Faculty of Economics Master Thesis
June – August 2007
Income Inequality and Openness to Trade
René van der Haar
2
Abstract
Both decreases in income inequality and economic growth can be important goals from a policy perspective. These goals are especially important for developing countries because both
phenomena can lead to alleviation of absolute poverty. There is widespread consensus among economists that openness to trade is an important policy instrument through which economic growth can be achieved. In addition, traditional trade theory, based on the Heckscher-Ohlin-Samuelson model of international trade, suggests that for countries abundant in unskilled labor, the returns to unskilled labor will increase in both relative and real terms as a result of trade. This proposition about changes in factor returns after trade liberalization only finds weak support in the empirical literature however. In fact, most of the available evidence points in the other direction; openness to trade is associated with increases in wage differences between skilled and unskilled and with increases in overall income inequality.
3
Acknowledgements
I would like to thank my thesis supervisors, drs. A.R.M. Gigengack and dr. E.H. van Leeuwen, for taking the time and energy to get me started on writing this thesis and for their support during the writing process.
4
1.Introduction
The neo-classical theories in economics argue that economic liberalization towards free markets lead to greater efficiencies, both within and between countries. This believe has a long tradition stemming back to 1776 when Adam Smith was the first to formalise the idea in his classic book The Wealth of Nations. The efficiency gains are present in both the production and the
consumption of goods. Regarding the production, free markets allow for efficient factor
allocation between countries and technology transfer. Through increased competition in the final goods markets, free trade causes prices to converge and monopolistic rents to decrease. If the conditions for perfect competition are met, aggregate welfare will rise for each country that liberalizes its economy by opening up for trade. As a result, within a country, the winners of the change to the free international market gain more than they need to compensate the ones that lose in the process of trade liberalization. Therefore, free markets are assumed to be an important catalyst of growth and anyone can benefit from it. The belief in the efficacy of free markets is the driving force behind the process of economic globalization.
Globalization is a phenomenon that embodies several simultaneous developments that reflect increased economic interdependence of countries. It refers to, inter alia, falling policy barriers to trade, declining transportation costs, outsourcing, immigration, improved means of
5
of some empirical studies testing the hypothesis of the HOS-model concerning the change in wage difference between skilled and unskilled worker, the skill premium. Chapter 5 will offer some modified HOS-models to provide theoretical explanations of the mixed empirical results found in the previous chapter. Chapter 6 will focus on the results of empirical studies that look at the relation between trade openness and overall income inequality. Chapter 7 will have an
exclusive focus on empirical studies that take a country’s factor endowments into account when explaining the link between income inequality and trade. Chapter 8 will discuss some other aspects of the relationship between trade openness and income distribution that are not easily modelled. In Chapter 9 I formulate a testable specification linking inequality to trade. This specification will be tested on gathered data. Chapter 10 will conclude.
2. Poverty and inequality
The focus in this paper will be on developing countries’ income inequality and how it relates to openness to trade. Absolute income inequality describes the absolute differences in income between individuals or groups within societies. However, because absolute differences in income levels will depend on a country’s mean income, this measure is unsuitable for inter-country comparison. Relative income inequality compares the income of a group or individual to the income of other groups or individuals; it describes the dispersion of incomes within or between countries. Because this is a relative measure, comparing income levels with each other, a rich country can have a more equal income distribution while having larger absolute income inequalities than a less wealthy country. In the rest of this paper, when I use the term income inequality or simply inequality, I refer to relative income inequality.
2.1 The link between poverty and inequality
6
goes to the people below the poverty line, some of them may pass the poverty threshold and reduce absolute poverty within the country. Another instrument to measure poverty is the Poverty Index. The index takes both the number of people below the poverty line into account as well as the extent of their poverty. Obviously, the links between reductions in the poverty index and growth and inequality are the same as with the poverty line. Growth and inequality improvements are thus both important mechanisms in poverty alleviation. For a survey of recent literature on the relationships between, poverty, growth and inequality, see Lopez (2004).
According to the World Bank (2001b):
“For a given rate of growth, the extent of poverty reduction depends on how the distribution of income changes with growth and on initial inequalities in income, assets and access to opportunities that allow poor people to share in growth … how growth affects poverty depends on how the income generated by growth is distributed within a country” (p. 52).
As noted in the introduction, there is widespread agreement that trade openness is associated with economic growth. The discussion above makes clear however, that if economic growth is combined with increases in inequality the effects on absolute poverty are ambiguous. Therefore, it is important to look for systematic relationships between trade openness and inequality. For a review of the links between trade and poverty, see Winters (2000) and Winters et al. (2004).
2.2 Measures of inequality
Several measures of relative income inequality measuring the extent and the distribution of income inequalities have been developed.
The Theil-index decomposes total inequality in inequality between groups. The Theil-index is the weighted sum of inequalities within subgroups of the entire population. The weight is the income of the group relative to the total income within the country. This index allows one to determine the amount of inequality that a group contributes to the total inequality.
Another measure that is often used is the income share in total income of the richest 20% of the population relative to the income share in total income of the poorest 20% of the population.
7
the origin represents the line of perfect income equality. This means that every person has the same amount of income. The horizontal line on the axis in combination with the vertical line from the 100% of households point represents the line of perfect inequality. In this case all the income in a country is earned by 1 individual. The line in the middle represents the actual income distribution and is called the Lorenz-curve. The Gini-coefficient is a summary statistic that quantifies the extent of income inequality depicted in a Lorenz-curve. It measures the surface of the area between the actual income distribution and the perfect equality line relative to the total area below the perfect equality line and the perfect inequality line. The value is therefore between 0 and 1. A value closer to 0 means less inequality and closer to 1 represents greater inequality. The Gini-index is the Gini-coefficient expressed as a percentage; it is the Gini-coefficient multiplied by 100%. Because the Gini-coefficient is a summary statistic, it only provides a general image of the actual aggregate level of income inequality. Not all the movements between different income groups are readily observable when using Gini-values. To capture all those effects, data on deciles or quintile shares in total income should be used.
Graph by Simon Kisane
8
Europe have the lowest Gini-coefficients and the most equal distribution of household incomes. Nonetheless, the authors argue that, for example, in Latin-America the traditionally high Gini-coefficients might have come from a very unequal access to rather limited opportunites, which is undesirable (p.6).
