On Fiat Money

Why do people accept fiat money? It is, after all, intrinsically useless. David Glasner has written a thoughtful post on fiat money. Glasner writes:

Why should a fiat money not be able to retain value? Well, consider the following thought experiment. For a pure medium of exchange, a fiat money, to have value, there must be an expectation that it will be accepted in exchange by someone else. Without that expectation, a fiat money could not, by definition, have value. But at some point, before the world comes to its end, it will be clear that there will be no one who will accept the money because there will be no one left with whom to exchange it. But if it is clear that at some time in the future, no one will accept fiat money and will then lose its value, a logical process of backward induction implies that it must lose its value now.

So why are people willing to accept it? Glasner suggests that the reason is because the government accepts money as payment for taxes. This is the chartalist view and has been around for some time. Glasner traces this back to Wicksteed. Ross Starr provides an analysis within the context of general equilibrium theory. However, I would like to suggest another proposition. To do so, we need to talk about two different questions: (a) why does fiat money exist?, and (b) why do people accept it?

There are a number of ways that have been suggested to explain the use of money. However, I would like to focus on one in particular. As emphasized by Brunner and Meltzer (1971) and Ostroy (1973), money is a substitute for information. To illustrate this point, consider an example. Suppose that there are three people, three goods, and three time periods. Person 1 produces Good 1 in Period 1, but wants to consume Good 2. Person 2 produces Good 2 in Period 2, but wants to consume Good 3. Person 3 produces Good 3 in Period 3, but wants to consume Good 1. Notice that there is a basic double coincidence of wants problem. This problem, however, does not necessarily require money. For example, the three individuals could get together and make a multilateral arrangement in which each individual promises to deliver the desired good to the appropriate trader in each period. No money is necessary.

The example above, however, requires that the individuals can perfectly commit to those actions. In reality, the individuals have an incentive to renege on their promises. For example, Person 3 could promise to give Good 3 to Person 2 in Period 3 so long as Person 1 supplies Good 1. However, if there is disutility associated with production, Person 3 has no incentive to produce anything given that they have already received their consumption good. But even this isn’t sufficient to require money. If trading histories are perfectly monitored and costly available to all parties, this would provide an incentive for individuals to behave honestly.

This latter assertion, however, bears little resemblance to the world in which we live. We do not have costless access to the trading histories of every possible counterparty. As such, when individuals cannot perfectly commit and there is imperfect monitoring, money is essential in the language of Hahn in that it expands the possible allocations available to economic agents. The role of money in this context is as a substitute for information. Money is memory.

The informational role of money makes money essential and is therefore preferable to other arrangements. However, this does not resolve the solution by backward induction that Glasner suggested above. Thus, we need to answer question (b) above.

In search models of money it is standard to denote the price of money as . The price of money refers to the goods price of money rather than the money price of goods as we are usually accustomed to thinking. It is important to think about the price of money because it is possible that this price could be zero (i.e. we have a non-monetary economy). A condition for a monetary equilibrium in these models is that

This implies that it must be true that the price of money today is greater than or equal to the present discounted value of the price of money tomorrow. In more familiar language to monetary theorists, it must be true that the inflation rate is greater than or equal to minus the rate of time preference.

According to Glasner, we know that at some date, T, the world ends and therefore nobody would accept money. Through backward induction, nobody would accept money today. However, the solution by backward induction is contingent upon knowing the date at which the world ends. For example, suppose that the probability of the world ending tomorrow is where is the information available at time t. Thus, the expected future value of the price of money at time t is:

Thus, we can re-write the necessary condition for equilibrium as:

So long as the probability of the world ending is not equal to one, a monetary equilibrium obtains and is therefore not subject to the backward induction critique. Fiat money can therefore be thought of as a rational bubble. We do not need to appeal to irrationality as Glasner suggests. Rather as a trader I know that money increases the set of feasible allocations in trade and therefore I have an incentive to use it and accept it so long as I anticipate that others will accept it in the future. In addition, I know that it is intrinsically worthless, but so long as the future is not certain (or time is considered infinite) it can have positive value because of its role as medium of exchange. As a result, fiat money trades above its fundamental value.

