Tóm tắt Luận án Analysis of inclusion in Vietnam economic growth during the period 2004 - 2016

ronment and space. Among them, economics and society are the two most crucial fields. Specifically, economic refers to growth and employment while society is a combination of factors such as health, education, social safety nets and gender. The World Economic Forum (2015) suggested inclusive growth to be analyzed with seven contents: Education and skills development, Employment and labor compensation, Asset building and entrepreneurship, Financial intermediation of real economy investment, Corruption and rents, Basic services and infrastructure, Fiscal transfer. These were also the seven principal pillars of inclusive growth analysis given in the study of Sammans et al. (2015). There were several other studies focused on analyzing the income and labor pillars of inclusive growth. In this group, there were studies by Anand et al. (2013) and Hann and Thorat (2013) looking at the aspect of income in inclusive growth. Meanwhile, Hausman, 6 Rodrik and Velasco (2005), Ramos et al. (2013) all conducted study about employment, labor and income. However, since inclusive growth is a multidimensional concept, the pillars (main content) in previous research about this topic are also very diverse. From literature review of empirical studies, it is possible to summarize some of the key areas of growth including: (i) economic growth, (ii) poverty and inequality, (iii) employment, (iv) education, healthcare and demographic issues, (v) environment, (vi) gender / gender inequality, (vi) space, (vii) social safety nets and (viii) infrastructure. 1.2. Inclusive growth measurement methods 1.2.1. Concentration curve and concentration index method The concentration curve was developed by Kakwani (1977) based on the cumulative percentage of the measurement variable (vertical axis) compared to the cumulative percentage of the population (horizontal axis), households are arranged in ascending order in average income, beginning with the lowest and ending with the highest income per capita households. The concentration curve showed the cumulative percentage of measurement variables taken by the percentage of households with lowest income per capita. From the concentration curve, Kakwani (1977) calculated the concentration index to measure the level of socioeconomic inequality. The concentration index can be stated as follows: In which
 is the measure of inequality, µ is its mean, and  =   is the ranking order of the  household in its distributionbased on the average income, with i = 1 for households with the lowest average income, and i = N for households with the highest average income. 1.2.2. Social opportunity function method The social opportunity function method, first developed by Ali and Son (2007), applied to non-monetary indicators. Afterwards, Anand et al. (2013) developed this idea for monetary indicators into the social mobility curve method. The measurement method of inclusive growth reflects the increase in the social opportunity function, depending on two factors: (i) the average opportunity created; (ii) and how to allocate opportunities among households in the economy. Households with average income increase, and ȳp is the average opportunity taken by p percent of the lowest-income (Kakwani, 1977, 1980) (1) 7 households (in which p fluctuates between 0 and 100 and ȳ is the average opportunity available for the population), then ȳ will be equal to ȳ when p equals 100 (that means including the whole population). Because ȳ varies with p, we can draw a curve for each value of p. This is the curve that focuses overall opportunity when households are arranged in ascending order in average income. The index represents the area below the opportunity curve, and is expressed mathematically as follows: ȳ*=  ȳ   ȳ* is opportunity index (2) To consider the issue of equality in opportunity distribution, we can develop and calculate the Equality of Opportunity Index (EOI) index as follows:  = ȳ∗ ȳ (3) With the same principle, Anand et al. developed the social opportunity function method to calculate the social mobility index for income criteria (also called the social mobility function method). 1.2.3. The composite index method The composite index method is a method of calculating inclusive development index based on the indices of individual indicators and assigning weights to those indicators. The aggregate index is built on a scale of 0 to 10, according to the degree of achievement of each country in each of these measurement dimensions with criteria. The higher the score is, the greater the inclusion that country had in that component indicator. 