Total factor productivity (TFP) growth has been considered the engine of economic growth (Afridi & Farooq, 2019). It is defined as the growth in output that is not accounted for by input. It can be measured through two approaches: a frontier approach and a non-frontier approach (Mahadevan, 2003). TFP growth is crucial for sustained output growth, since input-induced growth is subject to diminishing returns and is insufficient for producing more output in the long run. Solow’s (1956) growth model shows that the probable reason for cross-country differences in per capita income could be the difference in TFP growth.

The Asia-Pacific countries, including South Korea and Singapore (considered advanced countries), have been experiencing low TFP growth over the past decades due to the global financial crisis of 2008 (Felix et al., 2020). The TFP growth of Asia-Pacific countries was -1% in 2016, while it was 0.5% in 2010 (International Monetary Fund, 2017). These countries have growing energy demand, but lack productivity growth (BP Statistical Review of World Energy, 2021). Thus, it is essential to examine the dynamics between different types of energy consumption and TFP growth in the Asia-Pacific countries.

After the Kyoto Protocol, many countries concentrated on increasing the share of renewable energy in the energy basket to reduce greenhouse gas emissions. In this context, many studies in energy economics have focused on investigating the linear and nonlinear impacts of renewable energy consumption (REC) and non-renewable energy consumption (NREC) on economic growth across different countries/regions and periods (Apergis & Payne, 2011; Iyke, 2015; Jafri et al., 2021; Shastri et al., 2020; Zafar et al., 2019). However, the number of studies exploring the impact of disaggregated energy consumption on TFP growth is meager. For instance, Tugcu (2013) reported for the Turkish economy that REC positively impacted TFP growth, while nuclear and fossil fuel energy consumption negatively affected TFP growth. For BRICS countries, Tugcu and Tiwari (2016) found bidirectional causality for NREC and TFP growth, but no noticeable causality between REC and TFP growth. In a study of 36 countries, Rath et al. (2019) documented that renewable energy enhanced TFP growth, whereas fossil fuel energy consumption reduced it.

To the best of the authors’ knowledge, none of the previous studies on the nexus between disaggregated energy consumption and TFP growth have examined the asymmetric impacts of REC and NREC on TFP growth. Moreover, while previous studies used REC and NREC as key explanatory variables, we instead use renewable energy intensity (REI) and non-renewable energy intensity (NREI). The justification behind this approach is that, as the economy grows, energy usage must increase, but the issues are how intensely the economy uses energy and which sources are used with greater intensity. The objective of this study is to explore both the long- and short-run asymmetric impacts of REI and NREI on TFP growth in 17 Asia-Pacific countries[1] using a multivariate framework. In this study, we use the nonlinear autoregressive distributed lag (NARDL) model introduced by Shin et al. (2014). We also use foreign direct investment (FDI), trade openness (TO), and carbon emissions (CO2) as control variables (Isaksson, 2007; Kalaitzidakis et al., 2007).

This study finds that REI has a positive impact on TFP growth in the long run, whereas NREI has a negative effect on it in the short run. However, positive and negative shocks of NREI have asymmetric impacts on TFP growth in both the long and short run.

The remainder of the paper is organized as follows. Section II presents the data sources and methodology. The empirical results are discussed in Section III. Section IV draws a conclusion and makes policy implications.


A. Variable Description and Data

The study uses panel data from 1990 to 2018 for 17 Asia-Pacific countries. The dependent variable of our model, TFP growth, is constructed as the logarithmic difference of TFP between two subsequent years. The two key explanatory variables, REI and NREI, are formed as the ratio of energy consumption to the real gross domestic product (GDP). The data for TFP and the real GDP (measured in millions of constant 2011 US dollars) are collected from the Penn World Table (version 10.0). The data for REC and NREC (measured in quadrillions of British thermal units) and CO2 emissions (measured in millions of metric tons) are collected from the U.S. Energy Information Association (EIA). The FDI data (as a percentage of the GDP) are obtained from UNCTAD, and the TO data, measured as the sum of imports and exports (as a percentage of the GDP), are obtained from the World Development Indicators.

