The economics of higher education

right it's four o'clock now so i think we can kick things off hello and welcome everyone i'm ellen drayton and i'll be chairing today's event which is the penultima ifs public lecture just to highlight that our final lecture will be on the same time next week and that will be titled why tax corporate income but on to today i'm delighted that we'll be joined by ben waltman who's a senior research economist at the iss he has interest in education particularly in returns to education and funding models of education today he'll be talking to us about the economics of higher education and he'll be presenting roughly for 40 minutes after that time we'll have a short q a session and you'll be able to submit any questions via the slido app or alternatively you can vote for the questions that you'd like to be asked so without further ado i'll pass on to ben hello so i'm going to talk about the economics of higher education today and in particular i want to address three questions the first is why do people go to university and i'm going to outline a little economic model of higher education choices to think about that secondly why do most governments intervene in the market for higher education so first i'm going to look at what different governments do and then think a little bit about what could be their reasons in the framework of the model presented in the first part and third i want to look at what are the financial benefits of higher education while touching the question how one could find out and then using those methods that i've presented i'll um show you some results for england and throughout um this lecture is probably a bit more mathematical than other lectures in this series um but i will try to highlight the intuition behind all the maths so if you can't exactly follow something don't worry about it i'm gonna try to make sense of it even if you can't follow the maths so um on to why do people go to university um so what i would encourage you to ask if you're an undergraduate uh ask yourself or indeed if you are you've already graduated from university uh why did you decide to go to university did you go to get a more fulfilling job did you go to earn more money or where you're just um interested in the pure joy of learning for example similarly you might ask why did other people not go to university maybe um they never wanted to sit an exam again maybe they didn't want to have to leave home or maybe they just enjoyed practical work more than academic work so to make sense of all these different reasons for going to university or not going um i want to outline a little model of higher education choices so suppose there are two periods t equals one and t equals two in the first period which i'm going to call youth people choose whether to go to university or whether to start work straight away in the second period which i'm going to call working life everyone goes to work in the first period person i's income is going to be minus fhe if they go to university so if you have fees um which they have to pay and um why one eye non-hg if they don't go to university because if you don't go to university you can earn an income straight away in the second period person i's income is going to be y to ihe if they went to university and y2i non-hg if they did not go to university and um just to visualize this um so here's the first period youth so if you go to university which is in pink here um you got minus fhe because you have to pay fees but if you don't go to university you can earn an income y1i non-he whereas in working life um you earn an income whether or not you go to university and typically we think that the uh earnings that you're going to have if you do go to university y to iit are going to be higher than if you don't go to university so part of what going to university is is an investment which pays off in later life and um throughout this lecture i'm going to assume that there's no uncertainty about earnings so prospective new students know what's going to happen later that's of course not quite true but it's a useful simplifying assumption here so i'm going to assume that something about people's preferences first and in particular what i'm going to assume is that people's utility function takes this following um cop douglas form so utility of person i is going to be phi i times c 1 i to the one half times c 2 i to the one half where c 1 i and c 2 i are consumption in periods t equals 1 and t equals 2 respectively so c 1 i is how much you can consume when you're young and c2i is how much you can consume in working life and fire is going to be the non-monetary value of each option what does that mean well finite could contain if you're going to university the value of a more fulfilling job the joy of knowledge um all of the good things about university that um aren't a higher income and if you're not going to university on the other hand it might contain things like the value of never having to take exams again the value of not having to move away from home or the joy of practical work furthermore i'm going to assume that people can borrow any amount bi i'll be for borrowing here from financial markets at the interest rate r so um then uh given what we've assumed before we have um some budget constraints so for people who um go to university but if you go to university the budget constraint is going to be consumption in the first period is going to be whatever you borrow minus the fees that you have to pay and in the second period it's going to be y to i h e your income in the second period minus paying back whatever you borrowed with interest so that's one plus r putting those two things together eliminating the bi ag gives us this consolidated budget constraint here so that just means that the future value of consumption is going to have to equal the future value of income so um essentially you can only spend your income um looked at over the two periods and then similarly for those who don't go to university we're going to have a consolidated budget constraint that looks very similar which i've put down here um so if we put those things together now we