Lecture 7. Empirical tests of the Heckscher-Ohlin model

Transcription

1 Lecture 7. Empirical tests of the Hecscher-Ohlin model The Hecscher-Ohlin model maes a series of strong assumptions in order to isolate the effects of different relative factor endowments on trade between two countries. These assumptions include: Identical production technologies with constant returns to scale Perfect competition The absence of factor intensity reversals Identical and homogenous preferences The absence of international factor mobility The absence of impediments to trade It is obvious that this set of assumptions does not hold in reality and that predictions of the model cannot be expected to hold literally. The tas is to assess the significance of endowment differences in explaining trade patterns. Is trade pattern really related to differences in factor endowments? The Leontief Paradox The Hecscher-Ohlin model is incapable of predicting that any country will export a particular good when there are more goods than factors. (For example, the capital abundant country may not export the most capital-intensive good when there are three commodities.) However, the implicit international trade in factor services must be obeyed by the H-O theorem, i.e. with two factors and two or more goods, the capital-abundant country will find that its bundle of exports is more capital-intensive than its bundle of imports. 1

2 This theorem on the factor content of international trade was first examined by an American economist Wassily Leontief (1953) in the most famous empirical study in economics. Leontief aided the US planning efforts during the WWII by developing a technique of accounting for all inputs required in the production of the GNP. This technique, called the INPUT-OUTPUT analysis, recognizes that the production of, say an automobile, requires primary inputs such as capital and labor, in addition to intermediate inputs such as steel, paint, glass and the lie. The prior production of these inputs also capital and labor in addition to other intermediate inputs which in turn require capital and labor in addition to other intermediate inputs, and so on. Leontief developed a method for assembling various inputs into an input-output table that could be used to compute the total labor and capital embodied in production of any bundle of goods. An obvious application of input-output analysis developed by Leontief was to discover how much capital and labor were required to produce US exports in comparison with US imports. Note that here we have an important methodological problem! While it is sensible to use an American input-output table to compute the factor contents of US exports, a computation of the factor content of US imports would require detailed and consistent data on production techniques in all foreign trading partners. However, this was not feasible for Leontief so he calculated the capital and labor required to produce US goods that are similar (or compete with) American imports. 2

3 This procedure is valid under either of two conditions: First, if the factor endowment model holds and international factor prices are equalized, each country shares the same production techniques and using the US input-output table will not bias the import computations. Second, if production function exhibit fixed unit factor requirement coefficients (our a s in the technology matrix), or a constant ratio of capital to labor regardless of the factor price ratio, use of US techniques similarly captures foreign production accurately. Leontief calculated the capital and labor requirements in the production of a representative bundle of $ 1 million worth of both exports and import-competing goods in In that year the US was unquestionably the most capital-abundant nation in the world and was certainly capital-abundant and labor-scare relative to the rest of the world. Thus, the expectation was that exports were capital-intensive and imports laborintensive. Nevertheless, Leontief discovered that the capital-labor ratio in the US imports exceeded that in the US exports by some 23%! This unexpected result has been termed the Leontief paradox. 3

4 Table 1. Leontief s early tests of factor content of US exports and import substitutes Exports Import substitutes Imports/Exports 1947 trade Capital $ 2,550,780 $ 3,091,339 Labor (worer years) Capital/worer years $14,010 $ 18, trade Capital $ 2,256,800 $ 2,303,400 Labor (worer years) Capital/worer years $12,977 $13, SOLUTIONS TO THE LEONTIEF PARADOX Leontief s famous result caused a great surprise among economists schooled in the H-O tradition. Leontief himself was puzzled by the finding and asserted that the issue was really one of measurement. In particular, his belief was that, because superior education and training in conjunction with better management techniques, American labor was perhaps three time more productive than foreign labor. Thus, in effective labor units, the US was really labor-abundant. While, this view anticipated later thining about labor sills (i.e. human capital), it was ad hoc and unconvincing of the time. The 3 to 1 ratio suggested by Leontief was not predicted on a careful evaluation of labor efficiency throughout the world; it was simply the ratio required to get the expected result. 4

