The effects of an abrupt increase in taxes on candy and soda in Norway: an observational study of retail sales.

BACKGROUND: Fiscal policies are used to promote a healthier diet; however, there is still a call for real-world evaluations of taxes on unhealthy foods and beverages. We aimed to evaluate the effect of an abrupt increase, of respectively 80 and 40%, in the excising Norwegian taxes on candy and beverages on volume sales of candy and soda. We expected sales to fall. METHODS: We analyzed electronic point of sale data covering approximately 98% of volume sales of grocery stores in Norway. In two pre-registered models with weekly (log-)sales of taxed candy and soda from 3884 individual stores, we modeled the difference between the jump (discontinuity) in the trend around the time of the increase in taxes and the corresponding jump in the trend in a control season from the previous years (Model 1). In addition, we modeled the difference between the intervention and the control season in their changes in average sales (Model 2). RESULTS: Model 1 showed a 6.1% (one-sided 95% CI: not applicable (NA), 23.4, p-value = 0.26) increase and a - 3.9% (95% CI: NA, 4.9, p-value = 0.23) reduction in the differences in the jump in the trends, for candy and soda, respectively. The second model showed a relative decrease of - 4.9% (95% CI: NA, 1.0, p-value = 0.08) in the average sales of candy and an increase of 1.5% (95% CI: NA, 5.0, p-value = 0.24) in sales of soda. Supplementary analyses suggested that the results were sensitive to clustering on the time dimension. CONCLUSIONS: When using two different quasi-experimental designs to model changes in volume sales of taxed candy and soda, we were not able to detect reductions in sales that coincided with an increase in the taxes. Variation across time makes it difficult to detect potentially small changes in sales even when using an entire country's worth of sales data on the level of individual stores. We speculate that the tax increases were too modest to affect the prices to alter sales sufficiently.

How much of the variation in mortality across Norwegian municipalities is explained by the socio-demographic characteristics of the population?

The goal was to find out whether much of the variation in mortality between the 430 Norwegian municipalities could be attributed to socio-demographic characteristics of the population - operating through individual- or aggregate-level mechanisms. Two-level discrete-time hazard models were estimated for women and men at age 60-89 in 2000-2008, using registers covering the entire population. Year, age and a municipality-level random term were included in the first step. When socio-demographic characteristics of the individual and others in the municipality were added, the variance of the random term was reduced by 73-80% almost exclusively because of aggregate-level effects. Policy implications of these findings are discussed.