The Average Premium and the People It Hides
With the bar-exam threshold providing a credible local comparison, the question becomes more concrete: what did access to legal practice do to earnings over time, and how broadly was that gain shared? This chapter follows the estimate from the first years after the exam through a 23-year observation window, then moves from the mean to the distribution. That shift matters because the study's roughly 50 percent average premium does not describe a typical salary increase; it summarizes different career paths, with the largest gains concentrated among highly successful lawyers.
A career effect, not an immediate bonus
The earnings effect does not appear as soon as a candidate writes “pass” on the written exam. The written test is followed by grading, the oral stage, registration, and the practical start of legal work. In the study's yearly estimates, the effect begins about two calendar years after the first written attempt. That delay is consistent with the institutional sequence: a candidate may take the written exam in December, complete the oral stage months later, enter the bar association the following year, and begin practicing in the next calendar year.
This timing is useful evidence. If marginal passers already earned more before the exam, the later gap could simply reflect a pre-existing difference. Instead, the study reports no earnings difference before the first attempt and a growing gap after the period in which licensing and practice can begin. The threshold is therefore not merely associated with a group that was already on a higher path; it changes the timing of access to a different occupation.
The gap grows during roughly the first six years after the exam and then settles at about 20,000 euros per year. Across career stages, the relative difference ranges from around 20 percent to 70 percent. When the full available period from years 2 through 23 is averaged, the instrumental-variable estimate is 19,825 euros. The corresponding mean earnings of non-treated compliers—the marginal candidates who would remain outside legal practice if they barely failed—are 39,195 euros. Dividing the estimated gap by that comparison level gives the paper's approximate 50 percent interpretation.
The denominator is as important as the numerator. The result does not mean that every lawyer earns 19,825 euros more than every non-lawyer, nor that a person can add 50 percent to their current salary by acquiring a license. It means that, for the local group whose career path was shifted by barely passing, the estimated average annual earnings with access to legal practice exceeded the estimated average annual earnings without that access by that amount over the observed window.
The study also calculates a present discounted value of 248,479 euros at a 5 percent annual discount rate over the available career horizon. This number translates a stream of yearly differences into a single value, but it is not lifetime wealth and it is not a forecast for a new applicant. It depends on the observed years, the discount rate, and the assumption that the estimated pattern is the relevant stream for the marginal candidates. A different horizon or discount rate would produce a different present value.
The annual path and the average answer different questions. The yearly estimates show when the legal-practice advantage begins, how quickly it develops, and whether it persists. The 19,825-euro figure compresses those years into one summary. A reader interested in early-career liquidity should care about the delay and the smaller initial gap; a reader evaluating a long career should care about persistence and volatility. Neither reader should mistake the discounted total for cash arriving at the moment the license is obtained.
The paper also finds little attrition in the linked sample: of the 1,972 matched individuals, only 46 are lost by year 23 because of deaths or early retirement. That makes the long-run comparison less vulnerable to a simple story in which only the successful remain observable. It does not remove every concern about measurement or career choice, but it supports reading the later estimates as a continuation of the same cohort rather than as a sequence of increasingly selective survivors.
The comparison becomes more meaningful when the alternatives are made visible. Nearly half of law graduates who never pass the bar work in the public sector. Licensed graduates, by contrast, receive about 80 percent of their income from practicing law, although some also have public-sector income. The groups are not choosing between “work” and “no work.” They are sorting into occupations with different earnings processes, employment arrangements, and levels of exposure to the market for legal services.
That sorting also explains why employment itself is not the whole story. Marginal passers are more likely to have positive recorded earnings in the early years, plausibly because marginal failers are still in apprenticeships or waiting to retake the exam. Later, the employment difference disappears as people who do not obtain the license move into other work. The persistent result is not simply that one group finds a job and the other remains unemployed. It is that the two groups settle into different kinds of work, with different potential earnings.
The path in the diagram should not be read as a deterministic ladder. Some people who enter the waiting state do not become practicing lawyers, and some people initially sent toward alternative work later obtain the license. The diagram represents the main comparison in the data: a threshold changes probabilities, those probabilities change occupational exposure, and the resulting earnings are observed over a long period rather than at one instant.
