Performance of manufacturing enterprises – vietnam case study

Le and Harvie 2010 used the 2002, 2005 and 2007 SME database and cross sectional stochastic frontier model to estimate firm efficiency.. The chapter two selects the measure of firm perfo

UNIVERSITY OF ECONOMICS INSTITUTE OF SOCIAL STUDIES

HO CHI MINH CITY THE HAGUE

VIETNAM THE NETHERLANDS

VIETNAM - NETHERLANDS PROGRAMME FOR M.A IN DEVELOPMENT ECONOMICS

PERFORMANCE OF MANUFACTURING

ENTERPRISES – VIETNAM CASE STUDY

BY

NGUYEN VIET CUONG

MASTER OF ARTS IN DEVELOPMENT ECONOMICS

HO CHI MINH CITY, DECEMBER 2013

UNIVERSITY OF ECONOMICS INSTITUTE OF SOCIAL STUDIES

VIETNAM - NETHERLANDS PROGRAMME FOR M.A IN DEVELOPMENT ECONOMICS

PERFORMANCE OF MANUFACTURING ENTERPRISES – VIETNAM CASE STUDY

A thesis submitted in partial fulfilment of the requirements for the degree of

MASTER OF ARTS IN DEVELOPMENT ECONOMICS

By

NGUYEN VIET CUONG

Academic Supervisor:

PHAN DINH NGUYEN

HO CHI MINH CITY, DECEMBER 2013

Acknowledgement

This thesis would not have been possible without the support of many people I wish to express my gratitude to my supervisor, Dr Phan Dinh Nguyen who was abundantly helpful and offered invaluable assistance, support and guidance Deepest gratitude to Vietnam-Netherlands programme for sharing the literatures, invaluable assistance and providing me a big opportunity to complete this study I wish to express my love and gratitude to

my beloved families for their understanding and endless love, through the duration of my studies

Table of content

Chapter 1: Introduction 7

1.1 Reason of chosen topic 7

1.2 Research objectives 9

1.3 Research questions 9

1.4 Research methodology 9

1.5 Scope of the study 10

1.6 Justification of study 10

1.7 Structure of thesis 10

Chapter 2: Literature review 11

2.1 Measurements of firm performance 11

2.2 Stochastic frontier analysis in transition countries 11

2.3 Vietnamese technical efficiency analysis 13

2.4 Factors impact firm efficiency 15

2.5 Conceptual framework 21

Chapter 3: Methodology overview 22

3.1 Efficiency measurement concepts 22

3.2 The stochastic production frontier 24

3.3 Production functions accounting for technical change 25

3.4 Decomposition of productivity change 26

3.5 Stochastic production frontier with panel data 29

3.6 The stochastic frontier model using a single-stage estimation 31

3.7 Analytical framework 33

3.8 Variables are used in production function 33

3.9 Stochastic frontier production function 36

3.10 Variables are used in inefficiency model 37

3.11 Testing hypotheses 40

Chapter 4: Data and empirical results 43

4.1 Data issues 43

4.2 Testing hypothesis 46

4.3 Empirical result 47

4.4 Sources of technical inefficiency 48

4.5 The estimate technical efficiency 50

4.6 Total factor productivity decomposition 52

Chapter 5: Conclusions and policy implications 56

5.1 Main findings 56

5.2 Discussions and policy recommendations 56

5.3 Limitation and further studies 57

Appendix 59

Reference 65

List of tables

Table 1-Average of some variables by ownership 44

Table 2-Pairwise correlation of continuous variables 44

Table 3-Summary of hypothesis testing 47

Table 4-Estimation of stochastic frontier function and inefficiency model 48

Table 5-Distribution of production efficiency by year 51

Table 6-Frequency distribution of efficiency estimated by year 51

Table 7-The average of productivity by ownership 52

Table 8-Summary of factors impact on technical efficiency 55

List of figures

Figure 1-Conceptual framework 21

Figure 2-Technical and allocative efficiencies from an input orientation 22

Figure 3-Scale efficiency 23

Figure 4-The stochastic frontier 25

Figure 5-Estimation and decomposition of productivity change 27

Figure 7 - Distribution of continuous variables 45

Figure 8-Scatter plot correlation of inefficiency explanatory variables 46

Figure 9-Technical efficiency in period 2000-2008 53

Figure 10-Technical progress in Vietnamese manufacturing firms 53

Figure 11- Change in technical efficiency by ownership 54

Figure 12 - Total factor productivity growth 54

Figure 13-Total factor productivity growth account return to scale 55

Abbreviations

SFA: Stochastic frontier analysis

DEA: Data envelopment analysis

ML: Maximum likelihood

OLS: Ordinary least squares

GLS: Generalized least squares

TFP: Total factor productivity

HHI: Herfindahl index

GSO: General Statistics Office

SME: Small and medium enterprises

FDI: Foreign direct investment

SOE: State owned enterprises

CSO: Central state owned enterprises

LSO: Local state owned enterprises

LTD: Private, private limited or private joint-stock enterprises COO: Cooperative, collective or partnership enterprises FIO: Foreign invested ownership enterprises

