Good variance, bad variance, and stock return predictability

Abstract

Abstract In the stock market, past winners and losers usually have different attitudes to variance risk. In light of this, we decompose stock variance into good and bad variances, and further construct a composite signed variance. The recursively estimated coefficient of standard variance is volatile over time, but the ones of good, bad, and signed variances are fairly stable. This suggests the high efficiency of our decomposition approach. We provide convincing evidence that bad variance and composite signed variance can significantly predict future stock returns both in‐ and out‐of‐sample. Furthermore, a mean–variance investor can realize substantial economic gains by using our bad and signed variances relative to the traditional variance predictor and simple mean benchmark. Our decomposition approach can outperform other similar but complex techniques, including threshold regression and Markov regime switching. The results are consistent across multiple robustness tests.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 26 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.