Goods Prices and Exchange Rates: What Have We Learned?

The paper opens with the 1994-95 yen appreciation (from 113 to 80 yen per dollar, a 34% move), which, given relative productivity and wage growth, would imply “that labor costs fell by about 30% in the U.S. relative to Japan,” yet Japanese traded-goods output did not collapse and the prices of specific Japanese exports (a Toyota Celica, a Sony Trinitron) barely moved or even fell in dollar terms. The paper’s short answer is that “foreign buyers of Japanese products did not experience substantial price increases in spite of the sharp increase in Japanese labor costs” – that is, there was incomplete, even negative, exchange-rate “pass-through,” which the rest of the survey traces to destination-specific markup adjustment (pricing-to-market) rather than to quantity collapse.

The paper defines an integrated market as one in which “geography and/or nationality do not have systematic effects on transaction prices for otherwise identical products” (citing gold as the paradigm case), and a segmented market as one where location “influences the terms of the transaction in a substantial way.” Following Pigou’s (1920) concept, they operationalize this with third-degree price discrimination: “two national markets are segmented if buyers in those markets face systematically different common currency prices for the same product” – and note that “any perfectly competitive market… must be integrated,” while “a segmented market implies the existence of market power,” though an integrated market need not be perfectly competitive.

All three literatures are organized around a generic regression, p_t = a + dX_t + yE_t + wZ_t + e_t, where p is the price of a product, X is the primary control (a cost or competing-price measure), E is the exchange rate, and Z is a vector of other controls. Law-of-one-price tests set X to a comparison price for the same product in the other country and examine the coefficient on E; exchange-rate-pass-through (ERPT) studies set p to the local-currency import price and X to the exporter’s cost, with the coefficient on E as the “pass-through coefficient” (full pass-through if it equals one); pricing-to-market (PTM) studies use the ratio of export to domestic price as the dependent variable, “blend[ing] the microeconomic foundation of ERPT research… with the study of multiple transactions required in testing the LOP.”

Kreinin (1977), using a natural-experiments design around the 1971 currency realignments, “estimates pass-through to U.S. import prices to be only 50%” (versus 60% for Germany, 70% for Japan, 90% for Canada and Belgium, and 100% for Italy); subsequent 1980s studies “seemed to be centered around 60% pass-through” for the U.S. Since most ERPT regressions include a cost control, incomplete pass-through conditional on cost is conventionally read as evidence of markup adjustment – “60% pass-through conditional on costs implies a 40% markup adjustment to offset exchange rate changes” – though the paper cautions that mismeasured marginal cost (cost indices capture average, not marginal, cost, and can miss changes in imported-input costs or output-driven cost shifts) “may” bias the pass-through coefficient downward and so overstate the apparent markup adjustment.

Knetter’s model regresses export prices (measured at the port of export) on common time effects, destination-specific country effects, and destination-specific exchange-rate slopes; under integrated markets (competitive or not, provided markups are common across destinations), both sets of destination-specific terms should be zero, since all variation beyond a common markup is captured by the time effects. Statistically significant destination effects therefore signal “both market segmentation and a rejection of the constant elasticity of demand hypothesis.” Across Knetter’s and related studies (Marston 1990 on 17 Japanese industries; Gagnon and Knetter 1995 on autos), “the findings on PTM are quite consistent”: estimated degrees of markup offset range widely across industries – from near-zero for goods like tires and amplifiers (prices move about one-for-one with the exchange rate) up to roughly 70% markup offset for Japanese autos and microwave ovens – with the “average degree of PTM found in the transport, equipment, and consumer goods industries studied by Marston… in the neighborhood of 50%.”

Because The Economist’s cover price is pre-set in each market’s local currency and adjusted only infrequently, common-currency price differences mechanically track exchange-rate movements between adjustments – a point Ghosh and Wolf (1994) raised as a potential source of “spurious” PTM findings. But the paper’s own comparison shows this mechanical effect resolves itself for France and Germany, consistent with European integration, while it does not resolve for Germany versus the U.S.: “from January 1981 to March 1985, the dollar appreciated by 49% against the DM… dollar cover prices rose nearly 10%, while DM cover prices rose only 6%,” and “prices in Germany were at least 30% above prices in the U.S. from July 1994 until July 1996.” The authors conclude “the signs of integration evident between Germany and France are completely lacking between Germany and the U.S. Rather, there appears to be conscious price discrimination.”

Building on Bresnahan’s (1989) insight that exchange rates act as a “demand rotator” allowing a markup parameter to be separately identified from marginal cost, studies such as Aw (1993) on Taiwanese footwear jointly estimate demand and supply (markup) parameters; Aw finds the hypothesis of zero markups “cannot be rejected” for the U.S., Hong Kong, and Saudi Arabia, but finds a significant positive markup for Germany “primarily explained by collusion among Taiwanese exporters,” and that U.S. voluntary export restraints let Taiwanese exporters “capture the rents” (an 18% markup from 1977-1981). However, “studies that employ data on export quantities often report imprecise estimates of export demand elasticities,” illustrated by German beer exports to Japan where “unit values are relatively stable… [but] the bilateral export quantities are extremely volatile” – a pattern the authors attribute partly to underreported export shipments and partly to markets that are still “new or in transition,” concluding that “progress in measuring market power in international markets will require either better quantity data or better understanding of how existing export flows are related to underlying demand.”

Engel and Rogers (1996), using detailed U.S.-Canadian city-level CPI data, find a pure border effect “equivalent to approximately 1,780 miles” beyond what distance alone explains, but that this border effect “does not seem to be influenced by the U.S.-Canada free trade agreement”; their sequel finds relative-price volatility is “better explained by nominal exchange rate volatilities than by measures of trade barriers,” pointing to nominal price stickiness combined with volatile exchange rates as the preferred explanation. Wei and Parsley (1995) similarly find deviations from PPP across 14 OECD countries related to transportation costs and exchange-rate volatility, with free-trade-area membership contributing “little” additional explanatory power. Yet this contrasts with micro-level industry studies – Verboven’s (1996) study of European car prices finds that “binding quota constraints… significantly increase prices of Japanese cars in France and Italy” – so the paper concludes it remains unresolved “whether aggregation problems have obscured the importance of trade barriers in the macro studies, or whether the micro studies are an unrepresentative sample of industries.”

Page’s (1981) survey, the most recent evidence the authors report, finds that “98% of U.S. exports were invoiced in dollars,” with German exporters using DM for “over 82%” of exports, British exporters the pound for 76%, and Japanese exporters splitting 62% dollars to 32% yen. When exports are invoiced in the exporter’s own currency and local prices adjust infrequently, common-currency relative prices are mechanically sticky in the short run, which the paper says biases studies of U.S., German, and British exports “against finding PTM, at least in the short run” – so the markup-adjustment estimates the survey reports, if anything, likely understate the true degree of destination-specific price discrimination in the short run, even though PTM is found to persist well beyond the period during which invoice prices are sticky.

The paper’s stated bottom line is that “deviations in the law of one price are not just an artifact of non-identical goods, and incomplete pass-through is not just a result of changes in world prices. Rather, they both appear to be largely a result of third-degree price discrimination,” so that “national markets for goods are better viewed as segmented than integrated.” It leaves two items “on the agenda”: first, establishing which factors (trade and regulatory policy foremost among candidates) actually cause segmentation, given that “research to date has not generated a consensus” between macro studies (which find little role for trade barriers) and micro studies (which do); and second, improving the quantity data needed to measure market power precisely, since the authors say that “progress in measuring market power in international markets will require either better quantity data or better understanding of how existing export flows are related to underlying demand.”