Feminization, Defeminization, and Structural Change ...

9 downloads 948 Views 1MB Size Report
changes in female shares of employment and technological upgrading was associated with ... tion] encourages employment growth in labor-intensive sec-.
World Development Vol. 64, pp. 569–582, 2014 0305-750X/Ó 2014 Elsevier Ltd. All rights reserved. www.elsevier.com/locate/worlddev

http://dx.doi.org/10.1016/j.worlddev.2014.06.033

Feminization, Defeminization, and Structural Change in Manufacturing DAVID KUCERA and SHEBA TEJANI* International Labour Organization, Geneva, Switzerland Summary. — The paper uses accounting decomposition methods to analyze changes in female shares of manufacturing employment for 36 countries at different levels of development from 1981 to 2008, for the manufacturing sector as a whole and within a group of laborintensive manufacturing industries for selected countries. For the majority of countries, feminizing and defeminizing, labor-intensive industries contributed most to changes in female shares of total manufacturing employment and within-industry effects were more important than employment reallocation effects. Within labor-intensive industries, textiles, and apparel were the largest drivers of changes in female shares of employment and technological upgrading was associated with defeminization. Ó 2014 Elsevier Ltd. All rights reserved. Key words — feminization, defeminization, structural change, manufacturing, decomposition analysis, technological upgrading

1. INTRODUCTION

labor-intensive industries that produce primarily for export” (2000, p. 33). 3 The wearing apparel and textile industries provide the most clearcut examples. Yet the point can be overstated. There is a remarkably strong similarity in female shares of employment among manufacturing industries between diverse pairs of countries at different levels of development. But the relationship between female shares of employment and the labor-intensity of production among manufacturing industries within countries is comparatively weak, as the next section of the paper describes. Caraway’s second premise also merits remark. Successful economic development is characterized by structural change, including shifts toward higher value-added activities for which labor productivity (value-added per unit of labor) provides a useful measure. As Ocampo, Rada, and Taylor (2009) write in their volume on structuralist economics, “Historically, labor productivity increases have been the major contributing factor to growth in real GDP per capita” (p. 42). But these shifts can occur both within sectors and through compositional shifts toward higher value-added sectors. Indeed, studies applying accounting decomposition methods generally find that within-industry effects are more important than reallocation effects in driving labor productivity growth, consistent with Kaldor’s theories of economic growth (Kaldor, 1968; Ocampo et al., 2009; Roncolato & Kucera, 2014). By the same token, changes in female shares of total manufacturing employment can result from the reallocation of employment among manufacturing industries as well as by changes in female shares within manufacturing industries. The literature on gender and industrialization suggests that shifts in the labor- and capital-intensity of production have different consequences for men and women’s employment. To our knowledge, though, no study has provided systematic evidence on the relative importance of within-industry versus employment reallocation effects on the feminization and defeminization of manufacturing employment as well as which industries are driving these changes. 4 Conversely, while the distinction between within-industry and employment reallocation effects

A 1999 United Nations (UN) report on the role of women in development stated that “It is by now considered a stylized fact that industrialization in the context of globalization is as much female-led as it is export-led” (UN, 1999, p. 29). Archetypes of such female-led and export-led growth are the first-tier newly industrialized countries (NICs) South Korea and Taiwan (China). Yet by the time of the UN report, female shares of manufacturing employment had already been declining in these countries for some time. In this sense, a pivotal development with respect to manufacturing employment is feminization and defeminization, defined in terms of changing female shares of employment. 1 In her assessment of the large literature on gender and industrialization, Caraway writes that “To this day, we lack a compelling story explaining. . .different gendered industrialization trajectories,” by which she refers to both patterns across countries and over time, particularly patterns of feminization and defeminization (2006, p. 26). Regarding patterns over time, Caraway provides a concise formulation: Trends over time – feminization versus masculinisation – [are] best explained by the balance of employment between sectors. Since primary EOI [export-oriented industrialization] encourages employment growth in labor-intensive sectors relative to capital-intensive sectors, there is a strong relationship between EOI and feminization. However, as EOI matures, masculinization ensues since employment usually expands more rapidly in capital-intensive sectors (Caraway, 2006, p. 41). Two premises in Caraway’s account are noteworthy. The first is that sectors using more labor-intensive production methods tend to employ higher shares of women workers, and the other way around for sectors using more capital-intensive production methods. The second premise is that feminization and defeminization can be largely explained by the reallocation of employment (female and male) among sectors rather than changes within sectors. 2 The first premise is commonly expressed in the literature. Regarding Asia, for example, Seguino writes that “within the manufacturing sector, women have been sequestered in

* The authors would like to thank Leanne Roncolato, Arianna Rossi and three anonymous reviewers for their helpful comments. Any errors or omissions remain ours alone. Final revision accepted: June 29, 2014. 569

570

WORLD DEVELOPMENT

is central to empirical studies of structural change, this literature has paid scant attention to the gendered impacts of these shifts. Our paper is situated at the intersection of these two literatures and endeavors to address some of these limitations by applying the tools of structuralist analysis to the debates on gender and industrialization. 5 It applies accounting decomposition methods to analyze changes in female shares of manufacturing employment for 36 countries at different levels of development and addresses which industries—labor-intensive, capital-intensive, or intermediate—contributed most to changes in the female share of manufacturing employment and whether within-industry or employment reallocation effects were more important in explaining these changes. The first-round of analysis encompasses the manufacturing sector as a whole and is conducted at the level of three industry groups classified by the labor-intensity of production. Motivated by the findings of this analysis, we subsequently apply the same methods to analyze changes in female shares of employment within the group of labor-intensive manufacturing industries for five countries of particular interest: South Korea, Taiwan, and Malaysia, which feature prominently in the literature on gender and export-oriented industrialization as well as on successful late-development and experienced the strong defeminization of manufacturing; and Morocco and Turkey, more recent entrants into export-oriented industrialization which experienced the feminization of manufacturing employment. 6 For the manufacturing sector as a whole, our main findings are that for 25 of the 36 countries in our sample—feminizing and defeminizing and at different levels of development— labor-intensive industries contributed most to changes in the female share of total manufacturing employment. For 33 of 36 countries, within-industry effects were more important than reallocation effects in accounting for these changes. In turn, changes in the female share of total manufacturing employment were most commonly driven by changes in the female share of employment within the group of labor-intensive manufacturing industries. Though at odds with Caraway (2006), these findings are unsurprising in light of the weak correlation between the labor-intensity of production and the femaleintensity of employment for the countries in our sample. In this context, even structural change characterized by sustained shifts away from labor-intensive industries would not have a decisive effect on female shares of manufacturing employment, given that both feminization and defeminization are generally driven by within-industry rather than reallocation effects. Within the group of labor-intensive manufacturing industries, we find that the textile and apparel industries played key roles in South Korea, Taiwan, Malaysia, Morocco and Turkey. In South Korea and Taiwan in particular, the single most important driver of the defeminization of manufacturing employment was the declining female share of employment within the textile industry, which was associated with significant technological upgrading in the industry. These findings are relevant to debates in the literature on global production networks on whether “economic upgrading” leads to “social upgrading,” with the latter defined in terms of improvements in the quality of employment for different groups of workers (Milberg & Winkler, 2011; Rossi, 2013). 7 Other studies have also observed defeminization in the context of technological upgrading within particular manufacturing industries. 8 Deepening our understanding of the phenomena of defeminization occurring alongside technological upgrading is essential for enabling women to benefit from processes of structural change, as regards new industrial jobs created through employment reallocation as well as within industries.

