The younger, the better? Age-related differences in academic performance at university

Abstract

In this paper, we investigate differences in academic performance among students of different ages within the same cohorts using a unique database of students at Bocconi University. Our data allow to control for potential selection effects as well as for differences in cognitive ability, as measured by an attitudinal entry test. Contrary to most of the existing evidence for younger pupils, we document that at the undergraduate level, youngest students perform better compared with their oldest peers. This finding is only partly explained by differences in cognitive ability and rather seems to be associated with differences in social activities.

This is a preview of subscription content, access via your institution.

Fig. 1
Fig. 2
Fig. 3
Fig. 4

Notes

  1. 1.

    Although, Crawford et al (2007) do find some effects on 18-year olds and Bedard and Dhuey (2006) document a higher probability of college enrollment for the relatively older students in British Columbia, Canada, and in the USA.

  2. 2.

    Although this is still a rather controversial result. For example, both McEwan and Shapiro (2008) and Elder and Lubotsky (2009) find opposite results.

  3. 3.

    The survey covers several countries but, for comparison with our Bocconi data, we limit our analysis to the Italian sample.

  4. 4.

    Notice that such a policy does not necessarily maximize average performance but it might still be desirable if the social objective function is aimed at guaranteeing equality of opportunities.

  5. 5.

    As we discuss later in this section, the main data used in our analysis include students born between 1976 and 1979. The school system described in this section is the one pupils born in these years were subject to. The system has been modified since then, with the most important change occurring in 1999, when compulsory schooling was increased by 1 year. None of the students in our data is affected by this change.

  6. 6.

    Generally, the pre-1999 legal requirement imposes all pupils aged between 6 and 14 to attend school. However, students can satisfy the requirement by either passing the junior high school final exam (thus, even before they turn 14) or by having attended at least 8 years of full-time schooling as they turn 14.

  7. 7.

    Some technical institutes offer 3- or 4-year degrees. However, only students graduating from 5-year high schools can enroll at university. Students who want to go to high school cannot not be refused a place (at least within the public system) regardless of their previous school performance.

  8. 8.

    In educational systems where students are streamed into classes on the basis of observed performance already at the very initial stages of their academic careers, early differences in childhood development are perpetuated over a long period of time. In fact, the oldest pupils typically end up in the highest tiers (i.e., the classes of the best performing students) simply because at very young ages they are substantially more developed (Allen and Barnsley 1993). This type of mechanism has been documented to be very important in sports, where a large fraction of professional players are indeed born in the earliest months of the year (Dudink 1994; Barnsley et al 1992; Helsen et al 2005).

  9. 9.

    The most complete version of our data covers all students enrolled from 1989 to 2004. We exclude the earlier cohorts because data on the admission tests are not available. Moreover, we also limit the sample to cohorts with at least 80% of students who have completed their degrees, thus leading to our focus on the 1995–1998 freshmen.

  10. 10.

    Although this issue was not formally regulated, it was very uncommon to admit a new student directly into a grade higher than 2.

  11. 11.

    The few drop-outs and missing values lead to a loss of about 8% of the original observations.

  12. 12.

    Although in the main text, we present results based on month of birth, all our findings are confirmed when we measure age in days. However, Bocconi does not allow us to distribute the exact day of birth to external users and, therefore, only results produced using information on month and year of birth can be replicated by external researchers.

  13. 13.

    Exams are graded on a scale of 1 to 30, with 18 being the minimum pass grade and graduation marks range from 66 (pass) to 110. Such a peculiar grading scale comes from historical legacy: while in primary, middle, and high school students were graded by one teacher per subject on a scale of 0 to 10 (pass equal to 6), at university each exam was supposed to be evaluated by a commission of three professors (11 for the graduation defense), each grading on the same 0–10 scale, the final mark being the sum of these three. Hence, 18 is pass and 30 is full marks. Apart from the scaling, the actual grading at Bocconi is performed as in the average US or UK university.

  14. 14.

    To the extent that we will use variation across Italian provinces in our analysis, a similar concern applies to our approach as well. An underlying assumption of our identification strategy requires the timing of births to be uncorrelated with province-level unobservables. Compared with other studies (Bedard and Dhuey 2006; Crawford et al 2007), however, we can argue that all the institutions are kept constant across Italian provinces (at least formal institutions). Additionally, the exact source of variation for our instrument is at the level of the interaction of provinces and years of birth.

