How Integrated are Regional Green Equity Markets? Evidence from a Cross-Quantilogram Approach

This paper investigates the integration among sub-sectors within the environmentally friendly stock market and the integration between these sub-sectors and other financial asset classes. Using the recently developed cross-quantilogram framework, we contribute to the literature by quantifying the cross-quantile directional spillovers among regional green equity markets and other financial assets. First, we find that within the green equity market, the U.S. sector is the main transmitter of shocks while the Asian sector is the main receiver of shocks, however, the integration among regional green equity markets dissipates in the long run. Second, we find that the relationship between green equity markets and other financial assets such as energy commodity and conventional stock varies across regions and market conditions. Our results imply that understanding the heterogeneity in the internal and external integration of the green equity market is crucial for the design of successful investment strategies and effective policy incentives to promote environmentally friendly investments.


Introduction
The transition towards a low-carbon economy requires a substantial level of financial flows to environmentally friendly sectors. While policymakers and investors around the world are increasingly recognizing financial instruments that support environmentally friendly investments as important drivers of sustainable growth, financial flows towards low-carbon economic activity are not growing at the required speed. 1 To successfully direct further funding towards low-carbon economic activity, it is crucial to provide investors and policymakers with a detailed picture of the financial markets for green or environmentally friendly investments. Yet, in the literature, the green financial market has been analyzed from an aggregate perspective, which overlooks the heterogeneous response of green financial instruments to other asset classes and the internal interaction among sub-sectors of the green financial market. This paper aims at investigating the integration of various regional green equity markets among themselves and with other segments of the global financial markets.
To this end, we ask the following questions. First, how integrated are regional green equity markets under normal and extreme market conditions? Second, how do the relationships between green equity and other asset classes vary across regions and market conditions? Third, what are the implications of these regional variations for investors and policymakers?
To answer the above questions, we employ the cross-quantilogram framework of Han et al. (2016) on daily time series data of stock indexes of regional green equity and other asset classes over the period between November 2010 and July 2019. The cross-quantilogram approach is a correlation-based approach which utilizes model-free measures of correlations between two time series across the quantiles of each distribution. The goal is to identify the sign and strength of the spillover among regional green equity markets and other segments of the financial markets under various market conditions. We choose to rely on the crossquantilogram method for several reasons. First, this framework is based on quantile hits and thus does not depend on any moment condition. Therefore, it can accommodate time series with heavy tails, a characteristic commonly found in financial data. Second, the cross quantilogram allows us to capture the nonlinearity and asymmetries in the interdependence among financial assets, thereby providing new insights into the performance of green equity markets across a wide range of market conditions. In this paper, we separate the green equity market into three regions, which are the U.S., Europe and Asia, and evaluate their integration with one another and with other financial asset classes.
Our empirical results suggest heterogeneous degrees of integration among the regional green equity markets and other asset classes. First, within the green equity universe, the U.S. market has a strong influence on the European and Asian markets, while spillovers in the opposite direction are only significant when the U.S. market is in a bearish condition. We also find that the Asian green equity market is under significant influence from its U.S. and European counterparts, especially in the short run. These results suggest the importance of the U.S. region in the green equity market, therefore, it is important to monitor this segment closely, as shocks in the U.S. market tend to spill over to other regions. At the same time, our results also indicate the vulnerability of the Asian green equity market to movements in other markets. Therefore, as this market continues to grow in the future, it is important that investors closely monitor their Asian green equity investment. Moreover, policy that stabilizes the Asian green equity market can help attract further funding to this region, which in turn plays an important role in reducing the substantial carbon footprint of the region in recent years.
Second, our results also shed light on the heterogeneous relationship between regional green equity markets and other segments of the general financial markets. Specifically, we find significant spillovers from the energy commodity market to the Asian green equity markets, while spillovers from the energy commodity market to the U.S. and European green equity markets are insignificant. This suggests that adding green equity from the U.S. or Europe to a portfolio of energy commodity such as oil and natural gas can provide some diversification benefits. On the other hand, Asian green equity provides limited diversification benefits to a portfolio of energy commodity. Additionally, we also find that technology stock significantly affects all regional green equity markets at the lower quantiles. In contrast, spillovers from technology stock to the green equity markets at the median and higher quantiles are only significant for the European and Asian green equity markets. Therefore, the inclusion of U.S. green equity may provide some hedging benefits for investors in the technology stock market during normal or bullish market conditions. However, such diversification benefits diminish during bearish market conditions. In contrast, European and Asian green equities provide little hedging benefits for a technology stock investment portfolio. Finally, our results suggest that directional spillovers among the markets tend to dissipate at longer time horizons, which implies that the inclusion of green equity may provide larger diversification benefits to an investment portfolio in the long run than in the short run.
Our paper extends the literature in several ways. First, a large literature focuses on documenting the relationship between renewable energy stock and other asset classes, such as fossil fuel prices and general stock prices. For example, Henriques & Sadorsky (2008) and Sadorsky (2012) find that oil price can influence clean energy stock, however, its effect tend to be smaller than that of technology stock prices. Managi & Okimoto (2013) studies structural changes in the oil-technology-clean energy stock relationship and finds a positive and significant relationship between oil price and clean energy stock. Reboredo (2015) documents systemic risk and dependence between oil and renewable energy markets using copula models and finds significant tail dependence between these variables. More recently, Ahmad et al. (2018) studies the optimal hedge ratios between renewable energy stock and various asset classes, while Ferrer et al. (2018) studies how the relationship between renewable energy stock and other assets changes across various time frequencies. Yet, in this rich literature, no consensus has been reached on the relationship between the renewable energy stock market and other financial markets. One possible explanation is that most previous studies have focused on analyzing this nexus at the aggregate level by using an aggregate stock index that represents the entire green equity market. By doing this, the literature has not been able to identify the heterogeneity in the response of each segment of the green equity market to other financial assets. Additionally, previous work in this area has mainly focused on documenting the mean-to-mean relationship between environmentally friendly stock and other assets. Thus, empirical evidence on the tail dependence between these financial markets has not been fully investigated. 2 Our paper contributes to this literature by providing the first empirical evidence on the heterogeneity in the relationship between regional green equity markets and other financial assets. Our results show that the responses of green equity to movements in other markets vary largely across regions, which suggests heterogeneous diversification benefits of different types of green equity in an investment portfolio.
Second, our paper contributes to the literature by unraveling the integration within the green equity universe. Many previous studies have documented the external integration between 2 To the best of our knowledge, we are only aware of two papers that investigate tail dependence between green equity and other asset classes. First, Reboredo (2015) uses copulas to study the dependence structure between oil and renewable energy market. While copulas can measure tail dependence, it can only measure a general dependence between variables and the magnitude of dependence depends largely on the specifications of the copula functions. Second, Uddin et al. (2019) employs cross-quantilograms to measure dependence between renewable energy stock and several financial assets. However, their empirical analysis is based on an aggregate index for the entire green equity market, therefore, they are unable to document the heterogeneity in the internal and external integration of green equity and other asset classes. green equity and other financial assets, however, the internal integration among various segments within the green equity markets has been unexplored. The reason is that by using an aggregate green equity stock index, these studies have treated the global green equity market as one composite market. 3 Our empirical evidence shows heterogeneous degrees of integration across various regions within the green equity markets. Therefore, our empirical approach is relevant for green equity investors who are looking for diversification opportunities while maintaining the environmentally friendly nature of their investment portfolio. While our results show that investing in a diverse set of green equity may provide limited diversification benefits in the short run, such a strategy may be more effective in the long run, as directional spillovers across regional green equity markets tend to dissipate as the time horizon expands.
Third, our paper also contributes to the scant literature that document the Asian financial market for environmentally friendly investments. As one of the world's fastest growing regions, environmental degradation in this region is a growing concern. Urbanization, changing lifestyles and increasing demand for resources and services are putting pressure on natural resources in the region, which threatens the sustainability of the region's development. 4 As the transition towards a low-carbon economy requires a substantial amount of funding, it is important to inform investors and policymakers about the performance of the green equity market in Asia, in order to attract additional funding to this region. Among the few studies that document the Asian financial market for environmentally friendly investments (for example, Wen et al. (2014); Reboredo & Wen (2015); Zhang & Du (2017); Zhu et al. (2019)), most of them have focused on documenting the mean-to-mean relationship between clean equity stock indexes and other asset classes, while the tail dependence between these variables has remained unexplored. Moreover, these studies have not explored the relationship between Asian green equity markets and other regional green equity markets, which may play an important role in the design of effective hedging strategies and policies to promote Asian green equity among investors. We extend this line of research by investigating the integration of the Asian green equity market with other green equity markets and other asset classes at both the median and the tail of the return distribution. Thus, the results from 3 For example, the literature have extensively studied renewable energy indexes such as the WilderHill ECO index, the WilderHill NEX index, the S&P Global Clean Energy index, and the ERIX index.
4 Since the early 2000s, Asia was the world's largest CO2 emission. Its emission in 2016 was twice the level of the Americas and three times that of Europe. China accounted for more than 50% of emissions in Asia in 2016, followed by India. Emission also increased significantly since 2000 for other countries in the region (International Energy Agency, 2018). our empirical analysis provide investors and policymakers with important information on the overall performance of the Asian green equity market.
The rest of the paper proceeds as follows. Section 2 and 3 describe our data set and empirical methodology. Sections 4 presents the empirical results and finally, section 5 provides a conclusion.

