DO HEALTH INSURERS CONTRACT THE BEST PROVIDERS? PROVIDER NETWORKS, QUALITY, AND COSTS

• Author: Christoph Schottmüller,Jan Boone
• DOI: http://doi.org/10.1111/iere.12383
• Published date: 01 August 2019
• Date: 01 August 2019

INTERNATIONAL ECONOMIC REVIEW

Vol. 60, No. 3, August 2019 DOI: 10.1111/iere.12383

DO HEALTH INSURERS CONTRACT THE BEST PROVIDERS? PROVIDER

NETWORKS, QUALITY, AND COSTS∗

BYJAN BOONE AND CHRISTOPH SCHOTTM ¨

ULLER1

University of Tilburg, The Netherlands; University of Cologne,Germany

We provide a modeling framework to analyze selective contracting in the health-care sector. Two health-

care providers differ in quality and costs. When buying health insurance, consumers observe neither provider

quality nor costs. We derive an equilibrium where health insurers signal provider quality through their choice

of provider network. Selective contracting focuses on low-cost providers. Contracting both providers signals

high quality. Market power reduces the scope for signaling, thereby leading to lower quality and inefficiency.

1. INTRODUCTION

Selective contracting in health-care markets is the practice where an insurer limits the choice

of providers who can be visited by the insured when they need treatment. Among the advantages

claimed for selective contracting is the potential to “weed out providers who would be poor

choices for plan members, for reasons of either quality or cost” (McGuire, 2011b, p. 366).

There is quite some evidence that selective contracting helps to reduce costs. When managed

care was introduced in California in the 1980s, hospital prices tended to fall (Dranove et al.,

1993). Chernew and Newhouse (2011, p. 30) conclude from a literature overview that the

“central findings from these studies were that hospital cost and revenue growth slowed markedly

following the introduction of selective contracting.” The economic rationale for this is intuitive.

Insured patients do not worry (much) about the cost of a treatment and hence do not shop

around for a low-cost provider. If an insurer allows its customers to visit any provider, at least

some of them will visit inefficient ones. The insurer can avoid this by selectively contracting the

most efficient providers while excluding inefficient providers from the network (Dranove, 2000,

pp. 72–74).

Why selective contracting should help to raise quality is not as clear compared to its effect on

costs. Indeed, unlike provider cost, the patient is directly interested in provider quality. If one

provider offers higher quality than another, an informed patient tends to choose the former.

Although quality information may not be easily available, patients can learn from others’

experiences or decide which specialist to visit based on the advice of a primary physician.

However, there is a difference here between deciding which provider to visit once you know

which treatment you need and deciding which providers to exclude at the moment you buy

insurance. We assume that consumers cannot observe provider quality at the moment they

buy insurance (Handel and Kolstad, 2015). When they need treatment, their primary physician

helps them choose the best hospital from their insurer’s network. In this framework, we ask

whether insurers can signal provider quality to consumers through the choice of their network.

∗Manuscript received April 2017; revised October 2018.

1We thank Rein Halbersma and seminar audiences at CPB Netherlands Bureau for Economic Analysis, Dutch

Ministry of Health (VWS), at the Universities of Copenhagen and Florence, the European Health Workshop in

Toulouse, the Meeting of the European Economic Association in Mannheim, and the European Meeting of the

Econometric Society in Geneva for comments. Financial support from the Dutch National Science Foundation (VICI

453.07.003) is gratefully acknowledged. Please address correspondence to: Christoph Schottm¨

uller, Department of

Economics, University of Cologne, Albertus Magnus Platz, K¨

oln, NRW 50923, Germany. Email: c.schottmueller@uni-

koeln.de

1209

C

(2019) by the Economics Department of the University of Pennsylvania and the Osaka University Institute of Social

and Economic Research Association

1210 BOONE AND SCHOTTM ¨

ULLER

For example, can excluding some providers credibly signal to consumers that the excluded

providers have lower quality than the ones in the network? Or does a consumer conclude

that the insurer excludes high-quality providers because they are more expensive, just to keep

costs down?

This debate has also reached the popular press. Pear (2014) writes “insurers say, when

they are selective, they can exclude lower-quality doctors and hospitals.” Terhune (2013) cites

Donald Crane, chief executive of the California Association of Physician Groups, asking: “We

are nervous about these narrow networks. It was all about price. But at what cost in terms of

quality?”

Evidence of the effects of selective contracting on quality is mixed (Cutler, 2004, chapter

8; Gaynor et al., 2015).2A number of authors have argued that selective contracting tends to

reduce quality. To illustrate, although selective contracting (and managed care more generally)

was seen as a success in the United States in the 1980s and early 1990s, it “deteriorated into a

zero-sum competition over cost shifting, with patients the ultimate losers as quality suffered”

(Porter and Teisberg, 2006, p. 77). Zwanziger et al. (2000) argue that it is unclear whether the

cost reductions mentioned above were due to increased efficiency or lower quality. Finally,

some have argued that financial incentives for doctors—like the threat to be excluded by the

insurer—tend to reduce quality (Stone, 1997).

Our article offers a framework to think about these issues. We introduce a model with two

providers who can have different treatment cost and quality levels. These providers bargain

with an insurer and the insurer can decide to contract both providers or only one. In the latter

case, there is an exclusive contract between the insurer and (contracted) provider; we refer to

this as selective contracting. In the former case, there are common contracts and the provider

network consists of both providers. The bargaining between insurer and providers is modeled

as in Bernheim and Whinston (1998; henceforth, BW).

In contrast to the literature, in our framework, both selective contracting and common

outcomes can be optimal/arise in equilibrium depending on nature’s draw of cost and quality

levels. In equilibrium, provider networks can be either excessively narrow or too broad. This

allows us to formalize worries by both policymakers (inefficient providers are contracted) and

the public (efficient high-quality providers are excluded). Thus, we can analyze public policies

to either stimulate or discourage selective contracting by health insurers.