The mean is built at the top
The average becomes less intuitive when the researchers examine earnings by percentile. The distributions of licensed and non-licensed law graduates are nearly identical below the median. They begin to separate in the upper part of the distribution, with the divergence becoming substantial around the 75th percentile. At the 90th percentile, the estimated earnings difference is 38,247 euros. The paper's central interpretation is that highly successful lawyers generate much of the average return, while less successful lawyers often earn no more than comparable law graduates in other occupations.
The selected quantile estimates make the pattern visible. At the 10th percentile, the estimated difference is about −5,073 euros; at the 25th percentile, about 1,080 euros; at the median, about −1,117 euros; at the 75th percentile, about 10,022 euros; and at the 90th percentile, about 38,247 euros. The lower-percentile estimates are imprecise, so a negative point estimate should not be turned into a claim that licensing reliably lowers earnings for those workers. The safer conclusion is that the study does not find a clear positive advantage below the upper tail, while it finds much larger gains at the top.
A percentile is not a label for a fixed person. It is a position in the earnings distribution of the relevant comparison, and the person who occupies that position can differ across career paths. The 90th-percentile estimate therefore does not mean that a particular newly licensed lawyer has a 90 percent chance of receiving 38,247 euros more. It says that the gap between the high-earning portions of the two potential-outcome distributions is much larger than the gap near the center. The estimate describes how the distribution moves, not where an individual will land.
This distinction also explains why the study does not need to identify every source of high performance to establish heterogeneity. Client networks, specialization, location, reputation, business organization, and personal preferences may all matter, but the available evidence does not assign causal weights to them. What the quantile analysis establishes is more disciplined: after the threshold-induced change in access, the earnings consequences are not constant across the distribution. Explaining that pattern requires a mechanism account, but it should not be replaced by an unsupported success story.
This is the difference between an average return and a distribution of returns. Suppose one group contains many people near the comparison group's earnings and a smaller group with exceptionally high earnings. The group mean can be far higher even though the median person does not experience the same advantage. A reader who sees only “50 percent” may imagine a uniform shift upward. A reader who sees the quantile pattern understands that legal practice offers an opportunity structure: it can produce very high earnings, but it does not distribute those earnings evenly.
The distribution also changes the meaning of “success.” If the legal occupation paid a uniform premium, then the average would be a reasonable shorthand for an ordinary worker's likely gain. Here, the upper-tail concentration means that occupational entry and occupational performance must be kept separate. The license opens the market, but it does not determine client acquisition, case quality, reputation, business development, location, or the worker's ability to convert practice into high income. The study does not identify which of those factors explains why some lawyers reach the top; it only shows that the overall premium is not spread uniformly.
The same pattern appears in the alternative-occupation comparison. A graduate who chooses public-sector work may give up some chance of exceptional legal income in exchange for a more predictable earnings path. A graduate who practices law may gain access to the upper tail while also accepting the possibility of years that look no better than the outside option. These are not two versions of the same job with a credential added to one résumé. They are different occupational packages, and the mean earnings gap summarizes the packages imperfectly.
The study's employment result helps prevent another misreading. In the first years, marginal passers are more likely to have positive recorded earnings, but that difference fades later. A person who fails may remain in an apprenticeship whose earnings are too low to appear in the records, wait for a retake, or move into another occupation. Once those transitions are complete, the central contrast is not employment versus non-employment. It is the earnings distribution and risk attached to practicing law versus the distribution attached to the alternatives.
That is why the result should not be used as a personal forecast. It is evidence that marginal access to legal practice raised average earnings in this setting, over a long horizon, relative to the available alternatives. It is not evidence that every person who passes will earn more, that every person who fails would have earned more if admitted, or that the upper-tail outcome is likely for a new candidate. Nor does it identify the risk preferences, financial cushions, or constraints that make the same distribution acceptable to one worker and unattractive to another.
The result does support a stronger statement than “some lawyers do well.” The upper-tail pattern is tied to a causal access margin, not merely to a collection of anecdotes about successful professionals. But it also makes the headline more demanding to interpret. If the premium is concentrated among people with high earnings and legal earnings are more volatile, then the upside may be compensation for something that the mean alone hides. The next question is what that something is—and whether restricted competition, longer hours, a credential signal, or income risk provides the best explanation.