Ảnh hưởng của nguồn gốc FDI đến tình hình việc làm tại Việt Nam

Working Paper 2021.1.3.04

- Vol 1, No 3

ẢNH HƯỞNG CỦA NGUỒN GỐC FDI ĐẾN

TÌNH HÌNH VIỆC LÀM TẠI VIỆT NAM

Bùi Tùng Dương¹, Nguyễn Phương Thảo, Lê Hồ Anh Thư

Sinh viên CTTT KT - K57 – Viện KT & KDQT

Trường Đại học Ngoại thương, Hà Nội, Việt Nam

Vương Hiển Minh

Sinh viên KTĐN – K57 – Viện KT & KDQT

Trường Đại học Ngoại thương, Hà Nội, Việt Nam

Cao Thị Hồng Vinh

Giảng viên Viện Kinh tế và Kinh doanh Quốc tế

Trường Đại học Ngoại thương, Hà Nội, Việt Nam

Tóm tắt

Trong bài viết này, chúng tôi tập trung vào những câu hỏi sau: Ảnh hưởng của FDI lên ba yếu tố

của người lao động bao gồm tiền lương, lượng công việc tạo ra và lợi ích bảo hiểm? Nguồn gốc của

FDI có ảnh hưởng đến mối quan hệ trên hay không? Nếu có, ảnh hưởng như thế nào? Dữ liệu điều

tra doanh nghiệp tại Việt Nam năm 2017 và 2018 được phân tích sử dụng mô hình tác động cố định.

Chúng tôi nhận thấy khi kiểm soát tương tác giữa tỷ lệ yếu tố nước ngoài và biến kiểm soát, ảnh

hưởng của FDI lên lao động trở nên đáng kể. FDI có xu hướng làm tăng lương trung bình và giảm

số lượng việc làm tạo ra cũng như số lượng lao động được nhận bảo hiểm. Xét thêm yếu tố nguồn

gốc nguồn vốn đầu tư, ảnh hưởng của FDI lên lao động là khác nhau khi FDI đến từ các quốc gia

khác nhau, tương tự như kết quả các nghiên cứu trước.

Từ khóa: FDI, nguồn gốc FDI, Lao động, Tiền lương, Việc làm, Lợi ích bảo hiểm

THE EFFECT OF FDI ORIGIN ON

ASPECTS OF EMPLOYMENT IN VIETNAM

Abstract

In this paper, we focus on these questions: What are FDI effects on three aspects of employment

including wage, job creation and insurance benefits? Could country of origin moderate these

effects? And if it could, to what extent could it affect? Firm level data in Vietnam are analyzed for

2017 and 2018 using fixed-effects model. We find that when controlling for the interaction between

foreign share and control variables, FDI impacts on employment aspects change from insignificant

¹Tác giả liên hệ, Email: tungduongbui167@gmail

FTU Working Paper Series, Vol. 1 No. 3 (06/2021) | 45

into significant. FDI exerts an upward trend on the average wages and a downward trend for jobs

created and the number of workers receiving insurance. Regarding to country of origin, these

impacts on employment factors vary across FDI source, corroborating previous studies.

Keywords: FDI, FDI origin, Employment, Wage, Job, Insurance benefit.

1. Introduction

Foreign direct investment (FDI) is defined as a form of international capital movement with the

purpose of establishing and maintaining permanent equity relations with the foreign company at

the same time to exercise a noticeable influence on the management of the company

(Golejewska, 2002). Despite being perceived as a source of foreign influence and of competition

with local enterprises (Blomstrom et al.,1997), the attraction of foreign investors is still an

important goal of policy makers worldwide, especially in less developed countries where lack

of capital is one of the key constraints to economic prosperity (Coniglio et al., 2015).

The dramatic increase in FDI flows throughout the globe has led to the attention on its impact

on the host countries (Ni et al., 2017). In the field of International Business, a lot of empirical

studies have been devoted to understanding the effects of foreign investors on host countries’ wage

and employment; yet, the results of these studies are mixed and their evidence is still far from

conclusive. The divergence in empirical findings can be partly attributed to the methodological

issues and the characteristics of the host countries, industries and firms (Fortanier, 2007).

However, one factor that contributes to the relationship between FDI and host countries' wage and

employment have so far received little attention: the heterogeneity of FDI itself (Chen, 2011).

In the field of Economics, FDI is usually assumed as homogeneous flows of capital, thus, the

wage and employment effects are the same for all types of FDI. Rather, FDI differs by various

characteristics, such as by the size and entry mode, the role in the global value chain, the aim of

investment... but mostly related to firm performances (Chen, 2011). This article examines whether

the heterogeneous characteristics of FDI in empirical study enhances our understanding of the

impact of FDI: i.e. whether and to what extent the origin of investors affects the wage and non-

wage rate and the job creation rate of the host economies.

In this study, we focus on the role of one characteristic of FDI: its country of origin. Early

research proves that the FDI country of origin’s market conditions, business and institutional

systems (Whitley, 1998) have substantial influences on the strategic and organizational

characteristics of multinationals including human resource management practices (Bae et al.,

1998). Therefore, we form a hypothesis that FDI from different countries should also have

different wage, non-wage and employment effects on the host nations.

To test this and contribute further to the literature, we investigate an important emerging

economy, Vietnam. Since the late 1990s, Vietnam has experienced a significant surge inward

foreign direct investment (FDI) owing to the adoption of a major economic reform known as Doi

Moi in 1986 followed by enactment of the Law on Foreign Direct Investment in 1987 and its

accession to the World Trade Organization (WTO) in 2006. The rapid growth in inward FDI to

Vietnam has a positive impact on the registered foreign capital (GSO, 2017) and net export (Anwar

and Nguyen, 2011), thus leading to the increase in economic growth (Hoang et al., 2010; Vu et

al., 2008).

