Connectedness of green investments and uncertainties: new evidence from emerging markets

Abstract

Purpose

This study aims to investigate the response of green investments of emerging countries to own-market uncertainty, oil-market uncertainty and COVID-19 effect/geo-political risks (GPRs), using the tail risks of corresponding markets as measures of uncertainty.

Design/methodology/approach

This study employs Westerlund and Narayan (2015) (WN)-type distributed lag model that simultaneously accounts for persistence, endogeneity and conditional heteroscedasticity, within a single model framework. The tail risks are obtained using conditional standard deviation of the residuals from an asymmetric autoregressive moving average – ARMA(1,1) – generalized autoregressive conditional heteroscedasticity – GARCH(1,1) model framework with Gaussian innovation. For out-of-sample forecast evaluation, the study employs root mean square error (RMSE), and Clark and West (2007) (CW) test for pairwise comparison of nested models, under three forecast horizons; providing statistical justification for incorporating oil tail risks and COVID-19 effects or GPRs in the predictive model.

Findings

Green returns responds significantly to own-market uncertainty (mostly positively), oil-market uncertainty (mostly positively) as well as the COVID-19 effect (mostly negatively), with some evidence of hedging potential against uncertainties that are external to the green investments market. Also, incorporating external uncertainties improves the in-sample predictability and out-of-sample forecasts, and yields some economic gains.

Originality/value

This study contributes originally to the green market-uncertainty literature in four ways. First, it generates daily tail risks (a more realistic measure of uncertainty) for emerging countries’ green returns and global oil prices. Second, it employs WN-type distributed lag model that is well suited to account for conditional heteroscedasticity, endogeneity and persistence effects; which characterizes financial series. Third, it presents both in-sample predictability and out-of-sample forecast performances. Fourth, it provides the economic gains of incorporating own-market, oil-market and COVID-19 uncertainty.

Citation

Ogbonna, A.E. and Olubusoye, O.E. (2022), "Connectedness of green investments and uncertainties: new evidence from emerging markets", Fulbright Review of Economics and Policy, Vol. 2 No. 2, pp. 136-160. https://doi.org/10.1108/FREP-04-2022-0028

1. Introduction

The last two years has been characterized by alarmingly high uncertainty levels across countries of the world. The uncertainty was occasioned; first, by COVID-19 pandemic that led to institutional redundancy (Johnstone, 2021) and crumbled international trades and state economic activities due to imposed movement restrictions; and second, by the cases of rising geo-political tensions, emanating from conflicts, that has significantly affected most countries’ basic supplies, with oil and gas being greatly affected. The impacts of uncertainty on any economy is both socio-economically and environmentally facetted (Nguyen, Schinckus, & Su, 2022). High market uncertainty triggers increased risk aversion among investors, thereby prompting the search for alternative portfolio options to hedge against losses. Extant studies (see Dutta, Jana, & Das, 2020; Lee, Lee, & Li, 2020; among others) have revealed recent trends where investors are moving their portfolios to green investments that is instrumental to promoting eco-friendly environments (see Wang & Zhi, 2016; Broadstock & Cheng, 2019; Kanamura, 2020; among others), while also being a more sustainable and stable investment option (Park, Park, & Ryu, 2020). However, Fuss, Szolgayova, Obersteiner, and Gusti (2008) argue that investment pattern may be affected by own-market uncertainty, thus hampering productivity. There are online reports that itemize the riskiness of going green [1], hence, our motivation to study the dynamics of green investment markets, and its response to its own-market uncertainty.

Several studies have investigated the dynamics of the green investments in relation to other markets (see Bondia, Ghosh, & Kanjilal, 2016; Dutta, 2017; Reboredo, 2018; Broadstock & Cheng, 2019; Xia, Ji, Zhang, & Han, 2019; Dutta et al., 2020; Park et al., 2020; among others). A common feat with all previous studies is the mis-specification of the distribution of the returns, wherein normality is erroneously assumed. Several studies (see Andersen, Todorov, & Ubukata, 2020; Ogbonna & Olubusoye, 2021; Salisu, Gupta, & Ogbonna, 2021a, b; among others) evaluated the predictive potential of a measure of uncertainty that is based on tail thickness information (Engle & Manganelli, 2004; Bollerslev & Todorov, 2011; Bollerslev, Todorov, & Xu, 2015; Cremers, Halling, & Weinbaum, 2015; Baker et al., 2020) for stock returns. Consequently, we hinge on a more realistic representation of uncertainty using tail risks of green returns and oil returns.

As at April 2022, economic, social and governance (ESG) investments of emerging markets (except China) accounts for approximately 18% of foreign financing, a feat that quadrupled the recent years’ average. Amidst the observed positive trend in emerging markets’ adoption of sustainable finance, there exist concerns as to the accuracy of the valuation of sustainable “green” assets, given the inconsistency in data, disclosure requirements as well as sustainable finance classifications, which are considered potential threats to financial stability in such fast moving economies. In 2021, emerging markets, for the first time, gained more market shares than the advanced economies, with the strongest traction been linked to countries in Asia and the Western Hemisphere. The recent feats was accelerated by the emerging markets’ regulatory mainstreaming of sustainable finance strategies, which was majorly driven by green borrowing strategies and pandemic-induced demand. There are also stances of active implementation of climate-related disclosure requirements by some emerging markets. However, green financing in emerging markets yet remains insufficient as a result of inadequate carbon pricing, private climate finance constraints, high upfront capital and transaction costs, and significant project and/or country risk/uncertainty (IMF, 2022) [2]. The afore-stated feats inform our consideration of the emerging markets.

From the foregoing, we make four main contributions. First, we obtain tail risks estimates for green investments’ returns of selected emerging economies and global oil prices, using daily data and then incorporate same in a predictive model for green investment returns; while controlling for oil tail risks as well as other uncertainty measures using recently developed COVID induced uncertainty (CIU) index (Olubusoye, Ogbonna, Yaya, & Umolo, 2021; Salisu, Ogbonna, Oloko, & Adediran, 2021). This is to assess the effect of own-uncertainty amidst oil-uncertainty and COVID-19 effect/geo-political risk (GPR). Second, the Westerlund & Narayan, 2012, Westerlund & Narayan, 2015 (WN)-type distributed lag model that simultaneously accounts for salient feat such as persistence, endogeneity and conditional heteroscedasticity effects, which are characteristics of most financial series, is employed. Third, we present in-sample predictability as well as out-of-sample forecast evaluations using the root mean square error (RMSE) and the pairwise Clark and West (2007) test statistic for nested models. Fourth, we draw from Liu, Ma, Tang, and Zhang (2019) to compute the economic significance of incorporating the oil-uncertainty and CIU or GPR in the predictive model for green returns.

Foreshadowing the results, green returns responds significantly to own-uncertainty (mostly positively), oil-uncertainty (mostly positively), COVID-19 effect (mostly negatively) and GPR (mostly positively). There is evidence of hedge potential against uncertainties that are external to green investments. Also, incorporating external uncertainties improves the in-sample predictability, enhances out-of-sample forecast precision and yields some economic gains. On the significance of the study, intending investors in green assets as well as policy-makers would find our results useful. For the investors, our finding beams the light on the hedge-potential of green assets in fast growing economies, with confirmatory stance of their choice in a portfolio of investments. The information on the resilience of emerging markets’ green assets to internal and external uncertainties would be useful for policy makers that are keen on driving sustainable financing and post COVID-19 recovery, in line with the sustainable development goals (SDGs) targets.

Following the introductory section; Section 2 presents a brief review of extant literature; Section 3 discusses the methodology and data employed in the study; Section 4 presents and discusses the empirical results bordering on the predictability, forecast evaluation and economic significance; while Section 5 concludes the study.

2. Literature review

We review the literature on stock prices of green companies, global oil prices and the COVID effect. Henriques and Sadorsky (2008) empirically show the existence of relationship among prices of alternative energy stocks, oil, technology stocks and interest rate using a vector autoregressive (VAR) model framework, with technology stocks exhibiting higher impact, on alternative energy stocks, than oil prices. Sadorsky (2012b) confirms the stance by Henriques and Sadorsky (2008) using comparatively four multi-variate GARCH models variants (Baba-Engle-Kraft-Kroner (BEKK) (Baba, Engle, Kraft, & Kroner, 1990), diagonal, constant- and dynamic-conditional correlations) to show the higher correlation of alternative energy stocks with technology stocks than with oil prices and show evidence of oil prices volatility spill over to clean energy stock markets. Sadorsky (2012a) however challenges the standpoint that oil price increases are responsible for high levels of uncertainty in clean energy equity markets. Some studies examine oil price shocks-renewable energy nexus (see Kumar, Managi, & Matsuda, 2012; Wen, Guo, Wei, & Huang, 2014; among others). Reboredo (2015) reveals that the nexus between oil prices on clean energy stock are time-varying, using a copula model framework.

Pham (2016) examines the green investment market volatility behaviour, using a GARCH model framework, in an attempt to ascertain the riskiness of going green. The author finds evidence of time-varying spill over from conventional bond markets to green investments market. Using a threshold co-integration test that explicitly accounts for regime shifts in the long-run relationship, Bondia et al. (2016) finds long-run relationship between energy equity and oil markets nexus, which imperatively suggests either the inclusion of a break dummy to account for the observed structural breaks or data sub-sampling using the observed break points. Reboredo, Rivera-Castro, and Ugolini (2017) employ wavelet-based (continuous and discrete) models to examine the nexus between renewable energy stock and oil prices, and finds weak (strong) short-run (long-run) co-movements in the prices. Their stance also agrees with other studies (see Ahmad, 2017; Xia et al., 2019; among others), wherein rising energy prices tend to stimulate green energy stock prices. Kocaarslan and Soytas (2019) verified the existence of dynamic conditional correlations among clean energy, technology stocks and oil prices, and thereafter employed the auto-regressive distributed lag (ARDL) model to examine the impact of United States (US) dollar changes on the observed dynamic conditional correlations, controlling for some financial variables. The US dollar appreciation dominantly drives the dynamic conditional correlations, with empirical evidence of higher dominance when asymmetry effects are factored into the model.

