If It Were Enabling, Where's the Evidence? Applying Moral Hazard and Risk Compensation Theories to Community Naloxone Distribution and Other “Safety Net” Public Health Interventions

Naloxone, the opioid overdose antidote, is perhaps the best tool to save the lives of people who use drugs (PWUD) (Centers for Disease Control and Prevention, 2024). Growing evidence attests to the lifesaving impact of community-based naloxone distribution programs that increase naloxone availability and, consequently, the likelihood that naloxone will be available at an overdose event (e.g., Irvine et al., 2022; Razaghizad et al., 2021). Distribution strategies prioritize PWUD because they are most likely to witness an overdose (Keane et al., 2018; Strang et al., 2000) and thus can act as “community first responders” to quickly administer naloxone and avert death (Substance Abuse and Mental Health Services Administration, 2025). Although these programs are effective at reducing overdose mortality, such efforts have been criticized for “enabling” PWUD: Common talking points include PWUD seeking “higher highs” and using recklessly when naloxone is nearby (Nelson, 2025; Saleh et al., 2021; Stoffers, 2015) or being less likely to seek treatment because of naloxone buffering against negative consequences associated with use (Bazazi et al., 2010; Gatewood et al., 2016). Such arguments reflect a belief in the presence of often conflated data and theories: moral hazard (requiring population-level evidence) and risk compensation (requiring individual-level evidence). The distinction between these characterizations is important, especially when evaluating the outcomes of any “safety net” intervention¹ that aims to improve public health.

Discussions of “moral hazard” in the context of naloxone notoriously entered public discourse when economists Doleac and Mukherjee (2018, 2022) contended that naloxone access amounted to a moral hazard in the United States by actually increasing illicit opioid use and opioid-related arrests without reducing opioid-related mortality. These publications were quoted in The Washington Post, Forbes, and other mainstream news outlets to suggest that community-based naloxone distribution programs were not only ineffective, but actually harmful for society. The concept of moral hazard is not novel but rather a longstanding economic theory that characterizes when “immoral” or “bad” behaviors are safeguarded by an implemented policy and subsequently result in negative consequences that outweigh the overall intended benefits (Baker, 1996; Stone, 2011). Primarily, the theory of moral hazard was applied to insurance systems; for example, individuals with health insurance were more likely than those without to use health care services, which allegedly shifted costs from patients to insurance companies and decreased the efficiency of health care systems overall (Dembe & Boden, 2000; The Editors of Encyclopaedia Britannica, 2024). In the economics of moral hazard, “moral” and “immoral” originally refer to the undertaking of odds; that is, “the ‘moral’ insured was honest, careful, chaste, thrifty, [hardworking], moderate in habits, [and] did not gamble [emphasis added]” (Baker, 1996, p. 249), whereas the “immoral” individual was “tempted” by their insurance and careless in avoiding risk. Moral hazard, importantly, is a population-level theory, indicating that unintended negative consequences must be observed from a large-scale perspective for moral hazard to be present. Indeed, Doleac and Mukherjee (2022) reached their conclusion of naloxone-related moral hazard by using publicly available data that comprised opioid-involved arrests at the jurisdictional level, opioid-related emergency room visits at the state level, and fatal opioid overdoses at the county level before and after states enacted legislation to broaden naloxone access.

What must be true for a moral hazard pertaining to a specific intervention implemented at a population level to be present? For a moral hazard to exist, a majority of distinct individuals must collectively behave in such a manner that suggests each individual compensates for a perceived decreased risk by increasing their risk in another way. This phenomenon is risk compensation, which refers to an individual increasingly engaging in unsafe behavior when certain safety or preventative resources are available to reduce other risks (Hogben & Liddon, 2008). For example, economist Peltzman asserted that reductions in driver deaths on highways because of new auto safety regulations, such as seatbelt requirements or established brake standards to reduce the risk and severity of injury during traffic collisions, were negated by increased numbers of pedestrian deaths and collision-related property damage stemming from people exhibiting “increased driving intensity” (Peltzman, 1975). Peltzman did not use the terms “moral hazard” or “risk compensation,” and although he technically used an analytic strategy and offered a conclusion consistent with moral hazard, this seminal article is considered the first on risk compensation theory for establishing “risk demand” as the individual-level mechanism for his finding: Peltzman hypothesized that vehicle drivers would accept a certain degree of risk based on their income, alcohol intoxication, age, and vehicle speed—before and after the implementation of automobile safety regulations—which would contribute to increases in highway death rates, pedestrian injury, and property damage (Peltzman, 1975). Although this “Peltzman effect” with seatbelts has since been methodologically criticized (Hedlund, 2000) and failed to be replicated (Carpenter & Stehr, 2008; Rock, 1993; Yu et al., 2021), the concept of risk compensation was eventually adopted in risk homeostasis theory, which posits that people broadly have a certain level of risk they are willing to maintain and adjust their behaviors accordingly (Hedlund, 2000; Wilde, 1998). Following seatbelts and driving, risk compensation theory has been applied to other safety measures and related behaviors, including helmets and unsafe bicycling (Büth et al., 2023; Hoye, 2018; Macpherson & Spinks, 2008), pre-exposure prophylaxis (PrEP) and unprotected sexual intercourse as it relates to human immunodeficiency virus (HIV) transmission (Marcus et al., 2019; Myers & Sepkowitz, 2013; Traeger et al., 2018), and most recently, naloxone and intentional overdosing—as indicated by media reports of “yo-yo-ing” and “Lazarus parties” characterized by individuals intentionally overdosing to be resuscitated by naloxone (Crabtree & Masuda, 2019). Because risk compensation is viewed to occur at an individual level, person-specific variable operationalizations are required. Individuals may complete laboratory tasks under behavioral observation (e.g., bicycling simulations) or self-report their history of behaviors (e.g., timeline followback of drug use). Indirectly, individuals might rate their agreement with beliefs or intentions regarding behaviors as a proxy for potential risk compensatory behaviors (e.g., bicycle speeding), consistent with the theory of planned behavior (Ajzen, 1991).

