Technical efficiency and its determinants : The case of manufacturing firms in VietNam

Pooled Ordinary Least Squares Research and Develop Random Effects Model Scale Economies Scale Efficiency Change Stochastic Frontier Production Function Small and Medium Enterprises State

VIETNAM- NETHERLANDS PROGRAMME FOR M.A IN DEVELOPMENT ECONOMICS

TECHNICAL EFFICIENCY AND ITS DETERMINANTS: THE CASE OF MANUFACTURING FIRMS IN VIETNAM

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

MASTER OF ARTS IN DEVELOPMENT ECONOMICS

Pooled Ordinary Least Squares Research and Develop

Random Effects Model Scale Economies Scale Efficiency Change Stochastic Frontier Production Function Small and Medium Enterprises

State-Owned Enterprises Technical Efficiency Technical Efficiency Change Total Factor Productivity Technical Progress Time Trend

TABLE OF CONTENTS

CHAPTER 1: INTRODUCTION

1.1 The problem statement 2

1.2 Objectives of the research S 1.3 Research questions 6

1.4 Research methodology 6

1.5 Thesis structure 7

CHAPTER 2: LITERATURE REVIEW 2.1 Introduction 8

2.2 Basic Concepts and Theoretical Review 8

2.2.1 The Production Function 8

2.2.2 Cobb-Douglas production function 9

2.2.3 Technical Efficiency 11

2.2.4 Technical efficiency measurement 12

2.2.5 The stochastic frontier production function (SFPF) 13

2.3 Empirical Studies 16

2.3.1 Studies in advanced countries 16

2 3 2 Studies in developing countries 19

2.3.3 Studies in Vietnam 22

2.4 Analytical framework for the research 29

CHAPTER 3: RESEARCH METHODOLOGY AND DATA COLLECTION 3.1 Introduction 31

3.2 Research methodology 31

3.2.1 The stochastic frontier model 31

3.2.2 The technical efficiency model 34

3.3 Testing Hypothesis 37

3.3.1 The stochastic frontier model 37

3.3.2 The technical efficiency model 37

3.4 Data Collection 38

CHAPTER 4: ANALYSIS RESULTS 4.1 Sample profile 39

4.2 Technical efficiency 41

4.3 Comparison of technical efficiency 44

4.4 Technical efficiency model 46

4.4.1 Testing for the most appropriate model 46

4.4.2 Testing for heteroskedasticity 47

4.4.3 Determinants of technical efficiency 4 7 4.5 Chapter Summary 50

CHAPTER 5: CONCLUSIONS, RECOMMENDATION AND LIMITATIONS 5.1 The conclusions 51

5.2 The recommendations 54

5.3 Limitations 55

REFERENCES 56

APPEND I CES 60

LIST OF TABLES & GRAPHS

Table 2.1: Summary of Empirical Studies

Table 3.1: Summary ofvariables in the frontier production function

Table 3.2: Summary of variables in the technical efficiency model

Table 4.1: Descriptive statistics of output, capital and labour of manufacturing firms

in the period 2000-2004

Table 4.2: Estimates ofti model and tvd model

Table 4.3: The statistical tests of some hypothesis

Table 4.4: Summary of technical efficiency between ti model and tvd model

Table 4.5: Determinants oftechnical efficiency

Graph 4.1: The structure of 1,645 manufacturing firms from other sectors

-LIST OF FIGURES

Figure 1.1: The share of manufacturing enterprises in all industries ofVietnam

Figure 2.1: Illustration of Technical Efficiency

Figure 2.2: Analytical Framework

•

CHAPTER 1: INTRODUCTION

1.1 The problem statement

Since the launch of renovation in 1986, Vietnam has successfully transformed the centrally-planned economy into a market economy and made great achievements in social and economic aspects In the period of 2000- 2010, the country's economic growth was relatively high and stable at an annual average rate of 7 2% In 2010, the real GDP was recorded 3.4 times as much as that in 2000; the state budget collection was 5 times; and the GDP per capita stood at US$1,168 (GSO, 2011) By achieving these, Vietnam has moved from the group of poorest countries to the group of middle-income countries In addition, Vietnam has been successful in poverty reduction, close to achieving universal primary education, improving maternal health, reducing child mortality, obtaining much progress in gender equality and empowering women, and etc

In contribution to economic and social development, Vietnamese enterprises play a crucial role Business activities of enterprises have made significant progress In

1995, enterprises contributed about 45.3% of GDP; in 2001 this share increased to 53.2% and in 2007 was over 60% (GSO, 2008) The development of enterprises in many different sectors and localities lead to the change of economy's structure which reduces the share of agriculture and increases those rates of industry and services

With regards to manufacturing enterprises, they made important contribution to dealing with social matters such as creating more new jobs, increasing income for employees, contributing more to the state budget, and etc In more details, manufacturing enterprises create 2.203 million jobs, accounting for 47.3% of total jobs in all enterprises (GSO, 2007)

However, many weaknesses are found in the process of the development of the economy in general and the manufacturing sector in particular The infrastructure has not been completed and needs to be improved comprehensively The shortage

of electricity and water which are common may reduce the productivity (Klause et al., 2005) So, the efficiency and competitiveness of the economy still is lower than its potential

