One constant in every earnings season is the capacity for surprise. An early example in the current first-quarter reporting was Goldman Sachs – the bank’s share price dropped by 5% after it disclosed weak trading during the period.
When fund managers boast of the size of their research teams, this is the kind of news they may reasonably be expected to anticipate. With traditional fundamental research now easier than ever – balance sheets and valuation ratios are only a click away – buy-side analysts must seek information that is not so readily available, whether through conversations with corporate management or investigations into footfall at a business’s stores.
A new paper – by George Jiang at Washington State University, Russ Wermers at the University of Maryland and Ke Shen and Tong Yao at the University of Iowa – examines how successful fund managers are at endeavors of this nature.
As a proxy for how heavily managers rely on stock research, they constructed a measure of ‘information intensity’ in equities. This reflects the magnitude of company-specific news on the firm’s share price, isolated from wider market beta. Stocks that spike or crash on earnings updates or other corporate filings have a high information intensity, signifying that astute investor research into them should be rewarded.
Drawing on fund and stock data from the Center for Research in Security Prices and Thomson Reuters, with a sample of more than 3,000 active US domestic equity funds between 1980 and 2014, the academics first assessed funds’ exposure to high information-intensity companies. This rose over time from around 7-8% in the 1980s and 1990s to consistently above 10% after 2005, which indicated that by the end of the sample period, on average 10% of the return variances of stocks held by the funds were due to large information surprises.
Digging deeper into the characteristics of funds with a high information intensity, they also found that such managers had higher active shares – from 0.77 for portfolios in the lowest information-intensity quintile to 0.94 for those in the highest quintile – as would be expected. Higher information-intensity funds furthermore charged higher fees – on average 1.32% for the top quintile and 1.11% for the bottom – and tended to hold smaller and more illiquid stocks. Finally, funds with a high information intensity at any point in time tended to remain so over the following five years.
Low information investors
So did managers who concentrated on stocks driven by their news flow outperform? Superficially, it seemed not. Funds with a first-quintile information intensity did on average generate 0.08% more alpha per month, on a four-factor model net of fees, than those in the lowest quintile, but this was not statistically significant.
‘The insignificant relation between information intensity and subsequent fund performance, and the large performance dispersion among the top information-intensity funds, lead us to the conjecture that although information-intense stocks attract many active funds, not all such funds can successfully produce information,’ Jiang, Wermers, Shen and Yao summarized.
‘An analogy is the great American Gold Rush of the mid-1880s – many aspiring gold seekers went to California, but only a few made a fortune.’
They therefore turned to whether managers with a proven positive alpha record and high information-intensity portfolios tended to outperform. This emerged as a more reliable indicator of future performance.
Funds in the top quintile for both past-alpha generation and information intensity delivered a significantly positive future alpha of 0.198% per month, or almost 2.5% annually. Funds with similarly strong historical alpha but lowest-quintile information intensity went on to yield negative risk-adjusted returns.
‘These funds invest in information-intense stocks, and they are skillful in producing information on such stocks,’ concluded Jiang, Wermers, Shen and Yao of the former category. Of the latter, they observed that ‘although these funds have good past performance, their past performance is not the result of intense information production efforts, and thus smacks of random chance that does not last long.’