Paper: arXiv 2405.15721

Authors: Adam Baybutt

Abstract

We develop novel estimation procedures with supporting econometric theory for a dynamic latent-factor model with high-dimensional asset characteristics, that is, the number of characteristics is on the order of the sample size. Utilizing the Double Selection Lasso estimator, our procedure employs regularization to eliminate characteristics with low signal-to-noise ratios yet maintains asymptotically valid inference for asset pricing tests. The crypto asset class is well-suited for applying this model given the limited number of tradable assets and years of data as well as the rich set of available asset characteristics. The empirical results present out-of-sample pricing abilities and risk-adjusted returns for our novel estimator as compared to benchmark methods. We provide an inference procedure for measuring the risk premium of an observable nontradable factor, and employ this to find that the inflation-mimicking portfolio in the crypto asset class has positive risk compensation.

Complexity vs Empirical Score

  • Math Complexity: 8.5/10
  • Empirical Rigor: 6.0/10 (capped from 7.0: transaction costs not modelled; reported Sharpe ratio of 3.3 is implausibly high for daily or lower-frequency data; see penalty flags)
  • Quadrant: Holy Grail — high math complexity, high empirical rigor (effective empirical rigor capped from 7.0)

Why this score: The paper presents advanced econometric theory with double selection Lasso, high-dimensional PCA, and asymptotic inference, demanding high math sophistication. It is data- and implementation-heavy with crypto asset characteristics, out-of-sample backtests, and a GitHub repository for replication code.

Research Flowchart

  flowchart TD
  A["Research Goal: Develop a novel estimation procedure for a dynamic latent-factor model with high-dimensional asset characteristics."] --> B["Methodology: Double Selection Lasso Estimator"]
  B --> C["Data: High-Dimensional Crypto Asset Characteristics"]
  C --> D["Computation: Regularization & Asymptotic Inference"]
  D --> E{"Findings / Outcomes"}
  E --> F["1. Superior Out-of-Sample Pricing & Risk-Adjusted Returns vs. Benchmarks"]
  E --> G["2. Inference Procedure for Risk Premium of Nontradable Factors"]
  E --> H["3. Inflation-Mimicking Portfolio in Crypto shows Positive Risk Compensation"]

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper introduces the Double Selection Lasso Factor Model (DSLFM), a new estimation and inference procedure for dynamic latent-factor models with high-dimensional asset characteristics.

Source sentence
The main contribution of this research is to de-velop new estimation and inference procedures, termed the the Double Selection Lasso Factor Model (DSLFM), to fit the latent factors f t+1 and loadings Γβ in (1) and to con-duct standard asset pricing tests under the novel setting of high-dimensional asset characteristics.

Assumes 4

  • The number of latent factors k is low-dimensional while the number of characteristics p is high-dimensional (on the order of N and T). model
    Source sentence
    nt loading parameter; and, zi,t ∈R p are time-varying asset-specific characteristics where p is high-dimensional, i.e., on the order of N and T. Crucially, we follow the established practice in the literature of assuming the number of factors k is low-dimensional, i.e., N, T, p >> k ∈{1, 2, 3, . . . }.
  • Both the cross‑sectional dimension N, time dimension T, and characteristic dimension p grow to infinity (p, T, N → ∞). statistical
    Source sentence
    First, we develop estimation procedures and large-sample theory that allows p, T, N →∞.
  • Characteristic portfolio returns are uniformly bounded by a constant M (Bounded Characteristic Portfolios). statistical
    Source sentence
    (i) Bounded Characteristic Portfolios: For a fi-nite absolute constant M and ∀t, j, ct+1,j = Γ⊤ β,j f t+1 < M. (ii) Sparse Loading: Loading matrix Γβ admits an exactly sparse form.
  • The loading matrix Γβ is exactly sparse, having at most s non‑zero rows. statistical
    Source sentence
    That is, for ∃s ∈ N+, i.e. p > s ≥1, Γβ has at most s nonzero rows: P p j=1 1 n Γβ,j 1 > 0 o ≤s.

Data

Backtest sample: 2022 – 2022 · crypto assets · crypto · weekly

  • Crypto asset returns panel · Coin Metrics, CoinMarketcap, Glassnode crypto-onchain commercial
    crypto assets
    Source sentence
    Thank you to Coin Metrics, CoinMarketcap, and Glassnode for providing academic research discounts on data purchases. We utilize the same novel and rigorous panel of crypto asset returns developed in XYZ.

Costs

Not modeled A trading or execution result is reported without costs.

Source sentence
Although the strategies replicated out of sample, we should note these results are again before transaction costs and are for a very small test period.

Finds 5

Reported performance (Dynamic Latent-Factor Model (DSLFM)): Sharpe 3.30 (benchmark 4.07)

  • The predictive R² for all three models and the benchmark were negative, indicating worse pricing ability than a random walk. performance
    Source sentence
    The predictive R2 for all three models and the benchmark model were all negative, performing worse in MSE pricing ability than a random walk.
  • The estimated inflation risk premium in the crypto asset class is statistically significant at 1.4 basis points (standard error 0.0097 bps), corresponding to a 7.3% annual excess return. statistical figure from elsewhere in the paper
    Source sentence
    The inflation risk premium was estimated to be a statistically significant 1.4 bps with a standard error of 0.0097 bps.
  • In low‑dimensional simulations, DSLFM yields smaller estimation errors for the loading matrix Γβ compared to IPCA. performance
    Source sentence
    In the low-dimensional setting of N = 500, T = 100, p = 10, s = 1, we find DSLFM to obtain smaller estimation errors for Γβ as compared to IPCA; however, IPCA has an order of magnitude lower estimation errors for F. DSLFM’s outperformance for the latent loading is driven by taking on higher bias yet substantially lower variance of the estimator; this is obtained from soft-thresholding many of the rows.
  • DSLFM slightly under‑covers the 90% and 95% confidence intervals in the low‑dimensional setting. statistical
    Source sentence
    DSLFM slightly under-covers the 90% and 95% confidence intervals in the low-dimensional setting.
  • All three multivariate observable‑factor models generate economically significant weekly excess return spreads of about 1% with Sharpe ratios between 1.31 and 1.72, beating the benchmark model. performance
    Source sentence
    However, although statistically insignificant, all three multivariate ob-servable factor models had economically significant weekly time series average excess return spreads of about 1% with associated Sharpe ratios of 1.31-1.72, which all beat the benchmark model with a return spread of 0.5% and a Sharpe ratio of 0.65.

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 21 related papers). Coverage is partial, so this is neither replication nor refutation.

Code

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.