Green preference, green investment

Gao, Zhenyu; Luo, Yan; Tian, Shu; Yang, Hao Working Paper Green preference, green investment ADB Economics Working Paper Series, No. 722 Provided in Cooperation with: Asian Development Bank (ADB), Manila Suggested Citation: Gao, Zhenyu; Luo, Yan; Tian, Shu; Yang, Hao (2024) : Green preference, green investment, ADB Economics Working Paper Series, No. 722, Asian Development Bank (ADB), Manila, https://doi.org/10.22617/WPS240238-2 This Version is available at: https://hdl.handle.net/10419/298168 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/3.0/igo/ ASIAN DEVELOPMENT BANK ASIAN DEVELOPMENT BANK 6 ADB Avenue, Mandaluyong City 1550 Metro Manila, Philippines www.adb.org GREEN PREFERENCE, GREEN INVESTMENT Zhenyu Gao, Yan Luo, Shu Tian, and Hao Yang ADB ECONOMICS WORKING PAPER SERIES NO. 722 April 2024 Green Preference, Green Investment This paper examines whether individual investors’ green preference will be reflected in their investment decisions. It provides compelling evidence that individuals with stronger green preference invest more in green mutual funds, influenced by concerns over the physical and regulatory risks of climate change. It suggests that this behavior is not driven by financial incentives as preference-related investments may not always lead to financial gains from trading. About the Asian Development Bank ADB is committed to achieving a prosperous, inclusive, resilient, and sustainable Asia and the Pacific, while sustaining its efforts to eradicate extreme poverty. Established in 1966, it is owned by 68 members —49 from the region. Its main instruments for helping its developing member countries are policy dialogue, loans, equity investments, guarantees, grants, and technical assistance. ASIAN DEVELOPMENT BANK The ADB Economics Working Paper Series presents research in progress to elicit comments and encourage debate on development issues in Asia and the Pacific. The views expressed are those of the authors and do not necessarily reflect the views and policies of ADB or its Board of Governors or the governments they represent. ADB Economics Working Paper Series Green Preference, Green Investment Zhenyu Gao, Yan Luo, Shu Tian, and Hao Yang No. 722 | April 2024 Zhenyu Gao ([email protected]) is an associate professor at Chinese University of Hong Kong. Yan Luo ([email protected]) is a professor and Hao Yang ([email protected]) is a PhD candidate at Fudan University. Shu Tian (stian@adb. org) is a senior economist at the Economic Research and Development Impact Department, Asian Development Bank. Creative Commons Attribution 3.0 IGO license (CC BY 3.0 IGO) © 2024 Asian Development Bank 6 ADB Avenue, Mandaluyong City, 1550 Metro Manila, Philippines Tel +63 2 8632 4444; Fax +63 2 8636 2444 www.adb.org Some rights reserved. Published in 2024. ISSN 2313-6537 (print), 2313-6545 (PDF) Publication Stock No. WPS240238-2 DOI: http://dx.doi.org/10.22617/WPS240238-2 The views expressed in this publication are those of the authors and do not necessarily reflect the views and policies ofthe Asian Development Bank (ADB) or its Board of Governors or the governments they represent. ADB does not guarantee the accuracy of the data included in this publication and accepts no responsibility for any consequence of their use. The mention of specific companies or products of manufacturers does not imply that they are endorsed or recommended by ADB in preference to others of a similar nature that are not mentioned. By making any designation of or reference to a particular territory or geographic area inthis document, ADB does not intend to make any judgments as to the legal or other status of any territory or area. This publication is available under the Creative Commons Attribution 3.0 IGO license (CC BY 3.0 IGO) https://creativecommons.org/licenses/by/3.0/igo/. By using the content of this publication, you agree to be bound bytheterms of this license. For attribution, translations, adaptations, and permissions, please read the provisions andterms of use at https://www.adb.org/terms-use#openaccess. This CC license does not apply to non-ADB copyright materials in this publication. If the material is attributed toanother source, please contact the copyright owner or publisher of that source for permission to reproduce it. ADB cannot be held liable for any claims that arise as a result of your use of the material. Please contact [email protected] if you have questions or comments with respect to content, or if you wish toobtain copyright permission for your intended use that does not fall within these terms, or for permission to use theADB logo. Corrigenda to ADB publications may be found at http://www.adb.org/publications/corrigenda. Notes: In this publication, “CNY” refers to yuan. ADB recognizes “China” as the People’s Republic of China and “Hong Kong” as Hong Kong, China. ABSTRACT Based on Alibaba’s renowned “green” initiative, the Ant Forest program, we develop a novel measure to reveal an individual investor’s nonpecuniary green preference and link it to an individual’s investment actions. We present compelling evidence that supports nonfinancial incentives for investing in green mutual funds while divesting from “brown” funds. Concerns over climate physical and regulatory risks further reinforce this influence. Individuals’ green preferences do not lead to financial gains from trading. Moreover, we mitigate the endogeneity issue by employing the development of a local subway network as a source of variations in green preferences. Keywords: nonpecuniary preference, revealed preference, sustainable finance, FinTech JEL codes: G11, G50, Q55 ______________________ This paper was previously distributed under the title “Climate Risk Awareness and Retail Trading.” We are grateful to Abdul Abiad, Chanik Jo, Wenhao Li, Wenxi Jiang, Paolo Sodini, Leilei Song, Denis Sosyura, Wei Xiong, Anthony Zhang, and participants at a 2022 Asian Development Bank workshop, Xiamen University seminar, Deakin University seminar, and 2023 Asian Economic Development Conference for helpful comments and discussions. All remaining errors are ours. The authors acknowledge and appreciate support from the Digital Finance Open Research Platform (www.dfor.org.cn). All data are sampled, desensitized, and stored on the Ant Open Research Laboratory in an Ant Group Environment, which is only remotely accessible for empirical analysis. Due to length limits of ADB working paper series, a more complete version of this working paper with additional results on tests addressing confounding factors and sensitivity issues can be accessed at SSRN: https://ssrn.com/abstract=4655538. 