Skip to main content

Gluten-Reduced Beer Consumer Survey, 2014–2015: The Tabulated Results

985 people living gluten-free, asked what happened when they drank beer brewed from barley and then processed to remove gluten. Nearly four in ten who tried it reported a physical reaction. More than four in ten did not know at the till what grain it was made from. Seven in ten said it should not be sold as gluten-free. This page is the tabulated record — every figure, with the CSA's published numbers and our own recomputations shown side by side.

This dataset supports: Gluten-Reduced Beer Survey Results — the narrative page that interprets it, and Gluten-Reduced Beer Is Not Gluten-Free.

This page is deliberately a data page. The argument — why liking the beer protected nobody, why the confusion was structural rather than incidental — lives on the narrative page. This page is the record itself.

Methods, briefly

The Celiac Support Association and Bard's Beer surveyed people living gluten-free between 22 December 2014 and 28 February 2015, hosted on FluidSurveys. 985 unique responses were collected, including 378 written comments. Respondents were asked about diagnosis, whether they had tried a barley-based beer processed to remove gluten, what happened when they did, whether they knew the grain at purchase, what gluten level they personally accept, and whether such products should be marketed as gluten-free.

Per-question denominators. Every percentage uses the number of people who answered that question as its base, which is why the bases differ — only respondents who had already tried the beer were asked whether they knew it contained barley, so that question has 832 answers rather than 985. Where a base could not be reconstructed from the published summary, the base column is blank rather than guessed.

Two provenances, both published here. The CSA's published summary is the source of record. Where Gluten Free Brewer recomputed a figure from the raw response export, both numbers appear, labelled. They differ slightly and in one direction that matters to nobody's conclusion — the published overall reaction rate is 39.8%, the recomputation is 38.9% — but the arithmetic belongs in the open.

What is and is not published here

Published — two files. The tabulated results (counts, bases, percentages for every measure below), and the response-level export: all 985 individual responses, de-identified, so anyone can recompute our figures or ask a question we did not.

Removed from the response file: IP address, browser, device, referrer, submission timestamps, and the survey platform's respondent IDs. Every row carries a sequential number that links to nothing.

Geography is kept but coarsened. Respondent location is state or country level — useful data, and the reason the file supports a map. Any location with fewer than five respondents is collapsed into Other US state or Other international, because a single respondent from one small territory, combined with a diagnosis and a reported reaction, stops being a statistic and becomes a person. That moves 60 of 822 rows; every state large enough to plot survives intact.

Held back: the 378 free-text stories and the 184 "other, please specify" entries. Those are people describing their own medical experiences in their own words. The consent language governing the 2015 survey is not in the export and we have not located it — until we know what respondents agreed to, republishing their words is a decision that is not ours to make. If you are researching this and need them, ask; we would rather handle that individually than publish it and hope.

On the brand columns. Respondents could tick more than one beer, so a reaction cannot be attributed to any single brand — a per-brand reaction rate computed from this file is not valid, however tempting the arithmetic. The brands are published because withholding them would be its own distortion, and because the rates cluster closely enough (29–43%) to say something useful: this reads as a property of the process, not of one company.

How to read this through our doctrine

Two figures carry the argument, and neither is the headline.

The first is the split by liking. Of the 578 respondents who enjoyed the beer, 209 reported a gluten-related reaction anyway — 36.2%, against 38.9% for everyone who drank one. Taste and safety came apart completely. A consumer cannot taste the risk, which is why self-assessment is not a control.

The second is severity. Of 279 reported reactions, 141 were strong rather than slight — 19.7% of everyone who drank one. Half the reactions here were the severe kind, not faint unease.

One honest note in the other direction: the people who disliked the beer reported reactions at 50.4%, meaningfully higher than those who liked it. Some of them plausibly disliked it because they reacted to it. That is a confound worth naming, and it does not weaken the finding — it sharpens why the liked-it group is the one to look at.

Explore this data

Sort, filter, and compare measures across sections. The full tables below remain the printable, citable record; this is the same data, interactive.

Interactive explorer30 tabulated measures

Loading the explorer…

The dataset

Percentages as recorded. "Base n" is the number of people who answered that question where it could be established; blank where the published summary does not state it.

Sample and exposure7 measures
MeasureBaseBase nCount%Provenance
Total unique responsesall respondents985985100.0CSA published summary
Written comments submittedall respondents985378CSA published summary
Diagnosis requiring a gluten-free dietanswered the diagnosis question87.0CSA published summary
Diagnosis requiring a gluten-free dietanswered the diagnosis question98088.3GFB recomputation from raw export
Diagnosed with celiac diseaseanswered the diagnosis question75.7CSA published summary
Diagnosed with celiac diseaseanswered the diagnosis question98075.5GFB recomputation from raw export
Had tried a barley-based beer processed to remove glutenall respondents65.7CSA published summary
Reactions13 measures

Among everyone who tried it and reported an outcome:

MeasureBaseBase nCount%Provenance
Reported a gluten-related physical reactionthose who tried it39.8CSA published summary
Reported a gluten-related physical reactiontried it and reported an outcome71727938.9GFB recomputation from raw export
No reactiontried it and reported an outcome71743861.1GFB recomputation from raw export
Slight reactiontried it and reported an outcome71713819.2GFB recomputation from raw export
Strong reactiontried it and reported an outcome71714119.7GFB recomputation from raw export

Split by whether they liked the beer:

