Brad, the market will tell you when monetary policy is easy

The IS/LM debate continues. Scott Sumner and Brad DeLong now debate how to define “easy money”. Here is my take on how to identify easy and tight money.

In a world of monetary disequilibrium, one cannot observe directly whether monetary conditions are tight or loose. However, one can observe the consequences of tight or loose monetary policy. If money is tight then nominal GDP tends to fall – or growth is slower. Similarly, excess demand for money will also be visible in other markets such as the stock market, the foreign exchange market, commodity markets and the bond markets. Hence, for Market Monetarists, the dictum is money and markets matter.

Furthermore, contrary to traditional Monetarists, Market Monetarists are critical of the use of monetary aggregates as indicators of monetary policy tightness because velocity is unstable – contrary to what traditional Monetarists used to think. As Scott Sumner states:

“Monetary aggregates are neither good indicators of the stance of monetary policy, nor good policy targets. Rather than assume current changes in (the money supply) affect future (aggregate demand) with long and variable lags, I assume current changes in the expected future path of (the money supply) affect current (aggregate demand), with almost no lag at all.”

Hence, contrary to Milton Friedman’s dictum that monetary policy works with “long and variable lags”, Scott Sumner argues that monetary policy works with “long and variable leads”. Hence, the expectation channel is key to understanding the impact of monetary policy.

Market Monetarists basically have a forward-looking view of monetary theory and monetary policy and they tend to think that markets can be described as efficient and that economic agents have rational expectations. Therefore, financial market pricing also contains useful information about the current and expected stance of monetary policy.

Market Monetarists therefore conclude that asset prices provide the best – indirect – indicator of the monetary policy stance. Market Monetarists would like to be able to observe the monetary policy stance from the pricing of a futures contract for nominal GDP. However, such contracts do not exist in the real world and Market Monetarists therefore suggest using a more eclectic method where a more broad range of financial variables is observed.

Generally, if monetary policy is “loose” one should see stock prices rise, the currency should weaken and long-term bond yields should rise (as nominal GDP expectations increase). For a large country such as the US, a loosening of monetary policy should also be expected to increase commodity prices. The opposite is the case if monetary policy is tight: lower stock prices, strong currency, lower long-term yields and lower commodity prices.

Market Monetarists only favour “looser” (tighter) monetary policy if NGDP expectations are below (above) the central bank’s policy objective. Hence, Market Monetarists would always conduct monetary analysis by contrasting the signals from market indicators with how far away the objective is from the “bull’s eye” (the policy objective). This is illustrated in this chart.


A Market Monetarist version of the McCallum rule

McCallum – a inspiration for Market Monetarists

Scott Sumner has often expressed his admiration for Bennett T. McCallum. I share Scott’s view of McCallum and think that his work on especially monetary policy rules has been much underappreciated.

Even though McCallum does not automatically qualify as a Market Monetarist there is no doubt that a lot of his work have similarities with views expressed by the main Market Monetarists.

McCallum is a Market Monetarist in the sense that he stresses the money base rather than interest rates as the key instrument for monetary policy. Furthermore, McCallum argues that the central banks should target nominal GDP (NGDP) rather inflation or the price level. However, contrary to Market Monetarists McCallum stresses the rate of growth of NGDP rather than the level of NGDP.

The McCallum rule – nearly Market Monetarist

McCallum’s work on monetary policy has led him to propose the so-called McCallum rule. Here I am quoting from his 2006 paper “Policy-Rule Retrospective on the Greenspan Era”:

“This rule specifies the growth rate of the monetary base that the Fed should generate, rather than the value of the FF interest rate. Although in fact the Fed does not control growth of the monetary base, it could do so if it chose to and, in any event, we can use this growth rate as an indicator of monetary policy ease or restrictiveness, even if the Fed is not operating so as to exert control of this rate. The rule can be written as

(1)                 ∆bt = ∆x* − ∆vt + 0.5(∆x* − ∆xt-1).

Here the symbols are: ∆bt = rate of growth of the monetary base, percent per year; ∆vt = rate of growth of base velocity, averaged over previous four years; ∆xt = rate of growth of nominal GDP; ∆x* = target rate of growth of nominal GDP. In rule (2) the target value ∆x* is taken to be the sum of π*, the target inflation rate, and the long-run average rate of growth of real GDP (which is presumably unaffected by monetary policy). I take the latter to be 3 percent per year, so with an inflation target of 2 percent, we have ∆x* equal to 5.”

