Working Paper: NBER ID: w12203
Authors: Andrew Ang; Li Gu; Yael V. Hochberg
Abstract: Recent studies suggest that the underperformance of IPOs in the post-1970 sample may be a small sample effect or "Peso" problem. That is, IPO underperformance may result from observing too few star performers ex-post than were expected ex-ante. We develop a model of IPO performance that captures this intuition by allowing returns to be drawn from mixtures of outstanding, benchmark, or poor performing states. We estimate the model under the null of no ex-ante average IPO underperformance and construct small sample distributions of various statistics measuring IPO relative performance. We find that small sample biases are extremely unlikely to account for the magnitude of the post-1970 IPO underperformance observed in data.
Keywords: IPO; underperformance; peso problem; Markov model
JEL Codes: G12; G14; G32
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| IPO underperformance is not merely a statistical fluke (G24) | IPO performance is statistically robust (G24) |
| observed underperformance is significant (D29) | rejection of small sample biases (C83) |
| small sample distributions do not encompass actual point statistics (C46) | rejection of small sample explanation for IPO underperformance (G24) |
| required degree of outperformance is implausibly high (D29) | necessity of significant proportion of IPOs to triple in value monthly (G24) |
| transition probabilities reveal persistent underperformance (D29) | contradiction of small sample effect (C83) |