Anyone weighing whether to stay in school for another year or head into the workforce faces a version of the same question that has occupied economists for decades: how much does that extra year of education actually pay off in future earnings? The answer matters not just for teenagers deciding on college, but for governments budgeting for schools and for parents thinking about their kids’ futures.
A new analysis published in PNAS attempts to nail down that number using an unusual source of information: the random shuffling of DNA that happens at conception. The researchers estimate that one additional year of schooling raises annual earnings by about 8%, a figure higher than what conventional statistical methods produce.
Why the question is so hard to answer
The core problem is that people who stay in school longer tend to differ from those who don’t in ways that are hard to measure. They may be more ambitious, come from wealthier families, or have different personality traits that would boost their earnings even without more education. Simple comparisons between earnings and years of schooling can conflate the effect of education itself with the effect of everything else that comes bundled with it.
Economists have tried various workarounds. One approach compares siblings or twins to hold family background constant. Another exploits policy changes, such as laws raising the minimum school-leaving age, to find groups that got more schooling for reasons unrelated to their personal traits. But sibling comparisons still leave individual-level differences unaccounted for, and policy-based studies only capture what happens when the law forces a specific extra year at a specific age.
Tarjei Widding-Havneraas of the University of Oslo led a research team that took a different route. They used a technique called Mendelian randomization, which treats the genetic variants a person inherits at conception as a kind of natural experiment. Because parents pass down a random subset of their DNA to each child, siblings end up with different genetic predispositions in ways that are essentially unrelated to family circumstances.
Turning DNA into a natural experiment
The team took advantage of a large international genome-wide study that had identified hundreds of small DNA variants associated with how far people go in school. They combined 335 of these variants into a single score, essentially a genetic index of predisposition toward more schooling.
The logic runs like this: if two people differ in this genetic index, and if the index only affects earnings through the schooling it encourages, then comparing their earnings tells you something about the causal effect of education. Random genetic inheritance stands in for the random assignment you’d get in a controlled experiment.
To test this, the researchers pulled together data on more than 1.25 million Norwegians born between 1959 and 1982, drawing from national tax and education registries. They also used genetic data from about 109,800 participants in the Norwegian Mother, Father and Child Cohort Study. Earnings were measured as the average of a person’s top three earning years between ages 34 and 40, a window that closely tracks lifetime earnings. Education was measured as the highest attainment by age 33.
The team then ran the analysis several ways to see how the answer shifted depending on the method used.
What the different methods produced
The simplest approach, a standard regression using the full population, found that each additional year of schooling was associated with 5.9% higher earnings, or roughly 34,600 Norwegian kroner (about $3,600).
Comparing siblings within the same family produced a slightly lower estimate of 5.3%. Comparing dizygotic (fraternal) twins gave 4.3%, and comparing monozygotic (identical) twins, who share all their DNA, gave the lowest figure at 3.3%. The pattern that emerges as you strip away more sources of family and genetic confounding is a downward drift in the estimate.
The Mendelian randomization approach broke that pattern. Using the genetic score across the full genotyped sample, the estimate rose to 8.0%, which corresponds to roughly 53,600 kroner in additional annual earnings. A version that restricted comparisons to siblings, further guarding against family-level confounding, produced 6.3%. A third version that built the genetic score from a Norwegian-only sample to address concerns that the international genetic data might not translate well to Norway produced 9.7%.
Life-cycle patterns and the return on investment
The researchers also looked at how returns evolve over a working career. In every method, an additional year of schooling actually reduced earnings at younger ages, because more schooling meant less time working. Returns turned positive around age 27 and kept growing as workers aged.
To ask whether education is worth it overall, the team calculated the internal rate of return, which is the effective interest rate that education earns as an investment when you weigh forgone early-career earnings against higher later-career earnings. Across every method, this rate came out between 6.8% and 10.1%, well above the real market interest rate of 2.3%. In the authors’ framing, education pays off financially regardless of which statistical method you trust.
Why the genetic method gives higher numbers
The gap between the Mendelian randomization estimates and the more conventional estimates prompted the researchers to investigate possible explanations. One possibility is that people with high earning potential who lack academic inclinations sometimes leave school early to pursue well-paying trades or entrepreneurial paths, particularly in Norway’s compressed wage structure and natural-resource-heavy economy. That kind of hidden pattern would drag down conventional estimates without affecting the genetic approach.
Another possibility is that the genetic method captures the returns for a specific subset of people, namely those whose genetic predisposition nudges them into more schooling. This group leans more toward advanced degrees, where returns are known to be higher, and toward individuals from lower socioeconomic backgrounds, who also tend to see bigger gains from additional schooling.
Differences by sex and family background
Both the conventional and genetic methods pointed to higher returns to schooling for women than for men. Patterns by family income were less consistent: conventional analysis found larger percentage returns among people from lower-income families, while the genetic analysis found larger absolute-dollar returns among people from higher-income families.
Caveats worth keeping in mind
The method rests on assumptions that cannot all be fully verified. The genetic variants must influence earnings only through education and not through other channels, a condition known as the exclusion restriction. The researchers ran a battery of tests designed to detect violations and conducted a sensitivity analysis showing that the estimated returns would remain positive and statistically significant even if a large fraction of the genetic-earnings link were due to something other than schooling. Still, they note that they cannot rule out smaller amounts of such interference.
The Norwegian setting also matters. Norway has nearly free education, a compressed wage structure, and a strong welfare state. The authors caution that findings from this context may not translate directly to countries with different institutional arrangements, and that returns can vary by field of study in ways their analysis did not capture.
