Banks shelving alternative beta products at increasing pace

Data suggests overfitting partly to blame for mothballing of two in five strategies

In the past 12 months, banks have decommissioned 582 alternative risk premia products

Fresh data shows banks are retiring defunct risk premia products at a quickening rate, raising questions about whether some of the products in question have been poorly designed.

Critics say offerings are being shelved due to overfitting – when strategies are engineered to look strong in backtests, but then struggle once launched.

“It appears retired strategies on average suffer a larger performance deterioration after their live date [compared with other strategies], which may explain – at least in part – why they have been terminated,” says Matthias Lennkh, founding partner at asset manager Clear Alpha, which collected the data.

Others argue the high retirement rates are simply a by-product of how banks operate in the area.

In the past 12 months, banks have decommissioned 582 alternative risk premia products, compared with around 464 in the calendar year 2016 and 320 the year before, the data shows.

Roughly two in five strategies have been withdrawn since 2010, with 2,434 strategies now live and 1,677 products retired. Grouping together products that offer the same strategy but in various currencies or with different amounts of leverage compresses the dataset to around 300 live strategies and 110 retired strategies.

Forex carry has the highest percentage of retired strategies.  

Investors’ interest in risk premia investing has spiked in recent years and banks have sought to meet the demand by offering so-called alternative beta products based on long and short exposures to risk factors such as value or momentum. But concerns persist about the risks of overfitting in this sector and in factor investing more widely.

Sharpe ratios for retired strategies fell 90% after going live compared with backtests, the data indicates. That compares with a – still high – deterioration in Sharpe ratios of 73% for equivalent live strategies, according to a 2016 study authored by Clear Alpha with Antti Suhonen, a professor of finance at Aalto University in Finland.

Investors typically allow for a 50% deterioration in Sharpe ratio for systematic strategies when looking at backtests as an indicator of probable live performance.

As well as overfitting, regulatory changes are also to blame for the increase in retirement rates, Lennkh says.

The International Organization of Securities Commissions (Iosco) issued tougher guidance in 2013 on administering benchmarks in response to the Libor fixing scandal, seeking to stamp out conflicts of interest and improve transparency in how benchmarks are managed.

“Banks realised it was going to become ever more expensive to maintain and manage indexes from a regulatory perspective, so strategies that didn’t have assets under management have come under greater scrutiny and may have been retired,” Lennkh says.

“Overfitting is a huge problem in the space,” says Tony Morris, head of quantitative investment strategies research at Nomura. “On the other hand, we shouldn’t beat up the banks without remembering that exactly the same behaviour happens in fund management companies, because the products are very much a function of the same quantitative process.”

He adds: “Many people have a difficult time distinguishing between a model that explains history very well as opposed to a model that will explain the future.”

Clear Alpha’s data does not include the buy side.

Pierre Mendelsohn, chief executive of independent advisory firm Alpima and a former global head of retail structured products at Morgan Stanley, says banks are taking “huge care” to ensure they produce sensible products.

“As a business, you experiment, you launch, you don't necessarily have a seed or an order to start, and if some products end up not raising sufficient assets or not performing, then naturally they die. [By comparison] the survival rate of companies and products created in the real world is much lower.”

However, Andrew Beer, managing partner at Beachhead Capital Management, says experimentation cannot explain the rate of retirement. “New products aren’t sold as experiments but rather as finished products that are the culmination of extensive research,” he says. “If the product underperforms expectations, someone made a mistake.”

Investors are in danger of misjudging the likely performance of risk premia products if they ignore strategies that have been decommissioned, he says. “Banks effectively say: ‘The battle wasn’t that bad. Look at all the survivors!’ But there are bodies littered across the battlefield. They want to pretend the bodies aren’t there.”

Clear Alpha’s database contains the quantitative investment strategies offered to clients by Bank of America Merrill Lynch, Barclays, BNP Paribas, Citi, Commerzbank, Credit Suisse, Deutsche Bank, Goldman Sachs, HSBC, JP Morgan, Morgan Stanley, Natixis, Nomura, RBS, Societe Generale, UBS and Unicredit.

The other banks covered in the research declined to comment for this article.