# I. INTRODUCTION

In this paper, we examine the potential of the Asia-Pacific Islamic stock market to serve as a good hedge against uncertainty due to pandemics and epidemics (UPE). The study’s motivation has roots in the extant literature, which suggests a strong connection between the COVID-19 pandemic and conventional stocks (see, for example, Al-Awadhi et al., 2020; He et al., 2020; Phan & Narayan, 2020; Salisu & Akanni, 2020). Yet, the responsiveness of Islamic stock markets to COVID-19 seems relatively understudied (Ashraf et al., 2020). This is surprising given the vast literature on Islamic stock markets and the relevance of Islamic stocks in active portfolios (see Narayan & Phan, 2019). The resilience of Islamic stock markets during the global financial crisis is another motivation for this study; see Masih et al. (2018) for a review. Inspired by these studies, we hypothesize that the Asia-Pacific Islamic stock market is likely to serve as a good hedge[1] against UPE. We explore the link between the two variables by using the recent datasets on UPE developed by Baker et al. (2020).[2] While the interest is on the Asia-Pacific Islamic stock market, for completeness, the empirical analysis is also replicated for the conventional market. Additional analysis is also carried out to establish the predictive content of UPE for out-of-sample predictability of stock returns. Thus, the study’s findings have implications for portfolio diversification strategies, particularly for investors seeking to maximize returns in the Asia-Pacific region.

Using a predictability model suitable for the variables of interest (see Westerlund & Narayan, 2012, 2015), we find strong evidence in support of the hedging potential of the Asia-Pacific Islamic stocks during pandemics and epidemics compared to conventional stocks. The results supporting this conclusion pass a battery of tests for predictability analysis. These findings make three contributions to the literature. First, it extends the literature on the response of financial markets to the COVID-19 pandemic to capture the Islamic stock markets, which surprisingly seems to have received limited attention to-date. The second contribution relates to the assessment of the hedging potential of Islamic stocks to UPE. Following our study, hedging potential during COVID-19 can be tested in other markets using different assets. Finally, we test whether exploiting the predictive contents of UPE can enhance the out-of-sample forecasts of stock returns. The rest of the paper is structured as follows. Section II discusses data and methodology. Results and discussions appear in Section III. The final section concludes.

# II. DATA AND METHODOLOGY

We utilize daily data for the predictor series based on the recently developed measure of UPE by Baker et al. (2020). Data are obtained from the FRED database of the Federal Reserve Bank of St. Louis. For the predicted series, we utilize daily data of two composite stock indices: (i) the conventional stock price index proxied by the Standard and Poor’s (S&P) Dow Jones Composite Average stock index (DJCA); and (ii) the Asia-Pacific Islamic stock price index proxied by the Dow Jones Islamic Market Asia/Pacific index (DJIM). The DJIM index measures the performance of equity stocks traded in 15 countries from the Asia/Pacific region. Daily data on the two predicted series are obtained from the Bloomberg terminal. The full data scope ranges from 8/31/2010 (post global financial crisis period) to 9/15/2020. From this sample, we draw the COVID-19 period. The analyses are conducted using both the pre-COVID-19 sample (8/31/2010 to 12/31/2019) and the sample covering the COVID-19 pandemic (01/01/2020 to 9/15/2020). We also control for an additional predictor, oil price (using the West Texas Intermediate crude oil price[3]), given evidence of its strong connection with the stock market (Salisu et al., 2019, among others; see Smyth & Narayan, 2018).

In line with the risk-return hypothesis, we construct a predictive model that quantifies the response of the Asia-Pacific Islamic stock market as well as the conventional market to UPE. During turbulence, a number of studies have established a negative correlation between risk and market returns for conventional stocks (see Al-Awadhi et al., 2020; He et al., 2020; Phan & Narayan, 2020; Salisu & Akanni, 2020) and a positive correlation for Islamic stocks (see Masih et al., 2018). To circumvent any potential endogeneity bias and persistence effect, the predictive model takes the form (see Westerlund & Narayan, 2012, 2015):

$r_{t}=\alpha+\sum_{i=1}^{5} \lambda_{i}^{a d j} U P E_{t-i}+\gamma\left(U P E_{t}-\rho_{0} U P E_{t-1}\right)+\eta_{t}\hspace{10mm}(1)$
where $$r_{t}$$ is the log return of stocks price index (of either Islamic or conventional market); UPE has been previously defined; $$\eta_{t}$$ is zero mean idiosyncratic error term; the coefficient $$\lambda_{i}^{a d j}$$ measures the response of stock returns to UPE; and we allow a maximum of five lags given the 5-day daily data frequency. Five lags also allow us to capture more dynamics in the estimation process. Thus, the hedging potential of a market against UPE is evaluated using the Wald test for joint significance. The considered market is likely to at least retain the value of its returns, on the average, in the face of UPE, if $$\sum_{j=1}^{5} \lambda_{i}^{a d j} \geq 0$$; otherwise, it is more likely to be vulnerable. All the variables are pre-weighted with $$1 / \hat{\sigma}_{\eta}$$ to account for conditional heteroscedasticity effect and we refer readers to Westerlund & Narayan (2012, 2015) on the estimation procedure for Equation (1).

