Becoming father negatively affects professional road cycling performance

Abstract

This study examines the impact of becoming father on the performance of professional road cyclists. Fixed effects panel regression is used to compare cycling performance over a period after having a child with the performance in the same period the year before. The sample includes 299 professional male road cyclists who had one or more children between 2001 and 2019, with a total of 496 children. After correcting for personal and team related factors, cycling performance is significantly lower after the birth of a child. This is a first indication that having a child can influence the performance of professional athletes in a negative way. The results of this study may help to provide better psychological and athletic support for recent fathers, and are potentially relevant to contexts other than professional (cycling) sport.

Keywords:

• Father athletes
• child birth
• parenthood
• professional road cycling
• performance
• statistical analysis
• panel regression

REVIEWING EDITOR:

• Mike Rayner, University of Portsmouth, Portsmouth, UK

Subjects:

• Sport Psychology
• Work & Organizational Psychology
• Sport and Exercise Science
• Gender Studies - Soc Sci
• Health & Society

1. Introduction

Road cycling has traditionally been one of the most popular sports in Western European countries. In recent years, the sport has also been gaining interest globally. Although much research has been conducted on performance indicators in sports, the literature focusing on professional road cycling is relatively limited (e.g. Prinz & Wicker, 2012; Rogge et al., 2013). Some studies have focused on physiological and morphological factors identifying performance indicators of (bio)mechanical, metabolic and physiological nature based on laboratory research (e.g. Moran & Pitsiladis, 2017; Schumacher et al., 2006). Performance during the junior category can be used to predict performance during elite competition (Mostaert et al., 2022; Schumacher et al., 2006; Svendsen et al., 2018). Examples of emotional and cognitive factors include pre-race anxiety, self-confidence and pain handling (Spindler et al., 2018). In addition, strong cognitive abilities enable a rider to better assess a race situation and better distribute energy throughout the race (Larson & Maxcy, 2015; Rodríguez-Marroyo et al., 2009).

Until recently, the impact on mothers’ lives and careers has been the focus of scientific research, while the influence on fathers has been largely disregarded (Kotelchuck, 2022; Pot & Keizer, 2016), particularly in the sports literature (Eggebeen & Knoester, 2001; Settersten & Cancel-Tirado, 2010). A limited number of qualitative studies found that especially balancing work and family is challenging, but at the same time, parenthood can be a motivation for greater commitment (McGannon et al., 2018; Palmer & Leberman, 2009; Pedersen, 2001).

Also a man’s physical health can deteriorate by having a child (e.g. Torche & Rauf, 2021). This is partly due to fatigue from disrupted sleep, reduced exercise activity, and lower testosterone levels during and after pregnancy (Gray & Crittenden, 2014). Low testosterone levels have been shown to negatively impact athletic performance (Lucía et al., 2001). The birth of a child can be stressful in a number of ways, for example through uncertainty, worrying and pressure to be a competent father (Garfield et al., 2006; Nyström & Ohrling, 2004). Fatherhood also has positive emotional influences, such as life enrichment, satisfaction, joy, feelings of competence and confidence (e.g. Garfield et al., 2006; Kotelchuck, 2022; Nyström & Ohrling, 2004). Fatherhood could add an extra dimension to identity and be a motivation to train harder (Cohen, 2016; McGannon et al., 2018).

This study investigates the impact of having a child on the performance of professional road cyclists making a first contribution of quantitative research on the influence of children on the professional sports performance of male athletes. Performance is assessed using the cycling quotient (CQ) score. The CQ-ranking uses a weighted points system which ranks cyclists worldwide on an annual basis according to their performance during that year. Inspired by Phillips & Hopkins (2020), the analysis also includes personal variables (i.e. number of race kilometers, age, proportion of races in the home country) and team related variables (i.e. team strength, relative role in the team and whether the rider is riding for the first year in a new team or not) in order to isolate the influence of having a child from other factors.

2. Material and methods

2.1. Data collection and sampling

The dataset for the empirical study was constructed entirely from secondary data that can be accessed online at the websites www.cqranking.com, www.procyclingstats.com, several news websites and social media. All cyclists who held a position in the top 200 of the CQ ranking at the end of the year between 2010 and 2019 are part of this study. The cycling seasons of the years 2020 and 2021 are not considered due to many races cancelled due to the COVID-19 pandemic. The initial selection based on the CQ ranking in the relevant years included 645 riders.

