Conditional Asset Pricing Model Research Articles

Dynamic ESG Equilibrium

This paper proposes a conditional asset pricing model that integrates environmental, social, and governance (ESG) demand and supply dynamics. Shocks in the demand for sustainable investing represent a novel risk source, characterized by diminishing marginal utility and positive premium. Green assets exhibit positive exposure to ESG demand shocks, hence commanding higher premia. Conversely, time-varying convenience yield leads to lower expected returns for green assets. Moreover, ESG demand shocks have positive contemporaneous effects on unexpected returns, contributing to large positive payoffs in the green-minus-brown portfolio over extended horizons. The model predictions align closely with evidence on return spreads between green and brown assets, further reinforcing the apparent gap between realized and expected spreads. This paper was accepted by Lin William Cong, finance. Funding: Y. Liu acknowledges funding by the Research Grants Council of the Hong Kong Special Administrative Region, China [Grant T35/710/20R]. A. Tarelli acknowledges funding by the European Union - Next Generation EU [Project Code 2022FWZ2CR]. The views and opinions expressed are only those of the authors. Supplemental Material: The online appendices and data files are available at https://doi.org/10.1287/mnsc.2022.03491 .

Conditional autoencoder asset pricing models for the Korean stock market.

This study analyzes the explanatory power of the latent factor conditional asset pricing model for the Korean stock market using an autoencoder. The autoencoder is a type of neural network in machine learning that can extract latent factors. Specifically, we apply the conditional autoencoder (CA) model that estimates factor exposure as a flexible nonlinear function of covariates. Our main findings are as follows. The CA model showed excellent explanatory power not only in the entire sample but also in several subsamples in the Korean market. Also, because of this explanatory power, it can better explain market anomalies compared to the traditional asset pricing models. As a result of examining investment strategies using pricing error, the CA model measures the expected return of stocks better than the traditional asset pricing model. In addition, the CA model indicates that the firm characteristic variables are important in asset pricing conditional on macro-financial states, such as the global financial crisis and the coronavirus disease 2019 pandemic. The result shows that the major variables considered in the explanation of stock returns through the CA model may vary depending on the time. This is expected to provide a broader perspective on asset pricing through the CA model in the future.

Understanding Cross-Sectional Variability in Equity Returns Using a Conditional Asset Pricing Model

A common practice in the literature examining Turkish equity premia involves the application of time-invariant models that assume the constancy of parameters across time. The present study examines whether the cross-sectional variability among equity returns can be explained by market, size, value and momentum factors and whether the parameters are time varying, utilizing a conditional asset pricing model formulated by Ferson and Harvey (1999). The study yields four main findings. Firstly, I find that the market dividend yield has a positive and significant effect on portfolio returns over the period from July 1989 to May 2021. Secondly, I reject the time-invariance in betas, while not rejecting it for alpha. Thirdly, none of the factors I examined have been priced, indicating that the four-asset pricing model is not sufficient to explain time-varying premia. Finally, the results are sensitive to the methodology employed for portfolio construction.

Determinants of time-varying equity risk premia in an emerging market

PurposeThis study aims to document the time varying risk premia for market, size, value and momentum factors for an emerging market using a sophisticated conditional asset pricing model. The focus of this study is Turkish stock market denominated in local currency with its peculiar risk premia.Design/methodology/approachThe authors employ Gagliardini et al.'s (2016) econometric method that uses cross-sectional and time series information simultaneously to infer the path of risk premia from individual stocks.FindingsUsing this methodology, the authors assess several conditioning information and conclude that local dividend yield, inflation and exchange rates have the most explanatory power. The authors document the time varying risk premia in Turkey over three decades.Originality/valueExisting studies on dynamic estimation of risk premia lack a consensus as to which state variables should be included and to what extent they impact the magnitude of the premium. The authors extend the conditioning information set beyond the ones existing in the literature to determine variables that are specifically important for an emerging market.

Do oil price shocks have any implications for stock return momentum?

The study examines the implications of structural oil price shocks on one of the most widely observed and pervasive momentum anomalies. We use a structural vector autoregression (SVAR) model to find the time-varying responses of momentum payoffs of India, Japan, the UK, and the US to different structural oil shocks—oil supply global aggregate demand shock, speculative shock, and other oil demand shocks. Structural oil shocks, particularly oil demand and speculative shocks, have important implications as conditioning information in predicting stock return momentum behavior according to the conditional asset pricing model. Overall, the results indicate that structural oil shocks have a significant impact on small cap and loser stock portfolios. Our findings have crucial implications to develop effective momentum investment strategies that consider structural oil shocks as conditioning information.

