socmobyrecent

Notes on: Ichino, A., Rustichini, A., Zanella, G. (2025). College, cognitive ability and socio-economic disadvantage: policy lessons from the UK in 1960 – 2004. Forthcoming in The Review of Economic Studies.

https://users.econ.umn.edu › ~rusti001 › Research › Empirical › IRZ_RESTUD.pdf

Dave Harris

[A very welcome update of UK social mobility data, reworking but not referring to the classic work by Halsey and Goldthorpe on the period. This article is actually quite forbidding for non-mathematicians like me in that it gives each variable a symbol and then connects these symbols into various mathematical equations to try and trace the relations between them {using a 'general equilibrium Roy model'}. I don't think the mathematics is particularly difficult. Nevertheless, I have not included any equations but relied on the ordinary English summaries. There are some interesting implications, for the UK and for OECD countries].

The abstract tells us, for example that the expansion of university access in the UK beginning in the 1960s led to the selection into college 'of progressively less talented students from advantaged backgrounds', counter to the stated policy of increasing access to both talented and disadvantaged students.

Enrolment in tertiary education has increased in all OECD countries since 1970, by a factor of three, and further growth is planned, up until say 45% participation by 2030. It is already 45% in the UK in 2007 [5% in 1960] They want to measure both 'precollege cognitive ability and socio-economic disadvantage' to see the effects. They are going to use the Roy model to see how these traits affect graduation probability and also to measure the effects of both technological change and higher education policy. These will change the 'incentives to pursue a college degree' [and later the costs of doing so]. They assume that 'productivity and wages depend directly on cognitive ability for a given educational attainment' and we can test actual policies against this assumption. The UK is an ideal case study.

Robbins identified a reserve of untapped ability located above all in the '"poorer sections of the community"' (1) and recommended increased opportunity. They denied that expansion would lead to a lowering of the average ability of students in HE, and this is a shared assumption — but they need to be investigated. The authors have acquired data to do so.

They found that average ability of students selected into college did decline 'by about 13% of a standard deviation between the 1960s and the 1990s' (2). The average ability of what they call 'high school graduates without a college degree' also declined, leaving college graduates with a higher average ability than high school graduates in the 1990s nevertheless. The 'college premium'can be assessed in terms of the present value of earnings across the cohorts: this remained 'flat' despite 'education biased technological change' (2) [only 1 of two types of technological change as we shall see]. The college premium at the top of the cognitive ability range declined but increased at the lower ability levels, explicable by 'a novel pattern… 'Ability biased technological change'. We can see demonstrated [after much specific argument below] and interaction between technological change and an increase in the number of college graduates, explicable by a reduction in 'non-tuition costs' and by lower qualification barriers at entry, without paying attention to cognitive ability.

Overall, untapped ability did exist, but policy was inefficient at showing this ability into universities. Instead it 'favoured primarily low ability students from high SES families'. Policies which focused instead on selection of students of sufficiently high ability, 'possibly combined with a need to the disadvantage' were feasible and would have increased tertiary education opportunities and increased college graduates ability. Instead, a 'social welfare function' was pursued. Although this may be valuable in its own right, it contradicts a policy to increase individual productivity through fostering ability and HE. Even if graduate premiums remain positive, a lower precollege ability means 'a higher study effort cost' (3) [borne by institutions as well as students? ] Finally, you can preserve the high ability of students on selection without increasing inequality, they argue.

They then go on to explain the general equilibrium model. This involves human capital investment decisions. Competitive firms will combine college and high school graduates with results affecting the college premium. Similarly workers productivity and wages will depend on education and cognitive ability. Technological change will affect productivity, but the ability of employees will have an effect — technological change may either favour education or ability as above, and thus require either more graduates or more able students [they have already separated these 2].