From this set all riders who had one or more children between 2001 and 2019 and whose date of birth can be determined with an accuracy of at least 14 days are retained. To determine the birth dates of the children, birth announcements, birthday congratulations and interviews are consulted on news websites, fan pages and the social media of the cyclists. This yields an unbalanced sample of 299 riders with a total of 496 children.

Finally, observations are removed because of complete inactivity of the cyclist (due to injury, suspension, lack of a contract or other reasons) during at least one of the periods relevant for the estimated models (see section 2.2). As a result, the final sample consists of 267 riders with a total of 403 children. Of these, 237 (58.81%) were first born children and 103 (25.56%) were born in the off season period.

2.2. Methodology

The cyclist’s performance in a certain period after a birth is compared with the performance during the same period in the preceding year (i.e. the reference period). The most drastic influences of having a child manifest themselves in a time span of six to twelve weeks after birth (John et al., 2005). However, this study chooses a slightly longer period with a duration of 120 days, in order to obtain sufficient comparability between two consecutive cycling seasons. In order to control even better for annual fluctuations in the competition schedule, an extra analysis is done using an evaluation period of 150 days (both for the reference and after-birth period). We first run a basic test of difference (Welch two sample t-test) to compare the difference in performance (measured both over 120 and 150 days) between the reference and after-birth period. For the 120 days evaluation period, we see a moderate effect (p = 0.07), for the 150 days evaluation period, we do not see an effect (p = 0.1997). To control for other factors that impact performance (see Section 2.3), a fixed effects panel data regression model is subsequently estimated, which includes a dummy variable for each cyclist and thus controls for all person-specific factors that remain constant over time.

2.3. Variables

Table 1 shows the descriptive statistics of all non-categorical variables, both for a period length of 120 days and for a period length of 150 days.

Table 1. Descriptive statistics of the non-categorical variables in the regression model.

	120 days				150 days
Variable	Average	SD	Minimum	Maximum	Average	SD	Minimum	Maximum
Dependent variable
Performance (CQ points rider)
Reference	173.68	217.09	0.00	1605.00	209.43	231.79	0.00	1605.00
After birth	148.01	185.56	0.00	1462.00	189.42	223.21	0.00	1935.00
Independent variables
Number of race kilometers
Reference	3941.87	2055.64	188.8	9162.5	4853.94	2211.57	202.5	10498.60
After birth	3802.59	2069.07	86.00	9173.9	4719.41	2296.26	239.7	10595.14
Age (years)
Reference	28.24	3.36	20.00	38.00	28.24	3.36	20.00	38.00
After birth	29.24	3.36	21.00	39.00	29.24	3.36	21.00	39.00
Relative role in the team
Reference	2.70	7.62	0.00	144.5	2.39	2.68	0.00	15.28
After birth	2.44	6.32	0.00	100.45	2.29	3.81	0.00	55.84
Team strength (CQ points team mates)
Reference	2263.29	1482.23	6.00	6893.00	2815.02	1696.74	65.00	8643.00
After birth	2337.56	1539.62	39.00	7841.00	2885.45	1751.78	51.00	9015.00
Proportion races in own country
Reference	0.22	0.29	0.00	1.00	0.23	0.27	0.00	1.00
After birth	0.23	0.3	0.00	1.00	0.23	0.27	0.00	1.00
Number of observations	403

2.3.1. Performance

The variable to be explained, performance, is defined as the number of CQ points obtained during the studied period of 120 (resp. 150) days. For both time spans, performance is slightly higher in the reference period, with 173.68 (209.43, respectively) CQ points compared to 148.01 (189.42, respectively) CQ points after child birth.

2.3.2. Childbirth

The central explanatory variable in the analysis, ‘childbirth’, tests whether the performance of cyclists during a period of 120 (resp. 150) days after the birth of a child differs from the performance during the same period in the previous year. This variable is 1 for the test period after the birth, and 0 for the reference period one year earlier.

2.3.3. Firstchild

This dummy variable indicates whether the child born to a cyclist is the first (1) or a subsequent (0) child. This variable is used to assess whether the impact of a first child on the performance is larger than for subsequent children.