2.3 Measures of trade openness
Splimbergo (1999, Appendix B) points out that there is no general accepted measure of openness. He divides the openness measures used in the literature in “incidence-based” measures and “outcome-based” measures. Incidence-based measures, such as the level of tariffs, are direct indicators of trade policy. They suffer from 2 shortcomings however. The first relates to the unavailability of detailed data on tariffs. The second shortcoming is the fact that non-tariff barriers (NTB’s), such as licences or quotas, remain unmeasured.
Because of the problems with incidence-based measures, the literature uses outcome based measures more frequently. This measure implicitly covers all the sources of distortion imposed by a country’s trade policy. In addition, data on the trade outcomes are much more readily available. The most commonly used outcome-based measure is simply the ratio of the sum of exports and imports over GDP. Other outcome-based measures control for country characteristics, such as size or endowments, in measuring trade openness. Other openness measures focus on the price outcomes such as Barro and Lee’s (1994) Black Market Exchange Rate and the Price Distortion Index by Dollar (1992). However, the data necessary to compute these openness measures suffer from the same general unavailability as incidence-based measures.
Sachs and Warner (1995) combine several indicators of openness and classify countries as either open or closed. They judge a country’s trade policy as closed if has at least one of the following five characteristics: 1. Non Tariff Barriers covering at least 40% of trade. 2. Average tariff rates of 40% or more. 3. A black market exchange rate that is depreciated by 20% or more relative to the official exchange rate, on average, during the 1970s and 1980s.
9
3. Traditional trade theory and income distribution
David Ricardo was the first to model international trade. His conclusion that free trade leads to gains of trade is still the dominant view today. Gains from trade are derived from specialization of production. According to Ricardo, each country should specialize its production in goods in which they have a comparative advantage. This means that countries should produce goods in which they have the smallest relative cost of production compared to other countries, as opposed to the smallest absolute cost of production. In Ricardo’s model, there are 2 countries, 2 industries and labor is the only factor of production. The productivity of labor is assumed to vary across countries, which implies a difference in technology between nations and this is the motivation for advantageous trade flows between nations. As there is only one factor of production, the model can only be used to explain trade flows but doesn’t say anything about income distribution.
Eli Heckscher and Bertil Ohlin extended the ricardian model to incorporate 2 factors of
production, labor and capital. Capital refers to physical capital such as machines and equipment used in production. The returns that the owners of these production factors get in return for their use in the production process are called “wages” for laborers and “rents” for capital owners. The factors of production can move freely between the two industries but are immobile
internationally. The Heckscher-Ohlin (HO) Model assumes that different industries, producing different goods use the production factors in different proportions. The ratio of the quantity of capital to the quantity of labor used in production is called the capital-labor ratio. The assumption of different capital-labor ratios between industries gives the HO-model its generic name: The Factor Proportions Model. An industry that uses a higher capital-labor ratio than the other is said to be capital-intensive. By definition, the other industry is labor-intensive. Countries possess different quantities, or endowments, of capital and labor that are available for production. A country with a higher aggregate ratio of capital to workers are said to be capital-abundant and labor-scarce relative the other country. The other country is then said to be labor-abundant and capital-scarce. Contrary to the Ricardian model, the factor proportion model assumes that
10
goods in which they have a comparative advantage. The Hechscher-Ohlin theorem concerns itself with the pattern of trade between nations. This pattern is determined by the endowments of production factors of nations. The theorem states that a capital-abundant country will export the capital-intensive good and import labor-intensive goods while the labor-abundant country will import capital-intensive goods and export labor-intensive goods.
The Stolper-Samuelson theorem (SS, 1941) links changes in goods prices to changes in the returns to the production factors within the context of the HO-model. By focusing on factor returns, the SS-theorem was the first theoretical formulation to explain the effects of free trade on income distribution among factors of production. The SS-theorem states that when the price of a good rises, the return to the factor that is intensive in that good will rise as well.
Thus if for whatever reason the price of a capital intensive good rises, then the rental rate of capital will rise while the wage rate will fall. The opposite happens to the factor returns when the price of a labor intensive good rises.
The theorem has been generalized by Ronald Jones. Jones (1965) constructed a magnification effect for prices within the HO-model. The magnification effect allows for analysis of the effects of price changes in the goods market on the real wages and real rents earned by factor owners. Real returns indicate the purchasing power of wages and rents after taking the price changes into account. The real returns are therefore a better measure of wellbeing than the wage rate and rental rate alone. This extension is very useful as trade liberalization will change prices. The result of the magnification effect is that a movement to free trade will cause the real return of a country’s relative abundant factor to rise, while the real return of the country’s relative scarce resource will fall. It doesn’t matter in what industry the factor is employed. Therefore, when the price of a capital-intensive good rises, the real rental rate for capital will rise in the labor-intensive sector by the same amount as in the capital intensive sector. The link between prices and factor returns is discussed next.
11
to a new equilibrium there will be excess demand for capital, increasing its returns, and excess supply for labor, bidding down its price. The wages for workers will fall more than proportionally to the decrease in the import competing good, lowering their real wages.
The theorem makes clear predictions about the effect of trade liberalization on changes in factor returns; the real return to the factor in which the country is abundant will rise and the real return to its scarce factor will fall.
Samuelson (1948, 1949) extends the SS-theorem to show that, under certain conditions, international trade may lead to an equalization of the returns to production factors across
countries. The theorem is called the factor-price equalization theorem. It states that international trade will lead to an equalization of goods prices between countries and this will lead to an equalization of real factor prices between countries. The result is thus that throughout the world, labor will earn the same wages per unit of labor and capital will earn the same rents per unit of capital. The theorem predicts that when countries with different relative factor endowments engage in free trade, the real wages and real rents will converge to a point somewhere in the middle of the pre-trade relative levels between the countries. As a consequence, in the labor abundant countries real wages will rise and real rental rates will fall while the opposite result holds for capital abundant countries
The assumptions necessary for complete factor price equalization, however, are stark. The assumptions include that the same technologies are available for all countries at the same costs, there are no import taxes or transportation costs, factors of production cannot move between countries and are qualitatively the same. When these conditions are met, the real prices of factors of production will be the same in equilibrium.
The Rybczynski theorem (1955) demonstrates the relationship between an increase in the supply of one of the production factors, while holding the supply of the other factor constant, and
changes in outputs of the final goods within the context of the HO-model. The theorem states that an increase in a country’s endowment of a factor will result in an increase of output of the goods which uses that factor intensively, and a decrease in output of the other good.
