New Keynesian models rely upon price and wage rigidity to generate movements in macroeconomic variables that match, approximately, movements in actual variables over time. In order to match the movements in actual data, a particular degree of price/wage rigidity must be assumed. Thus, it is important that estimates of the actual degree of price and wage rigidity match the degree of rigidity imposed upon these models.
What does the evidence say about prices? When econometricians look at aggregated prices, they tend to find a great deal of stickiness, enough to justify the degree assumed in the models. However, when disaggregated prices are examined, prices appear to be much more flexible. For example, this is from "Reference Prices and Nominal Rigidities," by Martin Eichenbaum, Nir Jaimovich, and Sergio Rebelo, March 2008:
[T]he recent literature ... uses micro data sets to measure the frequency of price changes. The seminal article by Bils and Klenow (2004) argues that prices are quite flexible. Using monthly CPI data, they find that median duration of prices is 4.3 months. This estimate has became a litmus test for the plausibility of monetary models. In contrast, Nakamura and Steinsson (2007) focus on non-sale prices and argue that these prices are quite inertial. When sales are excluded, prices change on average every 8 to 11 months.
Kehoe and Midrigan (2007) also examine the impact of sales on price inertia. They use an algorithm to define sales prices that they apply to weekly supermarket scanner data. They find that, when sales observations are excluded, prices change once every 4.5 months. When sales are included, prices change every 3 weeks.
So there is quite a bit of variation in the estimates of stickiness derived from disaggregated data, and even more variation when more aggregate prices are examined. As the paper notes, whether or not sales are included in the price data is a key issue for the results from disaggregated data:
Excluding 'sales prices' from the data has a major impact on inference about price inertia. Not surprisingly, there is an ongoing debate in the literature about how to define a sale and whether one should treat 'regular' and 'sales' prices asymmetrically.
The paper makes a contribution to this debate by using the concept of "reference prices":
An advantage of working with 'reference prices' is that we do not need to take a stand on what sales are or whether they are special events that should be disregarded by macroeconomists.
But that is not the only reason to use this concept. They find that although disaggregated prices are quite flexible, they are flexible around an inflexible reference point, and that inflexibility can produce substantial nominal rigidities:
We argue that our evidence is inconsistent with the three most widely used pricing models in macroeconomics: flexible price models, standard menu cost models, and Calvo-style pricing models. There is, however, a simple pricing rule that is consistent with our evidence. This rule can be described as follows. Prices do not generally change unless costs change. For any given good the nominal reference price is on average a particular markup over nominal reference cost. The retailer sets the frequency with which they reset the reference price so as to keep the actual markup within plus/minus twenty percent of the desired markup over reference cost. ... With this rule reference prices can exhibit substantial nominal rigidities even though weekly prices change frequently.
Maybe pictures will help - the variability of the prices and the inflexibility of the reference points are both evident. This is two of the 60,000 prices they examine:
And, finally, the conclusion:
We present evidence that is consistent with the view that nominal rigidities are important. However, these rigidities do not take the form of sticky prices, i.e. prices that remain constant over time. Instead, nominal rigidities take the form of inertia in reference prices and costs. Weekly prices and costs fluctuate around reference values which tend to remain constant over extended periods of time.
Reference prices are particularly inertial and have an average duration of roughly one year. So, nominal rigidities are present in our data, even though prices and cost change very frequently, roughly once every two weeks. We document the relation between prices and costs and argue that our findings pose a challenge to the most commonly pricing models used in macroeconomics.
The challenge is not to the idea that wage and price rigidity cause unintended price dispersion and resource misallocations that bring about short-run variation in the macroeconomy present in the New Keynesian framework. The challenge is to the particular form of price and wage rigidity that has been imposed upon New Keynesian models (e.g. the Calvo rule). The pricing rules have always been a weak part of these models, they are often imposed rather than derived from first principles, so perhaps this will motivate progress on this front.
[One final note. This is part of the debate described in the post on the natural rate of unemployment. If prices are sticky, then shocks can move the economy away from its long-run optimal (equilibrium) path, and this opens up the possibility for monetary or fiscal policy to intervene and make people better off by redirecting the economy back to its long-run optimum. But if prices are perfectly flexible (or close enough, and there are no other rigidities, including information problems), then all movements are changes in the equilibrium path for the economy rather than deviations from it. In this case there is nothing for policymakers to do since any change in output would take the economy away from its optimal level making people worse rather than better off. However, the results in this paper are consistent with the idea that policy intervention can provide positive benefits.]