Understanding the information structure in bankruptcy liquidations is important for designing the optimal bankruptcy and financing market. In this paper, I construct a novel comprehensive bid-level bankruptcy liquidation auction data and structurally estimate the buyers' information structure using an auction model. I show that different asset types rely on different information sources, with tangible assets depending on the appraisal price, intangible assets depending on private valuations, and financial assets depending on certain insider knowledge not captured by appraisers. I text-analyze the appraisal reports using a natural language processing model and show how heterogeneous information production and bankruptcy liquidation frictions (quality management, relocation cost, and misallocation) contribute to the heterogeneity of information structure. A counterfactual simulation shows that the cost of misallocation is high (16.89%) for intangible assets compared to tangible assets (5.6%). Overall, my results suggest that tangible assets should be liquidated promptly to avoid maintenance failure, while intangible assets require a longer liquidation period to avoid misallocation costs, and financial assets should be financed by investors with expertise. 

How much value has been lost in the Chinese bankruptcy system due to excessive liquidation of companies whose going concern value is greater than the liquidation value, and how much value has been created by establishing the central liquidation platform, which handles asset liquidation instead of mandating creditors to seize the assets? I compile new judiciary bankruptcy auction data covering all bankruptcy asset sales from 2017 to 2022 in China. I estimate the valuation of the asset for both the final buyer and creditor through the revealed preference method using an auction model. On average, excessive liquidation results in a 13.5\% welfare loss. However, solely considering the liquidation process, an 8\% welfare gain is derived from selling the asset without transferring it to the creditors. Firms that are (1) larger in total asset size, (2) have less information disclosure, (3) have less access to the financial market, and (4) possess a higher fraction of intangible assets are more vulnerable to such welfare loss. Overall, this paper suggests that policies promoting bankruptcy reorganization by introducing distressed investors who target larger bankruptcy firms suffering more from information asymmetry will significantly enhance welfare in the Chinese bankruptcy market. 

Presentations: AEA Annual Meeting (Scheduled, 2024 San Antonio), FMA Annual Meeting (2023 Chicago), CFRI & CIRF Joint Conference (2023 Shanghai), CES North America Conference (2023 Oklahoma), University of Michigan Brownbag Seminar (2023, 2022), IFPHD Seminar (2023, 2022)

Award: semifinalist for the FMA best paper award (2023)

The 21st century has witnessed the rise of data-producing companies, yet evaluating the value of corporate data remains challenging due to existing accounting policies. I construct a novel dataset combining corporate 10-Q documents and website traffic to measure the size of corporate data. Leveraging a competitor's data breach event as an exogenous shock, I discover that a 1% increase in corporate data correlates with a corresponding 1% increase in firm valuation. To delve deeper into how data influences firm valuation, I conduct text analysis on the business descriptions within the 10-K files. This analysis distinguishes whether a firm's dataset is generated internally or externally by its customers. In firms where data is produced internally, data contributes to firm value by directly boosting revenue through data sales. For firms where data is produced by their customers, data increases the firm productivity through the information channel and thus increases firm valuation. 

Presentations: University of Michigan Brownbag Seminar (2020, 2019), Entrepreneurship Research Boot Camp (2019)

Award: Pathway to Research Award (2022)

This paper studies the recent rise in negative profit companies and propose one potential factor contributing to such phenomena: killer acquisitions, where the purpose of the acquisition is to terminate the target project and reduce its cost to the acquirer due to business stealing. In normal acquisitions, the value of the target equals the stand-alone value of the target plus synergies, while for killer acquisitions, the value of the target is the value the target steals from the acquirer. I show empirically that killer acquisition encourages the entry of negative profit companies within a short event window (1 to 5 years), while normal acquisitions do not. A killer acquisition reveals the cost of business stealing for incumbents to entrepreneurs and VCs thus encourages entry, while a project-based acquisition fulfills the acquirer’s demand for a project (or technology) and thus does not encourage a short run entry. I further show that this process provides VCs an additional exit option, and thus affects the VC's contract with startups: for industries with previous killer acquisitions, VC emphasize more on voting and exit rights other than cash flow rights in their contract.

Presentations: University of Michigan Brownbag Seminar (2021)