World Development Vol. 34, No. 12, pp. 19972015, 2006 2006 Elsevier Ltd.

. All rights reserved 0305-750X/$ - see front matter www.elsevier.com/locate/worlddev

doi:10.1016/j.worlddev.2006.02.010

Economic Liberalization and Wage Inequality in India

RUBIANA CHAMARBAGWALA * Indiana University, Bloomington, IN, USA
Summary. We investigate Indias widening skill wage gap and narrowing gender wage dierential during the two decades that coincide with the economic liberalization in the country. Using the nonparametric methodology developed by Katz and Murphy [Katz, L. F., & Murphy, K. M. (1992). Changes in relative wages, 196387: supply and demand factors. Quarterly Journal of Economics, 107(1), 3578], we nd that relative demand shifts contributed to relative wage shifts and that increases in the demand for skilled labor were mostly due to skill upgrading within industries. In assessing the contribution of external sector reforms to demand for skilled labor, we nd that international trade-in manufactures beneted skilled men but hurt skilled women, whereas outsourcing of services generated a demand for both female and male college graduates. 2006 Elsevier Ltd. All rights reserved. JEL classication F16, J24, J31 Key words trade and labor market interactions, education, wages, Asia, India

1. INTRODUCTION During the 1980s and 1990s, India experienced two dramatic changes in its wage structure. First, there was a considerable widening of the skill wage gap, accompanied by large increases in both the supply of and demand for high school and college graduates. Second, as Figure 1 shows, the gender wage dierential narrowed considerably among high school and college graduates. In examining alternative explanations for these changes, we nd that an increased demand for skilled workers and especially for skilled women contributed signicantly to these wage shifts. The skillupgrading within industries was primarily responsible for these demand shifts. Since these two decades mark a period of widespread economic liberalization in India, we measure the contribution of external sector reforms to an overall demand and nd that international trade-in manufactures beneted skilled men but hurt skilled women, whereas trade-in services (outsourcing) beneted both male and female college graduates. There is a vast literature that documents wage shifts and its determinants in the United States during the past few decades. 1 The liberalization of trade and investment in several developing economies and increasing globaliza-

tion during this period suggests that a skillabundant rich country, such as the United States, would experience a widening skill wage gap. The literature conrms this and nds that the relative demand shifted toward skilled or educated workers and away from unskilled or uneducated workers. The changes in technology and product demand that shifted employment away from manufacturing and toward sectors that are education and female intensive are oered as explanations for wage shifts in the United States. In this paper, we focus on how economic liberalization and globalization aects the wage structure in a poor but rapidly developing economy that is abundant in unskilled labor. India serves as a particularly interesting case study since beginning in the mid-1980s, the Indian government implemented a range of far-reaching economic policy reforms in the domestic and external sectors. 2 These reforms marked a clear break from the countrys socialist strategy of state-directed, heavy-industry based, and import substitution industrialization, which beginning in the early 1950s, was implemented through a series of ve-year plans. During the rst three decades after Indias
* Final revision accepted: February 27, 2006.

1997

1998
Primary School

WORLD DEVELOPMENT
Middle School

1.20 1.00
Females

1.40 Relative Wage 1.20 1.00 0.80 0.60

1983 1988 Year 1993 1999
Females Males

Relative Wage

0.80 0.60 0.40

Males

1983

1988 Year
College

1993

1999

High School

2.20 Relative Wage

3.20 3.00 Relative Wage 2.80 2.60 2.40 2.20 2.00

1983 1988 Year 1993 1999
Females Males

2.00 1.80 1.60 1.40 1.20

Females Males

1983

1988 Year

1993

1999

Figure 1. Relative wages for males and females: 19832000.

independence in 1947, high levels of protection were provided to import-competing industries, which were mostly capital-, technology-, and skilled-labor intensive. This was complemented by a system of industrial regulation, complex licensing, and nancial repression. Domestic sector reforms consisted of a signicant relaxation of industrial controls. Industrial licensing was abolished except in the case of industries of strategic or environmental importance. Antitrust legislation was amended to facilitate expansion and diversication of capacity by large rms. Via privatization, the number of industries reserved for the public sector was drastically reduced. The taxation system was rationalized, while the tax rates were slashed to ASEAN levels. Finally, major reforms were introduced in the banking sector to put the banking system on a sounder footing. External sector reforms consisted of trade and investment liberalization. The trade policy reforms aimed at liberalizing and promoting both exports and imports. As a result of lower taris, the elimination of quotas and import license requirements, and liberalization of technology imports, total exports and imports increased dramatically during the 1990s. Exports were liberalized via the abolition of export subsidies and controls, while the liberalization of imports was implemented via a rapid reduction in the tari rates and the abolition of licensing and quantitative restrictions on most

imports except consumer goods. The average ad valorem tari rate fell from 125% in 1990 to 40% in 1999. Besides lower taris, nontari barriers were reduced by eliminating quantitative restrictionsquotas and import licensing requirementsparticularly on capital and intermediate goods. In addition, technology imports were liberalized by eliminating technology license requirements. Foreign direct investment (FDI) was liberalized to a limited extent, resulting in an increase of FDI from $233 million to $3.3 billion during the 1990s. Several hypotheses have been put forth and tested to explain the rising skill premium and demand for skilled workers in less developed countries (LDCs). 3 Hypotheses II, III, IV, and V together constitute what is commonly referred to as the Skill Enhancing Trade (SET) Hypothesis. 4 The rst hypothesis (Hypothesis I) is based on the StolperSamuelson (SS) theorem in the HeckscherOhlinSamuelson model of trade. The SS theorem predicts that trade liberalization will raise the demand for and returns to the abundant factor of productionthat is, unskilled labor in most LDCs. Even though at rst glance this theorem predicts a decrease in the wage gap between skilled and unskilled labor in LDCs, on closer inspection it shows that as protective import taris, quotas, and licenses are removed, the price of formerly protected goods will fall. By the SS Theorem, a decrease

ECONOMIC LIBERALIZATION AND WAGE INEQUALITY IN INDIA

1999

in the relative price of a good will decrease the relative price of the factor used intensively in the production of that good and increase the relative price of the other factor. Since in many LDCsnamely, Colombia, Mexico, Brazil, and Moroccothe most protected sectors were those that were intensive in unskilled labor, the SS theorem predicts that trade liberalization in these countries should lower unskilled wages. India, on the other hand, had its highest protection levels in human- and physical-capitalintensive sectors. Therefore, the rising demand for and returns to skilled labor in India contradict the predictions of the SS theorem for the countrywhich are an increase in the demand for and returns to unskilled labor and an expansion of unskilled-labor intensive sectors. The second hypothesis (Hypothesis II) relates to economic reforms in general and not specifically to trade liberalization in driving increased demand for and returns to skilled labor. According to this hypothesis, LDCs may experience higher returns to skilled-labor intensive occupationssuch as professional, managerial, and administrative jobsas a result of reforms that generate demand for individuals who can implement these reforms. In India, reforms in both the domestic and external sectors may have created more white-collar jobs. The empirical evidence is mixed: Cragg and Epelbaum (1996) nd support for this hypothesis for pre-NAFTA Mexico while Attanasio, Goldberg, and Pavcnik (2004) nd no changes in the occupational returns during 198698 in Colombia. Global production sharing or outsourcing is the third hypothesis (Hypothesis III) that has been provided to explain the rising skill premium and demand for skilled labor in LDCs. Feenstra and Hanson (1996, 2003) argue that trade and investment liberalization on the part of LDCs allows developed countries (DCs) to transfer the production of intermediate goods and services to LDCs. For LDCs these activities are skill intensive, which result in a greater demand for and returns to skilled labor. Therefore, the external sector reforms that promote trade-in manufactures and services and those that attract foreign direct investment can benet skilled workers in LDCs. Feenstra and Hanson (1997) nd empirical support for this hypothesis for the case of Mexico. The fourth hypothesis (Hypothesis IV) relates skill-biased technical change (SBTC) to an increased demand for skilled labor and a rising skill premium. Wood (1995) argues that

