My favorite quote: "These patterns suggest that retail prices vary largely as a consequence of dynamic pricing strategies on the part of retailers or manufacturers, rather than current demand and supply conditions."
Well, duh. Did Emi talk to anyone in the consumer packaged goods industry, or maybe a retail buyer? She's looking at short term changes (one year, 2004) for an item with huge gross margins (carbonated soft drinks). Over a long time period, and for products more closely tied to commodity prices, one would see more evidence of demand and supply conditions.
Here is the summary:
What Drives Retail Price Movements?
" Only 16 percent of price variation is common across all stores selling an identical product, while 65 percent of price variation is common to stores within a particular retail chain (but not across chains), and 17 percent of variation is completely idiosyncratic to the individual store and product."
Recent research shows that retail prices vary enormously over short time periods -- within a year, the price of a typical grocery product can vary by 20-30 percent. These movements are large relative to the factors that economists usually think of as driving prices, such as wages and firm productivity.
In Pass-Through in Retail and Wholesale (NBER Working Paper No. 13965), NBER Faculty Research Fellow Emi Nakamura studies how prices co-move across products, firms, and locations to understand what drives these large price movements. Nakamura finds that only 16 percent of price variation is common across all stores selling an identical product, while 65 percent of price variation is common to stores within a particular retail chain (but not across chains), and 17 percent of variation is completely idiosyncratic to the individual store and product. In other words, when the price drops on Diet Coke at the 125th street Pathmark store, chances are that the price will not drop at the Fairway market down the street.
For her analysis, Nakamura uses a new dataset from A.C. Nielson that consists of weekly price and quantity series throughout 2004 for about 7,000 grocery stores across the United States. These stores encompass 33 chains in 50 major cities, for 100 bar-coded (Universal Product Code) items, totaling some 50 million observations.
Price movements such as those she observes do not arise from manufacturer demand or supply shocks, Nakamura argues, because those factors presumably would lead to common price movements across all stores. Furthermore, she finds that only a small fraction (19 percent) of price variation is common to all products in a category at a given retail store. In other words, when the price of Diet Coke drops at the Pathmark, chances are that price of the Pepsi Max down the aisle will not change. Therefore, Nakamura argues, retail demand and supply shocks are not the likely "drivers" of the observed price movements: if most price movements for Diet Coke arise from changes in supermarket costs or demand, then the resulting price movements should be common across different soft drinks -- and this isn't the case in the data.
These patterns suggest that retail prices vary largely as a consequence of dynamic pricing strategies on the part of retailers or manufacturers, rather than current demand and supply conditions. Not surprisingly, temporary sales play an important role in this price variation, accounting for a large fraction of the observed price movements. Furthermore, products with many temporary sales, such as soft drinks, also have a disproportionately large fraction of price variability that is entirely idiosyncratic to both the store (for example, Pathmark) and product (for example, Diet Coke).
The large amount of idiosyncratic movement in prices implies that average price series for a particular UPC -that is, averaged across retailers -- behave very differently from the underlying store-specific prices. Following an individual product over time week by week, Nakamura finds that prices are highly variable, with frequent price movements on the order of 10-15 percent, and also highly transitory, so that a low price this week says little about whether prices will be low the next week. But, taking the same data for a given UPC and averaging over stores and within months leads to a price series that is far smoother --the variability of prices drops by two-thirds -- and far more persistent. While individual price series seem dizzyingly complex, the lower volatility and greater persistence of the nationwide averages seems to leave greater scope for a link to standard economic factors such as wages, productivity, and exchange rates.
-- Matt Nesvisky