We study timing decisions of firms to file for IPOs, and of institutional investors’ to participate in such IPOs. We model this "book-building" process as a dynamic game with asymmetric information.
- Corporate finance
- Industrial Organization
- Applied Econometrics
- Applied Game Theory
The returns on the first day of IPO are predictable. When a US underwriter files a prospectus with the Securities and Exchange Commission, the issuer reports an estimated filing range. If the stock market fares better after the filing date, the underwriter can adjust the expected IPO price upwards. Similarly, while marketing the "new firm" to the investors, underwriters may learn new (private) information and adjust the IPO price accordingly. The celebrated weak-form of the market efficiency hypothesis posits that these price adjustments between the filing date and the IPO date cannot predict the first-day returns. Yet, the empirical literature that studies IPO data finds that they do, and it is known as “partial adjustment.”
In this research project, we are interested in answering the following questions. Why do we observe partial adjustment? And what does that mean for the market efficiency hypothesis?
Several competing hypotheses try to explain this phenomenon, but each one of them answers only part of the "puzzle." For example, in an important paper, Benveniste and Spindt (1989, Journal of Financial Economics) suggest that the adjustment process reflects private information institutional investors have (about the IPO firm or the market) that have not yet been priced-in by the underwriters and so this information will positively correlate with the actual value of the firm. Alternatively, Edelen and Kadlec (1989, Journal of Financial Economics) suggest that search friction in locating willing buyers explain the "market failure."
Both these theories miss several salient data patterns. For instance, the former paper cannot explain partial adjustments to public information, and the latter article cannot explain the asymmetry in price adjustments. And both these papers ignore the fact that partial adjustment is a dynamic process by nature, where information revelation and participation occurs over time. We know from game theory, e.g., Banerjee (1992, Quarterly Journal of Economics), that in any sequential decision problem (like in the IPOs), a decision-maker looks at the decisions made by previous decision-makers in taking her own decision can lead to herd behavior. In herd behavior, investors and firms will "copy" what others are doing instead of using their information, leading to an inefficient outcome. Any static IPO model misses this crucial aspect of the book-building process, such as "riding the IPO waves," making inaccurate predictions that do not hold in the data.
Building on this prior research, in this paper, we study the dynamic decisions of both the firms’ timing to go public and institutional investors’ timing to participate in an IPO. The existing models do not address IPO timing decisions. Moreover, none of the papers study the optimal time an investor should choose to participate in an IPO (e.g., earlier vs. later), how private information endogenously reveals over time, and, in turn, affects firms’ going public decisions.
In summary, in our model, we tie all these three pieces together:
- partial adjustment to private information;
- partial adjustment to public information;
- information cascade and formation of IPO waves.
The model will combine ideas from Sequential Auctions and IPO pricing to determine (a) optimal timing for firms to initiate an IPO; (b) optimal timing for institutional investors to participate in an IPO and their bids (i.e., offers); and (c) price adjustment process by a market-maker (i.e., the underwriter). Using the model, we will determine all testable conditions on the data that will allow us to test for the different competing hypotheses. Besides testing, we will also estimate the distribution of the information that is dispersed across various participants. These rich empirical predictions shed light on the partial adjustments and will also help us understand to what extent investors' private information causes IPO underpricing, if at all.
Our desired outcomes are as follows:
- We hope to present our work in leading national and international conferences.
- In the end, we would like to publish our manuscript in one of the top peer-reviewed journals.
- We hope to engage at least one graduate student in this research project and train them on the subject matter.
We will use at least 50% of the funding to hire at least one undergraduate student and one graduate student as research assistants. Our research project is "data-heavy" and the undergraduate student will help us clean the data. The graduate student will take a more leadership role in supervising the undergraduate student and also help us in writing computer programs that can solve our dynamic game-theoretic model. We will pay them an hourly wage to clean the data and to write scripts in either MATLAB, R, or Stata. In the process, the students will learn new data skills and learn how to estimate dynamic game-theoretic (i.e., sequential auction) models. We will always guide the students and advice them as they help us with the project.