RRD: Red River Delta

NMM: Northern Midlands and Mountain Areas

NSCC: North Central Coast and South Central Coast

CH: Central Highlands

SE: South East

MRD: Mekong River Delta

Chapter 1: Introduction

Nowadays, macroeconomic developments, particularly macroeconomic stabilization is focused in many analyses of economies undergoing transition Recent studies suggest macroeconomic reforms through improving enterprise performance plays important role in sustaining macroeconomic stability for transition economies Besides, firms are assumed importance of accelerating economic growth in developing countries They promote capital formation, create wealth in the country, and help to reduce unemployment and poverty Studying the enterprise’s efficiency always plays an important role in economic research, especially in developing countries

1.1 Reason of chosen topic

Vietnam began re innovation in 1986 and transited from planned economy to market oriented economy Besides achieving many successes in economic development and reduce poverty, Vietnam still confronts with unsustainable development issues and middle income trap Specifically, the comparative advantages of Vietnamese manufacturing firms remain heavily upon cheap labor and foreign direct investment, without enhancing their productivity A low level of productivity has been observed in this sector, since they lack new technology, product and process innovation, financial access , skilled labor, raw materials, high value added production and managerial skills (UNIDO, 2011) The practices demonstrated that the sustainable development can be maintained only if increasing productivity is engine of growth rather than accumulation of resources Moreover, structural evolution and the input productivity, which compose the quality of economic growth, are solutions for the middle income trap problem

Through identifying sources affecting firm inefficiency, direct impact on the overall growth of the economy can be revealed The appropriate policies and recommendation can be learned from these analyses As a result, measuring technical efficiency to improve productivity and competitiveness over the long term

is urgently needed especially for the manufacturing sector For this kind of study,

there are two approaches to measure firm efficiency and examine firm inefficiency effects The former is the parametric stochastic frontier analysis and the latter is non parametric data envelopment analysis However, the stochastic frontier analysis approach is more relevant in this context The reasons are the stochastic frontier approach closer to reality while it considers both factors beyond the control of the firm and firm-specific factors Moreover, the stochastic frontier method separates effect of inefficiency and other random shocks Whereas, the data envelopment approach does not differentiate technical efficiency and statistical noise and it is a non-statistical technique

Despite of the stochastic frontier analysis advantages, the SFA studies about Vietnamese enterprise’s efficiency are still inefficient, scatter and rare For example, Vu (2002) analyzed focus on SOE with a database of 164 manufacturing SOE for 1996-1998 and using two stage stochastic frontier analyses He revealed the skilled workers, engaged in exports activities impact positively on SOE performance Nguyen (2005) estimated technical efficiency of 32 manufacturing sector in Hanoi and Ho Chi Minh cities using 2000-2002 industrial data with both SFA and DEA approaches He found that Vietnamese industries operate with labor-intensive way Nguyen et al (2007) studied panel data of 1,492 firms in 2000-2003 using both SFA and DEA In this study, they found Vietnamese manufacturing firms improve productivity by capital accumulation rather than increase productivity Tran et al (2008) investigated 800 SME in 1996 and 1,500 SME in

2001 using cross sectional stochastic frontier model They claimed that SME lack of management skill through their firm’s age and size affect negatively on performance Le and Harvie (2010) used the 2002, 2005 and 2007 SME database and cross sectional stochastic frontier model to estimate firm efficiency They observed that the cooperation, subcontract and product improvement are positive factors impact on technical efficiency Nguyen et al (2012) estimated the enterprise’s efficiency and decomposed TFP growth into technical progress and technical efficiency change However, they did not examine the inefficient effects in their model

In brief, “Performance of manufacturing enterprises – Vietnam case study” topic is chosen for revealing Vietnamese manufacturing firms’ technical efficiency over the period 2000-2008 which can answer the question “Did Vietnam develops sustainable?” Although there were various literatures about Vietnamese manufacturing firms technical efficiency but these studies still incoherent and failed

to investigate the impact of ownership, finance, technology on technical efficiency This thesis tries to overcome the shortcoming of previous studies

1.2 Research objectives

The main objectives addressed in this thesis are:

First, this thesis estimates the technical efficiency of Vietnamese manufacturing enterprises in the period 2000-2008

Further, this analysis tries to identify firm-specific and business environment factors, which significantly affect the inefficiency of Vietnamese manufacturing firms?

