2020年摩根大通亚太区宏观量化虚拟会议要点.docx

《2020年摩根大通亚太区宏观量化虚拟会议要点.docx》由会员分享，可在线阅读，更多相关《2020年摩根大通亚太区宏观量化虚拟会议要点.docx（27页珍藏版）》请在淘文阁 - 分享文档赚钱的网站上搜索。

1、Global Quantitative and Derivatives Strategy21 August 2020APAC Quant & Derivative Strategy Robert Smith, PhD AC(61-2) 9003-8808Bloomberg JPMA RSMITH J.P. Morgan Securities Australia LimitedBerowne Hlavaty AC(61-2) 9003-8602 Bloomberg JPMA HLAVATY J.P. Morgan Securities Australia LimitedEvan Hu AC(85

2、2) 2800-8508J.P. Morgan Securities (Asia Pacific) LimitedMixo Das AC(852) 2800-0511Bloomberg JPMA MDAS J.P. Morgan Securities (Asia Pacific) Limited/J.P. Morgan Broking (Hong Kong) LimitedAda Lau AC(852) 2800-7618J.P. Morgan Securities (Asia Pacific) Limited/ J.P. Morgan Broking (Hong Kong) LimitedH

3、aoshun Liu AC(852) 2800-7736Bloomberg JPMA HLIU J.P. Morgan Securities (Asia Pacific) Limited/ J.P. Morgan Broking (Hong Kong) LimitedYukun Zhang AC(852) 2800-5148J.P. Morgan Securities (Asia Pacific) Limited/ J.P. Morgan Broking (Hong Kong) LimitedGlobal QDSMarko Kolanovic, PhD(1-212) 622-3677J.P.

4、Morgan Securities LLCDubravko Lakos-Bujas(1-212) 622-3601J.P. Morgan Securities LLCJ. R MorganJ.P.Morgan APAC Macro QuantVirtual Conference 2020Summary of ProceedingsOur 22nd Quant Conference was held on 30 July 2020 and was part of our global Macro Quantitative Conference series; this time virtual

5、for Asia Pac. It brought together practitioners and investors focused on risk premia investing and machine learning/alternative data. In attendance were 343 delegates from 170 firms across 11 different countries. In this report, we summarize the topics and key insights from the conference.We hosted

6、two fireside chats as well as presentations from practitioners and academics who are all leaders in their fields. Topics included aspects of risk premia strategies describing their rationale on construction, past reflections and expected future performance. The use of machine learning in both tradit

7、ional and new alternative datasets was also covered.Once again we also took a live survey of attendees on their expectations of risk premia and use of alternative data. We asked about expectations for the economic recovery, which most respondents (52%) thought would be W9 shaped. That was supported

8、with a bullish outlook for the S&P500 to be at levels exceeding 3,400 selected by more than 29% of respondents, and with 28% expecting it to at least remain around current levels. Yield on the US10 Year was thought stay low to negative (68%).The outlook on Big Data/Machine Learning is positive with

9、most seeing it as an opportunity to enhance existing quant strategies (45%). That said, 21% of respondents have no plan to use it extensively. The most common number of new alt data sets evaluated was 1-3, with quite a few having not looked at any. Of those data sets looked at most respondents (46%)

10、 had yet to find any that yield alpha. The results would suggest some frustration in the effort to make alternative data work*. However, there were a handful of respondents (4%) that had found more than 50 that yielded alpha.In recent years, investors have paid increased attention to Alternate Risk

11、Premia as a source of returns uncorrelated with conventional equity and bond risk premia. In addition, the application of Big Data, Machine Learning and Artificial Intelligence to risk premia investing remains a heavily discussed topic. At J.P. Morgan, we have published extensively on these topics,

12、including detailed guides on Cross-Asset Systematic Strategies, Cross-Asset Momentum and Equity Risk Premia Strategies; a primer on Big Data and AI Strategies: Machine Learning and Alternative Data; US Factor Reference Book and recently: COVID-19 Composite, 2019 Alternative Data Handbook, Automated

13、Machine Learning, The Value Conundrum, Cross Asset Style Timing, Defensive Risk Premia, and The quest for pu】e equity factor exposure.See page 24 for analyst certification and important disclosures, including non-US analyst disclosures.J.P. Morgan does and seeks to do business with companies covered

14、 in its research reports. As a result, investors should be aware that the firm may have a conflict of interest that could affect the objectivity of this report. Investors should consider this report as only a single factor in making their investment decision.Our Key Takeaways Alphas in China A-share

