Who smokes in Bangladesh

A single national number is the least useful thing a survey can tell you. In the 2009 STEPS survey, 26.2% of Bangladeshi adults currently smoked tobacco. That figure is accurate, and on its own it would send a programme in the wrong direction — because almost nobody in the population actually looks like the average.

The same dataset holds three breakdowns of that one number. Each says something a tobacco control programme would want to act on.

Use climbs with age, then eases

The clearest pattern is age. Prevalence rises across every band from the mid-twenties onward, peaking in the 55–64 group, before easing slightly among the oldest adults.

The gap between the youngest and the highest band is about eleven percentage points — the oldest working-age adults are roughly half again as likely to smoke as those in their late twenties. The slight fall after 65 is the kind of shape that can reflect either people quitting or a survivorship effect; the survey cannot distinguish between the two, and this report will not pretend otherwise.

Almost entirely men

Age is a gradient. Sex is a cliff.

Men are more than forty times as likely to smoke as women. This is the single most important fact in the dataset, and it is invisible in the 26.2% headline. A campaign built around the national average would be pitched at a population that is, in practice, overwhelmingly male — and would carry no useful signal about the small number of women who do smoke.

A modest urban–rural difference

The third breakdown is the mildest, and worth knowing precisely because it is mild.

Rural prevalence runs a little above urban. It is a real difference but a small one, and much smaller than either the age gradient or the sex gap. For anyone deciding where to direct effort, that ordering is the finding: target by sex first, age second, geography last.

A note on reading this file yourself

Every chart above is computed from the published CSV when this page loads — no figures are typed into the text. If you open the dataset, note that it stores three rows for each breakdown: the estimate, plus a lower and an upper confidence bound. Any analysis has to keep only the estimate rows, or it will quietly average a number together with its own error bars. The charts here filter on Type of indicator: Estimated for exactly that reason.