Cortese Kolm Lindström 2023
“What drives cryptocurrency returns? A sparse statistical jump model approach” was published in Digital Finance (Springer) 5(3), pp. 483-518, by Federico P. Cortese, Petter N. Kolm and Erik Lindström (SSRN abstract 4330421, dated 19 January 2023). It applies the sparse variant of the Statistical Jump Model — the version from Nystrup Kolm Lindström 2021 that jointly performs feature selection, parameter estimation and state classification — to cryptocurrency returns, to infer which features actually drive crypto regime dynamics.
The central finding is that a three-state model — bull, neutral, bear — best describes the dynamics of the largest cryptocurrencies. Out of the candidate features, the paper identifies first moments of returns, features representing trends and reversal signals, market activity, and public attention as the key drivers of crypto market dynamics. The emphasis throughout is on interpretability and robustness: the sparse jump model selects a small, interpretable feature set rather than fitting an opaque high-dimensional model, which the authors argue is a virtue given the noisiness of crypto data.
It appears in this vault as the clearest extension of the jump model to a new asset class — the Cryptocurrency Market — and as partial independent uptake. Lead author Federico Cortese (University of Milano-Bicocca) sits outside the Nystrup/Princeton research groups, which strengthens the evidence relative to a fully in-network paper; however, Kolm and Lindström co-author, so it is not a clean independent replication. The work is reproducible (the sparse jump model is available in open-source form), hence replication_available: yes.
Critically, this is a regime-classification and feature-identification study, not a trading study. No trading strategy, asset-allocation backtest, transaction costs, or profitability result appears in the paper — its contribution is characterising what drives crypto returns across states. Its profitability grade is therefore inconclusive: it confirms that the sparse jump model produces interpretable, persistent regimes on a volatile new asset class, but it provides no evidence — for or against — a tradeable, cost-net edge in crypto. This places it alongside Aydınhan Kolm Mulvey Shu 2024 as a methodological extension of the jump-model family rather than a profitability proof point.
Connections
- Statistical Jump Model — detects_regime, 2023, source: https://doi.org/10.1007/s42521-023-00085-x
- Cryptocurrency Market — detects_regime, source: https://doi.org/10.1007/s42521-023-00085-x
- Regime Classification — detects_regime, source: https://doi.org/10.1007/s42521-023-00085-x
- Nystrup Kolm Lindström 2021 — relates, source: https://doi.org/10.1007/s42521-023-00085-x
- Petter Kolm — proposes_model, source: https://doi.org/10.1007/s42521-023-00085-x
- Aydınhan Kolm Mulvey Shu 2024 — relates, source: vault synthesis
Ontology
Cortese Kolm Lindström 2023 defines Statistical Jump Model Cortese Kolm Lindström 2023 supports Regime Classification Cortese Kolm Lindström 2023 relates Cryptocurrency Market
Sources
- Cortese, F. P., Kolm, P. N. & Lindström, E. (2023). “What drives cryptocurrency returns? A sparse statistical jump model approach.” Digital Finance 5(3), 483-518. https://doi.org/10.1007/s42521-023-00085-x
- SSRN preprint (abstract id 4330421, 19 January 2023). https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4330421