نقدینگی بازار و تجارت سازمانی در طول بحران مالی 2007-8

This paper shows that institutional sell-side herding increased bid–ask spreads and liquidity risk during the 2007–8 financial crisis. Such an impact on liquidity is most pronounced in firms with large numbers of institutions that sold the same stocks, that is, have correlated trades. For the same reason, we find institutional investors with a dedicated, buy-and-hold, investment style to be the least likely to herd; their trading activity did not affect stock market liquidity during the crisis. Our results are robust to alternative explanations, different test specifications and consistent with recent theories highlighting the negative impact of institutional trading activity on market liquidity during a crisis.

Recent theories highlight the role of institutional trading in reducing market liquidity during a financial crisis (see Brunnermeier and Pedersen, 2009, Garleanu and Pedersen, 2007, Huang and Wang, 2009, Kyle and Xiong, 2001 and Xiong, 2001). They predict that institutions' sell-orders during negative extreme market shocks exacerbate order imbalances. In this paper, we corroborate these theories by presenting empirical results which suggest that institutional sell-side herding is a key factor in impairing market liquidity and creating liquidity risk. Our paper makes two important contributions: First, we show that the number of institutional shareholders is a (noisy but readily available) proxy for the level of institutional trading and the potential institutions' sell-side herding during a market downturn (Chiang and Zheng, 2010, Nofsinger and Sias, 1999, Sias, 2004 and Zhou and Lai, 2009). Second, we illustrate the significant effect of institutional sell-side herding and correlated trades on both trading costs and liquidity risk during the 2007–8 financial crisis. Shares of firms with a greater number of institutional shareholders are more likely to suffer from institutional herding when they exit at the same time, creating excess order imbalances and wider spreads. We do not equate herding to irrational behavior and do not make any attempt to discern which of the possible herding reasons drives our results. Recent theoretical studies suggest that increased selling activity during a crisis could be driven by several factors, such as increased funding constraints (Brunnermeier & Pedersen, 2009), higher risk aversion (Huang & Wang, 2009), and/or tighter risk management (Garleanu & Pedersen, 2007). All or some of the herding explanations could contribute to the herding results presented here. Companies with smaller numbers of institutional shareholders but large institutional holdings do not suffer as much, possibly because these institutions have better information about the firms, and are more likely to offload their shares in a more orderly manner to minimize the price impact of their transactions. To further understand institutional characteristics and their trading patterns, we classify institutional investors into groups according to their ‘type’ and ‘investment style’. Different investor types and styles have varying degrees of correlated trades both within and across their groups (Sias, 2004). Thus, given that a higher degree of correlated trades is associated with a greater impact on order imbalances, and hence illiquidity, we expect to find systematic differences in the impact of the various groups' trading activity on market liquidity. Based on the SEC 13f filings, institutional investors are classified into bank trusts (BNK), insurance companies (INS), independent advisors (IA), public pension funds (PPS) and university and foundation endowments (UFE).1 Furthermore, we use the database constructed by Bushee (2001) to group institutional investors into quasi-indexers (QIX), transient (TRA), and dedicated (DED) based on their long-horizon trading pattern.2 In general, IA dominates our sample firms over the 2004–8 period, and over two-third of the institutional investor sector is quasi-indexers. We do not detect any significant correlation between ‘type’ and ‘style’; all three ‘styles’ appear in a similar sector representation within each ‘type’. UFE trades have high within-group standard deviation, and are least correlated with the other institutional trading patterns, which diverge even further in the crisis period. The same applies to DED trades, which during the crisis became less correlated with QIX and TRA. This lower correlation in the trading pattern of UFE and DED with all the other groups leads to an insignificant impact of their herding activity on market liquidity. In contrast, the herding activity of all other institutional types and styles affects market liquidity. Separately, Koch, Ruenzi, and Starks (2010) find a strong commonality among the Amihud price impact measure of stocks owned by mutual funds which have high turnover and liquidity shocks due to their investors' withdrawals. Similarly, we find that the impact of institutional trading on liquidity is the strongest among the Independent Advisor (IA) group, which includes mutual funds. Among the five institutional investor types, IA is the most likely to be subject to institutional constraints and the least able to stick to their trading strategies during a financial crisis, thus we expect them to have correlated trades. But unlike Koch et al. (2010), we find that our results are statistically significant only for institutional sell-side trades during the crisis. The very high correlated buy-side herding before the crisis did not have any impact on liquidity or liquidity systematic risk. We recognize that there might be spurious relations and omitted variables that might explain our results. Therefore, we run a battery of tests to try to alleviate such concerns. We include in our model specifications several market and accounting-based control variables, which the literature has identified as important determinants of trading cost and liquidity risk. In addition, we use industry and firm fixed effect specifications to account for omitted variables. In all our models, we cluster the standard errors at the firm level to control for time-series dependence, and add year dummies to capture the effect of cross-sectional dependence. We run regression models based on the changes in the variables instead of their levels. We also run separate tests to control for alternative explanations based on institutional blockholders, or the lack of retail investors. Our main findings remain the same. The remainder of the paper is organized as follows. In Section 2, we position our paper alongside the previous literature and present our hypotheses. In Section 3, we describe our data sources, variable definitions and provide some descriptive statistics. In Section 4, we report our results. Section 5 provides some discussion and concluding remarks.

