Today, we are fortunate to have a guest contribution written by Olivier Coibion (UT Austin), Yuriy Gorodnichenko (UC Berkeley), and Gee Hee Hong (Bank of Canada); it is based on The Cyclicality of Sales, Regular and Effective Prices: Business Cycle and Policy Implications.

Each new release of Consumer Price Index (CPI) inflation numbers is met with howls of derision by a number of online commentators. While some of the claims are baseless and already debunked by statistical agencies (e.g. Greenlees and McClelland 2008), others reflect the numerous practical and conceptual difficulties involved in measuring the price level and its changes over time. For example, the “substitution bias”—which reflects the reallocation of expenditures by households across different goods as their relative prices change—has long been emphasized as a potential source of long-term bias in the measurement of prices in the CPI (e.g. Boskin Commission report 1996). The substitution bias could also lead to cyclical mismeasurement of inflation if the properties of price changes vary over the course of the business cycle. For example, a greater frequency of sales in recessions could readily lead to an overestimation of the average prices paid if households switch brands in response to these sales.
More broadly, a key conceptual distinction is that between the prices charged by retailers (what the BLS tracks in constructing the CPI) versus the “effective” prices actually paid by households. The substitution bias is one mechanism which can drive a wedge between the two, but it is not the only one. A second is the reallocation of expenditures by households across retailers. That is, because consumers can switch from expensive stores like Whole Foods to cheap stores like Walmart, the effective price paid by consumers can decline even if consumers buys identical baskets of goods after the switch. Unlike the substitution bias, store-switching can drive a wedge between the two price concepts even at the level of an individual product (e.g. a pack of Saltine crackers).
The distinction between the prices charged by retailers versus the “effective” prices paid by households is important not just for the measurement of inflation but also for better understanding macroeconomic dynamics and the effectiveness of monetary policy actions. For example, the macroeconomics literature has long emphasized that if firms change their prices infrequently (i.e. have “sticky-prices”), then unexpected changes in monetary policy should have pronounced effects on the level of economic activity. But price-stickiness at the level of the firm need not imply that the “effective” prices actually paid by households are themselves sticky. Chevalier and Kashyap (2011), for example, argue that if households respond strongly to sales, then “effective” price flexibility due to consumers reallocating their expenditures across goods or time could potentially undo much of the macroeconomic effects of the underlying price rigidities commonly observed in regular prices. This therefore suggests that characterizing the degree of price flexibility requires more than just measuring the frequency at which prices are changed.

Quantifying the Cyclicality in Prices Paid versus Prices Charged

In a recent working paper (Coibion, Gorodnichenko and Hong 2012), we quantify these concepts using a panel dataset of both prices and quantities sold at the universal product code (UPC) level across different stores in 50 U.S. metropolitan areas from 2001 to 2007. Because our data includes both prices and quantities sold, we can characterize the cyclicality in both the prices posted by retailers as well as the effective prices actually paid by consumers. For example, we can track the average price of a packet of Saltine crackers charged by individual retailers in Atlanta and also measure the average price paid for that same packet of Saltine crackers across the same retailers in Atlanta, where the difference between the two is that in the latter case, the weights applied to different retailers in the averaging process will vary with the expenditure shares associated with each retailer for that specific good (measured at the UPC level). By working at the UPC level, we can therefore separate the store-switching margin from the cross-good substitution margin.
While we find little cyclical sensitivity in the inflation rate of prices posted by retailers, we document that effective price inflation in the prices actually paid by households is significantly more sensitive to business cycle conditions than inflation in posted prices. In other words, when economic conditions in Atlanta worsen, we may see little change in the average price of Saltine crackers at any given retailer, but the average price that households pay for Saltine crackers in Atlanta will tend to fall sharply (or at least grow more slowly). Figure 1 illustrates this at a very aggregated level (across all goods and areas in our data) by plotting the percentage difference between the average price paid by households and the average level of the prices posted by retailers. This differential is strongly countercyclical: the high rates of unemployment during and subsequent to the 2001 recession are associated with significant drops in the prices paid by households relative to the average prices charged by retailers. This pattern reversed itself in the second half of the decade as economic activity improved. Thus, there is indeed significantly more flexibility in the average prices paid by households than standard price indices such as the CPI (which focus only on the prices charged at the level of the retailer) would suggest.

Figure 1: The Countercyclicality of Prices Paid by Households Relative to Posted Prices. Notes: The figure plots the difference between the “effective” price index and the “posted” price index. The latter is a fixed-expenditure-weighted average of all UPC prices in each store and metropolitan areas in the data, where weights are average expenditure share of each UPC in each geographic area relative to total household expenditures. The former is the fixed-expenditure-weighted average of average price paid by households for each UPC across all retailers in a metropolitan area. See Coibion, Gorodnichenko and Hong (2012) for details.

