Explaining time variation in geographic price dispersion

ABSTRACT

The pattern of price dispersion significantly varies over time and across locations. Using a detailed dataset with product-level retail prices, we examine the role of time-varying factors in shaping the time variation of price dispersion. We find that price dispersion variation in an integrated region is mainly driven by oil prices, while the variation in a segmented region is attributed to dispersion in real income. We also find that dispersion in value-added tax rates explains a significant portion of price dispersion fluctuations in both geographic dimensions. This paper offers new evindence on the trade-off that exists for the role of time-varying factors as contributors to price dispersion variation by highlighting their relative importance across different dimensions of economic geography.

KEYWORDS:

• Price dispersion
• time variation
• oil prices
• income dispersion
• VAT rates dispersion

JEL CLASSIFICATION:

• E31
• F31

Acknowledgments

We would like to thank David Peel, Rachel Pownall, and Giorgio Motta for their valuable comments.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ Price dispersion is defined as the cross-sectional variation in relative prices, often viewed as a measure of price deviations. At the aggregate level, greater price dispersion implies larger deviations from the purchasing power parity (i.e., larger absolute values of the real exchange rates). At the disaggregate level, greater price dispersion implies larger deviations from the law of one price (i.e., larger absolute values of the relative prices of individual goods across locations).

² The data set is described in more detail at http://worldwidecostofliving.com/asp/wcol_HelpWhatIsWCOL.asp.

³ For the OECD and EU groups, we include 28 and 15 member countries, respectively, that became members prior to 2000.

⁴ We follow the traditional method of currency conversion by using the nominal exchange rate to convert prices into US dollar in each period. One may reasonably cast doubt on this method of currency conversion because, given the sharp devaluation or revaluation often experienced by emerging economies, one can observe changes in a country’s prices denominated by the US dollar even without any actual change in prices in the local currency. In light of the LOP deviation, however, this concern can be viewed as suggestive of an imperfect pass-through of the exchange rate changes to prices. Without sticky prices and market segmentation, sharp changes in exchange rates will not induce deviations from the LOP, while the positive pass-through of the exchange rate changes becomes significantly limited in the presence of these frictions. This is why the existing empirical work (Bergin and Glick (2007), Crucini and Shintani (2008), Gopinath et al. (2011), Andrade and Zachariadis (2016), etc.) uses the traditional method of currency conversion, with most of the numeraire currency being the US dollar, when examining how weak the connection actually is between exchange rates and national price. We thank the referee for pointing out this issue.

⁵ The website is: https://databank.worldbank.org/data/source/world-development-indicators for real income per capita, https://www.oecd-ilibrary.org/taxation/consumption-tax-trends_19990979 for VAT rates,

https://stats.oecd.org/Index.aspx?DataSetCode=IOTS for input-output data, and

https://stats.oecd.org/viewhtml.aspx?datasetcode=OIWORLD&lang=en for oil prices.

⁶ We use retail sales taxes in place of VATs for the US.

The website is: https://apps.bea.gov/iTable/iTable.cfm?isuri=1&reqid=70&step=1#isuri=1&reqid=70&step=1 for real income per capita, and https://www.taxadmin.org/state-tax-agencies for sales taxes. Because sales tax data are not available at the level of the individual city, we assign each city to a state and use the state-level tax rate in place of the city-level tax rate.

⁷ The traded input shares are computed by putting together the direct and indirect convolutions of traded input requirements. The sectors considered as traded inputs are food products, wood products, paper products, refined petroleum products, chemicals, rubber and plastic products, non-metallic mineral products, iron and steel, non-ferrous metals, fabricated metal products, office and computing machinery, electrical machinery, communication equipment, precision instruments, and transportation equipment.

⁸ This normalization avoids problems inherent in choosing an arbitrary numeraire location.

⁹ Although not shown here, prices are also less dispersed across EU cities than across OECD cities, reflecting that EU countries are both geographically close and economically integrated.

¹⁰ Real income differences are considered in light of the firm link between price and income levels projected in the context of the Harrod-Balassa-Samuelson hypothesis and the pricing-to-market. Moreover, real per capita income may induce retail price differences through the channel of local costs, such as distribution costs and rents. See Atkeson and Burstein (2008) and Alessandria and Kaboski (2011) for the mechanisms wherein real income differences affect price dispersion.

¹¹ Since we run the regression separately for each country-grouping, we do not consider group-specific effects. In Section 4, we report coefficients on year dummies with time-varying regressors excluded.

¹² Here, we consider services as the base category.

¹³ We follow Bergin and Glick (2007) in constructing the adjusted values.