Daisuke Hamaguchi says investors who rely on probability in distribution underestimate risk
The chief investment officer at Japan’s Pension Fund Association, Daisuke Hamaguchi, warned about over-reliance on statistics at an AsianInvestor forum yesterday, rubbishing the practice of forecasting and taking a swipe at quantitative methodology and smart beta.
Delivering a frank opening address at AsianInvestor’s fourth annual Japan Institutional Investment Forum in Tokyo yesterday, Hamaguchi argued that investors would be unwise to rely on standard deviation as a tool to measure risk because the numbers can't be depended on.
“It’s like focusing on images in the rear-view mirror, looking backwards when you are driving forwards,” said Hamaguchi, who has responsibility for investing PFA’s ¥12.7 trillion ($104.5 billion) in assets. Of this, the institution manages about ¥5 trillion in-house in domestic equities and domestic and global fixed income.
While conceding that some in the audience might think his comments were off the mark, he highlighted the first bullet point that PFA uses to explain its cultural philosophy: that markets are uncertain and cannot be predicted.
“In the field of investment management, especially for a CIO to assert this philosophy consistently is quite difficult,” said Hamaguchi.
He noted that pension beneficiaries assumed that investment professionals could forecast markets accurately. “From our side, to say we don’t know and can’t forecast, to put this in writing in the first bullet point on our list is quite courageous, but we feel it is important,” he stated.
The key point of his address was that standard deviation – which he described as a combination of asset distribution and probability – could not be relied on as a tool to measure risk.
According to standard deviation modelling, he noted, the chances of equity valuations halving within a year were put at once in several centuries, yet this had happened twice in the past 15 years.
“When you try to express the movement of the market statistically in the form of distribution, I think this starting point is where problems begin,” said Hamaguchi.
He noted the assumption was that the data points used to create a picture of standard deviation were independent and did not influence one another. But in reality market movement was continuous and when investors become aware of risk they all acted together, meaning there was a chain reaction.
“In reality data points are not independent,” he observed. “When you try to deal in everything to do with distribution and probability you go wrong in one place or another. We do not need to be digital all the time, we need to think in analogue terms.”
His argument was that investors who relied on the tradition of using probability in distribution tended to underestimate risk. In this he specifically referred to quantitative methodology and smart beta.
He noted how quant-based funds in 2007 had accumulated similar positions based on similar logic, then all of suddenly created confusion by disposing of assets.
“There was a downturn in the market and constant distribution and probability calculations were made,” he said. “But there were hidden risks and these led to a chain reaction. On a personal level I am quite concerned about smart beta.”
He expressed further concern that the market had become so grounded on the concept of standard deviation in terms of risk management and analytics.
“They [money-managers] calculate it because they want to sound as if they are doing a lot of work and they want to sound professional,” he added. “But I tend to vacillate. I say we should not base ourselves on such numbers because they are defective.
“If you are aware of that and you are careful in using these numbers, that's ok. But as a measure of absolute risk they should not be depended on.
“In our industry, in various forms and under different names, calculations come into our work. We see an overabundance of numbers, so we need to be careful and aware of what the calculations are based on.”
¬ Haymarket Media Limited. All rights reserved.