MeasureBaseBase nCount%Provenance
Any reactionliked the beer57820936.2GFB recomputation from raw export
No reactionliked the beer57836963.8GFB recomputation from raw export
Slight reactionliked the beer57810718.5GFB recomputation from raw export
Strong reactionliked the beer57810217.6GFB recomputation from raw export
Any reactiondid not like the beer1397050.4GFB recomputation from raw export
No reactiondid not like the beer1396949.6GFB recomputation from raw export
Slight reactiondid not like the beer1393122.3GFB recomputation from raw export
Strong reactiondid not like the beer1393928.1GFB recomputation from raw export

Among diagnosed celiac respondents:

MeasureBaseBase nCount%Provenance
Any reactiondiagnosed celiac respondents who tried it52921139.9GFB recomputation from raw export
Awareness, marketing, and acceptance thresholds10 measures
MeasureBaseBase nCount%Provenance
Not aware at purchase that the beer contained wheat or barleyanswered the awareness question83236143.4CSA published summary
Both reported a reaction and were unaware of the grainanswered both questions149GFB recomputation from raw export
Said such products should not be marketed as gluten-freeall respondents72.1CSA published summary
Will consume zero gluten onlyanswered the threshold question41.2CSA published summary
Accepts anything labelled gluten freeanswered the threshold question22.0CSA published summary
Accepts anything certified gluten freeanswered the threshold question19.9CSA published summary
Accepts less than 20 ppmanswered the threshold question10.8CSA published summary
Accepts less than 5 ppmanswered the threshold question3.5CSA published summary
Accepts less than 10 ppmanswered the threshold question2.7CSA published summary

Download

Response-level export, 985 rows (CSV) — one row per respondent, de-identified. This is the file to use for reanalysis.

Tabulated results, 30 measures (CSV) — bases, counts, percentages, provenance, and per-row notes.

Response file — data dictionary13 columns
ColumnValues
respondent_idSequential 1–985. Arbitrary; links to nothing.
locationUS state or country, as geolocated by the survey platform. Cells below n=5 collapsed to Other US state / Other international. Blank for 163 respondents.
reason_for_gluten_free_dietCeliac diagnosis · another diagnosis requiring a gluten-free diet · health reasons · does not follow one
consumption_habitZero gluten · accepts "gluten free" labelling · accepts certification · under 20 / 10 / 5 ppm
tried_gluten_reduced_beerYes · No
tried_brunehaut, tried_estrella_damm_daura, tried_omission, tried_prairie_path, tried_other_brandyes or blank. Multi-select — see the brand note above.
experienceSeven values crossing liking with reaction: liked / did not like × no reaction, slight, strong; plus "I have not tried one."
aware_of_grain_at_purchaseYes · No. Asked only of those who had tried it (832 answers).
should_be_marketed_gluten_freeYes · No

Blank means the respondent did not answer that question. Every percentage on this page uses the number who answered that question as its denominator, which is why bases differ.

Record

First published 2026-08-14 as v1.0. Corrections to date: none. This dataset is a frozen record — it changes only for corrections, and every correction is logged in this section with its date and what changed.

Zenodo deposit: pending. This dataset will be deposited under CC BY 4.0 and this section updated with the labeled DOI pair (this version and all versions) per the data-page standard.

What this dataset does and does not establish

Supports: that many gluten-free consumers did not treat gluten-reduced beer as equivalent to gluten-free beer; that reported reactions were common enough to demand follow-up; that liking the beer did not predict tolerating it; that label and retail confusion were real.

Does not establish: a clinical reaction rate, a biological mechanism for any individual reaction, that every person with celiac disease will react to every gluten-reduced beer, or that survey data replaces laboratory testing, medical research, or regulatory review. These are self-reported consumer responses collected online, with the selection effects that implies.

How to cite this dataset

Celiac Support Association and Bard's Beer. Gluten Reduced Beer Survey, 2014–2015. Tabulated results republished by Gluten Free Brewer, 2026. https://glutenfreebrewer.com/research/survey-data/gluten-reduced-beer-survey-2015

Source and Validation Notes

Assembled programmatically from two in-repo sources: the published figures, thresholds, awareness numbers, and reconciliation footnote on docs/gluten-free/gluten-survey-results.md, and the reaction cross-tab held in the SurveyReactionChart component, which was itself recomputed by GFB from the raw response export. A single generator emits both this page's tables and the CSV, so the download and the display cannot drift apart.

Every internal figure was reconciled before publication: the three reaction bands each sum to their stated totals (438+138+141 = 717; 369+107+102 = 578; 69+31+39 = 139); the liked and disliked groups sum to the whole (578+139 = 717); reported reactions sum to 279; and each stated percentage recomputes from its own count and base to the published decimal.

The response-level file was built directly from the archived platform export (UTF-16, tab-delimited, 985 rows × 24 columns) by a script that drops the identifying columns, coarsens geography, and renames the survey questions to short field names. It was verified by reproducing the published control totals from the output file rather than from the source: 717 respondents who tried the beer and reported an outcome, 740 celiac diagnoses, 700 against the marketing claim, 361 unaware of the grain at purchase.

A correction to what this archive was previously described as. The export was recorded as having "names and contact details removed." That is true but was incomplete: the file also carried an IP address for all 985 respondents, geolocation for 822, plus browser, device, and referrer. None of those reach this site. The narrative page's description has been corrected to match.

The originating organisations are named because the survey was published under their names and Bard's commissioned it. Percentages are as published by the CSA except where a GFB recomputation is explicitly labelled.