Recently, a team of researchers made a splash with a series of studies on the idea that it’s possible to have too much free time. Life satisfaction peaked between 2-5 hours of free time per day, suggesting that unhappiness might come from being too idle—not just too busy.
When I tweeted the research, some commenters immediately attempted to discredit it, complaining that free time only explained 0.3% of the variance in life satisfaction. They also resorted to name-calling and ad hominem attacks, leading me to wonder if they were prosecutors looking for a takedown rather than scientists searching for the truth. I won’t dignify those specific comments with a response, but if you do choose to weigh in below, I’d encourage you to aim for the top of Paul Graham’s hierarchy of disagreement.
Let's refute the central point of the attacks. If you think a small amount of variance explained is a problem, you’re making a common mistake. As the field of medicine has long recognized, a small coefficient of determination can be extremely consequential. To take a few classic examples from meta-analyses: on average,
Ibuprofen explains a whopping 2% of the variance in pain reduction
Taking statins to lower cholesterol accounts for about 0.55% of the variance in major cardiovascular events
Routine ultrasounds explain 0.01% of the variance in successful pregnancy outcomes
I don’t know about you, but these tiny findings wouldn’t stop me from taking Advil or statins—or encouraging women to get ultrasounds to check on their babies.
The lesson here is one that’s been stressed for decades: we shouldn’t confuse variance explained with practical significance. Like health, life satisfaction is influenced by many factors. Frankly, I’d be more surprised—and more skeptical—if free time alone accounted for 3% of the variance instead of 0.3%. “The observed effect of discretionary time on subjective well-being is small,” the authors acknowledged, because “there are a slew of other variables that play into people’s overall assessment of their satisfaction in life.” The 99.7% unexplained variance isn’t all random. Some of it is measurement error, and much of it is omitted variables—as the authors clearly showed by adding 8 control variables to explain 6% of the variance.
The quality of behavioral science isn’t determined by the amount of variance explained. It depends on the reliability and validity of the research.
Reliability is a question of consistency. The researchers established that: after documenting diminishing returns of free time in a nationally representative sample of over 13,000 working Americans, they replicated it with another nationally representative sample of over 20,000 Americans—this time a mix of employed and unemployed adults—using different measures of both free time and life satisfaction. Once again, there was a significant curvilinear relationship, and the variance explained by free time was virtually identical (0.4% of the variance in life satisfaction this time). This broader sample made it possible to expand the range of free time and detect the significant drop in life satisfaction above five hours a day.
Validity is a question of accuracy. In their two initial studies, the researchers ruled out some alternative explanations by controlling for variables that might influence both free time and satisfaction, such as age, income, and marital and parental status—and by running a series of statistical robustness checks. For example, free time had a curvilinear relationship with satisfaction on weekends as well as weekdays—and it didn’t predict a dip if people felt they were using it productively.
To move from correlation to causation, the researchers conducted two experiments. Even randomly assigning participants to imagine having 7 hours of free time a day led them to anticipate being less satisfied than those who imagined having half that free time. Of course we need richer experiments to test whether randomly assigning people to that level of free time actually decreases their happiness.
In the meantime, it’s worth noting that there’s a well-documented dip in life satisfaction in the first year after retirement. Along with a loss of meaning and community, could it be that a gain in free time is a contributing factor? Are there steps we can take to help people harness the benefits of free time while avoiding the costs?
These are the kinds of worthwhile questions about free time that this research opens up, which is why I was excited to share it. As new studies emerge, our knowledge will undoubtedly evolve. Last time I checked, that’s the point of science. Before you hurl vitriol at a set of studies, it might be a good idea to spend some of your free time brushing up on what statistics do and don’t tell us.