1.3. Theoretical foundation for inclusive growth determinants 1.3.1. Theoretical foundation of growth determinants Some of the theories of growth mentioned in this section are classical theories of Adam Smith (1776), David Ricardo and Karl Marx; theories of Keynesianism, including Harrod-Domar theory which focused on the role of capital in economic growth; Neoclassical theories, represented by the Slow Swan theory (1956) stated that, besides capital, labor and technology also affected growth. Finally, there are endogenous growth model in Arrow and Romer theory with emphasis on human capital’s effect on a nation's economic growth. In addition, the role of population growth was also mentioned in a variety of theories, including Thomas Malthus's population theory explaining the impact of the population on income per capita. 8 1.3.2. Theoretical foundation for the determinants of income inequality or income distribution Factors affecting income inequality are usually divided into 5 groups, namely: economic development, demographics, politics, culture and environment, and macroeconomic factors. Each group consists of different representative elements. So far, a number of studies have chosen to analyze the one or some of these factors 9 with income inequality. However, the findings of those studies revealed different impacts of differenr factors on income inequality. The relationship is either positive or negative, while many factors did not show clear impact on income inequality. Among those aforementioned factors, consistent results from past research were shown in: positive effects of technological progress, inequality in education and foreign investment on income inequality. Meanwhile, financial development level affected income inequality negatively. Additionally, some factors were researched by many scholars, although the impact might be heterogeneous, including GDP per capita, economic restructuring, education levels and spending, inflation, and unemployment. CHAPTER 2 RESEARCH OVERVIEW OF INCLUSIVE GROWTH 2.1. Foreign research 2.1.1. Research conducted for multi-country sample The full version of the thesis includes an overview of studies in which the dependent variable reflects inclusive growth in income, income per capita and income inequality measurement. However, this summary only focuses on research with inclusive growth in income as dependent variable conducted for multi-country sample. Specifically, studies in this group included research by Anand et al. (2013), Jalles and Mello (2019), Doumbia (2018), Javed et al. (2018), Aoyagi and Ganelli (2015), Sen (2014), Ravi et al. (2013). Although all of them focused mostly on the income aspect of inclusive growth, these studies still varied widely in the selection of measurement variables for dependent and independent variables. The main independent variables used were: GDP per capita, inflation, human development indicators such as education and healthcare, institutions and governance, investment, trade openness and government spending. 2.1.2. Research conducted within one country There was some research conducted in one nation with inclusive growth of income as the dependent variable, such as: Studies of Arabiyat et al. (2019), Munir et al. (2018), Khan et al. (2016), Pukuh and Widyasthika (2017), Oluseye and Gabriel (2017). The common point of these studies is that the dependent variable - inclusive growth of income was measured by the social mobility curve as proposed in Anand et al. (2013). The independent variables mostly used in these studies were: GDP per capita, inflation, population growth, government spending, trade openness, and money supply growth. 10 2.1.3. Other research on inclusive growth Besides inclusive growth studies using quantitative analysis, there were some other studies also investigated this topic but only stopped at analyzing the situation of growth on one or some aspects. The number of these studies is much higher than the quantitative analysis research. Some following studies can be stated as example: Yuwa (2014), Schmid (2014), Habito (2009), Ganesh and Ravi (2009), Osmani (2008), Fernando (2008), Norman et al. (2007), Afzal ( 2007), Afzal and Jazhong (2007), Afzal and Xianbin (2004) and Bolt (2004). The main content analyzed in these researches was: economic growth, poverty, employment, institutions and infrastructure. 2.2. Domestic research Inclusive growth is a relatively new concept in Vietnam, so there has not much research done on this topic. Some studies in this group include: Le Kim Sa (2014), Pham Minh Thai and Vu Thi Minh Ngoc (2014), Nguyen Duc Thanh and Pham Van Dai (2014), Do Son Tung and Ma Ngoc Nga (2014), Le Kim Sa (2008). The content of inclusive growth studies in Vietnam mainly analyzed the labor market of one or several enterprises in a certain industry, thereby made policy recommendations to improve the inclusion for that market. Based on the overview of overseas empirical studies, together with domestic research analyzing inclusive growth, the thesis found a big gap of inclusive growth in income, especially from household perspectives. Furthermore, no studies in Vietnam have conducted quantitative analysis to investigate the impact of factors on inclusive growth in income in Vietnam. 