We find that selective contracting signals a focus on costs, whereas common contracts signal

quality. In particular, we distinguish two types of provider cost–quality configurations. In the

insurer-critical configuration, letting the insurer choose the provider (through selective contract-

ing) leads to inefficiency: Quality is too low (from a social/efficiency point of view) because the

insurer cares more about costs than about quality. In the patient-critical configuration, letting the

patient choose the provider leads to inefficiently high costs because the patient is insured and

therefore more interested in quality than in costs. An efficient outcome requires an equilibrium

where both providers are contracted in insurer-critical configurations, whereas there is selective

contracting in patient-critical configurations. We show that with insurer competition, such an

efficient outcome is feasible. Market power, either on the insurer or the provider side, makes an

efficient outcome less likely. We extend this (static) result to dynamic efficiency in the following

sense. Suppose providers can invest to raise quality and reduce costs. With insurer competition,

there exists an equilibrium where investments are chosen efficiently. With a monopoly insurer,

there is underinvestment in quality by the providers.

There are two strands of literature related to our approach. First, there is a literature on

incentives under managed care that considers aspects of managed care that we do not focus

on. One aspect of managed care is a move away from simple fee-for-service contracts and to

2Himmelstein et al. (1999) warn that some of the early comparisons between Health Maintenance Organizations

(HMOs) using selective contracting and indemnity insurance used not-for-profit HMOs. They provide evidence that

for-profit HMOs tend to offer lower quality than not-for-profit HMOs and therefore lower quality than indemnity

insurance that does not exclude providers.

PROVIDER NETWORKS,QUALITY,AND COST 1211

allow for more elaborate supply side management. To illustrate, capitation contracts can be

used to reduce oversupply of medical services. A recent overview of physician agency can be

found in McGuire (2011a). In Ma and McGuire (2002), managed care plans ration treatments

by threatening to (partially) exclude physicians with treatment costs above the target. In our

model, there is no decision margin with respect to treatment: Patients get treated if and only if

they need it. Selective contracting (applied to drugs) is discussed in McGuire (2011b). There it

helps to reduce drug prices, but there is no quality aspect. Moreover, the framework is not rich

enough to allow for a common outcome where more than one (substitute) drug is contracted.

As BW allow for common outcomes, we follow their setup. Ho (2009) estimates how insurers

and providers share profits. In contrast to our model, consumers are assumed to know provider

quality in her setup (leaving aside the idiosyncratic error term) and therefore the focus of our

article—signaling provider quality through network choice—does not play a role.

Although the main model focuses on vertically differentiated providers (differing in cost and

quality), Capps et al. (2003) consider horizontal differentiation. They find that in equilibrium,

all providers are contracted, whereas in our model, insurers tend to exclude high-cost providers.

Gal-Or (1997) also analyzes selective contracting in the context of horizontal differentiation

between both providers and insurers. In terms of cost and quality, providers are symmetric

in her setup. The extent of differentiation between providers compared to insurers determines

whether there is selective contracting in equilibrium. Consumers prefer the selective contracting

outcome, as it leads to lower prices. With vertical provider differentiation (in both cost and qual-

ity), we show that the realization of these cost and quality levels determines whether selective

contracting is optimal or not. In an extension, we consider horizontal provider differentiation to

argue that such differentiation tends to reduce provider competition and therefore efficiency.

If we interpret selective contracting as a way to intensify (payer driven) provider competition,

our results are in line with Gaynor (2006). He finds that more intense provider competition tends

to raise quality if treatment prices are regulated. Indeed, in our model with regulated provider

prices, an insurer contracts exclusively with the high-quality provider. As prices are regulated,

the insured do not need to worry that the cheapest (low-quality) provider is contracted.

Finally, we assume that all insured are the same. Hence, there is no effect of network size

on the type of customers buying insurance. Bardey and Rochet (2010) analyze a managed care

organization as a two-sided platform: attracting both providers and customers buying insurance.

The size of the network is then determined by the following trade-off. On the one hand, there

is a demand effect: Consumers tend to prefer broader networks. On the other hand, there is an

adverse selection effect: Broad networks are particularly attractive for high-risk types. Narrow

networks are more profitable if the latter effect dominates the former.

Several empirical papers have recently analyzed the effect of narrow networks on health-

care costs and welfare; see Ho and Lee (2016). This renewed interest is partially due to the

possibility to use narrow networks as a cost control instrument in plans offered through the

health insurance exchanges created by the Affordable Care Act. Dafny et al. (2017) find that

plans with narrow networks have indeed lower costs and prices. This is in line with Gruber

and McKnight (2016) and Atwood and Lo Sasso (2016), who attribute lower costs of narrow

networks to several factors, including a reduction in the quantity of care, a focus on low-cost

providers, and self-selection of healthy individuals into narrow network plans.

Summarizing, narrow networks have many effects. Our model focuses mainly on using net-

work size as a signal to steer consumers to efficient providers. Our model also touches upon the

effect narrow networks have on the bargaining power of an insurer. We abstract from selection

issues and effects of horizontal differentiation that have been analyzed by earlier papers. We

briefly discuss how these effects interact with our analysis in Subsection 6.3 and Section 7.

The second strand is the industrial organization literature on exclusive dealing. Rey and Tirole

(2007) give an overview of different approaches toward exclusive contracting. As mentioned,

we follow the BW approach to model the bargaining between providers and insurer, where

providers offer simultaneously both exclusive and common contracts and the insurer decides

which to accept. This allows us to immediately capture a result on contracted prices by Cutler