The OxyContin epidemic had large demographic effects on communities in the United States. In new research, Carolina Arteaga, Victoria Barone, and Stephen Claassen find that Purdue Pharma’s marketing strategy targeted specific areas, causing college-educated residents to flee and increasing fertility rates among the most affected populations.
In 1996, Purdue Pharma launched OxyContin, a potent and addictive narcotic, with a deliberate commercial strategy. OxyContin would enter through the cancer-pain market, where physicians were already accustomed to prescribing opioids, and from there expand into the far larger market for chronic, non-cancer pain. The campaign that followed was aggressive and, as courts have since established, deceptive and criminal: Purdue understated the drug’s addiction risk and has twice pleaded guilty to federal criminal charges over its marketing of OxyContin. But its marketing strategy worked. Prescribing highly addictive narcotics became common practice for moderate and chronic pain, reshaping pain management across the United States. Three decades later, the imprint of that strategy is still visible in America.
The opioid epidemic is usually studied as a mortality crisis, and for good reason: more than 800,000 Americans have died of overdoses since 1999. In a new paper with Victoria Barone and Stephen Claassen, we document a different and less examined consequence. The epidemic reshaped the demographic map of the country. It changed where Americans lived, drawing college-educated residents out of the hardest-hit places while raising fertility rates among those who stayed.
Isolating the epidemic’s effects requires separating it from everything else reshaping struggling American towns over the same decades: declining manufacturing, automation, and import competition from China and Mexico. Purdue’s own strategy provides a way to do so. Internal records unsealed in litigation show that the company first concentrated OxyContin’s marketing on areas with large populations of patients taking cancer-pain medicine, where its older opioid, MS Contin, had already accustomed physicians to prescribing narcotics. MS Contin’s patent was expiring, and Purdue wanted a new, patented drug to keep earning royalties in that market. Cancer was the entry point, not the goal: from there Purdue planned to move OxyContin into the much larger market for chronic, non-cancer pain. As a result, communities with higher cancer rates in the mid-1990s, just as OxyContin launched, received disproportionately heavy marketing and, in turn, far more prescription opioids. From cancer patients, opioids then spread to non-cancer patients through the physicians they shared.
Behind that geographic pattern was an extraordinary marketing operation. Purdue fielded one of the industry’s largest and highest-paid sales forces, with representatives’ bonuses tied to the volume they sold, and it tracked individual physicians’ prescribing in order to target the heaviest prescribers. Those incentives operated against weak and slow public oversight. The Food and Drug Administration approved OxyContin’s original label, which asserted that its delayed-release design was believed to reduce its abuse potential. Federal prosecutors and physicians warned of widespread abuse as early as 2001, yet prescribing continued to climb with few restrictions. The principal regulatory responses arrived only after prescribing had nearly peaked in 2010, and oversight of pharmaceutical marketing remains limited today.
That initial targeting is why we can use a community’s cancer-mortality rate in the mid-1990s to gauge how hard the epidemic would later hit it. The measure works for a simple reason: it was fixed before the epidemic began, and it tells us nothing about where a community was otherwise headed. In our companion work, we show that areas with higher and lower cancer mortality in the mid-1990s had similar rates of opioid-related deaths, fertility, and reliance on safety-net programs that moved in parallel before 1996, and began to diverge only as the epidemic unfolded. The relationship is strong: communities with one standard deviation higher cancer mortality in the mid-1990s received about 65% more opioid doses per capita by 2012, and saw drug-related deaths run 46% higher by 2017.
The consequences for local populations were substantial. Places more exposed to the epidemic grew significantly more slowly than the rest of the country. By 2020, a one-standard-deviation increase in exposure had reduced working-age population growth by 2.4 percentage points. By comparison, the average American commuting zone (a cluster of counties that forms a local labor market) grew about 6% between 2000 and 2020. The hardest-hit areas therefore grew roughly 40% more slowly than they otherwise would have.
This decline was driven almost entirely by people moving out, and it was selective. The residents most likely to leave were the college-educated, who departed at more than twice the rate of their neighbors without a college degree. That runs against what economists usually find empirically. The major regional shocks of recent decades—the loss of manufacturing to import competition, automation, plant closures—tend to generate surprisingly little migration; workers often stay even as local labor markets deteriorate.
Why did the opioid epidemic move people when trade shocks largely did not? Because it degraded the everyday livability of places. In the most exposed communities, crime rose: a one-standard-deviation increase in exposure raised crime rates by 8.3%, and emergency and public-health services came under strain. For a sense of scale, naloxone—a drug that reverses overdoses—was administered 215,906 times by emergency services nationwide in 2019, 36% of them in public spaces, and drug-related calls made up 7.2% of all New York City emergency medical services dispatches, more than 110,000 calls, in 2018. Residents of communities highly exposed to the epidemic were nearly twice as likely to report worrying daily about neighborhood safety (10% versus 5%). The deterioration was capitalized into housing markets: by 2020, a one-standard-deviation increase in exposure had lowered local home values by nearly 8% and rents by 2.5%. This is what economists call a negative amenity shock: a decline in the quality of a place. Such a decline affects everyone who lives there, but the residents able to act on it are those with the most options and the strongest preference for local quality of life. The college-educated fit that description, and they responded by leaving.
Where did those who left go? Not to new destinations. Migrants from high-exposure communities largely followed the same routes their neighbors had long taken, toward familiar nearby cities and established destinations, but in greater numbers. The epidemic intensified existing migration flows rather than redirecting them.
Among those who stayed, we find a second demographic response: fertility rates rose. A one-standard-deviation increase in exposure raised local birth rates by about 5.5% by 2018, with the increase concentrated among women without a college education, particularly in their late twenties. This was not simply an artifact of migration: the departure of college-educated women, who tend to have fewer children, did lower their share of the population, but birth rates rose even among the same women who stayed. Two forces likely drove this. The epidemic’s employment losses fell disproportionately on women without a college degree. In the most exposed communities, women’s employment rates dropped while those of men and college graduates held steady, lowering their opportunity cost of childbearing. Second, declining housing costs made starting a family more affordable. The effect was large enough to matter in the aggregate: absent the fertility response, the population decline in exposed areas would have been roughly 40% steeper by 2020.
It is worth being precise about the role of mortality. Overdose deaths did rise sharply, and that increase fell mostly on Americans without a college degree. But the rise in overdose deaths was not proportionally large enough to account for the population changes we document in these communities. The demographic legacy of the opioid epidemic is, above all, a story of who moved and who gave birth.
The epidemic has not ended. It has changed form, as fentanyl and other synthetic opioids continue to harm many of the same communities ravaged by OxyContin. The demographic effects of these drugs will keep unfolding well beyond the epidemic itself, as migration out of these communities and shifts in fertility rates continue to reshape these places for years to come. These demographic changes matter because they shape whether communities grow, attract new investment, and have the resources to recover. A commercial strategy devised in the mid-1990s set these forces in motion, and reversing them will require far more deliberate public effort than the decisions that started them.
Authors’ Disclosures: The authors report no conflicts of interest. You can read our disclosure policy here.
Articles represent the opinions of their writers, not necessarily those of the University of Chicago, the Booth School of Business, or its faculty.
Subscribe here for ProMarket’s weekly newsletter, Special Interest, to stay up to date on ProMarket’s coverage of the political economy and other content from the Stigler Center.
