"In the early 1990s only one in nine past-month [marijuana] users reported using daily or near-daily," Davenport and Caulkins write. "Now it is fully one in three. Daily or near-daily users now account for over two-thirds of self-reported days of use (68%)."
These usage patterns are similar to what's seen among tobacco users. "What’s going on here is that over the last 20 years marijuana went from being used like alcohol to being used more like tobacco, in the sense of lots of people using it every day," Caulkins said in an email.
Adults with less than a high school education accounted for 19 percent of all marijuana use in 2012 and 2013 (compared to 13 percent of the total adult population), according to the survey. This is similar to their 20 percent share of all cigarette use, but considerably higher than their 8 percent share of all alcohol use.
Similarly, Americans of all ages with a household income of less than $20,000 accounted for 29 percent of all marijuana use and 27 percent of all cigarette use, compared to only 13 percent of all alcohol use and 19 percent of the total adult population.
The concentration of use among poorer households means that many marijuana users are spending a high proportion of their income on their marijuana habit. Users who spend fully one quarter of their income on weed account for 15 percent of all marijuana use.
One interesting finding is that over the past 10 years as many states have liberalized their marijuana policies, marijuana arrests are down while marijuana purchases are up. This means that the risk of getting arrested for marijuana use has fallen sharply since 2002. That year, there was one marijuana arrest for every 550 marijuana purchases, according to Davenport and Caulkins. By 2013, there was one marijuana arrest for every 1,090 purchases.
"The criminal risk per marijuana transaction has fallen by half," they conclude. Much of that risk is still born by non-white marijuana users.
Davenport and Caulkins stress that since the study was conducted over a period preceding the opening of recreational marijuana markets in Colorado and Washington, it doesn't offer any evidence on the merits or lack thereof of legalization.
"Our results can in no way be interpreted as evidence toward the successes or failures of marijuana legalization or even medical marijuana laws," they write.
However, they say their research presents a number of things to consider as states like California, Arizona and Maine vote on marijuana legalization this fall.
"Most people who have used marijuana in the past year are in full control of their use, and are generally happy with that use," Caulkins said in an email. But, "consumption is highly concentrated among the smaller number of daily & near-daily users, and they tend to be less educated, less affluent, and less in control of their use."
The median marijuana user, in other words, may be someone who indulges periodically but generally doesn't consume a lot of it. However, most of the marijuana consumed in the U.S. isn't consumed by the median marijuana user, but rather by the very heavy users who smoke daily or more.
"There is a sharp contrast between what policy is best for the typical user versus what is best for the people who consume most of the marijuana," Caulkins said. Legalization may not change much in the life of the typical marijuana user. But heavy users who overindulge may find it even easier to do so when marijuana is legal and cheaper to buy.
The findings support the argument for legalization measures to be accompanied by public health protections -- like treatment programs and public awareness campaigns educating people about the risks of overuse.