Costly Information Acquisition (昂贵信息获取)
The principle that an information signal is more credible — and more profitable — when producing it requires the agent to bear non-trivial costs. In equilibrium, costly signals self-select for the cases where the agent expects the largest payoff. The principle underwrites the empirical power of Investor Corporate Site Visits / Abnormal Investor Visits (AIV) and, more broadly, the value of "expensive attention" relative to "free attention".
The economic argument
If producing a signal costs c and yields expected payoff \pi, the agent will only acquire the signal when \pi > c. Therefore:
- Higher observed c ⇒ higher expected \pi conditional on acquisition.
- A signal whose acquisition is costly has more bite than one whose acquisition is free.
This is the costly-vetting story behind Abnormal Investor Visits (AIV): institutions voluntarily fly to a firm's HQ; in equilibrium they do this only when they think the firm is meaningfully underpriced. The cost is what gives the visit its information content.
Empirical signatures
The costly-acquisition prediction is testable by conditioning on the cost. In So Wang Zhang 2024 Investor Corporate Visits Table 5 Panel A, AIV's return-predictive power is stronger when:
- The visiting institution is a non-holder (cost = build the contact from scratch).
- The visit is an initial one (no visit in past 6 months) — first-meeting cost > follow-up cost.
- The visit requires higher travel time (Baidu Map API) — direct travel cost.
A second empirical signature is the COVID natural experiment. When Chinese travel restrictions forced visits online (post-Feb 2020), the abnormal-visit signal lost all return-predictive power and mutual funds that had been heavy site visitors suffered disproportionate alpha losses (Table 9 Panel B). The cost was exogenously removed → the information advantage disappeared.
Where the same idea shows up
- Hedge-fund information acquisition (Crane, Crotty & Umar 2023, Management Science): hedge funds that actively acquire information outperform non-acquirers; the cost of acquisition is what selects for skill.
- Traveling analysts (Ben-Rephael, Carlin, Da & Israelsen 2022): analysts whose forecasts depend on in-person information suffer disproportionately during COVID lockdowns.
- Sell-side analyst site visits (Cheng et al. 2016, 2019; Han, Kong & Liu 2018): improved forecast accuracy among visiting analysts; event-window market reaction. The buy-side analogue is the Investor Corporate Site Visits stream.
- Conference attendance and selective access (Bushee, Jung & Miller 2011, 2017; Green, Jame, Markov & Subasi 2014a,b; Kirk & Markov 2016): selective access to management is valuable precisely because it is selective.
- Private management interactions (Soltes 2014; Solomon & Soltes 2015): single-firm US evidence that investors benefit from private interaction.
Contrast with "free attention"
| Attention signal | Acquisition cost | Selection mechanism | Predictiveness |
|---|---|---|---|
| Bloomberg / EDGAR download | ~0 | Everyone downloads | Weak / diffuse |
| Google search volume | ~0 | Everyone searches | Inconsistent; retail-attention reverses |
| Web traffic to IR page | ~0 | Anyone with a browser | Noisy |
| On-site investor visit | High (travel + time) | Institutions vet underpricing | Strong and persistent |
| Paid expert network call | High ($$) | Paying experts to dig in | Strong but proprietary |
| Short seller due-diligence report | High (reputational) | Reputational skin in the game | Strong; concentrated in overpricing |
The pattern: as acquisition cost rises, signal specificity rises and the agent's selection effect (underpricing or overpricing) becomes sharper. The corollary for alpha research is that observed attention measures are a hierarchy — Bloomberg download counts are nearly free, on-site visits are expensive — and the expensive end of the hierarchy is where asymmetric alpha lives.
Connection to information asymmetry and neglect
The costly-acquisition story has a natural interaction with the information environment of the target firm. So Wang Zhang 2024 Investor Corporate Visits Table 5 Panel B shows that AIV's predictive power is stronger in firms with poor public information environments — low analyst coverage, low institutional ownership, recent losses, small size. The mechanism: when public information is plentiful, the marginal value of an additional (costly) private channel is low; when public information is scarce, the private channel is worth a lot, and the equilibrium cost is justified.
What this implies for AIV replication
- Cost proxy: track travel time, non-holder status, initial-vs-follow-up. Higher-cost sub-samples carry higher alpha.
- Channel check: if the visit channel is disrupted (e.g. COVID lockdown, regulatory change), expect alpha to vanish even if visit counts remain.
- Neglect interaction: condition on low analyst coverage / low institutional ownership / low turnover. Alpha is concentrated there.
- Crowding risk: as more institutions read the paper and crowd the channel, the cost of being early rises but the cost of being there does not change much. Expect in-sample alpha to compress; expect OOS alpha to remain where neglect is highest.
See also
- Investor Corporate Site Visits — concrete implementation of costly acquisition.
- Abnormal Investor Visits (AIV) — the cross-sectional signal.
- Neglected Stocks — where costly acquisition has the most bite.
- Investor Attention — broader family.
- So Wang Zhang 2024 Investor Corporate Visits — primary empirical reference.