What’s a nice boy from MIT know about narcotics?

During his doctoral student days at the Massachusetts Institute of Technology, Jonathan Caulkins’s professors told him that mathematical modeling methods used in operations research could solve any problem. So he whittled humanity’s 10 biggest challenges down to the one that was both quantifiable and not yet addressed by engineers: drug addiction. He then spent a summer rolling with the police in Hartford, Conn., so he could meet the city’s drug dealers in person and grill them about their business.

Some two decades later, Caulkins, 41, now a professor of operations research and public policy at Carnegie Mellon University, in Pittsburgh, is among a handful of engineers, statisticians, and computer scientists using math modeling to predict, treat, and contain drug addiction. Armed with degrees in computer science, public policy, and electrical and systems engineering from MIT and Washington University, in St. Louis, Caulkins has carved an award-winning career out of analyzing the best combinations of policies needed to tackle different types and stages of drug epidemics.

“It’s common to use engineering to address public policy in industries with strong technological components, like telecommunications and transportation,” he says. “But drug-, crime-, and delinquency-prevention issues have traditionally been addressed through classic social science methods, which tend to focus on comparing static situations and pay less attention to changes over time. Engineering incorporates more dynamic systems analysis that’s less familiar to social scientists.”

Caulkins’s unorthodox approach has yielded some provocative findings, the main one being that long prison sentences don’t curtail drug use and often create more serious problems, such as HIV infection, unemployment, and families breaking up. As a result, he often finds himself walking a fine line between factions.

“In trying to be objective, sometimes I find something different than the conventional wisdom—which rocks boats because it forces people to rethink things,” he says. “Innovative programs are costly in the political sense. It’s easier to say, ‘Let’s get tough and lock them up.’ Drug addiction and markets are more complicated—they intersect with health care, foreign policy, education, and environmental issues—which takes more than a sound bite to explain to voters.”

Caulkins’s research is bound to have increasingly timely implications as the United States gears up for next year’s presidential election. His modeling has shown that drug use responds to such market factors as price and that a drug’s evolution from introduction to maturity mirrors the pattern of a medical epidemic. His findings, he says, reflect the need for a fluid counterbalance, rather than simply increasing incarceration, which has had limited success.

For example, the beginning of the cocaine surge in the 1980s prompted a spike in drug-related homicides. But as a mature epidemic, cocaine is now primarily used by an older, more drug-dependent, less violent population. Each stage requires different containment methods. Supply-control interventions, like a police crackdown on drug dealers, might be more effective during a drug’s early explosive use; more treatment and less reliance on incarceration might be called for later, when the drug is used mainly by addicts. Caulkins’s research ascertains the most effective mix of interventions during the course of an epidemic.

Modeling relies on computer programs to track changes in a system over time based on the introduction of different variables. Since 1990, Caulkins has been evaluating how singular factors—such as border control, school-based prevention, inpatient/outpatient treatment, price, and incarceration—affect drug use among recreational, criminal, HIV-infected, and addicted users. He then layers the findings to track increasingly complex situations. (Other modeling studies are looking at treatment allocations, how to predict which populations are most likely to become addicts, and how social policies are combating addiction-spawned diseases such as HIV and hepatitis C.)

Even so, modeling epidemics has its limitations: it works on a macro level, while individual behavior is much harder to predict. “It’s akin to needing quantum mechanics to understand a single atom, and Newtonian physics to understand aggregation, because the law of larger numbers takes effect,” Caulkins says.

Still, he says, “modeling works when you try to figure out the right policy for a country and have drug-control choices that are enormously diverse.”

Colleagues regard Caulkins’s biggest contribution as his challenging—and then proving wrong—many assumptions. For example, he discovered that higher drug prices do not necessarily increase crime, as previously thought.

“What distinguishes Caulkins from most modelers is that he takes the time to understand the topic before throwing equations at it,” says Mark Kleiman, a professor of public policy at the University of California at Los Angeles, who was one of Caulkins’s thesis advisors and has been “a fan” ever since. “He doesn’t assume, as lots of modelers do, that merely having abstract modeling capacity is as substantive as actually understanding the real-world problem.”

Slowly, Caulkins’s ideas have been gaining a foothold outside the United States, particularly in Australia, and in some regional U.S. arenas. “State and local governments are more concerned about cost-effectiveness, because they have to balance budgets; the federal government does not,” he notes.

But changing minds takes time. “Every time we release a study, there’s always some aspect that ruffles feathers,” he says, laughing. "We’ve successfully offended every faction, from law enforcement to the treatment community.

“I like it when there’s a little controversy, because it means we’re challenging the status quo.”