Many of our readers are students, and some have come to us wondering whether they should start a university degree or complete one they have already started. One thing to consider in making this decision is what effect getting a degree will have on your lifetime earnings. So in this post we summarise our reading of some of the empirical literature on this question, mostly focused on the UK.
There appears to be a consensus in the empirical literature that getting a degree provides a large financial return on the costs in increased lifetime earnings (generally better than an investment with a 10% return and maybe closer to 15%).
The most common way of studying the question of economic returns is to use correlations in data containing information on education, earnings and other variables (performing “ordinary least square regression” on it).
The obvious worry with this method is that the same abilities that help earn a higher income might cause people to go to university rather than the other way around. This is called ability bias. The standard view in the literature, however, is that this issue only has as minor effect on estimates of the return to education.
The literature here supports the common sense position that an undergraduate degree is generally a good investment in your career.
What I did
I began by searching Google (UK) for popular news pieces on the value of a degree, and then some academic articles, I found further articles through seeking out relevant articles cited by or citing previously viewed articles. This is not a systematic literature review.
A note about what the question is
As with any decision, getting a degree has some costs and some benefits. To simplify here we are only considering some purely monetary costs and benefits. The costs considered are firstly the fees, but, more importantly, also the opportunity cost of spending time studying that could have been spent working and earning money. The benefit is just the increased earnings you can expect from getting a degree over your entire lifetime.
It is important to keep in mind that the costs in this case are mostly incurred immediately, whereas the benefits only begin after you graduate and continue far into the future. This is important because resources can be invested to generate compounding benefits.1 The way economists deal with this is to discount the value of future benefits and costs by some constant percentage each year (called the discount rate). We can then look at what discount rate we would need to have for us to be indifferent between getting a degree and not doing so, this is called the rate of return.
To give some simple real world examples for comparison, the rate of return of investing in the S&P 500 has been about 7% over the last 60 years and the rate of return on paying off credit card debt will typically be in the range of 15-20%. The below studies suggest that the rate of return on a degree is between these two but probably closer to the rate of return on paying off credit card debt.
What does the literature say?
I found a number of studies published within the last few years looking at the returns to a degree in the UK. They all put the return significantly above 10% and maybe closer to 15%. A number of studies that the average benefit of the a degree on income amounts to a return of more than 10% return on the cost of tuition in the UK (including the opportunity cost of not earning wages while at university).23
An important point about these studies is that they are all based on simple analyses of correlations. The most obvious worry about such methodology in this case is that the correlation between education and income might be explained by a variable not included in the analysis. In particular you might worry that university graduates earn more because the most employable high school graduates are the ones who decide to enrol in university (even out of those whose grades qualify them for university). This issue with regression estimates of the value of education is known as “ability bias” and the standard view within the literature is that this is likely to be small. The most commonly cited source in support of this is the very widely regarded4 review Card(1999) which concludes that ability bias may account for around 10% of the effect found by regression estimates (so if the estimated return to education is 10% then the true rate of return is 9%).5 Card based this assessment mainly on studies using identical twins but other studies have tried to attack this problem in other ways without finding a large ability bias in simple regression estimates.678
Thus the academic literature appears to strongly support the common sense position that a university degree is a valuable investment for increasing your lifetime earnings.
What About Other Countries?
We did not investigate returns in other countries as closely as the UK. However, there was some reference to other countries in the literature we looked at: Card (1999) is a review of international literature on the returns to education so his conclusions about ability bias appear to hold internationally. Harmon & Walker (2001) compare the returns on education in many European countries (p. 23, fig 2.5), and finds that the UK has the second largest return to education, with the European average significantly lower at about 6.5%. On the other hand Ashenfelter et al (2000, p. 9) found no evidence to suggest significant differences among non-US countries (including the UK), but did find the US to have slightly higher returns to education.9 Also see these posts by Dylan Matthews on this issue focussing on the US. He cites a range of studies generally putting the figure on the returns to a degree or education more broadly between 10 and 20 per cent.
I will conclude with some related questions relevant to 80,000 Hours but which this brief foray into the literature does not touch on:
The studies looked at here do not say why getting a degree increases one’s earnings. For instance is it because universities teach important skills or is it because it signals to employers that you are the sort of person who can get a degree? There is some literature on this (Dylan Matthews refers to it in the blog posts linked above) and we suspect that both are important but we haven’t looked into it yet.
Does the boost in income reflect generally higher output, or is the effect confined to income? This seems to depend on the previous question, if the value of a degree lies primarily in the signal it sends to the employer, then we wouldn’t expect it to have much direct effect on general productivity. However, we would guess that the skills and network you learn in university are also important. In this case it is plausible that the higher incomes of graduates reflects higher overall productivity.
Another question is how valuable different sort of degree are: what are the best disciplines? Is a degree from Oxford or Cambridge more valuable than one from University College London? How valuable is it to get good grades? Some of the literature10 looked at did give estimates for different degree disciplines and even the value of getting a 2.1 versus a 2.2 but it is unclear how robust these are11). On the value of going to a better university, Walker and Zhu (2013, p. 24) cite a small literature suggesting a small effect on wages in the UK and a larger one in the US.
