Factors

Nine factors that follow the factor definitions in Mählmann and Sukonnik (2022) and Mählmann and Reck (2024), adapted to the information available in public fund filings.

Code	Factor	Signal	Preferred side
`value`	Value (PD)	Spread-to-maturity divided by a model-implied 12-month default probability	High
`momentum`	Momentum	Cumulative excess return over months t−3 to t−1	High
`illiquidity`	Illiquidity	Number of funds holding the loan (fewer = less liquid)	Less liquid
`quality`	Quality (PD)	Model-implied 12-month default probability	Low
`low_risk`	Low Risk	Spread duration	Low
`carry`	Carry	Spread-to-maturity	High
`ivol`	IVOL	Idiosyncratic volatility from a 12-month market model	Low
`bab`	BAB	Market beta from a 12-month market model (Dimson beta)	Low
`combi`	Combi	Average of the standardized signals of the eight factors above	High
`market`	Market	All loans in the investable universe	–

Reading the low-risk factors. For IVOL, BAB and Quality the preferred quintile holds the safest loans. Their raw long-short returns are often negative, because riskier loans earn higher raw returns. The anomaly shows up in risk-adjusted terms: the safest loans earn more than their low market exposure implies. Judge these factors by their alpha in factor_summary.csv, not by the raw long-short spread.

How the premia unfold over time

Factors differ in how long their premia last after portfolio formation. Momentum peaks two to four months after formation and fades after about six months. Low volatility (IVOL) and illiquidity earn their premia persistently. This is why the database offers several holding periods.

Dimson alpha of selected factor portfolios by month after portfolio formation, months 1 to 12. — **Figure 2.** Dimson alpha (annualized, basis points) of the preferred quintile against the benchmark, by month after portfolio formation. Equal-weighted portfolios.

Portfolio construction

Each month, loans in the investable universe with an available signal are sorted into quintiles.
The long-short return is the preferred quintile minus the opposite quintile. The benchmark is the average of all sorted loans.
Longer holding periods use overlapping cohorts with weights fixed at formation.
The implementation lag delays the first holding month by one month. N-PORT data are published with about 60 days delay, so the lagged series approximate an implementable strategy.

Details are in the methodology.