Resampled Portfolio Optimization and More Stable Allocations
Summary
The document discusses resampling as an alternative to traditional mean-variance optimization (MVO). Its main practical benefit is that the resulting portfolio allocations may vary less over time, which can reduce how often a portfolio needs to be rebalanced. The author of the question reports seeing more stable weights in experiments, but no improvement in returns.
The answer frames resampling as a cost tradeoff rather than a guaranteed performance improvement. Less frequent rebalancing may lower transaction costs, but any savings need to outweigh the potential cost of using allocations that differ from those produced by conventional optimization. The document acknowledges debate over whether resampling has statistical justification and refers to differing views in research, but it does not review that evidence or explain a specific resampling procedure. Its guidance is therefore practical and conditional, not a conclusion that resampling is broadly superior.
Key ideas
- Resampled optimization may produce allocations that change less over time than traditional mean-variance optimization.
- More stable weights may reduce rebalancing frequency and transaction costs.
- The method is worthwhile only if cost savings outweigh any disadvantage from different allocations.
- The document notes disagreement about resampling but does not resolve its statistical justification.
Tags
Full text
# portfolio optimization averaging weights, what are benefits? # portfolio optimization averaging weights, what are benefits? I'm playing around with different portfolio optimization techniques. Amongst others I was also looking at the resampling method, especially the one described in Meucci. I have two general questions regarding this technique. Question I would like to know what more expierenced people think about resampling methods? I've noticed that there is a controversial discussion on this site, see this question, as well as in the research area. For example, Scherer critique. On the other Meucci also points out some advantages and mentions the wide usage of this technique in industry. In my opinion, or what I've seen running some expirements, there is no additional gain regarding return but weights are much more stable when you have to recalibrate them. Often it is argued that there is no statistical justification for doing it. There is also a newer paper which gives a justification depending on the "traders style". As you can see there is a variety of articles and I would like to hear from somone with a better overview his / her opinion on the topic. ## Answer by Logic9 (score 1) https://quant.stackexchange.com/a/34250 The benefit to using a resampled efficient frontier is a matter of practicality. Suggested allocations are more stable over time compared to the outputs of a traditional MVO and as a result can require less frequent rebalancing and transaction costs. If rebalancing frequency and transaction costs are an issue for a strategy then it might be worthwhile to consider the resampling method. If the cost savings from less frequent rebalancing are greater than any negative expected impact from a differential between the output from a resampled optimization versus a traditional optimization then use of the resampling method would be justifiable.
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.