You might also be interested in
Notes and References
Another reason to discount may be that you expect that as time goes on and problems get solved, the best uses of altruistic resources will become progressively less cost effective. ↩
The most recent studies I found which explicitly calculated an internal rate of return for degrees in the UK were Conlon, G. and Patrignani (2011) “The Returns to Higher Education” Departmen for Business Innovation & Skills, and Walker, I. and Zhu, Y. (2010) “Differences by Degree: Evidence of the Net Financial Rates of Return to Undergraduate Study for England and Wales”. Conlon and Patrignani (using data from 1996-2009) calculated an average rate of return of 14.9% (see page 14 and page 56 for returns broken down by subject and gender). Walker and Zhu (using data from 1994-2009) calculated returns by degree type and gender in the UK (table A7, p.24 and finds returns of over 15% for women in all degrees, but for men rates of return varying from 30% (for a 2:1 in law, economics or management to 4.2% for a 2:2 in humanities). The most recent study of the return to a degree in UK was Walker, I. & Zhu, Y. (2013) “The Impact of University Degrees on the Lifecycle of Earning: Some Further Analysis” Department for Business Innovation & Skills. They didn’t calculate an explicit rate of return, but using data from 1993-2010 formed an even more optimistic view of the value of education. Walker and Zhu (2013) calculated a net present value of £168k for men and £252k for women (at a discount rate of 3.5%), compared to £121k and £82k respectively calculated by Conlon and Patrignani. ↩
The most recent meta-analyses I found were Harmon, C. & Walker, I. (2001) The Returns to Education: A Review of Evidence, Issues and Deficiencies in the Literature and Ashenfelter, O., Harmon, C. & Oosterbeek, H. (2000) A Review of the Schooling/Earnings Relatioship, with Tests for Publication Bias. Harmon & Walker (2001, p. 26, Table 2.3) find a return to education of 9.6% for men and 12.2% for women. While Ashenfelter et al. (2001, p.16) find a return of 3.4% in 1974 increasing 2% per decade, meaning about 7.5% when in 2001 was published and about 10% now (if trends continued). Note that the results of these meta-analyses are not strictly comparable to the studies referred to in note 1. First, the meta-analyses look at the return to a year of education rather than the return specifically to getting a degree. Secondly, their methodology is different, the studies included in the meta-analyses calculate return to education as the change in log wages with an extra year of schooling. This is based on a calculation which relies on simplifying assumptions (such as no university fees and assuming everyone works the same number of years) in Willis, R.J (1986) “Wage Determinant: a Survey and Reinterpretation of Human Capital Earnings Functions” pp. 531-532 ↩
Google Scholar records 2769 citations (December 6 2013). ↩
Card, D. (1999) “The Causal Effect of Education on Earnings” p. 1855 “Consistent with the summary of the literature from the 1960s and 1970s by Griliches (1977, 1979) the average (or average marginal) return to education in a given population is not much below the estimate that emerges from a simple cross-sectional regression of earnings on education. The “best available” evidence from the latest studies of
identical twins suggests a small upward bias (on the order of 10%) in the simple OLS
Blundell, R., Dearden, L. & Sianesi, B. (2004) “Evaluating the Impact of Education on Earnings in the UK: Models, Methods and Results from the NCDS” “These model and methods are subsequently applied to high quality data the British 1958
NCDS birth cohort to estimate the wage returns to different educational investments and to
illustrate the sensitivity of the different estimators to model specification and data availability.
This data is sufficiently rich to allow the comparison of the various methods, and in particular
to assess the importance of test score and family background information in generating
reliable estimates The overall returns to educational qualifications at each stage of the educational process
remain sizeable and significant, even after allowing for selection and heterogeneity in the
education response parameters.” Executive Summary ↩
Walker, I. & Zhu, Y. (2013) “The Impact of University Degrees on the Lifecycle of Earnings: Some Further Analysis” Department for Business Innovation & Skills compares data from the British Quarterly Labour Force Survey (QLFS) from 1993 to 2010 to the richer, longitudinal but much smaller British Household Panel Survey (BHPS) Data, which includes family background and high school performance data. ↩
A quite different way is to use what is called an instrumental variable analysis. That is researchers look for cases where something having nothing to do with ability causes some people to undertake education while others do not. To take a random example Brunello, G. & Miniaci, R. (1997) “The Economic Return to Schooling for Italian Men. An Evaluation Based on Instrumental Variables”, Labour Economics vol. 6(4), use a change in Italian law which made it easier to attend college after the law was passed. Perhaps surprisingly instrumental variable estimates of the returns to education are typically higher than regression estimates [Card(1999)]((http://www.stanford.edu/group/scspi/_media/pdf/Classic_Media/Card_1999_Education.pdf) p. 1855 “IV [instrumental variable] estimates of the return to education based on interventions in the school system tend to be 20% or more above the corresponding OLS [regression] estimates.”and Ashenfelter, Harmon & Oosterbeek (2000) p. 8 “With respect to the returns and their corresponding standard errors/t-statistics we see the
pattern emerging clearly – average returns of 6/7% with corresponding IV and twins study
estimated returns of 9%.” Although Ashenfelter et al. (2000, p. 16) argue that most of this difference can be explained by publication bias. ↩
“The pooled results suggest little difference in the estimated returns by
geographical region – countries in this non-US grouping include Finland, Honduras, Indonesia,
Ireland, Netherlands, Portugal and the United Kingdom” ↩
As noted above Walker & Zhu (2010, p. 24) give rates of return broken down by gender, type of major and whether the graduate obtained a 2.1 or 2.2, and Conlon & Patrignani (2011, p. 56) give rates of return based on major and gender. ↩
Walker & Zhu (2013, p. 21) note at least two potential biases in simple estimates of the return to different degree types: 1. It is plausible that students studying different degrees have different levels of ability which could bias the estimates (unlike the possible bias in the estimates of the generic value of education this is unstudied). 2. Some degrees can lead to disproportionately high levels of self-employment among successful graduates whose earnings are not properly recorded in the Labour Force Survey data. 3. Different subject types are studied in different proportions in different universities, so correlations with degree subject might just be capturing differences in university quality. (p. 25) ↩