5 In fact there is little evidence that this huge superiority in US labor ever existed even in the early 1950s. Moreover, to the extent that labor was enhanced by better American management we would expect that the relative productivity of capital to be enhanced as well! Numerous attempts have been made to verify Leontief s results, with mixed results. Objection 1. U.S. is a very special country Because the U.S. may have been an unusual country similar computations were made for other countries with all manner of endowments, incomes and maret structures. In some cases the results of these studies were consistent with expectations under the endowment model. For example, Heller (1976), in his paper published in the Review of Economics and Statistics, demonstrated that Japan s international trade in the 1960s followed an interesting dual structure. Japan s exports to less developed nations were capital-intensive and its imports were labor-intensive, while its exports to more developed nations were relatively laborintensive in comparison with imports. Objection is not a good year for testing the endowment theory A more compelling objection to Leontief s finding was that 1947 was not a very attractive year for testing the endowment theory. The model relies on the specification of a long-run equilibrium without any maret distortions. It can be hardly argued that the economies of Europe and Japan were in equilibrium in 1947, rather they were beginning a process of rapid dynamic adjustment in production and factor supplies (post WWII reconstruction). 5

6 Baldwin (1971), in his paper published in American Economic Review, recomputed Leontief s ratios for the United States using the 1958 input-output table and 1962 international trade data, with the result that the Leontief paradox was still strongly in evidence. Table 2. Follow-up tests of factor content in US trade Exports Import substitutes Imports/Exports 1962 trade Capital $ 1,876,000 $ 2,132,000 Labor (worer years) Capital/worer year $14,200 $ 18, Capital/worer year, 1.04 excluding natural resources Capital/worer year, 0.92 excluding natural resources and including human capital The analysis by Stern and Masus (1981) published in the Journal of International Economics, showed that by 1972, American exports had become capital-intensive relative to imports. Thus, the paradox may have been reversed by the 1970s. However, it is unclear how the relative capital intensity of exports could have risen given the fact that in the post-war period American relative capital abundance seems to have declined. 6

7 Objection 3. the assumptions of the Hecscher-Ohlin model are too strict Another line of objections has come from noting that the assumptions of the H-O model are too strict to be believed. The most enduring and valuable outcome of the debate over the Leontief paradox is that it stimulated trade theorists to thin more fully about the implications of departures from these assumptions. The role of barriers to trade It is possible that international structure of trade barriers could partly explain Leontief s result. Our theoretical analysis showed that free trade would lower the real incomes of each country s scarce factor providing an incentive for that factor to lobby for import protection. Thus, the U.S. might be expected to have high trade barriers to labor-intensive imports, and some foreign countries might erect restrictions on capital-intensive imports. Consequently, the these protectionist policies could reduce the levels of trade below those expected from endowment-based comparative advantage. The role of consumer preferences Another possibility is that preferences across countries differ rather than being identical and homogenous. Certainly, if there are significant trade biases in the sense that some countries have strong preferences for goods in which they would otherwise have a comparative advantage, the pattern of trade could be reversed. 7

8 The role of technology Another possibility is that countries do not share access to identical technologies. This observation is at the core of the Vernon product cycle model and relates ideas about trade in a more dynamic context. The issue here is that trade data for a particular year many not reflect a long-run static equilibrium so much as a shortrun, dynamic transition under varying technologies. In the product life cycle, for example, American exports of new goods may seem to be labor intensive when in reality they mae relatively heavy use of new technological information through the employment of highly silled, technical labor inputs such as engineers and scientists. This taes us directly to the case of more factors of production and in particular the concept of human capital discussed below. The role of other factors and factor heterogeneity The most substantive objection to Leontief s procedure is simply that it is inadequate to suppose that there are only two primary factors of production: capital and labor. Various forms of land and natural resources also serve as sources of comparative advantage. They may be the most relevant factors for the Hecscher- Ohlin model because they are internationally immobile. Moreover, physical capital and labor exist in many different forms. Therefore, we should not expect the average worer in high income countries to share identical productivity characteristics with the average worer in low income countries. National labor forces consist of different endowments of worers of various sills with sills being higher in countries that invest more in education and training. Worers are often distinguished by their human capital or accumulated investments in education and training (years of schooling). 8

