Investigating the role of health, education, energy and pollution for explaining total factor productivity in emerging economies

This study aims to analyze the role of health, education, energy and pollution in affecting productivity in selected emerging economies. Industrial share to GDP, trade openness and Information Communication and Technology (ICT) are used as control variables. Various dimensions of health and education are explored that include malnutrition, access to clean water, HIV, life expectancy and years of schooling at several levels. A set of Fixed Effect models provide evidence that all of the variables of health and education are critical for productivity. Further, the negative effect of trade openness calls for attention of the policy makers to work out the possibility of technology transfer through integration of economies so that productivity may be enhanced. Another dimension is to revisit the energy mix because the positive effect of energy use is offset by an increase in pollution. Public policy makers should understand the importance of public investment in necessary provisions for improving productivity, economic growth and ultimately the welfare of society, then it can make a difference.


Introduction
W ith the widespread recognition that total factor productivity (TFP) accounts for substantial cross-country growth differences, there is ample and rigorous research in growth literature to identify the factors and policies which are conducive to TFP. In this regard, the inclusion of human capital by Barro (1991) and Mankiw et al. (1992) supported by endogenous growth models of Romer (1986) and Lucas (1988) presented the hallmark of this research. Human capital accumulation is multifaceted and affects TFP through a variety of channels including technological innovations (Romer, 1990a(Romer, , 1990bAghion and Howitt,1998), labor market efficiency (Cole and Neumayer, 2006) and complementing the productivity of other inputs to production.
The relationship between human capital and TFP is ascertained generally through the role of health and education. The early empirical literature focused on a single aggregate measure of health like life expectancy and education like average years of schooling (see, for instance, Knowles and Owen, 1995;Bloom et al., 1999;Gallup and Sachs, 2000;Mayer, 2001;Bhargava et al., 2001;Webber, 2002). Though this literature provided useful insights about the relationship between human capital and TFP but lately, it was found limited from policy making perspective at a disaggregated level. As WHO (2002) identified "even though life expectancy has been most commonly used by economists but it does not capture all the aspects of the individual's current health that may affect productive capacity." Similarly, an aggregate indicator of a country's educational status neither fully represents different dimensions of a country's educational system nor highlights the potential areas which can expedite the productivity process.
Over the last few decades, the relationship between pollution and economic growth is well documented. Among the most recent studies, empirical evidence for different countries and regions can be seen in the research conducted by Esteve and Tamarit (2012), Aslan and Gozbasi (2016), Dogan and Ozturk (2017), Mikayilov et al. (2018), Adedoyin et al. (2020), and Zhang et al. (2019). However, the unfortunate record increase in greenhouse gas emissions during the last century implies that different perspectives about the role of environmental degradation for economic development to sensitize the policy makers about aligning the governance structure to sustain the economy is inevitable. For instance, the investigation on the relationship between economic growth and pollution is to evolve continuously especially when caution is required in the selection of variables in the model. In this perspective, it would be critical to look into the relationship between TFP and pollution instead of economic growth. The role of TFP that indicates technological development is critical for sustainable economic growth (Solow, 1957;Romer, 1986;Lucas, 1988).
According to Rath et al. (2019), the relationship between TFP, energy and pollution is critical to analyze to correct the policy direction. More recently, the impact of energy on growth is available at a relatively wide scale but the relationship of energy with TFP is not significantly available. Ladu and Meleddu (2014), Ackah and Adu (2014) and Tugcu and Tiwari (2016) have recently investigated this relationship for different groups of countries but fails to provide a holistic economic model. Dogan et al. (2020) has provided evidence on the TFP-energy-pollution nexus for selected African countries however it focuses on these three variables only and ignored other relevant variables in the model. As a matter of fact, role of education at different levels, undernourishment, lack of access to water and HIV prevalence for affecting TFP are largely ignored in the literature. Lenkei et al. (2018), among a few, has investigated the impact of different levels of education, as indicators of human capital, on economic growth.
Apart from compositional indicators of measures of human capital and ignoring the TFP-energy-pollution nexus, there are certain variables that besides their direct effect on TFP also intermediate (enhance/moderate) the role of human capital for TFP through the quality of human capital. In this regard, environmental quality has been recognized to have a close association with health quality while information technology represents the technological penetration of the education system of the economy. From a theoretical point of view, the relationship between environmental quality and TFP is ambiguous. On one hand, deterioration in environmental quality can decrease TFP through its negative effect on health (Schlenker and Walker, 2016;Graff Zivin and Neidell, 2012). On the other hand, this deterioration can increase TFP through its relation with inputs of production (Diao and Roe, 1997).