Moreover, the performance of enterprises has different results because of their resource, types of ownership, type and scale of business, location and some other reasons Although the business environment has been more transparent and flexible for business operation, the business results of each enterprise might not grow steadily In general, Vietnam enterprises expose their own features

Firstly, the number of new enterprises especially private companies has grown sharply since 2000 when the Enterprise Law carne into effect In three years after the issue of the Law, more than 72,600 new private enterprises were established, creating around 1.6 - 2 million new jobs (ClEM, 2004) These figures are very impressive when compared with just 26,000 private enterprises operating by the end

of 1998

Secondly, enterprises located in big cities such as Hanoi; Hochirninh city may enjoy many favorable conditions such as ideal geographical location; advantage of telecommunication, transportation; abundant labor supply with high skill to apply new technology in production Consequently, the number of enterprises in these cities increases very fast and accounts for about 4 7% of total number of enterprises and 45% of total revenue of the whole country (GSO, 2007) On the other hand, these enterprises are still facing with a lot of problems such as non-synchronous infrastructure, un-skilled labor Especially, each enterprise in big cities has to compete fiercely with many other local and domestic companies located at the same city These problems in association with improper policies might cause the companies to slowly increase their effectiveness

Thirdly, as a multi-sector market model operating according to the market mechanism and the state regulations, Vietnam's enterprises include state, private and foreign-invested sectors where the former plays a leading role in the economy

The government uses the state enterprises as an important tool to stabilize the economic environment and market prices of essential commodities such as electricity, coal, transport, rice and rubber So, state enterprises have received a lot of support, priorities, and subsidies from government Therefore, the performance of state enterprises is questioned about the efficiency relative to other sectors in the economy

micro-For above reasons, some issues need to be clarified such as the performance of firms

in Vietnam; the production efficiency level of firms located in former Hanoi, Hochiminh cities and other places; firms of the state, foreign and other sectors; and the factors influencing the technical efficiency of firms

The purpose of this thesis is to identify the above issues And, the manufacturing sector is selected to research because of following reasons: The share of manufacturing enterprises in all industries accounts more than 20 percent of all kind

of activity (GSO, 2006) However, manufacturing enterprises contribute important shares of revenue (more than 30 percent), number of employees (about 50 percent) and export value (22 percent)

The thesis applies a stochastic frontier production model and technical efficiency model to analyze the technical efficiency of manufacturing firms and try to find the determinants that affect firms' technical efficiency

Figure 1.1: The share of manufacturing enterprises in all industries of Vietnam

Source: GSO

1.2 Objectives of the research

Basically, this thesis aims at four objectives as follows:

(1) To measure the level of technical efficiency of manufacturing firms in the period 2000 to 2004

(2) To compare the difference in technical efficiency between manufacturing firms located in former Hanoi and Hochiminh City and those located in other provinces; between firms of state-owned, foreign firms and other firms

(3) To identify factors influencing the technical efficiency of manufacturing firms

(4) To suggest appropriate policies for improving technical efficiency of manufacturing firms

*Note: Former Hanoi: Because the data applied in the thesis from 2000 to 2004 Since August

1, 2008 Hanoi has merged with Hatay province and parts of neighboring of Vinhphuc and Hoabinh provinces

1.3 Research questions

With the research objectives, the thesis is therefore going to answer the following questions:

(1) What is the level oftechnical efficiency of manufacturing firms?

(2) What are differences in technical efficiency of manufacturing enterprises located in former Hanoi*, Hochiminh city and other provinces; state-owned, foreign and other firms?

(3) What are factors affecting the technical efficiency of manufacturing firms?

of manufacturing firms located in former Hanoi, Hochiminh city and other places; between state and foreign with other groups of manufacturing firms

In the second stage, the thesis examines factors influencing the technical efficiency

of enterprises For the data with two dimensions time series and cross sections, the thesis uses the panel data analysis via the appropriate method from pooled Ordinary Least Squares (OLS), Random Effects Model (REM) and Fixed Effects Model (FEM)

The data set applied for this thesis comes from the Vietnam Enterprise Survey conducted by the General Statistic Office in the period 2000- 2004

1.5 Thesis structure

The thesis is presented in five chapters After this chapter the rest of this thesis will

be presented in four chapters Chapter 2 covers the literature about the production function, technical efficiency, the stochastic frontier production function and empirical studies Chapter 3 presents the research methodology and data Chapter 4 presents the research results Finally, chapter 5 gives conclusions, recommendations and limitations ofthe study

CHAPTER 2: LITERATURE REVIEW

2.1 Introduction

This chapter provides an overview of literature about the research problems in a logical manner in order to drive this thesis into a correct direction This chapter is divided into four sections The first section presents the basic concepts and theoretical review which includes production functions, Cobb-Douglas production function, technical efficiency, technical efficiency measurement and the stochastic frontier production function The second section gives various empirical studies about technical efficiency to provide the foundation for developing the analytical model of this thesis The final section summarizes theoretical and empirical review and proposes the applied models in this thesis

