At his blog, David Andolfatto has raised several questions for those who advocate nominal GDP targeting. His most recent post is an attempt to look at nominal GDP targeting in an overlapping generations model. I would like to use this post to address David’s model as well as nominal GDP more generally.
I should start by saying that the thing that I enjoy about reading his blog posts is that they are both insightful and inquisitive. There is a subset of economists (unfortunately only a subset of economists) who want to use the tools of economics to address certain questions and see where those tools get them. One of the biggest misunderstandings about the mathematics of economic models is that the math is designed to feign sophistication or give the illusion of science. (Certainly there are some who like to work with (and perhaps show off) highly complex models, but this is not the point of the models and nor do I believe that this path leads to much enlightenment.) In reality, math is used simply to keep our logic consistent. The models allow us to make some assumptions and generate logically consistent implications based on our framework and assumptions. David’s scholarly work and blog posts are a great example of how to properly use these tools. Curiosity is what makes David interesting to read. He likes to make certain assumptions, write down a model, and see what conclusions the model provides.
In reading David’s posts, he seems to have two main questions for advocates of nominal GDP targeting: (1) Why should we care about nominal GDP? (2) Why should we target the level of nominal GDP?
I would like to answer these questions in a roundabout fashion. First, I will argue that I believe there are two main reasons for advocating nominal GDP targeting. Second, I will argue that to the extent that level targeting is desirable, it is because of rigidities (which David and I both have our doubts about).
I would like to start by discussing what a central bank can control and cannot control. I will argue that what the central bank can control one of two things: the price level or nominal income. I will then discuss why the central bank might choose one over the other.
In virtually any model that one writes down in which money plays a meaningful role, the price level and nominal income are determined by the money supply. Thus, if the money supply is defined as the monetary base or if the central bank can use the monetary base to control a broader definition of money, it is possible for the monetary authority to target the price level (inflation, if you prefer) or nominal income.
Given this basic premise, I would like to make two arguments as to why one might prefer a nominal GDP target to the price level. The first argument relies on the idea that economists and policymakers have imperfect knowledge of short-run fluctuations in the economy. I simultaneously like and dislike this argument. The second argument is that nominal GDP is not particularly important in and of itself, but that fluctuations in nominal GDP are indicative of a broader issue that monetary policy can potentially resolve. Finally, the argument frequently made by Scott Sumner and others is that a nominal GDP target is important because of sticky or rigid wages. Others, such as David Beckworth emphasize the role of debt and contracts. Thus, to the extent that one would want to target the level of prices or nominal income, this conclusion seems to come from nominal or real rigidities.
A simple way to characterize the difference between price level and nominal GDP targeting concerns is to appeal to money non-neutrality. This argument is largely one that relies on the fact that we have an imperfect understanding of short-run economic fluctuations. This argument is somewhat of a paradox. In some ways, this is perhaps the weakest argument for targeting nominal GDP. However, in some ways, it is perhaps one of the best because it does not assume that we know more than we actually do. The argument is as follows.
If money is neutral, it would seem that there is no reason to be concerned with nominal GDP. The price level is all that matters. (Even this is not necessarily true, however. One might want the price level to reflect changes in productivity. In which case, a nominal GDP target is likely to be more effective than a price level target. For a discussion of this issue more broadly, see George Selgin’s Less Than Zero.)
If money isn’t neutral, then monetary factors will have an effect on both the price level and real economic output. If this is case, targeting the price level or inflation requires some corresponding theory about how the effects of the monetary shocks are divided between real factors and the price level. Macroeconomists do not have a widely accepted theory of such a division. The New Keynesian view has two equations, the IS equation and the NK Phillips curve. I don’t find this view satisfying as I don’t view the Phillips curve as structural (see this, for as amusing example why). It certainly doesn’t help that this “structural relationship” is derived by the use of an assumption akin to the can opener.
