Do retail mergers affect competition? Evidence from grocery retailing

• DOI: http://doi.org/10.1111/jems.12218
• Author: Luke M. Olson,Daniel S. Hosken,Loren K. Smith
• Date: 01 March 2018
• Published date: 01 March 2018

Received: 27 October 2015 Revised: 19 May 2017 Accepted: 25 July 2017

DOI: 10.1111/jems.12218

ORIGINAL ARTICLE

Do retail mergers affect competition? Evidence from grocery

retailing

Daniel S. Hosken1Luke M. Olson1Loren K. Smith2

1U.S. Federal TradeCommission, 600 Penn-

sylvaniaAve NW, Washington, DC 20580,

USA

2Compass Lexicon, Washington,DC, USA

Abstract

This study estimates the price effects of horizontal mergers in the U.S. grocery retail-

ing industry. We examine fourteen regions affected by mergers, including mergers in

highly concentrated and relatively unconcentrated markets. We identify price effects

by comparing markets affected by mergers to unaffected markets using difference-

in-difference estimation with three different comparison groups, propensity score

weights, and by using the synthetic control method. Our results are robust to the choice

of control group and estimation technique. Wefind that mergers in highly concentrated

markets are most frequently associated with price increases, and mergers in less con-

centrated markets are most often associated with price decreases.

1INTRODUCTION

Economists long have believed that, all else being equal, mergers resulting in highly concentrated markets reduce competition,

increase consumer prices, and reduce consumer welfare. This belief provides the basis for much of the world's antitrust policy.

The United States, United Kingdom, and European Union, for example, review horizontal mergers prospectively.1The problem

for regulators is determining which mergers are likely to result in reduced competition. Unfortunately, there is remarkably little

reliable systematic evidence relating measures of market concentration, such as the Herfindahl-Hirschman Index (HHI),2to

manufacturer markups or consumer prices.

In this paper, we examine how prices change following significant changes in market structure resulting from a relatively

large number of mergers in the supermarket industry.By focusing on how changes in market structure induced by mergers affect

consumer prices, we can implicitly control for endogenous factors that determined the premerger marketstr ucture. We estimate

the causal effect of mergers on supermarket prices using two related empirical techniques. We begin by following the extant

literature and estimate merger price effects using a difference-in-difference analysis: we compare prices in markets experiencing

a merger to those in similar markets not experiencing a major change in market structure. The major criticism of this method

is that the decision to merge itself may be related to market participants’ expectations about future prices resulting in biased

estimates of the merger's effect on price. To address this concern, we first estimate two variants of the difference-in-difference

estimator that use the relative apriorilikelihood of a comparison market experiencing a merger to either limit the comparison

group (as suggested by Crump, Joseph Hotz, Imbens, & Mitnik 2009) or by weighting comparison markets by their estimated

propensity score (Hirano, Imbens, & Geert 2003; Imbens 2004). Second, we estimate merger price effects using the synthetic

control method developed by Abadie, Diamond, and Hainmueller (2010).

We would like to thank PatrickDeGraba, Paul Pautler, David Schmidt, Steve Tenn, Mike Vita, Matt Weinberg,Nathan Wilson, seminar participants at Drexel

University, two anonymousreferees and the editor for their comments. The views expressed in this paper are those of the authors and do not represent those

of the U.S. Federal Trade Commission or any individual Commissioner. Corresponding author: Hosken, U.S. Federal Trade Commission, 6th and Pennsyl-

vania Ave, NW, Washington, DC 20580, dhosken@ftc.gov. Olson: U.S.Federal Trade Commission, 6th and Pennsylvania Ave, NW,Washington, DC 20580,

lolson@ftc.gov, Smith: Compass Lexecon, 1101 K Street, NW,Washington, DC 20005, LSmith@CompassLexecon.com.

