The Fourth Industrial Revolution brings both opportunities and challenges

is expected to clarify the theoretical framework of technological progress impacting labor productivity growth, using the DEA and SFA to evaluate the contribution of technological progress to increasing labor productivity in Vietnam. (3) On the basis of the selected model, the study will deeply consider the current Vietnam’s statistical data situation, study how statistical processing to best exploit source of current secondary statistical data. (4) On the basis of research on the relationship of the impact of technological progress on increasing labor productivity, the thesis is expected to highlight the significance and role of technological progress in Vietnam’s labor productivity increase and propose solutions for promoting S&T to increase labor productivity, thereby promoting economic development and improving the quality of life. At the same time, the thesis is also expected to propose the supplement of necessary statistical data for the purpose of studying the impact of technological progress on increasing labor productivity. CHAPTER 2: RESEARCH METHODOLOGY FOR STUDYING LABOR PRODUCTIVITY AND THE IMPACT OF TECHNOLOGICAL PROGRESS ON LABOR PRODUCTIVITY INCREASE 2.1 Definitions 2.1.1 Productivity definition Productivity is understood as the output created in relation to the inputs of the production and service delivery process, it demonstrates the efficiency of using input resources in creating outputs. 2.1.2 Labor productivity and labor productivity increase OECD (2001) defines labor productivity as: “quantity index of gross output divided by quantity index of labor input”. 7 In economics, output can be measured by gross value of production or value added. Labor input can be measured by the number of employees, the number of employees converted to full-time employees, working hours, the labor input adjusted by quality. Increasing labor productivity is to increase the quantity or value produced from a unit of labor used. Therefore, the increase in labor productivity is of great significance, is an important factor to increase more products for the society, is the basis for reducing costs, contributing to improving people's lives and increasing accumulation to develop production and increasing the competitiveness of the economy. 2.1.3 Technological progress Technical progress or technological progress are new and better ways to create products and services, new techniques to use scarce resources in a more productive way. 2.2 Evaluate the impact of technological progress on labor productivity increase The thesis presents an approach to assess the contribution of technological progress in labor productivity increase, using two groups of methods: non-parametric method and parametric method. 2.2.1 Non-parametric approach to estimate the technology progress Data Envelopment Analysis (DEA) based non-parametric approach. Malmquist productivity index based on DEA (DEA - based MPI) is divided TFP change into two components: "efficiency change" and "technology change". Suppose there are n DMUs with m inputs and s outputs. The symbol tijx , tijy and 1+tijx , 1+t ijy , are the inputs and outputs of DMUj at times t and t + 1 respectively, where i = 1,., M; r = 1,, s; and j = 1,, n. MPIo = 2/1 1 11111 ),( ),( *),( ),(       + +++++ t o t o t o t o t o t o t o t o t o t o t o t o yxD yxD yxD yxD Where, ),( tototo yxD and ),( 111 +++ tototo yxD measure the efficiency of DMUo ( { }no ,...,2,1∈ ) in times t and t + 1, respectively, ),( 11 ++ tototo yxD measure efficiency in time t + 1 using the production technology of time t, it is called the DMUo's growth index, ),(1 tototo yxD + measuring the efficiency of the DMUo in period t using the production technology of time t + 1. According to Făre et al., MPIo> 1 indicates an increase in productivity, MPIo = 1 indicates an unchanged productivity, and MPIo<1 indicates a decrease in productivity. 8 To remove the hypothesis of ),( tototo yxD and ),( 111 +++ tototo yxD are constant (by Caves et al.) and accept technical inefficiencies, Făre et al. decompose MPI into two parts: MPIo = 2/1 1 11111 ),( ),( *),( ),(       + +++++ t oo t o t o t o t o t o t o t o t o t o t o yxD yxD yxD yxD = 2/1 111 11 1 000 111 ),( ),( *),( ),( ),( ),(       +++ ++ + +++ t o t o t o t o t o t o t o t o t o ttt t o t o t o t o t o t o yxD yxD yxD yxD yxD yxD The first component measures the efficiency change (EC) of the DMU0: ECo = ),( ),( 000 1 0 1 0 1 0 ttt ttt yxD yxD +++ The second component measures the technological change (TC) of the DMUo from time t to t + 1 TCo = 2/1 111 11 1 ),( ),( *),( ),(       +++ ++ + t o t o t o t o t o t o t o t o t o t o t o t o yxD yxD yxD