A Summary of Published and Forthcoming Papers
Major Academic Contributions
 Machine Learning/Big Data/Bitcoin/Textual Analysis :

Rapach, Strauss and Zhou (2013) is perhaps the first academic study (published in a top finance journal)
that applies LASSO (a popular machine learning tool) in finance. They use it to select predictors from
a large set of candidates, "bigdata", for forecasting stock markets globally.
 Jiang, Tang and Zhou (2018) study firm characteristics in China, via “bigdata” methods such as PCA and PLS,
and find that trading frictions, momentum, and profitability are the most effective predictors of future stock returns.
 Rapach, Strauss Tu, and Zhou (2019), superseding an early 2014 version,
apply Lasso to forecast industry returns
with their lagged values, and find information diffusion may be responsible for the significant predictability.
 Jiang, Lee, Martin and Zhou (2019) apply textual analysis to construct a manager sentiment index from firm tone of conference calls and financial statements ("bigdata"). They find that manager sentiment drives both firm aggregate investments and market returns.
 Rapach and Zhou (2020) apply the new cross sectional ML methods of Han, He, Rapach and Zhou (2020) to time series forecasting, and find they deliver the best empirical results compared with the currently most effective ones.
 Detzel, Liu, Strauss, Zhou and Zhu (2021) propose perhaps the first equilibrium model that shows how technical analysis can arise endogenously via rational learning, providing
a theoretical foundation of using technical analysis in practice.
They document that ratios of prices to their moving averages forecast daily Bitcoin returns in and outof sample, and similar results hold for smallcap, youngfirm, and lowanalystcoverage stocks as well as NASDAQ stocks during the dotcom era.
 Huang, Jiang, Li, Tong and Zhou (2021) propose a sPCA, a new modification of the popular principal component analysis (PCA) by scaling each predictor
with its predictive slope on the target to be forecasted to increase the predictive power of the extracted factor.
 Dong, Li, Rapach and Zhou (2022) show, via machine learning methods, that
cross section anomalies are important predictors of the US stock market returns, linking crosssection and time series predictability
and useful to index managers who even do not trade firm factors.
 He, Huang, Li and Zhou (2022)
provide a reducedrank approach (RRA) to extract a few factors from a large set of factor proxies
to explain expected returns, outperforming PCA, PLS and LASSO.
 Guo, Lin, Wu and Zhou (2022) find, via both traditional and machine learning tools, strong evidence of crosssectional predictability of corporate bonds by yield predictors that capture the info of past 1, 3, and up to 48 months.
 Chen, Guo and Zhou (2022) identify, via machine learning tools,
firm fundamentals that explain the shape of the option implied volatility curve, providing insights into
some puzzling patterns in the options literature.
 Rapach and Zhou (2022) provide a survey of studies on time series predictability, with various recent
methods including machinelearningbased ones.

Huang, Jiang, Li, Tong and Zhou (2023) propose a scaled sufficient forecasting (sSUFF) method to account for the nonlinearity and weak
factors, showing that bond returns are predictable by realtime macro variables.
 He and Zhou (2023) is perhaps the first study
on the predictability of pricing errors, which provides a powerful test for asset pricing models, rejecting
all well known models such as FamaFrench 5factor model and those recently developed based on machine learning.
 Zaffaroni and Zhou (2023) provide a survey of studies on cross sectional predictability, focusing on firm predictors and useful for stock selections.
 Han, He, Rapach and Zhou (2024) provide a new MLbased FM regression (a dynamic average of short and longshort predictions) which is easy and fast for implementation, interpretable and competitive to the
best ML methods.
 CAPM, APT and Factor models :
 Harvey and Zhou (1990) provide Bayesian multivariate tests of the CAPM (The Capital Asset Pricing Model) and find the probability that the market
is meanvariance efficient is quite small for a range of plausible priors.
 Zhou (1991) provides the first exact test of the zerobeta CAPM which allows
for borrowing rates be higher than the lending rates (more complex than the usual CAPM, but more realistic) and finds even this extension will not explain the market inefficiency.
 Zhou (1993) provides a finite sample test of the CAPM, with the exact Pvalue
computed via simulations without unknown parameters
for elliptical data, including in particular normal and tdistributions.
 Harvey and Zhou (1993) provide GMM tests of the CAPM, which is robust to general distributions.
 Geweke and Zhou (1994) provide an exact Bayesian framework for analyzing the arbitrage pricing theory (APT) and find the pricing errors
are little changed with including more factors beyond the first one (the pricing errors there may be better weighted by
using the asset inverse covariance matrix so that they will be invariant to portfolio repackaging).
