Yet that's started to change in recent years. Because of the recession, caseloads have grown from 26 million in 2007 to 46 million four years later. SNAP expenditures have increased from $35 billion in 2007 to $80 billion in 2012. About two-thirds of this growth is due to the increase in people eligible for and taking SNAP benefits. Much of the remaining increase is associated with the boost in benefits provided by the 2009 stimulus, although this increase will end in November.
In response, House leaders are now proposing to cut $40 billion from SNAP over time. They also want to allow states to drug-test SNAP applicants. This additional barrier will deter some people from pursuing benefits.
What’s particularly strange about the drug-testing campaign is that if you’re trying to find people with substance-use disorders, your local sports bar, community college, or hospital ER would provide a more target-rich environment. Given the high rates of injury among intoxicated young adults, such efforts would arguably be a wiser use of public funds. Other measures such as increased alcohol taxes would also be valuable.
The drug testing of SNAP recipients is yet another ideological sideshow that disfigures substance-abuse policy. It falsely implies that substance use disorders are a widespread cause of welfare dependence. It also implies, again falsely, that these disorders are highly concentrated among recipients of public aid.
Using 2011 data from the National Survey of Drug Use and Health (NSDUH), we looked at the behaviors and circumstances of adults ages 18-64 whose households received SNAP.[i]. We examined whether respondents had used some illicit substance during the previous month or year. We then looked at whether they met screening criteria for abuse or dependence on alcohol or illicit drugs. These are the people who would be referred for treatment by mandatory drug testing.
The basic pattern is shown in Figure 1, which compares SNAP recipients ages 18-64 (the blue bars) with non-recipients (red bars) on various measures of substance use and actual use disorders. The green bars then show the additional risk associated with SNAP receipt, adjusting for gender, age, education, race/ethnicity, marital status, and the number of minor children in the home.[ii] Because SNAP recipients are poorer, less-educated, and younger than non-recipients, the adjusted risk associated with SNAP receipt is noticeably smaller than the unadjusted differences on virtually every measure.
Sure enough, SNAP recipients are somewhat more likely than others to use or misuse illicit substances. About 24 percent of SNAP recipients and 16 percent of those who don’t get SNAP have used at least one illicit substance in the past year. This drops to 13.1 percent for SNAP recipients and 7.5 percent for non-recipients if one excludes marijuana.
Note, however, that the actual prevalence of illicit substance use disorders remains quite low—only about 5.3 percent among SNAP recipients who are only about 1.7 percentage-points more likely to have such disorders than comparable non-recipients.
And if one excludes marijuana, then abuse or dependence of other illicit substances is rare within the SNAP population. By far the most common substance use disorders among SNAP recipients (and among the general population) arise from alcohol use—behaviors generally left undetected by drug-testing.
On every measure we examine, SNAP recipients are only slightly more likely than non-recipients to display substance use disorders. Yet the absolute risks associated with SNAP receipt are quite small. And some obvious socio-demographic subgroups display much higher prevalence of substance use disorders than SNAP recipients do.
As shown in Figure 2, simple youth is a much better predictor than SNAP receipt. Young adults of both genders show markedly higher prevalence of substance use disorders than does the population of SNAP recipients. In particular, among all non-Hispanic white men ages18-24, 24.7 percent meet the criteria for substance use disorders. That’s more than twice the rate among SNAP recipients (12.0 percent). Young white men are also much more likely to have alcohol disorders (19.5 percent vs. 9.1 percent) and marijuana disorders (7.8 percent vs. 2.8 percent) than are SNAP recipients.
The millions of people who receive SNAP benefits rank low on any high-priority list for special screening for substance use disorders. Not a lot of thought has gone into what we’d do when SNAP applicants test positive, either. Much of the time, social service agencies will dissipate scarce human and material resources chasing down what will often turn out to be casual or non-problematic substance use.
No doubt, we’ll deter some people from applying who more seriously misuse alcohol or illicit drugs. Even in such cases, it’s not so obvious that this is the right thing to do. After all, if a poor single mom smokes too much pot, her kids still need their food money. A better policy would focus more narrowly and specifically on public aid recipients who have identifiable difficulties linked with drug or alcohol misuse. One could then offer the kind of screening, assessment, and services that are actually likely to help.
Proposals to drug-test SNAP recipients don’t address the genuine challenges posed by drug and alcohol misuse in American society. Instead, poor families who seek a little help with the food money are being used as stage extras in a different, nasty ideological fight. It’s a depressing sight.
[i] NSDUH data aren’t perfect. For example, we cannot tell which household member actually qualifies for SNAP assistance. NSDUH does provide a large, nationally-representative sample with very detailed questions regarding substance use disorders. Data are also self-reported—a matter of considerable controversy within the addiction services research field.
[ii] To be more boring and explicit, all statistics adjust for the weighted and stratified nature of the NSDUH sample. Incremental risk is calculated by estimating a multiple logistic regression, and computing predicted probabilities at the sample mean for all variables except a dummy variable dSNAP capturing household SNAP receipt. For each dependent variable, the absolute difference in predicted probability between dSNAP=0 and dSNAP=1 is presumed to capture the incremental risk.
Sheldon Danziger is Distinguished University Professor of Public Policy, University of Michigan and president-elect of the Russell Sage Foundation. Thanks to Tedi Engler of the University of Michigan for valuable comparative analysis of Michigan data.