Springer Nature is making SARS-CoV-2 and COVID-19 research free. View research | View latest news | Sign up for updates

The health returns to schooling—what can we learn from twins?

Abstract

This paper estimates the health returns to schooling, using a twin design. For this purpose, I use data on monozygotic twins from the Midlife in the United States survey. The results suggest that completing high school improves health, as measured through self-reported health, chronic conditions, and exercise behavior, but that additional schooling does not lead to additional health gains. Controlling for certain early life factors that may vary within twin pairs does not alter the main conclusions of this paper.

This is a preview of subscription content, log in to check access.

Fig. 1

Notes

  1. 1.

    Currie and Moretti (2003) and Chou et al. (2010) consider the effect of parental education on child health.

  2. 2.

    Bound and Solon (1999) showed that any differences between the twins that are not removed in a twin-fixed effects model could potentially even increase the endogeneity bias compared to OLS estimates. To see this, first note that the ability bias is determined by the ratio of exogenous variation to total variation. If differencing reduces the fraction of exogenous variation, ability bias may increase.

  3. 3.

    I thank an anonymous referee for suggesting this possibility.

  4. 4.

    While full-time students were already excluded from the sample, measurement error may also be present if there are twins who are part-time students at the moment of the interview, since the question in MIDUS about schooling concerned “completed” schooling. In the twin sample, there were only seven part-time students, however. I will check the robustness of the results to excluding the part-time students from the sample in Section 5.

  5. 5.

    The reason that the variation in years of schooling is greater than the variation in educational categories is that certain ranges of years of schooling are aggregated into larger educational categories in the imputation of Jaeger (1997).

  6. 6.

    The number of chronic conditions is a constructed variable in the MIDUS database which, from a list of 29 chronic conditions shown to the respondent, counts the number of conditions that the respondent claims to suffer from.

  7. 7.

    Brim et al. (2003) do not present standard errors and only show comparisons for a selected number of variables from the CPS.

  8. 8.

    The regressions without twin-fixed effects allow for clustering within families when calculating the standard errors. The reason is that twins can be assumed to share common unobserved factors at the family level.

  9. 9.

    Self-reported health at age 16 was positively and significantly related to adult self-reported health. A one-unit increase in the former variable (measured on a 1–5 scale) was associated with a 0.4-unit increase in adult self-reported health (measured on a 1–10 scale). It should be noted that there is substantial variation in self-reported health at age 16, as 42 % of the twin pairs report different levels of health at age 16.

  10. 10.

    Since health status at age 16 is retrospectively given, it may contain measurement error. This may, for instance, be due to recall errors or to current health affecting the reports. Since there is a significant and positive association between self-reported health at age 16 and current health, the variable cannot consist of pure noise, however. If current health is affecting the reports, so that greater current health is associated with greater self-reported health at age 16, it should be noted that this may give rise to a downward bias in the estimated effect of schooling on self-reported health. The reason is that part of the effect of education will then already be captured by self-reported health at age 16, which is used as a control variable.

  11. 11.

    Behrman et al. (2006) used a five-point ordinal scale, where self-reported health ranged from very poor to very good. There will, thus, be less variation in this measure compared the ten-point scale used in this paper.

  12. 12.

    These results come from specifications where paternal and maternal education were entered separately. When including both paternal and maternal education, the estimates became smaller and were no longer significant, as shown in Table 9.

  13. 13.

    This interpretation only holds, however, if the returns to schooling are in fact greater for those coming from low-educated backgrounds. To check this, I reran the twin FE models, including interactions between own schooling and the parents’ schooling. In order to reduce the number of interaction terms, I created a one to four variable of parental schooling, measuring the average parental educational attainment on a linear scale. The steps correspond to the dummy variables measuring educational categories. In order to address measurement error in parents’ education, I followed the approach of Ashenfelter and Krueger (1994) and averaged the twin reports of their parents’ schooling before creating the variable. The results for self-reported health suggested that the health returns to educational categories and years of schooling are decreasing in parental schooling, as the point estimates of the interaction terms were always negative. Although the point estimates were insignificant in all cases, the results are in line with the hypothesis that the returns to schooling is decreasing in parental schooling.

  14. 14.

    There are no comparable results for chronic conditions in the previous twin literature.

  15. 15.

    Similar results were obtained when using alternative measures, such as moderate activity during the summer. Only for vigorous activity during the summer were the results from the twin FE not significant.

  16. 16.

    The coefficient indicating high school completion was reduced from 0.79 to 0.76. For the variables indicating some college or a university degree, the coefficients were reduced from 1.03 to 0.94 and from 1.06 to 0.95. The variable indicating physical exercise showed a positive relationship with self-reported health.

  17. 17.

    For the first question, the scale has six steps, ranging from agree strongly to disagree strongly; whereas for the two other questions, the scale has four steps and runs from a lot to not at all.

