Smoking prevalence by country: one number per economy-year

World Bank - Prevalence of Current Tobacco Use (% of adults) — indicator SH.PRV.SMOK — is the number public-health arguments default to, and it earns the position structurally. One row per economy-year, roughly 265 economies, territories and aggregate regions, annual observations from 2000 to 2024 — about 6,625 rows — all carrying one definition: the percentage of people aged 15 and over who currently use any tobacco product, smoked or smokeless, daily or non-daily, age-standardized to the WHO Standard Population so a young median-age economy and an ageing one sit on the same scale.

Three properties make it unusually workable. Consistency: one indicator code applying one definition everywhere, which is precisely what cross-country tobacco work usually lacks. Reach: small states that commercial panels skip sit in the same table as the G20, keyed on ISO-3 codes. Completeness of context: cigarettes, pipes, cigars, cigarillos, waterpipes, bidis, kretek, heated tobacco and every form of oral or nasal smokeless tobacco count toward the figure; e-cigarettes stay out because they contain no tobacco. Within Datadory's catalog of 1,744 datasets across 159 viable industries, this record scores 10 out of 10 against a slice-wide mean of 7.81 — the only perfect score in the tobacco slice. Get a sample cut to your economies and years before anything else.

The phrase hides three different claims, and mixing them is the most common error in cross-country work.

• The World Bank's SH.PRV.SMOK counts current use of any tobacco product among adults — smoked and smokeless, daily or non-daily, age-standardized. Broadest product set, widest geography.
• Eurostat - Daily Smokers of Cigarettes (sdg_03_30) counts the population aged 15+ who smoke boxed cigarettes, cigars, cigarillos or a pipe — vapes and snuff deliberately excluded, so vaping cannot blur the combustible trend — measured only in Eurobarometer survey waves rather than interpolated annually.
• WHO Global Health Observatory modeled estimates report current tobacco use with low/high uncertainty bounds attached to every value: Malaysia's 34.5% male estimate for 2022 ships as 34.5 [22.0-47.0], Cameroon's 9.1% for 2023 as 9.1 [6.5-11.8] — two bands that say more about underlying survey disagreement than any methodology note.

Pick one definition per analysis, document it, and never average across sources. A Eurostat share, a World Bank any-tobacco figure and a WHO modeled estimate can disagree about the same country in the same year without any of them being wrong — they are answering different questions.

One observation per economy-year, reproduced exactly as captured at research time:

# total-adults line
SH.PRV.SMOK     AFE  2023  11.8997926098489   # Africa Eastern & Southern
SH.PRV.SMOK     AFE  2022  12.1786708462139
SH.PRV.SMOK     AFE  2024  11.650004712267

# sex splits ride sibling codes on identical keys
SH.PRV.SMOK.MA  AFE  2023  19.9590625406671   # male adults
SH.PRV.SMOK.FE  AFE  2023   3.77511582966064   # female adults

Read the anatomy rather than the digits. Four coordinates fix every observation — indicator code, economy name, ISO-3 code, year — and the value carries full computational precision because each figure derives from harmonized survey estimates rather than rounded press releases. The gap between the male and female readings for the same region-year, nearly five to one here, is exactly the structural fact the sex-split variants exist to expose; the total line folds it away by construction. Economy-years nobody observed arrive null, and nulls ship explicit rather than imputed, so absence stays distinguishable from zero.

Yes — and build the split in from the start rather than deriving it later, because the sex gap is often the story itself.

The World Bank publishes dedicated indicators for each sex: Prevalence of Current Tobacco Use, Male (% of male adults) (SH.PRV.SMOK.MA) and the female counterpart SH.PRV.SMOK.FE, each with roughly 5,000 rows across the same 2000-2024 span as the 6,625-row total-population series, on identical economy-year keys — so a three-series panel assembles with a two-column join. The smaller row counts mean the disaggregated series carry more gaps; check balance before fitting a panel model.

Eurostat's sdg_03_30 splits every value into Total, Males and Females for 35 European entities aged 15+, roughly 620 observations across seven waves. In the 2014 wave the EU-27 read 27% overall — split 32% of men against 22% of women, a ten-point gap that is itself the headline for anyone forecasting combustible volume.

Three failure modes account for most bad comparisons, and all three survive contact with good data.

Leaving aggregates inside a country ranking. The roughly 265 World Bank codes include regional composites such as Africa Eastern and Southern and income-group rollups riding the same schema as sovereign states. Any ranking that fails to tag them ranks Germany against an entire region. Every row carries its ISO3 code precisely so the filter stays one predicate long.

Averaging over unobserved economy-years. Null means not measured, never zero smoking. Panel completeness per year should be measured rather than assumed; average across unfiltered nulls and the missing states silently drag the panel toward nothing.

Mixing definitions mid-analysis. A Eurostat daily-cigarette share, a World Bank any-tobacco figure and a WHO modeled interval estimate are three different claims. Joined into one chart without labeling the units, they draw a series that looks continuous and is not.

Five records cover most country-level prevalence work. What separates them is geography, definition and grain rather than brand — the columns below are what decide which one anchors a project. As cataloged by Datadory, August 2026:

Market researchers and consultants size nicotine demand from the total-adults line, then sharpen with the male and female splits — one definition doing the harmonization work that normally consumes a study's first month.

Investors and quant researchers read prevalence trajectories as the citable spine of tobacco-demand narratives across markets, on one definition rather than a patchwork of national statistics offices.

Data scientists and ML engineers drop a verified country-year panel with explicit nulls and stable keys into feature stores without a cleaning pass — the headline column arrives before the sex splits append onto the same keys.

Public-health and excise analysts set prevalence slopes beside policy scores and tax records, turning a league table into an evaluation of whether intervention moved behavior. WHO's Report on the Global Tobacco Epidemic 2023 contributes the MPOWER scores — every member state rated 1-to-4 on each demand-reduction measure back to 2008.

Every one of these jobs runs off the same typed rows — delivered daily, weekly, or hourly, your call — so the workflow is a sample request away rather than a pipeline build.

Because the numbers were never the hard part — shape is. Prevalence here means three sibling indicator codes sharing one spine, aggregate regions hiding inside the country list, nulls that mean unmeasured rather than zero, and four definitions of 'smoking' spread across the field's catalogs. Reconciling all of that into one analysis-ready table eats days before any modeling starts.

Files, feeds, or straight into your warehouse. Daily, weekly, or hourly - your call. Datadory handles the decoding upstream of you: aggregates tagged so a sovereign-state filter stays one predicate, nulls kept explicit, the male and female splits aligned onto identical keys, and the Eurostat, WHO and IHME companions delivered joinable without a crosswalk. Every delivery ships with the field dictionary, sample rows for validation and a coverage statement stating exactly where the panel ends.

Start with a sample: name the economies, sexes and years you need, and real rows come back cut to them before anything recurring starts.

This page is one thread of the tobacco stack. Open the tobacco data hub for the complete scored index of the industry slice, or the ranked best tobacco datasets shortlist for the top of it. The cigarette consumption per capita page covers the demand side this prevalence panel pairs with naturally; the US cigarette production statistics page covers the American supply side. Dataset pages document the total-adults panel, the male and female variants and the WHO Global Health Observatory legacy repository down to the field dictionary, and the trade-off between the male-split and total lines is laid out in the male vs adults comparison. Practical order of operations: load the total-adults line for the cross-country baseline, append both sex splits on the shared keys, then layer MPOWER scores or GBD burden once the panel is clean.