By: David M Byrne, Brian K Kovak, and Ryan Michaels from Carnegie Mellon University

Abstract—Many markets exhibit price dispersion across suppliers of observationally identical goods. Statistical agencies typically assume this dispersion reflects unobserved quality, so standard price indexes do not incorporate price declines when buyers substitute toward lower-price suppliers. We show that long-run price differences across suppliers can be used to infer unobserved quality differences and propose an index that accommodates quality-adjusted price dispersion. Using transaction-level data on contract semiconductor manufacturing, we document substantial quality-adjusted price dispersion and confirm that a standard index is biased above our proposed index. To read more, CLICK HERE. 

I. Introduction

Accurate measures of market prices are important in most persistent challenges in the practice of price measurement is accounting for differences in product quality. Such differences may involve products’ characteristics, as well as aspects of the overall transaction, such as customer ser vice or timely delivery (Carlton, 1983). Price indexes seek to measure average price growth, controlling for differences in product quality across goods and over time. This is quite challenging in practice because of difficulties in observing detailed physical product attributes and other less tangible characteristics of transactions. While these challenges have been known for decades, they have recently taken on particular significance in markets for intermediate inputs.1 Such markets are characterized by

While these challenges have been known for decades, they have recently taken on particular significance in markets for intermediate inputs.1 Such markets are characterized by increased internationalization of production chains and shifts toward relatively low-price suppliers in developing countries such as China (Hummels, Ishii, & Yi, 2001). Moreover, an increasing number of “factory less manufacturers” outsource product fabrication activities altogether (Bayard, Byrne, & Smith, 2015; Bernard & Fort, 2013). These developments have led to greater substitution across suppliers of intermediate inputs, both domestic and international. In this context, failure to accurately estimate differences in quality across suppliers will lead to biased import quantity and productivity measures.

In this paper, we study the problem of price index construction when new suppliers and incumbents may charge different prices for goods of identical quality. Standard price indexes, known as matched-model indexes, typically assume that the law of one price holds, which rules out the possibility of price dispersion for identical goods. In doing so, they omit price declines when buyers shift toward suppliers offering discounts, and hence are biased upward. We propose a simple means to infer differences in unobserved quality based on long-run price differences across suppliers. Early in a product’s life cycle, market frictions can impede arbitrage across incumbents and new suppliers, generating price dispersion for goods of similar quality. Yet the influence of these frictions tends to dissipate over time, so that in the long run, several years after entry, price differences largely reflect quality differences. We use this insight to infer unobserved quality and thereby construct a novel price index that both accounts for quality differences across suppliers and allows for deviations from the law of one price.

We apply our method to the contract semiconductor manufacturing industry, using new transaction-level data that include information on prices and all relevant physical characteristics of each product. These data allow us to compare prices for technically identical products across suppliers located in different countries. We find large price differences across suppliers; for example, Chinese producers charged 17% less on average than firms in market leader Taiwan for otherwise identical products. Moreover, the price differences are especially large early in each product’s life cycle but partially converge later on, consistent with a setting in which frictions bind less over time. Together, these patterns sug-gest the presence of cross-supplier price variation that would confound matched-model price indexes. In fact, a standard matched-model index falls almost 1 percentage point per year more slowly than our proposed quality-adjusted index. This substantial upward bias in the standard approach is large enough to meaningfully bias productivity measures and other government statistics.

Although our empirical setting focuses on substitution between suppliers of imported intermediates, the proposed index applies broadly to environments involving the entry of low-price sellers. Thus, our approach also applies to the domestic retail context, where outlet substitution bias resulting from omitting new entrants is a long-standing concern.2 Our proposed index is quite feasible for measurement agencies to implement, as it uses only price and quantity information, which can be collected using existing surveys.

Our empirical findings contribute to a growing literature investigating the measurement implications of the globalization of supply chains, showing that shifts to low-priced offshore suppliers drive systematic bias in standard import price measures.3 Our quality inference procedure also provides an alternative to a common approach in the inter-national trade literature, which infers relative quality across suppliers from differences in market share conditional on price.4 While this approach is suitable for measuring long-run trends in quality or differences in quality for aggregate industries across countries, one should be cautious when studying short time spans or narrowly defined product markets. In a setting where market frictions slow the rate of arbitrage across suppliers, new entrants may have low market shares even when offering the same price and quality as an incumbent supplier. In this setting, the share-based approach systematically understates the relative quality of entering suppliers. We empirically confirm this point in the semiconductor market.

The paper proceeds as follows. Section II introduces the measurement problem and proposes a price index using long-run price differences to infer quality differences. Section III provides background on the contract semiconductor manufacturing industry, which differs in important ways from the more commonly studied memory and processor markets. Section IV investigates price differences and price dynamics between suppliers, finding strong evidence for price differences across suppliers of identical quality goods. In section V, we calculate a standard matched model index and show that it is biased upward in comparison to our proposed index. Section VI concludes.