um can display um the budget constraint and the preferences on a diagram like this which um if you've done some microeconomics you might be familiar with um so uh first um thinking about the case where you don't go to university so that's the green budget constraint here um that is um the um that the you've got an endowment point here which is income in the first period um and income in the second period and um this person is going to choose to borrow if they if they don't go to university and so they end up with a consumption bundle here which is um the consumption bundle where the budget constraint is tangent to the highest indifference growth here and similarly for the higher education case the endowment is up here so unlike what you've seen before we've got um the kind of first period endowment and negative territory because you have to pay fees but basically that doesn't really change anything substantive then you've got your budget constrained here for if you go to university and again this is the highest indifference curve so the consumption bundle is where the highest indifference curve is tangent to the budget constraint um and again you see that this person is going to borrow money in the first period so consumption in the first period is higher than the endowment um and um the pink budget constraint here is outside the green budget constraint or higher um which means that if this person goes to university um she can consume more um that doesn't necessarily mean that she will go to university because as we've seen they're also non-monetary factors consumption isn't everything so if she enjoys practical work more and that outweighs the higher earnings that she might get if she goes to university she might still choose not to go okay so um if we solve this constraint optimization problem to get c1i star and c2i star um which we may have done before in a microeconomics class um for both the um university and non-university case and we substitute um those consumption levels back into the utility function to get uihd and ui non-hd then we'll find that people are going to choose university um so people choose university if and only if the utility of going is higher than the utility of not going um so that's by definition but if we plug it in we can put some meat on the bones of this so then we get the um inequality at the bottom here which tells you that um the non-monetary value of going to hg times the future value of earnings and so everything you earn basically over the life cycle when you go to um university is greater than the non-monetary value of not going times the future value of earnings if you don't go and if the left-hand side is greater than the right-hand side then um you'll go and if it's smaller than you won't go so that's our theory um i've reproduced the inequality here and i just wanted to highlight what we can learn from this so factors pushing people towards higher education are going to be a high non-monetary value of higher education um low tuition fees fhe see this terms negative or high earnings after university right to ihe but also a low non-monetary value of not going to university is going to push you towards higher education um as well as low non-ate earnings y1i non-hg or y2i non-hd here and maybe not as intuitively um in this model with a private market higher education uh interest rates also matter a lot so if interest rates are low um then um more people are going to go to university because as you can see it enters negatively here and positively there um so assuming um that these three things hold true so students can borrow from financial markets at the true social cost of funds higher education only benefits students themselves and young people have perfect foresight and make optimal choices from their own point of view then the higher education choices that prospective students are going to make in this model are going to be efficient um so um just to recap what we've learned in this section so we now have a theory of higher education choice when there's a private market in higher education and a functioning financial market people are going to choose higher education if and only if it is better for them whether it is better for them depends on a number of factors some to do with money others not and in particular in the modern life outline people are going to choose higher education if and only if this inequality here holds true so remember this is the non-monetary value of higher education this is basically the your lifetime earnings if you um attend higher edu if you attend university and that needs to be greater than the non-monetary value of not going times um the uh life your lifetime earnings if you don't go to university um for you to choose to go to university in this model and under some assumptions these choices are going to be economically efficient so um that's the first part in the next section i want to look at why might higher education choices not be efficient and what do governments in practice do about that so why do most governments intervene in the market for higher education so there are different ways of funding higher education that different countries have chosen we've looked at a model of an entirely private market in higher education in the last section but in fact most countries do not have an entirely private market in the u.s there's a partly private market but there's a substantial state sector so state university sector which offers subsidized tuition fees and there are large brand programs for poorer students in england and australia um this system relies mostly on income contingent loans so those are loans that are paid back sorry as a percentage of earnings above a certain threshold um and the outstanding balances on those loans are written off after some years of repayment um in england that's 30 years and that works out to a substantial subsidy for most graduates in germany and scotland for example we have near no tuition fees that are charged including for postgraduates and subsidized loans are available for living costs um whereas in denmark and sweden first of all nutrition fees are charged but students are also