Ebook Search, Costly Price Adjustment and the Frequency of Price Changes – Theory and Evidence
The goal of the paper is to provide a better understanding of nominal rigidities by analyzing micro-level pricing decisions. We establish a new empirical finding that the intensity of consumer search for the best price affects the frequency of price adjustment: price changes are more frequent and smaller in markets in which search is more intense. The relationship we document is both statistically and economically significant. We analyze an equilibrium model that explains the relationship. In our model, which is related to Bénabou (1992), customers are heterogeneous and search for the best price. We define the propensity to search as the expected return to search for a given relative price dispersion in the market. Competing firms face costs to adjust nominal prices. We show that equilibrium pricing strategies are affected by market characteristics related to consumer search and that the model predicts the patterns observed in the data.
The empirical relationship we discover holds in two very different data sets. The first data set consists of store-level, actual transactions prices for 55 products and services in Poland, each observed monthly in up to 47 stores, over 1992-96. Following the classic Stigler (1961) paper, we proxy the propensity to search with the value of purchases, the good’s importance in household expenditure (conditional on the household buying the good) and the frequency of purchases. The second data set (Bils and Klenow (2004)), consists of prices collected by the Bureau of Labor Statistics in 1995-7; these prices cover 70% of US CPI. The data are grouped into 350 ELIs (Entry Level Items); for each ELI we have the average monthly probability of price changes in a given month. We proxy the propensity to search for goods grouped in a given ELI with its weight in CPI expenditure. Despite the very different environments (for example, the average CPI inflation is about 30% in Poland and is below 3% in the US), we find strong support for the predictions of the model in both data sets.
We consider several alternative explanations of the observed correlations: Taylor-Calvo’s time-contingent model, Kashyap’s (1995) price-contingent model, Diamond’s (1993) sticker price model as well as temporary sales and argue that they cannot explain the patterns in the data. We also discuss Rotemberg’s (2002) customer resistance model. This model has the potential to fit the data under two additional assumptions: that consumer resistance leads to smaller and more frequent price changes and that it is stronger in markets we identify as characterized by higher propensity to search. If so, more research on the effects of aggregate variables is needed to differentiate the two models.
Our search-based explanation of the cross-sectional heterogeneity is consistent with large, and remarkably consistent across countries, differences in the frequency of price changes across broad good categories. In both the Polish and the US data we find that the price changes are least frequent for services, followed by manufactured products and durable foodstuffs and are the most frequent for perishable foodstuffs. The same differences hold in recently obtained, extensive data sets for several European countries. Notably, differences across goods greatly exceed differences across countries. Furthermore, Bils and Klenow (2004) report that price changes are less frequent for processed than for raw goods. We attribute these differences to heterogeneity within broad good categories. The propensity to search for the best price is likely to be affected by non-price differences between goods (for example quality differences across different sellers). The larger are non-price differences, the smaller is the propensity to search for the best price (for given price differences) and so, according to our model, the less frequent are price adjustments. On the average, services are the most heterogeneous, followed by manufactured goods, durable foodstuffs and perishable foodstuffs. Similarly, processed goods are more heterogeneous than raw products.
We begin by presenting the model in the next section. Empirical evidence is in Section 3. In Section 4 we discuss alternative explanations of the patterns in the data. In Section 5 we provide a simple extension of the model which can account for the large observed differences in the average frequency of price changes between broad good categories. The last section concludes.
Download
PDF Ebook Search, Costly Price Adjustment and the Frequency of Price Changes – Theory and Evidence
Posted in :
