Intercountry Cost and Price Comparisons

THE RECENT ADJUSTMENT of some major exchange rates has led to renewed interest in the “price aspect” of the balance of payments adjustment process. The need to assess the role of relative prices is probably increasing with the rapid integration of the world economy and the growing interdependence of economic policies. Unfortunately, given the quality of the existing data, precise measurement of shifts in relative costs and prices involves numerous statistical and methodological pitfalls. This paper surveys some of the main methodological, statistical, and index number problems involved in intercountry comparisons of costs and prices. A few of the cost-price indices most frequently used in analyses of international competitiveness are presented in Appendix I. These data, covering the manufacturing sectors of ten selected industrial countries, have been collated on a quarterly basis from 1961 through 1970.

Abstract

THE RECENT ADJUSTMENT of some major exchange rates has led to renewed interest in the “price aspect” of the balance of payments adjustment process. The need to assess the role of relative prices is probably increasing with the rapid integration of the world economy and the growing interdependence of economic policies. Unfortunately, given the quality of the existing data, precise measurement of shifts in relative costs and prices involves numerous statistical and methodological pitfalls. This paper surveys some of the main methodological, statistical, and index number problems involved in intercountry comparisons of costs and prices. A few of the cost-price indices most frequently used in analyses of international competitiveness are presented in Appendix I. These data, covering the manufacturing sectors of ten selected industrial countries, have been collated on a quarterly basis from 1961 through 1970.

THE RECENT ADJUSTMENT of some major exchange rates has led to renewed interest in the “price aspect” of the balance of payments adjustment process. The need to assess the role of relative prices is probably increasing with the rapid integration of the world economy and the growing interdependence of economic policies. Unfortunately, given the quality of the existing data, precise measurement of shifts in relative costs and prices involves numerous statistical and methodological pitfalls. This paper surveys some of the main methodological, statistical, and index number problems involved in intercountry comparisons of costs and prices. A few of the cost-price indices most frequently used in analyses of international competitiveness are presented in Appendix I. These data, covering the manufacturing sectors of ten selected industrial countries, have been collated on a quarterly basis from 1961 through 1970.

I. Methodology

The manufacturing sector as a proxy for the export and import-competing sectors 1

Most foreign trade models have adopted a high degree of aggregation—identifying an export, import-competing, and purely domestic sector of an economy. It is, of course, recognized that according to the law of comparative advantage discussions of competitiveness should be confined to units as small as specific commodities or individual firms. However, given the nature of existing data, this level of disaggregation is not practical. The manufacturing sector as a whole has conventionally been used as a proxy for the export and import-competing sectors of industrial economies, and international comparisons of prices and costs usually refer to manufactured products. This high level of aggregation entails certain problems of a theoretical nature, concerning the validity and relevance of the purchasing power parity theory (see below). The rationale for using the manufacturing sector as a proxy for the export and import-competing sectors is quite obvious: manufactured products, for which data are easily obtained and published regularly, account for more than two thirds of total trade. The service sector enters into international trade only on a rather indirect basis. Moreover, trade in agricultural products remains heavily influenced by government policy toward quotas, subsidies, and special preference systems. While it is true to say, from a theoretical point of view, that the manufacturing sector is too broad an aggregate, masking important intrasector or interindustry shifts over time, it is also true, from the point of view of coverage, that the manufacturing sector is too narrow to catch certain important shifts in comparative advantage. For example, most market share analyses indicate a weakening U. S. competitive position in manufactured goods over the postwar period. However, over this period, Europe’s industrial base has grown much faster than that of North America, so that Europe’s comparative advantage in manufactured products may well have strengthened. On the other hand, North America’s comparative advantage may have shifted to agricultural products. Consequently, a portion of the losses of North America’s share of its market in manufacturing trade could well reflect structural shifts in comparative advantage that are not caught in an analysis of manufacturing trade alone.

The purchasing power parity theory of exchange rates 2

There have, of course, been many factors underlying disturbances of international equilibrium. The causes of these disturbances have typically been separated into three categories. They are, first, changes on the monetary side, in the form of differential rates of inflation between countries; second, changes on the real side, in the form of unequal rates of technical change and growth of productive capacity, output, income, and productivity; and, finally, capital movements. It should be noted at the outset that for analytical purposes the purchasing power parity theory has been developed almost exclusively with respect to this first set of monetary factors.

The concept of purchasing power parity provides the theoretical underpinnings for almost all intercountry comparisons of costs and prices. Interest in the purchasing power parity has had its ebbs and flows over the past 50 years. Despite the obvious limitations of partial equilibrium analysis, the theory has survived, and over the past several years the adjustment of some major parities has stimulated interest in relative price differentials and the factors underlying them.