Information and communication technology has brought radical changes in almost every field of economic and social life with its most pronounced effects on production and education sectors. Shapiro et al. (1999) remarked that information and communication technologies provide a platform for innovations to generate synergies and enhance the technological level of production process, hence TFP. Technology-induced innovations and other network externalities improve TFP in technology facilitated business integrated supply chain (Kim and Narasimhan, 2002;Kim et al., 2011). Information technology is expected to affect TFP indirectly by improving the quality of both teaching and learning.
Given this background and gap in the literature, this study has three distinct features. We add to the literature by comparing the role of aggregated and compositional measures of health and education for TFP. For this purpose, we have taken undernourishment, waterborne diseases, HIV, and air pollution for the potential disaggregate of health and their impact on TFP. Life expectancy is taken as an aggregate measure of the health status of a country. Similarly, primary, secondary and tertiary education as disaggregated measures of education while average years of schooling as a conventional aggregate measure of education. The comparison is expected to serve two important purposes both for academic and policy making purposes. First, the comparison of aggregate and compositional measures is expected to reveal that how many aggregate measures over/ understate the impact of human capital on TFP. Secondly, this exercise is expected to help identify the sectors that require policy precedence to improve TFP.
The second distinct feature of our study is that we have analyzed the impact of those factors for TFP which play an integral role in determining the quality of human capital. Environment quality is such a variable and highly related to the quality of health outcomes in the economy. By taking pollution as an indicator of environmental quality we have ascertained its impact for both. Similarly, by taking information and communication technology (ICT) goods as an indicator of information technology the study captures the role of the quality of education for TFP. This exercise is conducted for both aggregated and compositional measures of human capital. The findings of this exercise are expected to reveal that which compositional measure is well through its quality counterpart. Thirdly, the study is distinctive in a way to investigate the nexus between TFP, energy consumption and air pollution that is largely missing, especially in the presence of complementary variables that require for correct specification of the model. Thirty emerging economies are selected that are struggling for environmentally sustainable economic development.
This remaining study is organized as; research method and data sources are described in section "Research method and data sources", section "Results and discussion" is about results and discussion, Section "Conclusion" describes the concluding remarks of the study.

Research method and data sources
Looking at the literature reveals that human capital can be included either in production function or in the total factor productivity equation. For a panel of 83 countries, Miller and Upadhyay (2000) estimated the Cobb-Douglas production function with and without human capital. They found a positive effect of human capital on TFP, which turns from negative to positive for developing countries with the increase in trade openness. However, it remained controversial to include human capital in the production function, especially for panel data studies. Mankiw et al. (1992) advocated that human capital is a better fit for the cross-sectional studies while Islam (1995) did not find human capital as significant for the output. Benhabib and Speigel (1994) also use human capital for estimation of the production function and they find it insignificant. Therefore, this study investigates the impact of different health and education related indicators on total factor productivity. By doing so, this study reaches on a more elaborative conclusion and communicates the significance of all indicators to the policy makers for better development of policies and mitigation strategies. This study obtained the data of TFP from Penn World Table (PWT), 10.0. The list of variables and data sources are given in theTable 1.
Further, most of the literature relied on secondary schooling or average years of schooling and did not give due importance to different levels of schooling which has varied impacts on productivity. Higher education is important for innovation of the new advanced technology and things, which cause to increase the productivity of labor and make easy life. In the same way, primary and secondary education levels are also important to equip with the level of understanding to use the latest machinery. Moreover, the gap between developed and developing countries is due to education. Self and Grabowski (2003) identify that primary, secondary and tertiary education have a positive impact on the economic growth in Japan after the post-war period. Barro (1991) also estimates that the initial period of primary secondary and tertiary education has a positive impact on GDP growth. Thus for stable economic development, it is necessary to focus on all levels of education. Among others, it is advocated by Knowles and Owen (1995), Asteriou and Agiomirgianakis (2001), Bassanini and Scarpetta (2002), and Bloom et al. (2004). Data for primary, secondary and tertiary education is obtained from World Bank (2020).
