestimate the potentcy of the stimulus.”

The Future of Too Big Too Fail

As we emerge from the financial crisis, it is important to develop a framework for dealing with failing institutions. In particular, the nature of the doctrine of “too big to fail” must be addressed and re-examined. Recently, John Taylor and Larry White have spoken out about the need for a rule of law rather than a discretionary authority. Their comments and my thoughts are below the fold.

Prisoner’s Dilemma

The prisoner’s dilemma illustrated via YouTube:

HT: John Taylor

UPDATE: Scott Sumner writes, “I’ve never been more proud to be human.”

Taylor

Why didn’t anyone tell me that John Taylor is blogging?

In any event, Taylor does some of the best work in the profession — thoughtful, careful, and persuasive. Definitely check out the blog.

There is No Such Thing as ‘Clutch’

Those who know me personally can attest to the fact that I am a big sports fan. More importantly, I am a sports fan who pays close attention to statistics and what those statistics mean — especially in baseball. One of my biggest pet peeves as a sports fan is when an announcer refers to a player as “clutch”. This bothers me because it is usually after pointing out that a particular batter is 6 for 7 with the bases loaded (or runners in scoring position or . . . ) thereby ignoring the relevance of sample size. It is with great pleasure that I discovered J.C. Bradbury’s recent post on clutch hitting. Here is an overview:

I used probit models to estimate the likelihood that a player would get a hit (1 = hit; 0 = otherwise), or get on base (1= hit, walk, or hbp; 0 = otherwise) controlling for the player’s seasonal performance in that area (AVG or OBP), RISP 1989–91 performance in that area, whether the the platoon advantage was in effect (1 = platton; 0 = otherwise), and the pitcher’s ability in that area. To test hitting power, I used the count regression negative binomial method to estimate the expected number of total bases during the plate appearance and used his RSIP SLG 1989–1991 as a proxy for clutch skill in this area.

[...]

In samples of this size, statistical significance isn’t difficult to achieve; therefore, it isn’t surprising that in all but two instances the variables are significant. The two that are insignificant are the past RISP performance in batting average and slugging average. Thus, clutch ability doesn’t appear to be strong here.

However, the estimate of a clutch effect is statistically significant for getting on base. Is this evidence for clutch ability? Well, let’s interpret the coefficient. Every one-unit increase in RISP OBP is associated with a 0.00018 increase in the likelihood of getting on base; thus, a player increasing his RISP OBP by 0.010 (10 OBP points) increases his on-base probability by 0.0000018. For practical purposes, there is no effect.

There is no such thing as “clutch”.

Dow 10,000!

Again.

Measurement Before Theory, Part 3: A Further Reply to Arnold Kling

Arnold Kling writes:

What happened one year ago that caused the economy to tank over the winter?

(a) a credit crunch. Banks would not lend to one another, and they cut back on credit to businesses, which in turn caused the contraction in economic activity.

(b) a recalculation. People found out that their housing wealth was lower, so they spent less. The home construction, real estate brokerage, mortgage lending, and securitization industries found out that their services were in much less demand than they had been, and they cut back. Finally, Ben Bernanke and Henry Paulson shouted “The Great Depression might come back!” in this crowded theater, and everybody ran for the exits. For example, law firms started telling new hires to go do something else for a while.

(c) people woke up to find that the elves and helicopters had left less money lying around.

(d) people woke up to find that the Fed had lowered its de facto inflation target.

The economists I consider to be most sensible are pushing some combination of (a) and (b). I differ from the consensus in that I push (b) exclusively and minimize (a). Lots of folks–defenders of Bernanke in particular–push (a) more than (b). What Scott Sumner and David Beckworth wish to defend is (c) and/or (d).

If given the choice between these four descriptions, I would also choose (a) or (b), but mostly because (c) and (d) are misguided caricatures of what David Beckworth, Scott Sumner, and myself have been discussing. Those who believe that monetary policy was tight do not believe that there was “less money lying around.” While David and Scott might disagree on method, I think that this phenomenon can best be understood by an understanding of the quantity theory of money.