entitled to living cost grants um so something like a salary for going to university and um looking at this as an economist you might ask why is the point i've got these government loans and subsidies if the benefits of higher education whether non-financial or financial accrue to the students themselves why does the government need to get involved at all doesn't the private market attain the efficient outcome well um these are the assumptions from the last section so assuming that we've assumed that students can borrow from financial markets at the true social cost of funds higher education only benefits students themselves and young people have perfect foresight and make optimal choices from their own point of view well then the higher education choices in this model in the model are going to be efficient but um it's kind of unclear whether any of these assumptions holds in reality so in fact um i'm going to try to convince you that because these assumptions don't perfectly hold the outcome of a private market might in fact not be efficient or unlikely to be efficient so first and perhaps most importantly um private financial markets may not work well in practice um leading to high cost of costs of borrowing for prospective students secondly higher education may have positive externalities so that means that it may not only benefit students themselves but also benefit others and third some students may choose not to go to university even if going to university would be optimal for them and the consequence of that is what economists call market failure so if the higher education market was completely private fewer young people would go to university then would be optimal so by subsidizing higher education governments may be looking to raise the higher education participation rate so how many people actually go to university to the optimal level so let's look at all of this in a bit more detail so first i want to start with financial market imperfections so what i'm going to claim is that um the interest rate are charged on student loans may in fact be much higher than the true social cost of funds and one reason for that is called limited enforceability so for student loans in contrast to say a mortgage there's no collateral so then if people look to escape repayment so they don't want to repay their loans the lender has a problem because there's no house they can seize or anything like that and the legal process of forcing someone to pay if they don't want to pay um and nothing can be seized is going to be costly and may not always be successful so the consequence of that is going to be um that the interest rate r is going to be much higher um than the true social cost funds so some people are going to be put off going to university by the high cost of loans so we can think of this in the terms of the model that i outlined in the last section um so here again is the key inequality from that um so if r goes up um the term on the left is going to go down and determine the right because this here is going to go up so fewer people are going to go to university um so in terms of the diagram um that i showed we can think of this as a pivot um so for those of you who've seen this sort of thing before um the kind of you can think of the relative price of first period consumption as one plus the interest rate um so if the interest rate goes up here and the budget constraint is going to pivot around the endowment point so you can see here and so what's happened here is that before it was financially lucrative for this person to attend university see the pink budget constraint was outside of the green budget constraint and after a pivot here around the endowment point this is reversed so now the green budget constraint is outside of the pink budget constraint and the this person is actually financially better off by going to university and by not going to university sorry okay second positive externalities so in fact higher education may not only benefit students themselves but may also benefit others and one important channel for this are so-called human capture spillovers so working and living alongside university-educated people is the idea here might create learning opportunities for others and there's a large economic literature on this a classic paper is moranti 2004 who finds that an increased share of college graduates in the city raises the productivity of other workers in that city a more recent contribution for example is chen and co-authors in 2020 who found that sending chinese university students to live and work in the countryside during the chinese cultural revolution raise the education levels of rural children another channel are higher tax revenues as a result of university graduates higher earnings so um in some of my own work at ifs um i found together with some core with us that um nearly half the financial gain for men and around a quarter of the financial gain for women from going to university actually accrues to the taxpayer and so these are likely the most important effects human couples as their human capital spillovers and um and uh fiscal actionalities but there may be others so for instance um education we know also leads to lower crime higher acidic engagements pillows from better health so that means that it may be socially optimal for more people to go to university than would be privately optimal so the solution to this for those of you um who've done some public economics is a pigouvian subsidy so basically what you want to do is the government is subsidizing education in order to raise higher education investment to the socially optimal level um just to flag this is not a closed case as it were there are also reasons to think that it might be socially optimal for fewer people to go to university then would be privately optimal um this is a complicated um case so um if for instance higher education is partly a costly signal of productivity to employers there may be over investment in higher education in a private market and um the classic reference to for that expense 1973. um i'm not going to dwell on that um very much because while that's kind of gets a lot of attention