1. Introduction In her presidential address at the 2023 American Finance Association (AFA) Laura Starks highlights the importance of understanding the motivations behind sustainable finance, particularly the distinction between “value” versus “values” motivations (Starks 2023). Nevertheless, unraveling individuals’ motives for investments poses two major challenges. First, it is difficult to measure individuals’ nonpecuniary preferences toward sustainability and differentiate them from their financial considerations. Second, it is challenging to study how investors’ “green” preferences shape their investment choices. In this paper, we tackle these challenges by utilizing a revealed preference methodology (Samuelson 1938, 1948) and introducing a novel proxy that unveils nonpecuniary preferences for sustainability among retail investors. We then map this proxy to these investors’ investment decisions. This approach aligns with the theoretical framework presented by Pástor, Stambaugh, and Taylor (2021), who underscore the crucial role of changes in sustainability preferences in driving investors’ demand for green assets and subsequently impacting asset prices, especially during recent periods of heightened public awareness regarding climate and environmental issues. With our proxy, we can effectively track dynamics of individuals’ green tastes and empirically examine the influence of their individual environmental preferences on their green investments. Our proxy for nonpecuniary preferences for greenness is based on the Ant Forest program, a popular green initiative in the People’s Republic of China (PRC). This program won the United Nations (UN) Champions of the Earth award in 2019, which is the UN’s flagship global environmental honor (UNEP 2019). Operating under the FinTech giant Ant Group, an affiliate of Alibaba, Ant Forest aims to promote sustainable practices among individuals by encouraging them to reduce their carbon footprint and protect the environment in their daily lives. This program is integrated into the Alipay app, the PRC’s largest third-party mobile and online payment platform. It tracks users’ daily eco-friendly activities, such as using public transportation or opting out of single-use cutlery for food deliveries, to promote their green lifestyles. Users can earn “green energy points” for these actions, and once they reach a certain number of points, the Ant Forest program 2 plants a tree on their behalf and provides real-time satellite images of their trees. Importantly, green energy points cannot be exchanged for financial rewards, therefore the Ant Forest Program reveals an individual’s nonpecuniary preference for environmentally friendly practices. We then link these Ant Forest users’ green profiles to their investment portfolios. Since 2014, the Ant Group offered mutual fund distribution services through its Ant Fortune program, which allows investors to easily invest in almost all mutual funds available in the PRC. According to the IPO prospectus of the Ant Group, it is the largest online investment services platform in the PRC, with a total of CNY4,099 billion in assets under management matched and distributed as of 30 June 2020. We randomly select a sample of 200,000 retail investors who both engage in the Ant Forest program and trade on the Ant Fortune platform via the Alipay app. Our sample period spans from October 2019 to September 2021, and we obtain monthly mutual fund trading and holding data for these investors from the Ant Group, along with data on their Ant Forest green energy points, as well as information about their location, gender, and age. Additionally, we obtained information on the Environmental (E) ratings of the mutual funds from WIND, a widely used financial database in the PRC. Our baseline results demonstrate that retail investors tend to favor mutual funds with higher E-scores if they earn more green energy points. When we classify the top 20 percent of funds as green funds and the bottom 20% funds as “brown” funds based on their E-scores, we observe that investors who collect more green energy points would invest in green funds and divest from brown funds. These results indicate that investors are more inclined to choose green portfolios when their nonfinancial preferences for ecofriendliness are positively affected. A crucial concern is that individuals may invest in green funds not solely based on their nonpecuniary preferences. The superior performance of green funds might attract investors, potentially confounding our results. To mitigate this concern, we conduct our empirical analysis at the investor-fund-time level and incorporate fund-time fixed effects into all our regressions. This approach enables us to account for the time-varying 3 performance of funds, such as returns and risks, as well as other unobservable characteristics that might influence investment decisions due to investors’ financial considerations. Furthermore, we include investor fixed effects to control for individual characteristics and focus on the preference dynamics within investors rather than across investors. By including these fixed effects, we can isolate the impact of changing sustainability preferences in driving investors’ demand for green funds. Physical and regulatory risks are regarded as the primary climate risks (Stroebel and Wurgler 2021; Philipp, Sautner, and Starks 2020). In this light, we further investigate the influence of green preferences in shaping the green portfolios of retail investors who have varying exposures to the physical and regulatory impacts of climate change. To capture shocks to investors’ physical climate concerns, we utilize abnormal local temperature, following the approach of Choi, Gao, and Jiang (2020). Our analysis indicates that the transition from green preference to green investment is more pronounced for retail investors residing in cities exposed to abnormally high temperatures. Regulatory risks have a similar effect. We leverage the PRC’s proposition of its Dual Carbon Targets (DCT) to capture a time-series shock to climate change awareness of investors related to regulatory risks. We find that retail investors who have earned more green energy points tend to increase their portfolio exposure to green mutual funds following the DCT. This provides further evidence that climate and environmental regulations strengthen the connection between green preferences and investments, highlighting the influence of regulatory risks on investor behavior. To further validate the nonpecuniary motives of retail investors to invest in green funds, we examine whether the net buying of green funds upon earning more green energy points leads to outperformance. Our analysis indicates that the net purchase of funds with higher E-scores in response to a positive shock to their nonpecuniary preferences does not generate significantly higher returns over 1-, 3-, 6-, and 12-month horizons. This again underscores the role of nonfinancial considerations in green investment. Last, we perform several robustness and sensitivity tests. While the inclusion of fund-time fixed effects and the non-results of outperformance suggest that financial considerations 4 may not confound the nonpecuniary motives of green investment, it is important to acknowledge that there could still be unobservable variables that we have not accounted for. For instance, local economic shocks could impact both individuals’ environmentally friendly actions in their daily lives and their investment decisions. To mitigate the endogeneity issue, we develop an identification strategy based on the development of a local subway network. This approach employs variations in green energy points accumulation through low-carbon or low-energy consumption modes of travel. The development of a subway network provides individuals with alternative travel options, which could potentially reduce carbon emissions by encouraging the use of public transportation. However, it does not directly affect their investment choices between green and brown funds. By utilizing the development of local subway networks as an instrumental variable, we find that residents in corresponding cities