The following quotation discusses one of the more perplexing aspects of quantum mechanics:
In 1935, several years after quantum mechanics had been developed, Einstein, Podolsky, and Rosen published a paper which showed that under certain circumstances quantum mechanics predicted a breakdown of locality. Specifically they showed that according to the theory I could put a particle in a measuring device at one location and, simply by doing that, instantly influence another particle arbitrarily far away. They refused to believe that this effect, which Einstein later called “spooky action at a distance,”1 could really happen, and thus viewed it as evidence that quantum mechanics was incomplete.
I don’t plan to explain this phenomenon (and please don’t write in with an “explanation,” as you’ll only convince me that you don’t understand it.) But regardless of whether there is action at a distance in particles, I am convinced that the concept does not apply to economics. To be more specific, I don’t believe in “inflationary time bombs” hidden in money supply increases. And I don’t believe in “long and variable lags” from monetary policy shocks.
Then what do I believe? These six conjectures:
1. Monetary shocks can affect sticky wages and prices with a distributed lag, but the lag is not variable.
2. Monetary shocks can affect output with a distributed lag, but the lag is not variable.
3. Monetary shocks are very difficult to identify using traditional indicators (interest rates, money supply, etc.)
4. The “long and variable lag” hypothesis comes mostly from previous misidentification of monetary shocks.
5. When monetary shocks are correctly identified they have a very powerful and immediate impact on all sorts of asset prices.
6. If one does not observe the predicted asset price reaction from an apparent monetary shock, then the shock never happened.
I came to these conjectures in several different ways. One strand of research looked at monetary shocks during the Great Depression. I decided to investigate a model of money based on its role as the medium of account. Thus in the 1920s and 1930s it should have been possible to explain price level movements in the U.S. by examining changes in the supply and demand for gold (and in 1933 the price of gold.)
When I applied the model to the US, I was shocked by how well it was able to explain the ups and downs of stock prices, commodity prices, and industrial production. And there seemed to be almost no lags. When major central banks started massively hoarding gold after October 1929, stocks, commodities and industrial production plunged. Private gold hoarding in late 1931, spring 1932, early 1933 and late 1937 had similar powerful and immediate effects. Gold dishoarding in the summer of 1932, and 1936 had the opposite effect. And increases in the price of gold had the most powerful and easily identified effect of all (as I discussed in my George Warren post.)
Then I went back and looked at the famous examples of Friedman and Schwartz. The bank panics had a contractionary effect, but it was hard to find evidence that any of the key Fed policy movies had the effect F&S hypothesized. The famous discount rate increases of October 1931 (from 1.5% to 3.5%) seemed to have no impact at all. The same for the famous reserve requirement increases of 1936-37. And the effect of the 1932 OMOs was ambivalent, and almost certainly not what F&S claimed.
Suppose you were a really smart guy, and you convinced yourself that monetary policy was the key factor driving NGDP growth. And suppose you also reached the conclusion that fluctuations in NGDP growth explained the business cycle (or at least the demand side cycles.) But you lacked the theoretical tools necessary to correctly identify monetary shocks. The base went up in the Depression, so that wouldn’t work. Interest rates are unreliable for all sorts of reasons. So you notice that M1 and M2 plunged in the Great Depression. There’s your indicator. Even better, M2 tracks NGDP to some extent. (Of course all sorts of other nominal aggregates also track NGDP, but let’s put aside that annoying complication.)
Now you’re ready for your grand monetary history of the U.S. Now assume that M2 is not truly the right indicator of monetary shocks, it’s just one you found that works reasonably well for the big shocks. So someone asks why M2 rose this year, but NGDP did not. You respond, “just wait, it will eventually increase as well.” Now you’ve been boxed into a corner. You say money affects NGDP with a lag (even though in your famous Great Depression example NGDP seemed to fall before M2.) People are going to ask you how long the lags are, and when can we expect NGDP to rise? You answer, “it depends, the lags are long and variable.”