First, it is important to understand what the quantity theory is not. For example, Arnold Kling has admitted that he might be taking his “anti-monetarism to extremes” and that his motivation is to free many of us from old habits. For example, he writes:

Another habit I want to try to break is the habit of thinking that nominal income is proportional to money. At any given moment, one can take the ratio of PY/M and say “there’s your proportion for you,” but you can do that if you define M as mackerel as easily as if you define M as the monetary base.

Indeed, I am in agreement that it is meaningless to discuss variables based on their proportion to nominal income unless there is good reason. In fact, Friedman and Scwartz raised this very point in “Money and Business Cycles” (p. 213 in the reprinted version in Friedman, 1969):

The stock of money displays a consistent cyclical behavior which is closely related to the cyclical behavior of the economy at large. This much the factual evidence summarized above puts beyond reasonable doubt.

That evidence alone is much less decisive about the direction of influence . . . It might be, so far as we know, that one could marshal a similar body of evidence demonstrating that the production of dressmakers’ pins has displayed over the past nine decades a regular cyclical pattern; that the pin pattern reaches a peak well before the reference peak and a trough well before the reference trough; that it amplitude is highly correlated with the amplitude of the movements in general business.

[...]

Most economists would be willing to dismiss out of hand the pin theory even on such evidence; most economists would take seriously the monetary theory even on much less evidence, which is not by any means the same as saying that they would be persuaded by the evidence. Whence the difference? Primarily, the difference is that we have other kinds of evidence.

What Kling really seems to be suggesting is that not all changes in nominal income (and perhaps prices) can be explained by changes in the money supply. Again, this is not something that a quantity theorist would argue with. In his New Palgrave article on the quantity theory, Milton Friedman writes:

Changes in prices and nominal income can be produced either by changes in the real balances that people wish to hold or by changes in the nominal balances available for them to hold. Indeed, it is a tautology, summarized in the famous quantity equations, that all changes in nominal income can be attributed to one or the other . . . The quantity theory is not that tautology.

This provides the perfect segue into describing what the quantity theory is and why it is important to understanding the recession. (It is important to note that this is not how the quantity theory has been traditionally described. The quantity theory has come in a variety of forms and what follows is broadly consistent the QT.)

Recall the equation of exchange:

MV = Py

where M is money, V is velocity, P is the price level, and y is real output. The money supply itself, however, is a multiple of the monetary base (the currency in circulation plus bank reserves). Thus, we can rewrite the equation of exchange as:

mBV = Py

where m is the money multiplier, B is the monetary base, and V is now the velocity of the monetary base and V remains velocity as described above. This distinction is important because it stresses the interaction of the money multiplier and the monetary base. The money multiplier is a function of the reserve-to-deposit ratio, r, and the currency-to-deposit, c, ratio:

m = m(r, c)

where m_r (.), m_c (.) < 0 (m_i denotes the derivative w.r.t. i). Further, it is important to note that given M = mB, a decline in the money multiplier would reduce broader money aggregates through multiple deposit destruction.

Given this information, it would now be prudent to discuss the recession in light of this framework. The recession can largely be viewed in two stages. The first stage ran from Dec. 2007 to around the end of August 2008. This first stage was somewhat mild, or at least on par with a typical recession. The second stage, however, began in late August and early September 2008. There are two major events that correspond with this change: the collapse of Lehman Brothers and the Bernanke-Paulson testimony and TARP debacle. (John Taylor's research suggests the latter was more important to understanding the crisis.)

There are two effects that followed. First, the currency component began to rise considerably (note that this is in percentage change from the previous year):

Second, reserves increased substantially:

The sharp increase in excess reserves shown above can be attributed to increased uncertainty and to the Fed’s decision to pay excess reserves beginning in October.

These increases in currency and reserves relative to deposits all serve to reduce the money multiplier, m, as well as lead to reductions in velocity as spending falls. David Beckworth has depicted this phenomenon quite well graphically.

