信息共享还是利益冲突?——基于买方单独调研与买卖双方联合调研的实证检验

基于买方(Buy Side)与卖方(Sell Side)存在的利益关系(交易佣金),本文考察了买方(基金)调研对于卖方证券分析行业信息精度的影响。结果发现,基金调研行为对卖方机构预测准确度的影响机制并不单一。一方面通过“信息共享”效应提高了券商的平均预测准确性;另一方面,两者之间存在明显的“利益冲突”效应,该作用降低了卖方分析师的预测公允性。此外,本文更具体地研究了两种效应的内在传递机制。从“信息共享”的角度发现基金调研吸引更多券商分析师对上市公司的关注,从而为证券分析行业带来“信息增量”;从“利益冲突”的角度,实证结果表明在中国的二级市场中佣金关系(交易佣金席位)更能代表中国市场买卖双方之间的利益关系,这种利益关系带来的“利益冲突”降低信息有效性。进一步,本文通过TF-IDF方法挖掘调研文本信息,分析了牛熊市、投资者情绪和调研报告文本情感色彩对调研信息质量的影响。

Information Sharing or Conflict of Interest: Empirical Study of Buy-Side Individual and Joint Visits

Information plays a key role in the efficiency of economy and finance. In the asset management industry, investment methods based on intrinsic value have become the mainstream approach of active institutional investors.Therefore, investors are increasingly focusing on uncovering information about listed companies.The Shenzhen Stock Exchange now forces listed companies to report on investor relationship management activities, such as communication behaviors. This further improves the information efficiency of the A-share market.
In China's securities market, refinements to the information disclosure system and advances in commercial institutions provide new empirical data for studying the relationship between information acquisition and asset prices.Based on this background, this paper uses investor relationship information from the China Stock Market and Accounting Research (CSMAR) database between 2012 and 2016 to study the impact of buy-side visit behavior on sell-side analysts' predictions from the perspectives of information transmission and conflict of interest.
Different from previous studies, the paper pays more attention to the information mechanism for the buy-side and the sell-side in visit activities rather than a single market subject. We can further refine the influence of buy-side visits by distinguishing between different types of visits,specifically individual visits and joint visits.We put forward two hypothesized effects: the information sharing effect and the conflict of interest effect. The empirical results show that buy-side individual visits can reduce the forecast bias of the brokerage industry, while joint visits cause a decline in average forecast accuracy; this confirms the existence of the two effects.
In addition, we further study the internal transmission mechanisms of the two effects. For the information sharing effect, we find that the number of brokers tracking the enterprise after the fund visit increases significantly, indicating that the disclosure of such behavior and text records can cause more brokers to pay attention to listed companies. This provides information channels for brokerage analysis and forecasting. Regarding conflicts of interest, we find that when brokers provide site visit arrangements, research reports, and other services for funds to obtain trading commissions, the resulting benefits-based relationship has an impact on the accuracy of the information predicted by the sell side.
We also consider three external factors: market condition (bull or bear), investor sentiment, and emotion in visit records. In particular, we introduce the TF-IDF method to statistically analyze the representative vocabulary in the text. The empirical results are as follows: (1) Information efficiency is higher during bull markets,when investor sentiment is high; in a bear market, when investor sentiment is depressed, institutional investors are more dependent on brokers,leading to a more prominent conflict of interest effect. (2) When the survey minutes of listed companies have obvious emotional signals, fund visits can reduce the average error of brokers' predictions. Further, when the text shows a negative signal, the information formed by the fund visit is more obvious, and it becomes harder for the interest relationship to intervene in securities analysts' forecasts.
The paper makes several important contributions. First, it splits the effect of buy-side (fund) visits on sell-side analyst predictions into two dimensions.Second, we open the “black box” of the internal information mechanism of corporate visits. It is particularly difficult to quantify interest relationships, as the institutions in a visit sample may involve more than one fund and broker. We first expand the data into a one-to-one relationship and then combine the calculation. Third, the paper extends the analytical approach to information in the text record by using the TF-IDF method.Finally, the CICSI index is more in line with China's national conditions than the BW index and therefore more suitable for studying the effect of investor sentiments.
Overall, the paper puts forward a unique perspective to analyze the role of institutional investors in market information transmission, thus providing new evidence for research on information efficiency in the capital market. This paper also provides clear policy implications related to information disclosure.