Stephan Seiler Marketing Professor Imperial College London
Professor of Marketing
Imperial College Business School
South Kensington Campus
London SW7 2AZ
Email: Stephan.a.seiler@gmail.com
Google Scholar profile
SSRN profile
LinkedIn profile
Twitter: https://twitter.com/SeilerStephan
I use data to understand how consumers make choices in settings
ranging from laundry detergent discounts to choosing a hospital for
a bypass operation. I am particularly interested in how consumers gather
information before making a purchase and what we can learn from data on
consumer search behavior.
I am an Associate Editor at Marketing Science, Management Science,
Quantitative Marketing and Economics, and the Journal of Industrial Economics.
I co-organize the European Quant Marketing Seminar (eQMS), and I am a Public Editor at @QME_Journal.
Recent Research:
New Companion Overview Papers on "Consumer Search" (both with Elisabeth Honka and Raluca Ursu)
(1) Consumer Search: What Can We Learn from Pre-Purchase Data
This paper provides a high-level overview of substantive research areas where search data is particularly valuable. We first provide background on different types of search data and model frameworks and then cover how search data can be used to better estimate preferences, to analyze how marketing variables affect search behavior, and to study the impact of search frictions on market outcomes.
(2) The Sequential Search Model: A Framework for Empirical Research (with Accompanying Code)
In this more technical paper, we provide a unified treatment of the sequential model in empirical work aimed at especially at researchers that are new to working with search data and models. We also provide a comprehensive code base that covers different methods for computing reservation utilities and 4 different estimation approaches. Click here for access to the code.
🚨🚨 New Working Paper: 🚨🚨 How Much Influencer Marketing is Undisclosed? Evidence from Twitter
In this new paper with Daniel Ershov and Yanting He, we develop a new method to detect undisclosed sponsored content on Twitter.
We gather a novel data set of over 100 million posts across 268 brands from 2014 to 2021 and find that 96% (!!!) of sponsored content
is undisclosed. Despite tightening regulation, the share of undisclosed content decreases only slightly over time. Undisclosed
content is more likely to originate from younger brands with a large Twitter following, suggesting that disclosure might remain low
in the future.
New Paper: Demand Estimation with Text and Image Data 👈👈
New machine learning tools allow researchers to process unstructured data from text and images more easily. In this new paper
with Giovanni Compiani and Ilya Morozov, we extract measures of similarity between products from product images and different
sources of text (product descriptions, reviews, etc.) and then feed them into a demand model so that higher similarity leads to larger
cross-price elasticities between products. We apply the model to data from multiple categories and show that it helps us recover
flexible substitution patterns.
Paper on Soda Taxes featured on the "How I Wrote This" podcast.
Our paper on soda taxes (with Anna Tuchman and Song Yao) was covered in the most recent
episode of the “How I Wrote This” podcast. We discuss what motivated us to pursue this research,
the various decisions that we took when working on the paper, and how we navigated the review process.
Find out more about my recent research projects on my Blog.
... or follow me on Twitter ...