This decline in m and V should lead to a sharp decline in nominal spending. Thus, when we are talking about tight money we are not referencing a sudden decline in the amount of money that is lying around, but rather the failure of policy to answer the decline in m with a corresponding increase in B (I have noted that the Fed has performed relatively admirably in this case — at least in comparison to history). In fact, this is an insight that can be gathered from Friedman and Schwart’z Monetary History of the United States in their discussion of the Depression as the Fed allowed the money supply to fall because it did not increase the monetary base to offset changes in c (and in some cases, r).

What’s more, it is not necessarily that the Fed lowered their de facto inflation target, but rather that tight monetary policy created expectations of lower inflation (and lower nominal spending). Given that there is some endogeneity with respect to inflation expectations, there are other ways to measure the stance of monetary policy. David Beckworth has broken out the VARs again to analyze whether monetary policy can explain the decline in nominal spending. He shows that monetary policy explains a quite sizable portion of the decline in nominal spending (the precise size depends on the model specification and the stance of monetary policy).

Taken together, I would think that the evidence and theory presented here are at least somewhat compelling. At least certainly more so than (c) and (d) as described above.

Measurement Before Theory, Part 2: A Reply to Arnold Kling

Arnold Kling has posted a lengthy reply to David Beckworth, Nick Rowe, and myself. I will first respond only to his criticism of my previous post. (Hopefully, if I have time later, I will respond to the broader arguments that he puts forward.)

Arnold writes:

This issue of causality is a natural segue to the empirical issues raised by Josh Hendrickson. What if we can show that fluctuations in GDP are preceded by fluctuations in money demand or money supply? Would that not be strong evidenced against my monetary theory? Hendrickson points to a long tradition of economists, including Friedman and Schwartz, Allan Meltzer, and others, who claim to have found such evidence. While I recognize that this work is formidable, I retain some influence from the late Franco Modigliani, who conducted the Money Workshop at MIT. Modigliani was so frustrated by the way that monetarists would search for a definition of money that correlated with nominal GDP that Modigliani mockingly referred to M1, M2, and so on as Milton1 and Milton2.

All of the different Miltons lead me to say this: there is no single medium of exchange. Instead, there is a lot of substitutability in media of exchange. Think of all the different ways you have to pay for stuff. The way I see it, there is a lot of substitutability among stores of value, including among temporary stores of value. Because substitutability is not perfect, the Fed can fiddle around in asset markets and change the relative values of some assets. By a little bit. For a little while. But I don’t equate this fiddling with being able to hit a precise GDP target.

I think that my theory leads to a view of inflation as a fiscal phenomenon. Certainly, that works for hyperinflations–you cannot have a hyperinflation without an out-of-control government budget. The question is (and I guess we’re about to find out) whether you can have an out-of-control government budget without a lot more inflation. The monetarist view would be that if you don’t monetize the debt, you don’t get more inflation. The view that assets are close substitutes would suggest that whatever liabilities the government issues to pay for its deficits will eventually cause inflation.

There are three main points that need to be addressed. First, Kling references Modigliani’s complaint that monetarists like to choose the measure of the money supply that fits the data best. This is a legitimate criticism, to some extent, of the early work of monetarists. Nonetheless, I would point out that Modigliani himself wrote that “the stock of money has a major role in determining output and prices” (1977). Of course, showing that Modigliani said as much is not a sufficient response to the criticism. Thus, going back to the empirical evidence, it is important to address how monetary policy is measured as the Fed has more control over base money than it does the broader money aggregates. The results from the VAR literature that I referenced in my previous post has taken this into consideration by using non-borrowed reserves and the federal funds rate as the measure of monetary policy. In doing so, they are directly measuring the change in monetary policy conducted by the Federal Reserve. In this respect, the results I referenced regarding monetary policy shocks are immune to criticism that they have been chosen to fit the data.

The second point that Arnold raises is in regards to money as a medium of exchange. Money is indeed a medium of exchange. Market interaction is costly. When there are costs associated with information and transactions, individuals have to try to identify the best possible process of exchange. Money is used as a medium of exchange because it reduces the costs associated with acquiring information and allocating time to searching for optimal trading partners and arrangements. The evolution process of money as a medium of exchange traces back to Menger, but the points emphasized above have been sufficiently argued by Armen Alchian, Jack Hirshleifer, and Karl Brunner and Allan Meltzer in developing the theory of exchange.

