International trade flows are to a great extent determined by the location of the affiliates of multinational firms with worldwide production facilities. A great deal of this trade consists of direct intrafirm transactions between affiliates of the same firm. For example, in 1970 manufacturing exports from parent U. S. companies to their affiliates abroad accounted for over 20 per cent of all U. S. manufacturing exports. In the same year, intrafirm trade within the U. S.-controlled multinationals alone accounted for about 10 per cent of all world trade in manufactures.
The importance of such internal transactions has been growing over time. This could have important implications for the response of a country’s trade to changes in economic activity and relative competitiveness. For instance, trade flows generated by the location decisions of a firm with large fixed investments in several countries may not respond as rapidly to shifts in relative prices as those of an independent producer that is unconcerned with the effect of its actions on the profitability of overseas affiliates. This paper reports on some attempts to test whether the response of such internal trade flows to shifts in aggregate demand and relative prices differs from the response of more conventional trade. A simplified version of the Fund’s World Trade Model was fitted to data on the transactions of U. S. multinational companies and their overseas affiliates, and the results were compared with those obtained for conventional trade flows.
The paper first discusses the size of intrafirm trade flows and variations in their relative importance across countries and industries. The next section considers possible implications of such trade for the response to relative price changes. There follows an outline of simple export and import equations that can be fitted to data on conventional and intrafirm manufacturing trade in order to test whether significant differences exist in the parameters for the two types of trade. Final sections discuss the results and summarize the major conclusions. An appendix discusses some of the economic problems in measuring price elasticities in international trade caused by the use of artificial transfer prices in intrafirm transactions.
I. Econometric Implications of Using Artificial Transfer Prices in Intrafirm Trade
If the true model of intrafirm imports (small letters denoting log terms and all variables measured as deviations from means, to eliminate the constant term) is
m = απ + βy + e
but observed transfer prices, p, differ from the true prices, π, by a component, v, so that
p = π + v
then the model actually estimated is
m = α(p − v) + βy + e = αp + βy + (e − αv)
If all the ordinary-least-squares assumptions continue to hold, except that the error term is now correlated with the price variable, then the expected value of the estimated price elasticity will be
E(α) = α − αbvp·y
where bvp·y denotes the partial regression coefficient of p in the auxiliary regression of v on y and p (see Griliches (1957) for a more detailed discussion of the terminology). This auxiliary regression is
v = bp + δy + u
v = b(π + v) + δy + u
cov (π, v) ≧ 0, then 0 < λ < 1 and E(α) is biased downward; if
−vari(v) < cov(π, v) < 0
then 0 < λ < 1 and E(α) is still biased downward; if
−var(v) − var(π) < cov(π, v) − var(v)
then λ < 0 and E(α) is biased upward; if
cov(π, v) < − var(v) − var(π)
then λ > 1 and E(α) is biased downward.
In the conventional measurement error case, cov(π, v) is usually regarded as zero, so 0 < λ < 1 and E(α) is biased downward, with the magnitude of the downward bias being greater, the larger is the variance of the measurement error relative to the variance of the true price. But in the case of measurement error owing to transfer prices, generally a systematic relationship exists between the measurement error and the true price, so that cov(π, v) ≠ 0 and the following cases must be distinguished:
(a) If there is an incentive to overstate transfer prices, then we would expect cov(π, v) > 0 (for instance, if transfer prices were a percentage markup on true prices). In this case, the estimated price elasticity will again be biased downward in absolute amount, and if
(b) If there is an incentive to understate transfer prices, so that cov(π, v) < 0 (for instance, if transfer prices were marked down to some fraction of true prices), the estimated price elasticity may be biased upward or downward, depending on the relative sizes of the covariances and variances of π and v. To investigate this further, it may be recalled that it has been shown that the estimated price elasticity is biased upward only if
− var(v) − var(π) < cov(π, v) < − var(v)
Dividing through by var(π) yields
− (1 + c) < bvπ < − c
and bvπ is the regression coefficient of π in the regression of v on π, that is,
v = b0 + bv π · + E
It is known that bv π > −1 (otherwise a rise in the true price would mean that the firm would reduce its transfer price—an unlikely move). Therefore, it is only in the range of −1 < bv π < −c that an upward bias in estimated price elasticities will result. For instance, if the variance of the true price were twice the variance of the error term, so that c = ½, the firm would have to follow a policy of changing its transfer price by less than half the amount of the true price change before an upward bias in the estimated price elasticity would result.
II. Data Sources
All data are on an annual basis. All time series in index form have 1970 as the base year.
Almon, Shirley, “The Distributed Lag Between Capital Appropriations and Expenditures,” Econometrica, Vol. 33 (January 1965), pp. 178–96.
Armington, Paul S., “A Theory of Demand for Products Distinguished by Place of Production,” Staff Papers, Vol. 16 (March 1969), pp. 159–78.
Buckley, P. J., and R. D. Pearce, “Overseas Production and Exporting by the World’s Largest Enterprises: A Study in Sourcing Policy” (unpublished, University of Reading, Department of Economics, September 1977).
Chow, Gregory C., “Tests of Equality Between Sets of Coefficients in Two Linear Regressions,” Econometrica, Vol. 28 (July 1960), pp. 591–605.
Deppler, Michael C., and Duncan M. Ripley, “The World Trade Model: Merchandise Trade,” Staff Papers, Vol. 25 (March 1978), pp. 147–206.
Great Britain Monopolies and Restrictive Practices Commission, “Chlordiazepoxide and Diazepam” (Her Majesty’s Stationery Office, London, 1973).