The purchasing power parity theory has been expounded in two principal forms—appropriately called the “absolute” and the “relative” versions of the theory. According to the absolute version of the theory, under fixed exchange rates the ratio between exchange rates at any moment in time must reflect the purchasing power parity between currencies, as long as parity is calculated by reference to price indices for identical bundles of commodities.3 The absolute version of the purchasing power parity theory has usually been rejected out of hand by the vast majority of economists mainly because absolute comparisons of cost and price levels between countries are regarded as either difficult or impossible to make, or because the concept of a general “cost or price level” is not an operational one.4

While the absolute interpretation of the purchasing power parity theory seems to be unsatisfactory, the “relative” interpretation of the doctrine can lay some claim to respectability. This interpretation of the theory can be most easily grasped in a comparative statics framework. Beginning with an initial “equilibrium” situation, the relative interpretation of the theory asserts that, over the comparison period, changes in relative price levels indicate the required readjustment in exchange rates to re-establish an equilibrium situation.

One conceptual difficulty with the purchasing power parity theory is that it does not fit neatly into the simplified framework of classical trade theory. For example, in most abstract trade models using two goods and two countries, the usual assumptions are made that commodities are perfect substitutes and that competitive conditions obtain in all markets. Under these circumstances, relative prices between countries can shift when demand and supply elasticities differ or when market imperfections, such as market ignorance, arise. While these factors should not be overlooked, the reliance upon them as the sole factors underlying changes in relative prices is not attractive either theoretically or empirically. A preferable approach is to recognize explicitly that products are generally imperfect substitutes, i.e., they are differentiated not only by kind but also by provenance. According to this procedure, French machinery, Japanese machinery, French chemicals, and Japanese chemicals are four distinct products, in the sense that they are competing but imperfect substitutes of varying degrees. Not only is each good differentiated in terms of substitutes but also each good can be assumed to be differentiated (from the buyer’s viewpoint) according to the suppliers’ place of production.5

Explanations of shifts in relative prices between countries are more realistic once one makes allowance for international trade in imperfect substitutes, and especially when the existence of a large domestic non-trading sector is recognized. Under these circumstances, differential price developments both within and between countries become plausible. For example, the domestic price level can rise more in a country experiencing cost inflation than in a noninflating country. Likewise, exporters in the inflating country may raise their export prices as a result of cost inflation, while exporters of similar goods in the noninflating country may keep their prices stable. Under fixed exchange rates, a comparison of price indices covering similar exportable and/or domestic goods will usually show that the purchasing power parity of the inflating country has fallen relative to the noninflating country. As a result, the exchange rate between the two countries, as long as it remains unaltered, will cease to be an equilibrium rate. This is how the theory would work under ideal circumstances; however, as is outlined below, there are many qualifications to these conclusions.

While the purchasing power parity theory has an important element of truth, it should be borne in mind that the theory is oversimplified. Once one makes allowances for a large domestic nontrading sector, the consumption-production relationships between export, import, and non-traded goods and prices become highly complex.6 For example, this opens up the possibility of differential productivity trends and discriminatory pricing patterns between the domestic and export sectors.7 Further, once one abandons the haven of perfect substitutes and competitive conditions, the theory of price determination rests upon vague nonquantitative institutional factors rather than market mechanisms. Taking these two qualifications into account, purchasing power parity calculations are subject to a good deal of uncertainty. First, does the existence of a large nontrading sector vitiate purchasing power calculations, which are typically confined to cost-price developments in the goods-producing sector? Second, since observed prices under imperfect competition are not equilibrium prices, can one use comparisons of these price changes as a legitimate basis for judging a country’s underlying international competitiveness?

The purchasing power parity theory has been developed almost solely within a monetary context. It is most applicable when the direction of causation runs directly from changes in money supply to changes in prices, ceteris paribus. This relationship, which was initially stated by the English economists Hume, Ricardo, and Thornton, also appears to be the basis of the classical price adjustment mechanism of the gold standard. It is apparent, however, that it is impossible to totally segregate or neglect considerations of such real factors as structural shifts in consumer taste, differential rates and biases of technological change, and underlying productivity trends. Perhaps even more important, although the theory can be framed within the context of comparative statics, it is essentially a static, partial equilibrium theory and ill-suited for consideration of the dynamic process of income determination. Under certain circumstances, shifts in real factors as well as the process of income determination can upset international equilibrium and yet leave the equilibrium exchange rate based on purchasing power parity calculations unchanged.8

The preservation of purchasing power parity is neither a necessary nor a sufficient condition for the maintenance of equilibrium exchange rates and/or external balance. However, criticism of the theory can easily go too far. As Haberler points out, the theory is an oversimplified tool of analysis, but, if cautiously used along with other evidence, purchasing power parity calculations can have considerable diagnostic value in periods of serious inflation.9 Further, Haberler and Lutz have argued that monetary factors are the most obvious causes of international disequilibria and, on the whole, are much more important than real factors.10 Hence, in a world characterized by relatively stable, yet differential underlying productivity trends and divergent rates of inflation, the relative interpretation of the purchasing power parity theory may have a fairly wide scope of application—especially over the short term and medium term when real factors can be assumed to be changing rather slowly.