The Sources of Effective Price Flexibility
What drives this greater flexibility in the prices actually paid by households relative to that observed in the prices charged by retailers? One possibility is sales. If households buy more Saltine crackers when they are on sale during downturns, then the average price paid by households would fall even if posted prices did not. In the same spirit, if there were more sales of Saltine crackers during recessions, then again the average price paid by households could fall even if the regular posted prices did not. An alternative explanation for the observed flexibility of prices paid by households is store-switching: if consumers reallocate their expenditures toward lower-price retailers during trying economic times (i.e. buy crackers at cheaper stores), then this could also lower the effective prices paid by households relative to those charged by retailers.
Using the variation in the frequency of sales across different categories of goods, time, and geography, we find robust evidence that sales are pro-cyclical, i.e. a deterioration in local economic conditions reduces both the frequency of sales and the share of goods bought on sale. Furthermore, the (pro-)cyclicality of sales is a phenomenon which occurs primarily in more expensive retailers. This is consistent with expensive retailers using sales primarily as a means of attracting price-sensitive consumers rather than as a more flexible way of adjusting regular prices. As a result, if price sensitive consumers migrate to cheap stores during recessions, expensive store can find it unprofitable to have sales to attract price sensitive consumers and thus they can reduce the size or frequency of sales in recessions. Hence, sales are an unlikely explanation for the observed flexibility in the effective prices paid by households.
In contrast, we document robust evidence that households reallocate their expenditures across retailers in response to variation in economic conditions. Specifically, using a detailed panel of expenditures at the household level, we measure the average price rank at which individual households do their shopping and find that these price ranks decline significantly when local economic conditions deteriorate. Figure 2 illustrates at a more aggregated level that the share of household expenditures going to more expensive retailers fell substantially during the downturn of the early 2000s, consistent with significant household reallocation of expenditures across retailers.

Figure 2: The Aggregate Cyclicality of Store-Switching. Notes: The figure plots the share of total revenues going to “high-price” retailers over time as well as the aggregate unemployment rate. See Coibion, Gorodnichenko and Hong (2012) for details.

Macroeconomic Implications of Store-Switching

 
The key message from our empirical results is therefore that while significant flexibility is indeed present in the prices paid by households relative to those charged by retailers, this flexibility appears to be driven primarily by store-switching on the part of households rather than sales. Because previous work has almost exclusively focused on sales as a potential source of effective price flexibility, we build on this literature by integrating store-switching into an otherwise typical New Keynesian model to assess its macroeconomic consequences.
The first question that we address is whether effective price flexibility stemming from store-switching undoes the effects of the underlying price rigidities in terms of monetary non-neutralities. While store-switching slightly reduces the contemporaneous effect of monetary policy shocks on output, it has little effect on the persistence of the output gap response. As a result, store-switching on the part of households has little effect on the degree of monetary non-neutrality despite yielding a dramatic increase in the flexibility of effective prices paid by households.
Nonetheless, incorporating store-switching on the part of households into the model has several novel business cycle and policy implications. First, time-varying shopping effort and store-switching imply that policymakers should place more weight on limiting fluctuations in output relative to fluctuations in inflation. Second, there will be a cyclical mismeasurement of the price of the final consumption bundle when using standard fixed expenditure-weight price indexes, but a price index with time-varying expenditure weights, as in Figure 1, will capture most of the variation in the ideal consumption price index. Thus, a new price index which tracks the average prices paid by households rather than those charged by retailers would go a long way in addressing the cyclical mismeasurement of inflation arising from store-switching. Third, under price-level targeting regimes, significant welfare improvements are achievable if policymakers target the “effective” price index paid by households rather than a fixed expenditure-weight price index. Hence, the construction of an “effective” price index by statistical agencies could ultimately lead to significant improvements in economic welfare.
Finally, our empirical results could potentially shed new light on the puzzle of the missing disinflation during the 2007-2009 downturn. As documented in Ball and Mazunder (2011), the magnitude of the output gap in the recent recession would have been expected to lead to significantly more disinflation than actually occurred, based on historical Phillips curve correlations. While some have advocated downward wage rigidity as a potential explanation, such a result could also follow from cyclical changes in price rigidities. For example, if prices become increasingly rigid due to the procyclicality of sales as economic conditions worsen, as suggested by our results, then this could lead to a flattening of the Phillips curve at lower levels of economic activity. Estimates of the slope of the Phillips curve during regular times would therefore lead to an over-prediction of the decline in inflation during periods of economic crisis. We leave this tantalizing possibility for future work.

References:

• Ball, Lawrence and Sandeep Mazumder, 2011. “Inflation Dynamics and the Great Recession,” Brookings Papers on Economic Activity (Spring), 337-381.
• Boskin, Michael J., E. Dulberger, R. Gordon, Z. Griliches, and D. Jorgenson, 1996. “Toward a More Accurate Measure of the Cost of Living,” Final Report to the Senate Finance Committee, Dec. 4.
• Chevalier, Judith A., and Anil K Kashyap, 2011. “Best Prices,” NBER Working Paper 16680.
• Coibion, Olivier, Yuriy Gorodnichenko and Gee Hee Hong, 2012, “The Cyclicality of Sales, Regular and Effective Prices: Business Cycle and Policy Implications,” NBER WP 18273.
• Greenlees, John S. and Robert McClelland, 2008. “Addressing misconceptions about the Consumer Price Index,” Monthly Labor Review, August, 1-17.

This post is written by Olivier Coibion, Yuriy Gorodnichenko, and Gee Hee Hong.