trade liberalization results in defensive innovationthat is, greater competition from foreign rms may induce domestic rms in LDCs to either engage in R&D or to adopt new and advanced technologies in order to secure their market share in the domestic and international markets. Because of technology-skill complementarities, adoption of modern technologies raises the demand for and returns to skilled labor. Acemoglu (2003) describes how after trade liberalization in LDCs, increased capital goods imports can lead to a greater demand for skilled workers as a result of capital-skill complementarities, thus increasing the skill premium. An empirical support for this hypothesis is found by Attanasio et al. (2004) for Colombia and by Harrison and Hanson (1999) for the case of Mexico. Gorg and Strobl (2002) nd an increase in the relative wages of skilled labor in Ghana, brought about by skill-biased technological change induced through imports of technology-intensive capital goods or export activity. However, Pavcnik (2003) rejects the SBTC hypothesis for Chilean plants. Finally, the quality-upgrading of rms (Hypothesis V) is also considered a mechanism whereby the demand for and returns to skilled labor can increase in LDCs. The basic argument is that trade liberalization may induce a quality upgrading of rms, where quality can mean either rm productivity or product quality. Verhoogen (2004) nds a strong support for this hypothesis for the case of Mexico where greater exports as a result of the peso crisis resulted in better quality products being produced by exporters. Because higher quality products require a higher proportion of skilled workers, the relative demand for and returns to skilled labor increased. Our analysis documents large overall demand shifts away from unskilled workers and toward skilled workers, which played an important role in widening the skill wage gap during 198384 and 19992000 in India. These demand shifts were primarily within sectorsthat is, skillupgrading within industries explains most of the rising demand for skilled workers during this period in India. This result provides a strong evidence in the support of Hypothesis IIthat the creation of white-collar jobs as a result of domestic and/or external sector reforms played a major role in generating demand for skilled labor and increasing the skill premium. The agricultural sector experienced a dramatic contraction, whereas manufacturing

2000

WORLD DEVELOPMENT

employment fell slightly during the 1980s and 1990s in India. On the other hand, the service sector expanded considerably and was primarily responsible for generating a demand for high school and college graduates. This nding, along with the rising demand for and returns to skilled labor, provide evidence to reject Hypothesis I. That a country abundant in unskilled labor would experience these changes as its economy becomes increasingly liberalized, suggests that the SS theorem does not hold for the case of India. 5 In evaluating the role of external sector reforms, we measure trade-induced demand shifts in contributing to overall demand shifts in the economy. We nd that trade-in manufactures a generated demand for skilled men and a decreased demand for skilled women whereas trade-in services generated a demand for both male and female college graduates but a decreased demand for men and women, without a college degree. Because our trade-induced demand shifts arise from changes in the implicit labor supply embodied in trade ows, we only measure how overall changes in the skill content of exports and imports result in changes in the demand for skilled and unskilled labor and cannot evaluate Hypotheses III, IV, and V separately. We therefore cannot attribute our trade-induced demand shifts to either the transfer of manufacturing and service jobs from DCs to India, defensive innovation, or quality upgrading by rms. Being unable to distinguish between Hypotheses III, IV, and V is a limitation which we cannot address in this paper due to the lack of appropriate data. Nevertheless, the overall eects of Indias trade liberalization in manufacturing and services on demand for skilled and unskilled labor provide a valuable evidence by itself. That external sector reforms have generated opportunities for skilled workers indicates that Indias economic liberalization has and can continue to create incentives for accumulating human capital, which is ultimately critical to the countrys economic development. The rest of this paper is organized as follows. Section 2 describes the data used in our analysis. Section 3 documents the changes in relative wages in India during 198384 and 19992000. In Sections 4 and 5 we investigate the extent to which relative supply and demand changes contributed to shifts in relative wages in India. Section 6 measures the contribution of international trade-in manufactures and services toward relative demand changes during

the 1980s and 1990s in India. Section 7 concludes. 2. DATA AND METHODOLOGY The individual level data used in this study comes from the Employment and Unemployment Schedule of the National Sample Survey Organization (NSSO), administered nationally by the Government of India. The Employment and Unemployment Schedules are administered approximately every ve years in four subrounds, each with a duration of three months. We use four rounds of the NSSO surveys that cover the two decades that span Indias economic reformsthat is, 198384, 198788, 199394, and 19992000. The NSSO survey includes household and individual level datahousehold size and composition, social group, religion, income, assets, indebtedness, demographic variables (age, gender, marital status), education participation and attainment, and a detailed employment section on principal and subsidiary activities (industry, occupation, type and amount of income earned, and intensity of each activity). For our empirical analysis, we use data for individuals aged 15 years and above and create two samplesa wage sample to measure hourly wages of workers by demographic group and a count sample to measure the amount of labor supplied by these demographic groups. We divide our data into 100 distinct labor groups, dened by two gender groups (male and female), ve education groups (less than primary, primary, middle, high school, and college), and 10 age groups (1520, 2025, 2530, 3035, 3540, 4045, 4550, 5055, 5560, and 60+ years). The wage measure we use throughout is the average hourly wage of workers within a gendereducationage cell. An individuals average hourly wage is computed as the total wages during the past week divided by the total hours worked during that week. We then adjust the individual wages to 1988 Rupees using statelevel CPI for agricultural and industrial workers for rural and urban wages, respectively. This adjustment makes wages comparable over time, across states, and across rural and urban sectors. Our wage sample includes regular wage and salary workers since wages are only reported for this group. Self-employed workers (both wage and nonwage earning) are excluded from our wage sample. The count sample in-

ECONOMIC LIBERALIZATION AND WAGE INEQUALITY IN INDIA

2001

cludes all individuals who worked either as regular wage and salary workers or as selfemployed workers (wage and nonwage earning). 6 The amount of labor supplied by each demographic group is measured as the total hours worked by each group as a proportion of the total hours worked by all groups in that year. The total hours worked by each group is computed as the sum of hours worked during the past week for all individuals within each gendereducationage cell. Rather than generate 100 labor groups for the rural and urban sectors separately, our analysis aggregates both sectors. Even though ruralurban dierences are potentially interesting, the reason for this aggregation is the scarcity of data (especially for wages during 198788) which results in the missing values for some cells. Since the NSSO data reports wages only for individuals who earn a regular wage or salary, wages are missing for a considerable number of individuals in the rural sector, making it necessary to aggregate both sectors for our analysis. We calculate relative wages, a (100 4) matrix Wr, and relative supply, a (100 4) matrix Xr, from our wage and count samples. Our wage data consists of a (100 4) matrix W, which consists of the average hourly wage (adjusted to 1988 Rupees) from the wage sample for each of the 100 demographic groups in each of four years. Our labor supply data consists of a (100 4) matrix X, which consists of the proportion of hours worked from the count sample for each of the 100 demographic groups in each of four years. From X we construct a 100element vector, N, of the average employment shares of each group over the four years. We use this vector of xed weights to construct wage indices for each year as N 0 W, a (1 4) matrix. Deating wages in each year (W) by the value of the wage index for that year (N 0 W) generates the relative wages for each demographic group in each year, denoted by a (100 4) matrix Wr. 7 From Wr we calculate a 100-element vector, X, of average relative wages of each group over the four years. The average of the relative wages of each demographic group over the four years provides a natural basis for aggregating quantities of labor supplied across groups in terms of eciency units. We weigh the employment share of each group (X) by the average relative wage of that group (X) and sum over all groups to construct a measure of the total labor supply in the economy in each year in e-