FTU Working Paper Series, Vol. 1 No. 3 (06/2021) | 46

We study the effect of FDI and FDI origin on three aspects of employment in Vietnam

including wage, job creation and insurance. Although a few studies have examined FDI effects on

wage and employment in Vietnam (Nguyen, 2019; Hoi & Pomfret, 2010), this paper provides

some of the first findings on the differences in the effects of FDI level on wage, job and insurance

of domestic firms under the perspective of investor origin heterogeneity. The findings from this

paper are expected to have significant implications for evaluating and selecting the suitable foreign

investors to attract based on their impact on domestic firms' employment and the economy at large.

The following empirical analyses employ a firm-level dataset in over 2000 industries in

Vietnam during a two-year period from 2017 to 2018. The data were obtained from comprehensive

surveys commissioned by the General Statistics Office of Vietnam (GSO), including

questionnaires collecting information on enterprises in Vietnam including State enterprises, non-

state enterprises, enterprises that have foreign investment, cooperatives/consortium cooperatives.

The rest of the paper is structured as follows. Section 2 reviews the literature on effects of FDI

and FDI origin on employment. Section 3 specifies the method of data collection, empirical model

and dataset description. Section 4 presents empirical results and relevant analyses. Finally, Section

5 gives concluding remarks.

2. Literature Review

The literature regarding FDI and its effect on host countries’ employment has been well

documented. In this section, we summarize the effect of foreign ownership and FDI origins on

host countries’ employment including wage, job and insurance as well as the moderating effect of

country of origin that have been so far taken into consideration.

2.1. The wage effects of FDI and its country of origin.

2.1.1. FDI Wage effects

There is a large body of literature on the wage effect of foreign investment in host countries,

broadly classified into two main grounds: (1) foreign wage differentials and (2) foreign wage

spillover (Brown et al., 2003; Lipsey, 2004). Regarding the former, empirical evidence suggests

that foreign firms pay higher wages than their domestic counterparts within host countries, within

industries and regions in these countries even after detailed differences in firm characteristics like

capital intensity, size, location, industry features and educational level of workers are taken into

account (Conyon et al., 2003, Görg et al., 2007; Huang et al., 2017). Studies conducted in

developed countries estimate that average wages paid by foreign establishments are approximately

6-22% higher in the United States (Feliciano & Lipsey, 2006; Lipsey, 1994) and 4-26% higher in

the United Kingdom (Conyon et al., 2002; Driffield & Girma, 2003; Girma et al., 2001).

Similarly, the foreign wage premia are also proved to exist and even emphasized to be higher

in developing countries (Egger & Kreickemeier, 2013). In Indonesia, the average wage in foreign

plants is about 50% higher than in private local plants and 60% higher including other types of

labour compensation, such as bonuses, gifts, social security, insurances and pensions (Lipsey &

Sjöholm, 2004). In Venezuela and Mexico, wages in foreign-owned manufacturing establishments

are higher than in domestically owned establishments by 30% (Aitken et al., 1996).

Whilst wage gap estimates between foreign and domestic firms are consistent across existing

literature, the explanations for such results are varied. One common reasoning is that foreign

FTU Working Paper Series, Vol. 1 No. 3 (06/2021) | 47

investors pay higher wages in order to reduce worker turnover and thus, to minimize the risk of

technology and knowledge diffusion through labour mobility (Fosfuri et al., 2001; Glass & Saggi,

2002, Aitken et al., 1996, Balsvik, 2011; Poole, 2013). Other authors have argued that

multinational enterprises offer higher wages to compensate for the possible disadvantages of

employment in an MNE, for example, greater pressure and labor demand volatility (Fabbri et al.,

2003; Gorg & Strobl, 2003), or higher foreign plant closure rate (Javorcik, 2015). Another

motivation for higher wages paid by foreign affiliates can also be explained by rent-sharing across

international borders (Budd & Slaughter, 2004) and between employers and employees (Budd et

al., 2005).

The second strand of studies on the FDI-linked wage effect focuses on the impact of foreign

ownership on the wage rate and wage growth of domestically owned firms. Theoretical studies

show that FDI wage spillovers to domestic firms may be generated through several channels. The

presence of foreign ownership may lead to positive wage spillovers due to the increasing

competition in the labour market. This shifts up the labor demand curve and thus, obliges local

firms to increase their wage rates to attract and retain workers, especially high-skilled labour

(Aitken et al., 1996). Technological spillovers are also an important channel of FDI wage

spillovers. Technological externalities transferred through imitation/demonstration effect, labour

mobility or horizontal and vertical linkages (Crespo & Fontoura, 2007; Hoi & Pomfret, 2010) may

increase productivity and possibly, wage level of domestic firms (Javorcik, 2004; Görg &

Greenaway, 2004).

On the other hand, FIEs may recruit the best workers from domestic firms or acquire high

wage local firms, thus lowering the labor quality and the wage rate of local firms (Lipsey &

Sjöholm, 2004). Also, foreign participation in product markets may lead to lower scale of

production of domestic firms, reduce their market shares and even crowd them out (Aitken &

Harrison, 1999; Kosová, 2010).

The empirical evidence from existing literature shows mixed results of FDI wage spillovers.

Some authors prove that local workers are better off by FDI wage spillovers in a wide range of

countries, such as Indonesia (Lipsey and Sjoholm, 2004), The United Kingdom (Driffield &

Girma, 2003), The United States (Aitken et al., 1996), Poland (Bedi and Cieslik, 2002). Some

other researchers find negative effects of foreign presence on wage levels of domestic firms (Barry,

et al., 2005; Hu & Jefferson, 2002). Finally, some studies find no evidence of wage spillover from

FDI to domestic establishments, for example, in the United States (Feliciano & Lipsey, 2006), in

Mexico and Venezuela (Aitken et al.,1996), and in the United Kingdom (Girma et al., 2001).

