IHP 340 Module 2 Article Analysis Short Paper example

A Penny an Ounce and a P Value of .046: Reading the Statistics in the Berkeley Soda Tax Study

[Student Name]

Southern New Hampshire University

IHP 340: Statistics for Healthcare Professionals

Module Two Short Paper

[Instructor Name]

[Date]

A Penny an Ounce and a P Value of .046: Reading the Statistics in the Berkeley Soda Tax Study

The Research Question

Sodas, sweetened teas and sports drinks supply much of the added sugar Americans consume, and heavy intake is associated with obesity, type 2 diabetes and tooth decay. In March 2015, Berkeley, California, became the first city in the United States to tax distributors one cent per ounce on these drinks. Falbe et al. (2016) asked whether consumption of sugar-sweetened beverages fell in low-income Berkeley neighborhoods after the tax, compared with similar neighborhoods in Oakland and San Francisco, where there was no tax. The comparison cities matter: without them, any decline in Berkeley could have reflected a regional trend rather than the tax.

Design and Sample

The study used a repeated cross-sectional design. Interviewers stopped people at busy intersections in low-income neighborhoods and administered a beverage frequency questionnaire to 990 participants before the tax and 1,689 after it, about four months after the tax took effect (Falbe et al., 2016). Because different people were interviewed at each time, this is not a cohort study, which would follow the same people. The design can show whether consumption in a population changed, but not whether any individual drank less. Recognizing that the before and after groups are different people is the single most important step in reading this study correctly. If, for example, the post-tax sample happened to include more health-conscious people, part of the change could reflect who was sampled rather than what they drank.

The sample was a convenience sample in public places rather than a random sample, which limits how confidently results can be generalized to all low-income residents. Its strength is that the same methods were used in Berkeley and the comparison cities at both times.

Variables and Levels of Measurement

The table below lists the main variables.

Table 1

Main Variables in Falbe et al. (2016)

The Statistics Reported

The main result is reported as percent change: consumption of sugar-sweetened beverages decreased 21 percent in Berkeley and increased 4 percent in the comparison cities, and the difference between those changes had a P value of .046. Water consumption rose 63 percent in Berkeley and 19 percent in the comparison cities, a difference with P below .01 (Falbe et al., 2016).

Three points help interpret these figures. First, the percentages are relative changes in how often people drank, not changes in ounces or calories, so a 21 percent drop does not tell us how much sugar was avoided. Second, the P value of .046 means that if the tax truly had no effect, a difference this large between the cities would appear by chance less than 5 percent of the time. It falls just under the conventional .05 cutoff, so the result is statistically significant, but narrowly; a slightly different sample could have produced a P value above .05. Third, statistical significance says nothing about size. A reader must look at the percentages themselves to judge whether the change matters for health. The American Statistical Association has warned against exactly this kind of overreliance, stating that a P value does not measure the size of an effect or the importance of a result and that scientific conclusions should not rest only on whether a P value passes a threshold (Wasserstein & Lazar, 2016). Reporting the percent changes alongside the P value, as the authors did, is the better practice.

A Contrasting Study

A second evaluation of the same tax used different methods and reached a different conclusion about consumption. It analyzed supermarket scanner data covering millions of checkouts and found that sales of sugar-sweetened beverages in Berkeley stores fell 9.6 percent in the first year relative to what would have been expected without the tax, while sales rose in comparison stores. But in its telephone survey of Berkeley adults, the reduction in self-reported daily intake of these drinks, about 20 percent, was not statistically significant, with a P value of .49 (Silver et al., 2017). The authors noted that Berkeley residents drank far fewer sugary beverages than the national average even before the tax, so a small absolute change was hard to detect against a high standard error.

The two studies do not necessarily contradict each other. One sampled low-income neighborhoods, where baseline consumption was higher; the other surveyed a broader population with low consumption. One measured frequency; the other measured grams and calories. Their different results illustrate how the population, the measure and the sample size shape what a study can detect.

Limitations

Falbe et al. (2016) relied on self-reported frequency, which may be affected by recall and by participants' awareness of the tax. The post-tax survey was only four months after implementation, too soon to know whether changes lasted. The convenience sampling and repeated cross-sectional design mean that differences between samples could account for some of the result. And the study measured behavior, not health outcomes such as weight or diabetes.

What a Health Manager Should Take Away

For a public health manager, the study provides early, statistically significant evidence that a beverage tax was followed by lower consumption in low-income neighborhoods, supported by sales data from a second study. It does not show how much sugar was avoided or whether health improved, and its P value sits close to the threshold. The appropriate conclusion is cautious support: the tax appears to change purchasing and drinking behavior, and longer studies with health outcomes are needed before claiming health benefits.

Conclusion

Reading the Berkeley study statistically means identifying a repeated cross-sectional design with a comparison group, recognizing ratio-level outcome variables reported as percent changes, and interpreting a P value of .046 as narrowly significant without implying a large effect. Comparing a second study shows how different populations and measures can produce different significance results for the same policy, which is a lesson that applies to any study a healthcare professional reads.

References

Falbe, J., Thompson, H. R., Becker, C. M., Rojas, N., McCulloch, C. E., & Madsen, K. A. (2016). Impact of the Berkeley excise tax on sugar-sweetened beverage consumption. American Journal of Public Health, 106(10), 1865-1871. https://doi.org/10.2105/AJPH.2016.303362

Silver, L. D., Ng, S. W., Ryan-Ibarra, S., Taillie, L. S., Induni, M., Miles, D. R., Poti, J. M., & Popkin, B. M. (2017). Changes in prices, sales, consumer spending, and beverage consumption one year after a tax on sugar-sweetened beverages in Berkeley, California, US: A before-and-after study. PLOS Medicine, 14(4), Article e1002283. https://doi.org/10.1371/journal.pmed.1002283

Wasserstein, R. L., & Lazar, N. A. (2016). The ASA statement on p-values: Context, process, and purpose. The American Statistician, 70(2), 129-133. https://doi.org/10.1080/00031305.2016.1154108