ciency units, X 0 X, a (1 4) matrix. We then deate the actual labor supply (X) by the total labor supply in the economy measured in eciency units (X 0 X) for each demographic group in each year to get a (100 4) relative supply matrix Xr. 8 3. SHIFTS IN RELATIVE WAGES Table 1 reports changes in average relative hourly wages for all workers, women, and men by education levels over four periods in Indiathat is, three sub-periods 198388, 198794, and 19932000 and the overall period 19832000. The average hourly wages are relative wagesthat is, each groups wage relative to the wages for a xed bundle of workers described in Section 2 and calculated from the (100 4) matrix Wr. The gures in Table 1 represent changes in the log relative wage, multiplied by 100, for each group over the relevant time period. For the period during 19832000, the relative wages of all workers (both men and women together) with less than primary, primary, high school, and college education increased while those of workers with middle schooling decreased. The decline in the relative wages of workers with middle schooling was large with even larger increases in the relative wages of high school and college graduates. We observe this same pattern in the relative wages of women during 19832000. The relative wages of all women except those with middle schooling increased, with a substantial decrease in the relative wages of women with middle schooling and large increases in the relative wages of high school and college female graduates. For men, we observe decreases in the relative wages of workers with less than primary, primary, and middle schooling and increases in the relative wages of high school and college male graduates. The rise in the relative wages of less-skilled women can be explained by the SS theorem. On the other hand, the fall in the relative wages of less-skilled men together with the rise in the relative wages of skilled women and men suggest that the SET Hypothesis might have played a dominant role during the period of Indias reforms. Can we attribute the increase in the relative earnings of skilled workers to rms higher demand for skilled labor? Moreover, if relative demand shifts can explain the widening skill wage gap and the narrowing gender wage gap for high school and college

2002

WORLD DEVELOPMENT Table 1. Shifts in relative wages in India: 19832000

Group 198388 All adults <Primary school Primary school Middle school High school College Women <Primary school Primary school Middle school High school College Men <Primary school Primary school Middle school High school College 23.34 (0.0309) 8.35 (0.0418) 24.16 (0.0617) 16.62 (0.0707) 10.77 (0.1795) 39.08 (0.0253) 9.82 (0.0215) 37.58 (0.0656) 19.31 (0.0586) 19.95 (0.1446) 13.71 (0.0214) 7.57 (0.0290) 13.89 (0.0534) 14.24 (0.0821) 3.04 (0.2111)

Change in log relative wages (multiplied by 100) 198794 19.91 (0.0304) 8.16 (0.0418) 16.85 (0.0668) 18.10 (0.0920) 3.13 (0.1886) 30.99 (0.0252) 11.70 (0.0197) 22.30 (0.0678) 19.81 (0.0958) 8.53 (0.1635) 12.88 (0.0221) 6.25 (0.0323) 12.96 (0.0594) 16.60 (0.0909) 1.58 (0.2159) 19932000 1.82 (0.0240) 1.28 (0.0430) 1.76 (0.0701) 10.75 (0.1115) 11.15 (0.1591) 1.26 (0.0045) 7.48 (0.0204) 2.14 (0.0725) 19.26 (0.1193) 20.15 (0.1524) 2.14 (0.0157) 2.32 (0.0401) 1.47 (0.0628) 2.47 (0.1083) 3.03 (0.1697) 19832000 1.61 (0.0246) 1.09 (0.0434) 9.07 (0.0653) 12.23 (0.0946) 18.79 (0.1483) 6.83 (0.0049) 9.36 (0.0222) 17.42 (0.0704) 19.76 (0.0921) 31.56 (0.1319) 1.32 (0.0148) 3.64 (0.0375) 2.41 (0.0572) 4.83 (0.1011) 7.65 (0.1636)
s2 t nt s2
t

The reported numbers are of the form D(log Wr) 100, where Wr represents relative wages. Figures in parentheses r are standard errors of the dierence in relative wages between years t and t 0 . These are calculated as SEtt0 nt00 , where st is the standard deviation of relative wages in year t and nt is the number of observations in year t, and t 0 > t.

graduates, how has Indias trade liberalization contributed toward these demand shifts? We now turn to answering these questions. 4. ALTERNATIVE EXPLANATIONS FOR RELATIVE WAGE SHIFTS The previous section documented large shifts in the relative wages among women and men during the 1980s and 1990s in India. While Indias trade liberalization could be responsible for these changes by altering the relative de-

mand for workers with dierent education or skill levels, other factorsnamely relative supply shifts and changes in wage legislation could have brought about these changes as well. We investigate relative supply and demand changes as potential determinants of the relative wage shifts in India. (a) Changes in relative labor supply Table 2 summarizes changes in the relative labor supply over the 19832000 period, where each groups supply is measured in eciency

ECONOMIC LIBERALIZATION AND WAGE INEQUALITY IN INDIA Table 2. Relative supply changes: 19832000 Group 198388 All adults <Primary school Primary school Middle school High school College Women <Primary school Primary school Middle school High school College Men <Primary school Primary school Middle school High school College 21.70 (0.0017) 3.83 (0.0010) 1.30 (0.0010) 18.07 (0.0009) 33.70 (0.0004) 11.00 (0.0009) 4.58 (0.0002) 27.42 (0.0001) 26.62 (0.0001) 52.87 (0.0001) 27.30 (0.0009) 5.02 (0.0008) 0.88 (0.0010) 17.24 (0.0009) 30.85 (0.0005) Change in log relative supply (multiplied by 100) 198794 13.73 (0.0013) 21.22 (0.0008) 2.91 (0.0010) 14.08 (0.0010) 14.25 (0.0005) 15.88 (0.0007) 6.43 (0.0002) 31.16 (0.0001) 30.58 (0.0001) 17.61 (0.0001) 12.53 (0.0008) 26.05 (0.0006) 0.27 (0.0009) 12.25 (0.0009) 13.68 (0.0006) 19932000 16.60 (0.0012) 13.01 (0.0007) 5.57 (0.0009) 10.65 (0.0011) 8.75 (0.0006) 14.67 (0.0006) 4.89 (0.0002) 17.44 (0.0002) 14.02 (0.0002) 9.64 (0.0001) 17.68 (0.0008) 14.75 (0.0005) 3.90 (0.0008) 10.23 (0.0010) 8.59 (0.0006)

2003

19832000 52.03 (0.0016) 38.06 (0.0009) 9.77 (0.0010) 42.79 (0.0010) 56.69 (0.0005) 41.56 (0.0008) 6.12 (0.0002) 76.03 (0.0001) 71.22 (0.0001) 80.12 (0.0001) 57.51 (0.0009) 45.82 (0.0007) 2.75 (0.0009) 39.72 (0.0010) 53.12 (0.0005)
s2 t nt s2
t

The reported numbers are of the form D(log Xr) 100, where Xr represents relative supply. Figures in parentheses are r standard errors of the dierence in relative supply between years t and t 0 . These are calculated as SEtt0 nt00 , where st is the standard deviation of relative supply in year t and nt is the number of observations in year t, and t 0 > t.

units and includes all workers in the count sample. Each groups supply is then measured relative to the total supply in eciency units in a given year. The gures in Table 2 represent changes in the log relative supply, multiplied by 100, for each group over the relevant time period. For the 19832000 period, the relative supplies of all workers (both men and women together) with less than primary and primary education declined considerably and those of workers with middle, high school, and college education rose substantially. We observe an

identical pattern for menthat is, large decreases in the relative supplies of men with less than primary and primary education, a small increase in the relative supply of men with middle schooling, and substantial increases in the relative supplies of high school and college male graduates. While the relative supply of women with less than primary education declined considerably, the relative supplies of women with primary schooling rose slightly, while those of middle, high school, and college educated women rose substantially.