2.1.2. Wage Effects of FDI Country of origin.

A growing body of research has been done to examine the moderating effect of country of origin

on FDI linked wage effects, returning mixed results. In brief, these papers show that FDI origin

may play an important role in determining wage impact of FDI.

In the UK, Girma & Görg (2007) find evidence for significant positive wage effects resulting

from acquisitions from US companies, however, EU counterparts may not bring any impact. In

China, Liu et al. (2015) argue that while takeovers from North America and Europe put an upward

effect on wages, this seems negligible for HMT (Hongkong, Macau, Taiwan) and JKS (Japan,

Korea, Singapore) subsamples. This is maybe because North American and European acquirers

FTU Working Paper Series, Vol. 1 No. 3 (06/2021) | 48

possess higher technological intensity thus avoiding labor turnover related technology diffusions

by paying premium.

Similarly, the studies on sub-Saharan African firms return findings supporting the moderating

impact of parents origin (Coniglio et al., 2015; Blanas et al.,2019). In terms of development level,

MNEs from developed countries may be correlated with higher average wages than those from

less-developed counterparts (Coniglio et al., 2015). Relating to geographic area, the average wages

paid by Chinese investors seems lower than other counterparts as Chinese MNEs may highly

demand low-quality workers, thus offering lower wages and/or compared with others, their

locations tend to be remote from urban areas taking less labor costs (Coniglio et al., 2015).

Meanwhile, Blanas et al. (2019) argue that those from outside sub-Saharan Africa may pay higher

than local counterparts, regardless of labor type. However, these results for those from sub-Saharan

Africa seem true for only managerial and non-production laborers.

In sum, the impact of FDI source on FDI wage effects has increasingly grabbed scholars’

attention with diverse findings supporting that a heterogeneity of FDI induced wage effect may

come from the difference of FDI origins. However, in Vietnam it still has been negligibly

considered and we expect that the wage impact of FDI differs by FDI source.

2.2. Job creation effects of FDI and its country of origin

2.2.1. FDI Job creation effects

Employment creation has been regarded as one of the potential contributions of inward FDI to host

countries. However, most analyses on the influence of FDI on employment identify both positive

and negative potential effects (Jenkins, 2006; Rama, 2003; UNCTAD, 1994). FDI can increase the

local labour demand directly by establishing new greenfield subsidiaries (Rama, 2003) or even

expanding existing ones (ILO, 1984). FDI can also lead to increased volume of employment

through spurring forward and backward linkages (Golejewska, 2002; Ernst, 2005; Liu et al., 2009).

These “crowding in” effects may endure if these foreign firms make long-term commitments to

the host countries.

On the other hand, there is evidence that FDI also generates negative effects on host

economies' employment (Jenkins, 2006; Rama, 2003; UNCTAD, 1994). FDI may crowd out non-

competitive local firms, leading to job losses for the host economies. According to Jenkins (2006),

the reduction in volume of employment may also be associated with FDI involving the acquisition

of local firms and application of labour-saving technologies. Moreover, as multinationals are

footloose and able to relocate production and employment between their affiliates in different

countries, jobs created are likely to be highly unstable.

There is a number of empirical literature on the effect of FDI on employment creation in

both developed and developing countries. Most of the findings about developed countries point

out that firm-level employment remains unchanged or increases after foreign acquisition, for

example in the UK (Girma, 2005), Sweden (Bandick & Karpaty, 2007), Norway (Balsvik &

Haller; 2010). Studies about developing countries show that foreign firms, on average, grew

more rapidly (Lipsey et al., 2010) and have larger numbers of workers than domestically owned

firms (Barthel et al., 2011).

2.2.2. Job creation effects of FDI Country of origin.

FTU Working Paper Series, Vol. 1 No. 3 (06/2021) | 49

The study of the relation between country of origin and employment effect of FDI has increasingly

flourished in recent years. Overall, the nationality of FDI is likely to play a pivotal role in

determining the quantity of jobs created via FDI.

Regarding to development level, acquirers from developing countries may be negatively

associated with labor demand of US acquired firms, however, those from foreign industrial

countries tend to put an upward effect (Chen, 2011). Conversely, in sub-Saharan Africa, investors

from developing countries tend to generate less-skilled labors compared with domestic firms while

this impact of those from developed countries seems insignificant (Coniglio et al., 2015).

Relating to geography, in sub-Saharan Africa, South FDI (from other African countries) seems

more beneficial for employment growth of targets than North FDI (from the remaining) as the

technological or business climate dissimilarity between recipients and African investors seems

smaller compared to South counterparts (Gold et al., 2017). In China, HMT and JKS acquirers

may enhance job creation of the acquired while the effect of North America and Europe seems

insignificant because compared to others, HMT and JKS tend to enlarge workforce to develop

their business (Liu et al., 2015). Furthermore, Coniglio et al. (2015) reveal that Chinese investors

may generate more jobs (mainly less-skilled) than other foreign counterparts.

Generally, a growing quantity of researches have proved that job creating impact of FDI

differs depending on investors’ nationalities. Hence, taking Vietnam into account, we argue that

country of origin may attribute to the dissimilarity of job creation effect of FDI.