Finally, total factor productivity growth of in Vietnamese manufacturing firms is decomposed to find which source mainly contributes

1.5 Scope of the study

This thesis will focus on Vietnamese firm technical efficiency for the period

2000 to 2008 This study will also examine the sources of TFP growth and investigate inefficient factors using one stage stochastic frontier analysis

1.6 Justification of study

At the practical level, the information of technical inefficiency and factors affect it is necessary for economic policies to improve production efficiency At the theoretical level, topics about technical inefficiency of firm in the multi sector economy, as Vietnam still rare and inappropriate This study aims to bring some contribution to this area

1.7 Structure of thesis

For completing the thesis objective, it has organized as follows The chapter two selects the measure of firm performance, reviews the stochastic frontier studies and factors affect efficiency used in the analysis Chapter three introduces efficiency concepts, development of stochastic frontier methods Chapter three also describes the econometric model, variable, and model testing hypotheses Chapter four discusses the estimation results, inefficiency factors of firms and decomposition of TFP growth And chapter five contains the main findings, discussions and limitations

Chapter 2: Literature review

This chapter selects measurement of firm performance, reviews empirical studies in stochastic frontier analysis area and inefficiency factors use in this study

2.1 Measurements of firm performance

Firm performance can be measured by accounting ratios, productivity and efficiency The accounting ratios have several disadvantages when used to measure firm performance First, they cannot effectively reflect the multidimensional characteristic of the production process of many industry sectors Second, their interpretation can provide a lot of misleading information due to the fact that they can be artificially modified by managers Finally, using accounting ratios are subjective when the analyst can choose ratios in order to assess the overall performance

Productivity is defined as the ratio of the firm’s outputs with inputs firm uses Productivity measure normally refers to total factor productivity which includes all factors of production The other partial productivity measure, which considers firm’s productivity with aspect of one production factor, can provide a misleading result of overall productivity Also, productivity measure does have meaningful units of measurement

Efficiency is measured by comparing firm’s actual output with maximum producible quantity from its observed inputs The efficiency measure provides the benchmark of how firm allocates resources for production Comparing with accounting ratio and productivity measure, efficiency reflects firm performance more precisely and consistently This study chooses efficiency as a measurement of firm performance due to these reasons

2.2 Stochastic frontier analysis in transition countries

Data envelopment analysis and stochastic frontier analysis are estimating technical efficiency methods However, the stochastic frontier method is commonly used to estimate efficiency and it applied in many studies about transition countries

For example, Liu and Liu (1996) used an unbalance panel of 283 Chinese SOEs in the period 1980 to 1989 in seven industries to examine the inefficiency effects They found the bonus system had speedy and impressive efficiency effects Other factors like a contract to permanent staff, dual tracking pricing system, firm size and, administrative affiliations, authorities, and provinces were not always substantial impact on firm efficiency across industries

Brada et al (1997) found the firm’s efficiency was positively related to firm size when using firm level data to estimate frontier production function for Czechoslovak industry in 1990 and Hungarian industry in 1991 They also revealed export orientation had no effects on efficiency Their most important report was efficient firms were more profitable in Hungary where economy more reformed While in Czechoslovakia, the profit redistribution by the center led to an inverse relationship between efficiency and profitability

Kong and Marks (1999) used Chinese state-owned enterprises in four sectors from 1990 to 1994 to investigate the technology change They found no experience

in three sectors (chemical textiles and building materials) while in machinery industry experienced a neutral technical change The corporate system, sharing holding and bonus system had a positive impact on the firm’s efficiency

Piesse and Thirtle (2000) used the Hungarian firm level data from 1985 to

1991 of 117 agricultural firms and 43 in the light manufacturing sector to examine the inefficiency effects They found labor and material contributed to output level while energy and capital did not Overcapitalization, excessive management costs and participating in export markets were factors impact firm performance

Gao (2010) used a data of 863 firms in 11 cities of Chinese in 1995 to 2001 period to estimate the stochastic production frontier He found the privatization, hardening budget constraint and relaxing the firms’ obligations to retirees and xiagang (workers no longer actively working but still receiving minimum benefits from firms) workers significantly improved a firm’s efficiency

2.3 Vietnamese technical efficiency analysis

Vietnam has begun transitioning progress since 1986 and still in the transition process As the result, in recent years there has been an increasing interest in estimating technical efficiency of Vietnamese firms, which used either stochastic frontier analysis or data envelopment analysis For example, Vu (2002) analyzed a database of 164 manufacturing SOE for 1996, 1997 and 1998 by using two stage stochastic frontier analyses He estimated the technical efficiency of all industrial SOE to be 0.788 for 1997 and 1998 The share of skilled workers, located

in Ho Chi Minh city and engaged in export activities were most important factors affect the SOE’s performance However, this study used two stage panel data stochastic frontier analysis which can lead to inefficiency effects biased due to omission of relevant variables in the first stage of the frontier (Kumbhabar and Lovell, 2000)

Then, Nguyen (2005) used both SFA and DEA approach to estimate technical efficiency of 32 industries of manufacturing sector in Hanoi and Ho Chi Minh City Using panel data from 2000 to 2002, the obtained results demonstrated that the average efficiency score of HCM and HN manufacturing industries were closely equal and about 65 percent for SFA approach and about 60 percent with DEA method He reported these industries still based on a labor-intensive way of production The paper also found that HCM city’s manufacturing industries had technical efficiency higher than in Hanoi The disadvantage of this study was it used inappropriate industrial data for short periods

Afterward, Nguyen et al (2007) performed a stochastic frontier analysis and data envelopment analysis using panel data of 1,492 firms from 2000 to 2003 Under the specification of variable return to scale, the mean technical efficiency of these small and medium firms was about 50 percent under SFA approach, and about

