
Mexico's Soda Tax Cut Purchases 6%, Not 12%: What the Study Found
November 22, 2026 · Frisian News
Photo: Wikimedia Commons
The 2016 BMJ study on Mexico's soda tax is widely cited at a 12% drop in sugary drink purchases. The paper's own annual average is 6%, measured against a statistical model, with causality explicitly unconfirmed and no health outcome recorded.
Twelve percent has become the number that follows Mexico's soda tax around the world. The figure comes directly from the BMJ study by Colchero, Popkin, Rivera and Ng published in January 2016 (h6704), and it is in the paper. It describes one month: December 2014, the last month of the first year of the tax.
The paper's annual average for 2014 is 6%. Using store purchase data from over 6,200 households in 53 Mexican cities with populations above 50,000, the researchers compared actual purchases with a statistical model of what purchases would have looked like without the tax. That gap averaged 6% across the year, working out to 4.2 fewer litres of taxed beverages per urban Mexican over twelve months.
The monthly gap grew through the year, reaching the December peak before the year closed. Purchases of untaxed beverages rose 4% above the modelled baseline over the same period, driven mainly by increased buying of bottled water. The data covered shop purchases only. Drinks from taps, food service outlets, and fountain machines were outside the dataset.
The researchers were explicit about what their design could not show. The ScienceDaily summary of the paper notes the team "emphasise that this is an observational study so no definitive conclusions can be drawn about cause and effect". The study logged no weight outcomes, no diabetes diagnoses, and no health measure of any kind. The stated aim was to track purchases, not to establish whether health improved.
A follow-up paper by the same four authors, published in Health Affairs in 2017 and drawing on data from 6,645 households through December 2015, found the effect held and deepened. Taxed beverage purchases fell 5.5% in 2014 and 9.7% in 2015, for an average decline of 8.2% across both post-tax years. The lowest-income households reduced their purchases most in both years. That longer record is evidence of a real response, not a one-year artifact.
What the two-year data do not show is a 12% annual reduction at any point. The December 2014 peak made the press; the full-year averages, both in the 6% to 10% range, appeared in fewer headlines.
A tax that reduces taxed-drink purchases by 6% to 10% a year, against a modelled counterfactual, with no recorded health outcome and a stated inability to establish causality, may still be defensible policy. The data do not resolve that question. They settle one point: the headline number is a December reading, and the annual average is half that.
Published November 22, 2026 · Frisian News