II. Measurement Problems

In analyzing shifts in international competitive positions one would prefer to have either indices of actual export prices or unit costs of production, the latter ideally covering the export and import-competing sectors of the economy at an industry level of disaggregation. Unfortunately, except for a limited number of special studies, these data are not available on either a sufficiently current or comprehensive basis. In the absence of adequate international cost-price data most analysts have used export unit values, wholesale prices, and unit labor costs of production—usually covering the manufacturing sector—as a complementary set of indicators for export prices and unit costs of production. Some of the major advantages and defects in these data (Chart 1) are presented in the three subsections that follow.

Chart 1.
Chart 1.

Selected Industrial Countries: Comparative Cost-Price Movements, 1964–70 1

(Index, 1963 = 100; in U.S. dollars)

Citation: IMF Staff Papers 1971, 002; 10.5089/9781451947335.024.A003

* Change in par value.** Strikes1 Indices of competitors are weighted according to the total size and geographic distribution of their trade in manufactures.
Chart (concluded).
Chart (concluded).

Selected Industrial Countries: Comparative Cost-Price Movements, 1964–70 1

(Index, 1963 = 100; in U.S. dollars)

Citation: IMF Staff Papers 1971, 002; 10.5089/9781451947335.024.A003

* Change in par value1 Indices of competitors are weighted according to the total size and geographic distribution of their trade in manufactures.

Before going on to the problems of individual index numbers, one question bears mentioning: Should comparisons be made on the basis of official parities or actual spot exchange rates? Actually, the use of one exchange rate series or another makes little difference for purposes of conversion, since movements within the “band” are typically quite small. The data presented in Appendix I have in fact been converted on the basis of official parities (with the exceptions of the Canadian and German “floats,” for which averages of spot rates were used).11

Export unit values of manufactures

In the absence of consistent export price data, analysts have generally used export unit values of manufactures (i.e., average value per unit of quantity) as the best available indicator of export prices.12 Two principal reasons are usually cited for the use of this indicator: first, it is the closest approximation to an export price index published in most countries; second, it has a superior record to alternative cost-price indicators in explaining export performance.13

There are, however, many criticisms of the use of export unit values as an indicator of international price competitiveness. Different countries use different index number formulas in compiling their data. As a result, identical price changes can be reflected differently by different index numbers. The United Kingdom, Canada, and Sweden use Las-peyres indices, while the Common Market countries use Paasche indices and Japan and the United States use Fisher’s ideal indices. Strictly speaking, the unit value data for each country should be reweighted in a consistent manner—preferably based on Fisher’s ideal formula.14 However, the choice of index number formulas in short-period intertemporal comparisons is relatively immaterial compared with the problems stemming from other recognized inadequacies of measurement.15 A sharp distinction should, however, be drawn between index number problems arising from intertemporal as opposed to intercountry comparisons. Although the choice of different index number formulas for purposes of intertemporal analysis makes little difference, this is not true of intercountry comparisons. Because the structure of relative prices, production, and trade differs substantially between countries, absolute comparisons between different countries based on different index formulas are impossible. As a consequence, intercountry comparisons of relative costs and prices must be confined to changes over time.

The shortcomings of export unit value data as an indicator of international price competitiveness have been explored most thoroughly by Kravis and Lipsey.16 They raise many interesting issues, the three most important of which are discussed in more detail below. Briefly summarized they are as follows: (1) the underlying data on unit values lack sufficient coverage and embody too many erratic fluctuations to be a good proxy for export prices; (2) identical price changes can be represented differently because of shifts in commodity mix; (3) the unit value data do not make sufficient allowances for changes in specification and, hence, do not adequately compensate for improvements in quality. (These limitations are not unique to the unit value data but are common limitations of virtually all cost and price indices.) Although these criticisms have been directed predominantly against U.S. data, they are also relevant—and sometimes even more so—to other countries’ data.

Turning to the first issue, one can judge how well export unit values explain the behavior of export prices for only three countries—the United States, Germany, and Japan, the only countries that have published export prices. With respect to the United States, the National Bureau of Economic Research (NBER) has computed export price data for two specific time periods, 1913–23 and 1953–64. In the earlier period a comparison between unit values and export prices indicates very close agreement.17 Unfortunately, the earlier data have become outdated. Over the postwar period, as the mix of U. S. exports has shifted toward finished and capital goods, there has been a marked divergence between the two series. For the total of all commodity groups covered in the Kravis-Lipsey study for the period 1953–64, the export unit value index increased substantially more than the NBER export price index: 28 per cent versus 15 per cent.18 There were systematic upward biases in the unit value data for iron and steel, nonelectrical and electrical machinery, and metal manufactures. The comparisons for transport equipment showed large fluctuations but no substantial bias (Table 1).19

Table 1.