40 percent under the DEA approach They revealed that firm age had a positive effect on firm performance due to capital accumulation rather than an experience of the firm Moreover, these firms were still operating with a labor-intensive way of production In this analysis, Nguyen et al (2007) failed to examine the technical progress of these firms

Later, Tran et al (2008) studied data collected from 800 firms in 1996 and

1500 firms in 2001 using cross sectional stochastic frontier model They showed that mean technical efficiency of SME is 0.87 and higher than the 0.78 found for SOE (Vu, 2002) The micro enterprise and metropolitan location variables had a significant positive affect on SME’s technical efficiency while firm’s age had a negative impact However, the estimation using cross sectional data may not be consistent since the variance of the distribution of technical efficiency does not vanish when the sample size increases

Whereas Hoang et al (2008) investigated panel data of 4600 firms from

2001 to 2005 using data envelopment analysis They found private ownership exhibited the highest robustness to both “level” and the “rate” of technical efficiency On the other hand, state ownership had lowest robustness to both “level” and “rate” of performance Implementing higher levels of capital intensity, operating in larger business scales, competing in high competitive industries and locating in high-income regions were factors that helping firms gained efficiency

In recent years, Le and Harvie (2010) used the non-state SME database in

2002, 2005 and 2007 to estimate the technical efficiency for manufacturing small and medium enterprises They reported the mean technical efficiency for manufacturing SME are 0.843, 0.925 and 0.923 in 2002, 2005 and 2007 respectively The factors had a positive effect on manufacturing SME’s technical efficiency were Cooperation, sub-contract and product improvement Against firm age, firm sizes, government assistant land at the start, government credit at operation were the negative factors impact on technical efficiency

Up to date, Nguyen et al (2012) decomposed the TFP growth into technical progress and changes in technical efficiency using data of 8,057 firms during 2003-

2007 The average estimated technical efficiency of the stochastic frontier method was 30.9 percent The decomposed TFP reported average annual rate of change in technical efficiency was 3% and average technical progress change was about 2.3%

As the result, the TFP of the manufacturing sector during the study period had grown at the rate of 5.2% although the rate of growth decreased continuously However, similar to Nguyen (2005) and Nguyen et al (2007), Nguyen et al (2012)

used only two inputs in the production function which is not appropriate for the manufacturing sector

2.4 Factors impact firm efficiency

To take full advantage of stochastic frontier analysis, the factors affect firm efficiency should review and introduce in inefficiency model This part will consider these factors and build the conceptual framework use in this study

2.4.1 Ownership structure and firm performance

A substantial volume of literature pertaining to corporate governance has identified and discussed the role of ownership structure as a determinant of enterprise performance This subsection will review relationship of various corporate governance and firm performance

State ownership: Empirical studies mostly revealed a negative link between

state ownership and a firm’s performance From agency problems, politicians controlled state-owned firms, and they can exploit the firm’s assets easily (Le and Buck, 2009) In other words, they may have an increased incentive to avoid maximizing a firm’s profitability for other minor shareholders who are not the government This can be explained by the lack transferable residual claims of the government, the choice of social and political objectives rather than profit maximization objectives and the government’s higher transaction costs Moreover, government employed officers and workers based on the political relationship rather than their performance capability

Public firms are highly inefficient since they adopt strategies that satisfy policy interests such as excess employment or they are likely to serve the public interest better than private firms are (Boyscko et al ,1996; Sheifer, 1998) Boysko

et al (1996) suggested that the inefficiency of state firm due to the agency problem with politicians rather than that with managers, since politicians aim to obtain voting support from employees of state firms and labor unions by raising higher labor spending In this perspective, privatization may be implemented for public firms to improve their efficiency On the other hand, Sheifer and Vishny (1997) suggested that public firms, which are privatized without increasing the number of

investors, are likely face the agency problem due to insufficient investors to monitor firms Zeitun and Tian (2007) also suggested reducing government ownership increases firm performance for 59 Jordanian listed firms during 1989-2002

In contrast, a number of empirical studies discovered a positive association between state ownership and a firm’s performance State ownership can increase a firm’s performance since it produces a helping hand based on efficiency or state power (Le and Buck, 2009) In efficiency terms, the government may act as a controlling shareholder, and control managers more efficiency than widely dispersed ownership In state power terms, the government may use its power by providing supportive environments to improve a firm’s performance Le and Buck (2009) revealed a positive impact of state ownership on firm performance with 1,000 Chinese listed firms during period 2003-2005 Consequently, this issue is a complex and can depend on a number of factors, such as the stage of economic development, how performance is measured

Private ownership: Private owned firms present the opposite picture with

state owned enterprises They are usually being small but quick to seize, they often good at market orientation but poor in R&D Private owned enterprises mostly are family owned and have distinctive objective from SOEs Tan (1996, 2001) discovered Chinese start-up firms usually have a simple and flexible structure They often choose aggressive strategies since their simple structure allows them to react quickly to opportunities or proactively out maneuver

