2
$\begingroup$

In portfolio allocation literature there is lot of effort made in obtaining 'better' portfolio weights, for example via improving parameter estimates, introducing Bayesian approaches, incorporating constraints in the optimization procedure or by incorporating other predictive variables. However, in most empirical horse races (for example the famous work of DeMiguel et al) monthly reallocation is considered. Especially when taking transaction costs into account, most of the models fail as they loose performance due to estimation risk and the extreme weights cause high turnover. Therefore I am primarily interested in approaches that specify when to rebalance your portfolio instead of searching for fancier optimization techniques. What approaches do you know that monitor portfolio weights for 'structural breaks' in the allocation and help to specify the timing of reallocation?

$\endgroup$
1
$\begingroup$

vega captures the two most common solutions to this problem.

There are some valid criticisms of corridors as well. Because assets are correlated within a portfolio the decision to trade a particular asset should actually depend on the movements of other assets rather than having a corridor per asset. Also, finding the right corridor is often done using various forms of single period analysis, but this is really a multi-period problem (if I trade now what are the chances I'll have to trade next period).

Multi period dynamic balancing problems can be solved using dynamic programming. This solution really gets at the main tradeoff between tracking and costs. I realize this is to some extent this sounds like trading one "fancy optimization" for another, but the cost and correlations inputs are much more stable than return estimates so optimization can be expected to perform closer to the goal. Kritzman et. al. found a great trick to make dynamic programming scalable for larger portfolios as well.

$\endgroup$
2
$\begingroup$

I think generally there are two approaches: "calendar rebalancing" (such as monthly as you mention) and "optimal corridor width".

For the first option, the danger is the portfolio could stray considerably from your benchmark between rebalancing dates.

For the second option, track tactical deviation on a continuous basis. When you are outside the corridor, then rebalance. The corridor size depends on:

  • Transaction cost (higher cost -> wider corridor)
  • Correlations (higher correlation amongst portfolio assets -> wider corridor)
  • Volatility (higher volatility -> narrower corridor)
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.