1. statement of the problem

Within the tradition of the elasticity approach to the balance of payments,¹ exchange rate adjustment falls into the general category of switching policies (in the “old” Johnson sense). Along with tariff changes and a variety of other measures, exchange variations are expected to change the relative prices of importables and exportables, and thereby induce shifts in production and consumption mixes. These changes, in turn, are expected to restore or to maintain equilibrium in the balance of payments. But the expected salutary effects of exchange rate adjustments can occur only if they are fully or partially reflected in the prices of traded goods, rather than being offset by proportional changes in domestic prices. The extent to which exchange rate changes are transformed into changes in the prices of imports (denominated in the local currency) and exports (denominated in foreign currencies) is known as the “pass-through” effect of the exchange adjustment. Although the “monetary” approach (which centers attention on long-run equilibrium and focuses on the demand for, and the supply of, money as a stock) denies that devaluation can have a lasting effect on the balance of payments, even this approach admits to a transitory change in the terms of trade occurring over the short run (without, however, specifying how short the short run is).

This paper is devoted to an empirical estimation of pass-through effects, and is limited to a time horizon of three to four years (i.e., the relatively short run). Consequently, this paper belongs in the tradition of the elasticity approach. In that context, the pass-through effect has important implications, both for the balance of payments and for the propagation of inflation. Specifically, the paper estimates the following: (a) the pass-through effect of exchange rate changes of eight major currencies under the Smithsonian Agreement of 1971, with a partial investigation of the changes occurring in 1973; (b) the effect of the above price changes on the volume of trade flows and the implied import-demand elasticities; and (c) the elasticity of substitution in each of several major countries as between third country suppliers over two time periods: 1970-72 and 1970-73.

2. SOME A PRIORI REASONING

In theoretical, partial equilibrium terms, the pass-through effect depends on the elasticities of export supply and import demand of the country and its trading partners. Although these elasticities are known to vary between countries and over time,² a few generalizations concerning the pass-through effect can be made on a priori grounds. A small country, which can be assumed to face an infinitely (or very highly) elastic supply of exports from its trading partners, is likely to experience a nearly complete pass-through on the import side. In turn, only a partial pass-through can be expected with respect to the import prices of a large country, which presumably faces upward-sloping export supply curves. In other words, such a country can, by its own actions, affect its terms of trade. The reverse may be the case with respect to the supply of exports. Here, a large country is likely to have a more elastic export supply than a small one (because exports constitute a smaller proportion of output in most industries in large countries) and is, therefore, less likely to change export prices denominated in its own currency following an exchange adjustment. Consequently, exporters in a large country are likely to pass through a greater proportion of a devaluation or revaluation than exporters in a small country.³

But this argument assumes that the small country specializes in the export of a few commodities and that it exports a substantial portion of domestic output of each commodity—an assumption that may be incorrect. Viewed from the demand side, the foregoing observations reflect the fact that a small country is a price taker on the international market, unable to influence its terms of trade. But this statement leaves open the question of how large a country has to be before it acquires some control over its terms of trade, as well as the different degrees of market power associated with varying economic size. Consequently, the above a priori statement must be regarded as tentative. In essence, the pass-through question is an empirical one; it can be handled either through precise knowledge of the elasticities involved, or by a method specifically designed to measure the pass-through.

In many empirical studies of the domestic impact of commercial policy conducted in the 1950s and early 1960s, there was a tendency on the part of researchers to assume (implicitly or explicitly) a 100 per cent pass-through. More specifically, it was commonly assumed that changes in, say, tariffs were fully reflected in changes in import and/or export prices. In recent years, the pendulum has swung almost completely in the opposite direction (especially in some theoretical discussions). For example, in the view of some economists the law of one price ensures the same price on the world market for each internationally traded good (including “differentiated” products). Consequently, given sufficient time, exchange rate adjustments would be fully compensated for by changes in domestic prices.

Furthermore, during the recent period of fluctuating exchange rates, the issue has become a key factor in a certain theory that purports to explain world-wide inflation. While the basic hypothesis involved, the Mundell-Laffer (M-L) hypothesis, was developed by (or is attributed to) Professors Robert A. Mundell and Arthur B. Laffer,⁴ who are also closely associated with the monetarist approach, the M-L hypothesis is not an essential ingredient of that approach. Mundell and Laffer assert that not only do exchange fluctuations fail to equilibrate the balance of payments but they also contribute to worldwide inflation. The argument runs roughly as follows: The law of one price guarantees that, given sufficient time for adjustment (and abstracting from transport costs), all internationally traded goods will command the same price everywhere; this applies to homogeneous and differentiated products alike. Thus, a currency devaluation cannot, over time, change a country’s prices relative to those of its competitors; either its prices would rise or foreign prices would decline until prices were fully equalized internationally. Here, Mundell and Laffer introduce a second supposition—namely that the price response to exchange rate adjustment is not symmetrical. Export prices (denominated in local currency) rise in the devaluing country, but import prices fail to decline in the revaluing one. This asymmetry is often referred to as the “ratchet effect”. As a consequence, the equalization of international prices is accomplished strictly through price increases in the devaluing country. Since, in a regime of fluctuating exchange rates, some currencies depreciate and others appreciate over one time period, while the reverse tends to occur during some subsequent period, and because domestic price changes (i.e., increases) occur only in the depreciating countries and not in the appreciating ones, the net effect is a world-wide increase in the prices of traded goods.⁵

Both links in the M-L argument can be questioned. First, there is no a priori reason for the law of one price to hold in the case of differentiated products. Even a brand name can account for a persistent price differential. The elasticity of substitution between different suppliers of a manufactured product having similar characteristics is less than infinite, even in the long run.⁶ And, in any case, it makes a considerable difference whether the period required for price equalization following a currency devaluation is long or short. If it is very protracted (as implied in some versions of the M-L hypothesis), then the argument that devaluation does not improve a country’s competitive position holds only in the long run. Apart from the question of how long the long run is, it is clear that improvement could occur, and persist, during the years in which the price equalization process takes place. And that may be sufficient for exchange rate adjustments to perform their traditional function of improving the country’s competitive position and its balance of payments.⁷ By the time the relevant period was over, other exchange rate changes would undoubtedly occur.

