Long-term relative age effects: the
Case of English professional soccer
players using Market Values.
Laurens van der Ziel, 10653384
This document is written by Laurens van der Ziel who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
1. Introduction
In the context of school education and sport education the impact of being born in a certain month has been an ongoing debate in academics. Although the month of birth is not
considered to influence intrinsic talent, it influences the development of children (Bedard & Dhuey, 2006; Musch & Grondin, 2001). These differences in development stem from the practice of grouping children by age based on a cut-off date. This age-grouping leads to differences in maturity between children within groups. These differences in maturity then lead to differences in performances called relative age effects. These Relative age effects then possibly translate into productivity differences in adulthood (Plug, 2001; Ashworth & Heyndels, 2007).
This thesis explores and tests the long-run effects of relative age effects on productivity by using estimated market-value as a measure of productivity. In the second section the literature of the relative age effects in a school context are discussed. Building on the literature in the context of school education, the relative age effects with respects to sports education are discussed. Both the theoretical aspects as well as the empirical results with respect to the relative age effects are discussed. Relative age effects in terms of
representativeness and productivity are discussed in the literature. In Section 3 the data and the institutional context is discussed. Section 4 presents the hypotheses based on the literature review. Section 5 presents the empirical analysis. And finally, Section 6 concludes.
2. Literature Review
In many studies Economists have documented the relationship between month of birth and outcomes such as test scores, wages and educational attainment. Because of the
exogenous nature of date of birth the variation in educational attainment between quartiles caused by administrative cut-off dates has been a popular instrument in education production functions to estimate the returns of schooling (Angrist & Krueger, 1991; Plug 2001; Leuven et
al. 2004; Bedard & Dhuey, 2006).
Angrist and Krueger (1991) use quarter of birth as an instrument for educational attainment to estimate the impact of compulsory schooling on earnings. An individual born relatively early after the cut-off date, starts school at an older age and is therefore eligible to dropout after fewer years of schooling. They argue that compulsory school laws alone accounts for the relation between quarter of birth and earnings because quarter of birth is unlikely to be correlated with omitted determinants of earnings. In response to Angrist and Krueger (1991), Bound and Jaeger (1996) call the validity of any causal inferences regarding compulsory schooling made by Angrist and Krueger into question by showing that while the association between quarter of birth and educational attainment was weaker for more recent cohorts in the dataset of Angrist and Krueger, the relationship between quarter of birth and earnings was not weaker in those cohorts. Furthermore, they present evidence that quarter of birth also has a relationship with earnings and other labor market outcomes for cohorts that were not bound by compulsory schooling laws. Bound and Jaeger (1996) argue that there is evidence for different associations between quarter of birth and various outcomes that are unrelated to educational attainment.
In a school context there are multiple relations. Carroll (1992) provides evidence that season of birth is related to attendance rates. Mortimore et al. (1988) finds a relation between quarter of birth and the likelihood that a student will be evaluated as having behavioral difficulties. Dhuey and Lincomb (2008) find that relatively older students are more likely to being high-school leaders.
Furthermore a large body of literature outside of economics describes a relation between seasonality of birth and different health outcomes. For example; schizophrenia (Tochigi et al., 2004), autism (Gillberg, 1990), dyslexia (Livingston et al., 1993), suicide (Rock et al, 2006) and survival of centenarians (Doblhammer et al., 2005).
A possible explanation for these relationships is that the timing of births depends on unobserved characteristics of the parents (Bound et al., 1995). Buckles and Hungerman (2013) explores the association between month of birth and different maternal statistics. They find seasonality in the share of teenage mothers, married mothers, white mothers and high school educated mothers. Buckles and Hungerman (2013) test the hypothesis that seasonal variation in outcomes such as years of schooling, share of dropouts and wages comes from the fact that children born in different seasons are conceived by different groups of woman by regressing these outcomes on quarter of birth and a set of family controls. The size of the effect of quarter of birth on years of schooling is reduced by 25% to 40% when family controls are added. For the share of dropouts and wages the size of the effect of quarter of birth is also lower when family controls are added. In all three estimations the set of control variables are significant. They argue that while the magnitude is smaller, the persistence of seasonality after controlling for family background shows that other
explanations involving natural and social phenomena after conception or birth also play a role.