2.2 Basic Concepts and Theoretical Review

2.2.1 The Production Function

A production function is the technical relationship between the quantities of productive factors used and the amount of products obtained from every combination of factors, assuming that the most efficient available methods of production are used

A general production function can be written as:

(2.1)

Where:

Q is the quantity of output

etc

•

Technically, the production function depicts the maximum output that can be produced by the input combination, given the technology in use In production, the inputs can be changeable and substitutable

Although the relationship between output and inputs is fundamentally physical, production function often uses monetary values The production process uses several types of inputs that cannot be aggregated in physical units It also produces several types of output Goint production) measured in different physical units One

of the ways to deal with the multiple output case is to aggregate different products

by assigning price weights to them (Mishra, 2007)

There are many kinds of production function that can be used in empirical studies as follows:

- Linear production function is a function that assumes a perfect linear relationship between inputs and total output

- Leontief production function is a function that assumes the inputs are used in fixed proportions

- Cobb-Douglas production function is a function that assumes some degree of substitutability between inputs

- Other production functions such as quadratic, transcendental-logarithm (translog), and etc

2.2.2 The Cobb-Douglas production function

The most widely utilized functional form for econometric modeling is the Douglas form This form is the most popular one in applied research, because it is easy to handle mathematically

Cobb-The Cobb-Douglas production function with two inputs of labor and capital is as follows:

•

Where

Y is total production (the monetary value of all goods produced in a year)

L is the labor input

: constant returns to scale;

: decreasing returns to scale;

And if a + ~ > 1: increasing returns to scale

There some methods estimating the parameters of a Cobb-Douglas production function and the typical estimation is based on the linear equation From equation 2.2, taking the logs of both sides, the function is transferred as log-linear form as follows:

log Yi = log A + a log Li + ~ log Ki (2.3)

Where

Y, A, Land K are as defined earlier

The residual from estimation of function 2.3 is a random error term or a disturbance term named Ui The disturbance U is different for each firm and assumed to have normal distribution

And, figure 2.1 shows more clearly about the relation of technical efficiency and scale efficiency The ABC line represents the frontier for the production process Points are on the frontier showing the maximum pure technical efficiency Meanwhile, the through-origin line expresses the scale efficiency and the points are

on the line have constant returns to scale Scale efficiency is used to indicate whether or not a firm is operating at an optimal scale A firm, that has a technical efficiency, may be due to purely technical efficiency or scale efficiency

Points A, B, C, D and E present the combination of a certain levels of input and output Observations of A, B and C are on the frontier line, so they have purely technical efficiency Meanwhile, observations D and E are below the frontier line The tangential line at point B expresses the constant returns to scale of technology The point B shows the relative technical efficiency In other words, at this point, firm obtains both purely technical efficiency and scale efficiency due to its location

on the frontier and the constant returns to scale

Observations A and C are on the frontier so they are purely technical efficiency However, these points are not efficient in scale

Observation D is inefficient in both scale and technique Theoretically, the same level of input could be used to achieve a higher level of output, at point D the firm can move to the frontier between points B and C

Observation E is purely technical inefficient because it lies below the frontier; but it

is scale efficient because it produces at input level of x2 the scale efficient level of input at the same level of scale efficiency of point B

Figure 2.1 Illustration of technical efficiency

Output (:y)

y;

y~~ · ·

-x, Input tx)

2.2.4 Technical efficiency measurement

So far the firm's technical efficiency has been analyzed by many researchers Among various methods, the two following ones are frequently applied: the parametric approach called stochastic frontier production function - SFPF and the non-parametric approach or data envelopment analysis - DEA

A stochastic frontier production function approach is used to estimate a production function when the specification of technology is given In other words, the estimated production function can be used for all firms of the same sector In SFPF, the residuals or disturbance term is assumed to consist of two components The first component is assumed to have a nonnegative distribution that called inefficiency or technical inefficiency The technical inefficiency is defined as the difference

between the actual production level of a firm and the frontier (Minh and Dong, 2005) The other component is assumed to have a symmetric distribution which refers to as random components

The data envelopment analysis (DEA) was first developed by Farrell (1957) with a deterministic non-parametric frontier which is constructed by using mathematical programming methods from observed input-output data of sample firms

According to Ray and Ping (200 1 ), DEA makes only a few fairly weak assumptions about the underlying production technology and does not need a functional specification Based on those assumptions a production frontier is empirically constructed using mathematical programming methods from observed input-output data of sample firms

DEA has some advantages when it is used to estimate efficiency The specification

of production technology and the statistical distribution of inefficiency residuals are not required In addition, DEA can deal with multiple outputs easily and does not require any assumption about the functional form of production (Minh and Long, 2007) So, DEA focuses on taking into account and classifying variables, which can

be inputs or outputs of the production function However, because DEA does not decompose the residuals as the stochastic parametric estimations, Llewelyn and Williams (1996) therefore argued that nonparametric estimations could give biased estimate of inefficiency of firms

2.2.5 The stochastic frontier production function (SFPF)

With a given technology, the production of firms is combined and limited by the frontier production line which is formed by maximized production points Firms try

to achieve their maximizing outputs But not all firms can reach their frontier production For a firm that the production is below the frontier, the distance

calculated by their actual production and maximizing output presents the level of production inefficiency

