COVID-19 and Insurance market returns in emerging and developed markets: A comparative study based on Wavelet methods

This study investigates the co-movement nexus between COVID-19 and insurance industry returns for emerging and developed markets using a wavelet-based framework. Analysis on the daily observations from 22 nd January 2020 to 14 th September 2020 reveals that insurance returns ( INS ) responded strongly and negatively, right after the onset of the global COVID-19 outbreak, but asymmetrically later. Additionally, the devastation brought to INS is comparatively more severe but short-lived for emerging markets. The wavelet-based Granger causality and correlation confirm the robustness of our results. These findings are important for policymakers and investors in the insurance industry in the aftermath of COVID-19.


Introduction
Recently, we are passing through an unprecedented time in human history.Despite all the technological and economic progress we made, and the abundance of resources we have, everyone felt helpless at least for some period of time, in front of nature during the COVID-19 outbreak (Goodell, 2020).The global number of confirmed cases and deaths has surpassed 30 million and 1 million (as of 24 th September 2020), respectively, according to the World Health Organization (WHO, 2020).The devastation brought by the COVID-19 has reminded us about the dire challenges of pandemics that we may have to face in the future also.While it may be impossible to avoid these kinds of risks to human lives in entirety, there are ways to mitigate the shattering effects of such events.The insurance industry is the prime channel for hedging such risks in today's world (Chen et al., 2020).Extensive research is being carried out on multiple (financial and nonfinancial) aspects of the COVID-19 pandemic.For example, several studies have discussed the stock markets' reaction to the COVID-19 epidemic (Akhtaruzzaman et al., 2020;Azimli, 2020;Baek et al., 2020;Cepoi, 2020;Ciner, 2020;Erdem, 2020;Li et al., 2020;Mazur et al., 2020;Okorie & Lin, 2020;Topcu & Gulal, 2020;Zaremba et al., 2020;Zhang et al., 2020).Others have explored the impact of this shock on digital assets and commodity markets (Conlon & McGee, 2020;Corbet et al., 2020;Goodell & Goutte, 2020;Ji et al., 2020;Mensi et al., 2020;Mnif et al., 2020).However, the insurance industry's response to the pandemic still needs vivid exploration.
Figure 1 shows the time trend of new COVID-19 infections worldwide.

here>>
Because of heterogeneous market characteristics and varying business models with diverse customer bases in emerging and advanced countries (Mnif et al., 2020), insurance industries of both types of economies are expected to behave differently in response to the COVID-19 pandemic.The Swiss-Re reports that total premium income in developed markets (life and nonlife insurance) will fall by 4% due to COVID-19 this year, and is expected to recover with a positive growth of 2% in 2021.In contrast, the insurance premium growth in emerging markets will remain positive, up 1% and 7% in 2020and 2021, respectively (Swiss Re, 2020).Figure 2 compares the insurance premium (life and non-life) growth with respect to GDP growth in emerging and developed economies.Figure 3  The insurance industry could be affected by the COVID-19 in two broad categories; premiums and claims.While it is clear that the claims might have skyrocketed due to increased unemployment (payments to unemployment insurance buyers), hospitalizations (increased payments in case of health insurance policyholders), deaths (huge lump sum and annuity payments for life insurance), and business closures (payments against natural disasters), the impact on the premium is again expected to be twofold.First, there may be a huge loss of premium income due to reduced family, health, and life insurance sales resulting from wide-scale lockdowns with no direct interaction between salespersons and potential insurance policy buyers.Second, online insurance sales might boom due to the increased health and life risks brought by the pandemic.
People might want to hedge their huge payment risks due to potential illnesses caused by the epidemic or overall deteriorating health scenario.Correspondingly, increased unemploymentinsurance sales, and corporate-insurance sales may increase the premium income of the insurance companies.
The elasticity of corporate insurance demand after a catastrophic incident can explain if the insurance industry returns will be affected positively or negatively.One of the very few empirical papers on this important topic uses the data on 1800 large U.S. corporations and concludes that the insurance industry gests a boom in premium income after catastrophic incidents (Michel-Kerjan et al., 2015).Another study using 43 large catastrophic-insured incidents since 1970, finds a significant increase in industry revenues and stock returns of insurance brokers, right after such incidents (Ragin & Halek, 2016).The COVID-19 has had a severe negative impact on the Chinese insurance industry and caused a reduction in all kinds of inflows, including both commercial and individual premiums.The growth rate of gross premium fell by 9.53% as compared with 2019 (Wang et al., 2020).By performing a qualitative study on the opinions of industry experts and professionals in Ethiopia, Worku and Mersha (2020) find that COVID-19 had a catastrophic effect on the Ethiopian insurance industry.However, an interesting question has emerged in the context of COVID-19.Whether an event like this can be insured or not (Goodell, 2020)?
The possibility of positive changes in insurance returns after a catastrophic incident as supported by the literature, and the recent negative effects of COVID-19 shock for this industry in few countries as evidenced by few studies, calls for a fresh and definitive inquiry into this gravely important area.As it's difficult to collect the data on premiums collected, and claims handled for a vast majority of firms from all the emerging and advanced economies during COVID-19, we can gauge the overall impact from the changes in the daily market prices/returns of publicly traded insurance companies.More specifically, we use the daily insurance returns indices for advanced and emerging markets, representing the broad market performance of insurance industry stocks.