Several studies have wrestled with an explanation of employers’ seeming preference for men workers in the context of technological upgrading (Barrientos et al., 2004; Joekes, 1999; Seguino, 2005; Seguino, Berik, & Rodgers, 2010; Sundaram, 2009; Tejani & Milberg, 2010; Tejani, 2012). These studies invoke the lesser importance of low-wage women’s labor in more capital-intensive production; gender norms designating men as breadwinners and women as secondary workers, with men more likely to be hired for higher paying jobs; and the different skills requirements of new industrial jobs combined with the purportedly different skills of men and women workers and whether these differences are real or perceived. A useful overview is provided by Seguino (2005) in the East Asian context, who writes: “An additional problem for East Asia has been that the new industrial jobs that are emerging are gender-typed. . .. The reason for gender-typing such jobs is not clear” (p. 7, italics added). 9 As a complement to our decomposition analysis of changes in female shares within labor-intensive manufacturing industries, we provide preliminary evidence on these debates for South Korea, Taiwan, Malaysia, Morocco, and Turkey. In particular, we document for the textile and apparel industries the association between technological upgrading and defeminization as well as industry expansion and feminization. We also describe patterns of female shares of employment within services and agriculture and for the economy as a whole. 2. DATA AND PATTERNS OF GENDER REPRESENTATION Our analysis is based on manufacturing employment data from the United Nations Industrial Development Organization’s (UNIDO) Industrial Statistics Database (UNIDO, 2011a). These data refer to employees and exclude “home workers. . .working proprietors, active business partners and unpaid family members” (UNIDO, 2011b, p. 33). Sorting out the countries with adequate data left us with the 36 countries in Figure 1, showing female shares of total manufacturing employment from 1981 to 2008. Data are provided at the ISIC (International Standard Industrial Classification) two-digit level, Revision 3, for which there are 22 manufacturing industries. 10 Motivated by Caraway’s (2006, p. 41) discussion of labor-intensive and capital-intensive industries and to facilitate presentation and the construction of consistent time series (given missing data at the two-digit level), the data for the 22 industries are compiled into three industry groups classified by the labor-intensity of production. We refer to these as laborintensive, intermediate, and capital-intensive industry groups, as shown in Table 1. 11 Labor-intensive manufacturing industries are similar to those characterized as such by other studies and include textiles (ISIC 17) and apparel (ISIC 18) and other light manufacturing for which there are not prohibitive technological barriers, at least for many products within these industries. 12 Intermediate manufacturing industries include all machineryproducing industries, a set of closely allied industries making rubber, plastic, and non-metallic mineral products, as well as printing and publishing and paper and paper products. Capital-intensive manufacturing industries are those for which the product lends itself to production by highly automated “continuous-process” technologies (Chandler, 1977). The textile and apparel industries play central roles in the literature on gender and structural change as well as on global production networks, and both tend to be female-intensive and labor-intensive. Yet other female-intensive industries

FEMINIZATION, DEFEMINIZATION, AND STRUCTURAL CHANGE IN MANUFACTURING

571

Austria

Azerbaijan

Bangladesh

Bulgaria

Canada

Chile

Croatia

Cyprus

Egypt

Eritrea

Ethiopia

Germany

India

Indonesia

Iran

Ireland

Japan

Jordan

Kazakhstan

Kenya

Kyrgyzstan

Lithuania

Macao

Malaysia

Malta

Morocco

New Zealand

Panama

Philippines

Portugal

Puerto Rico

South Korea

Sri Lanka

Taiwan

Turkey

United Kingdom

80 70 60 50 40 30 20 10 0

80 70 60 50 40 30 20 10 0

80 70 60 50 40 30 20 10 0

80 70 60 50 40 30 20 10 0

Figure 1. Female share of manufacturing employment (%).

Table 1. Manufacturing industry groups ISIC (Rev. 3)

Labor-intensive manufacturing industries

17 18 19 20 28 36

Textiles Wearing apparel; dressing and dyeing of fur Tanning and dressing of leather; luggage, handbags, saddlery, harness, and footwear Wood and cork products, except furniture; articles of straw and plaiting materials Fabricated metal products, except machinery and equipment Furniture; manufacturing (not elsewhere classified) Intermediate manufacturing industries

21 22 25 26 29 30 31 32 33 34 35

Paper and paper products Publishing, printing, and reproduction of recorded media Rubber and plastic products Other non-metallic mineral products Machinery and equipment (not elsewhere classified) Office, accounting, and computing machinery Electrical machinery and apparatus (not elsewhere classified) Radio, television, and communication equipment and apparatus Medical, precision, and optical instruments, watches and clocks Motor vehicles, trailers, and semi-trailers Other transport equipment Capital-intensive manufacturing industries

15 16 23 24 27

Food products and beverages Tobacco products Coke, refined petroleum products, and nuclear fuel Chemicals and chemical products Basic metals

2004

2008

2000

1992

1996

1984

Year

1988

1980

2008

2004

1996

2000

1988

1992

1984

1980

2004

2008

1996

2000

1992

1984

1988

1980

2008

2000

2004

1992

1996

1984

1988

1980

2004

2008

2000

1992

1996

1984

1988

1980

2008

2004

2000

1992

1996

1984

1988

80 70 60 50 40 30 20 10 0

1980

Percentage

80 70 60 50 40 30 20 10 0

572

WORLD DEVELOPMENT

featuring in this literature are not evidently labor-intensive, notably the electronics industries which we classify as intermediate (ISIC 30 and 32). Conversely, wood products (ISIC 20) and fabricated metal products (ISIC 28) are labor-intensive industries that tend to be disproportionately male. Indeed there is only a weak correlation within countries between female shares of employment and the labor-intensity of production. For the countries in our sample, the average correlation coefficient (Pearson) between the two measures at the two-digit ISIC level is just 0.14 (with labor-intensity measured as the ratio of employment to value-added, both from the UNIDO database). 13 Moreover, while the female share of employment is highest for the labor-intensive industry group, it is actually somewhat higher for the capital-intensive than intermediate industry group. 14At least for our sample, then, there is no neat correspondence between female-intensity and labor-intensity. These observations need to be borne in mind when considering the relationship between the shift to higher value-added manufacturing industries and the feminization and defeminization of manufacturing employment. Looking instead at cross-industry correlations between pairs of countries in our sample, we tend to find stronger correlations for female shares of employment than for the laborintensity of production, even though the former is determined socially and the latter technically. Specifically, the average correlation coefficient (Pearson) is 0.75 for female shares of employment and 0.56 for the labor-intensity of production. 15 In sum, between the diverse pairs of countries in our sample, women’s patterns of representation across manufacturing industries is broadly similar, yet these patterns do not closely match with the labor-intensity of production.

The female share of manufacturing employment is defined as female manufacturing employment divided by total (female and male) manufacturing employment, or q = F/T, and industry-level female shares within manufacturing are correspondingly defined as qi = fi/ti. This can be also expressed as: ð1Þ

Taking first-order differences with respect to time, changes in the female share of manufacturing employment can be expressed relative to the initial share as follows: n ¼ R½/i0 ðgi  ni Þ þ ð/i0  ðq1 =q0 Þsio Þni 

ð3Þ

In words, the within-industry effect is the difference between industry-level female employment growth and total employment growth weighted by the distribution of female employment in the industry (relative to female employment for the manufacturing sector as a whole). In this sense, positive within-industry effects result when industry-level female employment grows faster than industry-level total employment. The interaction term is represented by q1/q0, the result of first-order differencing in discrete time steps, while the interaction effect is the difference between changes in the female share of manufacturing employment and the sum of within-industry and reallocation effects. Leaving the interaction term aside, the reallocation effect on changes in the female share of manufacturing employment is represented by the right-hand bracketed term in Eqn. (2), that is: nr ¼ R½ð/i0  si0 Þni 

ð4Þ

The reallocation effect is the difference between the distribution of female employment in the industry (relative to female employment for the manufacturing sector as a whole) and the distribution of total employment in the industry (relative to total employment for the manufacturing sector as a whole), multiplied by industry-level employment growth. Positive reallocation effects result when industry-level employment grows in industries for which the difference between the distribution of female employment and the distribution of total employment is positive, that is, in industries with above average female shares of manufacturing employment. 16 4. DECOMPOSITION RESULTS: MANUFACTURING AS A WHOLE