9 Regarding natural resources Jaroslav Vane (1963) made an observation that the Leontief paradox is consistent with the possibility that the U.S. was abundant in capital and labor but scarce in natural resources, such as oil, the extraction of which is quite capital-intensive. Thus, the act of importing scarce naturalresource services, the U.S. was importing the services of capital as well. Some economists accounted for this possibility in their wor by excluding certain natural resource-intensive products from their computations. For example, in the study by Stern and Masus (1981) this exclusion reinforced the finding that Leontief paradox was reversed in Keesing (1965) was the first to show that a disaggregation of U.S. labor by sill type was important in explaining the factor contents of trade. American net exports were clearly intensive in highly silled worers. This finding has proved to hold strongly in examinations of other datasets and definitions of human capital. Most trade economists agree that a fundamental determinant of U.S. comparative advantage is a relatively abundant supply of highly-silled labor. 9

10 Generalization of the Hecscher-Ohlin model: Many goods, Many factors In the Hecscher-Ohlin model we have assumed that there were only two goods and two factors of production. The natural question is to as whether and to what extent the H-O model can be generalized to higher dimensions. One possible generalization is to assume that products X and Y are groups of products such as all manufactures or all agricultural products. If relative commodity prices within the two groups are unchanged at all times, all the preceding conclusions will hold as relative prices and outputs of the two groups change. MANY GOODS If there are more than two commodities difficulties arise even with two factors of production. In this case, factor intensity ratios need not produce unique results. Suppose that a third good Z is added to the model and that: KY K X KZ > > L L L Y X What would be the pattern of trade in a two-country world? One might expect that the country well endowed with capital would export Y, the country well endowed with labor would export Z, and some indeterminacy could exist about good X. However, this is not necessarily true! It is possible for either country to export both Y and Z and import X. Z 10

11 What can be shown instead is that the bundle of goods exported by the capital abundant country will be capital-intensive in the sense of embodying a higher ratio of capital to labor than its import bundle. However, the capital abundant country s export bundle could consist either of some both Y and Z, two extreme goods, or of good X, the good of intermediate capital intensity. MANY FACTORS With more than two factors, additional problems arise if we insist on raning pairs of factor intensities. For example, with a third factor of natural resources called R, it is possible to have K X /L X > K Y /L Y but K X /R X < K Y /R Y which good is capital-intensive in a case lie this? It obviously depends on how the comparison is made. Bilateral comparisons do not have much meaning in a world with more than two factors. Results such as those given by the Stolper-Samuelson theorem and the Rybczyńsi theorem, which mae use of bilateral comparisons, do not easily generalize to higher dimensions. The Hecscher-Ohlin theorem does generalize in an important way, however, if we modify our definition of relative factor endowments. Suppose we can define products and factors so that there is an equal number of each, as in the 2x2 model. There can be any number of countries. We retain assumptions of identical CRS technologies, identical preferences and the absence of distortions. It is possible to ran endowments of any country by computing its share of each endowment in the global supply with the most abundant factor being the one with the highest relative share and the most scarce factor the one with the lowest relative share. In general any factor in which a country s share of the global supply exceeds that country s share of global income is raned as abundant, and the others are raned as scarce. If then we compute the amounts of each factor that are used to produce the bundle of exports, imports and gross national product, we can derive a theorem about the factor content of each nation s trade. 11

