Technical efficiency and its determinants the case of manufacturing firms in vietnam

The problem statement

Objectives of the research

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

This study aims to analyze the differences in technical efficiency among manufacturing firms based in former Hanoi and Ho Chi Minh City compared to those in other provinces Additionally, it will examine the efficiency variations between state-owned enterprises, foreign firms, and other types of businesses.

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

(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

Research questions

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

(1)What is the level of technical 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

Research methodology

The descriptive statistics, quantitative analysis are used to solve with the research questions.

The stochastic frontier production model in the form of Cobb-Douglas production function is applied to estimate and measure the technical efficiency of

This article examines the technical efficiency of manufacturing companies, specifically comparing firms located in former Hanoi and Ho Chi Minh City to those in other regions Additionally, it analyzes the differences in efficiency between state-owned and foreign manufacturing firms, highlighting the performance of various groups within the industry.

The thesis explores the factors affecting the technical efficiency of enterprises by analyzing panel data that incorporates both time series and cross-sectional dimensions Utilizing methods such as pooled Ordinary Least Squares (OLS), Random Effects Model (REM), and Fixed Effects Model (FEM), the research aims to provide insights into the efficiency dynamics within various enterprises.

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

Thesis structure

This thesis is structured into five chapters, with the subsequent chapters focusing on key areas of research Chapter 2 reviews the literature on production functions, technical efficiency, and the stochastic frontier production function, alongside relevant empirical studies Chapter 3 outlines the research methodology and the data utilized in the study.

4 presents the research results Finally, chapter 5 gives conclusions,recommendations and limitations of the study.

LITERATURE REVIEW

Basic Concepts and Theoretical Review

A production function defines the technical relationship between the quantities of productive factors utilized and the resulting output from various combinations of these factors, assuming the use of the most efficient production methods available.

A general production function can be written as:

Q is the quantity of output

X • › n are the quantity of factor inputs such as capital, labor, raw materials,

The production function illustrates the highest output achievable through a specific combination of inputs, considering the available technology In this context, inputs are flexible and can be substituted for one another.

The production function typically employs monetary values to represent the relationship between output and inputs, despite being fundamentally physical in nature This process involves various types of inputs that cannot be easily quantified in physical units and results in multiple outputs, often referred to as joint production, which are measured in different physical units To effectively manage this complexity, one approach is to aggregate the diverse products by applying price weights to each, as noted by 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 Cobb-Douglas functional form is the most commonly used model in econometrics, favored in applied research due to its mathematical simplicity and ease of use.

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

( t h e monetary value of all goods produced in a year)

A is total factor productivity or the technology state o and § are the output elasticities of labor and capital, respectively These values are constant, determined by available technology.

Output elasticity measures the responsiveness of output to a change in levels of either labor or capital used in production.

If o § = 1 : constant returns to scale;

Several methods exist for estimating the parameters of a Cobb-Douglas production function, with the most common approach utilizing a linear equation By applying logarithms to both sides of the equation, the function is transformed into a log-linear form, which facilitates easier analysis and interpretation of the relationship between output (Y), total factor productivity (A), labor (L), and capital (K).

Y, A, L and K are as defined earlier

The residual from estimation of or a distur bance term name d U,. The distur bance

U is differ ent for each firm and assu med to have norm al distri butio n.

According to Farrell (1957), total economic efficiency includes two components that are technical efficiency and allocative efficiency.

Technical efficiency refers to a firm's capacity to optimize output from a specific set of inputs or to reduce the inputs required for a certain level of output In contrast, allocative efficiency indicates a firm's proficiency in utilizing inputs in the most effective proportions based on market prices and the production technology employed.

Figure 2.1 illustrates the relationship between technical efficiency and scale efficiency in the production process The ABC line represents the frontier of maximum pure technical efficiency, while the through-origin line indicates scale efficiency, where points on this line demonstrate constant returns to scale Scale efficiency serves as a measure to determine if a firm is operating at its optimal scale.

efficiency, may be due to purely technical efficiency or scale efficiency

Points A, B, C, D, and E represent varying levels of input and output, with A, B, and C lying on the frontier line, indicating pure technical efficiency In contrast, points D and E fall below this frontier The tangential line at point B illustrates the constant returns to scale of technology, where the firm achieves both technical and scale efficiency due to its position on the frontier.

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

Observation D demonstrates inefficiency in both scale and technique, indicating that the same level of input could yield a higher output By shifting from point D, the firm has the potential to reach the production frontier between points B and C, maximizing efficiency and output.

Observation E demonstrates technical inefficiency as it operates below the production frontier; however, it is considered scale efficient, producing at an input level that is double the scale-efficient input level, while maintaining the same scale efficiency as point B.

Figure 2.1 Illustration of technical efficiency

Numerous researchers have analyzed the technical efficiency of the firm, employing various methodologies Two commonly used methods are the parametric approach known as the stochastic frontier production function (SFPF) and the non-parametric approach called data envelopment analysis (DEA).

A stochastic frontier production function approach is employed to estimate a production function based on a predetermined technology specification This methodology allows for the derived production function to be applicable across all firms within the same sector.

In SFPF, the disturbance term is comprised of two components, with the first representing technical inefficiency, characterized by a nonnegative distribution Technical inefficiency is defined as the gap between a firm's actual production level and the optimal frontier, as noted by Minh and Dong.

2005) The other component is assumed to have a symmetric distribution which refers to as random components.

Data Envelopment Analysis (DEA), introduced by Farrell in 1957, utilizes mathematical programming techniques to create a deterministic non-parametric frontier based on the observed input-output data of sample firms.

Data Envelopment Analysis (DEA) operates on minimal assumptions regarding production technology and does not require a specific functional form As outlined by Ray and Ping (2001), DEA empirically constructs a production frontier through mathematical programming techniques, utilizing observed input-output data from selected firms.

Empirical Studies

Research on technical efficiency has been extensively conducted in Vietnam and various other countries, primarily focusing on evaluating the technical efficiency within specific industries or comparing it across different sectors and firms in diverse locations Additionally, analyzing the factors that influence technical efficiency remains a compelling area of study This section will review several previous studies on the topic.

Oleg et a1 (2006) uses the panel data set with total of 35,000 firms in the period

Between 1992 and 2004, a study analyzed 256 industries using data from the German Cost Structure Census to evaluate technical efficiency The researchers aimed to uncover the relationship between various inputs, including material resources, labor compensation, energy consumption, capital, and external services, and the output of the production function, which was defined as the value of gross production minus subsidies and excise taxes.

The analysis indicates that the industry-specific effect is the primary determinant of variations in technical efficiency, followed by firm size and location as the second and third most significant factors Notably, smaller firms demonstrate greater operational efficiency compared to their larger counterparts While R&D intensity, outsourcing activities, and legal structure have a relatively moderate impact on technical efficiency, R&D intensity is found to negatively affect efficiency due to the time lag associated with R&D investments before yielding improvements Additionally, the study concludes that technical efficiency remains consistent over time, though it does not clarify the influence of yearly fluctuations on average efficiency levels.

This research provides valuable insights applicable to other studies, particularly through its use of panel data and a transcendental logarithmic (translog) production frontier function However, its findings may not be relevant for other research lacking essential input variables Notably, the study identifies a range of determinants affecting technical efficiency, encompassing both internal factors—such as firm size, outsourcing activities, and ownership structure—and external factors like industry affiliation, location, year effects, and market share.

Donghyun et al (2009) analyzed the productivity growth of the Swedish economy by examining panel data from 5,893 manufacturing and service firms between 1992 and 2000, resulting in a comprehensive dataset of 38,000 observations To estimate technical change and productivity growth, they utilized a production function, where the output (y) represents the firm's value-added, and the inputs (x) consist of various resources Additionally, the perpetual inventory method was employed as a proxy for capital stock.