yxD Measuring efficiency change can be decomposed into pure technical efficiency change and scale efficiency change. Pure technical efficiency change = ),( ),( 1011 t o t o t ov tt o t ov xqD xqD +++ Scale efficiency change = 2/11111111111 ),(/),( ),(/),( *),(/),( ),(/),(       ++++++++++ t o t o t oc t o t o t ov t o t o t oc t o t o t ov t o t o t oc t o t o t ov t o t o t oc t o t o t ov xqDxqD xqDxqD xqDxqD xqDxqD In which Dov is an assumed function of distance with a variable return on scale and Doc is an assumed function of distance with constant return on scale. In the above model, to solve the requirement of assessing technological progress by decomposing the Malmquist productivity index, the suitable variable to represent y (the dependent variable) are the added value. The input variable x includes 2 variables: labor and capital. Then, we have a DEA model with 2 inputs and 1 output. 2.2.2 Parametric approach Coelli, O’Donnell and Battese (2005) explain the innovative technologies that bring about economic changes over time. If observed over time, technological changes can be calculated by an econometric model. For example, the following models are used: Linear: n N n no xty ∑ = ++= 1 βθβ 9 Cobb-Douglas: n N n no tAy lnln 1 ∑ = += βθ Translog: mn M m nm N n N n nno xxxtty lnln2 1lnln 111 2 21 ∑∑∑ === ++++= ββθθβ Where t is a time trend; and θ, θ1, θ2 are unknown parameters to be estimated. When economist include time trends in their models, they are making implicit assumptions about the nature of technological change. To see this, take the above specifications and consider the percentage change in y in each period due to technological change. This is given by the derivative of ln y with respect to t: Linear: yt y θ = ∂ ∂ ln Cobb-Douglas θ= ∂ ∂ t yln Translog t t y 21 2 ln θθ += ∂ ∂ Neoclassical theory assumes that enterprises make full use of technology and full efficiency. But in reality, production does not achieve this, Farrell (1957) suggests determining the output of the most efficient firms as the production frontier, firms that have not reached the frontier are inefficient firms. Based on Farrell's studies, Aigner and Chu (1968) developed a production function and then Aigner, Lovell and Schmidt (1977) and Meeusen and van den Broeck (1977) proposed the stochastic frontier production function model of the form: ii t ii uvxq −+= βln In which, qi represents the output of the i-th firm; xi is logarithm of inputs; β is a vector of unknown parameters, vi is a symmetric random error to account for statistical noise and ui is a non-negative random variable associated with technical inefficiency. Once the equation has been determined, the unknown parameters will be estimated by econometric techniques. Developing further studies by Farrell, Aigner et al., Battese and Coelli (1995) put the model as follows: iiiii uvxq −++= lnln 0 ββ or )lnexp( 0 iiiii uvxq −++= ββ )exp(*)exp(*)lnexp( 0 iiiii uvxq −+= ββ Deterministic component noise ineffciency 10 Assumption that two inputs are capital and labor, the Cobb-Douglas stochastic frontier model as follows : ittitkitlitit vtkluy ++++=+ ββββ lnlnln 0 Where, yit is the output, lit is labor, kit is capital and t represents technological progress. Translog stochastic frontier model as follow : ln yit + uit = β0 + βtt + βllnlit + βklnkit + 0,5βttt2 + 0,5βkk (lnkit)2+ 0,5βll (lnlit)2 + βtl t lnlit + βtk t lnkit + βlk lnkitlnlit +vit, In which, yit is the output lit is labor, kt is capital, t is the trend of time which is a variable representing technological progress and βs is the coefficient to be estimated, random error, exp (v), and ineffective, exp (u). The production function consists of three main variables k, l and t, (capital, labor and technology change) that affect other variables and interact with each other. t2 represents technological change in itself that will further push technological change in the coming years. tlnk: technical change pushing capital increase, and tlnl is technical change pushing labor increase. First, output growth is interpreted into three generating factors: input change, scale efficiency changes and TFP change. tfpllkky gggg ++= εε In which, g is the symbol of the growth rate, ɛk and ɛl are output elasticity for capital and labor, respectively. P. W. Bauer (1990) has decomposed the increase of TFP into 3 components: technological progress, technical efficiency change and scale efficiency change. i. Technological progress: ittlittkttt lktt tlkf lnln2),,(ln ββββ +++= ∂ ∂ ii. Scale effect: ( ))(1 l lk l k lk k lk gg εε ε εε ε εε + + + −+ iii. Change in technical efficiency: . t uu ∂ ∂ −=− • Thus, increasing TFP can be written as follows: 11 • −+ + + + −++ ∂ ∂ = ugg t tlkfg l lk l k