The relative age effect in European elite soccer: A practical guide to Poisson regression modelling

Abstract

Many disciplines of scholarship are interested in the Relative Age Effect (RAE), whereby age-banding confers advantages on older members of the cohort over younger ones. Most research does not test this relationship in a manner consistent with theory (which requires a decline in frequency across the cohort year), instead resorting to non-parametric, non-directional approaches. In this article, the authors address this disconnect, provide an overview of the benefits associated with Poisson regression modelling, and two managerially useful measures for quantifying RAE bias, namely the Indices of Discrimination and Wastage. In a tutorial-like exposition, applications and extensions of this approach are illustrated using data on professional soccer players competing in the top two tiers of the “Big Five” European football leagues in the search to identify paragon clubs, leagues, and countries from which others may learn to mitigate this form of age-discrimination in the talent identification process. As with OLS regression, Poisson regression may include more than one independent variable. In this way we test competing explanations of RAE; control for unwanted sources of covariation; model interaction effects (that different clubs and countries may not all be subject to RAE to the same degree); and test for non-monotonic versions of RAE suggested in the literature.

The Beginning of Senior Career in Team Sport Is Affected by Relative Age Effect

“…In Italy, the youth categories are based on calendar years, for which all young players born from 1st January to 31st December of a calendar year are grouped together. Thus, the birth week (W B ) and the time of birth (T B ) of each player were computed according to previous studies (Brustio et al, 2018, 2019; Doyle and Bottomley, 2018, 2019). W B indicates the week of the year of player’s birth while T B indicates how far from the beginning of the year a player was born (range score between 0 and 1).…”

“…In line with recent studies on RAE (Brustio et al, 2018, 2019; Doyle and Bottomley, 2018, 2019; Rađa et al, 2018), a separate Poisson regression was applied to investigate the RAE in each team sport following the formula y = e (b0+b1x) . Specifically, y represents the frequency of birth in a given week while x represents the T B .…”

“…Specifically, y represents the frequency of birth in a given week while x represents the T B . Moreover, the Index of Discrimination (ID), which provided the relative odds of being selected for a player born in the first versus last week of the competition year (Doyle and Bottomley, 2018, 2019), was calculated as e -b1 . Finally, likelihood ratio R 2 was computed (Cohen et al, 2013).…”

Relative Age Influences Performance of World-Class Track and Field Athletes Even in the Adulthood

“…As recently suggested, the birthdate data were analyzed using Poisson regressions (Brustio et al, 2018; Doyle and Bottomley, 2018, 2019; Rada et al, 2018). Separate Poisson regressions were performed considering disciplines and gender.…”

“…Thus, in this study it has been calculated how the frequency of birth in a given week (y) was explained by the T B (x). Additionally, the Index of Discrimination (ID), which provides the relative odds of being selected for an athlete born in the first vs. the last week of the competition year, was calculated as e −b1 (Doyle and Bottomley, 2018, 2019). Likelihood ratio R 2 was computed according to Cohen et al (2013).…”