Griliches, Zvi, “Specification Bias in Estimates of Production Functions,” Journal of Farm Economics, Vol. 39 (February 1957), pp. 8–20.
Hogan, W. P., “British Manufacturing Subsidiaries in Australia and Export Franchises,” Economic Papers, the Economic Society of Australia and New Zealand, New South Wales and Victorian Branches (July 1966), pp. 10–27.
Lall, Sanjaya, “Transfer-Pricing by Multinational Manufacturing Firms,” Oxford Bulletin of Economics and Statistics, Vol. 35 (August 1973), pp. 173–95.
Lall, Sanjaya, “The Pattern of Intra-Firm Exports by U. S. Multinationals,” Oxford Bulletin of Economics and Statistics, Vol. 40 (August 1978), pp. 209–22.
United Nations Conference on Trade and Development, “Restrictive Business Practice: Studies on the United Kingdom of Great Britain and Northern Ireland, the United States of America and Japan,” Report No. TD/B/390 (New York, 1973).
United States, Department of Commerce, Bureau of Economic Analysis, Special Survey of U.S. Multinational Companies, 1970 (Washington, November 1972).
United States, Tariff Commission, “Implications of Multinational Firms for World Trade and Investment and for U. S. Trade and Labor: Report to the Committee on Finance of the U. S. Senate and its Subcommittee on International Trade on Investigation No. 332–69, Under Section 332 of the Tariff Act of 1930” (Washington, February 1973).
- Search Google Scholar
- Export Citation
)| false United States, Tariff Commission, “Implications of Multinational Firms for World Trade and Investment and for U. S. Trade and Labor: Report to the Committee on Finance of the U. S. Senate and its Subcommittee on International Trade on Investigation No. 332–69, Under Section 332 of the Tariff Act of 1930”( Washington, February 1973).
Mr. Goldsbrough, economist in the Special Studies Division of the Research Department when this paper was prepared, is currently an economist in the African Department. He is a graduate of Cambridge University and received his master’s degree and doctorate from Harvard University.
Mr. Smith W. Allnutt of the International Investment Division, Bureau of Economic Analysis, U.S. Department of Commerce, provided some unpublished statistics from the Special Survey of U. S. Multinational Companies, 1970 that were needed to calculate the weights used in the demand and price terms of the estimating equations.
See Trade and Industry (U. K. Department of Industry) (April 13, 1972), Table 3, p. 50.
Similar results are obtained by Lall (1978). A possible explanation of why intrafirm trade is more important in technology-intensive industries is discussed later.
A number of other examples of such restrictions on the export trade of overseas affiliates of multinational companies are given in United Nations Conference on Trade and Development (1973).
The economics of the specialized versus the standardized input are discussed in greater detail in Porter and Spence (1977).
Export equations for conventional and intrafirm trade are estimated for the United Kingdom, Belgium, France, the Netherlands, Switzerland, Canada, and the Federal Republic of Germany.
This paper is concerned with testing for significant differences in parameters for conventional and intrafirm imports, using the same model. More fundamental objections could be raised to the whole concept of applying a model based on consumer demand theory to intrafirm trade, which is essentially determined by decisions taken within the firm. Some attempts have been made to construct a completely different model for intrafirm trade, based on a choice-of-location framework, but these have not succeeded to any great extent.
Belgium, Canada, the Federal Republic of Germany, France, Italy, Japan, the Netherlands, Switzerland, the United Kingdom, and the United States. The World Trade Model covers these same ten countries plus Austria, Denmark, Norway, and Sweden.
The F-test for this is F(R*,R) = [(SSRr − SSRun)/R*]/[SSRun/R] where SSRr and SSRun are the sum of squared residuals in the restricted and unrestricted regressions, respectively; R* is the number of restrictions; and R is the number of degrees of freedom in the unrestricted model. For this test, the result was F(2, 20) = 13.4, which is significant at the 1 per cent level of confidence.
The F−statistics on the test for significant differences in the price coefficients were
Chemicals: F(2,20) = 8.03, significant at the 1 per cent level
Electrical products: F(2,20) = 2.52, significant at the 10 per cent level
Transport equipment: F(2,18) = 10.01, significant at the 1 per cent level
A breakdown by country of origin of manufacturing exports by U. S. majority-owned foreign affiliates (MOFAs) is available only from 1966.
Equations were also estimated for Canada and the Federal Republic of Germany. However, the price terms in both the intrafirm and conventional export equations were generally insignificant and of the wrong sign. No equation could be estimated for Japan because exports from U. S. affiliates located there were negligible.
The F-statistics were
United Kingdom: F(2,14) = 0.21, not significant
Belgium: F(2,14) = 3.70, significant at the 5 per cent level
France: F(2,14) = 1.44, not significant
Netherlands: F(2,14) = 2.64, significant at the 10 per cent level
Switzerland: F(2,14) = 10.37, significant at the 1 per cent level
Strictly speaking, the argument here should be in terms of probability limits, because the expressions concern products and ratios of moments of random variables, but these terms have been avoided for ease of exposition.
See R. David Belli, Smith W. Allnutt, and Howard Murad, “Property, Plant, and Equipment Expenditures by Majority-Owned Foreign Affiliates of U. S. Companies: Revised Estimates for 1966-72 and Projections for 1973 and 1974,” Survey of Current Business, Vol. 53 (December 1973), pp. 19–32, especially pp. 21–22, for further details.
Canada, the United Kingdom, Japan, Belgium, France, the Federal Republic of Germany, Italy, the Netherlands, and Switzerland.
Again, the markets included in the weighting scheme are the United States, Japan, Canada, the United Kingdom, Belgium, France, the Federal Republic of Germany, Italy, the Netherlands, and Switzerland.