McDonald's 2022
Due process. Among other things, it means that when someone's accused of a crime, the accusation must be addressed fairly, with full respect for that person's legal rights, even if they're a known criminal.
I thought about due process when I saw a study, published last month, that accuses McDonald's of using Instagram to disproportionately target children in poorer countries. Although I tend to think of McDonald's as a blight on humanity, this particular study is deeply flawed, statistically speaking. I don't think it shows that McDonald's has done anything wrong.
In this newsletter I will follow the principle of due process in defending McDonald's against this new study. But I won't be able to sleep tonight if I stop there. I will also present some stats showing that the company is far from being a benign provider of food and fun, and I'll close with a moderately cynical discussion of the company's plans for 2022.
The Mcdonald's Instagram study
The study in question has all the trappings of credibility. The journal (British Medical Journal (BMJ) Nutrition, Prevention, and Health) is peer-reviewed and highly esteemed among medical researchers. The authors are all NYU professors in the schools of population and public health. And, in many respects, the study methodology was strong.
For this study, the researchers examined McDonald's official Instagram posts in 15 countries. All of the posts were ads. (Apparently I'm the last person in America to realize that major corporations maintain Instagram accounts solely for the purpose of advertising.) Since fast food companies are known to adjust their marketing strategies according to country, the NYU researchers decided to take a close, intensive look at how the largest of these companies market their products differently according to a country's affluence.
The 15 countries were randomly chosen from three World Bank categories: high-income (Australia, Canada, Panama, Portugal, UAE, UK, USA), upper-middle-income (Brazil, Lebanon, Malaysia, Romania, South Africa), and lower-middle-income (Egypt, Indonesia, India). The NYU team examined all of McDonald's official Instagram posts in each country between September 1, 2019 and December 31, 2019. The content of each post was coded on dimensions such as whether or not the ad was targeted to children, whether or not it included price promotions, whether or not it promoted healthy habits, and so on.
Findings
During the time period under study, McDonald's posted a total of 849 ads to its more than 10 million Instagram followers across the 15 countries. Content analyses of these 849 ads revealed the following:
—More ads were posted in poorer countries
—More ads targeting children were posted in poorer countries
—More ads for price promotions were posted in poorer countries
—More ads promoting healthy habits were posted in high-income countries
"Poorer" countries is the researchers' term for those classified by the World Bank as "lower-middle-income" (Egypt, Indonesia, India), which is to say they're poor countries, but not the poorest in the world.
Two well-documented facts provide a context for the findings: (a) Diet-related chronic illnesses are more prevalent in poorer countries, and (b) consumption of fast food promotes diet-related chronic illnesses. Thus, the researchers concluded that by targeting poorer countries and their children, "McDonald’s social media ads may exacerbate healthcare issues in the most vulnerable countries in the world."
I admire studies like this, because the researchers call out public health risks without getting sanctimonious. They don't accuse McDonald's of callousness or greed. They don't mention Eric Schlosser ("Fast Food Nation") or Morgan Spurlock ("Super Size Me"). They simply link McDonald's marketing strategies to the affluence of each country where it maintains a social media presence.
I also agree with the authors' general conclusion – i.e., that McDonald's jeopardizes public health in poorer countries. However, I think this conclusion is more clearly demonstrated by other studies than by this one. Sad to say, the NYU study has awful statistical flaws. And so, in the interest of due process, I will take a moment to defend McDonald's against the researcher's findings. After that, I'll turn to other, more incriminating data.
Limitations
As this study is the first to link McDonald's social media activity to a country's affluence, I tried to think of the study as exploratory and not hold it to the highest standards of methodological rigor. Even so, the flaws are easy to spot. (It's like coming home and finding Ronald McDonald in your kitchen. You can't help noticing...)
