|
Article Excerpt
Summary. When economic agents have diverse private information on the fundamentals of the economy, prices may serve as a poor aggregator of this private information. We examine the information value of prices in a monopolistic competition setting that has become standard in the New Keynesian macroeconomics literature. We show that public information has a disproportionate effect on agents' decisions, crowds out private information, and thereby has the potential to degrade the information value of prices. This effect is strongest in an economy with keen price competition. Monetary policy must rely on less informative signals of the underlying cost conditions.
Keywords and Phrases: Differential information, Price inertia, Common knowledge.
JEL Classification Numbers: E31, E32, E58.
1 Introduction
One of the often-cited virtues of a decentralized economy is the ability of the market mechanism to aggregate the private information of the individual economic agents. Each economic agent has a window on the world. This window is a partial vantage point for the underlying state of the economy. But when all the individual perspectives are brought together, one can gain a much fuller picture of the economy. If the pooling of information is effective, and economic agents have precise information concerning their respective sectors or geographical regions, the picture that emerges for the whole economy would be a very detailed one. When can policy makers rely on the effective pooling of information from individual decisions?
This question is a very pertinent one for the conduct of monetary policy. Central banks that attempt to regulate aggregate demand by adjusting interest rates rely on timely and accurate generation of information on any potential imbalances in the economy. The role of the central bank in this context is of a vigilant observer of events to detect any nascent signs of growing imbalances. In particular, most central banks focus on the development of inflationary pressures. Signs of such pressures can be met by prompt central bank action through the use of monetary policy instruments.
It is possible to make a case that the information value of prices has improved in recent years. In his Jackson Hole paper, Rogoff (2003) notes that in very competitive sectors, like agriculture or semi-conductors, prices are significantly more flexible than in less competitive or highly regulated sectors. To the extent that globalisation and deregulation has increased competition across a wide range of industries, one could argue that firms are operating in a more competitive environment, and hence their prices will behave more like the prices in competitive sectors.
It is not our purpose here to assess the empirical claim that industries have, in fact, become more competitive. (3) Rather, we will assess the conditional statement which claims that if imperfectly competitive economies become more competitive, prices will become more responsive to changes in the fundamentals. Although it is undeniable that very competitive sectors such as agriculture and semi-conductors have very responsive prices, the question is whether the relationship between competition and price responsiveness is a continuous one.
There is some reason for doubting that the relationship between competition and price responsiveness is always continuous. Ball and Romer (1990) show that when firms worry about their market share, they will be reluctant to be first to raise prices (prices exhibit "real rigidities") so that even a small menu cost may induce firms not to change their prices. As the degree of competition increases, the real rigidities will become more severe, and this effect may become more potent.
We will show in this paper that even in the absence of menu costs and other nominal rigidities, greater competition may actually reduce the responsiveness of prices to changes in the underlying fundamentals. Our case is built on the importance of distributed information in decentralised economies where firms have access to information that is local to their region or industry, as well as to publicly available information that is available to firms economy-wide. Firms have their own "window on the world". In such a setting, when firms try to defend their market share, this will entail some degree of second-guessing the pricing strategies of their competitors. Even when there are no nominal rigidities, the outcome of navigating through the higher-order beliefs entailed by the second-guessing of others leads firms to set prices that are far less sensitive to firms' best estimates of the underlying marginal costs.
Our conclusion is that the relationship between competition and price responsiveness is highly discontinuous. As an imperfectly competitive economy becomes more competitive, the price responsiveness falls. In the limit, as markup falls to zero, prices become completely unresponsive to the fundamentals. Thus, it is critical to distinguish between a perfectly competitive economy and the competitive limit of an imperfectly competitive economy.
The notion that equilibrium outcomes, particularly prices, are affected by imperfect common knowledge is not new. The macroeconomics literature on the forecasting the forecasts of others begun by Townsend (1983), Phelps (1983) and Sargent (1991) has examined the quantitative impact of "symmetrically uninformed" agents. The issue has recently been revisited by Woodford (2003a) and others (4) in the context of an imperfectly competitive economy. The conclusions drawn from this literature to date have largely relied on numerical simulations of fully-fledged macroeconomic models modified to incorporate private information. The value of such exercises lies in their ability to inform debates on the numerical time series properties of macroeconomic aggregates. For instance, Woodford (2003a) has shown how the combination of strategic behaviour and private information can induce greater persistence in macroeconomic variables in response to shocks relative to a common information benchmark.