that story um i think the most um researchers in the higher education space would agree that um on balance the positive externalities probably outweigh these these signaling effects and and similar um finally uh privately some optimal decisions so prospective students may um underestimate the benefits of higher education that's one reason why they might um not decide to go and that would be so optimal um so that could be because they have improved information about what are the um financial and non-financial benefits of going to university and that as you can imagine is likely to especially affect people from disadvantaged backgrounds because perhaps um they are the first in their family to go to university and perhaps their parents also don't um have lots of friends who are going to university and um for those reasons um that it might be very difficult for them to know what it's actually like to go to university and what the labor market is like um if you're a graduate um secondly apart from those information problems there might be so-called internalities and so students may know that university would be best but maybe they just can't bring themselves to apply or actually go to university um so well the policy options here well so the government obviously can um provide more information and um in the uk um that's been very high on the policy agenda um and secondly um they can again make university more attractive by subsidizing it um which is uh what what we observe in practice that many governments do so how might government intervention look in the model that i outlined in the first part of this lecture so suppose the main problem is that the market for student loans does not work well so that's the first issue that i mentioned then a natural solution might be for the government to borrow on students behalf so how could that look like so we could have in period one so the government might borrow money from financial markets and use that money to pay for students fees and living costs um when they're while they're at university and in period two the government could tax graduates to pay off what it borrowed so this sort of tax is often called a graduate tax that's not going to restore the same allocations as meaning that students aren't going to make the same choices as they would in a completely private market because the government doesn't know how much each individual would have borrowed from the financial markets but it could come relatively close and it also in this sort of system the government can subsidize uh these um the um can subsidize the system um of higher education finance um in order to further encourage higher education participation so how can we capture this policy in the model well instead of having students pay fees fhe students would get living cross grants and which i'm going to call lhe in the first period and then students have to pay a graduate tax tower hg of their working life earnings and their working lives um and i'm going to assume that individuals can't borrow at all to make this a bit simpler or their interest rates are just privately high so nobody chooses to borrow so then we don't have to actually solve a concerned optimization problem here because everyone is just going to consume their endowment so people just um spend what they get in each period so for students um consumption in the first period is going to be the living cross grant that they get and consumption in the second period is going to be equal to one minus tau h e so the gradient tax times the second period income and for non-students we're just going to have people consuming their earnings in both periods so if we substitute um those consumption levels into the utility function and we get for the higher education case this equation and for the non-hd case we get that equation and then again people are going to choose higher education if and only if the utility of going to university is greater than the utility of not going which is then going to be this expression so that looks pretty similar to what we had before instead of lifetime earnings we here have um the um geometric mean of um earnings in um in the first period and earnings and net earnings in the second period um but if that means nothing to you don't worry basically the point is um it's pretty similar uh to what we had in the in the first part of the lecture um and so basically uh you might think that um the um the government has nearly restored the um the optimal allocation here uh so another thing to note about this is that the government as i said can um adjust this tau hg and the lhg in order to make more or less make make university more or less attractive um so that could kind of counteract or adjust for some of the externalities here so we can build the subsidy into the system um so arguably this model captures the higher education funding system in england quite well um so english students pay no fees upfront and receive money for living costs from the government um and then graduates pay back a percentage which is currently nine percent of their later life earnings above an earnings threshold which is currently 27 300. um so you might ask well hang on a minute doesn't the english system rely on student loans and that's right but in fact um the majority of students are never going to pay off their loans so for most the system is actually just like um this tax um as i described it um so just to recap um what i've talked about so a completely private market in higher education is quite rare most countries instead subsidize higher education but systems vary a bit there are three justifications for the government to intervene in the market for higher education that i mentioned the first is financial market imperfections the second is positive functionalities and third is privately suboptimal decisions um and if we adapt our model for government funding paid for by graduate tax students are going to choose higher education if and don't leave this inequality at the bottom here holds um and uh so that's um the non-monetary value of going to university times the geometric mean of the consumption levels in both periods um being greater