Green investments have also been shown to respond negatively (see Ahmad, Sadorsky, & Sharma, 2018; Dutta, Bouri, Das, & Roubaud, 2019) and positively (see Dutta, 2017) to oil volatility, with evidences of high volatility spill overs from oil market to green investments markets (Dutta, 2018a). There are evidences of safe-haven and/or hedge potentials of clean energy amidst uncertainty in other associated markets (see Dutta (2018a, b); Dutta et al., 2019; Bouri, Jalkh, Dutta, & Uddin, 2019; among others). These features are subjected in this study to the COVID and post-COVID period for new pieces of evidence. There are evidences of time-varying correlations between green investments markets and other markets, which are also sensitive to oil price changes as well as changes in other financial uncertainty measures (Broadstock & Cheng, 2019). Findings from Gormus, Nazlioglu, and Soytas (2018) reveals that the energy markets impact the bond market, irrespective of whether prices or volatilities are considered.

Troster, Shahbaz, and Uddin (2018) finds lower-tail-dependence of oil price to renewable energy spending, while Lee et al. (2020) establishes that there exists a bi-directional relationship between green investments and oil price for lower quantiles. Dutta et al. (2020) examines the reaction of green investments to oil shocks using Markov regime switching model and finds higher susceptibility of green assets to oil market volatilities than oil price fluctuations. This suggests possible suitability of oil market uncertainty/volatility for modelling and predicting green investment returns. Kanamura (2020) shows that green investment returns responds positively to declining market uncertainty; with evidence of the superiority of green investments to conventional investments.

Yousaf, Suleman, and Demirer (2022) examined the diversification and hedging potential of green investments for conventional stocks in the context of the recent COVID-19 pandemic and provides empirical evidence of the safe-haven feature of green investments during COVID-19 pandemic, while stating that green investment are more of a financial necessity for financial stability and performance than being a luxury good. Some research studies (see Tu, Mo, Liu, Gong, & Fan, 2021; Guérin & Suntheim, 2021) have suggested that green investments policies have the potential to effectively mitigate the adverse effect of COVID-19 pandemic. Tiwari, Abakah, Gabauer, and Dwumfour (2022) investigated the dynamic spill over effect among green bond, renewable energy stocks and carbon market during COVID-19 pandemic in a bid to ascertain the implications for hedging and investments strategies, using both the time varying parameter - vector autoregressive [TVP-VAR] and least absolute shrinkage and selection operator (LASSO) dynamic connectedness model frameworks. The authors show green investments, in addition to being potential diversifiers for renewable energy, to be shock receivers, rather than shock transmitters, given their negative net spill over values, hence, their plausible stability potential amidst surrounding markets’ uncertainties.

From the foregoing, there are established nexuses between oil prices/volatility and green investments returns, and evidences of the green investment returns’ stability to market uncertainty. Most of the studies neglect the tail information inherent in market returns series; thus, mis-specifying or wrongly assuming normality of returns that are supposed to be non-normally distributed. Therefore, the resulting estimates of the green investments markets and other market uncertainties nexus are biased. There are also evidences of regime switches, which suggest a consideration of sub-sample periods that capture the waves of the recent pandemic. However, there is still a dearth of literature investigating the wave-dependent connectedness of green returns and the uncertainties in own-market and other markets, with adequate consideration of the COVID-19 effect. We therefore adopt a WN-type model that accommodates salient features; such as persistence, endogeneity and conditional heteroscedasticity, with tail risks information to model the impact of own- and oil-market uncertainties and COVID-19 effect on green returns.

Intuitively, internal and external uncertainties are expected to either positively and/or negatively affect returns on green investments; since the latter, as with most financial assets, is news/sentiment-driven. Under high market uncertainty (risky market situation), risk taker investors may increase their investments in a bid to take advantage of future rise in market prices and make more gains, whereas a risk averse investor may reduce his investments to mitigate against possible losses. While the former may increase the available capital, induce productivity and subsequently lead to higher returns; the latter may inhibit productivity and consequently lead to lower returns, since re-allocation of resources may be slowed down and projects are more expensive (Christiano, Roberto, & Massimo, 2014). Also, investors may seek optimal diversifications of portfolio investments if there are several available investment options, with adequate information on the associated uncertainties of the portfolio assets’ make-up. The above are likely to lead to either positive or negative co-efficients of the associated uncertainty measures. Hence, we herein test the hypothesis of the significance of uncertainty in predicting green investments’ returns.

3. Methodology and data

We draw from Westerlund & Narayan, 2012, Westerlund & Narayan, 2015 [WN]-type distributed lag model that allows for simultaneous accommodation of some salient data features (endogeneity, persistence and conditional heteroscedasticity) that are inherent in financial and economic series. Again, given that our variables are either returns or volatilities, accounting for these salient features cannot be overemphasized. The model accounts for endogeneity and persistence by the inclusion of a differencing term, while the heteroscedasticity is resolved by pre-weighting the model variables with inverse of the standard deviation of conventional GARCH(1,1) model residuals.

The WN-type model is given by

(1)ginvt=α+∑k=1Kβkunct−1k+∑k=1Kγk(unctk−ρkunct−1k)+εt

where unctk=[trtginvtrtoilciut] is a K− vector of uncertainties at time t that reflects own-uncertainty, oil-uncertainty and CIU/GPR, where K=3; α is the constant; βk denotes the slope coefficients associated with the incorporated predictors; [∑k=1Kγk(unctk−ρkunct−1k)] represents the endogeneity/persistence adjustment terms for corresponding uncertainty measures, with ρk - the autoregressive coefficient of each predictor variable, indicating the corresponding degrees of persistence; while εt is the residual term that follows a white noise process.

We extract three models from equation (1) by restricting some model parameters to zero. Model-1 is obtained by imposing the restrictions, β2=0, β3=0, γ2=0 and γ3=0; hence, considering the nexus between green returns and its own-uncertainty. Model-2 imposes the restrictions, β3=0 and γ3=0 to assess the simultaneous impact of own-market and oil-market volatilities on green returns. Model-3 is the unrestricted model specified in equation (1), where we consider the effect of own-uncertainty and oil-uncertainty on green returns, while controlling for the recent COVID pandemic and GPRs .

The study employs the Clark and West [hereafter, CW] (2007) test to formally compare pairs of the three WN-type model variants, with a restricted variant serving as the benchmark within any paired comparison. The statistic is well suited for nested models, testing if the differences in forecast errors of contending models are statistically zero. The estimation equation for the CW statistic is given in (2):

(2)f^t+h=(rt+h−r^1t,t+h)2−[(rt+h−r^2t,t+h)2−(r^1t,t+h−r^2t,t+h)2]

where h is the forecast period; (rt+h-r^1t,t+h)2 and (rt+h-r^2t,t+h)2 are the squared residuals from the restricted and unrestricted variants, respectively, of our WN-type distributed lag model; while (r^1t,t+h-r^2t,t+h)2 is an adjusted squared residual that is peculiar to the CW test and incorporated as a corrective measure for the noisy forecasts of the larger model. The term, f^t+h is defined as MSE1-(MSE2-adj.), where MSE1=P-1∑(rt+h-r^1t,t+h)2, MSE2=P-1∑(rt+h-r^2t,t+h)2, adj.=P-1∑(r^1t,t+h-r^2t,t+h)2 and P represents the number of averaged forecast points. The test is based on the regression of f^t+h on a constant and the determination of equality, or otherwise, of paired contending forecast errors using the t-statistic of the estimated constant. Significant t-statistic would imply outperformance of the unrestricted model over the restricted model.

Table 1 presents a brief description of the green investments series from nine developing economies (Brazil, Russia, India, Indonesia, China, South Africa, South Korea, Turkey and Vietnam; oil price proxies (WTI and Brent); GPR and the recently developed CIU measure. On the country selection, we considered globally recognized emerging markets’ classification – BRICS (Brazil, Russia, India, China, and South Africa) and next eleven developing countries, but was restrained to nine emerging markets based on data availability. The selected stocks are essentially stocks of renewable-energy– and carbon-efficient–based firms that have the highest market capitalization in the corresponding countries. We consider daily data that spans 01.03.2000 to 09.12.2022, to accommodate periods of different financial crises as well as the inception, peak and flattening of the curve of the COVID-19 pandemic and the recent geo-political tension (Russian-Ukraine war); which amounts to 5921 data points. The data for corresponding countries do not all have the same start dates; hence, the differences in the number of data points. The data availability for the selected green investments defines the start date for the corresponding country. All the green series and oil variables are sourced from a financial database (www.investing.com); GPR, which is a measure of economic uncertainty, is sourced from Caldara and Matteo (2022) dataset (https://www.matteoiacoviello.com/gpr_country.htm); while the CIU is sourced from Olubusoye et al. (2021) and Salisu, Ogbonna, Oloko, and Adediran (2021). The green stocks series are transformed into returns, as a way to circumvent the problem of unit roots; while the market uncertainty measures (for stocks and oil prices) are obtained as the conditional volatility from an asymmetric ARMA(1,1)-GARCH(1,1) model with Gaussian innovation. The intuition here is to adopt the tail information inherently in the corresponding series, which gives a representation of risks or uncertainties that characterize financial series.