But even that isn’t enough. In the 1980s rapid money growth leads you to warn about a return of high inflation. The years go by, then decades, but the high inflation never comes. Now your whole model becomes discredited. But here is the ultimate irony. The best parts of your model are discarded (the monetarist transmission mechanism, skepticism about interest rates, the monetary theory of the business cycle), and yet economists of widely varying schools of thought hang on to the weakest part of the model, the slight of hand that pathetically tried to cover up for the model’s failure to identify monetary shocks, the “long and variable lags.”
You may recall in an earlier post I discussed how believers in astrology react when I ask if their model could be tested by comparing the average psychological profile of people born in different months. And the response is always, “well, it’s more complicated than that.” Yeah, perhaps there are long and variable lags.
I am pragmatist, so it was my Depression research that had the greatest impact on my views of long and variable lags. But after a while I also realized that the concept was totally at variance with rational expectations and the EMH. If monetary injections were inflationary time bombs, why didn’t they impact commodity prices, or TIPS spreads, or any of the other asset prices that theory predicts should be impacted. Then I started reading Svensson’s work on targeting the forecast, and realized the only meaningful indictor of monetary shocks was expected changes in the goal variable (inflation, NGDP, etc.)
Now I understood what had really happened in the Great Depression. The monetary shocks must have occurred right about at the same time as stocks and commodities crashed. But they crashed at roughly the same time industrial production plunged. Not just in 1929, but each time throughout the 1930s. There were no long and variable lags.
Monetarism came about long before the Ratex and EMH revolutions, and thus lacked the proper tools to identify monetary shocks. Then there was the long fruitless detour into equilibrium macroeconomics. By the time sticky-price models came back into vogue, the new Keynesians were driving the agenda. Models did incorporate rational expectations, but they still had trouble identifying monetary shocks. Woodford understood that what mattered wasn’t nominal interest rates, but rather the policy rate relative to the underlying (and unobservable) neutral rate of interest. And not the current interest rate, but the expected future path of interest rates.
But this was all so hopeless vague that real world macroeconomists went back to the old-fashioned Keynesian practice of equating low rates with easy money and high rates with tight money. When low rates didn’t produce fast growth in 2002, it was attributed to those mysterious “long and variable lags.” (Very few seemed to notice that money wasn’t easy at all in 2002, it was tight.) Lags had become an all purpose excuse not just for monetarists, but for macroeconomists of all stripes. Here are a few recent examples of the inflation “time bomb” approach:
Monetarist Allan Meltzer:
Keynesian Brad DeLong:
Austrian Bob Murphy:
I’m not trying to single out these three guys; this is the standard way that almost all macroeconomists look at things. And I don’t see how it is consistent with the EMH.
Here’s how I view the monetary transmission mechanism. The Fed does something to increase the future expected path of the base, or to reduce the future expected demand for the base. This raises the future expected path of NGDP growth. This raises all sorts of current asset prices. It raises the TIPS inflation forecast. It lowers the Baa/Aaa yield spread. And all of these things happen right away. This sets in motion changes in sticky wages and prices that occur over a longer period. But even those changes start right away. There is no mysterious “action at a distance” that violates the laws of economics. No time bombs waiting silently to go off at a future date. It is those immediate changes in expected NGDP growth and asset prices that cause the delayed changes If the immediate changes don’t happen, neither will the delayed changes.
How could everyone have gotten it wrong? Consider the following example. Suppose we have spent a long time on a certain monetary regime, and people are used to that regime. When change finally comes it might well be unexpected. Thus in the 1960s we began to experience faster than normal money growth. In past decades that would not have lasted long, as the price of gold was fixed. But now the higher money growth begins to persist, and the public starts to catch on, and AD and prices start rising more rapidly. It looks like the monetary expansion impacted the economy with a lag. But what really happened is that the initial expansion was viewed as temporary and thus had little effect (see Krugman 1998). Only when people began to understand that the policy change was permanent, did they revise their forecasts of expected future money and NGDP growth. And that change in the forecast of future nominal growth is the real monetary “shock”, the relevant shock not just in my view, but in ‘state of the art’ new Keynesian models. And there are no lags between that shock and asset prices.