1、Journal of International Economics 50 (2000) 5171www.elsevier.nl/locate/econbaseTechnology, trade and factor prices*Paul R. KrugmanDepartment of Economics, E52-383a, Massachusetts Institute of Technology, Cambridge,MA 02139, USAReceived 1 June 1998; accepted 3 March 1999AbstractThe view that recent
2、changes in the distribution of income primarily reflect technologyrather than trade may be the majority opinion, but has been harshly criticized by some tradeeconomists. This paper will argue that the critique in fact misses the point, essentiallybecause the critics undertake the wrong thought exper
3、iments. Trade volumes are notirrelevant: if one poses the question correctly, one immediately realizes that small tradevolumes are inconsistent with a story that attributes large distributional effects to trade. Thefactor bias of technological change is not immaterial, except in the case where such
4、changetakes place in a small open economy (as opposed to one that can affect world prices), andwhere technical change occurs only in that economy (rather than occurring simultaneouslyin other economies as well); since the real situation does not meet either criterion, factorbias definitely does matt
5、er. Most surprisingly, the much maligned use of a factor contentapproach to infer the effects of trade on factor prices turns out to be an entirely justifiedprocedure when carefully applied. 2000 Elsevier Science B.V. All rights reserved.Keywords: Factor prices; Technology; TradeJEL classification:
6、F11; F161. IntroductionOver the last decade or so, the StolperSamuelson theorem, that classic piece oftrade theory which asserts that changes in goods prices have magnified effects on*Tel.: 11-617-253-1551; fax: 11-617-253-4096.E-mail address: krugmanmit.edu (P.R. Krugman)0022-1996/00/$ see front ma
7、tter 2000 Elsevier Science B.V. All rights reserved.PII: S0022-1996(99)00016-152 P.R. Krugman / Journal of International Economics 50 (2000) 51 71factor prices, has moved from midterm exams into the heart of real-world debatesover economic policy. The reason is that an expansion of world trade, ande
8、specially of manufactures exports from low-wage countries, has coincided with afall in the real wages of less-skilled American workers (and with rising unemploy-ment in other advanced countries). It is natural to suspect a link between trade anddeclining wages; indeed, many commentators, including s
9、ome economists, havenot hesitated to assert flatly that growing trade is the principal cause of wagedecline.It is probably fair to say, however, that the majority view among seriouseconomic analysts is that international trade has had only a limited impact onwages. Skepticism about the effects of tr
10、ade on wages rests essentially on theobservation that despite its growth, trade is still quite small compared with theeconomies of advanced nations. In particular, imports of manufactured goods fromdeveloping countries are still only about 2 percent of the combined GDP of theOECD. The conventional w
11、isdom is that trade flows of this limited magnitudecannot explain the very large changes in relative factor prices that have occurred,in particular the roughly 30 percent rise in the wage premium associated with acollege education that has taken place in the United States since the 1970s. Lowestimat
12、es of the impact of trade on wages are often, though not always, based on amethodology that tries to compute the factor content of trade, and divides thetrade-induced changes in relative effective factor supplies by some estimated orassumed elasticity of substitution.If trade does not explain the bu
13、lk of the change in factor prices, what does? Theconventional answer is that technology is the culprit; in particular, that there hasbeen a pervasive skill-using bias in recent technological change, which has shifteddemand toward skilled and away from unskilled labor.But while the view that recent c
14、hanges in the distribution of income primarilyreflect technology rather than trade may be the majority opinion, it has beenharshly criticized by some trade economists, especially by Leamer (1998, 2000).The critique by Leamer and others may be summarized as follows:1. The observation that the volume
15、of trade between low-wage and high-wagecountries is small is irrelevant: prices rather than quantities are what matter, andprices are determined on the margin.2. The attempt to estimate the impact of trade by looking at its factor content is anonsensical exercise, betraying a failure to understand b
16、asic trade theory.3. The factor bias of technological change is also irrelevant: trade theory tells usthat what matters is the sector in which technical progress occurs, not the factorbias of that change.This critique would be very serious, if correct. However, this paper will arguethat the critique
17、 in fact misses the point, essentially because the critics undertakethe wrong thought experiments. Trade volumes are not irrelevant: if one poses theP.R. Krugman / Journal of International Economics 50 (2000) 51 71 53question correctly, one immediately realizes that small trade volumes are inconsis-
