<< Back To Home

Beverage purchases from stores in Mexico under the excise tax on sugar sweetened beverages: observational study

Thursday, 7th of January 2016

This article contradicts the results of industry sponsored research,

at http://cie.itam.mx/sites/default/files/cie/15-04.pdf

Beverage purchases from stores in Mexico under the excise tax on sugar sweetened beverages: observational study

BMJ 2016; 352 doi: http://dx.doi.org/10.1136/bmj.h6704 (Published 06 January 2016)Cite this as: BMJ 2016;352:h6704

1. M Arantxa Colchero, researcher1,
2. Barry M Popkin, professor2,
3. Juan A Rivera, director3,
4. Shu Wen Ng, associate research professor2

Author affiliations

Correspondence to: S W Ng shuwen@unc.edu

Accepted 24 November 2015

Excerpts below; full text is at http://www.bmj.com/content/352/bmj.h6704

Abstract

Study question What has been the effect on purchases of beverages from stores in Mexico one year after implementation of the excise tax on sugar sweetened beverages?

Methods In this observational study the authors used data on the purchase of beverages in Mexico from January 2012 to December 2014 from an unbalanced panel of 6253 households providing 205 112 observations in 53 cities with more than 50 000 inhabitants. To test whether the post-tax trend in purchases was significantly different from the pretax trend, the authors used a difference in difference fixed effects model, which adjusts for both macroeconomic variables that can affect the purchase of beverages over time, and pre-existing trends. The variables used in the analysis included demographic information on household composition (age and sex of household members) and socioeconomic status (low, middle, and high). The authors compared the predicted volumes (mL/capita/day) of taxed and untaxed beverages purchased in 2014—the observed post-tax period—with the estimated volumes that would have been purchased if the tax had not been implemented (counterfactual) based on pretax trends.

Study answer and limitations Relative to the counterfactual in 2014, purchases of taxed beverages decreased by an average of 6% (−12 mL/capita/day), and decreased at an increasing rate up to a 12% decline by December 2014. All three socioeconomic groups reduced purchases of taxed beverages, but reductions were higher among the households of low socioeconomic status, averaging a 9% decline during 2014, and up to a 17% decrease by December 2014 compared with pretax trends. Purchases of untaxed beverages were 4% (36 mL/capita/day) higher than the counterfactual, mainly driven by an increase in purchases of bottled plain water.

What this study adds The tax on sugar sweetened beverages was associated with reductions in purchases of taxed beverages and increases in purchases of untaxed beverages. Continued monitoring is needed to understand purchases longer term, potential substitutions, and health implications.

Funding, competing interests, data sharing This work was supported by grants from Bloomberg Philanthropies and the Robert Wood Johnson Foundation and by the Instituto Nacional de Salud Pública and the Carolina Population Center. The authors have no competing interests. No additional data are available.

Introduction

Myriad studies suggest that added sugar in beverages is linked with obesity and many cardiometabolic problems and have recommended that efforts to reduce consumption of sugar sweetened beverages to obtain meaningful improvement to health would require a tax that leads to price increases.1 2 3 4 5 6 7 Aside from industry funded studies, the consensus from a large literature of randomized controlled trials,8 longitudinal cohort studies, and smaller clinical studies is that humans do not reduce food intake when consuming caloric beverages. The lack of dietary compensation is hypothesized to be due to form (liquid versus solid), beverage type (for example, carbohydrate content, fat content), and resultant release of hormones such as ghrelin and insulin.9 Therefore, reducing the intake of sugar sweetened beverages could reduce body weight and many cardiometabolic problems.5 10 11 12

