Long-Term (Multi-Year) Purchase and Substitution Effects of the Mexican SSB Tax

What the evidence says

Multiple independent data sources agree on the core direction: Mexico's 2014 SSB excise tax reduced purchases of taxed sugar-sweetened beverages, and the reduction grew larger from the first year to the second. Falbe (2016), citing the original Colchero data, reports a 12% reduction in purchases of taxed SSBs one year after implementation, with the reduction increasing over that year (magnitude unspecified). Nakhimovsky (2016) reports a 6.1% average reduction across the first year that reached 12% by December 2014. Colchero (2016) reports a 6.2% decline in 2014 and 8.7% in 2015 (model 2), or 7.3% averaged across 2014–15 against a 2007–13 baseline (model 1). Miracolo (2021), synthesizing Colchero et al. (2017), reports 5.5% in 2014 and 9.7% in 2015, averaging 7.6%. A structural demand-supply model of the tax as currently implemented (Salgado Hernández, 2021) estimates a larger 19% reduction in both volume and sugar purchased, alongside a price increase of MP $1.24 per liter. The estimates cluster in the mid-single digits to low double digits depending on data source (household purchase panels vs. retail/structural models) and time window, but every source that measures the trajectory across more than one year finds the reduction accelerating rather than fading.

The mechanism runs through price pass-through and price elasticity. Nakhimovsky (2016) finds the price of taxed sugary soda rose 1.3–1.4 pesos per liter in 2014–15, more than the 1-peso-per-liter tax itself, while prices of other SSBs rose only 0.6 pesos per liter — less than the tax — and that smaller increase was no longer statistically distinguishable from zero by the first quarter of 2015. Own-price elasticity of demand for soda in Mexico grew more negative over time in repeated cross-sectional surveys: -0.61 in 1989, -0.85 in 1998, -1.085 in 2006. Salgado Hernández's post-tax structural estimate of -1.2 to -1.3 is consistent with a continuation of that decades-long trend toward greater price sensitivity.

The purchase reduction is not uniform across income groups. Falbe (2016) reports low-socioeconomic-status households cut purchases by 17%, compared to 12% overall. Nakhimovsky (2016) reports a quasi-experimental estimate of 9.1% reduction in the lowest socioeconomic third versus 5.5% in the highest third. Miracolo (2021) reports the same qualitative pattern — the largest reduction concentrated in the lowest socioeconomic group — without a comparable number of its own. All three sources agree on direction and rank ordering.

On substitution, plain water purchases increase: Colchero (2016) finds a 5.2% rise in per capita water sales in 2014–15 versus 2007–13 (model 1), or 11.8% in 2015 alone (model 2, significant). Miracolo (2021) reports the same 5.2% figure for the same period. Other untaxed categories do not show a comparable increase: Colchero (2016) finds no increase in low-calorie carbonated beverages (27.9 ml/capita/day in 2013 versus 2013 versus 27.7 ml/capita/day in 2014–15) and no significant increase in 100% fruit juice. On cross-price relationships, Nakhimovsky (2016) reports milk as a substitute in cross-sectional studies (cross-price elasticity 0.19–0.52), with the absolute own-price elasticity larger than the cross-price elasticity in every study reviewed, meaning any milk-consumption increase would be smaller than the SSB-consumption decrease. Candies and snacks, by contrast, behave as complements (cross-price elasticity -0.23 to -0.44 across two Mexico studies), meaning the SSB tax's effect on total energy intake is smaller than the SSB purchase reduction alone would suggest.

On long-term health outcomes, the evidence is thin to absent. Nakhimovsky (2016) states directly that existing evidence is insufficient to determine the long-term effect of the tax on obesity prevalence, based on only one quasi-experimental and two modeling studies, and calls for research using stronger designs. Separately, Miracolo (2021) reports that some studies project tax effects on consumption "may be more substantial in the long-term rather than the short-term" because habit formation is gradual — but this is explicitly a projection rather than an empirical, multi-year measurement, and no source in this set supplies data beyond the roughly two-year window (2014–2015) to test it.

Where the evidence splits

Sources disagree on what happens to purchases of untaxed beverages generally. Nakhimovsky (2016), using household purchase panel data, finds non-taxed beverage purchases "increased but not significantly from zero." Miracolo (2021), citing a separate segmentation study, reports that unhealthy-consumption households reduced taxed-beverage purchases while simultaneously increasing untaxed-beverage purchases. Salgado Hernández (2021), using market-share data from a structural demand-supply model, finds the opposite for formal untaxed beverages: their market share fell from 20.27% to 18.19%, alongside the fall in taxed beverages' share (19.61% to 17.56%); only the "outside option" — mainly informal, home-delivered water — gained share, from 60.11% to 64.25%. Salgado Hernández offers its own explanation for its result (a possible reduction in purchasing power for branded beverages), but the sources do not reconcile this with Nakhimovsky's and Miracolo's findings of rising untaxed purchases; the difference in data source (household purchase panel vs. retail/structural market-share model) and in how "untaxed" is defined (all untaxed beverages vs. formal retail vs. informal outside-option water) is stated, but no source explains why the two approaches point in different directions for formal untaxed beverages specifically.

Nakhimovsky (2016) reports milk as a substitute for SSBs in cross-sectional Mexican studies (positive cross-price elasticity, 0.19–0.52), but the same source notes that a quasi-experimental study from Mexico found milk to behave as a complement instead. The source does not give the magnitude of that complement estimate or resolve which finding should be trusted; it attributes the discrepancy only to the difference between cross-sectional and quasi-experimental designs, without further detail.

Within Colchero (2016) itself, the size of the increase in plain water sales depends on model specification: model 1 shows a 5.2% increase across 2014–2015 combined, while model 2 shows a larger, statistically significant 11.8% increase in 2015 alone. The paper does not state why the two models diverge beyond differing in specification.

Thresholds and dose-response

No threshold established in the current sources.

Open empirical questions

• No source in this set provides quantified purchase or substitution data beyond the roughly two-year window (2014–2015); whether the accelerating reduction continues, plateaus, or reverses in years 3+ is not measured here.
• Whether the purchase and substitution shifts translate into measured obesity or other health outcomes is explicitly flagged as unresolved (Nakhimovsky).
• Why household purchase panels (Nakhimovsky, Miracolo) find flat-to-rising untaxed beverage purchases while a structural market-share model (Salgado Hernández) finds formal untaxed beverage share falling is not reconciled by any source here.
• Whether milk functions as a substitute or complement for SSBs in Mexico specifically — cross-sectional vs. quasi-experimental designs disagree, and no source quantifies the complement estimate.