GENDER INEQUALITY AND LABOUR MARKET SEGMENTATION UNDER TRADE REFORM:

PEP research project 11029 INTERIM REPORT This version 9 March 2008 GENDER INEQUALITY AND LABOUR MARKET SEGMENTATION UNDER TRADE REFORM: Evidence of the gender wage gap from Vietnam, 1993 2004 Pham Thi Thu Tra & Le Thai Thuong Quan Nguyen Thai Thao Vi Ngo Quang Thanh Abstract In this study, we examine the gendered wage effect of trade reform vis-à-vis the employment distribution of men and women in the formal and informal labour sector in Vietnam for the period 1993-2004, using a series of household data surveys up to date. We find that the gender wage gap in favour of men persists in both sectors of wage employment. Among other factors relevant for the wage determination, trade openness appears to be a significant stimulus to an increase in individual wages in the period studied, and this effect is slightly higher for women. Our decomposition result reveals that discrimination against women is more prominent in the formal sector, despite a more severe gender gap manifest in the informal sector. For both sectors, our result indicates a narrowing, albeit small, impact of trade on the gender wage gap. 1

1. Introduction In the episode of trade liberalization in developing countries, men and women tend to be influenced differently. The effect of trade liberalization on the gender wage gap and other gender discriminations however, is unclear, both theoretically and empirically (see Fontana (2003) for a comprehensive review). Conceptually, the gendered wage impact of trade is mediated though a number of channels: the changing employment opportunities in the labour market and the changing earnings position of women relative to men s. In this framework, empirical studies reveal evidence of an improved role of women with respect to both employment conditions and earnings (Pham and Reilly, 2007, Reilly and Dutta, 2006, Fleck, 2001, Standing 1999, Pearson, 1999 etc.). Alternatively, other studies report evidence of a widening gender wage gap under trade reform (Nicita and Razzar, 2003, Paul-Majumder and Begum, 2000). While empirics on the trade - gendered earnings linkage is abundant and mixed, it has been silent on the pervasiveness of the informal labour sector, which notably characterizes the labour market in developing countries. As a result, the linkage between the gendered earnings effect of trade and labour market segmentation appears to be under researched. Vietnam was one of the countries that embarked on significant economic reforms over the past decades. One of the key reforms is to initiate and expand trade liberalization. In the period 1990 2004, Vietnam witnessed an impressive growth of more than 7% in GDP and of approximately of 19.3% and 13% in export and import, respectively (see table A1). This outcome was a result of a more outward looking trade policies accompanied by various export promotion measures. Concerning tariff evolution, customs tariffs were first introduced in 1988, and the number of tariff lines and tariff rates increased (table A2). It is also documented that the average effective rate of protection (ERP) for the tradable sectors experienced a considerable decline while there has been a marginal reduction of the nominal rate of protection (NRP) in the 1992 2003 period (Pham, 2007). Although the tariff reductions were not significant by international standards, the substantial removal of non-tariff barriers marked a remarkable step in the process of trade liberalization in Vietnam. 2

Vietnam s labour market also underwent impressive changes in the period 1993-2004 As in other developing countries, the informal labour sector captures a dominant share of employment. This share has declined from 90% to 82.81% between the years 1009 and 2004. The active role of women in the overall labour market is widely acknowledged, given a rising participation rate of women in economic activities (see section 4 below). Women constituted 45.07% of total labour force in Vietnam in the year 1993 and this figure increased to 50.35% in the year 2004. This trend was accompanied by a rise in women s earnings relative to men over the period 1993 2002 (Pham and Reilly, 2007). Within the formal wage employment, earnings disparity seems to persist over the period and is largely attributable to the within-sector differences rather than between-sector differences (Liu, 2004a). Another remarkable observation is ascribed to the over presentation of women in the informal labour sector. In the period 1993-2004, in spite of a declining share of the informal sector, the participation rate of women in this sector remains as high as around 86% over the period. Despite certain benefits of the informal sector, it is evident for developing countries that the informal sector is often associated with lower quality jobs and thus lower earnings as compared to its formal counterpart (Goldberg and Pavnick, 2003). In Vietnam informal female workers are reported to earn less than their male counterpart, and this disparity seems to grow (see table 2). As a result, the incidence of the large informal labour sector is expected to provide certain gender implications to an analysis of gender wage inequality under trade reform. This issue has not, however, been empirically examined for the context of Vietnam, and thus clearly merits further empirical considerations. There are a number of theoretical arguments as to why trade reform has a different impact on gendered earnings and hence on gender wage gap. Trade reform may lead to a contraction in the gender wage gap as labour market discriminations are eventuated as a response to increasing competitive forces induced by trade liberalization (Becker, 1971). This taste-based approach suggests an improvement in earnings of women of comparable skills to men, narrowing the gender wage gap. In addition, a conventional trade theory, based on the Heckscher-Ohlin model, points to the factor-price equalization effect of trade, which results in a relative increase in wages of unskilled labour as compared to wages of skilled labour. This effect may advantage women and accordingly lower the 3

