(Mis)measuring prices in an era of globalisation

David Byrne, Brian Kovak, Ryan Michaels 03 April 2014

a

A

China’s integration into world markets over the past 35 years has been dramatic both in pace and scope. An important aspect of this growth has been China’s expansion into relatively sophisticated product categories (Rodrik 2006, Schott 2008). In such cases one wonders whether the lower prices offered by Chinese firms reflect true discounts, or rather lower product quality (Schott 2008).

This question of how to identify quality differences in international markets is the latest manifestation of a long-standing challenge concerning how to interpret price dispersion across suppliers. The U.S. Bureau of Labor Statistics International Price Program generally resolves this by assuming that a low price charged by a new supplier must compensate for lower quality of the item or an inferior overall “shopping experience”.1  Otherwise, according to this reasoning, higher-priced competing suppliers could not sustain market share. Therefore, price reductions associated with entry are typically not incorporated into the price index.2

Although quality differences surely exist, price measures should also account for real price dispersion that can arise if arbitrage is impeded. To this end, our research illustrates a new way to identify real price dispersion when there are differences across suppliers in both product attributes and the quality of service (Byrne, Kovak, and Michaels 2013). We find that real price dispersion is substantial in the market we study. Critically, failing to account for its presence can skew official measures of inflation: omitting price declines due to the entry of lower-priced suppliers results in an upward biased price index. In turn, growth in real imported inputs will be understated and as a result, the growth of productivity – how much is made per unit input – will be overstated (Houseman, Kurz, Lengermann, and Mandel 2011).

Contract semiconductor wafer manufacturing

We focus on the market for semiconductor wafers, which contain the integrated circuits used in modern electronics. This market has played an outsized role in aggregate labor productivity growth (Byrne, Oliner, and Sichel 2013). Our data cover arms-length transactions, in which contract manufacturers of semiconductor wafers implement the designs of “factoryless” firms specialising in design and marketing.3  This pattern of vertical specialisation has grown substantially in the semiconductor industry, paralleling its growth throughout manufacturing (Bayard, Byrne, and Smith 2013, Bernard and Fort 2013). 

Products in this market have a few observable attributes that influence their prices. The most important are the wafer’s size (measured in milometers, mm) and the ‘line width’ – the size of each transistor (measured in nanometers, nm). Together, these determine the quantity of transistors that can be implanted on a wafer. Semiconductor technology proceeds discretely, with a new smaller line width arriving roughly every two years. Figure 1 illustrates this product cycle, showing the share of total sales for each line width at the largest contract semiconductor wafer manufacturer, Taiwan Semiconductor Manufacturing Corporation.

Figure 1 Technology cycle – TSMC sales by line width

Sources of price dispersion

Since our data report the price and physical attributes of semiconductor wafers in each transaction, we can directly account for the contribution of several product characteristics to price dispersion. Focusing on the two largest locations of production, we find that Chinese products are on average 26 percent cheaper than those produced in Taiwan, the market leader. However, when controlling for differences in product characteristics such as wafer size and line width, the average price difference falls to 17 percent. Thus, one third of the average price difference is accounted for by differences in observable product characteristics.

But there are other quality-related differences across suppliers that are harder to observe. These include more subtle product attributes as well as the overall quality of service. The latter encompasses features such as product reliability, timeliness of delivery, or the quality of customer service, which should be reflected in equilibrium price differences (Carlton 1983). The structure of the wafer market allows us to infer the role of unobserved quality based on price differences across suppliers late in the life of a product technology.

Our approach relies on the fact that buyers in this market face large costs of switching suppliers, reflecting investments and calibrations that are required for each new design and cannot be transferred across suppliers. These costs represent a substantial trading friction, such that a buyer will only switch suppliers if the price difference outweighs the switching cost. In our application, this gives Taiwanese firms an advantage, since they enter the market for each new technology (a wafer size, line width pair) at least two years before Chinese producers. This means Taiwanese producers can initially charge a high price to exploit their customer base, which is partially locked in by the switching cost. Then, as their original customers exit and Taiwanese market power erodes, they lower their prices. As a result, the influence of the switching cost on price dispersion abates, and late in the product cycle, price dispersion reflects only differences in unobserved quality across suppliers. We confirm this intuition using a dynamic pricing game with switching costs (see Klemperer 1995 for a survey).

The price data reflect this pattern, as illustrated in Figure 2. The solid line plots the price difference between China and Taiwan for observationally identical products in each quarter following China’s entry into the market for the relevant technology.4  China’s price upon entry is roughly $600 below Taiwan’s price for observationally identical products, amounting to a 35 percent gap. As the technology matures, the price differential narrows. Five years following Chinese entry, the price gap stabilises around $150. Following our discussion above, we interpret this difference as reflecting unobserved quality differences across suppliers. We can then subtract this difference from price gaps earlier in the product life cycle to identify the contribution of pure price dispersion. For instance, 2.5 years after China’s entry, frictional dispersion accounts for 60 percent of the observed gap.

Figure 2 Closing China-Taiwan price gap following Chinese entry

Implications for price measurement

As noted, statistical agencies regularly omit the effects of frictional price dispersion when calculating price indexes. Given our evidence for substantial real dispersion, we demonstrate how this can bias standard price indexes.5  Figure 3 shows three price indexes for semiconductor wafers. The quality-adjusted index treats a portion of dispersion as frictional using the estimation of unobservable quality just described. The resulting index falls by 11.2 percent per year.

