Who Steals Halloween Candy?
Halloween. The evening that 21% of adults pretend they're not at home, and 65% of parents admit to stealing candy from their children's haul.
Embarrassingly, I've done both. I'll probably burn in hell for it, surrounded by tiny goblins and witches singing Baby Shark off-key.
In this newsletter I want to share with you one of my favorite studies, carried out in Seattle on a chilly Halloween evening. It's a minor classic in social psychology, and it has a bit of everything: Clever methods, amusing anecdotes, ethically questionable procedures, and a message of importance for everyone.
The study was actually published in 1976. This is the first newsletter in which I'm not discussing new data.
Why delve into a nearly 50-year-old study? Not just because it's Halloween-themed and clever. This is one of those studies that reveals something of timeless importance about human beings, not to mention some of the not-so-obvious ways that statistics have shaped scientific research, for better and for worse.
The Halloween Study
The study was published in the Journal of Personality and Social Psychology. The lead author, Ed Diener, was a young social psychologist who became well-known in later decades for happiness-related scholarship.
For this particular study, Diener persuaded 27 homeowners in Seattle to help out. Here's how he and his co-authors described the setup:
"The entrance area in each of the 27 homes was arranged in the same basic pattern. Inside the front door, facing the entrance was a low table approximately 5 feet (1.5 m) long. On one side of the table was a large bowl full of bite-sized candy bars individually wrapped in brightly colored tissue paper. About 2 feet (.6 m) from the candy bowl rested a money bowl filled with pennies and nickels. Both bowls were periodically replenished during the evening. Within full view of both bowls was a decorative backdrop with a peep hole that camouflaged an unobtrusive observer."
You can guess where this is going. The "unobtrusive observer" was there to make note of which children stole candy and/or money.
Specifically, on Halloween evening, between 5 and 9 p.m., female research assistants were hiding behind the backdrop at each of the 27 houses in order to spy on the trick-or-treaters. The kids didn't know it, but Big Sister was watching! (This seems amusing or creepy, depending on your perspective.)
So, what was "Big Sister" looking for?
Hidden-observer studies often focus on individual characteristics, showing, for instance, that kids who cheat the most tend to be male, older, and more academically successful. (Watch out for those smart-looking teenage boys!)
Diener's study wasn't concerned with type of child who steals, but rather about the social conditions that promote stealing. Specifically, this was a study on deindividuation, the loss of personal identity that can occur in group settings.
Deindividuation
A person experiencing deindivduation temporarily forgets their individual characteristics and values while assuming the identity of a group.
We often equate this state of mind with an extreme loss of inhibition (mob violence, wild behavior at raves, etc.), but it can be as low-key as simply donning a uniform at work and acting like your colleagues – or stealing candy because that's what your friends are doing.
What causes deindividuation? Diener theorized that it's the presence of others, anonymity, and a lack of personal responsibility. A rioter smashes a store window because other rioters are present, the other rioters don't know him, and he feels no particular responsibility for the store window.
The purpose of Diener's study was to to test this theory. Are trick-or-treaters more likely to steal something when they're with others, anonymous, and have no responsibility for the theft?
Study procedures
While the hidden observer sat in silence behind the backdrop, a second research assistant greeted each individual trick-or-treater or group as they arrived, then told each child or group:
"You [or each of you] may take one of the candies. I have to go back to my work in another room."
With that, the child or children were left alone at the table, gazing through their little masks at a bowl of candy bars and, beside it, a bowl of pennies and nickels. (Tempting, right?) Children weren't told anything about the bowl of coins.
In this field experiment, there were three conditions, each testing one element of Diener's theory:
–Presence of others: Trick-or-treaters either came to the house alone or in groups.
–Anonymity: The second research assistant asked approximately half the trick-or-treaters for their names and addresses. The rest were not asked any personal questions.
–Diminished responsibility: Roughly half the time, when trick-or-treaters arrived in groups, the research assistant told one child in the group "I will hold you responsible if any extra candies are missing." Presumably, other members of the group would now feel less personal responsibility if they chose to steal something.
(Most of us would consider it intrusive, if not downright threatening, to ask kids for their names and addresses, or to hold them responsible for missing candy. Welcome to pre-1980s social science research. Studies like this couldn't be conducted now, as ethical standards for human experimentation have become much stricter.)
Main findings
Data analysis focused on 1,352 trick-or-treaters who showed up at the 27 homes.
That's a lot of kids – an average of 50 per home – but it underestimates the total number of trick-or-treaters, because data from children were excluded if they came to the door with a parent, came in groups of 7 or more, or showed up before 5 p.m. or after 9 p.m.
There were 414 "thefts" and – you guessed it – candy bars were most frequently stolen (65.4% of the time). 13.9% of the time children took money; 20.7% of the time they took both. Kids are smart: Most of them know that candy is way better than pocket change.