Data
To investigate the internal integration within the green equity market, we use stock indexes from the NASDAQ OMX Green Economy Index Family to represent the regional green equity markets. Specifically, we consider the NASDAQ OMX Green Economy U.S. Index (GEUS), the NASDAQ OMX Green Economy Europe Index (GEEU) and the NASDAQ OMX Green Economy Asia Index (GEASIA) as proxies for the performance of the U.S., European and Asian green equity markets. These indexes include companies domiciled in the U.S., Europe and Asia across the spectrum of industries closely related to the economic model around sustainable development. Our choice of the NASDAQ OMX Green Economy Indexes is based on the following reasons. First, unlike other stock indexes that mainly focus on the clean energy sector 5 , the NASDAQ OMX Green Economy Indexes include a broader set of green equities in every economic sector, thereby providing a more comprehensive picture of the green equity markets (NASDAQ, 2015). Second, these indexes allow the consideration of the Asian green equity market, which has not been extensively studied in the previous literature. As the rapidly growing economies in Asia are contributing an increasing share in global greenhouse gas emission, energy consumption and environmentally friendly investment 6 , incorporating the Asian market in the analysis of global green financial investment may have important implications for investors who are interested in further diversifying their green investment portfolio. We sourced daily data on the spot values of these indexes from Quandl for the period 10 November 2010 (the index launch date) to 08 July 2019 and convert them to daily returns by taking the log-difference of the daily index values.
Since another goal of the paper is to study how the dependence structure between green 5 Examples of such indexes include the WilderHill Clean Energy (ECO) Index or the European Renewable Energy (ERIX) Index. These indexes have been extensively studied in the clean energy finance literature. 6 According to the World Energy Outlook, Asia makes up half of global growth in natural gas, 60% of the rise in wind and solar PV, more than 80% of the increase in oil, and more than 100% of the growth in coal and nuclear (International Renewable Energy Agency, 2018). equity and other financial assets varies across regions, we also collect data on other asset classes in addition to the regional equity stock indexes. Specifically, we consider two groups of financial assets that have been found to strongly influence green equity in the previous literature: the energy commodity market and the conventional stock market. In this paper, we use the S&P GSCI Energy Index as a proxy for the performance of the global energy commodity market, and the S&P GSCI Crude Oil and Natural Gas Indexes as a proxy for the oil and natural gas commodity markets. These indexes track the performance of the global energy commodity market, thereby providing a common benchmark to compare the dependence structure between the regional green equity and energy commodity markets. To proxy for the stock market, we use the S&P Global BMI Index as a proxy for the performance of the general stock market and the NYSE Arca Tech 100 (PSE) Index as a proxy for the performance of the technology stock market, a sector that has been found to strongly correlate with green equity in the literature. 7 Finally, to account for the impact of market and policy uncertainty on the dependence structure among regional green equity, energy commodity and stock markets, we use the VIX, OVX and EPU indexes as proxies for the market, energy and economic policy uncertainty. Data for the VIX and OVX are sourced from Yahoo! Finance while data for the EPU index are available from the St. Louis FRED website. All these variables are log-differenced to ensure their stationarity.
Figures 1 and 2 plot the daily closing prices and returns of the variables described above. All three regional green equity indexes experience a decline during the oil price collapse of 2014-2016, however, the GEUS and GEEU indexes seem to recover more quickly than the GEASIA index. The S&P GSCI Energy index co-moves with the S&P GSCI Oil index, where both indexes exhibit a sharp decline during the oil price collapse of 2014-2016. the S&P GSCI Natural Gas index reaches its minimums in 2012 and 2016, while peaking in 2015. Higher production output and lower demand due to warmer temperature are the main reasons for low natural gas price in 2012 and 2016, while extreme cold weather explains the spike in natural gas price in 2015 (U.S. Energy Information Administration, 2013, 2017). The PSE and S&P Global BMI indexes co-move, where both indexes experience a decline between June 2015 and June 2016. This period corresponds to the 2015-2016 stock market selloff, where stock prices decline globally. Table 1 provides the summary statistics of the asset returns. Among the three regional green equity indexes, the GEUS index experiences the highest average returns while the GEASIA market experiences the lowest average returns during the sample period. All three energy commodity indexes experiences a negative average returns during the sampling period, potentially because of the 2014-2016 oil price collapse and the lower natural gas prices in response to lower demand, as explained above. Compared to the green equity and stock market indexes, the energy commodity indexes have higher standard deviations, which is expected because of the large movements in the oil and natural gas markets during our sample period (figure 1). All series have negative skewness, except for the uncertainty measures (i.e. the VIX, OVX and EPU indexes). All variables have kurtosis greater than 4, which implies fatter tails than a normal distribution. The Ljung-Box statistics on the returns and squared returns suggest that all series experience serial correlation and volatility clustering and the Jarque-Bera tests indicate that the series do not follow a normal distribution. Finally, the ADF unit root tests indicate that all return series are stationary.