12
labor. This can lead to a price reduction of labor intensive goods around the world, which in turn, will affect the return to labor everywhere.
The Heckscher-Ohlin model and the theorems derived from it is the work horse of international trade economists when assessing the long run distributional consequences of trade liberalization (Markusen, 2005). The traditional HO-trade model outlined above assumes that labor is a homogenous factor of production. More recent literature concerning the link between income distribution and trade liberalization extends the HO-analysis by considering capital, skilled labor and unskilled labor as the relevant factors of production. The main reason for this extension is the assumption of the complementarities between capital and skilled labour.
Labor is no longer seen as a single production factor. Instead, the labor force is divided in skilled and unskilled labor. International trade can affect the returns to unskilled and skilled labor differently. The conclusion of the HO-model remains unchanged however: Opening of trade will increases the return to the country’s abundant production factor.
If one assumes that developing countries are abundant in unskilled labour, the HO-model in combination with the Stolper-Samuelson theorem (from here on called the HOS-model) predicts a reduction in wage inequality between skilled and unskilled workers in those countries.
According to the World Bank (1999):
“Unskilled labor is relatively abundant in developing countries. In the context of the Heckscher-Ohlin model, trade reform can be expected to increase the overall demand for such labor in the long run. This follows since such countries have a comparative advantage in goods that use unskilled labor intensively. Removing policies that favor import-competing sectors at the expense of (labor-intensive) export sectors ultimately results in an expansion of the latter and contraction of the former. Any increase in the demand for unskilled labor results in a combination of higher wages and employment for this segment of the population”.
13
Several methodological approaches have tried to test the predictions of the HOS-theorem. One such approach is time series studies examining the change in wage gap between skilled and unskilled workers in the face of trade liberalization, see Chapter 4. The arrival of improved data on income inequality in the mid 1990s made cross country studies concerning the link between aggregate income inequality and trade openness possible, see Chapter 6. The latest class of studies also looks at aggregate income inequality and trade openness but explicitly account for countries’ endowments of production factors, see Chapter 7.
4. Empirical results from wage studies.
Studies testing whether trade openness has lead to a decrease in skill premiums, the difference between skilled and unskilled wages, also called the wage gap, use time series analysis. Though these studies have the advantage of being able to directly look into the factor proposition of the HOS-model, these studies have been critized based on the fact that they only concern middle-income countries and are confined to the manufacturing sector (Gourdon, 2006). Wage inequality studies tend to focus on one of two groups of nations: the East Asian tiger economies that entered the global market in the 1960s and 1970s or the Latin American countries that engaged in
significant trade liberalization reform in the 1980s and 1990s.
The evidence presented for these 2 regions and time spans are very mixed; where the Asian countries, on average, saw falling wage gaps, the Latin American countries tended to experience surges in skill premia. Anderson (2005), Arbache (2001) and Wood (1997) review the available studies.
4.1.1 Latin America
14
workers. According to Anderson, in many of these studies the so called “supply-and-demand” approach is used. This approach assumes that the supply of skilled and unskilled workers is unaffected by their relative wages. With
E
s/E
ulabeled as the fixed supply of skilled to unskilledlabor,
a
as the relative demand for skilled labor,W
s/ W
uas the relative wage, andσ
as theelasticity of substitution between skilled and unskilled workers, the following expression can be used:
ln (
W
s/ W
u) = ln (a
/σ
) – (1 /σ
) ln(E
s/E
u)With this expression it is possible to econometrically test whether a proxy for openness to trade has a significant effect on the relative wage, controlling for relative supply. If so, then it can be inferred that openness to trade affected relative demand.
This approach has been used by Robbins (1996b) who studied the trade liberalization experiences of Argentina, Costa Rica, Colombia, Chile and Uruguay during the 1980s and the first years of the 1990s. He showed that in all five countries during each country’s period of trade liberalization, skill differentials in wages, measured by education, increased. In all five countries the supply of relative skilled workers has risen, indicating that the widening of the skill premium was caused by an increase in demand for skilled workers.
Subsequent studies using household survey data, on Chile by Beyer et al.(1999), and on Costa Rica by Robbins and Gindling (1999) using the same approach confirmed Robbins’ results. Hanson and Harrison (1999), using plant level data, studied the consequences of the 1985 trade liberalization in Mexico by examining changes in both employment and wages for skilled and unskilled workers. Although employment effects were very small, they found a significant increase in the wages of skilled labor. According to these authors, one of the factors that contributed to the wage increase for skilled workers was the fact that both domestic firms engaged in exports and foreign firms tend to pay higher wages to skilled workers than other firms. Beyer et al. (1999) found a positive relationship between openness and the wage gap. Using regressions for Chile in the 1960-1996 period, they showed that the wage gap between college graduates and workers with primary education rose with openness to trade, decreased with the share of college graduates in the labor force, and decreased with the relative price of tradable goods. Cragg and Epelbaum (1996) discovered that after trade liberalizations the returns to education rose in Mexico and this contributed to the increase in skill premium. These
15
Attanasio, Goldberg and Pavcnik (2004) studied the case of Colombia. They reported an increase in the skill premium of 20 percent following trade reform.
Robbins (1994a) studied the changes in the structure of wages after trade liberalization in Chile using the factor content method. He found that although more skilled labor is embodied in Chile’s imports than in its exports, the returns to skilled labor rose after trade liberalization. Both Abache et al. (2004) and Green, Felstead and Gallie (2001) studied the returns to education in Brazil after trade liberalization. Both studies found an increase in marginal returns to college education caused by rising relative demand for college educated workers.
4.1.2 East Asia
Wood (1994, 1999) reviews the available literature available from Asian countries that opened up in the 1960s and 1970s. The lack of observations for this region and for this time period prevents the use of sophisticated econometric techniques. The conclusions thus tend to rely heavily on judgement and general knowledge of the economy concerned (Wood 1994, p227).
Most of the time series evidence from this region refers to the four tiger countries. These countries are: Singapore, Taiwan, Hong Kong and the republic of Korea. According to Wood (1999), most of the evidence supports the, what he calls, “conventional view” that the adoption of more outward-orientated policies increases the demand for workers with only a basic general education relative to the demand for workers with more education and skills. In the 10 years after the implementation of the changed trade policy, the gap in wages between basic educated workers and skilled workers decreased in Korea and Taiwan in the 1960s, and in Singapore in the 1970s. The only exception is Hong Kong were the wage gap actually increased in the 1950s following trade policy change. Wood argues that this can be explained by considering the large influx of unskilled labor from China. Changes in the wage difference between uneducated workers and basic educated workers are not well documented, except for Singapore. The wage difference between these two groups actually increased in the decade after trade liberalization. However, because of the high proportion of primary educated workers in the work force, this effect was not strong enough to raise overall income inequality.