lk k lkTFP ))(1( ),,(ln εε ε εε ε εε From Solow's model: g (Y/L) = α*g(K/L) + g(TFP), we have technological progress that affects labor productivity in the following way: g(Y/L) = α*g(K/L) + • −+ + + + −++ ∂ ∂ ugg t tlkf l lk l k lk k lk ))(1( ),,(ln εε ε εε ε εε By this approach, through the decomposition of factors in the TFP, it is possible to evaluate the contribution of technological progress to the increase in labor productivity based on the data on output (added value), input (labor and capital) over time. CHAPTER 3: LABOR PRODUCTIVITY AND THE IMPACT OF TECHNOLOGICAL PROGRESS ON INCREASING LABOR PRODUCTIVITY IN VIETNAM 3.1 Labor productivity assessment 3.1.1 Data processing to measure labor productivity and labor productivity increase The thesis calculates the per-worker labor productivity and per-hour labor productivity at economy level and economic sectors level. Data used include: GDP at current prices and constant prices, total number of employees and working hours (classified by industry). (1) GDP or added value From the General Statistics Office's data sources in the statistical yearbook, GDP at current prices and constant prices by economic sector are obtained. (2) Labor + Total number of employees: Data on the number of employees of the whole economy and for each economic sector can be obtained from the statistics source in the Statistical Yearbook. + Working hour: Labor productivity increased by technology progress Labor productivity increased by scale effect (SE) Labor productivity increased by technical efficiency (TE) Labor productivity increase Labor productivity increased by capital intensity 12 Working hour is not available in the Statistical Yearbook, so some additional data processing steps are required. From the annual employment report of the General Statistics Office, there are data on the average working hours per person per week of the economy, thereby calculating the total working hours as follow: Total working hours per year = Average working hours per person per week x Number of working weeks per year x Number of employees. + Labor by industry: The data on the working hours by kinds of economic sectors is not included in the Statistical Yearbook, so it should be processed and calculated by using the same method apply for the whole economy. From this data source, the total working hours by economic sector can be calculated by applying the above formula. 3.1.2 Overall labor productivity assessment of Vietnam's economy Labor productivity by worker in 2018 reached 102.0 million VND per person. Labor productivity by hour is 46.0 thousand VND per hour. The average growth rate of labor productivity in the period 2011 - 2018 in terms of the number of employees achieved is 4.87%, the average growth rate of labor productivity calculated by the working hour is 5.29%. Vietnam’s labor productivity is still low in comparing with some developing countries in Asia, and needs to be considered seriously to offer solutions to improve productivity. 3.2 Study the impact of technological progress on increasing labor productivity in Vietnam 3.2.1 Data and additional processing of capital data The data needed to use the model to evaluate the impact of technological progress on increasing labor productivity include: added value, labor and capital. The added value and capital are calculated at constant prices to eliminate price effect. For assessment of the economy, statistical data can be obtained for kinds of economic activity (level I - 20 economic sectors). The added value and labor data are mentioned above. Capital data are not available, because we use capital for production and business activities (often used as fixed assets), so some additional data processing is needed. According to OECD (2001), capital service is an appropriate measure to analyze productivity. Since capital for production is often not directly observable, it should be 13 estimated by assuming that capital service corresponds to a proportion of the capital stock that is transferred into production every year. Capital service ptktk SK ,, σ= In which, ptkS , is capital stock, σ is a ratio. For a given period σ is a constant. Calculating the capital stock requires long time series data on past investments and an initial amount of capital ( S ). Current capital stock 1 1 0 1 )1()1( − − = − ∑ −+−= t t i t t ISS δδ In which, St 1)1( −−δ is the remaining amount of original equity after deducting depreciation δ each year, 1)1( −− ti Iδ is the remaining investment after depreciation. Calculating the capital stock according to the above formula, it is necessary to have (i) data on investment (or asset accumulation) over time series, (ii) initial capital at the beginning of time series