However, the cost of complexity is that it is difficult to isolate the key forces at work. One of our tasks in this paper is to attempt to fill in this gap by presenting a theoretical framework that is simple enough to unpack the precise mechanism at work in the degradation of the information value of prices. For this purpose, we will concentrate on two examples--a simple Gaussian dynamic model, and a static model where we abstract from any intertemporal learning or allocation problems. The virtue of such simple examples is that we can employ arguments that rely on well-known results from elementary probability theory.
We show that when price competition between firms becomes more intense, the aggregate price level becomes extremely unresponsive to the underlying fundamentals. In a dynamic context, prices exhibit a great deal of inertia. Even though firms are rational and form prices based on forward-looking expectations, actual behaviour of prices have the outward signs of adaptive expectations. One of the enduring puzzles that macroeconomists have struggled with is how to explain the the apparent inertia in inflation without resorting to adaptive expectations (see Gali and Gertler, 1999). Our examples suggest that models of distributed information may be a promising line to pursue in tackling this problem.
There are potentially troubling implications of our results for monetary policy. The experience of monetary policy in the 1990s has posed challenges for the view that the central bank can rely on the rate of inflation to guide monetary policy. Even as overall inflation pressures eased during the latter half of the 1990s, economies expanded rapidly under conditions of strong demand, accompanied by surging asset prices. The subsequent downturn in economic activity in the major industrial economies, and especially the United States, has fuelled debates about the information value of goods price inflation as an indicator of overall economic imbalances. (5)
We begin in the next section with a simple Gaussian dynamic price setting model, and then follow up with a general finite state setting in a static context that does not entail any distributional restrictions. We then present a small numerical example. We conclude by discussing the implications of our results for the conduct of monetary policy.
2 Price inertia
We will be concerned with the pricing rule for firms of the form:
[q.sub.i] = [E.sub.i]q + [xi][E.sub.i][chi](1)
where [q.sub.i] is the (log) price set by firm i, q is the average price across firms, [chi] is marginal cost (in real terms)--our "fundamental variable"--and [xi] is a constant between and 1. The operator [E.sub.i] denotes the conditional expectation with respect to firm i's information set. Pricing rules of this form have been discussed for some time. Phelps (1983) derived a similar pricing rule in a competitive economy, and compared it to the 'beauty contest' game discussed in Keynes's General Theory (1936), in which the optimal action involves second-guessing the choices of other players. Townsend (1978, 1983) also discussed similar pricing rules. (6). However, our discussion in this paper has most in common with Woodford (2003a), who has revived interest in pricing rules of this form by showing how they arise naturally in macro models with differentiated goods and imperfectly competitive markets. The parameter [xi] is related to the elasticity of substitution between goods, and becomes small as the economy becomes more competitive. Appendix A presents an illustrative derivation in a partial equilibrium setting.
Rewrite (1) in terms of nominal marginal cost, defined as z [equivalent to] [chi] + q, yielding [q.sub.i] = (1 - [xi])[E.sub.i]q + [xi][E.sub.i]z. Taking the average across firms,
q = (1 - [xi])[bar.E]q + [xi][bar.E]z (2)
where [bar.E](dot) is the "average expectations operator", defined as [bar.E](dot) [equivalent to] [integral] [E.sub.i](dot)di.
Hence,
q = [[infinity].summation over (k=1)][xi] (1 - [xi])[.sup.k-1] [bar.E.sup.k]z (3)
where [bar.E.sup.k] is the k-fold iterated average expectations operator. With differential information, the k-fold iterated average expectations do not collapse to the single average expectation. (7)
Let us now embed the price setting decision in a dynamic context. Time is indexed by t [member of] {1, 2,...}, and suppose that z follows an AR(1) Gaussian process {[z.sub.t]} where
[z.sub.t] = a + [phi][z.sub.t-1] + [eta]t
[eta]t is Gaussian noise, and < [phi] < 1. The unconditional expectation of [z.sub.t] is [mu] = a/(1 - [phi]). There is a continuum of firms, and none of them ever observe the true value of the fundamentals [z.sub.t]. Instead, at date t, firm i observes the...
|
