The Impact of Date of Birth
on Dutch Professional
Football Players
Bachelor Thesis
Economics
University of Amsterdam
Guido Reincke
13 January 2014
Table of contents
Chapter title Page
1. Introduction 1-2
2. Related literature 2-4
3. Research question and hypotheses 4-6
4. Research method and results 6-12
5. Conclusion 12-13
Abstract
In this thesis, the relationship between date of birth and professional Dutch football players is investigated. According to previous research on this subject, there is a significant bias in the distribution of date of birth of professional football players. Here, three main questions are researched: 1) Is the distribution of date of birth of professional football players uniform compared to the Dutch population? 2) Which characteristics influence the measurement of success of football players? 3) Does the area of birth influence the measurement of success? Using a chi squared goodness-of-fit test to compare the distribution of the quarter of birth of Dutch football players to the Dutch population, the distribution of quarter of birth of football players is proven to be biased. In addition, a regression is performed containing a measurement of success as the dependent variable and age, BMI, position, quarter of birth and place of birth as independent variables. Although the regression itself is significant, none of the independent variables has a significant effect on the level at which a Dutch professional football player is playing. Lastly, area of birth does not significantly influence the measurement of success either.
1 Introduction
Football is one of the biggest sports in the world, always catching worldwide attention. But past research shows that there are problems with the date of birth distribution of professional football players, more professional football are born in the beginning of the year compared with the rest of the year. This causes a loss of talent and unfair opportunities.
Barnsley, Thompson & Legault (1992) found a strong bias in the distribution of birth towards the beginning of the activity year in junior World cups. Mujika, Vaeyens, Matthys, Santisteban, Goiriena & Philippaerts (2009) found a similair bias in the distribution of month of birth of Basque football players. Vaeyens, Philippaerts & Malina (2005) stated that the FIFA-regulations for the cut-off date selections changed in 1995. After this all countries were obliged have the same cut-off date, the bias changed in the same way. Simons & Paul (2001) and Ostapczuk & Musch (2013) had similar findings.
However, so far no research has been done on this subject on professional Dutch football players. In this thesis, it is investigated whether the quarter of birth influences the likelihood of becoming a professional football player. In addition, the correlation between the quarter of birth and the level at which a Dutch professional football player is playing is tested. Lastly, the influence of area of birth on success is assessed.
The main research question in this thesis is: What is the impact of date of birth on professional Dutch football players? This is going to be investigated by answering three questions: 1) Are you more likely, as a Dutch guy, to become a professional football player if
you are born in the first quarter of the year? 2) Are you more likely to play professional football at a higher level if you are born in the first quarter of the year? 3) Could other factors such as age, body mass index (BMI) or place of birth affect the level at which you are playing professional football? To answer these questions, the following hypotheses will be tested:
H0.1: The quarter of birth of professional Dutch football players is uniformly distributed compared to the distribution of quarter of birth of the Dutch population.
H0.2: There is no correlation between the level at which Dutch professional football players are playing and their quarter of birth.
H0.3: Age, BMI and place of birth do not have an impact at the level at which a professional Dutch football player is playing.
To perform the necessary regressions and tests, a database is created containing all the Dutch professional football players, making a distinction between professional and semi-professional football players. The quarter of birth distribution of these Dutch semi-professional football players is tested using a chi squared goodness-of-fit test. In addition, a linear regression is performed with a measurement of success as a dependent variable and age, BMI, place of birth, position and quarter of birth as independent variables. With this regression, it is tested if there is a correlation between the level at which a Dutch professional football player is playing and their month of birth. The regression is tested using a F-test and the variables are tested using a T-test.
The second part of this thesis contains an overview of research that has been done on this subject so far. In the third part, the research questions and hypotheses are further explained. In the fourth part the research method is explained and the results are presented. Finally, the fifth part contains a summary of the thesis and a discussion about the results.
2 Related literature
In this thesis, the date of birth distribution of professional football players and the factors which could possibly influence this distribution are investigated. As told in the introduction, there is a bias in the date of birth distribution of professional football players. This causes a 2
loss of talent, which probably could be avoided. This subject has been investigated for over 20 years on different continents and in different sports. Here, an overview of literature related to this topic is provided.
