Working Papers
"Strategic Advertising and Release in the Movie Industry" (2019) - Job Market Paper
Setting the release date and advertising for a blockbuster movie are the most important decisions a studio makes in movie distribution. I propose a model which combines studios’ release date (entry) and advertising choices under incomplete information. Studios optimize by setting release dates followed by advertising budget, and have payoffs structurally determined and dependent on demand. I adopt an estimation method which accounts for advertising and choice set endogeneities, the latter leading to a selection bias if not corrected. Model estimates, in particular the effect of advertising, are significantly different from the traditional estimates. I show that ignoring advertising and/or choice set endogeneity significantly underestimates the impact of advertising on box office. In light of the recent Disney-Fox merger, I simulate how this event will impact the movie industry.
Full draft available by request
This paper combines an entry model under incomplete information with Bertrand competition for differentiated products. Our model can be also viewed as an extension of Berry, Levinsohn and Pakes (BLP, 1995) that allows for endogenous consumers' choice sets as firms choose which products to offer. We show that the demand and cost parameters as well as the entry costs are identified. We then propose a semiparametric modification of the BLP estimator that accounts for the endogeneity of the choice set. A Monte Carlo study shows the inconsistency of the usual estimators and the good performance of our estimator in finite samples. Extensions of our methodology include endogenous product characteristics, a random price coefficient and incomplete information in both entry and pricing stages.
"Strategic Release of Blockbuster Movies" (2017) - 3rd Year Paper
The US blockbuster movie market is characterized by oligopolistic competition amongst major Hollywood studios. I examine those studios’ strategic concerns regarding the release dates of their movies by estimating an incomplete information timing game. I find the presence of a negative externality, namely that studios’ profits suffer when they release in the same week as others. This can be explained by theater capacity constraints. More over, perhaps more surprisingly, I also find no significant genre effect on studios’ timing decisions, suggesting that demand for blockbuster movie may be capped in any given week and/or that many movie goers have preference for a specific genre.
Work In Progress
"Do Instagram Influencers Influence the Shopping Cart?" (2019) - with Luis Cabral and James Nesbit
Data collected, analysis in progress
This paper models how Instagram influencers’ posts drive sales. Instagram is a social media platform that is increasingly becoming the go-to platform for brands to advertise on. Rather than posting the advertisements themselves, brands look to "influencers" who specialize in certain niches, and possess large amounts of followers who are potentially interested in the advertised products. According to AdWeek, posts on Instagram receive on average significantly more likes than their Facebook counterpart. Arguably the most engaging social media platform, Instagram’s users tend to be younger, and use the platform to discover new products. As a consequence, brands are increasingly spending big bucks to sponsor Instagram influencers.
We collect a unique dataset of influencer activities (engagement) and sales of a food and beverage company which predominantly sells its product online. We model the process in which engagement drives sales, through the evolution of a latent variable related to brand capital. We employ several innovations in cleaning up the raw engagement dataset, reflecting concerns regarding fake followers/likes, and other ways an influencer may artificially boost engagements. In addition, we introduce a new way of measuring traffic/interest in the brand, in light of the uniqueness of Instagram. Our model aids brands in analyzing the success of their social media campaigns without needing various data on impressions and traffic that are only accessible by Instagram.
"Modeling How Influencers Bid for Campaigns - a Non-Parametric Approach" - with Isabelle Perrigne and Quang Vuong
Data collection in progress
We estimate a multi-product scoring auction with an unknown scoring rule, with exogenous and endogenous bidder characteristics. We apply the model to the context of influencer marketing on Instagram. Bidding for campaigns is a key part of influencer marketing, with important implications for campaign profitability and relationship building. Evidence from the industry reveals that even controlling for following and engagement, influencers' rates for campaigns vary significantly. Our contribution is two-fold: first, we provide identification and estimation strategies for such multi-unit scoring auctions with endogenous bidder characteristics; second, we apply our estimator to a unique dataset of influencer bids for multiple Instagram campaigns. Our analysis informs brands about the distribution of bidders' valuations controlling for their observable characteristics, thereby guiding brands on how to select influencers for social media campaigns.
"Designing the Optimal Brand-Sponsored Post on Instagram" - with Luis Cabral and James Nesbit
Data collection in progress
We analyze what Instagram post characteristics lead to high engagement, and whether the (lack of) alignment of the sponsoring brand and the influencer niche matters. We consider both image and caption characteristics and analyze a large set of sponsored influencer posts on Instagram in the food and beverage space, across multiple influencers and campaigns.
"Sponsorships, Authenticity and Future Earnings: A Dynamic Model of an Instagram Influencer" - with Luis Cabral
We develop a theoretical model of reputation, in which an Instagram influencer faces the following trade off: she may accept an offer to post a sponsored ad on her channel, earning income in the current period, but potentially slowing down her follower growth by annoying her followers with the ad. Alternatively, she may choose to post a piece of organic content this period, which may facilitate her follower growth and thus her future earning capability. Influencers may be a good type or an action type (higher probability of recommending a bad product). Consumers Bayesian-update their belief about the probability that the influencer is a good type, given the history of good/bad products. We analyze the equilibrium outcome and how the action type influencers' probability of promoting a bad product depends on the initial belief of consumers of that they are a good type and the history of recommendations.