On the supply side obtaining a college degree depends on an individual's cognitive ability and socio-economic disadvantage, primarily determined by family background. These combined to affect the cost of study effort and therefore the graduation probability. The government can affect this combination and thus affect access, and this will produce 'higher and lower graduation probability regions' (4) apparently which can be studied by the Roy model. Specifically the correlation between ability and disadvantage when selected into college will affect the development of graduate characteristics. Robbins and others thought that SES disadvantage outweighed differences in ability, with the implication that increasing opportunities for the disadvantage will increase high ability graduates. However, they argue that there is 'a negative correlation between precollege ability and disadvantage', which they explain in terms of 'the economics of skill formation' [I am struggling to understand that. I don't know the literature cited. I think the implication is that in this case the disadvantaged still had independent sources for their lack of ability? I had thought the argument was that the advantaged took major advantage of the University expansion regardless of their ability].

The next few sections establish the model in this abstract way which I personally don't feel to be very helpful, mostly because I am too lazy to remember the values of each variable. Some interesting and controversial variables are introduced by fiat, either a good advantage of abstract theory which exceeds the usual empirical description, or theoreticism. One such is that human capital is defined by both 'cognitive skills and study effort levels' [which reminds me of the old definitions of 'merit' — talent plus effort]. Cognitive ability and social advantage affects the rewards of effort as well as the other way around [if we consider study cost]. Effort is arranged rather conventionally in terms of an alternative to consumption and leisure, I think. The model proceeds by suggesting optimal equilibrium levels of ability and advantage and how this will affect wages. This in turn enables the calculation of probabilities of success for a given level of effort taking into account ability and advantage. It is a complex relation, however. The assumption is that firms will also act rationally and competitively in their wage setting policies — that they choose labour to maximise profits, for example and that there is a transparent market for each skill and ability group.

They can also go on to specify the most rational calculative HE policy, assuming a government can control actively or passively for key parameters to manage student demand. For example they can offer means tested grants, or complex financial packages 'that disadvantaged households can hardly navigate' more passively, they can build new universities, either in response to a specific demand or, more passively, in terms of a more general policy. They can grant scholarship based on ability or rank order applicants. In the UK, the government favoured the construction of universities and polytechnics and the introduction of 'less stringent admissions criteria' (9).

These policies can be tested in terms of whether they achieved the Robbins goal to tap ability from poor sections of the community. The possibilities are stated formally in equation form first. The English language version specifies that policy must be outcome-based rather than procedural based, positive selection in other words. If the reserves of untapped ability run out, meritocracy will lead to a lowering of the ability of college graduates, but Robbins thought that those fears were unfounded.

Nevertheless, there are different variants. Policies will enhance abilities overall if students average ability is sufficiently high, for example if it is not lower than those or close to who obtain qualifications at high school anyway [if there is an increase in those qualifications?]. In other circumstances, the average ability of college graduates may only weakly increase. There are additional considerations if we wish to tackle disadvantage because study effort cost is still higher for low SES high ability students. Generally, particular combinations revealed in their equations 'are cumbersome', allowing for considerable differences between the aims of policy and outcomes [I do not claim to have understood fully this argument.

[Even more challenging diagrams ensue. This time, they are starting different types of ability and disadvantage in the college population]. From the English language descriptions, I gather that societies can differ in terms of the size of the pool of talented but disadvantage students who are excluded which will affect the chances of government intervention without reducing the quality of graduates and without unintentionally favouring high SES students.

They take education biased technological change as a contravening variable, which can itself introduce higher study effort because it increases the college premium across ability groups [they estimate by about 2% on average]. It seems that it is these circumstances that will reduce average ability and increase average disadvantage in the college population [because talented proles face higher study costs?].

There are further unintended effects. If the supply of graduates increases, the college wage premium will decline [about 0.6% they calculate] and this can affect the policy and 'ultimately reduce the average ability of individuals selected into college' (13). It is possible to conceive of a policy that makes a low SES a positive advantage in terms of college access [means tested grants?], but this risks ability enhancing policies, and will only work if ability is not correlated with social advantage.. This has been seen in policies that try and attract the most able by adjusting 'effort cost shifts' (14) to the extent that SES can even become irrelevant. This does produce increased chances of obtaining a college degree for those at the higher levels of ability and will increase the average ability of college graduates. [I would be very interested to see what these policies actually looked like]. It also 'mitigates' disadvantage.

The main source of data is from the University of Essex Inst for Social and Economic Research 2019, a 'representative longitudinal study of UK households' (14). The sample 2011 to 2013, wave three, had 49,692 people in it and asked about their cognitive ability. They further restricted this sample to individuals 'with nonzero cross-sectional weights' [pass — that left 38,223], white born in the UK, 31,132, education information, 31,072, born between 1940 and 1984, 23,288. A sub- sample of 22,175 also had information about their cognitive ability, and they further restricted it to 17,890 people who also had 'the required high school credentials to apply for college admission'. There is a table on page 15 to summarise the data

They used other statistics to 'predict the Discounted Present Value of lifetime earnings' (16) which suggest that wage profiles are flat in the final years of a career, between ages 51 and 60, which enables them to suggest that any increases in that period will be the result of only time effects. They then compared wage growth and cognitive information to predict the wage of each subject at different ages, and therefore the resulting discounted present value. They also used data in the 1970 British Cohort Study and the University Statistical Record [presumably on the UK].

Before 1992 there were two HE options — traditional universities and polytechnics [colleges of education ignored] with different funding, 'target population, organisations, subjects and admissions criteria'. In 1992 polytechnics were enabled to obtain university status leading to a new definition of a college graduate as obtaining tertiary education degree of any kind, that is combining traditional universities and polys.

They divided individuals into '"college cohorts"' (17 with 15 year windows, groups of individuals, actual college graduates or potential high school graduates of college attendance at age, year of birth +20. 1960 to 74, 1975 to 89, 1990 to 2004. They defend this division in terms of combinations of similarities across education and ability, with suspected differences in labour market, wages and employment.

The cognitive ability test in wave three had six subtests 'immediate word recall, delayed word recall, subtraction, number series, verbal ability and numeric ability' (17) [further explained in McFall 2013] Apparently, age was also available, either specific questions or 'generic material aid'and this was taken into account. The result was 14 cognitive ability variables — six sub- test scores and 'eight dummies for whether help was received'. They used apparently standard component analysis to get a general cognitive ability measure: the first principal component explains 18% of the variability, which they take to be indicative of a satisfactory measure of the respondents' cognitive ability.

They then seem to have standardised the measure to produce a mean of 100 and a standard deviation of 15 to enable a comparison with 'the widely used IQ measure' (18) this will enable them to make comparisons with the more familiar scale, but they argue that the particular scale is irrelevant since they are measuring changes.

The particular cognitive tests were administered after potential or actual college attendance 'and so may be endogenous to tertiary education', but this is 'not a concern'. The unique thing is that there are scores for verbal and mathematical tests of the same individuals 'at ages five, 10 and 34'. They can construct 'a normalised cognitive ability measure' to get the average ability that these three ages of high school graduates' and then see if they did or did not obtain a college degree. They had different ways of dealing with missing test scores.

In all the panels, the ability gap between two groups increases until age 10, but between age 10 and 34 the ability of college graduates does not increase, neither relative to high school only graduates nor in absolute terms, and this is apparently consistent with the literature. It seems that cognitive ability is 'unlikely to be malleable beyond infancy' and that the effect of schooling on cognitive skills 'is not mediated by general cognitive ability which instead seems to be largely unaffected by education' (19 – 20). Overall, they are measuring pre-college cognitive skill. It does reflect disadvantage. These skills 'are fixed relatively early in life' (20).

There are two issues, however: Flynn observed an apparent improvement in IQ scores during the 20th century in 14 nations, although that is been reversed in recent years. Salthouse thought that different types of cognitive skills involving different ways during the life-cycle. They thought they would counter this by normalising the ability measure. The Flynn effect would also be removed, but this is been explained anyway as unlikely to be generalised.

Turning to socio-economic disadvantage, students from 'low income low education s or ingle parent families are less likely to enroll and complete college' (20). They aggregate eight relevant variables from the Essex survey: 'parents years of schooling and six dummies (referring respectively to when the respondent was 14) for whether either parent was employed, whether the respondent was living with only one parent, and whether either parent was deceased'. The component test explains 22% of the variability and again the conclusion is that they have summarised a measure of advantage, which can be turned into disadvantaged by simply inverting the sign. They also want to change the distribution so that disadvantage now has a minimum of zero.

Again they want to estimate changes across cohorts. There are some difficulties including the lack of information on family income when respondents were 14, meaning that 'this measure is relatively coarse but arguably captures long-term determinants of a young person's cost of study effort' (21). However, unlike ability, there are difficulties with normalising since standards have certainly improved in the UK and this effect has to be fully considered.

[Now to fit the 'empirical facts' against the models]

The Essex sample is representative of the UK population. College graduates increased from 18% to 33% between 1960 – 74, and 1990 – 04 cohorts. In the same analysis, subjects with at least a high school degree were 28% and 35% respectively.

We can then show the joint distribution of ability and disadvantage by cohorts [in the form of slightly confusing distributions of dots, in the diagram on page 23]. The correlation between the two is negative in all college cohorts and is 'statistically and quantitatively stable over time [although it is low -0.14. At least it's not positive!]. It fits best the model that suggest that high ability parents have better education and higher income and they are able to transmit to their children higher ability 'via genes and better nurturing' as well as higher SES. This represents 'a particular social equilibrium' in the UK between 1960 and 2004 (22), but this is not 'policy invariant'

Given this fit, they can now go on to investigate three 'sets of facts' that will help them infer the features and impact of tertiary education expansion in the UK. These are: 'the evolution of mean cognitive ability and mean SES disadvantage of students who graduate'; 'the way the expansion was enacted'; 'the evolution of the college premium'. They choose the 'terciles of the distributions' for particular attention, to produce three ability groups according to whether ability is below the first tercile, between the first and second and above the second. The same for disadvantaged groups.

The average ability of college graduates has declined by two points '(13% of a standard deviation)' from 110 in the first cohort to 108 in the second, where 100 is the 'value to which the average ability the entire UK population is normalised in all cohorts' (23). For those high school graduates without tertiary education, ability declined by almost 5 points, '(about 30% of a standard deviation)', from 102 to 97. Overall, the tertiary education expansion in the UK 'was not ability enhancing' although it did favour high ability students compared to the ability of those previously excluded from college — but these were not able enough relative to the existing pool to affect the average values. Generally, 'the share of higher stability students declines across cohorts while the shares of less able students increase'.

For SES disadvantage, we have to start by thinking of average disadvantage is not constant but rather showing 'a declining trend that reflects the improving socio-economic status of the UK population during this period' [we are talking about disadvantage here]. In both cohorts above, high school graduates are 'very similar to the overall UK population in terms of disadvantage, which declines by about 15%' (24). The expansion of tertiary education had only marginal effects on the average background of college and high school graduates relative to the UK population, with a more substantial effect for the later cohort — although average disadvantage declined by 13% in that period, it declined by 23% among college graduates but only 5% among high school graduates. This is evidence that the recent expansion did bring into college high school graduates who were relatively advantaged [we are still talking about disadvantage increasing? Is the argument that disadvantage decreased anyway by 13%?]. 'The UK tertiary education expansion has not been disadvantage – mitigating' (25). The evidence also shows that 'the incidence of the most disadvantaged students in the college population is the lowest in all cohorts and barely changes' between the 60s and the 90s. By contrast, 'the share of the most advantaged students is always the highest and actually increases over this period' .

We can now bring ability and disadvantage together for the college population [in what is in effect a 2 x 2 or actually 3 x 3 table, with a levels of ability across the top and social disadvantage down the left. It is 3 x 3 because we are measuring terciles. [The cells seem to actually report 'the share of students in the college population'. A quick and inexpert inspection indicates that the largest scare in each case, each cohort, belongs to the top left cell indicating the strongest connection between ability and advantage, although, equally mysteriously as throughout, this is somehow referred to as disadvantage. Perhaps it is simply that the vertical column is upside down, with the most disadvantaged at the top?].

Overall, those college graduates with the higher stability and the highest disadvantage 'declined from 13% to 10% between the first and last cohort [figures rounded]. However those with the lowest ability and the lowest disadvantage increased from 5 to 7%. Overall, those who were relatively more able and relatively more disadvantaged 'actually ended up with reduced opportunities to obtain a college degree' compared to their 'lower ability and lower disadvantage counterparts. [A really astonishing finding, showing that disadvantage actually cancels out even superior ability?].

They argue the same by constructing a linear model for the probability of college graduation, using functions that describe ability, disadvantage, and their possible interactions. What this means is that during the expansion 'cognitive ability boosted' the probability of individual graduation more for advantaged students and for disadvantaged ones, while 'high ability but disadvantaged students' faced a more severe penalty in the later cohort than in the earlier one.

Policy interventions that were crucial include ending the divide between traditional universities and polytechnics; increasing the number of academic institutions 'without much consideration of quality, and by reducing ability requirements at entry' (26). Between 1966 and 1992, numbers of students enrolled in universities more than doubled, it was smaller for polytechnics but still substantial. Traditional universities expansion resulted from 'more subtle policy changes', for example 'reducing graduation costs'by bringing academic institutions closer to potential students [the average distance from the closest university dropped by about 6 km between 1960 and 2005, and in that period, with no fees, mobility costs were an important component]. This policy also disproportionately affected affluent families.

There is also evidence that the 'fraction of students admitted with weaker high school credentials' increased between 1970 and 1993 for Oxbridge, Russell and the less prestigious institutions, as a result of 'less demanding admission criteria' (28) including those with less than three A-levels and those admitted on the basis of 'HNC/HND/ONC/OND qualifications' (28). Admissions to Oxbridge still had higher best A-level scores than those admitted to Russell groups, who dominated students in the other institutions, and this indicator 'increases significantly over the period of observation'. This might be 'grade inflation' in high schools, or it might be that high school students improved over time with A-levels. 'We are unable to establish which scenario is the correct one' (28). However the average cognitive ability of graduates is not increased, indicating that universities have become less selective.

GCSEs replacing CSEs and O levels were also significant, and did lead to '"an increase in educational attainment at the secondary level… And hence an increase in the proportion of the young with sufficient academic credentials for potential admission to universities"' (28). This may also have 'contributed this well to the decline of high school graduates cognitive ability' as documented [very controversial!].

It is possible to estimate the consequences of these changes on the shares of high school and college graduates by ability. The share of high school graduates in the population has declined and this has been driven 'by a drop in the intermediate top ability groups', and a corresponding 'increase in college graduates shares for these groups' [?]. Fewer students drop out and more of them acquire the credentials to apply for college, which increases the odds of college graduation. However, the top ability group particularly benefits, but this is not enough 'to make the expansion ability enhancing' on its own. (29).

They estimate the college premium through the 'DPV of lifetime earnings' [DPV is not defined from what I can see-- Google defines it as 'It is a financial calculation used to estimate the total worth of all future earnings an individual will receive over their working life, expressed in today's currency. They 'pooled DVP but also noted 'polarisation the underlyinng evels by ability group'— It is decreasing for the intermediate group, but increasing for both bottom and top groups, although with a slight decline between the two cohorts, probably resulting 'from a remarkable increase for the least able that is more than offset by declines of the two more able groups' (30). This is in sharp contrast to the US where the college premium has increased since the 1980s — US technology dominated education, but not in the UK. Expansion in the UK increased participation of the least able student's compared with their more able peers.

We now need to consider 'technology parameters in the model'. Three of these refer to technological productivity ratios, one for each ability group'. There is also an issue of the 'elasticity of substitution between high school and college graduates' (31) in terms of TFP ( total factor productivity, Google suggests). [The sophisticated stuff follows estimates the difficulties of estimating the differences in these two groups] they also consider 'effort cost', which they estimate using [some conventional parameters, incomprehensible to me, I fear, 31]. The point is to estimate the distance between the odds of college graduation and the college premium for each ability group, and to calculate the 'average ability and disadvantage of college and high school graduates'. Higher education policy or technology can alter the labour market equilibrium.

[They calculate the relations using complex equations and apply them to particular percentiles of the sample. I'm afraid it is largely incomprehensible to me, PP 32F]. The English language explanation follows on 33. Estimated productivity ratios 'increase across ability groups within each cohort', as a result of technology available which will increase graduates productivity advantage over high school graduates, and also increase the premium between first and last cohorts. This is 'education biased technological change' already discussed in lots of literature, apparently.

It is also possible to discuss 'ability biased technological change' (34). Apparently, at any combination of ability and educational levels there is a decline in the ratio, meaning that 'technological change was biased against ability'. This is because 'computers and off shoring have progressively replaced routine workers in advanced economies, while at the same time increasing workers productivity in nonroutine and cognitive jobs'. This will probably produce 'job and wage polarisation'. One implication is that graduates productivity study increase relative to high school graduates, which explains the 'educational biased' technological change, but this effect is not found within educational groups, for example 'for workers with higher and lower cognitive ability': these changes do not reward more able workers. Computers do increase individual ability and complement cognitive skills 'in tasks that require them', which does bring a benefit from computerisation Ford low ability but equally educated workers — high ability individuals 'are less in need of aid from a computer' (35) [the example is computer-aided design in architecture, or the adoption of robots in surgery, or AI in customer support]. Note that there are also 'unobservable skills within education levels' (36). Overall within each education group, computers will have the effect of both replacing workers and enhancing the productivity of those engaged in nonroutine cognitive tasks, especially if they have low ability.

They looked at workers occupations in the Essex sample, and classified them according to 'task types' [nonroutine, analytical, interpersonal, routine cognitive, routine manual and so on]. They did find that computers favoured low ability workers, but that 'technological change was education – biased for both nonroutine and cognitive jobs'. So the productivity of the lowest ability workers did increase from introduction of computers, but those of the highest ability college graduates 'remains essentially flat', additionally from their declining share in the population. Overall, this has produced a 'flat evolution of the overall college premium in the UK'.

Overall, the evidence so far has shown that high ability college graduates increased as a result of UK university expansion, but this still resulted in 'a declining average cognitive ability and declining relative disadvantage of college graduates'. This is what they had estimated in their model building phase before. They are now going to discuss the economic meaning of these changes, in terms of effort costs and how they changed across the two cohorts. Basically, costs are 'decreasing and increasing, respectively with ability and disadvantage' (37) and this is been produced by educational policies in the UK.

In more detail there has been an 'increase in the opportunity cost [NB] of study effort… At all levels of disadvantage… [But] that was more pronounced for more disadvantaged students'. There is also been a 'large downward shift in the direct cost of effort, and this was more pronounced for lower ability students, and this produced a large increase in college attendance 'by progressively less able and less disadvantaged students'. Overall, 'the college expansion policy was neither ability enhancing nor disadvantage mitigating'.

This is different from the case in the US, and this is seen in the differences in the evolution of the college premium. The usual assumption is that the UK follows the US in terms of technology adoption and changes in industrial organisation. In that case, a flat college premium would best indicate 'a switch from a centralised to a decentralised organisational mode by firms' and that would in turn 'neutralise the effects of technological changes and of increasing college attendance' (38). Further, technological change seems to be more education biased and 'also biased against cognitive ability', due to computerisation. In the USA, the vast majority of colleges have become less selective, but this has increased college premium, possibly because it enabled the selection of 'an even larger fraction of low ability students whose productivity was actually favoured by workplace computerisation' [if I have understood this correctly, this is the most ironic conclusion of all, that technological change favours low ability workers, so it makes sense to cram them into universities, presumably so that this will direct them into suitable industries? That would imply that the US system is better at connecting universities with firms adopting technological change — maybe they are more open in that sense?]

It could have been possible to 'have reached the "reserves of untapped ability… In the poorer sections of the community"' [citing Robbins]. They then go on to describe two possible policies, which must also 'deliver an increase of the aggregate odds of college graduation similar to the observed one' — that is allow for the increased proportion of graduates that the current system delivers?].

