Mysteries

Big open puzzles that haven't been resolved yet.

Project maintained by ege-erdil Hosted on GitHub Pages — Theme by mattgraham

Nominal shocks have real effects?

This is a mystery that is at the heart of macroeconomics, and one that has still not been resolved satisfactorily. If you’ve ever heard about aggregate supply curves or Phillips curves, you likely already know what this puzzle is about, but if not here is a short explanation:

In principle, thinking of money as a unit of account, we might expect that changes in the value of money should not have much real impact on the economy. While how fast or how slowly money loses its value should affect how much of it people choose to hold through an “inflation tax”, there should be no particular reason that a one-time change in the value of a currency should have any impact on economic activity, since the prices of goods and services would just adjust accordingly so that the same fraction of the total money stock still has the same purchasing power as before the revaluation.

Indeed, sometimes this seems to be the case. When Turkey knocked off six zeroes from the Turkish lira in 2005, the change in the unit of account had no effect on Turkish economic life. Prices and wages merely fell by a factor of 1000 and life continued as before. There are many other examples of such currency revaluations, none of them having any noticeable effect on anything beyond the number of zeroes in the bills people carry in their wallets.

However, this flies in the face of the common belief that monetary policy is powerful and has a lot of impact on economic life. If all the impact monetary policy had on the economy is the inflation tax borne by people who hold non-interest bearing currency, the impact would be miniscule compared to any other tax on the books. The way macroeconomists have historically gotten around this problem is the postulate of nominal rigidities: the idea that since the currency used in a particular economy serves as the unit of account, a lot of quantities in an economy are indexed to the value of the currency (in other words, they are nominal) without any correction for inflation or deflation, and on top of that they are also typically slow to adjust in response to changes in the value of the currency. As Ball and Mankiw state in A Sticky Price Manifesto:

We believe that sticky prices provide the most natural explanation of monetary non-neutrality since so many prices are in fact, sticky… based on microeconomic evidence, we believe that sluggish price adjustment is the best explanation for monetary nonneutrality.

This phenomenon, if real, allows monetary policy shocks to have a much more significant impact. For example, a big unexpected increase in inflation now obviously creates problems for financial institutions who hold onto long-term debts while leveraging through short-term borrowing; and a big deflation can cause an economic depression since nominal wages might be slow to adjust, causing real wages to be too high for a period in which the only recourse for employers is to fire workers since they can’t afford to pay them the new, higher real wage.

The fact that non-inflation indexed debt contracts, wage contracts, et cetera are common even in countries with relatively high inflation rates is not under dispute, though as Stephen Williamson points out in New Monetarist Economics: Models, there is some dispute about the extent to which price or wage stickiness beyond what is present in these contracts is a real phenomenon. The mystery here is why people wouldn’t arrange debt and employment contracts, product orders and so on in such a way that they are naturally indexed against inflation, and why it would take them a long time to adjust when there is a change in the value of the currency. The Italian experiment cited above shows that economies are certainly easily capable of doing such a thing, so why don’t they if these frictions impose such large costs? Or is the truth that the frictions don’t impose large costs, and the nominal rigidity-based macroeconomic paradigm is simply mistaken?

The mysteries in this case are manifold. Do nominal shocks in fact have measurable real effects or not? If they do, then why don’t people change their behavior to correct for nominal shocks, for instance by indexing contracts on inflation? If they don’t, then why do people care so much about what central banks do, and why are models with nominal rigidities still dominating contemporary macroeconomics, and what is the actual cause of recessions?

Nominal shocks have real effects: New Keynesian school

This is the mainstream view, though it has significant weaknesses. As Paul Krugman helpfully explains,

When Keynes argued against the “classical economists”, he was to a large degree arguing against the view that there is no such thing as involuntary unemployment — a view often defended, then and now, by an appeal to the usual logic of supply and demand. If we’re looking at the market for, say, wheat, and there’s an excess supply — sellers want to sell more than buyers want to buy — we expect to see the price fall rapidly to clear the market. So if there were really a large excess supply of labor, shouldn’t we be seeing wages plummeting?

And the answer is no — wages (and many prices) don’t behave like that. It’s an interesting question why, one that has to be answered in terms of psychology and sociology, but it’s simply a fact that actual cuts in nominal wages happen only rarely and under great pressure. So wage stickiness is an essential part of a demand-side story about what’s going on with the economy; it’s how you answer the question of why wages aren’t falling.

It is indeed an interesting question, if we believe it to be true, and in fact the precise explanation has significant implications for macroeconomics as a discipline.

Two popular ways of formalizing the idea of price and wage stickiness are menu costs and Calvo pricing. In a menu cost model, price-setters can change the price they charge for their product or labor whenever they wish, but they incur a certain cost for doing so, which dissuades them from doing it too frequently. In a Calvo pricing model, price changes follow a Poisson process with a fixed arrival rate over which the individual producers have no control: they must simply hold their current price until they get lucky and are finally allowed to change it.

We might think that a small menu cost would be incapable of producing a lot of price rigidity, but in fact this is not so: since at a profit-maximizing point the derivative of profit with respect to price is zero, first order changes in the optimal price of a business only result in second order losses from sticking to your current price. This fact means even a relatively small menu cost can dissuade businesses from making frequent price changes. However, menu costs are not capable of explaining why people would stick to a particular wage or price even when it’s clearly causing them enormous losses in utility or profits, for instance if they are forced to remain unemployed as a consequence of them insisting on a high real wage. As a result, menu cost models of price stickiness consistently find only small welfare losses from the phenomenon of prices and wages being sticky, which is at odds with attempts to explain the Great Depression or the Great Recession using nominal rigidities.