1. Uneven sampling
The sample was small (15 countries) and unevenly distributed across country type. For example, there were seven high-income countries, but only three poorer ones. Statistically speaking, the consequence is that each of the poorer countries (Egypt, India, Indonesia) was disproportionately represented in the dataset. This doesn't invalidate the findings, but it reminds us that any conclusions we draw may only pertain to Egypt, India, and Indonesia, rather than all countries that could be called "lower-middle-income" or "poorer."
2. Poor reliability
Although some features of an ad may be indisputable (e.g., whether or not it contains a price promotion), others are more subjective and therefore a matter of interpretation. The researchers used a standard approach for dealing with subjectivity: They asked more than than one person to rate the ads on each dimension, then calculated the extent of agreement across raters. This type of agreement is called inter-rater reliability, and it can range from 0% to 100%. Ordinarily, data analysis won't proceed unless inter-rater reliability is high.
The researchers asked five individuals to rate the ads. Most of the ratings were dichotomous (each ad was rated as targeting or not targeting children, promoting or not promoting healthy habits, etc.). However, inter-rater reliability only exceeded 90% for two of the raters. The researchers discarded the ratings from the other three raters, and, for their final analyses, they alternated between ratings provided by the two raters who were in close agreement. This is amazing and wrong. If, out of five people, only two agree on what they're looking at, there's a problem, and data analysis should not proceed. Perhaps the instructions given to the raters were unclear. Perhaps the variable is too subjective. Other explanations are conceivable. But most experts would agree that forging ahead with data analysis is a terrible idea.
I suspect the problem here is that the researchers' approach to measurement wasn't sufficiently nuanced. For example, some ads clearly target children and some don't, but others don't seem to fit this neat dichotomy. Does the ad pasted in below target children? Well, maybe. If I had to say something definitive, I would argue that it's child-friendly, because it's a cartoon, but it doesn't specifically target children. (I don't know...maybe some children would disagree with me?)
Bottom line: I don't trust the researcher's data. That is, I don't trust the descriptive stats they provided (e.g., numbers of child-targeted ads). They gave no explanation for why most pairs of raters couldn't reach 90% agreement on what they were looking at, and they didn't indicate how much lower than 90% each pair was. (There’s nothing magical about that 90% figure, by the way; inter-rater reliability above 80%, for example, is often considered strong. The problem here is we just don’t know what the stats were for the majority of the pairs of raters, except that they agreed less than 90% of the time.)
3. Lack of significance testing
Both the researchers and media reports emphasized the quantitative findings. For example, the researchers found that McDonald's posted 154% more ads in poorer countries than in high-income countries. This would be troubling, except that (a) the actual difference between countries, for the four-month period studied, was an average of 38 posts vs. 28 posts (i.e., a difference of only about 1 post every 12 days), and (b) the researchers didn't do significance testing.
In other words, the researchers reported a small difference, and we don't know whether it's statistically significant or not. (A "significant" finding is one that a researcher presents as trustworthy, based on statistical tests suggesting a very low probability that the finding is a fluke. If you're a stats person, you may not like what I just wrote, but you'd probably agree that this is the way most researchers treat significance.)
Another finding – the one that got the most media attention – was that McDonald's ads disproportionately targeted children in poor countries. 22% of the ads in poorer countries, versus 14.5% in upper-middle-income countries and 12.4% in high income countries, were identified as targeting children. I find this result a little underwhelming. Although it's clear that McDonald's selectively markets to children, in this particular study we only see about a 10% difference in targeting between the poorest and richest countries, and we don't know whether the difference was significant or not.
Bottom line
It's not clear to me, from this study, that McDonald's disproportionately targets children in the Instagram ads it posts to poorer countries. I'm not convinced by any of the researchers' claims about adjustments to marketing in poorer countries. So, to be blunt, I don't think differently about McDonald's than I did before I read this study.
