The impact of trade and investment liberalization on

Kien Trung Nguyen

October 2014

Working Paper No. 2014/20

Arndt-Corden Department of Economics

Crawford School of Public Policy

ANU College of Asia and the Pacific

Page 2

Page 3

The impact of trade and investment liberalization on the

wage skill premium: evidence from Vietnam

Kien Trung Nguyen*

The Arndt-Corden Department of Economics

Crawford School of Public Policy

Australian National University

Email: kien.nguyen@anu.edu.au

* Acknowledgements: I am grateful to Prema-chandra Athukorala, Chris Manning, and Hal Hill for valuable

The impact of trade and investment liberalization on the

wage skill premium: evidence from Vietnam

Kien Trung Nguyen

The Arndt-Corden Department of Economics

Crawford School of Public Policy

Australian National University

Email: kien.nguyen@anu.edu.au

Abstract: This paper examines the impact of trade and investment liberalization on the wage

1. Introduction

The impact of an outward-oriented development strategy on wages remains a significant focus

in the ongoing debate about workers’ welfare in developing countries. Opening the market to

international trade and investment will affect the wage skill premium between skilled and

unskilled workers (herein forth referred as to the wage premium). However, there is no

unanimity in the theoretical literature on the impact of industrialization on the wage premium

in the manufacturing sector in developing countries. The Lewis-Fei-Ranis model1 does not

make a distinction between skilled and unskilled labour, but it implies that the wage premium

could continue to increase as long as surplus labour conditions prevail in the economy. The

standard Heckscher-Ohlin-Stolper-Samuelson (HOSS) theory predicts that in a labour-

abundant economy there will be a rise in manufacturing wages of unskilled workers

associated with an expansion in manufacturing exports. By contrast, the Feenstra-Hanson

extension (Feenstra & Hanson 1996) to the HOSS theory postulates that the engagement of

developing countries in global production sharing could result in increasing the wage

premium in manufacturing wages in these countries. This effect also rests on skill-biased

technological change, which accounts for the increased demand for skilled workers following

a rise in imports of capital goods and technology into the developing economies (Acemoglu

2003; Robbins 1996; Wood 1995).

Outward-oriented liberalization reforms in Vietnam since the early 2000s have

generated special interest in how the role of international trade and investment has affected

wage inequality in that country. In particular, these liberalization reforms of trade, investment

and enterprise policy regimes have intensified since 2006, and should have significant effects

on the wages of unskilled workers, where Vietnam’s comparative advantage lies. Given the

nature of these outward-oriented reforms, Vietnam provides an ideal case study of the wage

skill premium in the manufacturing sector.

Despite its importance in the debate on gains from global economic integration, the

issue of a wage premium has received little attention in the studies on Vietnam. Although

Page 6

In this paper, we examine the determinants of the wage skill premium in relation to

trade and investment liberalization in the Vietnamese manufacturing sector using a firm-level

dataset derived from the Vietnam’s Enterprise Surveys. The findings show that trade

liberalization, in particular reductions in tariffs on inputs, through its contribution to export

expansions and employment growth has served to reduce the wage premium in Vietnam. The

presence of foreign invested enterprises has contributed to an increase in the wage premium.

These findings make a significant contribution to the ongoing debate in the literature on

whether and how trade liberalization affects wages and inequality in developing countries.

The structure of the paper is as follows. Section 2 surveys the empirical evidence of the

wage premium following outward-oriented liberalization in developing countries. Section 3

establishes a model relating a wage premium to tariff variations and firm ownership in order

to determine the wage premium using a cross-section of firm-level data. Section 4 discusses

data compilation and econometric method. Section 5 presents and discusses the estimation

results. As a comparison, the following section investigates determinants of average wage

rates. The final section provides concluding remarks on the significant effect that export-

oriented foreign investment has on the wage premium.