12 Factor Content (Hecscher-Ohlin-Vane) Theorem: For an arbitrary but equal number of goods and factors, a raning of the content of any factor in net exports divided by its content in total output will duplicate the raning of relative factor endowments. This theorem is often called the Hecscher-Ohlin-Vane theorem because of the contribution of Jaroslav Vane (1986). Its practical importance is in showing that under reasonably general circumstances differences in relative factor endowments still determine comparative advantage. However, comparative advantage in this sense refers to the pattern of trade in factor services rather than in goods. It does not really matter that we cannot predict whether a country will export or import a particular commodity because commodity trade is merely the means for implicitly trading factors. Factor price equalization theorem also generalizes for any number of factors and goods as long as the number of goods is equal to the number of factors and all countries produce all these goods (incomplete specialization in production). More generally, both the FPE theorem and factor content theorem hold if there are more goods than factors so long as the number of goods produced in common by each nation is at least as large as the number of factors. The role of additional factors of production is properly incorporated into the factor endowment model only through the factor content (HOV) theorem. 12

13 We retain all the assumptions of the Hecscher-Ohlin model, except allowing there to be I factors and J goods, with J I (the number of factors must be smaller or equal the number of goods). For any country let s be its proportion of the world income. Finally, let FE i and FE i W denote endowments of factor i for country, and the world respectively. Let us define a raning of factor abundance and scarcity for country by virtue of its share of world endowments of each factor: FE 1 FE2 FEi FE > > I > > > >... s W... W W W FE1 FE2 FEi FEI Under the HOV theorem, country s consumption share lies somewhere in the middle of this chain. It also follows that country s net exports of the services of any factor are positive if its abundance raning for that factor lies above the consumption share and its net exports are negative if its raning lies below that share. That is, a country exports the services of its abundant factors and imports services of its scarce factors when factor abundance is measured relative to a global standard. SUPPLY SIDE To see this we derive Vane s (1968) factor content equation. We start with specifying the set of full employment conditions for each factor in country. Let FE and Q denote country s vector of factor endowments and country s output vector, respectively. The full employment condition can be written in the matrix notation as: 13

14 J FE 1 a11... a j j j Q a Q j = = =... J 1... FEI ai aij QJ aijq j j = 1 1 J 1 I I J where: a ij coefficient indicating the number of units of factor i used in the production of a unit of good j (unit input requirement) FE i = J [ a ]' Q = i j = 1 a ij Q j interpretation: the endowment of factor i = i-th row of technology matrix production vector = the sum of the amounts used (i.e. factor demands) in the production of goods 1, 2,, J. I 1 Using short-hand matrix notation we have: FE Q = = A Q 1 [ A ] FE In other words, we expressed the output vector of country in terms of its factor endowments and combinations of input-output coefficients. 14

15 DEMAND SIDE Assume that demand preferences are homothetic and identical across countries. In this case the demand (consumption) vector D representing country s consumption demand for each good can be expressed as a country s share in the world income s, times world output Q W of each good: D = s Q W = s K K c c Q = s [ A ] c= 1 c= 1 1 FE c where D, Q W, Q are the J 1 vectors. The share of country s income in world income is: s Y Y = = W K [ w ]' c [ w ]' c= 1 FE FE c c c c where: [ w ] [ w,..., w ] ' 1 = is the row vector representing country c s factor rewards. I 15

16 THE FACTOR CONTENT EQUATION The exports and imports of trade goods correspond to the gap between production and consumption. T = Q J 1 D (vector of country s net exports of goods) The value of net exports for a particular good j good j is imported. T j = Q D is positive if good j is exported and negative if j j The content of each factor i in the vector of goods j is given by the corresponding element of I 1 vector FC. FC = A T = A ( Q D FE ) = FE How do we measure the factor content of imports? A D = FE A K [ A ] s FE c= 1 D 1 The Hecscher-Ohlin model assumes that the input-output matrix is the same in each country A = A c. Hence, Vane s factor content equation can be written as: FC = A T = FE s K FE c= 1 = FE s FE W 16

17 Interpretation of factor content equation A typical element of the factor content matrix equation is: [ FC = a ] i i ' 1 1 T = 1 J J 1 J a T j = 1 ij j = FE i s FE W i The factor content equation views a country as implicitly exporting or importing factor services through the factor content of goods trade (the difference between factor content in output and factor content in consumption). KEY IDEA: Goods trade represents trade in factor services used to produce the exports sold abroad and the products imported from abroad. W [ a ]' T > FE s FE FC i = i 0 i > i if factor i tends to be used more intensively in exports (than in imports) W [ a ]' T < FE s FE FC i = i 0 i < i if factor i tends to be used more intensively in imports (than in exports) The practical importance of the H-O-V theorem is that it can be used to ran a country s factor endowments in terms of their relative abundance as revealed by trade. Once the ranings are compiled they may be compared to actual data on national and world endowments as a full test of the HOV theorem. 17

