Export Pricing Behavior of Manufacturing: A U.S.–Japan Comparison
Author:
Kenichi Ohno
Search for other papers by Kenichi Ohno in
Current site
Google Scholar
Close

The pricing behavior of U.S. and Japanese manufacturers is compared by using domestic and export price data and a framework of markup over cost. Major export industries in Japan have higher productivity growth and lower pass-through coefficients than American exporters, who tend to price to domestic cost. Japanese firms seem to price-discriminate between domestic and export markets. Although pass-through of major Japanese exports declined in the 1980s, evidence on the statistical significance of the change is mixed.[JEL 212, 227, 431]

Abstract

The pricing behavior of U.S. and Japanese manufacturers is compared by using domestic and export price data and a framework of markup over cost. Major export industries in Japan have higher productivity growth and lower pass-through coefficients than American exporters, who tend to price to domestic cost. Japanese firms seem to price-discriminate between domestic and export markets. Although pass-through of major Japanese exports declined in the 1980s, evidence on the statistical significance of the change is mixed.[JEL 212, 227, 431]

THE PURPOSE of this study is to compare export pricing behavior of U.S. and Japanese manufacturing industries under the floating exchange rate regime.

The concept of pass-through is crucial for this study. It is associated with how prices of internationally traded goods are affected by changes in exchange rates. Roughly speaking, pass-through is said to be complete when the exporter does not adjust prices in his home currency, so that exchange rate fluctuations are reflected entirely in local import prices abroad. By contrast, if import prices in local currencies remain stable, it is prices received by exporters that must adjust to exchange rate shocks. In this case, pass-through is said to be zero. Many of the manufacturing industries considered here are characterized by imperfect competition. Thus, the pass-through coefficient is likely to be the result of conscious price-setting behavior of the export firm.

Pass-through is sometimes defined as the elasticity of import prices with respect to the nominal exchange rate. However, one cannot uniquely determine the pass-through coefficient this way. Suppose, on the one hand, that a nominal depreciation is accompanied by a proportional inflation at home. In this purely nominal depreciation, nothing real is changed: the real exchange rate and competitiveness remain the same, exports will be priced the same abroad, and pass-through will be zero in the absence of money illusion. On the other hand, if there is no inflation differential and therefore a depreciation is both nominal and real, one might expect the export firm to adjust its prices to the new situation.

In the statistical tests that follow, pass-through will be defined with respect to the real effective exchange rate. By so doing, in effect one is testing a joint hypothesis of no money illusion and a particular behavior of interest in each case.

Previous empirical studies suggest that dollar prices of U.S. manufactured exports seem insensitive to changes in the real exchange rate and, therefore, that the movement of the dollar is almost completely passed through to foreign prices. In contrast, foreign manufacturers often “price to market” by revising export prices in their home currencies so that (for example) dollar prices of Japanese products remain relatively stable even when the dollar appreciates or depreciates. 1

To estimate the pass-through coefficient correctly, however, two statistical problems must be overcome.

First, one must control for changes in production cost. Observed changes in output prices may merely reflect exogenous changes in production cost rather than changes in markup. Furthermore, the exchange rate itself could systematically affect production cost by lowering and raising the price of tradable inputs. Therefore, even when the yen appreciates against the dollar, dollar prices of Japanese exports may not rise as much as the yen—a nominal yen appreciation systematically raises Japanese costs relative to American costs (that Is, the real yen exchange rate) only to the extent that inputs are nontradable.

There are several ways to capture production cost. In some studies, the domestic prices of similar goods are used as a proxy for cost. In others, direct measures of cost—such as unit labor cost—are used. In some cases, one may also infer cost changes from the nonstructural, error-component model.

The method of correcting for cost adopted in this study is a straight-forward one: a cost function with two inputs—labor and raw materials— is directly estimated. This gross-input approach contrasts with the value-added approach taken by Marston (1987a, b) where only the labor (and capital) employed in each industry is considered. For estimation and comparison of competitiveness, the gross-input approach is superior to the value-added approach because competitiveness depends not only on the productivity of an export industry but also on the productivity of “upstream” industries from which it buys intermediate products.

Second, a researcher must choose between aggregate data and disaggregated data. One could avoid the aggregation problem by looking at a number of highly disaggregated products. But conclusions obtained from such studies cannot be easily generalized because of a very limited coverage of industries. In contrast, using aggregate data such as export prices of all manufactured goods runs the risk of incompatibility, since the product mixes of U.S. and Japanese industries are not identical.

This study uses price data disaggregated to SIC (Standard Industrial Classification) two-digit and four-digit levels, with a large coverage of manufactured exports (73 percent for the United States and 84 percent for Japan). The extent to which differences in U.S. and Japanese export pricing behavior result from different product mixes, and the extent to which these differences are apparent at the sectoral level, will be examined.

In the next section, alternative theoretical models of pass-through are reviewed. Section II introduces the model. Section III reports the export pricing parameters. Section IV explores nonlinearity in Japanese export pricing. The final section summarizes the results.

I. Existing Models of Pass-Through

Many recent theoretical models attempt to explain export pricing behavior across countries, across industries, or over time. Three of these will be briefly reviewed in this section. 2 They alt assume imperfect competition, whereby the export firm sets rather than takes the price.

Static Profit Maximization

The first type of model reduces different pass-through coefficients to different parameters determining demand and cost in the framework of static profit maximization. In the most simple form, an oligopolistic foreign firm with a constant marginal cost faces a downward-sloping demand curve. For (current) profit to be maximized, marginal revenue must be equal to marginal cost, which dictates the firm’s pricing strategy. According to this theory, the pass-through coefficient is critically dependent on the shape of the demand curve. In particular, pass-through is different from unity (complete pass-through) unless the demand curve has a constant price elasticity. This and similar models have been presented by Knetter (1989), Krugman (1987a), and Mann (1986).

