The effects of taxes on purchases of sugar-sweetened carbonated soft drinks: a quantile regression approach

Abstract

The consumption of large quantities of Sugar-Sweetened Carbonated Soft Drinks (SSCSD) may lead to obesity, which is associated with health problems such as diabetes, cardiac diseases and mental health problems. The effects of increasing the Value Added Tax (VAT) on purchases of SSCSD are estimated. Obesity is more likely among heavy drinkers than among light or moderate drinkers. Therefore, the effects on high- and low-purchasing households are estimated by using Quantile Regressions (QRs) on Norwegian household purchase data. Since many households did not purchase SSCSD during each survey period, censored as well as ordinary QRs are used. Our results suggest that a VAT increase from 13 to 25% will have the highest percentage effect among low-purchasing households but the absolute effect is highest among high-purchasing households. Low-purchasing households will reduce their purchases by about 5 L while the reduction is almost 20 L among high-purchasing households. However, the effects among high-purchasing households are not statistically significant from zero. A reduction of 5 L corresponds ceteris paribus to an annual reduction of about 0.3 kg of body weight.

Acknowledgements

The Research Council of Norway, Grant no 134018/110, sponsored this research. The authors thank Liselotte Schäfer Elinder for help with the nutritional calculations.

Notes

¹ In the household consumption expenditure survey data, the per capita purchases of each household are multiplied by 26 to approximate annual per capita consumption. The wholesale data from the NBA are annual data. Purchases away from home may explain most of the differences in mean purchases reported by the two data sources, but some underreporting is also likely in the expenditure surveys.

² The Tobit model is used as our reference model. This model is frequently referred to as a censored model. In some applications, including ours, the optimal choice may be a corner solution. As pointed out by Wooldridge (2002, p. 518), it may be more appropriate to term the resulting model a ‘corner solution model’ rather than a censored model. Furthermore, zero purchases may arise for several reasons. First, in survey data with short recording periods, zero purchases may arise because of infrequency of purchases. Second, some households may not purchase SSCSD because they do not like the product. Third, zero purchases may arise because of standard corner solutions that result in a Tobit model. We do not have information that enables us to identify the main reason for the reported zero purchases in our sample.

³ A demand-system framework would allow for different price and income effects for different soft drinks and could (given available data) incorporate substitution effects between SSCSD and, for example, artificially sweetened soft drinks, bottled water, fruit juices, and fluid milk. However, across-equation restrictions such as symmetry cannot be imposed on QR models. Therefore, a system framework is less meaningful and a single-equation model is estimated.

⁴ The head of the household is defined as the household member with the highest income. Note, however, that household income data are unavailable.

⁵ For households surveyed on days in two different months, we used a weighted average of the index for those 2 months. The numbers of survey days in each month were used as weights.

⁶ The censoring is due to Buchinsky's algorithm as described above. In this algorithm, all observations are used to calculate the predicted values. Observations associated with negative predicted values are deleted and the model is re-estimated on the trimmed sample.

⁷ The pseudo R ² value for the Tobit model is calculated in a different way than are the pseudo R ² values for the QRs. Hence, comparisons are not meaningful and we do not report the Tobit model's pseudo R ² value.

⁸ Details concerning the calculation of the t-tests and the detailed results of the tests are available from the authors upon request.