2004

WORLD DEVELOPMENT

Table 2 documents a trend toward rising education levels during the 1980s and 1990s in India, which could be the result of educational policies such as higher public expenditure on education, more schools, better accessibility to schools, higher quality of education, and other incentives such as provision of meals in schools. (b) Changes in relative labor demand Changes in the demand for labor with dierent education levels can result from changes in the sectoral composition of output, which can be attributed primarily to changes in the product demand. In Table 3 we dene 18 industries and three occupation groups, while Table 4 reports average industry and occupation distributions for ve education groups each for men and women. The gures for each gendereducation group represent the share of employment (measured in hours worked during the preceding week) of that gendereducation group in the corresponding industry or occupation averaged over the four survey years. The large dierences in employment shares by gender

education groups suggest that shifts in labor demand across industries and occupations should have a signicant eect on the relative wages of these groups. While industry groups cannot be classied as exclusively low-, medium-, or high-skill because of varying education levels of its workers, it is possible to divide occupations into three skill groups. High-skill occupations include professional, technical, administrative, executive, and managerial workers; medium-skill occupations consist of clerical and sales workers; and low-skill occupations include production and service workers. Table 5 shows employment shifts into high- and medium-skill occupations during 19832000that is, the share of workers employed in professional, technical, administrative, executive, and managerial occupations increased by 5.28% points, while that of clerical and sales workers rose by 2.06% points. During the same period there was an employment shift out of low-skill occupationsthat is, a decrease of 7.34% points in the share of workers employed in production and service occupations. Table 5 demonstrates that skilled

Table 3. Industry and occupation groups Industry/occupation Industry 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 Occupation 1 2 3 Agriculture, hunting, forestry, and shing Mining and quarrying Manufacture of food, beverage, and tobacco products Manufacture of textiles, leather, fur, wearing apparel, and footwear Manufacture of wood and wood products Manufacture of paper, paper products, printing, and publishing Manufacture of chemicals, rubber, plastic, petroleum, and coal products Manufacture of nonmetallic mineral products Manufacture of basic metals, metal products, and metal parts Manufacture of machinery and transport equipment and parts Other manufacturing industries Electricity, gas, steam, water works, and water supply Construction Wholesale trade, retail trade, restaurants, and hotels Transport, communications Storage, warehousing, repair services Financing, insurance, real estate, business services Community, social, personal services, except repair services Professional, technical, administrative, executive, and managerial workers Clerical and sales workers Production and service workers

Industry 4 includes cotton, wool, silk, manmade and synthetic ber, and jute and other vegetable ber textiles. Occupation category 3that is, production and service workersincludes farmers, shermen, hunters, loggers, and related workers.

ECONOMIC LIBERALIZATION AND WAGE INEQUALITY IN INDIA Table 4. Average industry and occupation distributions for men and women: 19832000 Industry/ occupation 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 1 2 3 Women education level <Primary 27.31 15.69 28.32 15.82 14.11 6.00 9.04 21.05 2.64 1.03 9.41 0.72 12.94 9.26 1.23 1.05 10.98 11.59 2.88 6.51 22.32 Primary 1.81 0.82 4.88 3.74 1.60 2.24 3.63 0.71 0.38 0.44 1.87 0.11 0.57 1.22 0.20 0.14 1.43 0.00 0.70 0.81 1.81 Middle 0.59 0.08 1.22 2.26 0.38 2.17 1.41 0.34 0.33 0.29 0.69 0.07 0.22 0.68 0.15 0.27 1.52 0.00 1.84 0.62 0.65 High 0.15 0.45 0.46 1.00 0.00 2.79 1.43 0.28 0.33 1.12 0.50 0.98 0.15 0.32 0.69 0.53 5.01 0.00 6.95 1.44 0.22 College 0.02 0.16 0.04 0.13 0.08 0.12 0.44 0.08 0.27 0.59 0.00 1.32 0.08 0.08 0.47 1.65 3.37 1.49 5.25 0.92 0.03 <Primary 49.09 52.76 32.63 35.87 46.88 15.34 17.13 51.48 37.04 22.03 29.38 21.12 52.56 30.34 39.90 20.87 18.45 29.57 8.87 22.01 47.61 Men education level Primary 10.55 9.50 13.11 18.62 19.31 17.72 14.60 11.27 16.32 15.56 23.42 16.20 14.64 18.90 17.58 14.52 9.56 9.81 7.03 14.87 12.81 Middle 6.98 9.53 9.36 13.07 11.90 22.04 16.82 7.72 19.82 16.81 20.08 21.96 9.65 19.02 17.50 17.73 12.74 15.45 11.29 17.92 9.25 High 3.05 8.35 8.00 7.78 5.10 22.65 20.26 5.26 16.51 30.70 11.01 26.04 6.76 15.79 17.41 20.84 22.85 16.64 28.73 24.35 4.64

2005

College 0.44 2.67 1.97 1.71 0.64 8.92 15.24 1.81 6.36 11.42 3.64 11.50 2.41 4.40 4.85 22.40 14.10 15.45 26.47 10.56 0.67

Table 5. Changes in industry and occupation distributions: 19832000 Industry/occupation Change in employment share (in percentage points) 198388 198794 19932000 19832000 Agriculture, hunting, forestry, and shing Mining and quarrying Manufacture of food, beverage, and tobacco products Manufacture of textiles, leather, fur, wearing apparel, and footwear Manufacture of wood and wood products Manufacture of paper, paper products, printing, and publishing Manufacture of chemicals, rubber, plastic, petroleum, and coal products Manufacture of nonmetallic mineral products Manufacture of basic metals, metal products, and metal parts Manufacture of machinery and transport equipment and parts Other manufacturing industries Electricity, gas, steam, water works, and water supply Construction Wholesale trade, retail trade, restaurants, and hotels Transport, communications Storage, warehousing, repair services Financing, insurance, real estate, business services Community, social, personal services, except repair services Professional, technical, administrative, executive, and managerial workers Clerical and sales workers Production and service workers 2.31 0.21 0.16 0.24 0.05 0.05 0.09 0.10 0.12 0.14 0.09 0.01 0.32 2.36 0.18 0.57 0.41 0.01 1.49 2.22 3.72 1.95 0.30 0.09 1.58 0.21 0.06 0.10 0.10 0.13 0.13 0.06 0.08 0.67 0.78 0.36 0.31 2.95 0.00 1.33 0.31 1.02 4.15 0.17 0.17 0.15 0.06 0.07 0.08 0.13 0.12 0.01 0.25 0.06 0.94 2.06 0.67 0.78 1.13 0.16 2.46 0.14 2.60 8.41 0.07 0.10 1.18 0.23 0.06 0.27 0.07 0.12 0.00 0.40 0.01 1.29 3.63 0.85 1.66 1.41 0.16 5.28 2.06 7.34

