Evaluating the Impact of Singapore’s Nutri-Grade Beverage Labels on Prepackaged Beverage Purchases and Product Nutrient Composition: A Before–After Study, 2019–2024

Abstract

In December 2022, Singapore implemented mandatory Nutri-Grade front-of-pack labels for beverages. Using household scanner data (April 2019–March 2024), we evaluated changes in sugar purchases and product formulation. Households reduced sugar purchases by 3.1 grams per day (−18%), and purchases of unhealthy beverages fell 44%. Reformulation and reassortment lowered sugar content by 1.16 grams per 100 milliliters (−21%) but increased sweetener and lactose use. Higher-income and nutritionally literate households showed larger reductions. Findings support mandatory labeling but highlight equity concerns and supplier substitution risks. (Am J Public Health. 2026;116(7):893–897. https://doi.org/10.2105/AJPH.2026.308476)

To address Singapore’s burden of noncommunicable diseases,^1,2 the government implemented the Nutri-Grade (NG) policy on December 30, 2022, mandating front-of-pack nutritional labeling for all prepackaged nonalcoholic beverages.³

INTERVENTION AND IMPLEMENTATION

Under this policy, beverages are assigned a color-coded grade from A (healthiest) to D (least healthy), based primarily on free sugar content, and further downgraded if saturated fat thresholds are exceeded (Appendix Figure A, available as a supplement to the online version of this article at http://www.ajph.org). Products graded C or D must display NG labels, and D-graded beverages are prohibited from advertising.

This mandatory labeling scheme builds on the existing voluntary Healthier Choice Symbol program. A two-year transition period (late 2020–2022) was provided to enable supplier reformulation before full enforcement. The rollout of NG labeling created a natural policy window to evaluate both consumer purchasing shifts and industry reformulation or product assortment changes.

PLACE, TIME, AND PERSONS

This evaluation was conducted in Singapore, using NielsenIQ HomeScan data from a nationally representative panel of 1000 households (Appendix I Table A, available as a supplement to the online version of this article at http://www.ajph.org). We analyzed purchases of prepackaged beverages in two time periods: prepolicy (April 2019–March 2020) and postpolicy (April 2023–March 2024). Analyses focused on the 799 households with purchase data in both periods. To examine variation in responses, additional data on primary shoppers’ educational attainment and nutritional knowledge were collected through an add-on survey in March 2023. Nutritional content data for 1949 beverage products was compiled from government nutrient databases, manufacturer and retailer websites, and in-store audits conducted by the research team.

PURPOSE

The study aimed to evaluate the real-world effectiveness of Singapore’s NG beverage labeling policy using a pre– post quasi-experimental design. Unlike most prior studies, which focused on short-term or laboratory-based effects,^4,5 this study examined both sustained consumer behavior and industry responses, including reformulation and product reassortment. This study addressed two unresolved gaps: first, there is limited policy-scale evidence on mandatory, nutrient-specific front-of-pack labels in Asian contexts; and second, few evaluations jointly quantify consumer purchasing changes and producer adaptation over time under a mandatory labeling regime.⁶ By leveraging longitudinal scanner data and fixed-effects regression models, the study provided rare, policy-relevant evidence from an Asian context on the population-level impact of mandatory nutrient-specific labeling. We hypothesized that, relative to a counterfactual projected from prepolicy trend, NG implementation would reduce household sugar purchased and shift purchases toward healthier beverages (grades A and B), and that manufacturers would respond via reformulation and reassortment to lower free sugar and saturated fat across products.

EVALUATION AND ADVERSE EFFECTS

For the demand-side analysis, the primary outcome was the daily average total sugar (in grams [g]) purchased per household. This was calculated by aggregating the sugar content of beverages purchased at the household-month level and dividing it by the number of days in each calendar month. Secondary outcomes included the daily average saturated fat (g) purchased per household, as well as the daily average volume (in milliliters [mL]) of healthier (grades A and B) and less healthy (grades C and D) beverages purchased.

For the supply-side analysis, key outcomes included the monthly average free sugar (g) and saturated fat (g) content per 100 mL of beverages purchased. Additionally, we examined changes in the proportion of purchased beverages containing sweeteners and changes in lactose content (g) per 100 mL.