Role of Liquidity Risk in Asset Pricing: Evidence from Sri Lanka

Securities liquidity varies over time, which leads to equity return volatility. It implies that the liquidity in the capital markets is a significant source of risk. Therefore, liquidity risk in securities is difficult to diversify and contributes to the systemic market risk. This study aims to analyze the relationship between securities returns and liquidity risk while taking into account the time-varying characteristics of illiquidity on the Colombo Stock Exchange from 2015-2019 and taking into account the effect of liquidity level, using the Generalized Method of Movements (GMM) framework model to assess the persistence of illiquidity securities contributions of the updated version of the Amihud illiquidity (Amihud, 1986) proxy to represent across time market illiquidity and to research the time-series relationship between liquidity and returns. The pricing of liquidity risk and its implications for expected returns are empirically tested using the conditional liquidity adjusted capital asset pricing model (LCAPM), where stock returns are cross-sectionally dependent on market risk and three additional betas (β1, β2 , β3 ) that capture different aspects of illiquidity and its risk. The findings reveal some support for the conditional capital asset pricing model (CAPM), but results are not robust to alternative specifications and estimation techniques. The total effect of liquidity risk is 0.11%, and illiquidity is 2.5% per year. Illiquidity premium depends on the expected transaction cost at the end of the holding period for investors' 2.5%. This makes the overall illiquidity premium of 2.61%. These estimates and the overall importance of liquidity level and liquidity risk depend on the model implied restrictions of a constant market risk premium and a fixed transaction cost. However, LCAPM has constructed conditionally; it can relax these model-implied constraints and estimate different liquidity risk premiums while also allowing transaction costs to be a free parameter. The overall liquidity risk characterized by liquidity betas with a single market risk premium is relatively small and barely significant in the restricted model. Using this unrestricted model, find that the overall illiquidity premium corresponds to 2.61%. The empirical results shed light on these channels' toal and relative economic significance and provide evidence of flight to liquidity.
 
 Keywords: Capital Asset Pricing Model, Liquidity Risk, Liquidity beta, Generalize Method of Movement, Sri Lanka

Nonparametric Specification Testing of Conditional Asset Pricing Models

This article presents an adaptive omnibus specification test of asset pricing models where the stochastic discount factor is conditionally affine in the pricing factors. These models provide constraints that conditional moments of returns and pricing factors must satisfy, but most of them do not provide information on the functional form of those moments. Our test is robust to functional form misspecification, and also detects any relationship between pricing errors and conditioning variables. We give special emphasis to the test implementation and calibration, and extensive simulation studies prove the functioning in practice. Our empirical applications show a conditional counterpart of a well-known problem of unconditional models. The lack of rejection of consumption based conditional models seems to be due to a poor conditional correlation between consumption and stock returns.

Evaluating the Conditional CAPM using Consumption-based State Variables: Evidence from the Korean Stock Market

In this study, we evaluate the empirical performance of conditional asset pricing models using consumption-based measures as state variables. We incorporate three consumption variables known to forecast the equity risk premium as conditioning variables to capture time variations in the risk premium. These three variables are the consumption-aggregate wealth ratio, the surplus consumption ratio, and the labor income to consumption ratio. The asset pricing models evaluated in this study are the CAPM, the CAPM with human capital, the consumption CAPM, and the Fama-French three-factor model. We compare the unconditional and conditional specifications of these four asset pricing models using the two-pass cross-sectional regression methodology, using the size, book-to-market, turnover, and idiosyncratic risk sorted portfolios and sector portfolios as test assets. We demonstrate that the conditional CAPM with human capital performs far better than the unconditional specifications and about as well as the Fama and French three-factor model in explaining the crosssection of average stock returns in Korea.

Information Choice, Uncertainty, and Expected Returns

Abstract We investigate how information choices affect equity returns and risk. Building on an existing theoretical model of information and investment choice, we estimate a learning index that reflects the expected benefits of learning about an asset. High learning index stocks have lower future returns and risk compared to low learning index stocks. Analysis of a conditional asset pricing model, long-run patterns in returns and volatilities, other measures of information flow, and the information environment surrounding earnings announcements reinforce our interpretation of the learning index. Our findings support the model’s predictions and illustrate a novel empirical measure of investor learning.

Deep Learning for Conditional Asset Pricing Models

We propose a new pseudo-Siamese Network for Asset Pricing (SNAP) model, based on deep learning approaches, for conditional asset pricing. Our model allows for the deep alpha, deep beta and deep factor risk premia conditional on high dimensional observable information of financial characteristics and macroeconomic states, while storing the long-term dependency of the informative features through long short-term memory network. We apply this method to monthly U.S. stock returns from 1970-2019 and find that our pseudo-SNAP model outperforms the benchmark approaches in terms of out-of-sample prediction and out-of-sample Sharpe ratio. In addition, we also apply our method to calculate deep mispricing errors which we use to construct an arbitrage portfolio K-Means clustering. We find that the arbitrage portfolio has significant alphas.

Autoencoder asset pricing models

We propose a new latent factor conditional asset pricing model. Like Kelly, Pruitt, and Su (KPS, 2019), our model allows for latent factors and factor exposures that depend on covariates such as asset characteristics. But, unlike the linearity assumption of KPS, we model factor exposures as a flexible nonlinear function of covariates. Our model retrofits the workhorse unsupervised dimension reduction device from the machine learning literature – autoencoder neural networks – to incorporate information from covariates along with returns themselves. This delivers estimates of nonlinear conditional exposures and the associated latent factors. Furthermore, our machine learning framework imposes the economic restriction of no-arbitrage. Our autoencoder asset pricing model delivers out-of-sample pricing errors that are far smaller (and generally insignificant) compared to other leading factor models.