Second, there is no a priori reason to expect a ratchet effect in the case of exchange rate changes. Even if internal prices are inflexible in a downward direction, import prices (expressed in terms of the home currency of a revaluing country) can decline following an upward adjustment in the exchange rate. Indeed, empirical studies have shown many instances of such price reductions, on both a quarterly and an annual basis, in the postwar period.⁸

In sum, the pass-through effect of exchange rate adjustment has important implications, both for the balance of payments and for the propagation of inflation. But its extent is an empirical question and cannot be determined by a priori considerations. Such an empirical determination is the main purpose of the present study. Since this is essentially a short-run investigation, it does not settle the theoretical issues raised above. Nevertheless, it does shed light on the strength of these factors over a three-year period.

Since the data available for this investigation are annual and not quarterly, no time lags were built into the study—that is, it does not investigate the lag of price movements behind exchange rate variations or the lag of quantity changes behind prices changes. On the other hand, because percentage changes in both quantity and price were generated, their ratios would yield an elasticity of substitution between suppliers in each of the countries being studied.

3. METHODOLOGY—A GENERAL DISCUSSION

While other students of the subject have employed econometric models⁹ to examine the question at hand, this study relies on a “control country” approach designed specifically for this purpose. In a sense, this approach fulfills the role of a laboratory experiment or an economic model in “holding other things constant.” For, in any given year, a multitude of factors affect the import and export prices of a country. Yet, in estimating the pass-through effect, it is necesary to isolate the impact of exchange rate changes on the prices of traded goods (as well as on the volume of trade). In other words, the pass-through effect is the difference between the change in prices that actually has taken place and the hypothetical change that would have occurred in the absence of exchange rate changes (but with all other influences allowed to have their full impact).

The problem is to obtain the hypothetical change. The approach employed here uses a control country (or countries) to arrive at that change. Ideally, such a country should be similar in all or most respects to the country with which the control country is being compared, except that its exchange rate did not change. Because such an ideal is rarely, if ever, found in practice, we shall use several (usually three or four) control countries for each country under study. (Additionally, we shall examine the nature of possible biases and develop alternative formulas to deal with them.) In each case, the hypothetical price changes in the investigated country are inferred from the average change that actually occurred in the control countries. The control country (countries) vary from one investigated country to another. They are selected in each case so as to best hold “other things” (i.e., variables other than the exchange rate changes) constant, and to isolate the effect of exchange rate changes. A detailed description of the estimating procedures is offered in the next section.

It is not claimed that the control country approach is better (or, for that matter, worse) than alternative methods used to investigate the same problem; only that it is different. When an issue is sufficiently important, then—given the well-known possibility of errors in any empirical investigation—it is desirable to employ as many approaches as possible to investigate it. Similar results yielded by diverse methods would add confidence to the estimates.

Although the control country (or, alternatively, a control product group) approach is not often used in international economics, it has been employed on several occasions, and by now has acquired a respectable tradition in the field. Its use was originally stimulated by Orcutt’s seminal article on elasticity measurement,¹⁰ in which he demonstrated the strong downward bias imparted by the traditional regression model and urged the use of alternative methods. Since that time the control country or control group approach has been employed to deal with such problems as the effect of tariff changes on trade flows; trade and colonialism; the effect of tariff concessions on the exports of developing countries; and the effect of regional economic integration on the volume of imports.¹¹ In all these cases the control country approach yielded fruitful results. It has certainly proved robust enough to warrant its application to the present problem.

1. THE USE OF CONTROL COUNTRIES

Under the 1971 Smithsonian Agreement, there was a realignment of all major currencies. The exact change in the exchange rate of a given currency depends on the weight assigned to each of that country’s trading partners. Various indices have been constructed¹² to estimate the average changes in the exchange rate of each major currency. These indices, which generally employ 1970 as a base period, were developed for use in the period of fluctuating exchange rates. They differ from each other in the nature of the weights assigned to each bilateral change in the particular currency’s exchange value.¹³ Regardless of the weighting scheme, the indices show the following (approximate) exchange rate changes of major currencies between 1970 and 1972: U. S. dollar −10 per cent; Japanese yen +15 per cent; deutsche mark +6 per cent; and French franc unchanged. Similar averages are available for other currencies of the industrial countries. Most empirical investigations of the effects of exchange rate changes utilize these average changes.

In contrast, the present study makes use of the fact that the average change in the value of each currency is made up of differential degrees of bilateral changes vis-à-vis different individual currencies. For example, various currencies were revalued by different amounts relative to the U.S. dollar in the Smithsonian Agreement of 1971. The Smithsonian changes approximated the changes in exchange rates that took place between (average) 1970 and (average) 1972. But because other adjustments have occurred in the value of some currencies, the two sets of changes are not identical. Table 1 shows the average 1972 exchange rate changes, relative to the dollar, of 15 main countries of the Organization for Economic Cooperation and Development (OECD). For the sake of consistency with subsequent computations, mean dollar exchange rates in 1970 and 1972 $(i\mathrm{.}e\mathrm{.},\frac{{er}_{1970}+{er}_{1972}}{2})$ were used as a base in computing the (percentage) exchange rate variations.

Table 1.

Sixteen Oecd Countries: Percentage Changes in U.S. Dollar Exchange Rates, 1970-72, and Indices of Industrial Production and of Consumer Prices, 1972

Sources: Exchange rate changes were calculated from the Federal Reserve Bulletin, Vol. 59 (January 1973), p. A93. Consumer price indices were taken from Organization for Economic Cooperation and Development (OECD), Main Economic Indicators (August 1975). Indices of industrial production were taken from United Nations, Monthly Bulletin of Statistics, Vol. 30 (April 1976), Table 10, pp. 22-33.

Table 1.

Sixteen Oecd Countries: Percentage Changes in U.S. Dollar Exchange Rates, 1970-72, and Indices of Industrial Production and of Consumer Prices, 1972

	Percentage	1972	1972
	Change in	Index of	Consumer
	U.S. Dollar	Industrial	Price
	Exchange Rate	Production	Index
Country	(1970-72)	(1970=100)	(1970=100)
Australia	7	108	112
Austria	11	114	111
Belgium	12	109	110
Canada	5	113	108
Denmark	8	111	112
Finland	1	114	114
France	9	112	112
Germany, Fed. Rep.	13	106	111
Italy	7	104	111
Japan	17	110	111
Netherlands	12	111	116
Norway	8	110	114
Sweden	9	104	109
Switzerland	12	104	114
United Kingdom	4	102	117
United States	—	108	108

Sources: Exchange rate changes were calculated from the Federal Reserve Bulletin, Vol. 59 (January 1973), p. A93. Consumer price indices were taken from Organization for Economic Cooperation and Development (OECD), Main Economic Indicators (August 1975). Indices of industrial production were taken from United Nations, Monthly Bulletin of Statistics, Vol. 30 (April 1976), Table 10, pp. 22-33.