Table 2 presents some summaries and preliminary analyses to highlight the inherent data features (measures of location, variability and spread; as well as formal tests for evidence of conditional heteroscedasticity and serial correlation). This is to adequately incorporate observed data feats into the study’s adopted model. The average stock returns are positive for all the countries (except China) with the most and least volatile green returns corresponding to Russia and Turkey. The green returns are negatively (China, India, Russia and Vietnam) and positively (Brazil, Indonesia, South Africa, South Korea and Turkey) skewed and leptokurtic; which is an indication that the series are not normally distributed. All the green returns are stationary as revealed by the augmented Dickey-Fuller (ADF) unit root test; with evidence of the presence of conditional heteroscedasticity and serial correlation at specified lags, with low persistence. The green markets’ uncertainties are normally distributed, with evidence of ARCH effect, serial correlation and high persistence. A similar feat is exhibited by oil-uncertainty proxies, CIU and GPR. From the foregoing, the most appropriate model should incorporate these observed salient features simultaneously within its framework – the justification for our choice of the WN-type distributed lag model. The co-movement of the green returns with its own- and oil-market uncertainties are displayed in Figure 1. While the green returns are mostly characterized with volatility clustering and jumps, the oil-uncertainty appears to exhibit a significant jump around the period of the COVID-19 pandemic.

4. Empirical results

We adopt the WN-type distributed lag model framework to assess the nexus between green returns and (own- and oil-) market uncertainties, while controlling for the effect of the other global uncertainties (COVID pandemic as well as geo-political risks). We present the in-sample predictability results (in Table 3) when the extracted conditional standard deviation (volatility) of West Texas intermediate (WTI) and Brent are used to proxy the oil market uncertainty; and the forecast evaluation (Table 4 and A1 in appendix) at 15-, 30 and 60- day out-of-sample forecast horizons. For the three considered models, the coefficients associated with own-market uncertainty, oil-market uncertainty, geo-political risks and the CIU are reported. Model-1 (baseline model) is a subset of Model-2 (Model-1 augmented with oil-uncertainty) and Model-3 (Model-1 augmented with oil-uncertainty and CIU/GPR). We consider the GPR to account for global uncertainties other than those emanating from the green and oil markets. Consequently, Model-3 is the unrestricted model for Model-1 and Model-2; while Model-2 is an unrestricted model for Model-1. We consider the full, pre-COVID, COVID and Russia-Ukraine war [hereafter, RUW] sample periods to establish the green returns – uncertainty nexuses at specified sample periods. This is to ascertain whether the nexus is period invariant. GPR was employed as the other measure for global uncertainty across the four specified sample intervals while CIU is only used during the COVID period. The resulting model forecasts are compared using relative RMSE [hereafter, RRMSE] and the conventional CW test.

4.1 In-sample predictability

In Table 3, where volatility in WTI oil was used to capture oil uncertainty, we find predictability of own-market, oil-market and other global uncertainties (CIU and GPR) for green returns. Considering all three model variants in the full sample period, green returns appears to consistently respond negatively to own-uncertainty in the cases of Brazil (except under Model 1), Russia and Turkey; while it consistently responds positively in the cases of China, India, Indonesia, South Africa, South Korea and Vietnam. Similar feats (except for Russia) are observed under the pre-COVID period. This implies that high risk – low returns characterizes the green investments in Brazil, Russia and Turkey; while high risk – high returns characterizes the remaining six green investments. The economic intuition of the high risk – high returns is that of resilience to own-market shocks, while the high risk – low returns is that of vulnerability to its own shocks. In the former, investors risk the loss of their invested funds if the market instability lingers for a longer period; while in the former, prolonged market instability yield more gains for the investors.

Under COVID period, the green returns – own-market nexus is significantly negative (except for Indonesia, South Africa, South Korea and Vietnam). However, we find the nexus to be consistently positive for all the green investments under the RUW sub-period, which implies the profitability and resilience of green investments to rising geo-political tensions. This is not unexpected, seeing that the geo-political tension has had its toll majorly on fossil fuels, since the countries at war account for a large chunk of global fossil fuels supplies. For a positive green – uncertainty nexus, green returns are higher when own-market is risky/highly uncertain. Intuitively, the market is characterized by forward looking risk-taker investors that invest more when future prices are expected to be high and reduce investments when the future prices are expected to be low. A negative green – uncertainty nexus implies that the returns are lower when own-market uncertainty heightens. This is the immediate consequence of increased financial (borrowing) costs, which leads to reduced investments and consequently low productivity. The green investments that exhibit the high risk – high returns features are resilient to uncertainty that is occasioned by trading activities within own-market, while green investments with high risk – low returns feats are less resilient to own-uncertainty.

The effect of oil uncertainty on green returns are mixed (see Table 3: Model-2 (column 4) and Model-3 (column 6)). We find predictability (statistically significant coefficients) of oil-uncertainty for green returns; which is an indication of the connectedness of the green investment and oil markets. Under full sample, positive green – oil-uncertainty nexuses are observed in the cases of Brazil, Indonesia, Russia, South Africa, Turkey and Vietnam; which connotes higher turn-over for green investors during oil crises. Green investments (Brazil, Indonesia, Russia, South Africa, Turkey and Vietnam) could therefore serve as hedge for oil investments during oil crises. Imperatively, investors may allocate a higher share of their portfolio (comprising green and oil stocks) investments in green whenever the oil market is unstable to mitigate losses. In a sub-period examination; while oil uncertainty, during the COVID-19 pandemic, seem to cause green returns to rise, it caused green returns to decline in the RUW period. This indicates that the green – oil-uncertainty nexus may be dependent on the event that triggered the oil uncertainty; and intuitively implies that different domestic and/or global events could structurally alter the nexus. Controlling for other global uncertainties (COVID pandemic and geo-political risks using the CIU and GPR, respectively) does not alter the predictability of oil-uncertainty for the considered green returns except in the case of China. The green returns – oil-uncertainty nexus remains the same when Brent is used to proxy oil uncertainty (see Table A1 in appendix) except for a few cases like China where we find significantly positive (Model-3 under full sample) predictability of oil-uncertainty for green returns. In sum, the predictability of oil-uncertainty for green returns, with or without control variables, is confirmed.

Besides using the CIU and GPR, separately, as control variables to capture the effect of the COVID pandemic and geo-political tension; we also observe their direct impact on green returns. Green investments in China, Indonesia, South Africa and Vietnam are vulnerable to COVID effect, since lower returns were associated with worsening COVID pandemic. The case is different for green investments of Brazil, India, Russia, South Korea and Turkey; which were found to be resilient to the COVID effect, given that returns continued to rise despite the worsening COVID pandemic. Essentially, while these resilient green investments are potential safe havens for investors in the advent of a global health-induced uncertainty; the vulnerable green investments may require additional green financing policies to mitigate the impact of such global health-induced uncertainty. For the global uncertainty occasioned by geo-political tension (the RUW period), all the green investments (except Russia that is directly affected) were resilient to geo-political uncertainty originating from the RUW. The severity of the impact of geo-political uncertainty on green investments may be dependent on the extent of the economy’s involvement.

From the foregoing, the highest capitalized green investments are vulnerable to own-uncertainty (Brazil, Russia and Turkey); oil-uncertainty (India and South Korea); COVID uncertainty (China, Indonesia, South Africa, South Korea and Vietnam) and geo-political tensions (Russia). Brazil and Turkey are resilient to both oil and COVID uncertainties, but vulnerable to their own uncertainty. These predictability stances transcend the oil price proxies.

4.2 Forecast evaluation

Having ascertained the predictability of own- and oil-uncertainties for green stocks, with and without controlling for COVID effect or geo-political risks, we further subject the models to forecast evaluation using RRMSE and CW statistics. Given that we consider different sub-samples, we compute and report the proportion of out-performances for each model and oil-uncertainty proxies under the in-sample and three out-of-sample forecast periods. This is to ascertain that the predictability stance transcends the in-sample period; while also ascertaining the insensitivity, or otherwise, of our results to choice of forecast horizons. The RRMSE results are presented in Table 4. The RRMSE, which is a ratio of the unrestricted (numerator) model to the restricted (denominator) model, could be “less than unity”, “equal to unity” or “greater than unity”. “Equal to unity” indicates the equality of the forecasts from the contending models; while “less (greater) than unity” indicates that the numerator (denominator) model has more precise forecasts. We expect RRMSE values to be less than unity for our unrestricted model that incorporates additional variables to be adjudged the preferred model in any given pair of contending models.

From Table 4, there are relative out-performances of Model-2 over Model-1; Model-3 over Model-1 and Model-3 over Model-2. Under the columns labelled, “WTI”, Model-2 consistently outperformed Model-1 in the in-sample and across the out-of-sample forecast horizons (15-, 30- and 60-days ahead) except in the cases of China, Russia and South Africa (for longer out-of-sample periods). The incorporation of oil-uncertainty (Model-2) seem not to improve over the green investments’ returns’ forecasts of Model-1 that does not include oil-uncertainty as a predictor variable for China, Russia and South Africa. Model-3 outperformed Model-1 across the three forecast horizons and across all the green returns. In like manner, Model-3 consistently outperformed Model-2 in all cases except Brazil (out-of-sample) and Russia (60-days ahead forecast horizon only); and indicates that accounting for global uncertainty improves the forecast precision. Also, the incorporation of oil-uncertainty as a predictor variable, with and without controlling for other global uncertainties such as COVID effect and geo-political risk, does improve the forecast precision of green returns. The stance of performances is similar when Brent oil volatility is used, in place of WTI, to proxy oil-uncertainty. We subject our models to a more formal evaluation test – CW that shows whether the observed outperformances are statistically significant.