# III. MAIN FINDINGS

We begin the analysis with basic preliminary statistics and the results are presented in Table 1. We see higher returns and lower volatility in Islamic stocks relative to the conventional stocks during the COVID-19 pandemic while both index returns are stationary. The presence of salient features in the data, such as persistence, endogeneity bias, and conditional heteroscedasticity as depicted in Table 1 further validates the choice of our predictability approach and model. In the main analysis, we begin interpretation with the predictability results which show the response of both Islamic and conventional stock markets of the Asia-Pacific region to UPE (see Panel A of Table 2). We find a significant positive relationship between Islamic stock returns and UPE during the pre-COVID-19 period implying that higher returns can be achieved with higher incidence of pandemics and epidemics. However, the coefficient is not significant for the conventional stock market during the same period and the outcome remains the same even after controlling for oil price. The outcome further confirms the evidence in the literature about the resilience of the Islamic stock market to crises relative to the conventional stock market. Nonetheless, the hedging effectiveness of the Islamic stock market declines during the COVID-19 pandemic given the positive but insignificant coefficient of the UPE index while the sign is negative for the conventional stock market. Overall, on average, we find evidence of higher hedging potential for the Asia-Pacific Islamic stocks against UPE relative to the conventional market.

Table 1:Preliminary results
Statistics Pre-COVID sample (8/31/2010 to 12/31/2019) COVID-19 sample (01/01/2020 to 9/15/2020)
Islamic stocks Conventional stocks UPE Islamic stocks Conventional stocks UPE
Mean 0.022 0.042 0.386 0.082 -0.009 20.956
Std. Dev. 0.856 0.851 0.837 1.392 2.664 15.000
ADF -31.447a*** -50.774a*** -13.079a*** -11.7120a*** -7.359a*** -11.742b***
GARCH -44.169a*** -46.956a*** -55.871a*** -11.009a*** -12.782a*** -22.887b***
ARCH(2) 48.582*** 212.141*** 31.197*** 14.658*** 22.937*** 4.707**
ARCH(4) 40.704*** 119.030*** 27.079*** 7.603*** 13.067*** 2.367*
ARCH(6) 29.356*** 83.602*** 25.157*** 5.641*** 12.355*** 1.820*
Q-stat (2) 8.3683** 0.2196 31.551*** 4.467 11.571*** 7.993**
Q-stat (4) 20.244*** 5.9172 104.13*** 6.762 14.157*** 10.109**
Q-stat (6) 22.676*** 19.272*** 188.54*** 8.950 20.101*** 13.381**
Q2-stat (2) 107.32*** 440.60*** 68.541*** 32.881*** 44.548*** 10.067**
Q2-stat (4) 219.17*** 671.60*** 140.93*** 46.850*** 61.188*** 10.232**
Q2-stat (6) 276.68*** 842.98*** 225.76*** 62.724*** 85.091*** 11.469*
Persistence - - 0.610*** - - 0.903***
Endogeneity -0.002 (0.023) 0.003 (0.023) - -0.026** (0.013) -0.014 (0.025 ) -
Nobs 2366 2366 2366 178 178 178

Notes: Symbols ***, ** & * denote statistical significance at 1%, 5% & 10% levels, respectively; superscripts a and b indicate that the variable is stationary in levels and first differences, respectively. The GARCH-based unit root test is the one proposed by Narayan and Liu (2015) and it is considered an alternative to the Narayan & Popp (2010) test due to the data frequency used in this study (see also Salisu & Adeleke, 2016). Both unit root tests are conducted with a constant and a time trend. Details about the persistence and endogeneity tests are contained in Westerlund & Narayan (2012, 2015).

Additional analysis is rendered to evaluate the out-of-sample predictive content of the UPE index and this is done for multiple out-of-sample forecast horizons, namely a 10-day, a 20-day and a 30-day ahead forecast horizon using a 75%/25% data split. We use the historical average (constant returns) model that ignores the UPE index as the baseline model, and its relative performance with the UPE-based models (“Without control” and “With control”) is evaluated using the Clark & West (2007) test. The results are presented in Table 2 (Panel B for the in-sample forecasts and Panels C & D for the out-of-sample forecasts). We find that exploiting the information contained in the UPE index as well as oil price can improve the out-of-sample predictability of stock returns. In this paper we do not test the economic significance of these forecasts. We set this aside for future research.