2.3. Insurance benefit effects of FDI and its country of origin.

While the effects of FDI on wage compensation have been examined in a growing body of papers,

there has been still a lack of studies relating to FDI-linked impacts on non-wage compensation as

well as the correlation between FDI source and these effects. To the best of our knowledge, the

study done by Eren & Peoples (2013) is presumably the only one researching this issue. Using

information from non-manufacturing industries in the US, they argue that FDI activity is generally

associated with higher likelihood of laborers acquiring non-wage compensation including pension

and health insurance from employers regardless of their education level. However, this result

seems significant for male workers and high-educated female counterparts. To sum up, could FDI

affect insurance benefits of workers? And if it could, to what extend the country of origin would

moderate this relation? The answers have been still unclear, leaving room for more research.

3. Methodology

3.1. Data collection

We obtain data from the 2017 and 2018 surveys of the General Statistic Office of Vietnam. The

survey includes questionnaires collecting information on enterprises in Vietnam including State

enterprises,

non-state

enterprises,

enterprises

that

have

foreign

investment,

cooperatives/consortium cooperatives. The survey was randomly distributed to firms all around

Vietnam.

3.2. Models and Methodology of analysis

3.2.2. Statistical model

3.2.2.1. Models

FTU Working Paper Series, Vol. 1 No. 3 (06/2021) | 50

This paper aims to answer 2 questions:

What is the effect of FDI on aspects of employment including wage, job, and insurance

benefits?

And what is the effect of FDI from a specific country on the aspect of employment?

To address the questions, we estimate panel regression models based on the previous empirical

studies (Abor.J., and Harvey.S. K., 2008; Nguyen., 2019; Aitken et al., 1996; Ahmad Seyf., 2000;

Robert E. Lipsey et al., 2010). Therefore, the model of specifications of this paper is presented as

follows:

lnwage_it= αit + β1 FDIpercentage_it+ ∑βjControls_it+ µ_it(1)

lnjob_it= α_it+ β1 FDIpercentage_it+ ∑βjControls_it+ µ_it(2)

lnworker_insurance_it= α_it+ β1 FDIpercentage_it+ ∑βjControls_it+ µ_it(3)

(1) lnwage measured as the average wages paid to one labor in natural log, (2)

lnworker_insurance measured as the number of workers that receive insurance measured in natural

log, and (3) lnjob measured as the number of jobs measured in natural log.

3.2.2.2. Variable of interest

To address the first question, our variable of interest is FDI percentage which is defined as total

charter capital of FDI divided by total charter capital of the firm multiplied by 100% (Thanapol,

2012)

For the second question, the variable of interest is the dummy variable showing whether firms

receive FDI from a specific country or not. In our analysis, we regress on 6 countries that pour the

FDI in the most firms in our sample. Those include China (CN), Singapore (SG), Hong Kong

(HK), Taiwan (TW), South Korea (KR), Japan (JP)

3.2.2.3. Control variables

Although our data is collected through a randomization process. This can suggest our result will

be externally valid which means that the results can be applied for all firms in Vietnam in general.

However, for internal validity, the serious problem that may arise in our result is omitted variable

bias. Hence, we will have to control for variables that are associated with the change in FDI origin

and our 3 dependent variables. These include:

Lnage - Firm age - number of years firm in operation up to 2021 in natural log. The well-

established firms are expected to have a positive impact on wages as they gain a more secure

foothold in the product and labor markets, indicating their business success and strong paying

capacity. (Nguyen, 2019; Bullon et al., 2013),

Lnsize - firm size – previous studies measure size by total sales (Nguyen, 2019) or the number

of employees (Feliciano et al., 2006). However, different industries have unique characteristics in

sales so when comparing firms across industries, the measure do not correctly reflect the size of

the firms. Therefore, we use size which is measured by the total sales of the firm in a year divided

by the average sale in that industry in a year in natural log (lnsize) to correctly reflect the size of

the firm in a specific industry. The idea is similar to the relative size of firm within its industry.

Some firms have sales below the industry sales. The problem of this measure is that the size will

FTU Working Paper Series, Vol. 1 No. 3 (06/2021) | 51

not be correct in a small sample size as we may wrongly measure the average sale in the industry.

However, our sample is large enough so we can ignore the problem.

Lncapital_int - The capital intensity of the firm which is measured by fixed assets divided by

total employment (Nguyen, 2020). Fixed asset is calculated as the original price minus

depreciation. This variable is in natural log.

Year dummy - whether it is in 2017 or 2018.

Region dummy - at which province the firm is located, province code is collected from the

General Statistic Office of Vietnam (province dummy),

Industry dummy - which measure the main industry that firm operate in, Industry code is

collected according to VSIC 2007 _ 3 - level digit (Industry_3digit Dummy),

Subscript i is the firm index, subscript t is the year index, and µ is the error term.

3.2.2.4. Econometric strategy

Stata 14 is used for our data analysis.

Due to the characteristic of our data set being 2 years of panel data, we can use the Fixed-

effect model (FEM), Random effect model (REM), and Pooled OLS model. Fundamentally, FEM

and REM are developed to control the effect of time-invariant variables that are often unobservable

or too complex to measure.

In choosing what is the best model for our estimations of panel data. We follow the process

of Dougherty (2011).

Figure 1. Process of choosing regression model for panel data

Source: Dougherty (2011)

We run 3 different models which are the Fixed-effect model (FEM), Random effect model

(REM), and Pooled OLS model. Then, since the observations in our sample are randomly selected,

FTU Working Paper Series, Vol. 1 No. 3 (06/2021) | 52

we use the Hausman test to test whether there is a significant difference between the FEM and

REM. If the result rejects the null hypothesis that there is no systematic difference between the 2

approaches, we will use the FEM. If not, we continue to run the Breusch and Pagan Lagrangian

Multiplier test to choose between the Pooled OLS model and REM.

