Lost in Transit: Product Replacement Bias and Pricing to Market

Abstract

The microdata underlying U.S. import and export price indexes exhibit frequent product turnover and highly rigid prices. As a consequence, 40% of products are replaced before a single price change is observed and 70% are replaced after two price changes or less. An aggregate price index that focuses on price changes for identical items over time may, therefore, miss an important component of price adjustment occuring at the time of product replacements. We provide a model of this "product replacement bias" and quantity its importance using U.S. microdata on improt and export prices. We show that, accounting for product replacement bias, long-run exchange rate "pass-through" into U.S. import and export price indexes is almost twice as high as conventional estimates suggest, and changes in the terms of trade are roughly 75% more volatile. Our adjustment makes pass-through statistics easier to account for with existing general equilibrium models. (including the U.S. import and export price indexes) are close approximations of a "matchedmodel index," in which all price changes used to construct the index are for identical items and product replacements are "linked-into" the index; meaning that the price comparison between the first observation of the new product and the last observation of the old product is dropped when changes in the index are calculated. 2 A matched model index will remain constant throughout in our example since prices only change at the time of product replacements and price comparisons between old and new products are dropped when the price index is constructed. Estimates of exchange rate pass-through using this price index will yield zero pass-through irrespective of what the true degree of pass-through is. 3While this is obviously an extreme example, it captures important features of the actual data underlying the U.S. import and export price indexes. Price rigidity and frequent product turnover imply that about 40% of expenditure weighted price series in these data have no price changes and roughly 70% have two price changes or less. 4 Even products that do have price changes while they are in the index, typically exit the index after a prolonged spell of price rigidity. If the prices of new products entering the index have already adjusted to exchange rate movements over this interval (as in our simple example above), the response of these prices to movements in exchange rates over this interval will be "lost in transit" (i.e., neither picked up by an observed price change of the exiting nor entering products). In this case, the price index will never fully reflect the true comovement of prices and exchange rates, even in the long run.We develop a model of this "product replacement bias" and show how it depends on observable features of price data. We estimate our model using BLS micro-data on import and export prices.Our "corrected" measure implies that pass-through of changes in the trade-weighted U.S. exchange rate into U.S. import prices for the period 1982-2007 was 0.64-almost twice as...

Exchange Rate Pass‐through Into Retail Prices

“…Long‐run pass‐through is about 25% when a balanced panel of products is used. The estimates thus do not differ greatly from the baseline case (as a comparison to Figure (a) makes clear), although the fact that the latter is lower is consistent with the analysis in Nakamura and Steinsson ().…”

“…To assess the extent to which product replacement can significantly bias exchange rate pass‐through estimates (see Nakamura and Steinsson, ), we additionally estimate specification (1) using a balanced panel of products, that is, only the products that were available at a store in all five years . The sample size is substantially reduced, as the number of product–outlet combinations drops from 138,137 to 17,275.…”

A multi-country DSGE model with incomplete exchange rate pass-through: An application for the Euro-area

“…Rather, the main motivations of retaining a normative approach in this paper are twofold. First, recent studies using data at firm level have prompted renewed interest in the role of nominal rigidity as an important factor that influences the extent of the ERPT (Gopinath and Rigobon, 2008;Gopinath et al, 2010;Gopinath and Itskhoki, 2011;Nakamura and Steinsson, 2012;Burstein and Gopinath, 2013.). In particular, the Inflation Persistence Network (IPN) team at the European Central Bank conducts numerous studies that focus on the issue of nominal rigidity and its implications for the Euro-area.…”

Is there a quality bias in the Canadian CPI? Evidence from microdata

“… Additional difficulties stem from estimating demand or supply shifts associated with price change during model changeovers. For example, consumers may spend time searching for better deals in other stores (Kryvtsov and Vincent ) or stores may synchronize large price increases with occurrences of product substitutions (Nakamura and Steinsson ). The International Labour Office () discusses challenges for quality measurement related to sampling and collection issues. …”

“… Nakamura and Steinsson () show that product replacement bias lowers the long‐run exchange rate pass‐through in the US import and export price data. …”