Active management leaves fingerprints
Active Share as a portfolio measure – roughly, the share of portfolio holdings that differ from those in the benchmark index – was introduced by Cremers and Petajisto (2009) to quantify active management risk. However, holdings might be disclosed with a considerable delay or not be available at all. Research by, for example, Kacperczyk, Sialm and Zheng (2008) indicates that mutual fund holdings that are disclosed quarterly are not representative of actual investments.
Traditionally, active management risk has been quantified via tracking error. However, this metric does not distinguish between the effects of security selection and market timing. Hence, the same tracking error figure can be achieved with entirely different types of active management.
My paper in the Journal of Empirical Finance (2012) shows that different types of active management leave unique ‘fingerprints’ on portfolio returns. Perhaps somewhat surprisingly, it is possible to measure the effects of security selection and market timing – just by examining returns. It is actually quite easy to see why this is true.
Security selection and market timing affect returns differently
Firstly, think about a portfolio manager who engages solely in active security selection. Pure security selection risk is unrelated to the market risk, which means that portfolio returns are scattered fairly evenly around market returns. A 3% annual tracking error generated by security selection produces the daily returns shown below:
Source: A. Ekholm, 2015Now, consider a portfolio manager who engages in pure market timing. The variance between portfolio and market returns increases with market returns due to the excess market risk of the portfolio. The daily portfolio returns shown below also involve a 3% annual tracking error, but the difference with security selection is obvious to the eye:
Source: A. Ekholm, 2015Consequently, we can infer that the first portfolio manager has generated active risk by security selection and the second by market timing. This is exactly what my method does – and it only requires a simple regression analysis. Most importantly, it also works for the more realistic situation where active risk is generated by a combination of security selection and market timing:
Source: A. Ekholm, 2015Active risk and performance
My analysis of actively managed US equity mutual funds shows that performance has been positively related to past stock picking and negatively associated with past market timing. The positive effects of stock picking strengthen prior evidence based on portfolio holdings, while the negative effects of market timing are new.
I am currently working on another paper, confirming my earlier findings with new data. Furthermore, I have introduced an intuitively appealing break-up of portfolio risk: in my sample of actively managed US equity mutual funds, systematic risk factors accounted for 94%, security selection for 5% and market timing for only 1% of total return variance.
In summary, this method can be an efficient tool for monitoring active portfolio management and avoids the challenges associated with disclosed portfolio holdings. It even opens up the possibility to investigate portfolios that do not disclose their holdings at all such as hedge funds.
This is the second of two articles on different aspects of active portfolio management. Click here to access the first of these two articles.