From frontier production function, the authors estimate the error term, u,t Then it is specified as a two-way error component model as follows:

Where: p„ and v.t are firm-specific effects, time-specific effects and statistical noise, respectively In order to avoid over-paramiterization, firm-specification effects, p, are replaced by industry-specific effects, rd

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

Small firms tend to operate near their optimal production scale, while medium and large firms can improve efficiency by downsizing Additionally, estimates of technological change in production may be biased due to the influence of specific inputs, as changes in input proportions are linked to advancements in technology.

Elina (2006) analyzes the technical efficiency and factors contributing to inefficiency within the Finnish information and communication technology (ICT) manufacturing sector The study utilizes unbalanced panel data from 1990 to 2003, focusing on firms with a minimum of 20 employees in the ICT equipment manufacturing industry.

Key determinants of inefficiency include R&D investments, the firm-specific Lemer index (the ratio of operating profit to gross output), leverage ratios, ownership status (domestic vs foreign), exporter status, as well as firm size and age Findings indicate that the average firm operates at only about 56% of the technical efficiency level of the most efficient firms Moreover, technical efficiency varies significantly among firms, with time-varying efficiency averaging just over 40% of the reference rate of the most efficient firms.

This research highlights the use of a stochastic frontier model employing four distinct approaches to assess both time-invariant and time-varying efficiency levels Among these methodologies, the Battese-Coelli maximum likelihood model proves to be superior to the ordinary least squares model, while the translog production function is identified as the most effective option.

Alvarez and Gonzalez (1999) introduce a method that integrates panel and cross-sectional data to assess technical efficiency in dairy farming By analyzing a balanced panel of 82 dairy farms alongside cross-sectional data on input quality, they estimate the technical efficiency at an impressive 72% Their findings highlight the significance of input quality in determining efficiency levels within the dairy industry.

A key finding of the study is that technical efficiency is significantly influenced by the quality of inputs, particularly land and cattle This relationship was not identified in earlier research Additionally, while the initial analysis shows a positive correlation between technical efficiency and farm size, this relationship turns negative after adjustments are made.

Unobservable factors play a crucial role in clarifying the variations in technical efficiency, particularly when analyzed through the Corrected Ordinary Least Squares method This approach is effective only when pertinent information is available.

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:

Where: t is a time trend which captures the Hicks-neutral technical change; Y is output; K and L are capital and labour, respectively.

The study analyzes an unbalanced panel of 14 EU member countries from 1970 to 2005, employing the Kumbhakar and Lovel (2000) model to decompose Total Factor Productivity (TFP) growth into four components: technical change, scale component, technical efficiency change, and allocative inefficiency The findings highlight the trends in technology change, average TFP growth, and its individual components across the 14 EU countries during this period However, the research is limited by the constraints of the database used.

Estimating technical efficiency across all member countries requires a sufficiently large data set, which is challenging to obtain for countries outside the EU Therefore, the analysis involving 14 member countries may not provide a comprehensive assessment.

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

The study analyzes 5,160 observations using a two-stage approach to assess technical efficiency In the first stage, a standard frontier production function is utilized to estimate technical efficiency rates specific to various regions and sectors The second stage employs Tobin models to explore how regional and sectoral factors influence these technical efficiency rates.

Analytical framework for the research

The analysis of production technical efficiency through the stochastic production frontier function has gained popularity and proven effectiveness, rooted in the output-oriented technical efficiency concept introduced by Farrell in 1957 and further developed by Aigner et al and Meeusen and Broeck in 1977 This two-stage analytical method is widely utilized in various studies, as summarized in Table 2 and discussed in this chapter.

The econometric model can be utilized to analyze production functions, including forms like Cobb-Douglas and transcendental-logarithm From the estimated frontier model, technical efficiency is assessed, often using software such as Frontier version 4.1 or Stata Among various production functions, the Cobb-Douglas function is predominantly favored due to its simplicity This thesis specifically adopts the Cobb-Douglas production function for the 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.

Determinants of Technical Efficienc De endent variable.

Technical efficiency (TE) Ihde endent variables.

Total capital-labour ratio (K21) Years of operation (Year/Yeah)

- Liquidity ratio (Liq) Size of firm (Size) State-owned enterprise (StaEnt) Foreign-owned enterprise (ForEnt) Location in former Hanoi (Loc I) Location in Hochiminh City (Loc2)

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

This study aims to analyze the differences in technical efficiency among manufacturing firms in former Hanoi and Ho Chi Minh City compared to those in other provinces, as well as between state-owned enterprises, foreign firms, and private companies.

To identify factors influencing the technical efficiency of manufacturing firms.

To suggest appropriate policies for improving technical efficiency of manufacturing firms.

Labour (L): Number of employees Capital (K): Owner’s equity Output: Net turnover (Y)

Technical Efficienc TE TE= Real Output/Potential Output

RESEARCH METHODOLOGY AND DATA COLLECTION

Introduction

This chapter is organized into four key sections: the first introduces the models utilized in the thesis, while the second addresses the handling of proxy variables, data collection methods, and data analysis techniques The third section focuses on hypothesis testing, and the final section provides a detailed explanation of the data collection process.

Research methodology

This thesis employs a two-stage methodology to analyze technical efficiency Initially, it utilizes the stochastic frontier model to quantify technical efficiency, defined as the ratio of observed output to optimal output Subsequently, the study regresses technical efficiency against various factors and attributes that influence efficiency performance.

1n(Y;,) is the logarithm of the output for the i firm in the time t;

Total factor productivity (A) represents a common mean value for the intercept, while lnK represents the logarithm of capital input for firm i, and lnL denotes the logarithm of labor input for firm i Additionally, individual firm characteristics, such as location and ownership, are denoted by ×.t, with parameters o, §, and 8 being unknown values that require estimation.

V;a represents random variations in a firm's output caused by uncontrollable external factors, assumed to follow a normal distribution with a zero mean Additionally, the variable t indicates non-negative random variables that reflect technical inefficiencies in the firm's production process The measurement of Uit is determined through two specific methods.

- The time-invariant model (ti): ×.i is assumed to have a truncated-normal random distribution and has constant value over time within panel U:t'×

The time-varying decay model (TVD), also known as the Battese-Coelli model (1995), posits that the inefficiency term is represented as a truncated-normal random variable that can change over time This model incorporates a specific time-dependent function that influences the behavior of the inefficiency variable, allowing for a more dynamic analysis of efficiency in various contexts.

T corresponds to the last time period in each panel;

The decay parameter, rJ, is crucial for understanding inefficiency trends over time When rJ is greater than zero, it indicates a reduction in inefficiency as time progresses Conversely, if rJ is less than zero, inefficiency is on the rise A value of rJ equal to zero signifies that inefficiency remains constant throughout the observed period.

U, is assumed to be independent and identically distributed non-negative random variables that are obtained by the truncation at zero of the N(g,o') distribution.

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

‹i', is variance of noise and

If the value of O'u is zero, it indicates that U is also zero, signifying that firms operate at full efficiency The variable y reflects the total variation in actual output compared to the optimal output level, ranging from 0 to 1 A y value of 0 indicates that deviations from the frontier output are due to technical inefficiency, while a y value of 1 suggests that these deviations are the result of random error.

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

The output value (Y) represents the total net turnover at the end of the year, measured in millions of Vietnamese dong (VND) This figure is derived from the revenue generated by an enterprise through the sale of its products, after deducting taxes and other reductions such as discounts, price reductions, and returned goods Importantly, net turnover excludes revenue from financial activities or special activities, such as asset sales or compensation received from partners for contract violations.

