Health

One Month or One Year? Reading a Cannabis Survey

·

A researcher comparing monthly and annual calendars with survey forms.

In 2025, 21.2 percent of Americans aged twelve or older in the population covered by a federal survey reported using marijuana in the past year. The same survey put past-month use at 15.1 percent. The difference is large enough to change a headline from roughly one in five to roughly one in seven. Both figures come from the same national report. What changed was the stretch of time that qualified a person to be counted.

Ten answers and two totals

Imagine ten people interviewed today. Two last used cannabis this week. A third last used it three weeks ago, a fourth nine months ago, and a fifth several years ago. The remaining five have never used it. In this deliberately small example, three people count toward past-month use. Four count toward past-year use, because the three recent users also qualify for the longer window. Five count toward lifetime use.

Adding the monthly and annual counts would turn four people into seven entries. Subtracting them has a sensible interpretation. One person used cannabis in the preceding year but last used it more than a month ago. The subtraction tells us nothing about why that happened. The fourth person might have stopped after years of use or tried cannabis once at a wedding. Those different histories produce the same answer to a question about recency.

Nor does the monthly total tell us how often the three recent users consumed cannabis. One might report a single occasion and another nearly every day. They each contribute one person to the count. Questions about days of use can distinguish these patterns. Convenience terms such as “current user” can obscure the distinction unless the writer gives the survey’s time window alongside the label.

Whose month is it?

The month is the period leading up to each person’s interview, rather than one named month such as July. The year works the same way. January and November interviewees therefore describe different stretches of the calendar. A past-year estimate in the 2025 report should not be read as a register of everyone who consumed cannabis between January 1 and December 31, 2025.

The survey glossary sets out how answers on lifetime use and most recent use feed into the measures. Its definition covers more than smoking. Respondents can report cannabis consumed in foods and beverages, for example. The 2025 questionnaire lets a reader follow the route through the interview. A question about ever having used cannabis leads to a question about when that last happened. Separate questions explore the ways people used it. A participant who both smoked and ate cannabis appears once in the total for users but can appear in both categories in a table of methods. Percentages in that second table can add to more than a hundred.

From a set of interviews to a national estimate

In the ten-person example, everyone counted equally. A national survey uses weights so that respondents collectively represent the population the sampling design set out to cover. For the National Survey on Drug Use and Health, that population is civilians aged twelve and older outside institutions. The methodology documentation describes the machinery behind the estimates, including sampling and weighting. A reported figure in the millions is an estimate based on interviews, with uncertainty attached.

A change between years requires another piece of care. The relevant comparison is past month against past month, for the same age group, using years the survey considers comparable. Comparing this year’s monthly figure with last year’s annual figure builds a difference into the question before anyone has answered it. Even in the ten-person example, a switch between windows would make use appear to drop by a quarter, without a single person changing their behavior. The handful of words after the percentage carry much of the story.