Foreign ownership: Foreign ownership has become the most important

ownership in developing economies Foreign ownership has superior technology, managerial expertise, good corporate governance, and strong foreign-market network A number of empirical studies suggested that foreign-owned firms have a positive association with a firm’s technical efficiency Zhang et al (2001) revealed

a strong ownership effect on a firm’s technical efficiency while studied 1,989 Chinese industrial firms during 1996 to 1998 Foreign-owned enterprises exhibit the highest efficiency scores, but SOE firms exhibit the lowest Goldar et al (2003) analyzed 63 Indian engineering firms during 1990 to 2000 and they found that foreign-owned firms have higher efficiency than domestically owned firms

Hypothesis 1: Foreign and private ownership have a significant and positive

effect on the technical efficiency of Vietnamese manufacturing enterprises; owned firms perform best in terms of technical efficiency relative to other ownership type for Vietnamese manufacturing enterprises

foreign-2.4.2 Firm specification and firm performance

The effect of firm size on firm performance differs across countries and sectors, and it is still inconclusive The high degree of concentration in an industry

is specific characteristic of socialist economies Most sectors of industry are dominated by large firms and these firms face little competition from the foreign producers Hence, large firms have few incentives for efficient operation, and should display a high level of inefficiency On the contrast, the large firms can invest in production technology and R&D that leads to improved firm efficiency Oczkowski and Sharma (2005) found that firm size is positively associated with firm efficiency for 121 Nepalese manufacturing firms during period 2000-2001 On contrast, Admassie and Matambalya (2002) revealed that firm size might be negatively associated with efficiency if large firms faced with management and supervision problems While Limpaphayom and Ngamwutikul (2004) found firm size has no significant association with operating performance for Thai listed firm during period 1991-1994

Capital intensity has been identified in a number of empirical studies as a determinant of technical efficiency (Ramaswamy, 1996; Mahadevan, 2000; Wadud, 2004) Capital intensity is a proxy for differences in technological choices amongst firms Firms may gain efficiency by applying appropriate labor intensive or capital intensive technology in their production Capital intensity may have positive (Wu et al., 2007) or negative (Mahadevan, 2000; Wadud, 2004) relationship with the technical efficiency

Hypothesis 2: Firm size and capital intensity have a significant and positive

effect on the technical efficiency

2.4.3 Workforce and firm performance

Many studies showed that firm with well educated workforce are likely to be more efficient due to educated workers are more productive in performing given tasks An educated workforce has a greater capability to absorb and effectively utilize new technology Tan and Batra (1995) revealed that training provides workers the skills to perform a wide variety of tasks and upgrade job skills as new technologies are introduced Well educated workforce is also considered as high quality labor

On contrast, using a large number of female workers may reflect forms of simple assembly, handicraft, seasonal work and low salary (Batra and Tan, 2003) These organizations where low skills and job turnover limits learning are likely have low a efficiency level

Hypothesis 3: Labor quality has a significant and positive effect on the

technical efficiency and female ratio has a negative relationship with firm performance

2.4.4 Finance and firm performance

Liquidity is a firm’s ability to convert its assets into cash in order to meet its coming debt payments The concepts of liquidity can measure by some financial ratios, such as working capital and current ratio Financially constrained firms may have difficulty in operating their businesses efficiently The firm is likely to face less financial distress if it has more liquidity (Ross et al., 2007) Goldar et al (2003) found the liquidity ratio has significant and positive effect on the technical efficiency of the Indian engineering firm during 1997 to 2000, but not found such result during the period 1990 to 1997 They concluded the liquidity is an important factor to facilitate production operations, since the liquidity ratio indicates the ability of a firm to meet its financial liabilities in a short run of one year

The relationship between financial constraints and firm’s performance can link to the agency problems Agency costs can arise from the conflicts of interest among managers, shareholders, and debtors (Jensen and Meckling, 1976) Managers have incentives to purse only safe project with yield low return to shareholder (firm

owners) because their reputation threatened if they undertake projects that fail; shareholder is likely to invest in risky value project while creditors normally prevent borrowing firms from investing in high risky value project Leverage can decrease agency costs because firms with a higher level of leverage are likely to induce managers to improve their managerial performance to avoid bankruptcy and liquidation For this reason, there exists a positive effect between leverage and a firm’s performance For example, debt constrained firms are forced to reduce their operating costs and other operating expenditures due to hardly obtain external finance Besides, workers face a high risk of losing their jobs leading to an increase

in workers’ efficiency (Sena, 2006)

Mok et al (2007) revealed a positive association between leverage and firm’s technical efficiency They also found the positive impacts of technical efficiency and firms’ profitability for 238 largest Chinese foreign-invested toy-manufacturing firms in 2002 Dilling-Hansen et al (2003) found that financial solvency has a negative effect on firm performance, as measured by technical efficiency for 2,370 Danish firms in 1997 Weill (2008) also observed a negative relationship between leverage and firm performance, as measured by technical efficiency for 4,403 Italian manufacturing firms and 2,312 Spanish manufacturing during period 1998 to 2000 Chang and Shin (2007) also found a negative effect between leverage and firm performance, as measured by firm market value for 15 Korean chaebols