In addition to indicators of health and education, we include the other determinants that affect the TFP. These variables are trade openness, ICT, energy consumption per capita, and Industrial share to GDP. The various growth and productivity models suggest these variables. Miller and Upadhayay (2000) include trade openness to check the impact on TFP. It shows that trade openness has a positive impact on productivity. To check the impact of structural transformation on total Factor productivity, we took Industrial share to GDP. The Data of these variables is taken from World Bank (2020).
The prevailing macroeconomic structure is important to include in the model of TFP, which is captured by trade openness, energy consumption, ICT and industrial share to GDP. Trade openness, a proxy of transfer of knowledge, enhance the transfer of technology and leads to an increase in TFP (Isaksson, 2007;Harrison, 1996). Macro-instability, uncertainty and regulatory quality is captured through inflation, following (Daude and Fernández-Arias, 2010). Structural change is captured through agriculture share to GDP and industrial share to GDP. According to Loko and Diouf (2009) and Nawaz and Alvi (2017), along with others, high value addition leads to more productivity whereas with the increase in industrialization, agriculture share is decreased and industrial share in GDP is increased over time.
Equation (1) is an empirical form of TFP, for a panel of countries.
where tfp represents the total factor productivity, X denotes the indicator of health and education, trad represents the trade openness, en represents energy consumption per capital indus represents the industrial share to GDP and ICT represents information and communication technology service. This study is important in this perspective, that is, it focuses on morbidity-relevant health indicators as well as different levels of education for investigating their impact on total factor productivity. In order to check for the multicollinearity problem, the correlation matrix was computed for all the variables under consideration and is given in Table 2. There is no evidence of high multicollinearity problems among all variables. The correlation among all variables is <0.9, so there is no problem with multicollinearity (Farrar and Glauber, 1967). Since the r value for none of the above-mentioned pair of variables is >0.95, so interpretation of the relationship among the variable is correct. The negative sign shows that an increase in one variable will decrease the other variable and the positive sign shows an increase in one variable the other also increase. The results of descriptive statistics, including the mean, standard deviation, minimum, maximum, skewness, and kurtosis for each variable is provided in Table 3. Results show that three variables are including life expectancy, access to water, and air pollution, are negatively skewed i.e. more of the observations lying to the right of the average value of the series while other variables are positively skewed. Kurtosis tells us about the peakedness of the data. Only 4 series are mesokurtic, ten series are leptokurtic, while the rest is platykurtic. The total number of observations is 900, (30 × 30) panel observations. The first variable is TFP average value is 0.65 with 0.30 standard deviation, the minimum is 0.16, maximum is 2.40, the median is 0.60, positively skewed and leptokurtic. Figure 1 shows the average total factor productivity of selected 30 emerging economies in descending order from the years (1990-2019), China is shaded with red color indicating that the average total factor productivity in China is the highest of 1.41, in all selected thirty developing countries, the second highest average total productivity is Zimbabwe with 1.18 average total factor productivity and so on which can be seen clearly in the below figure.
The lowest average total factor productivity country is India, which is shown at the bottom of the figure, with an average TFP of 0.31.

Results and discussion
In the present study, we first identify the panel data model that is better between Random and Fixed Effects model. We apply the Hausman test that confirms the Fixed Effects model is more appropriate, as p-value is < 0.05. The null hypothesis of Hausman test is "Random effect is appropriate" (Wooldridge, 2019) so rejecting the null hypothesis means that the fixed effect model is appropriate. The estimated results of Fixed Effects models are presented in Table 4. Three models of TFP capturing different levels of education are estimated for primary, secondary and tertiary education whereas the fourth model used average years of schooling. The importance of education as a measure of human capital for determining the TFP is well established in the literature. However, an important insight, this paper shares, is the evidence on how the skillset is relevant for TFP for the selected countries. All estimated models have theoretically corrected signs for all the explanatory variables. According to Kennedy (2005), omitted explanatory variables and wrong selection of estimation techniques can be two of the possible reasons for wrong signs. So implicitly, this study has put the efforts to correctly specify the model and select the estimation technique. It is revealed that the coefficients of primary, secondary and tertiary schooling are 0.038, 0.045, and 0.091, respectively in the first three models and the variable of energy use, industrial share and ICT service have a positive relation with total factor productivity. Furthermore, the positive signs of variables indicate that increase in energy use, industrial share and ICT service will increase total factor productivity by different magnitudes.
