The accuracy of hospital merger screening methods

• DOI: http://doi.org/10.1111/1756-2171.12215
• Published date: 01 December 2017
• Author: Christopher Garmon
• Date: 01 December 2017

RAND Journal of Economics

Vol.48, No. 4, Winter 2017

pp. 1068–1102

The accuracy of hospital merger screening

methods

Christopher Garmon∗

This article analyzes the accuracyof various prospective hospital merger screening methods used

by antitrust agencies and the courts. The predictions of the screening methods calculated with

pre-merger data are compared with the actual post-merger price changes of 28 hospital mergers

measured relative to controls. The evaluated screening methods include traditional structural

measures(e.g., Herfindahl-Hirschman Index), measures derived from hospital competition models

(e.g., diversion ratios, willingness-to-pay, and upward pricing pressure), and hospital merger

simulation. Willingness-to-pay and upward pricing pressure are found to be more accurate at

flagging potentially anticompetitive mergers for further investigation than traditional methods.

1. Introduction

The hospital industry is one of the largest and most dynamic sectors in the United States

economy. In 2015, hospital services accounted for 5.7% of US Gross Domestic Product (GDP),

more than any other category of health expenditure.1A large fraction of US hospital expenditures

(40%) are financed with private health insurance or patient out-of-pocket payments. In recent

years, the growth of privately financed hospital expenditures has been driven almost entirely

by hospital price increases.2In most states, hospital prices charged to private health insurance

companies are unregulated and determined by negotiations betweenhospitals and health insurance

companies. The negotiated prices are determined in large part by local competitive conditions

and the ability of health insurance companies to substitute with competing hospitals in their

managed care networks. Hospital antitrust enforcement plays a significant role in US healthcare

∗University of Missouri—Kansas City; garmonc@umkc.edu.

This research was conducted while I was employedby the Bureau of Economics, Federal Trade Commission. The views

expressed in this article are the author’s and not necessarilythose of the Commission or any individual Commissioner.

I thank anonymous reviewers, Keith Brand, Leemore Dafny, Dan Hosken, Devesh Raval, Sean May, Seth Sacher,

David Schmidt, Loren Smith, Michael Vita, Nathan Wilson, Zenon Zabinski, conference and seminar participants at

the International Industrial Organization Conference, FTC Microeconomics Conference, DC Industrial Organization

Conference, KenyonCollege, and RAND Health for their suggestions, and Michael Bohne, Chris Carman, Laura Kmitch,

and Jordan Rhodes for their research assistance.

1Centers for Medicare and Medicaid Services, Historical National Health Expenditure Data, www.cms.gov/

Research-Statistics-Data-and-Systems/Statistics-Trends-and-Reports/NationalHealthExpendData/

NationalHealthAccountsHistorical.html (accessed on 6/6/2017).

22015 Health Care Cost and Utilization Report, Health Care Cost Institute, www.healthcostinstitute.org(accessed

on 6/6/2017).

1068 C2017, The RAND Corporation.

GARMON / 1069

cost containment by preserving hospital competition and limiting hospital price growth, while

also promoting quality and access to healthcare.

Over the past 25 years, significant changes have occurred in US hospital antitrust enforce-

ment. Between 1993 and 2000, during the largest hospital merger wave in US history, federal and

state antitrust authorities challenged eight proposed hospital mergers in federal court and failed in

each attempt. This string of setbacks led to an explosion of research on hospital competition and

the effects of hospital mergers. One branch of the literature retrospectively studied the effects of

past hospital mergers and found that the tools and assumptions upon which courts relied during

the 1990s often led to incorrect conclusions about the likely effects of hospital mergers. Another

branch of the literature attempted to model price formation in hospital markets and developed a

set of tools to directly assess the lost competition from hospital mergers and predict their price ef-

fects. These tools (e.g., diversion ratios, willingness-to-pay, upward pricing pressure, and merger

simulation) were used by the federal antitrust agencies in recent hospital merger challenges and,

unlike the 1990s, many of these challenges have been successful.

The use of tools that directly measure lost competition in hospital antitrust enforcement has

occurred alongside the general evolution of merger review in differentiated product markets and

the increasing reliance on direct measures of lost competition by the federal antitrust agencies in

the United States. The Federal Trade Commission (FTC) and the Department of Justice (DOJ)

revised their Horizontal Merger Guidelines (HMG) in 2010 to emphasize direct measures of

competition (e.g., diversion ratios and the value of diverted sales) and de-emphasize traditional

concentration measures (e.g., the Herfindahl-Hirschman Index [HHI]) in differentiated product

markets.3

With the recent use of newscreening tools in hospital antitr ust enforcement and the emphasis

on similar direct measures of competition in the review of mergers in other differentiated product

markets, it is important to evaluate whether these newtools are accurate in predicting post-merger

price changes. The original articles that developed the hospital screening tools did not assess the

accuracy of their predictions against actual post-merger outcomes. This article offers the first

comprehensive comparison of the predictions of a wide range of screening tools against the

actual post-merger price changes of a relatively large sample of hospital mergers. The actual

post-merger price changes (measured relative to controls) of 28 hospital mergers are compared to

the predictions of various screening methods. The screening methods include direct measures of

the competition between the mergingpar ties (i.e., diversionratios, upward pricing pressure [UPP],

and willingness-to-pay [WTP]), merger simulation, and traditional concentration measures (i.e.,

Herfindahl-Hirschman Index [HHI]) calculated with various market definitions and market share

metrics.