Empirical methodology
To examine the dependence between regional green equity markets and other financial assets, we employ the cross-quantilogram (CQ) approach by Han et al. (2016). Compared to other methods for directional spillovers, the CQ has several advantages. First, the method is based on quantile hits and does not require moment conditions, therefore, it works well for heavy-tailed variables such as financial time series. Additionally, since the CQ estimates correlation across quantiles, it is able to account for asymmetries in the dependence structure and to capture the relationship between time series across all parts of the distributions (for example, extreme negative, central, extreme positive observations). Finally, compared to traditional regression type methods, the CQ approach can accommodate long lags, thereby allowing simultaneous detection of the direction, magnitude and duration of dependence. In other words, the CQ approach offers a more complete picture of the dependence between the variables.
Let y it be stationary time series, where index i represents stock returns and t represents time (i = 1, 2, t = 1, ..., T ). Let F i (·) and f i (·) be the distribution and density functions of The cross-quantilogram between two events {y 1t ≤ q 1t (τ 1 )} and {y 2t−k ≤ q 2t−k (τ 2 )}, where k denotes the lag length (k = ±1, ±2, ...), for a pair of τ 1 and τ 2 is defined as: where ψ a (u) = 1[u < 0]−a is the quantile-hit process. The cross-quantilogram captures serial dependence between two series at different quantile levels and is invariant to any strictly monotonic transformation applied to both series, such as the logarithmic transformation.
In the case of two events, To test the null hypothesis H 0 : ρ τ (1) = ... = ρ τ (p) = 0 against the alternative hypothesis that ρ τ (k) = 0 for some k, Han et al. (2016) suggests the following Ljung-Box type test statistic: whereρ τ (k) is the sample cross-quantilogram, which is given as: whereq it (τ i ) (i = 1, 2) denote the estimated quantile function for each time series. Han et al. (2016) proposes using the stationary bootstrap procedure to approximate the null distribution of the cross-quantilograms and the Q-statistic above while avoiding any dependence on the nuisance parameters of the asymptotic distribution. The stationary bootstrap is a block bootstrap method with blocks of random lengths. Let {K j } j∈N be a sequence of iid random variables, which are drawn from a discrete uniform distribution on {k + 1, ..., T } and which are independent of the original data and {L j } j∈N . {L j } j∈N is a sequence of iid random block lengths with a geometric distribution.
be the blocks of length L j starting with the K j th pair of observations, where The stationary bootstrap procedure generates the boostrap samples {(y * t,k )} T t=k+1 , which are then used to estimate the conditional quantile functionq * The cross-quantilogram based on the bootstrapped resample is: In this paper, we consider 1,000 bootstrapped estimates ofρ * τ (k) to construct the confidence intervals for the test statistic in equation (2).
The correlation matrix of the quantile hit process and its inverse matrix are defined as: ] be an l × 1 vector of the quantile hit process. For i, j ∈ [1, .., l], let rτ ij and pτ ij be the ij-th element of Rτ and Pτ . Note that the cross-quantilogram is rτ 12 / √ rτ 11 rτ 22 . The partial cross-quantilogram is defined as: ρτ |z can be regarded as the cross-quantilogram dependence between y 1t and y 2t conditional on the control variables z.