Data description and sources
The COVID-19 outbreak is represented by "daily new infections worldwide", and data is sourced from the Bing COVID-19 Tracker1 .Insurance market returns in developed and emerging economies are represented by the "Nasdaq developed markets insurance index" (INS-DM)2 and "Nasdaq emerging markets insurance index" (INS-EM)3 , respectively, and collected from the Nasdaq official website.The sample period ranges from 22 nd January 2020 to 14 th September 2020.
The INS total returns index is derived from the following Equation (1).

PRI IDP TRI TRI PRI
Where, t TRI is the value of "total return index" on the current day, and 1 t TRI − shows its closing value on the previous day.

Methodology
To study the asymmetric covariance nexus between COVID-19 and insurance market returns in a different time and frequency domains, we have employed the novel methods of time series analysis, such as "continuous wavelet transforms" (CWT) and "wavelet coherence" (WTC).
Further, we check the robustness through "wavelet based Granger causality and correlation".The wavelet analysis has been frequently used in contemporary economics, finance, and environmental studies (Al-Yahyaee et al., 2019;Fareed et al., 2020;Goodell & Goutte, 2020).This technique does not have a prerequisite for the data to be stationary (Ng & Chan, 2012).The other advantages of using it include extraction of associations in the short-run, medium-term, and long run, and at different frequencies and time scales in the same estimation (Iqbal et al., 2020).

Continuous wavelet transforms (CWT)
To decompose the data series into time-frequency space, we employ the continuous wavelet transform.Such a decomposition helps to get the information from the local neighborhood by using the 'Morlet Wavelet' function.The main purpose of using wavelet transformation is to decompose and then rebuild the function Where n=1,…..N, and s represents the set of scale and t δ denotes the time step (Ng & Chan,   2012).

Wavelet Transform Coherence (WTC)
The wavelet coherence provides the localized correlation coefficient of two time series X= [Xn] and Y= [Yn] in the time-frequency domain.The WTC is calculated as the absolute square value of smoothed "cross-wavelet spectra" for both time series (X and Y) divided by the product of the smoothed individual "wavelet power spectra" of X and Y.
Where S indicates the smoothing parameter, and when there is no smoothing, the WTC will be equal to 1.Moreover, the coefficient of squared wavelet coherence satisfies the inequality i.e. 0 ≤ R 2 (r, s) ≤ 1.A value close to 1 means strong correlation, while a value close to zero shows weak correlation.Wavelet coherence between all possible combinations of our pairs of variables for both developed and emerging markets is given as under; R means coherence, while R2 specifies the squared coherence.Monte Carlo simulation is used to check the level of significance in WTC (Grinsted et al., 2004).rejecting the null hypothesis of normality and symmetry.Therefore, the wavelet technique is an appropriate selection that explores the co-movements over time and frequency domains instead of evaluating the average statistical relationship between the series (Grinsted et al., 2004).