gender wage gap since women are often over presented in the lower-wage, lower-skilled jobs than men. An alternative argument by Greenaway and Nelson (2000) emphasizes that trade may result in a premium on skills due to the increasing demand for high-skilled labour in the era of constant technology changes and globalization. The bargaining power of highskilled labour will then rise while that of low-skilled labour may not. Given that in developing countries the average man has a higher level of labour market skills and does the average woman, this skill-favourable scheme may widen the gender pay gap. Linked to the incidence of the informal sector under trade reform, it is commonly acknowledged that an influx of unskilled labours, which is largely comprised of women, into the formal wage employment, especially the export-oriented industries, could intensify the gender wage gap towards women disadvantage. With respect to the impact of trade featured in the informal sector of employment, Goodfriend and McDermott (1995) claim that the informal sector is likely to absorb the negative effects of trade expansion. Trade liberalization results in market expansion, leading to a larger scope for specialization and inducing gains from specialization. Accordingly, production from households the informal sector, will be shifted towards specialized firms the formal sector. Wages in the formal sector will be increased more than the value of household s labour marginal product. Given the overrepresentation of women in the informal sector, the gender wage gap tends to be more severe under trade reform. In this paper, we examine the gendered earnings effect of trade liberalization vis-à-vis the segmentation of men and women in the formal and informal labour sector in Vietnam for the period 1993-2004. Using a series of household data surveys for Vietnam, we particularly investigate to what extent trade reform will exert an earnings effect on both gender groups given their participation in the formal and informal labour sector. The first aim of this study is to discover characteristics that are relevant for explaining the earnings position of women relative to that of men under the era of trade liberalization. The second aim is to distinguish the gender earnings differentials between the sectors of employment and accordingly detect the impact of trade in each sector, using decomposition techniques. Our empirical examination will therefore provide further insights into gender implications of trade reform in Vietnam. 4

The paper proceeds as follows. In the next section, we review some related empirical literature on the gender wage gap. In section 3, we highlight some stylized facts on gender issues in relation to the labour market in Vietnam. The discussion on data and methodology will be presented in section 4, which is followed by section 5 with a discussion on empirical results. We conclude the paper with some remarks in section 6. 2. Literature review There is a growing empirical literature that explores the relationship between trade and gendered earnings in developing countries. Evidence on the topic, however, remains inconclusive with respect to both trade measures used and the employment context investigated. Using a sample of both developed and developing countries, Oostendorp (2002) found a negative association between trade openness, which is measured as the share of export plus import over GDP or the share of foreign direct investment net inflows of GDP), and the size of gender wage gap within occupational groups in the years 1983 1999. This empirical finding, however, does not explicitly attribute the narrowing effect as pertaining to an effect in male or female earnings. Santos and Arbache (2005) examined the gender wage gap for Brazil in the period 1982 1999, using an alternative measure of trade openness a transformation of sectoral tariff rates. Their empirical analysis points to a conclusion that increasing openness is associated with narrowing wage gap, which is largely mediated through a relative decline in men s wages. A few other studies discern the gender wage effect of trade with a particular interest in possible variations of such effects across different industries. For the case of Mexico, Artecona and Cunningham (2002) found that the residual gender wage gap over the period 1987-1993 declined more intensively in concentrated industries than in nonconcentrated industries. Alternatively, based on the same dataset, Ghiara (1999) proved that the wage differential increased substantially in the manufacturing sector, mainly due to changes in endowments of human capital between women and men. The study reports that these changes were not observed in the non-tradable sectors, however. Evidence found by Fleck (2001) further confirms that trade impact varies greatly between industries. 5

The adverse effect of trade on labour earnings has also been broadly documented in studies of the gender pay gap. Nicita and Razzar (2003) analyzed the extent to which the poor benefit from trade liberalization, which is notably reflected in the growth of the export-led sector in Madagascar. While the key empirical finding of their study suggests a rise in the income of poor household as a result of export-driven growth, it also indicates that the benefits are adversely distributed towards female workers, who appear to gain less from trade expansion. In Bangladesh, Paul-Majumder and Begum (2000) showed a similar picture of trade expansion manifest in the export-led industry. An increase in the gender gap in favour of men is mainly because males are increasingly employed in the skilled jobs while female workers are driven out from that job and concentrated more and more in low-skilled jobs. In pursuit of the trade-gender linkage inducted in the Vietnam s context, the empirical literature to date has remained modest, although studies on gender inequality have been growing. Liu (2004a, 2004b) examined gender inequality vis-à-vis gender earnings gap in Vietnam. In the former study, she considered the heterogeneity across the state and private sector in relation to gender wage gap and concluded that within-sector differences contribute more than between-sector differences in explaining the gender earnings gap. The latter study analyzes the decomposition of earning differentials across time: between 1993 and 1998. This study reveals some evidence of a wage disparity between male and female workers in Vietnam and attributes this disparity to an adverse change in discrimination. The findings of both Liu s studies are interesting but indifferent toward the presence of trade liberalization in Vietnam. In a field study on female workers in the garment sector, Kabeer and Tran (2006) characterized working conditions of and problems faced by women employed in this sector in 2001. They concluded that women were crowded into the garment industry as a path out of rural poverty as other sectors are not open to them. While this finding helps clarify the employment status of women and necessitate policies for a sustainable employment solution for women, it does not exclusively address the impact of trade on gendered employment and earnings. Using individual and household data obtained from VLSS 1992/1993 and 1997/1998, Rama (2001) pinpointed a sharp decline in the gender pay gap, and attributed this decline to the effects mediated through the economic reforms under Doi Moi. Pham and Reilly (2007) 6