16
with the expansion of the labor-intensive export sector, aggregate demand shifted in favour of less-skilled workers. Robbins also analysed movement in wages by education level in The Philippines for the period between 1978 and 1988. No clear trend was found in the skill differential in wages and the relative importance of demand and supply pressures on relative wages was inconclusive.
Arbache (2001) concludes that overall, the available evidence on wage inequality does not confirm the HO- and SS-theorems. In fact, it goes in the opposite direction. He goes on to say that: “A tentative summary of empirical evidence would show a common feature of the impact of trade liberalization on labor markets in developed and developing countries, i.e., a change in the structure of labor demand in favour of skilled workers” (p. 11).
4.2 Some explanations for rising wage gaps
The apparent rise in the wage gap between skilled and unskilled workers after trade liberalization has led economists to look into other possible phenomena that might have caused the skill premium to rise. One of the reasons could be the fall in industry wage premiums. Another is the link between trade and technology which could explain the increase in skilled labour demand. Finally, given the pre-reform trade conditions, the rise in wage premium in the Latin-American countries might have been exactly what the HOS-model predicted. In the next section, these arguments will be reviewed.
4.2.1 Industry wage premiums and trade
If the assumption of perfect competition is dropped, trade liberalization could have another effect on the skill premium by changes in the industry wage premium. The industry wage premium is the difference in wages between individual workers that cannot be explained by normal observable traits such as, age, skill, work experience or education. These differences can only be explained by the industry in which they work.
17
Depending on whether the industries that face the largest tariff cuts are skilled labor intensive or unskilled labor intensive, the economy-wide wage gap can either decrease or increase as a result of changes in industry premiums. If it is mostly unskilled labor that benefits from an industry’s tariff protection a fall in that industry’s wage premium caused by trade liberalization will lead to an increase in the wage gap between skilled and unskilled workers.
Another mechanism through which trade policy can affect industry wage premiums is through its effect on industry productivity. Empirical studies show that in developing countries trade liberalizations are associated with growth in productivity (Topalova, 2004). If these productivity gains are shared with the workers in form of higher wages, an industry’s wage premium will increase after tariff cuts.
Revenga (1997) found that in Mexico trade protection led to the generation of rents that was passed on to the workers in the form of a wage premium. When trade became liberalized and protection barriers were removed, these wage premia were greatly reduced and this affected the wage structure. Since Mexico protected its industries that were intensive in unskilled labor more than other industries, the barrier reduction led to an increase in Mexican wage inequality.
Although relatively small in comparison to overall changes in skill premiums, Goldberg and Pavcnik (2004) find a positive correlation between industry wage premiums and trade protection in Colombia. No connection between tariffs and industry wage premium is found for Brazil however (Pavcnik et al, 2004).
4.2.2 Trade induced skill biased technological change
Other explanations as to why the wage gap has increased have emerged in the academic literature. Skill biased technologic change is seen as a very possible cause for skill premia to rise following trade liberalization in unskilled labour abundant countries.
Technology, capital and skilled labor are assumed to be complements. If the technological level of an economy is increased, so will the demand for skilled labor. The effects on wage inequality will be an increase as the ratio of the wages paid to skilled versus unskilled labour increases. It is possible that skill biased technological change increased wages for skilled labor independently of trade, so that the wage gap would increase without trade reform. On the other hand, if the technological change is a direct consequence of more openness, a link between openness to trade and wage inequality becomes apparent.
18
transfers occur either by a Southern country’s efforts to imitate technologies or through importing capital goods that are available in the North. Both these activities rely exclusively on skilled labor in the South. The liberalization of trade allows both easier imports of technology intensive capital goods from the North and it allows Southern firms to export their products to the North where it will come into contact with Northern technologies which it can learn and imitate. The introduction of new technologies in the South requires skilled labor. As the demand for these workers will increase, skilled wages will rise. According to Pissarides, the rise in skill premium might only be temporary. The speed of technology transfer slows down as the South learns more and the technology difference becomes smaller. Only if the transferred production technology is skill-biased will there be a positive long-term effect on the wages of skilled labor compared to unskilled wages.
19
in the medium to long run in the Southern countries. The inflow of technology may increase the demand for skilled labour by such a level that skill premia may rise in both the South and in the North.
Attansio, Goldberg and Pavcnik (2004), examining Columbia, found evidence for the endogenous response of skill biased technological change to trade liberalization. Using regression analysis they showed that demand for skilled labour increased the most in the sectors that faced the largest tariff cuts.
Gorg and Strobl (2002) used panel data of manufacturing firms in the 1990s for Ghana. They studied the effects of skill-biased technological change that were induced by imports of technology-intensive capital goods or by export activities on the changes in skilled relative wages. They estimated the relative demand function for skilled labor using a translog cost function. They found that greater inflow of foreign machinery did in fact change the technological structure in ways that are consistent with permanent skill-biased technological change. They found that firms that import more foreign equipment employ a higher ratio if skilled to unskilled workers.
4.2.3 The structure of the trade policy instruments
The implicit assumption in the HOS-model is that countries protect their import-competing sector in the pre-liberalization period. The import-competing sector is the sector in which the scarce factor is used intensely. The opening of trade will then lead to a relative and real rise in return of the abundant production factor. If unskilled labor is the abundant production factor, wage inequality will fall as a result. However, Wood (1999) argues that by the time the Latin American markets opened up for trade, these countries were not abundant in unskilled labor on a global level at all. In fact, the Asian countries that entered the global markets in the decades before, were still vastly more unskilled abundant than the Latin American economies. This means that Latin America did not have a comparative advantage in the manufacturing of low skilled products at all and the opening of trade must have led to a contraction of the unskilled labor sector and a rise in the wage premium.
20
the labor intensive sector had the highest import tariffs. The trade reform involved cutting tariffs to a new uniform rate across industries, so that the most protected industries experienced the largest loss of protection. The result is that instead of rising prices and factor return in the labor intensive sector after the opening of trade, these sectors now experienced increased foreign competition and lower prices. The factor returns of unskilled labor declined and the wage gap increased as a result. Roberston (2000) calculated that 50% of the rise in wage gap between skilled and unskilled labor can be explained by the trade reform-induced fall in the relative price of labor intensive products.