A couple of Market Monetarist modifications 

So far so good. The target for NGDP growth is by the way the same as suggested by Scott Sumner and seems more less to be what the Federal Reserve implicitly was targeting during the Great Moderation and McCallum has in a number of papers demonstrated empirically that the Federal Reserve policies during the Great Moderation more or less were in line with the McCallum rule.

Even though the McCallum rule is rather Market Monetarist in nature it is still not the real thing. I would especially highlight two weaknesses in the McCallum rule that need to be corrected to make it truly Market Monetarist.

First of all, the McCallum rule does not take into account changes in the money multiplier, which obviously is a serious defect in the present situation where the money multiplier has decreased significantly. McCallum implicitly assumes a constant money multiplier. This is obviously problematic, as both traditional monetarists as well as Market Monetarists would argue that is it the development in broader monetary aggregates rather than the money base, which is important for NGDP.

David Beckworth and Josh Hendrickson have both suggested modifying the equation of exchange (MV=PY) to take into account changes in the money multiplier. From Beckworth:

“Note first that since the money supply (M) is a product of the monetary base (B) times the money multiplier (m), MV=PY can be expanded to the following:

BmV = PY

In this form, the equation says (1) the monetary base times (2) the money multiplier times (3) velocity equals (4) nominal GDP or total nominal spending (i.e. aggregate demand). The Fed has complete control over the monetary base, B, which is comprised of bank reserves and currency in circulation.”

As McCallum’s starting point is the traditional equation of exchange it is straightforward to incorporate the Beckworth’s and Hendrickson’s insight. So the first step modification would be to re-write the McCallum rule to:

∆bt = ∆x* − ∆vt − ∆mt + 0.5(∆x* − ∆xt-1).

Note that in this version of the McCallum rule ∆vt is the rate of growth of broad money (for example M2) velocity averaged over the previous four years. In the original version v was velocity of base money rather than of velocity of M2. ∆mt is the rate of change in the money multiplier averaged over the previous four years. This modification therefore mean that we indirectly are targeting broad money rather the money base.

My second Market Monetarist objection to the original McCallum rule is that it fundamentally is backward looking in nature. Market Monetarists stress that market prices provides the best information for the tightness of monetary policy and therefore the best information for forecasting NGDP. Hence, instead of using the lagged development in NGDP the expected growth for NGDP should be used.

My own – quite preliminary – empirical work indicates that NGDP growth one quarter ahead can be forecasted quite successfully based on relatively few financial variables (stock prices, bond yields, the dollar index and commodity prices).

I suggest forecasting the following model for expected NGDP growth model on US data:

E(∆xt+1)=a0+a1∆st+a2∆rt+a3∆et+a4∆ct+a5 ∆xt

Where, ∆st the quarterly change in S&P500, ∆rt is the quarterly change in 30-year or 10-year US Treasury bonds (whatever works best), ∆et the quarterly change in an index for a nominal trade weighted dollar-index and ∆ct is the quarterly change in global commodity prices (for example the CRB index). The expected signs of the coefficients would be a1,a2,a4,a5>0 and a3<0 (higher e is assumed to mean stronger dollar).

Estimating this model should be straightforward even though there obviously could be problems in terms of multiple-correlation, but those challenges should be relatively easy to overcome.

Obviously it would not be necessary to estimate an equation for expected NGDP if there was a tradable NGDP future, but as we all know such a things does not exist in the real world.

Using the estimated equation for expected NGDP growth one quarter ahead we get the following modified McCallum rule:

∆bt = ∆x* − ∆vt − ∆mt + 0.5(∆x* − E(∆xt+1))

Now we are pretty close to having a Market Monetarist version of the McCallum rule, which takes into account changes in the money multiplier and is forward-looking in nature.

The McCallum-Christensen rule

The only issue that we have not discussed is McCallum’s focus on the growth of NGDP rather than the level of NGDP. However, it should be noted that the McCallum rule is a feedback rule and if in NGDP growth (lagged/forecasted) falls below ∆x* then the money base will be expanded. This should ensure that NGDP returns to a stationary path, but it is not given that it would return to the pre-shock trend if a shock where to hit the economy.

The size of the coefficient on the feedback part of the modified McCallum rule does therefore not necessarily have to be 0.5. It could be bigger or smaller.

Therefore, I suggest writing the rule in a more generalized form:

∆bt = ∆x* − ∆vt − ∆mt + beta(∆x* − E(∆xt+1))

Where beta is a coefficient bigger than zero. Further empirical work and simulations of the rule will have to show what would be the appropriate size for beta. I hope some of my readers will take of the challenge of the further empirical work on the McCallum-Christensen rule.

%d bloggers like this: