Why Verbal Skill Opens the Education Door
This chapter follows mathematical and verbal skills through the education system, from A-level participation to college entry, graduation, degree quality, and postgraduate study. The focus is the first half of the math–verbal divide: why verbal skill has the larger estimated relationship with most educational milestones, and why that result must be separated from what happens inside a particular course. That distinction helps readers interpret educational success as a sequence of gates and environments, rather than as one number that already contains a student’s future earnings.
The central finding is striking but consistent. In the English cohort study, verbal skill is more strongly associated than mathematics skill with taking A-levels, attending college, graduating, earning a good degree, and going on to postgraduate study. The advantage remains even after college entry is considered. But when the question changes from “Who graduates?” to “Who graduates from the same kind of course?” the verbal relationship becomes much smaller, while mathematics remains positively related to performance. The change is a clue about selection, not a contradiction in the data.
Verbal skill has the larger relationship at most educational gates
The educational pathway begins before university. In England, A-level examinations are an important step toward university admission. The study finds that verbal skill is more strongly associated with taking A-levels than mathematics skill. That early difference matters because the opportunity to enter higher education is partly created before a student makes a college application. If a skill affects whether a person reaches the qualification stage, looking only at people who eventually attend college misses part of its role.
The same pattern appears at college entry. In the study’s instrumental-variables estimates, the coefficient on mathematics rank for college attendance is 0.276, while the coefficient on verbal rank is 0.700. These are coefficients on within-cohort percentile ranks, not direct percentage effects. The paper gives a more accessible translation: a ten-percentile-point increase is associated with an increase of about three percentage points in college attendance for mathematics and about seven percentage points for verbal skill.
That translation is useful because it shows what “larger” means without pretending that a rank is a natural unit. A move from, for example, the 50th to the 60th percentile is not the same educational change for every student or every cohort. It is a comparison within the study’s distribution. Still, the relative size of the estimates makes the result hard to reduce to a small difference in measurement: verbal skill has a much stronger connection to the chance of entering college in this specification.
The contrast continues after entry. For college graduation, the IV coefficients are 0.220 for mathematics and 0.521 for verbal skill. For graduation conditional on enrolment, the estimates are 0.104 and 0.166. The gap is narrower once the analysis is limited to people who have already entered college, but verbal skill remains the larger coefficient in that table. For a “good degree”—an upper-second-class degree or better—the coefficients are 0.124 for mathematics and 0.691 for verbal skill. For a first-class degree, the estimates are much closer: 0.189 for mathematics and 0.173 for verbal skill.
These results describe several different forms of educational progress. College attendance asks whether someone crosses the entry threshold. College graduation asks whether they complete a degree. Degree classification asks about the level of performance among graduates. Postgraduate participation asks whether education continues after the first degree. Verbal skill is not merely associated with the first decision. In the reported estimates, it is also strongly related to good-degree attainment and to entering postgraduate study within three years of completing a first bachelor’s degree. The postgraduate coefficients are 0.067 for mathematics and 0.243 for verbal skill.
The sequence matters more than any one coefficient. A skill can affect education by changing the probability of reaching the next gate, the probability of remaining in the system, or the outcome achieved after entry. Those routes can overlap, but they are not identical. A person who does not attend college cannot graduate from college; a person who attends but chooses a demanding or poorly matched course faces a different completion problem from someone who never reaches the application stage.
The estimates also show why correcting measurement error changes the substantive story. If the researchers use the observed GCSE subject grades directly, the difference between the mathematics and verbal coefficients is often smaller. When they use earlier Key Stage 2 scores as instruments for the GCSE ranks, the verbal coefficients generally increase while the mathematics coefficients decline. The exception is graduation conditional on college entry, where the verbal coefficient falls slightly, though it remains larger than the mathematics coefficient in the educational-attainment table.
This is a familiar problem in less technical form: a noisy measure can blur the distinction between two related things. Suppose two students’ grades do not perfectly reflect their underlying subject performance. If the errors are unrelated to one another, the estimates may be pulled toward zero. When both skills are positively related and both are measured imperfectly, the error can also make their relative contributions harder to separate. The study’s earlier assessments provide a way to reduce that problem, which is why the corrected results are the main evidence used in the paper.
The correction does not prove that verbal skill has a single, isolated cause of educational success. It strengthens the credibility of the comparison by addressing one important source of distortion. The estimates still describe the English cohorts, their institutions, their measures, and the assumptions of the instrumental-variables design. The right conclusion is therefore that verbal skill has the stronger estimated educational pathway in this setting—not that every form of literacy training will produce the same result in every education system.
There is another reason not to line up the coefficients as if they were scores in one competition: each outcome has a different denominator. College attendance is measured across the broad population. Graduation is measured among people who entered college, and postgraduate participation is measured among people who completed an undergraduate degree. A skill that affects an earlier gate changes who appears in the later sample. The later coefficient then describes a selected group, not the same population viewed repeatedly without alteration.
This selection does not make the later estimates useless. It tells us what question they answer. The college-attendance estimate asks how skill rank relates to crossing the education threshold. The conditional-graduation estimate asks how skill rank relates to completing a degree among those who crossed it. The degree-classification estimates ask about performance among people who reached graduation. Comparing these results is informative when the reader keeps the changing population in view; it becomes misleading when a larger coefficient is treated as a universal measure of importance.
The distinction is especially important for degree quality. A “good degree” and a first-class degree are not simply two points on a single smooth scale in the evidence. The study reports a much larger verbal coefficient for the former and similar mathematics and verbal coefficients for the latter. That difference suggests that the skill relationship can depend on the threshold being examined, the people who reach it, and the course environments in which performance is produced. It is a reminder that educational systems reward more than one kind of achievement, and that the route to each achievement can differ.
The diagram separates movement through the system from performance within an environment. That separation becomes essential in the second half of the story.
Graduating somewhere is not the same as graduating from the same course
The study’s aggregate results show a positive relationship between verbal skill and graduation. It would be tempting to read that as evidence that verbal skill is more important than mathematics for completing any course. But students do not enter courses at random. Their subject strengths help shape where they apply and what they choose, and the courses themselves differ in their completion rates.
The researchers examine this by adding fixed effects for institution-and-field-specific courses. A course fixed effect compares students within a more narrowly defined educational environment—for example, the same subject at a particular university—rather than treating all college entrants as if they faced the same programme. Once this course context is held constant, the mathematics coefficient for graduation is 0.100. The verbal coefficient is 0.016 and is statistically insignificant in the reported IV specification.
The change in the verbal estimate is informative. It suggests that a substantial part of the aggregate verbal–graduation relationship may come from selection into programmes where completion is more likely. A student with stronger verbal skills may be more likely to choose a humanities or social-science programme, or another course, with a higher average completion rate. The student then appears more likely to graduate in the broad population partly because of the environment entered. Once the comparison is made within the same course, the verbal advantage in completion is no longer clearly present in the estimate.
This is not evidence that verbal skill is irrelevant inside a course. It is evidence that two questions had been combined:
- Sorting across courses: Which students enter programmes with different completion environments?
- Performance within a course: Among students facing a more comparable course context, which skills are related to graduating?
The study’s results assign different roles to those two margins. Verbal skill is strongly related to the probability of reaching college and is associated with entry into courses where completion rates are higher. Mathematics remains positively related to graduation after course fixed effects are included. That pattern is compatible with mathematics supporting performance within some course environments, but the study does not test a randomized change in course difficulty or student support. The selection explanation is strongly suggested by the estimates and the course-return patterns; it is not a complete causal decomposition.
The distinction also changes the meaning of “educational success.” A completion rate can answer whether students finish. It does not answer whether they gained access to a high-return field, whether their institution offered strong labour-market connections, or what they earn after graduation. A field can have a higher completion rate and a lower average earnings return, or a different completion profile alongside attractive earnings prospects. The study does not establish that any such difference is caused by course difficulty. Treating graduation as the final outcome hides that trade-off.
This is why the stronger verbal relationship at educational gates does not settle the math–verbal divide. Verbal skill appears to help more people pass through the education system and may help them reach environments with higher completion rates. Mathematics is more closely tied to the fields that later carry higher average payoffs. The two findings can both be true because “entering and finishing education” and “choosing a destination with a high earnings return” are different decisions.
The same reasoning applies to postgraduate study. Among people who complete an undergraduate degree, stronger verbal skill is more strongly associated with entering postgraduate education. But conditional on entering a postgraduate programme, mathematics and verbal skill again point toward different fields: mathematics is associated with postgraduate STEM, while verbal skill is associated with postgraduate humanities, arts, and social-science fields. Continued education therefore adds another choice point rather than erasing the earlier pattern.
There is a practical temptation to use one educational indicator as a shortcut. If the goal is to widen college access, attendance may be the relevant outcome. If the goal is to improve completion, the comparison should distinguish selection into courses from performance within the same course. If the goal is economic mobility, degree completion must be connected to field, work, and earnings. Each measure answers a real question, but none answers all of them at once.
The careful conclusion is not that verbal skill “wins” education and mathematics “wins” everything after it. It is that the education system contains several gates and environments, and the two skills have different relationships with them. Verbal skill has the larger estimated role in reaching and progressing through most of those gates. Mathematics remains relevant to performance within a comparable course and, as the next part of the pathway will show, to the destinations students choose. The education result is therefore the beginning of the economic story, not its conclusion.