5. Modified HO-models to try to explain the anomalies.
Despite these possible explanations more complex models have emerged as a result of the mixed empirical performance of the standard model. These models extend the Heckser-Ohlin and Stolper-Samuelson models to try to explain how it is possible for wage inequality to increase after an increase in trade. In each of these models discussed below, one or more of the assumptions underlying the HOS-theorems is relaxed.
5.1 Heterogeneity in labour
21
NO-EDs and BAS-EDs will increase, causing an increase in wage inequality. If the proportion of uneducated workers is high enough compared to the basically educated workers, trade liberalization can actually increase an unskilled labour abundant country’s wage inequality. However, Wood argues that countries with a large supply of NO-EDs generally do not have a comparative advantage in manufacturing. Therefore, the net effect of increased manufacturing exports is likely to be a reduction of wage inequality even though the wage gap between uneducated labour and skilled labour remains more or less the same. The model makes the prediction that adopting a manufacturing export-oriented strategy causes a step change, or once and for all change, in relative wages that is spread out over a period of about ten years.
Gourdon (2006) presents empirical support for Wood’s model. He finds that openness lowers income inequality in countries in which the workforce is predominantly primary educated. In countries that are mostly endowed with either uneducated or higher educated workers openness has an inequality increasing effect.
5.2 Factor intensity reversals
Minhas (1962) was the first to analyze factor intensity reversals in the HOS-model. By relaxing the stringent assumption of “no factor intensity reversals” in a 2x2x2 HOS-model he is able to show that ratio of the return to skilled over unskilled labor can increase in both developed and underdeveloped countries when they start trading with each other.
In extending the HO-theorem, Samuelson (1948, 1949) made two crucial assumptions about the nature of technology:
1. That for each commodity there is a common production function everywhere, and that these production functions are mathematically “homogeneous of the first order,” and 2. That whatever the ratio of wage rate to capital cost (say, between two countries), the
optimal ratio of capital to labor in a given industry i is always greater or always less than in any other industry j. This is the so called strong factor-intensity assumption.
The HO-model also has the first assumption. Samuelson needed to include the second assumption in order to prove the stronger theorem of relative and absolute factor price equalization between countries. The assumption of identical production functions between countries is necessary to explain the trade flows in the HO-model. With consumer preferences assumed identical and similar technologies, all that is left to cause differences in prices, and thereby trade flows, is the difference in factor endowments.
22
intensity assumption, it is sufficient that all industries be subject to either a Cobb-Douglas production function, in which σi=σj=1 or to a production function with fixed coefficients, or perfect complements, so that σi=σj=0.
Minhas uses a constant-elasticity-of-substitution (CES) production function, or what he calls a homohypallagic production function. The CES production function was introduced just 1 year before this article by Arrow et al. (1961) and Manhis was one of the co-authors. The CES production function exhibits a constant elasticity of substitution, σ, for capital and labor. The CES allows for interindustry differences in the relative ease or difficulty with which factor inputs can substitute each other in production, so that σi>σj or σi<σj. .
23
Figure 1.
From Bagisha S. Minhas, 1962.
The figure concerns 2 counties, U (underdeveloped and labor abundant) and D (developed and capital abundant) and 2 industries, 1 and 2. The optimal capital to labor ratios employed in both industries, implied by the CES-production function, is shown by the lines 1 and 2 and is represented on the vertical axis. From this figure, it is clear that the relative factor intensity of industry 1 with respect to industry 2 is completely dependent on the w/r ratio. To the left of w* industry 2 is capital intensive, while to the right of the factor intensity reversal point industry 1 is capital intensive. The endowment ratio Xu and Xd of both countries is represented on the vertical axis. The curve in the bottom half of the figure depicts the non-monotonic relationship between the relative commodity prices (p1/p2), as implied by their relationship with optimal capital-labor ratios, and relative factor prices (w/r).
24
abundant country has a comparative cost advantage in industry 2, which is its relative capital intensive commodity. Differences in the comparative costs of 1 and 2 will give rise to trade. As both countries export their capital intensive product, one of the two countries must be in contradiction with the HO-theorem. So, when factor intensity reversals are allowed, a labor abundant country may export its relatively capital intensive commodity and a capital abundant country may export its labor intensive commodity. This pattern of trade is not compatible with the HO-predictions. The effect of trade is to raise the price of the commodity in which the country holds a comparable cost advantage and thereby it will raise the relative factor price of the factor intensely used in the production of this commodity. As it turns out, both countries use the same factor intensely in the production of their export product and relative factor prices will move in the same direction in both countries. The post-trade equilibrium relative price of commodity 1, P*e will lead to lower wage-rental ratios in both countries, see figure.
Minhas empirically tested the second assumption by Samuelson using input-output tables from the USA and Japan. The industries in the 2 countries are ranked according to their capital intensity. If the assumption is true then both countries must have ranked the industries identically. The Spearman rank correlation between the ordering of industries by capital intensity was only .328 based on ratios of total capital and labor requirements. The Spearman rank correlation based on direct capital and labor requirements was .730. Both these correlations are significantly different than unity.
In addition, by analyzing American industries he found significant differences in the parameter value of the elasticity of substitution between industries at the 95 percent level of confidence.
25
5.3 Specific factor model
The specific factor model is a variant of the Ricardian model. It was first discussed by Jacob Viner and is therefore sometimes referred to as the Ricardo-Viner model. Jones (1971) formalized the model mathematically and he referred to this model as the 2-good-3-factor model. The specific factor model can be used to analyze the effects of trade on labor allocation, output levels and factor returns within an economy in which 1 factor of production is specific to an industry. A specific production factor is a factor of production that is stuck in an industry and therefore immobile between industries. The specific factor cannot switch to another industry in response to changing market conditions. A factor can be immobile between industries for a number of reasons. For example, capital goods may be specifically designed for that industry’s needs or labor may have had specialized training for one sector that is completely unusable in the other sector.