More generally, marginal cost can be increasing or decreasing. Feenstra (1987) and Marston (1989) have developed a model in which the shapes of demand and cost curves jointly determine the pass-through coefficient. They showed that pass-through of greater than unity is possible in some cases.

These models predict that pass-through will be different in each market and that as market conditions change, so will pass-through, Yamawaki (1988) tried to correlate different pass-through coefficients with different characteristics of individual industries. These models also predict that firms with similar technology that share the same market will have similar pass-through coefficients—say, Japanese and U.S. firms selling automobile tires in the third market.

Hysteresis

The second model of pass-through is based on hysteresis. Recent studies concerning the entry-exit decision of foreign exporters suggest that the pass-through relationship may be path dependent or “hysteretic”: for example, if—on the supply side—there are unrecoverable or “sunk” costs associated with investment in, say, a service and distribution network (Baldwin (1988b); Baldwin and Krugman (1986); Foster and Baldwin (1986)) or if—on the demand side—consumer demand is sticky because of brand loyalty (Froot and Klemperer (1988)).

In the Baldwin-Krugman model, the number of active foreign firms in the domestic market will remain unchanged while the exchange rate fluctuates within a certain range. Once the exchange rate (even temporarily) moves out of such a range, however, entry or exit will occur, and the number of foreign firms—and therefore the industry supply curve—will be permanently altered. Modeling this phenomenon as a stopping problem in stochastic dynamic programming—thus introducing uncertainty about the future exchange rate—Dixit (1987, 1989) has shown that the hysteretic range of the exchange rate can be much wider than Baldwin and Krugman suggested. Similarly, Froot and Klemperer (1988) contended that pass-through depends on the expected future exchange rate—or how permanent the current depreciation or appreciation will be.

Models of hysteresis predict that pass-through coefficients tend to change after an extreme movement in the exchange rate. Thus, if one detected such a structural break around 1985 (when the dollar peaked), it would be strong evidence in support of these models. This evidence should preferably be accompanied by the data on the actual number of foreign exporters.

The Nature of Shocks

Finally, some authors have regarded pass-through as a function of the stochastic nature of the macroeconomy. The basic idea is that there is no reason to expect the same amount of pass-through when shocks driving the exchange rate are different. It was argued above that pass-through is likely to be zero when an appreciation is purely nominal but positive when it is not. One can generalize this principle to various other shocks.

Klein (1988) and Murphy (1988a, b) added a signal-extraction problem to this idea; that is, foreign firms infer the current domestic price level (a proxy for the prices of domestic rival firms) from the exchange rate. These authors argued that monetary shocks dominated exchange rate movements in the 1970s, and thus the exchange rate was highly correlated with subsequent inflation. In the 1980s, in contrast, most shocks to the exchange rate have been nonmonetary (mostly fiscal). Deterioration in the signal-to-noise ratio should lower the extent to which foreign firms react to the exchange rate as a signal for future inflation. (See also Daniel (1987).)

These models are interesting because of their general implication that pass-through can vary depending on economic structure. In addition, the relationship between the two endogenous variables—the exchange rate and export prices—presumably changes when there is a shift in economic structure. Although empirical verification of this hypothesis may be difficult, the idea appears worth pursuing.

The model presented here is one of variable markup over cost, with the possibility of a parameter shift over time. This framework is sufficiently general to nest different theoretical models within it, although the derivation of the price equation will be based on the static profit-maximization model. The empirical model provides only a partial equilibrium framework of analysis because it treats such variables as the exchange rate, wages, material prices, and business cycles as exogenous. These macroeconomic variables should be endogenous in a general equilibrium model, but regarding them as given is perhaps less objectionable when one deals with individual industries separately. Moreover, the instrumental-variables method will be used in estimation to take account of what simultaneity might remain.

II. The Model

In this paper, the pass-through coefficient is defined to be the elasticity of f.o.b. (free on board) prices in importers’ currency (actually, a basket of importers’ currencies) with respect to the real exchange rate, after adjusting for cost changes. Thus, the definition of pass-through reflects changes in markup but not changes in production cost. It also excludes changes in tariffs, surcharges, transportation and insurance costs, and distribution costs incurred in importing countries. In actual estimation, what will be estimated is unity minus the pass-through coefficient thus defined—which will be called θx or θx*.

First consider the home country (the United States). Assume that each industry uses two inputs, labor and materials. Labor represents nontradable inputs directly or indirectly employed by this industry, whereas materials are assumed to be internationally tradable. Production technology is characterized by constant returns to scale and Hicks-neutral technical change.

In each period, firms are assumed to minimize the unit cost of production by choosing the best combination of labor and materials. The cost function is given by a translogarithmic form:

ln c = ln α 1 ln w + α 2 ln q + 1 2 γ 11 ( ln w ) 2 + 1 2 γ 22 ( ln q ) 2 + 1 2 γ 12 ln w . ln q ϕ t , ( 1 )

where c is unit cost, w is wages, q is materials prices, and is the rate of technical change. The translog cost function is consistent with various degrees of input substitutability. In the special case in which all γ Yij are zero, equation (1) reduces to a Cobb-Douglas cost function with unitary elasticity of substitution.

As already pointed out, the exchange rate affects the price of tradable goods in two ways: by changing the domestic currency price of importable inputs and by the firm’s decision to revise its profit margin in light of the changed competitive position. Equation (1) captures the first effect by the inclusion of q (material price). The second effect—of primary interest here—will be estimated from the price equation introduced later. If one wished to obtain the combined effect of the exchange rate on output prices, however, it would be necessary to separate the movement in q attributable to the exchange rate from other movements. In this paper, the decomposition of q for each industry is not attempted; instead, the same observed q (inclusive of movement unrelated to the exchange rate) will be used for all industries. This is one area for possible improvement in subsequent studies of pass-through.