Comment on: Pseudo-True SDFs in Conditional Asset Pricing Models

Comment on: Pseudo-True SDFs in Conditional Asset Pricing Models

A reappraisal of luck versus skill in the cross-section of mutual fund returns

The first contribution we make to research on measuring U.S. mutual fund performance is to show that the cross-section bootstrap procedure used in one prominent publication on this topic can easily accommodate conditional asset pricing models. Using this result, we reestimate US fund performance using the conditional asset pricing model of Ferson and Schadt (1996) for the period of time covering January 1984 to September 2006, the same period of time used in this prominent publication. Unlike most of the current research on the performance of U.S. mutual funds, including the latter research, the estimates of fund performance produced here suggest that some funds, on a net return basis, performed very well over this period. Our second contribution is to update the U.S. mutual fund data to the end of 2018 and then to re-estimate fund performance over this longer period of time. These results show that fund performance, on a net return basis, is poor, confirming the findings of previous research that used data ending before 2018. The third contribution is to provide a detailed guide on how the mutual fund data widely used in this literature is constructed. Until now much of the information required to do this has not been made widely available. Our last contribution is to make some policy recommendations that should better align fund manager incentives with the interests of investors, an alignment which current practices have hindered.

Comment on: Pseudo-True SDFs in Conditional Asset Pricing Models

Comment on: Pseudo-True SDFs in Conditional Asset Pricing Models

Comment on: Price Discovery in High Resolution, Comment on: Pseudo-True SDFs in Conditional Asset Pricing Models, and Comment on: Pseudo-True SDFs in Conditional Asset Pricing Models.Comparing Fixed-versus Vanishing-Bandwidth Estimators of Pseudo-True SDFs

Comment on: Price Discovery in High Resolution, Comment on: Pseudo-True SDFs in Conditional Asset Pricing Models, and Comment on: Pseudo-True SDFs in Conditional Asset Pricing Models.Comparing Fixed-versus Vanishing-Bandwidth Estimators of Pseudo-True SDFs

Estimation of conditional asset pricing models with integrated variables in the beta specification

We introduce a methodology which deals with possibly integrated variables in the specification of the betas of conditional asset pricing models. In such a case, any model which is directly derived by a polynomial approximation of the functional form of the conditional beta will inherit a nonstationary right hand side. Our approach uses the cointegrating relationships between the integrated variables in order to maintain the stationarity of the right hand side of the estimated model, thus, avoiding the issues that arise in the case of an unbalanced regression. We present an example where our methodology is applied to the returns of funds-of-funds which are based on the Morningstar mutual fund ranking system. The results provide evidence that the residuals of possible cointegrating relationships between integrated variables in the specification of the conditional betas may reveal significant information concerning the dynamics of the betas.

Information Choice, Uncertainty, and Expected Returns

We investigate how information choices impact equity returns and risk. Building on an existing theoretical model of information and investment choice, we estimate a learning index that reflects the expected benefits of learning about an asset. High learning index stocks have lower future returns and risk compared to low learning index stocks. Analysis of a conditional asset pricing model, long run patterns in returns and volatilities, other measures of information flow, and the information environment surrounding earnings announcements reinforce our interpretation of the learning index. Our findings support the model's predictions and illustrate a novel empirical measure of investor learning.

Expected and realized returns in conditional asset pricing models: A new testing approach

Abstract We develop a new approach for testing conditional asset pricing models that avoids the issues in using realized returns as a proxy for expected returns. Testable restrictions are developed by asking what realized returns we would observe, given the pricing model under scrutiny. The new reverse testing approach is used to test the Merton ICAPM and a long-standing risk–return puzzle: the price of market risk has often turned out to be insignificant and at times even negative. The results from the new testing approach on US data give strong support for a positive relationship between conditional variance and the equity premium.

Comment on: Pseudo-True SDFs in Conditional Asset Pricing Models. Comparing Fixed- versus Vanishing-Bandwidth Estimators of Pseudo-True SDFs*

Comment on: Pseudo-True SDFs in Conditional Asset Pricing Models. Comparing Fixed- versus Vanishing-Bandwidth Estimators of Pseudo-True SDFs*

Autoencoder Asset Pricing Models

We propose a new latent factor conditional asset pricing model. Like Kelly, Pruitt, and Su (KPS, 2019), our model allows for latent factors and factor exposures that depend on covariates such as asset characteristics. But, unlike the linearity assumption of KPS, we model factor exposures as a flexible nonlinear function of covariates. Our model retrofits the workhorse unsupervised dimension reduction device from the machine learning literature—autoencoder neural networks—to incorporate information from covariates along with returns themselves. This delivers estimates of nonlinear conditional exposures and the associated latent factors. Furthermore, our machine learning framework imposes the economic restriction of no-arbitrage. Our autoencoder asset pricing model delivers out-of-sample pricing errors that are far smaller (and generally insignificant) compared to other leading factor models.