Empirical results
In this section, we present the empirical evidence on the internal integration among regional green equity markets and the external integration between green equity markets and other asset classes, using the CQ and PCQ framework discussed in section 3. Compared to the previous literature, our empirical approach offers several new insights into environmentally friendly financial markets. First, the CQ and PCQ method allow us to document the directional spillovers among the return series across different lag lengths and quantiles, which is important for the selection of hedging and diversification strategies under various market conditions. Second, by separating the global green equity market into regional indexes, we are able to investigate the internal integration of various segments within the green equity market and the heterogeneity in the integration between regional green equity and major asset classes. Our empirical results are helpful to identify new diversification opportunities that have been overlooked by the use of aggregate green equity indexes in previous studies.
We present our results in the form of heat maps for different lag lengths, where the axes correspond to a quantile of green equity and/or other asset returns. Heatmaps provide a graphical representation of the dependence among two series across the whole range of quantiles, therefore, they offer a visual and intuitive way to capture the entire dependence structures between the variables. In this paper, we consider 11 quantiles, specifically, q = (0.05, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 0.95), and thus the bivariate quantile combinations of each pair of variables are presented by 121-cell heatmaps. We use the Box-Ljung test described in section 3 to determine the statistical significant of each correlation. Any statistically insignificant correlation is set to 0. Figure 3 displays the CQs on the directional spillovers among regional green equity markets at four different lags: daily (k = 1), weekly (k = 5), monthly (k = 22) and quarterly (k = 66). Overall, the figure indicates varying degrees of integration among regional green equity markets across the time horizons.

