Using daily data from the German equity market, the authors provide evidence that suggests institutional investors exhibit herding behavior. Much of this behavior appears to be unintentional and could be attributed to common risk measurement and management practices—in particular, value at risk modeling—that are followed by most institutional investors and are often advocated or imposed by many regulators.
What’s Inside?
Using data drawn from the German stock market, the authors examine herding behavior in institutional investors. They emphasize that this behavior can be either intentional or unintentional. The authors examine the causes and consequences of such behavior, as well as its impact on market efficiency.
Herding is, according to the authors, “the tendency of investors to ‘bunch up’ on one side of the market.” Intentional herding occurs when traders deliberately follow the crowd, ignoring their own private information or analyses. The authors suggest that intentional herding tends to occur when there is uncertainty caused by lack of information about an asset, as tends to be the case with small-cap and illiquid stocks. Unintentional herding, by contrast, tends to occur with highly traded, large-cap stocks, which provide detailed public disclosure and information. In these cases, herding occurs because institutional investors rely on very similar risk models (in particular, value at risk models). Ironically, market instability, the very issue that regulatory authorities are trying to avoid with these risk management techniques, can be one of the consequences of implementing these techniques.
How Is This Research Useful to Practitioners?
The authors provide a useful review of the literature on herding behavior and build on this empirical literature to measure the presence of institutional herding. In particular, they use a measure proposed by Lakonishok, Shleifer, and Vishny (Journal of Financial Economics 1992) to test for the existence of institutional herding. But unlike most previous research, which tends to be based on low-frequency data and data that are not investor specific, the authors use daily, investor-specific data, which enable them to undertake a panel econometric analysis of issues associated with herding.
One of their main premises is that if investors’ herding behavior drives prices away from their fundamental values, then such effects should be mirrored in reversals of subsequent returns, as detailed by Choi and Sias (Journal of Financial Economics 2009). The results obtained from the research using panel data suggest that destabilizing effects do occur and are particularly prevalent with sell herds around large-cap and highly liquid stocks. The authors argue that this result is at least partly caused by homogenous risk management practices among institutional investors; a survey of risk management practices reports that all major investment banks use value at risk models based on historical data. Because of this finding, the authors suggest that regulators be aware of how their risk measurement and management systems can endogenize the risk of market instability.
They test the data on medium- and small-cap stocks for indications of herding behavior. Their results do not support the hypothesis that intentional herding occurs. Thus, the authors conclude that herding behavior is mainly unintentional.
They also analyze herding measures for the period before the financial crisis and find that they are typically higher than the measures drawn from the periods after the crisis (with the exception of sell herding during the crisis, which is marginally higher). They believe that this somewhat counterintuitive result can be explained by the construction of the herding measure, which accounts for trading in excess of the overall trend.
How Did the Authors Conduct This Research?
The authors conduct the analysis using data from July 2006 to March 2009, which allows them to test herding behavior under different conditions. They focus on institutional investors and include trades from 1,120 institutions that undertook proprietary trades. They split the analysis into trades involving stocks that are members of the DAX 30 (the German stock index of the 30 largest and most liquid stocks), the MDAX (a German mid-cap index of 50 stocks ranked lower than the DAX 30 in size and liquidity), and the SDAX (a German small-cap index of 50 stocks ranked lower than the MDAX in size and liquidity).
The analysis involves 1,044 trades of DAX 30 stocks, 742 trades of MDAX stocks, and 512 trades of SDAX stocks. On average, about 25 institutions trade each day, and institutional traders’ average daily market share of the DAX 30 is 46%. The authors are thus able to assess herding behavior across stocks with different levels of information and liquidity. More specifically, they are able to test hypotheses around intentional and unintentional herding.
The data are sourced from the German Federal Financial Supervisory Authority (BaFin), which identifies each trade by institution, end client, stock, time, price, and volume. The focus of the study is institutional trades undertaken for the institutions’ own accounts and excludes mandated liquidity providers, enabling the authors to identify institutional herding behavior as opposed to general market momentum.
Abstractor’s Viewpoint
The authors provide a reminder of the dangers of “groupthink”: If we all follow the same models, then we are likely to behave the same way and exhibit herding behavior, albeit unintentionally. They suggest that institutional behavior may have led to the endogenization of systemic risk. Perhaps this provides credence to the idea of gains stemming from contrarian trading—at least in the case of highly liquid stocks. But it does make me question whether we can truly say that a stock has digressed from its “correct” value. Basic economics suggests that price is determined by demand and supply. We could pose the argument that if we all believe the stock should be sold and a sell herd occurs, then this result is simply a manifestation of demand and supply. Furthermore, at what point can we say with certainty that a stock is under- or overvalued unless we are all using the same valuation models in the first place?