A policy could expand tertiary education opportunities for the most able [defined in terms of changing their particular variables — I am translating from the Greek here, and I'm not sure all the Greek names for the variables are present anyway, but it seems that you achieve this goal by setting some measure of costs to a negative value for those with above average ability — 'for example financial aid for university studies only to students who perform sufficiently well in a cognitive ability test', and other words the old scholarship system. The second policy expands opportunities for the most able and provides aid to the disadvantaged: apparently you need to reduce costs. Apparently, this is the policy in Chile.

A model, with diagram, (p.39) puts this in terms of their formal values.

Notes: Effects of two counterfactual expansion policies that would have achieved odds of college graduation similar to the observed ξ = 0:54, reporting for comparison the effect of the actual policy (left panel). In the central panel, a θ-target expansion { relative to the 1960-1974 status quo { decreases τ to −6:3 for students above average ability (while setting τ = 0 for those below average), and sets β = 0, γ = 9:1, and δ = 0. In the right panel, a (θ; λ)-target expansion that, relative to the θ-target policy, decreases γ only for students whose disadvantage is above the median and, for these students, in proportion to their ability: γ = 9:3 − I[λ > med(Λ)](0:235 + 0:135θ). The density in the north-eastern portion of the cloud is not sufficient for the (θ; λ)-target policy to expand the odds of college graduation beyond 0.53 without violating the constraint of positive cost shifts Ω(·) ≥ 0 and Γ(·) ≥ 0. Students marked with a \×" graduate from college under the counterfactual θ-target policy but not under the counterfactual (θ; λ)-target policy; for students marked with a \+" the opposite happens. In the scatter plots, each point is an individual in the 1990-2004 college cohort in our sample. Sample: USoc, 17,890 white respondents born in the UK in 1940-1984 with non-missing education and ability information and with at least a high school degree that qualifies for college enrollment (see the right panel of Table 1).

Overall, the first counterfactual policy will produce 'a decline in the fraction of college graduates' with low ability and low disadvantage, and an increase in those with fire ability and high disadvantage. This will increase the average cognitive ability of college graduates. Note that a confounding variable is the 'declining average disadvantage of the UK population' during the period between the two cohorts. Still, this policy would enhance the ability of graduates and mitigate the disadvantage better than the actual policy.

The second policy directs aid to high ability and high disadvantage students [both], and thereby raises 'graduation barriers' for others such as lower ability but more advantaged students, on the average, that is. There are limits to the [pool of talent?], and to exceed them would increase costs. However, this policy would still mean an increase in student ability above the actual average for the 1990 – 2004 cohort. This is what Robbins intended, but it is 'the opposite of what actually happened in the UK' [some further obscure findings indicate that the current policy also means that high ability students are less likely to graduate compared to this counterfactual policy — I wish I knew why]

So we introduce 'the systematic consideration of individuals cognitive ability in addition to the more conventional measures of socio-economic disadvantage' (41). We also identified two dimensions for technological change — 'education biased but also ability biased' we can use these to explain 'the flat evolution of the college wage premium in the UK between the two cohorts. Robbins did suggest that ability was a scarce resource and it needed to be reallocated, but such 'consideration and analysis are instead conspicuously absent in the current debate', including EU policy to increase the share of EU residents with tertiary educational attainment [apparently a document written by the EU Council 2021]. This document does not mention cognitive skills that students ought to have. This 'may be inspired by equity aspirations' but a better policy would better reconcile graduate quality with better opportunities for the disadvantaged.

Refs

NB couple of pieces by Boliver 2011 and 2013 on soc class. Higher education 61 (229 – 42) and the British Journal of sociology 64 (344 – 64). The EU Council document proposed future cooperation in the 'European Education Area'. There is also Pratt (1997) on the polytechnic experiment 1965 – 92. The Essex study is University of Essex, Institute for Social and Economic Research (2019): \Understanding Society: Waves 1-9, 2009-2018 and Harmonised BHPS: Waves 1-18, 1991-2009. [data collection]. 12th Edition.