The capital input (K) is defined as the total equity of an enterprise at the end of the fiscal year, expressed in million VND This capital is attributed to the enterprise's proprietor, members of joint-venture companies, shareholders in joint-stock companies, and funds allocated to the parent company by its subsidiaries.

Labour input (L) refers to the total number of employees or the overall compensation of employees within an enterprise In this model, it represents the aggregate of individuals that the business employs and compensates with wages.

This thesis utilizes Stata software to implement a stochastic frontiers model on panel data from 2000 to 2004, aiming to assess the technical efficiency of manufacturing firms The analysis distinguishes between firms located in former Hanoi, Ho Chi Minh City, and other provinces, as well as categorizing them into state-owned, foreign-owned, and other sectors for a comprehensive comparison.

Table 3.1 Summary of variables in the frontier production function:

Y Numeric Output value, measured by net turnover (million dong) Million VND

K Numeric Capital input measured by Owner's equity at the end of the year Million VND

L Numeric Labour input measured by the number of employees at the end of 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

The technical efficiency for the •th company in the time t is calculated as the

Then, the technical efficiency of a firm is modeled to analyze its determinants In this thesis, the selected determinants are as follows.

K21 represents the total capital-labour ratio, indicating the technical intensification of workers in manufacturing firms A higher K21 ratio signifies that a company is more capital-intensive, suggesting that its employees are equipped with more machinery Consequently, this leads to the expectation of increased output and enhanced technical efficiency within the firm.

The age of a business is determined by subtracting its establishment year from the year of the survey, indicating its operational longevity Generally, firms with more years in operation benefit from greater experience, improved management, and a more skilled workforce, which often leads to enhanced technical efficiency.

- Age2: It is the square of year of operation The relationship between year of operation and technical efficiency is expected as not a normal linearity.

- Size (Size of firm): There are six categories: firm has less than 10 employees= 1; 10-200 employees = 2; 201-300 employees = 3; 301-500 employees = 4; more than

The liquidity ratio, or Liq, is calculated by subtracting inventories from a company's current assets and then dividing by current liabilities This ratio reflects a firm's capability to fulfill its short-term financial obligations within a year A positive correlation is anticipated between the liquidity ratio and technical efficiency, indicating that higher liquidity may enhance operational performance.

The StaEnt variable identifies whether a firm is a state enterprise, assigning a value of 1 for state enterprises and 0 for non-state enterprises According to the Vietnam Enterprise Survey, state enterprises encompass central state, local state, and collective enterprises.

ForEnt, or Foreign Enterprise, is a binary variable indicating whether a firm is a foreign enterprise, assigned a value of 1 for foreign firms and 0 for domestic firms According to the Vietnam Enterprise Survey, foreign enterprises encompass those with 100% foreign capital, joint ventures with foreign partners, and other forms of collaboration with foreign entities.

- Locl (Location 1): It is a dummy variable which has two values as 1 if the firm is located in former Hanoi; and 0 if the firm is not located in former Hanoi.

- Loc2 (Location 2): It is a dummy variable which has two values as 1 if the firm is located in Hochiminh City; and 0 if the firm is not located in Hochiminh City.

Where: TE is technical efficiency of the firm i in the time t; and explanatory variables are explained in the table 3.2 below:

Table 3.2 Summary of variables in the technical efficiency model:

Capital-labor ratio represented by total capital of the firm divided by total number of employees

Age Numeric The years of operation Year

Age2 Numeric The square of years of operation

Six categories: firm has less than 10 employees= 1;10-200 employees = 2;

201-300 employees — 3;301-500 employees = 4; more than 500 employees = 5

Liquidity ratio is the ratio of current assets of a company minus inventories to current liabilities StaEnt Dummy State enterprise

Locl Dummy Firms located in former Hanoi

Loc2 Dummy Firms located in Hochiminh City

Testing Hypothesis

- The first hypothesis is to identify the technical inefficiency effects in the model that can be formulated as: : y=0 and H;: y>0

If H, is accepted, there are no technical inefficiency effects in the model In case the hypothesis is rejected, it’s concluded that there are technical inefficiency effects in the model.

- The second hypothesis is to identify whether manufacturing sector has constant returns to scale: : o §=1 and H;: o+§ J1

If He is accepted, it confirms the manufacturing industry is constant returns to scale.

- The third hypothesis is to identify whether technical inefficiency effect vary over time Hi: q=0 and H;: rJf 0

If Hi is accepted, it confums the technical inefficiency effect is time-invariant. And the alternative hypothesis is accepted proving that the technical inefficiency effects vary in the period.

This thesis utilizes panel data and requires the execution of various tests to determine the most suitable model among pooled Ordinary Least Squares (OLS), Fixed Effects Model (FEM), and Random Effects Model (REM).

The fourth hypothesis examines the relationship between technical efficiency and various selected factors, including the capital-labor ratio, years of operation, the squared years of operation, firm size, liquidity ratio, ownership type (state-owned versus foreign-owned), and the geographical location of firms, specifically those situated in major cities like former Hanoi or Ho Chi Minh City.

Data Collection

The data used in this thesis is the panel data from the Vietnam Enterprise Survey (VES) conducted by the General Statistic Office (GSO) in the period 2000 — 2004.

The dataset of manufacturing enterprises is derived from original data and categorizes firms into various sub-sectors based on their distinct production processes, material inputs, production equipment, and employee skills These sub-sectors include the manufacture of food products and beverages, tobacco, textiles and apparel, leather, wood, paper, chemicals, rubber, machinery, electronic products, motor vehicles and transport, furniture, and recycling.

In accordance with statistical regulations, each firm is assigned a unique code in the original dataset The manufacturing sector data, which includes 10,255 enterprises in 2000, shows a steady annual increase, reaching a total of 20,532 enterprises by 2004 From this dataset, only firms meeting specific criteria are selected for analysis.

- The firm has enough data of output, labour, owner’s equity, total capital, current asset, year of operation, etc in each year in period 2000 to 2004.

- The value of output, labour, owner’s equity, total capital, current asset is positive.

The firm operates under a unique code, but when two companies share the same code, additional criteria such as province code and year of establishment are used for differentiation, resulting in the removal of one firm from consideration.

The data set, extracted annually using Excel, is compiled into panel data covering the years 2000 to 2004 Manufacturing firms lacking information on key metrics such as output value, capital, labor, owner's equity, total capital, or current assets for any year within this period are excluded from the analysis.

Finally, the data set includes total 3,079 manufacturing companies that have

ANALYSIS RESULTS

Sample profile

Table 4.1 presents the descriptive statistics of production factors, categorized by manufacturing firms in former Hanoi, Ho Chi Minh City, and other provinces, as well as distinguishing between state-owned and foreign-owned enterprises.

Table 4.1: Descriptive statistics of output, capital and labour of manufacturing firms in the period 2000-2004:

From table 4.1, we have some statistical descriptions as follows:

Manufacturing firms in Ho Chi Minh City lead in capital and labor, with an average of 37,031 million VND and 417 employees, respectively, resulting in a high revenue of 84,749 million VND In contrast, former Hanoi's manufacturing firms exhibit the lowest averages in capital, labor, and output, which is surprising given the advantages typically associated with a major city like Hanoi.

Foreign-owned firms lead in owner capital with an average of 81,711 million VND, significantly surpassing State-owned firms, which average 37,031 million VND Additionally, foreign-owned firms employ an average of 521 employees, closely matching the 464 employees found in State firms.

Meanwhile, other sector with total of 1,645 firms has the lowest mean of labour

With a workforce of 168 employees and a capital of 6.030 million VND, many manufacturing firms are characterized by their small to medium size, indicating a low scale of labor and capital Graph 4.1 illustrates the employee structure across 1,645 firms, highlighting this trend in the industry.