This idea of money as a medium of exchange is rendered moot if we assume perfect information, no uncertainty, or that money and other assets are perfect substitutes. Arnold argues the latter that there is a lot of substitutability in the medium of exchange. Really? I don’t think so. Sure, one can pay for transactions with credit or checkable deposits, but these payments ultimately are settled through the exchange of money. As Brunner and Meltzer (1993: 68n) note, “Credit cards, for example, reduce a seller’s cost of acquiring information about the buyer and encourage the separation of payments and purchases. This increases (relatively) the use of deposits as a medium of exchange but does not eliminate the use of money.” Although he believes that money cannot be seen as a medium of exchange, he does concede that substitutability is imperfect and that monetary policy can affect asset prices, but that this impact is minor and short-lived. This concession, however, is important because it would seem to fly in the face of two of his main points: (1) that money is not a medium of exchange, and (2) that monetary policy cannot have real effects.

In Arnold’s final point above, he advocates the fiscal theory of the price level. Some variation of this theory has been used by economists like Sims, Cochrane, and Woodford. Carlstrom and Fuerst summarize the fiscal theory as follows:

Weak-form FT [ed. note: FT = fiscal theory] begins with an obvious link between monetary and fiscal policy. Since seignorage (revenue from money creation) is a possible revenue source, long-run monetary and fiscal policy are jointly determined by fiscal budget constraints. Whether monetary or fiscal policy determines prices involves an assumption about which policymaker will move first, the central bank or the fiscal authority. Weak form FT assumes that the fiscal authority moves first by committing to a path for primary budget surpluses/deficits, forcing the monetary authority to generate the seignorage needed to maintain solvency. Sargent (1986) describes this as a “game of chicken.”

[...]

This version of the fiscal theory predicts that fiscal policy determines future inflation as well. Although this is true, it does so only by determining future money growth. The traditional version of the FT, therefore, is not at odds with
the quantity theory, in the sense that prices are still driven by current or future money growth.

[...]

More recently, a stronger version of the fiscal theory has been posited. Strong-form FT maintains that fiscal policy determines future inflation, but independent of future money growth. Unlike the weak theory, where inflation is still (ultimately) a monetary phenomenon, strong-form FT maintains that fiscal policy affects the price level and the path of inflation independent of monetary policy changes.

It is unclear to me which theory Kling is advocating, but he sounds like he is advocating the strong-form of the fiscal theory as he seems to suggest that we have to have high inflation with an out of control budget. If that is indeed the case, it is problematic for Arnold’s view as Carlstrom and Fuerst’s examination concludes that “this is little more than an intellectual curiosity.”

Ultimately, I do not believe that we can reconcile Arnold’s theory with the empirical evidence unless he is willing to make serious concessions regarding the role of monetary policy. I would argue that it is possible to minimize the role of money and monetary policy without eliminating it from analysis. This, however, puts Arnold’s theory in a bit of a bind. If money is included, I fail to see how his theory can be distinguished from the natural rate hypothesis. Couldn’t the process of recalculation simply cause the natural rate of unemployment to rise for a period thereby limiting the ability of monetary policy to alleviate unemployment without causing inflation? Arnold theory differs only in the sense that he wholeheartedly rejects the ability of monetary policy to have any effect on unemployment and inflation in the short run.

Measurement Before Theory

Arnold Kling has been developing a macroeconomic theory that he refers to as “Recalculation”. It is somewhat of a takeoff of real business cycle theory (Kling himself has referred to this as “not your father’s real business cycle theory). David Beckworth highlights three main features of Kling’s analysis:

(1) Monetary policy has no effect on expectations in the short-run.

(2) Monetary policy has no effect on nominal economic activity in the short run.

(3) Monetary policy has no effect on real economic activity in the short run.

What is perhaps most ironic about Kling’s theory is that he has largely taken a position on monetary policy that is consistent with real business cycle theorists, but rather than doing so by assuming rational expectations (as the New Classicals did), he does so by assuming strong habit formation:

Like most economists, I view real GDP in the long run as determined by real factors, such as the supply of factors of production, the state of technology, and the nature of economic and cultural institutions. However, I view average prices in monetary units as reflecting habits. The government can change people’s habitual price behavior only by making significant, long-lasting changes in the amount of deficit that it finances by printing money. On the other hand, changes in money-printing that are modest and short-term have essentially no effect.