View Table

Table 1.

Sixteen Oecd Countries: Percentage Changes in U.S. Dollar Exchange Rates, 1970-72, and Indices of Industrial Production and of Consumer Prices, 1972

	Percentage	1972	1972
	Change in	Index of	Consumer
	U.S. Dollar	Industrial	Price
	Exchange Rate	Production	Index
Country	(1970-72)	(1970=100)	(1970=100)
Australia	7	108	112
Austria	11	114	111
Belgium	12	109	110
Canada	5	113	108
Denmark	8	111	112
Finland	1	114	114
France	9	112	112
Germany, Fed. Rep.	13	106	111
Italy	7	104	111
Japan	17	110	111
Netherlands	12	111	116
Norway	8	110	114
Sweden	9	104	109
Switzerland	12	104	114
United Kingdom	4	102	117
United States	—	108	108

Sources: Exchange rate changes were calculated from the Federal Reserve Bulletin, Vol. 59 (January 1973), p. A93. Consumer price indices were taken from Organization for Economic Cooperation and Development (OECD), Main Economic Indicators (August 1975). Indices of industrial production were taken from United Nations, Monthly Bulletin of Statistics, Vol. 30 (April 1976), Table 10, pp. 22-33.

Our interest lies in estimating the pass-through effects of these exchange rate changes. In order to illustrate the method employed, we begin by using the United States as an example; that is, we ask what effect did the dollar devaluation of 1971 have on the prices (and quantities) of U. S. imports and exports? Each of the 15 main OECD countries revalued its currency relative to the dollar between 1970 and 1972. Corresponding to each exchange rate adjustment, there was a change in the dollar prices of U. S. imports from, and the foreign currency prices of U. S. exports to, the partner country in question. But the observed (bilateral) price changes cannot be fully attributed to the revaluations. Caused by general inflation, differential monopoly power, and other factors, they would have occurred—at least in part—even in the absence of the exchange adjustments. To isolate the effect of the dollar devaluation on foreign trade prices, one must estimate the hypothetical price change that would have occurred anyway. The difference between the hypothetical and the actual price change is the effect of the exchange adjustment on foreign trade prices.

The control country approach is used to estimate the hypothetical price change for each revaluing country. Assume, for example, that Finland and Norway are identical countries in all respects (including the bundle of goods they export to the United States), except that Norway revalued by 8 per cent relative to the dollar between 1970 and 1972, while Finland did not change its dollar exchange rate. If the U.S. (dollar-denominated) import price index from Norway increased 16 per cent, while that from Finland increased 14 per cent, the effect of Norway’s revaluation on its dollar export prices to the United States is taken to be (16 per cent - 14 per cent =) 2 per cent, and its pass-through effect on U. S. imports is then (2 ÷ 8 =) 25 per cent. It other words, the export price change of Finland (the control, or c, country) is used as a proxy for the hypothetical price change of Norway (the i country in which follows) in the absence of revaluation. Schematically (with all figures representing percentages):

				Estimated
	Norway	Finland	Differential	Pass-Through
${\dot{ER}}^{\$}$	+8	0	8
${\dot{P}}_M^{U\mathrm{.}S\mathrm{.}}$	+16	+14	2	25

View Table

				Estimated
	Norway	Finland	Differential	Pass-Through
${\dot{ER}}^{\$}$	+8	0	8
${\dot{P}}_M^{U\mathrm{.}S\mathrm{.}}$	+16	+14	2	25

where $\dot{E}{R}^{\$}$ is the percentage change in the dollar exchange rate between 1970 and 1972, and ${\dot{P}}_M^{U\mathrm{.}S\mathrm{.}}$ is the percentage change in U.S. import prices from each country over the same period. In the general notations that follow, the country under investigation (the United States in our example) will be referred to as the K country. Estimates will be constructed for eight such countries. In the study of each K country, there will be 3 to 5 c countries (the criteria used for their selection will be spelled out below), and 9 to 11 i countries (defined as all countries other than the c country selected that trade with the K country), for a total of 33 to 45 observations. The estimated pass-through of the K country is an average of these observations, with the averaging procedure to be described in Section II.5 below.

2. BASIC ESTIMATION FORMULA FOR BILATERAL IMPORT PASS-THROUGH

For greater generality, we introduce the following notation: $\mathit{iE}{\dot{R}}^K$ and $\mathit{cE}{\dot{R}}^K$ are the 1970-72 percentage changes in the exchange rates of the i and the c countries, respectively, relative to country K. These exchange rates are defined as units of the K currency per unit of the i or c currency. For example, if K is the United States, the expressions refer to dollar exchange rates, defined as cents per unit of foreign currency. (In what follows, we shall use the U.S. as an example of the K country whenever an example is needed for the purpose of clear exposition.) $c{\dot{P}}_X^K;i{\dot{P}}_M^K$ are the 1970-72 percentage changes in the prices of the K country’s imports from the c and the i countries, respectively (on products imported into country k from both c and i countries). ${\dot{P}}_x$ and ${\dot{P}}_d$ refer to the 1970-72 percentage changes in export prices and domestic prices, respectively.

For any country K whose currency was devalued with respect to country i:

That is, the percentage change in the prices of country K’s imports from country i is a fraction α_i of the percentage change in i’s exchange rate relative to K, where α_i is the measure of the pass-through. Considering $c^{{\dot{P}}_M}$ as a proxy for what _iP_M would have been in the absence of exchange rate adjustment, the pass-through effect of the i country’s revaluation is estimated as:

In all calculations, the percentage changes of prices and exchange rates are average 1970 to average 1972, with mean dollar exchange rates and mean price levels in 1970 and 1972 used as bases; the commodity bundles included in each comparison of the K country imports from the c and the i countries are those common to the two exporting countries, with the matching done at as high a level of disaggregation as possible.