The likelihood of obesity among Mexicans of all ages is high.13 14 The prevalence of overweight and obesity is more than 33% for young people aged 2-18 years (about the same across all age groups) and around 70% for adults (half of whom are obese).15 16 17 The prevalence of diabetes in Mexico (based on hospital admissions) is the highest among the Organization for Economic Cooperation and Development countries,18 and ischemic heart disease and diabetes are the two leading causes of mortality in Mexico.19 Additionally, the prevalence of overweight and obesity increased by 12% between 2000 and 2006 and reached 72% among adults in 2012.14 Concomitant with the rise in obesity and diabetes in Mexico are large increases in the consumption of sugar sweetened beverages20 21–Mexico had the largest per capita (163 liters) intake of soft drinks in 2011. Several studies showed that before the debate over this tax the intake of sugar sweetened beverages was rapidly increasing in Mexico.20 21 22 Reducing such consumption has been an important target for obesity and diabetes prevention.23 24 A Ministry of Health beverage guidance panel had proposed a tax years earlier and it was endorsed, among others, by many medical societies.24

In September 2013, as part of the federal budget, the Mexican congress passed an excise tax on sugar sweetened beverages and a sales tax on several highly energy dense foods.25 A specific excise tax of 1 peso/L (approximately a 10% price increase based on 2013 prices) on non-dairy and non-alcoholic beverages with added sugar and an ad valorem tax of 8% on a defined list of non-essential highly energy dense foods (containing ≥275 calories (1151 kJ) per 100 g) came into effect on 1 January 2014. Agencies collect the excise tax on sugar sweetened beverages from the manufacturers, and other research indicates that this tax is entirely passed on to consumers at the point of sale. Prices of sugar sweetened beverages increased on average by 1 peso/L in 2014 (exactly the amount of the tax), and these changes in prices, which began in the taxs first month, were observed throughout the year.26 27 Using scanned and recorded food purchase data from a representative group of Mexican households in cities with more than 50 000 residents from January 2012 through December 2014, we evaluated changes in the purchases of consumer beverages after the implementation of the excise tax.

Methods

We obtained data on purchases from January 2012 through December 2014 from Nielsen Mexicos Consumer Panel Services, which is equivalent to the data from the US Nielsen Homescan panel.28 In the US, Nielsen Homescan data have been used in several studies, including some that have linked purchases to data on nutrition labels to determine the caloric content of purchases and to evaluate industry efforts.29 30 However, linking purchases to nutrition data is currently not possible in Mexico owing to the lack of comprehensive data sources related to labeling. Therefore we focused on changes in the volumes of beverages purchased.

Each year the Nielsen Mexico Consumer Panel Services samples Mexican households in 53 cities (in 28 states plus Mexico City) with more than 50 000 inhabitants. Based on government statistics, this sample represents 63% of the Mexican population and 75% of food and beverage expenditures in 2014.31 The original dataset contained 205 827 household-month observations from 6286 households. We used complete case analysis; 715 observations (0.3%) were dropped because of missing information on the highest educational attainment of the heads of the households. Consequently, our analytic sample included 205 112 household months across 6253 households, of which 86% participated in all rounds. Each household is weighted based on household composition, locality, and socioeconomic measures through iterative proportional fitting to match demographic estimates from the National Institute of Statistics and Geography (Instituto Nacional de Estadística Geografía e Informática, INEGI). Enumerators visited the households every two weeks to collect diaries, product packaging from special bins provided for this study (scanned by the enumerators), and receipts, and to carry out pantry surveys. Bar code information provided all other data.

For descriptive purposes, we categorized the sample into the six regions used by INEGI: central north, central south, Mexico City, north east, north west, and south. The variables we used in the analysis included demographic information on household composition (age and sex of each household member) and socioeconomic status; information that is updated annually. Socioeconomic status groups (low, middle, and high) were based on a six category measure derived from annually updated questions on household ownership of assets (for example, number of bathrooms, number of bedrooms, number of vehicles owned) and education attainment of the head of the household. Onto the Nielsen Mexico Consumer Panel Services data we overlaid two contextual measures: the states quarterly unemployment rate from INEGI,32 and the two economic minimum daily salary for each year from Mexicos National Commission of Minimum Salaries33 (after adjusting for state and quarter specific inflation from INEGIs consumer price indices, www.inegi.org.mx/est/contenidos/proyectos/inp/inpc.aspx).