Health Care and Health Insurance Are Not the Same

Jeffrey Toobin writing in the New Yorker:

The main argument that opponents of the health-care law have come up with is that the mandate regulates economic inactivity—i.e., not buying insurance—and the Commerce Clause allows only the regulation of economic activity. In the first appellate review of the law, last summer, the Sixth Circuit demolished that argument. The court pointed out that there are two unique characteristics of the market for health care: “(1) virtually everyone requires health care services at some unpredictable point; and (2) individuals receive health care services regardless of ability to pay.” Thus, there was no such thing as “inactivity” in the health-care market; everyone participates, even if he or she chooses not to buy insurance.

I don’t understand this argument. Health care is not health insurance. There is no inactivity in the health care market, but there is inactivity in the health insurance market. The law mandates health insurance purchases, not health care purchases. This does not seem to be a valid comparison.

Also, Ezra Klein writes regarding Paul Ryan’s budget plan:

It’s Medicaid and other health spending, which includes the Affordable Care Act, where Ryan really brings down the hammer: That category falls by 1.25 percent of GDP. So Ryan’s cuts to health care for the poor are almost twice the size of his cuts to health care for the old.

The plan cuts government-provided health insurance to the poor, not health care to the poor. The distinction is important because there is evidence that in some cases being uninsured is comparable to being on Medicare (see here, here, and here).

Schools of Thought, Rules, Discretion, and Rational Expectations

Arnold Kling has written a piece in which he implies that Keynesians and Monetarists are all victims of confirmation bias. They see things around them that reinforces their priors and they view this as evidence that their worldview is correct. He also concludes by suggesting that he thinks that both schools of thought are wrong. There are a couple of points that I want to assess.

First, I have pointed out confirmation bias before when Paul Krugman described why he is a Keynesian. However, as I discuss in that post, the problem is not that Krugman is wrong, but rather that there are other theories that predict observationally equivalent observations. The key is to develop a null hypothesis to differentiate between the two. Confirmation bias is only a bias if you are incorrect, otherwise it is evidence. (I am reminded of the idea that the first stage of being an alcoholic is the same as that of someone who is not alcoholic: denial.)

The problem, I think, is with falsifying schools of thought. You cannot falsify Keynesianism, Monetarism, or any other school of thought. Each of these schools — to the extent to which they even exist today in the profession — is made up of diverse members. There are a broad range of issues in which there is agreement. Some of these positions might be well-founded, others not. However, the idea is to develop testable hypotheses. If my model is correct, I would expect to observe X. Do I observe X? Are there other possible explanations? If so, can I differentiate between these explanations?

One of the most dubious aspects of modern macro, in my view, is the role of sticky prices. One can track prices and determine that they are slow to adjust, but does that mean that they are important from a macroeconomic perspective? Not necessarily. In order to test this, one needs a better way to test this hypothesis than simply observing that prices adjust slowly or asking business people how, when, and why they adjust prices. If sticky prices don’t have macroeconomic effects, this is an IO story. The conventional way to test the hypothesis is to write down a model without sticky prices and the identical model with sticky prices, simulate the models and see which one better predicts what we observe in reality. However, does this necessarily tell us that sticky prices are important? No. It could be that something is missing from the model and that sticky prices are picking up that effect. For example, we know that monetary policy shocks have persistent effects on real output. A basic RBC model cannot replicate this observation. Modifying the RBC model to have sticky prices (i.e. a New Keynesian model) is better able to explain the persistence. But so what? The RBC model cannot pick up this characteristic because all markets are centralized. In reality, the true data generating process might have segmented markets, decentralized trading, or a variety of other features. The fact that models with sticky prices perform better than models without does not necessarily tell us anything.