Comparison of Price Indices: Change in Export Unit Values and International Prices, 1953–64

(In per cent)

article image
Sources: Derived from national statistical sources and data presented in Irving B. Kravis and Robert E. Lipsey, Price Competitiveness in World Trade, National Bureau of Economic Research, Studies in International Economic Relations, No. 6 (Columbia University Press, 1971), Chapter II, pp. 20–24, and Chapter VIII, p. 187.

SITC categories above are weighted by 1963 export values.

The evidence on the relationship between export prices and unit values for the period 1961–69 in Germany and Japan is rather mixed. Unfortunately, we can make comparisons only at a rather aggregate level covering manufactured products in general. For the German data, unit values overstated export prices in the early 1960s. More recently, the relationship has been reversed. On balance, unit values do not appear to have been a very close proxy for German export prices. The Japanese data, on the other hand, indicate a very close fit between unit values and export prices. The two series have fluctuated within a very narrow spread of only 7 percentage points over the past decade.20

Kravis and Lipsey note three principal reasons for the discrepancies between unit values and export prices, viz., deficiencies in coverage and in sampling techniques and inadequate allowances for quality changes, especially with respect to capital goods. These problems have become increasingly important as the proportion of highly finished, technologically sophisticated exports has grown.

Coverage problems

Export unit values for important groups of manufactured products are often simply not available, and the proportion of exports directly represented in the indices has tended to fall. For example, in recent years, from 23 to 25 per cent of finished imports and from 15 to 25 per cent of finished manufactured exports were covered in the U. S. unit value indices.21 Low sample coverage is not necessarily a serious deficiency if the sample is representative. However, the selection of prices for inclusion in the U. S. indices is not made by random methods. Instead, commodities are chosen to obtain the greatest coverage at the lowest cost.22 The sampling problem is especially difficult in the capital goods sector.23 Even the best-covered commodity divisions are, in the opinion of Kravis and Lipsey, inadequately covered;24 the coverage of machinery and equipment is completely inadequate.

In summary, one can conclude that the coverage of the U. S. export unit value index could be greatly improved, especially with respect to capital equipment and the introduction of new products. How serious a problem inadequate sampling is can be tested by comparing movements in covered and uncovered components of the series. As noted above, the test with earlier data, which yielded tolerable errors, is outdated. The data for other countries appear to have broader coverage; however, they probably suffer from many of the shortcomings in coverage embodied in the U. S. data. Finally, since coverage is inadequate specifically in the capital goods classifications, which are more likely to be experiencing rapid technological change, this may introduce a systematic upward bias to observed price increases.

Specification problems and quality change

The major defect of price indices is their failure to account adequately for the improvement of quality. Over the past several years some rather interesting empirical attempts at quantifying quality changes through regression techniques have been made.25 It is standard practice in constructing price indices to adjust for those quality changes to which a price can be attached. For example, the introduction of automatic transmissions on the market at $200 each should not raise the price of automobiles in conventional price indices, even though this feature eventually becomes incorporated as standard equipment.

The problem arises when quality change takes on other dimensions that are impossible or difficult to price.26 Under these conditions a general approach to both time series and interspatial comparisons is the construction of “hedonic” price indices using multiple regression techniques upon known, common specifications, such as weight, horsepower, and revolutions per minute. According to this approach, the regression coefficients can be interpreted as prices for specific quality characteristics and recorded price changes adjusted for identified quality changes. Although this approach is not without difficulties, Kravis and Lipsey have applied this technique to a wide range of capital goods, including locomotives, aircraft engines, diesel and outboard engines, tractors, chemical reactors, automobiles, trucks, and ships. In general, they obtained good results with R¯2s in the vicinity of 0.90. Although this is an improvement in treating quality change, it should be noted that this technique measures quality predominantly with respect to specific, tangible characteristics. Quality changes may, of course, take intangible forms, such as increased durability and more functional design.