But this procedure assumes that the export prices in the control country are unaffected by the revaluation of the K country’s currency. In fact, the rise in $i^{{\dot{P}}_M^K}$ would generate at least some switch in the K country demand from country i to the control country, and would therefore raise $c^{{\dot{P}}_M^K}$. Although this effect is unlikely to be large¹⁴ (especially over a relatively short time span), it biases the estimate downward. The result of the calculation must, therefore, be regarded as a lower bound, with the actual pass-through effect being somewhat higher. Such a bias is inherent in the control country, or the control group, approach.

Equation (2) is the basic estimating formula for the bilateral import pass-through. For each K country there are 3 to 5 control countries and 9 to 11 i countries, yielding a total of 33 to 45 αi bilateral observations. The import pass-through of K is the average of these observations; this will be discussed in Section II.5. But, prior to that, the next section will dwell on various biases in the formula and will offer adjustments, as well as alternative formulas.

3. BIASES, ADJUSTMENTS, AND ALTERNATIVE FORMULAS

In essence, then, the control country represents developments that affect the prices of the i country’s exports to the K country in the absence of revaluation. In addition to being an important trading partner of the K country, the “ideal” control country (or countries) should fulfill two conditions: its exchange rate relative to the K country should remain unchanged; and it (they) should adequately represent economic conditions in the i countries. In particular, the control country (countries) should—on the average—experience the same rate of growth and of domestic inflation as the i countries over the period being considered. It is not always possible to find control countries that meet these conditions.

Starting with the first condition and using the United States as an example of a K country, it may be seen in Table 1 that each of the 15 countries revalued to some degree against the dollar. As a consequence, it was necessary to select as control countries those that revalued least. This led to the selection of Finland, Canada, and the United Kingdom.¹⁵ Each was used separately to estimate the pass-through of a revaluation of each of the 12 other countries, yielding 36 α_i bilateral observations. But the use (as a control) of a country that revalued against the dollar requires the introduction of two alternative assumptions, with the results establishing lower-bound and upper-bound estimates of the pass-through effect.

Consider the following illustrative scheme with respect to the dollar exchange rates:

	Germany, Fed. Rep.	United Kingdom
${\dot{ER}}^{\$}$	13	4
${\dot{P}}_M^{U\mathrm{.}S\mathrm{.}}$	16	12

View Table

	Germany, Fed. Rep.	United Kingdom
${\dot{ER}}^{\$}$	13	4
${\dot{P}}_M^{U\mathrm{.}S\mathrm{.}}$	16	12

Under the first assumption, none of the United Kingdom’s (export) price increase was owing to revaluation of the pound; the entire 12 per cent rise would have occurred anyway. The hypothetical increase in the Federal Republic of Germany’s export price in the absence of mark revaluation is 12 per cent, and the pass-through is estimated at $(\frac{16-12}{13}=)$ 31 Percent. This is the estimate generated by equation (2). Presumably, the number subtracted from the German price increase is excessive, and the estimate should be regarded as a lower bound. The closer the (dollar) exchange adjustment of the control country is to zero, the closer the lower-bound estimate is to the true estimate. As a second alternative, assume that the control country (the United Kingdom) experienced a 100 per cent pass-through. Then only 8(12 – 4) per cent of the U. K. price rise would have occurred in the absence of a sterling revaluation. The German pass-through is then estimated at $(\frac{16-8}{13}=)$ 61 per cent. This is a reasonable assumption to generate the upper-bound (α_i,u) estimate

As the revaluation of the control country relative to the K country approaches zero, the lower and upper bounds converge. Because of the downward biasdnherent in the control country approach (discussed previously), the actual estimate is probably closer to the upper bound than the lower bound.

While the selection of control countries is dictated mainly by the stability of their exchange rates relative to the K country, the second condition requires that “on the average” they be representative of the i countries with which they are to be compared. This is one reason for selecting several (rather than one) control countries and averaging the results. However, whenever there is a difference between the average inflation rates of the c countries and the i countries,¹⁶ an adjustment is deemed necessary.

Specifically, the change in the i country’s export prices P_x is a function of the change in its exchange rate and the change in its domestic prices:

Since our interest centers on the effect of the exchange rate, we wish to hold constant the impact of domestic inflation. But ${\dot{P}}_d$ may be different in the control and the i country. To account for the difference, the price index of K country imports from the control country was adjusted in each comparison (with an i country) by the difference between the two countries’ 1972 domestic prive indices (1970= 100).

where $i{\dot{P}}_d\text{and}c{\dot{P}}_d$ are the domestic price indices of the i and K countries, respectively. An ideal index for this purpose would be one that included all potentially traded goods but excluded imports. However, since such an index was not available, the consumer price index was employed.¹⁷ The assumptions underlying this adjustment are that domestic price changes are fully reflected in the prices of export goods, and that the effect of exchange adjustment is superimposed upon them. Although they are somewhat unrealistic, the bias introduced by these assumptions is probably rather small, for the following reasons: first, because the discrepancy between domestic prices and export prices is likely to be similar in the two countries; and second, because the differences between $i{\dot{P}}_d\text{and}c{\dot{P}}_d$ (and, therefore, the adjustments themselves) are usually small. In other words the control countries were usually representative of the i countries with respect to their domestic rates of inflation (as well as their growth rates). Equation (4) was inserted appropriately into equations (2) and (3) to generate estimates adjusted for differential inflation rates. But, in most cases, these did not differ from the unadjusted figures.

4. ESTIMATION FORMULAS FOR BILATERAL EXPORT PASS-THROUGH

A symmetrical analysis was undertaken with respect to exports. Again, using the United States as the K country in a schematic illustration, assume that U. S. export prices $\left({\dot{P}}_X^{U\mathrm{.}S\mathrm{.}}\right)$ of identical bundles of goods destined for the control country and the i countries changed as follows from 1970 to 1972:

	i Country	c Country
${\dot{ER}}^{\$}$	10	0
${\dot{P}}_M^{U\mathrm{.}S\mathrm{.}}$	+12	+8

View Table

	i Country	c Country
${\dot{ER}}^{\$}$	10	0
${\dot{P}}_M^{U\mathrm{.}S\mathrm{.}}$	+12	+8

Following the previous reasoning, the 10 per cent revaluation in the i country caused U. S. export prices to rise by (12 - 8 =) 4 per cent. Thus, (4 ÷ 10 =) 40 per cent of the revaluation was absorbed by a rise in U. S. dollar export prices, and the pass-through effect is estimated at 60 per cent. More generally, the pass-through effect of the i country’s revaluation relative to the K country is estimated from the following equation:

As in the case of imports, a downward bias is introduced into $c^{{\dot{P}}_X^K}$ because of substitution in demand for K country products from the control country to the i country. The resulting estimate must be regarded as a lower bound.