offered further insights into a sizable contraction in the gender pay gap by examining the degree to which the gender pay gap varies across the conditional wage distribution. The key results point to stability in the gender pay gap in most wage distributions over the period 1993-2002, and assign this stability to much of the effects exerted by the economic reforms in Vietnam. Like Rama (2001), Pham and Reilly (2007) were not able to distinguish components of Doi Moi, which may be accountable for an improved wage position of women. A very recent study by Nguyen et al. (2007) examined the impact of trade liberalization, which is measured by sectoral trade intensity, on gendered earnings and on returns to education in particular. Consistently with previous studies on Vietnam, Nguyen et al. (2007) reported a contraction in the gender wage gap and associated this contraction with export-driven changes in female labour demand rather than with importdriven changes in competition. In addition, the study reveals that the returns to education have increased, but the increase in the returns to education cannot be explained by increases in trade openness. Admittedly, these studies do not provide conclusive evidence of the relationship between trade and the gender pay gap. Moreover, has none of the existing studies examined the trade-gendered earnings linkage intersected with distinctive features of the labour market in developing countries, notably the incidence of the informal sector, under the presence of trade liberalization. Further empirical investigations on this field for Vietnam as a developing country case are therefore clearly warranted to grasp a better understanding on the nature of the relationship between trade and gender pay gap. 3. The Vietnam s labour market: gender and sectoral distribution of employment In this section, we highlight some issues relating to Vietnam s labour force participation and segmentation in the period 1993 2004. We also compare some key employment indicators between different categories formed by gender and sector. Vietnam s labour market is characterized by a high participation rate among both men and women as shown in table 1. Different patterns in the changes of gender-specific participation rates emerged in the period, with an increase in the participation rate of women from 79.34% to 80.28% between the years 1003 and 2004. Table 1 also depicts the dynamics of the labour market, differentiated into sector of employment and gender. 7

In terms of sector, self-employment represented a largest component of Vietnam s labour market. The share of self-employment, however, exhibited a contraction over this period, i.e. from 79.95% in 1993 to 68.98% in 2004. By contrast, the shares of formal wage employment and of informal wage employment were both increased. A rise of formal wage employment from 2.44% of the labour force in 1993 to 16.69% in 2004 was indicative of a remarkable growth of the formal labour sector in Vietnam during this period. This outcome seems to accords with the introduction of the new Enterprise Law enacted in 1999. On the other hand, the development of the informal wage sector was inclined with the prosperity of household enterprises in the period studied (Vijverberg and Haughton, 2002). The shares of men and women in the labour force also describe some interesting patterns across sectors. Self-employment category was constantly marked by the active role of women, except for the year 1998. In contrast, women were less active than men in the category of informal wage employment, evident from a lower share of women relative to that of men. Moreover, this category also exhibited a falling share of female labours, in contrast to a steady extension of male labours. A similar trend was observed with respect to formal wage employment, though the lowest proportion of women was exhibited in the year 1998 and increased in the year 2004. Not surprisingly, women appeared to be less attached to and/or less chosen to join the formal sector as compared to men. So, although women have been actively involved in the Vietnam s labour market under the era of economic reforms, the majority of working women was absorbed in self-employed jobs. This outcome gives an indication of gender employment disadvantage, as also pointed out by other studies (Kabeer et al. 2005). 8

Table 1. Labour force participation and labour force segmentation 1993-2004 1993 1998 2004 Aggregate Male Female Aggregate Male Female Aggregate Male Female Labour force participation 81.24 83.38 79.34 79.62 80.83 78.52 81.78 83.31 80.28 (%) Labor force segmentation by gender (%) Overall - 48.21 51.76 53.48 46.52 50.23 49.77 Self-employment 47.16 52.84 51.91 48.09 44.99 55.01 Informal wage employment 59.43 40.57 62.24 37.76 68.43 31.57 Formal wage employment 51.92 48.08 63.35 36.65 56.27 43.73 Labour force segmentation by sector (%) Self-employment 79.95 78.16 81.63 72.12 70.00 74.56 68.98 61.78 76.25 Informal wage employment 13.16 16.21 10.32 10.59 12.32 8.59 14.34 19.53 9.09 Formal wage employment 2.44 2.62 2.26 8.70 10.30 6.85 16.69 18.69 14.66 Public sector & others 4.45 3.01 5.79 8.59 7.37 9.99 - - - 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 Source: authors calculations from the VLSS 1992/1993, VLSS 1997/1998 and VHLSS 2004. Note: a. Figures are calculated as a participation rate of the group of interest over the employed people, for men and women separately. b. Labour force covers people aged from 15 to 65 years old. c. Employment is identified as having jobs over the past 12 months d. Self-employment consists of agricultural self-employment and non-farm self-employment. Informal wage employment is confined to paid work for other households or small household enterprises. These two categories constitute the informal sector. e. Formal wage employment refers to those employed in formal enterprises of different kinds of ownership. This category is identified as the formal sector. f. Government sector refers to working for the government and other state organizations. Table 2 provides detail of the employment picture in Vietnam with indicators of average hourly wage rates and average working hours on a daily basis. This information is differentiated into gender participation in each sector. Regarding the allocation of working hours per day, the average woman devoted approximately 5% less of her time to work as compared to her male counterpart. This gender difference was small and has been stable over time across sectors. Whereas men and women seem to contribute equal time to income-generating activities, but they have been rewarded differently given the fluctuations in hourly wage rates between the gender groups as shown in table 2. Hourly wage rate has grown on average at approximately 21% per annum, increasing from 2,120VND to 5,530VND, and from 1,800VND to 4,890VND between the years 1993 and 9