The model uses some straightforward assumptions. It is assumed that the economy produces 2 goods and is endowed with two factors of production, capital and labor, in a perfect competitive market. One factor of production is assumed to be completely immobile between industries. The other factor can move between industries freely and without costs. Usually capital is assumed to be immobile, while labor is considered to be the mobile factor. As both industries use capital in their production, the immobility of capital implies that capital is not homogeneous and cannot be substituted between industries. In this perspective one can understand why Jones referred to the model as a 2-good, 3-factor model. The 3 factors being labor, specific capital in industry A and specific capital in industry B. The endowments of factors is fixed and fully employed in the economy. Firms maximize profits, taking wages and output prices as given. The industries continue to employ labor until the wage is equal to the marginal product of labor. As the capital stock is fixed in both industries, labor is assumed to exhibit diminishing marginal rates of return in both industries. The allocation of labor between industries will be uniquely determined if all firms behave rationally.
26
the import competing sector, this sector will have to contract by reducing labor input. The increase in output in the export sector and the decrease in output in the import competing sector will continue until the wage in the economy reaches a level that equalizes the marginal product in both industries once again.
The return to labor has increased as the wages are now at a higher level. The return to capital will have increased for the export industry specific capital while the owners of the import competing industry specific capital will see their returns decrease as a result of lower revenues and higher wages. The effects of a change in output prices on the real return to the production factors in the specific factor model is described by the magnification effect derived by Jones (1971). The magnification effect for the specific factor model predicts the following effects on the factor returns after trade liberalization by a capital abundant country:
The real return to capital in the export industry will rise with respect to the prices of both imports and exports. The real return to capital in the import-competing industry will fall with respect to both import and export prices. The real wage to workers in both industries will rise with respect to the price of the import goods and will fall with respect to the export good. Whether or not workers are better off from a welfare perspective thus depends on their individual consumption preference with respect to the two goods. Some workers may be better off, while other are worse off.
For reason of completeness and to show the dynamics of distribution I will shortly discuss the distributional outcomes of the immobile factor model.
The immobile factor model is based on the Ricardian model. The differences are the possibility to extend the model to multiple factors and factor mobility between sectors is no longer assumed. In the model by Ricardo, labor, as the sole production factor, is assumed to be freely mobile between sectors. In the immobile factor model moving production factors between industries is assumed to be prohibitively costly. Factors of production are therefore confined to the industry that they are employed in.
27
will now have negative profits because their output price has been lowered. The assumptions of full employment and zero profits imply that factor prices must adjust to lower levels. The effects on factor rewards of this model are straightforward: The return to factors employed in the export sector increases while the return to production factors in the import-competing sector fall.
The HOS-model, the specific factor model and the immobile factor model all predict changes in factor rewards after an economy opens up for international trade. The immobile factor model predicts that factors in the export sector will have higher returns to their input and factors in the import-competing sector will have lower returns. The specific factor model predicts that export-specific factors will have higher returns while import-competing export-specific factors see their returns lowered with an ambiguous effect on the real return to the mobile factor. The HOS-model predicts that real income will be re-distributed from the country’s scarce factor to the country’s abundant factor.
28
5.4 Cones of diversification
Davis (1996) developed a HO-model with three goods, two factors and many countries to try to explain the anomaly of the deteriorating relative wage position of unskilled workers in developing countries following trade liberalization. However, unlike the HO- and SS-theorem, Davis no longer uses the global economy as the reference set to measure relative factor abundance. Instead, factor abundance matters only relatively to a smaller set of countries with similar endowment proportions. The relevant reference set to analyze the distributional effects of trade for any country is composed of those countries within the same “cone of diversification”. Countries within the same cone have similar endowments and produce the same range of goods. He assumes that both goods and factor markets are perfectly competitive. Technologies exhibit constant return to scale and employ fixed coefficients. The factors are capital and labor and are available in fixed supply in each country. The three goods are X, Y and Z, in decreasing order of capital intensity, so that Kx> Ky> Kz. Furthermore, global factor price equalization does not occur. Rather, he assumes that the capital abundant countries (the North) have a higher wage to rental ratio than the labor abundant countries (the South). The difference in global endowment will lead to countries in the North to produce only goods X and Y, while southern countries only produce Y and Z. These are the 2 cones of diversification.
29
and very little of Y. Imports of X and Y are therefore financed by exports of Z. Finally, the region at the intermediate level of capital abundance satisfy their own needs in terms of Y and Z. They therefore export Y and Z in exchange for X.
Considering a reduction in the import tariff of product Z in the southern most capital abundant region, Davis shows that this will cause the zero profit frontier of product Z to shift in. This, in turn, causes lowered wages and an increase in rental rates. This occurs despite the fact that this region is labor abundant in a global sense. Thus, trade liberalization may not lead to an increase in the return to the factor that is abundant in a global sense, as predicted by the SS-theorem. The crucial element is that the region is capital abundant in a local sense. As such, trade liberalization will benefit the factor that is abundant locally. Note that following trade liberalization, the wage to rental ratio also increases in the labor abundant region in the North, so despite being capital abundant globally, the return to Northern labor could increase as a result of trade. The wage to rental ratio will fall in the capital abundant regions in both the North and the South. It will rise in the least capital abundant regions in both the North and the South and will remain virtually unchanged in the intermediately endowed regions. The basic result of the model is that in order to determine whether the return to a factor of production will rise or fall from trade liberalization, its country’s relative factor abundance in the SS-setting is completely irrelevant. All that matters is a country’s relative endowment within its own cone of diversification. Another result of this model is that under free trade, and with identical constant returns to scale technologies there will be factor price equalization among members of the same cone, but not between cones. In his paper he generalizes the model to multiple factors. All the results are robust to the extension of more factors.
5.5 Metzler Paradox
Lloyd Metzler (1949a) demonstrated how the imposition of an import tariff may lead to a reduction in the domestic relative price of the protected commodity. This is called the Metzler Paradox. He analyzed the effect of a tariff on the domestic relative price in a general setting and examined what conditions could cause the paradox. In his own words:
30
Metzler explains, that if this was to happen, the country’s scarce factor of production, which is intensely used in the import-competing sector, will suffer both a relative decline in its share of the national income and an absolute decline in its real return. The implication is that if a tariff is cut, and protection for the import-competing industry cancelled, a nation’s scarce production factor will have a relative increase in factor returns. This outcome is the opposite of what the SS-theorem suggests.