  18. 18.

    In the twin sample, 58 % disagreed somewhat or disagreed strongly to the statement, whereas 20 % agreed somewhat or agreed strongly. Twelve and 10 % agreed a little or disagreed a little, respectively, to the statement.

  19. 19.

    As noted by a referee, it is also possible that time preferences are not stable over time and that education affects the rate of time preference. This possibility was proposed by Becker and Mulligan (1997). If this hypothesis was true, however, one would still expect a positive relationship between time preferences and schooling, which I do not obtain.

  20. 20.

    The question used to assess parental time investments was the following: “How much time and attention did your mother/father give you when you needed it?” The scale went from 1 (none at all) to 4 (a lot). In the sample used, 32 % reported “a lot”, whereas 10 % reported “none at all” regarding paternal time investments. Fifty-eight percent reported “some” or “a little.” For maternal time investments, 56 % reported a lot and only 3 % reported none at all.

  21. 21.

    Not all twins participated in the follow-up survey and my estimate is based on a sample of 541 identical twins.

  22. 22.

    One worry would be that the reliability ratio varies by the level of education. I therefore estimated the reliability ratio separately for low-educated (less than 13 years of schooling) and high-educated (more than 12 years of schooling). The estimated reliability ratios were 0.93 and 0.90, respectively, causing no concern about heterogeneity in reliability ratios by education. Using different cutoffs for defining low and high-educated did not change the results.

  23. 23.

    I thank the anonymous referee for suggesting this possibility.

  24. 24.

    In Lundborg (2008), however, no effect of education on health insurance participation was found using the MIDUS twin sample.

References

  1. Aigner D (1973) Regression with a binary independent variable subject to errors of observation. J Econom 1:49–60

  2. Amin V, Behrman J, Spector TD (2010) Does more schooling improve health behaviours and health outcomes? Evidence from U.K. Twins Working paper: Royal Holloway

  3. Amin V (2011) Returns to education: evidence from U.K. twins: comment. Am Econ Rev 101:1629–35

  4. Arendt JN (2005) Does education cause better health? A panel data analysis using school reforms for identification. Econ Educ Rev 24:149–160

  5. Ashenfelter O, Krueger A (1994) Estimating the returns to schooling using a new sample of twins. Am Econ Rev 84:1157–1174

  6. Ashenfelter O, Rouse C (1998) Income, schooling and ability: evidence from a new sample of identical twins. Q J Econ 113:253–284

  7. Becker GS, Mulligan CB (1997) The endogenous determination of time preference. Q J Econ 112:729–758

  8. Behrman JR, Rosenzweig MR (2004) Returns to birthweight. Rev Econ Stat 86:586–601

  9. Behrman JR, Kohler H, Myrup-Jensen V et al (2011) Does more schooling reduce hospitalization and delay mortality? New evidence based on Danish twins. Demography 48:1347–1375

  10. Behrman JR, Li H, Suen W, Yang Z, Zhang J (2006) Does more schooling improve health behaviors and health outcomes? Evidence from Chinese twins (Working paper). Hong Kong: Chinese University of Hong Kong

  11. Belman D, Heywood J (1991) Sheepskin effects in the return to education. Rev Econ Stat 73:720–724

  12. Black SE, Devereux PJ, Salvanes KG (2007) From the cradle to the labor market? The effect of birth weight on adult outcomes. Q J Econ 122:409–439

  13. Bonjour D, Cherkas L, Haskel J et al (2003) Returns to education: evidence from UK twins. Am Econ Rev 93:1799–1812

  14. Brim OG, Baltes PB, Bumpass LL et al (2003) National survey of midlife development in the United States (MIDUS), 1995–1996 [Computer file]. 2nd ICPSR version. Ann Arbor, MI: DataStat, Inc./Boston, MA: Harvard Medical School, Dept. of Health Care Policy [producers], 1996. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor]

  15. Bound J, Solon G (1999) Double trouble: on the value of twin-based estimation of the return to schooling. Econ Educ Rev 18:169–182

  16. Card D (1999) The causal effect of education on earnings. In: Ashenfelter O, Card D (eds) The handbook of labor economics, Volume 3A. Elsevier Science, Amsterdam, pp 1801–1859

  17. Chou S, Liu J, Grossman M et al (2010) Parental education and child health: evidence from a natural experiment in Taiwan. Am Econ J: Applied 2: 33–61

  18. Currie J, Moretti E (2003) Mother’s education and the intergenerational transmission of human capital: evidence from college openings. Q J Econ 118:1495–1532

  19. de Walque D (2007) Does education affect smoking behavior? Evidence using the Vietnam draft as an instrument for college education. J Health Econ 26:877–895