  15. 15.

    If schooling s i is interpreted as actual taught material, then it is indeed constant within cohort for all students in our sample. However, if s i is intended as mere schooling time, than it does vary for the few re-takers, who repeated the same grade one or more times and, thus, spent more time in school although they were taught the same material.

  16. 16.

    One may question this assumption if parents try to target the actual month of birth for reasons related to work, parental care or even directly to school entry. However, since conception and pregnancy still are largely stochastic events, it is unclear whether such targeting can go beyond the seasonal frequency.

  17. 17.

    For comparison with Bedard and Dhuey (2006), note that they define their instrument as “assigned age” or − m i in our terminology.

  18. 18.

    Notice that the presence of non-compliers does not necessarily affects monotonicity. The instrument may have no effect on the behavior of some units (the non-compliers) but it should not have effects of opposite signs across units, which is what makes monotonicity fail.

  19. 19.

    Bedard and Dhuey (2006) are more concerned with endogeneity due to grade retention (which leads to late enrollment, in our terminology) than with early schooling. Moreover, they seem to believe that the parents of children born on the left side of the cut-off date (those born in December in our setting) would “hold them out of school so that they enter kindergarten a year late (who are positively selected)” (page 1442). Figure 1 shows that in our setting early school entry is the most important source of selection and the descriptive statistics in Table 3 suggest that early enrollees are positively selected. In general, it is unlikely that the relationship between relative age and month of birth is monotonic in all the cases considered in Bedard and Dhuey (2006).

  20. 20.

    We define the province of origin as the province where the student attended high school. Very similar results are obtained using the province of birth. We simply choose the variable with the fewer missing values.

  21. 21.

    One caveat with our instrument comes from data limitations. In fact, we only have information on private schools currently recognized by the government and it is impossible to reconstruct which ones were already active before the year 2000. We are told by officials at the Ministry of Education that turnover of schools is very limited.

  22. 22.

    Notice that this is argument is very similar to those used to justify the use of the initial stock of migrants as an instrument for current migration, as in Altonji and Card (1991) or in Card (2001, 2005).

  23. 23.

    We also have data on private primary and secondary schools but including these in the instrument set reduces efficiency, probably because of the high correlation between the incidence of these various types of schools (0.98).

  24. 24.

    In the period covered by our data, Bocconi offered six types of bachelor degrees with different specializations. The most popular ones were management (with acronym CLEA) and economics (CLEP). A third option was a more academic version of the BA in economics (DES). Other programs specialized in financial markets (CLEFIN), public administration (CLAPI), law, and economics (CLELI).

  25. 25.

    Clustering the standard errors at the province level would allow for a more flexible structure of heteroskedasticity but would still be subject to the Moulton problem (Moulton 1990) and to a more substantial small sample, given that clustered standard errors are consistent as the number of clusters goes to infinity. In any event, in unreported results we find that clustering at the province level yields similar results to those reported in the main text.

  26. 26.

    Specifically, we found that the estimated effect is very similar across income groups and larger among the low ability. Results are available from the author upon request.

  27. 27.

    A good fraction of students (about 10% in our data) graduate after the official duration of the degree (4 years for all but one program). Late graduation is a well-known problem of the Italian university system, see Garibaldi et al (2011) for an analysis of this phenomenon on our same data.

  28. 28.

    Only very recently, a degree program in law has been introduced.

  29. 29.

    Notice also that any process that is driven merely by ability should already be taken care of by our estimates that condition on the entry test score.

  30. 30.

    These alternative versions of Fig. 4 are available from the authors upon request.

  31. 31.

    We define seasons exclusively on the basis of months so that results can be replicated by external researchers. Namely, December, January, and February are coded as winter; March, April, and May as spring; June, July, and August as summer; and September, October, and November as autumn.

  32. 32.

    When we focus on regular students only, variation in relative age is, indeed, perfectly collinear to variation in dates of birth.

  33. 33.

    Note that many studies focus on estimating the effect of early schooling, i.e., age0 i . However, given the collinearity issues that we discussed in Section 3, this can be equivalently interpreted in terms of relative age.

  34. 34.

    The ISAS survey was conducted within a project by the University of Padua and the Max Planck Institute for Demographic Research.

  35. 35.