Internal integration among green equity markets
First, in the case of lag 1, directional spillovers are positive and statistically significant from GEUS to GEEU and from GEUS to GEASIA across all the quantiles. Thus, when returns on the GEUS index are low (high), it is more likely to observe low (high) returns on the GEEU and GEASIA indexes the following day. Regarding the directional spillovers in the opposite directions, figure 3 shows that the GEEU and GEASIA indexes significantly influence the GEUS index only when the GEUS is in its lower quantiles. Second, we also find that at lag 1, the Asian market is the main receivers of shocks from other markets. Specifically, directional spillovers from both the GEUS and GEEU indexes to the GEASIA index are positive and significant across all the quantiles. In contrast, directional spillovers in the opposite direction are only positive and significant when all returns are in the lower quantiles. Altogether, our results suggest that in the short run (when k = 1), crashes are bidirectional within the green equity universe. In other words, the regional green equity markets are more integrated during turbulent times, where low returns in any region often spill over to other regions the following day. In contrast, booms are unidirectional. Specifically, high returns in the U.S. market often lead to high returns in the European and Asian markets the following day. However, high returns in the European and Asian markets are not necessarily associated with high returns in the U.S. market the following day. This indicates the significance of the U.S. market in the overall green equity market, where high returns in the U.S. green equity market send a positive signal to investors, which encourage them to expand their environmentally friendly investments to other regions. Thus, closely observing the U.S. green equity market is helpful in predicting stock returns in other regional green equity markets. Our results also indicate the vulnerability of the Asian green equity market to the performance of the U.S. and European green equity markets, therefore, it is important that investors closely monitor the Asian portion of their green equity portfolio.
Second, figure 3 shows that the directional spillovers among the green equity indexes become smaller and less statistically significant as the lag length increases, which implies that regional integration within the green equity universe dissipates over time. This indicates some hedging opportunities within the green equity market in the long run, where investors can hedge their green equity from one region with green equity from other regions.
To estimate how general market conditions affect the integration among regional green equity markets, we incorporate several control variables and estimate the partial crossquantilograms described in section 3. Specifically, our set of control variables includes the VIX index (a proxy for overall stock market uncertainty), the OVX index (a proxy for overall energy market uncertainty) and the EPU index (a proxy for economic policy uncertainty). Figure A-1 displays the PCQs among the regional green equity markets after incorporating these control variables. Since the results from our PCQs estimates are similar to the main results in figure 3 across the lag lengths and since the directional spillovers dissipates at longer lags, we only present the PCQs for lag 1 in figure A-1. 8 Overall, the inclusion of these control variables does not qualitatively change the dependence structure among regional green equity markets, although we observe a slightly stronger spillovers at the lower quantiles from the GEEU and GEASIA indexes to the GEUS index. Note that our results do not imply that the VIX, OVX, EPU have no influence on the green equity market. They only indicate that these control variables carry little information on the dependence structure among the regional green equity markets.
Finally, to examine how directional spillovers change over time, we perform a recursive subsample estimation of the cross-quantilogram with the first window length of 252 days. Specifically, we estimate the cross-quantilogram for the first window period. Then we increase the length of the window by one day and re-estimate the cross-quantilogram and we continue this process until the end of the sample period. As our above discussion indicate that directional spillovers are the most significant at lag 1 and weaken at longer time horizons, we focus our discussion on the recursive cross-quantilograms for lag 1. 9 Figure A-2 presents the results, where the first, second and third rows show the recursive CQs when both return distributions are at the 5%, 50% and 95% quantiles. The horizontal axis represents the starting year of the recursive window. The blue lines are time-varying cross quantilograms in the recursive subsamples and the red lines indicate the 95% confidence interval for the no-predictability null hypothesis. We derive our confidence intervals by using 1000 bootstrap iterations in this paper. Overall, figure A-2 shows time-varying directional spillovers across the regional green equity markets. However, our recursive estimates are consistent with our previous findings that directional spillovers are larger from the GEUS to the GEEU and GEASIA indexes and from the GEEU to the GEASIA index, while spillovers in the opposite directions are smaller.  Figure 3: Cross-quantilogram heatmaps among regional green equity markets This figure reports the cross-quantilogram between the markets (Market 1 → Market 2) at four time horizons: daily (k = 1), weekly (k = 5), monthly (k = 22), quarterly (k = 66). In each heatmap, the vertical axis represents the return quantiles of Market 2, while the horizontal axis represents return quantiles of Market 1.

Integration between regional green equity markets and other asset classes
In this section, we will discuss how the integration between green equity markets and other asset classes vary across the regions. In this paper, we focus on two main asset classes: energy commodity and general stock markets. These assets have been found in the previous literature to have strong influence on the green equity market. However, in contrast to the previous literature that treats the global green equity market as one composite market, our empirical approach aims at identifying the regional variations in the relationship between green equity market and other asset classes both at the median and tails of the return distributions. We will discuss the details of our findings in the following sections. Figure 4 displays the CQs heatmaps between regional green equity markets and the energy commodity market, which is proxied by the S&P GSCI Energy index. One main finding from the figure is that, the energy commodity market does not have a significant impact on the GEUS and GEEU indexes while it has a positive and statistically significant impact on the GEASIA index across all the quantiles. Thus, green equity in Asia crashes and booms with the energy commodity market, while there is little integration between the energy commodity market and the European and U.S. green equity. One possible explanation is that compared to the Asian market, the U.S. and European green equity markets are more established. As a result, investors in the U.S. and European markets have started to decouple the green equity market from the energy commodity market, therefore, their green equity investment decisions are not significantly affected by changes in energy commodity returns. In contrast, the less established Asian green equity market have only started to gained attention in recent years. Since the behavior of the Asian green equity market are not as well understood as its U.S. and European counterparts, investors' decisions to invest in this market are largely influenced by the performance of alternative investment options. Since our data set contains more recent periods, our results provide evidence about the nexus between energy commodity and green equity across the regions in recent years. Figure 4 also indicates that none of the regional green equity market has a significant influence on the energy commodity market across all lag lengths and quantiles.