J Econ Manage Strat. 2018;27:3–22. © 2017 WileyPeriodicals, Inc. 3wileyonlinelibrary.com/journal/jems

4JOURNAL OF ECONOMICS & MANAGEMENTSTRATEGY

A strength of our study, in contrast to much of the literature,3is that we estimate the price effects of a number of merg-

ers in the same industry affecting different geographic markets with different levels of market concentration at roughly the

same time. Our approach follows the suggestion of Carlton (2009) that researchers should examine the price effects of all

mergers (those likely and unlikely to result in price increases) to more fully understand how mergers affect the competitive

process. Our sample contains eight mergers that took place in highly concentrated markets and six mergers that took place

in moderately concentrated or unconcentrated markets. Our results are largely consistent with the presumptions of antitrust

regulators as stated in the Horizontal Merger Guidelines.4We find that five mergers resulted in estimated price increases of

more than 2%, and that four of those mergers were in highly concentrated markets. Five mergers resulted in estimated price

decreases of more than 2%, and only one of those occurred in a highly concentrated market. The remaining four mergers were

associated with relatively little change in price. These findings are robust to the choice of comparison group and estimation

technique.

Our paper adds to the literature that evaluates the efficacy of antitrust enforcement by examining how prices change fol-

lowing mergers of competing firms. Roughly 60 published studies have estimated the price effects of mergers, with the major-

ity finding that mergers have resulted in price increases.5The ability to draw general conclusions regarding the efficacy of

horizontal merger policy from the published literature, however, is limited. Only a tiny fraction of the thousands of mergers

filed with the U.S. antitrust agencies have been studied, and among those the majority of studies estimate the price effects

of mergers taking place in one of only four industries: banking, airlines, hospitals, and petroleum. Of particular interest to

our paper are two recent studies that each estimate the price impact of a single merger in the supermarket industry. Huang

and Stiegert (2009) found that the merger of grocery retailers Kohls and Copps in Madison, Wisconsin did not raise prices in

the months immediately following the merger, but did increase prices relative to a control market two years after the merger.

Allain, Chambolle, Turolla, and Villas-Boas (2013) examined a largemerger of French supermarket retailers that affected many

cities, and found that prices increased in markets directly affected by the merger relative to a comparison group of unaffected

markets.

The remainder of this paper is organized as follows.Section 2 describes our data sources, and Section 3 presents t he methodol-

ogy used to construct our merger and comparison markets. Section 4 describes our estimation strategy and presents the empirical

findings of the study. Section 5 concludes.

2DATA

Our study uses three data sources. The first is A.C. Nielsen's Trade Dimensions retail database. Each year Trade Dimensions

creates a census of retail outlets operating in the United States for a number of retailing industries, including supermarkets, club

stores, liquor stores, conveniencestores, and restaurants. In this study, we focus on the primary formats used for grocery retailing:

conventional supermarkets, supercenters, and club stores.6Our dataset consists of annual observations, including the location,

size, estimated sales, the store's banner (the name the store operates under), and corporate ownership of each supermarket,

supercenter, and club store in the United States from 2004 through the fall of 2009. An additional feature of the dataset is that

every store location has a unique identification number that allows us to track store ownership over time.

The price data we use consists of the prices used to construct the ACCRA Cost of Living Index, produced by the Council for

Community and Economic Research (CCER). The ACCRA price index is designed to compare the cost of living for moderately

affluent professional and managerial households in different U.S. metropolitan areas at a point in time.7The price data assembled

by CCER are collected by the staff of roughly 350 local U.S.Chambers of Commerce.8In t hisstudy, our primary dataset consists

of the prices collected for the 26 grocery products in the ACCRA sample.9These prices typically correspond to a distinct food

product, such as a pound of T-Bone steak or a 2-liter bottle of Coca-Cola, sold at a specific retail outlet on a given day.

CCER reports its data at the level of the Core Based Statistical Area (CBSA). A CBSA is defined as a set of adjacent counties

connected by commuting ties to a common urban core of at least 10,000 residents and is designed to capture the political

jurisdictions in a region that are closely connected by commerce.10 The population of the markets (CBSAs) included in the

CCER data varies dramatically from medium sized markets such as Lima, Ohio (106,000) to the largest U.S. CBSA of New

York City (19,800,000). Smaller markets in the CCER data tend to have fewer price quotes per item than large markets.11

Because we observe the retail banner that a price quote corresponds to but not the specific retail location that was visited, we

treat the CBSA in which the price was collected as the geographic unit of observation.

The CCER data is particularly well-suited to our study.First, it contains prices on a broad set of supermarket products designed

to measure the typical “market basket” of consumers’ food purchases. Second, the data covers more geographic regions within

the United States than any other publicly accessible pricing dataset of which we are aware.This allows us to study many mergers