RATIONAL DISINHIBITION AND EXTERNALITIES IN PREVENTION

• Date: 01 November 2019
• DOI: http://doi.org/10.1111/iere.12402
• Published date: 01 November 2019
• Author: Flavio Toxvaerd

INTERNATIONAL ECONOMIC REVIEW

Vol. 60, No. 4, November 2019 DOI: 10.1111/iere.12402

RATIONAL DISINHIBITION AND EXTERNALITIES IN PREVENTION∗

BYFLAVIO TOXVAERD1

University of Cambridge, U.K.

This article studies a model of disease propagation in which rational and forward-looking individuals can

control their exposure to infection by engaging in costly preventive behavior. Equilibrium outcomes under

decentralized decision making are characterized and contrasted to the outcomes chosen by a social planner. In

general, individuals overexpose themselves to infection, leading to suboptimally high disease prevalence. The

model is applied to study the welfare effects of preexposure prophylaxis, which reduces transmission between

serodiscordant couples and causes disinhibition. It is shown that a decrease in the induced infection risks increases

disease prevalence and can lead to decreases in overall welfare.

1. INTRODUCTION

Historically, infectious diseases have posed a major obstacle to human well-being, causing

both morbidity and mortality to human populations. Despite significant progress in understand-

ing diseases and developing medicines to combat their spread, epidemic outbreaks continue to

be an issue of first-order importance for public policy. The recent introduction of preexposure

prophylaxis, such as the drug Truvada, has been lauded as a “game changer.” Such drugs de-

crease the probability that infection is transmitted during an unprotected encounter and have

raised expectations that it may radically alter the calculus of HIV control.2

Regrettably, there is a considerable lack of clarity about both the likely effects and the

desirability of such medical innovations, and the issue of so-called risk compensation remains

controversial. Risk compensation, or disinhibition, is said to occur if a decision maker responds

to decreased risk by engaging in more risky behavior.

In this article, I consider second-best interventions such as preexposure prophylaxis and their

relation to disinhibition in a susceptible-infected-susceptible (SIS) type model. My analysis con-

tributes to this important policy debate by addressing the following two fundamental questions:

Does the introduction of preexposure prophylaxis to the uninfected (i) cause disinhibition and/or

does the introduction of preexposure prophylaxis to the uninfected (ii) decrease social welfare?

To preview the answer to the first question, my analysis confirms that a permanent decrease

in the infectiousness of the disease will prompt an increase in exposure and in steady-state

disease prevalence. The reason is that such a change alters the trade-offs faced by decision

makers in favor of increased exposure. Although the decrease in infectiousness decreases the

rate of transmission per exposure, the exposure itself increases so much that the net effect is

∗Manuscript received January 2017; revised January 2019.

1I am grateful to the editor, Rakesh Vohra, an associate editor, and two anonymous referees for very useful comments

and suggestions on how to improve the manuscript. I also thank David Bardey, Jess Benhabib, Frederick Chen, Drew

Fudenberg, Pierre-Yves Geoffard, Mark Gersovitz, Chryssi Giannitsarou, Christopher Harris, David Levine, Michael

Mandler, David Myatt, Tomas Philipson, Joshua Ross, Marzena Rostek, Robert Rowthorn, Larry Samuelson, Jakub

Steiner, Habtu Weldegebriel, and Peyton Young for constructive conversations and feedback. Finally, I also thank

seminar participants at Columbia University, and Wake Forest University and participants at the 2009 Annual Meetings

of the American Economic Association, San Francisco, the 2011 North American Winter Meetings of the Econometric

Society, Denver, and at the Journ´

ees Louis-Andr´

eG

´

erard-Varet, Marseille (2010). Please address correspondence to:

Flavio Toxvaerd, Faculty of Economics, University of Cambridge, U.K. E-mail: fmot2@cam.ac.uk.

2See, for example, The Guardian, August 2, 2016: “Judge to Rule on NHS Funding for ‘Game-Changing’ HIV

Treatment” and BBC, February 25, 2015: “Analysis: An HIV ‘Game Changer’?”