Furthermore, Difference and Difference (DiD) is a common approach that some articles have

used for analyzing the effect of FDI on aspects of employment, prominently Girma (2005) and

Hijzen et. al (2013). However, since our dataset is only in 2 years, it would be difficult to prove

the parallel trend assumption of DiD in which firms with FDI and firms without FDI need to show

similar trends throughout history. Most importantly, one assumption of DiD is there needs to be

no spillover between the 2 groups: firms with FDI and firms without FDI. However, there is likely

to have spillover especially in wages among firms.

3.3. Data description

This survey provides information about the firm’s characteristics and operation indicators

including location, the amount of foreign investment, the country from which investment comes,

sales, employment, labor policy, firm size, firm age and assets.

The data consists of 969,227 observations in 2 consecutive years: 2017 and 2018. There are

376,043 observations in 2018 and 593,184 observations in 2017. After cleaning our data set, we

have the final sample of 207,847 observations across 2-year period.

Table 1 represents the descriptive statistics. The figures in the table suggest considerable

variations within continuous variables, indicating the heterogeneity of firms in the sample.

Regarding the level of FDI poured into domestic firms, the average FDIpercentage among all firms

in our sample is 10.86%. The maximum amount of FDI percentage is 100 percent, indicating a

completely foreign-invested firm, and the minimum level is 0 percent, implying no FDI presence.

Since some firms have their sales below the industry average sales, this will make the relative size

less than 1 and make the value of firm size in natural log below 0.

Table 1. Summary of key variables

Variables

Obs

Mean

4.0411

2.7598

2.8216

10.8666

-1.2911

2.2713

11.4204

Std. Dev.

0.8997

1.593

Min

Max

11.8983

11.1604

11.1409

100

lnwage

204, 166

207, 792

114, 825

207,795

206,735

207,847

136,723

0

lnjob

0

lnworker_insurance

1.7312

30.714

2.2703

0.5299

1.8318

FDIpercentage

0

lnsize

-14.2112

1.0986

0.3087

8.2763

4.8441

20.1167

lnage

lncapital_int

Source: Calculations from software

Table 2 present which country of origin firm often receive FDI the most. The top 6 countries

are Singapore, Taiwan, China, South Korea, Hong Kong, and Japan. In raking the total value of

FDI country pour into firm in our sample. The order slighly change when Thailand and UK lies in

top 6.

FTU Working Paper Series, Vol. 1 No. 3 (06/2021) | 53

Table 2. Top 10 countries with the highest FDI poured into Vietnam.

FDI in VN

Country

Numbers of firms

(millions VND)

Korean

48,240,681

5,501,000

901,000

2,741

2,352

1,857

1,013

664

Taiwan

Japan

China

2,739,764

13,241,188

7,529,000

274,870

Singapore

Hong Kong

USA

479

249

Thailand

6,045,000

866,467

246

Malaysia

199

The United Kingdom

4. Results

4.1. Correlation matrix

4,216,301

167

We will begin by checking the multicollinearity problem in our model. If our independent

variables are highly correlated with one another, then it would be difficult to measure the actual

effect of the FDI percentage on wage, job, and insurance benefits holding constant other variables.

We will then imprecisely estimate our coefficients. The Correlation Matrix is presented in table 5

below. We can see that all our regressors are barely correlated with the highest correlation is 21.2%

which is lower than 80%. Therefore, we come to conclusion that the multicollinearity problem

does not pose a serious threat to our estimations.

Table 3. Correlation Matrix

FDIpercentage

1

lnage

lnsize

lncapital_int

FDIpercentage

lnage

-0.0707

0.1423

1

lnsize

0.212

0.063

1

lncapital_int

0.0446

0.0045

1

Source: Calculations from software

4.2. Regression results

To understand the effect of FDIpercentage on wage, job, and insurance benefit, the multiple

regression model is adopted. Respectively, we run the Pooled OLS model, Fixed Effect Model,

and Random Effect Model with the result shown in table 6.

4.2.1. Effects of FDI on wages, job creation and insurance benefits

FTU Working Paper Series, Vol. 1 No. 3 (06/2021) | 54

Considering the effect on wage, the coefficient of the variable of interest FDIpercentage is

significant in all three models and all have a positive sign. In the FEM, the coefficient is significant

at 10% level.

When the dependent variable is job and insurance benefits, the coefficient of FDIpercentage

is significantly positive in Pooled OLS and REM while it is negative and insignificant in the FEM.

For lnsize and lncaptial_int, the coefficients are consistent in all three models when measuring

the effect on wage, job creations and insurance benefits.

However, lnage become insignificant and change its direction in FEM when measuring its

effect on wage, job creations and insurance benefit.

FTU Working Paper Series, Vol. 1 No. 3 (06/2021) | 55

Table 4. Estimation results (Pooled OLS Model, Fixed-effect Model, Random effect Model)

Wages

Job

Insurance benefits

FEM

Independent

variables

Pooled

OLS

Pooled

OLS

Pooled

OLS

FEM

REM

FEM

REM

0.0034***

(0.00006)

0.0013*

(0.00007)

0.0034***

(0.00007)

0.0023***

(0.00008)

-0.00009

(0.00058)

0.0033***

(0.0001)

0.0041***

(0.0001)

-0.0007

(0.0007)

0.0048***

(0.0001)

FDIpercentage

lnage

-0.0438***

(0.0039)

0.0286

(0.0236)

-0.0429***

(0.0043)

0.2653***

(0.0049)

0.0215

(0.0182)

0.2789***

(0.0055)

0.3418***

(0.0071)

-0.0358

(0.0252)

0.3426***

(0.0079)

0.1543***

(0.00096)

0.169***

(0.0048)

0.1563***

(0.0010)

0.5231***

(0.0012)

0.1961***

(0.0037)

0.4774***

(0.0013)

0.5732***

(0.0019)