18 W A country should export ( FC i > 0) the factor services offered by its abundant factors ( FE i s FEi > 0). In other words, both sides of the factor content equation should have the same sign (wea test, does not have a null hypothesis). Two empirical approaches to testing the H-O-V theorem have been employed: The first technique is based on statistical tests that relate measures of net exports across different industries to factor intensities in production. For example, Baldwin (1971) found that capital intensity was negatively related to net exports for the US (lie in the Leontief paradox) but human capital was positively related to net exports. The second technique is based on the regression analysis that tests whether net exports are a linear function of the difference between a country s vector of factor supplies and its vector of factor demands. Assuming identical and homothetic tastes the demand for factors is simply the product of the country s global income share and global factor endowments. For example, Leamer (1980) used the regression analysis to estimate this linear relationship. THE SIGN AND THE RANK CORRELATION TESTS: Sign proposition: If trade is balanced, a country is a net exporter of the services of its absolutely abundant factors and a net importer of the services of its absolutely scarce factors. Ran proposition: If a country is abundant in factor h relative to factor z, then the country s proportionate net trade in the services of factor h exceeds its proportionate net trade in the services of factor z. 18

19 The raw correlation test considers the measured and predicted factor content as an observation and computes the correlation between the measured and the predicted factor content of trade. The ran correlation test orders the measured and the predicted contents according to their values and the computes the correlation between the rans. In both cases correlations should be near one (no null hypothesis). Such tests have been performed for the US by Masus (1985) and for a collection of countries by Bowen, Leamer and Sveiausas (1987). TABLE 3. Tests of the H-O-V sign proposition (country by country). Country Proportion of sign matches (%) Argentina 33 Australia 33 Austria 67 Belgium-Luxembourg 50 Brazil 17 Canada 75 Denmar 42 Finland 67 France 25 Germany 67 Greece 92 19

20 Hong Kong Ireland Italy Japan Korea Mexico Netherlands Norway Philippines Portugal Spain Sweden Switzerland UK US Yugoslavia Source: Bowen et al. (1987)

21 TABLE 4. Tests of the H-O-V sign proposition (factor by factor) Factor Proportion of sign matches (%) Capital 52 Labor (total) 67 Professional/Technical 78 Managerial 22 Clerical 59 Sales 67 Service 67 Agricultural 63 Production 70 Land Arable 70 Pasture 52 Forest 70 Source: Bowen et al. (1987) 21

22 TABLE 5. Tests of the H-O-V ran proposition (country by country) Country Ran correlation Proportion of correct pairwise ranings (%) Argentina Australia Austria Belgium-Luxembourg Brazil Canada Denmar Finland France Germany Greece Hong Kong Ireland Italy Japan Korea Mexico Netherlands Norway Philippines Portugal Spain

23 Sweden Switzerland UK US Yugoslavia Source: Bowen et al. (1987) TABLE 6. Tests of the H-O-V ran proposition (factor by factor). Factor Ran correlation Proportion of correct pairwise ranings (%) Capital Labor (total) Professional/Technical Managerial Clerical Sales Service Agricultural Production Land Arable Pasture Forest Source: Bowen et al. (1987) 23

24 Bowen et al. (1987) found that exporters of net factor services are predicted as well by a random process similar to flipping a coin as by the factor endowments model! Both the correlations and the proportion of correct pairwise ranings suggest that ran proposition holds more often than does sign proposition. Yet, the ranings are not exact, and since the choice of a critical value for the proportion of correct ranings is largely subjective, the extent to which ran proposition is to be considered validated is also subjective. Unfortunately, the results of these early tests indicate that the ranings of factor abundance revealed by trade do not convincingly replicate actual endowment ranings, suggesting that the theorem does not hold literally in real world situations. Bowen et al. (1987) investigated a number of alternatives to the assumptions underlying the HOV model. These included technological differences and non-proportional consumption. They could reject the identical technologies assumption in favor of neutral technological differences but could not reject identical and homogenous preferences. More recent tests by Trefler (1993) that correct for technological differences across countries seem to provide more support for the HOV theorem. He found that the data favored the assumption of neutral technological differences. 24