Hypothesis 4: Financial constraints (leverage) have a significant and positive

relationship with the technical efficiency of Vietnamese manufacturing enterprises Vice versa, the more liquidity the lower is the technical efficiency of Vietnamese manufacturing enterprises

2.4.5 Export and firm performance

A large body of empirical evidence showed that export firm performs better than non export firm in terms of productivity The literature proposes two mechanics to explain the positive correlation between exporting and productivity First mechanism is the self-selection which state only the most productive firms survive in the highly competitive export market while the low productivity

exporters will be forced to exit Another mechanism is learning by exporting, in which export firm gains more knowledge in the export market and gain access to expertise from their buyer in product designs and production methods that non-export firm do not have

Blalock and Gertler (2004) investigated an evidence for the self-selection hypothesis by using panel data for the Indonesian manufacturing period 1990-1996 Their results contradicted previous result from developed countries that the productivity of export firms increases from 2 to 5 percent after exporting Hansson and Lundin (2003) revealed the continuing exporting firms are significantly more productive than non-exporters Yet, they found no significant differences in TFP growth between exporters and non-exporters Hallward-Driemeier et al (2002) found that foreign owned firms and exporting firms have significantly higher productivity in Indonesia, Korea, Malaysia, the Philippines, and Thailand during period 1996-1998

Hypothesis 5: Firms participate in export markets has a significant higher

technical efficiency (the self selection hypothesis exist)

2.4.6 Market competition and firm performance

There are many evidences showed that competition enhances productivity Green and Mayes (1991) stated the competition extension is an important variable for explaining the efficiency difference As the results, they recommend policy to raise productivity of firms is deregulation, which expected to intensify competition

Li (1997), Djankov and Hoekman (2000) and Grosfeld and Tressel (2001) reported the increasing firm productivity after deregulation for China, Bulgaria and Poland, respectively Heskel (1991) found a negative link between high levels of market concentration and market share with total factor productivity when using UK panel data from 1980-1986 Nickell (1996) also reported low rent firms had consistently higher productivity growth than high rental firm

Hypothesis 6: High levels of market concentration have impacted negatively

on firm’s technical efficiency

2.5 Conceptual framework

Specialization Size

Figure 1-Conceptual framework

After reviewing sources of firm inefficiency, the conceptual framework of inefficiency model can be formulated as follows: four factors affect the technical efficiency are technological progress, economic scale, resource allocation and human capital The capital intensity, female ratio describes the firm technology progress The capital intensity expected impact on firm efficiency positively While female ratio which present for simple and labor intensive production is expected have a negative effect on firm performance The firm economy scale can be proxies

by firm size and firm specialization (compete in export markets) Both proxies are expected have a positive effect on firm productivity The resource allocation effect can use finance and market competition variables to examine Finance is expected negative sign while market competition variable are expected positive sign In the final, the proxies for human capital expected have a positive effect on technical efficiency Particularly, labor quality is expected have a positive effect

Chapter 3: Methodology overview

Before conducting the analysis, this chapter will introduce efficiency concepts, stochastic production frontier and development of stochastic frontier analysis methodology The analytical framework, variables and hypothesis testing are also introduced here

3.1 Efficiency measurement concepts

of CRS, technical efficiency of firm, which use two input factors to produce a unit

of output, can illustrate in figure 2 The isoquant line PP’ is the minimum combination of the two inputs the firm might use to produce a unit of output The firm is considered technically efficient if it can combine of the two inputs along the isoquant line Technical inefficiency is defined by any point located above and to

the right of the isoquant line The point S represents the technical inefficiency since the firm can reduce input without reducing output At point R, the firm can be technically efficient than at point S by producing the same unit but it uses less

OS

zero and one The ratio TE will equal one if firm is technically efficient

When input price information is given, allocative efficient is measured by the ratio 0Q/0R Firm production is technically efficient at the point R but it can be allocative inefficient because the firm can reduce production costs by the distance

QR Firm production is allocative and fully technical efficiency at the point R’

Scale efficiency

Figure 3-Scale efficiency

Source: Coelli et al (2005, p59)

Coelli et al (2005) demonstrated that a fully technical efficient firm may not have the optimal scale of operation For example, a firm might operate with increasing returns to scale if its scale of production is too small under the specification of variable return to scale (point C) In this case, firm C can increase the scale toward point A to increase its production In contrast, a firm can operate with decreasing returns to scale if its scale of production is too large (point B) So

firm B can decrease the scale towards point A to be more productive At point A, the firm cannot change its production scale because it operates at the most productive scale size or at the technical optimal productive scale (Coelli et al., 2005) It is possible to use the distance measure to estimate different types of efficiencies For example, the ratio of the slope 0F to the slope of the ray 0C is equivalent to TE VRS DC DF/ , which is the technical efficiency of firm at point F based on variable returns to scale technology