are more likely to amass more green energy points through low-carbon and low-energy green travels. The instrumented green energy points further increase their holding with green funds, ensuring the causal effect of the shock to individual green preferences on their green investment behaviors. Moreover, we find our results are particularly pronounced for young investors and residents in the cities facing higher levels of air pollution. Our results are also robust using an alternative textual-based measure to identify green funds, after we exclude the sample from cities during the coronavirus disease (COVID-19) pandemic lockdown periods, when we consider a subsample of individuals who redeem green energy points for tree-planting, when we adopt alternative retail trading measures, and when we utilize cumulative green energy points. Overall, our study provides compelling evidence for the nonpecuniary motivations behind green investment. We develop a novel metric that timely and precisely reveals individuals’ nonpecuniary preferences for embracing environmental sustainability and link it to their investment portfolios. We mitigate the endogeneity concerns by introducing a new instrumental variable to identify the role of nonfinancial motives in driving investment decisions. We find that both climate physical and regulatory risk concerns reinforce the 11 In April 2015, the Ant Group acquired Shumi platform, the mutual fund distribution license of which allowed the group to enter the platform business. According to the IPO prospectus of the Ant Group, by the end of June 2020, it had become the largest online investment services platform in the PRC by assets under management matched and distributed through its platform, which totaled CNY4.1 trillion. It has partnered with approximately 170 asset managers, including the vast majority of mutual fund companies as well as leading insurers, banks, and securities companies in the PRC. Such a wide partnership allows the Ant Group to offer more than 6,000 products through its platform, covering fixed income, equities, and balanced mutual funds. In 2021, the Asset Management Association of China started to publicize the non-money market mutual fund distribution size by each distribution channel—including banks, brokerage firms, and independent tech and FinTech platforms, among others. The Ant Group ranked first on the list at the end of the first quarter of 2021 with a distribution size of CNY890.1 billion. It maintained its top position and reached a distribution size of CNY1.20 trillion by September 2021, which coincides with the end of our sample period, when it surpassed the second channel (China Merchants Bank) by around 40%, and the third channel (Tiantian Fund Distribution) by over 100%.1 4. Data and Measures This study was conducted by the Ant Open Research Laboratory in a remote operating interface of Ant Group Environment.2 All data were sampled and desensitized, and analyzed by the Ant Open Research Laboratory. The laboratory is a sandbox environment where the authors remotely conduct empirical analysis and individual observations are invisible. The main regression variables include basic variables, investment variables, and consumption variables. We combine data from multiple sources. Our sample period spans from October 2019 to September 2021. We randomly select 200,000 retail investors, who have traded at least once during this 24-month sample period, from the online mutual fund distribution platform under the Ant Group (i.e., Ant Fortune). We add 1 Asset Management Association of China. Fund Sales Industry Data. https://www.amac.org.cn/researchstatistics/datastatistics/fundsalesindustrydata/. 2 See https://www.deor.org.cn/labstore/laboratory. 12 the trading requirement to assure that non-traders are excluded from this study. The number of unique non-money market funds traded by investors in this initial sample is 7,744, which is compatible with the Ant Group’s disclosure in its IPO prospectus that its mutual fund distribution platform offered more than 6,000 products to users at the end of June 2020. And according to the Asset Management Association of China, the average number of non-money market funds during our sample period is around 7,217. 3 Collectively, these statistics confirm that (i) the Ant Fortune covered almost the entire universe of mutual funds in the PRC at the time of our study, and (ii) our sample investors could trade a wide range of funds that are representative of the whole mutual fund market. To quantify a fund’s greenness, we utilize the E (environmental performance) scores provided by Wind, a widely used financial database in the PRC, which offers the most extensive coverage of E-scores for mutual funds. Our methodology results in a total of 3,087,120 trades during the sample period, encompassing 4,414 unique mutual funds that have been assigned E-scores by Wind. Among these funds, 3,053 are mixed funds, 817 are index funds, 543 are equity funds, and 1 is bond fund. (Bond funds typically do not receive E-scores as part of their evaluation.) In addition, we collect sample investors’ demographic information—such as age, gender, and location—from the Ant Fortune platform. For each sample investor, we also obtain her data from Ant Forest, including the monthly green energy points earned and redeemed for environmental protection initiatives (e.g., tree-planting). For each mutual fund traded by our sample investors during the sample period, we obtain additional information from Wind, including the fund’s monthly returns. 4.1 Green Energy Points For each retail investor i in month t, we obtain the total green energy points granted to them for participating in low-carbon activities. We take the logarithm of one plus the total 3 The number of non-money market funds was 5,898 at the end of September 2019 and 8,536 at the end of September 2021. The data can be found at https://www.amac.org.cn/researchstatistics/datastatistics/mutualfundindustrydata/. 13 green energy points earned by investor i in month t, and denote it as GreenPointit. We expect an investor with a higher GreenPointit to a have greater green preference. As mentioned in 2.1, there are 40 ways for individuals to earn green energy points during our sample period, which could be classified into five major categories. For each investor, we obtain not only her monthly total green energy points earned, but also the points earned under each of the five categories. In the robustness check, we construct an alternative measure for sample investors’ green preference, Cum_GreenPointit, which is the cumulative green energy points that investor i has earned from the beginning of the sample period (October 2019) to the end of month t. While GreenPointit captures an investor’s green preference conditional on her participation in daily eco-friendly activities in month t, Cum_GreenPointit is constructed conditional on her participation over a longer period. We also set a dummy variable, EarlyUseri,t, to be one for investor i if she has joined the Ant Forest program for a period longer than the sample median, and zero otherwise. 