0.1566***

(0.0062)

0.5200***

(0.0021)

lnsize

0.0438***

(0.0011)

0.1618***

(0.0034)

0.0506***

(0.0012)

-0.1581***

(0.0014)

-0.3346***

(0.0026)

-0.1884

(0.0014)

-0.0630***

(0.0020)

-0.1205***

(0.0039)

-0.067***

(0.0020)

lncapital_int

Time

(Dummy)

Yes

Province

(Dummy)

Industry

(Dummy)

3.4435***

(0.0433)

1.8766***

(0.2378)

3.3771***

(0.0466)

4.5738***

(0.0530)

6.8335***

(0.1834)

4.7472***

(0.0567)

2.6685***

(0.0870)

4.5152***

(0.2589)

2.5197***

(0.0932)

Constant

Observation

R-Squared

135,486

0.38

135,486

0.101

135,486

0.38

136,365

0.73

136,365

0.106

136,365

0.73

90,970

0.68

90,970

0.027

90,970

0.68

Source: Authors

FTU Working Paper Series, Vol. 1 No. 3 (06/2021) | 56

In conclusion, when we implement different models, the effect of FDIpercentage on aspects

of employment changes in terms of statistical significance and direction. Therefore, the coefficient

of the variable of interest is not consistent.

We continue to implement some tests as mentioned in the Methodology. First, we ran the

Hausman test and witnessed the p-value<0.05 for all 3 cases with different dependent variables.

This means that we will reject the null hypothesis that there is a systematic difference in the

estimation of FEM and REM. Therefore, the FEM is an appropriate choice. To be more certain,

we continue to run the Breusch-Pagan Lagrange Multiplier test, the random effect model is our

choice since the p-value is lower than 0.05 which means we reject the null hypothesis that the

Pooled OLS Model was appropriate. We will then choose to use the random effect model. In

conclusion, FEM is the best choice for our estimation and according to table 6, FDIpercenatge

seems to have no effect on aspects of employment in the FEM. In addition, for the FEM, the

autocorrelation problem is irrelevant since our dataset include only 2 years, we then witness no

endogenity problem through autocorrelation.

When choosing the fixed-effect model, we witness no clear impact of FDIpercenatge on each

of the 3 aspects of employment. We doubt that when countries pour FDI into firms, the effect of

FDI may interact with firm existing characteristics and show no impact on wage, job creation and

insurance beneft. Therefore, we try to exclude those effects to see what the actual effect of FDI on

aspects of employments is.

After including three interaction terms of FDIpercentage with respectively lncapital_int,

lnsize, lnage, the coefficient of FDIpercentage changes from statistically insignificant to

significant even with different dependent variables in table 7. The coefficient of the interaction

term FDIpercentage with lncapital_int and lnsize is significant while the interaction term with

lnage is insignificant. The actual impact of FDI was clear after we included the interaction terms.

For wage, an increase in FDIpercentage by 1% increases wage by 1,09%. When lncapital_int

and lnsize increase by 1%, it reduces the effect of FDIpercentage on wage by respectively 0.07%

and 0.08%. These changes are quite small numerically. The change in age has no clear impact on

the effect of FDIpercentage on wages.

For jobs, an increase by FDIpercentage by 1% reduces the number of jobs by 24,68%. When

lncapital_int increases by 1%, the effect of FDIpercentage on jobs will increase by 0.21%. For

1% increase in lnsize increase the effect of FDIpercentage on job by 0.07%. The effect of the

interaction term with lnsize is quite small.

For insurance benefits, an increase in FDIpercentage by 1% reduces the number of workers

receiving insurance by 0.44%. The coefficient of the interaction term with lncapital_int and lnsize

is significantly positive by quite small in magnitude.

FTU Working Paper Series, Vol. 1 No. 3 (06/2021) | 57

Table 5. Estimation results adding interation terms (Fixed-effect)

Fixed - effect

Job

Independent

Wage

Worker insurance

variables

(1)

(2)

(1)

(2)

(1)

(2)

0.0013*

(-0.00007)

0.0286

0.0109***

(-0.002)

-0.00009

(-0.00058)

0.0215

-0.2468***

(-0.0015)

-0.0007

(-0.0007)

-0.0358

-0.0044**

(-0.0021)

-0.0434

FDIpercentage

Lnage

0.0393***

(-0.0247)

0.1842***

(-0.0052)

0.1743***

(-0.0038)

0.1632***

(-0.01877)

0.1780***

(-0.0039)

(-0.0236)

0.169***

(-0.0048)

0.1618***

(-0.0034)

(-0.0182)

0.1961***

(-0.0037)

-0.3346***

(-0.0026)

(-0.0252)

0.1566***

(-0.0062)

-0.1205***

(-0.0039)

(-0.0265)

0.1346***

(-0.007)

Lnsize

-0.3726***

(-0.0028)

-0.1260***

(-0.0044)

lncapital_int

-0.00067***

0.0021***

0.0002**

FDIpercentage*lncapital_int

(-0.00009)

(-0.00007)

(-0.00009)

-0.00083***

0.00072***

0.00087***

FDIpercentage*lnsize

FDpercentage*lnage

(-0.0001)

(-0.00009)

(-0.0001)

-0.00096

(-0.0007)

0.00018

(-0.00054)

0.00062

(-0.00072)

Time (Dummy)

Yes

Province (Dummy)

Industry (Dummy)

1.8766***

(-0.2378)

135,486

1.7574***

(-0.2388)

135,486

6.8335***

(-0.1834)

136,365

7.1738***

(-0.1815)

136,365

4.5152***

(-0.2589)

90,970

4.5744***

(-0.2609)

90,970

Constant

Observation

R-Squared

0.101

0.103

0.106

0.1

0.027

0.026

Note: ∗∗∗p < .01, ∗∗ p < .05, ∗ p <0.1

Source: Authors

FTU Working Paper Series, Vol. 1 No. 3 (06/2021) | 58

4.2.2. Effect of FDI origin on aspects of employments

After analyzing the general effect of FDI on aspects of employments. We want to take a deeper

look into how FDI from a certain country affect wage, job, and insurance benefit as they change

their level of charter capital in firms

We choose 6 countries that pour FDI in the most firms in the sample. These countries include

China (CN), Singapore (SG), Hong Kong (HK), Taiwan (TW), Korea (KR). Japan (JP).