Table 5 presents the CW test results for in-sample and out-of-sample forecast horizons (15-, 30- and 60-day ahead) under the WTI oil-uncertainty (see Table A2 for Brent). The contending model pairs being considered are indicated on each horizontal pane of the corresponding tables. We expect a statistically significant and positive statistic for the first (unrestricted) model to be preferred over the second (restricted) model. A significantly negative statistic indicates preference of the restricted (second) model over the unrestricted (first) model. From the above, we compute the proportion of out-performances from the stances observed across different sub-samples. Table 5 shows that there are predominantly more positive than negative CW statistics, and more significant positive than negative CW statistics. This indicates that models that account for oil-uncertainty, with or without controlling for COVID or geo-political effects, yield more precise forecasts than the baseline model that is a function of own-uncertainty. This further highlights the relevance of the oil-uncertainty as a good predictor, which does not only result in significant in-sample predictability but also yield more precise out-of-sample forecasts. Similar feats are observed for Brent, which implies that our results are robust to choice of employed oil-uncertainty proxy and also confirms the predictability of oil-uncertainty for green returns. Intuitively, incorporating oil-uncertainty in the predictive model for green returns improves the forecast precision of the model with own-uncertainty.

4.3 Economic significance

In addition to the already established statistical significance of our predictive models, we test the economic significance of incorporating oil-uncertainty and COVID effects as predictors in the predictive model for green returns. This follows from the work of Liu et al. (2019), which allows for the determination of the economic gains of incorporating certain predictors in a baseline model. Essentially, the study tests whether our WN-type distributed lag models that incorporate oil-uncertainty and COVID effects do offer economic gains over a variant that has only own-uncertainty as a predictor.

Available portfolio to a typical mean-variance utility investors is optimally allocated by shares among choice investment options in contrast to a risk free asset; with an optimal weight, wt, defined as in equation (3)

(3)wt=1γθr^t+1+(θ−1)r^t+1fθ2σ^t+12

where γ is a risk aversion coefficient; θ is a leverage ratio (Zhang, Ma, Shi, & Huang, 2018), which is set to 6 and 8, premised on investors’ maintenance of a 10% account margin; r^t+1 is the stock returns forecast at time t+1; r^t+1f is a risk-free asset (Treasury bill rate); and σ^t+12 is 30-day moving average return volatility estimate. The certainty equivalent returns (CER) associated with investors’ optimal weight, wt, is defined in equation (4)

(4)CER=R‾p−0.5(1/γ)σp2

where R‾p and σp2 are the out-of-sample period portfolio returns’ mean and variance; and portfolio returns is given as Rp=wθ(r-rf)+(1-w)rf. Given the excess return volatility (σ2), the portfolio return variance is defined as Var(Rp)=w2θ2σ2. The determination of the economic significance is achieved by maximizing an objective utility function as in equation (5)

(5)U(Rp)=E(Rp)−0.5(1/γ)Var(Rp)=wθ(r-rf)+(1-w)rf-0.5(1/γ)w2θ2σ2

In presenting the economic significance of the contending models comparatively, we tabulate the models’ estimate of portfolio returns, its associated volatility, its certainty equivalent returns and Sharpe ratio (SP), SP=(Rp-rf)/Var(Rp). A model with the highest returns, CER and SP; and least volatility (see Liu et al., 2019) yields the most economic gains. The result is presented in Table 6 under the classification of leverage parameter set to 6 and 8, with risk aversion level set as 3.

There are mixtures of high returns – high volatility and low returns – high volatility stances, which aligns with the predictability results. There are a number of cases where Model-3 and Model-2 yielded higher economic gains than Model-1, given that the SPs of both models are higher than that of Model-1. This reveals the economic importance of incorporating oil-uncertainty as a predictor for green returns. The economic significance result may be sensitive to the considered green returns. Adjusting for endogeneity in all three models appears to mask the distinction in their performances; however, there are observable economic gains in favour of incorporating oil-uncertainty and CIU. The stance is not markedly different for the leverage ratios (6 and 8) and risk aversion combinations. Imperatively, oil-uncertainty is a relevant predictor for green returns, having satisfied both statistical and economic significance feats. This confirms the connectedness between green investments and uncertainty. Intuitively, green investment responds to its own shocks and other external shocks especially shocks emanating from the oil market. Conclusively, incorporating oil-uncertainty in the predictive model for green returns statistically improve forecast precision and yields economic gains; hence, qualifying oil-uncertainty and CIU as relevant predictors.

5. Conclusion

The study examines the green investment market dynamics of emerging markets from a predictability perspective, to ascertain the existence of connectedness between green returns and global oil market, while controlling for COVID-19 effect and geo-political tension. Essentially, the response of the green returns to own-uncertainty, oil-uncertainty and other global uncertainties, such as COVID-19 pandemic and GPR, are analysed. Nine different renewable-based and carbon-efficient-based stocks are selected, each from an emerging economy (Brazil, China, India, Indonesia, Russia, South Africa, South Korea, Turkey and Vietnam) based on their position as the most market capitalized stock in the corresponding country. The data spans 01.03.2000 to 09.12.2022, which covers periods of financial market uncertainties, COVID pandemic as well as Russian-Ukraine war. The study employed the WN-type distributed lag model with three variants that are distinguished by the predictor variables they comprise. Model-1 is the baseline model with own-uncertainty as the sole predictor. Model-2 is an augmented model that incorporates oil-uncertainty in addition to own-uncertainty. Model-3 further augments Model-2 by incorporating CIU/GPR. The three model variants are nested, which necessitated the adoption of CW test, in addition to RRMSE, to evaluate their forecast performances. Also, to fully understand the connectedness, we consider predictability and forecast performances under different sub-sample periods – full, Pre-COVID, COVID and RUW periods.

The study establishes significant connectedness between green investments and oil-uncertainty, given that there exist significant predictability of oil-uncertainty for green returns. Green returns respond significantly to own-uncertainty, oil volatility and COVID effect. There are mixed stances of predictability wherein we find some stances of negative (positive) relationships between green returns, and own-uncertainty and oil-uncertainty. Our result of negative relationship between green returns and oil-uncertainty aligns with the findings of Ahmad et al. (2018) and Dutta et al. (2019), while the stance of positive nexus aligns with Dutta (2017). The nexus does not differ markedly even when the predictive model takes COVID-19 effect or geo-political risks into account. While some green investments are vulnerable to own- and oil-uncertainty, some green investments exhibit hedging potentials against oil crises. The generality of green investments in emerging countries can be said to exhibit mixed behaviour in the midst of global market uncertainty. On the forecast evaluation, model that incorporate oil-uncertainty and CIU/GPR yield more precise forecasts, with higher economic gains; though high returns are associated with high risk. The result will provide quality information to guide investors in their portfolio strategies to mitigate loss of invested funds and policy makers’ financial policy decisions to focus on improving green financing.

This study tried to cover as many green investments as possible, but was restricted by data availability, since the advocacy for green investments is only beginning to gain popularity. For future research, it may be interesting to assess the spill over among green investments plausible trading strategies with green stocks among a portfolio of financial assets. Research could also be conducted to ascertain whether on green investments’ returns respond asymmetrically to external sources of uncertainty (good and bad volatilities).

Figures

Figure 1

Co-movement of stock returns and market uncertainty measures

Table 1

Variable definition

Country	Green stock	Acronym	Start date	End date	Data points
Green stock
Brazil	Magazine Luiza SA	MGLU3	05.04.2011	09.12.2022	2964
China	Wintime Energy Co. Ltd	600157	05.17.2012	09.09.2022	2692
India	Tata Power Co. Ltd	TTPW	01.04.2000	09.12.2022	5920
Indonesia	Adaro Energy	ADRO	03.01.2012	09.12.2022	2748
Russia	Gazprom PAO	GAZP	01.25.2006	09.12.2022	4339
South Africa	Exxaro Resources Ltd	EXXJ	11.28.2001	09.12.2022	5424
South Korea	SK Innovation	96770	08.31.2011	09.08.2022	2877
Turkey	ODAS Electrik Uretim Sariayi Ticaret AS	ODAS	05.23.2013	09.12.2022	2428
Vietnam	Gia Lai Electricity JSC	GEG	03.23.2017	09.12.2022	1428
Global variables
	Brent Crude oil	BRENT	01.05.2000	09.06.2022	5915
	West Texas Intermediate	WTI	01/05/2000	09.06.2022	5915
	COVID Induced Uncertainty	CIU	03.03.2020	01.27.2022	498
	Geo-Political Risk	GPR	01.03.2000	09.12.2022	5921

