My major research interests lie in technology and entrepreneurship. Specifically, I examine how information technologies such as social media and AI technologies facilitate startup scaling. I use empirical methods including econometrics and machine learning.

In my previous and ongoing projects, I find information technologies like social media and analytical tools can help startups better scale in financing, institutions and innovations. Traditionally disadvantaged groups of entrepreneurs (e.g., female and less-connected ones) benefit more from using such IT tools.

While researching above questions, I also generate new empirical questions. For example, whether and why companies adopt multiple social platforms? Do firms lie on social media? These questions have also developed into a number of new research projects.
Job Market Paper

Artificial Intelligence, Lean Method and Startup Product Scaling

3 Puzzles about AI and innovation:

- Why most companies have not seen benefits from their AI investment?

- The greater the novelty of an innovation, the less likley there is reliable data available. So can AI really work for innovation?

- Does AI replace experimentation methods in creating innovation, or complement?

This paper is to answer the above questions.

Abstract: Despite variability in returns, many startups have adopted artificial intelligence (AI) to develop new products. We examine the impact of AI capability on startup product innovations, as well as the role of lean methods in explaining some of the variabilities in the return on AI investments. We identify the startups’ internal and external AI capabilities based on the company’s usage of AI in processing internal or external information. Using a novel dataset of about 2,000 startups in China from 2011 to 2020, we find that companies that adopt general AI capability (either internal or external AI) create more innovative products, including both novel products (i.e., products that are new to the market) and incremental products (i.e., new versions of existing products). Moreover, AI investments complement lean methods in product innovations. Separating general AI into internal and external AI capabilities, we find that lean methods complement external AI capabilities in developing novel products, and internal AI capabilities in developing incremental products. Our further analyses show that using lean methods to provide experimentation data and using AI to improve the efficiency of experimentations creates a virtuous cycle that can help alleviate market uncertainties in developing novel products and facilitate product iterations in developing incremental products. These findings are consistent in both software companies and companies that develop physical products.

Publications / Accepted Manuscripts

Management Science, Forthcoming
Kauffman Best Paper Award, International Conference on Information Systems, 2019
Under Review / Prepare for Submission

Can Social Media Change Institutions? Evidence from Chinese Entrepreneurship (with Lynn Wu)
CIST 2022, SCECR 2022, DRUID 2022
Abstract: Institutional environments play an important role in entrepreneurial activities. In this study we examine whether startups’ social media activities could improve the institutional environments that they face. Using comprehensive firm information and social media data from WeChat for about 2,000 Chinese startups from 2011 to 2020, we find that social media activities can help companies with few political connections to obtain institutional resources. They are more likely to receive administrative approvals from the government to conduct their primary business activities. They are also more likely to receive positive media coverages. We find that social media can not only help these companies broadcast quality signals, but also direct public opinions and mobilize public support in their favor, all of which are essential to establish public legitimacy. These effects are especially strong for firms that face greater institutional barriers and have scarce resources. In addition to social media engagement, the content of social media post also plays a critical role. We find that posts about corporate social responsibility (CSR) play a bigger role than others in helping firms receive favorable treatment and establish public legitimacy. Consistent with our proposed mechanism, we find social media’s effect is significantly reduced after implementing policies that lower institutional barriers, such as after the Mass Entrepreneurship and Innovation Policy.

You Should Diversify, but Why? Multi-Platform Social Media Strategy and E-Commerce Performance (with Yakov Bart, Serguei Netessine and Lynn Wu)
SCECR 2022, CIST 2021
Abstract: Over last few decades, many large social media platforms emerged, gradually becoming indispensable channels for e-commerce retailers to build relationships, maintain communications and advertise to consumers. While some online retailers choose to concentrate their engagements on the fewer social platforms, others adopt a more diversified social media strategy by placing eggs in multiple baskets. Which approach is preferable and why? Using a longitudinal dataset on e-commerce social media metrics and performance indicators, we find that holding other factors constant, companies adopting a more diversified social media strategy perform better than those using a more concentrated strategy. This is because maintaining a more diversified social media strategy is positively associated with having website visitors with greater purchase intentions. Specifically, we show that diversifying social media strategy is associated with higher conversion rates, pages per visit and visit duration, and lower bounce rate, leading to 2~5% increase in total web sales. We also find the benefits of social media diversification are more pronounced for less popular retailers who has greater need for high-purchase-intention visitors. We further show that having website visitors with higher purchase intentions is more likely due to the synergistic effects generated by overlapped impressions from multiple social media sites, instead of different platforms offering distinct capabilities for driving traffic.