Calvo pricing has dubious microfoundations: it’s clear that a business can change their price whenever they want and they don’t actually have to wait for Fortune to smile upon them so that they would finally be allowed to do so. The reason Calvo pricing has been so popular is that we can simply estimate the empirical frequency of price changes from the data and then put this as a fixed Poisson process into our model. This approach has a significant weakness to the so-called Lucas critique: the fact that businesses make price changes at a fixed frequency in your sample doesn’t mean that this frequency is a constant of nature and that it wouldn’t change if central banks tried to exploit it by monetary policy, for instance.

Calvo pricing also has other problems. The New Keynesian Phillips curve that is predicted by Calvo pricing has extremely poor fit with data. People have tried to explain this lack of fit by an identification problem: essentially we observe no Phillips curve in equilibrium because the central bank adjusts monetary policy in response to changes in the economy, meaning that a simple regression trying to confirm the Phillips curve is going to suffer from endogeneity problems. This sounds vaguely plausible until you realize that a consequence of this is that models with such Phillips curves consistently give the result that almost all of the variation in inflation is explained by price and wage markup shocks, and almost none of it is explained by monetary policy shocks. In other words, we have to believe that the rise of inflation in the US in the 1970s, and its subsequent suppression in the 1980s, was all driven by mysterious and unidentified “shocks”. This doesn’t seem like a fruitful approach.

Regardless, price stickiness is a real phenomenon if we set aside concerns about its magnitude and the costs it imposes. I think the single most decisive piece of evidence in favor of it is the following graph:

US real narrow exchange rate

This graph shows the monthly change in the US real narrow exchange rate from 1964 until today. As you can see, there is a marked increase in real exchange rate volatility between 1971 and 1973, and since then we’ve continued with the new higher exchange rate volatility. This transition corresponds exactly to the timing of the abandonment of the Bretton-Woods system of fixed exchange rates and the transition to a floating exchange rate regime in the US and in many of its trading partners.

Why is this surprising? Well, if nominal prices weren’t sticky, then it wouldn’t matter if we had fixed or floating exchange rates, since a nominal currency appreciation and an increase in domestic prices have the exact same effect on the real exchange rate. However, we know intuitively that businesses don’t change their prices by about 0.25% every day, which is about the post-1973 volatility we see in this plot: real exchange rates are much more volatile under floating exchange rate regimes than fixed exchange rate regimes, and this gives strong support to the idea that there is at least mild price stickiness, perhaps driven by small but nonzero menu costs.

It’s important to note that this graph does not support the idea that prices are so sticky that we could get a Great Recession-type event from merely a contractionary monetary policy combined with sticky prices, because overall monthly volatility in real exchange rates is much too small for us to draw such a conclusion.

There is an obvious reason why people insist that the evidence for highly sticky prices and Phillips curves is so much stronger than it actually is: these mechanisms are the primary way in which modern macroeconomic models produce multipliers for fiscal stimulus and large effects of monetary policy on real economic variables. If we cast them aside, as the real business cycle research program attempted to do in the 1980s, there isn’t much left for the government or the central bank to do in order to fight recessions, and we’re once again forced to rely on unidentified productivity shocks to explain booms and recessions.

Nominal shocks have real effects: Austrian school

The Austrian school also maintains that nominal shocks have real effects, but their proposed mechanism is different. In their view, nominal interest rate manipulation by the central bank not only has large effects on the real interest rate for a short time, which is a mainstream view, it also successfully deceives entrepreneurs and businesses into thinking that an interest rate change now will persist for many years or decades into the future. If you believe interest rates will be low for a prolonged period of time, you’ll be more sympathetic to engaging in long-term projects with a low rate of return, which will drive an investment-oriented boom. However, since in fact interest rates will not be low for a prolonged period of time, there will eventually be a reckoning moment in which all of these projects will have to be abandoned and liquidated, resulting in recession.

The problems with the Austrian school view are even more glaring:

Moreover, even if none of these problems existed, the Austrian view would predict that unexpected changes in the corporate income tax rate or the capital gains tax rate should be driving boom and bust cycles, which they don’t. The Austrian story about why nominal shocks should have real effects is therefore quite shaky.

Nominal shocks don’t have real effects: Real business cycle school

The real business cycle approach to macroeconomics became popular in the 1980s, and while much of the infrastructure of their approach has survived the decades, their basic assumption that business cycles are driven by productivity shocks has fallen out of favor. A good reference for exploring this is Shocks by John H. Cochrane.

The basic idea of the real business cycle school was simply to explain variations in employment and output throughout the course of the business cycle by variations in productivity without any market frictions. If this were the case, it would mean that “countercyclical policy” by governments and central banks is suboptimal: unemployment during a recession is simply a natural response to changes in current or expected future productivity, and it can possibly be made worse by bad government policies such as extending unemployment insurance duration, forcing employers to raise wages, putting in high effective marginal tax rates, et cetera.

There are many counterintuitive conclusions that come out of the real business cycle models. For instance, such models have the feature that people react to news of a coming productivity boom in the future by reducing labor supply today, resulting in a “recession” that might more appropriately be called a “vacation”. The intuition for it is simple - people aren’t willing to work as much for the same wage if they are overall wealthier - but it’s hard to swallow the idea that good news about future economic growth would cause people to work less hours or quit their job today. These models also have a hard time producing the observed correlation structure between investment, consumption, hours worked, et cetera. More details on these problems can be found in Cochrane’s paper.

The end?

I might add more perspectives and attempted explanations to this page in the future, but for now, the question of whether nominal shocks have real effects and how big these effects are remains perhaps the most important open problem in macroeconomics, with no resolution in sight.