An explanation involving a natural phenomenon is the „Fetal origins hypothesis‟ which proposes that the period of gestation has significant impact on the health development of an individual throughout its life (Barker, 1986). An example of a social phenomenon that
explains seasonality is compulsory schooling as described by Angrist and Krueger (1991). Finally, a social phenomenon explaining the influence of month of birth on outcomes like tests scores, years of schooling and earnings are relative age effects. Relative age effects are the effects of differences in maturity between children that are generated by grouping children in classes based on cut-off dates. The literature shows that in any grade older children perform better than their younger peers but this difference decreases over time as relative maturity differences naturally decreases (Bedard & Dhuey, 2006; Leuven et al., 2004). Cunha et al. (2006) argues that early human capital investments and skill attainment facilitates later human capital investment and skill attainment. Consequently, early relative age differences may have long lasting effects on productivity.
The empirical work by Angrist and Krueger (1991) estimate the net effect of relative age on educational attainment and thus fails to identify the relative age effects from compulsory schooling effects. Plug (2001) similarly estimates the net effect of relative age on educational attainment with data from the Netherlands. Plug exploits a change in policy in the
Netherlands that says that students are no longer allowed to dropout when reaching their compulsory schooling age and must finish their current education, thus assumes the compulsory schooling effect to be zero. Bedard and Dhuey (2006) importantly note that interpreting the causal impact of relative age is difficult because age enters into educational
decisions in different ways. For example, when cut-off dates are not strict parents may defer entry for their child by a year making them the oldest student. This would bias the relative age effect downward. Second, relative younger children at school entry are more likely to repeat a grade. Finally, when children compete for different programs yielding different human capital returns relatively older children have a higher chance of being selected for a program yielding a high human capital return because they have a maturity advantage. When selection happens earlier the relative age effect coming from program selection is larger. This effect may be self-reinforcing because younger students that are selected in a low yielding program may become less engaged and motivated. Relative age effects found by for example Plug (2001) therefore are net effects which reflect school entry deference, grade retention, program placement and maturity differences.
Cascio and Schanzenberch (2012) estimates the causal effects of having more mature peers, called peer effects. They exploit the random assignment of children to different
kindergarten classrooms. They find that exposure to more mature children generates positive peer effects measured in test scores in school and the probability of taking a college-entry exam. This is consistent with the literature on positive spillovers from having higher scoring peers, see for example (Zimmerman, 2003).
Prior to the discussion of the existence of relative age effects in education, the existence of relative age effects in the context of team sports education was convincingly showed. In contrast to school education sports education is voluntary. Because of the voluntary nature of team sports relative age effects can be measured in terms of representativeness by looking at the divergence between the expected number of players and observed number of players born per month (Musch & Grondin, 2001). Musch and Grondin (2001) review the literature on relative age effects in sports context and conclude that the literature unambiguously shows relative age effect in terms of representativeness in favor of relative early-born players in all team sports, including soccer. Schorer et al. (2009) additionally finds evidence that the distribution of players across birth quartiles differ per player positions in handball. Relative age effects in sports context differ from relative age effects in school context in different aspects causing it to be more prevalent in sports. Firstly, in addition to initial
differences in cognitive development (Ward & Williams, 2003) maturity differences in physical development bring an additional disadvantage for relative younger children. In the context of soccer, athleticism and motor skills are important skills. Both of which are strongly correlated to chronological age. Musch and Grondin (2001) argue that, given the individual variability of physical development, a relative age disadvantage coupled with a late maturation makes it hard to compete at high levels. Secondly, competition is fiercer in sports education compared
to school education. While in school children compete for different tracks, in sports children constantly compete for places on a team both intra club, where different teams within a club within the same age category play at different levels, and inter club, where different clubs offer sports education at different levels. In the general context of soccer, professional soccer clubs with highly professional academies select children at early ages from a big pool of amateur teams based on their perceived talent, increasing the effect of competition (Fumarco, 2015). Third, children can drop out at any given moment. While school is compulsory and it is only possible to drop out after reaching the compulsory learning age. As described by the literature on relative age effects in school, the strictness relating cut-off dates might influence the size of relative age effects. In the literature regarding sports this is not regularly discussed. One might expect a similar downward bias in the size of relative age effects as in school context (Bedard & Dhuey, 2006) because when a pupil with a relative age disadvantage participates in a lower age category he becomes relatively old in their new age group.