Moral hazard and risk compensation are sometimes referenced together or used interchangeably. Still, in general, moral hazard tends to be used by economists for large-scale topics such as insurance and environmental risk at a population level, while risk compensation tends to be used by psychologists to describe individual-level behavioral changes attributed to the presence of safety measures. However, zooming in on the purported moral hazard can readily identify what compensatory risky behaviors might occur if moral hazard truly exists. Returning to the insurance example, Croghan (2001) commented on the “moral hazard” of insurance coverage in which the benefit of lower medication costs for patients is allegedly outweighed by wasteful prescriptions and higher costs that insurers bear. This extensive shift in cost (purportedly indicative of moral hazard) could, if examined through a lens of risk compensation, be understood as a majority of unique individuals over-relying on the safety provided by their insurance and neglecting to adopt nonpharmaceutical lifestyle changes that would otherwise advance their health goals (Mathur et al., 2016). Although these theories are inherently related or even intertwined, they are not synonymous: Intuitively, one individual who exhibits risk compensation does not establish moral hazard. On the other hand, evidence of moral hazard does not necessarily mean that every individual is engaging in compensatory behaviors, but rather enough individuals are behaving in riskier ways such that ensuing harms are widely observed. Accordingly, these propositions depict the relationship between the two: For population-level moral hazard to be present, individual-level actors must collectively engage in risk compensatory behaviors, whereas the reverse is not necessarily true, as the observation of individual-level risk compensation does not immediately evince moral hazard.

Why is it important to recognize the conceptual distinction between moral hazard and risk compensation? First, this novel conceptualization bridges economics and psychology, distinct branches in the social sciences, in a meaningful and practical way that promotes the demand for multidisciplinary expertise and consultation in evaluating safety net strategies. Second, distinguishing between the two helps establish the kinds of questions that can be asked and answered with specific evidence: population-level data for moral hazard, which can be difficult to identify and attain and frustratingly imprecise, or individual-level data for risk compensation, which are limited by its operationalization (self-report vs. observed, behaviors vs. intentions vs. beliefs). Third and relatedly, this insight proposes a “burden of proof ” (really, a “burden of risk compensation proof ”) before considering a question of moral hazard. Last, this conceptualization makes it clear: The extent to which moral hazard and/or risk compensation are evident is ultimately an empirical question and should be recognized as such.

With this understanding, we return to the original question: Is greater public availability of naloxone a moral hazard? The first step in observing moral hazard is discovering evidence of individual-level compensatory risky behaviors by PWUD that are protected by naloxone access. Because the primary intention of naloxone is to reduce the fatality risk of opioid overdoses, we must look for other “risky” behaviors that are reasonably pertinent to naloxone. However, naloxone provision is actually associated with no change in experiencing a nonfatal opioid overdose (Tse et al., 2022) and with a decrease in opioid-related emergency department visits (Coffin et al., 2016). Also, naloxone access may contribute to either no change or a decrease in opioid use frequency (Doe-Simkins et al., 2014; Jones et al., 2017; Seal et al., 2005; Tse et al., 2022; Wagner et al., 2010) and no change in injection drug use frequency (Colledge-Frisby et al., 2023). Although some PWUD share anecdotes of witnessing riskier drug use and diminished concerns when naloxone was around compared with when it was not (Heavey et al., 2018; Kline et al., 2020), the preponderance of existing evidence suggests that PWUD do not behave in riskier ways (e.g., using more drugs, intentionally overdosing) when provided with naloxone.

