Sugar-sweetened beverage consumption data: who drank what, measured three ways

Consumption data on sugary drinks arrives at three different units of analysis, and mixing them up is the single most common way this research goes sideways. Person-level intake records capture what one individual drank yesterday. Market-level volume series track what a country consumed as a whole. Supply-side proxies measure how much sweetener moved into the food system - which is adjacent to consumption without being it.

CDC NHANES - Dietary Intake & Beverage Consumption Data owns the first grain, and nothing else in the slice competes there. Its two-day dietary recall files (DR1TOT and DR1IFF) record every food and beverage a nationally representative American sample consumes: one row per participant, per item, per recall day, carrying gram weights, kilocalories, total sugars and caffeine on each item. A two-year release covers about 10,000 participants; the 2017-2018 individual-foods file alone holds 112,683 rows, of which roughly 32,700 are beverages (food codes beginning with 9). Continuous cycles run from 1999-2000 through the August 2021-August 2023 release.

The other two grains split between UNESDA - Soft Drinks Europe Market Data (industry-reported volumes and litres per capita for 29 European markets, 2015 through 2025) and USDA ERS Sugar and Sweeteners Yearbook Tables (US caloric-sweetener deliveries since 1966). Three layers, three questions: who drinks it, how much a market moves, and how much sweetness entered the system.

The day-one individual-foods extract carries ten verified fields, and the identifying spine is short: SEQN names the person, DR1IFDCD names the item consumed, and everything else hangs off that pair. Four rows exactly as delivered:

SEQN     : 93704
DR1IFDCD : 94000100
DR1IGRMS : 240.0

SEQN     : 93704
DR1IFDCD : 91708030
DR1IGRMS : 26.0
DR1IKCAL : 95.0
DR1ISUGR : 13.88

SEQN     : 93705
DR1IFDCD : 92101000
DR1IGRMS : 870.0
DR1IKCAL : 9.0
DR1ICAFF : 348.0

Read the third row aloud and the design shows itself: 26 grams consumed, 95 kilocalories, 13.88 grams of total sugars - one beverage item, one recall day, one person. Read the fourth and the other pattern appears: 870 grams of a near-zero-calorie drink carrying 348 milligrams of caffeine. Sugary drinks, diet drinks, juices and water all live in the same file, separated only by food code, which is what makes category comparisons possible without stitching sources together.

WTDRD1, the dietary day-one sample weight, rides along on every participant. It converts the surveyed sample into a national estimate; unweighted averages inherit the survey's sampling design and will mislead on any demographic cut.

Three clocks tick at once, and they start in different centuries. The intake clock is the youngest: NHANES dietary releases run continuously from 1999-2000 forward, in two-year steps, which still yields more than two decades of person-level trend. The supply clock runs longest - USDA ERS has tracked caloric-sweetener deliveries since 1966, with prices since 1980, capturing the HFCS era from its beginning. The price spine outdoes both: BLS nonalcoholic beverages indexes reach January 1947, with carbonated drinks broken out from juices.

Europe's ledger is shorter but denser. UNESDA's 29 market files cover 2015 through 2025 for the EU aggregate and 2016 through 2025 for most countries - annual, national, split by category and again by calorie rating. The UK adds a second industry account: British Soft Drinks Association's Annual Industry Insights Report covers 2025 calendar-year volume and value sales by category, with archive publications reaching back to 2014.

The practical consequence: any claim about 'rising consumption since the nineties' rests on the supply proxies or the price series, not on person-level records, because intake microdata simply did not exist in continuous form before 1999. Label which clock a number comes from and most apparent contradictions between studies dissolve.

Six records do nearly all the work behind this query, drawn from the fourteen that carry Soft Drinks & Non-Alcoholic Beverages as their primary tag in Datadory's catalog:

• CDC NHANES - Dietary Intake & Beverage Consumption Data (quality score 9) - the person-level layer. Every beverage a representative American sample consumes, grams and sugars attached, demographics joinable on SEQN.
• UNESDA - Soft Drinks Europe Market Data (7) - the European market layer. Volume in million litres and litres per capita across five reporting lines, plus the no/low-calorie share of soft drinks, for 29 markets.
• WHO Global Health Observatory Indicators (8) - the policy and outcome layer. SSB-tax status waves for about 194 countries and adult obesity estimates from 1980 through 2024, one country-year fact per row.
• USDA ERS Sugar and Sweeteners Yearbook Tables (9) - the ingredient ledger. Caloric-sweetener deliveries since 1966, sugar and HFCS prices since 1980, tariff-rate quota detail and port-level Mexican imports.
• Open Food Facts (9) - the product-composition layer. 4,697,746 barcode-level products worldwide with brand, category, transcribed ingredients and sugars per 100 grams or millilitres.
• American Beverage Association - Industry Data & Insights (5) - the industry's own account: Balance Calories Initiative sugar-reduction reports from the 2014 baseline through 2024, beside a member directory of roughly 171 companies resolved facility by facility.