15、s market are higher than those in developed markets. High hedging costs and funding costs, regulatory restrictions and increasing money inflows are some of the hurdles for quant hedge funds in China to maintain a high alpha of over 25% as in the past.Speaker BiographyWith more than 19 years of inves

16、tment experience, Mr Qiu has served as portfolio manager in top investment banks and hedge funds such as UBS, Credit Suisse, Deutsche Bank and Millennium. Mr Qiu has extensive experience in high frequency trading, statistical arbitrage and quant macro strategies. Mr Qiu received an advanced degree i

17、n physics from the University of Pennsylvania and a bachelors degree in physics from Fudan University. Mr Qiu oversees the investment process as well as the general management of the firm.A Framework for Downside ProtectionCharles Wu5 Deputy CIO and GM DC Investments, State SuperSummaryThe speakers

18、talk focused on a framework for downside protection, utilizing a combination of Portfolio Hedging and Portfolio Design. The particular situation faced by State Super (NSW Australia) is of net outflows of 3% p.a. to meet forecast superannuation (pension) payments. Mr Wu suggested that a positive cash

19、 flow portfolio can consider the right tails (shooting for the stars) but he flagged that it doesnt work so well if you have negative cash flow, as there is a clear switch in the size and shape of the return distribution curves with a fat left tail under the outflow model. Mr Wu stated that the left

20、 shoulder (-5% to -15%) occurs more frequently than one might expect - about 10% of the time, but this isnt enough to form the base case of portfolio construction. He also stated that losses typically occur after increased correlations and Betas of a 70:30 portfolio. The speaker then introduced the

21、holistic framework for downside protection, broken down into fund level and asset class level investments. At the Fund Level this includes; dynamic asset allocation (DAA/TAA), Diversification, Derivatives (protection). The Asset Class covers; Liquid Growth (low vol, low beta equity), Property, Infra

22、structure (yield driven), Liquid Defensives (duration), Alternatives & Currency. Importantly, each asset classs expected value of downside protection in bps are estimated for a 10% draw-down for the next month. When required, this is compared to actual performance, such as the during Feb/2020 drawdo

23、wn. Defensiveness is built into the portfolio using an overlay (portfolio composition and construction as well as statistical) along with more statistical approaches of asset allocation and active management. Contractual (Derivatives) hedging can add value during down-turns, but is expensive to keep

24、 in-place. So one needs to balance this cost with participation in the upside by including some leverage to increase exposure (covering say 2/3 of the costs) can be achieved according to Mr Wu. Portfolio Design is also a critical aspect of downside protection. Diversification for a fund of fund mana

25、ger is tested to ensure managers realized performance clusters match expectations. If, for example a Fixed Income fund is clustered with Equities you have to question their risk profile and inclusion in the portfolio would be reviewed, according to Mr Wu.Conclusions Once the downside is managed, the

26、 upside will sort itself out. Contractual hedging is effective but expensive so make the portfolio at least break even. Building a diversified portfolio is essential, as is understanding the drivers of the statistical hedging.Q&AQ: In this approach/framework, would you invest OR divest OR neither im

27、mediately after a sudden equity market correction accompanied by a vol spike? A: We dont use our framework in such a fashion, although there is scope for this. We do have a valuation and sentiment framework which influences our risk level, which we used to rotate to defensive just before the pandemi

28、c. During the pandemic our sentiment index was in fear9 mode so we de-risked further, but then missed some of the rebound. Q: Do you make wholly systematic changes to your portfolio or do you adjust with an overlay?A: We use input from underlying managers and a systematic rigor in how we approach th

29、e portfolios, with discretion overlay added after. Q: How often do you run the dendrogram and affinity charts for your risk monitoring? Is it weekly or monthly?A: It depends on the data series. On market data we use daily data and test on a weekly basis. Relationships change such as a period when th

30、e USD correlated with equity markets and it is important to observe and understand these market dynamics and regime shifts. Other data such as portfolio performance is analyzed monthly, matching the frequency of input data. Q: Have you considered using a fully latent factor decomposition on your por

31、tfolio?A: Yes, but it comes with challenges. You have to be aware that the data and holdings are not as simple or pure as listed equities, such as airports. Then the question becomes what you do with the output, because it is not so simple to trade an airport, for example. We continue to do research

32、 in this space. Q: A question on ML, do you use clustering on any other areas of research or do you use other ML techniques?A: We believe that data and data-science will give an edge relative to other market participants. We use models to forecast market direction, supervised and un-supervised techn

