International Knowledge Spillovers and Wage Inequality in Developing Countries
？Ming-cheng Wang, Chen-ray Fang and Li-hsuan Huang
In this paper, international knowledge spillovers are incorporated in a horizontal innovation model, designed to explain the observed uncertain effects that openness of trade can have on wage inequality in small developing countries. Openness of trade can produce two different effects: an increase in the relative price of less-skilled labor-intensive products and a wider skill discrepancy due to knowledge spillovers from the more to less developed country. The former triggers a fall in the wage premium, while the latter widens the wage premium gap in the developing country. These two opposing forces explain the observed uncertain effects of openness to trade on wage inequality in developing countries.
JEL Clasification: F110, F430, J310, O300, O410.
Keywords: international knowledge spillovers, wage inequality, developing country,
endogenous growth model
？ Ming-cheng Wang, Associate Professor, Department of Economics, National Central University, Chungli 320, Taiwan, R.O.C. Tel.: 011-886-3-4227151 ext. 66307, E-mail: firstname.lastname@example.org.
Chen-ray Fang, Associate Professor, Department of Economics, National Taipei University, 151, University Rd., San Shia, Taipei, 237, Taiwan, R.O.C. Tel.: 011-886-2-86747132, Fax: 011-886-2-26739880, E-mail: email@example.com.
Corresponding author. Li-hsuan Huang, Professor, Department of Economics, National Central University, Chungli 320, Taiwan, R.O.C. Tel.: 011-886-3-4227151 ext. 66306, Fax: 011-886-3-4222876 E-mail: firstname.lastname@example.org.
1The conventional Stolper-Samuelson theorem in the trade literature predicts that
openness of trade will lead to a decrease in wage premium in a country if it is relatively unskilled-labor abundant. Increased openness should lead to less wage inequality in developing countries where less-skilled labor is often relatively abundant. The empirical evidence for this prediction is however mixed. Despite the fact that increased openness did lead to a decrease in wage inequality in the 1960s in Taiwan and South Korea and in the 1970s in Singapore [Robbins (1996); and Wood (1996)], this was not the case in some
2of the Latin American countries in the 1980s and 1990s. We can thus find different
evidence of the effects of openness on wage inequality in developing countries. The Stolper-Samuelson mechanism seems to ignore factors that increase wage inequality when developing countries open up to trade. This leaves space for theorists to fill the void and look for theoretical models that are capable of explaining the above diverse wage inequality movements.
In the literature, studies on determination of the wage inequality between skilled labor and less-skilled labor either generally focus on developed countries or neglect certain trade issues [Aghion, Caroli and Garcia-Penalosa (1999); Galor and Moav (2000);
3Aghion (2002) and Acemoglu (2002); Fang, Huang and Wang (2008)]. Until recently,
one line of trade research has focused in particular on explaining the wage inequality in developing countries. In these studies, a variety of mechanisms are proposed to model the impact of trade openness on the observed diverse wage inequality movements in developing countries such as outsourcing [Feenstra and Hanson (1996)], nontraded goods
1 It refers to the restrictive version of the Stolper-Samuelson theorem in Deardroff (1994). 2 For instance, as cited in Zhu and Trefler (2005), just over half of 20 developing and newly industrialized countries in the Freeman and Oostendorp (2001) wage database experienced rising wage inequality throughout the 1990s. 3 For instance, Acemoglu (2002) develops a model that stresses the market size effect and the endogenous technology choice, reaching the conclusion that increased openness leads to an increase in wage inequality in the northern countries.
with tariff reductions [Xu (2003)], preference heterogeneity [Glazer and Ranjan (2003)], product cycles with exogenous northern (developed country) innovations [Zhu (2004)],
4southern (developing country) catch-up [Zhu and Trefler (2005)], trade fragmentation
[Marjit, Beladi and Chakrabarti (2004)], and differences in trade policies before trade
5 liberalization [Ripoll (2005)].
In this study we examine the above issue but offer a different mechanism. We incorporate both international knowledge spillovers (hereafter IKS) and intra-national (sectoral) knowledge spillovers into Romer’s (1990) horizontal innovation model.