Possible season of birth effects as described in the context of school education must also be considered in sports education. Seasonality in socioeconomic background of parents in birth rates might also influence quarter of birth effects through the character development,
selection of a specific sport, dropout rates and entry of soccer academy that is not free (Baxter & Maffuli, 2003; Yang et al., 1996). In the literature regarding relative age effects in sports seasonal effects are considered to have a very small, if not zero, effect. Musch and Hay (1999) compare relative age effects in Australia, Brazil, Germany and Japan which have different cut-off dates, reversed seasons and different socio-cultural factors. In all countries relative age effects in terms of representativeness were observed. In addition Musch and Hay (1999) show that a shift of the cut-offs date in Australia is followed by a reduction of the representativeness of players that were relatively old under the previous cut-off date and an increase of the players that were relatively young under the previous cut-off date. Helsen et
al. (2000) also study the effect of a shift in the cut-off date in Belgium and find a similar
adjustment in the representativeness of players born in a certain month. Furthermore seasonal effects in two adjacent months are considered to be very small (Barnsley & Thompson, 1998).
Relative age effects in terms of productivity in the long-term in a sport context are not widely analyzed in the literature on relative age effects. Ashworth and Heyndels (2009) observe a wage gap in the advantage of relative late-born players in the German first league. A
hypotheses explaining this result is that relative late-born players represent a more selective subset of players. Relative late-born players have survived stronger selection so the relative
late-born professional players are on average more talented. As the relative maturity
differences disappear when players become professional the average talent differences are translated in productivity differences. A second hypothesis formulated by Ashworth and Heyndels (2007) is that relative late-born players enjoy from training with more mature peers similarly as in a school context described by Cascio and Schanzenberch (2012). Fumarco (2015) finds no evidence supporting these hypotheses as he finds a wage bias in favor of relative early-born players.
3. Data Collection
The data source of this paper is the German based website transfermarkt.de.
Transfermarkt.de is a company that collects and publicizes individual player data. Our sample consists of all players that are born in England and have completed their soccer education in England with a market valuation higher than zero. The vast majority of these players are under contract for a UK based team. An English sample is chosen since there is no existing literature on long-term relative age effects in terms of productivity in England. Relative age effects are also expected to be relative large in England due to the physical nature of the English competitions (Elliot & Weeden, 2010). No data is included from players that were also part of youth setup outside England because this could bias our results. In England the cut-off date is September 1st, which is different than most other European countries that normally have January 1st as the cut-off date. The study of Ashworth and Heyndels (2007) in Germany use a sample where the relevant cut-off date is August 1st. Fumarco (2015) use an Italian sample with January 1st as the relevant cut-off date. The youth system is very strict in the UK. Young soccer players only qualify for dispensation when they have a diagnosed- disability or significant development delay.1 It is allowed for an individual to compete at one age-category above its original category. The strict nature allows for the assumption that entry deterrence does not play a significant role. When talented relative older players are more likely to compete at one age-category above its normal category relative older players may also enjoy from peer effects similar to relative younger players.
1
http://www.thefa.com/~/media/files/thefaportal/governance-docs/equality/disability-and-mental-health/fa-dispensation-policy.ashx
The data consist of individual- market valuation, birth-date and player position. Player positions are categorized in the following four categories.