B. Methods

Following Rath et al. (2019), we express the relations between NREI, REI, FDI, TO, CO2, and TFP growth (TFPG) in the following functional form:

\[\begin{align}\text{TFPG}_{\text{it}} =&\ \alpha_{0} + \alpha_{1}\text{lnNREI}_{\text{it}} + \alpha_{2}\text{lnREI}_{\text{it}} \\ &+ \alpha_{3}\text{lnFDI}_{\text{it}} + \alpha_{4}\text{lnTO}_{\text{it}} + \alpha_{5}{\text{lnCO}_{2}}_{\text{it}} \\ &+ \epsilon_{\text{it}}\end{align}\tag{1}\]

where \(\epsilon_{\text{it}}\) is the disturbance term.

The NARDL representation of our model is as follows:

\[\begin{align}{\mathrm{\Delta}TFPG}_{\text{it}} =&\ \alpha_{0} + \tau\text{TFPG}_{it - 1} + \theta_{1}^{+}\text{lnNREI}_{it - 1}^{+} \\&+ \theta_{2}^{-}\text{lnNREI}_{it - 1}^{+} + \theta_{3}^{+}\text{lnREI}_{it - 1}^{+} + \theta_{4}^{-}\text{lnREI}_{it - 1}^{-} \\&+ \theta_{5}\text{lnFDI}_{it - 1} + \theta_{6}\text{lnTO}_{it - 1} \\&+ \theta_{7}{lnCO_2}_{it - 1} + \sum_{j = 1}^{p - 1}\pi_{j}{\mathrm{\Delta}TFPG}_{it - j} \\&+ \sum_{j = 0}^{q - 1}{({\varphi_{1}}^{+}{{\mathrm{\Delta}lnNREI}_{it - j}}^{+} +}{\varphi_{2}}^{-}{{\mathrm{\Delta}lnNREI}_{it - j}}^{-}) \\&+ \sum_{j = 0}^{q - 1}{({\varphi_{3}}^{+}{{\mathrm{\Delta}lnREI}_{it - j}}^{+} +}{\varphi_{4}}^{-}{{\mathrm{\Delta}lnREI}_{it - j}}^{-}) \\&+ \sum_{j = 0}^{q - 1}{\varphi_{5}{\mathrm{\Delta}lnFDI}_{it - j} + \sum_{j = 0}^{q - 1}{\varphi_{6}{\mathrm{\Delta}lnTO}_{it - j}}} \\&+ \sum_{j = 0}^{q - 1}{\varphi_{7}{\mathrm{\Delta}lnCO_2}_{it - j}} + \mu_{\text{it}}\hspace{15mm}(2)\end{align}\]

where \(\text{lnNREI}_{\text{it}}^{+}\) and \(\text{lnNREI}_{\text{it}}^{-}\) represents the partial sum decomposition of positive and negative changes in\(\ \text{lnNREI}_{\text{it}}\), while \(\text{lnREI}_{\text{it}}^{+}\) and \(\text{lnREI}_{\text{it}}^{-}\) are the same for \(\text{lnREI}_{\text{it}}\). Moreover, the φ terms are the short-run coefficients, and the long-run coefficients can be calculated as \(\lambda' = \frac{\theta'}{\tau}\). To test the asymmetric impacts of REI and NREI on TFPG, the Wald test is employed.


This study finds the presence of slope heterogeneity and cross-sectional dependence in the variables studied.[2] First-generation unit root tests (Breitung, 2001; Im et al., 2003; Levin et al., 2002; Maddala & Wu, 1999) and second-generation unit root tests (Pesaran, 2007) are employed to check the order of integration of the variables. Table 1 shows the results of the unit root tests (both first and second generation).

Table 1.Panel unit root tests results
Variables Levin, Lin & Chin t* (LLC)
(Constant& Trend)
(Constant& Trend)
Im-Pesaran-Shin (IPS)
(Constant& Trend)
Fisher ADF Chi-Squared
(Constant& Trend)
(Constant& Trend)
I(0) I(1) I(0) I(1) I(0) I(1) I(0) I(1) I(0) I(1)
TFPG -7.119*** -8.994*** -10.9328*** 320.940*** -4.552***
lnNREI -0.587 -5.761*** 1.406 -7.6369*** -2.1298 -10.9934*** 34.317 98.620*** -2.212 -4.808***
lnREI -1.595* -0.290 -10.391*** -4.0483*** 21.804 126.335*** -3.091**
lnFDI -5.683*** -5.139*** -7.9774*** 99.370*** -3.696**
lnTO -2.135** 0.279 -11.542*** -3.3532*** 29.305 81.869*** -2.871**
lnCO2 -2.986*** 2.965 -8.704*** -1.4962 -11.4108*** 35.979 59.767*** -2.416 -4.688***

