Too late for talent to kick in? The relative age effect in Argentinian male football

Abstract

Does a young football (soccer) player’s birthdate affect his prospects in the sport? Scholars have found a correlation between early births in the competition year among young players within the same cohort and improved chances in sports as they advance to other stages. This article is one of the first studies to ask this question about a male premier league in Latin America – the Argentinian ‘A’ league. It uses a large-N data-set of all players in the period 2000–2012, around 3000 players. The article finds a large effect of the player’s relative age on his prospect to become a professional, though the effect is only present in the case of Argentinian-born players. The effect evaporates once a set of measures are employed to compare professional players with one another. The article contributes to the discussion of the biased effect of seemingly neutral institutional policies, and its conclusions may shed light in other areas.

Acknowledgements

I am very grateful to Patricio Saldaña, who helped compile the players’ data, and to Viridiana Islas and Georgina Timossi, who assisted in other stages of the research. I thank Ernesto Calvo, Matthew Hersh and two anonymous referees for their very insightful comments; Matthew also offered helpful editing suggestions. Any remaining errors are my own.

Notes

1. The next section summarizes the literature.

2. Department of Education, UK, ‘Month of Birth’.

3. Navarro et al., ‘The Relative Age Effect’.

4. Grondin et al. ‘Trimestres de naissance et participation au hockey et au volleyball’; Saavedra García et al., ‘Relative Age Effect in Lower Categories of International Basketball’.

5. Pérez Jiménez et al., ‘Relative Age Effect in Spanish Association Football’; Salinero Martín et al., ‘El efecto de la edad relativa’; Corredoira, ‘El efecto de la edad relativa’; Sedano et al., ‘The Relative Age Effect in Spanish Female Soccer Players’; González Aramendi et al., ‘El Efecto Relativo de la Edad’; González Aramendi et al., ‘Mes de nacimiento y éxito en el fútbol’.

6. Jullien et al., ‘Influence de la date’.

7. Cobley et al., ‘Relative Age Effects in Professional German Soccer’.

8. Augste et al., ‘The Relative Age Effect’.

9. Vaeyens et al., ‘The Relative Age Effect in Soccer’.

10. Van den Honert, ‘Evidence of the Relative Age Effect’.

11. Williams, ‘Relative Age Effect in Youth Football’.

12. Helsen et al., ‘The Relative Age Effect in European Professional Soccer’.

13. Costa et al., ‘Relative Age Effect in Brazilian Soccer Players’.

14. Saavedra García et al., ‘Relative Age Effect in Lower Categories of International Basketball’.

15. Till et al., ‘The Prevalence, Influential Factors and Mechanism of Relative Age Effects’.

16. Coutts et al., ‘Relative age effects in Australian Football League’.

17. Delorme et al., ‘The Relative Age Effect in Elite Sports: The French Case’.

18. Grondin et al. ‘Trimestres de naissance et participation au hockey et au volleyball’; Hurley, ‘A Proposal to Reduce the Age Discrimination’; Nolan et al., ‘Hockey success and birth date’; Wattie et al., ‘Tracking Relative Age Effects over Time in Canadian NHL Players’.

19. Okazaki et al., ‘The Relative Age Effect among Female Brazilian Youth Volleyball Players’.

20. Edgar et al., ‘Season of Birth Distribution of Elite Tennis Players’.

21. Agricola, ‘Relative Age Effect in Junior Tennis (Male)’.

22. Ulbricht et al., ‘The Relative Age Effect’.

23. Delorme et al., ‘Relative Age Effect in Elite Sport: The French Case’.

24. Vincent et al. ‘Gender Differences in the Relative Age Effect’; Delorme et al., ‘Relative Age Effect in Elite Sport: The French Case’; Wattie et al., ‘Tracking Relative Age Effects over Time in Canadian NHL Players’; Van den Honert, ‘Evidence of the Relative Age Effect in Football in Australia’.

25. Romann et al., ‘Relative Age Effects in Swiss Junior Soccer and Their Relationship with Playing Position’; Till et al., ‘The Prevalence, Influential Factors and Mechanism of Relative Age Effects’; Saavedra García et al., ‘Relative Age Effect in Lower Categories of International Basketball’; Schorer et al., ‘Influences of Competition Level, Gender, Player Nationality’.

26. Romann et al., ‘Relative Age Effects in Swiss Junior Soccer and Their Relationship with Playing Position’; Saavedra García et al., ‘Relative Age Effect in Lower Categories of International Basketball’; Grondin et al., ‘Date de naissance et ligue nationale de hockey’.

27. Saavedra García et al., ‘Relative Age Effect in Lower Categories of International Basketball’; Schorer et al., ‘Influences of Competition Level, Gender, Player Nationality’ (handball).

28. Schorer et al., ibid.

29. The league’s teams are non-profit civil associations or ‘clubs’.

30. When a team playing in the premier league hires a player as a professional, he can play either in that league or in the so-called ‘Reserva’, a league developed to give field minutes to professional players who do not make the cut to be in the initial line-up or as substitutes in the premier league. Some players go back and forth between the two leagues, even during the season.

31. The data-set includes professional players hired as professionals even if they only played in the ‘Reserva’ league. See ibid.

32. I collected these data by processing both team rosters’ and individual players’ entries from the Webpage futbolpasion.com. I double-checked the birthdates through online searches.

33. The lists are available in Wikipedia.

34. The relevant variable computes any games played by the national team, a friendly game counting as much as a game in the World Cup – doing otherwise would have meant having only a small handful of observations.