When the control country experiences a revaluation relative to the K country, equation (5) becomes the upper-bound estimate, while the lower-bound estimate is given by:

As in the case of imports, a price adjustment was deemed necessary. For the change in the prices of the i country’s imports from the K country (or in the prices of the K country’s exports to the i country) is a function of the change in the exchange rate, as well as of the domestic rate of inflation in the i country. In employing the control country approach, we wish to hold the second factor constant, so as to concentrate on the first. Consequently, similar price adjustments were introduced into equations (5) and (6) as in the case of imports; that is, the $c^{{\dot{P}}_X^K}$ was adjusted upward by $c{\dot{P}}_d-i{\dot{P}}_d\mathrm{.}$.

5. AVERAGING OF BILATERAL α_i OBSERVATIONS FOR EACH K COUNTRY

No restrictions are imposed on the individual comparisons; only average results of all bilateral i country observations for each K country are relied upon in arriving at the estimated pass-through. In addition to unweighted averages, a weighted average was calculated in two steps. First, an unweighted mean of each α_i estimate (generated by the three to five control countries) was computed for each of the i countries. These were then averaged by assigning each i country a weight proportional to the K country’s imports from it (or export to it) in 1970 and 1972. We have

Yet another method of averaging the individual αi observations is by estimating a weighted regression of the form

where Ṗ represents the differential percentage price change between the i and the control country $(i{\dot{P}}_M^K-c{\dot{P}}_M^K)$, E

View Expanded

• Download Figure
• Download figure as PowerPoint slide

View Full Size

• Download Figure
• Download figure as PowerPoint slide

• Download Figure
• Download figure as PowerPoint slide

To illustrate this method, consider the resulting equation for Japan’s imports

when $\dot{ER}$ = 15 per cent, Ṗ = 11.63, and the pass-through is estimated at (11.63 ÷ 15 =) 77.5 per cent. Similarly, for the United States, this method yielded an import pass-through of between 35 per cent (lower bound) and 70 per cent (upper bound; and an export¹⁹ pass-through of 100 per cent.

In most cases, the various methods of arriving at the estimate yielded similar (though not identical) results. Consequently, after several alternative estimates are discussed for the United States, only one (best) estimate will be presented for each of the other countries.

6. estimation of trade volume and elasticities

With respect to the quantity Q of trade, the control country approach suggests that the percentage change in the K country’s imports from (exports to) the control country (whose exchange rate relative to the K country did not change) serves as a proxy for the percentage change in the K country’s imports from (exports to) the i country in the absence of exchange rate adjustment. The estimated effects of an i country revaluation on the K country’s imports from, and exports to, the i country would then be (respectively):

These would then be aggregated over all i countries using a weighted average as indicated earlier for prices. However, because of expected substitution between sources of K country imports (and between destinations of its exports) caused by the exchange rate adjustments, such estimates must be regarded as upper bounds.

Implied in the quantity and price figures are estimates of the import demand elasticity for each K country and of the elasticity of world demand for each K country’s exports.

Finally, the data generated will be used to estimate the elasticity of substitution between various OECD supply countries in each K country market. In most cases, these data contain 30 to 40 paired source country observations of percentage changes in import prices and quantities between 1970 and 1972, and between 1970 and 1973. Each observation pertains to a common bundle of goods exported by both countries in the pair. A regression of quantity change on price change (cross section)—both in percentage terms—would yield an estimate of the elasticity of substitution in each K country market. The specific form of the regression is:

for each pair of suppliers i and j (covering matched products), where b is the estimated elasticity.

1. period covered

All observations employed in this paper relate to percentage changes between 1970 and 1972. These two years are well suited for this study’s purposes. The economies of all OECD countries were on a steady expansion path, with the indices of industrial production (1970 = 100) of most European countries and Japan in the 105-112 range in 1972. Domestic prices of the OECD countries advanced at a roughly similar and relatively moderate pace. With 1970 equal to 100, the consumer price indices of most industrial countries were in the 108-113 range in 1972, although greater variations were exhibited in their indices of unit labor costs in manufacturing.²⁰ For some of the countries these indices show a quantum jump in 1973; in all countries, an even greater jump occurred in 1974, the year of double digit world-wide inflation. This behavior is also evident in the export and import price indices. The point of this discussion is that in 1970-72, and to some (but a lesser) extent in 1970-73, the exchange rate adjustments were major developments, whose effect was unlikely to be swamped by that of the other price changes. The same can hardly be said of 1970-74. The 1970-72 comparison is therefore considered reliable.

2. DATA SOURCES

Because the study focuses on the behavior of unit value and quantity (or volume²¹), it requires the use of a set of commodity trade statistics that are as disaggregated as possible. Only high disaggregated products are homogeneous enough for the quantity information to be meaningful. For the United States (as the K country), this study draws on the U. S. Treasury’s Trade Statistics U.S.A. data tapes for 1970 and 1972. Based on the seven-digit Schedule A (import) and Schedule B (export) classifications, they include value and volume data for individual products imported to, and exported from, the United States by source country and destination country, respectively. For all other countries, the somewhat more aggregated (four-digit or five-digit Standard International Trade Classification (SITC) of commodities) OECD trade data tapes were used as the primary data source. To check on the comparability of the data, the pass-through effect of U. S. imports was estimated twice in the same manner using both sets of data. The results were, indeed, very similar: in the 35-65 per cent range using the U. S. census data, and in the 40-70 per cent range using the OECD data.