McStats
Even when stats are calculated and interpreted correctly, they can be misleading if there are many to choose from. For a large company like McDonald's, you could marshal statistical support for almost any story. For example, consider the stats on McDonald's charitable contributions: 368 Ronald McDonald Houses, where families can stay for free near their hospitalized children. 50 Ronald McDonald Care Mobiles that bring health care to underserved communities. Scholarships for ethnic minorities, including half a million dollars per year for students attending HBCUs. And, during the pandemic, over 12 million meals donated to frontline workers. The company may be villainous, but it's not a cartoon villain.
When I think of McStats, what comes to mind illustrates three themes: McDonald's is everywhere, it's bad for you, and it harms the environment. (I freely admit to cherrypicking these themes as well as the accompanying details.)
1. Ubiquity
McDonald's is inescapable. Here are a few stats that illustrate the point:
—For over a decade, McDonald's has served an average of 50 to 70 million people per day worldwide, with only about a 25% reduction in sales during the pandemic thus far.
—If McDonald's were a country, it would have a larger economy than half the countries in the world. (Specifically, it would rank 90th out of 195, thanks to annual revenues exceeding 20 billion dollars for the past 15 years.)
—McDonald's is the world's largest distributor of toys (over 1.5 billion per year – more even than Toys-R-US) and it's the largest private owner of playgrounds in the U.S.
—In the U.S., there are more McDonald's restaurants than there are hospitals (about 14,000 vs. about 10,000).
2. Health
Eating at McDonald's on more than rare occasions is bad for your health. The stats that demonstrate this are endless. Here's how I simplify things:
(a) Individual-level data
When you eat at McDonalds, you get a lot of calories, fat, sodium, and sugar, but not much nutrition. Admittedly, the stats we could use to define "a lot" and "not much" are subject to false precision, because there's so much variety in what you could order at McDonald's and how much you eat of it – and you have a range of options for what you eat the rest of the day when you're not there. Also, it's clear now that what you choose to eat McDonald's is at least as important as the fact that you're eating there in the first place, because there’s huge variability in the nutritional content of different menu items. (See Appendix 1 for more details.) I suggest that if you eat at McDonald’s regularly, you should consult charts like this, consider various nutritional guidelines, and choose carefully.
(b) Population data
Studies have shown a strong correlation between the number of McDonald's restaurants available to people and obesity rates. For example, check out the following graph drawn from a study that examined 75% of the world's population:
Roughly speaking, this graph suggests that the more McDonald's restaurants there are per capita, the greater the obesity rates. Of course, places with more McDonald's probably also have more fast food restaurants generally, along with other differences that contribute to what you see in the graph. But the strength of the correlation is startling. (That R2 value of 0.948 in the upper right corner of the graph tells you that almost 95% of the variability in obesity rates per million people can be predicted by the number of McDonald's restaurants where they live. In other words, if you only know how many McDonald's restaurants there are in a particular region, you can predict the obesity rates for that region with a high degree of accuracy.)
3. Other
I don't mean to downplay other examples of McEvil by lumping them into an "other" category; I just don't want to wear out your patience. As discussed in Appendix 2, concerns about work conditions and environmental impact have also been statistically documented.
McDonald's 2022
Over the years, McDonald's has made many changes in response to criticisms of its impact on health and environment. Although the cynical view of these changes is that they’re just good PR, the changes McDonald’s has been making are beneficial, regardless of what motivates them. Here are three on the way in 2022:
1. Safer packaging
In 2022, McDonald's will continue to remove toxic chemicals from its food packaging. Removal of BPA, BPS, and phthalates was completed in 2021. Next to be removed are PFAS (per- and polyfluoroalkyl substances), the "forever chemicals" that turn up in our food and water and have been linked to a variety of serious health problems including cancer, hormonal disruptions, and impaired immune system functioning. McDonald's announced that it would begin removing PFAS from its packaging in 2022, with the goal of finishing by 2025. Many experts say that the company is moving too slowly on this issue. I would agree. Federal, state, and local government bans or regulation of PFAS have been in place since 2018, and the evidence of PFAS toxicity is clear.