Linear homogeneity in input prices implies that3

α1+α2=1and γ12=γ11=γ22.(2)

Thus, equation (1) can be simplified to

lnc=lnα0+αlnw+(1α)lnq12γ(lnwlnq)2ϕt.(3)

By Shephard’s lemma (logarithmic version), the share of each input in total cost is

SL=lnc/lnw=αγlnw+γlnqSM=lnc/lnq=1α+γlnwγlnq.(4)

With these notations, the elasticity of substitution is defined to be 4

ΣLM=(γ+SLSM)/SLSM.(5)

From equation (3), the rate of change in unit cost is 5

c.=αw.+(1α)q.γ(lnwlnq)(w.q.)ϕ.(6)

Let us next turn to the pricing behavior of the firm. Consider an export firm competing with a foreign rival firm in the foreign market. It is assumed that each firm maximizes profit in terms of its home currency, taking the other firm’s price as given. 6 The profits of the export firm and the rival firm are

π=(ep*c)z(7a)
πr*=(pr*c*)zr(7b)

where p* and p* are foreign-currency prices of the export firm and the rival firm, respectively; z and are quantities sold; and e is the nominal exchange rate. The first-order conditions for maxima are

p*=(c/e)[11/η]1(8a)
pr*=c*[11/η]1,(8b)

where ŋ and ŋ* are the price elasticities of demand facing each firm.

The domestic-currency price of exports p(=ep*) is, from equation (8a), a markup over domestic cost c:

p=c[11/η]1,(9)

where the markup depends on the price elasticity, which in turn is a function of relative price and foreign income y (which are determinants of demand):

η=η(p*/pr*,y).(10)

Divide equation (8a) by equation (8b) to see that relative price itself is a function of the real exchange rate and the two elasticities:

p*/pr*=(ec*/c)1f(η,η*),(11)

where ŋ* also depends on relative price and real income.

From equations (9)-(11), p is ultimately a function of real income, the real exchange rate, and domestic cost:

p=h(y,s)ckyλsθc.(12)

The last line is the log-linear approximation of the nonlinear function h; λ and θ are long-run elasticities of price with respect to the cyclical factor and the real exchange rate; and s is the real exchange rate deflated by relative costs (ec* / c). A real depreciation is shown as a rise in s. In equation (12), 1 - θ is the pass-through coefficient (exclusive of the cost effect of exchange fluctuations). If θ = 0, one has p = Kyλc. Exports are priced to domestic cost, and pass-through is complete. In contrast, if θ = 1, one has p = Kyλ(ec*), implying that price is set to match foreign cost (converted to domestic currency). In this case, pass –through is zero.

In terms of rates of change, equation (12) can be written as

p.=λy.+θs.+c..(13)

A tilde (–) is attached to p because equation (13) is the desired price, which may differ from the actual price because of slow adjustment.

Two different measures of the real exchange rate are proposed. First, in the main part of the study industry-specific effective exchange rates, deflated by normalized unit labor costs, will be used. Second, for subsequent sensitivity analysis, these measures will be replaced by an aggregate real effective exchange rate index. 7

Assume that the actual price of output adjusts slowly because of the existence of long-term contracts and menu costs. The partial adjustment mechanism can be described as

p.=(1μ)p.1+μp.+ϵ,(14)

where μ(0≤μ≤1) is the adjustment speed and ϵ is an error term.

Allowance is made for the possibility of price discrimination between domestic sales and exports. In other words, parameters pertaining to pricing strategies (μ,λ,θ) can differ depending on whether the buyer is a domestic national or a foreigner, although parameters pertaining to technical constraints (α,γ,ϕ)) are the same.

Using subscript d for domestic sales variables and parameters and subscript X for those for exports, one can write the price equations, combining equations (6), (13), and (14), as follows:

U.S. Price Equation (Domestic Sales)

p.d=(1μd)p.d,1+μd[λdy.d+θds.+αw.+(1α)q.γ(lnwlnq)(w.q.)ϕ]+ϵd;           (15)

U.S. Price Equation (Exports)

p.x=(1μx)p.x,1+μx[λxy.x+θxs.+αw.+(1α)q.γ(lnwlnq)(w.q.)ϕ]+ϵx;           (16)

Both equations are expressed in the exporter’s currency (that is, U.S. dollars). Note that the cyclical variables, ẏd and ẏx, are different in equations (15) and (16).8 All other explanatory variables are common to both equations. The pass-through coefficient, or the proportion of exchange rate fluctuations reflected in local import prices (f.o.b.), is 1 - θx,- (Actually, pass-through in the model is θx - 1. Here, the sign is reversed, following the convention that pass-through is normally expressed as a positive fraction; had been defined in terms of importers’ currency, the pass-through coefficient would be simply θx.)

The pass-through coefficient could be a function, alternatively, of (1) shapes of demand and cost curves (the static profit-maximization theory), (2) the number of foreign firms in the domestic market (the hysteresis theory), or (3) the stochastic property of the macroeconomy (the nature-of-shocks theory).

Similar price equations can be derived for the other country. With an asterisk (*) used to denote the foreign variable, one has, symmetrically:

Japanese Price Equation (Domestic Sales)

p.*d=(1μ*d)p.*d,1+μ*d[λ*dy.*d+θ*ds.*+α*w.*+(1α*)q.γ*(lnw*lnq*)(w.*q.*)ϕ*]+ϵ*d;           (17)

Japanese Price Equation (Exports)

p.*x=(1μ*x)p.*x,1+μ*x[λ*xy.*x+θ*xs.*+α*w.*+(1α*)q.γ*(lnw*lnq*)(w.*q.*)ϕ*]+ϵ*x;           (18)

Both equations (17) and (18) are expressed in yen.