Barnsley et all. (1992) collected all the birth dates of players competing in the 1990 World Cup and in the 1989 Under-17s and Under-20s World Tournaments in Football. Players born early in the “activity year” (from August 1 to July 31 for football) were over-represented compared to those born later. This effect was particularly large in the Under-17s and Under-20s tournaments. These results were interpreted as the “relative age effect”, which means that people born just after the cut-off date are older and more full-grown. Generally, they play football at a higher level in their youth and thus they have more chance of becoming a professional football player.
Building on the relative age effect, Mujika et all. (2009) researched the date-of-birth distribution of professional Basque football players. It was concluded that the birth-date distributions of all groups of players have a significant bias towards early birth in the selection year compared to the reference population. Between group-comparison revealed that the relative age effect increases at a higher level of youth football. This bias represents a significant loss of potential youth football talent.
Vaeyens et all. (2005) stated that the FIFA-regulations about the “cut-off date” selections for countries changed in 1995. Before, each country could pick its own “cut-off date” – most countries used the beginning of the football season, August 1th. After 1995 all countries were obliged to start using the start of a new calendar year, January 1th, as their “cut-off date”. Vaeyens et al. (2005) compared the birth distributions after and before the off date” changed. It was found that in all situations footballers born just after the “cut-off date” were over-represented compared to players born late in the new selection period. However, players with birthdays at the start of the old selection year (August) were still over-represented after the “cut-off date change”. In addition, Vaeyens et al. (2005) investigated differences in the mean number of selections and in playing minutes between players born at the start or the end of the selection year, but without significant results. Vaeyens et al. (2005) suggest that match-based variables may provide a more reliable indication of the relative age effect in football.
Simmons & Paull (2001) found the same bias in English football. They also had similar findings towards the shift in the “cut-off date” in English football. Despite reducing the influence of body mass, the selection bias towards the older players remained.
Ostapczuk & Musch (2013) stated that the relative age effect is proved to be significant in professional football in almost all European countries, Japan, Brazil, Australia and the USA. This effect did not exist in non-contact sports such as male gymnastics, table tennis, field hockey, golf (Medic, Starkes, Weir, Young & Grove, 2009) and dancing (van Rossum, 2006), because the physical benefits of relative older youth players are not very worth full in these sports. This finding is different from the finding of Simmons & Paull (2001) that body mass has no influence on the relative age effect. Ostapczuk & Musch (2013) found a biased distribution of birth in the German Bundesliga, which is in line with findings in previous literature.
In line with this, Noland & Howell (2010) researched the distribution of the date of birth of professional ice-hockey players in the highest American league, the NHL. The findings were almost exactly the same as in professional football: there is a strong bias in the distribution of the date of birth in professional ice-hockey.
In short, all research on this subject concludes that there exists a significant bias in the distribution of the date of birth. Therefore, it is expected that this research leads to similar conclusions about the distribution of the date of birth of Dutch professional football players, assuming nationality does not alter the results. The influence of body mass on the date of birth distribution is also further explored in this thesis because of its relevance. In addition, research on the possible influence of match-based variables (Vaeyens et al., 2005) is done, to identify what these variables are and what their influence on the success of a professional football player could be.
3 Research question and hypotheses
The research question in this thesis is: What is the impact of date of birth on professional Dutch football players? This is investigated by answering three questions: 1) Are you more likely, as a Dutch guy, to become a professional football player if you are born in the first quarter of the year? 2) Are you more likely to play professional football at a higher level if you are born in the first quarter of the year? 3) Could other factors such as age, body mass index (BMI) or place of birth affect the level at which you are playing professional football?
As told in the related literature, this topic has been investigated in the past: Barnsley et all. (1992), Mujika et all. (2009), Vaeyens et all. (2005), Simmons & Paull (2001) and Ostapczuk & Musch (2013) all found a strong bias in the date of birth distribution in professional football. The same result is expected answering the first question. There has not
been done that much research about the second and third question, so a clear expectation cannot be given. But Ostapczuk & Musch (2013) and Simmons & Paull (2001) have had different findings about the influence of body mass on professional footballers.