Some notion of the importance of quality adjustments can be obtained from Table 1.27 In the U. S. data, a comparison between the NBER international export price index and the export unit value series gives an approximation of the importance of quality change (since the former series explicitly allows for quality change while the latter makes only implicit allowances).28 Unfortunately, the comparison is not unambiguous, since the NBER data are collected from different sources and changes in these indices may reflect a number of factors, such as shifts in commodity composition. Nonetheless, the classifications and coverage are consistent and imply that over the 11-year period 1953–64 the failure to adjust for quality improvement has biased the capital goods and metal products component of the unit value index upward on average by slightly less than 1 per cent a year. Since these quality adjustments apply to roughly three fourths of U. S. manufactured exports, one could infer that the published U. S. unit value index of manufactures is biased upward by some two thirds of 1 per cent a year.29

Given that our interest lies in relative price developments, the question that immediately arises is whether the failure to adjust for quality change introduces a greater bias in the U. S. unit value data than in other industrial countries. On this question there is unfortunately little information. On a priori grounds, one might argue that the quality improvement factor should be of greater importance in other countries than in the United States. This view has usually been developed with reference to the technological gap between Europe and North America and the catching-up process. Briefly stated, the hypothesis is that Europe and Japan have made rapid strides in technology and have reduced the U. S. lead in industrial application. In addition, the growth of U. S. manufacturing and licensing operations abroad, incorporating the latest U. S. technology, would tend to enhance the international diffusion of technology.

Unfortunately, a testing for differential rates of quality adjustment bias between industrial countries is hampered by the shortage of current disaggregated export unit value data. For example, the only readily available published data covering Standard International Trade Classification (SITC) 7 are those of the United Kingdom and those of the United Nations covering all developed countries as a group.30 However, Kravis and Lipsey have constructed detailed commodity “price level” indices for the period 1953–64, based primarily on data from industrial and government contracts and tenders for the United States, the United Kingdom, the European Economic Community (EEC), and Germany (see Appendix II, Table 16). Consequently, one can test for quality adjustment biases within this group of major industrial countries by comparing changes in export unit values with changes in the NBER international export price indices for these countries. (As noted above, the former series makes no explicit allowance for quality change while the NBER indices do.)

The two most striking aspects of these “quality biases” are the close ranking between countries and the relatively uniform pattern within major commodity groups over time (Table 1). Rather surprisingly, one finds that the quality adjustment bias for the United States (9½ per cent) over the period 1953–64 was marginally higher than that of other industrial countries. Perhaps reflecting in part differential rates of technical progress and commodity composition of trade, the German bias (9¼ per cent) was eight tenths of 1 per cent higher than that of the EEC excluding Germany. On the other hand, the U. K. adjustment was lower than the average for industrial countries as a group but about in line with the EEC excluding Germany. However, given the relative uniformity in these biases over time, both between countries and within commodity groups, as well as the errors in measurement of these adjustments—these intercountry biases can be regarded as of roughly the same magnitude and direction.

The very stable pattern of bias within major commodity groups is also striking. This could reflect a variety of factors including a wide international diffusion of technological knowledge. There also appear to have been more pronounced biases in the export unit values for iron and steel and nonelectrical machinery than in the unit values for other categories. Consequently, the bias imparted to individual countries’ total export unit value indices is also dependent upon commodity composition.

In summary, the failure of export unit value indices to account adequately for quality change combined with differential rates of quality improvement are commonly viewed as major problems distorting intercountry comparisons of costs and prices. On a priori grounds one might have expected quality change to have been of greater importance in industrial countries other than the United States. However, the analysis presented above suggests that the biases between the United States, the United Kingdom, Germany, and the EEC appear to have been on average of the same magnitude and direction over the period 1953–64.31 Although these quality adjustments have obvious statistical and conceptual limitations—for example, there could be marked variations in quality change in particular years—the uniformity of these adjustments over time both between major industrial countries and within major commodity groups lends credence to the use of export unit value data. Intercountry comparisons of relative prices for this group of countries based on changes in export unit values need not be systematically vitiated by differential rates of quality change.32

Commodity composition

Even if export unit value indices correctly represented price movements of exports, there would still be ambiguities in their use as indicators of international competitiveness. One difficulty in interpreting price data is that the commodity composition of trade differs from country to country. Hence, it is difficult to say whether apparent shifts in price differentials are a reflection of actual price changes or differences in weights for identical price movements. This is an obvious problem when countries specialize in the production and export of specific commodities whose prices fluctuate more than the average. For example, a country’s export price index could increase because of an autonomous increase in demand for the products in which it specializes. A simplistic comparison on the basis of purchasing power parities could indicate a weakening in competitive position, when the opposite is true. Several authors regard this problem as a serious limitation to the purchasing power parity theory.33

One possible way of compensating for variations in commodity composition would be to use equal weights in the price indices of all countries. There are, however, various objections to this approach. First, sufficiently disaggregated price data for such a reweighting are not available. Moreover, to justify a statistical modification of this kind one must assume that substitution and supply elasticities are the same for all countries, so that individual countries experience identical price changes. Such assumptions seem unduly restrictive, and this method has not been widely accepted or attempted.

