Can mergers increase output? Evidence from the lodging industry

• Author: Steven Tschantz,Arturs Kalnins,Luke Froeb
• Published date: 01 March 2017
• DOI: http://doi.org/10.1111/1756-2171.12172
• Date: 01 March 2017

RAND Journal of Economics

Vol.48, No. 1, Spring 2017

pp. 178–202

Can mergers increase output? Evidence

from the lodging industry

Arturs Kalnins∗

Luke Froeb∗∗

and

Steven Tschantz∗∗

We find that hotel mergers increase occupancy. In some specifications, price also rises. Because

these effects occur only in markets with high capacity utilization and high uncertainty, we reject

simple models of price or quantity competition in favor of models of “revenue management,”

where firms price to fill available capacity in the face of uncertain demand.

1. Introduction

For mergers,the predictions of theor y—anticompetitivemergers increase price and decrease

output, whereas procompetitive mergers do the opposite—are so well accepted that they are

enshrined in both law and policy.1However, there are particular circumstances where the standard

predictions do not apply (e.g., Froeb, Tschantz, and Crooke, 2003) and evidence on mergers

remains thin. This has led to calls for more and better empirical work, especially from those

who work at the enforcement agencies, for example, Froeb et al. (2005), Carlton (2007), Farrell,

Pautler, and Vita (2009), and Bailey (2010). These policy makers want enough data on mergers,

and enough variation across observed merger effects,to deter mine whichtheoretical model should

be used to evaluate the competitive effects of a particular merger in a particular industry (FTC,

2005).

In this article, we estimate the competitive effects of mergers in the US lodging industry,

and use the results to distinguish between different classes of oligopoly models. The peculiar

features of this industry—capacity is fixed, marginal costs are small relative to fixed or sunk

costs, and price must be set before demand is realized—imply that firms maximize profit by

“managing revenue,” that is, by pricing to match expected demand to capacity (Anderson and

Xie, 2010; Talluri and Van Ryzin, 2004; provide many examples). Under these conditions, the

standard merger predictions of increased price and decreased output need not apply. Indeed, in

the empirical work that follows, we test these predictions and reject them.

∗Cornell University; atk23@cornell.edu.

∗∗Vanderbilt University; luke.froeb@vanderbilt.edu, tschantz@math.vanderbilt.edu.

Wewish to acknowledge useful comments from colleagues at Colgate, Columbia, Delaware, Federal Trade Commission,

Lehigh, INFORMS Revenue Management Conference, Ithaca, NY, 2010, US Dept. of Justice, and Vanderbilt.

1See, for example, the US Dept. of Justice and Federal TradeCommission, Horizontal Merger Guidelines, 1997.

178 C2017, The RAND Corporation.

KALNINS, FROEB AND TSCHANTZ / 179

In general, estimating the competitive effects of mergers presents at least twoproblems. The

first is censoring. Most big mergers must obtain regulatory clearance from competition agencies.

This means that we observe onlysmall mergers, those that were thought not to be anticompetitive,2

or those which were allowed to proceed after court or regulatory challenges failed.3The second

problem is that mergers may have a variety of effects, many of them unrelated to their effect on

competition. For example, an acquisition may induce higher levels of effort among employees of

the acquired firm as they try to impress new management. This could lead to an observed merger

effect that has nothing to do with competition.

To address these difficulties, we examine a large sample of US lodging mergers in many

different local markets. Our sample of mergers is the largest that we know of.4Almost all of the

898 lodging mergers are too small to raise anticompetitive concerns, so data censoring, and its

attendant bias, are expected to be small.

The relatively small size of the merging firms means that we do not expect to see big

anticompetitive effects in the data. However, due to our large sample size, we expect to have

enough statistical power to estimate small merger effects,if they exist. To isolate the competitive

effects of mergers, we measure the effects of within-market mergers (in the same geographical

tract) relative to out-of-market mergers. This comparison removes merger effects unrelated to

competition, using a difference-in-difference estimator (premerger versuspostmerger and within-

market versus out-of-market).

Our main empirical finding is that hotel mergers raise occupancy,without reducing capacity.

In some regressions, price also appears to increase. These effects are small, but statistically and

economically significant. They occur only in markets with high capacity utilization and high

uncertainty.

We identify two theoretical mechanisms that could explain the empirical results. The first is

that mergers reduce uncertainty about demand. Less uncertainty means better forecasts, which

means fewerpricing er rors—and higher occupancy—as the hotels are better ableto match realized

demand to available capacity.

A second mechanism—that the merged hotels are better able to compete for group and

convention business—could also account for price and/or quantity increases. If the merged

hotels are large enough to host groups, but the premerger hotels are not, then the merger could

increase convention demand for the mergedhotels. However, our finding that occupancy increases

only in markets with high levels of uncertainty and capacity utilization leads us to prefer the

former explanation, although we cannot rule out the possibility that increases in group demand,

particularly during low-demand periods, could account for the results.

In Section 2, we motivate the topic by reviewing three antitrust investigations, in three

different industries, where firms manage revenue. In Section 3, wecharacterize mergers in several

different classes of oligopoly models to developtestable hypotheses. We test these hypothesesand

interpret the results in Sections 4 and 5. In Section 6, we conclude by discussing the implications

of our results for oligopoly modelling and antitrust policy.

2. Motivation: antitrust enforcement in industries where firms

manage revenue

In two of three antitrust decisions reviewed in this section, it appears that enforcement

decisions were guided byintuition gleaned from standard models of price or quantity competition.

2See Pautler (2003) and Danzon, Epstein, and Nicholson (2007) for summaries of empirical work in the banking,

pharmaceutical, and biotech industries.

3Werden, Joskow, and Johnson (1991) estimate the effects of the Northwest/Republicairline merger; and Farrell,

Pautler, and Vita(2009) repor t on hospital mergers.

4The largest published study is that by Dranoveand Lindrooth (2003), who study the cost effects of 122 hospital

mergers.

C

The RAND Corporation 2017.