than the non-monetary value of not going to university times the geometric mean of the consumption levels if you don't go to university okay final issue i want to talk about what are the financial benefits of higher education um so what we want to know here is the what's often uh called the financial return to higher education so how much on individual eyes working life earnings in percent as a result of going to university so the comparison is what person i would have earned had she not going to university um so that's important information both for prospective students who are thinking about whether or not to attend university and it's important information for the government thinking about the design of the higher education funding system however i should emphasize that financial returns of course are only one part of the benefit of higher education for individuals so um you know going to university also may mean that you will be healthier or happier later in life um as well as for society so we've talked about externalities in the previous section um so if you think about in terms of our model what we're going to want to estimate is the later life percentage earnings return which i'm going to call row i which is going to be y 2 i h e minus y 2 in energy divided by non-he earnings so um yeah how much higher in percent of your non-age earnings are going to be your second period earnings if you go to university um so just some more notations i'm going to call you define this indicator variable here d i h e which is going to be 1 if person r goes to university and otherwise it's going to be 0. and suppose we've got data on working live earnings y2y this dihg so we know whether someone went to university or not and the vector background characteristics which i'm going to call x i and i'm then i'm going to assume that second period earnings for both um h e and non energy so whether or not you've got university are given by the following minsa type earnings equation um so that's y2y will equal e to the power of alpha zero plus alpha one d i h e plus x i beta so x i remember is our vector of observables plus a ui which is an unobservable component that we as the econometricians don't observe so don't worry if you haven't seen this before um we're gonna get to something kind of more manageable in a minute so um but first note that i've assumed here that the alpha one is fixed across individuals which we're gonna relax to some extent later so um if we substitute this expression now into the expression for rho i we're gonna get row y equals rho so the rho is going to be fixed and that's going to be e to the power of alpha one minus one um which is actually if you um are familiar with this sort of thing almost the same thing as rho so row and alpha one are going to be pretty near each other um in fact alpha one is going to be um uh what we might call that a log point return um and um that knowing that alpha one is going to allow us to get at the row um so if we take logs of both sides of this earnings equation we're going to get this which might look um more familiar to people um so we have log earnings on the left hand side and then something that very much looks like a regression equation and ui is going to be a random error term from the point of view of the econometrician i.e us as the researchers um so can we estimate alpha one by overlaps regression um so which is you know if you've done any eclipse you'll have seen that um so the crucial assumption here is the central or less assumption um so the expected value of ui given the x i um and d i h e is going to be zero and that's what we have to assume so that implies that conditional and the other regresses each regressor has to be uncorrelated with ui and as we're interested in the alpha one the crucial condition for us is this one down here so um conditional on the observables our um error term ui has to be uncorrelated with this choice indicator diag and otherwise you're going to get what's called limiting fiber bias or selection bias so what does this assumption mean for people who haven't done very much econometrics so for instance this ui may contain some unobserved factors that are valuable in the labor market so you might think of charm or self-confidence or strong work ethic or curiosity as things that are valuable in the labor market i.e allow you to get a high income but which a researcher can't observe and this assumption then is going to say that conditional on what we observe about people these unobserved factors are uncorrelated with uh whether someone goes to university so that means looking at people with the same observable characteristics that we've got the same marks at school um who are from a similar socioeconomic background and so on do people who are are people who are uh charming self-confidence etc and no more likely or unlikely to go to university and if that's plausible then um then i think we're good um if it's not true the ols estimate so the estimate that comes out of this method of um estimating the financial returns higher education is going to be picking up differences in the effect of these unobservable factors um which will then falsely be attributed um to going to university um so how can we make progress in thinking about this assumption um well one possible guide is the theory from the previous sections we showed that in the model with government subsidies funded by graduate tax we're going to have um people going to university if and only if this inequality holds so if we then substitute in the equation for the second period earnings we're going to get this thing here and as you can see the a lot cancels out here including the ui because background factors equally affect the utility if you go to university and if you don't go so in the model the dihe does not directly depend on ui um so we could still have it be correlated um if the phi ih e the finite non hg or the y one i known hd are correlated with the ui even conditional on our vector observables x i so again to think about the intuition here so we said well maybe curiosity is part of ui and that may have positively affect earnings and might be positively