Does Academic Research Destroy Stock Return Predictability?

We study the out‐of‐sample and post‐publication return predictability of 97 variables shown to predict cross‐sectional stock returns. Portfolio returns are 26% lower out‐of‐sample and 58% lower post‐publication. The out‐of‐sample decline is an upper bound estimate of data mining effects. We estimate a 32% (58%–26%) lower return from publication‐informed trading. Post‐publication declines are greater for predictors with higher in‐sample returns, and returns are higher for portfolios concentrated in stocks with high idiosyncratic risk and low liquidity. Predictor portfolios exhibit post‐publication increases in correlations with other published‐predictor portfolios. Our findings suggest that investors learn about mispricing from academic publications.     

APPLICABILITY OF THE FAMA‐FRENCH THREE‐FACTOR MODEL IN FORECASTING PORTFOLIO RETURNS

For the model‐based estimation of the equity cost of capital, evidence shows that the common practice of using the average historical factor premiums as the estimates of the next‐period factor premiums generates inaccurate estimates. I propose an alternative way to estimate factor premiums by using the structural variables that are important predictors of future asset returns. Based on the out‐of‐sample results from a trading strategy with four in‐sample model‐selection criteria, I find that my estimation procedure performs better than the common practice even when transaction costs are considered.     

Information diffusion and the predictability of New Zealand stock market returns

This paper examines the impact of international predictors from liquid markets on the predictability of excess returns in the New Zealand stock market using data from May 1992 to February 2011. We find that US stock market return and VIX contribute significantly to the out‐of‐sample forecasts at short horizons even after controlling for the effect of local predictors, while the contribution by Australian stock market return is not significant. We further demonstrate that the predictability of New Zealand stock market returns using US market predictors could be explained by the information diffusion between these two countries.     

Market Illiquidity and Conditional Equity Premium

We examine the time‐series relation between aggregate bid‐ask spreads and conditional equity premium. We document that average marketwide relative effective bid‐ask spreads forecast aggregate market returns only when controlling for average idiosyncratic variance. This control allows us to document the otherwise elusive relation between illiquidity and returns. The reason is that idiosyncratic variance correlates positively with spreads but has a negative effect on conditional equity premium, causing an omitted variable bias. Our results are robust to standard return predictors, alternative illiquidity measures, and out‐of‐sample tests. These findings are important because they provide strong support for the literature's conjecture that marketwide liquidity is an important asset pricing risk factor.     

Can machines learn capital structure dynamics?

Yes, they can! Machine learning models predict leverage better than linear models and identify a broader set of leverage determinants. They boost the out-of-sample R² from 36% to 56% over OLS and LASSO. The best performing model (random forests) selects market-to-book, industry median leverage, cash and equivalents, Z-Score, profitability, stock returns, and firm size as reliable predictors of market leverage. More precise target estimation yields a 10%–33% faster speed of adjustment and improves prediction of financing actions relative to linear models. Machine learning identifies uncertainty, cash flow, and macroeconomic considerations among primary drivers of leverage adjustments.     

Evaluating risk‐based capital regulation

This paper evaluates the effectiveness of risk‐based capital (RBC) regulation and challenges some evidence from the well‐known study by Haldane and Madouros (2012). We reconsider the evidence on the relationship between RBC ratios and failures of US banks from Haldane and Madouros (2012) and find their results are not robust to changes in the sample period or regression model. Using data on US commercial banks from 2000 through 2015 and an improved regression model, we compare banks’ RBC ratios and simple capital ratios as predictors of bank risk. We find simple capital ratios to be significantly better than complex RBC ratios as predictors of bank risk.     

Variable selection and corporate bankruptcy forecasts

We investigate the relative importance of various bankruptcy predictors commonly used in the existing literature by applying a variable selection technique, the least absolute shrinkage and selection operator (LASSO), to a comprehensive bankruptcy database. Over the 1980–2009 period, LASSO admits the majority of Campbell et al. (2008) predictive variables into the bankruptcy forecast model. Interestingly, by contrast with recent studies, some financial ratios constructed from only accounting data also contain significant incremental information about future default risk, and their importance relative to that of market-based variables in bankruptcy forecasts increases with prediction horizons. Moreover, LASSO-selected variables have superior out-of-sample predictive power and outperform (1) those advocated by Campbell et al. (2008) and (2) the distance to default from Merton’s (1974) structural model.     