Figure 3 Price indexes with various quality dispersion assumptions

This ideal quality-adjusted index is bracketed by two other indexes. The “technology-country” index follows BLS procedures by assuming all price dispersion is compensated by lower (unobserved) quality. Hence, the index omits any (real) price declines that occur when Chinese producers enter the market for a particular technology. As a result, the technology-country index falls more slowly than the quality-adjusted index. The “technology-only” index makes the other extreme assumption: that there are no unobserved quality differences, so all price gaps reflect frictional price dispersion. This index includes the full price decline when Chinese producers enter, and hence falls the fastest. 

We draw two conclusions from Figure 3 regarding price measurement practice. First, the official BLS approach likely omits price declines resulting from shifts to producers with lower quality-adjusted prices. Second, statistical agencies can easily calculate an alternative index, following the technology-only index, to bound the true quality-adjusted price change. This does not require the analysis of price dynamics used to calculate our quality-adjusted measure. The results for the wafer market suggest that this bound can be tight – as long as one can first control for differences in observable product characteristics.

Though the BLS does not generally collect such detailed data on goods’ attributes, it is important to emphasise that the bounding procedure can still be very instructive. A similar bounding exercise can first be implemented using statistical agencies’ existing transaction-level price data. This analysis would indicate where bias is potentially large and a more time-consuming effort to construct a quality-adjusted index may then be in order. Kim and Reinsdorf (2013) illustrate how one can augment the BLS micro data with additional information on product attributes to control for these differences. We encourage further research in this area and hope that the statistical agencies will receive support to expand these efforts.

Authors’ note: The views presented in this article are the authors’ alone, and do not necessarily reflect those of other members of the Federal Reserve System.

References

Bayard, Kimberly, David Byrne, and Dominic Smith (2013), “The Scope of U.S. Factoryless Manufacturing,” http://www.upjohn.org/MEG/papers/baybyrsmi.pdf.

Bernard, Andrew and Theresa Fort (2013), “Factoryless Goods Producers in the US,” NBER Working Paper 19396.

Byrne, David, Brian Kovak, and Ryan Michaels (2013), “Price and Quality Dispersion in an Offshoring Market: Evidence from Semiconductor Production Services,” NBER Working Paper 19637.

Byrne, David, Stephen Oliner, and Daniel Sichel, “Is the Information Technology Revolution Over?” International Productivity Monitor, 25, p. 20-36.

Carlton, Dennis (1983), “Equilibrium Fluctuations when Price and Delivery Lag Clear the Market,” Bell Journal of Economics, 14 (2), p. 562-572.

Feenstra, Robert, Benjamin Mandel, Marshall Reinsdorf, and Matthew Slaughter (2013), “Effects of Terms of Trade Gains and Tariff Changes on the Measurement of U.S. Productivity Growth,” American Economic Journal: Economic Policy, 2013, 5 (1), p. 1-38.

Hausman, Jerry and Ephraim Leibtag (2009), “CPI Bias from Supercenters: Does the BLS Know that Wal-Mart Exists?,” in E. W. Diewert, J. S. Greenlees, and C. R. Hulten, eds., Price Index Concepts and Measurement, National Bureau of Economic Research.

Houseman, Susan, Christopher Kurz, Paul Lengermann, and Benjamin Mandel (2011), “Offshoring Bias in U.S. Manufacturing,” Journal of Economic Perspectives, 25(2), p. 111-32.

Kim, Mina and Marshall Reinsdorf (2013), “A Test of Hedonic Price Indexes for Imports,” http://www.upjohn.org/MEG/papers/Kim%20and%20Reinsdorf%20w%20revisions.pdf.

Klemperer, Paul (1995), “Competition when Consumers have Switching Costs: An Overview with Applications to Industrial Organization, Macroeconomics, and International Trade,” Review of Economic Studies, 62, p. 515-539.

Nakamura, Emi and Jon Steinsson (2012), “Lost in Transit: Product Replacement Bias and Pricing to Market,” American Economic Review, 102(7), p. 3277-3316.

Reinsdorf, Marshall (1993), “The Effect of Outlet Price Differentials on the U.S. Consumer Price Index,” in M Foss, M Manser, and A Young (ed), Price Measurements and Their Uses, Chicago: University of Chicago Press, p. 227-258.

Rodrik, Dani (2006), “What’s so Special about China’s Exports?” China & World Economy, 14(5), p. 1-19.

Schott, Peter (2008), “The relative sophistication of Chinese exports,” Economic Policy, 23, p. 5-49.


1 The IPP implements this by defining a “product” as a contract between a buyer and seller (Nakamura and Steinsson 2012). Therefore, the price of a product does not necessarily fall if the buyer switches to a lower-price seller.
2 There is a domestic analogue to this known as outlet substitution bias: BLS would not interpret a new outlet store’s low prices as a quality-adjusted decline in the average product price. The classic assessment of this approach is Reinsdorf (1993). See also Hausman and Leibtag (2009).
3 The data are compiled by a non-profit trade association called the Global Semiconductor Alliance.
4 See Byrne, Kovak, and Michaels (2013) for a detailed discussion. The gray dashed line plots the raw price differences for a particular technology – 200mm wafer size and 180nm line width. The black solid line takes similar profiles for all technologies and averages across technology for each year since Chinese entry. The dotted line fits a quadratic trend in a model with individual technology fixed effects.
5 See Houseman et al. (2011) and Feenstra, Mandel, Reinsdorf, and Slaughter (2013) for other examples of the price measurement challenges resulting from globalisation.
 

a

A

Topics:  Industrial organisation

Tags:  imperfect competition, switching costs