Equations (15)-(18) are to be estimated simultaneously to take advantage of possible cross-equation correlation of error terms. This procedure also allows imposition and statistical tests of cross-equation parameter restrictions.

III. Estimation and Tests

Equations (15)–(18) were estimated using the iterative three-stage least-squares method, with a constant and once-lagged dependent and independent variables of the entire system as instruments.9 Quarterly data were used. The beginning of the sample period differs from one industry to another, depending on the availability of U.S. export prices, and ranges from 1977:4 to 1983:3. The end of the sample period, however, is uniform (1987:3).

Although the model is written in instantaneous rates of change for conciseness in notation, actual estimation used the discrete time approximation of continuous change. Thus, in what follows the dotted variable in fact is the logarithmic first difference in the corresponding variable.

Table 1.

Technical Parameters

article image
Note: Boldface italic type or * indicates significance at the 10 percent level; ** indicates significance at the 5 percent level. LLRT is the log-likelihood ratio lest, and SIC is Standard Industrial Qassification.

In the case of precision instruments, historical U.S. domestic prices were unavailable. Consequently, a system of three equations excluding the U.S. domestic price equation was estimated. Separately, for primary metal products and passenger cars, estimation produced incredibly low or high α-coefficients for the United States. To obtain more reasonable estimates, conditions α = α* and γ= γ* were imposed for these industries.

In estimation, the adjustment speed was not restricted to be less than unity, and some estimates of μ indeed exceeded unity (that is, over-correction). Note, however, that the system is stable as long as all μs are between zero and 2 (note also that first differences in logarithms are used).

The problem of serial correlation in error terms does not occur—in fact, that is the main reason for estimating the system in rates of change. Of the 75 equations estimated, 7 (3) have statistically significant first-order serially correlated errors at the 10 percent (5 percent) level, which is roughly what should be expected from the type-I error. Higher-order correlations are also absent.

Detailed descriptions of the data and estimated results by industry are provided in the Appendix. In addition to individual industries, the aggregate equations are also reported there. In what follows, the main results are reported, classified by topic.

Technology

Table 1 presents estimated technical parameters for 7 SIC two-digit industries and 12 SIC four-digit industries.

The first two columns report the α coefficient for the United States and Japan, respectively. The α coefficient is related to value-added generated both directly and indirectly—through upstream industries— by the industry. In two cases (chemicals and paperboard for Japan), α seems too low. In two other cases (precision instruments and power-driven hand tools for Japan), α is greater than unity, which clearly cannot be. Nonetheless, these latter two estimates are not significantly different from unity and thus are consistent with true values close to but smaller than unity. Otherwise, estimated values of a appear reasonable.

In explaining the low pass-through coefficients for Japanese manufactured exports in the United States, one argument points to the fact that a rise in the yen makes raw materials and intermediate inputs cheaper for Japanese manufacturers, and this tends to offset the direct adverse effect of yen appreciation. However, the values of α close to unity suggest that such an offsetting effect is empirically negligible for many industrial producers. Although intermediate inputs frequently occupy more than half the reported production cost of individual firms, most of these intermediate inputs already contain domestic labor by way of domestic input-output relationships. The (high) value of a is an estimate of the “true” domestic value-added component of production cost. A relatively high a also implies that the direct effect of exchange rate fluctuations on material costs is small—and production cost is dominated by labor cost. This is not surprising for machinery and equipment embodying the latest technology, which Japan mainly exports. The results also confirm that using only unit labor costs in measuring international competitiveness is a reasonable approximation for these industries.

Table 2.

Export Price Elasticities and Price Discrimination

article image
Note: See note to Table 1.

The next two columns in Table 1 test whether γ is zero—that is, technology is Cobb-Douglas—for the United States and Japan, respectively, using t–statistics. Results are mixed, with some industries rejecting the hypothesis and others not rejecting it.

The fifth column tests, by the log-likelihood ratio test (LLRT) method, whether the two technical parameters, α and 7, are identical across countries. Again, outcome depends on individual industries. At the least, there seems to be no overwhelming evidence that these two technical parameters are different for the United States and Japan.

What is most striking about the bilateral comparison of technology is prominent gaps in the rates of technical change, (ϕ —ϕ*, as shown in the last three columns of Table 1. In all industries examined here, Japan has higher rates of Hicks-neutral technical change than the United States—and in ten instances the difference is statistically significant by LLRT (see the last column). This finding is in accord with that of Marston (1987a, b) and HatsoPoulos, Krugman, and Summers (1988), who reported similar differentials in labor productivity growth between the United States and Japan. It supports the view that many U.S. manufacturing industries are lagging behind Japanese competitors in productivity—despite the recent rise in U.S. manufacturing productivity after the stagnant 1970s (see United States (1988)),

Export Price Elasticities and Pass-Through

In the columns denoted (a) in Table 2, estimates of export price elasticities with respect to the real exchange rate for the United States (θx) and Japan (θx*) are reported. Recall that the definition of the pass-through coefficient (controlled for cost changes, business cycles, and adjustment speed) is 1 - θx and 1 - θx*, respectively. Therefore, θx and θx* can be expected to be normally between 0 and 1, which, however, is not the case for all industries.10

Anomaly seems to arise in industries that rely heavily on one input (for example, pulp in paper and paperboard industries) as well as in some of the SIC four-digit industries (that is, internal combustion engines and printing machinery). It is probable that, at disaggregated levels, θx and θx* are picking up the effects of industry-specific shocks (product innovations, taxes, trade barriers, price fluctuations of major inputs, and the like) that are spuriously correlated with the exchange rate.

