Will Kinlaw and Jay Moore discuss how pension funds can avoid traffic jams, road construction and the associated costs of delays as they get portfolios back on the road to strategic asset allocation weights.

Whether it's a built-in navigator, a portable navigation device or a smart phone app, today's drivers frequently turn to navigation systems to help them arrive at their chosen destination safely and on time.

While driving, navigation systems can ‘recalculate' a planned route to avoid delays or correct wrong turns. When asset owners set out to maintain the asset mix in their portfolios, as defined by their strategic asset allocation, they face a similar — but far more complex — task.

Car navigation systems have a limited number of ‘inputs' to consider when planning a route. By contrast, the decision to rebalance a portfolio encompasses hundreds of variables. The potential costs are also greater. These include transaction costs such as commissions, bid/offer spreads and taxes, as well as market impact and opportunity costs. For asset owners, the most significant of these are opportunity cost — in the form of tracking error between the legacy and target portfolios — and market impact on pricing from the purchase and sale of securities.

Most institutional investors have fully embraced the notion that periodically rebalancing their portfolio is necessary to maintain the strategic asset allocation. However, many use sub-optimal approaches for doing so. The need for portfolio rebalancing can be caused by a variety of factors such as asset valuation changes, manager reallocations, or short-term tactical views. Sub-optimal rebalancing techniques often incur upwards of 50% greater costs than a more sophisticated approach available with the improved technology at our disposal.

In theory, investors should rebalance portfolios whenever the cost of sub-optimality (from drifting from the target asset allocation) exceeds the cost of restoring the optimal weights. But many ignore this trade-off, falling back on standard rebalancing approaches.

Two traditional approaches to rebalancing include calendar-based strategies (which rebalance to target allocations on a periodic, pre-determined schedule), and tolerance-band approaches (which trigger rebalancing whenever a portfolio's asset mix drifts outside of a pre-determined band, above or below a set target). Both approaches ignore the trade-off between the costs of rebalancing and the materiality of the deviations from the target portfolio. They also ignore the fact that current actions affect future decisions and costs.

Since the global financial crisis, institutions have become more cognisant of the fact that return, liquidity and cost distributions among different asset classes can be highly non-normal. For example, strategies involving derivatives and certain hedge funds may exhibit highly non-normal returns with inconsistent liquidity (hence higher trading costs). Investors have also come to realise that they are much more sensitive to downside deviations from mean returns than they are to upside deviations. This sensitivity — or investor utility — as well as the possibility of non-normal returns and higher trading costs, should be accounted for when rebalancing the portfolio. These types of risks provide meaningful insight to portfolio rebalancing, yet are often accompanied by significant computational challenges.

Fortunately, advances in computer processing have provided a solution. Investors can use an alternative optimisation methodology called full-scale optimisation, which seeks to maximise expected portfolio utility over multiple periods in a sample. Multi-period optimisation technology can consider as many portfolio trading scenarios as necessary to identify decision points that yield the highest expected investor utility.

Driven by massive parallel processing, this dynamic optimisation process can comprehensively analyse thousands of investments and millions — or even billions — of hypothetical investment scenarios. Full-scale optimisation takes into account all features of costs and return distributions, including skewness, fat tails and correlation asymmetries, thereby providing an improved picture of the inherent risks. It then generates trading rules for a specified time horizon. Essentially, the algorithm creates a roadmap that can be used until the assumptions on which the portfolio is built change, just as with a car's navigation device.

Full-scale optimisation presents a particularly attractive option for institutional investors with an aversion to losses below a specified threshold — for example, those facing reserve requirements, loan covenants, or risk of insolvency or termination. As many investors substantially underestimate within-horizon exposure to (or tolerance for) loss, this category may encompass many institutions. Of course, it is important that investors fully understand their liquidity requirements and liability schedules, taking on appropriate risk to maximise wealth and grow income. Minimising ‘alpha decay' between rebalancing dates may also be a goal, particularly for quantitatively driven strategies.

The vast scale of this kind of scenario building requires extraordinary technological resources. It uses grid computing and parallel processing to achieve complex computations. Grid computing relies on multiple networked computers to distribute process execution across a parallel infrastructure, enabling faster processing of large-scale computation. But, even using 28-processor grid computing and parallel processing to speed up the computations, one can only tackle a portfolio with less than 10 assets. Fortunately, in 2009, Harry Markowitz and quantitative investment manager Erik van Dijk created an algorithm that reduces the complexity of the problem and makes rebalancing feasible for up to 100 assets.

Not only is such a program flexible in terms of the number of assets it can analyse and for which it can provide a roadmap, it is also customisable. Optimal rebalancing can be customised to accommodate investors' tracking-error target, cash inflows, benefit payments, and maximum allowed deviations. The goal of optimal rebalancing is to trade less often and more meaningfully to ensure minimal tracking error against pre-determined preferences. Additionally, a transition manager can further manage the short-term portfolio deviations through the use of futures overlay and cash equitisation strategies when proper rebalancing thresholds are not met.

As quantified by State Street's experience, optimally rebalancing can reduce transaction costs by as much as 50%, compared with a simple 2% tolerance-band approach. For example, by using dynamic processing to set a rebalancing schedule, an institutional investor with a $1bn portfolio allocated among four assets could save some $600,000 in trading costs.

Optimal rebalancing achieves this result while also delivering lower sub-optimality costs. Plan sponsors and their transition managers may also consider using futures to rebalance assets, hence avoiding the substantial unwanted costs associated with the trading of physical securities.

Today's institutional investors face the challenge of comprehensively rebalancing portfolios in the face of dramatic increases in the scope, detail and timeliness of financial data, not to mention a proliferation of asset and security types. Such an environment cries out for an alternative to simple, mechanistic approaches, one that allows an investor not only to explicitly weigh the tradeoff between the costs of sub-optimality and of transactions but also to account for the fact that a rebalancing decision made today affects the rebalancing decisions available in the future.

A multi-period optimisation program addresses this urgent need among institutional investors. Using such a program powered by grid computing and parallel processing, asset owners can ‘recalculate' their route as needed and receive a customised roadmap that spells out the best response to any deviation from the optimum target asset mix.

William Kinlaw is managing director and head of portfolio and risk management research at State Street Associates in Cambridge, Massachusetts; Jay Moore is managing director and head of currency management and portfolio solutions strategy at State Street Global Markets in Boston