A nominal income target, by contrast, does not require are particular theory or assumption about the effects of monetary shocks are divided between real output and the price level in the short run. This argument was first made by Bennett McCallum in the 1980s and the premise of imperfect knowledge remains. The strength of this argument is its simplicity. If monetary factors have real effects and if there is uncertainty about how these effects are transmitted, then nominal GDP targeting is preferable to targeting the price level or inflation.
The weakness of this argument is that it is unlikely to convince those who are skeptical of nominal GDP targeting — especially those armed with a particular model. (For example, frequent criticism of nominal GDP targeting by those of the New Keynesian variety is that IF monetary factors effect only with a lag, THEN nominal GDP targeting can be destabilizing. But what reason do we have to believe this is true? Apparently, just models — non-microfounded models to boot — that assume this is true.)
The second argument is one that I know David is familiar with as it is something that I have been working on. According to this view, there is nothing special about nominal GDP. However, the behavior of nominal GDP is indicative of something else — fluctuations in the liquidity of transaction assets. Suppose that there are a wide variety of transaction assets (e.g. currency, checking accounts, repurchase agreements, T-bills). We can generate an endogenous mechanism through which the relative liquidity of these assets fluctuates or, for simplicity, we can assume that the liquidity of each asset is exogenous. If we take this latter assumption, adverse liquidity shocks leave the holders of these assets liquidity constrained. This constraint reduces both real and nominal spending.
Why do we care about these liquidity shocks in reference to nominal GDP? We care for two reasons. First, it is possible that the central bank can relax this liquidity constraint by increasing the supply of perfectly liquid base money. Second, the liquidity shock is empirically observable through fluctuations in nominal GDP. It is possible to write down a model in which monetary policy and liquidity are the primary sources of fluctuations in nominal GDP. As such, a decline in nominal GDP signals that there is a shortage of transaction assets and monetary policy can be used to relax the liquidity constraint. By targeting nominal GDP, this type of policy has the potential to mitigate some of the effects of liquidity shocks.
Finally, it is important to consider the implications of targeting the level rather than the growth rate of nominal GDP or the price level. Like David, I am skeptical of sticky price/wage stories, but nonetheless this is one argument that is advocated by some supporters of nominal GDP targeting and there are explicit theoretical models.
Suppose that we start with the basic neoclassical growth model with a constant rate of productivity growth augmented with labor market search. Firms use capital and labor to produce economic output. In addition, firms use a small subset of their labor force to recruit new workers. Wages are rigid. In particular, it is assumed that when firms and workers are matched, the wage is given as
where is the real wage, and is labor productivity. This is the basic setup of the model in Robert Shimer’s new paper.
Now suppose that there is a transitory 1% reduction in the capital stock relative to the steady state. It follows from the model that investment, consumption, real GDP, and employment decline (see Shimer’s paper for details). After the initial shock investment, consumption, and real GDP begin to recover. However, employment does not. The reason that employment doesn’t recover is because of the assumption about the behavior of the real wage. Firms are willing to offer the higher real wage consistent with productivity growth despite the fact that the real wage is “too high”, but they reallocate their workers away from recruiting and towards production. As a result, less unemployed workers are matched with firms. A jobless recovery results.
This model has no analysis of monetary policy. However, the implications are consistent with those made by Scott Sumner and others. It is therefore possible — although, again, the model does not discuss monetary policy — that by targeting the price level, monetary policy could return real wages to the previous trend level. An interesting exercise would be to add money into this labor search model a la Berentsen, Menzio, and Wright and determine what the optimal monetary policy might look like and whether a price level would be consistent with optimal policy.
One can make similar arguments for nominal income level targeting with regards to debt contracts. David has discussed the role of nominal contracts extensively with David Beckworth on his blog and so I will simply refer readers to that conversation here.
I suspect this post will do little to change David’s mind. In fact, I don’t know how useful this post actually is in that these topics probably require more thought than blog posts allow or at least are able to convey. However, it is my hope that it will add something useful to the conversation — contrary to what has happened in the comment section of his recent blog post.