The Right Mix
IN PRINT ARCHIVE CIR Winter 2007
By Sébastien Page, senior managing director, State Street Associates / State Street Global Markets
Most institutional investors employ asset-liability optimization to determine optimal portfolio weights. These optimizations invariably begin decaying upon implementation as changes in asset prices drive portfolios’ component parts away from optimal targets. This process is made more complex by the fact that different investments, in various asset classes, move away from their optimal targets at different speeds as their performance varies. In an idealized world—with no transaction costs—investors would continually rebalance to their optimal weights in real time. But because of various transaction costs, including commissions and taxes, market impact and opportunity costs, investors must balance the cost of portfolio suboptimality against the cost of restoring the optimal weights.
Two suboptimal approaches to portfolio rebalancing currently prevail in our industry—calendar-based and tolerance-band rebalancing. The problem with calendar-based rebalancing is that markets might shift substantially between periodic rebalancing or, conversely, that rebalancing on a given date might be unnecessary as assets have not drifted enough to merit the rebalancing.
Tolerance-band rebalancing is generally thought to be an improvement over calendar-based rebalancing. The problem with tolerance-band rebalancing is that different portfolio components have different levels of elasticity. Certain volatile investments might see their prices shift substantially through the course of normal trading, while others might move less dynamically. A 3% drift for one investment might ring alarm bells, while a 5% drift in another might be a routine fluctuation meriting no action whatever. As with calendar-based rebalancing, the problem is identifying the sweet spot between tolerating and over-reacting to portfolio drift, taking into account the impact of each over- or under-weight position on the entire portfolio’s return-risk profile.
The construction of this roadmap takes into account all possibilities, and determines optimal decisions by a recursive process. While it is impossible for investors to predict how their portfolio will drift over time, this process works backwards through billions of scenarios to determine future optimal decisions for each possible future portfolio state. The probability-weighted future costs implicit in the roadmap are then discounted, and optimized against the expected suboptimality and transaction costs in the current period, for every possible portfolio state, until the entire roadmap is constructed. In this way, the optimal rebalancing roadmap indicates the course of action for every possible portfolio state in every time period.
As might be expected, the vast scale of this kind of scenario building requires extraordinary mathematical and computational resources. It uses grid computing and massive parallel processing to achieve these vast computations. Grid computing relies on multiple networked computers to distribute process execution across a parallel infrastructure, enabling faster processing of large-scale computation problems. As quantified in substantial back-testing and simulations, the process is expected to substantially reduce transaction costs by as much as 50% compared to, for example, a simple 2% tolerance-band approach. Moreover, the algorithm achieves this result while at the same time delivering lower suboptimality costs.
Providers use futures to rebalance assets, thereby avoiding substantial unwanted costs associated with the trading of physical securities. Throughout the process, portfolios are kept in the market, thus avoiding opportunity costs that can arise when overweight positions are converted to cash when physical securities are sold. Fund managers can focus on implementing strategy, not on raising cash. Clients’ portfolio data are downloaded daily from custodians, with rebalancing performed by transition management teams as needed. Portfolios are constantly monitored and documented through monthly performance and cost analytics reports.
Sébastien Page's presentation, click