    The original question reads: “How often have you had sexual intercourse during the last three months?” Possible answers are: never, less than once a month, once a month, two or three times a month, once a week, two or three times a week, four or five times a week, almost every day. Based on self-reported interval-coded answers, we construct a continuous estimated frequency of intercourses taking the mid-point of each interval. Results are robust to alternative specifications.

References

  1. Allen J, Barnsley R (1993) Streams and tiers: the interaction of ability, maturity and training in systems with age-dependent recursive selection. J Hum Resour 28(3):649–659

    Article  Google Scholar 

  2. Altonji J, Card D (1991) The effects of immigration on the labor market outcomes of less-skilled natives. In: Abowd J, Freeman RB (eds) Immigration, trade and labor. University of Chicago Press, Chicago

    Google Scholar 

  3. Anger S, Heineck G (2010) Do smart parents raise smart children? The intergenerational transmission of cognitive abilities. J Popul Econ 23(3):1105–1132

    Article  Google Scholar 

  4. Angrist JD, Pischke JS (2009) Mostly harmless econometrics: an empiricist’s companion. Princeton University Press, Princeton

    Google Scholar 

  5. Barnsley RH, Legault P, Thompson AH (1992) Family planning: football style, the relative age effect in football. Int Rev Sociol Sport 27(1):77–88

    Article  Google Scholar 

  6. Bedard K, Dhuey E (2006) The persistence of early childhood maturity: international evidence of long-run age effects. Q J Econ 121(4):1437–1472

    Google Scholar 

  7. Berlinski S, Galiani S, Manacorda M (2008) Giving children a better start: preschool attendance and school-age profiles. J Public Econ 92(5–6):1416–1440

    Article  Google Scholar 

  8. Berlinski S, Galiani S, Gertler P (2009) The effect of pre-primary education on primary school performance. J Public Econ 93(1–2):219–234

    Article  Google Scholar 

  9. Billari FC, Caltabiano M, Dalla Zuanna G (2007) Sexual and affective behaviour of students. An international research. Cleup Editrice, Padova

  10. Black SE, Devereux PJ, Salvanes KG (2009) Too young to leave the nest? The effects of school starting age. NBER working paper: 13969

  11. Buckles K, Hungermann DM (2008) Season of birth and later outcomes: old questions, new answers. NBER working paper: 14573

  12. Card D (2001) Immigration inflows, native outflows and the local market impacts of higher immigration. J Labor Econ 19(1):22–64

    Article  Google Scholar 

  13. Card D (2005) Is the new immigration really so bad? Econ J 115:F300–F323

    Article  Google Scholar 

  14. Carneiro PM, Heckman JJ (2003) Human capital policy. In: Heckman J, Krueger A (eds) Inequality in America: what role for human capital policy. MIT Press, Cambridge

    Google Scholar 

  15. Crawford C, Dearden L, Meghir C (2007) When you are born matters: the impact of the date of birth on child cognitive outcomes in England. Institute for Fiscal Studies, London

    Google Scholar 

  16. Cunha F, Heckman JJ (2007) The technology of skill formation. Am Econ Rev 97(2):31–47

    Article  Google Scholar 

  17. Cunha F, Heckman JJ (2010) Investing in our young people. NBER working paper: 16202

  18. Dhuey E, Lipscomb S (2006) What makes a leader? Relative age and high school leadership. Econ Educ Rev 27(2):173–183

    Article  Google Scholar 

  19. Dudink A (1994) Birth date and sporting success. Nature 368:592

    Article  Google Scholar 

  20. Elder TE, Lubotsky DH (2009) Kindergarten entrance age and children’s achievement: impacts of state policies, family background and peers. J Hum Resour 44(3):641–683

    Google Scholar 

  21. Fredriksson P, Ockert B (2005) Is early learning really more productive? The effect od school starting age on school labor market performance. IZA discussion paper: 1658

  22. Garibaldi P, Giavazzi F, Ichino A, Rettore E (2011) College cost and time to complete a degree: evidence form tuition discontinuities. Rev Econ Stat (in press)

  23. Goodman A, Sianesi B (2005) Early education and children’s outcomes: how long do the impact last? Fisc Stud 26(4):513–548

    Article  Google Scholar 

  24. Heckman JJ, Masterov DV (2007) The productivity argument for investing in young children. Rev Agric Econ 29(3):446–493