3. DECOMPOSITION METHOD

q ¼ F =T ¼ Rfi =Rti

nw ¼ R½/i0 ðgi  ni Þ

ð2Þ

where: n ¼ ðq1  q0 Þ=q0 ni ¼ ðti1  ti0 Þ=ti0 gi ¼ ðfi1  fi0 Þ=fi0 /i0 ¼ fi0 =F 0 sio ¼ ti0 =T 0 Changes in the female share of manufacturing employment can be decomposed into within-industry versus reallocation effects based on the reallocation of either female employment or total employment. Our decomposition is based on the reallocation of total employment, consistent with our interest in the changing industry-level composition of employment in the process of development. The within-industry effect on changes in the female share of manufacturing employment is represented by the left-hand bracketed term in Eqn. (2), that is:

Country-level decomposition results for changes in the female share of manufacturing employment are shown in Table 2, based on the labor-intensive, intermediate, and capital-intensive industry groups. The 36 countries in our sample are grouped by seven regions: developed, Central and Southeast Europe and the Commonwealth of Independent States (CIS), and—for developing countries—East Asia, South Asia, Latin America and the Caribbean (LAC), Middle-East and North Africa (MENA) and Sub-Saharan Africa (SSA). 17 Note that decomposition results are mainly based on annual (rather than endpoint) changes in data (see Appendix: Data notes for exact years of data used for each country). To facilitate presentation, decomposition results are converted to percentages by dividing by the absolute value of the period average change for each country in the female share of manufacturing employment and multiplying by 100. When period averages of changes in female shares of manufacturing employment are small, percentages can be greater than 100, with Bulgaria an extreme case. Regional patterns of feminization and defeminization are addressed by column (2), which shows whether the sum of each industry groups’ contribution to changes in the female share of manufacturing employment (which always add to 100) is positive or negative, with shading used for positive sums. Though this does not show the magnitude of changes in the female share (see Figure 1 for this), it does provide an immediate sense of overall directions of change. Taking developed and Central and Southeast Europe and CIS countries together, there is a general pattern of defeminization, with negative values for 13 of 17 countries, with Canada, Ireland, Portugal, and Turkey the exceptions. Taking developing countries together, there is a more even split between countries

FEMINIZATION, DEFEMINIZATION, AND STRUCTURAL CHANGE IN MANUFACTURING

experiencing feminization and defeminization of manufacturing employment, with negative values for eight of 19 countries. Much of the literature on gender and structural change focuses on the late-developing, export-oriented countries of East Asia, in particular the so-called Asian Tigers. For the three countries in East Asia that experienced overall declines in the female share of manufacturing employment—Malaysia, South Korea, and Taiwan—the declines over spans of 15– 20 years were strikingly large. In Malaysia, the female share of manufacturing employment declined from 51% to 39% between the early-1990s and 2007; in South Korea, the decline was from 45% to 30% between the early-1980s and 2002; in Taiwan, the decline was from 49% to 41% between the early-1980s and 2001. Each industry group’s sum contribution (the sum of withinindustry, reallocation, and interaction effects) to changes in female shares of manufacturing employment is shown in columns (3) through (5), with shading used to indicate the largest absolute contribution. For the majority of countries, 25 of 36, labor-intensive manufacturing industries contributed most to changes in the female share. This compares with seven and five countries for intermediate and capital-intensive manufacturing industries, respectively (counting equal contributions from these two industry groups for Canada). For developed countries, a notable pattern is that for the three countries experiencing the feminization of manufacturing employment—Canada, Ireland, and Portugal—this was driven by positive contributions from intermediate and capital-intensive manufacturing industries. Conversely, for the seven developed countries experiencing defeminization, this was driven by negative contributions from labor-intensive manufacturing industries. This pattern is unsurprising in light of the strong regional shift in labor-intensive manufacturing exports from developed to developing countries, particularly in East Asia, suggestive of the limited prospects for workers in developed countries in these industries. The decisive role of labor-intensive manufacturing industries holds for other countries as well. Most notably, for eight of nine countries in East and South Asia, patterns of feminization and defeminization were predominantly driven by labor-intensive manufacturing industries. The exception is the Philippines, where the rising female share of manufacturing employment was predominantly driven by intermediate manufacturing industries. As for whether within-industry or employment reallocation effects were more important in accounting for changes in the female share of overall manufacturing employment, the answer is clear and holds for all regions. This is shown by comparing columns (9) and (13), with shading used to indicate the larger absolute value. For 33 of the 36 countries in our sample, within-industry effects were stronger than employment reallocation effects, with Cyprus, Ethiopia, and Turkey the exceptions. Taking these findings together suggests that changes in female shares of employment in the manufacturing sector as a whole were often driven by changes in the female share of employment within the group of labor-intensive manufacturing industries. Note that Table 2 shows industry-group breakdowns for within-industry effects in columns (6) to (8) and employment reallocation effects in columns (10) to (12). Another way of putting this, then, is that column (6) often carries the most weight of these six columns in accounting for changes in female shares of manufacturing employment. Indeed, for 18 of the 36 countries in our sample, column (6) is largest of the six columns in absolute value (and matches column (2) in sign). Among these 18 countries are Malaysia, South Korea, and Taiwan, which experienced the defeminiza-

573

tion of manufacturing employment, but also Morocco, which experienced the feminization of manufacturing employment, all countries to which we will return at greater length. Changes within labor-intensive industries are followed in prevalence by changes within intermediate industries, shown in column (7), which is largest in absolute value of the six columns for nine of the remaining countries in our sample (and matches column (2) in sign). 18 5. DECOMPOSITION RESULTS: LABOR-INTENSIVE MANUFACTURING INDUSTRIES In our prior analysis, we observed that the group of laborintensive manufacturing industries contributed most to changes in female shares of total manufacturing employment for 25 of 36 countries. To dig more deeply into this finding, we apply the same decomposition methods to evaluate changes in female shares within the group of labor-intensive manufacturing industries. We do this for five of the abovenoted 25 countries of particular interest: South Korea, Taiwan, and Malaysia, early export-oriented industrializers which experienced the strong defeminization of manufacturing employment; and Morocco and Turkey, later export-oriented industrializers which experienced the feminization of manufacturing employment. 19 It is perhaps worth making explicit that we focus on this handful of countries not because they are representative but rather because we believe they merit further consideration as a way of more fully interpreting our results, particularly in light of the literature on the possible gender implications of export-oriented development paths. These results are shown in Table 3 based on data at the two-digit ISIC level. 20 Our discussion of these results will show that the textile and apparel industries acted as the common driver of both the defeminization and feminization of employment in the group of labour-manufacturing industries in South Korea, Taiwan, Malaysia, Morocco, and Turkey, largely through the withinindustry effect for the first four countries and the reallocation effect for Turkey. Consequently, we also provide short profiles of key developments in these industries for these five countries, particularly regarding the relationship between technological upgrading and defeminization as well as industry expansion and feminization. 21 A summary of these industry profiles is provided in Table 4. Here it is also useful to refer to female shares of employment in these industries, shown in Figure 2 for these five countries. (a) Defeminizing countries For South Korea, Taiwan and Malaysia at the level of the three industry groups, we noted that the most important driver of the declining female share of total manufacturing employment was the declining female share within the group of labor-intensive industries (Table 2). In South Korea and Taiwan at the two-digit level for labor-intensive industries, the within-industry effect is weaker, inevitable in the context of any substantial employment reallocation within this group of industries, yet remains somewhat stronger than the reallocation effect (53% and 47% respectively for South Korea, and 53% and 46% respectively for Taiwan (Table 3, column 7)). The patterns for South Korea and Taiwan are strikingly similar. For both countries, the most important driver of the declining female share of employment for the group of labor-intensive industries was the declining female share of employment within the textile industry. But also important

574

WORLD DEVELOPMENT Table 2. Decomposition of changes in female shares of total manufacturing employment by three industry groups (%)

Notes: Shading indicates positive value for col. (2); largest industry-group contribution to changes in female shares for cols. (3) to (5), and larger absolute magnitude for cols. (9) and (13). When period averages for changes in female shares of manufacturing employment are small, % can be greater than 100.