25 TABLE 7. HOV estimates of technological differences Country 1983 GDP per capita relative to the US Productivity relative to the US Austria Bangladesh Belgium Canada Colombia Denmar Finland France Greece Hong Kong Indonesia Ireland Israel Italy Japan Netherlands New Zealand Norway Paistan Panama Portugal T-statistic for null hypothesis that relative productivity equals

26 Singapore Spain Sri Lana Sweden Switzerland Thailand Trinidad UK Uruguay West Germany Yugoslavia Source: Trefler (1993) Table We should expect that relative GDP per capita numbers be highly correlated with the estimates of relative productivity. In fact this correlation is 0.89 for

27 We should not be, however, very much surprised by the inability of actual data to conform to rigorous assumption of the H-O-V model. No one should be surprised that the world is not perfectly competitive and that there exist maret distortions such as taxes or trade barriers. Therefore, rather than testing the truth of the theory based on unrealistic assumption it is much more illuminating to measure how strongly factor endowments seem to affect actual trade flows. REGRESSION ANALYSIS For each of ten broadly defined commodities, such as raw materials and labor-intensive manufactures, Leamer (1980) estimated this relationship using a sample of 47 countries with measures of 10 factor endowments in The regression estimates were taen as measures of the lin between endowments and trade. Table below shows computations of relative factor endowments (i.e. shares of each country in the world supply of factors) for selected countries. Table 8. Relative factor endowments, 1975 (% of world endowments). Factor Canada Germany India Japan Mexico Korea UK US GNP Capital Prof labor Lit. labor Illit Labor Trop

28 Land Dry land Temp land Coal Minerals Oil Source: Leamer (1980) The entries in the first row are national/country shares in the world GNP, or s j terms. The remaining rows show each country s share in the world supply of factor endowments. In theory these may be compared to the GNP shares to measure relative factor abundance. For example, Japan had percent of the world capital stoc, which exceeded its GNP share of 11.4 percent suggesting that Japan was capital abundant. On the other hand, Japan was very scarce in land and natural resources. The U.S. was abundant in dry land, coal, and oil and scarce in lower-silled labor and tropical land (in fact the U.S. is a major importer of oil, suggesting that its tastes are biased toward oil consumption in comparison with other countries!). The question is how the factor endowments relate to actual trade flows. Leamer s (1980) regression estimates for selected groups of products are shown in the table below: 28

29 Table 9. Estimates of the effects of factor endowments o trade, Factor Raw animal Cereals Laborintensivintensive Capital- Chemicals Machinery materials Capital Prof. labor Lit. labor Illit. Labor Trop. Land Dry land Temp. land Coal Minerals Oil Source: Leamer (1980). Each entry indicates a change in net exports that is estimated to result from a unit increase in the associated endowment. For example, a $ 1 million increase in the capital stoc of the average country (average over all countries) would reduce net exports of raw materials by $ 8,800 and cereals by $ 4,300. At the same time, it would increase net exports of animal products by $ 40, of labor intensive products by $ 1000, of capital intensive products by $ 16,500, of chemicals by $ 3800 and of machinery by $ 29,

30 Increases in land and resource supplies tend to expand exports of raw materials and agricultural goods while reducing exports of manufactured goods. Expansions of the stoc of literate labor (blue collar worers and educated worers) raise net exports of most manufactured goods but reduce net exports of materials and agricultural commodities. Professional labor is a source of comparative advantage for chemicals. Overall, the equations fit the data well, each equation typically explains as much as 50 to 60 % of observed trade patterns. Main conclusion: Factor endowments seem to have an extremely important influence on comparative advantage in world trade. 30