The Better-Than-Average Effect from 1985 to Today
Psychologists tried replicating a 34-year-old result. What's changed since 1985?
Posted July 8, 2019
Some 34 years ago, Alicke (1985) published one of my favorite studies in psychology. Now considered a classic, this study revealed not only the immense variability in the perceived social desirability of common traits, but it also laid the groundwork for the popular egoistic result that most people tend to think of themselves as having more “good” (and fewer “bad”) traits than the average person.
The design was simple: University students were asked to rate 154 traits1 on how characteristic each trait was of themselves (“self-ratings”), and of the average college student (“other-ratings”).
A separate group of students had previously rated how socially desirable these traits were. The results were simple, too: As traits increased in social desirability, college students were more likely to make favorable social comparisons on those traits. In other words, participants claimed that descriptors like “intelligent,” “friendly,” and “reliable” were more characteristic of themselves than of the average college student. The same was true on the opposite end: participants claimed to be less “lazy,” “dishonest,” and “vain” than average.
Does the better-than-average effect replicate 34 years after Alicke’s work?2 And how have perceptions of trait desirability changed since then? With psychology’s recent (and encouraging) trend of revisiting the so-called classics, I get the privilege of seeing what has changed since I first learned about popular phenomena like this one in college.
Recent advances in methodology and culture within the field of psychology, including pre-registered replication attempts and open-science practices, now make it easier than ever to ask whether classic findings hold up and how they might have changed.
Mok and Feldman (2019) recently conducted a pre-registered replication and extension of Alicke (1985) and posted a write-up of the results online. The authors summarize their results concisely: “The better-than-average effect has appeared to remain robust since more than three decades when the original study was conducted, despite [some changes] in the design and statistical analyses in the present replication” (p. 40).
This study used rigorous methods, a preregistered analysis plan across two waves of sampling on Amazon’s Mechanical Turk recruitment platform, and some simple methodological changes that help ensure the results’ robustness across design choices.
Mok and Feldman found that American research participants continue to claim an abundance of desirable traits like resourcefulness, attractiveness, and kindness. These same participants claimed that negative traits like “mean,” “snobbish,” and “jealous” were much more representative of others than of themselves. There was a remarkably strong positive correlation (r = .77) between trait desirability and social comparison scores (self-ratings minus ratings of the average person), and this project reported one of the strongest empirical correlations I have ever seen in the field: r = .92 between trait desirability and simple self-ratings.3
Mok and Feldman took the research one step further by adding an “extension” to their replication attempt. In addition to social desirability (and controllability, which I do not discuss in this post), they also measured trait commonness, or how common people think each trait is in the general population.
Perceived commonness was positively associated with desirability, suggesting that people believed desirable traits to be more common in the population; I found this to be an optimistic result. Common traits were thought to be representative of the average person (no surprise there), but not representative of the self (which is both theoretically novel and interesting).
Finally, I counted the five most and five least socially desirable traits from the 1985 study and today (traits displayed in the table). Note that here we are comparing college students’ ratings to online workers’ ratings, so this rough comparison is between apples and oranges (at best).
Interestingly, today’s MTurk workers rated “intelligent” as the most desirable trait of all, though this trait barely cracked the top ten in Alicke’s results. And perhaps even more surprisingly, “attractive” was the 5th most desirable trait among online participants in 2019, though this trait wasn’t even in the top 50 among 1985’s University of North Carolina students.
I am particularly impressed by this work because not only was the replication attempt done right (i.e., with a pre-registered plan, complete transparency including open data, and a theoretically interesting extension), but also because the project was led by a master’s student. This is good evidence that rigorous replications can be run by researchers who do not yet have a Ph.D. (though there should probably be at least one Ph.D. level researcher involved). This is encouraging both for the future of the field and for the tests of classics in psychology that have yet to be run.
In closing, I can’t help but point out that “Reliable” was the most socially desirable trait out of all 154 (it was rated #1 in 1985 and #2 in 2019). This may give us some insight into the field of psychology’s thriving interest in reliability, but that's a topic for another post. Fortunately for social psychologists, one of the field's most popular and commonly cited results has another piece of evidence that points toward reliability.
1. Another benefit of direct replications (or replication and extensions) is that they provide careful scrutiny of old papers. Mok and Feldman (2019) pointed out that Alicke’s (1985) JPSP paper “reported using 154 traits, [but that] a detailed examination of the list revealed a total number of only 149” (p. 17). Whatever happened to those 5 leftover traits may simply be lost to time.
2. My colleagues and I recently found a robust BTAE in a representative sample of Americans, but we only studied a single trait: intelligence (Heck, Simons, and Chabris, 2018). What about the other 153 traits?
3. Krueger and I (2015) found r = .96 between trait desirability and self-ratings, but only over 19 items.
Mok, P. & Feldman, G. Alicke (1985): Pre-registered replication and extension. DOI: 10.13140/RG.2.2.30198.04164. Preprint retrieved from https://www.researchgate.net/publication/334119885_Alicke_1985_Pre-regi…
Heck, P.R., & Krueger, J.I. (2015). Self-enhancement diminished. Journal of Experimental Psychology: General, 144, 1003-1020. http://dx.doi.org/10.1037/xge0000105
Heck, P.R., Simons, D.J., & Chabris, C.F. (2018). 65% of Americans believe they are above average in intelligence: Results of two nationally representative surveys. PLoS ONE. https://doi.org/10.1371/journal.pone.0200103