was the reallocation of employment toward the disproportionately male fabricated metal products industry and away from the disproportionately female apparel and leather goods industry. South Korea’s export competitiveness in the textile and apparel industries has been argued to result from a conducive industrial policy along with a large gender wage gap (Amsden 1991; Jomo, 2009; Seguino, 1997). South Korea and Taiwan both shifted their trajectories of export-oriented industrialization toward more capital-intensive production and higher value-added exports in the 1980s, in particular from apparel toward electrical and electronics products, machinery and transport equipment (Berik, 2000; Caraway, 2006). During this period, apparel producers began to relocate from first-tier NICs including South Korea and Taiwan to China and second-tier NICs, particularly Indonesia, Malaysia, and Thailand (Kim, 2000). A number of factors made South Korea and Taiwan less attractive for apparel production: wage increases, currency appreciation, and the withdrawal of the Generalized

System of Preferences (GSP) from the first-tier NICs under the Generalized Agreement on Tariffs and Trade (GATT) (Gereffi & Memedovic, 2003; Jomo, 2009). Apparel exports as a share of total exports in South Korea declined from 17% to 3% and in Taiwan from 12% to 2% between 1980 and 2000 (Gereffi & Memedovic, 2003). At the same time, textile and apparel industry firms in both countries upgraded their position in global production networks and began to manufacture higher value-added products. South Korean apparel firms shifted from low value-added assembly to full-package production and original brand manufacturing by creating global sourcing networks, leveraging business ties with US buyers, and “triangle manufacturing” (that is, by positioning themselves as intermediaries between US buyers and producers in lower cost countries) (Gereffi & Memedovic, 2003). South Korea’s ongoing success in textile exports also attests to its ability to continuously upgrade and invest in cutting-edge technologies in order to maintain international competitiveness (Keane & Velde 2008; Singleton, 1997).

FEMINIZATION, DEFEMINIZATION, AND STRUCTURAL CHANGE IN MANUFACTURING

575

Table 3. Decomposition of changes in female shares of employment for labor-intensive manufacturing industries, 2-digit ISIC (Rev. 3) (%) (1) Country Sum contribution Malaysia South Korea Taiwan (China) Morocco Turkey Within-industry effect Malaysia South Korea Taiwan (China) Morocco Turkey Reallocation effect Malaysia South Korea Taiwan (China) Morocco Turkey

(2) Textiles (ISIC 17)

(3) Wearing apparel & Leather goods (ISIC 18 & 19)

(4) Wood products (ISIC 20)

(5) Fabricated metal products (ISIC 28)

(6) Furniture (ISIC 36)

(7) Total

21.0 39.7 48.1 42.3 154.1

16.0 30.6 25.7 57.0 507.6

24.9 1.1 2.7 3.2 69.9

25.2 19.5 14.6 3.3 149.5

45.0 11.3 14.3 0.6 173.9

100.0 100.0 100.0 100.0 100.0

21.4 33.6 39.0 38.4 104.9

49.1 8.0 12.3 44.2 18.6

13.0 1.2 0.1 1.3 8.1

4.7 1.6 10.4 1.7 12.2

30.0 12.1 12.2 1.7 1.6

92.1 53.4 53.3 84.6 88.7

0.2 5.0 8.1 3.6 64.4

66.0 22.8 13.1 15.6 534.1

42.6 2.3 3.0 1.6 58.9

23.1 22.2 26.0 1.4 171.2

17.1 1.1 1.9 1.3 175.6

17.0 46.6 46.1 17.7 181.8

Notes: See Table 1 for full industry headings. When period averages of changes in female shares of labor-intensive manufacturing employment are small, % can be greater than 100.

Table 4. Summary of textile and apparel industry profiles Textiles

Apparel

Defeminizing countries Malaysia Negative within-industry effect, consistent with preference for men workers in context of technological upgrading

South Korea

Taiwan (China)

Strong negative within-industry effect, consistent with preference for men workers in context of technological upgrading Strong negative within-industry effect, consistent with preference for men workers in context of technological upgrading

Negative within-industry and reallocation effects, consistent with preference for men workers in the context of technological upgrading combined with a shift of production to lower cost countries

Feminizing countries Morocco Strong positive within-industry effect, consistent with employment of women workers as part of a cost-cutting flexible employment strategy Turkey

Strong negative within-industry effect and strong positive reallocation effect, consistent with preference for men workers in context of technological upgrading combined with rapid growth of total employment Negative reallocation effect, consistent with shift of production to lower cost countries

Negative within-industry effect, consistent with preference for men workers in context of technological upgrading

Strong positive within-industry effect and positive reallocation effect, consistent with employment of women workers as part of a cost-cutting flexible employment strategy combined with growth of total employment Strong positive reallocation effect, consistent with rapid growth of total employment

Notes: “Strong” is used to describe a reallocation or within-industry effect greater than 30%; effects less than 10% are not addressed. Turkey is treated differently in this regard, given that percentages shown in Table 3 are inflated as a result of the period average of changes in female shares of labor-intensive manufacturing employment being small.

The textile industry in Taiwan also witnessed rapid restructuring in the last two decades. Many firms relocated to China while others not able to innovate and compete closed down, reflected in the over 40% drop in both textile exports and the number of textile companies from the late-1990s to present (Yang, 2012). Yet the Taiwanese textile producers that remained managed to upgrade and found a niche in the high-tech and higher value-added textile market. Taiwanese textile companies own a substantial number of patents for new materials, and the unit price of Taiwanese fabric exports rose markedly after the late-1990s, reflecting the shift to higher value-added products (Yang, 2012).

This upgrading may be related to the defeminization observed in the textile and apparel industries in South Korea and Taiwan, insofar as higher technology production leads to a stronger preference for male workers. The literature suggests several possible reasons, as noted in the introduction to this paper. Looked at in these ways, women’s manufacturing employment in South Korea declined as a result both of offshoring of production to lower cost countries, particularly for apparel, and of technological upgrading, particularly for textiles. This interpretation is in line with the results of our decomposition analysis for South Korea, showing stronger negative reallocation than within-industry effects for apparel

Malaysia

South Korea

Morocco

Turkey

1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008

WORLD DEVELOPMENT

1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008

576

Taiwan

1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008

Percentage

100 90 80 70 60 50 40 30 20 10 0

100 90 80 70 60 50 40 30 20 10 0

Year

Textiles

Apparel/ Leather Goods

Figure 2. Female share of employment in textiles & apparel/leather goods (%).

and stronger negative within-industry than reallocation effects for textiles. In Taiwan, the share of female employment declined in apparel and even more so in textiles, possibly for similar reasons regarding the stronger preference for men workers in the context of upgrading. Consistent with this narrative are our results showing similar negative contributions from reallocation and within-industry effects for apparel and a particularly large negative within-industry effect for textiles. For Malaysia at the two-digit level for labor-intensive industries, the within-industry effect remains much stronger than the reallocation effect (92% and 17%, respectively (Table 3, column 7)). The within-industry effect was driven by declining female shares in the apparel and leather goods, furniture, and textiles industries, in order of importance. Though the total reallocation effect within the group of labor-intensive industries is weaker, underlying this are large positive and negative effects that offset each other. These result from the reallocation of employment toward the disproportionately female apparel and leather goods industry (contributing positively to changes in female shares within the group of labor-intensive industries) which are offset most importantly by the reallocation of employment toward the disproportionately male wood products and fabricated metal products industries (contributing negatively to changes in female shares). Malaysia was one of the Southeast Asian countries that received large FDI inflows in the textile and apparel industries, as production shifted from first-tier NICs like South Korea and Taiwan. These industries grew strongly and became significant contributors to value-added, exports and employment between the late-1960s and 1990s (Ong, 1987; Rasiah, 2009). As wages increased in the 1990s and the end of the Multi-Fibre Arrangement approached, firms began to automate and adopt new-process technologies (Rasiah, 2009). Malaysia was able to consolidate its position as one of the world’s top 25 apparel exporters, with apparel exports as a share of total exports increasing from 1% to 2% between 1980 and 2000 (Gereffi & Memedovic, 2003). In the results of our decomposition analysis for Malaysia, the contribution of the apparel industry was positive in spite