The scale efficiency of firm at point F can be estimated from the distance of constant returns to scale technology DE over the technically efficient data point DC (Coelli et al., 2005)

Scale Efficiency (SE)  DE DC/

3.2 The stochastic production frontier

Aigner, Lovell and Schmidt (1977) and Meeusen and Van de Broeck (1977) independently introduced the basic stochastic production frontier within a cross-sectional context These models contained error component capture the effect of

model can express as follow:

0 1

0 1

0 1 Deterministic component Noise Inefficiency

is positive While the frontier output of firm B lies within the deterministic frontier

attains maximum feasible output Otherwise, TEi < 1 shows a shortfall of firm observed output from its maximum feasible output

Figure 4-The stochastic frontier

Source: Coelli et al (2005, p244)

The estimation of stochastic frontiers is complicated since there are two random error terms And the preferred method to estimate is maximum likelihood (ML) since ML estimators have asymptotic properties Aigner, Lovell and Schmidt (1977) applied the method of maximum likelihood under the assumptions of a half-

~ (0, )

v iid N  and inefficiency

u iid N 

3.3 Production functions accounting for technical change

The Cobb-Douglas and Translog production functions are common used functional forms for production functions in stochastic frontier analysis While Cobb-Douglas form is first order flexible and provide enough parameters for first-order differential approximation, the Translog provide a second order flexible

approximation The restrictive properties of Cobb-Douglas function form is constant return to scale and unitary elasticity of substitution of the production function Hence, the Cobb-Douglas functional form can not be applied in case firms have variable constant to scale and its elasticity varies across data point

Technological advances can lead to changing production functions over time Thus, the model can include a time trend to reflect industry-specific technological development and account for technological change Account for technical change, Cobb-Douglas and Translog production functions can be given as follows:

The Cobb-Douglas production function (restricted model)

Where y is the level of output;

x is a set of input whose elements are xi and xj;

T is a time trend representing technical change

Technical change can be calculated as the percentage change in y in each period:

3.4 Decomposition of productivity change

Figure 5 illustrates the firm’s production function with single input is used to produce a single output, and producer expands from (x,y) to (x+1,y+1) The convex curves characterize the decreasing return to scale of firm production technology

Technical progress has occurred between period t and t+1, since f x t ,  1; f x t , ; The production has been technically inefficient in

problem is to attribute output growth to input growth and productivity growth Then the productivity growth must be decomposed into the contribution of returns to scale, technical change and change in technical efficiency

Figure 5-Estimation and decomposition of productivity change

Source: Kumbhakar and Lovell (2000), p282

With t denotes a time trend as a proxy for technical change and u represent technical inefficiency And technical change is not restricted to be neutral with respect to the inputs

The rate of technical change can be defined by

According as technical change shifts the production frontier up, leaves it unchanged, or shift it down

The rate of change in technical efficiency can be provided by

u TE

t



  

Divisia index of productivity change is defined as:

n n n

TFP y X  y S x

Where ,

n n n

and ww1 , ,w n 0is an input price vector

Total factor productivity change is also provided as follows:

 

 

= 1

n n n n

respect to each of the inputs,

is the allocative inefficiency

When price information is unavailable, the allocative inefficiency component cannot be calculated, and the decomposition simplifies to:

1/  /  ln /

y y dy dt d y dt

3.5 Stochastic production frontier with panel data

Panel data used in measuring a firm’s technical efficiency to relax the strong distributional assumptions when using cross sectional data Since maximum likelihood estimation of the stochastic frontiers models use cross sectional data suffer three serious problems First, the firm efficiency can be estimated but may not be consistent Second, strong distribution of technical inefficiency is required to separate technical efficiency from statistical noise Third, the technical inefficiency effect is not dependent on the independent variables, but it may be correlated with input vectors

Panel data can avoid the disadvantages of cross sectional data listed above Also, panel data can be used to examine the underlying production technology over time First, the independence and strong distributional assumptions can be relaxed when using panel data Second, more information when adding more observations for each firm, the technical efficiency can be estimated consistently when size of the sample begins to increase According to the random error term assumption, the analysis is expected that the efficient firms maintain their efficiency level while inefficient firms to enhance their efficiency levels overtime This assumption is not realistic for several industries As the result, two structures on the inefficiency effects were imposed and they are time-invariant and time varying

Time-invariant inefficiency models

Technical change does not allow in time-invariant inefficiency models In this model, the inefficiency effects are introduced, assumed to be nonnegative and the technical efficiency effects are constant over time

Besides, there is no distributional assumption on the technical efficiency effects The assumption for statistical noise is iid with zero mean and variance and not correlates with the independent variables The estimates of the coefficients may not

be reliable if the number of firms is small since this model measures a firm efficiency relative to the most efficient firm This model captures variation across firm time-invariant technical efficiency and time-invariant the effects of all phenomena

The random effects model has more attention due to the disadvantage of the fixed effects model The technical efficiency is assumed to be uncorrelated with independent variables and the statistical error in the random effects model This model can be estimated either by the maximum likelihood method or the standard two-step generalized least squares