Table 2

Summary statistics and preliminary analyses

Country	Mean	CV	SK	KT	Unit root test	Conditional heteroscedasticity			Serial correlation						Persistence
					ADF	ARCH(1)	ARCH(5)	ARCH(10)	Q(1)	Q(5)	Q(10)	Q2(1)	Q2(5)	Q2(10)
Stock returns
Brazil	0.08	53.78	0.33	11.72	−59.46***I(0)	160.27***	56.83***	30.40***	0.04	2.91	20.31	152.34***	413.70***	550.79***	−0.09***
China	−0.04	−73.43	−0.01	8.80	−46.86***I(0)	159.65***	51.06***	27.14***	0.04	12.68**	13.93	151.03***	384.55***	494.75***	0.10***
India	0.06	41.36	−0.11	9.54	−55.19***I(0)	878.47***	204.47***	105.64***	0.02	20.00**	36.77***	765.65***	1416.50***	1691.20***	0.05***
Indonesia	0.03	104.94	0.26	5.75	−50.66***I(0)	22.64***	17.81***	26.37***	0.00	7.73	16.61*	22.50***	167.70***	279.17***	0.03*
Russia	0.00	1213.46	−1.10	31.64	−62.42***I(0)	76.94***	39.26***	23.38***	0.00	3.51	9.97	75.70***	272.83***	391.29***	0.05***
South Africa	0.06	40.97	0.04	7.04	−67.55***I(0)	204.24***	54.18***	38.16***	0.03	11.03*	27.54***	197.04***	343.48***	587.70***	0.09***
South Korea	0.00	493.06	0.66	12.08	−52.72***I(0)	289.84***	108.63***	56.11***	0.00	13.27**	19.43**	263.82***	806.27***	934.86***	0.02
Turkey	0.09	34.15	0.13	7.55	−32.33***I(0)	67.29***	24.41***	13.57***	0.01	13.97**	17.89**	64.99***	149.12***	181.92***	0.03*
Vietnam	0.03	74.67	−0.90	15.94	−36.95***I(0)	6.33**	2.14*	0.99	0.00	13.45**	22.49**	6.32***	13.16***	13.30	0.02
Stock market uncertainty
Brazil	17.17	1.11	3.49	17.65	−7.16***I(0)	5.78**	8.49***	5.42***	16.30***	44.93***	84.30***	5.79**	50.48***	69.33***	0.98***
China	8.07	1.16	2.65	11.78	−9.41***I(0)	1.53	0.56	0.29	19.21***	21.41***	32.61***	1.54	2.95	3.11	0.94***
India	6.07	1.03	4.58	33.74	−10.55***I(0)	205.97***	42.67***	21.43***	347.01***	374.85***	397.86***	199.24***	216.35***	219.90***	0.98***
Indonesia	7.78	0.58	3.82	26.83	−11.59***I(0)	2.34	26.06***	14.62***	8.92***	31.17***	40.36***	2.34	127.34***	166.25***	0.89***
Russia	5.53	1.62	4.84	29.95	−4.97***I(0)	0.00	0.04	0.02	1.86	20.75***	23.77***	0.00	0.20	0.23	0.99***
South Africa	6.17	0.69	3.79	25.81	−8.42***I(0)	11.08***	3.52***	2.28**	52.51***	75.27***	87.83***	11.07***	19.02***	25.56***	0.98***
South Korea	5.77	0.93	4.27	30.75	−5.64***I(0)	3.94**	5.79***	3.04***	96.66***	326.77***	341.59***	3.95**	30.27***	32.72***	0.99***
Turkey	9.08	0.70	3.41	21.43	−12.19***I(0)	0.24	0.73	0.40	4.11**	21.41***	28.79***	0.24	3.81	4.21	0.88***
Vietnam	6.63	1.48	17.44	426.25	−22.09***I(0)	0.00	0.00	0.00	0.00	0.54	0.84	0.00	0.01	0.02	0.49***
Global oil uncertainty
WTI	7.69	2.90	12.91	203.33	−10.08***I(0)	639.16***	136.24***	71.56***	511.22***	531.00***	624.00***	577.42***	680.19***	723.79***	0.98***
Brent	7.27	3.40	16.74	349.47	−9.37***I(0)	1.12	0.27	0.24	108.36***	158.32***	203.60***	1.13	1.42	2.46	0.96***
Other global uncertainty measures
CIU	21.57	0.67	2.06	7.90	−15.54***I(1)	10.42***	4.40***	1.67*	14.65***	23.79***	26.64***	10.32***	23.61***	23.70***	0.95***
GPR	113.57	0.61	4.51	38.10	−9.84***I(0)	83.38***	50.55***	56.74***	507.83***	678.41***	838.96***	82.30***	347.10***	816.30***	0.82***

Note(s): CV – Coefficient of variation; SK – Skewness; KT – kurtosis; ADF – Augmented Dickey-Fuller; CIU – COVID induced uncertainty; GPR –Geo-political risk. The ***, ** and * denote statistical significance of the corresponding tests at 1%, 5% and 10%, respectively. The ADF statistic tests the null of unit root, where significance indicates that the series are stationary. Significant ARCH statistics indicates evidence of ARCH effect in the series; while the significant Q(.) and Q²(.) statistics show evidence of series correlation at the specified lag order

Table 3

In-sample predictability (WTI as proxy for oil uncertainty)

Sample period	Model-1	Model-2		Model-3
	Stock market uncertainty	Stock market uncertainty	Oil (WTI) market uncertainty	Stock market uncertainty	Oil (WTI) market uncertainty	Other global uncertainties
Brazil
FULL	0.0005*** [0.0001]	−0.0010*** [0.0002]	0.0029*** [0.0005]	−0.0001 [0.0004]	0.0046*** [0.0004]	0.1195*** [0.0219]
Pre-COVID	0.0071*** [0.0012]	0.0043*** [0.0011]	0.0153*** [0.0018]	0.0009*** [0.0002]	0.0154*** [0.0006]	0.2810*** [0.0063]
COVID	0.0057*** [0.0018]	−0.0088*** [0.0013]	0.0088*** [0.0003]	−0.0075*** [0.0028]	0.0097*** [0.0011]	0.1225** [0.0536]
RUW	0.0829*** [0.0076]	0.1346*** [0.0041]	0.1262*** [0.0132]	0.2011*** [0.0140]	0.0692*** [0.0145]	2.6577*** [0.4931]
China
FULL	0.0015 [0.0010]	0.0000 [0.0002]	0.0000 [0.0002]	0.0018** [0.0007]	−0.0007*** [0.0003]	0.0453** [0.0182]
Pre-COVID	0.0017* [0.0009]	0.0113*** [0.0021]	0.0064*** [0.0010]	0.0139*** [0.0015]	0.0058*** [0.0009]	0.2976*** [0.0068]
COVID	−0.0449*** [0.0047]	−0.0470*** [0.0065]	−0.0005** [0.0002]	−0.0473*** [0.0051]	0.0011*** [0.0003]	−0.4794*** [0.0159]
RUW	0.0154 [0.0131]	0.1177*** [0.0426]	−0.0238** [0.0097]	0.1698*** [0.0366]	−0.0492*** [0.0146]	0.0618 [0.2280]
India
FULL	0.0041*** [0.0008]	0.0071*** [0.0009]	−0.0028*** [0.0004]	0.0036*** [0.0002]	−0.0009** [0.0004]	0.0234*** [0.0032]
Pre-COVID	0.0026*** [0.0009]	0.0066*** [0.0015]	−0.0055*** [0.0003]	0.0099*** [0.0015]	−0.0065*** [0.0003]	0.1285*** [0.0126]
COVID	−0.0200 [0.0135]	0.0108 [0.0085]	0.0003 [0.0008]	−0.0140** [0.0058]	0.0002 [0.0011]	0.3170*** [0.0708]
RUW	−0.0627 [0.0738]	0.0345* [0.0177]	0.0968*** [0.0072]	0.0101 [0.0215]	0.0512*** [0.0122]	0.4387*** [0.1536]
Indonesia
FULL	0.5114*** [0.0201]	0.5080*** [0.0209]	0.0025*** [0.0003]	0.5105*** [0.0210]	0.0033*** [0.0007]	0.1929*** [0.0064]
Pre-COVID	0.5338*** [0.0275]	0.5354*** [0.0239]	0.0051*** [0.0011]	0.5383*** [0.0247]	0.0093*** [0.0023]	0.1022*** [0.0221]
COVID	0.4310*** [0.0383]	0.4338*** [0.0376]	−0.0008 [0.0010]	0.3947*** [0.0369]	0.0041*** [0.0005]	−0.3551*** [0.0115]
RUW	0.3158*** [0.0300]	0.3641*** [0.0406]	−0.0081 [0.0124]	0.0913* [0.0476]	−0.0221*** [0.0023]	2.0392*** [0.2211]
Russia
FULL	0.0002 [0.0007]	−0.0029*** [0.0010]	0.0033*** [0.0004]	−0.0002 [0.0010]	0.0018*** [0.0004]	0.1073*** [0.0069]
Pre-COVID	−0.0015** [0.0006]	0.0042*** [0.0014]	0.0010 [0.0014]	0.0007 [0.0015]	−0.0018* [0.0010]	0.2155*** [0.0045]
COVID	−0.0207 [0.0142]	−0.1329*** [0.0290]	0.0035*** [0.0006]	−0.1754*** [0.0339]	0.0043*** [0.0008]	0.0254 [0.0315]
RUW	0.0529*** [0.0076]	0.0727*** [0.0079]	−0.0262*** [0.0065]	0.2269*** [0.0209]	−0.0443*** [0.0100]	−4.9998*** [0.4580]
South Africa
FULL	0.0057*** [0.0004]	0.0045*** [0.0009]	0.0023*** [0.0004]	0.0031*** [0.0006]	0.0024*** [0.0005]	0.1064*** [0.0117]
Pre-COVID	0.0037*** [0.0012]	0.0063*** [0.0007]	−0.0035*** [0.0011]	0.0071*** [0.0009]	−0.0079*** [0.0014]	0.0270*** [0.0085]
COVID	0.0448*** [0.0053]	0.0514*** [0.0111]	0.0043*** [0.0010]	0.1091*** [0.0085]	0.0016*** [0.0004]	−0.6044*** [0.0144]
RUW	0.0134 [0.0138]	0.0795*** [0.0178]	0.0037 [0.0035]	0.1155*** [0.0362]	−0.0321*** [0.0021]	1.1064*** [0.0938]
South Korea
FULL	0.0233*** [0.0019]	0.0347*** [0.0024]	−0.0004 [0.0004]	0.0387*** [0.0008]	−0.0020*** [0.0003]	0.0798*** [0.0051]
Pre-COVID	0.0325*** [0.0021]	0.0184*** [0.0038]	0.0125*** [0.0029]	0.0308* [0.0167]	0.0121 [0.0132]	0.0430 [0.1464]
COVID	0.0356*** [0.0016]	0.0567** [0.0282]	−0.0028 [0.0035]	0.0546*** [0.0039]	0.0008 [0.0007]	−0.4291*** [0.0171]
RUW	0.0031 [0.0133]	−0.0103 [0.0240]	0.0258** [0.0100]	−0.0342 [0.0223]	0.0042 [0.0064]	0.5698*** [0.1761]
Turkey
FULL	−0.0097*** [0.0005]	−0.0142*** [0.0002]	0.0028*** [0.0002]	−0.0141*** [0.0008]	0.0035*** [0.0003]	0.2000*** [0.0117]
Pre-COVID	−0.0057** [0.0028]	−0.0049*** [0.0002]	0.0056*** [0.0004]	−0.0020** [0.0009]	0.0083*** [0.0005]	−0.2556*** [0.0111]
COVID	−0.0070 [0.0069]	−0.0536*** [0.0016]	0.0043*** [0.0005]	−0.0363*** [0.0060]	0.0057*** [0.0007]	0.2404*** [0.0250]
RUW	0.1038*** [0.0111]	0.1059*** [0.0193]	0.0077 [0.0051]	0.1037*** [0.0250]	0.0223** [0.0106]	−0.3109 [0.3551]
Vietnam
FULL	0.0499*** [0.0128]	0.0184*** [0.0021]	0.0003 [0.0002]	0.0204*** [0.0020]	0.0000 [0.0001]	0.0001 [0.0048]
Pre-COVID	0.1137*** [0.0170]	0.0914*** [0.0154]	0.0088 [0.0067]	0.0290*** [0.0047]	0.0127*** [0.0015]	0.0141** [0.0063]
COVID	0.0393*** [0.0081]	0.0508*** [0.0073]	0.0006*** [0.0001]	0.0439*** [0.0022]	0.0028*** [0.0001]	−0.5200*** [0.0075]
RUW	0.0807*** [0.0122]	0.0639*** [0.0098]	−0.0455*** [0.0055]	0.0665*** [0.0119]	−0.0677*** [0.0068]	0.6907*** [0.0551]