2. Trade and investment liberalization and wage skill premium: empirical issues

As with the theoretical literature, the findings of empirical studies remain inconclusive. There

is significant evidence to support the narrowing wage premium between skilled and unskilled

workers in labour-abundant economies in East Asia following export-oriented

industrialization. Throughout the 1960s-1970s export expansion narrows the wage premium

in East Asian economies (Galenson 1992; Kim & Topel 1995; Kuo 1989; Wood 1997). In

Taiwanese manufacturing, the wage inequality between white-collar employees and blue-

collar ones reduced over the process of industrialization (Kuo 1989). Similarly, a reduction in

Page 7

Foreign investment coupled with technological advances is also important for affecting the

relative demand for skilled workers, and has with mixed effects on the wage premium. There

is evidence of a widening of the wage premium resulting from foreign investment and

technical change in Mexico’s maquiladoras (Feenstra & Hanson 1997; 1999) while the

opposite has been observed in Indonesia (Suryahadi, Chen & Tyers 2001). In this country, the

relative demand for unskilled labour rose while the wages of unskilled relative to skilled

workers decreased in the manufacturing over the period 1970s-1980s following the trade

reform. The Indonesia’s finding appears consistent with the prediction of the trade models;

whereas, contrary to Feenstra and Hanson (1996), foreign participation raised the relative

demand for unskilled workers.

Despite its prominence in empirical works, the impact of trade and investment

liberalization on the wage premium still remains a sparsely research subject of a Vietnam case

study. Gallup (2004) examines the wage inequality in Vietnam over the 1990s using two

rounds of the Vietnam Living Standards Survey (VLSS) between 1992/93 and 1997/98.

Employing the same rounds of household-level dataset, Liu (2004) investigates the changing

wage structure following the economic reform, focusing on gender wage gap and the overall

wage inequality between skilled and unskilled workers. Both these studies suggest a moderate

decline in wage inequality between these two time points, however, they do not take the

specific characteristics of trade and investment liberalization into account such as firm

ownership or tariff reduction.

The study by Fukase (2013) explores the two rounds of Vietnam Household Living

Standards Survey (VHLSS) in 2002 and 2004 in order to examine the wage premium in the

aftermath of the 2001 Bilateral Trade Agreement (BTA) with United States. In this study,

potential endogeneity of export intensity is addressed by using the province-tariff reduction as

an instrument. The findings show that increasing exports lead to growth in wages of unskilled

workers in those provinces that experience more exposure to trade liberalization. This rise in

wages is consistent with the East Asian economies in the process of export-oriented

industrialization. A major limitation of this study is that it examines the impact of only a

single trade agreement on industrial wages, without appropriately controlling for the impact of

comprehensive trade and investment liberalization in the overall economy during the same

period. Another limitation is that the study does not explore determinants of inter-industry

variations in wages, which is important for assessing the wage outcome of the

industrialization process.

3. Empirical model

There has been an increased interest in addressing the impact of international trade and

investment on the wage premium in firm heterogeneity models (Goldberg & Pavcnik 2007).

The Melitz model of firm heterogeneity and international trade (Melitz 2003) proposes that

the labour market outcomes of trade mainly rely on the degree to which the firm is involved

with globalization.2 Recently, a study by Amiti and Davis (2011) explores a Melitz-type

model with an incorporation of a fair wage approach (Egger & Kreickemeier 2009) and

establishes a link between firm heterogeneity and the wage premium. Their key proposition is

that large productive firms that tend to be involved with exporting and importing are likely to

Page 9

The dependent variable (Ws/Wu) is the wage premium which in turn is the ratio of an average

wage of skilled workers to that of unskilled workers. As is standard in many previous studies

(Hanson & Harrison 1999; Pavcnik et al. 2004), non-production workers are a proxy for

skilled labour and similarly, production workers for unskilled labour. Then, the ratio of the

average wage of non-production workers (white-collar) and production workers (blue-collar)

is used as a measure of the wage premium. Although the white-collar/blue-collar

classification does not capture perfectly skill levels measured by education attainment, the

usage of either measure as a dependent variable on the wage premium model brings about

comparable results in many empirical studies (Krueger 1997; Slaughter 2000).

