Reply to Scott Sumner's post 'Michael Hatcher on NGDP targeting'

Michael Hatcher, University of Southampton. Posted on September 19th 2016.

I recently came across Scott Sumner’s comments on my first post on nominal GDP targeting (see here). Today I finally got round to writing a reply. I focus on just two points which relate to my argument about demand shocks posing particular problems for NGDP targeting.

Sumner starts out by arguing that demand shocks are not truly exogenous but instead the result of poor monetary policy:

The main purpose of NGDP targeting is to reduce the severity of demand shocks. For instance, in mid-2008 inflation in the US had risen well above target, and hence the Fed tightened monetary policy, causing NGDP to fall 3% over the next year. This was a powerful negative demand shock, caused by the Fed’s tight money policy. This shock made the financial crisis much worse, and also sharply increased unemployment. Under NGDP targeting the Fed would have had a much more expansionary monetary policy in late 2008, and hence the demand shock would have been much smaller.

The first point to make is that many demand shocks are unrelated to domestic monetary policy. An example would be a fall in the demand for UK exports due to a slowdown in China or the Eurozone. The same would be true of an unanticipated cut in government spending or an unanticipated increase in taxes. Thus, regardless of whether monetary policy is causing fewer shocks or not, demand shocks will be present and will be a force to reckoned with. It is therefore important to think about how the economy would respond to such shocks under NGDP targeting and other regimes.

Sumner then goes on to argue that NGDP targeting would anyhow be stabilizing for demand shocks:

Hatcher might reply that even with the best of intentions there would still be negative demand shocks under NGDP targeting, as monetary policymakers are not perfect. I agree. But the make-up required to reach the old trend line would actually be stabilizing under NGDP targeting. For instance, if the Fed makes a mistake and NGDP growth overshoots the target, then they need to gradually reduce NGDP to bring it back to the trend line. Normally a policy of reducing NGDP growth might cause a recession. But if you start from a position where NGDP has overshot the target, then you are starting from a position where output and employment are above their natural rates. So the contractionary monetary policy is actually stabilizing, as it brings you closer to the natural rate.

I agree that regimes based on level targets can be stabilizing -- provided that the target is credible. In fact, Patrick Minford and I have explained the intuition in the context of price-level targeting here. The basic mechanism is as follows, for the case of a nominal GDP target. Starting from a position where the economy is at or near the NGDP level target, a negative demand shock would push the economy below the target and hence call for looser future monetary policy in order to make up for the lost ground. Because the target undershoot is corrected rather than treated as a bygone (as it would be under inflation targeting), the economy receives the stimulus it needs to return output to its full capacity level.

A lot of stimulus would come from the reduction in nominal interest rates. But there is also a second mechanism: the expectations channel. The idea is that the private sector will anticipate higher future inflation and higher output, as this is what is needed to return nominal GDP to target. Higher future inflation will drive down ex ante real interest rates and encourage households to bring forward purchases of durable goods (as prices will rise in the future). Higher expected output, on the other hand, will raise household income prospects. The joint effect is to drive forward the recovery. In terms of this expectations channel, however, I expect that price-level targeting would clearly outperform NGDP targeting.

To see why, recall that NGDP targeting promises to make up for a target undershoot through a higher price level, higher output or some combination of the two. In other words, it does not provide clear guidance about what the future inflation rate will be. Consequently, the expectation that nominal GDP will rise in the future need not translate into much of an increase in expected inflation. Inflation expectations become very important at the zero lower bound because, with nominal interest rate cuts no longer an option, the only way to lower short-term ex ante real interest rates is by credibly promising higher future inflation. Since nominal GDP targeting fails to give a firm promise to this effect, it is likely to be a less effective mechanism for escaping the zero lower bound than price-level targeting.

The key difference with price-level targeting is that it promises to make up all of the target shortfall through higher future inflation -- not just some part of it. As argued by Eggertsson and Woodford, this is an extremely effective mechanism for stabilizing the economy at the zero lower bound because it drives down ex ante real interest rates when stimulus is needed. In fact, Eggertsson and Woodford (p. 182) show that optimal monetary policy at the lower bound involves a state-contingent price level target, and that standard price-level targeting provides a good approximation to this optimal policy (pp. 185-189). This is precisely because it promises to undo any deflation with subsequent inflation of equal magnitude. As argued above, nominal GDP targeting does not share this desirable feature.