2004, for men and women, respectively. This overall improvement in labour earnings was accompanied by an improved wage position of female relative to male wage position. As can be seen from table 2, the gender wage gap in favour of men persisted, though it has been gradually narrowed. As different sectors of employment are considered, some interesting patterns emerged. First, formal wage earners were better paid than their informal counterparts irrespective of gender. It is important to note that in 1997 the Vietnamese government commenced the policy on minimum wage for employees working in Vietnamese formal enterprises. This policy might have certain implications for the levels of wage in the formal sector. Second, and more strikingly, a contrary between informal wage employment and formal employment was observed in the gender wage gap. While women received a better reward in the formal sector, their position became relatively poorer in the informal sector, i.e. the gap was narrowed from 0.668 to 0.879 for the formal sector, while it was widened from 0.846 to 0.732 for the informal wage sector. Table 2. Employment indicators: average working hours/day and nominal hourly wages Working hours/day Nominal hourly wage rate (1000VND) 1993 1998 2004 1993 1998 2004 Male 7.042 7.966 7.16 2.115 5.292 5.525 Female 6.611 7.546 6.809 1.796 4.673 4.887 F/M Ratio 0.939 0.947 0.951 0.849 0.883 0.885 Self-employment Male 6.801 8 6.616 - - - Female 6.467 7.22 6.485 - - - F/M ratio 0.951 0.903 0.980 - - - Informal wage employment Male 8.317 8.115 8.135 2.266 5.553 4.265 Female 7.781 8.027 7.725 1.917 4.253 3.122 F/M ratio 0.936 0.989 0.950 0.846 0.766 0.732 Formal wage employment Male 6.707 8.275 7.941 1.956 6.149 6.837 Female 6.573 8.254 7.901 1.306 5.34 6.009 F/M ratio 0.980 0.997 0.995 0.668 0.868 0.879 Source: authors calculations from the VLSS 1992/1993, VLSS 1997/1998 and VHLSS 2004. Note: a. Hourly wage rate includes all payments in cash and kinds. b. F/M ratio is an indication of the relative position, measured as the ratio of average female wage (working hour) to that of male counterpart. 10

In sum, the foregoing descriptive analysis points to a contraction of the informal sector, and on the contrary a growth of the formal sector in Vietnam s labour market in the period 1993 2004. In both sectors of wage employment, men were reported to earn more than did women during this period. While the differential was narrowed in the formal sector, it was substantially widened in the informal sector, implying a relatively poorer position of women in informal employment along the Doi Moi also noticed by Kabeer (2005). A question remains as to what extent the divergence in the gender pay gap across sectors can be explained using empirical models, which account for the presence of trade interlinked with gender and sector of employment. 4. Data and Model specification Data descriptions The paper makes use of several sources of available data. The major source of data refers to the three household surveys, two Vietnam Living Standard Surveys (VLSS, henceforth) conducted in 1992/1993 and 1997/1998 and the Vietnam Household Living Standard Survey (VHLSS, henceforth) conducted 2004 1. Another source of data from General Statistics Office of Vietnam offers information on export, import values and GDP at the provincial level, from which commune adjusted trade ratios will be derived to proxy for trade liberalization. Although the household surveys were slightly modified over time, the basic content between the two early surveys VLSSs and the later VHLSS are similar. At the household and individual level, the surveys provide a wide range of information on household characteristics including basic demography, employment and labour force participation, education, health, income, expenditure, housing, fixed assets and durable goods etcetera. The sample size of the VLSS 1992/1993 and of the VLSS 1997/98 amounts to a number of 4,800 and 6000 households, respectively. The full sample of VHLSS 2004 covers 45,000 households, of which 9,000 households were surveyed in terms of expenditure 1 We note that Vietnam Household Living Standard Survey conducted in 2002 is also accessible. However, we decide to select the three surveys of our main interest. i.e. VLSS 1992/1993, 1997/1998 and VHLSS 2004 to carry out our empirical analysis for the 1993 2004 period. While the use of all the four household surveys will not change the essence of our study, it requires much more empirical work. The analysis can be extended to include VHLSS 2002 upon request. 11

and income. It should be noted that the two VLSSs construct a panel of 4300 households. It is, however, unable to link the two VLSS with VHLSS 2004. Through the three surveys data on employment including earnings are available for all household members aged from 15 and 65 in the preceding 12 months. Note that this is not the case for self-employment category. Details of employment information slightly differ between VLSSs and VHLSS in that the latter survey provides a more specific classification of employment category. The individual-level sample is restricted to 7,565 wage earners including 426, 513 and 6,626 for the years 1992, 1998 and 2004, respectively 2. We note that information on earnings is not available for the self-employment category and fairly missing for the informal category, warning us to interpret the results with some caution. In the present study, we exploit the information on the main job over the preceding 12 months, noticing that the earnings from the second job on average hold a negligibly small fraction in total earnings of the individuals. The nominal hourly wages include payment in cash and kind. The nominal wages are then adjusted to the year 2000 price using the CPI supplied by GSO. The natural logarithm form of real hour wage rates are incorporated as the dependent variable in the Mincerian earnings equations. The sample of wage earners is classified into three categories of employment: formal wage earners, informal wage earners and government officers. It is necessary to provide further explanations on this classification, especially with respect to the definition and the scope of informal wage employment. As recommended by ICLS(2003), informal employment consists of two groups: informal self-employment and informal wage employment. Given the insufficiency of the data used, some adaptations are needed for our study. First, 2 Note that, while individual earnings information on the preceding 12-month jobs is abundant in the VHLSS 2004, this is relatively poorer in the two VHLSSs. On the contrary, earnings information on the preceding 7-day jobs was fuller in the two VHLSSs, while it was not at all recorded in the 2004 survey. For the sake of consistency, in our analysis we decided to count on the job over the preceding 12 months, which are available across the three surveys. This leads to a relatively small sample size of the years 1993 and 1998 samples, and thus does not allow us to estimate separate regressions for these two years. We admit that this limitation certainly has an adverse impact on the rigorousness of our empirical results. However, we argue that the earnings impact of trade on will be better captured when taking into consideration the jobs engaged in the whole year rather than the jobs over the preceding 7 days. Pham and Reilly (2007) use the preceding 7-day jobs from the former two surveys and the preceding 12-month job from the VHLSS 2002. While this gives rise to the problem of inconsistency, it remains as a possibility for our consideration in the next stage of the research. 12