 Chou and Zhou (2006) provide bootstrap tests of the CAPM, which is robust to iid distributions and more accurate
than usual asymptotic tests.
 Kan and Zhou (2017) provide asymptotic tests the CAPM under tdistributions, in which the
parameter estimates are efficiently obtained by using the EMalgorithm rather than the standard inefficient OLS estimator.
 He, Huang, Li and Zhou (2022)
provide a reducedrank approach (RRA) to extract a few factors from a large set of factor proxies
to explain expected returns, outperforming PCA, PLS and LASSO.
 He and Zhou (2023) is perhaps the first study
on the predictability of pricing errors, which provides a powerful test for asset pricing models, rejecting
all well known models such as FamaFrench 5factor model and those recently developed based on machine learning.
 Chib, Zhao and Zhou (2024) identify, via Bayesian model scanning, a set of new factors out of
a large set well known ones plus twelve principal components of anomalies unexplained by the existing ones.
 Liu, Zhou and Zhu (2023) propose a 4factor model, with a
trend factor that highlights the important role of individual
trading in asset pricing, especially in China.
 FOMC and Other Announcements :
 Liu, Tang and Zhou (2022) propose a novel method to recover the Federal Open Market Committee (FOMC) risk premium and drift sizes:
the risk premium varies from 1 to 326 bps, upward drift is 101 bps, and downward drift 129 bps.
 GMM :
 Zhou (1994) provides the first GMM tests for patterned weighting matrices that allow analytically solutions
in many finance applications (cited by Matyas (1999) in his GMM book; Cochrane (2001) presents a similar GMM test in his asset pricing book).
 SDF :
 Kan and Zhou (1999) show that the usual SDF (stochastic discount factor) approach provides unreliable risk premium estimate,
later studies resolve this problem by adding factor moment conditions for which no more analytical solutions available.
 Kan and Zhou (2006) provide the tightest lower bound on the SDF to date, showing that well known asset models do not
have enough volatility in the SDF to pass this bound.
 He and Zhou (2023) introduce the idea of testing pricing error predictability including those from a general SDF, based which they reject the well known asset pricing models and well as those recently developed based on machine learning.
 Twopass Regressions :

Shanken and Zhou (2007) provide formal model misppecification tests in addition to a comprehensive theoretical and small sample study of the widely used Fama and MacBeth (1973) twopass procedure that is fundamental in understanding to what extent crosssectional expected returns/values are explained by certain factor attributes.
 Jagannathan, Schaumburg and Zhou (2010) provide a survey of the literature on the twopass procedure.
 Bai and Zhou (2015) provide both the asymptotic theory for Fama and MacBeth (1973)
twopass regressions and new biasedadjusted OLS and GLS estimators in the common N>T case.
 Predictability :
 Lamoureux and Zhou (1996) show that the market is a random walk if no conditional information is incorporated.
 Rapach and Strauss and Zhou (2010) provide the first empirical evidence that: the US market risk premium is consistently predictable outofsample, with macroeconomic predictors via a combination forecast approach.
 Zhou (2010) provides a theoretical upper bound on the degree of predictability, improving substantially Ross bound implied by asset pricing models when state variables are available.
 Kong, Rapach, Strauss and Zhou (2011)
analyze the predictability of market components.
 Rapach and Strauss and Zhou (2013) find that the US stock market leads the world markets even at the monthly frequency.
 Rapach and Zhou (2013) provide a survey of the literature on stock return predictability.
 Neely, Rapach, Tu and Zhou (2014) show further that the predictive power of technical indicators
matches or exceeds that of macroeconomic variables (note that technical indicators, such as moving averages of prices, can capture
fundamental information too, such as world political stability, that are reflected in prices and not yet reflected in common macro variables).
(Highlight: still a top predictor shown by Goyal et al (2023))
 Huang, Jiang, Tu and Zhou (2015) find investor sentiment is a powerful predictor of the stock market.
 Rapach, Ringgenberg and Zhou (2016) show that the aggregated short interest is another powerful predictor.
(Highlight: still a top predictor shown by Goyal et al (2023))
 Huang and Zhou (2017) provide substantially tighter bounds on predictability and find major
asset pricing models fail to explain observed predictability due to inadequate state variables.
 Lin, Wu and Zhou (2018) provide a new iterated combination forecast method (of which the PLS is a special case), and, with this new method, they find the predictability of corporate bonds is both economically and statistically significant.