and (iii) the depreciation rate. Annual investment (or asset accumulation) data are obtained from GSO’s data, initial capital amount and depreciation rates are not included in the statistical data, so it is necessary to estimate. The thesis refers to the Perpetual Inventory Method (PIM) to calculate the initial capital. The idea of the PIM is to interpret the capital stock of the economy as an amount of inventory. Inventory is increased with capital accumulation (or capital investments). The maximum amount of capital is right after the investment and gradually decreasing over time, the amount decreasing each period depended on the depreciation rate. Once the initial capital is available, the capital for the following years is calculated based on the increase in capital investment and the increase in fixed assets in the year obtained from the Statistical Yearbook. After calculating the capital stock of the whole economy, it is possible to allocate capital to economic sectors based on the capital structure. 3.2.2 Using non-parametric approach Based on processed statistical data of economic sectors (level I), with added value data, labor and capital of each economic sector from 2010 - 2018, using DEA software, calculating the Malmquist index Based on DEA - CRS (input oriented), the estimation results are summarized as follows: 14 Table 1: The results of the Malmquist index calculated from the added value data, the number of employees and capital stock of the economic sectors (2011- 2018) Year Efficiency change (Effch) Technical change (Techch) Pure efficiency change (Pech) Scale efficiency change (Sech) TFP change (Tfpch) 2011 1.032 0.920 1.001 1.031 0.950 2012 1.043 0.919 1.031 1.012 0.959 2013 1.016 0.967 1.008 1.008 0.983 2014 0.996 0.984 1.009 0.987 0.981 2015 0.958 1.007 1.007 0.952 0.965 2016 0.986 1.017 1.012 0.974 1.002 2017 0.972 1.053 0.991 0.982 1.024 2018 0.954 1.035 0.997 0.957 0.987 Average 2010-2018 0.994 0.987 1.007 0.988 0.981 The Malmquist TFP index is calculated to be 0.981, less than 1. During the period from 2010 to 2018, the TFP decreases due to a decrease in technological progress (0.987) and scale efficiency (0.988). However, in terms of trend, technological progress has gradually increased, the technological progress index is greater than 1 from 2015 to 2018. Table 2: The results of the Malmquist index calculated from added value data, total working hours and capital stock of economic sectors (2011-2018) Year Efficiency change (Effch) Technical change (Techch) Pure efficiency change (Pech) Scale efficiency change (Sech) TFP change (Tfpch) 2011 1.032 0.922 0.998 1.034 0.952 2012 1.041 0.928 1.027 1.014 0.966 2013 1.010 0.974 1.027 1.014 0.966 2014 0.985 1.001 1.009 0.976 0.986 2015 0.955 1.004 1.001 0.955 0.959 15 2016 0.987 1.015 1.009 0.978 1.002 2017 0.980 1.048 0.993 0.986 1.027 2018 0.955 1.033 0.993 0.962 0.986 Average 2010 – 2018 0.993 0.990 1.005 0.988 0.982 If we replace the data on employees for total working hours, then the Malmquist productivity index does not differ much. The trend also shows that from 2014 to 2018, the TFP index has results greater than 1, due to the improvement of technical change, and pure efficiency change. Scale efficiency change trends to decrease from 2014 to 2018, while technological progress trends to increase. Pure efficiency changes still trend to get better during this period. 3.2.3 Using the parametric approach - stochastic frontier production function When applying the stochastic frontier production function, it is necessary to choose the appropriate type of function. Applying Frontier 4.1 software, the tests were conducted as follows: • Test 1: Choose the type of function; • Test 2: Verifying whether there is a technical inefficiency or not; • Test 3: Verifying function having technical inefficiency with half-normally distributed; • Test 4: Verifying function having invariable technical inefficiency overtime; • Test 5: Verifying function having technological progress; • Test 6: Technological progress effect to increase capital and labor. Test method: Using likelihood-ratio obtained from the estimation of the above models to test the function form. The statistical test is LR (λ) = -2 [L (H0) - L (H1)], where L (H0) is maximised log-likelihood value of the model which is considered as the null hypothesis; and L (H1) is the maximized log-likelihood value of the stochastic frontier function which considered as alternative hypothesis. This statistical test has an approximate distribution of χ2 with degrees of freedom equal to the difference between the corresponding parameters in the null hypothesis and the alternative hypothesis. A. Application of the "stochastic frontier function" with data of