Chinese corporate distress prediction using LASSO: The role of earnings management

Motivated by recently increasing accounting manipulation cases and deteriorating economic condition in China, we investigate the importance of a set of earnings management predictors and develop up-to-date distress prediction model with earnings management consideration for the Chinese market. Employing annual firm-level data from January 2014 to December 2018, we find that real earnings management (REM) is robustly selected out as a key distress predictor via the variable selection technique LASSO. Our results consistently show that REM could improve early warning of distressed companies with a slight sacrifice of accuracy in predicting healthy companies. After considering the cost of misclassification, it is confirmed that REM contains incremental information about a forthcoming corporate distress risk. Meanwhile, our results also detect an interesting finding that in China, aggressive real earnings management signals the lower probability of corporate distress, indicating distressed firms have lower capacity to conduct REM.     

Out‐of‐sample stock return predictability in emerging markets

This article builds on the widely debated issue of stock return predictability by applying a broad range of predictor variables and comprehensively considering the in‐sample and out‐of‐sample stock return predictability of ten advanced emerging markets. It compares forecasts from models with a single predictor variable, multiple predictor variables and a combination forecast approach. The results confirm the findings of Welch and Goyal (2008) for US data that only a limited number of individual predictor variables are able to deliver significant out‐of‐sample forecasts. However, a combination forecast approach provides statistically and economically significant out‐of‐sample forecast results.     

Predicting the Failure of Developmental Gold Mining Projects

This paper investigates firm‐level financial and non‐financial information and their association with project failure for a sample of pre‐production gold development firms. Pre‐revenue generating ‘single project’ mining companies are chosen, since project failure is synonymous with company failure for these firms. The setting is interesting due to the high information asymmetry and limitations of the GAAP‐based Altman Z‐score in this context. A definition of project failure is applied and both financial and non‐financial predictors are compared. Failure is driven by whether the deposit is open pit or underground, and whether the cash cost of production is disclosed at feasibility completion.     

On the Stability of the Cross‐Section of Expected Stock Returns in the Cross‐Section: Understanding the Curious Role of Share Turnover

In this paper, we shed further light on cross‐sectional predictors of stock return performance. Specifically, we explore whether the cross‐section of expected stock returns is robust within stock groups sorted by past monthly return. We find that the book/market and momentum effects are remarkably robust to sorting on past returns. However, share turnover is negatively related to future returns for stocks with abnormally low stock price performance in the recent past, but postively related to returns for well‐performing stocks. This casts doubt on the use of turnover as a liquidity proxy, but is consistent with turnover being a proxy for momentum trading which pushes prices in the direction of past price movements. Our results are robust to both NYSE/AMEX and Nasdaq stocks, and also robust to stratifying the sample by time period.     

Ability of accounting and audit quality variables to predict bank failure during the financial crisis

We examine the ability of selected accounting and audit quality variables measured in a period prior to the financial crisis (i.e., the four quarters of 2006), to predict banks that subsequently failed during the financial crisis. We employ two sets of samples from the US: a troubled banks sample that includes banks that failed in or after 2007 as well as banks classified as being troubled based on profitability, loan quality, and balance sheet position in 2007, and a full sample that includes all banks with available required data. Using the troubled banks sample, we identify six reliable predictors of bank failure: auditor type, auditor industry specialization, Tier 1 capital ratio, proportion of securitized loans, growth in loans, and loan mix. For the larger full sample of banks, we identify the following ten predictors of bank failure: auditor type, Tier 1 capital ratio, proportion of securitized loans, nonperforming loans, loan loss provisions, growth in commercial loans, growth in real estate loans, growth in overall loans, loan mix, and whether the bank is a public bank.     

A Robust Unsupervised Method for Fraud Rate Estimation

If one is interested in managing fraud, one must measure the fraud rate to be able to assess the degree of the problem and the effectiveness of the fraud management technique. This article offers a robust new method for estimating fraud rate, PRIDIT‐FRE (PRIDIT‐based Fraud Rate Estimation), developed based on PRIDIT, an unsupervised fraud detection method to assess individual claim fraud suspiciousness. PRIDIT‐FRE presents the first nonparametric unsupervised estimator of the actual rate of fraud in a population of claims, robust to the bias contained in an audited sample (arising from the quality or individual hubris of an auditor or investigator, or the natural data‐gathering process through claims adjusting). PRIDIT‐FRE exploits the internal consistency of fraud predictors and makes use of a small audited sample or an unaudited sample only. Using two insurance fraud data sets with different characteristics, we illustrate the effectiveness of PRIDIT‐FRE and examine its robustness in varying scenarios.     