Graph 4.1: The structure of 1,645 manufacturing firms from other sectors e Size4: 301-500 employees

Technical efficiency

The maximum likelihood estimates of the parameters from the time invariance inefficiency (ti model) and the time-varying inefficiency (tvd model) stochastic frontier production function are presented in table 4.2:

Table 4.2: Estimates of ti model and tvd model:

Time-varying inefficiency (tvd model)

P- value Coef S.E value t- value P- lnK o 0.467 0.009 51.07 0.000 0.412 0.009 43.71 0.000

Firms in former Hanoi Be 0.154 0.054 2.83 0.005 0.175 0.0552 3 18 0.001 Firms in HCM City By 0.143 0.040 3.55 0.000 0.182 0.0411 4.43 0.000 State-owned firms By -0.076 0.043 -1.77 0.076 -0.026 0.0436 -0.61 0.543 Foreign-owned firms 84 0.194 0.050 3.89 0.000 0.353 0.0512 6.90 0.000 constant LnA 6.565 1.424 4.61 0.000 6.541 0.2309 28.33 0.000 mu p 4.087 1.423 2.87 0.004 3.548 0.2200 16.13 0.000 eta q 0.020 0.0015 14.02 0.000

Sigma v2 o', 0.344 0.0044 0.326 0.0042 c (residual error) 0.0284 0.0282 var (u)/var (c) 0.845 0.851

Table 4.3: The statistical tests of some hypothesis:

Null hypothesis Chi2 p-value Decision

Hi: ti model nested in tvd model

Before concluding on the estimation of the stochastic frontier model, it is crucial to determine the presence of technical inefficiency effects The initial test, as shown in Table 4.3, rejects the null hypothesis (y=0) for both the ti and tvd models This finding aligns with the average technical efficiency results of 0.710 for the ti model and 0.715 for the tvd model, indicating that manufacturing firms are operating at approximately 71 percent of their potential output on average.

In both models, most parameter estimates show statistically significant p-values The ti model indicates output elasticities of 0.467 for capital and 0.600 for labor, while the tvd model shows elasticities of 0.412 for capital and 0.607 for labor These findings suggest that the labor's share of output significantly exceeds that of capital.

The combined output elasticity of the manufacturing sector is slightly above one, with values of 1.066 in the TI model and 1.018 in the TVD model, indicating increasing returns to scale This finding aligns with the results presented in Table 4.3, which reject the null hypothesis that the production function demonstrates constant returns to scale.

The positive and statistically significant coefficients for the locl and loc2 variables in both models indicate that manufacturing firms situated in major cities like former Hanoi and Ho Chi Minh City can achieve higher output levels compared to those located in other provinces.

The analysis reveals that the coefficient for the foreign variable is both positive and statistically significant, with values of 0.194 in the ti model and 0.353 in the tvd model In contrast, the State coefficient shows a negative value of 0.076 at a 10 percent significance level in the ti model, while it lacks statistical significance in the tvd model.

The rejection of the hypothesis that the ti model is nested within the tvd model highlights the greater relevance of the tvd model in explaining technical inefficiency Consequently, the tvd model is chosen as the preferred framework for analyzing the technical efficiency of the manufacturing sector.

In the TVD model, the positive estimate for the rj parameter (rj = 0.020) indicates a decline in technical inefficiency effects over time, suggesting a favorable technological catch-up rate within the manufacturing sector This finding is further supported by the results presented in Table 4.3, which reject the null hypothesis that the eta value (rJ) equals zero, confirming that efficiency levels fluctuate over time.

The variance ratios for the ti and tvd models are similar, with var(u) accounting for 84.50% and 85.10% of the estimated variance in the residual error term, respectively This indicates that approximately 85% of the variation in manufacturing production output can be attributed to differences in technical inefficiency.

Comparison of technical efficiency

Table 4.4 presents the technical efficiency of manufacturing firms across various regions, including those in former Hanoi, Ho Chi Minh City, and other provinces It also compares the technical efficiency among state-owned, foreign-owned, and manufacturing firms from other sectors.

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

Coefficiency of Capital - L Returns to Scale

2 Manufacturing firms informer Hanoi (378firms)

3 Manufacturing firms in Hochiminh City (833firms)

4 Manufacturing firms in other provinces (1,868firms)

5 State-owned manufacturing firms (810firms)

6 Foreign-ou›ned manufacturing firms (624firms)

7 Other sector’s manufacturing firms (1,645firms)

Note IRS increasing returns to scale, DRS decreasing returns to scale Source Author’s calculation

The analysis reveals that, regardless of the manufacturing firms located in various provinces, the technical efficiency results are consistent across all groups Notably, the technical efficiency in the time-varying parameter (tvd) model surpasses that of the time-invariant (ti) model, with a positive and statistically significant coefficient of eta Therefore, we can conclude that the manufacturing sector demonstrates a higher level of technical efficiency when assessed through the tvd model.

Vietnam exhibits a positive technological catch-up rate across all firm categories, with varying technical efficiency that improves over time Notably, firms in Ho Chi Minh City lead with a catch-up rate of 0.0323, while those in former Hanoi have the lowest at 0.0002 Additionally, foreign-owned firms demonstrate the highest catch-up rate of 0.046, in contrast to firms from other sectors, which show a lower rate of 0.017.

Most manufacturing firms, excluding those in Ho Chi Minh City's sector, experience increasing returns to scale Additionally, all firm groups demonstrate a higher output elasticity with respect to labor compared to capital.

The analysis of technical efficiency among 3,079 firms reveals that the output elasticity concerning capital is 0.467, while the output elasticity related to labor stands at 0.600 This indicates that, with labor input held constant, a 1% increase in capital input leads to a proportional increase in output.

45 in 0.467% increase in the output; and keeping capital input unchanged, 1% in labour input brings in 0.6% increase in the output.

* Technical efficiency by location of firms:

By locality, firms in former Hanoi attain the highest level of technical efficiency, 0.789; manufacturing firms located in other provinces have the lowest level, 0.688;

Foreign-owned firms demonstrate the highest technical efficiency, achieving a score of 0.831, while state-owned manufacturing firms follow closely with a technical efficiency score of 0.80, placing them in the second position.

Manufacturing firms of other sectors which are small scale of labour and capital have the lowest technical efficiency, 0.683.

Technical efficiency model

- From tvd model, the technical efficiency level is estimated first and then it is used to execute some testing process and analyze factors impacting on technical efficiency.

4.4 1 Testing for the most appropriate model

When analyzing panel data, three common methods are utilized: Pooled Ordinary Least Squares (POLS), Random Effects Model (REM), and Fixed Effects Model (FEM) Conducting tests to determine the most suitable model is essential for accurate analysis.

First, testing for random effects (POLS vs REM) by Breusch-Pagan Lagrange multiplier rejects the null hypothesis that variances across firms are zero So, POLS is un-appropriate (appendix 21).

The Hausman test, which compares Random Effects Model (REM) and Fixed Effects Model (FEM), evaluates whether the differences in coefficients between the two models are systematic The null hypothesis posits that unique errors are not correlated with the coefficients However, since the test rejects this null hypothesis, FEM is determined to be the most suitable model for the analysis.

A test for heteroskedasticity is performed to check whether group is heteroskedasticity in FEM And, it rejects the null hypothesis which is homoskedasticity and concludes the presence of heteroskedasticity (appendix 22).

I apply the fixed effects model to estimate the determinants of technical efficiency with data of 3,079 manufacturing firms in the period from 2000 to

2004 In order to control for the heteroskedasticity, the option “robust” in Stata is seclected The result of factors influencing the technical efficiency is summarized in the table 4.4.

