Files in this item



application/pdfSCHNEIDER-DISSERTATION-2017.pdf (737kB)
(no description provided)PDF


Title:Essays on the Iowa liquor market
Author(s):Schneider, John William
Director of Research:Deltas, George
Doctoral Committee Chair(s):Deltas, George
Doctoral Committee Member(s):Perry, Martin; Marshall, Guillermo; Reif, Julian
Department / Program:Economics
Degree Granting Institution:University of Illinois at Urbana-Champaign
Random forest
Instrumental variables (IV)
Regression discontinuity
License fees
Sales tax
Industrial organization (IO)
Abstract:This thesis presents three essays in the field of empirical industrial organiza- tion, with a particular focus on the retailer market for liquor in Iowa. This work is divided into three chapters, with each focusing on a different aspect of retailer decision making and purchasing behavior. The first chapter presents new findings on the impact of retailer licensing fees on market structure and provision of services in market equilibrium for bottled liquor in Iowa. By utilizing a unique variable fee schedule with a large recurring license fee, Iowa provides an opportunity for determining the effects of licensing fees on market outcomes. Using a regression discontinuity approach, this paper finds no effect on the number of retailers operating, but provides some weak evidence for an increase in the volume sold per store. Two key implications follow. First, even relatively large regulatory burdens on retailers have a low impact on business outcomes. Second, increasing license fees is an inefficient way of restricting alcohol consumption. These findings suggest that higher per-volume taxes on liquor may not be a good policy for reducing alcohol consumption and abuse. We recommend future work to address the potential management structure effects of licensing fee regimes. The second chapter investigates relatively novel techniques for demand es- timation. Random forest regression is quite new to economics, but is very established as a predictive tool in other disciplines. Random forest regression is better able to capture high-level interactions and nonlinearities than most other estimation techniques, but is limited in its ability to be interpreted. Here we apply this technique to a unique dataset of liquor sales in the state of Iowa to estimate demand among retailers for the top-selling liquor prod- ucts. We observe much more accurate predictions from the random forest model given the same datasets, but random forest regression exhibits a few undesirable properties regarding error distribution. The final chapter focuses on the assortment decision faced by liquor retail- ers. Any retailer selling highly-diversified, imperfectly substitutable products must contend with difficult assortment selection decisions. Retailer assort- ment decisions depend on complex, interdependent cost and revenue func- tions along with strategic considerations regarding the assortment decisions of competitors or potential market entrants. This paper contributes a novel identification strategy to generate reduced-form estimates for the effects of competitor assortment decisions among liquor retailers in Iowa. We find that, overall, retailers try to match each other’s assortment strategies, but there is considerable heterogeneity across specific products.
Issue Date:2017-05-19
Rights Information:Copyright 2017 John W. Schneider
Date Available in IDEALS:2017-09-29
Date Deposited:2017-08

This item appears in the following Collection(s)

Item Statistics