In Support of Stimulus
Paul Krugman recently argued that the case for fiscal stimulus is weak, given December’s employment report. In this post, I offer a countervailing perspective. My discussion takes explicit account of model uncertainty. This is atypical among economists, but I see it as essential if we are to provide compelling analyses of policy questions.
By way of backdrop: We have gone through a long period of overly tight monetary policy (as measured by sustained high unemployment and low inflation). At the same time, total factor productivity (TFP) has been low. The question is: what do these data imply about the impact of an aggregate demand increase on TFP?
Theory 1: An increase in aggregate demand would not increase TFP.
The benchmark theory is that TFP evolves due to factors that are largely exogenous to economic policy. Under this theory, a large fiscal stimulus will lead demand to push up on scarce supply, and cause inflationary pressures. The Fed would need to tighten monetary policy in order to keep inflation near its 2% target. Because the economy is at or near potential, keeping inflation near-target will require keeping aggregate output close to current forecasts.
This is, I think, the conventional model of how stimulus will operate. It is this analysis (combined with the implicit eventual need to increase tax distortions) that leads many economists to argue against stimulus. But it’s worth noting that, even under this benchmark theory, there are two potential benefits of stimulus.
First, an increase in public investment would generate a re-allocation of aggregate activity from the private to the public sector. There are good reasons to see such a re-allocation as being desirable. The private sector apparently sees the risk-adjusted rate of return on its marginal investment as being lower than corporate borrowing rates. In contrast, it seems plausible that the risk-adjusted return to many forms of public investment is higher than historically low US government debt yields. These observations imply that it is optimal to shift economic activity from the private sector to the public sector.
Second, many, including myself, have argued that the low value of r* is due in part to a shortage of US government debt. Expanding the government debt would lead to a higher r*. This would be helpful to central banks around the world, including the Fed.
Theory 2: An increase in aggregate demand would increase TFP.
It does not seem unreasonable to hypothesize that low aggregate demand has led businesses to be reluctant about implementing new ideas and that has led, at least in part, to suppressed TFP. Under this hypothesis, a large fiscal stimulus could increase the level of TFP. (Its long-run growth rate would remain the same.)
In this case, a large fiscal stimulus will have much more limited inflationary pressures. The Fed will not have to raise rates all that rapidly. We should expect aggregate output, investment, and employment to increase relative to expectations.
Deciding what to do
Most macroeconomists would say at this point: look, the benchmark theory is that stimulus has no effect on TFP. You have to provide me with clear and convincing evidence that this benchmark theory is wrong. Arguably, this is the way science should proceed: start with a simple theory of what's known and only abandon that simple theory if there’s enough countervailing evidence.
But I think that this is the wrong way to approach policymaking. We have two theories of how fiscal stimulus will affect TFP after a prolonged period of suppressed aggregate demand. Necessarily, we don’t have strong evidence in support of either theory (because we haven’t had many prior episodes of this kind). (I did provide some limited evidence in favor of Theory 2 in a post I wrote early last year.)
Instead, we need to confront the model uncertainty head on. Here’s a rough table that encapsulates the risks.
Theory 1 Theory 2
Stimulus -L G
No Stimulus 0 0
Under both theories, we get an outcome of 0 (that is, “economy evolves according to current expectations”) if stimulus is not undertaken. In contrast, “stimulus” has two possible outcomes: -L and G (where L and G are both positive numbers).
The table does make a case for the “no stimulus” policy option. A cautious policymaker might seek to limit the downside risk associated with his or her choices. This policymaker would be averse to stimulus because it leads to a loss -L.
But this approach seems overly cautious, given the large possible gain G associated with stimulus. After all, as I argued above, there are good reasons to think that the loss L is relatively small. (Indeed, in creating my table, I hesitated about whether stimulus might actually be on net beneficial under Theory 1 as well.)
On net, I would say that its upside risk does make stimulus the right choice.
What data would convince me otherwise? I would be definitely be more leery of stimulus if inflation were stronger. Right now, year-over-year core PCE inflation is still running at 1.6%. If core PCE inflation was at 2% (and expected to stay at 2% or above over the medium-term forecast horizon), it would be much harder, I think, to sustain a pro-stimulus stance.
January 7, 2017