Tabel 4.5: Determinants of technical efficiency

Te Coefficients Std Err p-value

Note: *; ** and *** denote significance at 10%, 5% and 1% levels, respectively

The model's p-value of zero indicates a rejection of the null hypothesis, which asserts that all coefficients are jointly equal to zero Additionally, the R-squared (within) value of 0.3343 suggests that the chosen factors in the model account for approximately 33.43% of the total variation in technical efficiency.

From the result in table 4.5, the determinants of technical efficiency are expressed in more details as follows:

The coefficients for the K21 and Liq variables are not statistically significant at the 10% level, while the other coefficients show statistical significance This suggests that the period from 2000 to 2004 may be insufficient to adequately assess the impact of these variables on technical efficiency.

The analysis of the efficiency model reveals that the parameters Age and Age squared (Age2) are statistically significant at the 1% level The findings indicate an inverted U-shaped relationship between age and performance, which can be expressed by the equation Te = f(Age, Age2) = 0.08305*Age - 0.00083*Age² This suggests that as age increases, performance initially improves before eventually declining.

To determine the maximum level of technical efficiency, we take the first derivative with respect to age and set it to zero This calculation reveals that the critical point occurs at the age of 50, indicating the year when technical efficiency reaches its peak.

After a firm is established, its technical efficiency tends to improve with years of operation, peaking at around 50 years before declining Initially, new machinery and skill development among employees contribute to rising efficiency Over time, firms reach their optimal production and business levels, but eventually, efficiency diminishes due to factors such as labor fluctuations and management challenges.

The Size variable demonstrates a statistically significant positive impact on technical efficiency, with a coefficient of 0.42275 at the 1% level This indicates that larger companies operate more efficiently than smaller ones in terms of employee quantity, suggesting that firms with greater economies of scale in labor are more efficient than those with lower scales.

Large firms typically enjoy more substantial and stable trading orders due to their scale and superior management compared to smaller companies However, they often encounter challenges such as administrative management issues, difficulties in fostering business innovation, and conflicts of interest within their management teams.

Chapter Summary

Manufacturing firms in Vietnam have increasing returns to scale; the output elasticity of labour is higher that of capital.

From 2000 to 2004, the technical efficiency of manufacturing firms exhibited variability, with an average efficiency rate of 71.50 percent Additionally, the manufacturing sector demonstrated a positive technological catch-up rate of 0.020, indicating progress in enhancing efficiency.

Significant disparities in technical efficiency exist among various manufacturing firms, with those based in former Hanoi exhibiting superior levels Foreign-owned firms achieve the highest technical efficiency scores, while state-owned firms demonstrate a level of efficiency comparable to their foreign counterparts.

The size of a firm significantly enhances its technical efficiency, while the years of operation also play a crucial role Manufacturing firms tend to reach peak operational efficiency around 50 years of age, after which their production efficiency begins to decline Additionally, the study reveals no significant correlation between the capital-labor ratio or liquidity ratio and technical efficiency.

CONCLUSIONS, RECOMMENDATION AND LIMITATIONS

The conclusions

This thesis utilizes the stochastic frontier model to evaluate the efficiency of 3,079 manufacturing companies from 2000 to 2004 and investigates the factors influencing technical efficiency during this timeframe Key conclusions highlight significant insights into the determinants of efficiency in the manufacturing sector.

Vietnam's manufacturing sector exhibits an average technical efficiency of 71.50 percent, surpassing the mean efficiency of 60 to 70 percent found in other developing countries, according to Tybout (2000) Nevertheless, this figure is lower than the technical efficiency levels recorded for small and medium manufacturing enterprises in Vietnam during the years 2002 and 2005.

2007 which are 84.3 percent, 92.5 percent and 92.3 percent, respectively, Viet and Charles (2010).

There are some differences in technical efficiency among manufacturing firms from different location and ownership.

Manufacturing companies in Hanoi and Ho Chi Minh City exhibit impressive technical efficiency rates of 78.90% and 76.30%, respectively, significantly surpassing the 68.80% efficiency found in firms located in other provinces These major cities offer numerous advantages, including access to a skilled workforce, the ability to implement advanced technologies and modern management practices, and cost-effective information technology and transportation options However, challenges such as high land rental costs and issues related to pollution and waste management hinder the technical efficiency of these firms.

Foreign-owned companies demonstrate the highest technical efficiency level at 83.10%, followed closely by state-owned firms at 80.00% The advantage of foreign-owned firms lies in their substantial capital, averaging three times that of all manufacturing firms, and their effective labor management, which is 1.6 times greater This strategic approach enables them to achieve an output that is 2.5 times higher, reinforcing the importance of employing a skilled workforce for optimal efficiency.

Manufacturing firms from other sectors have the lowest performance score at 68.30 percent According to Graph 4.1, 91 percent of these firms are categorized as small and medium-sized based on their workforce Additionally, Table 4.1 indicates that there are a total of 1,645 firms in this sector, with an average capital of 6,030 million VND and a workforce to support their operations.

Between 2000 and 2004, small and medium firms exhibited a notably low level of technical efficiency, with a significant disparity compared to the findings of Le and Harvies (2010), which indicated that the technical efficiency of these firms ranged from 83 percent to 92 percent.

Small and medium enterprises frequently encounter significant challenges due to limited capital and management resources This financial constraint hampers their ability to invest in advanced technology and hire skilled personnel Additionally, these businesses often lack transparent governance structures, operating informally with family oversight and adhering to unwritten rules (Klause et al., 2005).

The manufacturing sector exhibits increasing returns to scale across all firm categories Notably, the output elasticity concerning labor is 0.604, surpassing the output elasticity related to capital, which stands at 0.441 However, this does not imply that manufacturing firms should prioritize hiring additional workers over investing in machinery and technology to enhance production.

The manufacturing sector in Vietnam plays a crucial role by employing a significant number of unskilled or low-skilled workers, thereby generating numerous job opportunities despite offering low compensation This sector contributes to the production of various goods for society, but it is characterized by low added value Many manufacturing firms continue to benefit from these conditions, highlighting the need for improvement in workforce skills and product value.

Among determinants of technical efficiency of manufacturing sector in Vietnam, the size of firm influences positively and greatly technical efficiency.

Large enterprises benefit from economies of scale, allowing them to lower average unit costs as production increases and facilitating operational expansion Consequently, these larger firms demonstrate greater production efficiency In contrast, small businesses often struggle due to limited capital and an unstable workforce, which hampers their operational effectiveness and leads to lower technical efficiency.

Manufacturing firms in Vietnam, with an average operational age of just 11 years, have significant potential to enhance their technical efficiency over time As these firms mature, they can improve management capabilities, streamline production processes, and elevate worker skills Estimates suggest that the peak level of technical efficiency can be attained around the age of 50, highlighting the importance of experience in driving operational success.

Finally, the thesis does not find the relationship between capital labour ratio and liquidity ratio of manufacturing firm and its efficiency.

The recommendations

From the results of the analysis, the thesis provides some recommendations for manufacturing enterprise and the government.

Manufacturing firms experience greater labor elasticity compared to capital elasticity, indicating that increasing the workforce is more effective for boosting output than investing in capital However, it is essential for firms to implement effective policies that focus on hiring and utilizing experienced and skilled workers rather than merely increasing employee numbers By fostering a stable and talented workforce, companies can enhance production efficiency and achieve higher output levels.

The manufacturing sector is highly competitive, posing significant challenges for new and small enterprises that can hinder their operational efficiency Despite these obstacles, these firms can leverage their status as newcomers to enhance production efficiency through learning-by-doing and learning-by-watching By imitating techniques from foreign-owned enterprises, domestic companies, particularly smaller ones, can minimize their own learning costs and adopt more effective practices.