There is no reason to expect the same pass-through across all industries, since industries face different demand and cost curves (see Section I). From the (a) columns in Table 2, one may see no evidence of Japanese export price elasticities systematically higher (that is, pass-through systematically lower) than those of the United States, since Japanese coefficients are not always higher. But this impression is deceptive. Let us aggregate estimated SIC two-digit U.S. and Japanese export elasticities, using alternatively U.S. and Japanese exports weights (adjusted to sum to unity):

article image

Results are unambiguous: U.S. estimates yield low export price elasticities (that is, almost complete pass-through) whether U.S. or Japanese weights are used, whereas Japanese estimates yield lower pass-through in the aggregate regardless of weights. This finding arises from the fact that general machinery, electrical machinery, and transportation equipment, which weigh heavily in the exports of both countries, have very different estimates—positive and significant in Japan, nonpositive and insignificant in the United States. 11 Thus, the aggregation problem does not appear to be the main reason for the observed asymmetry in export pricing behavior between the United States and Japan—at least at the SIC two-digit level. On average the United States has a pass-through coefficient of 0.95 (= 1 - 0.05) and Japan has a pass-through coefficient of 0.78 (= 1-0.22).

Figure 1.
Figure 1.

Changes in Domestic and Export Prices: U.S. General Machinery

(In percent, December to December)

Citation: IMF Staff Papers 1989, 003; 10.5089/9781451973037.024.A002

Note: Export prices before 1979 were unavailable

This can also be demonstrated graphically, in Figures 1 and 2. Figure 1 depicts yearly changes in domestic sale prices (in domestic currency) and export prices (in domestic currency and in the foreign currency basket representing export destinations) of general machinery for the United States; Figure 2 similarly depicts these changes for Japan. Domestic and export prices tend to move together in the United States, whereas Japanese export prices adjust systematically to the exchange rate.

Aggregation Problem in the Exchange Rate?

Can the use of an aggregate measure of the real exchange rate, rather than industry-specific ones, change these results substantially? Columns denoted (b) in Table 2 report export price elasticities using the common real exchange rate measure. Comparing columns (a) and (b), one discovers that different measures of the real exchange rate do not alter the result by much in most cases. However, there are notable exceptions. The coefficients for primary metal products, American passenger cars, and Japanese transportation equipment become smaller and lose significance when the aggregate measure is used.

Figure 2.
Figure 2.

Changes in Domestic and Export Prices: Japanese General Machinery

(In percent, December to December)

Citation: IMF Staff Papers 1989, 003; 10.5089/9781451973037.024.A002

Price Discrimination

Figures 1 and 2 raise another important issue. If major export industries of Japan adjust export prices but not domestic prices as the yen rises or falls, the discrepancy between the two develops systematically with the exchange rate. For instance, at the time of a strong yen, the same brands—and even models—of Japanese automobiles, cameras, stereo equipment, and the like will become cheaper abroad than at home. This generates periodical deviations from purchasing power parity at the most disaggregated level and prompts the allegation of “dumping.” Such violation of the law of one price, however, should be regarded not so much as an “unfair” trade practice as a natural consequence of a yen appreciation given Japanese export pricing behavior. In times of a weak yen, the reverse phenomenon of selling the same goods cheaper at home than abroad is observed.

The last two columns of Table 2 show LLRT results of price discrimination between domestic and export markets. None of the U.S. industries price-discriminates, but some Japanese industries do. However, the hypothesis of no price discrimination is not rejected for major Japanese industries, including general machinery (plotted in Figure 2). This apparent inconsistency between the visual impression and LLRT is puzzling—although one should be aware that the power of LLRT may not be very high.

IV. Nonlinearity in Pass-Through

Models of hysteresis predict certain nonlinearity in the relationship between the dollar and import prices in the United States. For example, export prices may respond differently to large and small changes in the exchange rate. Furthermore, pass-through coefficients may be permanently altered after an extreme but temporary appreciation or depreciation. Using aggregate U.S. import price data, Baldwin (1987) and Kim (1988) detected such a structural break sometime in the first half of the 1980s. However, such a break can also occur when the demand and cost curves shift, or when the nature of shocks changes.

In this section, the possibility of nonlinear relations between the exchange rate and Japanese export prices is explored for the sample period of 1975:4–1987:3. (American data are too short for such tests.) The general method used for this purpose can be described as follows. Equation (17)—the Japanese domestic price equation—is combined with the modified equation (18):

p . * x = ( 1 μ * x ) p . * x , 1 + μ * x [ λ * x y . * x + θ * x , 1 s . 1 * + θ * x , 2 s . 2 * + α * w . * + ( 1 α * ) q . γ * ( ln w * ln q * ) ( w . * q . * ) ϕ * ] + ϵ * x ;           (18a)

where s1* is equal to s* (real exchange rate) when the tatter satisfies a certain condition (for example, is positive) and is equal to zero otherwise, and s2* = s* — s1*. Equations (17) and (18a) were estimated by iterative three-stage least squares with these exchange rate variables entered separately. In this way one can examine whether export price elasticities are different depending on (1) whether the yen is rising or falling, (2) whether changes in the yen are large (greater than the sample standard deviation) or small, (3) before or after 1981:1, and (4) before or after 1985:2.

These various estimates of θx* are presented in Table 3, together with LLRT results for the hypothesis that two export price coefficients are identical. (However, most LLRT statistics are insignificant despite marked differences in numerical estimates, again raising the question of the power of LLRT.) The findings are discussed for the four major export industries of Japan: primary metal products, general machinery, electrical machinery, and transportation equipment.

