The traditional method of using aggregated monthly data to measure long run risk is flawed and inaccurate, according to important new research by State Street. Co-authors David Turkington, Will Kinlaw and Mark Kritzman have found that there is a huge divergence in risk and return over long periods, which is not visible when using measures such as volatility and correlation derived from monthly data.
Typical measures of risk over three year periods use estimates based on monthly data, however there is a time series effect which means that data is not an accurate reflection of reality.
Their research, which is the subject of a forthcoming paper in the Journal of Portfolio Management, looks at the performance measurement effect of this and measures three data sets: mutual fund performance, hedge fund performance, and risk parity strategies.
David Turkington, managing director and head of investment and risk research at State Street Global Exchange, says the research has important implications for performance measurement.
“We found that measuring across manager or asset allocation strategies, that what is superior depends on the time horizon,” he says.
Further the difference can be quite dramatic, as measured by the hedge fund universe where the quartile ranking of hedge fund performance changes dramatically when the denominator is changed.
“The numerator or return doesn’t change but the risk you think you’re exposed to is very different when you look at performance with monthly data versus yearly data,” he says.
“Risk parity is also found to have superior risk adjusted performance but that can be the opposite when you use three to 10 year data.”
The motivation for this performance divergence concept was the observation that certain asset classes, for example US and emerging market equities, are very correlated using monthly data.
“You wouldn’t expect that there is divergence over three years, but there is and it is meaningful. There are time series effects,” he says.
This is important as typical measures of risk, and the available technology to investors, uses three year data estimated on a monthly horizon.
“It isn’t recognised how bad an approximation of reality it is,” he says. “Clients have long run risk and return targets but they are not measuring the long term risk appropriately.”
The fact that risk measurement is not accurate has implications for portfolio construction.
“It may be that one portfolio cannot run or manage the short term and long term risk at the same time, investors might have to choose between the two,” Turkington says. “Most investors care about both long horizon and within horizon risk. There are a bunch of portfolios better suited to long term objectives that aren’t being evaluated.”
An example, he says, from the asset owner perspective is that on a month to month data set fixed income looks like a better hedge for liabilities, but over the long horizon that doesn’t have the growth aspect for hedging liabilities.
“Equities may be a better hedge for the growth of liabilities.”
The research has important implications for investors, and provides them with additional metrics to look at when assessing managers, strategies or asset allocation decisions.
“There are a striking number of examples where there is large divergence and it is not always in the same direction, divergence could be less or more than expected, so the effect for asset owners is dependent on their portfolios.”
State Street Global Exchange is developing a suite of web-based tools, called Investment Labs, which apply its research concepts and allow investors to analyse and monitor different regimes and risk signals.
The first of these is Risk Lab which pulls together a dozen or so years of research around market turbulence and absorption ratio as a measure of fragility and can compute the risk indices of price returns off any assets.
“This enables any data set to be loaded and evaluated over history, so asset owners can use their own real data. It is a more personalised way to monitor risk.”