    Article  Google Scholar 

  25. Helsen WF, van Winckel J, Williams MA (2005) The relative age effect in youth soccer across Europe. J Sports Sci 23(6):629–636

    Article  Google Scholar 

  26. Imbens GW, Angrist JD (1994) Identification and estimation of local average treatment effects. Econometrica 62(2):467–475

    Article  Google Scholar 

  27. Jones BF (2010) Age and great invention. Rev Econ Stat 92(1):1–14

    Article  Google Scholar 

  28. Lutz W, Skirbekk V (2005) Policies addressing the tempo effect in low-fertility countries. Popul Dev Rev 31(4):699–720

    Article  Google Scholar 

  29. Mayer SE, Knutson D (1999) Does the timing of school affect how much children learn? In: Mayer SE, Peterson PE (eds) Earning and learning: how school matter, chap 4. Brookings Institution, Washington, pp 79–102

  30. McEwan PJ, Shapiro JS (2008) The benefits of delayed primary school enrollment. discontinuity estimates using exact birth dates. J Hum Resour 43(1):1–29

    Google Scholar 

  31. Moulton BR (1990) An illustration of the pitfall in estimating the effects of aggregate variables on micro units. Rev Econ Stat 72(2):334–338

    Article  Google Scholar 

  32. Persico N, Postlewaite A, Silverman D (2004) The effect of adolescent experience on labor market outcomes: the case of height. J Polit Econ 112(5):1019–1053

    Article  Google Scholar 

  33. Rizzi EL, Dalla Zuanna G (2007) The seasonality of conception. Demography 44(4):705–728

    Article  Google Scholar 

  34. Salthouse TA, Schroeder DH, Ferrer E (2004) Estimating retest effects in longitudinal assessments of cognitive functioning in adults between 18 and 60 years of age. Dev Psychol 40(5):813–822

    Article  Google Scholar 

  35. Skirbekk V (2005) Why not start younger? Implications of the timing and duration of schooling for fertility, human capital, productivity and public pensions. International Institute for Applied Systems Analysis, Laxenburg

  36. Skirbekk V, Kohler H, Prskawetz A (2004) Birth month, school graduation and the timing of births and marriages. Demography 41(3):547–568

    Article  Google Scholar 

  37. Staiger D, Stock JH (1997) Instrumental variable regressions with weak instruments. Econometrica 65(3):557–586

    Article  Google Scholar 

  38. Thompson AH, Barnsley RH, Battle J (2004) The relative age effect and the development of self-esteem. Educ Res 46(3):313–320

    Article  Google Scholar 

  39. Wilson G (2000) The effects of season of birth, sex and cognitive abilities on the assessment of special educational needs. Educ Psychol 20(2):153–166

    Article  Google Scholar 

Download references

Acknowledgements

We would like to thank the Bocconi University for financial support through its “Basic Research Fund” and also for granting access to its administrative archives for this project. In particular, the following persons provided invaluable and generous help: Giacomo Carrai, Mariele Chirulli, Mariapia Chisari, Alessandro Ciarlo, Enrica Greggio, Gabriella Maggioni, Erika Palazzo, Giovanni Pavese, Cherubino Profeta, Alessandra Startari, and Mariangela Vago. We are also indebted to Elsa Artadi, Piero Cipollone, Christian Dustmann, Giacomo De Giorgi, and Matt Harding for comments and suggestions on earlier drafts. We would also like to thank seminar participants at the University of Lausanne, the University of Padua, and the C6 Csef-Igier Symposium. Michela Braga provided excellent research assistance. Data on the ISAS survey have been collected in a project of the University of Padua and the Max Planck Institute for Demographic Research, coordinated by Gianpiero Dalla Zuanna, whose willingness to share their data with us is gratefully acknowledged. Two anonymous referees provided extremely useful comments.

Author information

Affiliations

Authors

Corresponding author

Correspondence to Michele Pellizzari.

Additional information

Responsible editor: Junsen Zhang

Electronic Supplementary Material

Below is the link to the electronic supplementary material.

(PDF 71.9 KB)

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Pellizzari, M., Billari, F.C. The younger, the better? Age-related differences in academic performance at university. J Popul Econ 25, 697–739 (2012). https://doi.org/10.1007/s00148-011-0379-3

Download citation

Keywords

  • Age effects
  • Cognitive ability
  • Academic performance

JEL Classification

  • J1
  • I23
  • I24