Statistical Analysis

We employed fixed-effects linear regression with a binary variable for the policy period and its interaction with a linear time trend. This estimates postpolicy outcomes relative to a counterfactual projected from prepolicy trends. For household-level outcomes, fixed effects absorb time-invariant household characteristics. Only time-varying covariates, including a household size categorical variable (1–2, 3, 4, or ≥ 5 members) and an indicator for income (≥ Singapore $7001), were included in the model. Heterogeneous effects on the primary outcome were examined by stratifying models by (1) high household income (≥ Singapore $7001) versus otherwise, (2) minority race/ethnicity (Malay, Indian, others) versus Chinese, (3) high education (university and above) versus otherwise, and (4) high nutritional knowledge (median score and above) versus below median. For supply-side responses, fixed-effects estimation was conducted separately for stand-alone producer effects (changes among unique products) and net effects (results weighted by purchase quantities). The same analyses were conducted for all beverages combined and, separately, for dairy beverages. Additionally, to minimize overlap with early COVID-19 policies, we reestimated models using a stricter prepolicy period (April 2019–January 2020), with unchanged results (Appendix II).

Consumer Behavior

The NG policy shifted demand toward healthier beverages and reduced household sugar purchases (Figure 1). Compared with a counterfactual scenario without the policy, the NG policy was associated with a 3.11 g per day reduction in household total sugar purchases (−18%; 95% confidence interval [CI] = −4.99, −1.23) and no statistically significant reduction in saturated fat (−6%; 95% CI = −0.53, 0.25). The sugar reduction was both immediate and sustained over 12 months. Purchases of less healthy beverages (grades C and D) declined by 90 mL per day (−44%; 95% CI = −117.00, −63.75) and purchases of healthier beverages (grades A and B) increased by 79 mL per day (+82%; 95% CI = 59.05, 99.10), aligning with Singapore’s guidance to prefer water and lower or no-sugar beverages.⁷ This pattern suggests that consumers did not simply replace unhealthy beverages with healthier ones but may have increased their overall consumption of healthier beverages, potentially diluting health benefits if overall energy intake also rose.

FIGURE 1—

Mean Differences (and Percent Changes) in Nutrient Content and Volume of Household Purchases Between Postpolicy Purchases and Counterfactual Scenario Postpolicy Prepackaged Beverage Purchases for (a) Sugar, (b) Saturated Fat, and (c) Beverages: Singapore, April 2019–March 2024

Note. Overall model-estimated differences are derived from fixed-effects models comparing postpolicy nutrient content or volume of purchases to counterfactual postpolicy nutrient content or volume of purchases based on prepolicy trends. Purchase data were provided by Nielsen HomeScan data. Error bars indicate 95% confidence intervals. The x-axis of the left and middle graphs shows the number of months after Nutri-Grade (NG) policy implementation. On the x-axis of the right graph, NG A/B represents healthier beverages with a grade of A or B, and NG C/D represents less healthy beverages with a grade of C or D. The y-axis represents the amount of nutrient (in grams) or volume of beverage purchases (in milliliters) per household per day.

Heterogeneity

Relative to the counterfactual, policy effects were larger among higher-income, Chinese households than among minority households, where changes were small or statistically insignificant (Appendix I Table B). Effects differed by nutritional knowledge rather than education, highlighting the need for targeted strategies to strengthen nutrition literacy beyond Singapore’s “My Healthy Plate/Nutri-Grade” guidelines and school-based initiatives.^7,8

Supplier Reformulation and Reassortment

Manufacturers reduced free sugar and saturated fat but increased sweetener use and lactose, with evidence of threshold-targeting. Analysis of unique products across time periods showed that the average free sugar content declined by 1.16 g per 100 mL (−21%; 95% CI = −1.17, −1.16) and saturated fat by 0.10 g per 100 mL (−19%; 95% CI = −0.10, −0.09; Figure 2). Meanwhile, the share of products containing sweetener rose by 6% (95% CI = 0.061, 0.062), and lactose content increased by 0.15 g per 100 mL (34%; 95% CI = 0.152, 0.154), suggesting that manufacturers substituted free sugar with alternative sweet-tasting compounds to avoid label penalties. Products clustering just below NG thresholds (e.g., 5 g and 10 g free sugar) further indicate strategic reformulation (Appendix I Figure B). Dairy beverages showed particularly large reductions in sugar (−30%; 95% CI = −1.69, −1.67) and a modest decline in saturated fat (−9.6%; 95% CI = −0.12, −0.11), but also sharp increases in lactose (+78%; 95% CI = 1.22, 1.23).