6Incorporating sectoral knowledge spillovers is common in a two-sector model setup. The
justifications for incorporating IKS into the north-south trade model are twofold. First, empirical evidence for knowledge spillovers from developed countries to developing countries is significant. Coe and Helpman (1995) and Lichtenberg and Potterie (1998) consider that international R&D expenditure significantly enhances total factor productivity; the effects are more profound with the degree of the openness. Moreover, Coe, Helpman and Hoffmaister (1997) document that 1% of R&D expenditure in
7developed countries raised productivity in developing countries by 0.06%. Second,
recent theories of economic growth identify trade as a major vehicle for IKS, which makes available to the developing country products that embody foreign knowledge,
4 Unlike Zhu and Trefler (2005), the innovation rate is endogenous in our model. 5 Fischer (2001), Beaulieu, Benarroch and Gaisford (2004) and Long, Riezman and Soubeyran (2004) provide other explanations. Fischer (2001), for example, uses a dynamic specific factors model to explain the income inequality following trade liberalization and concludes that income inequalities in land (labor)-abundant countries increase (decrease) dynamically after trade liberalization. 6 See Fang, Huang, and Wang (2008) for examples of the sectoral spillover: the space-age technology of visco-elastic foam used in manufacturing memory foam pillows and beds to promote good quality sleep; and the medical lab-technology used in cosmetics services. The presence of sectoral knowledge spillover is also evident as shown in empirical studies by Bernstein and Nadiri (1988), Griliches (1992), Branstetter (2001), and Keller (2002). 7 Engelbrecht (2002) utilizes data from 77 developing countries to confirm the positive role of human capital in the absorption of international knowledge spillovers. Bernstain and Mohnen (1998) and Branstetter (2001) documents that international knowledge spillover exists in firm level data.
8thereby enhancing productivity. It is also likely to open communication channels that
stimulate the adoption of foreign ideas.
The main purpose of this paper is thus to propose a theoretical framework, incorporating trade and international knowledge spillovers, that is capable of predicting the above-mentioned uncertain effects of openness on wage inequality in developing countries. The logic is straightforward. On the one hand, openness of trade facilitates IKS, which in turn increases the relative demand for skilled labor and pushes up wage inequality in developing countries. New technologies obtained by the skilled (in the developing countries) will to some extent diffuse to less skilled workers thereby leading to an increase in the demand for less-skilled labor. The net effect of these two kinds of knowledge spillovers on the equilibrium wage inequality in the developing country is herein referred as the skill discrepancy effect. On the other hand, in a country that is
relatively unskilled-labor abundant, openness of trade can lead to a decrease in wage premium, which we refer as the price effect. These two opposing effects can both affect
wage inequality in a small developing country after it has been opened to trade.
If it is, in fact, the relative strength of the above two forces that determines wage inequality in small developing countries, then the effect of increased trade openness on wage inequality could vary at different time periods and in different countries. Since it is plausible that the effect of IKS was larger in the 1980s than in the 1960s, we interpret the widening of wage inequality that occurred in some Latin American countries (for example: from the mid-1970s to the early 1980s in Argentina and Chile; between the mid-1980s and the mid-1990s in Colombia, Costa Rica and Uruguay; from 1984 to 1990 in Mexico [Zhu and Trefler (2005)]) was a result of the dominance of the skill
8 Grossman and Helpman (1991), Chuang (1997), Murat and Pigliaru (1998) and Ben-David and Loewy (2000) incorporate international knowledge spillovers into theoretical models to examine the relationship between trade and growth. These models, however, consider only one type of labor and are therefore unable to examine the effects of the increased openness on wage inequality in developing countries.
discrepancy effect, while the declining wage inequality in Taiwan and South Korea, after trade was opened up in the 1960s, was a result of the dominance of the price effect. Our
proposed model can thus explain diverse wage inequality movements in the empirical evidence in developing countries that have been opened up to trade. This complements the explanation of the southern wage inequality mentioned in the literature above.