Position Categories
Goalkeeper Defensive Midfield Offensive
Goalkeeper Central Defender Central Midfielder Central Striker
Left Fullback Left Midfielder Shadow Striker
Right Fullback Right Midfielder Left Winger
Defensive Midfielder Attacking Midfielder Right Winger
To answer the question whether relative age effects in terms of productivity exist, a measure of individual productivity is needed. Individual productivity is usually measured by wages. In soccer however individual contract information is not publicly disclosed and clubs have no strategic advantage in disclosing them. Ashworth and Heyndels (2007) uses estimations on German player wages from kicker magazine and Fumarco (2015) uses estimations on Italian player wages from Gazetta dello Sport as a measure of productivity. This study uses market value estimations of individual players by transfermarkt.de as an alternative measure of productivity. Transfermarkt.de is a company that collects individual player data and provides player valuations. The market valuations are based primarily on algorithms and secondly on input from forum members. Transfermarkt.de provides a guideline for forum members on how to value players. The following five categories make up the market-value according to the guideline; qualities and potential, prestige of the club and the league, current and previous achievements, age, position. Similar to wages, market valuations have a parabolic relation with age because when players mature they become more skilled and qualities are revealed over time and when a player is nearing the end of its career the productivity will decline. Compared to wages, market valuations on average rise earlier with age because market valuations capture potential. They drop faster compared to wages when nearing the end of career because the potential re-sell value diminishes earlier than productivity. The market value estimations are expected to be less sticky than wages since market valuations can be updated as new information about a player‟s productivity is comes available.
The following table presents the summary statistics of the sample. Table2. Descriptive statistics Mean Std. deviation Market Value Total 1079.85 73.4843 Defensive 986.9639 137.3046 Goalkeeper 633.3333 154.3982 Offensive 1408.623 179.0444 Midfield 1017.481 105.4501 September - November 838.0651 103.9863 December - February 1058.247 122.6045 March - May 1278.506 177.7515 June - August 1298.948 220.4811 Observations Defensive .3897229 .0117218 Goalkeeper .0883372 .0068209 Offensive .2661663 .0106225 Midfield .2557737 .0104865 September - November .3192841 .0112053 December - February .2800231 .0107921 March - May .2222864 .0099935 June - August .1784065 .0092021 Other Variables Age 3849.101 40.15233 N 1731
4. Hypotheses
Based on the literature review three hypothesis are formulated.
Hypothesis 1: Among English based soccer players there exists a skewed month of birth distribution in favor of relative older players.
It is frequently observed in the literature that relative older players are overrepresented both at a youth level and professional level (Musch & Grondin, 2001). Relative older players benefit from being more mature. This increases their chances of being scouted thus having a higher probability to become a professional soccer player.
Hypothesis 2: Among English based soccer players the skewed birth distribution in favor of relative older players differs across positions.
Schorer et al. (2009) observes in youth Handball competitions that the monthly birth
distribution significantly differs across positions. This has to my knowledge not been studied in soccer. The monthly birth distribution might differ across positions because the importance of maturity differences differs across positions.
Hypothesis 3: Among English based soccer players there exists a productivity gap in favor of relative younger players.
This hypothesis was first formulated by Ashworth and Heyndels (2007). The intuition behind it is that relative younger players represent a more selective subset of players that survived a discriminatory bias in favor of relative older players. As maturity differences disappear
relative younger players are on average more skilled. Another explanation is that relative younger players benefit from training with more mature peers.
5. Empirical Analysis
As described in the previous section the presence of a relative age effect in terms of
representativeness has become a stylized fact in the sports literature. It is therefore expected that the birth distribution within each year is skewed in this sample. Relative age effect in terms of representativeness is usually measured using Spearman‟s rank correlation coefficient to compare the differences between the observed number of players and
expected number of players based on monthly birth rates for each month. Figure1 shows the monthly distribution of the observed and expected number of players and the difference between those. The expected number of players born in a specific month is based on the average monthly birth rates in England and Wales from the period 1992-1995. There was no data available of the monthly birth rates before this period. Since birth rates seasonality is very similar each year it is expected that this does not greatly influences the analysis.