The first question has been researched in the past, so previous findings are being tested. The distribution of the quarters of birth of Dutch professional football players is compared to the quarter of birth distribution of the Dutch population. Consequently, a chi-squared distribution test is applied to assess whether the distribution of the quarter of birth of Dutch professional football players is biased, which would be in line with previous research. The hypothesis addressing this question is H0.1: The quarter of birth of professional Dutch football players is uniformly distributed compared to the distributing of quarter of birth of the Dutch population.
The second question builds on the research of Vaeyens et al. (2005), which investigated whether the average percentage of minutes played and the average percentage of games played is equal for football players born in different parts of the year. The research did not result in significant findings. In this thesis, a linear regression is performed, by combining the average percentage of minutes played and that season’s UEFA coefficient of the football player’s club as a dependent variable. The quarter of birth is used as an independent variable and test statistics should indicate whether the quarter of birth has a significant impact on the level at which a professional Dutch football player is playing. The hypothesis that is tested is H0.2: There is no correlation between the level at which Dutch professional football players are playing and their quarter of birth.
The third question is related to the second question because the same dependent variable is used and the independent variables in question two and three are combined in a linear regression. First of all, BMI is used as an independent variable in this regression. There exist contradicting statements in the literature about this subject of Simmons & Paull (2001) and Ostapczuk & Musch (2013). The second independent variable which is used in this regression is age, which logically is expected to have a correlation with the level at which a football player is playing. Finally, a dummy variable expressing place of birth is added to the regression. This dummy variable divides Dutch football players between those born in “The Randstad” (the most urbanized part of The Netherlands) and those not born in “The Randstad”. This dummy is used to see, in case any of the other independent variables has an effect on a footballer’s success, whether this effect is consistent over the whole country. The final hypothesis is H0.3: Age, BMI and place of birth do not have an impact at the level at which a professional Dutch football player is playing.
In short, the first question is tested against related literature findings and a biased distribution of the quarter of birth is expected. The second question is an extension of the research of Vaeyens et al. (2005). In the third question, a closer look is taken at the discussion between Simmons & Paull (2001) and Ostapczuk & Musch (2013) about the impact of physiques on professional football players. This gives a better idea of which factors are related to the success of a professional football player, and what the role of quarter of birth is.
4 Research method and results
In order to be able to perform the necessary regressions and tests, a data set is created based on specialized football databases such as footbaldatabase.eu. The data set contains all the professional Dutch football players. It is sometimes difficult to distinguish between professional, unprofessional and especially between full-prof and semi-prof. To address this issue, only football players playing in the top division of a country whose competition was ranked in the top 30 of the UEFA country coefficient ranking at the end of the season 2012/2013 are used. A second condition is that the football player officially played at least one minute. Competitions ranked outside of the top 30 are not used due to the difficulty of assessing the difference between full- and semi-prof in those competitions. The same logic applies for competitions in countries outside of Europe and for second division in top 30 countries. The 2012/2013 coefficients are used because it is the most recent season which is fully completed.
For all these professional Dutch football players the date of birth, length, weight, position, place of birth, competition in which they were playing, percentage of minutes they played for their club, and the UEFA club coefficient of their club in 2012/2013 are collected and processed into the data set. The date of birth is necessary to collect the age and quarter of birth of the football players. The length and weight are used to calculate the BMI. The positions of the players, divided in goalkeeper, defender, midfielder and forwards, are used to create 3 dummy variables (Stock & Watson, 2012, p. 195). The place of birth is used to create a dummy variable born in “The Randstad”. The percentage of minutes played of each separate footballer have been multiplied with their club’s UEFA coefficient in 2012/2013 (if their club has any) to create a numerical measurement of the dependent variable success.