For Doleac and Mukherjee's (2022) claim of naloxone-related moral hazard to be present, we must also observe population-level negative consequences that can be plausibly attributed to naloxone access. However, their selection of population-level variables (“interest in drug treatment,” “demand for illicit opioids,” opioid-related hospital visits and thefts) and proxy data was inappropriate, and their broader study methods were notably flawed (Frank et al., 2018; Greene, 2018; Szalavitz, 2024). Recently, a rigorous effort failed to replicate their results of moral hazard (Alexeev, 2025), which is consistent with the convergence of the literature presented thus far. Furthermore, Doleac and Mukherjee lacked consideration of how PWUD feel about naloxone: Many qualitative studies report that PWUD find the experience of being revived by naloxone to be extremely noxious, painful, and something to be avoided (Lai et al., 2021; Neale & Strang, 2015; Russell et al., 2024; Rzasa Lynn & Galinkin, 2018). These accounts contradict the notion that anyone would willingly seek out opportunities to be administered naloxone, which is implied by endorsement of naloxone-related risk compensation and moral hazard beliefs.

Although multiple published findings demonstrate the lack of naloxone-related risk compensation behavior and moral hazard evidence, such studies are matched or even outnumbered by studies examining perceptions of naloxone-related risk compensation (e.g., Winograd et al., 2020). Indeed, endorsement of naloxone-related risk compensation beliefs has been identified among health professionals who provide services to PWUD, including physicians and other clinical substance use treatment providers (Bessen et al., 2019; Gatewood et al., 2016; Green et al., 2013; Winograd et al., 2017, 2020, 2023), emergency medical service and fire personnel who arrive at overdose calls (Bessen et al., 2019; Kruis et al., 2022; Tobin et al., 2005; Winograd et al., 2020, 2024), and pharmacists who stock and dispense naloxone (Bakhireva et al., 2018; Sisson et al., 2019; Thakur et al., 2020). These beliefs have also been observed in PWUD themselves (Banks et al., 2025; Bessen et al., 2019; Heavey et al., 2018; Kline et al., 2020). Naloxone-related risk compensation beliefs are also prevalent among the U.S. general public (Agley et al., 2022; Calabrese & Bell, 2019; Hohmann et al., 2022; Pyra et al., 2022; Schleinitz et al., 2018), which is problematic because voters endorsing negative views toward naloxone may disapprove of specific legislation or political candidates working to increase naloxone access. Unfortunately, the lack of empirical evidence of naloxone-related risk compensation behaviors alongside the prevalent belief that naloxone-related risk compensation exists illustrates a drastic mismatch between reality and perception of naloxone's “enabling” capability.

Of course, expectations of risk compensation surround safety net interventions well beyond naloxone. For example, there was immense public worry that the availability of an oral contraceptive (i.e., “the pill”) would “promote liberal sexual ethics” and increase premarital sex (Asbell, 1995; Watkins, 1998). In addition, a systematic review revealed that about one in five parents are hesitant to have their daughters receive the human papillomavirus (HPV) vaccine, which is effective in preventing certain cancers, fearing that it would increase riskier sexual behavior (Kasting et al., 2016). This condemning pattern toward pleasurable activities like having sex and using drugs is not surprising, given that the United States was founded on puritanical principles guided by norms of fastidious, doctrinal self-discipline (HISTORY.com Editors, 2023). Such principles drive how heavily these behaviors—and the people who engage in them—are vilified and stigmatized. Considering the lack of stigma toward other behaviors that are inherently risky, like driving cars and bicycling, these less “indecent” activities do not carry the “sinful baggage” of sex and drugs. Objections toward drug use specifically reflect a particular meaning of the word “moral” that is overtly religious (e.g., Prohibition in the early 20th century) and racialized (e.g., bans on Chinese immigrant-run opium dens and “Mexico-branded marijuana/marihuana”) (Ezell et al., 2024; Thompson, 2013), and with the 1914 Harrison Narcotics Act and ensuing War on Drugs, drug use became a “crime” warranting strict surveillance and punitive responses that disproportionately impact people of color (Cohen et al., 2022).