35. I used the Webpage buscardatos.com.

36. Delorme et al., ‘Relative Age Effect in Elite Sports’; Delorme et al., ‘Relative Age Effect and Chi-Squared Statistics’.

37. The authors may insist that relatively young players might wrongly believe that they would be adversely affected by the age difference when in fact they would not, so the two mechanisms would be different. Now, if relatively young players are able to at least indirectly observe (e.g. by watching a practice) the role of physical traits in the sport, it seems that they will be well placed to develop an informed perception.

38. Delorme et al., ‘Relative Age Effect and Chi-Squared Statistics’.

39. González-Víllora et al., ‘Relative Age Effect in UEFA’; Saavedra García et al., ‘Relative Age Effect in Lower Categories of International Basketball’; Helsen et al., ‘The Relative Age Effect in European Professional Soccer’; Augste et al., ‘The Relative Age Effect’.

40. The results of the four t-tests that take into account the month, two-month period, quarter and four-month period in which players were born, respectively, are as follows. The group of players from the expected distribution is always reported first, followed by the group of players from the observed distribution. The number of observations from both distribution is the same, N = 2557. Test numbers are in parentheses. (1) Month: expected (M = 6.66, SD = 3.54), observed (M = 5.53, SD = 3.37), t(5112) = 11.7, p < .001. (2) Two-month period: expected (M = 3.58, SD = 1.74), observed (M = 3.03, SD = 1.66), t(5112) = 11.5, p < .001. (3) Quarter: expected (M = 2.55, SD = 1.14), observed (M = 2.20, SD = 1.08), t(5112) = 11.2, p < .001. (4) Four-month period: expected (M = 2.00, SD = .83), observed (M = 1.79, SD = .79), t(5112) = 9.8, p < .001. The Kolmogorov–Smirnov tests show the same type of difference between the groups as the t-tests, at similar levels of significance. Results are shown in the appendix – test numbers 1′ through 4′.

41. One indicator of young players’ physical development – which would partly explain the observed effect – might be their height. Needless to say, it is the player’s height at one or several points before his draft in the premier league (a piece of information that is not available) what matters most, not current height. The reason is that, irrespective of their birthdate, players who meet the professional cut probably do not have strong physical limitations; physical differences due to the relative age gap probably even out over time; and players born in a given year typically play with those born in a different year. That current height is not important is clear from the data, since there is no correlation between height and birthdates. If it were the case that the young players’ height counts as one of the reasons creating the relative age gap – something that it is not possible to know from this study – it is evident that the correlation disappears when the player becomes professional.

42. That is, on average, Argentinian-born players would be younger.

43. T-tests: (9) Goalkeepers: expected (M = 6.59, SD = 3.6), observed (M = 5.34, SD = 3.34), t(446) = 3.8, p < .001. (10) Defenders: expected (M = 6.69, SD = 3.5), observed (M = 5.46, SD = 3.36), t(1480) = 6.88, p < .001. (11) Midfielders: expected (M = 6.7, SD = 3.55), observed (M = 5.57, SD = 3.4), t(1956) = 7.30, p < .001. (12) Forwards: expected (M = 6.56, SD = 3.53), observed (M = 5.62, SD = 3.36), t(1224) = 4.79, p < .001. The Kolmogorov–Smirnov tests show the same type of difference between the groups as the t-tests, at similar levels of significance. Results are shown in Table 5 in the appendix – test numbers 9′ through 12′.

44. Tournaments typically have a total of around 600–700 players, about 30–35 players per team if ‘Reserve’ players are included.

45. See in the appendix Table 3, test numbers 13 through 36 (t-tests) and Table 5, test numbers 13′ through 36′ (Kolmogorov–Smirnov). As in the following tests, and for the sake of brevity, I only report results for the tests employing the month of birth as the relevant time interval.

46. This is a conservative measure, since some of the individual tests in the case of the remaining seven clubs are significant as well. Employing only the month interval, the difference is significant at some level below .05 in the case of all 26 teams but 6 in the t-tests, and in all but 3 in the Kolmogorov–Smirnov tests.

47. See in the appendix Table 3, test numbers 37 through 62 (t-tests) and Table 5, test numbers 37′ through 62′ (Kolmogorov–Smirnov).

48. There are only four provinces or districts with more than 100 players, comprising 85% of unique players (the Buenos Aires province alone contributes 51% of players, and other 32% come from Santa Fe, Buenos Aires City and Cordoba, in descending order of magnitude). Concerning three of these provinces, the difference in the t-tests is significant at the .001 level. The difference is not significant in the case of Cordoba, the province with the fewer players among the 4 (139). Finally, the difference is significant at either the .05 or lower level if players from all other provinces (337) are grouped together. Kolmogorov–Smirnov tests show similar outcomes, except for the tests with players from other provinces, which are not significant. See in the appendix Table 3, test numbers 63 through 67 (t-tests) and Table 5, test numbers 63′ through 67′ (Kolmogorov–Smirnov).

49. Apart from the fact that a player may have multiple entries, the measure is not perfect since it does not compute the number of actual minutes in the field – it does not sort out those in the starting line-up that make it to the end from substitutes that play only for a few minutes.

50. See Table 4, test number 68 in appendix.

51. With such a large sample (around 15,000 observations), this effect might be due to chance. A clear effect does not arise if only keeping observations pertaining to the year of birth with the most observations in each tournament (and testing by tournament).

52. See Table 4, test number 69 in appendix.

53. See Table 4, test number 70 in appendix.

54. Hurley, ‘Equitable Birthdate Categorization’, 253.