State-owned manufacturing firms demonstrate operational efficiency, and as the government privatizes these enterprises, it is crucial to foster conditions that support their production expansion Larger firms tend to operate more effectively than smaller ones, particularly in capital-intensive manufacturing subsectors Therefore, any expansion efforts must be accompanied by enhancements in management capabilities and a highly skilled workforce.

The government encourages foreign investment in the manufacturing sector, allowing these companies to utilize resources efficiently in their production processes This influx of foreign enterprises not only creates numerous job opportunities for skilled workers but also provides valuable insights for domestic firms, enabling them to learn and improve their own production methods.

Limitations

The thesis reveals significant insights into the technical efficiency of the manufacturing sector and its determinants through the use of panel data However, the author acknowledges certain limitations in the econometric analysis of the findings.

The dataset, covering only five years from 2000 to 2004, yields limited and non-diversified estimation results To better understand the variations in technical efficiency and its determinants, a dataset spanning a longer time frame would be beneficial Additionally, the factors influencing technical efficiency in this analysis are restricted in scope.

To enhance our understanding of technical efficiency, it is essential to explore additional factors, particularly endogenous variables By categorizing the manufacturing sector into specific sub-sectors—such as food and beverage, textiles, and rubber and plastic production—we can achieve more accurate research outcomes This approach allows for a deeper analysis, as firms within the same sub-sector share more relevant characteristics, leading to more meaningful results.

With the fixed effects model, the thesis does not examine time-invariant variables such as foreign and state factors, location variables influence on technical efficiency '

There are not many researches on technical efficiency using panel data of Vietnam firms, then the results of this thesis is limited in results comparison and analysis.

Afriat, S.N (1972), “Efficiency Estimation of Production Functions”, International

Economic Review, Vol 3, No 13, PP 568—598.

Aigner D.J and Chu S.F (1968), “On Estimating the Industry Production Function”, The American Economic Review, Vol 4, No 58, PP 826-—839.

Aigner, D J., Lovell, C.A and Schmidt, P (1976), “A Contribution to the empirics of Economic Growth”, The Quarterly Journal ofEconomics, Vol 107, No 2, PP 407-437.

Aigner, D., Lovell, C A K and Schmidt, P (1977), “Formulation and Estimation of Stochastic Frontier Production Function Models”, Journal ofEconometrics, Vol 6, PP 21—37.

Alvarez, A.M and Gonzalez, E (1999), “Using cross-section data to adjust technical efficiency indexes estimated with panel Data”, Amerian Journal of Agricultural Ecnonomics, 81:894-901.

Amadou, N (2007), “Analysis of factors affecting the technical efficiency of Arabica coffee producers in Cameroon”, African Economic Research Consortium, Research Paper No 163.

Battese, G.E and Coelli, T.J (1987), Prediction of Firm-level Technical Efficiencies with Generalized Frontier Production Function and Panel Data”, Journal ofEconometrics, Vol 38, Vol 3, PP 387—399.

Battese, G.E and Coelli, T.J (1995), "A Model for Technical Inefficiency Effects in a Stochastic Frontier Production Function for Panel Data”, Empirical Economics 20(2): 325- 332.

Bauer, W (1990), “Recent Developments in the Econometric Estimation of Frontiers”,

Cattle, N and White, B (2007), “Efficiency and Scale Economies in the Western

Australian Wheatbelt”, School ofAgricultural and Resource Economics, University of Western Australia.

Chuc, D N., Gary, S., David, S., Anh, N.N and Ngoc Q.P (2008), “FDI Horizontal and Vertical Effects on Local Firm Technical Efficiency”, Depocen, Working Paper Series No 2008/17.

Coelli, T.J (1996), “A Guide to FRONTIER Version 4.1: A Computer Program for Stochastic Frontier Production Function and Cost Function Estimation.”, Centerfor Efficiency and Productivity Analysis (CEPA), University of New England, Working Paper

Den, D T., Ancev, T and Harris, M (2007), “Technical Efficiency of Prawn Farms in the Mekong Delta, Vietnam”, AusAid CARD Program.

Donghyun, 0., Heshmati, A.S and Hans, L (2009), “Technical Change and Total Factor Productivity Growth for Swedish Manufacturing and Service Industries”, CESIS

Electronic Working Paper Series, Paper No 193.

Elina, B (2006), “Technical Efficiency in an R and D intensive Industry: Finnish ICT manufacturing”, Vatt-Keskustelualoitteita Vatt Discussion Papers, No 389.

Farrell, M.J (1957), “Measurement of Productive Efficiency”, Journal of the Royal

Fritsch, M and Stephan, A (2004), “What Causes Cross-industry Differences of Technical Efficiency? - An Empirical Investigation”, German Institute for Economic

General Statistic Office of Vietnam (GSO), (2001), Statistical Yearbook 2000, Hanoi Statistical Publishing House.

Goldar, B N., Renganathan, V S., and Rashmi, B (2003), “Ownership and Efficiency in Engineering Firms in India”, Indian Councilfor Research on International Economic be/‹if/onx, Working Paper No 115.

Gounder, R and Xayavong, V (2004), “Stochastic Frontier Analysis of New Zealand's Manufacturing Industries: Some Empirical Results”, Department of applied and international Economics, Discussion Paper No.04.07.

Greene, W.H (1980), “On the Estimation of a Flexible Frontier Production Function Model”, Journal ofEconometrics, vol 13, issuel: 101-115.

Hung, N.V (2006), “Measuring Efficiency of Vietnamese Commercial Banks: An Application of Data Envelopment Analysis (DEA)”.

Kim S and Shafi I M., (2009), “Factor Determinants of Total Factor Productivity Growth in Malaysian Manufacturing Industries: a decomposition analysis”, Crawford school of Economics and Government, The Australian National University, Canberra.

Klaus, E.M., Yen, T.T.T and Hung, V.N (2005), “Doing Business in Vietnam”, Center for East European Studies, Department of International Econocmics and Management.

Koopmans, T.C (195 l), “An Analysis of Production as Efficient Combination of Activities, Activity Analysis of Production and Allocation”, Cowles Commission for

“ Research in Economics, Monografia no 13, New York.

Llewelyn, V.L and Williams, J.R (1990), “Production Frontier, Panel Data, and Time- varying Technical Inefficiency”, Journal ofEconometrics, vol.46, issues 1-2:201-21 1 Mankiw, N.G., Romer, D and Weil, D.N (1992), “A Contribution to the empirics of Economic growth”.

Marco, F (2010), “Estimation and Decomposition of total factor productivity growth in the EU Manufacturing Sector: A Long Run Perspective”, ISAE, Italy, Journal of Applied Economic Sciences, Volume V/Issue 3(13)/Fall 2010.

Minh N.K and Long G.T (2005), “Efficiency of Construction Firms in Vietnam”, MPRA

Minh, N.K and Dong, V.Q (2005), “Parametric and Non-parametric Estimates ofTechnical Efficiency for Aquaculture-processing Firms in Vietnam”.

Mishra, S.K (2007), “A Brief History of Production Functions”, MPRA Paper, No.5254, posted 07 November 2007.

Parameswaran, M (2002), “Economic Reforms and Technical Efficiency: Firm Level Evidence from Seclected Industries in India”, the Centre ’s website, Working Paper No 339.

Ray, S.C and Ping, Z (2001), “Technical Efficiency of State Owned Enterprises in China (1980-1989): An Assessment of the Economic Reforms ”, Department ofEconomics, University of Connecticut, Working Paper No 2001-05.

Richmond, J (1974), “Estimating the Efficiency of Production”, International Economic

Robert, S.P and Daniel, L.R (2009), “Microeconomics”, 7th ed., Upper Saddle River, New

Sang, V.N and Kokkelenberg, E.C (1991), “Measuring Total Factor Productivity, Technical Change and the rate of Returns to Research and Development”, Centerfor

Economic Studies, U.S Bureau of the Census, Washington, D.C., Discussion Papers,

Shephard, R.W (1970), “Theory of Cost and Production Functions”, Princeton University Press, Princeton, NJ.