SEP OCT NOV DEC JAN FEB MAR APR MAY JUN JUL AUG OBSERVED 195 184 174 154 185 146 145 125 115 84 109 116 EXPECTED 144 134 148 142 150 148 151 147 148 145 136 138 DIFFERENCE 51 50 26 12 35 -02 -06 -22 -33 -61 -27 -22 -65 -15 35 85 135 185
Observations per month
Figure1.
Figure1 visually shows that the month of birth rate distribution is skewed. Players born in the first five months of the admission year are overrepresented, whereas players born in the last seven months are underrepresented. This observation is consistent with the hypothesis that relatively old players have an advantage compared to relative young players and have a bigger chance to become a professional soccer player. To formally test this hypothesis the Spearman rank correlation coefficient is calculated between the ranking of the months following the administration year (September = 1, October =2…) and the ranking of the months with respect to differences between observed and expected number of players starting with the highest negative difference (June = 1).
Figure2 reports the Spearman‟s rank correlation which has a value of -0.881 (p<0.0001) so there is a significant negative correlation between the months order and representativeness.
Month
Observed Players
Expected
Players Difference Rank in Administration year
June 84 144.8 60.8 10 May 115 148.0 33.0 9 July 109 135.6 26.4 11 April 125 147.4 22.4 8 August 116 138.3 22.3 12 March 145 151.4 6.4 7 February 146 148.3 2.3 6 December 154 142.4 -11.6 4 November 173 147.8 -25.2 3 January 185 150.1 -34.9 5 October 184 134.0 -50.0 2 September 195 144.0 -51.0 1 Figure2. rs = -0.8881, p < 0.0001
Schorer et al. (2009) finds evidence that the distribution of players across birth quartiles differ for player positions in handball. We also test this for our sample using the Pearson‟s chi-squared test (goodness of fit). The players in the sample are categorized by main playing position. The players with defensive, midfield and offensive position are pooled together as described in section 3.
The following table shows the number of players in each position per quarter of birth and the corresponding chi-square value and p-value.
Figure3.
Position Quarter 1 Quarter 2 Quarter 3 Quarter 4 Statistics
Goalkeeper 48 39 37 29 0.7958 Defensive 240 184 145 105 8.4472 Midfielder 141 128 90 84 1.5674 Offensive 123 134 113 91 8.3029 Total 552 485 385 309 13.1416
The table shows that there is a relative age effect in term of representativeness in all categories of player position. The results also show that among relative late-born players a higher portion of players are defenders compared to the total sample and among relative early-born players a higher portion of players are forwards compared to the total sample. Regarding goalkeepers and midfielders there is no evidence that position affected the distribution of players across the birth quartiles. The relative age effect in terms of
representativeness is thus larger for defensive players and smaller for offensive players. The average market value of offensive players is highest among the position categories while the average market value of defenders is less than the average of the sample. A possible way that relative age effects in terms of representativeness translates in productivity differences is thus through the difference of the magnitude of relative age effects across positions.
The summary statistics in section 3 suggest that a relative age effect in terms productivity in favor of late-born players exists. The following figure3 shows the average market-value for each month. The vertical lines are the corresponding confidence intervals.
An upward trend can be observed in the average market-value within an administration year which is in line with the hypothesis that relative younger players have on average a
productivity advantage. However the observed drop in average market-value in August is not in line with this hypothesis. The average market-value in August is very similar to the
average market-value in September while the hypothesis described by Ashworth and Heyndels (2007) predicts the difference between those months to be the largest.
To analyze the productivity gap more systematically an earning function is estimated. While value does not represent individual earnings we explained in section 3 that market-value has similar properties as wage in that it is increasing with age at a decreasing rate and as the end of a career reaches it is decreasing with age. In a typical Mincerian earning function measures for years of schooling and experience are included. We don‟t have data on years of schooling. The relative age effect thus represents the net effect of relative age on market-value, including the effect of relative age on years of schooling. Therefore, the
variable years of schooling is not included because adding years of schooling would make the interpretation of the parameter regarding relative age more difficult. There is also no data on experience. Experience is expected to be highly correlated with age. Since players may enter competition at different ages, as a result of the exclusion of experience the relative age parameter also measures the effect that relative age has on the age of breakthrough.