The remainder of the paper is organized as follows. In Section 2 we outline the model for determining wage inequality under free trade. The home country is specified as being a small country, with a relative abundance of less-skilled labor. International knowledge spillovers are built into the R&D sectors. In Section 3 we compare results for the international trade regime with those under autarky. In the final section some conclusions are offered.
2. The Model
Consider an economy, endowed with less-skilled denoted by and skilled labor L
, which produces two final goods, traditional goods X and advanced goods Y. Both H
types of labor are either combined with specific capital goods to produce final goods or hired by R&D firms to design blueprints to produce the associated capital goods. Specific factors are utilized to produce both types of final goods, in the sense that the traditional final goods are produced by less-skilled labor and the capital goods that are associated with this less-skilled labor; the advanced final goods likewise. In the same vein as in the final goods sectors, specific labor is hired to produce blueprints in the two R&D sectors. Less-skilled labor is the sole input in the traditional R&D sector, while skilled labor is the only input for the advanced R&D sector. Both the traditional and advanced capital goods sectors manufacture the associated capital goods.
The following simplifying assumptions are made. First, the home country is a small
open economy; the foreign country (i.e., the rest of the world) is essentially a closed economy. All the foreign variables are denoted with the superscript *. The home economy is thus a price taker of the world’s final output and global interest rate. Second, both X
and are tradable. The home country has relatively abundant less-skilled labor in the Y
***sense that with, meaning that under free trade the home country HH！L/H(L/H
has a comparative advantage in the traditional output. All other parameters are the same for both countries. Third, capital goods are not traded, labor does not move across the border, and international knowledge only spills over from the advanced R&D sector of the foreign country to its counterpart in the home economy. Finally, traditional goods X
are treated as the numeraire.
The production of final goods is specified as
Z1，；；XLudj，！！(), 01,； (1) Xj?0
A1，；；YHvdj，(), (1)' Yj?0
Lwhere H and are the associated labor inputs for producing and , XYXY
uvrespectively; and are the employment of the th type of capital goods for jjj
producing and , respectively; and and are the measures of the varieties of XYZA
capital goods in the traditional and advanced sectors, respectively, with specified to A
be larger than to ensure the positive skill premium. The evolutions of the measures of Z
varieties for the home country are specified as follows:
(dZ；；1， )，~, ~(0, 01,！！； (2) ZAZLZdt
(*?1，?A，：(A)AH,：(0,0！?！1. (2)’ A
In equation (2), intra-national (sectoral) knowledge spillover is allowed in that technology spills over from the advanced R&D sector to the traditional R&D sector with
as the sectoral spillover efficiency and . The IKS is allowed in equation (2)’ ；0！；！1
in that technology spills over from the advanced foreign R&D sector to the advanced
9home R&D sector with being the IKS efficiency and . Note that and 01 ！！??~
indicate the production efficiency of the two R&D sectors, respectively. The term L：Z
() is the only paid input in the traditional (advanced) R&D sectors. Intra-national HA
spillover from skill intensive industries and/or firms to their less skilled counterparts is
10；(/)AZevident. It is noted that in equation (2) captures the one-way positive sectoral spillover, and the effect increases with the skill discrepancy between the advanced and
11the traditional R&D sector, i.e.,. AZ/
*?(/)AALikewise in equation (2)’ captures the IKS effect in the skilled home
*R&D sector. contributes to the evolution of so as to echo both the theoretical AA
and empirical findings that openness to trade enables the developing country’s products
to embody foreign knowledge and to enhance productivity [Grossman and Helpman (1991), Ben-David and Loewy (2000), and Branstetter (2001)]. Despite missing a direct effect for the IKS on the traditional R&D sector, equations (2) and (2)’ together do imply
*that the evolution of is indirectly affected byA. Z
Assume that markets for final goods are competitive. The objective function of the
ZZ1，；；?，，，()(),LudjPuudjWLrepresentative firm in sector is where XXXjjjLX??X00
P(u)W is the rental price of the traditional capital good; and is the wage rate of jjLX
uu，L. Under a symmetric setup (), the optimality conditions are jX
9 is the case with no IKS. ?，010 See for example, Fang, Huang, and Wang (2008); Bernstein and Nadiri (1988); Griliches (1992), Branstetter (2001); and Keller (2002) for theoretical concerns and empirical support. 11; We can alternatively specify equation (2) to be . The results are, however, Z，~[；A？(1，；)Z]Lz
similar. It is also noted that the model reduces to the case with no sectoral spillover if . ；，0
，；； and (3) WLZu，，(1)()；LXX
11，，；；PuLu()()，； . (4) X
Similarly, the optimality conditions for the representative firm in sector are Y
，；； and (3)’ WPHAv，，(1)()；HYYY
11，，；；PvPHv()(),，； (4)’ YY
Wwhere is the wage rate for ; is the price of ; and is the rental PP(v)HYHYYY
price of the advanced capital good.