4.2 Retail Trading For each retail investor i in month t, we calculate her net purchase of each sample mutual fund, which equals the difference between her purchase and sales value of the fund scaled by the sum of the two values and expressed as a percentage: NetBuyi,j,t = (BuyValuei,j,t - SellValuei,j,t) / (BuyValuei,j,t + SellValuei,j,t) × 100%. (1) The variable NetBuyi,j,t conveys information about investor i’s active trading, and a higher value of the variable indicates investor i’s greater investment allocation to mutual fund j in month t. For the robustness checks, we construct two alternative measures to capture investor i’s active trading of fund j, NetBuy_Alt1i,j,t and NetBuy_Alt2i,j,t, using alternative denominators. For NetBuy_Alt1i,j,t, we replace the scaler in equation (1) with investor i’s holdings of fund j at the end of month t-1. The scaler for NetBuy_ Alt2i,j,t is the value of the total fund portfolio holdings of investor i at the end of month t-1. 14 4.3 Mutual Funds’ Environmental Performance For each mutual fund traded by sample investors, we obtain its E-score from Wind, which is released on a semiannual basis. The E-score specifically measures a fund’s environmental performance, with a higher value indicating better performance. It is plausible that retail investors are not sophisticated enough to have information on the specific value of the E-score of each fund they trade. Therefore, we set two dummies to capture the salient features of a fund’s environmental performance: E-Top20%j,t and EBot20%j,t. E-Top20%j,t (E-Bot20%j,t) equals one for fund j if its E-score is in the top (bottom) 20th percentile in the Wind fund universe, and zero otherwise. Although a retail investor may not know the exact value of a fund’s E-sore, she is likely to know whether a fund is positioned more toward green (E-Top20%j,t = 1) or brown (E-Bot20%j,t = 1) in the environmental performance spectrum. The abovementioned environmental performance measures are all constructed based on the environmental score assigned to each fund by Wind. However, it is possible that individual investors may not be aware of or pay sufficient attention to the environmental rating. We therefore propose an additional set of environmental performance measures by conducting textual analysis of the investment philosophy descriptions of each fund. This alternative methodology allows us to effectively capture qualitative information pertaining to the environmental performance of funds. The investment philosophy section of a fund introduces its investment targets, investment principles, and strategies used to achieve its investment goals. As the investment philosophy section summarizes the important aspects of a fund, in addition to its past performance which is described separately, it is prominently featured on the Ant Fortune app, positioned just below the introduction of the fund manager. The content is excerpted from the fundraising report. The position of this section on the app ensures easy accessibility for investors to read and comprehend. Based on the investment philosophy section of each mutual fund, we set a dummy variable, E-Fundj, which equals one for funds that have referenced the word “environment” 15 in this section, and zero otherwise. For each fund, we also tally the frequency of occurrences of “environment” in the investment philosophy section, and then scale it by the length of the section. This normalized count is denoted as E-Countj. Funds that discuss environmental issues in their investment philosophy section are arguably demonstrating a heightened focus on these issues. Relative to their counterparts, such funds are more likely to take investment targets’ environmental performance into consideration in constructing their portfolios. We therefore expect a non-zero E-Fundj and a higher E-Countj to be indicative of funds that are greener. To perform placebo tests, we also construct a variable S-Fundj (G-Fundj), which equals one if the word “social” (or “governance”) is mentioned in the fund’s investment philosophy section, and zero otherwise. Similarly, S-Countj (G-Countj) is constructed based on the scaled count of “social” (or “governance”) in the investment philosophy section. To further isolate the green feature of funds, we also identify green funds using a more stringent requirement. We set E ⊥ -Fundj to be one only for funds that mention “environment” exclusively in their investment philosophy section, without the mention of “social” or “governance.” For funds with a non-zero E ⊥ -Fundj, we further calculate the normalized frequency of occurrences of “environment” in its investment philosophy, in a similar spirit to E-Countj, which is denoted as E ⊥ -Countj. E ⊥ -Countj is automatically set to be zero for funds with a E ⊥ -Fundj that equals zero. By analogy, S ⊥ -Fundj, S ⊥ -Countj, G ⊥ -Fundj, and G ⊥ -Countj are constructed for placebo tests. 4.4 Physical Impacts of Climate Change Choi, Gao, and Jiang (2020) demonstrate that investors revise their beliefs about climate change upward and divest carbon-intensive stocks after experiencing warmer-than-usual temperatures. We follow their study to use abnormal local temperature to capture shocks to investors’ green preference brought by heightened physical climate concerns. We obtain city-month level historical temperature data from China Meteorological Administration, and follow Choi, Gao, and Jiang (2020) to construct an abnormal temperature measure, AbTmpi,t, which is the temperature of the city where investor i resides in month t minus the city’s average temperature in the same month of the year 16 over the past 10 years. 4 A higher AbTmpi,t is indicative of warmer-than-usual local temperature. 4.5 Regulatory Impacts of Climate Change We capture the effect of regulatory impact of climate change using the PRC’s proposition of its DCT. The existing literature has suggested that governmental environmental commitment is likely to raise investors’ climate-risk awareness. For instance, Seltzer, Starks, and Zhu (2022) and Bolton and Kacperczyk (2021) use the Paris Agreement as a shock to investors’ awareness about carbon risk, and show that the carbon premium increases following the Paris Agreement. In a similar spirit, we expect the PRC’s proposition of its DCT to boost Chinese investors’ climate-risk awareness. The DCT, the targets of achieving a carbon peak by 2030 and carbon neutrality by 2060, was initially proposed at the 75th UN General Assembly in September 2020. However, the detailed measures and action plans were unveiled in the annual plenary session of the National People’s Congress, the top legislature of the PRC, in March 2021. The figure below plots the monthly search volume for the term “carbon neutrality,” the ultimate goal outlined in the PRC’s DTC proposition, on the country’s dominant search engine, Baidu. The search volume for “carbon neutrality” was low when the DCT were initially proposed in September 2020 but surged in March 2021 when the detailed action plans were released. It suggests that the DCT caused much public attention only after the specific measures and action plans were announced. Accordingly, we set March 2021 to be the event month and set a dummy variable Postt to be equal to one for all months after the event month. 4 More information about the data can be found at https://data.cma.cn. 17 Search Volume for “Carbon Neutrality” on Baidu Notes: This figure plots the monthly search volume for “carbon neutrality”, the ultimate goal outlined in the People’s Republic of China’s (PRCO proposition of the Dual-Carbon Targets (DCT) on Baidu, the dominant search engine in the PRC. The search volume surges in March 2021, coinciding with the PRC's announcement of detailed measures and action plans for achieving DCT during the annual plenary session of the National People's Congress, the PRC's top legislative body. Source: Authors’ calculations. 