Are producer prices good proxies for export prices?

Monthly Labor Review, Oct, 1997 by Bill Alterman

Results of a recent study show that producer price indexes remain imperfect measures of export price trends; major conceptual and methodological differences underlie the differences between the two series

Accurate export (and import) price indexes have long been considered crucial for the construction and analysis of such critical economic data as measures of inflation-adjusted net exports, price elasticities of goods in world trade and estimates of the impact of exchange rate fluctuations on domestic inflation and U.S. "competitiveness." Historically, price indexes reflective of total output for U.S. companies, such as the Bureau's producer price indexes (PPt), were frequently used as substitutes for export price indexes. In order to assess the similarity between these two series, the Bureau of Labor Statistics (BLS) recently undertook a project to compare price trends of the Bureau's export price indexes (XPI) with producer price indexes.

The Producer Price Index is designed to measure price changes in a domestic industry's total output, regardless of the destination of that output. The Export Price Index is designed to measure price changes of only those goods that are physically shipped out of the country. Thus, the export price index would appear, at first glance, to be a subset of the producer price index and arguably derivable from it. In general, nearly all of the companies, products, and prices that make up the universe of U.S. exporters also are found in the universe of U.S. domestic producers.(1) However, there are significant differences between the two indexes, including differences in the underlying concepts and in the operational methodology used to construct the indexes.

Concepts and methods

From a production theory point of view, the conceptual export price index underlying the XPI is similar to the conceptual output price index underlying the PPI. Both generally focus on a revenue' maximizing firm for given product prices, input quantities, and technology. There is, however, one significant difference. The PPI normally will not price goods that are being "sold" from one division or plant of a firm to another arm of the same firm. Under the definition of a net output index, these so-called intracompany transfers are considered out-of-scope of the PPI when the different branches of the same company are considered to be in the same industry. In contrast, the XPI represents everything that is exported.

In order for the m and the XPI to move similarly, two conditions must hold. The first is that identical goods in different countries sell for the same price when their prices are expressed in a common currency. This is usually referred to as the "law of one price." For example, if we were to assume that markets were competitive and that complete arbitrage were possible (that is, that there were zero price discrimination), there should not be any difference between the export price of a Chrysler minivan and its domestic price.(2) A number of empirical studies, however, have found that the law of one price does not hold. Possible explanations for these deviations include transportation costs, tariffs, and nontariff barriers, as well as producers' attempts to maximize revenue by varying the price of a good based on destination; this is usually referred to as "pricing to market."(3) Short-term economic trends, such as exchange rate fluctuations or differences in aggregate growth rates, could also play a part. For example, if the dollar were to appreciate sharply against the British pound, the corresponding price of a minivan in terms of the pound would also rise sharply. As a consequence, the exporter might choose to lower his dollar price during the short term in order to maintain market share.(4)

Even if the law of one price holds between U.S. export prices and U.S. domestic prices, a second condition would have to be met in order for producer price indexes to be adequate proxies for changes in export prices: similar goods would have to have the same rate of price change. Because the XPI and the PPI are based on limited samples, it is likely that the actual products priced for the PPI and XPI indexes differ. For example, while the PPI may price a Chrysler minivan, the XPI may sample a Ford minivan. In order for the indexes to trend together, prices of the Chrysler minivan need to trend similarly to prices of the Ford minivan. However, because of differences in consumer tastes and other factors, the minivan that is exported may be very different from the one sold for domestic consumption. For example, a minivan sold to Great Britain would have the steering wheel on the right, while one sold in the United States would have the wheel on the left; or a minivan exported to Europe might be smaller and have different pollution control restrictions. To take the extreme example, suppose that the PPI index for a particular product group contained no exported products, but that such products had, in fact, been exported and are the basis for the XPI. Applying an export weight to the PPI index as part of an estimation of export price change would make little sense without the assumption that the price movements of similar products are the same. In short, to the extent that m indexes consist of products that are not in the XPI, and given that the two conditions necessary for index parity are unlikely to be satisfied, one might expect that many producer price indexes would be poor proxies for the export price indexes representing the same product groups.