FIGURE 2—

Mean Differences (and Percentage Changes) in Nutrient Profile Between All Unique Beverages: Singapore, April 2019–March 2024

Note. Overall model-estimated differences are derived from fixed-effects models comparing the 12-month postpolicy nutrient profile of all unique beverages to the counterfactual postpolicy nutrient profile of all unique beverage purchases based on prepolicy trends. Purchase data were provided by Nielsen HomeScan data. Error bars indicate 95% confidence intervals.

Net Effects Weighted by Volume Purchases

Purchase-weighted changes amplified sugar reductions but not saturated fat (Appendix I Figure C: free sugar, –1.5 g per 100 mL [−30%]; saturated fat, −0.1 g per 100 mL [−12%]). Among dairy beverages, free sugar fell 2.7 g per 100 mL (−63%), but saturated fat rose 0.1 g per 100 mL (+8%) and lactose rose 2.6 g per100 mL (+199%). This reflects persistent demand for full-fat fresh milk (because of its taste and creaminess,⁹ although it is not recommended for those older than two years¹⁰), which limited saturated fat impact. These findings underscore that consumer preferences ultimately shape the reach and effectiveness of labeling policies.¹¹

Adverse Effects

Despite its overall effectiveness, the policy revealed several unintended consequences:

• •Threshold-targeting near NG cutoffs⁶;
• •Increased use of sweeteners and lactose¹²: because grades penalize free sugar but not nonsugar sweeteners, suppliers can improve grades by replacing sugar with sweeteners—consistent with the observed rise in sweetener use;
• •Equity concerns (smaller or null effects in some groups).¹³

SUSTAINABILITY

A mandatory, nutrient-specific label can sustain demand- and supply-side adjustments while preserving autonomy. Preparation and public communication facilitated compliance, and continued point-of-purchase exposure reinforced behavior change. In Singapore, NG was paired with public communications and a ban on grade D advertising across online and offline media. Sustaining impact will require tightening marketing restrictions on unhealthy foods, deploying targeted equity measures, and conducting regular, evidence-informed monitoring of product composition.

PUBLIC HEALTH SIGNIFICANCE

Relative to a counterfactual projected from prepolicy trends, NG produced measurable, policy-scale changes in both purchasing and product composition, addressing our core research question in a real-world setting. Two limitations merit note: our data capture household beverage purchases for home consumption (out-of-home purchases are not observed), and fixed-effects estimates do not fully control for unobserved time-varying factors that may influence purchasing behaviors. Even with these limitations, our findings indicate that mandatory labeling can meaningfully reduce sugar purchased through both consumer behavior changes and industry reformulation. Importantly, this study highlights the complexity of food-industry adaptation, including unintended consequences such as manufacturers’ substitution of free sugar with sweeteners. Many front-of-pack labels—including Nutri-Grade and Nutri-Score—do not constrain nonsugar sweeteners. Labels should either build sweetener controls into the algorithm (e.g., disclosure plus penalties or thresholds, or overall “sweetness” caps) or be paired with complementary restrictions and active monitoring. The observed disparities in consumer response also underscore the importance of tailoring public health policies to reduce, rather than reinforce, existing inequities in diet-related disease risk.¹⁴ For settings weighing front-of-pack labels (e.g., the United States), sugar gains alongside substitution call for adopt-if-robust designs—explicit sweetener controls with equity-focused oversight and evaluation. Taken together—and consistent with the World Health Organization’s Global Action Plan for noncommunicable diseases (2013– 2030)¹⁵—Singapore’s experience with NG suggests that well-designed labeling policies, when supported by equity-focused measures, regulatory oversight, and ongoing evaluation of sustained responses, can play a critical role in improving nutritional environments while preserving consumer autonomy.