Neglected Stocks (被忽视股票)
Stocks with thin public information environments — low analyst coverage, low institutional ownership, low trading volume, recent losses, small market cap. Across multiple empirical settings, alpha concentrates in neglected names: their public-information channels are weak, so private costly-information channels (Investor Corporate Site Visits, paid research, expert networks) carry disproportionate weight.
Definition
There is no single canonical definition. Common proxies:
| Proxy | Typical threshold | Source |
|---|---|---|
| Low analyst coverage | ≤ 2 analysts covering | Lee & So (2017); So, Wang & Zhang (2024) Table 5 |
| Low institutional ownership | Below cross-sectional median | So, Wang & Zhang (2024) Table 5 |
| Recent losses | Negative net income in last fiscal year | So, Wang & Zhang (2024) Table 5 |
| Small market cap | Bottom tercile by circulation market cap | So, Wang & Zhang (2024) Table 5 |
| Low trading volume | Below-median share turnover | So, Wang & Zhang (2024) Table 6; many microstructure papers |
| No subsequent analyst revision | No FY1 EPS revision in past month | So, Wang & Zhang (2024) Table 5 |
| Low abnormal trading volume | Below-median residual turnover | So, Wang & Zhang (2024) Table 6 |
The proxies are correlated but not identical; a multi-factor "neglect score" is the most defensible definition for empirical work.
Why neglect predicts alpha
Neglect is a state variable for the information environment. When public information is thin, the marginal value of additional (private) information is high, so:
- Informed agents have an information advantage that takes longer to be impounded into price.
- Post-earnings-announcement drift is stronger (see Post-Earnings Announcement Drift).
- Cross-sectional momentum and post-event reversals last longer.
- Costly-acquisition signals (Abnormal Investor Visits (AIV)) carry more weight.
In So Wang Zhang 2024 Investor Corporate Visits Table 5 Panel B, AIV's return-predictive coefficient has significantly positive interactions with all four neglect proxies (low coverage, low institutional ownership, recent losses, small size). Table 6 shows the AIV hedge return nearly doubles when the long leg is restricted to high-AIV / low-abnormal-turnover and the short leg to low-AIV / high-abnormal-turnover: ~1.49% / month DGTW α vs. ~0.66% unconditional.
The classical line of work
- Arbel & Strebel (1982, 1983) — original "neglected-firm effect"; small, under-analysed stocks earn excess returns.
- Arbel, Carvell & Strebel (1983) — "glamour vs. value" — glamour (high analyst attention) underperforms value (neglected).
- Hong, Lim & Stein (2000) Bad news travels slowly, JF — low-analyst-coverage firms have slower price adjustment to bad news; provides the formal mechanism for why neglect generates return predictability.
- Hong & Stein (2007) — disagreement-as-a-channel refinement.
- Lee & So (2017) Uncovering expected returns, JFE — abnormal analyst coverage (A_{TOT}) is itself a positive cross-sectional alpha proxy, even after controlling for the standard set; the buy-side analogue of A_{TOT} is Abnormal Investor Visits (AIV).
- Crane, Crotty & Umar (2023) — hedge-fund information acquisition proxies interact with neglect.
A-share specifics
On the SZSE / A-share universe:
- The So Wang Zhang 2024 Investor Corporate Visits sample has 108,874 firm-months over 2012-07 to 2019-12 — a large cross-section in which ~50% of firms host a visit in any year, but neglect is still very much present (median 4 institutions per visit; ~77% of mutual-fund visits are non-holder).
- Combine low coverage with low trading volume and the AIV hedge alpha nearly doubles vs. unconditional.
- "Small cap" on A-share needs care: there are large state-owned enterprises that are small by market cap but politically connected and well-covered; pure size neglect is a noisier proxy than on US data.
Implementation notes
- Use multiple neglect proxies jointly — don't rely on any single one.
- For A-share data, combine CSMAR (analyst coverage, institutional ownership) with Wind or Tushare (turnover, market cap).
- Be careful with look-ahead bias when constructing the interaction terms: use only information available at month-end m when forecasting m+1 returns.
- Neglect proxies themselves decay slowly — they are persistent characteristics, not signals — so a quarterly re-estimation is usually sufficient.
被指数纳入 = 强制脱离冷门（2026-10-20 补记）
Greenwood & Sammon 2022 提供了"脱离冷门"这条通道的反向证据，但结论与早期文献相反：
| 时期 | 标普 500 纳入的平均总收益 |
|---|---|
| 1980 年代 | 3.4% |
| 1990 年代 | 7.6% |
| 2010—2020 | 0.8%（与零无统计差异） |
早年的指数纳入确实是一次"冷门 → 热门"的强制转换——被动基金买入把一只没人研究的股票推入研究覆盖，机构调研随之跟上。但这个溢价在 2010 年代消失了，尽管被动化程度和机械需求冲击都比当年更大。
论文给出的主因是"迁移"：近年超过 80% 的纳入是从标普中盘 400 升入标普 500，强制买入与强制卖出同时发生，净需求冲击被对冲。这意味着——"进入大盘指数"本身已经不再是"脱离冷门"了，因为这些股票本来就有一个不算冷的身份。
对 neglect 类因子的直接含义：如果用"是否会被纳入重要指数"来构造 neglect 的反向信号，在梯度化的指数体系下这个信号会大幅失效。A 股的沪深 300 / 中证 500 / 中证 1000 同样是梯度结构，迁移抵消机制可能同样存在——这是本 wiki 记录的一个待检验假设。
Limitations
- A-share "neglect" interacts with policy / state ownership: some firms are de facto under-covered because they are politically connected and the policy-relevant news channel is different from the equity-research channel.
- "Neglect" is endogenous: when an alpha paper on neglected stocks is published, more analysts cover them. Like all attention-based factors, the signal decays. 索引纳入是这条衰减曲线上最陡的一段——参见 Greenwood-Sammon 2022，纳入溢价从 1990 年代的 7.6% 掉到 2010 年代的 0.8%。
- The neglect premium is partly an illiquidity premium: small-cap, low-turnover stocks have higher transaction costs that may offset the gross alpha. Sharpe ratios after costs are sometimes unimpressive.
See also
- Abnormal Investor Visits (AIV) — the cross-sectional signal that exploits the neglect / private-information gap.
- Investor Corporate Site Visits — costly private channel.
- Post-Earnings Announcement Drift — release channel through which neglect is impounded.
- Investor Attention — broader family.
- So Wang Zhang 2024 Investor Corporate Visits — primary reference.
- Greenwood & Sammon 2022 — 指数纳入效应为何消失