Note(s): The figures in each cell are the estimated coefficient with their corresponding standard errors in square brackets. The ^***, ^** and ^* denote statistical significance at 1%, 5% and 10%, respectively

Table 4

Forecast evaluation using RRMSE

Countries	In-sample	Out-of-sample			In-sample	Out-of-sample
		h=15	h=30	h=60		h=15	h=30	h=60
		WTI				Brent
Model-2 relative to Model-1
Brazil	100.0	100.0	80.0	80.0	40.0	40.0	40.0	40.0
China	20.0	20.0	20.0	20.0	60.0	60.0	60.0	20.0
India	80.0	80.0	80.0	80.0	20.0	20.0	20.0	20.0
Indonesia	60.0	60.0	60.0	60.0	60.0	60.0	60.0	60.0
Russia	20.0	40.0	40.0	40.0	80.0	80.0	40.0	40.0
South Africa	60.0	60.0	40.0	40.0	60.0	60.0	40.0	40.0
South Korea	80.0	80.0	80.0	80.0	80.0	80.0	80.0	80.0
Turkey	80.0	80.0	80.0	80.0	100.0	100.0	100.0	80.0
Vietnam	100.0	100.0	100.0	100.0	40.0	40.0	40.0	40.0
Model-3 relative to Model-1
Brazil	100.0	80.0	60.0	60.0	80.0	60.0	60.0	60.0
China	80.0	80.0	80.0	80.0	60.0	80.0	80.0	60.0
India	80.0	80.0	80.0	80.0	60.0	60.0	60.0	40.0
Indonesia	60.0	60.0	60.0	60.0	80.0	80.0	80.0	80.0
Russia	60.0	60.0	80.0	80.0	100.0	100.0	100.0	80.0
South Africa	80.0	80.0	80.0	60.0	60.0	60.0	40.0	20.0
South Korea	100.0	80.0	80.0	80.0	100.0	80.0	80.0	80.0
Turkey	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0
Vietnam	80.0	80.0	80.0	80.0	100.0	80.0	80.0	100.0
Model-3 relative to Model-2
Brazil	60.0	40.0	40.0	40.0	100.0	80.0	80.0	80.0
China	60.0	60.0	60.0	60.0	80.0	100.0	100.0	80.0
India	80.0	80.0	60.0	60.0	40.0	40.0	40.0	40.0
Indonesia	80.0	80.0	80.0	80.0	60.0	60.0	60.0	60.0
Russia	100.0	80.0	80.0	40.0	100.0	80.0	60.0	60.0
South Africa	100.0	100.0	100.0	80.0	80.0	80.0	60.0	60.0
South Korea	80.0	80.0	80.0	80.0	60.0	60.0	60.0	60.0
Turkey	60.0	60.0	60.0	60.0	100.0	100.0	80.0	80.0
Vietnam	60.0	60.0	60.0	60.0	80.0	60.0	60.0	80.0

Note(s): The figures are the percentage of occurrences of a RRMSE value that is greater than unity, when the ratio of the first (unrestricted) model to the second (restricted) model in the pair of contending models labelled on each pane is obtained. The higher the percentage number, the more the preference for the unrestricted model over the restricted model

Table 5

Forecast evaluation using CW test (WTI as proxy for oil uncertainty)

	In-sample				Out-of-sample
					h=15				h=30				h=60				Average
Country	N	SN	P	SP	N	SN	P	SP	N	SN	P	SP	N	SN	P	SP	N	SN	P	SP
Model-2 versus Model-1
Brazil	–	–	60.0	40.0	–	–	40.0	60.0	–	–	60.0	40.0	–	–	60.0	40.0	–	–	53.3	46.7
China	–	–	80.0	20.0	20.0	–	60.0	20.0	20.0	0.0	60.0	20.0	20.0	0.0	60.0	20.0	20.0	0.0	60.0	20.0
India	–	–	20.0	80.0	–	–	20.0	80.0	–	–	20.0	80.0	–	–	20.0	80.0	–	–	20.0	80.0
Indonesia	20.0	20.0	–	60.0	20.0	20.0	0.0	60.0	20.0	20.0	0.0	60.0	20.0	20.0	–	60.0	20.0	20.0	–	60.0
Russia	–	–	80.0	20.0	–	–	100.0	–	–	–	100.0	0.0	–	–	100.0	–	–	–	100.0	–
South Africa	40.0	–	60.0	–	40.0	–	60.0	–	40.0	0.0	60.0	0.0	40.0	–	60.0	–	40.0	–	60.0	–
South Korea	–	–	60.0	40.0	20.0	–	40.0	40.0	20.0	0.0	40.0	40.0	–	–	60.0	40.0	13.3	–	46.7	40.0
Turkey	–	–	20.0	80.0	0.0	–	20.0	80.0	0.0	0.0	20.0	80.0	–	–	20.0	80.0	–	–	20.0	80.0
Vietnam	–	–	60.0	40.0	0.0	–	80.0	20.0	0.0	0.0	60.0	40.0	–	–	80.0	20.0	–	–	73.3	26.7
Model–3 versus Model–1
Brazil	–	–	40.0	60.0	–	–	60.0	40.0	–	–	80.0	20.0	–	–	80.0	20.0	–	–	73.3	26.7
China	20.0	–	40.0	40.0	20.0	–	40.0	40.0	20.0	–	40.0	40.0	20.0	–	60.0	20.0	20.0	0.0	46.7	33.3
India	–	–	40.0	60.0	–	–	40.0	60.0	–	–	40.0	60.0	–	–	40.0	60.0	–	–	40.0	60.0
Indonesia	–	40.0	–	60.0	–	40.0	–	60.0	–	40.0	–	60.0	–	40.0	–	60.0	0.0	40.0	–	60.0
Russia	–	–	80.0	20.0	–	–	100.0	–	–	–	100.0	–	–	–	100.0	–	–	–	100.0	–
South Africa	–	–	100.0	–	–	–	100.0	–	–	–	100.0	–	–	–	100.0	–	–	–	100.0	–
South Korea	–	–	40.0	60.0	–	–	40.0	60.0	–	–	40.0	60.0	–	–	40.0	60.0	–	–	40.0	60.0
Turkey	–	–	–	100.0	–	–	20.0	80.0	–	–	–	100.0	–	–	–	100.0	–	–	6.7	93.3
Vietnam	–	–	40.0	60.0	–	–	40.0	60.0	–	–	40.0	60.0	–	–	40.0	60.0	–	–	40.0	60.0
Model–3 versus Model–2
Brazil	–	–	100.0	–	40.0	–	60.0	–	40.0	–	60.0	–	40.0	–	60.0	–	40.0	–	60.0	–
China	–	–	80.0	20.0	20.0	–	60.0	20.0	20.0	–	60.0	20.0	20.0	–	40.0	40.0	20.0	–	53.3	26.7
India	20.0	–	80.0	–	20.0	–	80.0	–	20.0	–	80.0	–	20.0	–	80.0	–	20.0	–	80.0	–
Indonesia	–	20.0	–	80.0	–	20.0	–	80.0	–	20.0	–	80.0	–	20.0	0.0	80.0	–	20.0	–	80.0
Russia	–	–	40.0	60.0	–	–	80.0	20.0	–	–	80.0	20.0	20.0	0.0	60.0	20.0	6.7	–	73.3	20.0
South Africa	–	–	60.0	40.0	–	–	60.0	40.0	–	–	60.0	40.0	–	–	60.0	40.0	–	–	60.0	40.0
South Korea	–	–	100.0	–	–	–	100.0	0.0	–	–	100.0	–	–	–	80.0	20.0	–	–	93.3	6.7
Turkey	20.0	–	60.0	20.0	20.0	–	60.0	20.0	–	–	80.0	20.0	–	–	80.0	20.0	6.7	–	73.3	20.0
Vietnam	–	–	80.0	20.0	–	–	60.0	40.0	–	–	60.0	40.0	–	–	60.0	40.0	–	–	60.0	40.0

Note(s): This presents a summary of the CW statistics obtained from the different sample intervals considered. Consequently, the figures in each cell are the proportion (%) of observed feats of N (negative but not significant), SN (negative and significant), P (positive but not significant) and SP (positive and significant) CW statistics. The first model in the pair of contending models is adjudged the most preferred if the proportion of the positives surpasses the proportion of negatives CW statistics