This table reports the results of panel unit root tests. *, **, and *** represent the significance level at 10%, 5%, and 1% respectively. The values in the parentheses represent standard errors. ADF denotes Augmented Dickey-Fuller, while CIPS denotes Cross-sectional Augmented IPS.

Table 1 shows that the variables are integrated in mixed order, that is, I(0) or I(1). This is why the panel NARDL model is used in this study, since it allows one to estimate whether the variables are I(0) or I(1) (but not I(2)) and to test for asymmetry in both the short and long run.

Table 2.Results of NARDL model
Variables Coefficients
Long run
poslnNREI 0.733
neglnNREI 0.777
poslnREI 1.147***
neglnREI 1.185***
lnFDI -2.421**
lnTO 0.401
lnCO2 0.801***
ECT -0.942***
D.poslnNREI -39.18***
D.neglnNREI -39.20***
D.poslnREI 0.422
D.neglnREI 0.507
D.lnFDI 7.164
D.lnTO 1.632
D.lnCO2 32.35***
Constant -9.090***
Diagnostics test Log Likelihood = -845.1183
Long-run and short-run asymmetric test (Wald test)
Variables Long run Short run
χ2 Prob. χ2 Prob.
NREI 9.54 0.002 8.19 0.004
REI 0.09 0.765 0.000 0.948

This table shows the results of the NARDL model. *, **, and *** represent the significance level at 10%, 5%, and 1% respectively. The values in the parentheses represent standard errors.

Table 2 reports the results of the NARDL estimation and shows that, in the long run, a positive shock in REI increases TFP growth, whereas a negative shock in REI decreases TFP growth. On the other hand, a positive shock in NREI raises TFP growth in the short run, while a negative shock reduces it. So, the empirical results reveal that renewable energy positively impacts TFP growth (in the long run) and non-renewable energy negatively affects TFP growth (in the short run), which is along the same lines as the findings of Tugcu (2013) and Rath et al. (2019). In addition, FDI hampers TFP growth in the long run. This finding contradicts the finding of Rath et al. (2019), but is supported by Binh et al. (2014), who stated that, if FDI concentrates more on natural resource extraction than manufacturing, it will tend to reduce TFP growth. On the other hand, CO2 emissions positively affect TFP growth in both the long and short run, which is consistent with the finding of Kalaitzidakis et al. (2007). Due to concerns about growing emissions and their negative impact on the environment, countries are introducing new cleaner technology, which in turn is boosting TFP growth.

Lastly, the statistical significance of the error correction term (ECT) confirms the asymmetric long-run relation between the concerned variables. However, Wald test results reveal that NREI asymmetrically impacts TFP growth, while REI symmetrically impacts TFP growth in both the long and short run.


This study aims to investigate the asymmetric impact of renewable and non-renewable energy on TFP growth throughout 1990–2018 for 17 Asia-Pacific countries. It concludes that. in the long run, an increase (decrease) in REI leads to a rise (fall) in TFP growth, whereas an increase (decrease) in NREI depresses (expands) TFP growth in the short run. Moreover, FDI negatively affects TFP growth in the long run, while CO2 emissions positively impact TFP growth in both the long and short run. The result of the Wald test confirms the asymmetric linkage between only NREI and TFP growth.

In view of these findings, this study’s policy implication is that, to achieve sustained TFP growth as well as economic growth, the Asia-Pacific countries need to focus on using renewable energy sources more intensely, which, in turn, will help to reach the goal of reducing carbon emissions and achieving sustainable development.

  1. Australia, China, Fiji, Hong Kong, India, Indonesia, Japan, Macao, Malaysia, Mongolia, New Zealand, Philippines, Russia, Singapore, South Korea, Sri Lanka, and Thailand. The countries are selected on the basis of availability of data.

  2. The results are not provided in this paper to conserve of space, but they are available from the authors upon request.