The SFPF model is proposed and developed by Aigner and Chu (1968), Afriat (1972), Richmond (1974) They use the conventional production functions like the Cobb-Douglas or the transcendental logarithmic (translog) as the stochastic frontier production Then, the technical efficiency of firm is measured

The parameters of SFPF are proposed to estimate by different approaches Among parametric approaches, the method of Battese and Coelli ( 1995) is applied most of studies

The method permits that output is specified as a function of controllable factors of production, technical efficiency term and other random shocks Other random shocks affect output and they are outside the control of producers The general form

of a stochastic frontier production function can be defined as follows:

(2.4)

Where

i= 1,2,3, ,Nfirms

Yi is the real output of firm i;

xi is the vector of inputs of firm i;

~ is a vector of production parameters to be estimated;

Residuals Ui and Vi are estimated and assumed to follow some particular probability distributions

Vi represents the effect of all random factors and it may be positive or negative Residual Vi is assumed to be independently and distributed as normal random variables with zero mean (0) and constant variance cr2v: vi~ N(O,cr2v)

Ui represents technical inefficiency of the firm and is always positive It follows positively normal distribution and is truncated at zero (0): ui ~ N\J.ti,a2 u)

The firm's technical efficiency can be measured as the ratio of actual output against potential output, as follows:

(2.5)

Where

Yi is the actual output of the i firm

Yi * is the potential output of the i firm

In order to obtain the specific factors that affect each firm's technical efficiency (TE), following Battese and Coelli (1995), the mean ofTE can be specified as:

(2.6)

Where

TEi is technical efficiency of the i firm

zi is a vector of specific socio-economic variables that may influence the technical efficiency of the firm

o is a vector of unknown parameters to be estimated

By utilizing the parameterization proposed by Battese and Corra ( 1977), a2 and a2 v

can be replaced with a 2 = a 2 u + a 2 v andy= a 2 ula 2 that can be done with calculation

of maximum likelihood estimates

Where

a 2 is the variance of noise and

a 2 is the variance of inefficiency effects

It means the production uncertainty (cr ) comes from two sources: pure random factors and technical inefficiency

y is the proportion of uncertainty, having a value between one to zero

If the value of y is zero, the deviations from the frontier are attributed to random

error If it has the value of one, deviations are due to technical inefficiency

2.3 Empirical Studies

So far, the technical efficiency has been widely examined in Vietnam and many other countries The studies often focus on estimating the technical efficiency of a specific industry or comparison of the technical efficiency between some industries

in an economy and among firms located in different location In addition, the determinant of technical efficiency is an attractive object to be analyzed This part

of study is going to review some previous studies

2.3.1 Studies in advanced countries

Oleg et al (2006) uses the panel data set with total of 35,000 firms in the period

1992 to 2004 in 256 industries from the German Cost Structure Census to estimate the technical efficiency Researchers try to find the relationship between inputs of materials, labour compensation, energy consumption, capital, external services, the value of gross production minus subsidies and excise taxes as output of the production function and the technical efficiency

The result of analysis reveals that industry specific effect is the most important determinant on technical efficiency variation Firm size and location of firm are the second and third most important factor respectively Small firms operate more efficiency than larger firms However, R&D intensity, outsourcing activity and the legal form affect the technical efficiency relatively fairly R&D intensity negatively

influences on technical efficiency that can be explained by a time lag between R&D

spending, then it lately makes improvements of technical efficiency And, the study

also finds that technical efficiency is time invariant, but it fails to indicate whether

year effects make an increase or a decrease of average efficiency

Some important lessons learned from this research are useful for other studies as

follows The research is employed the advantage of the panel character It uses

many inputs in the production frontier function which is in form of a transcendental

logarithmic (translog) However, it is not applicable in other research if there is a

lack of some useful input variables It is the same thing in the second model to

measure the determinants of technical efficiency It's remarkable that the authors

select many important factors as determinants on technical efficiency including both

internal and external factors to the firms The external factors include industry

affiliation, location, year effects and share in industry And internal factors

comprise firm size, outsourcing activities and ownership (legal form) of firm

Donghyun et al (2009) research productivity growth of Swedish economy by using

the panel data of 5,893 manufacturing and services firms in the period 1992-2000

with total of 38,000 observations In order to estimate the technical change and

productivity growth of firms, the production function is applied having the output Yit

as value-added of firm and inputs as a series of inputs Xit· And, the perpetual

inventory value is a proxy of the capital stock

From frontier production function, the authors estimate the error term, Uit· Then it is

specified as a two-way error component model as follows:

(2.7)

Where: IJ.i, A.t and Vit are firm-specific effects, time-specific effects and statistical

noise, respectively In order to avoid over-paramiterization, firm-specification

effects, I-ii are replaced by industry-specific effects, lld·

One of the important findings is that the returns to scale positively correlate with

firm size The smaller firms exploit the labour force relatively more efficient than

the capital stock, and vice versa for larger firms And small firms operate close to their optimal scale of production meanwhile small-medium, medium and large firms can raise their efficiency when they reduce their scale Another interesting finding is that the estimate of the technological change can be biased The authors point out the reasons of bias that are certain inputs In more details, the technological change in production caused the changes in the proportion of inputs Elina (2006) estimates the technical efficiency and determinants of inefficiency in Finnish information and communication technology (ICT) manufacturing sector The research used the unbalanced panel data of ICT equipment manufacturing in the period 1990-2003 with firms having at least 20 employees