The estimated equation is the following;
Where Market Value is players i‟s estimated market value as of 20 December; AGE is the player‟s age measured in days relative to that of the youngest players in the sample (the youngest players in the sample is born on 11-7-1999). Our analysis includes different measures of relative age. RA is measured in terms of days born after the cut-off date (RA = 1, if the birth-date is September 1st and RA=2, for September 2nd…) and Quarter born after the cut-off date using three dummies (Q2, Q3 and Q4) with the first Quarter as the reference (September-November). The represent the estimated parameters, is the random error term. The estimated sign of the coefficient is positive and the estimated sign of the coefficient is negative. The estimated sign of is positive using days born after the cut-off date, since later-born individuals are relatively younger in their cohort. Similarly with Quarter dummies the coefficients are positive and increasing with later Quarters.
robust robust Variable ln Market Value ln Market Value ln Market Value ln Market Value
Age 0.00077*** 0.00077*** 0.00076*** 0.00076***
(0.00008) (0.00007) (0.00008) (0.00007)
Age2 -8.77e-08*** -8.77e-08*** -8.74e-08*** -8.74e-08***
(9.77e-09) (8.57e-09) (9.77e-09) (8.60e-09)
Quarter2 0.17* 0.17** (0.09) (0.09) Quarter3 0.19** 0.19** (0.09) (0.10) Quarter4 0.23** 0.23** (0.10) (0.10)
Days after cut-off date 0.00072** 0.00072**
(0.00033) (0.00033)
Constant 4.11*** 4.11*** 4.10*** 4.10***
(0.15) (0.14) (0.14) (0.14)
Observations 1,731 1,731 1,731 1,731
R-squared .06 .06 .06 .06
Standard errors in parentheses *** p<.01, ** p<.05, * p<.1
These estimations show clear support for a relative age effect in terms of productivity in favor of relative younger players. The coefficients regarding Age have the expected sign and are statistically significant with alpha < 1%. Relative Age, measured in days born after cut-off date is positive and significant with alpha < 5%. This is in line with hypothesis that relative
younger players have a higher productivity on average. Players are in this sample, after controlling for age, on average 0.0072% more worth per day they are born after the cut-off date. It is more intuitive comparing the first quarter born- with the last quarter born players; after controlling for age players born in the last quarter are on average 23% more worth than first quarter born players.
Defensive Offensive Goalkeeper Midfield Variable ln Market Value ln Market Value ln Market Value ln Market Value
Age 0.00081*** 0.00071*** 0.00094*** 0.00071***
(0.00013) (0.00019) (0.00018) (0.00015)
Age2 -9.55e-08 *** -8.12e-08 *** -1.00e-07 *** -7.81e-08 ***
(1.74e-08) (2.55e-08) (1.87e-08) (1.93e-08)
Days after cut-off date 0.00058 0.00014 0.00066 0.00144**
(0.00052) (0.00070) (0.00095) (0.00065)
Constant 4.11*** 4.43*** 3.40*** 3.99***
(0.24) (0.32) (0.40) (0.29)
Observations 674 461 153 443
R-squared .06 .04 .17 .06
Standard errors in parentheses *** p<.01, ** p<.05, * p<.1
The figure above estimates relative age effects in terms of productivity for different subsets of the sample categorized by position. The relative age effects in terms of productivity are highest among midfielders and lowest among offensive players. These estimations are however only statistically significant for midfielders. An explanation could be that Midfielders benefit relatively more from peer effects and offensive players relatively less or in the case of offensive players the low estimate may reflect that they represent a less selective subset of players compared to other positions. The fact that the parameter is not the largest for defensive players, while representing the most selective subset of players, may indicate that defenders benefit less from peer effects compared to midfielders.