correlated with diage so might make you go to university because if you're curious you might have a higher non-monetary value of higher education phi iag then what we would overestimate we would overestimate the financial benefit of um going to university alpha one um because we would falsely attribute to going to university what really is a result of curiosity um in contrast we might think that you know strong work ethic for instance is also part of ui and that positively affects earnings but is negatively correlated with diage through y1i non-hg so people who have strong work ethic can earn more if they go to universe if they don't go to university in while they're young then we would underestimate the alpha one so more generally the ui is only going to cancel out from the ac decision rule in our particular model that i've outlined um basically because i've rigged it up in this way so um in that depends on the douglas preferences that i've assumed and it depends on the earnings function so the general takeaway here is that and the ui could affect the ag participation decision diage either directly or indirectly as i've outlined um but um now um finally moving on to kind of what actual estimates are we're going to assume that this works so um the ols the yeah the oxygenated assumption holds um that i've shown you um and the data set where this was most plausible for england is the leo data set which contains linked school records university records and tax records for everyone who took gcses in england since 2002 and what you can do with this data set is look at annual earnings at age up to age 29 because basically the um people who took gccs in england in in 2002 aren't actually much older than that yet so the oldest year we catch them in the data is age 29 um and then the argument basically is that because we observe so much background information about them their higher education choice might be as good as random conditional on those characteristics so the x i and the kind of theoretical discussion earlier um and just to give you a sense of what background information we've got in leo so one is prior attainment which is really important so and gcse results and a level choices and results and so on um but we also have a socio-economic background so things like free school meals um deprivation of the neighborhood where people grew up whether they went to independent school and so on and finally demographic characteristics so things like ethnicity your region gender and whether someone speaks english as an additional language and so here are some results um from work we've done at ifs um so uh ignore those um firstly columns but um in the final column here we see kind of what happens if you um control for all of these factors um and as you can see these results are done separately from men and women and you see that that kind of alpha one you get here is reported in the table so that's 0.04 for men and 0.23 for women and that translates into a percentage return of 4 for men and 26 for women so the idea here is that um people who went to university uh or men who went to university at age 29 um earn on average four percent more and women earn on average 26 more as a result of going to university so because they went to university um [Music] that's that's what comes out of this um now so far we have assumed that the return to higher education is the same for everyone um in in the theoretical sections i've just shown that that's clearly not true between men and women um but actually returns also vary usually by subject studied by institution and of the life cycle so in the last few slides i'm going to show you some evidence of heterogeneity along these dimensions so first here are financial returns by subject so in black you have the um what you can call the raw difference so that's just the difference between um how much people earn who study these subjects at age 29 um compared to people who um got um at least five days start to see at gcse but didn't go to university at all and that's the blue line here and you can see that everyone who went to university all women who went to university um earn more than on average than in all subjects than women who didn't go to university which is in the blue here but then when you control for all the background characteristics so most importantly um school attainment then you see that the kind of estimated returns so how much of these differences are actually um truly the effect of higher education um goes down quite a lot um but we still see that medicine and economics which have the highest earnings also have the highest returns where some other subjects such as creative arts and social care the returns are pretty small this is the same for men so uh the kind of it all looks pretty similar to and the results for women but interestingly here it's all kind of shifted down a little bit so there are actually some men who study some subject or who study creative arts actually earn less at age 29 um than men who don't go to university at all but have five at least five they start to see at gcse um and um if you um if you look at the returns actually returns are negative for at age 29 for quite a few subjects [Music] so we've looked at subjects this is institutions um so what you see here is that uh people who went to russell group universities tend to get higher returns than people who went to um some non-russell group universities so especially um some of the kind of newer universities have relatively low returns sorry this is women um if you look at men again the picture is similar but it's kind of all shifted down a bit um because in general men have lower returns by age 29 um but um actually this changes over the life cycle so if you do some simulation i said to you you can't actually um observe people in the data beyond age 29 but um you can do some simulation based on older cohorts and what we see is that um or what we'd expect is that uh men appear to catch up quite a lot so um by age 40 the returns are actually similar to those of women around 30 percent um so to conclude there are many reasons why people do or do not go to university financial or otherwise