Table 6

Economic significance

Model	WTI								Brent
	Returns	Volatility	CER	SP	Returns	Volatility	CER	SP	Returns	Volatility	CER	SP	Returns	Volatility	CER	SP
BRAZIL
Model-1	0.0982	2.5328	−4.8969	−0.0084	0.0983	2.5340	−4.8968	−0.0084	0.0982	2.5328	−4.8969	−0.0084	0.0983	2.5340	−4.8968	−0.0084
Model-2	0.1039	2.6149	−4.8678	−0.0048	0.1040	2.6165	−4.8677	−0.0047	0.1014	2.5687	−4.8916	−0.0064	0.1015	2.5701	−4.8915	−0.0063
Model-3	0.1205	2.5256	−5.0155	0.0056	0.1206	2.5275	−5.0154	0.0057	0.1255	2.6000	−4.8651	0.0086	0.1257	2.6020	−4.8650	0.0087
CHINA
Model-1	0.1050	3.6014	−0.4042	−0.0035	0.1043	3.5949	−0.4049	−0.0039	0.1050	3.6014	−0.4042	−0.0035	0.1043	3.5949	−0.4049	−0.0039
Model-2	0.1287	3.7761	−0.3830	0.0088	0.1278	3.7677	−0.3839	0.0083	0.1364	3.8082	−0.3809	0.0127	0.1355	3.7994	−0.3818	0.0123
Model-3	0.1909	4.3345	−0.3442	0.0381	0.1899	4.3241	−0.3453	0.0376	0.2103	4.4207	−0.3235	0.0469	0.2092	4.4096	−0.3247	0.0465
INDIA
Model-1	0.0739	4.0030	−0.6548	−0.0188	0.0737	4.0014	−0.6550	−0.0190	0.0739	4.0030	−0.6548	−0.0188	0.0737	4.0014	−0.6550	−0.0190
Model-2	0.0915	3.8258	−0.6814	−0.0103	0.0911	3.8230	−0.6817	−0.0105	0.0996	3.9741	−0.6686	−0.0060	0.0992	3.9711	−0.6690	−0.0062
Model-3	0.0592	4.2428	−0.6868	−0.0254	0.0590	4.2412	−0.6870	−0.0255	0.0621	4.2892	−0.7141	−0.0239	0.0617	4.2856	−0.7145	−0.0241
INDONESIA
Model-1	0.0926	3.9009	−0.8471	−0.0096	0.0926	3.9017	−0.8471	−0.0096	0.0926	3.9009	−0.8471	−0.0096	0.0926	3.9017	−0.8471	−0.0096
Model-2	0.0993	3.8681	−0.8400	−0.0063	0.0993	3.8695	−0.8400	−0.0063	0.0963	3.8731	−0.8564	−0.0078	0.0963	3.8747	−0.8564	−0.0078
Model-3	0.0986	3.8473	−0.7229	−0.0067	0.0985	3.8489	−0.7229	−0.0067	0.0942	3.7034	−0.7662	−0.0091	0.0940	3.7035	−0.7663	−0.0092
RUSSIA
Model-1	0.0037	4.8304	−1.2623	−0.0491	0.0035	4.8288	−1.2626	−0.0492	0.0037	4.8304	−1.2623	−0.0491	0.0035	4.8288	−1.2626	−0.0492
Model-2	−0.0108	4.7748	−1.3264	−0.0560	−0.0110	4.7733	−1.3266	−0.0561	−0.0021	4.8478	−1.2696	−0.0516	−0.0023	4.8465	−1.2698	−0.0517
Model-3	−0.0029	4.9472	−1.2573	−0.0515	−0.0032	4.9448	−1.2576	−0.0516	−0.0040	4.9277	−1.2837	−0.0521	−0.0043	4.9254	−1.2840	−0.0522
S_AFRICA
Model-1	0.1351	3.5240	−0.4790	0.0125	0.1354	3.5268	−0.4787	0.0126	0.1351	3.5240	−0.4790	0.0125	0.1354	3.5268	−0.4787	0.0126
Model-2	0.1153	3.4096	−0.5464	0.0020	0.1155	3.4118	−0.5462	0.0021	0.1206	3.5370	−0.5142	0.0048	0.1210	3.5401	−0.5139	0.0050
Model-3	0.1541	3.2386	−0.5544	0.0236	0.1542	3.2399	−0.5543	0.0236	0.1433	3.2680	−0.5603	0.0175	0.1434	3.2696	−0.5602	0.0176
S_KOREA
Model-1	0.1594	4.7148	−1.2531	0.0220	0.1592	4.7129	−1.2533	0.0219	0.1594	4.7148	−1.2531	0.0220	0.1592	4.7129	−1.2533	0.0219
Model-2	0.1571	4.7664	−1.2224	0.0208	0.1569	4.7645	−1.2226	0.0207	0.1619	4.7709	−1.2321	0.0230	0.1617	4.7687	−1.2323	0.0229
Model-3	0.1575	4.6355	−1.2336	0.0213	0.1573	4.6333	−1.2338	0.0212	0.1768	4.5825	−1.2544	0.0305	0.1766	4.5805	−1.2546	0.0304
TURKEY
Model-1	0.0687	1.3292	−1.3057	−0.0372	0.0685	1.3284	−1.3059	−0.0374	0.0687	1.3292	−1.3057	−0.0372	0.0685	1.3284	−1.3059	−0.0374
Model-2	0.0612	1.3235	−1.2980	−0.0438	0.0610	1.3232	−1.2981	−0.0440	0.0703	1.3332	−1.2982	−0.0358	0.0700	1.3322	−1.2984	−0.0360
Model-3	0.0476	1.4636	−1.2552	−0.0529	0.0475	1.4636	−1.2553	−0.0530	0.0450	1.4801	−1.2268	−0.0547	0.0449	1.4797	−1.2269	−0.0549
VIETNAM
Model-1	0.1627	6.0269	−1.3374	0.0208	0.1626	6.0265	−1.3374	0.0208	0.1627	6.0269	−1.3374	0.0208	0.1626	6.0265	−1.3374	0.0208
Model-2	0.1463	6.0827	−1.3740	0.0141	0.1463	6.0827	−1.3740	0.0141	0.1537	6.0287	−1.3603	0.0171	0.1537	6.0287	−1.3603	0.0171
Model-3	0.1466	5.9675	−1.3017	0.0143	0.1464	5.9655	−1.3019	0.0142	0.1229	6.0135	−1.3499	0.0046	0.1228	6.0125	−1.3500	0.0045

Note(s): CER – certainty equivalent returns, SP – Sharpe ratio. Models with the highest returns, CER and SP; and least volatility are adjudged to yield the highest economic gains

Table A1

In-sample predictability (Brent as proxy for oil uncertainty)