2006

WORLD DEVELOPMENT

jobs (those employing high school and college graduates) in professional, technical, administrative, executive, and managerial occupations (occupation category 1) are dominated by men (28.73% and 26.47%, respectively) rather than women (6.95% and 5.25%, respectively). This suggests that employment shifts into high-skill occupations may have disproportionately beneted high-skilled men rather than high-skilled women. The actual employment levels and changes are determined by the interaction of both the demand for and supply of dierent labor groups. However, if changes in Indias occupation distribution can be considered a rough measure of shifts in rms demand for workers with dierent education levels, these occupational shifts show a simultaneous increase in the demand for high-skilled workers and decrease in the demand for medium- and lowskilled workers, suggesting that the SET Hypothesis might have played a signicant role during Indias reforms. 5. NONPARAMETRIC ANALYSIS The nonparametric methodology proposed by Katz and Murphy (1992) provides a simple framework for decomposing the extent to which relative supply and demand changes contributed to the relative wage changes in India. We rst test whether relative labor supply changes alone can explain changes in the relative wages by education levels, or, instead, relative labor demand changes must have been non-neutral or factor biased. We then evaluate between- and within-sector changes in relative labor demand. (a) Relative supply versus relative demand changes If we assume a stable relative demand for our labor groups, the relative supply shifts presented in Table 2 cannot explain the relative wage shifts observed during the 19832000 period for most groups. With a stable relative demand, we should observe a negative correlation between the changes in relative supplies and wages. We observe this negative correlation for only two groups during the 1983 2000 periodthat is, all workers and men with primary education. For all other groups this correlation is positive, implying that the relative demand for these groups must have chan-

ged during this period. Using our measures for relative wages (Wr) and relative supply (Xr), we rst test whether relative labor supply changes alone can explain the changes in relative wages by education levels, or, instead, relative labor demand changes must have been non-neutral or factor biased. In the Katz and Murphy (1992) framework, the aggregate production function consists of K types of labor inputs. 9 The vector of associated labor demands can be written as X t DW t ; Z t ; 1:1 where Xt is a (K 1) vector of labor inputs employed in the market in year t, Wt is a (K 1) vector of market prices for these inputs in year t, and Zt is a (K 1) vector of demand shift variables in year t. The demand shift variables in Zt embody the eects of technology, other nonlabor inputs such as capital, and product demand on the demand for labor inputs. Eqn. (1.1) can be written in terms of dierentials as dX t Dw dW t Dz dZ t : 1:2 Under the assumption that the aggregate production function is concave, the (K K) matrix of cross-price eects Dw is negative semi-denite which implies that dW 0t dX t Dz dZ t dW 0t Dw dW t 6 0 1:3

which says that the changes in factor supplies (dXt) net of demand shifts (Dz dZt) and the changes in wages (dWt) must negatively covary. We can therefore test whether or not supply shifts alone can explain the changes in relative wages. If factor demand is stable (i.e., Zt is xed or dZt = 0) then Eqn. (1.3) implies that dW 0t dX t 6 0. If we compare two years s and t, and nd that W t W s 0 X t X s 6 0; 1:4

then the observed changes in relative wages can potentially be explained solely by supply shifts. In other words, if the inequality in Eqn. (1.4) holds then the period between years s and t could have experienced a xed factor demand which would have had no impact on relative wages. If the inequality in Eqn. (1.4) does not hold, then supply shifts alone cannot explain the relative wage changes. Instead, non-neutral or factor-biased demand shifts must also have played a role in explaining relative wage changes.

ECONOMIC LIBERALIZATION AND WAGE INEQUALITY IN INDIA

2007

We test the stable relative demand hypothesis by computing the inner products of changes in relative wages and the changes in relative supplies for the 100 gendereducationage groups (Wt Ws) 0 (Xt Xs) for our four time periods. The results are presented in Table 6, where three of the four comparisons are positive and therefore reject a stable relative demand hypothesis. During 198388 there is a possibility that relative demand was stable and therefore did not aect relative wages. For all other periods, we nd that shifts in the relative demand played an important role in the relative wage changes in India. The positive relationship between the changes in relative wages and relative supplies is the strongest for the overall period 19832000 (0.0218) with the sub-period 198794 being the largest (0.0170) of the three sub-periods. Figure 2 illustrates our results by plotting the changes in log relative wages against changes in log relative supplies for the 100 labor groups for each of the four periods. The lines drawn in the gures are predicted values from the weighed least squares regressions of changes in log relative wages on the changes in log relative supplies for each time period, using employment shares of each group in the initial period as weights. The graphs for the three subperiods 198388, 198794, and 19932000 reinforce our ndings from the inner products, showing a strong negative relationship in relative supplies and wages during 198388, a strong positive relationship during 198794, and a weak relationship during 19932000. Even though the inner product for the overall period 19832000 is positive, the graph for this period shows a weak relationship between relative supplies and wages. (b) Between- and within-sector demand changes To explore the role of relative demand changes in relative wage changes, we focus on two types of demand changesthose that

occur between industries (shifts that change the allocation of labor demand between industries at xed relative wages) and those that occur within industries (shifts that change the allocation of labor demand within industries at xed relative wages). Important sources of both between- and within-sector demand changes include skill-biased technology transfer, changes in the prices of nonlabor inputs such as capital, changes in product demand, and changes in the composition of domestic output. The between- and within-sector demand shift measures proposed by Katz and Murphy (1992) are based on the xed coecients manpower requirements index (Freeman, 1980). This index measures the percentage change in the demand for a demographic group as the weighted average of the percentage employment growth by industry, where the weights are the industrial employment distribution for the demographic group in a base period. The index can be written as P X Ejk DEj  j ajk DE j d ; 1:5 DX k Ek Ej Ek j where k indexes demographic groups and j indexes sectors. DX d is the change in demand k for group k, Ej is the total labor input in sector j, DEj is the change between years of total labor input in sector j, Ek is the base year employE ment of group k, and ajk Ejk is the group ks j share of total employment in sector j in the base period. The employment measures Ek and Ej are in eciency units. 10 We turn Eqn. (1.5) into an index of relative demand shifts by normalizing the employment measures Ek and Ej so that the total employment in eciency units in each year sums to one. We use the average of the four survey years to be our base period. Thus, we use the average share of total employment in sector j of group k over the 19832000 period as our measure of ajk and the average share of group k in total employment over the 19832000 period as our measure of Ek.

Table 6. Relationship between changes in relative wages and changes in relative supplies: 19832000 198388 Inner product Coecient Standard error 0.0289 0.6157*** (0.1592) 198794 0.0170 0.4471* (0.2310) 19932000 0.0031 0.0461 (0.0435) 19832000 0.0218 0.0147 (0.0270)

The reported numbers for inner products are the inner products of changes in relative wages and changes in relative supplies. The coecient estimates and standard errors are from regressions of log relative wages on log relative supplies, using employment shares of each group in the initial period as weights. A * and *** represent regression coecients that are signicant at the 10% and 1% levels respectively.

2008

WORLD DEVELOPMENT
1
1 lnw8388/Fitted values -.5 0 .5

-1.5 -.5

-1

-1

lnw8794/Fitted values -.5 0 .5

-1

-.5 lnw8388

0 lnx8388

.5

0 lnx8794 lnw8794

.5 Fitted values

Fitted values

1.5

lnw9300/Fitted values 0 .5 1

-.5

-.4

-.2

-5.55e-17 .2 lnx9300 lnw9300 Fitted values

.4

.6

-1 -1

lnw8300/Fitted values -.5 0 .5

-.5

0 lnx8300 lnw8300

.5 Fitted values

1.5

Figure 2. Wage and supply changes for 100 groups: 19832000.