2.3.4. Number of race kilometers

The number of race kilometers during the periods under study is an important control variable in the regression. Logically, a higher number of race kilometers provides more opportunities to score points, and so performance should be higher for riders who have ridden more kilometers. As Table 1 indicates, the average race kilometers for both time periods is slightly lower after having a child than in the previous year. However, a T-test shows that the null hypothesis of a zero difference cannot be rejected (p > 0.1).

2.3.5. Age

Research from ProCyclingStats has shown that riders with a career of ten years or more reach a performance peak on average at the age of 28, and their performance begins to steadily decline from the age of 30 (Donlevy, 2020). In the broadest sample from this study, the mean age of the rider at the date of birth of his child is 29.24 years (and 28.24 years at the start of the reference period), with a standard deviation of 3.36 years, which is mainly in the peak period, before performance starts to decline.

2.3.6. First year with current team

According to previous research, the number of years a rider has been riding with a particular team does not impact his performance (Prinz & Wicker, 2012; Rodríguez-Gutiérrez, 2014). However, Rodríguez-Gutiérrez (2014) does find a significant negative effect for riders who have just changed teams and thus are riding for the first year with a particular team. Therefore, we include the variable ‘first year with current team’, a dummy variable that equals 1 when the rider rides with a different team during the time period under study than in the year before. This is the case for just over a quarter of the observations in the dataset. Table 2 shows the frequencies of all categorical control variables in more detail.

Table 2. Descriptive statistics of the categorical variables in the regression model.

	120 days		150 days
	Frequency	Percentage	Frequency	Percentage
First year with team
Reference
No new team	294	72.95	291	72.21
New team	109	27.05	112	27.79
After birth
No new team	298	73.95	297	73.7
New team	105	26.05	106	26.3
Olympic games
Reference
No Olympics	365	90.57	360	89.33
Olympics	38	9.43	43	10.67
After birth
No Olympics	369	91.56	365	90.57
Olympics	34	8.44	38	9.43
Number of observations	403

2.3.7. Relative role in the team

There exists a certain hierarchy in cycling teams, whereby the higher-ranking team leaders (‘leaders’) are supported by lower-ranking helpers (‘domestiques’) who sacrifice their own chances of winning. For this reason, it can be assumed that riders perform better the stronger their leader’s role is. This study measures relative leadership (‘Relative role in the team’) as the CQ score of a rider divided by the average CQ score of his teammates (excluding the rider) during the period under study. This variable thus measures how many times a rider is stronger than the average of his teammates over the considered period.

2.3.8. Team strength

A logical hypothesis is that having a stronger team positively influences the performance of a rider. This hypothesis assumes that the riders under study are team leaders. Nevertheless, the performance of team leaders can also be negatively affected by a higher team quality if this means that there are multiple leaders, for which helpers have to divide their support among these different leaders (Prinz & Wicker, 2012). The variable ‘Relative role in the team’ will control for these issues. In line with the approach of Rodríguez-Gutiérrez (2014), a proxy variable for team strength is constructed. This variable is defined as the total number of CQ points accumulated by the rider’s teammates (excluding the rider’s own CQ points) over the period under study.

2.3.9. Proportion of races in home country

The variable ‘proportion of races in home country’ was calculated as the number of races organized in a rider’s home country in which he participated during the studied period, in relation to the total number of races in which he participated during that period. The theoretical assumption, which was empirically confirmed by Rodríguez-Gutiérrez (2014), is that riders are more familiar with the terrain in their home country and/or more supported by local supporters there, so they perform better when they participate in more races in their home country.

2.3.10. Olympic Games

In general, different cycling seasons in the 21^st century can be compared well based on the CQ scores, since the races with a lot of CQ points to earn are organized at about the same time each year. An important irregularity are the Olympic Games. They only take place once every four years, but they do offer a lot of CQ points (400 for the winner of the road race and 250 for the winner of the individual time trial). The event may also have an impact on riders who do not participate in the Games, since the field of participants will on average be less strong in the other races taking place around this period. To control for this, a dummy variable ‘Olympics’ is added, which takes the value 1 when an Olympic cycling competition takes place during the relevant period.

2.3.11. Offseason

This dummy variable allows testing whether the impact of children born in the on- or off- season is different. The off-season is defined as a 3-month period from 16^th October until 15^th January. It is expected that the timing of becoming a father may influence the training cycle of the rider.