Regional green equity markets and energy commodity markets
Next, to investigate how the integration between green equity markets and energy commodity varies with the type of energy commodity, we separate the energy commodity market into the oil and natural gas markets and re-estimate the CQs between each regional green equity market and the oil and natural gas markets. In this paper, we use the S&P GSCI Crude Oil Index as a proxy for the oil commodity market and the S&P GSCI Natural Gas Index as a proxy for the natural gas commodity market. 10 Figures 5 and 6 present our results. First, we find insignificant directional spillovers between the regional green equity markets and the natural gas commodity market. Second, there is insignificant spillovers from the regional green equity markets to the oil commodity market. Third, returns in the oil commodity market only have a significant impact on the Asian green equity market while their impact on the European and U.S. green equity markets are insignificant. Overall, these results are consistent with the findings of a weakening relationship between energy commodity markets and the stock prices of new energy companies in several previous studies (Ahmad, 2017;Sadorsky, 2012;Henriques & Sadorsky, 2008).
Figures A-3-A-5 display the PCQs between the regional green equity markets and the energy commodity markets after controlling for the VIX, OVX and EPU. Overall, our discussion above still applies, however, the oil commodity market exhibits a slightly more significant impact on the U.S. and European green equity market at the lower quantiles. Figures A-6-A-8 present the recursive CQs between the regional green equity markets and the energy commodity markets. In the case of the oil market, our recursive estimates confirm our findings above that the oil commodity market only has a significant impact on the GEASIA index. Additionally, there was an increase in directional spillovers from the oil commodity market to the GEASIA index during the oil price collapse between late 2014 and early 2016 at the 0.05 and 0.95 quantiles. This suggests that the oil commodity market and the Asian green equity market are more integrated during turbulent times when returns are extreme. Regarding the nexus between the green equity market and the natural gas commodity market, our recursive CQ estimates still show insignificant directional spillovers between these markets across the quantiles. However, we find an increase in the CQ from the natural gas market to the GEUS index, particularly at the 0.05 and 0.95 quantiles. This reflects the increasing importance of natural gas in the U.S. after the shale gas revolution in the early 2000s.
10 An alternative proxy for the oil market is the WTI crude oil prices, the Brent crude oil prices, and the NYMEX continuos contract crude oil future prices. An alternative proxy for the natural gas market includes the Henry Hub natural gas prices or the NYMEX continuous contract natural gas future prices. We find that the correlations between these alternative measures and out main measures of the oil and natural gas markets are more than 0.90. Thus, we choose to rely on the S&P GSCI Crude Oil and Natural Gas indexes, because the global scope of these indexes provides a commom benchmark to study the heterogeneous dependence structure between green equity and energy commodity across regions.
To summarize, our results suggest that the nexus between the energy commodity markets and the green equity markets varies across regions and types of energy commodity. First, our results show larger vulnerability of the Asian green equity market to the energy commodity market, compared to other regional green equity markets. Specifically, the oil commodity market has a significantly positive impact on the Asian green equity market, while its impact on other green equity markets is insignificant. Additionally, we find insignificant spillovers from the green equity market to the oil commodity market across regions and quantiles. This implies that investors in the oil market can hedge or diversify by investing in the U.S. or European green equity markets, however, diversification benefits from the Asian green equity market are limited. Second, with respect to the natural gas commodity market, our results show an overall insignificant integration between this market and the regional green equity markets. Therefore, investors in the natural gas commodity market can diversify their investment with green equity. However, since we also find an increase in the spillover from the natural gas commodity market to the U.S. green equity market over our sample period, the diversification benefits provided by the U.S. green equity market may have declined in recent years. Finally, our results also indicate that hedging energy commodity with green equity may be more effective in the long run, as the CQs become insignificant as the lag length increases.  Figure 4: Cross-quantilogram heatmaps between regional green equity markets and energy commodity market This figure reports the cross-quantilogram between the markets (Market 1 → Market 2) at four time horizons: daily (k = 1), weekly (k = 5), monthly (k = 22), quarterly (k = 66). In each heatmap, the vertical axis represents the return quantiles of Market 2, while the horizontal axis represents return quantiles of Market 1.

21
Electronic copy available at: https://ssrn.com/abstract=3488402 Figure 5: Cross-quantilogram heatmaps between regional green equity markets and OIL commodity market This figure reports the cross-quantilogram between the markets (Market 1 → Market 2) at four time horizons: daily (k = 1), weekly (k = 5), monthly (k = 22), quarterly (k = 66). In each heatmap, the vertical axis represents the return quantiles of Market 2, while the horizontal axis represents return quantiles of Market 1.

22
Electronic copy available at: https://ssrn.com/abstract=3488402 Figure 6: Cross-quantilogram heatmaps between regional green equity markets and GAS commodity market This figure reports the cross-quantilogram between the markets (Market 1 → Market 2) at four time horizons: daily (k = 1), weekly (k = 5), monthly (k = 22), quarterly (k = 66). In each heatmap, the vertical axis represents the return quantiles of Market 2, while the horizontal axis represents return quantiles of Market 1. Figure 7 displays the CQs between the regional green equity market and the stock market, which is proxied by the S&P Global BMI index. 11 Our results show that the GEUS index has a significant and positive influence on the BMI index across all the quantiles. This may be due to the importance of the U.S. stock market in the global stock market. Since the GEUS index consists of firms domiciled in the U.S., it has a significant impact on the global stock market. We find similar results for the GEEU index, however, the impact of the GEEU index on the BMI index is smaller than that of the GEUS index across all the quantiles. Finally, the GEASIA index has an insignificant impact on the BMI index, potentially because of the less dominant role of the Asian stock market in the global stock market.