4.6 Trading Performance For each investor i’s trading of fund j in month t, we assess her trading performance 1, 3, 6, and 12 months after the trade. Specifically, we calculate investor i’s net purchase, in value of fund j in month t, and multiply it by the fund’s return in the following periods:5 Profit1Mi,j,t = (BuyValuei,j,t, - SellValuei,j,t) × Retj,t-t+1, (2) where Profit1Mi,j,t is the profit that investor i could obtain one month after her trading of fund j in month t, BuyValuei,j,t and SellValuei,j,t are the values of the purchase and sale of 5 Due to the availability of only 2 years of data on individual investors’ holdings and transactions, we are unable to comprehensively track the dynamics of portfolio performance by considering the timing of buying and selling over an extended period. Therefore, we focus our analysis on the “buy and hold” performance of investors across various time horizons. While this approach provides valuable insights into the investors’ performance within the given timeframe, we acknowledge the limitations imposed by the data availability. 0 50000 100000 150000 200000 250000 300000 350000 400000 450000 18 the fund, respectively, and Retj,t-t+1 is the return of fund j at 1 month after t. We similarly calculate Profit3Mi,j,t, Profit6Mi,j,t, and Profit12Mi,j,t,, by replacing Retj,t-t+1 in equation (2) with Retj,t-t+3, Retj,t-t+6, and Retj,t-t+12, respectively, which represent the profits that the investor could obtain 3, 6, and 12 months after her trading of fund j in month t. We also assess investors’ trading performance based on the abnormal returns of funds being traded. Specifically, we replace Retj,t-t+n in equation (2) with the abnormal return of fund j, which is calculated as its raw return over the n-month period in excess of its sector return during the concurrent period. The abnormal return that investor i earns 1, 3, 6, and 12 months after trading fund j in month t is denoted as AbProfit1Mi,j,t, AbProfit3Mi,j,t, Profit6Mi,j,t, and Profit12Mi,j,t,,, respectively. In investigating the trading performance of investor i for fund j in month t, we control for her historical trading performance in regressions, which reflects her trading ability and might affect our results. The investor’s historical trading performance, or Investor_Profiti,t, is the cumulated profit that investor i has earned by trading mutual funds on the Ant Fortune platform from the beginning of the sample period until the end of month t. 4.7 Instruments Based on Local Subway Development We propose an instrumental variable for green energy points earned by an individual based on the development of a local subway network. It is natural to expect that an individual living in cities with more developed subway networks will be more aware of and more willing to adopt the low-carbon travel option, which constitutes an important part of her low-carbon lifestyle choices. In constructing the instrumental variables, SubwayStationi,t and SubwayKmsi,t, we calculate the ratios of the number of subway stations and the total mileages of subways, respectively, in the city where investor i resides at the end of month t, to the city’s population at the end of the previous year. The instruments SubwayStationi,t-1 and SubwayKmsi,t-1 are expected to be positively correlated with local individuals’ GreenPointsi,t and GreenTraveli,t. At the same time, these two instruments are less likely 19 to be directly associated with individuals’ green investment decisions, making them suitable instrumental variables for the alleviation of potential endogeneity concerns. 5. Empirical Results In this section, we examine whether an investor’s green energy points earned in the Ant Forest program, which is representative of her green preference, would affect her net purchase of mutual funds conditional on the funds’ environmental performance. Table 1 presents the summary statistics of our sample investors and funds. The average age of our sample investor is 33.9 years old, and 44.7% of them are female. The GreenPoints of sample retail investors averages 6.99 with a standard deviation of 1.93. The average of the environmental performance score (E-score) of sample mutual funds is 3.01 with a standard deviation of 1.98. Table 1: Summary Statistics Panel A: Main variables Variables No. Obs. Mean Std Retail trading NetBuy 3,087,120 40.8533 83.4425 NetBuy_Alt1 5,468,225 8.2217 0.9628 NetBuy_Alt2 6,204,409 4.2685 32.8742 Investors’ demographic information Female 3,087,120 0.4471 0.4971 Age 3,087,120 33.8928 9.3290 Green energy points’ variables GreenPoints 3,087,120 6.9947 1.9266 Cum_GreenPoints 3,087,120 9.5011 2.1836 GreenTravel 3,087,120 6.3534 2.8073 TravelReduction 3,087,120 2.3606 2.7929 P&PReduction 3,087,120 3.5030 1.8682 EnergySaving 3,087,120 0.0577 0.5272 Recycle 3,087,120 0.0552 0.6029 Mutual funds’ environmental performance E 3,087,120 3.0084 0.9766 E-Top20% 3,087,120 0.1518 0.3588 E-Bot20% 3,087,120 0.1872 0.3901 Other variables AbTmp 3,087,120 0.5350 1.2606 Profit-1M 3,087,120 –0.5240 1.2114 Continued on the next page 20 Panel A: Main variables Variables No. Obs. Mean Std Profit-3M 3,087,120 0.0096 1.8588 Profit-6M 3,087,120 0.0274 3.0279 Profit-12M 3,087,120 0.0429 4.1385 AbProfit-1M 3,087,120 –0.0046 1.0146 AbProfit-3M 3,087,120 –0.0045 1.4499 AbProfit-6M 3,087,120 –0.0079 2.3295 AbProfit-12M 3,087,120 –0.0017 2.9700 Investor_Profit 3,087,120 6.6780 28.8767 SubwayStation 3,087,120 126.6060 146.7259 SubwayKms 3,087,120 195.4650 247.5886 Polluted 3,081,662 0.5311 0.4990 EarlyUser 3,087,120 0.5185 0.4997 Panel B: GreenPoints of sample investors by subgroup By age No. Obs. Mean Std. Young (age below median) 1,685,465 7.0608 1.8020 Old (age above median) 1,401,655 6.9153 2.0636 By gender Male 1,706,861 7.1389 1.9139 Female 1,380,259 6.8163 1.9272 By the date of joining Ant Forest EarlyUser (joining date below median) 1,600,974 7.3482 1.4997 LateUser (joining date above median) 1,486,146 6.6140 2.2378 Notes: Panel A reports summary statistics of the main variables used in the empirical tests. Panel B presents the distribution of green energy points earned by sample investors conditional on age and gender. All variables are defined in the Appendix Table. Source: Authors’ calculations. 5.1 Baseline Tests We perform the baseline test at the investor-fund-year-month level, as specified below: NetBuyi,j,t = β0 + β1GreenPointsi,t-1 × Ej,t-1 + ∑i Investori + ∑j,t Fundj × Timet + ε i,j,t, (3) where retail investor i’s net purchase of fund j in month t is regressed on the interaction term between the green energy points that she earns in month t-1 and the E-score of fund j. We control for both investor and fund-year-month fixed effects in the tests to isolate the effects of individual heterogeneity or time-varying fund characteristics (e.g., fund returns and risks). The results are reported in Table 2. 