The point that I want to make is not that we should forsake sticky price models or a belief in the macroeconomic implications thereof. Rather, my point is that we need to generate better testable hypotheses to examine these questions. In other words, the reason that I am skeptical of the role of sticky prices is not because I believe that there is no evidence, but rather that it is not clear how this evidence stacks up against possible alternatives, which have not been examined. The evidence in favor of these views lies in the confirmation of a basic hypothesis. It does not, however, mean that those who accept the important macroeconomic role of sticky prices have committed confirmation bias, but rather that they have tested the hypothesis against insufficient alternatives.

What’s more, this is only one issue. Keynesians believe in a lot more than sticky prices. As such, it should take a lot more evidence to overturn an entire school of thought. And examining each issue is hard. Economics don’t get to use natural experiments. As such, one needs to develop clear, testable null hypotheses against reasonable alternatives.

This brings me to my second point. Who are the stubborn Monetarists and the stubborn Keynesians? The view of stubborn Monetarists and Keynesians suggests that those who see the world through a similar lens to these historical schools of thought are somehow standard bearers for this line of thinking and that nothing much has changed. New Keynesianism is different from Old. New Keynesians use microfoundations, they accept rational expectations, etc. Modern monetarists are largely divided into two categories: Market Monetarists and New Monetarists. If we take out nominal income targeting, there is very much agreement among the two groups. The blend of modern Monetarists is different from Old Monetarism, although both advocate the importance of money, both from a micro and a macro perspective. Nonetheless, there is a difference between modern Monetarism and Old Monetarism. Gone are the days of advocating constant rates of money growth, for example.

Arnold Kling seems to suggest that the primary difference between Monetarists and Keynesians is one of policy, with Keynesians favoring fiscal policy and discretion and Monetarists favor monetary policy and rules. This is a mischaracterization and is, in fact, undermined by Kling’s own article as he implies that John Taylor is a stubborn Monetarist advocating the Taylor rule. But there is nothing Monetarist about John Taylor or his monetary policy rule. Taylor is a New Keynesian and his monetary rule is of central use in New Keynesian models. While it is true that the most vocal advocates of fiscal stabilization are what one would consider Keynesians, the consensus of the profession prior to the recession was not one in favor of fiscal stabilization. See, for example, James Bullard’s “Death of a Theory” paper.

Finally, tucked into this article is a mild critique of rules and rational expectations. Kling states:

The rational expectations hypothesis virtually ensures that macroeconomic performance will be better under rules than under discretion. If policy follows predictable rules, then individuals in the private sector will not make mistakes that come from guessing incorrectly about the path of policy. If such mistaken guesses are the main cause of recessions, then the use of rules could eliminate recessions.

First, the literature on rules versus discretion with rational expectations was really about the inflationary bias of the central bank, not about recessions. This mixes the view of Friedman and other Old Monetarists that the Fed tended to either stimulate too much, causing inflation, or contract too much, causing a recession with the rational expectations literature that focused on inflation bias.

Second, Kling seems to offer a veiled criticism of rational expectations. What rational expectations suggests is that people do not make systematic errors. As such the literature on rules versus discretion suggests that if the central bank is able of systematically misleading the public, then rules are better than discretion. In that sense the rational expectations hypothesis is important to the debate, but what is the alternative? Even under adaptive learning, the solution will converge to the rational expectations solution.

Overall, I don’t see schools of thought dominating the literature. I see a lot of loud voices in the blogosphere and I think that Arnold Kling views these voices as a representative sample of the profession. I think this is very misleading.

Forecasts and Standard Errors

Via John Whitehead, I was led to this discussion of climate change by the EPA. The EPA repeats the following claim from the IPCC:

The average surface temperature of the Earth is likely to increase by 2 to 11.5°F (1.1-6.4°C) by the end of the 21st century, relative to 1980-1990, with a best estimate of 3.2 to 7.2°F (1.8-4.0°C) (see Figure 1). The average rate of warming over each inhabited continent is very likely to be at least twice as large as that experienced during the 20th century.

Does the standard error on this century-long forecast seem small to anyone else? (I’m not being facetious, I’m genuinely asking in the hopes that those will greater knowledge of the issue will answer.)

