Consequences of firm-specific stock price crashes on analyst forecasts: Evidence from China

The paper investigates how firm-specific stock price crashes (SPCs) affect sell-side analysts’ forecast accuracy. Prior research has extensively studied determinants of SPCs but rarely their consequences; Kim et al. (2022) is the only prior event-focused study, showing SPCs can trigger investor attention. The authors posit two competing hypotheses: analyst attention theory predicts SPCs draw limited analyst attention to the affected firm, improving the information environment and forecast accuracy; analyst–management collusion theory predicts executives facing negative governance signals from SPCs may lobby analysts for optimistic forecasts, reducing accuracy. The study empirically evaluates which effect dominates in the context of Chinese listed firms, where analysts play a key information intermediary role especially under uncertainty.

Consequences of SPCs literature splits into: (1) studies on crash risk (crash-proneness) using lagged risk measures (negative skewness, down-to-up volatility) to examine one-year-lag consequences; and (2) event-focused studies identifying SPC events and using event methods (DID) to assess short-run effects (Kim et al., 2022). Analyst forecast accuracy depends on the firm’s information environment (e.g., disclosure transparency, audit quality, report readability) and on information acquisition channels (media, transportation improvements, site visits). Analysts also exhibit behavioral biases (management catering, culture effects, herding). The paper formulates two competing hypotheses: H1A (SPCs stimulate analyst forecast accuracy via attention and an improved information environment) and H1B (SPCs deteriorate accuracy via collusion incentives). It further develops cross-sectional hypotheses that SPC-induced accuracy improvements are stronger for analysts with information disadvantages—those without pre-SPC site visits (Hypothesis 2) and those lacking geographical proximity (Hypothesis 3).

Data come from CSMAR for analysts and firm financials, covering 2001–2020, excluding financial firms. SPC identification follows Kim et al. (2011a, 2011b, 2022): compute firm-specific weekly returns from market-model residuals, transform to w_it = ln(1 + r_it), and flag an SPC when w_it falls at least 3.20 standard deviations below the firm’s mean over a 12-month estimation period. The authors identify 2,781 SPC events, then use propensity score matching (nearest neighbor within industry and year) to match each SPC firm to a non-SPC firm using lagged predictors: detrended turnover (DTURN), negative skewness (NCSKEW), mean and volatility of firm-specific returns (RETLAG, SIGMA), sales growth (SALESG), firm age (AGE), asset tangibility (TANG), and size (SIZE). The matched sample includes 2,127 SPC firms and 2,127 controls. The DID window spans t = −6 to +6 months around the SPC month (t = 0). For control firms, the same timing as their matched treated firm is used. Final panel comprises 56,837 firm–month–analyst observations. Primary DID specification: FERROR_ijt = α0 + α1 CRASH_jt + α2 POST_jt + α3 (CRASH_jt × POST_jt) + Controls + ε_jt, where CRASH = 1 for treated firms, POST = 1 for post-SPC months (0 to +6), and FERROR is analyst forecast error. Analyst forecast error is defined as (FEPS_ijt − AEPS_ijt) / AEPS_ijt; higher values imply lower accuracy. Controls include analyst-level (experience, forecast horizon, firm-specific years, number of firms followed, team indicator, brokerage size), firm-level (size, ROA, leverage, board size, independent director ratio, largest shareholder stake, institutional ownership, SOE indicator), and audit-level (audit fee, audit committee, Big 4 auditor) covariates. Year and matched-pair fixed effects are included; standard errors clustered at the firm level. Robustness and identification checks: (1) Parallel trends test with event-time coefficients; (2) Alternative accuracy measure scaled by stock price, FERROR_P = |FEPS − AEPS| / |P|; (3) Instrumental variables (2SLS) with mutual fund hypothetical sales (MFHS) as an instrument for exogenous crash pressure (first-stage probit for CRASH, then fitted values interacted with POST); (4) Placebo tests using fictitious crash dates six months before and after the true SPC month. Validation and mechanism analyses: Cross-sectional DID by information access proxies—pre-SPC site visits (visit vs non-visit), geographical proximity (local same-province vs non-local), and analyst reputation (star vs non-star, based on New Fortune rankings). Mechanism (effort) measured by PRIVACY, the amount of private information in analyst forecasts following Altschuler et al. (2015) and Park et al. (2017), with higher PRIVACY implying greater private-information effort.