27 Table 5: Influence of Climate Regulatory Risk Shock (1) (2) (3) NetBuyi,j,t NetBuyi,j,t NetBuyi,j,t GreenPoints i,t-1 × Ej,t-1 × Postt 0.3818*** (0.0000) GreenPoints i,t-1 × Ej,t-1 0.0254 (0.5819) GreenPoints i,t-1 × E-Top20%j,t-1×Postt 0.8266*** (0.0000) GreenPoints i,t-1 × E-Top20%j,t-1 -0.0141 (-0.9009) GreenPoints i,t-1 × E-Bot20%j,t-1 × Postt -0.3685*** (0.0073) GreenPoints i,t-1 × E-Bot20%j,t-1 0.0082 (0.9130) GreenPoints i,t-1 × Postt -0.4635** 0.6909*** 0.9114*** (0.0371) (0.0000) (0.0000) GreenPoints i,t-1 -2.1665*** -2.0910*** -2.0936*** (0.0000) (0.0000) (0.0000) Investor FE YES YES YES Fund × Time FE YES YES YES Adj. R2 0.1729 0.1729 0.1728 No. of Obs. 3,087,120 3,087,120 3,087,120 Notes: This table examines the influence of climate regulatory risk shock, captured by the PRC’s unveiling in Mach 2021 of its measures and action plans to achieve the Dual Carbon Targets (i.e., carbon peak by 2030 and carbon neutrality by 2060), on the translation of investors’ green preference into their trading for green funds. The variable Postt is set to one for periods after March 2021, and zero otherwise. The regressions are performed at the investor-fund-time level, with the dependent variable being investor i’s net purchase of fund j in month t, as defined in equation (1). GreenPoints i,t-1 refers to Ant Forest green energy points earned by investor i during month t-1, and Ej,t-1 is the environmental performance score (E-score) of fund j provided by Wind. The variables E-Top20%j,t-1 and E-Bot20%j,t-1 are dummies indicating whether fund j is in the top or bottom 20th percentile, respectively, of the fund universe in terms of its E-score. All variables are defined in the Appendix Table. Investor fixed effects and fund-time fixed effects are controlled. Standard errors are clustered at the fund level. The p-values are reported in parentheses. *, **, and *** represent significance at 10%, 5%, and 1% level, respectively. Source: Authors’ calculations. 5.4 Investment Performance In all previous tests, we have controlled fund-year-month fixed effects to isolate the influence of funds’ past performance on investors’ investment decisions, and green energy points still exhibit a strong and robust influence on investors’ tendency to invest in funds with a better environmental performance. The evidence supports our conjecture 28 that nonmonetary motives could be driving the connection between retail investors’ preference for green initiatives and their green investments. We perform additional tests to examine the investment performance of sample retail investors, conditional on their green preference and the environmental performance of the funds being traded. If these investors indeed invest in green funds for nonpecuniary motives, we should observe no significantly positive profits earned by these investors after trading green funds. Again, we perform investor-fund-month level regressions. For each fund j traded by investor i in month t, we calculate the profits the investor could earn 1, 3, 6, and 12 months after month t, which are denoted as Profit1Mi,j,t, Profit3Mi,j,t, Profit6Mi,j,t, and Profit12Mi,j,t,,, respectively. We also calculate the abnormal profit that could be earned by investors in a similar way—by only considering fund returns in excess of its sector returns. The abnormal profits measured 1, 3, 6, and 12 months after investor i’s trading of fund j in month t are denoted as AbProfit1Mi,j,t, AbProfit3Mi,j,t, Profit6Mi,j,t, and Profit12Mi,j,t,,, respectively. The calculation of these profit measures is specified in equation (2) and described in detail in section 4.6. We perform the following regression: Profiti,j,t = β0 + β1GreenPointi,t × Ej,t-1 + ∑i,t Investori × YearMontht + ∑j,t Fundj × Timet + ε i,j,t. (6) where the dependent variable could be one of the six profit measures mentioned above, and the coefficient of interest is β1. A significantly positive β1 would suggest that investor i is rewarded financially by investing in fund j, conditional on its environmental performance, and vice versa. The results are reported in Table 6. Panels A and B of the table use post-trading profit and post-trading abnormal profit as the dependent variable, respectively. The results show that no matter which post-trading interval is examined or how trading profit is measured, retail investors do not earn a significantly positive profit for investing in funds with a better environmental performance. The results render further support to our conjecture that our green preference measure constructed based on retail investors’ Ant 29 Forest green energy points is nonpecuniary in nature and that it is linked to investors’ green investment behavior for nonfinancial motives. Table 6: Investment Performance Panel A (1) (2) (3) (4) Profit1Mi,j,t Profit3Mi,j,t Profit6Mi,j,t Profit12Mi,j,t GreenPoints i,t-1 ×Ej,t-1 0.0010 0.0006 –0.0010 0.0001 (0.2323) (0.5419) (0.4882) (0.9745) GreenPoints i,t-1 –0.0042* –0.0072** –0.0063 –0.0158*** (0.0719) (0.0259) (0.1720) (0.0020) Investor_Profit i,t –0.0026*** –0.0078*** –0.0062** –0.0061** (0.0077) (0.0002) (0.0472) (0.0498) Investor FE YES YES YES YES Fund × Time FE YES YES YES YES Adj. R2 –0.0069 –0.0157 –0.0239 –0.0184 No. of Obs. 3,087,120 3,087,120 3,087,120 3,087,120 Panel B (1) (2) (3) (4) AbProfit1Mi,j,t AbProfit3Mi,j,t AbProfit6Mi,j,t AbProfit12Mi,j,t GreenPoints i,t-1 ×Ej,t-1 0.0006 –0.0004 –0.0016 0.0010 (0.3913) (0.5611) (0.1043) (0.1834) GreenPoints i,t-1 –0.0023 –0.0005 0.0026 –0.0042 (0.2425) (0.8265) (0.3944) (0.9880) Investor_Profit i,t –0.0013* –0.0016 0.0002 0.0011 (0.0847) (0.1401) (0.9336) (0.4427) Investor FE YES YES YES YES Fund × Time FE YES YES YES YES Adj. R2 –0.0104 –0.0241 –0.0303 –0.0283 No. of Obs. 3,087,120 3,087,120 3,087,120 3,087,120 Notes: This table assesses the post-trading performance of investors conditional on their green preference captured by Ant Forest green energy points GreenPoints i,t-1, and the environmental performance of funds being traded captured by the E-score Ej,t-1 . The regressions are performed at the investor-fund-time level. The dependent variable Profit1Mi,j,t in column (1) of Panel A is the profit that investor i could obtain one month after her trading of fund j in month t, as specified in equation (2). Similarly, we assess the profits she could obtain 3, 6, and 12 months after the trading, which are used as the dependent variables in columns (2)–(4), respectively. In Panel B, the dependent variables are the abnormal profits that investor i could obtain 1, 3, 6, and 12 months after trading fund j in month t, where the fund’s abnormal return is calculated relative to the concurrent average return of funds in its sector. The variable Ej,t-1 is the most recently available environmental performance score (E-score) of fund j provided by Wind. Investor_Profit i,t is the cumulated profit that investor i has earned through trading mutual funds on the Ant Fortune platform by the end of month t. All variables are defined in the Appendix Table. Investor fixed effects and fund-time fixed effects are controlled. Standard errors are clustered at the fund level. The p-values are reported in parentheses. *, **, and *** represent significance at the 10%, 5%, and 1% levels, respectively. Source: Authors’ calculations. 