As regards the sign of exchange rate changes, primary metal products show similar export price responses to either appreciation or depreciation. For the other three machinery and equipment industries, however, there is a tendency to raise yen prices more readily when the yen depreciates (positive changes) than to lower them when the yen appreciates (negative changes). The results remained the same even in a simpler model (not shown here), in which cost, business cycles, and adjustment speed were not considered. This is quite contrary to the popular belief that Japanese exporters tend to keep U.S. dollar prices lower than what short-term profit maximization would dictate whether the yen is strong or weak.

As to the magnitude of changes in the yen, large changes prompt large adjustment in export prices in some industries but small adjustment in others—and generalization seems difficult. Ideally, one would like to look at large cumulative changes versus small cumulative changes in the exchange rate for such tests—whereas the criterion here for large changes is for each period. This is not as bad as it seems, however, because variance in the real exchange rate is highly serially correlated. Thus, the large changes here are indeed often cumulative.

There is an indication, as far as the three major export industries are concerned, that export price elasticities increased (and thus the degree of pass-through decreased) after the first quarter of 1981. This finding confirms other empirical studies that detect a structural break in U.S. import prices sometime in the early 1980s. 12 However, when the break point is shifted to the second quarter of 1985 when the dollar finally started to fall, one no longer has evidence of structural change. When one takes explicit account of cost changes, business cycles, and adjustment speed (as is done here), it may be that the discontinuity of the pass-through relationship in 1985—which some studies find—-may disappear.

The results do not explain why these nonlinearities are occurring. In particular, a break in pass-through in the early 1980s—when the dollar was neither extremely high nor low—cannot be regarded as strong evidence of hysteresis.

At the same time. Table 3 makes clear that the outcomes of nonlinearity tests depend heavily on the individual industries chosen for study, particularly at highly disaggregated levels. This is perhaps because other important variables specific to each industry are not properly captured. In this sense, aggregated data that cancel out idiosyncracies of individual industries may be more suitable for the study of pass-through, provided that the aggregation problem is not serious.

V. Conclusions

The main findings of this study can be summarized as follows.

  • Japan has higher rates of technical change in relation to the United States in all industries considered here. There is no strong evidence, however, that other technical parameters are dissimilar between the two countries.

  • In general machinery, electrical machinery, and transportation equipment, which are major export industries in both countries, U.S. pass-through coefficients are close to, and insignificantly different from, unity. Japanese pass-through coefficients are less than 0.8 and significantly different from unity. Evidence is less clear at a more disaggregated level or for materials industries.

  • At the SIC two-digit level, data do not support the hypothesis that aggregate U.S. and Japanese export prices behave differently because of different product mixes. In other words, the asymmetry is present even at the sectoral level.

  • Japanese major exporting industries appear to price-discriminate between domestic and overseas markets as the exchange rate changes. No such tendency is detected for U.S. manufacturers.

  • There seems to have been a structural break in pass-through of Japanese machinery and equipment exports in the early 1980s, although the results of statistical tests remain mixed.

Although much has been learned about pass-through, information is still insufficient to evaluate different theoretical models of pass-through. In particular, detection of a structural break does not necessarily support the hypothesis of hysteresis, since such a break could occur for various other reasons as well.

Table 3.

Export Price Elasticities and Nonlinearity for Japan

(Estimates of θx* under certain conditions)

article image
article image
Note: See note to Table 1.

APPENDIX

Data Definitions and Estimation Results by Industry

The data used are defined below; estimation resuhs by industry are given in Table 4.

U.S. domestic prices, published by the Bureau of Labor Statistics (BLS) of the U.S. Department of Labor, were retrieved from the DRI Database. For all SIC two-digit industries and SIC 3546 (pumps and pumping equipment), new producer price indices were too short for the analyses, and therefore wholesale price indices were used. For all other SIC four-digit industries, producer price indices were used. SIC-based U.S. export prices were obtained from BLS. The limited availability of these data was the main constraint on the coverage and sample periods of the analyses.

Japanese domestic and export wholesale price indices were obtained from the Bank of Japan (BOJ). Because BOJ data are not organized according to SIC, they were matched up with U.S. data, by aggregating multiple series if necessary. One outlier in the export price of precision instruments (1982:3) was ignored. Some products were excluded from the study because a reasonable concordance could not be established. All U.S. and Japanese price data are quarterly (last month of the quarter).

Exchange rate series were generated in two different ways. First, industry-specific real effective exchange rate series were constructed for individual industries in each country. Weights were derived from bilateral export shares of 16 industrial countries (Canada, the United States, Japan, Belgium, Denmark, France, the Federal Republic of Germany, Ireland, Italy, the Netherlands, the United Kingdom, Austria, Finland, Norway, Sweden, Switzerland, and Spain— minus the home country) published in the United Nation’s Commodity Trade Statistics (New York) for 1984. The International Monetary Fund’s normalized (that is, cyclically adjusted) unit labor costs for manufacturing were used as deflators. Second, in columns (b) of Table 2, real effective exchange rate series for the United States and Japan, not disaggregated by industry, were obtained from the Fund’s database. Weights were based on multilateral (that is, including third market effects) manufacturing export shares of the same 16 countries, and the same deflators were used (see McGuirk (1987)).

Cyclical variables were constructed as follows. For domestic price equations, the change in seasonally adjusted domestic real GNP was used. For export price equations, the weighted average of changes in the seasonally adjusted real GNP of the major industrial countries, excluding the home country, was used, with weights proportional to 1981 GNP. These cyclical variables were zero-mean adjusted for each sample period.

Wages were obtained from the Fund’s International Financial Statistics (IFS) and were seasonally adjusted. Material prices for the United States are BLS’s wholesale prices for nonfood materials (including fuel). Material prices for Japan are the import price index in IFS (Japanese imports in the sample periods were predominantly fuel and raw materials). Logarithms of wages and materials prices used in estimation were zero-mean adjusted for each sample period.

Table 4.