33、iques. We use reinforcement learning to analyze what was driving the market (in the past) and gives us a better understanding to use going forward. We use machine learning and data science to empower our decision making processes.Our Key Takeaways Downside protection is important, but especially so

34、for portfolios in draw-down. Diversification combined with leverage and derivatives need to be applied along with TAA and overlays to effectively manage a portfolio across asset classes.Speaker BiographyMr Wu joined State Super in 2015. In his role as Deputy Chief Investment Officer and General Mana

35、ger, Defined Contribution Investments, Mr Wu is responsible for formulating investment strategy to assist members to achieve return objectives on a risk-adjusted basis. He was previously an Investment Manager at Media Super and an analyst at Mercer. Mr Wu holds a Master of Commerce and a Bachelor of

36、 Computer Engineering and is a Chartered Financial Analyst holder. He also serves as Vice President and the Director of University Outreach at Chartered Financial Analyst Society Sydney.Drivers of the Disconnect Between Stock Markets and the EconomyProf. Talis J. Putnins, University of Technology Sy

37、dneySummaryAfter the longest bull market on record (12 years), global stock markets crashed mid-March and have largely recovered by mid-July. The speaker shared some news quotes reporting how stock markets are ignoring or disconnected from the real economy. Prof. Putnins suggested that there are thr

38、ee main drivers of the disconnect: macro vs micro efficiency, behavioral biases and Fed style intervention. Prof. Putnins suggests that at the micro-level markets are increasingly efficient at relative stock pricing, however the macro-level absolute prices are actually becoming less efficient. The r

39、eduction in macro-level informativeness was argued to be due to (i) the rise of passive, (ii) rise of delegated management, (iii) tight benchmarking and 60/40 mandates. This implies that market-wide valuations are now less closely linked to future earnings/output. The speaker suggests this is why fu

40、nd managers struggle to beat the market (high micro-efficiency for relative prices reduces alpha) yet the market seems disconnected from the economy (low macro-efficiency of absolute prices). While it might be difficult to gain from stock picking, he suggested that value can be added by strategic, d

41、ynamic asset allocation. Talis then explored the second factor was behavioral, The Post-Traumatic Stress Disorder” (PTSD) of markets. There is a perceptual bias following shift away from a strong stimulus until perceptions catch up with reality. Empirical evidence confirms that market participants u

42、nder-estimate risk immediately after a spike in volatility, leading to an over-estimate of asset values until perception catches up with reality. Prof. Putnins suggested we have moved away from free markets to Ted Markets7, citing a few quotes: US Federal Reserve: Well do all it takes to stop this!”

43、 and RBAs Philip Lowe: We too will transact in whatever quantities are necessary to achieve this objective”. The most aggressive unconventional monetary policy“ on record is being enacted by the Central Banks: Fed, ECB, RBA, BoJ, BoE which have become the biggest market participants now. For example

44、 the Fed has purchased $3Tn in assets in 3 months.ConclusionsMarkets efficiently price stocks relative to one another, but show inefficiency in market-wide valuations. The situation is becoming worse with the rise of delegation, tight benchmarking, and passive funds. Market-wide movements have becom

45、e less related to future economic conditions. Behavioral biases mean that risk is underestimated following extreme volatility, leading to excessive optimism, inflated valuations, and market bounces until perceptions converge to reality. Central banks are now playing a major role in setting overall m

46、arket price levels, driving markets up precisely when the economic outlook is dire, creating a disconnect between markets from the underlying economy,. but balance sheet expansion cannot continue indefinitely!Q&AQ: Another contributor for the wedge getting worse (slide 11) might also be rise of fact

47、or investing. Could you comment on that?A: Overall market wide price in formativeness and efficiency was studied, and the cheaper forms of investing such as smart-beta investing would be wrapped up in that broader trend of getting out of more expensive active investments that exploit the cross-secti

48、on of stocks, but we dont have any hard evidence specifically on that. Q: Re 2020 perceptual bias, wasnt implied volatility considerably above realized volatility in May, June? So, is that period “excess complacency (based on valuations) or “excess fear, (based on measured implied/realized volatilit

49、y)? A: I haven91 explored the recent realized volatility to implied but the extreme levels in March and reduced levels recently are still elevated which is when this perceptual bias exists. Now looking at realized and implied volatilities might be due to excessive fear at the height of the volatility period. Also remembe