30 5.5 Addressing Endogeneity As discussed in section 4.7, subway development is closely related with local residents’ green travel tendency, yet is less likely to directly affect their investment decisions toward green portfolios. It is thus a suitable instrumental variable for local individuals’ green energy points. We use SubwayStationi,t and SubwayKmsi,t, which are the scaled city-level number of subway stations and subway mileage, respectively, as instruments for local investors’ green energy points, GreenPointsi,t, and their points specifically earned for green travel, GreenTraveli,t. We estimate the following two-stage equations: GreenPointsi,t (GreenTraveli,t) = β0 + β1SubwayStationi,t-1 (SubwayKmsi,t-1)+ ∑I Investori + ∑t Timet + ε i,t, (7) NetBuyi,j,t = β0 + β1Pre (GreenPointsi,t-1 (GreenTraveli,t-1)) × Ej,t-1 + ∑i Investori + ∑j,t Fundj × Timet + ε i,j,t, (8) where Pre (GreenPointi,t-1) and Pre (GreenTraveli,t-1) in equation (8) is the predicted value of investor i’s GreenPoints and GreenTravel in month t-1 estimated using equation (7). Table 7 reports the results of the two-stage regressions. Table 7: Local Subway Development as Instrumental Variables Panel A: 1st stage (1) Greenpointi,t (2) Greenpointi,t (3) GreenTraveli,t (4) GreenTraveli,t SubwayStationsi,t-1 0.0021*** (0.0000) SubwayKmsi,t-1 0.0012*** (0.0001) SubwayStationsi,t-1 0.0032*** (0.0000) SubwayKmsi,t-1 0.0016*** (0.0003) Investor FE YES YES YES YES Time FE YES YES YES YES F statistic 23.065 14.825 26.446 13.038 Adj. R2 0.7233 0.7233 0.7586 0.7586 No. of Obs. 3,087,120 3,087,120 3,087,120 3,087,120 Continued on the next page 31 Panel B: 2nd stage IV = Stations IV = KMs IV = Stations IV = KMs (1) NetBuyi,j,t (2) NetBuyi,j,t (3) NetBuyi,j,t (4) NetBuyi,j,t Pre (GreenPoints i,t-1 ) × Ej,t-1 6.9409*** 7.5156*** (0.0020) (0.0029) Pre (GreenPoints i,t-1) –21.610 –9.3446*** (0.2450) (0.6842) Pre (GreenTravel i,t-1 ) × Ej,t-1 4.5356*** 5.5920*** (0.0020) (0.0029) Pre (GreenTravel i,t-1) –14.121 –6.9529 (0.2450) (0.6842) Investor FE YES YES YES YES Fund × Time FE YES YES YES YES Adj. R2 0.1723 0.1723 0.1723 0.1723 No. of Obs. 3,087,120 3,087,120 3,087,120 3,087,120 Notes: This table presents the findings of tests conducted using instrumental variables. The two instrumental variables are SubwayStationsi,t-1 and SubwayKmsi,t-1, which represent the number of subway stations and the total mileages of subways at the end of month t-1 in the city where investor i resides, scaled by the city’s population at the end of the previous year, respectively. Greenpointi,t denotes the total green energy points that investor i earns in the Ant Forest program in month t, and GreenTraveli,t is the green energy points she earns specifically for green travel, one of the five categories of daily low-carbon lifestyle choices in the Ant Forest program. Panels A and B report results of the firstand second-stage tests, respectively. In Panel B, the regressions are performed at the investor-fund-time level, with the dependent variable being investor i’s net purchase of fund j in month t, as defined in equation (1). Pre (GreenPoints i,t1) in columns (1) and (2) is the predicted value from columns (1) and (2) in Panel A, respectively. Pre (GreenTravel i,t-1) in columns (3) and (4) is the predicted value from columns (3) and (4) in Panel A, respectively. The variable Ej,t-1 is the most recently available environmental performance score (E-score) of fund j provided by Wind. All variables are defined in the Appendix Table. Investor fixed effects and fundtime fixed effects are controlled. Standard errors are clustered at the fund level. The p-values are reported in parentheses. *, **, and *** represent significance at the 10%, 5%, and 1% levels, respectively. Source: Authors’ calculations. Panel A shows the results of the first-stage regression. Across all four columns, the instrumental variables SubwayStationsi,t-1 and SubwayKmsi,t-1 are significantly and positively associated with GreenPointsi,t or GreenTraveli,t at the 1% level. The result in column (1) suggests that a one standard deviation increase in SubwayStationi,t-1 would contribute to a 4.4% increase in GreenPointsi,t relative to the sample mean (6.99). The F-statistics across all four columns are greater than 10, confirming that the number of local subway stations and subway mileage are not considered weak instruments for GreenPoints and GreenTravel of local investors. 32 Panel B reports the results of the second-stage regression. The coefficient on the interaction term between retail investors’ instrumented GreenPoints or GreenTravel and fund environmental performance is significantly positive at the 1% level across all four columns. The results render further support to our conjecture that retail investors’ green preference, revealed through their participation in daily eco-friendly activities, plays a role in shaping their investment preference. 6. Conclusions Our research builds on the success of the Ant Forest program, a well-known green initiative in the PRC. By using a new measure that reflects how individuals strengthen their nonpecuniary preferences with regard to environmental issues, we map this measure to their investment decisions. This approach provides compelling evidence of the nonfinancial drivers behind green investing. To address the challenge of endogeneity, we use the development of a local subway network as an instrumental variable, which causes variations in the accumulation of green energy points through eco-friendly travel. Our findings emphasize the strong impact of nonfinancial incentives on green investment decisions, including how individual concerns over the physical and regulatory risks of climate change further strengthen these incentives. Although individuals may update their nonmonetary green preferences, our research suggests that these updates may not necessarily generate financial gains from trading. 33 Appendix Table: Variable Definitions Variables Definitions Retail trading NetBuy i,j,t The difference between investor i’s purchase and sales value of fund j in month t scaled by the sum of the two values. The variable is defined in equation (1) and expressed in percentage. NetBuy_Alt1 i,j,t The difference between investor i’s purchase and sales value of fund j in month t scaled by the value of fund j in her portfolio at the end of month t-1 . The variable is expressed as a percentage. NetBuy_Alt2 i,j,t The difference between investor i’s purchase and sales value of fund j in month t scaled by the total value of her fund portfolio at the end of month t-1. The variable is expressed as a percentage. Retail investors’ demographic information Young i An indicator that equals one for retail investors aged below the sample median (32 years) at the beginning of the sample period, and zero otherwise. Age i Retail investors’ age at the beginning of the sample period. Female i An indicator that equals one for female investors and zero for male investors. Ant Forest program-related variables GreenPoint i,t The logarithm of one plus the total green energy points earned by investor i in month t in the Ant Forest program. Cum_GreenPoint i,t The logarithm of one plus the cumulative green energy earned by investor i from the beginning of the sample period until the end of month t. GreenTravel i,t The logarithm of one plus the green energy earned by investor i under the “green travel” category in the Ant Forest program, though activities such as traveling by public transportation and walking. TravelReduction i,t The logarithm of one plus the green energy points earned by investor i under the “travel reduction” category in the Ant Forest program, through activities such as using online services. P&PReduction i,t The logarithm of one plus the green energy points earned by investor i under the “paper and plastic reduction” category in the Ant Forest program, through activities such as requiring no single-use cutlery