of a strong negative within-industry effect that was more than offset by a stronger yet positive reallocation effect. This reflects the overall doubling of total employment in the apparel industry from 1983 to 2007, whereas total employment in the textile industry held steady. 22 The overall contribution of textiles to the female share of manufacturing employment was negative, the result of a negative within-industry effect on top of a reallocation effect that was effectively zero. As in South Korea and Taiwan, upgrading in apparel and textiles in Malaysia was associated with declining female shares of employment, perhaps due to the stronger preference for men workers in more technologically advanced production. (b) Feminizing countries Of the two countries considered that experienced the feminization of manufacturing employment, results for Morocco also provide evidence for the importance of changes in female shares of employment within the textile and apparel and leather goods industries. We noted that for Morocco at the level of the three industry groups, the most important driver of the increasing female share of total manufacturing employment was the increasing female share within labor-intensive industries (Table 2). At the two-digit level for the group of labor-intensive industries, the within-industry effect is much stronger than the reallocation effect (85% and 17% respectively (Table 3, column 7)), driven by increasing female shares in apparel and leather goods and textiles. Morocco undertook an export-oriented growth strategy in the 1980s, and apparel exports as a share of merchandise exports increased from 5% to 32% between 1980 and the early-2000s (Plank, Rossi, & Staritz, 2012). The textile and wearing apparel industries taken together are the fourth largest foreign exchange earner in the country (AEO, 2012). Both industries rely heavily on imported intermediate inputs, largely the result of outward processing trade and special trade agreements that allowed EU firms to assemble goods in partner countries by exporting inputs and re-importing the finished product under preferential conditions (Nordas, 2004; Plank et al., 2012).

FEMINIZATION, DEFEMINIZATION, AND STRUCTURAL CHANGE IN MANUFACTURING

Morocco has become a key supplier to the fast fashion market in the European Union (EU), with 85% of its apparel exports and 67% of its textile exports shipped to the EU-15 as of 2010 (Plank et al., 2012). The fast fashion market thrives on making the latest trends available quickly to customers and requires flexible yet reliable low-cost suppliers. Because buyers maintain thin inventories and lead times are short, much market risk is borne by suppliers. Suppliers employ two categories of workers to deal with this risk: skilled, regular workers who manage the quality requirements of orders; and unskilled, irregular workers who are engaged for more variable operations like packing and loading (Barrientos, Gereffi, & Rossi, 2011; Rossi, 2011). The latter are typically young women who are paid less than the minimum wage and lack formal contracts and social protection (Barrientos et al., 2011). This evidence suggests that the rising share of female employment in the apparel and textile industries reflects the strategy of Moroccan exporters of coping with the risk and cost pressures of supplying the fast fashion market in the EU. This reading accords with the results of our decomposition analysis, showing that while the reallocation as well as within-industry effects for textiles and apparel were positive, the within-industry effects were much larger. For Turkey, a strong negative within-industry effect for labor-intensive industries is offset by a stronger yet positive reallocation effect (89% and 182% respectively (Table 3, column 7)). The positive reallocation effect was driven by the apparel and leather goods industry while the negative within-industry effect was driven by textiles. Like Morocco, Turkey adopted an export-oriented growth model in the 1980s and saw a dramatic rise in apparel and textile exports,

577

becoming by 2005 the world’s second largest apparel exporter (after China) (Tokatli & Kizilgu¨n, 2009). Turkey was also similar to Morocco in that a large of its textile and apparel exports went to the EU, around 65% as of 2003 (Evgeniev, 2006). Turkey’s strong industrial base, a favorable exchange rate, and preferential access to the EU through a customs union as of 1996 were factors in this success (Tokatli & Kizilgu¨n, 2009). The apparel industry evolved into a network of small- and medium-sized establishments during the 1980s and continues to rely extensively on informal workers (Evgeniev, 2006). Turkey is argued to have maintained its success in apparel due to its ability to respond to short lead times in the context of the fast fashion market, relying on women to provide a flexible workforce (Evgeniev, 2006; Tokatli & Kizilgu¨n, 2009). At the same time, Turkey upgraded its position in the apparel value chain from full-package production to original design manufacture by leveraging its production abilities and networks with branded retailers as well as by promoting Istanbul as a fashion destination (Fernandez-Stark, Frederick, & Gereffi, 2011). The textile industry in Turkey has shifted since the 1980s away from a network of small establishments toward larger firms using more capital-intensive methods. Upgrading in the industry is likewise reflected in the growing share of workers in registered establishments paying higher wages than in the apparel industry (Evgeniev, 2006). In the results of our decomposition analysis for Turkey, we observe a large positive reallocation effect for apparel, reflecting the large increase in total employment in the industry. 23 In this sense, Turkey resembles Malaysia. As with South Korea, Taiwan, and Malaysia, the contribution of textiles to the

Table 5. Employment in services and agriculture 1980

1990

2000

2008

Female share of total employment (%) Malaysia 33.5 South Korea 38.2 Moroccoa NA Turkey NA

35.5 40.8 NA 31.5

34.7 41.4 25.2 26.9

35.7 41.9 26.8 26.4

Female share of employment in services (%) Malaysia 31.4 South Korea 38.4 Morocco NA Turkey NA

35.7 43.3 NA 14.0

40.0 46.4 16.9 17.7

42.5 47.7 18.2 22.5

Share of total employment in services (%) Malaysia 38.7 South Korea 37.0 Morocco NA Turkey NA

46.5 46.7 NA 32.4

49.5 61.2 35.5 40.0

57.4 67.9 37.2 49.5

Female share of employment in agriculture (%) Malaysia 39.4 South Korea 43.7 Morocco NA Turkey NA

34.5 46.2 NA 50.9

26.5 47.8 32.4 45.2

23.0 46.4 38.9 46.9

Share of total employment in agriculture (%) Malaysia 37.2 South Korea 34.0 Morocco NA Turkey NA

26.0 17.9 NA 46.9

18.4 10.6 44.4 36.0

14.0 7.2 40.9 23.7

Notes: NA indicates that data are not available. While employment data are available for earlier years for Morocco and Turkey, these only cover urban areas. Source: ILO (2013). a Data for Morocco are for 2002 rather than 2000.

578

WORLD DEVELOPMENT

female share of manufacturing employment in Turkey was negative, driven more by the within-industry than reallocation effect, and—consistent with the literature—we suggest this may result from a stronger preference for men workers in the context of industrial upgrading. (c) Employment in services and agriculture Our discussion has so far addressed only the manufacturing sector. Yet structural change in the process of economic development is also typically characterized by employment contraction in agriculture and expansion in services. Women may have a preference for or an advantage in services compared to manufacturing, and it could be hypothesized on these grounds that the defeminization of manufacturing employment in Malaysia and South Korea is the result of a shift of women’s employment to services rather than dynamics within manufacturing. 24 In Table 5, we present female shares of employment in services and agriculture as well as for the economy as a whole along with shares of total employment in services and agriculture—for Malaysia and South Korea as well as for Morocco and Turkey—to provide a preliminary sense of this hypothesis as well as to provide a broader context for developments in manufacturing (ILO, 2013). 25 The female share of employment in services was higher than the female share of total employment in Malaysia and South Korea but lower in Morocco and Turkey, indicating that the service sector was disproportionately female only in the former two countries. More relevantly, both the share of total employment in services and the female share of employment in services increased in all four countries (though less strongly in Morocco, for which data go back only to 2002). At least based on this cursory evidence, a shift in women’s employment to the service sector would not obviously account for defeminization in Malaysia and South Korea, given that employment trends in services were broadly similar in Morocco and Turkey, which experienced the feminization of manufacturing employment. As for agriculture, the share of total employment in agriculture was considerably higher in Morocco and Turkey than in Malaysia and South Korea, and agriculture was also more disproportionately female in the former two countries, as indicated by the higher ratios of the female share of employment in agriculture to the female share of total employment. 26 While all four countries experienced declining shares of total employment in agriculture, a much larger absolute number of female workers were released from agriculture in Turkey than in Malaysia and South Korea. In Morocco, however, a countervailing factor was the increase in the share of female employment in agriculture, such that the absolute number of female workers in agriculture increased from 2002 to 2008. 27 For Turkey, in any case, to the extent that this pool of female workers found jobs in manufacturing, developments in agriculture also come into play in accounting for the feminization of manufacturing employment. Though beyond the scope of this paper, developing a clearer sense of how developments in other sectors influences the feminization and defeminization of manufacturing employment is a worthwhile area for future research, for which it would be particularly insightful to track movements of male and female workers among the various sectors of the economy in the process of economic development. 6. CONCLUDING REMARKS This study addresses the gendered employment impacts of structural change in manufacturing for 36 countries at

