Uncategorised posts

UniSuper’s proprietary risk program challenges investment assumptions

UniSuper, the $23 billion Australian pension fund for those working in higher education and research, has developed an in-house risk budgeting and factor analysis program that monitors the extent to which the fund deviates from its strategic asset allocation, and ensure the fund’s active risk is allocated appropriately between managers.

Drawing on past academic research, the head of research and risk management David Schneider and head of public markets Dennis Sams, have extended conventional models to set a minimum excess return hurdle at which active risk is appropriate, and encapsulate the extent to which the active risk assigned to each of the fund’s managers is consistent with the expected performance of those managers.

“The new model is causing us to question a lot of our assumptions,” Schneider said, pointing to its use of currency  hedging as an example.

Traditionally UniSuper has used currency hedging across all of its portfolios, now it is considering adjusting the level of hedging depending on the investment option’s level of risk.

The UniSuper Risk Budgeting and Optimisation System (TURBOs) has been designed by the fund to determine how each manager generates their returns; identify market factor exposures for each manager and aggregate these to determine overall factor exposures; and assess the expected future alpha for each manager.

Traditionally risk budgets have been derived with reference to tracking error, however for UniSuper, which has both a defined benefit and a defined contribution plan, Schneider believed this was lacking.

“This approach is appropriate for asset managers whose mandates are often specified in terms of tracking error limitations. However, the concern with this approach for institutions with guaranteed liabilities is that there is no direct interaction between the maximum tracking error and the fund’s liabilities,” he said.

In addition, he said, the choice of an appropriate tracking error budget is subjective, so an alternative approach to risk budgeting was determined.

UniSuper sets its strategic asset allocation to meet the investment objectives of the accumulation options, and pay the liabilities for the defined benefit division as they fall, so any deviation from the SAA is a source of risk to the fund. Specifically, adding active management adds risk to the fund.

“The marginal increase to risk is only justifiable if the fund’s expected alpha exceeds the benefit that could be obtained by changing the fund’s SAA benchmark, and moving along the fund’s constrained efficient frontier,” Schneider said. “This idea provides an inequality that is used in our risk budgeting formulation, namely that each option’s ex-ante alpha needs to exceed a minimum hurdle to justify a departure from beta allocations.”

The work undertaken by UniSuper is built up from prior work on risk budgeting – including that of Mena (2007), Litterman (2003), Scherer (2000), Kozun (2001) and Sharpe (2002). However the authors extended these prior findings by, amongst other things, removing the simplifying assumption that excess returns between managers are uncorrelated; introducing the idea that to justify active risk, one needs to exceed a hurdle return in excess of 0 per cent.

“A lot of risks focus on attribution but I was more worried about risk allocation,” Schneider said. “It must give insight into how to actively change the portfolio.”

In order to meet these objectives six processes were outlined to be computed: TURBOs:

1.attributes each manager’s returns between a series of market factor exposures (beta) and an observed ex-post (historic) alpha component. This step is resolved using factor analysis and multiple regression, with appropriate adjustments to manage collinearity, heteroskedasticity and co-integration

2.determines the ex-post total risk (volatility) and tracking error, and assesses the marginal and proportional contribution to that risk from each manger

3.uses Bayesian techniques to determine an ex-ante estimate of each manager’s alpha

4.assesses the extent to which the beat exposure differs to the fund’s SAA benchmarks

5.uses risk budgeting techniques to set a minimum excess return hurdle at which active risk is appropriate and assess the extent to which the hurdle is expected to be achieved

6.employs reverse optimisation to confirm whether the active risk assigned to each of the fund’s managers is consistent with the expected performance of the manager.

Asset Owner:
Join the discussion