For simplicity, we assume that one unit of capital good is produced by one unit of the associated final good, and that the two capital goods markets are monopolistically competitive. The instantaneous profit functions for the capital goods producers are
?，，Puuru() and (5) u
?，，PvvrPv(), (5)’ vY
rwhere is the real interest rate. Using equations (4) and (4)’, the optimality conditions
of the capital goods producers are
211，，；；；()Lur， and (6) X
211，，；；；()Hvr，. (6)’ Y
Substituting equations (4) and (6) into equation (5) and equations (4)’ and (6)’ into
equation (5)’, respectively, we obtain
1，； ru and (7) ，?u；
；1，? rPv. (7)’ ，vY；
We assume in a competitive market that the R&D firm in each sector will sell its
12rblueprints to the capital goods producers at prices and . With a constant , PPPZAZ
and can be derived as follows: PA
?1，；rt， and (8) Petdtu，，()?Zu?0；
；，1?rt，? . (8)’ ，，Pe(t)dtPv?AvY0；
Due to free entry, these are the prices of the blueprints that make the monopolistically
competitive firms with zero economic profit.
As to the two R&D sectors, the representative firm’s profit functions are
?，，PZWL?，，PAWH and , ZZLZAAHAZA
WWLHwhere and are the wage rate for and , respectively. Using equations LHZAZA
(2) and (2)’, the optimality conditions are
；；1， and (9) WPAZ，~LZZ
？?1，? (9)’ W，P：(A)A.HAA
Equation (9) says that the real wage paid by the R&D firm is equal to the marginal
；(AZ)product of low skilled labor, which depends upon the sectoral spillover, i.e., , and
equation (9)’ says that the real wage received by the skilled labor depends upon the
？?(AA)international spillover, i.e., .
The equilibrium in the two labor markets requires
L？L，L and (10) XZ
H？H，H. (10)’ YA
Assume free mobility for each type of labor between the associated final goods sector and
the R&D sector. Arbitrage then implies that
12r is constant in a balanced growth equilibrium.
WWW，， and (11) LLLXZ
WWW，，. (11)’ HHHYA
Using equations (3), (9), (3)’ and (9)’, equations (11) and (11)’ give rise to
；1，；Z??，；； and (12) PLu，()ZX，?A~??
；1，A，；；? (12)’ ()().P，PHvAYY？：A
These are blueprint prices that allow R&D firms to employ labor to engage in their R&D activities. Labor arbitraging between the final goods sector and the R&D sector results in
；P(/)ZAa that rises with . The logic is as follows. The higher the skill Z
；(AZ)discrepancy is, the higher the productivity of the traditional R&D sector that can be obtained through sectoral spillover. Other things being equal, this causes the supply of
Pless-skilled blueprints to increase and thus lowers . The same reasoning applies to the Z
？?(AA)result that P rises with . A
PEquating from equations (8) and (12) and P from equations (8)’ and (12)’ ZA
~A；；；，1，()() and (13) uLXZ；
*：A；?；，1，()()vH. (13)’ YA；
These two equations demonstrate the relationship between the employment of capital goods and labor in the two final goods sectors. When the skill discrepancy is larger, unskilled labor benefits more from the sectoral spillover and thus less capital goods are needed to produce the same amount of traditional final goods.
The demand side of the economy is specified below. The optimization problem