Regardless, the points summarized by David Beckworth present testable hypotheses. Luckily, there is a voluminous literature on money and business cycles. Beckworth himself takes up the challenge on his blog and presents evidence against Kling’s theory. (UPDATE: Bill Woolsey shows that Kling’s theory would hold if money demand was not dependent on income. Of course, a large body of empirical evidence shows that it is dependent on income. Also, as Bill notes, it doesn’t make sense when carried to its logical conclusion either.) As a complement to Beckworth’s post, I would like to review some empirical evidence.

There is an abundance of evidence that suggests that Kling is wrong. Let’s begin with point number 2 as it is the easiest to refute. Kling argues that monetary policy cannot affect nominal variables (i.e. the price level) in the short run:

Another way to express my view is that there is a probability distribution for nominal GDP growth. By printing money much faster starting today and persisting for several years, the government can raise both the mean and the variance of the distribution of nominal GDP growth many years from now. However, in the short run, both the mean and the variance are determined by things that have happened in the past, including past monetary policy but also including Recalculations and other factors that affect real GDP as well as past habits of price-setting.

I suppose that this depends on one’s definition of the short-run. We know from the literature using vector autoregressive models that contractionary monetary policy does not affect the overall price level, initially, because prices are sticky. After about a year and half, the GDP deflator persistently declines. Consumer prices do, however, result in a reverse hump-shaped response that begins one quarter after the monetary shock. (For a summary of this literature, see Christiano, Eichenbaum and Evans, 1999.) It is also important to note that in the VAR literature, what one is measuring is a monetary policy shock and not the effect of systematic monetary policy. This evidence thus casts doubt not only on Kling’s hypothesis that monetary policy doesn’t effect nominal variables, but also that a regime shift is necessary to influence price behavior and therefore inflation as these “shocks” are enough to generate the behavior denied by Kling’s model.

The second point in question is whether or not monetary policy effects real variables in the short run. Kling argues that it does not:

So, the question is whether there is a medium run in which M affects Y. My bizarre monetary hypothesis is that the answer is “no.” That is, I believe that the medium run usually looks like the short run, in which changes in M show up as changes in V. The medium run only looks different if the central bank is engaging in a regime shift, changing the long-term trend of M and P.

There is an abundance of evidence, however, that money does affect real output in the short run. The pioneer monetarist, Clark Warburton, published a collection of work that demonstrated that causation flowed from money to output to prices and then to velocity, thereby both demonstrating that monetary policy has short term real effects and that changes in M are not offset by changes in V as Kling suggests. Friedman and Schwartz’s A Monetary History of the United States and Monetary Trends in the United States and the United Kingdom as well as the work of David and Christina Romer (here and here) are also particularly of interest here. What’s more the work on monetary policy shocks shows that a contractionary monetary policy shock results in a reverse hump-shaped response to real output that peaks 4 quarters after the shock. In Allan Meltzer’s paper in the JME in 1986, he uses a multi-state Kalman filter to compare direct and reverse causation of money and output and finds that the evidence is greater for direct causation of money to output. Edward Nelson recently published in the JME in 2002 a paper that demonstrated that changes in the real monetary base are an important and independent factor effecting changes in output. This corresponds with the previous literature of Meltzer and Evan Koenig that changes in the monetary base cause statistically significant changes in consumption (independent of a significant interest rate effect).

There is an overwhelming body of empirical evidence that suggests that Kling is incorrect about monetary policy. Nonetheless, I felt the need to respond to his posts because I think that recalculation is important. However, I think that Kling has taken the argument too far in terms of his discussion of monetary theory and policy. One could argue that monetary policy (and indeed fiscal policy as well) are limited in their ability to correct for unemployment when resources are being reallocated without making the claim that monetary policy is unimportant. Monetary policy clearly has effects on output and prices in the short run. Denying this leaves Kling’s theory precariously in contradiction to the facts.