Things are Different at the Zero Bound

Here's a summary of a recent exchange with Tyler Cowen and Scott Sumner:

1. Tyler Cowen says that (New) Keynesians such as DeLong, Krugman, and Thoma advocate fiscal policy, but he thinks monetary policy is a better choice.

2. I respond by saying that monetary policy suffers from time-consistency problems that fiscal policy does not have, and that's one of the reasons I prefer fiscal policy. There are ways to revive monetary policy, but they are uncertain. I then note that fiscal policy has uncertainties too, and therefore my choice is not to rely exclusively on either policy tool, instead we should pursue both types of policies.

3. Scott Sumner responds by saying fiscal policy has problems too.

4. I respond by noting that the problems he is talking about are not problems in the model I am using. They may be problems in other models, and other people are free to use those models if they think they are better, but that does not change the fact that within the class of models I am using the problems are not present.

5. Scott Sumner responds by saying he was using a different model, one he made up on the spot yesterday when writing the original post. Unfortunately, the model in the original post is buried in a large amount of text, and the point being made is not as clear as it might have been (in response to one of his complaints, it's not always the reading that's the problem).

Using a different model is fine. I've already noted that the models I have been using have their problems, and that there can be a legitimate debate over what type of model is best. But at some point you have to commit to a specific model and use it to answer your questions, one that has hopefully been carefully specified and thoroughly investigated, and that does not change daily. Full awareness of the model's weak spots and limitations should be used to qualify the answers you give, but you cannot avoid committing to a model of some sort. The policy advice I have been advocating is based upon these models and is fully consistent with them. That doesn't mean that other models won't give different answers, but those aren't the models I am using.

Now, assuming I've unearthed it correctly, let me turn to Scott's specific objection. One way to impose Scott's objection on the models I am using is to assume the Fed is a strict inflation targeter, i.e. that it never let's the price level deviate from target if there is any way at all to avoid doing so (so inflation is always zero in the model unless the zero bound for the nominal interest rate gets in the way). I base this interpretation on the following statement Scott made in a follow-up to his original post:

if people begin to believe the Fed intends to keep core inflation at 1% per year for the next 10 years, there really isn’t much fiscal policy can do. 

Here's Woodford's description of what happens in this case during normal times, i.e. when the Fed pursues a strict inflation target and the interest rate is above zero:

As an example of another simple hypothesis about monetary policy, suppose that the central bank maintains a strict inflation target, regardless of the path of government purchases. (For conformity with the assumption made above about the long-run steady state, suppose that the inflation target is zero.) In the case of the Calvo model of price adjustment,... maintaining a zero inflation rate each period requires that p_t = P_t each period... [U]nder this policy, aggregate output Y_t will be the same function of G_t as in the case of flexible prices, and the multiplier will be given by (1.7). Again, this result does not depend on the precise details of the Calvo model of price adjustment. ...

Equation 1.7 is the multiplier in the classical model with full price adjustment, and it is necessarily less than one. So if the Fed is a strict inflation targeter and prices are sticky, the result is the same as if prices are fully flexible. Thus, under strict inflation targeting fiscal policy will not be very effective in times when the interest rate is above zero. When government spending goes up during normal times, inflation increases, and sharp increases in the real interest rate are needed to return inflation to its target value. The sharp increase in the real interest rate offsets the increase in output brought about by the increase in government spending, and this is what makes the multiplier small in this case. Under reasonable parametrizations, there "really isn’t much fiscal policy can do."

[More particularly, with strict inflation targeting, the multiplier is less than one. When the Fed follows a Taylor rule instead of strict inflation targeting, the multiplier is larger, but still less than one (though not always, it could even be less than the multiplier for strict inflation targeting under some conditions). If monetary policy maintains a constant real rate instead of following a Taylor rule, the multiplier is equal to one.

This means that during normal times, sticky price models predict fiscal policy multipliers of a magnitude less than or equal to one, with the exact magnitude depending upon the rule the Fed follows, i.e. how the real interest rate responds to fiscal policy changes. These values are much like those we see from actual estimates using data from time periods when the interest rate is above the zero bound. Thus, contrary to what many people believe, estimates of multipliers less than one obtained using data from time periods where the nominal interest rate is above zero (e.g. war periods) actually support New Keynesian models.]

The results above are only applicable when the interest rate is unconstrained by the zero bound. But things change when the zero bound is a constraint (as Scott seems to acknowledge in a subsequent post). It's no longer true that fiscal policy multipliers are relatively small. In general, at the zero bound fiscal policy multipliers are greater than one, and this remains true under strict inflation targeting. Woodford explains:

Note that if, as in Eggertsson and Woodford (2003), it is assumed that the central bank pursues a strict zero inflation target as long as this is consistent with the zero lower bound, then the ... values computed here for the multipliers dY_L/dr_L and dY_L/dG_L are the same under that simpler hypothesis.

The larger multiplier occurs because the increase in government spending increases inflation (more precisely it reduces the rate of deflation). If the crisis is expected to last another period with some probability, as it will in the model, then government spending is expected to persist as well and expected inflation will rise. The increase in expected inflation lowers the real interest rate when the zero bound is a constraint (even with strict inflation targeting), and the lower real interest rate generates additional economic activity.

Note that the source of the increase in expected inflation is the expected increase in government spending in the next time period. All that's required for expected inflation to rise is that fiscal policy is expected to persist another period. However, the Fed won't do anything in response to the rise in inflation expectations because under the assumptions of the model the target interest rate remains negative (and to go back to an earlier point in the discussion, the expected inflation is credible due to observable changes in fiscal policy in the present time period).

The bottom line is that despite recent claims to the contrary, when the economy is at the zero bound fiscal policy is still effective, i.e. it has a multiplier greater than one, even under strict inflation targeting.

If I've mischaracterized anyone, I'm sure I'll hear about it.