4.2.2.1. Effect of FDI origin on wages

The table 8 represent how wage change with FDI from different origin. We can see that

FDIpercentage from Hong Kong (HK), Singapore (SG), Japan (JP) have a significantly positive

impact on wages in which HK have the biggest impact.

lnage continues to have no clear impact on wages, the results are consistent among countries.

lnsize and lncapital_int have significantly positive coefficients indicating that an increase in these

variables will increase wages.

For the interaction term, the change in lncapital_int and lnsize will affect how much

FDIpercentage impacts wage in some certain countries. SG, HK, KR, JP witness a small reduction

in the effect of FDIpercentage on wage when increasing lncapital_int since the coefficient of the

interaction term is statistically significant at 10% level.

Some witness a significant coefficient on interaction term of FDIpercenatge with lnsize.

Those include CN, TW, KR, JP, and increas in lnsize reduce the effect of FDIpercenatge on

wages.

The change in age has no clear impact on the effect of FDIpercentage on wage since all

coefficients in 6 cases are statistically insignificant.

FTU Working Paper Series, Vol. 1 No. 3 (06/2021) | 59

Table 6. Effect of FDI origin on wages

Wage

Indepenent variables

CN

SG

HK

TW

KR

JP

-0.00078

0.0111*

0.0153**

0.0017

0.0025

0.0141***

FDIpercentage of each origin

(-0.0037)

0.0279

(-0.0061)

0.0286

(-0.006)

0.0293

(-0.0035)

0.0295

(-0.0029)

0.0287

(-0.0036)

0.0282

lnage

(-0.0236)

0.1703***

(-0.0048)

0.1619***

(-0.0237)

0.1691***

(-0.0048)

0.1626***

(-0.0236)

0.1692***

(-0.0048)

0.1622***

(-0.0237)

0.1708***

(-0.0049)

0.1617***

(-0.0238)

0.1710***

(-0.0049)

0.1634***

(-0.0237)

0.1721***

(-0.0048)

0.1666***

Lnsize

lncapital_int

(-0.0034)

-0.00008

(-0.0002)

(-0.0035)

-0.0008**

(-0.0003)

(-0.0034)

-0.00065*

(-0.0004)

(-0.0035)

0.00005

(-0.0035)

-0.00028**

(-0.00014)

(-0.0035)

-0.0010***

(-0.00016)

FDIpercentage*lncapital_int

(-0.00019)

-0.0004**

(-0.0002)

0.0004

0.00004

(-0.0004)

0.00005

-0.0001

(-0.00038)

-0.003*

-0.0005***

(-0.00019)

-0.0008

-0.00036**

(-0.00018)

-0.00007

-0.0010***

(0.0026)

-0.0002

FDIpercentage*lnsize

FDpercentage*lnage

(-0.0011)

Yes

(-0.0019)

Yes

(-0.0017)

Yes

(-0.0011)

Yes

(-0.001)

Yes

(-0.00125)

Yes

Time (Dummy)

Province (Dummy)

Yes

Industry (Dummy)

Constant

Yes

1.8842***

(-0.2378)

1.8712***

(-0.2379)

1.8821***

(-0.2379)

1.8962***

(-0.2378)

1.8785***

(-0.2381)

1.8405***

(-0.2377)

Observation

135,486

0.09

135,486

0.09

135,486

0.09

135,486

0.09

135,486

0.09

135,486

0.09

R-Squared

Note: ∗∗∗p < .01, ∗∗ p < .05, ∗ p <0.1

Source: Authors

FTU Working Paper Series, Vol. 1 No. 3 (06/2021) | 60

4.2.2.2. Effect of FDI origin on job creation

In table 9, when comparing the effect of FDI origin on the number of jobs created in firms. The

impact of FDIpercenatge on jobs is significantly negative. In the case of FDI from CN and HK,

the coefficient of FDIpercentage is statistically significant at 10% level while other countries have

a coefficient significant at 1% level. The effect in FDIpercenatge from those 2 countries reduces

job the lowest compared among 6 countries.

FDI from JP and SG are the 2 that reduce jobs the most as FDIpercentage increases. An

increase in FDIpercentage of 1% reduce job by 2.88% in the case of JP and reduce job by 2.34%

in the case of SG.

For all 6 countries in the table, the increase in lncapital_int reduces job in the firm and the

increase in lnsize increase the number of jobs. Firm age still does not affect the number of jobs in

firms.

For the interaction term, the increase in lncapital_int increases the effect of FDIpercentage

on jobs in all 6 countries. The increase in lnsize also increases the effect of FDIpercentage on jobs

in most cases. In the case of JP, the increase in lnsize has no clear impact on the effect of

FDIpercentage on jobs. In the case of SG, the interaction between FDIpercentage and lnsize is

significantly negative at 10% proposing that an increase in lnsize results in a reduction in the effect

of FDIpercentage on jobs.