different levels of development. It differs from prior studies of structural change in applying accounting decomposition methods to identify drivers of feminization and defeminization rather than of productivity growth. As such, it aims to contribute to a better understanding of the role of gender in processes of structural change, within manufacturing if not more broadly. Conversely, our study differs from prior studies of gender and industrialization in systematically accounting for the relative importance of within-industry versus employment reallocation effects on changes in the female share of manufacturing employment. The study distinguishes between these effects for the manufacturing sector as a whole as well as within a group of labor-intensive manufacturing industries for selected feminizing and defeminizing countries. Though these effects have been discussed in the literature on gender and industrialization, the application of accounting decomposition methods enables a more definitive sense of their relative importance and of the industries driving feminization and defeminization. For the majority of countries in our sample, feminizing and defeminizing, we find that the group of labor-intensive industries contributed most to changes in the female share of total manufacturing employment. With but a few exceptions, we also find the greater importance of within-industry than employment reallocation effects for the manufacturing sector as a whole, most commonly resulting from changes within the group of labor-intensive industries. Consistent with these findings, there is only a weak correlation between the female-intensity of employment and the labor-intensity of production within the countries in our sample. Though structural change in the process of economic development is indeed characterized by shifts toward less labor-intensive, higher valueadded manufacturing industries, it is developments within industries that generally matter more in accounting for patterns of feminization and defeminization for the manufacturing sector as a whole. As noted, this finding is analogous to that of studies applying accounting decomposition methods to identify drivers of productivity growth (e.g., Ocampo et al., 2009; Roncolato & Kucera, 2014). The paper explores the defeminization of manufacturing employment in greater depth for South Korea, Taiwan, and Malaysia. A key factor in accounting for defeminization in these countries is the declining female share of employment within the group of labor-intensive manufacturing industries and particularly within the textile industry in South Korea and Taiwan and the apparel industry in Malaysia. In these three cases, the declining female share within these industries has also been associated with substantial technological upgrading. Consistent with the literature addressed in the introduction, we have suggested that this may result from employers’ preference for male workers in more technologically sophisticated jobs. Regarding the defeminization of manufacturing employment in the context of technological upgrading, Barrientos et al. (2004) write, “What explains these trends and what happens to the women workers who lose or fail to find jobs is obviously of considerable policy interest” (p. 5). To give an example, knowing the extent to which defeminization results from employment reallocation effects versus withinindustry effects—and within which specific industries—may usefully guide training programs aiming to better integrate women into manufacturing employment in the context of technological upgrading. At the same time, there is an acknowledgment in the literature that the causes of defeminization occurring alongside technological upgrading are not well understood, but ought to be given the policy issues at

FEMINIZATION, DEFEMINIZATION, AND STRUCTURAL CHANGE IN MANUFACTURING

stake. These issues thus merit further inquiry, for which it is important to have accurate assessments of the comparative skills of male and female workers. For these purposes, it is insufficient to rely on such commonly used indicators as years of educational attainment given that these provide poor measures of skills, particularly problematic for developing countries (Hanushek & Woessmann, 2008). Rather it would be useful to have direct assessments of workers’ skills, perhaps following a testing approach similar to the Program for International Student Assessment (PISA) regarding students’ skills. 28 A complementary way forward is suggested by Ong (1987), who undertook firm-level surveys in her study of Japanese multinationals producing electronics in Malaysian EPZs. Ong asked managers and engineers in these enterprises why the vast majority of employees engaged in semi-skilled assembly operations were female, and their responses boil down to one thing: For the type of work they were doing, women were more productive than men. In her study of Mexican EPZs, Salzinger refers to the “trope of

579

productive femininity,” which suggests that such stated views on women’s greater productivity in semi-skilled assembly operations not be taken too literally (2003, p. 10). Caraway elaborates the point, arguing that “whether women possess these traits is irrelevant; the crucial point is that employers believe that they do. As elegantly stated by Salzinger, these images are important not because they reflect reality but because they produce it” (2007, p. 30). It would be curious, though, if in the face of technological upgrading, the “trope of productive femininity” were somehow supplanted by the trope of unproductive femininity. Rather than answer the question of why technological upgrading is associated with defeminization, this provides another way of looking at the issue while raising yet other questions. Yet insofar as tropes are acted upon, not least to make hiring decisions, they have a more than figurative reality. This applies not just to skill differences of men and women workers, real and perceived, but also to gender norms designating men as breadwinners.

NOTES 1. See Standing (1989, 1999) for the application of a more wide-ranging definition of feminization. 2. In her study on Indonesia, however, Caraway does refer to “sectoral feminization,” by which she means changing female shares within industries (2007, p. 12). 3. For other examples, see Barrientos, Kabeer, and Hossain (2004) and Elson and Pearson (1981). 4. Caraway (2007) is a partial exception in that she identifies the sectoral drivers of feminization and defeminization in Indonesia but does not distinguish between within-industry and reallocation effects. Olivetti and Petrongolo (2011) are also worth noting in that they apply decomposition analysis to distinguish within-industry from between-industry components of gender bias in labor demand. However, while some of their decompositions are for done for different points in time, the decompositions are of differences between countries rather than changes over time and moreover do not address developing countries. 5. Useful studies of gender and structural change include Seguino and Braunstein (2012), Wamboye and Seguino (2012), Gaddis and Klasen (2013) and Rendall (2013), all of which differ from our paper in addressing gender differences at the aggregate level as well as in using econometric rather than decomposition analysis. More specifically, for countries in Latin America, Seguino and Braunstein (2012) address gender differences in employment using as measures of structural change the ratio of manufacturing exports to imports, fuel and ores as a share of merchandise exports, and the terms of trade. For countries in SubSaharan Africa, Wamboye and Seguino (2012) also address gender differences in employment using trade-related measures, as well as manufacturing and agricultural value-added as shares of GDP. For a large sample of countries, Gaddis and Klasen (2013) address gender differences in female labor force participation (FLFP) and in particular the “U”-shaped pattern of FLFP in the process of economic development, using data on sectoral value-added and employment to measure structural change. For Brazil, Mexico, India, and Thailand, Rendall (2013) addresses gender differences in wages, employment, and education using data on sectoral and occupational employment to measure structural change.

6. See Joekes (1999) and Barrientos et al. (2004) on Malaysia, Seguino (1997), Joekes (1999), and Murayama and Yokota (2008) on South Korea, and Berik (2000) and Barrientos et al. (2004) on Taiwan, as well as Papyrakis, Covarrubias, and Verschoor (2012) for a survey of the literature on gender and export-oriented development more generally. 7. As Rossi (2013) argues in her study of the apparel industry in Morocco, this may depend on whether economic upgrading takes the form of process, production, functional or chain upgrading, based on typologies developed by Gereffi (1999) and Humphrey and Schmitz (2000). 8. See Joekes (1999) regarding the machinery and automobile industries in Mexico as well as export processing zones (EPZs) in Malaysia, the Philippines, and South Korea; Murayama and Yokota (2008) regarding electronics and electrical goods industries in EPZs in South Korea; and Barrientos et al. (2004) regarding EPZs in Taiwan, Malaysia, and Mexico. 9. A similar sense of doubt on this very question is conveyed by Sundaram (2009), who writes that “In the absence of conclusive evidence, it is difficult to be certain as to why employers exhibited a ‘preference’ for male workers. . .” (p. 46). 10. Leaving aside recycling. 11. See Kucera and Sarna (2006) for details on the construction of the three industry groups. 12. See Cutler, Berri, and Ozawa (2003) for an overview of previouslyused classifications of labor-intensive manufacturing industries. 13. Not including Kazakhstan, given the unavailability of value-added data. Note that Caraway (2006, 2007) uses data from the UNIDO Industrial Statistics Database for the 1981–95 period and for a somewhat larger sample of countries and runs an OLS regression with the natural logarithm of the female percentage of employment as a dependent variable and the natural logarithm of the ratio of value-added to employment (inverse to the ratio used for our correlation coefficients) and country dummies as independent variables. She reports a coefficient estimate on the ratio of value-added to employment of 0.24, statistically significant at the 1-% level.