Before considering the modification of existing data, however, it would seem appropriate to investigate the possible size of aggregation bias introduced into export unit value data as a result of shifts in commodity composition. Distortions arising from this factor are directly dependent upon changes in individual country’s value shares of individual commodity groups. (This holds regardless of the degree of specialization exhibited by each country.) A priori, it would seem to be a plausible assumption that individual countries’ value shares of individual commodity groups remain relatively stable over time, given the existence of differentiated products and imperfect competition. This can also be shown from a statistical examination of existing data, although a good deal of empirical work is still required on this problem. It should be added that this form of aggregation bias is probably very slight for the EEC, the United States, and Japan, since these countries use Paasche and Fisher indices. Aggregation bias of this form is more likely to arise with the use of Laspeyres indices in Canada, Sweden, and the United Kingdom.

Although there may not be any serious aggregation bias in export unit value data arising from shifts in commodity composition, this problem does introduce ambiguities into the interpretation of these data. Therefore, particular caution is required in the interpretation of relative price developments of countries, such as Canada, that specialize in the production and export of commodities whose prices fluctuate more than the average.

Conclusions on export unit values

In concluding this brief survey of statistical and methodological problems, one might ask whether export unit values can be used legitimately in analyses of international competitiveness. No doubt, the quality of the data is relatively poor; improvements in commodity coverage, for instance, would be highly desirable. Other statistical weaknesses, as well as methodological difficulties concerning the exact interpretation of relative movements in export unit values, make these data relatively uncertain as indicators of countries’ relative competitive positions. However, many of the statistical and methodological problems—such as those arising from shifts in commodity composition and the weighting of commodity groups—are common to most or all price indices. Moreover, various pieces of evidence show that international comparisons of movements in export unit data are not systematically distorted as a result of aggregation bias or marked differences in quality adjustments between major countries. Finally, there is the pragmatic evidence of the relevance of export unit data as explanatory factors in countries’ export performance. It would, therefore, seem fair to conclude that, among the admittedly weak indicators of international price competitiveness, export unit values are reasonably acceptable and—if interpreted with caution—they might even constitute the best set of regularly published data in this field.

Wholesale prices of manufactures

One conceptual problem with confining an analysis of international competitiveness solely to international price indices is that these indices cover only, or mainly, goods that are actually exported. Consequently, if a country’s competitive position in a certain commodity weakens, the commodity may eventually drop out of the index altogether. Therefore, it is preferable to consider also costs of production, as well as the prices of domestically produced but not necessarily exported goods. Price indices of the latter kind have some advantages for analyses of international competitiveness, because it is important to gauge the competitive strength of not only the export sector but also the import-competing sector and industries dependent on these two sectors.

While the wholesale price index avoids many of the specification problems that plague the unit value index, it suffers from a whole set of other problems. The indices of different countries vary widely in coverage, method of construction, and weighting. Perhaps most important, except in a very few instances where the data have been reorganized into industry selling price indices,34 the data are not particularly relevant to any particular group of producers or consumers in an economy. Even if individual countries adopted consistent commodity coverage, intercountry comparisons of wholesale price indices would remain susceptible to differences because prices are measured at different stages of production and distribution.35 In addition, wholesale price indices do not accurately measure the movements of domestic prices, quite aside from the usual problems of quality change.36 Conceptually, the index should be based upon transaction prices; however, most of the data are obtained from manufacturers’ list prices. As a consequence, the observed short-run rigidity in the wholesale price index is rather spurious and introduces a systematic lag in measured price changes relative to actual price changes.

Wholesale price indices with international trade weights

Following the example of Lary, Kravis and Lipsey have reweighted the individual components of the wholesale price index with international trade weights rather than with domestic value-added weights.37 Lary reweights the wholesale price index for only the United States, while Kravis and Lipsey reweight the index for Germany as well. Unfortunately, the data for other countries are too fragmentary to draw any conclusions on the use of this technique.

Kravis and Lipsey consider the reweighted wholesale price index a slightly better indicator of international prices than export unit values. However, this conclusion may be relevant particularly for the United States, where the foreign sector is small and differential movements in domestic and export prices are not particularly prevalent. Even if the data were available to carry out the reweighting of the European and Japanese indices, the interpretation of the reweighted index would almost certainly be subject to serious ambiguities.38 Kravis and Lipsey, in a recent survey of 120 U. S. companies, found that roughly one half of the firms priced their export differently from domestic sales.39 The practice of price discrimination was even more prevalent in Western Europe and Japan.40 The explanations given for these differences in pricing behavior were that, first, U. S. firms have a steady stream of new products coming onto the market, facing virtually no competitors. Second, exports are not a large enough proportion of sales for many U. S. corporations to warrant differential pricing policies. For Western European countries, there is some fragmentary information that exporters follow discriminatory pricing techniques. For example, in the European Economic Community, despite the abolition of intra-Community tariffs and some progress toward standardizing taxation systems, very large price differences for rather standardized commodities still exist.41

In conclusion, wholesale price data provide useful information concerning domestic price developments. However, it is difficult to accept this index as the best available indicator of international price competitiveness. Nonetheless, wholesale price indices are relevant supplementary indicators for the competitive price position of not only the export sector but also the import-competing sector and industries dependent on these two sectors.

