Anthony Sanford, Ph.d.

Finance researchers keep producing increasingly complex and computationally-intensive models of stock returns. Separately, professional analysts forecast stock returns daily for their clients. Are the sophisticated methods of researchers achieving better forecasts or are we better off relying on the expertise of analysts on the ground? Do the two sets of actors even capture the same information? In this paper, I hypothesize that analyst forecasts and forecasts constructed using option prices will be different because they draw on different information sets. Using hypothesis tests and quantile regressions, I find that option-based forecasts are statistically significantly different from analyst forecasts at every level of the forecast distribution. Then, using cross-sectional regressions, I show that this difference originates in the distinct information sets used to create the forecasts: option-based forecasts incorporate information about the probability of extreme events while analyst forecasts focus on information about firm and macroeconomic fundamentals.

In statistics, samples are drawn from a population in a data generating process (DGP). Standard errors measure the uncertainty in sample estimates of population parameters. In science, evidence is generated to test hypotheses in an evidence generating process (EGP). We claim that EGP variation across researchers adds uncertainty: non-standard errors. To study them, we let 164 teams test six hypotheses on the same sample. We find that non-standard errors are sizeable, on par with standard errors. Their size (i) co-varies only weakly with team merits, reproducibility, or peer rating, (ii) declines significantly after peer-feedback, and (iii) is underestimated by participants.

In this paper, we use event studies to estimate the effects of changes to a public firm's board of trustees on stock returns. The goal is to determine whether the gender of an incoming board member is perceived differently by investors. Scholarly findings on gender and leadership have been mixed at best. Overall, the evidence seems to indicate that women and men in comparable leadership positions are much more alike than different. Yet, the number of women in leadership positions in the United States (and globally) is still disproportionately low-a phenomenon known as the "glass ceiling." Our study shows that women and men, at least in the United States, are still not created equal in the eyes of investors. Using BoardEx data on the composition of U.S. public firm boards for 1992-2017, we find that changes to a firm's board are consistently perceived as a negative information shock by investors, but the effect of incoming female board members is more than twice as negative as than of male counterparts.

How does the interaction of uncertainty and R&D impact corporate investment? We provide evidence that R&D significantly increases corporate investment responsiveness to PVGO news and uncertainty shocks. These results are consistent with predictions from the R&D-based real options model of corporate investment. To establish credible causal results we combine new measures of systematic and firm-specific PVGO shocks, for which we utilize stock price and option data, with exogenous measures of R&D capital stocks derived from panel variation in state R&D tax credits. We also rule out a number of potentially competing explanations for our results, including firm-level differences in lumpiness of investments, financial frictions, lifecycle growth opportunities or moral hazard-implied asset substitution or risk shifting.

In this paper, I construct an optimal portfolio by minimizing the expected tail loss (ETL) derived from the forward-looking natural distribution of the Recovery Theorem (RT). The RT is one of the first successful attempts at deriving an unparameterized natural distribution of future asset returns. This distribution can be used as the criterion function in an expected tail loss (ETL) portfolio optimization problem. I find that the portfolio constructed using the RT outperforms both the equally-weighted portfolio and a portfolio constructed using historical ETL. The portfolio constructed using the RT has the smallest historical tail loss, smallest maximum drawdown, smallest Sortino Ratio, and smallest Sharpe Ratio.

This article introduces a model to estimate the risk neutral density of stock prices derived from option prices. To estimate a complete risk-neutral density, current estimation techniques use a single mathematical model to interpolate option prices on two dimensions: strike price and time-to-maturity. Instead, this model uses B-splines with at-the-money knots for the strike price interpolation and a mixed lognormal function that depends on the option expiration horizon for the time-to-maturity interpolation. The results of this "hybrid" methodology are significantly better than other risk-neutral density extrapolation methods when applied to the recovery theorem.

This article uses public perceptions to forecast short-term fluctuations in asset prices. Based on four billion tweets scraped between 2009 and 2019, I perform textual analysis to construct daily sentiment indices. The sentiment indices allow us to forecast stock volatility jumps as well as expected jump levels. The implications of forecasting volatility jumps are substantive. First, volatility jumps have a significant effect on option prices. Second, changes in the volatility path lead to large (negatively related) changes in the prices’ future trajectory. Determining what information causes jumps allows for better risk management and more accurate asset pricing models.

Characterizing risk in financial markets has, at times, presented contradictory findings due to differing volatility estimation methodologies. Our research shows that the choice of time intervals in the volatility calculations distinctly captures different aspects of risk, depending on the stock's specific mean-reversion process. Short-interval volatilities predominantly capture idiosyncratic risk, while longer intervals capture systematic risk. This distinction, driven in part by characteristics like mean-reversion rates of the volatility processes, provides deeper insights into the broader conceptualizations of risk and volatility in the asset pricing literature, especially when it comes to our volatility-based definitions of market participants, like retail traders.

The portfolio theory literature has long strived for perfection when it comes to estimating expected returns. In this paper we develop a theoretical framework which shows that perfect (known) expected returns may not be necessary for (optimal) portfolio optimizations. Using simulated data, we show that results from a portfolio constructed using perfect expected returns do not outperform those whose results are "shocked." Our findings suggest that optimizations using correctly ranked expected returns outperform portfolios whose expected returns are known. We confirm these results empirically using option-recovered expected idiosyncratic returns which outperform other benchmark portfolio optimization models.

Predictors of excess returns exhibit persistence and time-varying variance, implying the need for heteroskedastic and autocorrelation consistent errors (HAC) in linear tests. Using simulations, we show that although they lead to important improvements, such corrections fail to provide adequate size properties under the null hypothesis of zero abnormal returns. Even optimally specified robust estimators suffer from size distortions, implying that the best HACs remain imperfect. We propose a standardization of the robust estimator that addresses the problem, albeit not completely. We find that between 2006 and 2021, more than 20% of a wide panel of predictors differ in significance status at the standard 5% level in comparing this estimator to ordinary least squares, and more than 30% at a more restrictive level.

In this paper, I redefine the prices derived in Ross’ Recovery Theorem (Ross, 2015) using a multivariate Markov chain rather than a univariate one. I employ a mixture transition distribution where the proposed states depend on the level of the S&P 500 index and its options’ implied volatilities. I include volatility because the transition path between states depends on the propensity of an underlying asset to vary. An asset that is highly volatile is more likely to transition to a far-away state. These higher transition probabilities should lead to higher state prices. The multivariate method improves upon the univariate RT because the latter does not include the volatility inherent in the state transition, which makes its derived prices less precise. The multivariate RT produces forecast results far superior to the univariate RT. Using quarterly forecasts for the 1996-2015 period, the out-of-sample R-square of the RT increases from around 12% to 30%. Moreover, using simulated data, I show that including the implied volatility in the multivariate Markov chain more closely captures the inherent risk in business cycles.