An import tariff has 2 effects, a direct effect and an indirect effect. The direct effect is to raise the domestic relative price of the import good relative to the price of the export good over and above relative world prices. The indirect effect of the tariff is the fact that the increased import price will shift the tariff imposing country’s demand curve for imports downwards, causing a lower world equilibrium price for these imports, improving the terms of trade (price export good/ price import good) in the country as a result. If the latter effect dominates the former, the domestic relative price of the import good will decrease and resources will be shifted to the export sector. Whether a tariff increases or reduces the real and relative returns of the scarce factors therefore depends upon the size of favorable movement in the terms of trade compared with the size of the tariff.
The size of the improvement in the terms of trade depends on the elasticity of the foreign country’s demand for the export good the tariff imposing country.
Metzler assumes that the revenue collected through the tariffs is used to lower income taxes. The increased private income will cause domestic consumers to spend more money on both domestic and foreign goods. As demand for foreign goods rises as a result of this increased income, the nation’s demand curve for foreign goods is shifted up again. The share of money spend on foreign goods in total additional income is called the marginal propensity to import, and is labeled k.
31
Figure 2.
From Lloyd A. Metzler, 1949. Graph has been slightly modified to include line OP.
32
Lloyd generalizes the conditions under which this situation might occur. The only parameters that have any importance in this issue are the elasticity of foreign demand for the domestic export product, labeled
η
, and the domestic propensity to import,k
.
He shows, using a technique first developed by Lerner (1936) that domestic prices will remain unchanged after a tariff only if:η
= 1- k.
So only if the foreign elasticity of demand for that country’s export is equal to 1 minus the domestic propensity to import, the tariff will not affect domestic price ratios and the relative factor rewards will remain unchanged. If
η
is smaller than1 – k
, the domestic price of imports will fall relatively to the domestic price of exports. Ifη
is larger than1 – k
, the tariff will cause the relative domestic price of imports to increase and the consequences for factor rewards will be in line with the SS-theorem. As1 – k
will by definition be smaller than 1, note that it is only possible for η to be smaller than 1 -k
if the elasticity of foreign demand is inelastic,η
< 1. However, if the inelasticity of the foreign demand is lower than 1 – k, Metzler’s analysis shows that the theoretical possibility of counterintuitive factor reward changes is possible.Minabe (1974) discusses the Metzler paradox and notes that the use of the point elasticity at the free trade point as done by Metzler is only appropriate if one is to consider small tariff changes. If a country levies a high tariff on import, then the point elasticity and marginal propensity to import at the original point are no longer relevant. Instead, one must take the average elasticity and the average propensity over the relevant range. This has an important implication because the higher the tariff, the higher the point elasticity of the foreign offer curve, and hence the higher the average elasticity. As the Metzler paradox becomes less likely if the elasticity is higher, the likelihood of the appearance of the Metzler paradox will depend on the size of the tariff.
6. Aggregate inequality
6.1 From functional to personal income and household production and consumption.
33
factors is called the functional distribution of income. Overall income inequality is concerned with the personal distribution of income.
White and Anderson (2001) point to the fact that individuals may own multiple factors of production. Households may have many sources of income. Households can derive income from capital, natural resources, labor, land and other factors of production. Their personal income is the sum of these individual factor returns. Their income thus depends on the return to each factor and how much of that factor they own. The personal income inequality in a country therefore not only depends on the distribution of incomes between factors of production, but also on the inequalities in the ownership of production factors, or assets. If the ownership of some factor v is distributed more equally than factor k, an increase in the share of factor v in national income relative to factor k will reduce income inequality and vice versa (Anderson, 2005). Assuming that trade openness will increase the relative return to the abundant factor, in HOS-models with many countries and many factors, the effect of trade openness on income inequality will depend on a country’s endowment in the various production factors. The consequences on income inequality must come from an assumption about the ownership distribution of production factors. Labour is assumed to be the most equally distributed production factor. Therefore, the higher is the endowment of any factor j to labour, the larger will be the increase in income inequality after the opening of trade.
34
As explained, personal income combines all income sources. Most studies on trade reform focus on wages of the formally employed. However, Rosenzweig (1988) argues that a large proportion of individuals in the poorest developing countries are not employed in the formal labor market. In fact, many individuals are self-employed and/or work in their family business or family farm and spend a significant amount of time to the production of goods and services for their own consumption. In many developing countries household production is a major contributor to the household income. Edmonds and Pavcnik (2006), for example, found for Vietnam in 1993 that only 19 percent of the adult population worked for wages and only 7 percent was employed in the manufacturing sector. In addition, 90 percent of the adult population report to work within their households.
According to Singh et al. (1986), because the majority of the poor in most developing countries is self-employed, the best way of thinking about poor households is in terms of the “farm household”. A farm household is defined as a household that produces goods and services, sells its labor and consumes. Because a farm household is both a consumer and a producer of certain, predominantly agricultural goods, price changes in such a good will impinge on their real income through both the production and consumption channel.
6.2 Cross country studies on aggregate inequality
Contrary to time series studies that have an exclusive focus on relative wages of the skilled and the unskilled, cross country inequality studies focus on aggregate inequality within a country. It therefore combines all the income sources, not just income from skilled and unskilled labor. Whereas wage studies tend to focus on middle income countries, this approach allows inclusion of low income countries into the study. These studies draw their data from household surveys. There reason that these kind of studies were not carried out until the late 1990s is that there was just not enough acceptable inequality data available. This changed in 1996 when Deininger and Squire (1996) published a data set which contained both gini-coefficient for a large number of countries as well as quintile shares in national income.
35
The general regression used in these studies is:
INEQit =
ά
0+
ά
1Y
it+
ά
2Open
it+
ά
3(
Open
it*
Y
it) +
ά
4Z
it+
ε
itZ denotes a set of control variables.
Support for the HOS-theorem would imply a negative sign for
ά
1and a positive sign for
ά
2.
6.2.1 Empirical results from cross country studies
The different studies yield quite different results. The results range from confirming the HOS-prediction to rejecting it completely. I will discuss them next, starting with the studies that found a positive relationship between trade and inequality (so that trade leads to more inequality).
Milanovic (2005) looked at the impact of trade openness on relative income shares across the entire income distribution for a sample that consisted of both developed and developing countries. He examined the income share of the various deciles separately. He finds strong evidence that in countries with low average income levels, the income share of the lower income deciles is smaller in countries that are more open to trade. It’s only when income levels rise between $5000 and $7000 per capita at international prices the situation reverses and the income shares of the lower and middle deciles start to rise compared with the highest deciles in their society.
