How to read habit and income research

Some UPSHIFT articles say “people who do X tend to have Y.” Others say “doing X increased Y.” The difference is not a matter of style. It comes from how the research was designed. Here are three examples that show how to tell them apart.

What we know

1. Correlation: exercise and wages (US)

A study using the US National Longitudinal Survey of Youth (NLSY) found that people who exercised regularly had wages 6–10% higher (Kosteas 2012). The researcher used propensity score matching, a statistical method that makes exercisers and non-exercisers look as similar as possible on measured traits.

Even so, these are observational data. People who stick with exercise may already be healthier or more organised. Differences that do not appear in the data cannot be balanced out. So what we can say is that people who exercise tend to have higher wages.

2. Natural experiment: sleep and earnings (US)

A study using the American Time Use Survey (ATUS) took a different approach (Gibson & Shrader 2018). It relied on the fact that, within the same time zone, people in places where the sun sets later tend to sleep less. Sunset time has nothing to do with a person’s character or ability. Using this kind of chance variation is called an instrumental variable (IV) approach.

The study estimated that when average weekly sleep in an area rises by one hour, earnings are 1.1% higher in the short run and 5% higher in the long run. Note that this is an area-average effect. It does not directly measure what happens when one individual sleeps an extra hour.

3. Randomised controlled trial (RCT): steps (US)

The clearest evidence of cause comes from experiments that assign people by lottery. In the US STEP UP trial, 602 people were randomly assigned, and the group with a competition element walked 920 more steps a day than the control group (Patel and colleagues, 2019). Random assignment means pre-existing differences cancel out on average. However, the study measured steps, not income.

Bonus: comparing people who chose to join (US saving)

In the “Save More Tomorrow” program, employees committed in advance to save part of their future pay rises. Of those offered the plan, 78% joined, and participants’ saving rates went from 3.5% to 13.6% over 40 months (Thaler & Benartzi 2004). That is a large change, but participants chose to join. They may have been more eager savers to begin with, so selection bias remains.

How to read it

When you see research in the news or on social media, check three things:

• Who was compared, and how were they split? A lottery points toward cause. Self-selected groups or simple comparisons point toward correlation.
• Whose data, and from which country? Results from US employees may not carry over directly to work in Japan or Korea.
• What was measured? Watch for studies of steps or motivation being presented as if they were about income.

At UPSHIFT, we write correlational findings as “people who do X tend to have Y,” and only use “doing X increased Y” for RCTs and natural experiments. All of this research describes averages. It does not predict your individual future.

A small step for today

Knowing the strength of the evidence can change which habits you pick. Sleep is one of the few topics where a near-causal design has linked it to earnings, so try logging “Slept about 7 hours.” In money, where automatic systems have strong support, set up “Paid myself first” just once. And while exercise research is mostly correlational, exercise is also linked to health and motivation, so keep going with “Moved for 5+ minutes.”

Habits chosen with a clear sense of what the research does and does not show are less likely to be swept away by the next trend.

Sources

• [A1] Kosteas (2012) The Effect of Exercise on Earnings: Evidence from the NLSY. Journal of Labor Research.
• [B1] Gibson & Shrader (2018) Time Use and Labor Productivity: The Returns to Sleep. Review of Economics and Statistics.
• [D6] Thaler & Benartzi (2004) Save More Tomorrow: Using Behavioral Economics to Increase Employee Saving. Journal of Political Economy.
• [A9] Patel et al. (2019) Effectiveness of Behaviorally Designed Gamification Interventions With Social Incentives for Increasing Physical Activity Among Overweight and Obese Adults Across the United States: The STEP UP Randomized Clinical Trial. JAMA Internal Medicine.