 Gao, Han, Li and Zhou (2018) discover an intraday predictive pattern: the first halfhour return on the market predicts the last halfhour return.
 Jiang, Lee, Martin and Zhou (2019) find that a new manager sentiment index predicts strongly aggregate investments and market returns.
 Huang, Li, Wang and Zhou (2019) show that timeseries momentum (TSM),
the predictability of the past 12month return on the next onemonth
return, is quite weak for the large cross
section of assets, altering conclusion of the literature.
 Chen, Tang, Yao and Zhou (2020) propose an investor attention index and find its predictive power on the stock market due to likely the reversal of temporary price pressure.
 Dong, Li, Rapach and Zhou (2022) find anomalies collectively have impressive predictive power on the market due to differences in long and short pricing error patterns.
 Rapach and Zhou (2022) provide a survey on time series predictability, with various recent
methods including machinelearningbased ones.
 Zaffaroni and Zhou (2023) provide a survey of studies on cross sectional predictability, which focus on firm predictors and useful for stock selections.
 Technical Analysis/Trend Factor/Momentum :
 Zhu and Zhou (2009) provide perhaps the first theoretical study to show that technical analysis, specifically the widely used moving averages,
can add value to asset allocation under uncertainty about predictability or
uncertainty about the true model governing the stock price.
 Zhou, Zhu and Qian (2012) provide an optimal asset
allocation strategy using technical indictors.
 Han and Yang and Zhou (2013) find that technical analysis, applied to portfolios sorted by volatility or other info
proxies, can outperform the buyandhold strategy substantially, and yield abnormal returns over 20% annually, which cannot be explained by market timing ability, investor sentiment, default and liquidity risks.
 Olszweski and Zhou (2014) show that combining both technicals and macro/fundamentals offers a significant improvement in riskadjusted returns.
 Han, Zhou and Zhu (2016) provide perhaps the first general equilibrium model on moving averages to justify their predictability, and to understand the role of technical traders;
they also propose a trend factor to
capture simultaneously all three stock price trends (the short, intermediate and longterm), which outperforms substantially existing factors, such as the momentum, by more than doubling their Sharpe ratios. (Highlight: the best crosssection predictor/factor from 19702021 shown by He et al (2023))
 Gao, Han, Li and Zhou (2018) discover perhaps the first intraday momentum pattern: the first halfhour return on the market predicts the last halfhour return.
 Liu, Zhou and Zhu (2023) propose a 4factor model, with a trend factor that highlights the important role of individual trading in asset pricing, especially in China.
 Anomalies :
 Chou, Li and Zhou (2006) study how anomalies help an investor to beat the market.
 The trend factor of Han, Zhou and Zhu (2016), based on technical analysis measures of trends across time horizons, is one of the greatest anomalies ever in terms of return and Sharpe ratio.
 Han, Huang and Zhou (2021) provide a simple dynamic monthly trading strategy to improve substantially existing anomalies that are formed on an annual basis.
 Han, Huang, Huang and Zhou (2022) uncover a volume amplification effect: expected return is related to trading volume positively among underpriced stocks but negatively among overpriced stocks.
 Dong, Li, Rapach and Zhou (2022) find anomalies collectively have impressive predictive power on the market due to differences in long and short pricing error patterns.
 Asymmetry :
 Hong, Tu and Zhou (2007) provide the first modelfree test for asymmetric correlations (and betas) to see if stocks move more often with the market when the market goes down than when it goes up.
 Jiang, Wu and Zhou (2018) provide a general asymmetry test based on entropy, and find significant asymmetry risk premium.
 Portfolio Choice :
 Tu and Zhou (2004) show how to select an optimal meanvariance portfolios under a realistic tdistribution and under asset pricing model priors, and they find that, tough the utility level does not change much vs normality assumption, but portfolio weights
are drastically different to achieve that.
 Kan and Zhou (2007) derive, for the first time, an explicit expression for the expected utility loss under parameter estimation risk for normallydistributed returns.
 Fabozzi, Huang and Zhou (2010) provide provide a survey of the literature on robust portfolios.
 Avramov and Zhou (2010) provide a survey of the literature on Bayesian portfolio analysis.
 Tu and Zhou (2010) show how economic objectives can serve as useful priors that yield superior portfolios, which, in particular, perform generally better than the well known 1/N rule.