# IV. CONCLUSION

In this study, we examine the resilience of the Asia-Pacific Islamic stock market as well as the conventional stock market to uncertainty due to pandemics and epidemics. Using a predictability technique that accommodates the salient features of the data, we find evidence in favour of the hedging potential of the Asia-Pacific Islamic stocks during pandemics and epidemics albeit with lower hedging effectiveness during the COVID-19 pandemic. Further results show the significance of exploiting the information about UPE as well as other global factors, such as the oil price, in the out-of-sample predictability of stock returns.

Table 2:Predictability and forecast evaluation results
Panels A: Predictability results
Pre-COVID sample (8/31/2010 to 12/31/2019) COVID-19 sample (01/01/2020 to 9/15/2020)
Islamic stocks Conventional stocks Islamic stocks Conventional stocks
Without control 0.3760*** (0.1276) [2.9455] 0.0884 (0.1134) [0.7793] 0.0207 (0.0478) [0.4328] -0.0177 (0.0631) [-0.2796]
With control 0.3273*** (0.1114) [2.9385] 0.1069 (0.1111) [0.9624] 0.0036 (0.0148) [0.2421] 0.0022 (0.0197) [0.1124]
Nobs 2366 2366 178 178
Panel B: In-Sample forecast evaluation
Pre-COVID sample (8/31/2010 to 12/31/2019) COVID-19 sample (01/01/2020 to 9/15/2020)
Islamic stocks Conventional stocks Islamic stocks Conventional stocks
Without control 1.1844 0.5785 2.1739*** 1.2422
With control 2.8730*** 3.2848*** 2.3774*** 2.3384***
Nobs 1775 1775 140 140
Panel C: Out-of-Sample forecast evaluation [Pre-COVID sample (8/31/2010 to 12/31/2019)]
Islamic stocks Conventional stocks
h=10 h=20 h=30 h=10 h=20 h=30
Without control 1.2089 1.2131 1.2170 0.5390 0.4698 0.4414
With control 2.8929*** 2.8922*** 2.9057*** 3.2675*** 3.2375*** 3.2305***
Nobs 1785 1795 1805 1785 1795 1805
Panel D: Out-of-Sample forecast evaluation [COVID-19 sample (01/01/2020 to 9/15/2020)]
Islamic stocks Conventional stocks
h=5 h=10 h=20 h=5 h=10 h=20
Without control 2.1228*** 2.0375*** 2.1386*** 1.1527 1.0643 1.0328
With control 2.3286*** 2.2346*** 2.4255*** 2.2738*** 2.2855*** 2.2912***
Nobs 150 160 170 150 160 170

Notes: “Without control” implies a predictive model that does not control for oil price return while “With control” is an extended predictive model that accounts for the oil price variable. The reported statistics in Panel A are obtained from the joint significance test of five lags of the UPE index. For the forecast evaluation, the performance of the UPE-based models [“With control” & “Without control”] is compared with the historical average that ignores the UPE index. Nobs denotes number of observations. The asterisks ***, ** & * imply statistical significance at the 1%, 5% & 10% levels, respectively. Values in parentheses represent standard errors while those reported in square bracketsare the t-statistics. For the Clark & West test, the null hypothesis of a zero coefficient is rejected if this statistic is greater than +1.282 (for a one sided 0.10 test), +1.645 (for a one sided 0.05 test), and +2.00 for 0.01 test (for a one sided 0.01 test) (see Clark & West, 2007).

# Acknowledgement:

I would like to thank the Editor and the anonymous referee for many helpful comments. However, the usual disclaimer applies.

1. The intuition behind the use of hedging here is drawn from the inflation hedging literature, where an asset that serves as a good hedge against inflation is expected to retain or increase in value as inflation increases (see Arnold & Auer, 2015). However, the interpretation here slightly differs from the conventional inflation hedging since the UPE indicator cannot play the role of inflation in terms of the relationship between real and nominal variables. Thus, in this paper, there are three possible outcomes in relation to UPE. Assume the coefficient on UPE is “b”, then the three possible outcomes are: (i) b>0 (high hedge, such that the return increases with UPE); (ii) b<0 (worst hedge, where return reduces with UPE); and (iii) b=0 (low hedge, where return remains unchanged with UPE).

2. Using the same measure, Salisu & Adediran (2020) find a strong connection between energy risk and pandemics. An alternative index by Salisu & Akanni (2020) confirms the vulnerability of conventional stocks to the COVID-19 pandemic.

3. This is considered a good reflector of movements in global oil prices (see Narayan & Gupta, 2015).