In this analysis, the prime explanatory variables are tariffs, ownership dummies, and

export orientation. First, tariff variables are used to examine the effect of trade liberalization

on the wage premium. Both input (IT) and output tariffs (OT) are incorporated separately

because a firm’s wage outcome for a specific tariff reduction depends on whether the firm has

been involved in exporting or importing (Bernard et al. 2007). The reduction in output tariffs

in a developing country will increase wages in export firms that have a higher proportion of

unskilled workers. Meanwhile, it will reduce wages in import substitution firms, resulting in a

decline in the wage premium. Put simply, these effects of output tariff reductions are in line

with the standard HOSS trade theory. In a similar manner, a lower input tariff will affect

wages in firms that depend on intermediate inputs because of their lower prices relative to

those of domestically produced inputs. Given the higher skill-intensive production of

intermediate inputs relative to those of final product production in the developing world,

reducing input tariffs is likely to increase an import of intermediate inputs. As a result, firms

will reallocate resources to higher unskilled-intensive production that helps to narrow the

wage premium. They also will lay off skilled workers, reducing the wage premium.

Moreover, the inclusion of the input tariff is very much relevant to the labour-abundant

identify a significant difference in wages among three investment forms of FIEs in

Vietnamese manufacturing. Thus, it is important to include these three dummies, rather than

one foreign ownership dummy in this analysis of the wage premium in Vietnam. This is

because FIEs generally have a tendency to employ more skilled workers in Asian developing

economies (Lipsey 2004; Lipsey & Sjöholm 2004; Ramstetter 2004; Ramstetter & Sjöholm

2006).

Aside from the firm ownership feature, the vector Z in Equation (1) controls for other

firm characteristics. Based on the increasingly important role of firm-heterogeneity (Melitz

2003), several firm-specific factors are incorporated in the model. These include real output

(RQ), capital intensity (KL), gender ratio (GR), and skill share (SS).

In addition to firm specific factors, two additional vectors are included. The first one is

a variable (REG) presenting regional feature. Generally, capital-intensive industries are likely

to be concentrated in a region with a high level of economic development, leading to the

expansion of the relative demand for skilled workers in that region. Accordingly, the wage

premium will increase. As suggested by the geographic pattern of wages, three regional

dummies for the Red River Delta, the South East Area and the rest of Vietnam (a base

dummy) are employed. In addition to the regional effects, a variable (INS) of 21 industry

dummies at the two-digit VSIC level is incorporated to capture specific-industry effects. This

controls for the possible differences of the relative supply of skilled labour across industries

as well as other unobserved industry heteroskedasticity.

4. Data and variable construction

The main data in this study is the firm-level dataset compiled from the unpublished returns of

the Enterprise Surveys undertaken by the General Statistical Office of Vietnam (GSO).

Estimating this wage premium model using a panel dataset by pooling cross-firm and time-

series data is ideal, especially when our key interest is the wage premium. Unfortunately, this

preferred data is not available, given the nature of the Enterprise Surveys of Vietnam. To date,

there is only one Enterprise Survey of 2009 that has comprehensive information on

employment and wages by education and production – non-production workers. In that

Page 11

skilled labour. Production workers are treated as unskilled labour. Owing to the data on wages

classified by occupations, the wage bill used for examining the wage premium includes only

regular salary and other cash payments made to employees. In other words, wages do not

cover the employer’s contribution to social and health insurance funds.

With regard to demographic characteristics of workers, the data provide information on

gender, educational attainment, and occupation. Using the educational information on the

highest completed grade, skill share (SS) is measured as the proportion of the total workers

who completed tertiary education (college/university degree) in each firm. It is important to

note that this measurement of skill share may not totally match the ratio of non-production

and production workers. This is because some skilled workers are hidden in the production

worker category. (It would be preferable if wages could be broken down into categories by

educational attainment but this kind of data is not available). Finally, a gender ratio (GR) is

measured by the ratio of female to male workers.