except for VHLSS 2004, the other two surveys do not allow to differentiate informal selfemployment from self-employment as a whole. As shown from the three household surveys, self-employment is largely assigned to agricultural activities (see table 2 below), which are considered to be informal according to the proposed indicators on employment by Millennium Development Goal 3 (Chen et al., UNIFEM 2005). So, we fit selfemployment in the informal sector alone. Second, informal wage employment is not well identified from the current surveys in that we do not have detailed information on different types of contract workers (employees of informal enterprises, casual or day labourers, temporary or part-time workers, paid domestic workers, unregistered or undeclared workers, called homeworkers). For this reason, we categorize work for small household enterprises (in VLSS 1997/1998), and work for other households (in the VHLSS) as informal wage employment. Although this categorization may be somewhat ad hoc, it is drawn from the ICLS s (1993) consensus which refers the informal sector as employment and production that takes place in small and/or unregistered enterprises. Third, in the VLSS 1992/1993, working for other households is merged in the same group with working for private enterprises. As a matter of fact, the number of private enterprises was reckoned to be very small in the period 1992/1993. Therefore, we claim working in private company/household in the VLSS 1992/1993 as pertaining to informal employment. Fourth, while employment in the government sector captures a considerable share in the labour market in Vietnam, its connection to trade reform may be limited. Moreover, significant gender wage disparities were not found within the government sector in the previous studies using the data from VLSSs (Liu, 2004a). Finally, the government sector was treated as a special type of employment according to the VHLSS 2004 survey and thus precluded from the formal labour work. To make it compatible among the three surveys, we therefore exclude individuals employed in the government sector in the two earlier surveys. Model specifications. In an attempt to identify the effect of trade on individual earnings, we adopt an extended Mincerian earnings equation, which controls for various determinants of earnings including individual endowments, geographical and regional attributes and industry affiliations. By convention, individual endowments such as age, marital status, education, 13

working experience, household headship and migrant status are included in the earnings model. Following Pham and Reilly (2007), Reilly and Dupta (2006), we use a set of educational dummies rather than years in schooling to capture human capital effects. Indicators of completion of primary school, secondary, high school and higher education are constructed with the reference group of individuals who are illiterate or without any schooling. To proxy for labour market experience, we use both age and its squared term, and also working experience and its squared term. As in Pisani and Pagan (2004), we argue that indicators of regional residence may also influence earnings and therefore include a set of seven regional dummies in the earnings models. Further, indicators for industry affiliation of individuals are introduced to control for industry-fixed effects. Finally, year dummies are incorporated to capture the impact of other economic policies than trade on individual earnings. The standard Mincerian wage equations are then augmented with the inclusion of trade variable L to capture the impact of trade on sectoral wages. Though some argue that in developing countries employment responses to trade policy are greater than wage responses (Goldberg and Pavnick, 2004), this remains as a strong assumption. Thus, it is reasonable to consider the effect of trade in wage equations, making it possible to detect a link between trade and the gender wage gap in the next empirical application. A critical concern in the estimation of earnings is attributed to the choice of individual between different sectors of employment. Econometric techniques can be used to correct for self-selection, and appear to be most effective if identification of the selection equation is achieved by the inclusion of one or several variables that affect the choice of a certain sector, but not the sectoral earnings. In this framework, the probability of working in a certain sector will be first estimated, assuming that an individual chooses the sector which maximizes their utility. The resultant coefficients from this selection equation are then used to compute the inverse of the mills ratio term for inclusion in the earnings equations. While a few studies on gender pay gap for Vietnam apply this two-step procedure and find a significant selection effect in regard to female earnings estimates (Liu, 2004ab), others argue that the use of self-selection correction techniques is highly sensitive to distributional assumptions and to the choice of appropriate instruments 14