3. COMPUTATIONAL PROCEDURES

Unit values (value ÷ volume) were calculated for every commodity imported into (exported by) the K country from (to) each of the industrial countries show in Table 1 except Australia (which was included only when the United States was designated the K country) for the years 1970 and 1972. These countries account for most of world trade in manufactured products. Trade flows that contained no volume information (primarily because even the five-digit or seven-digit product was too heterogeneous) were rejected—that is, not included in the study. Next, the changes from 1970 to 1972²² in (a) volume (or quantity) and (b) unit value were computed for every trade flow. Each change was then converted into a percentage change, using the average of 1970 and 1972 as a base.²³ All subsequent computations involving averaging and aggregation relate strictly to these percentages. The objective was to obtain paired source countries (for K country imports) and paired destination countries (for K country exports) for comparisons of average percentage changes in unit values and quantities that in each case covered only those products that were exported by (or to) both countries in the pair. This made comparisons possible between each i country and each control country with a commodity bundle common to both of them. The procedure is best described by the use of one illustration—U. S. (the K country) imports from the pair of source countries made up of Canada and the Federal Republic of Germany.

All disaggregated U. S. imports from both countries in the pair were matched by their seven-digit number. Commodities not imported from both countries were excluded from the comparison. For the accepted products (i.e., those imported into the United States from both source countries in 1970 and 1972), an average percentage change in unit value and volume was calculated over all commodities. In calculating the average for each country, the value of U. S. imports of that product (from the country) in the two years combined was used as a weight:

Similar indices were calculated for Canada.²⁴

A similar procedure was followed for each pair of source countries of U. S. imports. However, the commodity coverage differed between pairs, because in each case only products exported by both countries in the pair were included. Paired comparisons of quantity and unit values were also made for U. S. exports, with respect to the countries of destination. The same procedure was followed with respect to the trade of each K country. However, for the United States (only), in addition to averages covering all commodities, averages pertaining to all manufactures (SITC 5-8), and to certain groups of manufactures were calculated. It is to these average percentage changes that the formulas developed in Section II apply.

1. PASS-THROUGH EFFECT

Of the 15 countries for which indices of changes in price and quantities were computed, France, Denmark, Norway, and Sweden experienced either no change or only minor changes in their effective exchange rates between 1970 and 1972. Since some of the remaining countries were “used up” as control countries, estimates were prepared for eight industrial countries—the six countries appearing in Tables 2 and 3, plus two small countries discussed in the text. In selecting the countries to be studied, an attempt was made to include countries of different sizes, in terms of their gross national products and international trade volumes, and consequently of varying degrees of monopoly power on the international market. Table 3 presents several estimates for the United States (in addition to a regression-generated result mentioned in Section II.6), while Table 2 presents the “best” estimates for six countries. As can be seen in the second column of Table 2, the control countries differ for each country investigated. Generally speaking, the results display a measure of consistency regardless of which combination of control countries (out of those listed in each case) is used. In order to avoid an impression of undue precision, all pass-through estimates were rounded to the nearest 5 per cent. With two exceptions (listed in footnote 3 of Table 2), the estimates correspond well to theoretical expectations.

Table 2.

Six Industrial Countries: Estimated Effects of Exchange Rate Adjustment on Trade in All Commodities, 1970-72

¹A pass-through on the import side is defined as the percentage of the devaluation (revaluation) translated into an increase (reduction) in domestic prices. On the export side, it is defined as the percentage of the devaluation (revaluation) translated into a reduction (increase) in foreign import prices from the country under study, or 100 per cent minus the percentage change in that country’s domestic export prices.

²Figure appears to be unduly high.

³Volume and estimated elasticity of import demand are biased downward.

Table 2.

Six Industrial Countries: Estimated Effects of Exchange Rate Adjustment on Trade in All Commodities, 1970-72

			Value of
			Trade With Industrial		Percentage Change in		Percentage
		Direction	Countries	Estimated	Effective	Change in	Change in	Implied
K	Control	of	(average for	Pass-	Exchange	Terms of	Volume of	Elasticity
Country	Countries	Trade	1970-72)	Through 1	Rate	Trade	Trade	of Demand
			Billion U. S. dollars	Per cent
United States	Canada	Imports	32	50	−10	−5	−10	−2.00
	United Kingdom	Exports	28	100			+17	−1.00
	Finland
Germany, Fed. Rep.	Belgium, Netherlands, Switzer	Imports	25	60	+6	+3	no discernible effect	…
		Exports	30	902
	land, Austria
Japan3	Germany, Fed. Rep.	Imports	9	80	+15	+9.75	+15	−1.25
	Belgium	Exports	11	85			−14	−1.10
	Switzerland
Canada	United Kingdom	Imports	14	90	+6	+3	+13	−2.50
	Italy	Exports	16	60			−14	−2.30
Belgium	Austria	Imports	11	90	+3	+2	no discernible effect	…
	Netherlands	Exports	12	75
	Switzerland
Italy	France, Denmark,	Imports	11	100	−2	−2	no discernible effect	…
	Norway, Sweden	Exports	11	1002

¹A pass-through on the import side is defined as the percentage of the devaluation (revaluation) translated into an increase (reduction) in domestic prices. On the export side, it is defined as the percentage of the devaluation (revaluation) translated into a reduction (increase) in foreign import prices from the country under study, or 100 per cent minus the percentage change in that country’s domestic export prices.

²Figure appears to be unduly high.

³Volume and estimated elasticity of import demand are biased downward.

View Table

Table 2.

Six Industrial Countries: Estimated Effects of Exchange Rate Adjustment on Trade in All Commodities, 1970-72

			Value of
			Trade With Industrial		Percentage Change in		Percentage
		Direction	Countries	Estimated	Effective	Change in	Change in	Implied
K	Control	of	(average for	Pass-	Exchange	Terms of	Volume of	Elasticity
Country	Countries	Trade	1970-72)	Through 1	Rate	Trade	Trade	of Demand
			Billion U. S. dollars	View Expanded Per cent View Expanded
United States	Canada	Imports	32	50	−10	−5	−10	−2.00
	United Kingdom	Exports	28	100			+17	−1.00
	Finland
Germany, Fed. Rep.	Belgium, Netherlands, Switzer	Imports	25	60	+6	+3	no discernible effect	…
		Exports	30	902
	land, Austria
Japan3	Germany, Fed. Rep.	Imports	9	80	+15	+9.75	+15	−1.25
	Belgium	Exports	11	85			−14	−1.10
	Switzerland
Canada	United Kingdom	Imports	14	90	+6	+3	+13	−2.50
	Italy	Exports	16	60			−14	−2.30
Belgium	Austria	Imports	11	90	+3	+2	no discernible effect	…
	Netherlands	Exports	12	75
	Switzerland
Italy	France, Denmark,	Imports	11	100	−2	−2	no discernible effect	…
	Norway, Sweden	Exports	11	1002

¹A pass-through on the import side is defined as the percentage of the devaluation (revaluation) translated into an increase (reduction) in domestic prices. On the export side, it is defined as the percentage of the devaluation (revaluation) translated into a reduction (increase) in foreign import prices from the country under study, or 100 per cent minus the percentage change in that country’s domestic export prices.