The data from Enterprise Survey, which covers all registered formal firms, have been

cleaned and observations with non-positive values eliminated, along with outliers. For each

firm, the survey also provides information on gross output, and capital stock and profits. All

data in normal terms were converted to real term, using appropriate deflators. Additionally,

the dataset has four-digit industry classification codes (VSIC) consisting of 110

manufacturing industries that allow us to match industries with the tariff data. Note that many

small firms report implausible or unrealistic data due to their poor information and accounting

i=1

where tariffs are output tariff on the final product the four-digit

The weights aij are based on input coefficients from the Input-Output table (I-O table) of

2007, the latest I-O table in which comprehensive information on specific intermediate inputs

is available. Based on the concordance between I-O industry codes and industry classification

(VSIC) provided by GSO, we obtain the input tariff for each industry. Then, both the tariffs

Page 12

on final products and intermediate inputs are constructed at the four-digit VSIC level in order

to merge them with the firm-level data.

5. Results

This section presents the OLS estimation results to explore the determinants of the wage

premium. In order to control for any heteroskedasticity in the error terms, statistical

significance of the estimated coefficients is tested using standard errors based on the robust

variance-covariance matrix estimator (Wooldridge 2002). Summary statistics as well as a

correlation matrix of the variables are reported in Tables 2 and 3.

[Tables 2 and 3 around here]

The results for five alternative specifications are shown in Table 4. These specifications

are different from each other in the following features. The full model (Equation 1) is shown

in the first column, whereas the second column reports the specification excluding a skill

share variable. The third specification includes only a Red River Delta dummy, then the

fourth replaces that with the South East Area dummy. The specification without control of

industry dummies is presented in the last column.

[Table 4 around here]

A reduction in tariffs on intermediate inputs is likely to be associated with a narrowing

of the wage premium since the coefficients of input tariff (IT) are positive and significant at

conventional levels in most cases. Holding other factors unchanged, a ten-percentage point

reduction in input tariff is likely to result in a reduction of seven per cent in the wage

premium in the manufacturing sector. There is strong statistical evidence of an intra-industry

effect on the wage premium from lowering the input tariff. This finding is well in line with

Vietnam. As in our own analysis based on the I-O tables of Vietnam,6 a high proportion of

labour-intensive manufacturing exports in the Vietnamese economy over the period 2000-09

draws heavily on imported intermediate inputs which are viewed as more skilled labour-

intense. Lower input tariffs will stimulate a higher demand for imported inputs, leading to an

expansion in demand for unskilled workers in traditional labour-intensive industries such as

apparel, footwear, and furniture. For this effect, the wage premium is likely to decline.

Page 13

Next, lowering output tariffs also tends to narrow the wage premium. The positive sign

and significance of output tariff (OT) across various specifications implies that cutting output

tariffs tends to narrow the wage premium. The model predicts that a ten per cent fall in the

output tariff results in a three per cent decline in the wage inequality between skilled and

unskilled workers, holding other factors constant. This result is highly consistent with the

Heckscher-Ohlin trade model; opening up to international trade will cause each country to

specialize in industries that use the country’s most abundant factor more intensively. This

result is especially relevant in the case of the Vietnamese economy, where a massive pool of

unskilled workers has not been depleted. Trade liberalization adopted since the second half of

1990s has stimulated resources reallocation toward unskilled labour-intensive industries.

Consequently, export expansion resulting from the reduction and elimination of tariffs has

increased the price of unskilled labour-intensive goods, causing a rise in wages of unskilled

workers. All together, the significance of both tariffs implies a synergy in the effects of trade

liberalization on the wage premium.

In addition, export orientation appears to have widened the wage premium. The

coefficient of an export orientation (EO) is statistically insignificant in many cases, suggesting

that export orientation is less likely to be related to the wage premium. However, as shown in

Column 5, the coefficient on the EO variable is highly significant with a positive sign after

excluding two-digit industry dummies. Holding other effects constant, higher export

expansion is possibly associated with an increase in the wage premium between skilled and

unskilled labour. This finding runs counter to the prediction of the HOSS model but is

consistent with the abundance of unskilled labour in Vietnam. Export production in

Vietnamese manufacturing employs a large proportion of unskilled workers. As the country

has been abundant in unskilled labour,7 thus the wage premium can increase as implied by the

Lewis-Fei-Ranis model. Thus, there is the possibility that a rise in exports contributes to a

Page 14

consistent with the widely held proposition in the model of foreign investment and relative

wages (Feenstra & Hanson 1996; 1997).