2.3. Research framework The thesis was conducted by following several steps: (i) identifying research objectives, (ii) reviewing the research materials, (iii) developing analytical framework, (iv) collecting, analyzing and processing data, (v) doing research findings and (vi) making policy recommendations. Regarding the content of inclusive growth, this study analyzed six main groups: (i) economic growth, (ii) poverty and inequality, (iii) employment, (iv) education and health care, (v) space and (vi) infrastructure. Within each group, one or more indicators would selected for analysis. 11 Regarding quantitative analysis, the thesis developed analytical framework as follows: Figure 1: Quantitative analysis framework of the research CHAPTER 3 THE SITUATION OF INCLUSIVE GROWTH IN VIETNAM DURING THE 2004-2016 PERIOD 3.1. Situation of inclusive growth of income 3.1.1. Situation of economic growth and income distribution In general, Vietnam's economic growth in the past three decades since the Renovation implementation has achieved great results. As a result, the proportion of poor households has decreased significantly, GDP per capita increased (In average, Vietnam's income per capita for the whole period 2004-2016 doubled, based on purchasing power parity in 2011). Vietnam has been one of the countries with high economic growth rate in the region. In addition, the economic structure was also shifting in line with the trend of developing economies towards increasing the proportion of industry and services, reducing the proportion of agriculture. Income inclusive growth Human resources Labor quality Education level Years of schooling Healthcare Macro factors Inflation Crisis GRDP FDI Institutions and policies Ratio of trained labor to total labor Provincial Competitiveness Index Budget spending 12 However, while the poverty rate dropped, fluctuations recorded in the distribution of the poor across the country. The majority of poor people came from rural areas. Whereas, if based on regional criteria, most poor people originated from the Midlands and Northern Mountains (in 2010 and 2016). Some indicators reflecting the income inequality used were Gini, income between 20% of the richest population and 20% of the poorest population in the country, based on urban-rural areas and ethnicity showed different fluctuations. Considering both criteria, inequality in urban areas reduced and recorded more volatile than in rural areas, ethnic minorities group (including Hoa and non-Kinh people) also followed the same pattern compared to Kinh people. The average growth rate of income in both urban and rural areas decreased more in the last years of the study period, this rate was highest in the period 2010-2012 in all six geographical regions. The proportion of income in industry and services was still small, wages and salaries constituted the largest part in urban areas’ income, while in rural areas income mainly came from agricultural activities. 3.1.2. Inclusive growth of income in Vietnam Inclusive growth of income improved over all years throughout the study period, regardless of the scope of analysis. However, this improvement was mainly explained by the improvement in average income, which did not come from more equal income distribution. 2010 was the year witnessing the most unequal household income distribution in Vietnam, while the most equal distribution time was 2006. In terms of provinces and cities, there were also some notable changes coming from all three indicators: income growth, equality growth and inclusive growth. 3.2. The inclusive growth situation of some non-income indicators 3.2.1. Education, healthcare, labor and employment Education: The thesis analyzed the rate of joining school at the right age in all educational levels, by urban-rural area and by sex; and the highest number of schooling years of the member with the longest study time in the household to calculate the opportunity index. The results indicated that the rate of joining school at the right age was lower with the higher levels of education, this rate in urban areas was higher than in rural areas, and also the rate recorded in females was higher than males. Regarding the criteria of the longest time of schooling (by year), the equality was higher in urban areas than in rural areas and the same pattern shown among Kinh people compared to other ethnic groups. 