Aligning Factor Data Frequency with Model Labels and Rebalancing
Summary
This note explains how to handle low-frequency financial or analyst factors alongside daily price and volume factors when constructing nonlinear models. One approach is to merge monthly or quarterly fundamentals with daily market data and carry the latest available fundamental values forward. The resulting stock selections can then support daily rebalancing.
If a model uses only low-frequency fundamentals, the note recommends making the prediction labels and portfolio review schedule match that slower frequency instead of converting the inputs to daily observations. It cautions that this choice reduces the amount of training data and may raise overfitting risk. The document offers general guidance rather than empirical comparisons: it provides no model specification, dataset, validation results, or evidence identifying which approach performs better in a particular market.
Key ideas
- Monthly or quarterly financial factors can be merged with daily market factors by carrying values forward.
- Daily rebalancing is one possible schedule when model inputs include daily price and volume data.
- For models using only low-frequency data, labels and portfolio decisions should use a compatible frequency.
- Using low-frequency observations can reduce the training sample and increase overfitting risk.
- The guidance does not include a tested comparison of the alternative approaches.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.