Stock price crashes (SPCs) are significant market events, but research primarily focuses on their causes rather than their consequences. This study addresses this gap by examining how firm-specific SPCs affect analyst forecast accuracy. Analysts act as information intermediaries, and their forecasts are crucial, especially under uncertainty. Two competing hypotheses are proposed: the analyst attention theory, suggesting that SPCs, being rare events, attract analyst attention, improving the information environment and thus forecast accuracy; and the collusion view, positing that SPCs, signaling poor corporate governance, may lead to analysts colluding with management for optimistic forecasts, decreasing accuracy. The study uses a large sample of Chinese listed companies (2001-2020) and a difference-in-differences (DID) analysis to test these hypotheses. The choice between pandering to management or investors is key in determining the dominant effect.

Existing literature extensively discusses the determinants of firm-specific SPCs but lacks research on their consequences. Studies on SPC consequences are categorized into those focusing on SPC risk (measured by negative skewness, volatility, etc.) and those focusing on SPC events. The former uses lagged SPC risk measures in regression models to examine long-term consequences, while the latter, exemplified by Kim et al. (2022), employs DID analysis to investigate immediate consequences of SPC events. This study follows the latter approach, exploring the impact on analyst forecast accuracy. The literature also shows the influence of company information environments (disclosure transparency, audit quality, etc.) and analyst behaviors (biases, herding) on forecast accuracy. This study uniquely investigates the effect of SPCs on analyst behavior under uncertainty, particularly focusing on the interplay between analyst attention and collusion.

The study uses data from the CSMAR database on listed companies and financial analysts in China (2001-2020). SPCs are identified using the method from Kim et al. (2011a, 2011b, 2022), where a weekly return falling more than 3.20 standard deviations below the 12-month mean constitutes an SPC. Propensity score matching (PSM) is used to match SPC firms with similar non-SPC firms in the same industry and period. A DID analysis is employed, setting the SPC month as the baseline (t=0) and examining analyst forecast accuracy changes within a one-year window (t=-6 to t=+6). Analyst forecast accuracy is measured as the deviation of the analyst's earnings per share (EPS) forecast from the actual EPS, scaled by the actual EPS. The primary regression model includes CRASH (treatment group dummy), POST (post-SPC period dummy), their interaction (CRASH*POST), and several control variables at analyst, firm, and audit levels. Robustness checks include parallel trend tests, alternative accuracy measures (using opening stock price as a scaling variable), instrumental variable (IV) analysis using mutual fund flow redemption pressure as an instrument, and placebo tests.

The DID analysis reveals a significantly negative coefficient for CRASH*POST in the main regression, indicating that analyst forecast accuracy significantly increases in the post-SPC period for treatment (SPC) firms compared to control firms. This supports the analyst attention hypothesis. Control variables show that high ROA and high institutional ownership are associated with lower forecast errors. Robustness tests confirm this finding: the parallel trend test shows no significant difference in forecast accuracy before the SPC; using an alternative accuracy measure (scaling by opening stock price) maintains the significant negative coefficient for CRASH*POST; the IV approach using mutual fund flow redemption pressure confirms the causality; and placebo tests rule out the influence of contemporaneous unobserved factors. Further analyses demonstrate that the increase in accuracy after SPCs is more pronounced for analysts who did not conduct on-site visits before the crash, lacked geographical proximity to the firm, and were not "star" analysts (based on New Fortune magazine rankings). This emphasizes the role of information disadvantage in driving the attention effect. Finally, channel analysis, using a measure of private information in analyst forecasts as a proxy for analyst effort, shows that analysts with information disadvantages significantly increase their effort after SPCs, contributing to the improved accuracy.

The findings support the analyst attention theory, showing that SPCs can positively affect market efficiency by stimulating analysts' information gathering and improving forecast accuracy. This contrasts with the collusion view, suggesting that analysts prioritize investors over management in the face of SPCs. The study significantly contributes to understanding analyst behavior under uncertainty, highlighting the role of reputation and information acquisition in shaping their response to extreme market events. The results are particularly insightful for analysts with information disadvantages, showcasing their increased efforts to mitigate reputational damage after SPCs.

This study is the first to examine the consequences of SPCs on analyst behavior, providing evidence of positive externalities from the analyst perspective. It enhances our understanding of analyst behavior under uncertainty, showing how reputation concerns drive increased effort and improved accuracy after SPCs. The study also contributes to the literature on the determinants of analyst forecast accuracy by highlighting the role of analyst attention. Future research could explore the long-term impact of SPCs on analyst forecast accuracy and conduct cross-country comparisons to investigate the generalizability of these findings.

The study focuses on the Chinese stock market, limiting the generalizability of the findings to other markets with different regulatory environments and analyst roles. The one-year window for examining post-SPC effects might not capture long-term consequences. The reliance on self-reported data for analyst site visits could introduce measurement error. The use of a specific measure for private information and analyst effort could limit the scope of the study.