The determinants of inefficiency are selected including R&D investments, the specific Lerner index (ratio of operating profit to the value of gross output), the firm leverage ratio, ownership status in terms of domestic and foreign, exporter status, size and age The results showed the average firm enjoying only about half of the frontier firm's technical efficiency level (about 56%) The technical efficiency varies very much by firm, the time-varying efficiency averages at little over 40% of the most efficient firm's reference rate

firm-The outstanding point in this research is that the author uses a stochastic frontier model with four different approaches to estimate time invariant and time-varying efficiency levels Among these models, the Battese-Coelli maximum likelihood model is more appropriate than ordinary least squared model And, the translog production function is the best one

Alvarez and Gonzalez (1999) develop a method that combines panel and cross sectional data in the estimation of technical efficiency The researchers use the balance panel of 82 dairy farms and cross sectional data on input quality to estimate technical efficiency The predicted value of technical efficiency has value of 72% Then the authors use the cross sectional information of input quality to compute the corrected technical efficiency index The value of new one is unchanged, 72%

The important finding is that technical efficiency depends heavily on the information about input quality For example, the authors find the strong relationship between technical efficiency and land and cows Such a finding is not found in the previous examination And, technical efficiency is positively related with the farm size in the first analysis but negative relationship after adjustment So, the unobservable factors help to explain the variation of technical efficiency more clearly by the method of Corrected Ordinary Least Square However, this procedure

is only exploited once we have relevant information

In order to avoid the multi-collinearity, Marco (2010) uses a stochastic frontier production function in the form of Cobb-Douglas including a time trend to capture the Hick-neutral technical change:

2.3.2 Studies in developing countries

In researching the technical efficiency and its determinant of manufacturing firms in China, Wu (2002) uses data set of many firms in 30 regions in 1995, with total of

5,160 observations The author applies a two-stage approach The first stage is employed standard frontier production function to estimate regional and sectoral specific technical efficiency rates In the second stage, he applies Tobin models to investigate the impact of regional and sectoral specific factor on technical efficiency

Selected determinants of technical efficiency are: Depreciation measured by the ratio of net value of fixed assets over the gross value of fixed assets; Productive assets measured by the ratio of gross value of productive assets over gross value of fixed assets; Labour compensation measured by average wage; Incentive system as the ratio of bonus and allowance payment over total wage captures the effect of the incentive system on performance among the industries; Taxation system is measured by the ratio of tax over value-added And, the dummy variables are the location in the western and the central

The research shows that technical efficiency of Chinese manufacturing firms IS

about 80 percent in average and technical efficiency of all sub-manufacturing sectors In addition, factors of labour compensation, taxation incentive and agglomeration are found to be important determinants of technical efficiency However, the results are only for 1 year, 1995 We don't find the variation of technical efficiency, its trend and technology rate over time

Goldar et al (2003) try to analyze the effect of ownership on efficiency of engineering firms in India and compare the technical efficiency among three groups

of firms including foreign ownership; domestically owned private sector firms; and public sector firms The research applies two-stage approach, the first model is stochastic frontier production function and the second model is the regression of technical efficiency on some selected factors In order to select the most appropriate model among OLS, Fixed Effects and Random Effects Models, the Langrange multiplier test and Hausman test are applied

The estimation of the first model shows the technical efficiency of each sector from the lowest score of domestically public sector, 66.66 percent to the highest level of foreign owned firms, 79.40 percent In more details, domestically owned firms in Indian engineering industry have lower technical efficiency than foreign firms do Meanwhile, no significant difference in technical efficiency is found between private sector and public sector firms The study also expresses the indication of process of efficiency convergence that the domestically owned firms tend to catch

up with foreign owned firms in terms of technical efficiency Both export and import intensity is found to be significant variables in explaining technical efficiency However, the results are not consistent for most of estimation With 10 years from 1990 to 2000, the researcher divided into 3 periods and the models are estimated for three cross-sections So, the advantage of panel data of 10 years is not fully employed

Ray and Ping, (200 1) use the data envelopment analysis to study changes in levels oftechnical efficiency overtime in China's state-owned enterprises in the period of 1980-1989 From DEA model, the technical efficiency score is estimated, then it is used as the dependent variable in the second model And, the explanatory variables are including: the bonus wage ration; the type of management control; the tax ratio; the level of government owning the enterprise; the ratio of output and input prices; capital per worker; ratio of non-production to production workers, firm size and industrial classification

