Pitfalls in quantitative modelling according to S&P
Mitch Abeyta, managing director of Standard & Poors, said that quants did not use daily data 20 years ago. Today 76 percentage of company reports preliminary information – and all of this impact lagging assumptions.
Quarterly reporting lag for S&P500 companies in 1998 took 32 days, 60 days for annual reports. This year, it takes 22 days for quarterly reports, 45 for annual reports. 45 days was aggressive 20 years ago but it is getting better.
S&P found that conservative lagging leads to look behind bias and aggressive lagging leads look ahead bias although some months are more problematic.
He recommended the use of filing dates when possible, “it is clearly much better than making an estimate.” He observed that 4th quarter lags are greater than the other three, and that preliminary lags are increasing due to the increasing amount of information being recorded. But final lags are decreasing.
Estimated lags have some error built into them.
As for the restatements to data, companies who change (accounting, etc.) are very good short candidates. Between 1988 and 2008, 53% of the companies changed 2 to 6% of their data, and 5.8% changed more than 13%.
Alternative beta
“That hedge funds produce alpha is a widely held belief,” said Lars Jaeger, partner at Partners Group. But they have been trading in the most efficient markets.
There are different types of beta:
1. Traditional beta (exposure to equity markets, interest rates, credit risk, emerging markets). It does not take much to extract those betas.
2. Alternative beta (style factors, event risk, volatility, risk of commercial hedgers in futures markets, liquidity risk, spread risk). One needs various techniques (shorting, leverage, derivatives) to extract alternative beta. “This is pretty much what defines a hedge fund,” he said.
Source:
invest in HFs for the persistent benefits - alternative risk premium | QuantInvest conference...
Posts
No comments:
Post a Comment