for food-delivery services and requiring electronic receipts instead of printed copies. EnergySaving i,t The logarithm of one plus the green energy points earned by investor i under the “energy saving” category in the Ant Forest program, through activities such as purchasing energyefficient appliances. Recycle i,t The logarithm of one plus the green energy points earned by investor i under the “recycle” category in the Ant Forest Continued on the next page 34 Variables Definitions program, through activities such as recycling used clothes, cell phones, and appliances. EarlyUser i,t An indicator that equals one if, by the end of month t, investor i has joined the Ant Forest program for a period longer than the sample median, and zero otherwise. Mutual funds’ environmental performance based on ratings E j,t The E-score, issued by Wind, of fund j, conditional on its environmental performance. The score is updated on a semiannual basis. E-Top20% j,t A dummy variable that equals one if the E-score of fund j is in the top 20th percentile of the fund universe with Wind Escores, and zero otherwise. E-Bot20% j,t A dummy variable that equals one if the E-score of fund j is in the bottom 20th percentile of the fund universe with Wind Escores, and zero otherwise. Mutual funds’ nonfinancial performance based on textual analysis E-Fund j An indicator that equals one for fund j if it mentions “environment” in its investment philosophy, which is excerpted from its fundraising report and displayed in the Ant Fortune app, and zero otherwise. E-Count j The number of times that “environment” is mentioned in the investment philosophy section of fund j displayed in the Ant Fortune app, scaled by the length of the section. S-Fund j An indicator that equals one for fund j if it mentions “social” in its investment philosophy section displayed in the Ant Fortune app, and zero otherwise. G-Fund j An indicator that equals one for fund j if it mentions “governance” in its investment philosophy section displayed in the Ant Fortune app, and zero otherwise. S-Countj / G-Countj The construction is analogous to that of E-Countj. E ⊥ -Fund j An indicator that equals one for fund j if it mentions “environment” in its investment philosophy section displayed in the Ant Fortune app but does not mention “social” or “governance,” and zero otherwise. E ⊥ -Count j The number of times that “environment” is mentioned in the investment philosophy section of fund j, scaled by the length of the section, conditional on E ⊥ -Fundj not being zero. E ⊥ - Countj is set zero if E ⊥ -Fundj equals zero. S ⊥ /G ⊥ -Fundj The construction is analogous to that of E ⊥ -Fundj. S ⊥ /G ⊥ - Countj The construction is analogous to that of E ⊥ -Countj. Physicaland regulatory-shock-related variables AbTmp i,t The abnormal temperature measured following Choi, Gao, and Jiang (2020), which equals the temperature of the city where investor i resides in month t minus the city’s average temperature in the same month of the year over the past 10 years. Post t A dummy variable that equals one for months after March 2021, and zero otherwise. In March 2021, the PRC unveiled Continued on the next page 35 Variables Definitions its measures and action plans to achieve the Dual Carbon Targets in the annual plenary session of the National People's Congress, the top legislature in the PRC. Post-trading performance Profit1M i,j,t / Profit3Mi,j,t /Profit6Mi,j,t / Profit12Mi,j,t Profit1M i,j,t is the profit that investor i could obtain 1 month after her trading of fund j in month t. It is calculated by multiplying investor i’s net purchase of fund j (unscaled) in month t by fund return 1-month after t, as specified in equation (2). By analogy, we also construct performance measures for 3, 6, and 12 months after the trading, which are denoted as Profit 3M i,j,t, Profit 6M i,j,t, and Profit 12M i,j,t, respectively. AbProfit1M i,j,t /AbPro fit3Mi,j,t /AbProfit6Mi,j,t/AbPr ofit12Mi,j,t AbProfit1M i,j,t is the abnormal profit that investor i could obtain one month after her trading of fund j in month t. It is calculated by multiplying investor i’s net purchase of fund j (unscaled) in month t by the abnormal return of fund j 1 month after t. The abnormal return of fund j is its return over the average return of funds in its sector during the concurrent period. By analogy, we also construct abnormal performance measures for 3, 6, and 12 months after the trading, which are denoted as AbProfit 3M i,j,t, AbProfit 6M i,j,t, and AbProfit 12M i,j,t, respectively. Investor_Profit i,t The cumulated profit that investor i has earned through trading mutual funds on the Ant Fortune platform from the beginning of the sample period until the end of month t. City characteristic Polluted i A dummy variable that equals one for the city where investor i lives if its average monthly PM2.5 concentration is above the median of all cities in the PRC in 2018, and zero otherwise. Instrumental variables SubwayStations i,t The number of subway stations of the city where retail investor i lives in month t, scaled by the city’s resident population at the end of the previous year. SubwayKms i,t The total mileage of subways in the city where retail investor i lives in month t, scaled by the city’s resident population at the end of the previous year. PRC = People’s Republic of China. Source: Authors’ compilation. 36 REFERENCES Ardia, David, Keven Bluteau, Kris Boudt, and Koen Inghelbrecht. 2022. “Climate Change Concerns and the Performance of Green vs. Brown Stocks.” Management Science. https://doi.org/10.1287/mnsc.2022.4636. Alekseev, Georgij, Stefano Giglio, Quinn Maingi, Julia Selgrad, and Johannes Stroebel. 2022. ”A Quantity-Based Approach to Constructing Climate Risk Hedge Portfolios.” National Bureau of Economic Research (NBER) Working Paper No. w30703. Alok, Shashwat, Nitin Kumar, and Russ Wermers. 2020. “Do Fund Managers Misestimate Climatic Disaster Risk.” The Review of Financial Studies 33 (3): 1146–83. Anderson, Anders, and David T. Robinson. 2019. “Climate Fears and the Demand for Green Investment.” Swedish House of Finance Research Paper 19-14. _____. 2022. “Financial Literacy in the Age of Green Investment.” Review of Finance 26 (6): 1551–84. Bauer, Rob, Tobias Ruof, and Paul Smeets. 2021. “Get Real! Individuals Prefer More Sustainable Investments.” The Review of Financial Studies 34 (8): 3976–4043. Bolton, Patrick, and Marcin Kacperczyk. 2021. “Do Investors Care about Carbon Risk?” Journal of Financial Economics 142 (2): 517–49. Boermans, Martijin A., and Rients Galema. 2019. “Are Pension Funds Actively Decarbonizing their Portfolios?” Ecological Economics 161 (C): 50–60. Choi, Darwin, Zhenyu Tao, and Wenxi Jiang. 2020. “Attention to Global Warming.” The Review of Financial Studies 33 (3): 1112–45. Choi, Darwin, Zhenyu Tao, Wenxi Jiang, and Hulai Zhang. 2023. Carbon Stock Devaluation. Available at SSRN 3589952. Engle, Robert, Stefano Giglio, Bryan Kelly, Heebum Lee, and Johannes Stroebel. 2020. “Hedging Climate Change News.” The Review of Financial Studies 33 (3): 1184– 216. Giglio, Stefano, Mattero Maggiori, Johannes Stroebel, Zhenhao Tan, Stephen Utkus, and Xiao Xu. 2023. ”Four Facts about ESG Beliefs and Investor Portfolios.” National Bureau of Economic Research No. w31114. Gibson Brandon, Rajna, Philipp Krueger, and Peter Schmidt. 2021. “ESG Rating Disagreement and Stock Returns.” Financial Analysts Journal 77 (4): 104–27.