MULTIDIMENSIONAL DISTANCE‐TO‐COLLAPSE POINT AND SOVEREIGN DEFAULT PREDICTION

By focusing on sovereign defaults, this paper introduces a multidimensional distance‐to‐collapse point based on a two‐step procedure. The first step is nonparametric and provides an early warning system that signals a potential crisis whenever preselected leading indicators exceed specific thresholds. The second is parametric and incorporates the first‐step country default predictors within a probit specification. Such a two‐step procedure generalizes the distance‐to‐default à la Merton within a multidimensional setting, wherein we care about the distance of each indicator from its threshold. Empirical evidence about debt crises of emerging markets over the period 1975–2002 proves that our methodology predicts 80% of the total defaults and non‐defaults in and out of sample. Copyright © 2012 John Wiley & Sons, Ltd.     

Are cryptocurrencies homogeneous?

This article investigates if cryptocurrencies returns' are similarly affected by a selection of demand- and supply-side determinants. Homogeneity among cryptocurrencies is tested via a least absolute shrinkage and selection operator (LASSO) model where determinants of Bitcoin returns are applied to a sample of 12 cryptocurrencies. The analysis goes beyond existing research by simultaneously covering different periods and design choices of cryptocurrencies. The results show that cryptocurrencies are heterogeneous, apart from some similarities in the impact of technical determinants and cybercrime. The cryptocurrency market displays evidence of substitution effects, and design choices related explain the impact of the determinants of return.     

The Really Long-Run Performance of Initial Public Offerings: The Pre-Nasdaq Evidence

Financial economists have intensely debated the performance of IPOs using data after the formation of Nasdaq. This paper sheds light on this controversy by undertaking a large, out‐of‐sample study: We examine the performance for five years after listing of 3,661 U.S. IPOs from 1935 to 1972. The sample displays some underperformance when event‐time buy‐and‐hold abnormal returns are used. The underperformance disappears, however, when cumulative abnormal returns are utilized. A calendar‐time analysis shows that over the entire period, IPOs return as much as the market. The intercepts in CAPM and Fama–French regressions are insignificantly different from zero, suggesting no abnormal performance.     

Predictors of unrealistic optimism: a study of Norwegian risk takers

In areas of voluntary risk behaviour, as with other kinds of risk, people tend to be overly optimistic regarding not being injured. A study of risk perception and causal explanations of injury assessments was conducted on 199 respondents from three different sub‐groups in Norway; skydivers (n?=?88), fire fighters (n?=?73) and soldiers (n?=?38). Unrealistic optimism was studied by means of four demographic variables: the background of the subject (sub‐sample), gender, age, and education. In addition, three predictors of unrealistic optimism were taken into account—safety attitudes, control, and anxiety. These predictors were included in an Analysis of Linear Structural Relationship (LISREL) analysis. The results showed that optimism differed between the sub‐groups, and that different factors influenced risk perception depending on the group and depending on whether the assessment was of oneself or of others. These findings offers additional information that will help explain the inconsistent findings in the current literature of unrealistic optimism. Of the predictors investigated, safety attitudes were found to be the most important, which may be because respondents preoccupied with safety are more aware of potential dangers and thereby less optimistic.     

Predicting corporate policies using downside risk: A machine learning approach

This paper develops a text-based downside risk measure using corporate annual reports and assesses its ability to forecast future corporate policies. The forward-looking measure dynamically captures adverse firm conditions evolving from economic fundamentals. When the measure is below its sample average, leverage, investment, R&D, employment, and dividends consistently fall. When the measure rises, firms increase cash holdings. The proposed measure also delivers robust and persistent forecasts based on in-sample and out-of-sample LASSO regressions.     

On the Relative Pricing of Long‐Maturity Index Options and Collateralized Debt Obligations

We investigate a structural model of market and firm‐level dynamics in order to jointly price long‐dated S&P 500 index options and CDO tranches of corporate debt. We identify market dynamics from index option prices and idiosyncratic dynamics from the term structure of credit spreads. We find that all tranches can be well priced out‐of‐sample before the crisis. During the crisis, however, our model can capture senior tranche prices only if we allow for the possibility of a catastrophic jump. Thus, senior tranches are nonredundant assets that provide a unique window into the pricing of catastrophic risk.     

CEO Risk Preference and Investing in R&D

This study aims at: (1) developing an index to measure CEO risk tolerance using publicly available data, and (2) examining the association between this index and investment in risky projects. Using relative pay‐at‐risk as a proxy for risk preference (tolerance) is a new proposition and is supported by having significant association with CEOs' socio‐demographic variables—the variables often studied in connection with risk aversion. Furthermore, this risk preference indicator has a positive association with risk‐taking behaviour as proxied by R&D expenditures. The in‐sample estimation and out‐of‐sample predictions support (a) using relative pay‐at‐risk as a valid proxy for risk tolerance, and (b) finding statistically significant positive association between this measure and R&D expenditures. The association has different degrees of strength for nine out of 11 industries.