The Tobit regression analysis provides some significant results The bonus system has strongest positively impact on technical efficiency And, among various systems, the lease, stock, asset management systems have the most contribution to technical efficiency This research proves that technical efficiency can be estimated

by DEA However, we should selected factors influencing the technical efficiency

by checking their correlation

The relationship between economic reforms and technical efficiency at firm level is analyzed by Parameswaran (2002) in some selected industries of India The author applies the frontier production fuction of trans log form with inputs of capital stock constructed perpetual inventory method, labour measured by total wage, and total cost of material and energy In order to measure the shift of the frontier production function over time, he inputs the variable of time (t) In the second model of technical inefficiency, the selected factors are technology import intensity, export intensity, raw material import intensity, R&D intensity and some dummy variables

The remarkable point of the research is that the author performs the test to select an appropriate model of production frontier function The testing results express the trans log functional form is selected and the variable oft is significant in the model implying shift of the production frontier over time

2.3.3 Studies in Vietnam

Minh and Long (2007) analyze the enterprise's technical efficiency of construction firms using both the data envelopment analysis model and the stochastic frontier production function model in their study, "Efficiency of Construction Firms in Vietnam: Assessment by Parametric and Non-parametric Approaches" In the parametric approach, they exploit the translog production frontier function as follows:

lnYi = lnj{Xi) = ~ajlnxj + ~~j(lnxi + ~yj(lnxj) (lnxh)- Vi+ Ui

Where: Y and X are output and input of the i firm;

In the model : krl which in net capital-labour ratio of each construction firm; rand

r2 are net revenue and squared net revenue; the dummy variable loc presents firm located in Hanoi or Hochiminh city; the dummy variable dnnn indicates the firm is

a state-owned; and sis random error

The findings show that estimations from both approaches are consistent with pure technical efficiency of about 60 percent Furthermore, technical efficiency of non-state firms is lower than state firms And, it is positively influenced by firm location, Hanoi and Hochiminh City Meanwhile, the capital-labour ratio has impact

on the efficiency in the parametric approach, it does not influence the efficiency performance in the non-parametric approach

There are some shortages in the research such as: it uses the data of 1 year; selected variables are not adequate with the operation of construction firms So, the findings are not rich enough to explain the technical efficiency

Viet and Charles (20 1 0) study the performance of domestic non-state manufacturing small and medium enterprises (SMEs) in Vietnam The two-stage model is used to estimate and analyze the efficiency of production The stochastic frontier model is the Transcendental-logarithm (Translog) form:

Ln Yi = ~o + ~ 1lnKi + ~2lnLi + ~3lnMEi + ~4(lnKi + ~sOnLi

+ ~6(lnMEi + ~7lnKilnLi + ~slnLilnMEi +Vi-Ui (2.11) Where: Yi is output of firm i; Ki is value of capital of firm I; Li is labour of firm i; MEi is value of materials and energy for firm i; Vi is random error following identical and independently normal distribution; Ui is technical inefficiency in which Ui is followed positively independently normal distribution and truncated at zero; The technical inefficiency model consists of the firm-specific and external environment variables The findings show that manufacturing SMEs have relatively high average technical efficiency from 84.2 percent to 92.5 percent In addition, it also reveals that firm age, size, location, ownership, cooperation with a foreign partner, subcontracting, product innovation, competition and government assistance

significantly related with technical efficiency However, export has no impact on technical efficiency

The study reveals some limitations: the sample is cross-sectional data in 3 periods

of 2002, 926 firms; of 2005, 2,228 firms and of 2007, 2,050 firms So, the technical efficiency in each year is separately estimated The technical efficiency doesn't express the trend However, we can find the trend of technical efficiency and the rate of technology catch-up by using panel data

Minh and Dong (2005) utilize both parametric and non-parametric to analyze the technical efficiency of 135 aquaculture-processing firms in Vietnam in 2002 With regard to the parametric model, the authors use the production function under the translog form with the variables including the output (VA) of a firm that is the total value-added of the firm; two inputs are net capital (K) and total number of workers (L ) The study shows some important findings as following: the pure technical efficiency is low at about 41.2 percent (as DEA model) and 67.6 percent (as SFPF model); no relationship is found between ownership, firm size and technical efficiency Meanwhile region, capital-labour ratio and the wage are positively related to technical efficiency The result of technical efficiency and its determinants are limited in explanation The technical efficiency score IS

inconsistent between two models The sample of data is only for one year that can cause those limitations

Chuc et al (2008) investigate the spillover impacts of the foreign direct investment (FDI) to technical efficiency of domestic small and medium size enterprises (SMEs)

in the paper "FDI Horizontal and Vertical Effects on Local Firm Technical Efficiency" From the stochastic frontier production model in form of a Cobb-Douglas function, the authors estimate the technical efficiency Then, to analyze the spillover effects ofFDI on technical efficiency, the following model is used:

(2.12)

Where: Horizontal, Backward and Forward are used as proxies for the horizontal and vertical effects ofFDI on local firms

Another form of the reduced model basing on a Cobb Douglas production function

in the models

The technical efficiency and its determinants are widely researched in developed and developing countries including Vietnam The popular methods used to measure technical efficiency are DEA and SFPF And, the estimations of technical efficiency are not much different in these studies from developed and developing countries The most significant point from studies of developing countries is that the researchers concern on the difference of technical efficiency about local entities and the counterpart of foreign In addition, they also estimate the different influence of dependent factors on technical efficiency between local entities and foreign-owned ones or entities having cooporation with foreign partners The results show that foreign have higher technical efficiency than domestically-owned firms (Ray and Ping, 2001) or they might have a positive relation with technical efficiency (Viet and Charles, 2010)