Unit labor costs in manufacturing

Cost data—in particular, data relating to labor costs—have also been widely used in analyses of international competitiveness. Unit labor cost data may be of particular relevance in relatively open economies that are small or of moderate size. In these economies, export prices may to a great extent be autonomously determined in world markets and thus fail to reflect differential domestic cost and price developments. Similarly, the wholesale price indices, which are heavily weighted with prices of internationally traded goods, may respond quickly to external factors. As a result, a cost series that moves independently between countries may sometimes be a superior indicator of underlying international competitiveness. The causal relationship between unit labor costs and export prices (unit values) is very complex. In certain countries, domestic cost-price developments have differed from export price developments for extended periods. (These differences are clearly illustrated by Chart 1.) Nonetheless, trends in unit labor costs are useful as complementary indicators of a country’s competitive position, since in long-term competitive situations, shifts in unit labor costs must be reflected in either prices, profits, or resource allocation.

The statistical shortcomings and difficulties in making intercountry comparisons of unit labor cost series are well known. Because unit labor costs are generally regarded as important determinants of domestic prices, profits, and investment behavior, quarterly seasonally adjusted series for ten industrial countries have been computed from 1961. A detailed discussion of the assumptions, methodology, and statistical sources are presented in Appendix III. The data embody a large number of conceptual and measurement problems, the most important of which are summarized below. All four of the statistical series used in computations of unit labor costs in manufacturing (industrial output in manufacturing, employment, hours worked, and earnings) are subject to statistical weaknesses. Consequently, the final estimate may involve a number of biases that need not be offsetting.

The growth of industrial output tends to be systematically understated in most industrial countries. This reflects, in part, inadequate allowance for quality improvements with a consequent overstatement of price increases and a corresponding understatement of output changes over time. Further, since the industrial output index is aggregated on the basis of base period value-added weights, the index understates the growth of output because industries expanding more quickly than the norm receive too small a weight, while, conversely, relatively slow-growth industries receive too great a weight. This latter problem grows in importance, the more remote the bench-mark period and the greater the divergence of growth rates between industries.42

The employment data for manufacturing are quite good. The main problem is the inconsistent coverage of salaried workers, the fastest growing component of labor force in manufacturing. Generally speaking, salaried employees are covered in the indices of employment; however, they are not covered in the data on hours of work or earnings. However, this problem may not be overly severe since the salaried component makes up only 20–25 per cent of the labor force in manufacturing. In addition, the faster growth of the salaried component of the labor force is a universal phenomenon. The data on hours worked are very weak. In many countries these data are not available on a quarterly basis, and often they are of questionable reliability.

The concept of the wage bill, or labor remuneration, provides great difficulties. Ideally, one would prefer a standardized concept of labor remuneration covering all forms of pecuniary income, income in kind, and employers’ costs related to such fringe benefits as pensions, social security contributions, and health and medical insurance. Generally speaking, these fringe benefits are not covered in the data on earnings. Although fringe benefits have been growing faster than the payroll component of the wage bill, this is true in almost all countries, and consequently short period intercountry comparisons based on changes over time may not be overly distorted.

The majority of interindustry and intercountry cost studies have concentrated on labor costs rather than total costs, reflecting in part the inherent difficulties in measuring capital costs.43 This heavy concentration on labor cost comparisons can be partly rationalized by the high stability of factor shares over time within individual countries.44

In spite of all the difficulties of comparing data of different countries, there are very striking similarities among the manufacturing industries of developed countries. First, there is a strikingly stable relationship between value added and wages and salaries paid in manufacturing between industrial countries. Labor remuneration in manufacturing was equal to roughly 50 per cent of value added in 17 industrial countries for the period 1953–63, with a very diverse sample, ranging from the United States and the U.S.S.R. to the United Kingdom and Hungary.45 Second, the objection has often been raised that the manufacturing sector is too broad an aggregate and that comparisons could be made more legitimately if they were confined to specific industries. This is, of course, very difficult to do in view of the problems of isolating a homogeneous output and a relatively consistent industry coverage between countries.46 Nonetheless, there is a good deal of empirical evidence that an aggregate indicator, covering the manufacturing sector, is not unrepresentative of underlying industry developments. For example, it is a well-known fact that average hourly earnings or labor share of value added in different manufacturing industries show a remarkably similar hierarchy in the major industrial countries.47 In almost all countries, labor remuneration is lowest in the tobacco industry and highest in the transport equipment industry.48 The economic and sociological reasons for this uniform hierarchy of earnings and wage rates by industry are not germane to the present discussion. The important fact is that despite differences in technology and factor endowment intercountry comparisons of unit labor costs in manufacturing are not vitiated by marked differences in factor intensities of production in identical industries internationally.