Estimation Results by Industry

article image
article image
article image
article image
Note: These are the results of iterative three-stage least-squares estimation of equations (15)–(18); r–statistics are in parentheses; is the adjusted coefficient of determinatio. a Using aggregate data in Figure 1.

REFERENCES

  • Baldwin, Richard E, (1988a), “Some Empirical Evidence on Hysteresis in Aggregate U.S. Import Prices,” NBER Working Paper 2483 (Cambridge, Massachusetts: National Bureau of Economic Research, January).

    • Search Google Scholar
    • Export Citation
  • Baldwin, Richard E, (1988b), “Hysteresis in Import Prices: The Beachhead Effect,” American Economic Review (Nashville, Tennessee), Vol. 78 (September), pp. 77385.

    • Search Google Scholar
    • Export Citation
  • Baldwin, Richard E, and Paul R. Krugman, “Persistent Trade Effects of Large Exchange Rate Shocks,” NBER Working Paper 2017 (Cambridge, Massachusetts: National Bureau of Economic Research, September 1986).

    • Search Google Scholar
    • Export Citation
  • Berndt, Ernst R., and Laurits R. Christensen, “The Translog Function and the Substitution of Equipment, Structures, and Labor in U.S. Manufacturing 1929-68,” Journal of Econometrics(Amsterdam), Vol. 1 (1973).

    • Search Google Scholar
    • Export Citation
  • Berndt, Ernst R., and David O, Wood, “Technology, Prices, and the Derived Demand for Energy,” Review of Economics and Statistics(Amsterdam), Vol. 57 (August 1975).

    • Search Google Scholar
    • Export Citation
  • Berndt, Ernst R., B.H. Hall, R.E. Hall, and J.A. Hausman, “Estimation and Inference in Nonlinear Structural Models,” Annals of Economic and Social Measurement (New York), Vol. 3 (1974).

    • Search Google Scholar
    • Export Citation
  • Chadha, Binky, “Exchange Rate Shocks and Optimal Goods Price Adjustment in a T>o Country Setting” (unpublished; New York: Columbia University, September 1987).

    • Search Google Scholar
    • Export Citation
  • Daniel, Betty C, “Price Setting, Imperfect Information, and PPP Deviations,” paper presented at the conference on The Dynamic Behavior of PPP Deviations (Cambridge, Massachusetts: National Bureau of Economic Research, August 1987).

    • Search Google Scholar
    • Export Citation
  • Denny, Michael, and Doug May, “The Existence of a Real Value-Added Function in the Canadian Manufacturing Sector,” Journal of Econometrics (Amsterdam), Vol. 5 (January 1977).

    • Search Google Scholar
    • Export Citation
  • Dixit, Avinash, “Entry and Exit Decisions of Firms under Fluctuating Real Exchange Rates,” (unpublished; Princeton: Princeton University, October 1987).

    • Search Google Scholar
    • Export Citation
  • Dixit, Avinash, “Hysteresis, Import Penetration, and Exchange Rate Pass-Through,” Quarterly Journal of Economics (New York), Vol. 104 (May 1989).

    • Search Google Scholar
    • Export Citation
  • Dornbusch, Rudiger, “Exchange Rates and Prices,” American Economic Review (Nashville), Vol. 77 (March 1987).

  • Feenstra, Robert C, “Symmetric Pass-Through of Tariffs and Exchange Rates Under Imperfect Competition: An Empirical Test,” NBER Working Paper 2453 (Cambridge, Massachusetts: National Bureau of Economic Research, December 1987).

    • Search Google Scholar
    • Export Citation
  • Foster, Harry, and Richard E. Baldwin, “Marketing Bottlenecks and the Relationship Between Exchange Rates and Prices” (unpublished; October 1986).

    • Search Google Scholar
    • Export Citation
  • Froot, Kenneth A., and Paul D. Klemperer, “Exchange Rate Pass-Through When Market Share Matters,” NBER Working Paper 2542 (Cambridge, Massachusetts; National Bureau of Economic Research, March 1988).

    • Search Google Scholar
    • Export Citation
  • Fuss, Melvyn A, “The Demand for Energy in Canadian Manufacturing: An Example of the Estimation of Production Structures with Many Inputs,” Journal of Econometrics (Amsterdam), Vol. 5 (January 1977).

    • Search Google Scholar
    • Export Citation
  • HatsoPouios, George N., Paul R. Krugman, and Lawrence H. Summers, “The Problem of U.S. Competitiveness,” paper presented at the Fourth International Symposium of the Economic Planning Agency (Tokyo, March 1988).

    • Search Google Scholar
    • Export Citation
  • Helkie, William, and Peter Hooper, “An Empirical Analysis of the External Deficit, 1980-86,” in External Deficits and the Dollar: The Pit and the Pendulum, ed. by Ralph Bryant, Gerald Holtham, and Peter Hooper (Washington: The Brookings Institution, 1988).

    • Search Google Scholar
    • Export Citation
  • Hooper, Peter, and Catherine L. Mann, “The U.S. External Deficit: Its Causes and Persistence,” International Finance Discussion Paper 316 (Washington: Board of Governors of the Federal Reserve System, November 1987).

    • Search Google Scholar
    • Export Citation
  • Kim, Yoonbat, “Exchange Rates and Import Prices in the U.S.: A Varying Parameter Estimation of Exchange Rate Pass-Through,” Vandeveer Discussion Paper Series No. 88-1 (Illinois: Southern Illinois University at Carbondale, January 1988).

    • Search Google Scholar
    • Export Citation
  • Klein, Michael W., “Macroeconomic Aspects of Exchange Rate Pass Through” (unpublished; Worcester, Massachusetts: Clark University, February 1988).

    • Search Google Scholar
    • Export Citation
  • Knetter, Michael M., “Price Discrimination by U.S. and German Exporters,” American Economic Review (Nashville, Tennessee), Vol. 79 (March 1989), pp. 198210.

