Hi, I'm Zijun Tian

Fifth year Ph.D. student in the Economics Department

@ the University of Pennsylvania

Email: zjtian96@sas.upenn.edu

You can download my CV here.

Education

Ph.D. in Economics, University of Pennsylvania, 2017 - Now

B.S. in Mathematical Decision Science & Economics, University of North Carolina at Chapel Hill, 2014 - 2017

Research Interests (influencer marketing, social media, online digital marketing)

My research focuses on developing both theoretical and quantitative models to study different aspects of influencer marketing which provide managerial insights on 1) how firms optimize their cooperation with influencers and design of marketing campaigns on social media; 2) how platforms improve their content arrangement and topic recommendation given users' content browsing/creating preferences and the observed topic evolution. In particular, I propose and apply machine learning and deep generative models to explore the numerous, unstructured, multi-dimensional data from TikTok and Instagram and investigate interesting marketing questions, driven by consumer and diffusion theories.

Working Papers


  • Winning the Attention Race: Analyzing Content Popularity And Topic Evolution On TikTok (Zijun Tian, Ryan Dew and Raghuram Iyengar)

[Abstract]

In this project, we build a large-sized panel dataset of the hashtag-grouped short videos on the Discover page of TikTok and leverage that special hierarchical data structure to explore what content starts and grows to be more viral, what factors explain the popularity of both hashtags and videos within them, and how the content within hashtags evolves over time. This project is the first to thoroughly examine empirical regularities on TikTok and investigate the unique form of influencer marketing and video advertising via sponsored topics. These insights have implications for firms to better manage their social media campaigns and platforms to optimize their content recommendation toward the whole community to meet their desired marketing goals (e.g. boosting virality vs. encouraging content diversity). More specifically, we first predict the dynamic video-level impression growth by developing an 2-step predictive algorithm that automatically decomposes viewership of each video. In the first step, we establish a hierarchical Bayesian framework that evaluates the influencer’s impact on video impression growth and how that might be differed by some macro-level hashtag attributes. Then we further predict the unexplained impression growth left from the first step based on our proposed representation learning framework (VS-VAE) coupled with a neural network predictor. It encodes each unstructured, multi-dimensional video into a lower-dimensional latent representation and predicts its pure content attractiveness in a scalable nonlinear way. Finally, based on the learned video representations, we compare videos and measure the amount of innovation/imitation in terms of content similarity happening under each hashtag and trace it over a fixed time horizon to obtain their respective content evolution trajectories. They are converging to their own steady states, influenced by the hashtag design, and, together with the hashtag design, significantly affect the virality of a hashtag.

  • How Should Firms Cooperate with Honest Influencers? (Zijun Tian)

[Abstract]

In this paper, I strategically model the firm’s cooperation with honest influencers, analyze when the firm benefits from such cooperation and its optimal “design” of the influencer’s credible message through quality revelation and influencer selection. I address the influencer's concern of his/her reputation by setting his/her utility function to be a linear combination of the firm's payment and the negative expectation inflation from the "misled'' consumers. We find that the firm can asymmetrically gain from the uncertainties introduced through the influencer’s message that shift the consumers’ beliefs in its preferred way and achieve a positive profit even when the product is of fairly lower-than-expected quality. Moreover, analogous to the “signal jamming” literature in persuasion, we similarly find that with very favorable or unfavorable product quality, the firm does not distort the influencer’s signal and reveals the true quality to the influencer. However, the "jammed" message resulting from the influencer's imperfect learning of the product under no revelation does increase the firm’s profit when sponsoring mediocre products. Finally, we consider one real-life application that extends our previous analysis to the new vs. established firm whose new product differs in quality variance. The results show that higher quality variance increases the profitability of cooperating with an honest influencer when the product is not terribly low due to more uncertainties in the consumers' evaluation process, while lower quality variance increases the applicability of such cooperation due to the influencer's lower reputation risk of writing a sponsored review. They offer managerial implications on how firms should cooperate with honest influencers for higher economic returns. For example, to boost profits, they can sponsor influencers for their brand new products instead of merely new generations of older products or cooperate with influencers whose reviews are more descriptive rather than informative.

Works in Progress


  • What Content to Post by Amateur vs. Professional Influencers and Whom to Sponsor? (Zijun Tian)

[Abstract]

In this project, I consider a heterogeneity in influencers in terms of their qualification for sponsorships. Due to the easy and almost costless entry to social media platforms, more and more users are competing for firms’ sponsorships. In order to be attractive for advertisers, the amateur influencers need to post what they believe can help them gain a large following which makes them visible to firms' sponsorships, while the professional influencers want to profit from their posts while still keeping their follower loyal. Therefore, I first theoretically model the “amateur” influencer’s imitating vs. differentiating behavior, the “professional” influencer’s alternation and trade-off between posting ads. vs. organic stories, and their interactions. Then, I systematically collect all post data for a random set of the above-mentioned two types of influencers on Instagram. For the amateur influencers, I test whether they imitate in their posts, what they imitate, whether and to what degree imitation helps them transition to professional. For the professional influencers, I test their alternating posting behavior in response to the observed imitation from the amateurs and the attractiveness of the firm's sponsorship. Finally, I test whether the above results would be differed by product categories and potential editorial bias on Instagram. The results would help firms understand the attraction and repulsion among influencers and thus help them make better decisions on which influencers to sponsor in order to expand the eWOM of their sponsored content. For example, to leverage the free advertising and the eWOM among the imitating amateurs and their followers (in addition to the original paid advertising from the professionals), firms should sponsor ''more" influencers if the amateurs imitate popular topics. If they imitate successful idols instead, then firms should sponsor "more influential" influencers.