1737

C

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

and Economic Research Association

1738 TOXVAERD

to increase disease incidence. This outcome is not pathological and holds both for a utilitarian

social planner and for self-interested individuals.

To preview the answer to the second question, my analysis shows that a decrease in the

infectiousness of the disease may indeed lead to a decrease in social welfare under certain

conditions. The reason is that individuals make different choices than those preferred by a

social planner. There are two sources for this discrepancy. First, there is a pure externality

effect that arises because individuals do not internalize the benefits to others that flow from

the individual protecting itself. As a consequence, aggregate equilibrium protection is too low,

thus causing higher future disease prevalence. This means that the equilibrium future path of

infection faced by the individual is higher than the path preferred (and indeed chosen) by a

social planner. But since each individual takes the path of aggregate infection as given, this

means that the individual and the planner face different future paths of infection, and this

causes a second discrepancy between the optimal choices and the equilibrium choices. This

further drives a wedge between optimal and equilibrium infection paths. This second effect

stems from the fact that each (small) individual takes aggregate disease prevalence as given.

These effects mean that the introduction of preexposure prophylaxis may well lead to a

decrease in social welfare. That this possibility can materialize is verified numerically.

These findings point to two central policy conclusions. First, the customary focus on disinhibi-

tion in the policy debate is misplaced. When one decreases the infectiousness of a disease, it may

be perfectly desirable that individuals increase exposure and hence cause disease prevalence

to rise. After all, to the extent that individuals derive utility from such exposure, it is right that

policy should take that into account when trading off the costs and benefits of such exposure.

Second, the presence of external effects may lead policy interventions to have undesirable con-

sequences. In particular, a well-intentioned second-best policy to reduce infections may in fact

exacerbate the problem of overexposure and lead to socially inferior outcomes.

The present work contributes to a growing literature that studies the effects and desirability of

permanently decreasing the infectiousness of diseases through medical innovations. Lakdawalla

et al. (2006) consider the effects of simultaneously introducing antiretroviral therapy to suscep-

tible and infected individuals in a susceptible-infected (SI) type model. They note that there are

several confounding effects. Ceteris paribus, the susceptible individuals may benefit, because

the introduction of antiretrovirals effectively reduces the probability of transmission per risky

sexual act. On the other hand, antiretrovirals increase the survival probability of infected indi-

viduals, thereby increasing the source of infection. Unfortunately, Lakdawalla et al. (2006) do

not distinguish between equilibrium outcomes and socially optimal outcomes, and it is therefore

not a priori clear from their work that policy interventions are warranted in the first place.3Chan

et al. (2018) consider the effects of introducing antiretroviral therapy to susceptible individuals

and conclude that they are likely to benefit from reduced infectivity. Gersovitz (2010) considers

innovation in an infection context, but mainly offers results for technological improvements

that do not interact directly with individual’s decisions. The technological improvements he

considers cannot be interpreted as preexposure prophylaxis, and his analysis is therefore not

directly comparable to that in this article. Finally, Tamalas and Vohra (2018) consider the ef-

fects, if imperfect protection, in a network-theoretic model of partner choice and show that a

decrease in infectiousness can cause welfare-reducing changes in partnership formation.

Disinhibition and related offsetting effects have been noted in several different contexts. In

energy and conservation economics, the so-called rebound effect has been recognized as a real

possibility (see Gillingham et al., 2015, for a review of this literature). The rebound effect is

said to be present if an innovation that is intended to reduce the use of a resource instead

acts to increase the use of said resource. The rebound effect is also known as the Peltzman

effect, articulated in the context of safety equipment and driving (see Peltzman, 1975; Hoy and

Polborn, 2015; for a more general and abstract treatment). In traffic and network economics,

3A consequence of the analysis in the present article is that when no recovery is possible, that is, in a standard SI

model, equilibrium is socially optimal unless the population is heterogeneous.