Trestini, S (2006), “Technical Efficiency of Italian Beef Cattle Production Under a Heteroscedastic Non-Neutral Production Frontier Approach”, TESAF Department,

Tybout, J.R (2000), “Manufacturing firms in Developing Countries: How Well Do They

Do, and Why?”, Journal of Economic Literature.

Viet, L and Charles, H (2010), “Firm Performance in Vietnam: Evidence from

Manufacturing Small and Medium Enterprises”, School ofEconomics, University of

Wang, J.Y and Blomstrom M (1992), “Foreign investment and technology transfer: A simple model”, European Economic Review, 36: 137-1555.

APPENDICES APPENDIX 1: Results of summary of all manufacturing firms in the period 2000-2004

Variable Mean Std Dev Min Max | Observations overall 67902.01 257335.4 1 8467547 | N - 15395 between 244018.2 19.4 5775797 | n = 3079 within 81805.12 -2127963 2759652 | T - 5

APPENDIX 2: Results of summary of former Hanoi’s manufacturing firms in the period 2000-2004

Min Max | Observations overall | 52505.08 138473.2 15 1656817 | N - 1890 between | 129852.5 47.6 1503653 | n - 378 within 48464.9 -506259.5 821746.9 | T - 5

APPENDIX 3: Results of summary of Hochiminh city’s manufacturing firms in the period 2000-2004

Min Max | Observations overall | 84749.05 274212.4 6 4748254 | N = 4165 between | 260324.9 153 3822103 | n = 833 within | 86536.09 -1493375 2081994 | T - 5 overall | 417.1405 1637.762 1 49756 | N = 4165 between | 1561.198 4.4 35192.6 | n = 833 within | 497.2594 -17179.46 14980.54 | T - S overall | 37030.51 149262.3 10 3433053 | N - 4165 between | 144113.3 10.8 2488468 | n = 833 within | 39122.1 -1072638 981615.5 | T = 5

APPENDIX 4: Results of summary of other provinces’ manufacturing firms in the period 2000-2004

Variable Mean Std Dev Min Max | Observations overall 1 63505.03 267503.3 1 8467547 | N - 9340 between | 253709.8 19.4 5775797 | n - 1868 within | 84952.44 -2132360 2755255 | T - 5

1 overall | 291.855 807.8368 1 20028 | N = 9340 between | 784.9994 2 13234.8 | n - 1868 within | 191.4162 -4518.945 7085.055 | T = 5 k overall | 24907.01 101464.9 13 2444251 | N - 9340 between | 97229.9 33.8 1659482 | n - 1868 within | 29077.85 -755072.8 1055495 | T - 5

APPENDIX 5: Results of summary of State’s manufacturing firms in the period 2000-2004

Dev Min Max | Observations overall | 87121.53 263628.8 1 4731648 | N - 4050 between | 253133.2 19.4 3822103 | n = 810 within | 74074.62 -952958.9 1951564 | T - 5 overall | 464.1109 885.7782 2 11101 | N = 4050 between | 872.8569 5 10288.8 | n - 810 within | 153.2206 -1194.289 3214.111 | T - 5 k overall | 30944.01 132021.6 10 3433053 | N - 4050 between | 128895.3 10.8 2488468 | n - 810 within | 28846.73 -813227 975529 | T - 5

APPENDIX 6: Results of summary of Foreign’s manufacturing firms in the period 2000-2004

Variable Mean Std Dev Min Max | Observations overall | 169699 461699.2 30 8467547 | N = 3120 between | 434221.6 305.8 5775797 | n = 624 within | 157669.3 -2026166 2861449 | T - 5 l overall | 520.6596 1965.861 5 49756 | N - 3120 between | 1868.639 9.2 35192.6 | n - 624 within | 614.2256 -17075.94 15084.06 | T - 5 k overall | 81711.16 191266 147 2724186 | N - 3120 between | 181725.1 852.2 2024304 | n = 624 within | 60008.65 -1027957 1112299 | T - 5

“ APPENDIX 7: Results of summary of Other sector’s manufacturing firms in the period 2000-2004

Variable Mean Std Dev Min Max | Observations overall | 19823.51 49685.92 3 1004170 | between | 45528.58 28.2 485267.2 | n - 1645 within | 19921.02 -349597.1 615872.9 | T - 5

1 overall | 167.5474 495.271 1 11663 | N = 8225 between | 480.9453 2 9510 | n - 1645 within | 118.7329 -3239.253 2676.947 | T = 5 k overall | 6029.695 15868.91 11 287106 | N = 8225 between | 14462.39 53.2 213077.6 | n = 1645 within | 6539.36 -150830.9 199135.9 | T - 5

APPENDIX 8: Results of time-invariant inefficiency model of all 3,079 manufacturing firms in the period 2000-2004

Number of obs = 15395 Number of groups = 3079

484542 lnl | 5999489 010475 57.27 0.000 5794184 6204795 locl | 1534861 0542879 2.83 0.005 0470839 2598884 loc2 | 1433019 0404067 3.55 0.000 0641063 2224975 staetp | -.0760441 0428747 -1.77 0.076 -.160077 0079887 foretp | 1943188 0498907 3.89 0.000 0965347 2921029 cons | 6.565329 1.424144 4.61 0.000 3.774058 9.3566

Obs per group: min avg - - max =

’ Wald chi2(6) - 14678.43 Log likelihood - -17621.917 Prob > chi2 - 0.0000

APPENDIX 9: Results of time-varying decay inefficiency model of all 3,079 manufacturing firms in the period 2000-2004

Obs per group: min avg max

4304276 lnl | 6064609 0103161 58.79 0.000 5862418 6266801 locl | 1753239 0552174 3.18 0.001 0670997 2835481 loc2 | 1820449 0411297 4.43 0.000 1014321 2626577 staetp | -.0264891 0435719 -0.61 0.543 -.1118885 0589103 foretp | 3529977 0511649 6.90 0.000 2527163 4532792 cons | 6.540999 2308662 28.33 0.000 6.08851 6.993489

1) [ilgtgamma] cons - 0 chi2( 1) - 795.88 Prob > chi2 = 0.0000

( 1) [ilgtgamma]_cons - 0 chi2( 1) - 786.63 Prob > chi2 = 0.0000

Time-varying decay inefficiency model Number of obs 15395

Group variable: i Number of groups 3079

APPENDIX 11: Hypothesis test whether the manufacturing sector having constant returns to scale

1) [lny]lnk + [lny]lnl - 1 chi2( 1) - 45.22 Prob > chi2 - 0.0000

( 1) [lny]lnk + [lny]lnl = 1 chi2( 1) - 3.28 Prob > chi2 = 0.0702

APPENDIX 12: Hypothesis test whether ti model nested in tvd model

(Assumption: ti nested in tvd)

Model | Obs 11(null) 11(model) df AIC BIC ti | 15395 -17621.92 10 35263.83 35340.25 tvd | 15395 -17364.97 11 34751.95 34836.01

Note: N=Obs used in calculating BIC; see [R] BIC note

APPENDIX 13: Hypothesis test whether rJ=0 in tvd model

1) [eta]_cons = 0 chi2( 1) = 196.57Prob > chi2 - 0.0000

APPENDIX 14: Results of ti and tvd models of 378 former Hanoi’s manufacturing firms in the period 2000-2004