We use Eqn. (1.5) to calculate overall, between, and within sector demand shifts based on employment in 18 industries and three occupations (dened in Table 3). We dene our overall (industry-occupation) demand shift index for group k, DX d , as the index given in Eqn. (1.5) k when j indexes our 48 industry-occupation cells. We also decompose this index into betweenand within-industry components. The betweenindustry demand shift index for group k, DX b , k is the index given in Eqn. (1.5) when j indexes the 18 industries. The within-industry demand shift index for group k, DX w , is calculated as k the dierence between the overall demand shift index and the between-industry demand shift index (i.e., DX w DX d DX b ). The withink k k industry demand shifts reect shifts in employment among occupations within industries. Table 7 reports the relative demand shift estimates for 10 demographic groups over six time periods. During 19832000, between-sector, within-sector, and overall demand shifted away from men and women with less than primary, primary, and middle schooling in favor of men and women with high school and college education. Between-sector shifts are substantially smaller than within-sector shifts for all

demographic groups, indicating that while there was some expansion of sectors employing more educated labor and contraction of sectors employing less educated labor, overall shifts in relative demand were caused primarily by skillupgrading within industriesthat is, shifts away from low-skill occupations and toward high-skill occupations within industries. The occupational shifts reported in Table 5 support this nding for the 19832000 period. 11 These within-sector shifts in favor of skilled labor support Hypothesis IIthat is, the creation of white-collar jobs as a result of domestic and/ or external sector reforms raised a demand for more educated labor. In order to identify how various sectors contributed toward the overall demand shifts presented in Table 7, we calculate the disaggregated demand shift measures for three broad sectorsagriculture, manufacturing, and services. Table 8 shows that the decline in the relative demand for uneducated and less educated men and women is dominated by the agricultural sector. However, the increase in the relative demand for high school and college educated men and women, is driven by the service sector. These results imply that during the

ECONOMIC LIBERALIZATION AND WAGE INEQUALITY IN INDIA Table 7. Sector and occupation based relative demand shifts: 19832000 Group 198388 Between sector shift Women <Primary school Primary school Middle school High school College Men <Primary school Primary school Middle school High school College Within sector shift Women <Primary school Primary school Middle school High school College Men <Primary school Primary school Middle school High school College Change in log relative demand (multiplied by 100) 198794 19932000

2009

19832000

2.05 0.89 0.01 0.87 0.45 1.32 0.08 0.07 0.71 1.42

2.55 2.74 0.69 4.63 6.72 2.56 2.02 1.18 1.03 2.95

2.76 1.58 1.04 2.41 2.14 0.81 0.62 0.47 0.78 1.12

7.81 5.41 1.76 1.64 4.49 4.85 1.46 0.62 2.48 5.30

6.90 5.35 2.74 0.82 4.42 6.57 4.28 2.91 1.86 9.43

3.33 3.71 1.15 6.34 9.19 3.17 2.69 1.70 1.07 5.49

8.48 6.97 3.22 1.09 3.03 5.35 2.73 3.19 1.03 8.24

20.82 17.52 7.36 6.15 15.54 16.38 10.15 8.06 3.86 21.00

Overall shift (industry and occupation) Women <Primary school 8.96 Primary school 6.25 Middle school 2.76 High school 0.05 College 3.97 Men <Primary school Primary school Middle school High school College

5.88 6.45 1.84 10.97 15.90

11.23 8.55 4.26 3.50 0.89

28.63 22.93 9.12 7.79 20.03

7.89 5.73 6.15 21.23 4.36 4.72 2.11 11.61 2.84 2.88 2.72 8.69 2.57 2.10 1.81 6.34 10.84 8.45 9.36 26.30 The reported numbers are of the form log 1 DX s 100, where s represents between sector (b), within-sector (w), k and overall (d) demand.

1980s and 1990s Indias economy moved away from agriculture and toward services, generating lower demand for unskilled labor and higher demand for skilled labor. The contraction of agriculture and expansion of services, together with increased demand for

skilled workers is a strong evidence against Hypothesis I. The liberalization of the domestic and external sectors in India appears to have shifted employment away from unskilled-labor intensive and toward skilled-labor-intensive activities.

2010

WORLD DEVELOPMENT Table 8. Relative demand shifts in agriculture, manufacturing, and services: 19832000

Group 198388 Overall demand shift in agriculture Women <Primary school 6.97 Primary school 5.10 Middle school 2.79 High school 0.51 College 0.13 Men <Primary school Primary school Middle school High school College 5.75 3.48 3.47 1.34 0.69

Change in log relative demand (multiplied by 100) 198794 19932000 19832000

6.04 4.18 2.24 0.51 0.07 4.94 2.98 3.00 1.22 0.53

10.19 6.38 3.14 0.43 0.34 8.21 4.75 4.74 1.33 0.10

25.21 16.55 8.41 1.46 0.14 20.19 11.65 11.64 3.94 1.13

Overall demand shift in manufacturing Women <Primary school 0.65 Primary school 0.94 Middle school 0.35 High school 0.70 College 1.12 Men <Primary school Primary school Middle school High school College Overall demand shift in services Women <Primary school Primary school Middle school High school College Men <Primary school Primary school Middle school High school College 0.38 0.01 0.22 1.62 3.69

1.87 3.97 3.41 1.29 0.55 3.18 4.13 3.20 2.17 1.01

0.09 0.07 0.15 0.21 0.36 0.07 0.11 0.38 0.85 1.07

2.63 4.87 3.92 0.37 0.93 3.66 4.02 2.58 0.35 3.76

0.83 0.05 1.09 0.07 2.79

1.88 1.82 3.73 12.47 15.96

0.83 1.03 0.53 2.83 0.22

0.25 0.77 5.29 9.91 18.53

0.83 1.94 2.03 3.12 0.30 2.34 2.18 4.19 0.41 2.92 1.16 4.44 2.09 5.05 0.98 7.95 7.11 9.22 6.75 21.52 The reported numbers are of the form log 1 DX s 100, where s represents between sector (b), within-sector (w), k and overall (d) demand.

6. DEMAND SHIFTS ARISING FROM INTERNATIONAL TRADE The occurrence of a widening wage dierential between unskilled and skilled workers, to-

gether with large increases in the demand for skilled workers during the two decades that span Indias economic reforms suggest that the SET Hypothesis might have played a role as a result of trade liberalization. In this section

ECONOMIC LIBERALIZATION AND WAGE INEQUALITY IN INDIA

2011

we examine the extent to which international trade in manufactured goods and services contributed to relative demand shifts during the 1980s and 1990s in India. Following Katz and Murphy (1992) we estimate the labor supply equivalents of trade (i.e., the implicit labor supply embodied in trade) by transforming trade ows into labor supply equivalents on the basis of the utilization of labor inputs in the domestic manufacturing industries. We measure only the direct labor supply embodied in trade and ignore the inputoutput eects. Therefore, the implicit labor supply in trade is the labor input required to produce traded output domestically. This section essentially utilizes the factor content approach to measure the impact of trade on wages. As documented by Wood (1995), this approach has several major drawbacks according to Leamer (1994) but may nevertheless be the appropriate method of analysis due to data limitations. The major criticism is that trade aects wages only through prices, making it unnecessary to focus on the factor content in trade ows. Also, trade ows are a function of wages and therefore endogenous. Given these problems, a more appropriate empirical analysis would be to examine how plausibly exogenous industry taris aect the demand for skilled and unskilled labor. This would allow us to estimate the causal impacts of trade liberalization on relative demand using trade prices rather trade ows as the measure of trade reform. Despite the advantages of using taris rather than trade ows, we use the factor content approach in our analysis because even though data on taris is available at a reasonable level of industry disaggregation, such an analysis would require detailed rm-level or industrylevel data. Estimating the impact of industrylevel taris on demand for skilled and unskilled labor at the rm or industry level within the context of a regression requires information on the employment of skilled and unskilled labor at the rm or industry level. To the best of our knowledge, such data is available in the Annual Survey of Industries but only for a few years during the 1980s, which does not encompass the period of Indias external liberalization, making it inappropriate for our empirical analysis. An alternative method of analysis is to estimate the impact of taris directly on the industry wages for skilled and unskilled men and women, as in Dutta (2004), Reilly and Dutta (2005), and Kumar and Mis-