13 However, in all criteria, the opportunity index reflecting the inclusion of access to educational opportunities (measured by years of schooling) improved. Healthcare: Opportunity of accessing to healthcare was analyzed through access to health insurance and free health checks for citizens. Considering this opportunity, in all criteria, inclusion increased over the years, except for the period 2006-2008. Labor and employment: The labor force participation rate and the percentage of trained workers were higher in both males and females. 3.2.2. Some other non-income indicators In this section, the study investigated the ability to access social security through the social insurance access rate in 5 income groups and access to basic living conditions such as electricity, water and sanitation. The rate of social insurance coverage decreased gradually according to 5 income groups, with large disparities among groups. In each group, this rate reduced during 2006-2010, but improved over the rest of the time. In addition, regarding inclusion of the three basic types of access, the opportunity to access electrical grid was highest, followed by sanitation and finally tap water. 3.3. Some constraints to inclusive growth in Vietnam Despite many achievements in growth and poverty reduction, inclusive growth in Vietnam has been facing many challenges and limitations. One of them can be stated such as (i) uneven growth, (ii) low employment and labor productivity, (iii) large gaps in asset holdings and access to opportunities in life. CHAPTER 4 ESTIMATION OF FACTORS AFFECTING INCOME INCLUSIVE GROWTH IN VIETNAM 4.1. Model specification 4.1.1. Model building The thesis built an econometric model to analyze the impact of factors on inclusive growth of income for provinces and cities of Vietnam as follows:  =  +  +  (4) 4.1.2. Estimation method The thesis applied the Panel data regression estimation method: conducted with fixed / random effects estimator and spatial estimator. 4.1.2.1. The fixed and random effects estimator 14 Using Hausman test, the research chooses between two fixed and random effects models. 4.1.2.2. Spatial econometric panel data model The spatial estimation method was used when there was suspicion of the spatial relationship among entities. According to Le Gallo et al. (2003), in measuring economic relationships, ignoring spatial correlation may lead to biased and unreliable estimates. This was also the spatial autocorrelation shown in several studies such as Paraguas and Kami (2005) or Higazi et al. (2013). The thesis created the spatial matrix according to distance and conducted necessary tests to select the appropriate spatial model. After finishing the tests, the selected spatial model was the Spatial Autoregressive Model with Auto Regressive disturbances. In addition, the construction of different matrix types also tested the robustness check of the model. The results indicated that there was not much difference when selecting different types of matrices to make estimates. In other words, the robustness of the model was confirmed. 4.2. Data sources, data descriptions and variables used in the estimation model 4.2.1. Data sources 4.2.1.1. Characteristic of provincial data According to the government decision, from the beginning of August 2018, all Ha Tay, Me Linh district of Vinh Phuc province, and 4 communes of Luong Son district, Hoa Binh province were merged into Hanoi. Therefore, for data consistency, all data of Me Linh district in Vinh Phuc province and data of 4 communes in Hoa Binh province before 2010 was be calculated as Hanoi data. 4.2.1.2. Source of data identification The dependent variable is the inclusive growth of income, which was calculated as the social mobility index according to the social mobility function method of Anand et al. (2013), using data of average household income in VHLSS. Independent variable: The data for the independent variable used the secondary data source from the General Statistics Office, the Statistical Office of provinces, the Ministry of Finance, the provincial Department of Finance, and Vietnam Chamber of Commerce and Industry (VCCI). Only the inclusion index of education as independent variable was calculated from the data in VHLSS by using of the social opportunity function method of Ali and Son (2007). 15 All independent variables in the model were: GRDP per capita in the first period, inflation, dummy variable (shown for crisis factor), the ratio of investment to GRDP, Foreign Direct Investment (FDI), the ratio of trained labor to total labor, the inclusion index of education, the human resources for health per capita, the provincial competitiveness index (PCI), and local budget expenditures. 