Pet Peeves: Monetary Policy Edition

A recent op-ed in the WSJ posits the following claim:

A few critics of quantitative easing (QE) and the zero interest rate (ZIRP) have correctly pointed out that these policies weaken the dollar and thereby reduce the purchasing power of American paychecks. They increase the risk of future inflation, obscure the true cost of the unsustainable fiscal policy the federal government is running, and transfer wealth from savers to debtors.

But QE and ZIRP also reduce long-term economic growth by punishing savers, reducing saving and investment over the long run. They encourage the misallocation of resources that at a minimum is preventing the natural rebalancing of our economy and could sow the seeds of another painful boom-bust.

There is a lot going on in these statements. Also, I don’t like thinking about monetary policy in terms of interest rates. Nonetheless, here are my main points of rebuttal:

(a) Does the Fed control real interest rates in the long run? In other words, the author seems to think that by having the Fed target a federal funds rate near zero that this effects savings and investment (i.e. lower interest rates lead to lower savings and since savings=investment this means lower investment and lower growth). But don’t people make savings and investment decisions based on real interest rates whereas the Fed is targeting a nominal interest rate? Consider an example to illustrate my point. Suppose that the Fed came out tomorrow and said that they wanted to promote savings and investment and therefore were raising their target of the federal funds to 5%. Holding inflation expectations constant, the Fisher equation implies that the real interest rate would rise substantially. However, we cannot hold inflation expectations constant in this example because that policy would necessarily have an effect on inflation expectations. In fact, it is very likely that this policy would induce expectations of deflation. The policy would do nothing to raise real interest rates.

(b) The use of “boom-bust” is clearly a reference to the Wicksell-Hayek business cycle model. However, in that model the boom (which leads to a subsequent bust) is caused by malinvestment due to the fact that the market rate of interest is set below the natural rate of interest. Assuming that this view is empirically valide, what is the current natural rate of interest? One empirically testable hypothesis from this view is that when the market rate of interest is held below the natural interest rate, this results in ever-accelerating inflation. Do we observe ever-accelerating inflation? Do we at least observe rapidly growing productivity in the face of stable inflation? I think that the answer to those questions is no.

(c) Does inflation reduce the purchasing power of American paychecks? First, this depends on how we measure things and the time horizon. If American paychecks have less purchasing power, then we would expect to see real compensation declining. We don’t. Perhaps the author is referring to the purchasing power relative to other currencies. Have we seen a marked depreciation in the USD relative to the Euro or the Pound sterling? No.

Do not mistake this analysis for downplaying the costs of inflation. The costs associated with inflation can be significant. However, just because something costs more in nominal or real terms does not necessarily mean that individuals are worse off. For example, we have seen a large increase in the price level since 1970. Nevertheless, given that an 18 cubic foot refrigerator cost about $400 in 1971 and the same size refrigerator costs about $450 today, the average worker in 1971 would have had to work about 107 hours to purchase that refrigerator (based on average wages) whereas the worker in 2011 would only have had to work about 23 hours.

Rising prices aren’t a cost of inflation, they are the definition of inflation.

(The costs of inflation are the things that would have been avoided in the absence of inflation. For example, the inflation tax on money holdings, “shoe leather” costs, costs associated with shorter contract durations and negotiation, etc. Read Axel Leijonhufvud, for example.)

(d) Inflation reallocates wealth from savers to borrowers only when it is unexpected. Has inflation been higher than expected?

Money and Measurement

My colleague Michael Belongia and his co-author Peter Ireland have new NBER working paper entitled, “The Barnett Critique After Three Decades: A New Keynesian Analysis.” Here is the abstract:

This paper extends a New Keynesian model to include roles for currency and deposits as competing sources of liquidity services demanded by households. It shows that, both qualitatively and quantitatively, the Barnett critique applies: While a Divisia aggregate of monetary services tracks the true monetary aggregate almost perfectly, a simple-sum measure often behaves quite differently. The model also shows that movements in both quantity and price indices for monetary services correlate strongly with movements in output following a variety of shocks. Finally, the analysis characterizes the optimal monetary policy response to disturbances that originate in the financial sector

Measurement matters.