Table 1: Descriptive statistics
- quarter born Age length weight BMI ratio played season coefficient
Mean 2,347432024 24,5770393 181,392749 75,429003 22,916487 0,430905008 4070,289078 Standard Error 0,062085968 0,22227807 0,36170913 0,34908863 0,07673337 0,017839185 425,6904194 Median 2 24 181 75 22,9176098 0,402941176 2342 Mode 1 22 180 75 23,1481481 0,065359477 0,842 Standard Deviation 1,129555178 4,04399499 6,58072083 6,35111099 1,39604133 0,324555527 6080,075325 Minimum 1 16 165 60 18,5 0,000326797 0,45 Maximum 4 42 200 96 26,8744961 1 36585
Table 1 shows that there in total 331 professional Dutch football players, from which the mean quarter of birth is 2.347. Therefore it can be concluded that there is a bias in the distribution of the quarter of birth towards the beginning of the year, which is in line with expectations. This bias is more extensively tested hereafter. From those 331 footballers, 77 are playing in foreign competitions and 254 are playing in the Dutch competition. Most of those 77 footballers are playing in England and non in Croatia, Israel, Norway, Serbia, Czech Republic, Belarus and Sweden. The average age is 24 years and 7 months old, where the oldest footballer is 42 and the youngest 16. The average length is 181,4 centimeters and the average weight is 75,4 kilogram. On average, a Dutch professional football player played 43% of the minutes in competition games of their club. 167 footballers were born in “The Randstad” and 164 were not. Of these 331 players, there are 25 goalkeepers, 107 defenders, 109 midfielders and 90 forwards.
To address the first question, only the new cut-off date is used because the average Dutch professional football players (aged 24 years and 7 months) was around 5 years old when the cut-off date changed, so the old cut-off date should not have as much influence as in the paper wrote by Vaeyens et al. (2005). Because the quarter of birth distribution of the Dutch population is not equally spread over the four quarters, the average quarter of birth distribution of the Dutch population over the last 15 years has been used. A chi-squared goodness-of-fit test (Keller, 2012, pp. 577-595) is used to calculate whether the quarter of birth distribution of Dutch professional football players is different from the quarter of birth 7
distribution of the Dutch population. Because a relative large sample is used for this test, the data easily satisfy the “rule of five” condition (Keller, 2012, p 581). A chi-squared test uses the expected frequency (e), which is the quarter of birth distribution of the Dutch population and it
uses the real frequency (f), which is the quarter of birth distribution of Dutch professional football players.
As can be seen in figure 2, there is a strong bias in the distribution of the date of birth of Dutch professional football players. 102 players are born in the first quarter, 83 in the second quarter and 75 and 71 in the third and fourth quarters. The expected distribution would be, according to the distribution of the Dutch population; 80.67, 81.82, 87.13, 81.38. With a two-tailed chi-squared goodness-of-fit test you could reject H0.1: The quarter of birth of professional Dutch football players is uniformly distributed compared to the distributing of quarter of birth of the Dutch population. With a p-value of 0.0340. This could be expected because this result is similar to the results of research done about this subject in the past.
The same test has been done for football players who are and who are not born in “The Randstad” separately to see if the result will be the same for different parts of the Netherlands.
For the 167 footballers born in “The Randstad” the quarter of birth distribution is: 57, 38, 34 and 38. While the expected quarter of birth distribution was: 40.7, 41.3, 44.0 and 41. The p-value from the two-tailed chi-squared goodness-of-fit test in this case is: 0.0257. For the same goodness-of-fit test but with footballers born outside of “The Randstad” the p-value is: 0.4637. This is very interesting because the quarter of birth distribution of Dutch professional football players born outside of “The Randstad” is not significant different from the quarter of birth distribution of the Dutch population. There could be several reasons for this: the sample is a lot smaller or there could possibly be no effect at all. Further research on this subject is recommended, possibly with the same test for professional football players who are born in another country where you distinguish between those who are born in the most and least urbanized parts of country. Because it could possibly be that this effect only counts in the more urbanized parts of a country.