The remaining eight primaries - Census demand geography down to tract level, FoodData Central's federal nutrient reference, FAOSTAT's upstream crop accounts, the BLS price spine, BSDA's UK categories, Fresha's on-premise channel proxy and the two public-sector catalogs - fill specific jobs rather than headline ones.

The table lines up the six records a consumption study actually leans on, with the grain each one promises and the history it carries.

Notice what the comparison exposes: no single record spans both person-level detail and multi-country breadth. NHANES covers one country deeply; UNESDA covers thirty markets thinly; WHO covers the world but measures policy and outcomes rather than drinks consumed. Studies that need several of these axes at once are the norm, not the exception - which is why the join keys matter more than any single record's depth.

Within Datadory's catalog this subject sits in a slice whose mean quality score of 7.50 sits just under the catalog-wide 7.81, with five of fourteen records scoring 9 - unusually strong at the top for an industry whose flagship record is a government health survey rather than a commercial feed.

• Sugar-intake quantification. Total sugars per beverage item per person per day, summed across a recall and weighted, turns 'Americans drink a lot of soda' into a distribution with a mean, a tail and demographic cuts by age, sex and income band.
• Tax evaluation. Join WHO's three SSB-tax waves to UNESDA per-capita volumes for taxed and untaxed European markets, aligning years explicitly, to estimate whether volumes bent after implementation.
• Reformulation tracking. Open Food Facts' sugars-per-100ml values, timestamped per product, show whether manufacturers answered a tax by cutting sugar inside the bottle rather than losing volume.
• Supply-vs-intake reconciliation. ERS deliveries measure sweetener entering food manufacture; NHANES measures sweetener reported as consumed. The gap between the two, tracked over cycles, is itself a finding about waste, nonbeverage use and reporting bias.
• Portfolio and ESG positioning. Benchmark a company's sugar exposure by market against each market's obesity burden and policy trajectory, using the same country keys across all three panels.

Four personas take the lion's share: market researchers and consultants segmenting consumers from microdata, public-health and policy analysts running tax evaluations, competitive-intelligence and product teams tracking reformulation, and investors & quants overlaying policy risk onto consumer-staples positions.

The raw records arrive in incompatible shapes - SAS transport files keyed on sequence numbers, Excel workbooks with figures embedded in year columns, an indicator warehouse returning country-year facts with uncertainty bounds as strings. Every analytical mistake in this workflow starts with someone re-deriving a codebook mid-project. So the decoded version is the product: typed numerics, labeled categorical values, dates normalized, the field dictionary traveling unchanged across whatever channel you choose.

Start with a sample: name the markets, cycles, categories or years you actually touch, and the extract returns shaped like the question rather than like the publisher's filing cabinet.

Self-report bias leads the list. NHANES recalls rely on participants remembering two days of intake, and published validation work consistently finds under-reporting of snacking and sugary drinks - treat levels as floors and trends as the signal. The two-day window also cannot see seasonal variation within a person, and vintage moves in two-year steps, so the freshest wave always lags the calendar.

Timing mismatches bite when joining layers. WHO's tax status exists only in 2017, 2019 and 2021 waves while UNESDA volumes run annually from 2015 - deciding whether a tax year maps to the same calendar year of sales is a real modeling decision, and the answer differs by country implementation date.

Coverage holes follow the funding map. No intake survey equivalent to NHANES spans Europe, so European consumption claims rest on industry-reported market volumes rather than individual records; the UK adds tens of headline figures where a microdata file would sit. And proxy quality degrades quietly: caloric-sweetener deliveries measure sugar and HFCS moving into food manufacture, not soda consumed, and FAOSTAT balance sheets spread sweetness across every use. Both are legitimate proxies if your methods section says so.

Finally, weights are not optional on the intake side. WTDRD1 is what makes a participant sample into a national estimate; skipping it produces numbers that look plausible and describe nobody in particular.

This page is one workflow inside the soft-drinks stack. Country-level sizing worked in depth lives at soft drink market size by country; the upstream ingredient view sits at sugar cane production statistics dataset. The soft-drinks-non-alcoholic-beverages data hub keeps all fourteen records ranked, and best soft drinks non alcoholic beverages datasets puts the scorecard in one view.

For record-level detail, the product pages carry complete field dictionaries and samples: the CDC NHANES dietary intake record, the UNESDA Europe market record and the WHO Global Health Observatory record. The head-to-head between the person-level and supply-side views is written up in the NHANES vs FAOSTAT comparison.