²Figure appears to be unduly high.

³Volume and estimated elasticity of import demand are biased downward.

Table 3.

United States: Estimated Percentage Pass-Through of Dollar Devaluation for Imports and Exports, 1970-72

¹Uses the United Kingdom as control for the large i countries; Finland as control for the small i countries; and Canada as control for the medium-sized i countries.

²SITC denotes Standard International Trade Classification.

Table 3.

United States: Estimated Percentage Pass-Through of Dollar Devaluation for Imports and Exports, 1970-72

		Unweighted		Weighted		Selected	Country1
		Mean		Average		Weighted	Average
Commodity Group	Price Adjustment	Lower bound	Upper bound	Lower bound	Upper bound	Lower bound	Upper bound
		Imports
All commodities	Yes	20	55	35	65	30	60
	No	…	…	25	50	15	35
SITCs 5-8 2	Yes	20	50	30	55	30	55
	No	…	…	30	55	0	25
SITC 5	No	…	…	20	50	…	…
SITC 7	No	…	…	25	55	…	…
		Exports
All commodities	Yes	70	100	85	110	80	100
	No	…	…	95	125	90	115
SITCs 5-8	Yes	80	110	90	115	95	120
SITC 5	No	…	…	85	105	…	…
SITC 7	No	…	…	105	125	…	…

¹Uses the United Kingdom as control for the large i countries; Finland as control for the small i countries; and Canada as control for the medium-sized i countries.

²SITC denotes Standard International Trade Classification.

View Table

Table 3.

United States: Estimated Percentage Pass-Through of Dollar Devaluation for Imports and Exports, 1970-72

		Unweighted		Weighted		Selected	Country1
		Mean		Average		Weighted	Average
Commodity Group	Price Adjustment	Lower bound	Upper bound	Lower bound	Upper bound	Lower bound	Upper bound
		Imports
All commodities	Yes	20	55	35	65	30	60
	No	…	…	25	50	15	35
SITCs 5-8 2	Yes	20	50	30	55	30	55
	No	…	…	30	55	0	25
SITC 5	No	…	…	20	50	…	…
SITC 7	No	…	…	25	55	…	…
		Exports
All commodities	Yes	70	100	85	110	80	100
	No	…	…	95	125	90	115
SITCs 5-8	Yes	80	110	90	115	95	120
SITC 5	No	…	…	85	105	…	…
SITC 7	No	…	…	105	125	…	…

¹Uses the United Kingdom as control for the large i countries; Finland as control for the small i countries; and Canada as control for the medium-sized i countries.

²SITC denotes Standard International Trade Classification.

We start with the estimated pass-through in the United States that is shown in Table 3 and in the first row of Table 2. While the estimates vary, depending on the commodity coverage and on the weighing procedure used in calculating the averages, the figures do impart a distinct impression. For U. S. imports, the pass-through effect is between 30 and 55 per cent (somewhat less than that for imports of chemicals). Considering the downward bias imparted to these figures by the control country estimating procedure, the actual estimate is probably close to (somewhat less than) one half. Foreign exporters absorbed over one half of the dollar devaluation by lowering their export prices. Thus, if the effective exchange devaluation of the dollar was about 10 per cent,²⁵ then U. S. import prices increased by nearly 5 percentage points, while foreign export prices declined by more than 5 percentage points.

This is a relatively short-run estimate, where the period of adjustment allowed for is less than a year. The estimate is reasonably consistent with earlier findings concerning the effect of U. S. tariff reductions,²⁶ and used a similar methodology. It suggests that, at least in the short-run, the U. S. import demand curve is roughly similar in slope to the foreign export supply curve facing the United States.

In the case of U. S. exports, the estimates leave an unmistakable impression (again, without attaching undue significance to any individual figure) that the devaluation pass-through was complete or nearly so.²⁷ In other words, U. S. exporters failed to raise their dollar prices as a result of the devaluation; foreign currency prices of U. S. exports declined in proportion to the dollar devaluation. This estimate is consistent with an infinitely elastic (or nearly so) U. S. export supply curve which could result, at least in part, from the small share of exports in the total output of most U. S. industries. In sum, it appears that, in the short run, the U. S. commodity terms of trade deteriorated by about half the proportion of dollar devaluation, or by roughly 5 per cent.

Besides the United States, the Federal Republic of Germany and (to a lesser extent) Japan appear to be “price makers” on the buying side of the international market, with pass-through effects on the import side of 60 and 80 per cent for Germany and Japan, respectively.²⁸ For the remaining three countries in Table 2, the import pass-through rises to between 90 and 100 per cent. Not reported in the table are estimates for Austria and Switzerland, each of which show 100 per cent import pass-through. These results conform to theoretical expectations in terms of their absolute size; they conform especially well in terms of the country ranking.

On the export side, the estimated pass-through ranges from 60-100 per cent. The export pass-through appears to be invariably on the high side, meaning that export prices expressed in domestic currencies did not change much. For the Federal Republic of Germany and Italy, the results appear to be unduly high, and to contain a possible upward bias. The estimates for Canada undoubtedly reflect the importance of the United States as a trading partner of Canada, for they are nearly a mirror image of the U. S. results.

By applying the pass-through estimates in Table 2 to the respective changes in the effective exchange rates, we obtain the changes in the terms of trade of each country owing to the exchange adjustment. The largest positive impact occurred in Japan, while the largest negative effect took place in the United States. In sum, within the three-year time span under consideration, the estimates support the traditional view that (when the product of the supply elasticities exceeds the product of the demand elasticities) devaluation worsens a country’s terms of trade, while revaluation improves them.

