Decile Portfolio Sorts for Testing Cross-Sectional Return Signals
Summary
A portfolio sort groups assets by a characteristic, such as expected return, and compares the subsequent performance of the groups. Researchers commonly set periodic breakpoints from the cross-sectional distribution of the chosen variable, then form portfolios that span low to high values. Returns can be equal-weighted or value-weighted, and the return spread between the highest and lowest groups is often used to assess whether the characteristic relates to later returns.
The document frames this as an empirical research method rather than a standalone trading strategy. It notes that portfolio formation frequency can vary with the signal: annual rebalancing is common in some studies, while momentum sorts may be refreshed monthly. Timing matters when signals rely on accounting data, since using information before it was publicly available creates look-ahead bias. The approach is nonparametric, but practical implementation may face high turnover costs and constraints on shorting. A related answer mentions Fama–MacBeth regression as a way to test a proposed factor.
Key ideas
- A portfolio sort ranks assets by a selected characteristic and groups them using cross-sectional breakpoints.
- Researchers compare group returns and often examine a long–short spread between the highest and lowest portfolios.
- Sorting is a nonparametric way to study relationships between asset characteristics and subsequent returns.
- Rebalancing frequency depends on the signal, and information timing must avoid look-ahead bias.
- Turnover costs and short-sale constraints can limit the practical performance of a sorted portfolio.
Tags
Full text
# name of this portfolio optimization strategy # name of this portfolio optimization strategy I have come across a portfolio selection strategy that buys in equal amounts the top decile of expected earners, and simultaneously short sells the lowest decile in a similar fashion. What is this strategy called? I am looking for words more specific than "market/dollar neutral." ## Answer by skoestlmeier (score 1, accepted) https://quant.stackexchange.com/a/41209 Short answer: This 'portfolio sort' is rather a common approach in empirical finance research, than a portfolio strategy. Your mentioned strategy is called 'portfolio analysis' or in more detail a 'univariat portfolio sort'. The general approach is to form portfolios of stocks, where the stocks in each portfolio have different levels of the variable posited to analyze cross-sectional relations. As a nonparametric technique, it does not make any assumptions about the nature of the cross-sectional relations between the variables under investigation. The steps are as follows: - Calculate periodic (e.g. yearly) breakpoints that will be used to group the entities in the sample into portfolios. The breakpoints are usually determined by percentiles of the sort-variable $X$ at time $t$ of the cross-sectional distribution. - Group all entities in the sample into portfolios. Each time period $t$, all entities in the sample with values of $X$ that are less than or equal to he first breakpoint are put in portfolio one, i.e. the 'low' portfolio. Portfolio two holds entities with values of $X$ that are greater than er equal to the first and less than or equal to the second breakpoint, and so on. - Calculate the equal- or value-weighted return for each portfolio for the subsequent year. - In addition to calculating the average return for each portfolio, you often calculate the return-difference of the high- and low-portfolio, that is a strategy investing long in the highest and short in the lowest portfolio. This difference portfolio is the primary value used to detect a cross-sectional relation between the sort variable and the outcome variable (here: return). It is common practice, to update the portfolios yearly. If you sort on the 'momentum' variable, monthly reformation is common. For example, Fama/French (1992) form portfolios based on size or book-to-market ratio at the end of June each year and calculate the return for the subsequent year. The formation at the end of June is due to the 'look-ahead-bias', that is to avoid to use accounting data for the previous fiscal-year end, which may not be publicly available until end June of the next year. Be aware to try exploiting this strategy: - Portfolio reformation results in high transaction costs, which may decrease the return significantly. - Due to regularization it is often not possible to fully replicate the short-strategy. References: Bali/Engle/Murray (2016), Empirical Asset Pricing: The cross-section of stock returns, 1. ed. Fama/French (1992), The cross-section of expected stock returns, The Journal of Finance. ## Answer by PF Chang (score 0) https://quant.stackexchange.com/a/41210 Related to skoestlmeier's answer, you should also check out https://en.wikipedia.org/wiki/Fama%E2%80%93MacBeth_regression that is precisely the regression to test the strength of a proposed risk factor (= long-short portfolio return)
Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.