Thursday, April 19, 2012

Y2Care about Monetary Rules

Y2K as a natural experiment in macroeconomics

Y2K Ready
In macroeconomics, there is almost no such thing as a controlled experiment. Unlike in physics, where one could construct a trebuchet and determine the optimal intial angle of elevation for a given projectile, in economics, one can't build an entire economy and test the effects, say, of capital gains tax cuts on real output.

Macroeconomic models are far more complex than what's needed for physics, in most cases -- instead of gravity and drag coefficients, you need all sorts of parameters, not all of them known, many of them difficult to measure with certainty, and the model itself is uncertain.

In most cases, real-world economies tend not to volunteer themselves as subjects for experimentation. Policy error, moreover, would be tremendously costly.

And yet, from time to time, stable conditions exist in the real world, and a single macroeconomic policy changes -- thus creating what economists call a "natural experiment." Sure, the experimental conditions may not be perfectly controlled -- but the real-world test informs in a way that economic modeling cannot. Economist hunt eagerly for such instances through the historical record -- the Romers' paper on tax policy in the interwar era (highly recommended reading) is one such example, using the US as "a laboratory for investigating the incentive effects of changes in marginal income tax rates."

The Y2K scare is, in my opinion, an overlooked natural experiment with tremendous potential for economic inquiry. (There was only one paper, a NY-FRB staff report from 2003, which I could find which used Y2K as something resembling a test case.) It meets the essential experimental controls: a stable macroeconomy, a very limited nature of a policy shock. What's best about it, in fact, is that the real shock of Y2K was fully anticipated, but did not happen -- it allows us to study the preemptive policy response in near-total isolation.

But first, this is why it's not complete isolation. In the run-up to Y2K, when the operation of the financial system seemed in doubt, there was an exogenous increase in the desire to hold cash rather than in bonds, and in currency rather than in electronic deposits. There was also an increase in the overnight risk premium for banks. And I think last, one could argue there was some investment -- see the John Quiggin paper, replacing computers and related equipment -- which would have otherwise not occurred. All of those are important, but this is really as good as it gets.

Y2K was, in effect, a monetary policy experiment. The Fed planned for the worst. To calm financial markets, the Fed increased bank reserves, expanding the monetary base and doubling excess reserves; it also printed a whole lot of dollars. It did this promising that it would fully stabilize nominal aggregate demand in the event of a real shock; and that in the absence of a shock, it would fully pull back on the expansion of the money supply and monetary base.As we can see in the graphs above, which map a year from July 1999, there was a large-magnitude change in all of these monetary measures, also very cleanly separated from the secular trend -- it's not hard, in other words, to distinguish signal from noise.

And I was thinking that I should show you a graph of prices during this time, or in real output, or in interest rates, or in anything -- but there is nothing. Y2K, obviously, didn't happen, but there was no consequence of the monetary expansion seen in the data other than in monetary measures. That's amazing because, while economists have known since the 1980s that an increase in the money supply matters to prices only in the long run, Y2K lets us see the strength of expectations set by a central bank rule (the rule, in this case, is the bank's ad-hoc promise to stabilize nominal aggregate demand) and thus the power of expectational channels in monetary policy.

This matters today. As in Y2K, the Fed has taken dramatic action -- expanding the monetary base and money supply, first to calm financial turmoil and then in attempted quantitative easing, and the result has been the creation of far more excess reserves than in Y2K. What Y2K gives us is the extreme case of how a policy shock based in rules and expectations ends up working; the fact that the circumstances are so similar to today, except for the lack of such a rule, tells us that markets are expecting an eventual drawdown of the monetary base (obviously). Markets do reflect, but in a much more muted capacity, the monetary expansion -- interest rates have fallen across the board, inflation breakevens are seen some lift -- but we aren't using the expectational channel at all today. Markets do not expect nominal AD stability, and the rule-less monetary easing has been underwhelming if not ineffective.

Y2K tells us that there's another way.


  1. Great insight Evan. The Fed stabilizing AD expectations before Y2K but failing to do so now is a powerful contrast.

  2. Wow. Evan, I think you have an amazing ability to take so many things that I might "know of" and put them together in a way I would have never seen.

    I reckon you've probably also looked at the literature regarding daylight savings time and monetary policy. Quite the interesting analysis of nominal shocks and real effects.

  3. Although I think that these are very interesting examples, I don't think they reflect an "ad-hoc promise to stabilize nominal demand", which didn't exist in 1999 any more than it does today.

    The increase in the monetary base prior to Y2K was almost exclusively in the currency component. (Excess reserves doubled, but they were such a tiny part of the monetary base that this contributed almost nothing to the aggregate level.) This is unsurprising: fears that electronic assets might be wiped out caused people to temporarily stockpile paper assets instead. The increase in currency was demand-driven, not a result of any deliberate decision by the Fed. The Fed kept interest rates (its short-to-intermediate term target) roughly stable during this period, meaning that the relative cost of holding currency rather than other liquid assets didn't really change. Instead, the Fed passively accommodated a change in demand. In this environment, there is no reason to expect currency hoarding to have an effect on real variables in the economy, even if the Fed has very little credibility for nominal stabilization at all.