Sample period	Model-1	Model-2		Model-3
	Stock market uncertainty	Stock market uncertainty	Oil (Brent) market uncertainty	Stock market uncertainty	Oil (Brent) market uncertainty	Other global uncertainties
Brazil
FULL	0.0005*** [0.0001]	0.0000 [0.0003]	0.0030*** [0.0004]	0.0009* [0.0005]	0.0044*** [0.0006]	0.0964*** [0.0064]
Pre-COVID	0.0071*** [0.0012]	0.0043*** [0.0011]	0.0153*** [0.0018]	0.0009*** [0.0002]	0.0154*** [0.0006]	0.2810*** [0.0063]
COVID	0.0057*** [0.0018]	0.0091* [0.0046]	0.0041*** [0.0010]	−0.0002 [0.0056]	0.0046*** [0.0007]	0.0741 [0.0591]
RUW	0.0829*** [0.0076]	0.1648*** [0.0343]	0.1438*** [0.0139]	0.1310*** [0.0220]	0.0346 [0.0261]	1.9378*** [0.5281]
China
FULL	0.0015 [0.0010]	0.0009 [0.0007]	0.0003*** [0.0001]	0.0014** [0.0006]	0.0003* [0.0002]	0.0394*** [0.0149]
Pre-COVID	0.0017* [0.0009]	0.0113*** [0.0021]	0.0064*** [0.0010]	0.0139*** [0.0015]	0.0058*** [0.0009]	0.2976*** [0.0068]
COVID	−0.0449*** [0.0047]	−0.0603*** [0.0053]	−0.0020*** [0.0001]	−0.0473*** [0.0034]	0.0003 [0.0003]	−0.4668*** [0.0267]
RUW	0.0154 [0.0131]	0.1377*** [0.0162]	−0.0292*** [0.0040]	0.1617*** [0.0353]	−0.0533*** [0.0145]	0.1540 [0.2228]
India
FULL	0.0041*** [0.0008]	0.0059*** [0.0011]	−0.0008** [0.0003]	0.0031*** [0.0005]	−0.0003 [0.0003]	0.0300*** [0.0035]
Pre-COVID	0.0026*** [0.0009]	0.0066*** [0.0015]	−0.0055*** [0.0003]	0.0099*** [0.0015]	−0.0065*** [0.0003]	0.1285*** [0.0126]
COVID	−0.0200 [0.0135]	0.0124 [0.0148]	−0.0015*** [0.0004]	−0.0528*** [0.0094]	−0.0004 [0.0010]	0.3872*** [0.0803]
RUW	−0.0627 [0.0738]	0.0822*** [0.0283]	0.0860*** [0.0055]	−0.0167 [0.0298]	0.0723*** [0.0050]	0.5149*** [0.1454]
Indonesia
FULL	0.5114*** [0.0201]	0.5205*** [0.0206]	0.0005*** [0.0002]	0.5151*** [0.0205]	0.0015*** [0.0003]	0.1983*** [0.0096]
Pre-COVID	0.5338*** [0.0275]	0.5354*** [0.0239]	0.0051*** [0.0011]	0.5383*** [0.0247]	0.0093*** [0.0023]	0.1022*** [0.0221]
COVID	0.4310*** [0.0383]	0.4240*** [0.0392]	0.0005 [0.0003]	0.4177*** [0.0360]	0.0004 [0.0004]	−0.1404** [0.0575]
RUW	0.3158*** [0.0300]	0.3587*** [0.0440]	0.0106 [0.0125]	0.1354*** [0.0421]	−0.0310*** [0.0047]	2.0389*** [0.1765]
Russia
FULL	0.0002 [0.0007]	−0.0004 [0.0011]	0.0010*** [0.0001]	−0.0002 [0.0007]	0.0009*** [0.0002]	0.0956*** [0.0038]
Pre-COVID	−0.0015** [0.0006]	0.0042*** [0.0014]	0.0010 [0.0014]	0.0007 [0.0015]	−0.0018* [0.0010]	0.2155*** [0.0045]
COVID	−0.0207 [0.0142]	−0.0956*** [0.0162]	0.0014*** [0.0002]	−0.0697*** [0.0157]	0.0015*** [0.0001]	−0.0897* [0.0522]
RUW	0.0529*** [0.0076]	0.0563*** [0.0063]	−0.0139*** [0.0024]	0.2312*** [0.0311]	−0.0365** [0.0141]	−5.4211*** [0.4801]
South Africa
FULL	0.0057*** [0.0004]	0.0068*** [0.0006]	0.0010*** [0.0003]	0.0055*** [0.0006]	0.0024*** [0.0004]	0.0564*** [0.0027]
Pre-COVID	0.0037*** [0.0012]	0.0063*** [0.0007]	−0.0035*** [0.0011]	0.0071*** [0.0009]	−0.0079*** [0.0014]	0.0270*** [0.0085]
COVID	0.0448*** [0.0053]	0.0212*** [0.0070]	0.0030*** [0.0011]	0.1046*** [0.0075]	0.0020** [0.0010]	−0.6440*** [0.0331]
RUW	0.0134 [0.0138]	0.0810*** [0.0179]	0.0020 [0.0032]	0.1109*** [0.0316]	−0.0241*** [0.0044]	0.8672*** [0.1227]
South Korea
FULL	0.0233*** [0.0019]	0.0370*** [0.0021]	−0.0024*** [0.0005]	0.0405*** [0.0013]	−0.0021*** [0.0007]	0.0660*** [0.0063]
Pre-COVID	0.0325*** [0.0021]	0.0184*** [0.0038]	0.0125*** [0.0029]	0.0308* [0.0167]	0.0121 [0.0132]	0.0430 [0.1464]
COVID	0.0356*** [0.0016]	0.0568*** [0.0019]	−0.0020*** [0.0002]	0.0569*** [0.0040]	−0.0003 [0.0003]	−0.3741*** [0.0148]
RUW	0.0031 [0.0133]	0.0045 [0.0200]	0.0256** [0.0096]	−0.0297 [0.0225]	0.0059 [0.0066]	0.5391*** [0.1744]
Turkey
FULL	−0.0097*** [0.0005]	−0.0112*** [0.0010]	0.0032*** [0.0004]	−0.0135*** [0.0004]	0.0026*** [0.0001]	0.1656*** [0.0069]
Pre-COVID	−0.0057** [0.0028]	−0.0049*** [0.0002]	0.0056*** [0.0004]	−0.0020** [0.0009]	0.0083*** [0.0005]	−0.2556*** [0.0111]
COVID	−0.0070 [0.0069]	−0.0240*** [0.0088]	0.0018*** [0.0004]	−0.0435*** [0.0045]	0.0016*** [0.0002]	0.3464*** [0.0293]
RUW	0.1038*** [0.0111]	0.1340*** [0.0119]	0.0235*** [0.0042]	0.1360*** [0.0260]	0.0193** [0.0085]	−0.7501*** [0.1843]
Vietnam
FULL	0.0499*** [0.0128]	0.0493*** [0.0141]	0.0000 [0.0003]	0.0250*** [0.0033]	0.0003*** [0.0001]	0.0083* [0.0050]
Pre-COVID	0.1137*** [0.0170]	0.0914*** [0.0154]	0.0088 [0.0067]	0.0290*** [0.0047]	0.0127*** [0.0015]	0.0141** [0.0063]
COVID	0.0393*** [0.0081]	0.0508*** [0.0065]	0.0001*** [0.0000]	0.0381*** [0.0022]	0.0035*** [0.0004]	−0.5115*** [0.0331]
RUW	0.0807*** [0.0122]	0.0908*** [0.0107]	−0.0677*** [0.0032]	0.0912*** [0.0125]	−0.0730*** [0.0019]	0.6002*** [0.0427]

Note(s): The figures in each cell are the estimated coefficient with their corresponding standard errors in square brackets. The ^***, ^** and ^* denote statistical significance at 1%, 5% and 10%, respectively

Table A2

Forecast evaluation using CW test (Brent as proxy for oil uncertainty)

	In-sample				Out-of-sample
					h=15				h=30				h=60				Average
Country	N	SN	P	SP	N	SN	P	SP	N	SN	P	SP	N	SN	P	SP	N	SN	P	SP
Model-2 versus Model-1
Brazil	–	–	100.0	–	–	–	100.0	–	–	–	100.0	–	–	–	100.0	–	–	–	100.0	–
China	–	–	80.0	20.0	20.0	–	60.0	20.0	20.0	–	60.0	20.0	20.0	–	60.0	20.0	20.0	–	60.0	20.0
India	–	–	80.0	20.0	–	–	80.0	20.0	–	–	80.0	20.0	–	–	80.0	20.0	–	–	80.0	20.0
Indonesia	20.0	40.0	–	40.0	20.0	40.0	–	40.0	20.0	40.0	–	40.0	20.0	40.0	–	40.0	20.0	40.0	–	40.0
Russia	–	–	80.0	20.0	–	–	100.0	–	–	–	100.0	–	–	–	100.0	–	–	–	100.0	–
South Africa	40.0	–	60.0	–	40.0	–	60.0	–	40.0	–	60.0	–	40.0	–	60.0	–	40.0	–	60.0	–
South Korea	–	–	80.0	20.0	20.0	–	60.0	20.0	20.0	–	60.0	20.0	–	–	80.0	20.0	13.3	–	66.7	20.0
Turkey	–	–	20.0	80.0	–	–	–	100.0	–	–	–	100.0	–	–	20.0	80.0	–	–	6.7	93.3
Vietnam	20.0	–	60.0	20.0	60.0	–	40.0	–	60.0	–	40.0	–	20.0	–	80.0	–	46.7	–	53.3	–
Model–3 versus Model–1
Brazil	–	–	60.0	40.0	–	–	100.0	–	–	–	100.0	–	–	–	80.0	20.0	–	–	93.3	6.7
China	–	–	60.0	40.0	20.0	–	40.0	40.0	20.0	–	40.0	40.0	20.0	–	60.0	20.0	20.0	–	46.7	33.3
India	–	–	80.0	20.0	–	–	80.0	20.0	–	–	60.0	40.0	–	–	60.0	40.0	–	–	66.7	33.3
Indonesia	–	40.0	–	60.0	–	40.0	–	60.0	–	40.0	–	60.0	–	40.0	–	60.0	–	40.0	–	60.0
Russia	–	–	80.0	20.0	–	–	100.0	–	–	–	100.0	–	–	–	100.0	–	–	–	100.0	–
South Africa	–	–	100.0	–	–	–	100.0	–	–	–	100.0	–	–	–	100.0	–	–	–	100.0	–
South Korea	–	–	80.0	20.0	–	–	80.0	20.0	–	–	80.0	20.0	–	–	80.0	20.0	–	–	80.0	20.0
Turkey	–	–	20.0	80.0	–	–	20.0	80.0	–	–	20.0	80.0	–	–	20.0	80.0	–	–	20.0	80.0
Vietnam	–	–	40.0	60.0	–	–	40.0	60.0	–	–	40.0	60.0	–	–	40.0	60.0	–	–	40.0	60.0
Model–3 versus Model–2
Brazil	–	–	100.0	–	–	–	100.0	–	–	–	80.0	20.0	–	–	80.0	20.0	–	–	86.7	13.3
China	–	–	80.0	20.0	–	–	80.0	20.0	–	–	80.0	20.0	–	–	100.0	–	–	–	86.7	13.3
India	20.0	–	60.0	20.0	20.0	–	60.0	20.0	20.0	–	60.0	20.0	20.0	–	60.0	20.0	20.0	–	60.0	20.0
Indonesia	–	–	20.0	80.0	–	–	20.0	80.0	–	–	20.0	80.0	–	–	20.0	80.0	–	–	20.0	80.0
Russia	–	–	80.0	20.0	–	–	100.0	–	–	–	60.0	40.0	–	–	60.0	40.0	–	–	73.3	26.7
South Africa	–	–	80.0	20.0	–	–	80.0	20.0	–	–	80.0	20.0	20.0	–	60.0	20.0	6.7	–	73.3	20.0
South Korea	–	–	80.0	20.0	–	–	80.0	20.0	–	–	80.0	20.0	–	–	80.0	20.0	–	–	80.0	20.0
Turkey	–	–	80.0	20.0	–	–	80.0	20.0	–	–	80.0	20.0	–	–	80.0	20.0	–	–	80.0	20.0
Vietnam	–	–	100.0	–	–	–	80.0	20.0	–	–	80.0	20.0	–	–	80.0	20.0	–	–	80.0	20.0

Note(s): This presents a summary of the CW statistics obtained from the different sample intervals considered. Consequently, the figures in each cell are the proportion (%) of observed feats of N (negative but not significant), SN (negative and significant), P (positive but not significant) and SP (positive and significant) CW statistics. The first model in the pair of contending models is adjudged the most preferred if the proportion of the positives surpasses the proportion of negatives CW statistics

Notes

1. see https://www.investopedia.com/articles/investing/081115/green-bonds-benefits-and-risks.asp

2. https://www.imf.org/en/News/Articles/2022/04/26/sp-042622-shaping-the-frontier-of-sustainable-finance-in-emerging-markets