3. Results

Table 3 reports the results of the fixed effects OLS panel regression, for a period length of 120 days (left column) and 150 days (right column). Three true outliers were omitted from the analysis. These are riders who performed disproportionately better than their teammates in one of the studied periods, either because they rode for a smaller, weaker team that year or because they accumulated an exceptional amount of points during the winter period while most of their teammates were active little or not at all. Indeed, during the sampling, all riders who didn’t have a contract with a professional cycling team during at least one of the investigated periods were deleted, in order to be able to measure the explanatory variables related to the team (first year with the current team, relative role in the team and team strength) in a reliable way. The three outliers removed are very similar to such riders without a team contract¹.

Table 3. Regression results for the fixed effects model with a period length of 120 days after having a child and for the model with a period length of 150 days after having a child.

	120 days		150 days
Performance (CQ points)	Coefficient	SD (robust)	Coefficient	SD (robust)
Childbirth (reference: year before) Firstchild Childbirth*Firstchild interaction Number of race kilometers Age (years) First year with current team Relative role in the team Team strength Proportion races in own country Olympic Games Off-season Constant Number of observations Number of cyclists R² F-Test Sig. F-Test	−21.866 −39.350 22.185 0.015 −6.588 −2.563 43.933 0.036 −10.042 −3.074 −32.607 151.904 800 267 60.377 72.309 0.000	11.274* 19.908** 14.024 0.004*** 4.600 10.405 6.006*** 0.007*** 31.906 11.987 17.633* 128.517	−18.523 −42.227 25.871 0.017 −5.364 0.363 54.642 0.037 −40.375 −5.918 −27.146 86.221 800 267 63.875 83.908 0.000	10.081* 21.326** 12.942** 0.004*** 4.465 10.421 5.107*** 0.006 32.480 12.524 17.654 125.738

The reference period is the same period of 120 (resp. 150) days in the previous year. Robust standard errors are shown. Coefficients marked with *, ** and *** are significant at the p < 0.1 level, the p < 0.05 level and the p < 0.01 level, respectively.

The model does not encounter any major multicollinearity problems (the variance inflation factors are everywhere smaller than 3.5) and the influence of possible heteroscedasticity is controlled for by using robust standard errors. However, the residuals do not sufficiently follow a normal distribution. There is disagreement in the literature about the importance of this condition for panel regressions (e.g. Maas & Hox, 2004; McCulloch & Neuhaus, 2011). Given the large sample we are confident that the impact of non-normally distributed residuals is limited for this study.

The regression results report a globally negative impact (p = 0.053) of having a child on professional cycling performance. A professional cyclist scores on average 21.87 CQ points less during the 120 days after having a child than he achieved during the same period the year before, ceteris paribus. For the first child, the impact is measured by the variables Childbirth, Firstchild and the interaction between both. Taken together, the impact of a first child is estimated to be a decrease of 39.04 CQ points on average, ceteris paribus. In a Chi Square linear hypothesis test for the sum of these three parameters, the null hypothesis of a zero sum is rejected (p = 0.023). Over 150 days we observe a similar decrease in performance. The Childbirth and Firstchild variables as well as their interactions are significant, and in the linear hypothesis test the null hypothesis of a zero sum is rejected (p = 0.063). Both the overall impact of getting a child and the impact of a first child are smaller, yet still present (-18.52 and -34.88 CQ points on average, ceteris paribus).

Of the other person-related variables, only the number of race kilometers is significant (p < 0.01). In line with expectations, the more kilometers a rider rides, the higher his performance, with on average 0.015 (0.017 in the model with a period of 150 days) more CQ points per extra kilometer, ceteris paribus. The rider’s age, the proportion of races he rides in his home country and Olympic Games have no impact according to this model. Regarding team-related variables, ‘First year with current team’ has no impact, while ‘Team strength’ is strongly significant in both models (p < 0.01). A rider obtains on average 0.036 more CQ points for each extra CQ point gained by his teammates, ceteris paribus. If a child is born in the off-season, performance is estimated to be 32.61 CQ points lower, ceteris paribus. Over a period of 150 days, the off-season indicator is not significant. Finally, the relative role in the team has a significant positive impact (p < 0.01).