4.2.2.2. Expected sign of the variables Based on the overview of past research in chapter 2, the signs of variables were expected as follows: Variables with positive expectation: GRDP per capita in the first period, the ratio of investment to GRDP, the ratio of trained labor to total labor, the inclusion index of education, the provincial competitiveness index (PCI), the human resources for health per capita. Variables with negative expectation: Inflation Variables with unclear expectation: The dummy variable for crisis, local budget expenditures, Foreign Direct Investment (FDI). 4.3. Model results 4.3.1.1. Descriptive statistics and correlation matrix among variables The table showing descriptive statistics and correlation among variables was presented in the full text of the thesis. 4.3.2. Estimating the fixed and random effect model 4.3.2.1. Test results of fixed and random effect model The Hausman test results showed that the value Prob> chi2 = 0.0000, therefore the fixed effects estimation model was selected. The specific estimation results were as follows Table 1: Test results of FE model and RE model without space factor VARIABLES (1) FE (2) RE lgdppop2004 0.29*** 0.32*** (0.07) (0.04) i_gdp 0.48*** 0.11 (0.14) (0.12) lcpi1 -0.35*** -0.35*** (0.04) (0.03) 16 lfdi -0.01** -0.00 (0.01) (0.01) lpci -0.00 0.17** lchins ledu labor_tyle lyte2 Constant (0.08) 0.06*** (0.01) 0.32*** (0.10) 1.36*** (0.24) -0.05** (0.02) -1.32*** (0.40) (0.08) 0.06*** (0.01) 0.25*** (0.09) 0.99*** (0.22) -0.10*** (0.02) -1.71*** (0.37) Observations 365 365 R-squared 0.85 Number of mun 63 63 Hausman test Prob>chi2 = 0.0000 Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Positive effects were shown in variables: GDP per capita in the first period, the ratio of investment to GDP, the ratio of trained labor to total labor, the inclusion index of education, the provincial competitiveness index (PCI), local budget expenditures. Negative effects were presented in variables: Inflation, the human resources for health per capita, FDI. Both positive and negative effects recorded in: Dummy variable shown crisis 4.3.2.2. Tests in the fixed and random effects estimation model The thesis conducted heteroskedasticity test, multi-collinearity test, and test for time fixed effects. The results proved that there was heteroskedasticity in the fixed effects model, no multi-collinearity among variables in the model, and the model was affected by time factors. 4.3.3. Estimating spatial models The study built a sequence of steps to select the appropriate spatial model Step 1: Built a spatial matrix on Geoda software 17 Step 2: Estimated the Spatial Durbin Model (SDM) Step 3: Verified the model selection. In this step, there were some following minor steps: (i) performed Hausman test to choose between fixed or random effects models, (ii) Verified the choice of spatial matrix type Step 4: Estimated the selected spatial model: This step was done by two methods (i) estimating without effect from LeSega and Pace (2009), and (ii) estimating with effect from LeSega and Pace (2009). 4.3.3.1. Tests to estimate spatial models In this section, the study conducted some tests including: Hausman test for Spatial Durbin Model, test for spatial dependence of dependent variables, test for selecting suitable spatial model. Specifically, there were tests between Spatial Autoregressive Regression (SAR) and Spatial Durbin model (SDM) (consequently chose SDM model), between the Spatial Error model (SEM) and SDM (consequently chose SDM model), between SAC, SDM and Generalised Spatial Panel Random Effects Model (GSPRE) (with the result of selecting SAC model). 4.3.3.2. Results of estimating SAC spatial model The results of estimating the SAC spatial model were presented with two methods, with or without effects according to LeSega and Pace (2009). These two authors pointed out that there could only be direct and indirect effects in the spatial estimation model. Accordingly, the direct effect was used to measure the change in effect of the independent variable on dependent one in the same city, while the indirect effect was the cross-space effect used to measure the change in effect of one city’s independent variable on another city’s dependent variable. The total effect was the combined effect of the direct and indirect effects. The estimated results of spatial models without considering the effects according to LeSega and Pace (2009): The results were quite similar to the estimation model presented above (without considering spatial factors). Table 2: Test results of estimating SAC spatial model VARIABLES SAC lgdppop2004 0.09** (0.04) i_gdp 0.19*** (0.07) lcpi1 -0.10*** (0.03) lfdi -0.00 18 (0.00) lpci -0.00 (0.00) lchins 0.02*** (0.01) ledu 0.23*** (0.06) labor_tyle 0.63*** (0.13) lyte2 -0.05*** (0.01) rho 0.70*** (0.04) lambda -0.60*** (0.09) sigma2_e 0.01*** (0.00) Observations 441 R-squared 0.64 Number of mun 63 Standard errors in parentheses *** p<0.01, ** p&l