 Tu and Zhou (2011) propose portfolio strategies that beat the 1/N rule in almost all scenarios except when the true weights are happen to be close to 1/N.
 Kan, Wang and Zhou (2022) propose perhaps the first optimized portfolio rule (other than the standard maximum likelihood estimator)
that combines the global minimum variance portfolio with a zerocost portfolio (the rule converges to the true optimal portfolio in large sample
and yet outperforms the ML portfolio in small sample), and derive analytical expressions for the optimal portfolio and utilities, providing insights into the real world problem where the managers are fully invested in risky assets (no riskfree assets).
 Yuan and Zhou (2023) show that the naive 1/N rule can be surprisingly optimal in a onefactor model,
and may be beaten only with conditional information in general.
 Volatility :
 Zhou and Zhu (2010) show how to select portfolios under short and longterm volatility risks and find huge impacts
vs the commonly used one volatility factor model.
 Zhou and Zhu (2015) extend the longrun risks model of Bansal and Yaron (2004) by allowing both a long and a shortrun volatility components in the evolution of economic fundamentals. With the extension, the new model not only is consistent with the volatility literature that the stock market is driven by two, rather than one, volatility factors, but also provides significant improvements in explaining various puzzles of equity and options data.
 Liu, Tang and Zhou (2019) show that, contrary to some previous studies, volatilitytiming
strategies do not work when applied to the aggregate stock market, once correcting a lookahead bias.
 Chen, Guo and Zhou (2022) apply machine learning tools to identify
firm fundamentals that explain the shape of the option implied volatility curve, providing insights into
some puzzling patterns of the literature.
 Active Portfolio Management :
 Zhou (2008a, b) extends the fundamental law of active portfolio management pioneered by Grinold (1989) to the case of estimation errors
and the case of conditional performance.
 Zhou (2009) provides an extension of the popular
BlackLitterman model by incorporating information from the data (such as the dynamics of how the market moves) beyond combining views with equilibrium.
 Behavior Finance :
 Huang, Jiang, Tu and Zhou (2015) provide a new investor sentiment index that is aligned with the purpose of predicting the aggregate stock market. By eliminating a common noise component in sentiment proxies, the new index has much greater predictive power than existing sentiment indices both in and outofsample, outperforms well recognized macroeconomic variables and can also predict crosssectional stock returns sorted by industry, size, value, and momentum.
 Zhou (2018) provides a review on various measures
of investor sentiment based on
market, survey, and media data, respectively, and
discusses various potential extensions and a number of issues for
future research.
 Jiang, Lee, Martin and Zhou (2019) provide a manager sentiment index based on the aggregated textual tone of conference calls and financial statements. They find that manager sentiment differs from investor sentiment in driving firm aggregate investments and market returns.
 Chen, Tang, Yao and Zhou (2020) propose an investor attention index, and find it has strong predictive power on the stock market due to perhaps the reversal of temporary price pressure.
 Chen, Tang, Yao and Zhou (2023) propose an employee sentiment index, and find it has strong predictive power on the stock market, as high employee sentiment leads to high contemporaneous wage growth resulting in lower firm cash flow and lower stock returns.
 Household Finance :

Gormley, Liu and Zhou (2010) show both theoretically and empirically that insurance (of large wealth shocks) plays an important role in household investment and savings decisions.
 Chinese Financial Markets and Monetary Policy :
 Jiang, Rapach, Strauss, Tu and Zhou (2011) find that the Chinese stock market is twice as predictable as the US.
 Fan, Li and Zhou (2013) analyze its supply factor in the Chinese bond market.
 Han, Wang, Zhou and Zou (2014) show momentum exists in China, but on a shortterm basis only.
 Fan, Jiang and Zhou (2014) provide an overview of the Chinese bond market.
 Liu, Tu, Zou and Zhou (2018) examine impacts of China's unique monetary policies in the perspective of the DSGE Model and Taylor rule.
 Jiang, Tang and Zhou (2018) provide perhaps the first comprehensive analysis,
perhaps the first machine learning study too, of firm characteristics on explaining Chinese stock expected returns in the cross section, and
find that trading frictions, momentum, and profitability have strong forecasting power in China.
 Wang, Zhou and Zhu (2021) provide an option pricing model accounting for shortselling and margintrading constraints, and show that
Chinese warrant prices, save for the time around expiration dates, are better characterized as derivatives than as pure bubbles.
 Liu, Zhou and Zhu (2023) propose a 4factor model, with a
trend factor that highlights the important role of individual
trading in asset pricing, especially in China.
 