hra (2005) who use the same NSSO data as we do in this analysis. This direct approach is more informative and robust if the aim is to estimate the direct impact of trade liberalization on wage inequality and not distinguish between supply-side and demand-side factors in shifting wages. The focus of this section is to estimate the impact of external sector reforms on relative demand shifts not on wage inequality, making the factor content approach more appropriate and the only one given the data limitations. We measure Lk , the implicit labor supply of t demographic group k embodied in trade in year t as  ! X I it ek Eit Lk ; 1:6 t i Y it i where i indexes 16 manufacturing industries, k indexes 10 demographic groups (two gender and ve education groups), and t indexes four years. ek is the average proportion of employi ment in industry i among workers in group k over the 19832000 period, Eit is the share of P employment in industry i in year t i E it 1 , Iit is the net imports in industry i in year t (Importsit Exportsit), and Yit is the output in industry i in year t. Positive net imports imply that the country is importing more foreign labor than exporting domestic labor, which will result in a fall in domestic labor demand. 12 We measure T k , the eect of trade on relative t demand for group k in year t as  X  ! X   1 I it I it ; Tk k ek Eit Eit t i Y it Y it E i i 1:7 where E is the average share of total employment of group k during the 19832000 period. 13 The rst term in Eqn. (1.7) is the implicit labor supply of group k embodied in trade, normalized by base year employment of group k (Ek) with the sign reversed to convert the supply shift measure into a demand shift measure. The second term adjusts the demand shift measure so that trade aects only relative demands for labor. 14 We use data on imports, exports, and output by industry for the years 1983, 1988, 1993, and 1999 from the Trade & Production Database, provided by the World Bank. These data cover 3-digit ISIC manufacturing industries, which we aggregate into 16 industry groups, as shown in Table 9.
k

2012

WORLD DEVELOPMENT Table 9. Classication of manufacturing industries ISIC code Industry Manufacture of all food products Manufacture of beverages, tobacco, and related products Manufacture of all textiles Manufacture of textile products, including wearing apparel Manufacture of wood and wood products Manufacture of paper, paper products, printing, and publishing Manufacture of leather, leather products, fur, leather substitutes, including footwear Manufacture of chemicals and chemical products Manufacture of rubber, plastic, petroleum, and coal products Manufacture of nonmetallic mineral products Manufacture of basic metals Manufacture of metal products, metal parts, except machinery and equipment Manufacture of machinery and equipment, excluding transport Manufacture of electrical machinery Manufacture of transport equipment and transport parts Other manufacturing industries

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

311 313, 321 322 331, 341, 323, 351, 353, 361, 371, 381 382 383 384 385,

314

332 342 324 352 354, 355, 356 362, 369 372

390

Table 10. Relative demand shifts predicted by changes in international trade-in manufactures: 19832000 Group Women <Primary school Primary school Middle school High school College Men <Primary school Primary school Middle school High school College 198388 0.046 0.155 0.319 0.094 0.023 0.043 0.021 0.027 0.034 0.035 198794 0.078 1.462 2.718 0.905 0.195 0.115 0.003 0.051 0.114 0.117 19932000 0.019 0.388 0.527 0.150 0.056 0.008 0.009 0.000 0.017 0.017 19832000 0.013 2.006 3.564 1.149 0.274 0.080 0.027 0.025 0.097 0.100

The reported numbers are of the form DT k 0 T k0 T k , where t and t 0 represent dierent years and t 0 > t. tt t t

Table 10 presents the estimated changes in relative demand predicted by the changes in international trade-in manufactures for 10 demographic groups over six time periods. Several points are worth emphasizing. First, even though the magnitude of trade induced demand shifts are small compared to the overall demand shifts presented in Table 7, the units of these shifts are not comparable. Second, even though the overall demand for female high school and college graduates increased substantially during 19832000, trade reduced the demand for these groups. Third, trade generated the demand for men with middle, high school, and college education. Perhaps this gender dierential can be explained by the fact that skilled jobs in the manufacturing sectorsuch as technicians, supervisors, and managerstend to be domi-

nated by men. Fourth, the largest manufacturing trade-induced demand shifts occurred during 198794 for both men and women. This was the period immediately following Indias trade liberalization reforms, which were initiated during the late 1980s and strengthened from 1991 onwards. Therefore, manufacturing trade-induced demand shifts were an immediate response to external sector reforms. We next turn to measuring the contribution of trade-in services to relative demand shifts. Using data for payments, receipts, and output for three broad groups of servicesnamely transportation, insurance, and miscellaneous services (communication, construction, nancial, software, news agency, royalties, management, and other services)we estimate the contribution of international trade-in services to relative de-

ECONOMIC LIBERALIZATION AND WAGE INEQUALITY IN INDIA Table 11. Relative demand shifts predicted by changes in international trade-in services: 19832000 Group Women <Primary school Primary school Middle school High school College Men <Primary school Primary school Middle school High school College 198388 0.006 0.007 0.008 0.006 0.002 0.005 0.005 0.005 0.004 0.003 198794 0.010 0.012 0.012 0.010 0.000 0.007 0.007 0.008 0.007 0.001 19932000 0.053 0.063 0.064 0.052 0.012 0.041 0.038 0.044 0.035 0.019

2013

19832000 0.037 0.044 0.044 0.036 0.010 0.029 0.027 0.031 0.025 0.015

The reported numbers are of the form DT k 0 T k0 T k , where t and t 0 represent dierent years and t 0 > t. tt t t

mand shifts between 1983 and 2000 in India. The data are published by the Reserve Bank of India as trade-in invisibles. Table 11 shows that for the period during 19832000, trade-in services decreased the relative demand for men and women without a college degree and increased the demand for male and female college graduates. Again, even though these demand shifts are small compared to the overall demand shifts for these groups, the units are not comparable. The largest service trade-induced demand shifts occurred during 19932000the period when trade and investment were increasingly liberalized. Thus, there was a delayed response of service trade-induced demand shifts to the liberalization of trade and investment, which only started generating demand for college graduates during the mid- and late-1990s. 7. CONCLUSION Together, our relative wage and demand shifts for the overall period 19832000 support Hypothesis II and reject Hypothesis I. 15 We nd large increases in within-sector demand for skilled men and women, indicating the creation of more white-collar jobs, which occurred mostly within the service sector. These increases, together with a shrinking of the agricultural sector and expansion of the service sector reject Hypothesis I. Even though the SS theorem predicts that trade liberalization should benet the abundant factor and hurt the previously most protected sectors, during the 1980s and 1990s India experienced both liberalization of trade and a rising skill premium. Our tradeinduced demand shifts nd some support for

Hypotheses II, III, IV, and V together, even though we cannot decompose the trade-induced demand increases for skilled labor as arising from either skill-upgrading within industries, global production sharing, skill-biased technical change, or quality-upgrading. We nd relative demand shifts for the overall economy that are explained neither by trade-in manufactures nor services. During 19832000, the demand for uneducated women employed in manufacturing fell, but trade-in manufactures raised the demand for this group. The overall demand for female college graduates increased within the manufacturing sector even though manufacturing trade-induced demand fell for this group. The demand for men without any education and with a middle school education fell in the manufacturing sector. However, manufacturing trade increased demand for both these groups. Similar dierences are found for the service sector as wellwhereas overall demand for men and women with less than primary, primary, middle, and high school education increased within the service sector, outsourcing of service activities from foreign countries to India (i.e., trade-in services) actually decreased demand for all these groups. These discrepancies suggest that skill-deepening in the Indian economy was not solely generated by the increased trade-in manufactures and services brought about by trade liberalization, but was also the result of changes within the economy that were not related to trade. Perhaps domestic sector reformssuch as deregulation and delicensing of industry, privatization, and possibly even tax reformswere responsible for generating some of the increased demand for skilled labor. This suggests that in addition

2014

WORLD DEVELOPMENT

to external sector reforms, domestic market reforms may have been responsible for widening the skill wage gap in India. 16 What explains the narrowing gender wage dierential for high school and college graduates during the 1980s and 1990s in India? Our nding that international trade-in manufactures beneted skilled men but hurt skilled women implies a widening gender wage gap. Trade-in services, on the other hand, beneted both skilled men and women, implying a constant gender wage gap. Thus, our ndings for trade-induced demand shifts appear to be inconsistent with a narrowing gender wage gap. We provide the following explanation to reconcile this inconsistency. As demand for skilled labor in the tradable sector expanded

as a result of trade liberalization, skilled men, relocated from the nontradable to the tradable sector, making it necessary for skilled women to replace them in the nontradable sector. 17 Our argument holds even if skill-deepening did not occur within the nontradable sector but is strengthened if the nontradable sector became increasingly skill intensive. If one accepts this explanation then our trade-induced demand shifts do indeed support the narrowing gender wage dierential. By generating more skilled employment, the skill-enhancing trade may have directly beneted skilled men more than skilled women. However, the indirect benets together with the direct benets of trade liberalization may have made educated women in India gained truly from trade.