added value, the number of employees and capital stock in the period 2010 - 2018 The stochastic frontier function is used with the output is added value, the input is the number of employee and capital stock, with the help of Frontier 4.0. After testing, the chosen function is a half-normally distributed translog function, with technical 16 inefficiency changes over time, technology progress does not affect labor intensity and capital intensity: ln yit + uit = β0 + βllnlit + βklnkit + 0,5βll (lnlit)2 + 0,5βkk (lnkit)2 + βlk lnkitlnlit + βtt + 0,5βttt2 +vit. Table 3: production function’s parameters estimation results with the data of added value, number of employees and capital stock of the economic sectors in the period 2010 – 2018 β Standard error t-ratio β0 15.979 1.604 9.957 βl -0.463 0.186 -2.478 βk -0.637 0.256 -2.479 βll 0.062 0.021 2.834 βkk 0.037 0.012 3.084 βlk -0.012 0.022 -0.587 βt -0.004 0.010 -0.466 βtt 0.002 0.0006 3.449 From the production function, the ratio of increase in labor productivity sourced from technological progress is: t t tlkf ttt ββ 2),,(ln +=∂ ∂ Table 4: Estimates of technological progress based on data of added value, capital stock and number of employees of economic sector (2011-2018) Year The ratio of increase in labor productivity by technology progress 2011 0.0047 2012 0.0094 2013 0.0142 2014 0.0189 2015 0.0237 2016 0.0285 2017 0.0334 2018 0.0382 Average 2011- 2018 0.0179 17 The change in labor productivity due to technological progress (average 2011 - 2018) that estimated from the data of economic sector is 1.79%. The trend shows that, the technological progress increases gradually from 2011 to 2018, this result is quite consistent with the results of the Malmquist index over the years. The contribution ratio of technological progress to labor productivity increase can be calculated by dividing the degree (or ratio) of technological progress change by the degree (or ratio) of change in labor productivity. With an average growth rate of labor productivity of 4.87% in this period, technological progress 1.79% is estimated to contribute 36.7% to increase labor productivity. The average technical efficiency of this period is 35.1%. The technical efficiency is still low but trends to increase during this period. A. Application of "stochastic frontier function" to data of added value, total working hours and capital stock in the period 2010 - 2018 Using data on added value, total working hours and capital stock, after doing test, the appropriate production function is chosen. It is the form of half-normally distributed translog function, with technical inefficiency changes over time. The change in labor productivity due to technological progress (average 2011 - 2018) is estimated 2.90%. With an average growth rate of labor productivity 5.29% in this period, technological progress increases by 2.90%, technological progress is estimated to contribute 54.8% to increase labor productivity. The average technical efficiency of this period is 41.4%. Economic sector (level I)’s data are suitable for a general assessment of the economy, because economic sectors (level I) aggregates different business activities including enterprises and types of cooperatives, households and individual business. 3.2 Impact of technological progress on productivity in manufacturing enterprises Manufacturing has been a fast-growing economic sector in the past 10 years and is a good performer of the contribution of technology progress to the output growth. Therefore, the thesis will study manufacturing enterprises to illustrate the impact of technological progress on increasing labor productivity. The data used include added value (constant prices) from 2010 to 2018, number of employees and fixed assets as of December 31 from 2010 to 2018 (converted to constant prices) of enterprises in the manufacturing sector (24 secondary economic sector) obtained from the GSO data source. 18 From the data, added value in the period from 2010 to 2018 average increase of 14.0% is a rapid growth. Labor growth in this period was also quite high 6.0% and the labor productivity increase 6.8% per year. The results of calculating TFP Malmquist index are as follows: Table 5: Results of the Malmquist index calculated from data of manufacturing (2010-2018) Year Efficiency change (Effch) Technical change (Techch) Pure efficiency change (Pech) Scale efficiency change (Sech) TFP change (Tfpch) 2011 1.178 0.940 1.024 1.151 1.107 2012 1.066 0.969 1.046 1.019 1.032 2013 0.808 1.223 0.975 0.829 0.988 2014 0.929 1.099 1.027 0.905 1.021 2015 1.022 1.001 0.915 1.117 1.023 2016 1.012 1.110 0.864 1.172 1.123 2017 1.063 0.926 1.012 1.051 0.984 2018 1.016 0.862 1.082 0.938 0.984 Average 2010 -