  20. Fuchs VR. (1982) Time preference and health: an exploratory study. In: Fuchs VR (ed) Economic aspects of health. University of Chicago Press, Chicago, pp 93–120

  21. Fujiwara T, Kawachi I (2009) Is education causally related to better health? A twin fixed-effect study in the USA. Int J Epidemiol 38:1310–1322

  22. Griliches Zvi (1979) Sibling models and data in economics: beginnings of a survey. J Polit Econ 87:37–64

  23. Grimard F, Parent D (2007) Education and smoking: were Vietnam War draft avoiders also more likely to avoid smoking? J Health Econ 26:896–926

  24. Groot W (2000) Adaption and scale of references bias in self-assessments of quality of life. J Health Econ 19:403–420

  25. Grossman M (1972) On the concept of health capital and the demand for health. J Polit Econ 80:223–55

  26. Holmlund H, Lindahl M, Plug E (2011) The causal effect of parents’ schooling on children’s schooling: a comparison of estimation methods. J Econ Lit 49:615–51

  27. Hungerford T, Solon G (1987) Sheepskin effects in the return to education. Rev Econ Stat 69:175–177

  28. Idler EI, Benyamini Y (1997) Self-rated health and mortality: a review of twenty-seven community studies. J Health Soc Behav 38:2137

  29. Isacsson G (2004) Estimating the economic returns to educational levels using data on twins. J Appl Econ 99:99–119

  30. Jaeger DA (1997) Reconciling the old and new census bureau education questions: recommendations for researchers. J Bus Econ Stat 15:300–309

  31. Jürges H, Reinhold S, Salm M (2011) Does schooling affect health behavior? Evidence from the educational expansion in Western Germany. Econ Educ Rev 30:862–872

  32. Kane TJ, Rouse C, Staiger D (1999) Estimating returns to schooling when schooling is mismeasured. National Bureau of Economic Research Working Paper No. 7235

  33. Kenkel D, Lillard D, Mathios A (2006) The roles of high school completion and GED receipt in smoking and obesity. J Labor Econ 24:635–660

  34. Knowles J, Postlewaite A (2005) Wealth inequality and parental transmission of savings behavior. Mimeo, University of Pennsylvania

  35. Lleras-Muney A (2005) The relationship between education and adult mortality in the U.S. Rev Econ Stud 72:189–221

  36. Lochner L (2011) Non-production benefits of education: crime, health, and good citizenship. NBER Working Paper No. 16722

  37. Lundborg P (2007) Smoking, information sources, and risk perceptions. New evidence on swedish data. J Risk Uncertain 34:217–240

  38. Lundborg P (2008) The health returns to education—what can we learn from twins? IZA Working Paper 3399

  39. Lundborg P (2010) Why are identical twins not identical? Parental treatment and schooling among monozygotic twins. Mimeo, Department of Economics, Lund University

  40. Nichols RC, Bilbro WC (1966) The diagnosis of twin zygosity. Acta Genet 16:265–275

  41. Oreopoulos P (2006) Estimating average and local average treatment effects of education when compulsory school laws really matter. Am Econ Rev 96:152–175

  42. Powdthavee N (2010) Does education reduce the risk of hypertension? Estimating the biomarker effect of compulsory schooling in England. J Hum Cap 4:173–202

  43. Pronzato C (2012) An examination of paternal and maternal intergenerational transmission of schooling. J Popul Econ 25:591–608

  44. Rosenzweig MR, Schultz TP (1982) The behavior of mothers as inputs to child health: the determinants of birth weight, gestation, and rate of fetal growth. In: Fuchs VR (ed) Economic aspects of health. University of Chicago Press, Chicago, pp 53–92

  45. Royer H (2009) Separated at girth: US twin estimates of the effects of birth weight. Am Econ J Appl Econ 1:49–85

  46. Webbink D, Martin NG, Visscher PM (2010) Does education reduce the probability of being overweight? J Health Econ 29:29–38

Download references

Acknowledgements

The author would like to thank the seminar participants at the American Society of Health Economists, European Association of Labour Economists, Erasmus University, European Economic Association, University of Mannheim, Uppsala University, Lund University, and Swedish Institute for Social Research for their valuable comments and suggestions.

Author information

Correspondence to Petter Lundborg.

Additional information

Responsible editor: Alessandro Cigno

Appendix

Appendix

Table 8 Fixed effects regressions on binary indicators of self-reported health
Table 9 Regressions on the probability of observing a difference in education

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Lundborg, P. The health returns to schooling—what can we learn from twins?. J Popul Econ 26, 673–701 (2013). https://doi.org/10.1007/s00148-012-0429-5

Download citation

Keywords

  • Health production
  • Education
  • Schooling
  • Twins
  • Returns to education
  • Ability bias

JEL Classification

  • I12
  • I11
  • J14