    • Search Google Scholar
    • Export Citation
  • Krugman, Paul R. (1987a), “Pricing to Market When the Exchange Rate Changes,” in Real-Financial Linkages Among Open Economies, ed. by S.W. Amdt and J.D. Richardson (Cambridge, Massachusetts; MIT Press).

    • Search Google Scholar
    • Export Citation
  • Krugman, Paul R. (1987b), ed.. Strategic Trade Policy and the New International Economics (Cambridge, Massachusetts: MIT Press).

  • Krugman, Paul R. and Richard E. Baldwin, “The Persistence of the U.S. Trade Deficit,” Brookings Papers on Economic Activity: 1 (1987), the Brookings Institution (Washington).

    • Search Google Scholar
    • Export Citation
  • Mann, Catherine L., “Prices, Profit Margins, and Exchange Rates,” Federal Reserve Bulletin (Washington), Vol. 72 (June 1986).

  • Marston, Richard C. (1987a), “The Relative Competitiveness of U.S. and Japanese Manufacturing,” paper presented at the NBER-Ministry of Finance International Symposium (Tokyo, September 1987).

    • Search Google Scholar
    • Export Citation
  • Marston, Richard C. (1987b), “Real Exchange Rates and Productivity Growth in the United States and Japan,” in Real-Financial Linkages Among Open Economies, ed. by S.W. Arndt and J.D. Richardson (Cambridge, Massachusetts; MIT Press, 1987).

    • Search Google Scholar
    • Export Citation
  • Marston, Richard C. “Pricing to Market in Japanese Manufacturing” (unpublished; Philadelphia; University of Pennsylvania, March 1989).

    • Search Google Scholar
    • Export Citation
  • McGuirk, Anne K., “Measuring Price Competitiveness for Industrial Country Trade in Manufactures,” IMF Working Paper, WP/87/34 (unpublished; Washington: International Monetary Fund, 1987).

    • Search Google Scholar
    • Export Citation
  • Murphy, Robert G. (1988a), “Import Pricing and the Trade Balance in a Popular Model of Exchange Rate Determination” (unpublished; Chestnut Hill, Massachusetts: Boston College, March 1988).

    • Search Google Scholar
    • Export Citation
  • Murphy, Robert G. (1988b), “Import Pricing and the Macroeconomy” (unpublished; Chestnut Hill, Massachusetts: Boston College, March 1988).

    • Search Google Scholar
    • Export Citation
  • United States, Office of the President, Economic Report of the President, Together with the Annual Report of the Council of Economic Advisers (Washington: Government Printing Office, 1988).

    • Search Google Scholar
    • Export Citation
  • Woo, Wing T., “Exchange Rates and the Prices of Nonfood, Nonfuel Products,” Brookings Papers on Economic Activity: 2 (1984), The Brookings Institution (Washington).

    • Search Google Scholar
    • Export Citation
  • Yamawaki, Hideki, “Export Pricing Behavior and Market Structure” (unpublished; Berlin: International Institute of Management, February 1988).

    • Search Google Scholar
    • Export Citation
*

Mr. Ohno is an economist in the Middle Eastern Department of the Fund; he was an economist in the Research Department when this paper was written. He holds a doctorate from Stanford University as well as degrees from Hitotsubashi University, Tokyo.

2

Other explanations of pass-through emphasize (1) the dollar’s role as a dominant invoice currency; (2) U.S. firms’ global market power; (3) difference in the export dependency ratio; (4) aggregate demand conditions; (5) the size of firms—Japanese export firms are targe and have deep pockets; and (6) differences in the profit-maximization horizon—Japanese are long-term maximizers, whereas Americans are short-term maximizers. These last differences may be due to differences in corporate culture, capital cost, the role of the stock market, productivity growth, or direct investment. There are also eame-theoretic models of pass-through in which, for example, no or little pass-through is Pareto superior to the Cournot-Nash solution (Chadha (1987). See also Dornbusch (1987), Hooper and Mann (1987), and Krugman (1987a) for additional models.

3

Totally differentiate equation (1) and set Δ Inc = Δ ln w = i In q = k and Δt = 0. The resulting equation must hold for any ln w or ln q, hence equation (2).

5

From now on, all equations are in rates of change. The primary reason for this is to eliminate high serial correlation in the error terms. An alternative way is to model the structure of error explicitly, but this would increase computational difficulty.

6

The markup equation derived below is based on what is called the static profit-maximization approach in Section I.

7

For consistency, one would ideally like to make use of the cost equation (6) in constructing real effective exchange rate indices. This cannot be done, however, because no such equations exist for trading partners other than the United States and Japan. The alternative strategy taken here is to use multilateral real effective exchange rates computed separately.

8

The variable yd is defined as the zero-mean adjusted rate of change in domestic real gross national product (GNP), whereas yx is a weighted average of zero-mean adjusted rates of change in real GNP of the major industrial countries other than the home country. Because the price equations do not have free constant terms, it is necessary to subtract the mean from these independent variables so as not to bias the estimate of 0. (Note that’s already has a mean very close to zero and does not need such adjustment.)

9

The MINDIS, or minimum-distance estimation, command of the RAL statistical package was used. The algorithm is from Berndt and others (1974).

10

However, θx < 0 cannot be ruled out under certain assumptions about cost and demand—see Feenstra (1987) and Marston (1989).

11

However, LLRT results in Table 2 are less conclusive. Only paper-related industries and electrical machinery are seen to have export price elasticities that differ statistically between the two countries.

12

Curiously, this is not apparent from Figures 1 and 2. Nor did a simpler model with no correction for cost, business cycles, and adjustment speed detect such a break.

  • Collapse
  • Expand