These results cast doubt on the strength of the so-called ratchet effect. Foreign exporters to the United States met the revaluation of their currencies by lowering their export prices by more than half the revaluation, more than matching the increase in import prices in the devaluing country. Conversely, exporters to the Federal Republic of Germany (and, to a lesser extent, to Japan) raised their prices by only 40 per cent of the German revaluation.

Similarly, U. S. export prices, expressed in foreign currencies, declined roughly in proportion to the U. S. devaluation,²⁹ while Japanese and Canadian export prices rose proportionately less than their revaluations. Nor is there support for the M-L argument that domestic price increases would fully erode any competitive gain from devaluation, or that currency devaluations constitute the main force propelling the worldwide inflation. On that latter point, additional (though highly tentative) evidence can be gleaned from the data. The general impression one receives is that domestic price movements in the countries of destination influence the pricing policies of foreign exporters.

Export prices of any exporting country tended to rise more, or decline less, in countries of destination that were plagued by relatively high inflation, than in countries with a relatively stable price level, where local competition was presumably stiffer. This suggests that the international transmission of inflation following exchange rate adjustments depends on domestic inflationary pressures in the importing countries, rather than the converse. However, it should be stressed that this is a short-run analysis, whereas the thesis attributed to Mundell and Laffer (as well as the monetary approach to the balance of payments) is said to hold only in the long run.

2. TRADE VOLUME AND ELASTICITIES

A marked effect on the volume of trade was observed in three of the six countries investigated. For Belgium and Italy, the small size of the exchange adjustment may account for the absence of change in volume, but for the Federal Republic of Germany, the explanation must be sought elsewhere. In the remaining three countries, the results conformed to theoretical expectations. For the United States, imports are estimated to have decreased by 10 per cent and exports to have increased by 17 per cent as a result of the devaluation. Given a 10 per cent dollar devaluation and a 50 per cent pass-through on the import side, this yields a U. S. import demand elasticity of (10 ÷ 5 =) 2.³⁰ With a 100 per cent pass-through on the export side, the estimated foreign elasticity of demand for U. S. exports is 1.7. Similarly, the import demand elasticities are 1.25 and 2.5 for Japan and Canada, respectively, while the foreign demands for their exports have elasticities of 1.1 and 2.3, respectively. Not shown in the table are estimated elasticities of Austria’s demand for imports of -0.9, and of foreign demand for Austria’s exports of -3.5. Finally, it should be noted that the volume results for Japan, and therefore the implied elasticity of demand, are strongly biased downward because the control countries devalued by 4 to 5 per cent relative to the yen. Most elasticities generated by exchange rate changes may be lower than tariff elasticities, because traders may consider exchange rate changes reversible in making their pricing decisions, while the General Agreement on Tariffs and Trade rules make tariff changes irreversible for the most part. However, most volume changes shown here are of sufficient magnitude to meet the Marshall-Lerner stability conditions.

Corresponding to the calculation of the average differential change (1970-72) in unit values of each K country’s imports from pairs of source countries is a similar calculation of the differential percentage change in the volume of imports from the same pair over the same period, including an identical bundle of goods. Since changes in K country income and domestic prices can be assumed to affect both source countries to the same degree, the differential change in quantity is attributable to the differential change in foreign prices.

For U. S. imports, the linear regression (of percentage change quantity differentials on percentage change price differentials) fitted to the paired observations is

where the figures in parenthesis represent t-statistics. The size of the price coefficient and its level of significance (as well as the R² and the D-W coefficient) remain virtually unchanged if each observation is weighted by the share of U. S. imports from all industrial countries.

The elasticity of substitution in the U. S. market for all commodities is thus estimated to be −1.8. This result corresponds well to short-run elasticity estimates arrived at by other methods, and falls considerably short (are about half the size) of estimated long-run elasticities³¹ obtained when a ten-year period of adjustment (of quantity to price) was allowed for.

Table 4 presents estimates of substitution elasticities for several countries based on price and quantity changes between (a) 1970 and 1972 and (b) 1970 and 1973. For the United States (for period (a) only), estimates are also provided for some first-digit SITC groups.

Table 4.

Eleven Industrial Countries: Estimated Substitution Elasticities, 1970-72 and 1970-73¹

¹Empty cells indicate results that are either statistically insignificant or that have the wrong sign. The estimates for Belgium yielded an incorrect (positive) sign.

²SITC denotes Standard International Trade Classification.

Table 4.

Eleven Industrial Countries: Estimated Substitution Elasticities, 1970-72 and 1970-73¹

Country	Commodity Group2	1970-72	1970-73
Canada	All commodities	−1.0	−0.5
Denmark	”	−1.1	−1.4
Finland	”	−0.8
France	”	−1.0	−0.4
Germany, Fed. Rep.	”	−0.5	−0.5
Italy	”	−1.2	−1.3
Japan	”	−0.8	−0.2
Netherlands	”	−0.4
Norway	”	−0.8
Sweden	”		−0.6
United States	”	−1.8	−0.8
”	SITC 5	−1.2	…
”	SITC 6	−0.5	…
”	SITC 7	−1.5	…
”	SITC 8	−1.8	…

¹Empty cells indicate results that are either statistically insignificant or that have the wrong sign. The estimates for Belgium yielded an incorrect (positive) sign.

²SITC denotes Standard International Trade Classification.

View Table

Table 4.

Eleven Industrial Countries: Estimated Substitution Elasticities, 1970-72 and 1970-73¹

Country	Commodity Group2	1970-72	1970-73
Canada	All commodities	−1.0	−0.5
Denmark	”	−1.1	−1.4
Finland	”	−0.8
France	”	−1.0	−0.4
Germany, Fed. Rep.	”	−0.5	−0.5
Italy	”	−1.2	−1.3
Japan	”	−0.8	−0.2
Netherlands	”	−0.4
Norway	”	−0.8
Sweden	”		−0.6
United States	”	−1.8	−0.8
”	SITC 5	−1.2	…
”	SITC 6	−0.5	…
”	SITC 7	−1.5	…
”	SITC 8	−1.8	…

¹Empty cells indicate results that are either statistically insignificant or that have the wrong sign. The estimates for Belgium yielded an incorrect (positive) sign.

²SITC denotes Standard International Trade Classification.

What is generally striking about the estimates is their relatively small size. With the exception of the United States, practically all countries have a substitution elasticity of one or less, even when the period allowed for is three years (1970–73).