In this analysis we used the purchase of beverages by each household between 1 January 2012 and 31 December 2014. Data from the Nielsen Mexico Consumer Panel Services include the number of units purchased and the volume and price of each unit. From these we totalled the monthly volume and beverage categories each household purchased across each of the 36 months. Then we calculated the volume per capita per day for interpretability. Our beverage categories followed the 2012 National Health and Nutrition Survey (Encuesta Nacional de Salud y Nutrición) groupings for beverage intake as much as possible22 34; these were further grouped into larger categories or subgrouped as described in supplemental table 1. We classified products into beverage categories in 2014 based on product descriptions and sources available on the internet and in stores. In this study we focus on the top level taxed and untaxed beverages. Our two categories for taxed beverages were carbonated sodas and non-carbonated sugar sweetened beverages, and our three categories for untaxed beverages were carbonated drinks such as diet sodas; sparkling, still, or plain water; and other drinks, including unsweetened dairy beverages and fruit juices. The Consumer Panel Services did not collect information on purchases of dairy products from all of the sampled households until October 2012 (personal communication). Therefore we limited our analyses of the categories “other untaxed drinks” and “overall untaxed beverages” to October 2012 through December 2014.

Patient involvement

No patients were involved in setting the research question or outcome measures, nor were they involved in the design and implementation of the study. There are no plans to involve patients in the dissemination of results.

Descriptive statistics

We present descriptive statistics of the households in the analytic data. Then we present the unadjusted trends in household purchases as reported during the period January 2012 through December 2014, which includes the first year of the post-tax period (beginning 1 January 2014). We conducted simple t tests to determine whether the volume of beverages purchased in each post-tax month was statistically different from that of the same month in 2012 and 2013. Stata 13 was used for all analyses.35

Difference in difference fixed effects analyses

As the tax was implemented nationally, it was not possible to construct a true experimental design to study the association between the tax on sugar sweetened beverages and purchases. Therefore we applied a pre-post quasiexperimental approach using difference in difference analyses along with fixed effects models,36 37 with fixed effects at the household level. Fixed effect models have several advantages, mainly that they account for non-time varying unobserved characteristics of households (for example, preference for certain types of beverages). As such, non-time varying measures (for example, region of households residence) are omitted in the model.

As the distribution of beverage purchases per capita were skewed and not normally distributed, we used the logarithm of beverage purchases as outcomes in the models. The model adjusts for the seasonality of beverage purchases using a variable for each quarter of the year and demographic information on household composition, socioeconomic status, and contextual factors (unemployment rate and minimum salary).

To allow for interpretability, we back transformed the logged outcomes into milliliters per capita by calculating and applying Duan smearing factors.38 Specifically, Duan smearing ensures that in the presence of non-zero variances in the volume purchased, the back transformed predicted outcome is not downward biased.38 This also allowed us to compare in absolute and relative terms the estimated post-tax volume of beverages purchased in January through December 2014 to the estimated counterfactual post-tax volume assuming a pretax trend. We did consider presenting predicted values that also detrended seasonality, by setting all quarters to the same quarter, but these seasonal trends are interesting and more accurately reflect the changing demand for beverages over the course of the year. We also corrected the standard errors by clustering the analyses at household level.

The model also takes into consideration periods of non-purchases of beverage categories, when more than 10% of the observations using inverse probability weights were non-purchases. We calculated inverse probability weights by modeling the probability of purchasing, adjusted for the same covariates as the main regression for the log of purchases (the inverse of the predicted values obtained from this model being used in the main regression as a weighting factor).39 40 We conducted analyses for the full sample and stratified the analyses by socioeconomic status (low, middle, high), using separate models to determine if there were differences for these subsamples. The supplemental materials provide additional details on the analytic approach. We used Stata 13 for all analyses.35

Sensitivity analysis among untaxed beverages

We conducted sensitivity analysis for the untaxed beverages modeled from October 2012 to December 2014. Given the large number of missing values for dairy beverages from January to September 2012, imputation was not an adequate option. Instead, we repeated the models excluding dairy beverages and compared the results from January 2012 to December 2014 with those from October 2012 to December 2014.