Number of obs Number of groups

Obs per group: min - avg - max -

Wald chi2(2) - 2106.99 Log likelihood - -1863.4897 Prob > chi2 - 0.0000 lny lnk

- Time-varying decay inefficiency model Group variable: i

Obs per group: min avg = max -

/inu 1 52 6 2 634 194 97 13 2 6 9 9 18 0.000 525.8813 526.6455 /eta | 000175 0000158 11.09 0.000 0001441 0002059 /lnsigma2 | 0304769 0658885 0.46 0 644 — 0986622 1596159 /ilgtgamma | 1.316424 0956989 13.76 0 000 1 128857 1.50399 sigma2 | 1.030946 0679275 9060487 1.17306 gamma | 7885861 0159547 755628 8181688 siqma_u2 ] 8129897 0683548 6790168 9469626 sigma v2 J 2179563 008121 2020394 2338732

' APPENDIX 15: Results of ti and tvd models of 833 Hochiminh city’s manufacturing firms in the period 2000-2004

Log likelihood = -3915.1078 Prob ằ chi2 = 0.0000 lny lnk

/mu | 2.644587 2853867 9.27 0.000 2.08524 3.203935 /eta | 0323194 0035119 9.20 0.000 0254362 0392026 /lnsigma2 | -.1096812 0430941 -2.55 0.011 -.1941441 -.0252184 /ilgtgamma | 1.169187 063546 18.40 0.000 1.044639 1.293735 sigma2 | 896l197 0386175 8235392 9750969 gamma 7629981 0114912 7397442 7847787 sigma u2 6837376 0386702 6079454 7595298 sigma_v2 212382l 0052841 2020254 2227388

Obs per group: min avg

Time-varying decay inefficiency model Number of obs = 4165

Group variable: i Number of groups = 833

APPENDIX 16: Results of ti and tvd models of 1,868 other provinces manufacturing firms in the period 2000-2004

Obs per group: min - 5 avg - 5 max - 5

Time-varying decay inefficiency model

Number of obs - Number of groups -

Obs per group: min = avg - max -

/inu | 3 4 8 8 37 S 2 4 00 1 4 6 1 4 53 0.000 3.017955 3.958795 /eta | 0179619 001724 10.42 0.000 0145829 021341 /lnsigma2 | 2495572 0260966 9.56 0.000 1984088 3007057 /ilgtgamma | 7894952 0423517 18.64 0.000 7064873 872503 sigma2 | 1.283457 0334939 1.219461 1.350812 qaiama I 68 7 7 2 2 9 00 90 9 55 6696245 7052663 sigma_u2 | 8826628 0333337 8173299 9479957 sigma v2 | 4007942 0066287 3878023 4137862

Number of obs - 9340 Number of groups = 1868

Wald chi2(2) - 8312.06 Log likelihood - -11465.588 Prob > chi2 - 0.0000 lny lnk

' APPENDIX 17: Results of ti and tvd models of 810 State’s manufacturing firms the period 2000-2004

Time-invariant inefficiency model Group variable: i

Number of obs - Number of groups = Obs per group:

, Time-varying decay inefficiency model Group variable: i Time variable: t

Number of obs - Number of groups -

Obs per group: min = avg = max -

/mu | 6.656491 3.223679 2.06 0.039 3381963 12.97479 /eta | 0124029 0057868 2.14 0.032 0010609 0237449 /lnsigma2 | -.078794 0511712 -1.54 0.124 -.1790878 0214998 /ilgtgamma | 1.385229 072335 19.15 0.000 1.243455 1.527003 sigma2 | 9242303 047294 8360325 1.021733 gamma | 7998295 011581 7761648 8215674 sigma_u2 | 7392266 047665 645805 8326483 sigma_v2 | l850037 0047297 1757337 1942737 min = max avg -

APPENDIX 18: Results of ti and tvd models of 624 Foreign’s manufacturing firms in the period 2000-2004

Obs per group: min = 5 avg - 5 max - 5

Time variable: t Obs per group: min - avg - max -

Time-varying decay inefficiency model Number of obs - 3120

Group variable: i Number of groups = 624

APPENDIX 19: Results of ti and tvd models of 1,645 other sector’s manufacturing firms in the period 2000-2004

Obs per group: min = 5 avg - 5 max - 5

, Time-varying decay inefficiency model Group variable: i Time variable: t

Number of obs = 8225 Number of groups - 1645

” Appendix 20: Descriptive statistic of efficiency factors

85 I T - 5 liq overall | 3.575912 41.7722 0 4258.33 | N = 15395 between | 21.18007 024 858.52 | n = 3079 within | 36.00609 -854.6141 3403.386 | T = 5 staetp overall | 2630724 4403157 0 1 | N = 15395 between | 4403729 0 n - 3079 within | 0 2630724 2630724 | T - 5 foretp overall | 2026632 40l9967 0 1 | N - 15395 between | 4020489 0 1 | n = 3079 within | 0 2026632 2026632 | T - 5 locl overall | 1227671 3281804 0 N - 15395 between | 328223 0 n - 3079 within | 0 1227671 1227671 | T - 5 loc2 overall | 2705424 4442545 0 N - 15395 between I 4443122 0 n = 3079 within | 0 2705424 2 7 0 5 4 2

Appendix 21: Testing OLS vs REM and testing REM vs FEM

• - Testing OLS vs REM xttest0 Breusch and Pagan Lagrangian multiplier test for random effects te[i,t] = Xb + u[i] + e[i,t]

Var sd = sqrt(Var) te | 2.856562 e | 0858215 u | 5711535

Test: Var(u) = 0 chi2(1) - 19437.05 Prob > chi2 = 0.0000

- Testing REM vs FEM hausman fem rem

(b) (B) (b-B) sqrt(diag(V_b-V_B)) fem rem Difference S.E. k21 | —.0000309 000052 -.000083 aqe ! 0830454 0686352 0144102 0008111 age2 | -.0008323 -.0010049 0001726 000 6 size | 4227513 5829311 -.1601798 002 4 7 4 liq | -.000015 -.0000217 6.69e-06 b - consistent under Ho and Ha; obtained from xtreg B = inconsistent under Ha, efficient under No; obtained from xtreg

Test: Ho: difference in coefficients not systematic chi2(5) - (b-B)' [(V_b-

Appendix 22: Test for heteroskedasticity for FEM xttest3

Modified Wald test for groupwise heteroskedasticity in fixed effect regression model H0: sigma(i)’2 - sigma’2 for all i chi2 (3079) - l.0e+07 Prob>chi2 = 0.0000

Appendix 23: The best model of determinant on technical efficiency xtreg te k2l age age2 size liq staetp foretp locl loc2, fe vce(robust) Fixed-effects (within) regression Number of obs = 15395

Group variable: i Number of groups - 3079

(Std Err adjusted for 3079 clusters in i) te | Coef.

Interval] k2l | -.0000309 000051 -0.61 0.544 -.0001309 0000691 age | 0830454 0038507 21.57 0.000 0754952 0905956 age2 | -.0008323 0001051 -7.92 0.000 -.0010383 -.0006263 size | 4227513 0l17802 35.89 0.000 3996534 4458492 liq | -.000015 0000529 -0.28 0.777 -.0001187 0000888 staetp foretp

(omitted) (omitted) locl | (omitted) loc2

(omitted) 10.86161 0395404 274.70 0.000 10.78408 10.93914 sigma u | 1.3132341 sigma_e | 29295314 rho | 9525954 (fraction of variance due to u i)

Tiêu đề	Technical Efficiency And Its Determinants: The Case Of Manufacturing Firms In Vietnam
Tác giả	Tran Van Khue
Người hướng dẫn	Dr. Nguyen Trong Hoai, Dr. Pham Le Thong
Trường học	University of Economics Ho Chi Minh City
Chuyên ngành	Development Economics
Thể loại	thesis
Năm xuất bản	2011
Thành phố	Ho Chi Minh City

Định dạng
Số trang	84
Dung lượng	498,82 KB