2. Healthier Happy Meals
Last year, McDonald's promised that in 2022, more than half of its Happy Meals would contain less than 600 calories and 650 mg of sodium, with saturated fats and sugar each contributing less than 10% of total calories. That’s progress, although the cynical interpretation is that McDonald's is saying, in effect: "We're going to take some of the unhealthy crap we feed your kids, and make it it a little less unhealthy." Anyway, I'm not sure how impressed we should be with the statistical goals. For instance, capping Happy Meals at 600 calories sounds worthwhile (that's 1/3 to 1/2 of the recommended daily intake for 4-8 year old girls, for instance), but what about the Happy Meals that aren't reduced below this value?
3. The McPlant
McDonald's has partnered with Beyond Meat to produce the McPlant, a vegan alternative to hamburgers. The McPlant was rolled out in the UK in October, and it's on the way to the U.S. this year. Excited?
The McPlant has gotten favorable reviews, and both public as well as environmental health would benefit if it catches on. But I think we can expect some resistance from the carnivores among us. In addition, consumer research shows that when healthy food options become available, people actually become more likely to engage in indulgent ones. This research predicts that the mere presence of the McPlant on a menu might induce some people to order something more less healthy than they would've otherwise ordered!
And finally...
I couldn't find a place in this newsletter to share my favorite McDonald's anecdote, so I'll include it here: Queen Elizabeth owns a McDonald's. It's located in Oxfordshire, roughly 60 miles northeast of Buckingham Palace. (Chalk and cheese, as they say over there…)
Appendix 1: The McMonth
Many of you are familiar with the 2004 documentary "Super Size Me", in which Morgan Spurlock ate nothing but McDonald's for one month. By the end of the month, Spurlock had gained almost 25 pounds and developed physical and mental health problem. Less well-known are demonstrations such as the one conducted in 2018 by Nyree Dardarian, a Drexel University professor, who also ate nothing but McDonald's for a month. By the end of the month, Dardarian had gained only 1 pound, and she reported no differences in blood chemistry or state of mind (other than occasionally craving foods you can't get at McDonald's).
Why did Spurlock and Dardarian have such different McMonths? Presumably owing to the rules they chose to follow. Spurlock intentionally ate every item on McDonald's menu and supersized whenever he could; as a result, he took in about 5,000 calories per day (roughly twice the recommended amount for a person of his size), and he wasn't shielded from the unhealthiest items on the menu. In contrast, Dardarian limited her daily intake to 1,400 calories and chose judiciously.
These anecdotes remind us that what you choose to eat at McDonald's is at least as important as whether you choose to eat there in the first place. This is consistent with the huge variability in the nutritional composition of different menu items.
Appendix 2: McDonald's inside and out
Anecdotal evidence tells us that working in a McDonald's restaurant is unpleasant and sometimes unsafe, and workers across the world are becoming increasingly disgruntled about their “McJobs”. The pay is low, and the income gap between the average McDonald’s worker and its CEO has more than doubled in the past decade. (For example, in 2015, McDonald’s employees earned $7.73 an hour on average, while the the CEO earned $9,247 per hour.) As a corporation McDonald's maintains some distance from the working conditions in its restaurants, because it operates mainly on a franchise system. As the company notes on its own website, "at the end of 2020, we employed approximately 200,000 people worldwide, including our Company staff and crew at our Company-owned and operated restaurants. In addition, more than 2 million people worked at franchised McDonald’s restaurants globally." In other words, over ten times as many people work for McDonald's franchises as for the company itself.
As for environmental impact, the most pressing scientifically-established problems include plastic pollution (a large proportion of the 1.5 million tons of waste McDonald’s generates per year) and the environmental impact of reliance on beef (1 billion pounds per year, obtained from nearly 6 million heads of cattle), a problem Michael Pollan described compellingly in The Omnivore's Dilemma. McDonald’s maintains a set of web pages, entitled Our Planet, which describe the company’s commiment to environmental health. These pages are a masterpiece of mendacious PR that I will address in a separate newsletter later this year.