4. Discussion

The decrease in cycling performance in the period after birth can be explained from different points of view. Previous research pointed out the importance of physical, emotional and hormonal factors for a rider’s performance (e.g. Lucía et al., 2001; Spindler et al., 2018). For example, raising young children is usually associated with higher levels of fatigue and sometimes physical symptoms such as headaches and nausea (Hardy et al., 2018; Kotelchuck, 2022). Second, stress and depression can arise from uncertainty about one’s own role in raising the child, from tensions in the relationship with one’s partner and from tensions concerning the work-life balance (Gremillion et al., 2018; Nyström & Ohrling, 2004). Also, on average, fathers are less likely to take risks than childless men (Settersten & Cancel-Tirado, 2010). In a cycling context, it is therefore possible that riders with children show more careful riding behavior than they did before having a child. Our dataset, however, does not include trait characteristics such as risk-taking or sensation-seeking, measured by, e.g. the Sensation Seeking Scale (DuRoy, 2000), or commitment to compete, measured by, e.g. the Revised Competitiveness Index (McEwan et al., 2024). These kinds of personality traits could (and probably do) play an important role in performance, and should be included as control variables in future research studies to better isolate and explain the impact of having a child. Finally, during and after their wives’ pregnancy, men have lower testosterone levels on average, which is related to hormonal fatigue and may therefore negatively affect their athletic performance (Gray & Crittenden, 2014; Lucía et al., 2001). The fact that the impact of giving birth to a first child is estimated to be even higher, can be explained along the same lines. This study only examined cycling performance during the months following child birth, but according to this argument, an effect could also occur during pregnancy. Further research could therefore study whether cycling performance already begins to decline during the months preceding birth, or whether in this period positive effects are more likely to prevail.

It was mentioned earlier that the panel regression does not control for differences in race schemes between different cycling seasons. Only the number of race kilometers is included, but not the nature and difficulty of the races, since this is considered part of the effect of having a child and thus should be absorbed by the variables ‘Childbirth’ and ‘Firstchild’. Over longer periods in the cycling season there is normally more stability in the rider’s race schedule, since it is more likely that the schedule will include, for example, a Grand Tour or an important classic in both years. In addition, the pressure to perform is likely to be higher if the rider has had bad results for a few months, or bad results may lead him to adapt his schedule to races in which his chances of winning are higher. These factors may also cause a significant effect over 120 days but not over 150 days. If this is the case, the negative impact of having a child on cycling performance is more likely to be due to lighter schedule choices during the first months after the birth of a child than to the physical or mental condition of the rider. Related to racing and training schemes, the model shows a negative impact of giving birth in the off-season (if measured over 120 days). In the off-season, it is perhaps easier to give in to the training scheme, whereas in the on-season the rider will remain focused on the races and the competition.

All in all, the empirical results from this study provide cautious initial support for the idea that having a child can negatively influence the performance of professional cyclists, and that a first child can be considered even more impactful. This finding is of practical importance, because it shows that it is important to pay attention to the developments in the private lives of professional athletes. A better understanding of these developments can contribute to a better mental and/or sports management of professional athletes.

Besides the central focus on the impact of having a child, this study found, in line with earlier literature, that the number of race kilometers, the strength of the teammates and the relative strength of the rider compared to his average teammate all have a positive influence on cycling performance. For the age of the rider, as in the study by Torgler (2007), no significant influence on cycling performance was reported. Rodríguez-Gutiérrez (2014) showed that riders perform significantly better the more they proportionally participate in races organized in their home country, and significantly worse if they ride for the first year with a new team. The current study, however, couldn’t confirm these effects.

The empirical research in this study has some shortcomings. Firstly, Cabaud et al. (2016) point out the dangers in modelling cycling performance based on general year rankings. Due to the way CQ score scales are constructed, it is likely that they generally reflect the actual performance of a good rider. However, it may happen that a rider during a season concentrates on one specific main goal (e.g. World Championships or Olympic Games) and takes part in few other races, or uses them as training rather than trying to win them. In that case, his CQ score will not accurately reflect his actual strength. Also for mere helpers the CQ score is not an accurate measure, as they systematically sacrifice their own chances of winning to support their team leaders. Because this study focused only on riders in the top 200 of the CQ annual classification, and included both race kilometers and relative leadership role, these issues were largely controlled for.