(Rama, 2001, Pham and Reilly, 2007). Following the latter argument, we resign from a selection correction procedure in this paper 3. Measures of trade openness In pursuit of identifying the impact of trade on individual earnings, we introduce a set of trade variables L in both the Mincerian equations to proxy for trade openness. As a phenomenal problem in the trade literature is pertaining to the lack of a clear definition of what is meant by trade openness, there is little consensus on the measures of trade in existing empirical studies (Yanikkaya, 2003). Broadly, trade openness measures can be divided into two categories: measures of trade volume and measures of trade restrictions. While the former mostly refers to the ratios of exports plus imports to GDP, the import penetration ratios, and export shares in GDP, the latter contains tariff rates, export taxes, total taxes on international trade and indices of non-tariff barriers. Not only does each trade measure encounter its own problem, it also provides different implications to the issues studied. Trade volume is a standard measure and often suffers from criticism, which argues that a highly open economy may well be accompanied by export subsidies and/or by trade barriers to protect its import sector. From a policy point of view, a comprehensive form of trade barriers appears to be ideal in ascertaining the impact of trade policy (Yanikkaya, 2003, Harrison and Hanson, 1999 among others). Unfortunately, such as a measure is hardly available for developing countries. Moreover, for the context of Vietnam the heavy use of non-tariff barriers as a substantial trade restriction during the early nineties may weaken the effectiveness of tariff rates as a proxy of trade liberalization (Pham, 2007). At the current stage of this study, we are therefore confined to the use of trade intensity ratios as an indicator for trade openness in our empirical applications. Two proxies for trade openness will be in place. The first proxy refers a standard measure of the degree of openness - the share of exports and imports as a percentage of GDP at the provincial level. Although this ratio is warned as a loose measure of trade liberalization, it does reflect trade intensity, which is believed to have an impact on labour earnings as reported in other studies on the gender pay gap (Oostendorp, 2002). We adopt this measure while allowing for some caution with result 3 As a matter of fact, we re-estimated the Mincerian earnings equations using the Heckman selectivity correction and found no evidence of significant selection-bias 15

interpretations. Regarding the level of trade exposure captured by this trade index, another limitation of this proxy arises because earnings are estimated at the individual level while trade intensity is measured at an aggregate level the provincial level. The second proxy of trade openness is accordingly derived from the first proxy, aiming to translate the impact of trade, which occurs at macro level, into the effect of trade on wages and employment decision at the individual level. In this way, we construct trade indices at the commune level the lowest administrative level, using the information on exports, imports and GDP at the provincial level. We argue that the level of trade exposure for individuals/households should be adjusted according to the relative contribution of each commune to the provincial trade performance 4. The provincial aggregated export (import) over GDP data are, therefore, adjusted to the r commune level, generating two trade openness indices: rural trade openness index T c u for rural communes and urban trade openness index T c for urban communes. (1) T r c Q EX p IM c p C GDPp Qc i 1 (2) T u c E EX p IM c p C GDPp Ec i 1 where Qc is the output value of all farming activities at commune c; C is the number of communes in province p; EX p, IM p, GDP p are aggregate values of export import, and GDP of province p. r Tc is the trade exposure adjusted by the agricultural output weight at each rural commune. Ec is the number of people working in tradable sectors in each urban commune; and weight of each urban commune. u Tc is the trade exposure adjusted by the employment The use of this proxy rests on the assumption that individuals are tied to their commune to make contribution to their provincial trade performance. As such, labour mobility is 4 Pham (2006) suggests to combine the tariff data and the household surveys to construct a proxy of trade exposure the commune level. The nationally aggregate sectoral tariff data are, therefore, adjusted to the commune level, generating two trade openness indices: agricultural openness index and nonfarm openness index. In our application, trade intensity is used rather than trade restrictions, and similar weights as used by Pham are applied here. 16

assumed to be minimal across communes. Despite this strong assumption, the trade indices as applied appear to be a practical choice for the surveys 1997/1998 and 2004 given the coverage of a reasonably large number of communes in the three surveys, 150 communes in VLSS 1992/91, 156 communes in VLSS 1997/98, and 3081 communes in VHLSS 2004. 5. Empirical Results Wage equations Given the two employment outcomes of our interest, i.e. formal vs. informal employment, we carry out the estimation for each sector separately. Note that the estimation of the pooled sample of both gender groups (not reported here) for both sectors indicates that the sectoral dummy is significant. Informal wage earners in general earn 10% less than their peers employed in the formal sector. The results for the pooled sample of each gender group with respect to both proxies of trade are presented in table 3. We first examine the effects which are reported as the most relevant in determining individual earnings. We notice a significant wage differential between men and women across both sectors of employment. Indeed this wage differential appears to be more pronounced in the informal sector where men earn approximately 30% higher then women while this figure is 10% in the formal sector. This finding is consistent under both trade measures used. The effect of age is evident and significant in most estimation. Apparently, age and its squared term describe the expected inverted-u shaped relationship between wage rates and the human capital attribute. An exception is accredited to female earnings equation in the formal sector, where the impact of age follows the same pattern but appears to be insignificant. Unlike the effect of age, reward to working experience is only significant in the formal sector, indicating experience as an important criterion in wage determination in this sector. The returns to experience, however, appear to be relatively low for both males and females. For instance, an increase of one year in working experience only offers a rise of approximately 4% in the wage rate. This finding is also found by other studies on Vietnam gender pay gap (Liu, 2004ab and Rama, 2001). 17

