Escaping the zero bound. NGDP, PLT vs raising the inflation target.

Post Post Script on the zero bound and raising the inflation target.

It's put to me that NGDP targeting is a better way to avoid the zero bound than raising the inflation target.

This question can't be answered definitively without reference to a model, or class of models.  So, as ever, let's take the standard monetary, sticky price model, of which there are several variants, used in central banks to simulate the effects of monetary policy.  It may not be right, but it is beholden on market monetarist NGDP fanatics to write down a different and better one if they don't like it, or its conclusions.

This model's steady-state obeys the Fisher relation.  [Steady state singluar there for simplicity].  Which is to say that investors in nominal assets will require compensation for inflation.  Namely, interest rates in the long run will rise one for one with a rise in the average inflation rate.  For a given variance of shocks to the economy requiring offsetting action by the central bank [at this point sticky price assumption invoked], the higher the rate of inflation associated with the monetary regime, the lower the frequency of zero bound episodes, and the less missing stimulus.

There are a whole set of regimes that would deliver higher long run inflation.  Regimes that target a weighted average of deviations of inflation from a target, and deviations of real activity from potential.  Or regimes that target a weighted average of deviations of the price level from a rising trend and deviations of activity from potential.  If you choose equal weights in the former, you get an NGDP growth target.  If you choose this weight in the latter, you get an NGDP level target.  These are optimal in some special cases, but generally not optimal.

That doesn't mean NGDP targeting is out.  It could be rescued by arguments that it is easy to communicate, defend, and likely to endure.  But then you can't really argue with any conviction that inflation targeting is inferior on these grounds.

Once you have chosen how much you want to increase the long run average inflation rate to escape the zero bound - using one of the infinite variety of regimes to deliver it, including NGDP targeting if you like - there's then the issue of whether your target should be specified as a rising levels target or a growth target.

The objective that it is optimal to assign to the central bank, assuming it can commit to it, is clear in the models.  It's a growth rates target, with a very high weight on inflation.  [About 20:1].  Curiously, and somewhat confusingly, in many varieties of this model, if you assign this growth rates target to the central bank, and it can commit to pursuing it, and, of course, there are rational expectations, then the path for the price level delivered is often, but not always, stationary.  And so it will resemble what much of the informal discussion about central bank targets means when it refers rather vaguely to a 'price level target'.  Inhabiting this model world, we can observe that if we drop the assumption of commitment, and presume instead (more realistically), that central banks can't tie their hands each period, they will suffer more frequent and or larger episodes at the zero bound, for a given long run average inflation rate, on account of not being able to stabilise the economy with the 'extra' instrument of future expectations of current interest rates.

If you are still with me, we then get to the literature that says that when central banks can't commit, you can, somewhat paradoxically, get them to deliver the best outcomes according to the original, optimal, growth rates target ['inflation targeting'] by assigning them variants of a levels target [eg a 'price level target'].  So, applied to the current question, we could, having chosen the long run average inflation rate, squeeze a bit more by way of zero bound avoidance by assigning a levels target of sorts to a central bank condemned to act discretionarily.

That said, my own work with Blake and Kirsanova points out that this is not nearly as reliable a result as once was thought.  For those interested:  rational expectations models under discretion typically generate multiple equilibria if there are state variables like debt or capital.  And then it becomes ambiguous what the welfare benefits are to assigning a discretionary central bank a levels target [of whatever variety] since one doesn't know immediately which equilibrium one is jumping from and to.

Analytical drawbacks like this aside, readers should imagine the communication somersaults required to get over to the general public that, in order to get over the fact that policy suffers from [dynamic] lack of commitment, the government, which would ideally like a growth rates based target to be pursued, is actually going to assign a levels-based target to its central bank.

And we have yet to deal with the fact that the benefits presume that agents have forward-looking rational expectations, an assumption that is highly unrealistic.

Taken together, the case for levels based targets - including NGDP levels targets - is, both practically and analytically, extremely weak.  This was why the Bank of Canada rejected moving to a PLT.

So, based on this, I don't see that following a levels target of any variety, NGDP or otherwise, is an alternative to raising the long run average inflation rate, however implemented.  There's no choice [reform to negative rates aside] but to bite the bullet and figure out a different trade-off, between the costs of long run higher inflation, and the costs of higher volatility [imposed by missing stimulus at the zero bound].

In so far as we choose to go for higher average inflation, the choice of whether this should be delivered by a regime that weights equally inflation and real growth [NGDP growth] or  something else, like 'inflation targeting', resolves in favour of the status quo on practical [it's a version of status quo] and theoretical grounds [models point to IT].