Given the greater participation of FIEs in labour-intensive industries, it is likely that

export-oriented foreign investment, particularly wholly owned FIEs in export-oriented

manufacturing, has played a significant role in increasing the wage inequality between skilled

and unskilled labour. As revealed in trade orientation in Figure 1, joint ventures with private

firms and wholly owned FIEs tend to be concentrated on labour-intensive manufacturing for

exports, as above 40 per cent of their output in 2009 was in export-oriented manufacturing

industries. On the other hand, nearly 80 per cent of SOE total output is involved with

domestically oriented industries. Within export-oriented manufacturing, one half of

employment has been attributed to the wholly-owned FIEs which account for over a third of

total manufacturing output in 2009 (Figure 2). In summary, export-oriented foreign

investment has been the backbone for output growth and employment expansion in

Vietnamese manufacturing.

[Figure 1 and Figure 2 around here]

Other determinants of wage premium such as KL, RQ, and GR are statistically

significant with the expected signs. As a representation of firm’s characteristics, firm output –

a proxy for firm size – is positive and significant at the five per cent level or better in all

cases. Ceteris paribus, larger firms tend to have a widening wage premium. Moreover, the

coefficient of capital intensity (KL) is positive and significant at conventional levels. This

result is straightforward as firms with a higher proportion of capital are likely to recruit more

skilled workers. As a result, an expansion of demand for skilled workers results in a widening

of the wage premium.

In regard to worker’s characteristics, the coefficient of the gender ratio (GR) is highly

significant with an expected (positive) sign, suggesting a higher share of female workers is

likely to cause a rise in the wage premium. This is explained by the fact that a high gender

ratio is associated with a higher proportion of unskilled workers, implying that they will have

As shown in the first column in Table 4, the coefficient of skill share (SS) is significant but

carries an unexpected (negative) sign in the full model. This result is contrary to the normal

expectation that this skill ratio would contribute to the wage premium. As previously noted in

the discussion on the data compilation, this unexpected finding could be the result of

Page 15

measurement error. Wages for production workers – a proxy for unskilled workers – may

partly include those of skilled workers in the production work, leading to a magnification of

the wage for unskilled workers in the wage premium. Testing was undertaken for the possible

effects of this issue on the other explanatory variables. The results in the second specification

are significantly robust to the exclusion, given that it does not affect the significance level and

sign of coefficients on various explanatory variables.

The coefficients of the regional dummies are highly significant in the full model. The

positive coefficient of the South East Area (Sea) including Ho Chi Minh City implies that

firms operating in that region have a higher wage premium than those located in other less

urbanized regions, holding other effects constant. A similar comment also holds for those

firms located in the Red River Delta (Rrd). A plausible reason for these effects is that these

two regions have many advantages as compared to the rest of Vietnam in terms of market

competition, availability of resources, and business environment. It is important to note that

there is a notable difference between these two most urbanized regions. We conduct a test of

the effect of that difference on the wage premium, as shown in the Columns 3 and 4.

Controlling for the South East Area, the coefficient is still positive and significant while that

for the Red River Delta becomes insignificant. Clearly the evidence of the higher wage

premium for firms operating in the South East Area may be associated with the fact that the

region has been more involved with market-based disciplines than has the North.

6. Comparison results

The preceding section provides the estimation results examining the effect of trade and

investment on the wage premium. For the purpose of comparison, an estimation of the model

that relates manufacturing wages to the similar explanatory variables is also undertaken. This

analysis is based on the following model.

����������������=β0+β1ITij+β2OTij+β3DPij+β4JVSij+ β5JVPij+ β6WFIEij