An interesting finding emerges as regards to returns to education. For both men and women employed in the formal sector, returns to education appear to be significantly higher at more advanced levels of education, i.e. completion of high school and completion of higher education. For the informal sector, low levels of education are also found to be significant, especially for women. While the evidence of an increased returns to education is not surprising as it is also documented by other studies for Vietnam (Pham and Reilly, 2007, Nguyen et al, 2006, Liu 2004ab), this finding highlights an interesting divergence in returns to education between the two sectors of employment across gender groups. Although education attainment is higher for both men and women in the formal sector than in the informal sector, discrimination against women is manifest only in the formal sector, where reward to education is in favor of men. Men are reported to earn approximately 10% ceteris paribus higher than women of a comparable level of education. By contrast, educated women are more appreciated in the informal sector given a pay premium of 24% over men. Table 3 also reports evidence of the role of regional attributes on labour earnings. Across estimations, variations in earnings describe a geographical pattern in which workers in the south on average earn more than in the centre and the north of Vietnam, irrespective of sector of employment. As a matter of fact, this finding is consistent with a popular stylized fact in Vietnam, and also well confirmed by previous studies (Pham & Reilly, 2007, Liu, 2004ab). Similar to the regional effect, the impact of industry affiliation appears to be exciting. It should be noted that non-tradable sectors including construction and services are taken as the reference group in our estimation. For men, manufacturing and food processing sectors are among the most well paid in the formal sector, while at the same time these sectors absorb the highest share of male workers. Agriculture and textile sectors, as commonly observed, offer a relatively lower wage to male workers. A different story applies for women, as only is the negative effect of industry affiliation evident. This finding implies that women less from trade when employed in the tradable sectors, as also suggested by Nicita and Razzra (2003) for Madagascar, and by Paul- Majumder and Begun (2000) for Bangladesh. In particular, textile and food processing industries are found to be attached with a lower pay in both sectors of employment. 18

Table 3. Regression results of earnings for gender and sector of employment Dependent variable: the natural log of real hourly wage. The estimates based on the first proxy of trade are presented in the first four columns, and the last four columns refer to the estimates with the second proxy of trade Formal Informal Formal Informal Male (1) Female (2) Male (3) Individual endowments Marital status -0.0273-0.1015-0.0666 0.0326-0.0218-0.1158-0.0518 0.0437 (0.0478) (0.0372)*** (0.0358)* (0.0498) (0.0542) (0.0404)*** (0.0357) (0.0508) Age 0.0249 0.0002 0.0388 0.0540 0.0261 0.0110 0.0404 0.0661 (0.0116)** (0.0124) (0.0081)*** (0.0140)** (0.0132)** (0.0138) (0.0085)*** (0.0154)*** The squared term of age -0.0003 0.0000-0.0005-0.0007-0.0004-0.0002-0.0005-0.0009 (0.0001)** (0.0002) (0.0001)*** (0.0002)*** (0.0002)** (0.0002) (0.0001)*** (0.0002)*** Migrant status -0.0259-0.0158 0.0184 0.0467-0.0182-0.0364 0.0519 0.1051 (0.0571) (0.0513) (0.0707) (0.1051) (0.0754) (0.0580) (0.0783) (0.1114) Household headship -0.0107-0.0805-0.0497-0.0001-0.0027-0.0728-0.0468-0.0217 (0.0322) (0.0401)** (0.0325) (0.0779) (0.0401) (0.0560) (0.0345) (0.0870) Working experience 0.0387 0.0374 0.0031-0.0153 0.0399 0.0324 0.0034-0.0190 (0.0065)*** (0.0069)*** (0.0062) (0.0120) (0.0078)*** (0.0080)*** (0.0061) (0.0124) The squared term of working experience Education dummies Completion of primary school Completion of secondary school Female (4) Male (5) Female (6) -0.0006-0.0006 0.0001 0.0007-0.0006-0.0003 0.0001 0.0008 (0.0002)*** (0.0002)*** (0.0002) (0.0005) (0.0003)** (0.0003) (0.0002) (0.0005) -0.0654-0.1114 0.0297 0.2505-0.0307-0.0178 0.0353 0.2098 Male (7) Female (8) (0.0993) (0.0937) (0.0781) (0.0975)** (0.1108) (0.0985) (0.0787) (0.0983)** -0.0014-0.0454 0.0985 0.2576 0.0067 0.0293 0.1123 0.2291 (0.0907) (0.0923) (0.0806) (0.1090)** (0.1010) (0.0978) (0.0801) (0.1108)*** Completion of high school 0.1881 0.1425 0.2165 0.4975 0.2187 0.2194 0.2219 0.4294 (0.0925)** (0.0898) (0.0851)** (0.1283)*** (0.1046)** (0.0965)** (0.0892)** (0.1355)*** Higher education 0.5207 0.4587 0.0520 1.1957 0.5865 0.5477-0.2621 0.8903 (0.0941)*** (0.0929)*** (0.3547) (0.2469)*** (0.1068)*** (0.1014)*** (0.5309) (0.1547)*** 19