580

WORLD DEVELOPMENT

14. For the 36 countries on average, the female shares for the labor-intensive, intermediate, and capital-intensive industry groups are 42%, 26%, and 28%, respectively, based on annual averages for all years of data after 1980. 15. Not including Kazakhstan for the labor-intensity of production, given the unavailability of value-added data. For female shares of employment, the average correlation coefficient is 0.75 with or without Kazakhstan. 16. The method follows Syrquin (1986), Pieper (2000), and Ocampo et al. (2009) in their decompositions of labor productivity growth. 17. Note that years used for the decomposition analysis differ somewhat from the years shown in Figure 1, since there are problematic discontinuities in the data resulting from changes in industry definitions and survey types as well as other observed jumps, as described in the Appendix: Data notes. 18. For two countries in our sample, we also did the decomposition analysis for a restricted set of years, to address questionable volatility in the data, as follows: India, leaving out 1993–98; and Malta, leaving out changes between 1998–99 and 2004–05. For India, this made no substantive difference to results. For Malta, labor-intensive industries remained the biggest sum contributor to changes in the female share of manufacturing employment, but this was driven more by employment reallocation than within-industry effects and, more generally, employment reallocation effects were larger than within-industry effects. 19. In Morocco, the female share of manufacturing employment increased from 30% to 48% between the early-1990s and 2008; in Turkey, the increase was from 20% to 22% between 1992 and 2001. 20. Note that some two-digit industries are combined in Table 3, to facilitate comparability among countries and over time.

21. We leave aside the leather goods industry in these profiles, given the focus of the literature on the textile and apparel industries. 22. According to the UNIDO (2011a) data for Malaysia, employment in the apparel industry totaled about 30,000 in 1983 and 72,000 in 2007, and in the textile industry about 33,000 and 35,000, respectively, in those years. It is worth remarking, though, that more than one-third of firms in textile, apparel, and electronics industries in Malaysia subcontract to homeworkers (Ghosh, 2002; Sim, 2009). Homeworkers are not included in the UNIDO database used for our decomposition analysis, and results must be interpreted with caution in this regard. 23. According to the UNIDO (2011a) data for Turkey, employment in the apparel industry totaled about 93,000 in 1992 and 153,000 in 2001 and in the textile industry about 181,000 and 220,000, respectively, in those years. 24. We thank an anonymous reviewer for suggesting this line of inquiry. 25. Data with gender breakdowns are not available for Taiwan in ILO (2013). 26. For 2008, for example, the ratios were 0.6 for Malaysia, 1.1 for South Korea, 1.4 for Morocco, and 1.8 for Turkey. 27. From 1990 to 2008, the number of female workers in agriculture declined from 600,000 to 343,000 in Malaysia, from 1,495,000 to 783,000 in South Korea, and from 4,760,000 to 2,354,000 in Turkey; in contrast, the number of female workers increased from 1,363,000 to 1,621,000 in Morocco from 2002 to 2008 (ILO, 2013). 28. For further information, see: http://www.oecd.org/pisa/aboutpisa/.

REFERENCES AEO, 2012. (African Economic Outlook) (2012). Morocco. Abidjan: African Development Bank; Paris: OECD; New York: United Nations Development Programme; Addis Ababa: United Nations Economic Commission for Africa. Amsden, A. (1991). Diffusion of development: The late-industrializing model and greater East Asia. The American Economic Review 81(2). Papers and Proceedings of the Hundred and Third Annual Meeting of the American Economic Association (pp. 282–286). Barrientos, S., Gereffi, G., & Rossi, A. (2011). Economic and social upgrading in global production networks: A new paradigm for a changing world. International Labor Review, 150(3–4), 319–340. Barrientos, S., Kabeer, N., & Hossain, N. (2004). The gender dimensions of the globalization of production. Policy Integration Department working paper 17. Geneva: International Labor Office. Berik, G. (2000). Mature export-led growth and gender wage inequality in Taiwan. Feminist Economics, 6(3), 1–26. Caraway, T. L. (2006). Gendered paths of industrialization: A crossregional comparative analysis. Studies in Comparative International Development, 41(1), 26–52. Caraway, T. L. (2007). Assembling women: The feminization of global manufacturing. Ithaca: ILR Press. Chandler, A. (1977). The visible hand: The managerial revolution in American business. Cambridge, MA: Harvard University Press. Cutler, H., Berri, D., & Ozawa, T. (2003). Market recycling in laborintensive goods, flying-geese style: An empirical analysis of East Asian exports to the US. Journal of Asian Economics, 14(1), 35–50. Elson, D., & Pearson, R. (1981). “Nimble fingers make cheap workers”: An analysis of women’s employment in third world export manufacturing. Feminist Review, 7(Spring), 87–107. Evgeniev, E. (2006). Industrial and firm upgrading in the European periphery: The textile and apparel industry in Turkey and Bulgaria (Doctoral dissertation). Budapest: Political Science Department of the Central European University.

Fernandez-Stark, K., Frederick, S., & Gereffi, G. (2011). Skills for upgrading: Workforce development and global value chains in developing countries. Durham, NC: Duke University Center on Globalization, Governance and Competitiveness. Gaddis, I., & Klasen, S. (2013). Economic development, structural change and women’s labor force participation. A reexamination of the feminization U hypothesis. Department of Economics working paper 02/2103. New York: New School for Social Research. Gereffi, G. (1999). International trade and industrial upgrading in the apparel commodity chain. Journal of International Economics, 48, 37–70. Gereffi, G., & Memedovic, O. (2003). The global apparel value chain: What prospects for upgrading by developing countries?. Vienna: United Nations Industrial Development Organization. Ghosh, J. (2002). Globalisation, export-oriented employment for women and social policy: A case study of India. Social Scientist, 30(11/12), 17–60. Hanushek, E., & Woessmann, L. (2008). The role of cognitive skills in economic development. Journal of Economic Literature, 46(3), 607–668. Humphrey, J., & Schmitz, H. (2000). Governance and upgrading: Linking industrial cluster and global value chain research. IDS working paper series 120. Brighton: Institute of Development Studies. ILO (2013). Key indicators of the labor market. Geneva: International Labor Organization. Joekes, S. (1999). A Gender-analytical perspective on trade and sustainable development. In Trade, gender and sustainable development (pp. 33–59). Papers prepared in support of the themes discussed at the PreUNCTAD X Expert Workshop on Trade, Sustainable Development and Gender. New York and Geneva: United Nations. Jomo, K. S. (2009). Export-oriented Industrialisation, Female Employment and Gender Wage Equity in East Asia. Economic and Political Weekly, 44(1), 41–49.