How much candy did the kids steal? Not much. Across the 27 houses, the average number of extra candy bars taken ranged from 1.6 to 2.3 per child. This leads me to one of the most delightful remarks you'll ever find in a peer-reviewed experimental research article:
"It appeared that if a child transgressed by taking extra candy, he took the number of extra candies that his hand would hold."
You can almost see a little hand cautiously dipping into the bowl...
Here are the main findings:
–Theft was more common among trick-or-treaters who showed up in groups rather than alone.
–Theft was more common among trick-or-treaters who were anonymous than among those who were asked for their names and addresses.
–Theft was more common among groups when one trick-or-treater in the group was deemed responsible for extra candies being taken.
This table shows the specific percentages of each group of children who stole something:
You can see that anonymity made a big difference, because the percentages in the left column are much bigger than the corresponding percentages in the right column.
You can also see that being in a group made a difference, because the percentages for the Group row are much bigger than the percentages above them in the Alone Row.
In addition, that 57.2% illustrates what's called an interaction effect. Being part of a group and anonymous led to more theft than you'd expect based on the percentages for each alone.
We could stop here and call this a pretty good demonstration of how people work. All sorts of antisocial behavior – from mob violence to the tragedy of the commons to stealing Halloween candy – can be attributed to groups of individuals who aren't personally identifiable.
However, there's one more important finding: By far the biggest effect comes from being in a group, being anonymous, and knowing someone else will be responsible for any rule-breaking. 80% of children in this group stole something.
What I find striking is not that theft was most prevalent among this group – I suspect most of us would predict that – but that the percentage was so high. Evidently, there are circumstances in which a clear majority of children will break the rules. (The same can be said for adults, according to other classic studies, such as the Milgram obedience to authority experiments.)
Ethical implications
Studies have shown that variables like gender, age, and personality influence whether people break the rules.
The Halloween Study illustrates that situational variables play a role too. Other research has documented these effects among adults. Sure, you choose your own behavior, and some people do tend to behave more badly than others, but whether a person breaks the rules or not is also influenced by social context.
In short, the world doesn't just consist of thieves and honest people. It includes folks who might steal under some circumstances but not others.
This kind of talk makes some people uncomfortable, because it seems to downplay personal responsibility for rule-breaking. If we blame circumstances, not people, for their crimes, it seems to follow that we shouldn't hold people accountable for those crimes.
I don't think that conclusion necessarily follows. You can acknowledge that bad behavior is caused in part by situational variables, yet still hold individuals accountable for what they do wrong.
Thinking this way allows us to be more compassionate toward rule-breakers, without constraining how strict or lenient we are in dealing with them.
I leave you with a hypothetical moral question: What would you do if you caught a trick-or-treater stealing Halloween candy – or, for that matter, a parent doing the same thing?
I won't offer advice here, other than to call for compassion. Some of that candy does look pretty tempting (unless it's circus peanuts or candy corn, which, according to an analysis published last month, are the Halloween candies people hate the most. Candy corn is truly awful; if you steal it, you're just mean, and eating it later will be your well-deserved karmic punishment!)
Happy Halloween, and thanks for reading!
Appendix: Scientific progress?
Almost 50 years later, the venue for the 1976 Halloween study (Journal of Personality and Social Psychology, or JPSP) continues to be one of the premier psychology journals, but the way researchers go about their business has radically changed.
What's striking about Diener's study, and many other classics in psychology, is what might now be called statistical minimalism. A lot of effort went into procedure and measurement (setup at 27 houses; observation of more than a thousand children), but the data were barely touched.
(Note for stats people: Diener just ran a few Chi-squares, piecemeal.)
Nowadays, in JPSP and other premier journals, you often see the opposite: Simple procedures and measures; intimidatingly complex statistics. (Stats people: We would say now that Diener's data call for multilevel logistic regression, since the outcome is dichotomous (steal vs. not steal) and the data are nested by house and, partly, by group.)
What happened?
In a word, computers. The emergence of PCs and statistical software programs radically transformed data analysis. From the 1980s on, as computational power increased and the programs became more sophisticated, statistical analyses in peer-reviewed studies became correspondingly more complex.
Cultural standards have evolved in response to these trends. Just as Diener could publish what we would call under-analyzed data, given the standards of his day, so researchers in many fields now are expected to make use of arcane statistical procedures, even in cases where, arguably, they're not needed.
I'm not a statistical Luddite. There's nothing inherently wrong with complicated statistics. They're preferable when they can do everything simpler ones do, and more.
My concern – which comes up a lot in my newsletters, and in the book I'm writing – is that as the statistics have become more complex, the approaches to measurement, at least in some fields, have grown weaker. Crudely speaking, measurement is now less about understanding the world and more about generating data that the statistical procedures can handle.
Unfortunately, powerful statistics can't save weak measures. If the measures yield flawed data, statistical analyses usually can't fix things. Garbage in, garbage out.
This isn't a terribly serious problem if we're merely asking who steals Halloween candy. Of greater concern are the studies of immediate relevance, such as recent ones tracking the cardiovascular effects of sugar consumption. I may be covering one of these studies, currently in press, in an early-spring newsletter.