Male (1) Formal Informal Formal Informal Female Male Female Male Female Male (2) (3) (4) (5) (6) (7) Region dummies Red River Delta -0.0091-0.1911 0.0442-0.2487 0.0202-0.1842 0.0401-0.1682 (0.0471) (0.0476)*** (0.0457) (0.0776)*** (0.0523) (0.0540)*** (0.0468) (0.0715)** North Central Coast -0.0934-0.1753 0.0558-0.3113-0.1125-0.1161 0.0493-0.2668 (0.0538)* (0.0565)*** (0.0541) (0.1033)*** (0.0628) (0.0684)* (0.0574) (0.1017)*** South Central Coast 0.0903 0.0054 0.1834-0.0407 0.1055-0.0009 0.2002 0.0040 (0.0501)* (0.0506) (0.0521)*** (0.0827) (0.0600) (0.0620)* (0.0558)*** (0.0794) Central Highland 0.2771 0.0671 0.1562 0.0020 0.2790 0.1415 0.1875 0.1618 (0.0751)*** (0.0816) (0.0747)** (0.1080) (0.0935)*** (0.1083) (0.0792)** (0.1109) South East 0.2607 0.2640 0.3555 0.1422 0.3643 0.3842 0.3851 0.2408 (0.0535)*** (0.0525)*** (0.0540)*** (0.0921) (0.0574)*** (0.0557)*** (0.0539)*** (0.0881)*** Mekong River Delta 0.1960 0.2147 0.2691-0.0315 0.2428 0.2512 0.2739 0.0523 (0.0496)*** (0.0471)*** (0.0488)*** (0.0839) (0.0578)*** (0.0552)*** (0.0513)*** (0.0812) Urban 0.2239 0.1046 0.0630 0.0599 0.1533 0.0997 0.0453 0.0980 (0.0288)*** (0.0279)*** (0.0312)** (0.0507) (0.0351)*** (0.0329)*** (0.0354) (0.0585)* Industry affiliations Agriculture -0.0714-0.0338-0.1517 0.0006-0.0163-0.0469-0.1495-0.0157 (0.0718) (0.0874) (0.0323)*** (0.0642) (0.0733) (0.0896) (0.0327)*** (0.0669) Manufacturing 0.2091 0.0547-0.0305 0.0274 0.2650 0.0929-0.0189 0.0434 (0.0362)*** (0.0465) (0.0324) (0.0640) (0.0401)*** (0.0512) (0.0326) (0.0644) Food processing 0.0783-0.1271-0.0776-0.1366 0.1170-0.1284-0.0608-0.1516 (0.0512) (0.0605)** (0.1082) (0.1057) (0.0545)** (0.0651)** (0.1080) (0.1014) Textile and garment -0.1014-0.1148-0.3178-0.1294-0.0387-0.0870-0.3074-0.1383 (0.0712) (0.0422)*** (0.1031)*** (0.0644)** (0.0756) (0.0492)* (0.1038)*** (0.0663)** Paper and other light goods -0.0438-0.1342-0.2094-0.1682 0.0178-0.0944-0.1819-0.1311 (0.0773) (0.0959) (0.0842)** (0.1019)* (0.0794) (0.0977) (0.0810)** (0.1024) Year 2004 0.9828 0.8756 0.4439 0.3755 1.0916 0.8800 0.4478 0.3506 (0.1470)*** (0.0927)*** (0.0647)*** (0.0910)*** (0.1626)*** (0.0963)*** (0.0683)*** (0.1019)*** Female (8) 20

Formal Informal Formal Informal Male (1) Female (2) Male (3) Female (4) Male (5) Female (6) Male (7) Female (8) Year 1998 1.1139 1.0552 0.5263 0.6126 1.1302 1.0035 0.4780 0.5618 (0.1444)*** (0.1258)*** (0.1018)*** (0.1217)*** (0.1601)*** (0.1255)*** (0.0953)*** (0.1248)*** Trade openness [(Ex + Im)/GDP] Trade index at the provincial level 0.0711 0.0759 0.0266 0.0442 (0.0139)*** (0.0141)*** (0.0156)* (0.0232)* Trade index at the 0.2528 0.2455-0.1083-0.4522 commune level (0.1832) (0.1329)* (0.0865) (0.3169) Constant -0.5402 0.0764-0.1280-0.6029-0.6789-0.1270-0.1577-0.7337 (0.2825) (0.2307) (0.1714) (0.2479)** (0.3150)** (0.2453) (0.1810) (0.2521)*** Observations 2068 1615 1903 780 1579 1207 1722 707 R-squared 0.36 0.44 0.19 0.23 0.34 0.41 0.19 0.22 Robust standard errors in parentheses *significant at 10% level; ** significant at 5% level; *** significant at 1% level 21

The results also indicate that marital status profoundly reduces female earnings in the formal sector, while this does not apply in the informal sector. It is interesting to note that although this factor is reasonably assumed not to be related to labour productivity (Liu, 2004a), it turns out to be a determinant in wage determination of the formal sector. This may give rise to an additional warning on discrimination in the formal sector. Consistent with other studies on Vietnam, we do not find evidence for the effect of migrant status and household headship on earnings. We turn now to examine the impact of trade openness on individual earnings in both sectors of employment. Measured by the percentage of export plus import over GDP at the provincial level, trade openness is shown to have a gendered effect in the period studied. Irrespective of sector, the results indicate that greater openness is associated with higher wages, evident from significant coefficients in estimations for both gender groups. For example, in the formal sector, 1% markup in the trade index corresponds to a rise of 7.1% and of 7.6% in male and female earnings, respectively. Regarding the informal sector, the positive effect is, however, rather marginal and moderately significant, i.e. an increase of 2.7% and of 4.4% in the wage rates for males and females, respectively. As the second trade proxy, i.e., the commune adjusted trade index, is applied, the trade effect on informal earnings is no longer significant, though has the same sign. The reported coefficients also reveal a difference in gendered response to trade openness. Slightly higher coefficients attached with openness in female earnings equations give a rough indication of a narrowing effect of trade on gender wage disparity. This finding seems to support the hypothesis on trade premium skills (Greenaway and Nelson, 2000), and needs to be further examined by a decomposition analysis. Decomposition analysis Given the estimated wage distributions, we decompose the wage gap to investigate whether the wage differentials are caused by wage discrimination, or from a disparity in skill endowments between men and women in each sector of employment. To do so, we follow the conventional decomposition Blinder-Oaxaca (1973) approach, which has been commonly used in wage gap studies. This approach isolates the wage gap not explained by endowments but by a difference in endowments rewards the so-called treatment effect. Let w m and w f be the means of the natural logs of male (m) and female ( f ) wages. The wage gap can be written as: 22