The focus of the analysis is on evaluating methods that can be implemented with data

that are likely available to regulators during the initial preliminary investigation of a merger.

It is at this stage that delineating between possible anticompetitive mergers and beneficial or

innocuous mergers is most useful and imposes the least regulatory cost. Although a full-phase

investigation can provide the regulator with detailed data and other evidence to increase the

precision of its estimates, a full-phase investigation imposes significant costs on the merging

parties and the regulator. The ideal screen for an initial investigation avoids “casting a wide net”

and instead focuses the regulator on the mergers most likely to be anticompetitive. All of the

screening methods evaluated in this article can be calculated with data often available without

a full-phase investigation: patient discharge data and other public data sets. It is important to

note that this excludes merger simulations calibrated with health insurance claims data, such as

Gowrisankaran, Nevo,and Town (2015). This article evaluates onlymerger simulations calibrated

with less-detailed hospital and discharge data.

3For example, from Section 6.1 of the 2010 HMG: “The Agencies rely much more on the value of divertedsales

than on the level of the HHI for diagnosing unilateral price effects in markets with differentiated products.”

C

The RAND Corporation 2017.

1070 / THE RAND JOURNAL OF ECONOMICS

Any evaluation of mergerscreening methods is complicated by active antitrust enforcement.

In an era of active and effective hospital antitrust enforcement, most mergers that are likely to

be anticompetitive on balance are blocked or never proposed. Thus, a sample of consummated

mergers taken from a period of active antitrust enforcement may be truncated and biased toward

mergers with limited reductions in competition and significant procompetitive effects (e.g., cost

savings) (Carlton, 2009). To address this issue, our sample of consummated hospital mergers

includes 12 mergers in North Carolina and Missouri that occurred between 1997 and 2001. This

period was at the tail end of the federal and state hospital antitrust losing streak and before the

successful hospital merger challenges of recent years. In addition, North Carolina introduced a

hospital Certificate of Public Advantage (COPA) regulatory program in 1995 that gave merging

hospitals participating in the program antitrust immunity. Only one pair of merging hospitals

participated in North Carolina’s COPA program,4but the option to participate, coupled with

recent court rulings favoring hospital mergers, likely contributed to an environment in which

competing hospitals felt safe to merge with less risk of an antitrust challenge. The Missouri

hospital mergers in the sample include one merger challenged by the FTC, but allowed by the

courts. We also address the bias from antitrust enforcement by directly measuring post-merger

variable cost changes, as well as price changes, and focusing on the mergers that were not

associated with significant efficiencies.

Analyzing hospital mergers from North Carolina and Missouri in the late 1990s and early

2000s may lessen the truncation problems caused by antitrust enforcement. However, the hospital

industry has undergone many changes since the early 2000s, potentially limiting the applicability

of findings from that period. Some have argued that methods used in hospital merger review

and enforcement should evolve and account for the changes in healthcare delivery and finance

that have occurred since the passage of the Affordable Care Act (Guerin-Calvert, Maki, and

Vladeck, 2015). To address these concerns and test the accuracy of hospital merger screens in

this potentially new regime, our sample of hospital mergers also includes 16 recent transactions

from 2007–2012.

The comparison of actual post-merger price changes against the pre-merger predictions

of the screening tools reveals that, apart from merger simulation, the new screening tools (in

particular, WTP and UPP) are more accurate than traditional concentration measures at flagging

potentially anticompetitivehospital mergers for fur ther review. However,the relationship between

the new screening tools and the post-merger price changes is not precise or robust to alternate

price change measurements, so care should be taken when using the tools to screen mergers for

further investigation. Merger simulation performs poorly, but this may be due to the limited data

available to calibrate the simulation in the initial investigation. Among the traditional concen-

tration measures, those that employ market shares based on flexible geographic boundaries are

more accurate at predicting post-merger price changes than concentration measures based on

fixed boundaries.

The article is arranged as follows. Section 2 reviews hospital antitrust enforcement over

the past 20 years, and the hospital competition literature that developed alongside it. Section 3

describes the evaluated screening tools in detail. Section 4 describes the data, the criteria for

merger selection, price measurement and price change estimation, and the construction of the

screening tools. Section 5 compares the screening tools to the post-merger price changes, and

Section 6 concludes.

4In December 1995, Memorial Mission Hospital and St. Joseph’sHospital—the only two short-term, general acute

care hospitals in Asheville, NC—entered into a joint management agreement to form Mission Health and simultaneously

entered into a COPAagreement with the state of Nor th Carolina, granting the merger antitrust immunity in exchange for

regulation of Mission Health by the state. This merger is not included in the sample of hospital mergers analyzed in this

article.

C

The RAND Corporation 2017.