Table 2.1 Summary ofEmpirical Studies

1 Oleg et 2004; 35,000 determinants of of owners working in the firms; R&D

- The production function; - 1st model: output, inputs of capital stock and

Donghyu Panel data of decomposed - The dependent variables in the 2nd model:

2 net al Swedish components of capital intensity growth rate; market

(2009) 1992-2000 the error from the competition index; number of employees

1st model having at least bachelor's degree; ratio of including equity to total assets; wage growth rate technical change

Unbalanced Applying 4 panel data of different models R&D investments, the firm-specific Lerner

3 Elina ICT with both OLS index; the firm leverage ration; ownership (2006)

equipment and Maximum status of domestic and foreign; export status; 1990-2003 Likelihood size and age

estimations

- SFPF model

- Cross sectional Milk production (thousands of liters); total Alvarez Balance Data data on input

cost of labour; total farm area; number of

4 and of82 dairy quality is used to milking cows; total amount of feedstuffs; Gonzalez farms from adjust the

(1999) 1987to 1991 individuals roughage that accounting total expenditure to

effects from the produce forage crops

first stage Panel data of

SFPF model with Inputs comprise capital decomposed into ICT

14 member

5 Marco countries of time trend to and non ICT related capital; Labour

(201 0)

the EU from measure decompose into high, medium and low-skilled

1970 technical change labour; energy, meterials and services; output

The data in - 1st model: gross output; inputs: value of

1995 ofa

- SFPF model working capital, gross value of assets and

Wu large group - Determinants of number of employees

6

(2002) of firms technical - 2nd model: Depreciation; Productive assets;

located in 30 Agglomeration effect; Labour compensation; Chinese efficiency Incentive system; Taxation system; Dummy

Panel data of

- 1st model: the output is Gross value added;

63 firms in - SFPF model the -The inputs: gross fixed assets' Salary and Wages

7 Goldar et engineering determinants of - 2nd model: Export intensity; Import intensity;

al (2003)

industries, in technical vertical integration; R&D intensity; Adv

10 years, efficiency intensity; Liquidity Ratio; Excise duty rate;

- 1st model: inputs include labour; capital, materials and energy; gross industrial outputs;

A sample of - DEA model -2nd model: selected factors to study are the Ray, S.C 769 SOEs -The bonus wage ration; the type of management

8 and Ping, during 1980- determinants of control; the tax ratio; the level of government

z (2001) 1989 in technical owning the enterprise; the ratio of output and

China efficiency input prices; capital per worker; ratio of

non-production to non-production workers, firm size and industrial classification

- I st model: inputs of capital stock constructed Unbalanced by perpetual inventory method, labour

panel data of - SFPF model measured total wage and total cost of material Parames

640 firms for -The and energy; output of firm

- I st model and 2nd model: output is net

- SFPF model revenue; inputs are the average laborers in the Data of2,298

- DEA model year and the net capital Minh construction - 3rd model: the selected factors are the net

10 and Long firms in -The capital-labour ratio of each firm; net revenue

determinants of (2007) Vietnam in

technical and squared net revenue as proxies of the firm

2002

efficiency size; the dummy variable for firms located in

Hanoi and Hochiminh City and dummy variable for state-owned firm

Data of926

- I st model: inputs are the value of capital, firms in - SFPF model

Viet and 2002; 2,228 -The labour and Value of Materials and Energy

11 Charles firms in 2005 determinants of -2nd model of technical inefficiency, the (2010) and 2,050 technical selected factors: Age (year of operation); Size

firms in inefficiency (wage of worker); and some other dummy

- I 51 model: inputs are net capital and total Data of 135 - SFPF model number of workers

Minh and aquaculture- - DEAmodel - 2nd model of technical efficiency, the

12 Dong processing -The selected factors: firm size (average number of (2005) firms in determinants of workers); capital-labour ratio, wage per

Vietnam in technical employee, external-total cost ratio; capital

2002 inefficiency structure (total equity to total capital); and

region

- SFPF model -The model exploring the spillover effects Panel data of on technical - ]

51 model: output is total revenue affirm; 1,000 efficiency inputs are total asset and total permanent

13 Chuc et manufacturin - The reduced employees

2.4 Analytical framework for the research

From the above theoretical and empirical review, it's popularized and effective to analyze the technical efficiency of production by the stochastic production frontier function The analytical method is related to the concept of output-oriented technical efficiency first proposed by Farrell (1957) and popularized by Aigner et al., (1977) and Meeusen and Broeck (1977) The method following a two-stage approach is applied in many studies summarized on table 2.1and reviewed in this chapter

In the first stage, the econometric model can be applied to the production functions

in different forms such as the Cobb-Douglas, transcendental-logarithm, etc From the estimated frontier model, the technical efficiency is measured The technical efficiency can be estimated by the support of various soft-wares such as Frontier version 4.1 or Stata And, in many kinds of production function, the Cobb-Douglas function is used mostly because it is easy to understand This thesis also employs the Cobb-Douglas as the production function for manufacturing sector

