White Hat Bias

Recently, I came across a new term—White Hat Bias (WHB). Although the article that introduced it was published back in 2010, I had never encountered it before.

The term was coined by Cope and Allison in their paper “White Hat Bias: Examples of Its Presence in Obesity Research and a Call for Renewed Commitment to Faithfulness in Research Reporting” (2010). They defined WHB as "bias leading to distortion of research-based information in the service of what may be perceived as righteous ends".

The authors explore this type of bias using two obesity-related topics:

The consumption of nutritively sweetened beverages (NSBs) as a presumed risk factor.
Breastfeeding as a potential protective factor.

As an example, they examined how two key studies on these topics were cited in scientific literature up to 2008 (when their analysis was conducted):

James et al., 2004 – Preventing Childhood Obesity by Reducing Consumption of Carbonated Drinks: Cluster Randomized Controlled Trial (195 citations).
This study investigated an intervention aimed at reducing children’s consumption of sweetened beverages and whether it led to lower obesity rates. The authors analyzed data using two approaches: Dichotomous (overweight/obese vs. normal weight) and
Continuous (change in Body Mass Index, BMI)
They found no significant change in BMI after 12 months. However, in the dichotomous analysis, the percentage of overweight and obese children increased by 7.5% in the control group, while in the intervention group, it decreased by just 0.2% (mean difference: 7.7%, 2.2–13.1%).
Ebbeling et al., 2006 – Effects of Decreasing Sugar-Sweetened Beverage Consumption on Body Weight in Adolescents: A Randomized, Controlled Pilot Study (45 citations).
This randomized experiment examined whether reducing sweetened beverage consumption led to weight loss. The primary analysis found no significant differences between the control and intervention groups. However, subgroup analysis revealed that among participants with a high baseline BMI, weight changes in the intervention group differed significantly from the control group.

How These Studies Were Misrepresented

The authors analyzed how other scientific papers cited these two studies and found that:

84.3% (James et al.) and 66.7% (Ebbeling et al.) of citing papers misrepresented the results, exaggerating the impact of reducing NSB consumption on obesity.
Only 12.7% and 33% of papers accurately conveyed the overall findings.
Almost no papers (3.5% for James et al., 0% for Ebbeling et al.) downplayed the significance of the results.

In other words, the majority of scientific articles overstated the harmful effects of NSBs on obesity—even when the data did not support a clear conclusion. Why? The authors suggest that researchers may have been driven by a desire to emphasize the dangers of these drinks for the public good.

Distortions in Press Releases and the Media

The study also examined how research findings are distorted in press releases from universities and medical centers. For example:

The press release for Ebbeling et al. claimed that “a simple beverage-focused intervention led to weight loss” but failed to mention that the main analysis found no statistically significant difference between control and experimental groups.
The press release for James et al. stated that “discouraging children from drinking fizzy drinks can prevent excessive weight gain”—even though the study did not analyze weight changes per se, and that effect on BMI was statistically insignificant.

The exaggeration intensified when findings moved from press releases to mainstream media. Compare:

UCLA press release:

“Research released today provides the first scientific evidence of the potent role soda and other sugar-sweetened beverages play in fueling California’s expanding girth.”

Media coverage:

“For the first time, we have strong scientific evidence that soda is one of the—if not the largest—contributors to the obesity epidemic.”

Another Example: Breastfeeding and Obesity

The article also examined how breastfeeding is presented as a protective factor against obesity. The authors analyzed a WHO report on the subject and found that:

Selective reporting was used—only the strongest results supporting a link between breastfeeding and lower obesity risk were included.
Weaker or less impressive findings were excluded without explanation.

The Term "White Hat Bias" in Other Fields

The term White Hat Bias was later used in a 2013 study (Young & Xia, “Assessing Geographic Heterogeneity and Variable Importance in an Air Pollution Data Set”). The authors examined distortions in air pollution risk analysis and found that:

The impact of air pollution on life expectancy was minimal in the eastern U.S. and nonexistent in the west.
The health risks of smoking were much greater than those of air pollution.

Current Citations of These Studies

Curious about how these obesity studies are cited today, I checked Google Scholar:

Ebbeling et al., 2006 now has 822 citations.
James et al., 2004 has 1,250 citations.

Some newer studies cite these works correctly. For example, a 2013 meta-analysis on soft drinks and obesity reported:

“For the study by James et al. (2004), although the difference in BMI change did not reach significance, there was a significant difference in the prevalence of childhood overweight and obesity between intervention (0.2% reduction) and control clusters (7.5% increase).”

But misrepresentation still persists. A 2022 study (Prevalence of Childhood Obesity in the United States in 1999–2018: A 20-Year Analysis) cited James et al., 2004 to support this claim:

“Inculcating a correct perception of childhood obesity and reducing the high consumption of nutrient-rich carbonated drinks can help to reduce obesity.”

Again, we see a slight exaggeration—strengthening the study’s conclusions to reinforce a socially beneficial message.

Why Isn't White Hat Bias Discussed More?

Despite its relevance, White Hat Bias is rarely mentioned in discussions about research distortions. Why?

Is there White Hat Bias in talking about White Hat Bias? Maybe people avoid discussing it to protect the greater good?
Or perhaps the reason for misrepresentation doesn’t matter—whether you distort results to confirm your hypothesis, strengthen your argument, or promote public health, a distortion is still a distortion.