and a model of higher education choices can help clarify how these factors together influence choices um secondly choices in a completely private higher education market may lead to under investment in higher education and i've given three factors of why that might be so one is financial market imperfections and externalities but also private sub optimal decisions and um these are reasons for governments to provide loans and subsidize higher education which is what we observe in practice and finally financial returns to higher education can be estimated um using ordinary least squares regression and provided that the exogenetic assumption holds as i discussed um or with other methods um uh the um just to show you give you some um oils as results for england um which i um discussed so um to recap the age 29 financial return is um around 25 for women and 5 for men um [Music] and but there is large heterogeneity in these returns first of all by institution but also by subject and over the life cycle where we see men catching up um that's all for me thanks very much for listening great thanks ben um that was really interesting talk and i think quite topical given the ongoing debates in about funding in in higher education we're getting in quite a few questions now um so i think i'll just jump into those and please everyone continue to submit your questions or vote for questions that you want to be asked on sido app so the first one we have is what are the advantages and disadvantages of the government sticking with income contingent loans rather than switching to a graduate tax um okay so um the um yeah so i guess i've sort of equated the two in my lecture i guess so i said well for most students in england um the the system already acts like a graduate tax um and so these two options are actually more similar than some people um than one might think um so that's the first thing to say um so the difference comes in for people who have very high earnings so um we currently estimate that slightly less than 80 percent of people are going to um pay off their loans um over the so it's slightly less than 80 people are not going to pay off their loans in the current system so for them the system is um just like a graduate tax at the moment so that leaves slightly more than 20 of people for whom it's not um who are the highest earners um now for them a graduate tax would mean that um they would have to pay more um because well obviously they couldn't um pay off their loans um under that system so um yeah so then the the sort of other difference here is that currently the repayment period is limited to 30 years and presumably a graduate tax would apply over the whole life cycle but one could design that in different ways why the advantages and disadvantages well um so with the graduate tax in principle um you could shift more of the burden of paying for higher education to the highest earners and some might see this as an advantage um there's also kind of a potential disadvantage in the sense that you might think it's unfair um in in in the sense that some people are going to pay way more for their education than it it cost um i wouldn't want to take a stance on that um but i think the most difficult thing with a graduate tax is actually operational um so um you know how do you think about people who complete partial degrees how do you think about people who change subject and so on um how how are you going to deal with say i mean now it's kind of easier with with eu students because they now pay international fees but um what about say are students who are still eligible for the same types of loans um you know how are they going to be taxed in ireland um so so i think there are some kind of real operational difficulties that make it make it hard to have a granular tax great thanks so our next question which has been voted up is why are men's estimated returns so different from women's at age 29 um so this is kind of a difficult one because we um we basically can't tell from the data that we have um we um strongly or i strongly suspect that it has a lot to do um with fertility choices that come at a different time for um graduate women typically than for those who don't go to university so um basically women who don't go to university tend to have children earlier um and then tend to kind of drop out of the labor market for some time and or would use their hours so that's why at the sort of early stage in the life cycles like age 29 the returns look very large um for um for women because um well the kind of comparison group many of them are working part-time or only partly in in the labor market for part of the year um and we don't have that kind of same effect for men because um having you know becoming a father basically doesn't tend to have the same labor market effects and become as becoming a mother yeah that makes sense um so the next question we've had in is is there an optimal level of tuition fees and how would this relate to the current level um well so i guess um well basically it's hard to know um so in theory you know there's an optimal subsidy um in any model that you're going to write down um in practice um what you want to achieve is um from from sort of purely economic point of view um you know there might be lots of other kind of non-economic reasons for having a student loan system a certain way but from purely a non-economic view from a pretty economic point of view if you want um to um you want those people to attend university um for whom it's socially optimal to do so um and i've sort of argued that it's probably like it's probably socially optimal for more people to go then it's privately optimal to go um because of externalities and so on um so that kind of would suggest that you don't want to um you don't want to put the whole um cost of going to university on students you want to subsidize it as a government and that's likely to be optimal and that's in fact what we see in practice um now how much you want to subsidize it whether it's um whether it's you know socially optimal to subsidize it as much as say the english government do or does or maybe it's optimal to subsidize it as much