This paper investigates the role of institutional investor trading in escalating market illiquidity during the recent fi nancial crisis. We measure the intensity, scale and fl ow of institutional trading by the number of institutional investors holding shares, their proportional ownership and the ratio of institutions' buy trades to all trades. We study two aspects of market liquidity: trading cost as measured by the quoted proportional spread and commonality in liquidity as mea- sured by R 2 and the beta of the cross-sectional spread variations. Our results suggest that institutions' sell-side herding together with their correlated trades contributed to the increased trading costs and li- quidity risk during 2007 – 8. Our fi ndings are robust to various different model speci fi cations and measures. They provide empirical support for theoretical predic- tions on the role of trading in exacerbating liquidity shortages during a crisis. In particular, prior theoretical literature predicts the existence of increased selling activity by institutional investors during a crisis. The rationale is that, during a market downturn, institutional investors face high selling needs due to increased funding constraints ( Brunnermeier & Pedersen, 2009 ), high risk aversion ( Huang & Wang, 2009 ), tight risk management ( Garleanu & Pedersen, 2007 ), or of fl oading of assets by convergence traders ( Kyle & Xiong, 2001 ). A common prediction is that these pressures lead to excess order imbal- ances and inventory costs, hence greater illiquidity. In this paper, we provide evidence that fi rms with a larger number of institutional investors are more vulnerable to sell-side herding. We report that institutional investors did indeed increase their selling activity substantially during the fi nancial crisis. More importantly, we show that this sell-side trading activity led to an increase in fi rms' trading costs and liquidity risk during the crisis. We do not investigate which of the above theoretical predictions drives our results. During a crisis, all of them might play a signi fi cant role in raising the selling needs of institutional investors. Our results collectively indicate a signi fi cant role for institutional investors' sell-side herding in increasing a fi rm's trading costs and liquidity risk. They con fi rm the speculation by Chordia et al. (2000) that commonality in liquidity could be due to institutional herding (p. 14), and provide direct evidence to support the conjecture of Koch et al. (2010) that correlated trading increases commonality in liquidity. Our study extends Koch et al. by showing the impact of in- stitutional herding on a fi rm's quoted spread, as well as demonstrat- ing the prevalence of these phenomena across different institutional investor types during a crisis. Moreover, our results indicate that only few institutional investors, mainly these with a dedicated invest- ment style, did not contribute to the increase in liquidity risk during the 2007 – 8 fi nancial crisis. Controversial as it may sound, our fi ndings also suggest that restricting the number of institutional shareholders holding a fi rm's equity might help to reduce the fi rm's liquidity risk during a fi nancial crisisargues that, during a crisis, market participants fi rst sell the most liquid stocks in their portfolios. Similarly, Anand et al. (2010) report that institutional investors tilt their sell- ing activity towards less liquidity-sensitive stocks during a crisis. These fi ndings imply that our results for II_count and institutional investors' sell-side herding cannot be driven by reverse causality. Still, we accept that the endogeneity issues cannot be completely resolved. For example, one may argue that institutional investors sold stocks in anticipation of a liquidity drain out. Until we can successfully fi nd an instrument to control for such explanation, the endogeneity problem cannot be eliminated. 25 But, our results indicate a strong connection between sell-side herding, II_count and liquidity risk even if the direction of causality remains as a conjecture. It is the fi rst empirical evidence that is in line with the theoretical predictions of Xiong (2001) , Brunnermeier and Pedersen (2009) , Garleanu and Pedersen (2007) and Huang and Wang (2009) . In 2007, institutional investors held 80% of the S&P 1500 equity and accounted for the majority of the trading volume in the markets. Given the size of institutional investors' participation in the economy, institutional investors' trading behavior has a large impact on liquid- ity and trading costs and can easily induce market-wide systemic liquidity risk. So this paper not only provides important policy impli- cations but also some empirical support to theoretical models of extreme liquidity (e.g. Huang & Wang, 2009; Kyle & Xiong, 2001 ), in particular regarding the role of (institutional investors') trading activity in exacerbating the liquidity drain-out during the recent fi nancial crisis.