18、tent with a story that attributes large distributional effects to trade. The factor biasof technological change is not immaterial, except in the case where such changetakes place in a small open economy (as opposed to one that can affect worldprices), and where technical change occurs only in that e
19、conomy (rather thanoccurring simultaneously in other economies as well); since the real situation doesnot meet either criterion, factor bias definitely does matter. Most surprisingly, themuch maligned use of a factor content approach to infer the effects of trade onfactor prices turns out to be an e
20、ntirely justified procedure when carefully applied.This paper is in six parts. It begins with the impact of technology on factorprices. Section 2 reviews the standard analysis of technology and factor prices in aone-good economy, while Section 3 turns to a multiple-good economy, discussingthe relati
21、ve roles of sector and factor bias of technological change. Section 4 offersa discussion of the relevance of trade volumes to the assessment of trades impacton factor prices. Section 5 develops a geometric approach to the relationshipbetween factor prices and factor supplies. Section 6 then uses thi
22、s machinery toshow the validity of the factor content approach when making but-for analysesof the effects of trade on income distribution. Finally, Section 7 reviews the debateand asks how it can have gotten so far off track.2. Technology and factor prices in a one-good economyA useful starting poin
23、t for any discussion of the impact of technology on factorprices is the analysis first introduced by Hicks, which showed how the effects oftechnical progress in a one-good economy depend on its factor bias.Consider a constant returns, competitive economy that produces a single1aggregate output using
24、 two inputs, skilled labor (S) and unskilled labor (U ). InFig. 1 curve II is the initial unit isoquant. The slope of the ray OE is the aggregateratio of unskilled to skilled labor in the economy, and the slope of ww is the ratioof skilled to unskilled wages.Now consider the effects of technical pro
25、gress, which can be represented as aninward shift of II, say to I9I9. If the relative supplies of skilled and unskilled laborremain unchanged, the new relative wage rate will be determined by the slope ofI9I9 where it crosses OE.Clearly, the effect of technological change on factor prices depends on
26、 the biasof that change. If technical progress is Hicks-neutral, that is if the unit isoquant1Throughout this paper I will think in terms of a two-factor economy in which the two factors areskilled and unskilled labor. Why not capital and labor? Because the empirical fact is that while the skillprem
27、ium has risen sharply, the share of compensation in national income has been quite stable. Ideallywe would work with three or more factors, but to do so would obscure the important issues treated laterin the paper.54 P.R. Krugman / Journal of International Economics 50 (2000) 51 71Fig. 1. Technology
28、 and factor prices in a one-good economy.simply shifts radially inward, there will be no change in relative factor prices. Iftechnical progress is skill-biased, that is if the ratio of skilled to unskilledemployment will rise at any given wage ratio, then the effect of this technicalprogress will, a
29、s shown in Fig. 1, be to raise the skill premium to a level indicatedby the slope of w9w9.Notice that it is quite possible that technical progress will actually lower the realwages of some workers. The intercept of w9w9 with the vertical axis measures theamount of unskilled labor necessary to purcha
30、se one unit of output, i.e. the inverseof the real wage of unskilled labor. As drawn in Fig. 1, this real wage has clearlydeclined.This is all standard theory, more than 60 years old. What makes it relevant isthat there is overwhelming evidence that recent technological change has indeedbeen strongl
31、y skill-biased. The essential point was made clearly by Lawrence andSlaughter (1993): although the wage premium associated with education has risensharply since the 1970s, which should other things equal have led to a substitutionaway from skilled labor, in fact there has been a rise in the college-
32、educated shareof employment, not only in the economy as a whole, but within almost everyindustry.Does skill-biased technological change, then, explain the rise in the relativewage of skilled workers? I will turn to critiques of this idea in the next section, andto the claim that trade rather than te
33、chnology is the culprit later in the paper.Before getting to these questions, however, it may be worth mentioning onepotentially worrisome issue that arises even if one is willing, for the sake ofP.R. Krugman / Journal of International Economics 50 (2000) 51 71 55argument, to think of the economy as