In two-step generalized least squares method, all parameters were obtained in the first step with OLS, then the intercept and coefficient are re-estimated by GLS

in a second step As the large number of firms and long time-series data are applied, the more consistent estimates of GLS The assumption of this method is that the technical efficiency effects are uncorrelated with the independent variables and the statistical error This assumption improves efficiency in estimation

When the strong distribution assumptions or the strong independence assumptions are held, maximum likelihood estimation is feasible The distribution

of technical efficiency may be half-normal distribution (Pitt and Lee, 1981) or truncated normal specification (Battese and Coelli, 1988)

These three approaches have different properties and assumptions The random effects model based on GLS is preferred when the number of firms is large and time period is small The maximum likelihood estimation is more referable when independence assumptions are held since it has distribution assumptions In fact, the time-invariant technical efficiency assumption is likely can not hold if the panel data becomes longer

Time-varying inefficiency model

The technical efficiency levels can change over time, since firms expect to learn from their learning by doing effect As the panel becomes longer, the technical efficiency effects would change Coelli et al (2005) proposed the form of time varying technical efficiency

( ) *

u  f t u

Where f(t) is a function of technical efficiency over time Kumbhakar (1990)

and Battese and Coelli (1992) introduced the time varying technical efficiency model using the maximum likelihood technique in a random effect framework Technical inefficiency is assumed to have a truncated normal distribution in both models

decreasing and concave or convex, depending on the signs of these two parameters ( and )  and 0  f t( ) 1 

Battese and Coelli (1992) introduced the stochastic frontier method for unbalanced panel data where technical inefficiency effects are assumed to be distributed as truncated random variables and vary systematically with time Their model can be expressed asy it x itv it u it, where f t( )  exp (t T ) The technical inefficiency function (uit) is less flexible due to involve with only one

if  0and f t( )  0

3.6 The stochastic frontier model using a single-stage estimation

The inefficiency effects can be estimate with one-step process or a two step process With two-step process, the production function is estimated and technical efficiency is predicted in the first step Then predicted technical inefficiency is regressed against a set of explanatory variables As a consequence, the obtained inefficiency effects are biased due to omission of relevant variables in the first

stage To overcome the shortage of two-step approach, Battese and Coelli (1995) suggested one-stage stochastic frontier analysis method This approach allows for estimation of both technical change and time-varying technical inefficiencies one stage Consider the stochastic frontier production function:

it it it it

y  x  v u

Where yit is production (output) of the ith firm;

(0, v)

iid N  random errors, identical independently distributed of normal distribution;

independently distributed by truncation (at zero) of the normal distribution with mean z it , and variance u

it it it

u z  w

technical inefficiency effects;

wit is the unobserved random variables, obtained by truncation of a normal

(i.e., w it  z it )

The strong independence assumption of maximum likelihood method is now

= 1,2, , T, and i = 1,2,…,N The likelihood function expressed in terms of the

inefficiency in the overall residual variance And the technical efficiency of the iith

firm defined as follows:

model misspecification arise The second non-negative random variables

2 1

component links business environment and firm specific variables with the inefficiency effects or the non-negative random variables

3.8 Variables are used in production function

Coelli et al (2005) suggested that input and output quantities, prices and quality characteristic are important for the measurement of efficiency and

productivity This subsection will discuss common used input and output variables and select appropriate variables to be used in this empirical analysis

Output

Value added and gross outputs are commonly used as the output in most empirical studies for manufacturing sector For example, Nguyen et al (2007) used value added to measure firm efficiency for a manufacturing firm in 2000-2003 In contrast, Tran et al (2008) and Nguyen et al (2012) applied gross outputs to estimate technical efficiency for manufacturing sector in 2003-2007 If value added

is adopted as the output, only two inputs are used When value added is not adapted, value of all output produced as the output is used For the purposes of the analysis, this study use firm annual sales revenue which can be obtained from the income statement, as a proxy for output values The reason of not using value added as an output is its value can be negative and become missing when take natural logarithm Moreover, gross output allows the analysis used production function with three inputs, which this study focus to assess the materials usage of manufacturing firms The nominal output of a firm is deflated by the relevant product price index of the given industry and the given year which provided by GSO

Intermediate materials

For this kind of empirical analysis, there are five input categories commonly used – capital (K), labor (L), energy (E), material input (M), and purchase services (S) The energy (E), material inputs (M), and purchased services (S) can often be combined to form a single other input category (Coelli et al., 2005) Since the intermediate inputs energy, material inputs, purchased services, and other administrative and production costs cannot be separated individually due to the limitation of data in income statements and balance sheet, all those ones in aggregate are used as intermediate materials input Separate transaction relating to production and non-production costs are not provided in the firms’ balance sheets and financial statements over the period 2000 to 2008 To solve this problem, intermediate input costs can be obtained as follows: subtract the profit before tax of sales goods and services from sales revenue of goods and services to have the cost

of good sold The intermediate material equals the cost of good sold subtracts fixed