Another shortcoming is the way the variable ‘Relative role in the team’ is defined since its encoding is strongly related to the dependent variable, which makes the relationship between these two variables self-explanatory. For future studies, it may be appropriate to construct a kind of smart index for the leadership position of a rider in his team. This can be done for example based on his official leader positions (starting with a rider number ending with the number ‘1’) in different races.

The pool of professional top cyclists who became father and about whom sufficient reliable data is available, is too limited to include, for example, only the spring classics of a given cycling season. Consequently, observations that are very different from each other have to be included (from the years 2000 to 2019 and from different periods in those years, including the less active winter months). The data set contains many outliers, which makes the interpretation of the results difficult. The most influential observations were kept out of the regression analysis, but even then the results remain subject to strong variability and influential observations.

5. Conclusion

This study examined the impact of having a child on the performance of professional cyclists. A fixed effects panel regression showed a negative impact of becoming father on the performance, measured by comparing the CQ score of a rider during a period of 120 days after having a child with his CQ score during that same period of 120 days in the previous year. Also when a longer period of 150 days was considered, the negative effect of having a child remains apparent. The results provide tentative empirical support for the idea that cycling performance is negatively affected in the first weeks after having a child, with an enhanced impact of having a first child. This was a first contribution to the quantitative research on the influence of fatherhood in a professional sports context.

Furthermore, some other factors which might influence the performance of a cyclist were examined, namely the number of kilometers raced, his age, whether or not it was his first year with his cycling team, his relative role in that team (and thus the degree of support he received from his teammates), the strength of his teammates, the proportion of races he rode in his own country and whether or not the Olympics took place during the period. As expected, the number of race kilometers, the relative team role and the strength of the teammates each had a positive impact on performance.

An alternative research approach would be to compare the performance of cyclists who have had a child with the performance of cyclists without children. However, this method requires a sufficiently large similarity between the riders with and without children, which is very difficult to ensure in cycling due to the wide variety of races and rider profiles. Another promising avenue for future research is a qualitative study on how cyclists experience fatherhood and its impact on performance. The dataset as well as results obtained in this study can be used to identify under- or overperforming father cyclists for in-depth interviews.

Currently, the impact of childbearing on the lives of fathers is still under-recognized. Further research on this topic would be very useful to gain more in-depth insight, in order to promote both the personal well-being and the professional performance of father cyclists. The research described in this paper was limited to professional road cycling. An open question is whether the findings of this research also hold for other racing formats of professional cycling, for instance cyclocross, mountain biking, BMX, which all have differing characteristic natures (McEwan & Muller, 2020), or for other formats of cycling that are scored aesthetically.

Disclosure statement

The authors report there are no competing interests to declare.

Data availability statement

The data that support the findings of this study are openly available in figshare at https://figshare.com/s/9d21d94a04bb5c06bf1d

Additional information

Funding

The authors report there are no funding agencies to declare.

Notes on contributors

Anke Baetens

Anke Baetens holds a master degree in Business Administration from KU Leuven, Faculty of Economics and Business. She currently works as Financial Analyst for the Flemish government.

Jeroen Belien

Jeroen Belien is Campus Dean and Full Professor at the Faculty of Economics and Business at KU Leuven Campus Brussels where he teaches courses on Operations Research and Operations Management. His research passion is about applying mathematics to improve processes and decision-making with applications to scheduling, routing, education, and sports. Besides having published in various journals including European Journal of Operational Research and Transportation Science, he holds editorial positions in INFORMS Transactions on Education, INFORMS Journal on Applied Analytics, Journal of Scheduling, and OR Spectrum.

Filip Van den Bossche

Filip Van den Bossche is Vice Dean for Education and Professor at the Faculty of Economics and Business at KU Leuven where he teaches courses on Statistics, Predictive Business Analytics, Econometrics, and Quantitative Methods. His main research interest is applied econometrics, including time series analysis, forecasting, and composite index methodologies. He has published in various journals such as Journal of Statistical Software, Journal of Business Research, Decision Support Systems, and Research in International Business and Finance.

Notes

1 If these three outliers are included anyway, the coefficient of determination of the model (R²) drops to 0.257. In the model with a period duration of 150 days, R² drops to 0.439.