In the second stage, the technical efficiency is first estimated from the production function And then a set of selected factors are selected to research their impact on technical efficiency of the firms

In summarizing the technical efficiency of manufacturing firms and its determinants, according to the theoretical and empirical evidence, the analytical framework is presented in the figure 2.2 below

Figure 2.2: Analytical Framework

l

Technical Efficiency (TE):

TE= Real Output/Potential Output

- Total capital-labour ratio (K2l)

- Years of operation (YearN ear)

- Liquidity ratio (Liq)

- Size of firm (Size)

- State-owned enterprise (StaEnt)

- Foreign-owned enterprise (ForEnt)

- Location in former Hanoi (Loc 1)

- Location in Hochiminh City (Loc2)

(3) To identifY factors influencing the technical efficiency of manufacturing firms

(4) To suggest appropriate policies for improving technical efficiency of manufacturing firms

CHAPTER 3: RESEARCH METHODOLOGY AND DATA COLLECTION

3.1 Introduction

This chapter includes two main sections The first section will introduce the models which are used in the thesis The second section will discuss how the thesis deals with proxy variables, data collection and data analysis The third section presents some hypothesis testing In the last section, data collection is explained

3.2 Research methodology

The thesis follows a two-stage approach In the first stage, the stochastic frontier model is applied to measure technical efficiency which is defined as the ratio of the observed output to the best practice output And in the second stage, technical efficiency is regressed against a set of factors and attributes that affect efficiency performance

3 2.1 The stochastic frontier model

(3.1)

Where

ln(Yi1) is the logarithm of the output for the i firm in the timet;

A is the the total factor productivity that has a common mean value for the intercept;

lnKit is the logarithm of the capital input for the i firm;

lnLit is the logarithm of the labour input for the i firm;

Xi1: are some individual characteristics of the firm such as location, ownership

a, p and 8 are unknown parameters to be estimated;

Vit is a random variation in output due to factors outside the control ofthe firm It's assumed to be independently and identically distributed as normal random variables with zero mean;

Uit is a non-negative random variable accounting for technical inefficiency in the production of the firm And, Uit is measured by two following methods:

-The time-invariant model (ti): Uit is assumed to have a truncated-normal random distribution and has constant value over time within panel Uit=Ui

- The time-varying decay model (tvd) or Battese-Coelli (1995) model: Ui is assumed a truncated-normal random variable and it may vary over time So, the inefficiency term is a truncated-normal random variable multiplying by a specific function of time

(3.2)

Where

T corresponds to the last time period in each panel;

11 is the decay parameter to be estimated If 11 > 0, the degree of inefficiency decreases over time; If 11 < 0, the degree of inefficiency increases over time; if 11 is zero, the inefficiency is constant during the period

Ui is assumed to be independent and identically distributed non-negative random variables that are obtained by the truncation at zero of the N(!J.,cr2) distribution

The maximum likelihood estimation is used to calculate technical efficiency as follows

Where

cr2 v is variance of noise and

cr2 is variance of inefficiency effects

if the value of a is equal to zero, then Ui is also zero which means the firms are fully efficient

y represents a total variation of actual output deviating from the frontier level of

output and has a value between 0 to 1 If the value of y is 0, the deviations are due

to technical inefficiency If the value of y is 1, the deviations from the frontier are

attributed to random error

The next part will present the measurement ofY, K and L variables

- The output value (Y): The output is total net turnover at the end of the year measured in millions of Viet Nam dong (VND) Total net turnover of enterprise gains by selling its products after subtracting taxes and other reduction such as discounting, reducing selling price, returning goods Net turnover doesn't include turnover gaining by financial activity or turnover gaining by special activity like selling off asset, getting money due to partner violating contract, etc

- The capital input (K): It is selected as total equity of enterprise at the end of the year and measured in million VND The capital belongs to proprietor of the enterprise, to member of joint-venture company or of shareholders in joint stock company, fund that is submitted to parent company by child companies, etc

- The labour input (L): The labour input can be taken as number of employee or total compensation of employees In the model, the labour input is total of persons that enterprise uses and pays wage

And in this thesis, the Stata software is used to run the stochastic frontiers model for panel data in the period 2000 to 2004 In order to compare the technical efficiency

of manufacturing firms, the estimation is carried out separately for the group of firms located in former Hanoi, Hochiminh city and other provinces; similarity with the group of state-owned firms, foreign-owned firms and other sectors

Table 3.1 Summary ofvariables in the frontier production function:

Name

Variable

variable type

y Numeric Output value, measured by net Million VND

turnover (million dong)

K Numeric Capital input measured by Owner's Million VND

equity at the end of the year Labour input measured by the

Number of

L Numeric number of employees at the end of

employees the year

StaEnt Dummy State-owned firms

ForEnt Dummy Foreign-owned firms

Locl Dummy Firms located in former Hanoi

Loc2 Dummy Firms located in Hochiminh city

3.2.2 The technical efficiency model

The technical efficiency for the ith company in the time t Is calculated as the conditional expectation of e-uit with respect to Cit (Cit= Vit - Uit):