    For instance, suppose that there were a few small outages on Y2K, and public jitteriness about electronic markets led the pre-Y2K increase in currency to be permanent. Would there have been any change to the real economy, conditional on the same state-contingent interest rate policy? Would the Fed's failure to "draw down" the monetary base have lead to a change in nominal demand, or any other macroeconomic outcome of significance? Not really.

    Incidentally, this is related to one of the reasons while "Old Monetarism" lost favor; it focused too much on monetary aggregates, which were subject to demand shocks unrelated to anything happening in the real economy. (This became particularly problematic in the 80s, as the universe of money-like instruments expanded and traditional aggregates ceased to obey relatively predictable patterns.) Thinking about monetary policy in terms of setting interest rates removed the sensitivity to a certain kind of demand shock---namely, changes in the relative desire for base money vs. other near zero-maturity liquid instruments. This meant that a central bank could unblinkingly follow an interest rate rule while experiencing massive shocks to the public's desire to hold currency, and nothing much would happen to the real economy---there would just be large changes in the base, as happened prior to Y2K. In a sense, the Y2K experience is an excellent demonstration of the benefits of interest rate targeting.

    Of course, there are many other problems with a simple interest rate rule, and its invulnerability to shocks along the base money/T-bill margin does not mean that it is similarly invulnerable to other shocks to asset demand---for instance, changes in the desire to hold liquid vs. illiquid assets. Meanwhile, "market monetarism" overcomes the classic critique of "old monetarism" by using macroeconomic outcomes (usually nominal GDP) as its target of choice, not some unstable monetary aggregate. In a narrow sense, therefore, Y2K experience as a vindication of both interest rate rules and market monetarism over their more primitive monetarist counterparts. It highlights the importance of a policy framework that's resilient to shocks to currency demand.

  4. continued...

    This is not to say that the experience doesn't hold some important lessons. We were lucky that our policy framework at the time (which was definitely not nominal GDP targeting of any form) was well-designed to accommodate shocks to currency demand. But what about all the other shocks to liquidity preferences, ones that don't necessarily happen along the base money/liquid assets margin? What about shocks to financial intermediation that simultaneously increase the demand for and lower the supply of liquid assets in general, not just currency? Then a simple interest rate rule does very poorly indeed.

    There are many ways in principle to fix this. You can try using a richer interest rate rule, or you can change to a fundamentally different approach like nominal GDP targeting. But certainly some kind of change is necessary. In this light, I think the Y2K example is best-viewed as a consciousness-raising exercise, demonstrating how our policy rule has protected us from a certain kind of shock, but reminding us that shocks exist and that our policy rule might not respond in the right way to all of them.

    Back in my active blogging days I emphasized the analogy between currency demand and other forms of liquidity demand, and I used a fairly silly example to illustrate my point. Your Y2K example is infinitely better, particularly because it's interesting to think about what economic pundits would have said if we did have a base money target and currency hoarding had plunged us into recession. You can easily imagine them making generic statements about how rates were high because people lacked "confidence", coupled with some vaguely moralistic claims that recession was necessary until we could fix some fundamental problems. But really, the recession would have been entirely preventable under a slightly improved policy regime, one that didn't cause interest rates to spike whenever currency demand changed.

    By the way, your blog is great. I also started blogging at roughly the end of high school, but I was nowhere near as far along as you. My only caution is to not let blogs influence too much of your intellectual development. There is a lot of wonderfully insightful thinking online, but there is also a lot missing---somewhat paradoxically, the "mainstream" of economic thought is almost nowhere to be found, probably because mainstream thinkers have a lot of other appealing venues to air their views. Although it is often justly attacked by bloggers, the mainstream has many good elements too, and you don't want to miss out on it.

  5. No, the Fed wasn´t "promising" to do anything. Just as it wasn´t (oficially) targeting inflation. But stabilize nominal spending was in effect what the Greenspan Fed was doing.
    You will see the same pattern of Fed action less than 2 years later at the time of 9/11. Contrary to Y2K it wasn´t expected AND happened. Monetary wise, the Fed reacted by increasing excess reserves to compensate for the drop in the base (reserve) multiplier, thus maintaining monetary equilibrium and a stable level of NGDP!

  6. "In economics, there is almost no such thing as a controlled experiment."

    Maybe you should put a "macro" in front of economics there. Randomized experiments regularly occur in micro studies and especially in program evaluation. Just ask Mathematica Policy Research, J-PAL or even better yet read this article ( by Banerjee and Duflo.