For the linear regression an OLS-estimator has been created with the following assumptions (Stock & Watson, 2012, pp. 221-247):
1. The standard error has a conditional mean zero given X1i, X2i, X3i, X4i…., Xki; that
is, E(ui|X1i, X2i, X3i, X4i,…. , Xki) = 0
This implies that the regression has no omitted variable bias, which is very difficult to conclude because it still is very hard to measure quality of football players. But all the absolute values which could affect the dependent variable in this regression are included.
2. (X1i, X2i, X3i, X4i,…. , Xki, Yi), i=1,….,n are independently and identically
distributed draws from their joint distribution.
This assumption holds automatically if the data are collected by simple random sampling, which is the case in this regression.
3. Large outliers are unlikely: X1i, X2i, X3i, X4i,…. , Xki and Yi have nonzero finite
fourth moments.
This can be seen at the descriptive statistics in table 1 where the kurtosis of all variables is finite and nonzero.
4. There is no perfect multicollinearity
In this regression, the effects of the independent variables are quite different from the effect of the dependent variable so it can be concluded that this regression has no multicollinearity either.
As dependent variable in this regression the measurement of success, percentage of minutes played in competition games times 2012/2013 UEFA club coefficient, has been used. Only 209 out of the 340 professional Dutch football players were playing at a club with a UEFA coefficient so the other 131 footballers could not be used in this part of the thesis. As independent variables age, age squared (because the effect of age is most likely to be non-linear (Delorme & Champely, 2013)) and quarter of birth have been added and an F-test has been used to check if the regression is significant (Keller, 2012, pp. 394-396). All the separate independent variables have been tested with a t-test to see if they are significant as well.
Regression Statistics
Coefficients Standard Error t Stat P-value Lower 95% Upper 95% Multiple R 0,209914
Intercept -10,9061 5,99339 -1,8197 0,070264 -22,7227 0,910441 R Square 0,044064 Age 0,88687 0,461364 1,92228 0,055957 -0,02276 1,796497 Adjusted R Square 0,030075 Age squared -0,01439 0,008787 -1,63784 0,102989 -0,03172 0,002933 Standard Error 3,087053 Quarter born -0,046 0,191287 -0,24045 0,810221 -0,42314 0,331147 Observations 209
This regression shows some interesting results. The regression has a significant F value. But none of the T statistics is significant nor the adjusted R square and the standard error of the regression is very large which can cause type 2 error (Stock & Watson, 2012, p 119) In the next regression the BMI has been added to find out of this has a significant impact on the independent variable.
Coefficients Standard Error t Stat P-value Lower 95% Upper 95% Regression Statistics
Intercept -7,62374 7,250735 -1,05144 0,294299 -21,9197 6,672253 Multiple R 0,21702 Age 0,855005 0,463449 1,844875 0,066505 -0,05876 1,768769 R Square 0,047098 Age squared -0,0137 0,008835 -1,55109 0,12243 -0,03113 0,003716 Adjusted R Square 0,028413 Quarter born -0,03562 0,191883 -0,18564 0,852914 -0,41395 0,342708 Standard Error 3,089696 BMI -0,12917 0,160293 -0,80586 0,421264 -0,44522 0,186871 Observations 209
After adding BMI to the regression, the significance of the F value has decreased, the adjusted R squared has decreased as well as all the T statistics and the standard error has increased. The variable BMI has a P value of 0.42, so it can clearly be said that BMI has no influence on the measurement of success. Now the dummy variable place of birth has been added to the regression to find out of this has a significant impact on the independent variable.
Coefficients Standard Error t Stat P-value Lower 95% Upper 95% Regression Statistics
Intercept -11,0956 6,038682 -1,83742 0,067602 -23,0018 0,810623 Multiple R 0,210951 Age 0,893353 0,462875 1,930008 0,054993 -0,01928 1,805986 R Square 0,0445 Age squared -0,01447 0,00881 -1,64198 0,102135 -0,03183 0,002904 Adjusted R Square 0,025765 Quarter born -0,04702 0,191741 -0,24525 0,806508 -0,42507 0,331023 Standard Error 3,093904 Dummy
Randstad 0,134332 0,440133 0,305208 0,760519 -0,73346 1,002125 Observations 209
This regression shows the same result as the regression where BMI was added, all the P-values have worsened and the T statistic is extremely low. So for the dummy variable Randstad can be said that it has no influence on the independent variable. To see if the position at which a footballer is playing has an impact on the measurement of success, three dummy variables have been added to the regression. These dummy variables express goalkeeper, defender and midfielder. No dummy variable for forwards has been added to avoid the dummy variable trap (Stock & Watson, 2012, p 243).
