In practice, I have found that as a portfolio manager I only spent about one-third to one-half of my time figuring out my picks. All that other “messy” stuff, if improperly managed, can lead to distressing results. Some examples are:
- Our diagnostics show that our selection process worked, but why did we underperform?
- We got fired over a misunderstanding???
- I’d hate to tell you this but John the portfolio manager and Mary the trader are continuously at each others’ throats…
This is one in a series of posts on all that "messy" stuff: What do you do after you’ve made your picks. I would emphasize that there is no one size fits all answer. Your mileage will vary. (See part 1 on Reading your client here).
Portfolio Construction: how much to buy and sell
If the selection process is about deciding on what to buy and sell, portfolio construction is about deciding on how much to buy and sell. I would break down this process into the following steps:
- Deciding on your benchmark
- Deciding on what your bets are: minimizing your un-intended bets and properly sizing your intended bets
What are your bets?
You should only make bets only when you have an edge. What is the essence, or the underlying themes, of your selections and how confident you are about them?
Risk models can help and I am a big fan of them. A portfolio manager with a risk model can see more easily see his bets and therefore eliminate or minimize his un-intended bets and properly size his intended bets. Size the intended bets according to Grinold’s principles: a manager’s value-added (Information Ratio) is a function of his selection skill (Information Coefficient) and the number opportunities (N) he has.
There is no one size fits all solution in choosing risk models. It depends on your selection process. A top-down manager should probably use a risk model that focuses mainly on macro-economic risk factors to analyze his portfolio. A traditional bottom-up stock selector or sector rotator might want to use a fundamental factor model, such as the one pioneered by Barra. Traders with shorter term time horizons may be better served by principal component models, as offered by firms such as APT and Northfield.
Should you optimize your portfolio?
Some managers use risk models just to analyze and understand their risk exposures. Others take the additional step of asking the risk model to construct the portfolio for them through an optimization process. This quantitative technique may not be suitable for investors with fundamentally driven processes as this group often have trouble numerically specifying many of the inputs to the optimizer.
Portfolio Optimization: What kind of painter do you want to be?
Managers who use optimization need to understand the nuances of the optimizer and how it interacts with the forecast alphas. I would use the analogy of being a painter and knowing what you want to paint. A quant with an index-plus, or a low tracking error active mandate, will keep the risk aversion parameter high with fairly low forecast alphas. This would be the equivalent of painting a series of subtle colors with smooth transitions between colors.
One of the frustrations of the optimizer output from index-plus style optimizations is that the optimizer will often replace one stock ranked "hold" with another that is ranked "hold" for risk control reasons. If the intent is a to build a "pedal-to-the-metal" portfolio, then the manager needs to take steps to emphasize the tails, or extremes, of the forecast alpha score distributions. In other words, only buy stocks ranked "buy" and sell stocks ranked "sell". This would be the equivalent of painting a bright colorful mosaic, compared to the dull but subtle colors of the index-plus mandate.