Similar results are obtained by Barro (2000) for a panel of 84 countries for the period 1960-1990. He also found statistically significant non-linearities in the relationship between openness and equality. As in the case in Milanovic’s research, openness is associated with increased inequality in poor countries and decreases in inequality in rich countries. Barro, using Gini coefficients for the inequality measure, finds a turning point at a level of $13000 per capita income.
Savvides (1998) examined the change in Gini-coefficient in 34 countries between 1974 and 1994. He analysed changes in tariffs, in non tariff barriers and in the Sachs-Warner indicator. He found a positive relationship between openness to trade and income inequality for developing countries and a negative relationship for developed countries.
36
There are some studies however that do not find such a negative relationship between openness and inequality for poor countries. Dollar and Kraay (2000) find that openness is positively associated with per capita income growth. In addition, they cannot reject their null hypothesis that the income of the poor, measured by the lowest quintile’s income share, rise equiproportionately with average income. In other words, all income shares benefit the same relatively from the gains of international trade. Their main conclusion is therefore that economic growth is the most important way in which to alleviate absolute poverty. The data from Dollar and Kraay include 92 countries from 1950 until 1999 and they used several measures of trade openness, none of which were significant.
Li, Squire and Zou (1998) find no statistically significant effect of openness on the Gini coefficient in their panel of 49 countries for 1960-1990.
Edwards (1997) uses time series observations for 43 countries for the 1970s and 1980s and decade averages. He did not find any significant relationship between changes in trade openness, measured by Sachs-Warner and tariff rates, on changes in inequality.
Finally, Calderon and Chong (2001) did find support for the HOS-theorem. For their panel of 102 countries for 1960 through 1995 they found that trade openness did not affect inequality in developed countries but it did cause inequality to decline in developing countries.
7. Inequality and endownments
While the aggregate inequality studies of the previous chapter included all sources of income, the theoretical justification of these studies is the HOS-prediction that trade openness improves the relative welfare position of the unskilled in developing countries as articulated by the quote of the World Bank, see page 12. As it does not take factor endowments into account, the implicit
assumption is thus that developing countries are abundant in unskilled labour and unskilled workers are overrepresented among the poor. The opening of trade will then lead to a price increase of unskilled labour intensive goods, an expansion of the unskilled labour intensive industries and to a corresponding rise in unskilled labour demand and in the return to unskilled labour, which in turn lowers income inequality.
37
may be one of a country’s most prominent endowments, this does not mean it is the only
abundant endowment. Many countries, for example Latin American countries, are relatively well endowed with natural resources. The HOS-model predicts that the opening of trade will lead to a rise in the factor return of natural resources. The effects on income inequality then depend on the distributional equality of this factor, which in the case of natural resources is likely to be positive so that inequality might rise. In addition, there may be complementarities between factors. If the exploitation of natural resources is skilled labour intensive, the wage gap may increase as well, possibly leading to further income inequality.
In fact, a country’s endowments can influence the income distribution regardless of trade. This can be illustrated by a general observation by Splimbergo et al. (1999):
”Some factors such as land or capital can be concentrated in the hands of few people because there is no upward limit to their accumulation; other factors of production such as skills cannot be concentrated to the same extent because there is a natural upward limit to the amount of
education that an individual can accumulate. (…) Consequently, if an economy is mostly
endowed with land and capital, there is no limit to the concentration of wealth. If an economy is endowed mostly with education, the distribution of income is expected to be more egalitarian, keeping constant the other factors” (p.81).
The country’s factor endowments therefore matter a great deal if one examines the effects of trade on income distribution. If one considers multiple factors then a relative rise in the return to a factor which is relatively more equally distributed than the other factors, will lead to a fall in income inequality. Conversely, if trade causes the relative return to a factor without a natural limit to their accumulation to rise, income inequality is expected to deteriorate.
Leamer (1987) presented a three factor, n-good Heckscher-Ohlin model that includes natural resources as a factor of production. In this model greater openness to trade will increase income inequality in natural resource abundant countries. The return to this factor will increase as a result of trade liberalization and as natural resources are typically less equally distributed in terms of asset inequality of ownership than other factors, overall inequality will increase.
38
Capital is assumed to be less equally distributed than labor; therefore a fall in capital’s share in GDP will improve a nation’s income distribution.
7.1 Empirical studies with endowments
Studies that take endowments into account test the hypothesis that the effect of trade openness on income distribution depends on the country’s relative endowments.
The regressions used in testing the hypothesis:
INEQit = β0 + β1OPENit + β2jOPENit*RFEijt + β3Zit + εit,
Where Z is a set of control variables and RFE is a set of relative factor endowments. β2j therefore measures the change in income inequality caused by a change in the country endowment of factor j.
39
Splimbergo et al. (1999) studied the empirical links between factor endowments, trade and personal income distribution. The authors based their model on the one developed by Bourguignon and Morrison (1990). They used the Gini’s from the Deiniger-Squire database as their measure of income inequality of the dependent variable in equation (3). They had no systematic data on the structure of ownership of the various factors and therefore it is left out of the analysis. They computed the effective relative endowments by comparing a country’s endowment to the world average endowments weighted by the degree of openness of other countries, reflecting that relatively closed countries do not compete completely on the global market. In addition they used an index of openness that is controls for factor endowments, size and distance to trading partners. They had observations on a total of 34 countries, both developed and developing, between 1965 and 1992.
They found that land and capital abundant countries have a less equal income distribution while skill abundant countries have a more equal income distribution. They argue that the reason for this is that capital and land can be accumulated by a relatively small number of people within a society while skill cannot. Therefore, countries that are relatively well endowed with skill tend to be more egalitarian in terms of their income distribution. All the endowment parameters were significant.
Trade openness is associated with higher inequality, keeping constant the factor endowments. Interestingly, they found that once nations start to trade, the effects are reversed. Openness seems to undo the effect of factor endowments on equality because the coefficients of the endowment and of the interaction term of the endowment with openness have opposite signs. Splimbergo et al. find this a direct contradiction with the HO-framework. They argue that the HO-model suggests that the income distribution in a land abundant country, which is already bad in a closed country, would worsen after opening up because land is a scarce global factor and its return will increase as a result. Opposite signs are found for skill and capital as well. Even though skill abundant countries are still more equal, they become less equal as a result of trade while capital and land abundant countries become more equal. Finally, it must be noted that the interaction term between openness and land was not significant while the other two interaction terms were significant.