Coefficients Standard Error t Stat P-value Lower 95% Upper 95% Regression Statistics
Intercept -10,7369 5,977831 -1,79613 0,073969 -22,5239 1,050015 Multiple R 0,260971 Age 0,923374 0,460658 2,004468 0,046355 0,015059 1,831688 R Square 0,068106 Age squared -0,01503 0,008798 -1,70823 0,08913 -0,03238 0,002319 Adjusted R Square 0,040426 Quarter born -0,02766 0,192803 -0,14345 0,886076 -0,40782 0,352506 Standard Error 3,070537 Goalkeeper -0,67859 0,899088 -0,75476 0,451275 -2,45139 1,09421 Observations 209 Defender -0,76542 0,558255 -1,3711 0,171865 -1,86618 0,335332
Midfielder -1,2493 0,54869 -2,27688 0,023841 -2,3312 -0,16741
The F-value of this regression is significant and the dummy variable midfielder and the independent variable age have a significant P-value. But the adjusted R squared is still very small and the standard error is still very large. Because the other two dummy variables have very large P-values, the significant P-value of the dummy variable midfielder is not very useful.
Thereby none of the hypotheses following hypotheses is rejected; H0.2: There is no correlation between the level at which Dutch professional football players are playing and their quarter of birth and H0.3: Age, BMI and place of birth do not have an impact at the level at which a professional Dutch football player is playing. Age certainly has a lot of influence but further research, maybe involving professional football players from other, similar organized, countries, should be done to be able to reject the hypothesis. With the regression performed in this thesis there should be no reason to suggest that the quarter in which a professional Dutch football player is born has any influence on the measurement of success. Neither do BMI nor place of birth. There could be a couple of reasons for this: 1) The sample could be too small, 2) The measurement of success could be not accurate enough or 3) The effect could simply not exist. For further research, a larger sample should be used and there should be tried to find a more accurate measurement of success. According to the 11
research done in this thesis Simmons & Paull (2001) should win their discussion with Ostapczuk & Musch (2013), because no connection has been found between BMI and success of a Dutch footballer.
5 Conclusion
In this thesis, the impact of quarter of birth on Dutch professional football players was explored. Three questions were answered: 1) Are you more likely, as a Dutch guy, to become a professional football player if you are born in the first quarter of the year? 2) Are you more likely to play professional football at a higher level if you are born in the first quarter of the year? 3) Could other factors such as age, body mass index (BMI) or place of birth affect the level at which you are playing professional football?
In the past, this subject was researched by several academia. Mujika et all. (2009), Barnsley et all. (1992), Vaeyens et al. (2005), Ostapczuk & Musch (2013) and Simmons & Paull (2001) all found a strong bias towards the cut-off date in the date of birth distributions in professional football.
To answer the research question: What is the impact of quarter of birth on professional Dutch football players?, three hypothesis have been tested: H0.1: The quarter of birth of professional Dutch football players is uniformly distributed compared to the distributing of quarter of birth of the Dutch population., H0.2: There is no correlation between the level at which Dutch professional football players are playing and their quarter of birth. and H0.3: Age, BMI and place of birth do not have an impact at the level at which a professional Dutch football player is playing. The first hypothesis was expected to be rejected because of the research done about in the past.