Results and discussion
<<Insert Table 1 here>> In Figure 5(a) there are three significant and noticeable red contours.The big one contains several arrows [ indicating that COV19 is leading insurance returns in emerging markets with an outphase coherence in the short-run (0-4) and medium-term (4-8) during April and May.This finding is in line with the recent studies concluding a significant negative impact of COVID-19 shock on the insurance industry of China, Turkey, Ethiopia, and Ghana (Babuna et al., 2020;Öztürk et al., 2020;Wang et al., 2020;Worku & Mersha, 2020).A tiny red contour in the 0-4 frequency band, during mid of April, shows few arrows Z that mean COV19 is leading insurance returns with inphase coherence.This may be due to the dominant and positive effect of Chinese recovery from COVID-19 in April, which is the biggest emerging market.This result supports the literature citing a quick rebound in the market prices after a swift decline due to catastrophic incidents (Brounen & Derwall, 2010).Finally, a little anomaly can be seen in the third red contour in the short-run during mid-September, where a few arrows ] indicate that COV19 is lagging insurance returns with in-phase coherence.Actually, this is the time when most of the countries ease restrictions on the limited movement of goods and people, which in turn promotes the business and finance activities, including insurance.
Similarly, in Figure 5(b), there are four noticeable red contours.The one with prominent size contains several arrows [ indicating that COV19 is leading insurance returns with significant out-phase coherence in the short-run (4-8) during mid-April to mid-May for the developed markets.The negative response of insurance returns to COVID-19 shock in developed markets is similar to the findings of other studies reporting a high financial contagion and significant insurance returns decline in response to the crisis events; like the terrorist attacks of 9/11 and the global financial crisis of 2007-08 (Baluch et al., 2011;Cummins & Weiss, 2009;Marović et al., 2010;Ramiah et al., 2010).However, our results do not support the idea of a low correlation of insurance stock returns with the market portfolio after a financial crisis, as reported by Thomann (2013).Three small red contours exist in the short-run (0-4) during mid-February, and mid-April to mid-May, showing a few arrows ]^ indicating a mixed trend.
The swift negative response right after the COVID-19 is also in line with the literature citing high contagion in both emerging and developed markets following a financial crisis event (Aloui et al., 2011).In a nutshell, the impact of COVID-19 on insurance returns is more severe but short-lived for emerging markets.The findings are in line with Marović et al. (2010), suggesting the negative impact of financial disaster on the insurance industry.Moreover, emerging markets have comparatively limited resources to deal with the severe impact of the pandemic and therefore are expected to suffer the worst.Additionally, in developed markets, the low-interest rates throughout the world during the last two decades have been putting a negative strain on the market returns of insurance companies which rely heavily on fixed income markets for investments (Killins & Chen, 2020).Therefore, the recent decline in insurance returns should be of great concern as the interest rates are still persistently low.

4.1.Robustness check
We further apply wavelet-based correlation and Granger causality tests for the robustness check.
In this regard, we first decompose all the raw series into different frequencies (D1, D2,…., D6) and a smoothed variable (S6) by employing MODWT based "Least Asymmetric (LA) Wavelet Filter" (Daubechies, 1992).The MODWT follows the Multi-resolution analysis (MRA) of the J=6 pattern for all time series.Figure 6 shows the MODWT decomposition of variables on J = 6 wavelet levels.
<<Insert Figure 6 here>> After decomposing the variables, we employ the wavelet correlation.<<Insert Figure 7 here>> Table 2 reports the results of wavelet-based granger causality for emerging and developed markets.
COV19 Granger causes insurance returns in the frequency ranges of d2 (short-run) and d3 (medium-run) for emerging markets.In contrast, COV19 Granger causes INS(DM) in the shortrun (D1), mid-term (D2), and long-run (D6) for developed markets.These findings are robust to the results of wavelet coherence (WTC).The insignificance of results for other frequency ranges shows the asymmetry in associations between our variables at different scales.

Concluding remarks
In the short run and mid-term, COVID-19 caused a decrease in the insurance industry returns of both the emerging and advanced economies.Although both types of economies show some kind of rebound in returns recently, the overall drop in emerging markets is more significant than the developed markets, due to the strong growth prevailing in emerging markets before COVID-19.
Further, the impact of COVID-19 is found to be asymmetric on insurance returns in both emerging and advanced economies in varying time-frequency combinations.Immediate policy measures should be taken to support the insurance industry's potential decline, which is already bearing low profitability caused by the persistently low-interest rates in developed markets recently.Our results represent the scenario from COVID-19 until now (September 14, 2020), and the long-term impact will be unraveled in future studies with the passage of time.

Disclosure statement
No potential conflict of interest was reported by the author(s).Life and non-life insurance premium growth with respect to GDP growth in emerging and developed markets.Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries.This map has been provided by the authors.
illustrates the insurance market returns indices in emerging and developed economies.
− represents the value of the "price return index" on present day and the previous day (closing value), respectively.Finally, t IDP indicates the "index dividend points".
L ∈ R .The CWT can be expressed mathematically as under;

Figure 4 Figure 5
Figure4shows the CWT of all the variables by decomposing the series into time and frequency Figures 7(a) and 7(b) show the wavelet correlations between COV19 and INS for emerging and developed markets, respectively.We observe a negative/positive correlation between COV19 and INS at different timefrequency bands for emerging and developed markets.However, we find a bit stronger wavelet correlation in emerging markets.

Figure 3 .Figure
Figure 3. Insurance market returns in emerging and developed markets

Figure
Figure 6 MODWT decomposition of variables on J = 6 wavelet levels Figures

Figure 5 Wavelet
Figure 5 Figure 6 Where COV19 is the daily new cases worldwide, INS_EM and INS_DM are the Nasdaq insurance market indices for emerging and developed markets, respectively.S is the smoothing parameter.
Table 1 presents the summary statistics of our variables.It is clearly evident that INS(EM), INS(DM), and COV19 exhibit skewing and heavy-tailed distributions.The Shapiro-Wilk and Jarque-Bera statistics illustrate that the series have leptokurtic and asymmetric distributions,