Given this meaning of “morality,” consider Table 1, in which four behaviors and corresponding safety net strategies varying in perceived immorality are compared on two dimensions: evidence and perceptions of risk compensation. Car seatbelts are not associated with more dangerous driving (Carpenter & Stehr, 2008; Hedlund, 2000; Rock, 1993; Yu et al., 2021), and the overall public safety benefit of seatbelts (Carpenter & Stehr, 2008) precludes the belief that seatbelts “make things worse.” Similarly, bicycle helmets are not stigmatized or labeled as “enabling” more dangerous bike riding despite mixed empirical evidence suggesting that riders do engage in compensatory risk when wearing helmets (Kang et al., 2021; Phillips et al., 2011; Schleinitz et al., 2018). (Interestingly, vehicle drivers drive closer to bicyclists wearing helmets compared with those who are not [Walker & Robinson, 2019], insinuating that the mere observation [rather than direct use] of the safety resource may encourage compensatory behaviors among individuals who are not engaging in the targeted behavior at all.) Despite evidence of some degree of individual-level risk compensation, the overall public health benefit of bicycle helmets is substantiated in their reduction in the severity of head injuries (Büth et al., 2023; Hoye, 2018; Macpherson & Spinks, 2008), which fails to illustrate moral hazard. As such, and because of the imperceptible public stigma surrounding helmets, most U.S. localities have mandatory bicycle helmet legislation (Bachynski & Bateman-House, 2020). In comparison, the evidence of PrEP-related compensatory behaviors is similarly inconsistent regarding number of sexual partners, unprotected sexual intercourse, and rates of sexually transmitted infections (Milam et al., 2019; Murchu et al., 2022; Powell et al., 2019; Traeger et al., 2018), signaling some occurrence of PrEP-related risk compensation. However, the well-established effectiveness of PrEP in preventing HIV is overshadowed by prevalent, exaggerated concerns of riskier sexual behavior held by physicians (Blumenthal & Haubrich, 2014; Calabrese et al., 2017; Marcus et al., 2019; Myers & Sepkowitz, 2013; Norwood & Zuñiga, 2023), and this imagined “sexual moral hazard” likely drives a very low PrEP prescription rate (Karris et al., 2014; Norwood & Zuñiga, 2023). Last, for naloxone, although there is a lack of evidence of risk compensation, the perception of its existence persists: Recent surveys report that 10%–25% of American adults believe naloxone is enabling and increased access causes increased drug use (Hohmann et al., 2022; Schlosser et al., 2022), which may explain why only 58% of Americans support making naloxone available without a prescription (Pyra et al., 2022).

Table 1.

Reality versus perception of risk compensatory behaviors occurring related to three “safety net” interventions

“Safety net” intervention	Reality of risk compensation	Perception of risk compensation
Seatbelt	Not occurring ×	Not occurring ×
Helmet	Occurring ✓	Not occurring ×
Pre-exposure prophylaxis (PrEP)	Mixed ?	Mixed ?
Naloxone	Not occurring ×	Occurring ✓

Discussing the benefits of safety net interventions such as seatbelts, helmets, PrEP, and naloxone is incredibly relevant in today's policy environment. Indeed, a preliminary federal budget document reports that the U.S. Department of Health and Human Services aims to cut significant funding for naloxone and overdose response (Bedigan, 2025), which worries many well-known scholars and advocates (Boness, 2025; Dingman & Mackesy-Amiti, 2025; Frederique, 2025). Also, the U.S. Supreme Court considered arguments for Medicare coverage of preventative services, including PrEP prescription (Barrón-López & Buhre, 2025; Munz, 2025) earlier this year, with petitioners arguing that PrEP “facilitate[s] and encourage[s] homosexual behavior, prostitution, sexual promiscuity, and intravenous drug use” (Barrón-López & Buhre, 2025). Pervasive beliefs that these safety nets are enabling, that “the cure is worse than the disease,” and that “tough love” is needed for individuals to face the consequences of their “immoral” behaviors contrasts a “safety net” attitude that rescues against a “sink or swim” approach that leaves the vulnerable to drown.

Conclusions

As outlined in this article, the population-level phenomenon of moral hazard and the individual-level phenomenon of risk compensation should be conceptualized and evaluated separately. When doing so, one should understand that the latter must be evident on an overwhelming scale for there to be evidence of the former and, further, that belief in either phenomenon is associated with the societal stigma attached to the specific behavior at issue. The unsupported belief in “enabling” effects of interventions is perhaps most publicized for naloxone but has been applied to other drug-related harm reduction strategies, such as sterile syringe programs, drug checking, and overdose prevention centers—not because there is any evidence that such strategies increase risk, but because illicit drug use is a heavily stigmatized behavior. Looking ahead, the evaluation and messaging of safety net strategies designed to reduce the risk of certain behaviors should focus on the empirical evidence of those strategies, taking care not to play into unfounded beliefs driven by stigma and moral judgment.

Funding

No direct support was provided for this manuscript. Brandon Park's and Rachel P. Winograd's work was supported by the Substance Abuse and Mental Health Services Administration (H79TI087738) and the National Institute on Drug Abuse (R15DA059765).