Because of the observed similarity in average market value between August and September the following figure, comparing August and September empirically, is included.
Players born in either September or August.
robust Variable ln Market Value ln Market Value
Age 0.00066*** 0.00066***
(0.00015) (0.00011)
Age2 -6.84e-08*** -6.84e-08***
(1.83e-08) (1.21e-08) August 0.08 0.08 (0.16) (0.15) Constant 4.22*** 4.22*** (0.27) (0.23) Observations 311 311 R-squared .07 .07
Standard errors in parentheses *** p<.01, ** p<.05, * p<.1
A productivity bias in favor of players born in August is observed which is consistent with hypothesis 3. A player born in August is given their age, on average 8% more worth than a player born in September. It is expected that the difference between September and August are largest in terms of productivity. This is not observed because a player born in the 4th Quarter is given their age, on average 23% more worth than a player born in the 1st Quarter. This comes from the variability of market values and the fact that the models above only explain 7% of the variability of market values.
6. Conclusion
In school and sports education children are grouped according to chronological ages, based on cut-off dates. This thesis analyzes the long-term effects of relative age differences within groups on productivity among English professional soccer players. In line with the existing literature a skewed birth distribution in favor of relative older players in found. In addition to existing literature evidence is provided that the overrepresentation of relative older players is highest among defensive players and smallest among offensive players. The literature on the long-term effects of relative age differences in terms of productivity has been small and inconclusive. Analyzing market-value estimations on English soccer players, a significant productivity bias is found supporting the hypothesis formulated by Ashworth and Heyndels (2007) that there exists a productivity bias in favor for relative younger players. The strict nature of the English youth system allows for the assumption that entry deterrence does not play a significant role. Market-value estimations provide a good alternative to wages as a measure for productivity as they are less sticky and publicly available. Market valuations can be used in future studies on long-term relative age effects. Analyzing the relative age effects in terms of productivity across positions, a productivity bias is observed in all position
categories. This indicates that the relative age effect in terms of productivity cannot be fully explained by the fact that offensive players are relatively overrepresented among relative younger players. Explanations for the productivity bias are that relative younger players are part of a more selective subset of players and younger players benefit from training with more mature peers. More research is needed to identify the mechanisms of how peer effects possibly translate into productivity differences.
Bibliography
Angrist, J. D. and Krueger, A. B. (1991). Does compulsory school attendance affect schooling and earnings? Quarterly Journal of Economics, 106(4):979– 1014.
Ashworth, J., & Heyndels, B. (2007). Selection bias and peer effects in team sports the effect of age grouping on earnings of German soccer players. Journal of Sports Economics, 8(4), 355-377.
Barker, D. J. (2001). Fetal and infant origins of adult disease. Monatsschrift Kinderheilkunde, 149(1), S2-S6. Barnsley, R. H., & Thompson, A. H. (1988). Birthdate and success in minor hockey: The key to the
NHL. Canadian Journal of Behavioural Science/Revue canadienne des sciences du comportement, 20(2), 167. Baxter-Jones, A. D. G., & Maffulli, N. (2003). Parental influence on sport participation in elite young
athletes. Journal of sports medicine and physical fitness, 43(2), 250.
Bedard, K., & Dhuey, E. (2006). The persistence of early childhood maturity: International evidence of long-run age effects. The Quarterly Journal of Economics, 121(4), 1437-1472.
Bound, J., Jaeger, D. A., and Baker, R. M. (1995). Problems with instrumental variables estimation when the correlation between the instruments and the endogenous explanatory variable is weak. Journal of the American Statistical Association, 90(430):443–450
Bound, J., & Jaeger, D. A. (1996). On the Validity of Season of Birth as an Instrument in Wage Equations: A Comment on Angrist & Krueger's" Does Compulsory School Attendance Affect Scho (No. w5835). National bureau of economic research.
Buckles, K. S., & Hungerman, D. M. (2013). Season of birth and later outcomes: Old questions, new answers. Review of Economics and Statistics, 95(3), 711-724.