- Main DID: CRASH × POST coefficient is significantly negative (coef = −0.186, t = −4.471), indicating that analyst forecast errors decline (accuracy improves) more for SPC firms than matched non-SPC firms in the post-event window. Controls behave as expected in several cases: ROA and institutional ownership associate with lower errors; longer forecast horizon associates with higher errors; larger brokerage size and team involvement associate with lower errors. - Parallel trends: No pre-trends in event time; negative and significant effects begin in the SPC month and post-period. - Alternative accuracy measure: Using FERROR_P, the interaction term (reported as CRASH+POST) is significantly negative (coef = −0.002, t = −2.433), confirming improved accuracy post-SPC. - Instrumental variables: First stage shows MFHS positively predicts crashes (coef ≈ 0.002, z ≈ 1.976). Second stage shows fitted CRASH × POST is negative and significant (coef = −0.874, t = −1.967), supporting a causal interpretation that exogenous crash pressure increases analyst accuracy. - Placebo tests: No significant CRASH × POST effects in windows using fictitious SPC dates (−12 to 0; 0 to +12), mitigating concerns of confounding contemporaneous shocks. - Validation analyses (heterogeneity): The accuracy improvement is concentrated among analysts with information disadvantages: non-visit group (CRASH × POST significantly negative), and non-local group (significant), while visit and local groups do not show significant improvements. By reputation, the effect is significant only for non-star analysts; star analysts’ accuracy is not significantly affected, consistent with higher baseline effort and information acquisition. - Mechanism (effort): Post-SPC increases in PRIVACY (proxy for private information effort) are significant for analysts with information disadvantages (non-visit, non-local) and for non-star analysts, indicating SPCs spur greater effort among these groups, mediating the accuracy improvement. - Sample scope: 2,127 treated and 2,127 control firms; 56,837 firm–month–analyst observations over 2001–2020; 2,781 SPC events identified before matching.

The evidence supports the analyst attention view over the collusion view: SPCs, as rare salient events, draw limited analyst attention toward affected firms, improving the information environment and thus forecast accuracy. Improvements are concentrated among analysts lacking pre-existing information advantages (no site visits, geographically distant) and those with lower reputational capital (non-stars), consistent with attention reallocation and incentive to protect reputation following investor wealth losses. Mechanism tests suggest that enhanced private-information effort mediates the effect for disadvantaged analysts. These findings clarify how analysts respond to heightened uncertainty and adverse shocks, demonstrating positive externalities of SPCs through market information intermediation.

This study provides the first comprehensive evidence on the consequences of firm-specific stock price crash events for analyst behavior. Using a DID framework with matched controls in China (2001–2020), we show that analyst forecast errors decrease after SPCs, with robust results across parallel trend checks, alternative accuracy measures, IV estimation, and placebo tests. Cross-sectional and mechanism analyses indicate the effect is driven by analysts with information disadvantages and lower reputation, via increased private-information effort. The findings highlight positive externalities of SPCs through improved analyst attention and accuracy, and offer insights into how reputation and information access shape analyst responses to shocks. Future research should examine long-run dynamics of forecast accuracy post-SPCs and assess external validity in cross-country settings with different market maturities and analyst roles.

The analysis focuses on short-run effects within a one-year window around SPCs and does not assess long-term persistence or reversal of accuracy changes. The setting is China’s relatively less mature market; generalizability to developed markets may differ due to institutional and analyst role differences. Data constraints limit public availability and may affect replication.