NOTES
1. See, for example, Katz and Murphy (1992), Bound and Johnson (1992), Katz and Revenga (1989), and Murphy and Welch (1991). 2. Domestic reforms consisted primarily of industrial deregulation, which were implemented during the midand late-1980s. The external sector reforms, such as trade and investment liberalization though introduced during the late-1980s, were more thoroughly implemented starting from the early-1990s. 3. See Goldberg and Pavcnik (2004) for an overview of the literature on trade and wage inequality. 4. The SET Hypothesis is attributed to Robbins (1996). 5. Alternatively, these results could mean that the SS theorem does in fact hold in India, though it is overwhelmed by support for the SET hypothesis. 6. We conducted the same analysis dening the wage and count samples identicallythat is, including only regular wage and salary workers in both samplesand obtained very similar estimates as reported in this paper. Thus, focusing solely on the formal labor market does not generate dierent results. 7. Each groups wage is indexed to the wages for a xed bundle of workers (all workers who earned a regular wage or salary). Thus, the relative wage for each group is measured as each groups wage relative to the wages for a xed bundle of workers. 8. Each groups labor supply is measured relative to the total labor supply in the economy in eciency units. 9. In our analysis K represents 100 distinct labor groups, dened by two gender groups (male and female), ve education groups (less than primary, primary, middle, high school, and college), and 10 age groups (1520, 2025, 2530, 3035, 3540, 4045, 4550, 50 55, 5560, and 60+ years). 10. For the employment measures Ek and Ej, we weighed the total labor input (hours worked) of each group or sector by the average relative wage of that group or sector to construct measures of labor demand in eciency units. 11. In assessing these results, two factors must be taken into account. First, as noted in Katz and Murphy (1992), the between-sector demand shift index is a biased measure because it does not measure demand shifts only at xed relative wages, but also includes the demand shifts brought about by changing the relative wages. Second, because the within-sector demand shifts proposed by Katz and Murphy (1992) measure shifts in employment only between three occupation groups, they might not capture the full eect of within-sector changes in the relative demand. We would require a more detailed information on the skill content of occupations within industries to obtain more precise within-sector demand shift estimates. 12. ek and Eit are measured in eciency units by i weighing the total labor input (hours worked) of each group or sector by the average relative wage of that group or sector. 13. Refer to Murphy and Welch (1991) for details of this demand shift index.

ECONOMIC LIBERALIZATION AND WAGE INEQUALITY IN INDIA 14. Ek is measured in eciency units by weighting the total labor input (hours worked) of each group by the average relative wage of that group. 15. As mentioned in 1, our results could also mean that the SS theorem does in fact hold though it is overwhelmed by support for the SET hypothesis. 16. This hypothesis is denitely worth testing, but would require detailed rm- or industry-level data in

2015

addition to data on deregulation and privatization. While this is beyond the scope of the current paper, we intend to pursue this in future research. 17. This is similar to what occurred in developed countries during World War II, when women increasingly entered the labor market to replace the men who went to war.

REFERENCES
Acemoglu, D. (2003). Patterns of skill premia. Review of Economic Studies, 70, 199230. Attanasio, O., Goldberg, P., & Pavcnik, N. (2004). Trade reforms and wage inequality in Colombia. Journal of Development Economics, forthcoming. Bound, J., & Johnson, G. (1992). Changes in the structure of wages during the 1980s: an evaluation of alternative explanations. American Economic Review, 82(3), 371392. Cragg, M., & Epelbaum, M. (1996). Why has wage dispersion grown in Mexico? Is it the incidence of reforms or the growing demand for skills? Journal of Development Economics, 51, 99116. Dutta, P. V. (2004). Trade protection and inter-industry wages in India. PRUS Working Paper No. 27, Poverty Research Unit at Sussex, University of Sussex. Feenstra, A. D., & Hanson, G. H. (1996). Foreign investment, outsourcing and relative wages. In R. C. Feenstra, G. M. Grossman, & D. A. Irwin (Eds.), Political economy of trade policy: Papers in honor of Jagdish Bhagwati (pp. 89127). Cambridge, MA: MIT Press. Feenstra, A. D., & Hanson, G. H. (1997). Foreign direct investment and relativewages: evidence from Mexicos Maquiladoras. Journal of International Economics, 42, 371393. Feenstra, A. D., & Hanson, G. H. (2003). Global production sharing and rising inequality: a survey of trade and wages. In E. Choi, & J. Harrigan (Eds.), Handbook of international trade (pp. 146185). Malden, MA: Blackwell. Freeman, R. B. (1980). An empirical analysis of the xed coecients manpower requirements model, 1960 1970. The Journal of Human Resources, 15(2), 176199. Goldberg, P. K., & Pavcnik, N. (2004). Trade, inequality, and poverty: what do we know? Evidence from recent trade liberalization episodes in developing countries. In S. M. Collins, & D. Rodrik (Eds.), Brookings trade forum 2004: Globalization, inequality, and poverty (pp. 223269). Washington, DC: Brookings Institutions Press. Gorg, H., & Strobl, G. (2002). Relative wages, openness, and skill-biased technological change in Ghana. Unpublished paper. Harrison, A., & Hanson, G. (1999). Who gains from trade reform? Some remaining puzzles. Journal of Development Economics, 59, 125154. Katz, L. F., & Murphy, K. M. (1992). Changes in relative wages, 196387: supply and demand factors. Quarterly Journal of Economics, 107(1), 3578. Katz, L. F., & Revenga, A. L. (1989). Changes in the structure of wages: the United States versus Japan. Journal of the Japanese and International Economies, 3, 522553. Kumar, U., & Mishra, P. (2005). Trade liberalization and wage inequality: evidence for India. International Monetary Fund Working Paper. Leamer, E. E. (1994). Trade, wages, and revolving door ideas. NBER Working Paper No. 4706. Murphy, K. M., & Welch, F. (1991). The role of international trade in wage dierentials. In M. Kosters (Ed.), Workers and their wages (pp. 3969). Washington, DC: AEI Press. Pavcnik, N. (2003). What explains skill upgrading in less developed countries? Journal of Development Economics, 71, 311328. Reilly, B., & Dutta, P. V. (2005). The gender pay gap and trade liberalization: Evidence for India. PRUS Working Paper No. 32, Poverty Research Unit at Sussex, University of Sussex. Robbins, D. (1996). HOS hits facts: Facts win; Evidence on trade and wages in the developing World. Unpublished paper. Verhoogen, E. (2004). Trade, quality upgrading and wage inequality in the Mexican manufacturing sector: Theory and evidence from an exchange-rate shock. Unpublished paper, University of California, Berkeley. Wood, A. (1995). How trade hurt unskilled workers. Journal of Economic Perspectives, 9(3), 5780.