Cascio, Elizabeth U. and Diane Whitmore Schanzenbach. 2007. “First in the Class? Age and the Education Production Function.” NBER Working Paper 13663.
Chotai, J., Forsgren, T., Nilsson, L. G., & Adolfsson, R. (2001). Season of birth variations in the temperament and character inventory of personality in a general population. Neuropsychobiology, 44(1), 19-26.
Cunha, F., Heckman, J. J., Lochner, L., & Masterov, D. V. (2006). Interpreting the evidence on life cycle skill formation. Handbook of the Economics of Education, 1, 697-812.
Dhuey, E., & Lipscomb, S. (2008). What makes a leader? Relative age and high school leadership. Economics of Education Review, 27(2), 173-183.
Doblhammer, G., Scholz, R., & Maier, H. (2005). Month of birth and survival to age 105+: Evidence from the age validation study of German semi-supercentenarians. Experimental Gerontology, 40(10), 829-835.
Elliott, R., & Weedon, G. (2010). Foreign players in the English Premier Academy League:„Feet-drain‟or „feet-exchange‟?. International review for the sociology of sport.
Fumarco, L. (2015). Relative Age Effect on Labor Market Outcomes for HighSkilled Workers: Evidence from Soccer.
Gillberg, C. (1990). Do children with autism have March birthdays?. Acta Psychiatrica Scandinavica, 82(2), 152-156.
Helsen, W. F., Starkes, J. L., & Van Winckel, J. (2000). Effect of a change in selection year on success in male soccer players. American journal of human biology, 12(6), 729-735.
Leuven, E., Lindahl, M., Oosterbeek, H., & Webbink, D. (2004). New evidence on the effect of time in school on early achievement.
Livingston, R., Adam, B. S., & Bracha, H. S. (1993). Season of birth and neurodevelopmental disorders: summer birth is associated with dyslexia. Journal of the American Academy of Child & Adolescent Psychiatry, 32(3), 612-616.
Mortimore, P., Sammons, P., Stoll, L., Lewis, D., & Ecob, R. (1988). School matters: The junior years. Open Books.
Musch, J., & Hay, R. (1999). The relative age effect in soccer: Cross-cultural evidence for a systematic discrimination against children born late in the competition year. Sociology of Sport Journal, 16(1), 54-64.
Musch, J., & Grondin, S. (2001). Unequal competition as an impediment to personal development: A review of the relative age effect in sport. Developmental review, 21(2), 147-167.
Plug, E. J. (2001). Season of birth, schooling and earnings. Journal of Economic Psychology, 22(5), 641-660. Rock, D., Greenberg, D., & Hallmayer, J. (2006). Season–of–birth as a risk factor for the seasonality of suicidal behaviour. European Archives of Psychiatry and Clinical Neuroscience, 256(2), 98-105.
Russell, R. J. H., & Startup, M. J. (1986). Month of birth and academic achievement. Personality and Individual Differences, 7(6), 839-846.
Schorer, J., Cobley, S., Büsch, D., Bräutigam, H., & Baker, J. (2009). Influences of competition level, gender, player nationality, career stage and playing position on relative age effects. Scandinavian journal of medicine & science in sports, 19(5), 720-730.
Tochigi, M., Okazaki, Y., Kato, N., & Sasaki, T. (2004). What causes seasonality of birth in schizophrenia?. Neuroscience research, 48(1), 1-11.
Ward, P., & Williams, A. M. (2003). Perceptual and cognitive skill development in soccer: The multidimensional nature of expert performance. Journal of sport and exercise psychology, 25(1), 93-111.
Yang, X. L., Telama, R., & Laakso, L. (1996). Parents' physical activity, socioeconomic status and education as predictors of physical activity and sport among children and youths-A 12-year follow-up study. International Review for the Sociology of Sport, 31(3), 273-291.
Zimmerman, D. J. (2003). Peer effects in academic outcomes: Evidence from a natural experiment. Review of Economics and statistics, 85(1), 9-23.
