The correlation structure of anomaly strategies

Relative Valuation with Machine Learning

“…To rule out the possibility that the discovered return predictability is a known anomaly, we use a 13‐factor model, which is the FF6 model augmented with seven “cluster portfolios” of anomaly based trading strategies (shown as FF6 + 7 clusters of anomalies in table 3 (value‐weighted) and table IA.7 in the online appendix (equal‐weighted)). Geertsema and Lu [2020] show that this augmented model renders insignificant 80 mean‐significant anomalies previously identified in the literature. The return predictability from machine valuation errors is unlikely to be a prominent known anomaly, because the 13‐factor model ( FF6 + 7 clusters of anomalies ) only reduces the abnormal returns slightly and all abnormal returns remain highly significant.…”

“…Several recent papers have applied machine learning techniques to topics in accounting and finance. These include stock return predictability (Gu, Kelly, and Xiu [2020]), analysis of the textual structure of news (Bybee et al [2020]), bond yield predictability (Bianchi, Büchner, and Tamoni [2021]), credit risk forecasting (Khandani, Kim, and Lo [2010]), dimensionality of risk factors (Geertsema and Lu [2020]), fraud detection (Bao et al [2020]), and director selection (Erel et al [2021]). Our application of machine learning to relative valuation is novel.…”

“…One is the Fama-French six-factor (FF6) model. The other is the FF6 model augmented by seven additional factors that are shown to subsume all mean-significant anomalies accumulated to date byGeertsema and Lu [2020]. Refer to table 2 for a summary of the valuation models.Means and abnormal returns are in percentage points.…”

Clustering-Based Sector Investing

“…Greengard et al (2020) employ t-distributed stochastic neighborhood embedding (t-SNE) to cluster risk factors into 6 groups. In similar spirit, Geertsema and Lu (2020) use agglomerative clustering to group anomalies based on correlation-based dissimilarity. In the context of industrial organization, Hoberg and Phillips (2016) group firms into industries using a clustering algorithm on the text of 10-K product descriptions, and Hoberg and Phillips (2018) document momentum effects using text-based industries.…”

Precious Metals Comovements in Turbulent Times: COVID-19 and the Ukrainian Conflict

“…Multivariate analysis and alternative correlation methods have also been used by other authors in the literature for examining the associations between various assets classes. Examples include the Pearson correlation coefficient (Geertsema and Lu 2020), time-varying correlation methods (Chiang et al 2007), partial-correlation coefficients (Kenett et al 2015;Jung and Chang 2016), Fisher correlation (Krishnan et al 2009), dynamic conditional correlation (Engle and Colacito 2006), partial-distance correlation (Creamer and Lee 2019), de-trended cross-correlation (W ątorek et al 2019), and the multiscale partial-correlation coefficient (Michis 2022).…”