FEMINIZATION, DEFEMINIZATION, AND STRUCTURAL CHANGE IN MANUFACTURING Kaldor, N. (1968). Productivity and growth in manufacturing industry: A reply. Economica, 385–391 (November). Keane, J., & Velde, D. W. te. (2008). The role of textile and clothing industries in growth and development strategies. London: Overseas Development Institute. Kim, S. (2000). Effects of outward foreign direct investment on home country performance: Evidence from Korea. In T. Ito, & A. Krueger (Eds.), The role of foreign direct investment in East Asian economic development. Chicago: University of Chicago Press. Kucera, D., & Sarna, R. (2006). Trade union rights, democracy, and exports: A gravity model approach. Review of International Economics, 14(5), 859–882. Milberg, W., & Winkler, D. (2011). Economic and social upgrading in global production networks: Problems of theory and measurement. International Labor Review, 150(3–4), 341–365. Murayama, M., & Yokota, N. (2008). Revisiting labor and gender issues in export processing zones: The case of South Korea, Bangladesh and India. IDE discussion paper 174. Japan: Institute of Developing Economies. Nordas, H. K. (2004). The global textile and clothing industry post the agreement on textiles and clothing. World Trade Organization discussion paper 5. Geneva: World Trade Organization. Ocampo, J. A., Rada, C., & Taylor, L. (2009). Growth and policy in developing countries: A structuralist approach. New York: Columbia University Press. Olivetti, C., & Petrongolo, B. (2011). Gender gaps across countries and skills: Supply, demand and the industry structure. NBER working paper 17349. Cambridge, MA: National Bureau of Economic Research. Ong, A. (1987). Spirits of resistance and discipline: Factory women in Malaysia. Albany, NY: State University of New York Press. Papyrakis, E., Covarrubias, A., & Verschoor, A. (2012). Gender and trade aspects of labor markets. Journal of Development Studies, 48(1), 81–98. Pieper, U. (2000). Deindustrialization and the social and economic sustainability nexus in developing countries: Cross-country evidence on productivity and employment. Journal of Development Studies, 36(4), 66–99. Plank, L., Rossi, A., & Staritz, C. (2012). Workers and social upgrading in “fast fashion”: The case of the apparel industry in Morocco and Romania. Austrian Research Foundation for International Development working paper 33. Vienna: Austrian Research Foundation for International Development. Rasiah, R. (2009). Malaysia’s textile and garment firms at the crossroads. Journal of Contemporary Asia, 39(4), 530–542. Rendall, M. (2013). Structural change in developing countries: Has it decreased gender inequality?. World Development, 45, 1–16. Roncolato, L., & Kucera, D. (2014). Structural drivers of productivity and employment growth: A decomposition analysis for 81 countries. Cambridge Journal of Economics, 38(2), 399–424. Rossi, A. (2011). Economic and social upgrading in global production networks: The case of the garment industry in Morocco (Doctoral dissertation). Brighton, UK: Institute of Development Studies, University of Sussex. Rossi, A. (2013). Does economic upgrading lead to social upgrading in global production networks? Evidence from Morocco. World Development, 46, 223–233. Salzinger, L. (2003). Genders in production: Making workers in Mexico’s global factories. Berkeley: University of California Press. Seguino, S. (1997). Gender wage inequality and export-led growth in South Korea. Journal of Development Studies, 34(2), 102–132. Seguino, S. (2000). Gender inequality and economic growth: A crosscountry analysis. World Development, 28(7), 1211–1230. Seguino, S. (2005). Gender inequality in a globalizing world. In Paper presented at United Nations Research Institute for Social Development Conference, New York, March 7, 2005. Seguino, S., Berik, G., & Rodgers, Y. M. (2010). An investment that pays off: Promoting gender equality as a means to finance development. Berlin: Friedrich Ebert Stiftung International. Seguino, S., & Braunstein, E. (2012). The impact of economic policy and structural change on gender employment inequality in Latin America, 1990–2010. MPRA Paper 43261. Mu¨nich: Mu¨nich Personal RePEc Archive. Sim, H. C. (2009). Globalisation and its challenges to gender mainstreaming in Malaysia. Akademika, 86(1), 123–129. Singleton, J. (1997). The world textile industry. London: Routledge.

581

Standing, G. (1989). Global feminization through flexible labor. World Development, 17(7), 1077–1095. Standing, G. (1999). Global feminization through flexible labor: A theme revisited. World Development, 27(3), 583–602. Sundaram, J. K. (2009). Export-oriented industrialisation, female employment and gender wage equity in East Asia. Economic and Political Weekly (January 3). Syrquin, M. (1986). Productivity growth and factor reallocation. In H. Chenery, S. Robinson, & M. Syrquin (Eds.), Industrialization and growth. New York: Oxford University Press. Tejani, S., & Milberg, W. (2010). Global defeminization? Industrial upgrading, occupational segmentation and manufacturing employment in middle-income countries. SCEPA working paper 2010-1. New York: Schwartz Center for Economic Policy Analysis. Tejani, S. (2012). Global defeminization? Industrial upgrading, sex segmentation and manufacturing employment in developing countries. In Tejani, S., Three essays on employment and productivity growth (unpublished dissertation). New York: New School for Social Research. Tokatli, N., & Kizilgu¨n, O. (2009). From manufacturing garments for ready-to-wear to designing collections for fast fashion: Evidence from Turkey. Environment and Planning, 41(1), 146–162. UN (1999). World survey on the role of women in development: Globalization, gender and work. New York: United Nations. UNIDO (2011a). Industrial Statistics Database, 2-digit level of ISIC code (Revision 3). Vienna: United Nations Industrial Development Organization. UNIDO. (2011b). User’s guide, INDSTAT2 Industrial Statistics Database, 2-digit level of ISIC code (Revision 3). Vienna: United Nations Industrial Development Organization. Wamboye, E., & Seguino, S. (2012). Economic structure, trade openness, and gendered employment in sub-Saharan Africa. Processed. Yang, L. (2012). Textile makers in Taiwan create a high-tech niche. New York Times, B1 (December 26).

APPENDIX A. DATA NOTES 1. To emphasize longer run developments in female shares of manufacturing employment, only countries having 10 or more years of data on female and total employment for most industries in the UNIDO database (2011a) were included in the sample. After data cleaning, the number of years of data dropped to nine for Canada, Egypt, and Japan. 2. Raw data and female shares at the industry level were carefully screened for problematic discontinuities, particularly when there were changes in industry definitions and survey types or other observed jumps, and we did not evaluate changes over such problematic discontinuities. 3. The UNIDO database is mainly provided at the ISIC two-digit level, Revision 3, but older data are also provided based on earlier classification systems. Whenever possible we include these older data by combining them into one of the three industry groups classified by the labor-intensity of production. 4. The decomposition analysis is mainly based on annual changes in data, though for nine countries, we also evaluated changes over 2 years (that is, with one-year gaps in the data) and for two countries we also evaluated changes over 3 years (that is, with two-year gaps in the data), though only in cases for which there were no problematic discontinuities over the data gaps. 5. The years of data used in the decomposition analysis are listed below in the same order as in Table 2.

582

Austria Canada Cyprus Germany Ireland Japan Malta New Zealand Portugal United Kingdom Azerbaijan Bulgaria Croatia Kazakhstan Kyrgyzstan Lithuania Turkey Indonesia Macao (China)

WORLD DEVELOPMENT

1983–89, 1990–94 1981–89 1981–94, 1999–2007 1981–90 1988–90, 1991–95 1994–2002 1981–85, 1990–2008 1988–93, 1995–2001 1981–87, 1990–94 1981–92, 1993–2000 1997–2008 1993–2005 1990–2007 1998–2007 1998–2001, 2003–08 1996–99, 2003–08 1992–2001 1993–2007 1984–85, 1988–98, 2001–03

Malaysia Philippines South Korea Taiwan (China) Bangladesh India Sri Lanka Chile Panama Puerto Rico Egypt Iran Jordan Morocco Eritrea Ethiopia Kenya

Available online at www.sciencedirect.com

ScienceDirect

1983–94, 1995–2007 1992–95, 1996–2006 1982–89, 1991–2002 1981–2001 1981–88, 1990–92 1993–98, 1999–2007 1990–92, 1993–2001 1985–98, 2001–06 1982–85, 2001–05 1990–2000 1991–95, 2002–06 1991–93, 1994–2005 1985–93, 2000–08 1992–98, 2000–08 1993–2007 1991–2008 1987–94, 1996–98