as it's subsidized in scotland which is essentially more um or as much as you see in say denmark which is even more um i think that's an open question and we have a slightly related question which is is there any data on how tuition fees would rise or fall if determined by the market rather than being set by government um well it's well so the short answer is not really um so you know or as i kind of alluded to basically all systems that we see um in the world uh have a substantial role for government and higher education funding so there isn't a purely private market at all um one thing we did see in the uk is that um you know when um the cap for um the fees that could be set was um raised from around three thousand pounds to nine thousand pounds in 2012 and basically all universities charge the maximum amount so that kind of suggests that if you remove this part now fees would be higher than 9 000 still at many universities and um you know that same thing is also kind of suggested by looking at um fees for international students which also tend to be much higher than fees for domestic students so um so yeah i think um it's safe to assume that if we didn't have a cap on fees fees would be higher than they are now um and now we actually have a question on not on c caps but student number caps so how would student number caps affect tuition fees and the graduate earnings premium okay oh that's a hard one so um the so for tuition fees well so as i said most universities charge the maximum level of tuition fees so you know if the cup is not changed basically the answer for tuition fees is not at all um so um so yeah we would expect that if the if the caps stay the same tuition fees same whatever is done to um to student number numbers um the other thing is um the other question was oh the other part of the question was how would it affect the graduate earnings premium um i think that is a very difficult question um basically we don't know um so you know economic theory might suggest which is kind of what um the question probably alludes to that you know if you increase the supply of graduates um maybe the graduate premium is going to fall and so similarly if you restrict the supply of graduates through these number caps the graduate premium might be higher than it otherwise would be um that argument makes a lot of sense the only thing is that um this is not what we've observed in the data so you know in the in england and in other countries we've seen a massive expansion of the higher education sectors so many more people now go to university than used to and yet the graduate premium has been remarkably stable so another question we have is can we use the subject and institution combinations i think that means the returns estimates uh to determine where an individual should or shouldn't attend university um well i mean it's it's it's certainly useful information i think so you know there's a lot of different factors that individuals should consider when making their university choice and you know returns are clearly not everything you know and um and especially the returns for other people so every individual may have a different experience but i think looking at past cohorts so you know how people have done who've attended this course in the past um is is one kind of um piece of the puzzle and one important piece of information that people may want to look at maybe just to get an idea and certainly that's kind of what we're hoping with our research to achieve that it would provide useful information for prospective students as well as for policymakers great um so i think we're in the final couple of minutes so i'll try and ask a couple more questions um while we still have you here um so given the dependency on remote education to what extent has coven 19 affected the quality of the outcome of higher education um yeah so that's kind of too early to say i think so you know given that uh the people who've kind of most been most affected by this remote education aren't in the labor market yet um we don't know anything yet about um what their outcomes are going to be and it's going to be a while because um people tend to take a while to settle into the labor market um so you know by the time these people are in their late twenties and the people who will kind of finish or are finishing university now and we're gonna have a clearer picture but that's still um several years away i think sure uh so i think we'll go for a final question which is we'll go for a technical question what sources of potential bias are you most concerned about in your ols estimates of returns to higher education um yeah so i think um the i it's a great question i think um so the um the um the most concerning ones i think um are that um there's this kind of um unobservable factors um like um you know i mentioned curiosity but there's many others um that may um that may um help you in the labor market and um that would also make you want to go to university so maybe you know in general motivation might be one of those so if you're a more motivated universal individual maybe um you want to both go to university and um and and uh get a and you're going to get a better job or like ambition is also another one of those um so those are all factors that might lead us to overestimate um their returns to higher education on the other hand you know as i mentioned it doesn't all necessarily go that way so there might also be other factors that might may kind of um pull in the other direction um so i think yeah i i as you allude to i think um because of those potential unobservables uh these estimates should be taken with some um grain of salt but um we do believe that they're the best estimates available and we do think they are useful for people too um both for policy makers thinking about the design of the system and for uni for individuals making decisions great perfect timing we're now on five o'clock so i think we'll wrap up now um big thanks again to ben and for our audience for joining and participating in the discussion another reminder that our final ifs public lecture is this time next week and thanks again for everyone for attending