34、 if it produced only one good. The issue isthe following: has the growth in total factor productivity been sufficient to be2consistent with the large changes we have actually seen in factor prices?To see why this might be a problem, consider Fig. 2. Here, II represents anestimate of the unit isoquan
35、t at some initial date, say 1973, and E shows the unitinputs of skilled and unskilled labor at that date. At some later date, say 1989, weobserve the unit inputs described by E9, and the factor prices indicated by w9w9.The situation shown here is one in which the growth in average labor productivity
36、has not been very large, indeed in which the input of skilled labor per unit ofoutput has actually increased; but factor prices have changed substantially. Is thisoutcome consistent with a technological explanation?The answer is no. One thing we know about technical progress is that oldtechnologies
37、remain available; geometrically, that means that the new unitisoquant cannot lie outside the old one at any point, and therefore also that anyfactor price line tangent to that new unit isoquant cannot cross the old isoquant.This criterion is clearly violated in Fig. 2. In fact, in this case output w
38、ouldliterally be cheaper to produce at 1989 factor prices using the 1973 inputcoefficients. If the real data looked like this, we would therefore be entitled toFig. 2. When technology cannot be the explanation.2This concern was suggested to me by Kenneth Arrow. It should not be confused with the arg
39、umentsometimes made that technological change cannot explain wage changes because the rate of growth oftotal factor productivity has not accelerated. This argument involves a crude confusion between the rateof TFP growth and the bias of that growth.56 P.R. Krugman / Journal of International Economic
40、s 50 (2000) 51 71conclude that technology is not a sufficient explanation for the change in factorprices. (Even if it is not literally cheaper to produce using the old inputcoefficients, we would still reject the technological explanation if the new factorprice line crossed an estimate of the old un
41、it isoquant based on a reasonableelasticity of substitution.)We see, then, that a technological explanation of changes in factor prices is nota tautology: even before we get to issues posed by international trade, we mustface the possibility that a technology explanation will be internally inconsist
42、ent.This is most likely to happen if there are large factor price changes over a periodof small improvements in productivity, which sounds qualitatively like a gooddescription of the last 20 years in the United States. Before we proceed, then, wehad better make sure that the data do not reject a tec
43、hnological explanation out ofhand.Fig. 3 shows some relevant data for the United States, based on wage data fromBernstein and Mishel (1994) and on productivity data from the Economic Reportof the President. Skilled workers are identified with college-educated workers;unskilled workers with all other
44、s. The 1973 isoquant is an estimate of the unitisoquant based on an elasticity of substitution of 1. It turns out that thetechnological explanation passes this test: although US productivity growth hasbeen disappointing, it has been large enough that even with a reasonably largeelasticity of substit
45、ution the data are consistent with a factor-bias explanation ofchanging factor prices. While this by no means demonstrates that a technologystory is correct, it does show that it is feasible.But there are some trade economists who assert that the whole issue of factorbias in technology is irrelevant
46、, that while factor bias may matter in a one-sectorFig. 3. A feasability test of the technology story.P.R. Krugman / Journal of International Economics 50 (2000) 51 71 57model, when we consider trading economies with multiple sectors it ceases to haveany impact on factor prices. To assess this claim
47、, we must now extend the model.3. Technology and factor prices in multi-good modelsTo understand the objections of Leamer and others to analyses that stress thefactor bias of technological change, let us now consider an economy that usesskilled and unskilled labor to produce two goods, a skill-inten
48、sive good X and an(unskilled) labor-intensive good Y. Let us initially assume that relative goodsprices may be taken as given. We will see shortly that this is a very misleadingassumption, but it is important to understand the logic.Fig. 4 uses a Lerner diagram to represent the equilibrium of this e
49、conomy. Thetwo curves XX and YY are not unit isoquants: they are equal value isoquants,each corresponding to the same value at world prices as the other. (Thus eachmight represent $1 million worth of its respective good.)If the country is to produce both goods, factor prices must be such that $1million of X and $1 million of Y cost the same amount to produce. Thus therelative wage must equal the slope of the line ww that is tangent to both isoquants.And we can then check to confirm that the country will actually produce bothgoods: it will do so if and only if its e