Another aspect of unit labor cost comparisons warrants further discussion: whether unit labor costs should be measured on a national currency basis rather than on a U. S. dollar basis. Although most comparisons are usually made on a U. S. dollar basis, there are also legitimate reasons for making comparisons on a national currency basis. In converting the unit labor cost data to a U. S. dollar basis when a devaluation takes place, the least objectionable procedure would seem to be to use the official parities—the old one and the new. However, this probably means that the effects of par value changes are exaggerated—certainly in comparison with the export unit value series.49 Another problem with this technique is that it is difficult to draw the line for quasi-changes in exchange rates, arising from such measures as changes in border taxes, subsidies, and temporary surcharges. The data in Appendix I are presented on both a U. S. dollar basis and a national currency basis.

In summary, the unit labor cost data are perhaps best suited to the analysis of domestic cost-price developments. However, these data can be a useful supplementary indicator of international competitiveness, especially in relatively open small or moderate-sized economies, where export and wholesale prices are strongly influenced by developments in world markets. As a result, a cost series that moves independently between countries may sometimes be a superior indicator of international competitiveness.

III. Conclusions

Given the quality of the existing data, the measurement of shifts in relative costs and prices involves numerous statistical and methodological pitfalls. Consequently, it is impossible to say whether small observed changes in relative costs and prices indicate significant shifts in international competitive positions or whether they reflect purely random factors.50 Over the longer term, however, many of these erratic factors can be expected to even out. Cumulative changes in relative costs and prices are thus likely to be both larger and statistically more significant, especially if they tend to run in a consistent pattern.

Most observed shifts in relative export prices are comparatively modest, probably reflecting the fact that export prices are to a great extent determined in world markets. Thus, even though export unit values, on statistical grounds, may be the best single indicator of international competitiveness, these data are often difficult to interpret because—apart from variations related to par value changes—a significant proportion of the observed semiannual or annual changes are quite small. It can therefore be argued, on a priori grounds, that despite their various limitations unit labor costs series might be a superior gauge of international competitiveness, as these data tend to move independently of one another. (These relationships are clearly indicated by the means and standard deviations of the frequency distributions presented in Chart 2 and Table 2.)

Chart 2.
Chart 2.

Frequency Distribution of Changes in Export Unit Values of Manufactures and Unit Labor Costs in Manufacturing Relative to Those of Competitors

(In per cent)

Citation: IMF Staff Papers 1971, 002; 10.5089/9781451947335.024.A003

Table 2.

Frequency Distribution of Changes in Export Unit Values of Manufactures and Unit Labor Costs in Manufacturing Relative to Those of Competitors

(In per cent)

article image

Semiannual data appear to provide the best basis for intercountry comparisons of costs and prices.51 The choice of a time interval, of course, involves some judgment of the average period for consumers and entrepreneurs to react to shifts in relative prices. A priori, it seems reasonable to assume that consumers and entrepreneurs react to changes in price relationships only with a time lag and that changes in export (or import) behavior will not be influenced in as short a period as a few months. A period as long as a calendar year, on the other hand, may mask important adjustments within the year. Hence, it may be reasonable to assume a six-month lag between shifts in relative cost and price relationships and changes in export (or import) behavior. While a half year may be the minimum time period conducive to meaningful analysis of cost and price trends, there is also a maximum time horizon beyond which cumulative comparisons lose significance. For example, beyond a period of, say, seven to ten years, the industrial structures of individual economies are likely to undergo substantial change. Consequently, long-term cumulative comparisons of relative costs and prices may reflect structural shifts in consumption, production, technology, and resource allocation arising from changes in comparative advantage, rather than differential rates of inflation arising from monetary factors.

Finally, given the numerous statistical and conceptual shortcomings embodied in individual indicators noted above, it is preferable to use several indicators, along with other evidence, to form an overview of cost and price developments in individual countries.

APPENDICES

I. Comparative Cost-Price Movements

Table 3.

United States: Comparative Cost-Price Movements, 1961–701

(1963 = 100)

article image
Source: See Appendix III.

Data for all countries pertain to manufacturing sectors.

Note: The digits to the right of the decimal points in these tables are provided for the sole purpose of reducing rounding errors and should not be regarded as statistically significant.
Table 4.

United Kingdom: Comparative Cost-Price Movements, 1961–70

(1963 = 100)

article image
Table 5.

Japan: Comparative Cost-Price Movements, 1961–70

(1963 = 100)

article image

From 1969, industrial products; prior to 1969, manufactured goods.

Table 6.

Canada: Comparative Cost-Price Movements, 1961–70

(1963 = 100)

article image
Table 7.

Sweden: Comparative Cost-Price Movements, 1961–70

(1963 = 100)

article image
Table 8.

Belgium: Comparative Cost-Price Movements, 1961–70

(1963 = 100)

article image