Regional green equity markets and the stock market
Regarding the impact of the BMI index on the green equity indexes, figure 7 indicates that the BMI index has a positive and significant impact on the GEUS index only when the GEUS index is at the lower quantile. This suggests that investors in the U.S. green equity market move from green equity to regular stock when the U.S. green equity market is in a bullish condition. In contrast, the BMI index exhibits a positive influence on the GEASIA index across all the quantiles. In other words, the Asian green equity market booms and crashes with the global stock market. This is consistent with the fact that compared to other regional green equities, Asian green equity represents a newer class of investment, therefore, it tends to be more vulnerable to movements in other financial markets. Our results also suggest that most of the directional spillovers in figure 7 dissipates at longer lag lengths, however, they are more persistent than the directional spillovers between green equity and energy commodity markets. One possible explanation is that green equity is a part of the stock market, therefore, it is more integrated with the stock market than with the energy commodity market.
In addition to considering the general stock market, we also consider the directional spillovers between regional green equity and technology stock, a class of equity investment that has been found to be highly correlated with green equity in the literature (for example, Sadorsky 11 We choose to rely on the S&P Global BMI index as a proxy for the general stock market because its global coverage allows us to have a common benchmark across all the regions. In unreported results, we also calculate the CQs between regional green equity and and its corresponding regional stock index. While our results at the tails are similar for the global stock index and the regional stock index, we find insignificant median-to-median dependence between green equity and general stock market within each region. This implies that during normal market conditions, regional green equity markets are more integrated with the global stock market than with their regional stock market. This implies that investors view the green equity market as a part of the global stock market, rather than as a part of the local stock market. (2012); Managi & Okimoto (2013); Reboredo (2015); Bondia et al. (2016)). Figure 8 presents the CQ heatmaps between regional green equity markets and the technology stock market, which is proxied by the NYSE Arca Tech 100 index. We find that the technology stock market and the U.S. green equity market significantly influence each other when both markets are at the lower quantiles. In the case of the European green equity market, our results show significant spillovers from the technology stock market, however, spillovers in the opposite direction are only significant when the technology stock market is at its lower quantiles. Similar results are obtained for the Asian green equity market, however, we find a stronger dependence between the Asian green equity market and the technology stock market. A possible explanation for the varying dependence of the regional green equity markets on the technology stock market relies on the heterogeneity in green financing activity across the regions. According to Bloomberg Energy Finance, in recent years, Asia has grown to be the leading region in renewable energy finance, followed by Europe and the U.S. (Bloomberg New Energy Finance, 2018). Thus, together with the fact that Asian green equity is a relatively new segment of the global green equity market, it is more vulnerable to movements in the technology stock market. In contrast, more developed regions such as Europe and the U.S. have lost their role as the main destination of new energy finance, therefore, they are less integrated with the technology stock market (Bloomberg New Energy Finance, 2018). Since our data set starts from November 2010, it is able to capture these recent trends in the nexus between green equity and technology stock markets. We also find that the dependence structure between the green equity markets and the technology stock market persists for a longer period than that between the green equity and energy commodity markets. These results are consistent with the previous findings of a stronger correlation between green equity and technology stock than between green equity and energy commodity in the literature (Sadorsky, 2012;Managi & Okimoto, 2013;Reboredo, 2015;Bondia et al., 2016).
To test the robustness of our results, we incorporate the VIX, OVX and EPU indexes into our analysis and estimate the PCQs between the regional green equity indexes and the stock market indexes. The results for these PCQs are reported in figures A-9 and A-10. We also estimate a recursive CQs to account for the time-varying dependence structure between these markets (figures A-11 and A-12). Overall, these alternative specifications are consistent with our findings discussed above.  Figure 7: Cross-quantilogram heatmaps between regional green equity markets and BMI commodity market This figure reports the cross-quantilogram between the markets (Market 1 → Market 2) at four time horizons: daily (k = 1), weekly (k = 5), monthly (k = 22), quarterly (k = 66). In each heatmap, the vertical axis represents the return quantiles of Market 2, while the horizontal axis represents return quantiles of Market 1.