FTU Working Paper Series, Vol. 1 No. 3 (06/2021) | 61

Table 7. Effect of FDI origin on job creation

Job

Independent variables

CN

SG

HK

TW

KR

JP

-0.0052*

(-0.0028)

0.0217

-0.0234***

(-0.0047)

0.0211

-0.0090*

(-0.0049)

0.0211

-0.01***

(-0.0027)

0.0223

-0.0154***

(-0.0023)

0.0192

-0.0288***

(-0.0027)

0.0223

FDIpercentage of each origin

Lnage

(-0.0182)

0.1946***

(-0.0037)

-0.3353***

(-0.0026)

0.0005***

(-0.00017)

0.0005***

(-0.00016)

-0.0003

(-0.0182)

0.1965***

(-0.0037)

-0.3364***

(-0.0026)

0.0019***

(-0.00026)

-0.00063*

(-0.00035)

-0.00008

(-0.0015)

Yes

(-0.0182)

0.1965***

(-0.0037)

-0.3350***

(-0.0026)

0.00084***

(-0.0003)

0.00088***

(-0.00029)

-0.000029

(-0.0013)

Yes

(-0.0182)

0.1939***

(-0.0037)

-0.3370***

(-0.0026)

0.0009***

(-0.00015)

0.0006***

(-0.00015)

-0.00048

(-0.00085)

Yes

(-0.0183)

0.1923***

(-0.0037)

-0.3412***

(-0.0027)

0.0013***

(-0.0001)

0.00049***

(-0.00014)

0.00031

(-0.0008)

Yes

(-0.0182)

0.1936***

(-0.0037)

-0.3459***

(-0.0027)

0.0025***

(-0.0001)

0.00017

(-0.0002)

0.00003

(-0.00095)

Yes

Lnsize

lncapital_int

FDIpercentage*lncapital_int

FDIpercentage*lnsize

FDpercentage*lnage

(-0.0009)

Yes

Time (Dummy)

Province (Dummy)

Industry (Dummy)

Yes

6.8436***

(-0.1833)

135,365

0.105

6.8619***

(-0.1833)

135,365

0.104

6.8381***

(-0.1834)

135,365

0.106

6.8428***

(-0.1833)

135,365

0.105

6.8868***

(-0.1832)

135,365

0.105

6.9558***

(-0.1823)

135,365

0.102

Constant

Observation

R-Squared

Note: ∗∗∗p < .01, ∗∗ p < .05, ∗ p <0.1

Source: Authors

FTU Working Paper Series, Vol. 1 No. 3 (06/2021) | 62

4.2.2.3. Effect of FDI origin on insurance benefits

The table shows that in most cases, the effect of FDIpersence on insurance benefits is unclear

since the coefficient is not statistically significant. Only FDI from JP and TW have a significant

coefficient at 5% level in which JP have a negative impact while TW have a postive impact.

lncapital_int keep on having a negative effect on insurance benefit and lnsize keep on having

a positive effect in all 6 cases. Lnage appear to have no effect on the number of workers receiving

insurance.

For the interaction terms, the effect of changes in lnsize, lncapital_int show a mixed result in

6 cases. The interaction terms in the cases of CN, SG, HK present statistical result. Change in

lnsize will change the effect of FDI on insurance benefit in CN, SG, HK. Chang in lncapital will

change the effect of FDI on insurance benefit of TW and JP. In general, it seems that the effect of

interaction term is not consistent among the 6 countries for insurance benefits.

FTU Working Paper Series, Vol. 1 No. 3 (06/2021) | 63

Table 8. Effect of FDI origin on insurance benefit

Insurance benefits

Independent variabbles

CN

SG

HK

TW

KR

JP

0.0038

-0.0042

0.0029

0.0073**

-0.00068

-0.0071**

FDIpercentage of each origin

(-0.0037)

-0.035

(-0.0061)

-0.0371

(-0.0062)

-0.0364

(-0.0035)

-0.0354

(-0.0029)

-0.0347

(-0.0035)

-0.0351

lnage

lnsize

(-0.0252)

0.1543***

(-0.0063)

(-0.0252)

0.1567***

(-0.0063)

(-0.0252)

0.1553***

(-0.0062)

(-0.0253)

0.1563***

(-0.0063)

(-0.0254)

0.1546***

(-0.0063)

(-0.0253)

0.1545***

(-0.0063)

-0.1201***

(-0.0039)

-0.1207***

(-0.0039)

-0.1204***

(-0.0039)

-0.0002

-0.1187***

(-0.004)

-0.1208***

(-0.004)

-0.1247***

(-0.004)

lncapital_int

-0.00027

0.000098

-0.00053***

(-0.00019)

0.000087

-0.000049

(-0.00015)

0.00029

0.00065***

(-0.00016)

0.0004

FDIpercentage*lncapital_int

(-0.00023)

0.00069***

(-0.00032)

-0.00025***

(-0.00038)

0.0015***

FDIpercentage*lnsize

(-0.00023)

-0.000098

(-0.0011)

(-0.00052)

0.0014

(-0.0019)

(-0.0003)

0.0001

(-0.0017)

(-0.0002)

0.0001

(-0.0011)

(-0.00018)

-0.0005

(-0.001)

(-0.00025)

0.0002

(-0.0012)

FDpercentage*lnage

Time (Dummy)

Yes

Province (Dummy)

Industry (Dummy)

Yes

4.5008***

4.5046***

4.5042***

4.4773***

4.5035***

4.5501***

Constant

(-0.2586)

90,970

0.03

(-0.2588)

90,970

0.03

(-0.2856)

90,970

0.03

(-0.2588)

90,970

0.03

(-0.2592)

90,970

0.03

(-0.2589)

90,970

0.03

Observation

R-Squared

Note: ∗∗∗p < .01, ∗∗ p < .05, ∗ p <0.1

Source: Authors

FTU Working Paper Series, Vol. 1 No. 3 (06/2021) | 64