The estimated results of Fixed Effects model (IV) presented in Table 4 indicate that the coefficients of an average year of schooling are 0.087, the variable year of schooling is positive relation with total factor productivity while trade openness has negative, furthermore, energy use, industrial share, ICT service and year schooling have a positive relation with total factor productivity. The positive signs of variables indicate that increase in energy use, industrial share; ICT service and year schooling will increase on average total factor productivity by the different magnitude Tertiary education has more impact on TFP compared to secondary education. It may be due to specialization in their field. As people enter tertiary education they become specialized in their field, due to this specialization they become more productive (Kocourek and Nedomlelová, 2018). In this perspective, Boldrin (1) TFP 1.000 (2) Trade Openness  2004) has intensively reviewed the human capital policies of four Asian tigers that support the finding of this study. If we compare primary and secondary education, then we find that the function is diminishing return to scale. Skill mismatch may be one of the possible reasons for more contribution of low education level along with the low demand of high skills. Further, the production methods take less educated labor force as more obvious for the desired level of productivity then obviously, returns to education will be less for more education that requires the structure of the economy to change. In the context of the fourth industrial revolution, it is possible to witness a steep rise in the level of significance of the level of education. However, according to Hanushek and Woessmann (2015), both the length of education and its quality are important for productivity.
Higher education facilitates technological innovation whereas the basic level of education shows the effectiveness of labor in the production process (Daude and Fernández-Arias 2010;Benhabib and Spiegel 1994). Manuelli and Seshadri (2014) argue that an increase in the years of schooling adds less to the quality of human capital in developing countries. It is one of the prime reasons for less contribution in productivity of secondary education than primary education. Keeping this perspective in mind, the current study adds to the literature by signifying the returns to scale through human capital by segregating the levels of formal education. Important learning for the developing countries is to understand the dynamics of industrial requirements so that the education system may be transformed according to the requirements. It will also be held to produce productive youth into the economy and not the unemployed.  The majority of the empirical literature captures health as either expenditures on health, life expectancy or the diseases like malaria, cholera, etc. and ignores the roots of all the diseases specifically malnutrition and lack of access to clean drinking water. These studies include but are not limited to Tompa (2002) and Bloom et al. (2004). From a public policy viewpoint, it is important to control the problem of non-availability of clean drinking water, and malnutrition, which simply required optimal development expenditures by the government to ensure the availability to nutritious food and clean drinking water in every part of the country. By giving the importance of non-availability of clean drinking water, malnutrition, HIV, air pollution and as well as life expectancy, the present study estimate the Fixed Effects models and the result are given in Table 4 (V-IX). It is indicated that the coefficient of life expectancy is 0.11, the variable life expectancy has a positive relation with total factor productivity. However, the coefficients of undernourishment, lack of access to water, air pollution and HIV are −0.140, −0.039, −0.073, and −0.146, respectively.

Conclusion
This study reveals that priority must be given to human capital which is not only important for the health and education status of the society but also critical for enhancing productivity, economic growth and meeting the international development agenda. So, if the proper attention is paid to the human segment and the appropriate level of immunity then things might have been changed more easily. As a matter of the international development agenda, i.e. Sustainable Development Goals, it is also important to ensure universal access by 2030, which cannot be made possible without development expenditures for the provision of safe drinking water and nutritious food. In this perspective, food poverty or malnutrition is also required due consideration which is a very serious problem in a majority of the countries included in the sample of this study. It can be made possible to control this problem by employing a targeted approach through the provision of social safety net and with the awareness campaign through all means of communication. The network of basic health units or any other network available at the grass-root level can be better used to sensitize all age groups to have better nutrition. Another missing aspect is the non-availability of education on nutrition at any level of education. Further, the spread of HIV/AIDS is not controlled yet which requires due attention of the policy makers for educating people. This all can be made possible within available financial resources however there is no alternative to prioritization. Understanding the importance of these indicators for improving TFP and economic growth can make it possible for policy makers. The bottom-up approach according to the true spirit of SDGs, if adopted, and the local governments are empowered then there will be no reason to deal with the challenge of access to clean drinking water and malnutrition.

Data availability
All datasets analyzed are included in the paper.