Implications of the results
Our empirical results on the internal integration among regional green equity markets and the external integration between these markets and other asset classes have implications for both investors and policymakers.
First, our CQ and PCQ estimates suggest that most of the integration among regional green equity markets and other asset classes is more significant at shorter lag lengths. This implies a fast processing of information by these markets. These results are consistent with several recent papers that documented the short-lived nature of connectedness among various financial markets. For example, Lau et al. (2017) and Tiwari et al. (2018) find larger connectedness across precious metals and several financial assets in the short term than in the long term, and similar results were also obtained in Ferrer et al. (2018), which focuses on the relationship between several financial assets and renewable energy stocks. Thus, our results suggest that the benefits from hedging green equity with financial assets such as energy commodity or conventional stock are larger in the long run than in the short run.
Second, our CQ and PCQ estimates show that regional green equity markets are more integrated among themselves and with the energy commodity and stock markets at the lower quantiles, which implies increasing connectedness among these markets during bearish market conditions. This result is consistent with the stylized fact that market interconnectedness are stronger during turbulent periods (Okimoto, 2008). Thus, the effectiveness of different hedging and diversification strategies may change during different market conditions. Third, our results on the internal integration among the regional green equity markets show the significant influence of U.S. green equity on European and Asian green equity. Thus, it is important to pay close attention to the U.S. green equity market, as its performance is helpful in predicting the performance of green equity in other regions. On the other hand, Asian green equity is the main receiver of shocks within the green equity market. Therefore, as Asia gains its dominance in the global market for green financing and given the rapid increase in Asian energy consumption and carbon emission in recent years, closely monitoring this segment of the green equity market plays an important role in global carbon emission reduction.
Fourth, our empirical evidence shows that the energy commodity market, particularly the oil commodity market, can significantly influence green equity stock returns in Asia, thus, it plays an important role in shaping the profitability of environmentally friendly investment in Asia. In contrast, other regional green equity markets exhibit an insignificant dependence structure with the energy commodity market, which suggests the decoupling of green equity in these regions with the energy commodity market. This means that investors in the energy commodity market can benefit from hedging their investment with green equity from the U.S. and Europe, while the diversification benefits from Asian green equity are limited, especially in the short run. In addition, price regulations of the energy commodity market can significantly increase financial flows towards the Asian green equity market, while they may have insignificant impact on investors' incentives to invest in the U.S. and European green equity market.
Finally, our results indicate that green equity markets are more integrated with the stock market than with the energy commodity market. Therefore, investors view green equity as an equity investment rather than as an energy investment. Additionally, the dependence between green equity and the stock market is more significant at the lower quantiles of the return distributions, while there exists substantial regional variations in the dependence structure among these variables at the median and upper quantile of the return distributions. Thus, green equity investors can adopt similar investment strategies across all regions during turbulent time when returns in all markets are low. On the other hand, it is highly recommended that they take into account the regional heterogeneity in the response of the green equity market to the general stock market when designing their investment strategies during normal or bearish market conditions.

Conclusions
Concerns over climate change have sparked substantial interests among policymakers and investors in environmentally friendly financial instruments. As the green financing market expands in its scope and size, heterogeneity in the performance of various sub-segments in this market will emerge, which have important implications for policymakers and investors. This paper aims at investigating how the integration among green equity, energy commodity and stock markets varies across regions and market conditions. Our empirical results show significant integration among regional green equity markets. Specifically, the U.S. market is the main transmitter of shocks to the European and Asian markets while the Asian market is the main receiver of shocks from other regions. Our results also indicate varying degrees of integration between regional green equity markets and other financial assets. Specifically, the oil commodity market significantly influences the Asian green equity market, while its impact on other regional green equity markets is insignificant. The gas commodity market exhibits an insignificant dependence structure with the green equity market, however, our findings indicate an increasing dependence between the natural gas commodity and the U.S. green equity market over the sample period. Finally, the technology stock market are more integrated with the Asian and European green equity markets than with the U.S. green equity market.
Our findings of a significant regional variations in the integration within the green equity market and between the green equity market and other financial assets highlight the importance of investigating the environmentally friendly financial sector at a disaggregate level. To the best of our knowledge, our paper provide the first empirical evidence of the quantile directional spillovers among various sub-sectors within the green equity market, and between these sub-sectors and other financial assets. Our findings of a heterogeneous relationship among regional green equity and other asset classes imply that green equity investors should take into account the regional origins of their investments and adopt a diverse set of trading strategies. At the same time, investors in the energy commodity and general stock market who are interested in diversifying their investment with green equity should proceed with caution, as the amount of diversification benefits depends on the type of green equity. Finally, policy must take into account these regional variations in the green equity market, since the same policy may be effective in attracting additional funding towards environmentally friendly investments in one region but turn out to be ineffective in another region. Cross-quantilogram heatmaps between regional green equity markets and energy commodity market with control variables

A.2 Recursive cross quantilograms
This figure reports the partial cross-quantilograms (PCQs) between the markets (Market 1 → Market 2) after controlling for market uncertainties. In each heatmap, the vertical axis represents the return quantiles of Market 2, while the horizontal axis represents return quantiles of Market 1. The PCQs are calculated for lag 1.

35
Electronic copy available at: https://ssrn.com/abstract=3488402 Cross-quantilogram heatmaps between regional green equity markets and OIL commodity market with control variables This figure reports the partial cross-quantilograms (PCQs) between the markets (Market 1 → Market 2) after controlling for market uncertainties. In each heatmap, the vertical axis represents the return quantiles of Market 2, while the horizontal axis represents return quantiles of Market 1. The PCQs are calculated for lag 1.

36
Electronic copy available at: https://ssrn.com/abstract=3488402 (a) GEUS → GAS VIX OVX EPU (b) GAS → GEUS (c) GEEU → GAS (d) GAS → GEEU (e) GEASIA → GAS (f) GAS → GEASIA Figure A-5: Cross-quantilogram heatmaps between regional green equity markets and GAS commodity market with control variables Cross-quantilogram heatmaps between regional green equity markets and BMI commodity market with control variables This figure reports the partial cross-quantilograms (PCQs) between the markets (Market 1 → Market 2) after controlling for market uncertainties. In each heatmap, the vertical axis represents the return quantiles of Market 2, while the horizontal axis represents return quantiles of Market 1. The PCQs are calculated for lag 1.