To answer the questions, a data set containing all professional Dutch football players was created. This data set contains all footballers who at least played a game in the 2012/2013 season for a club which was playing at the highest level of a competition which was ranked in the top 30 of the UEFA coefficient ranking. In total, the data set contained 331 footballers from which 254 were playing in the Dutch competition. The age, the place of birth, the BMI, the percentage of minutes played in competition games and the 2012/2013’s UEFA coefficient of their clubs were included in the database as well. The date of birth was used to determine in which quarter of the year each football player was born. The distribution of the quarter of birth of Dutch professional football players was tested with a chi-squared goodness-of-fit test against the average distribution of the quarter of birth of the Dutch
population. With a p-value of 0.0340, the hypothesis, H0.1: The quarter of birth of professional Dutch football players is uniformly distributed compared to the distributing of quarter of birth of the Dutch population, was rejected. The UEFA coefficient and the percentage of minutes played were used to create a measurement of success for the level at which a professional football player was playing. This measurement of success was used as a dependent variable in a regression where age, the dummy variable born in “The Randstad, BMI and another dummy for the position at which a footballer was playing were used as independent variables. The P-values of the T statistics of these variables were calculated as well as the significance of the F values of the regressions. The results were interesting, the F values of the regressions were significant but none of the T statistics were, although the variable age nearly was. Because of this neither the hypothesis H0.2: There is no correlation between the level at which Dutch professional football players are playing and their quarter of birth nor the hypothesis H0.3: Age, BMI and place of birth do not have an impact at the level at which a professional Dutch football player is playing could be rejected.
The findings in the first part of this thesis are in line with the findings Mujika et all. (2009), Barnsley et all. (1992), Vaeyens et al. (2005), Ostapczuk & Musch (2013) and Simmons & Paull (2001) had in the past. The quarter of birth of professional football players is not uniformly distributed and biased towards the “cut-off” date. There is no connection found between the physiques of a football player and their success, which is similar to the findings Simmons & Paull (2001) had in the past.
The results of the chi squared goodness-of-fit test are clear but the findings from the regressions are limited. To be able to have a better view on this subject, further research is recommended. A more accurate measurement of success could help future research and a similar research project in a larger country with a similar background. A larger sample size than used in this thesis is recommended to obtain more reliable results.
6 Preliminary reference list
Barnsley, R.H., Thompson, A.H., Legault, P. (1992). Family planning: Football style
The relative age effect in football. International review for the Sociology of Sport, 27, 77-87.
Centraal bureau van de statistiek (2013). Distribution date of birth of the Dutch population, http://statline.cbs.nl/StatWeb/publication/?VW=T&DM=SLNL&PA=37296ned&D1= a&D2=0,10,20,30,40,50,60,(l-1),I&HD=130605-0924&HDR=G1&STB=T .
Delorme, N. & Champely, S. (2013). Relative Age Effect and chi-squared statistics. International Review for the Sociology of Sport.
Football database (2013). Dutch footballers, http://www.footballdatabase.eu/ .
Keller, G. (2012). Managerial Statistics. South Western Cencage Learning, 9th edition.
Medic, N., Starkes, J.L., Weir, P.L., Young, B.W., & Grove, J.R. (2009). Relative age effect in masters sports: Replication and extension. Research Quarterly for Exercise and Sport, 80, 669-675.
Mujika, I., Vaeyens, R., Matthys, S.P.J., Santisteban, J., Goiriena, J. & Philippaerts, R. (2009).The relative age effect in a professional football club setting. Journal of Sports Sciences, 27, 1153-1158.
Nolanm, J.E. & Howell, G. (2010). Hockey success and birth date: The relative age effect revisted. International Review for the Sociology of Sport, 45, 507-512.
Ostapczuk, M. & Musch, J. (2013). The influence of relative age on the composition of professional soccer squads. European Journal of Sport Science, 13, 249-255.
van Rossum, J.H.A. (2006). Relative age effect revisited: findings form the dance domain. Perceptual and Motor Skills, 102, 302-308.
Simmons, C. & Paull, G.C. (2001). Season-of-birth bias in association football. Journal of Sports Sciences, 19, 677-686.
Stock, J.H. & Watson, M.W. (2011). Introduction to Econometrics. Person Education, 3th edition.
Vaeyens, R., Philippaerts, R.M. & Malina, R.M. (2004). The relative age effect in soccer: A match-related perspective. Journal of Sports Sciences, 23, 747-756.
