The effect of absenteeism in the

workplace: An example of NFL injuries

and team performance

Strategy Economics Thesis

DANIEL ERGAS
533390

1
Table of Contents
1. Introduction ..........................................................................................................................3
1.1Overview, Relevance, and Contribution .....................................................................3
1.2 Hypothesis and Research Structure ............................................................................4
2. Literature review ................................................................................................................5
3. Setting, Background, and Data ..........................................................................................6
3.1 Setting.........................................................................................................................6
3.2 Background Knowledge of American Football .........................................................8
3.3 Dataset ........................................................................................................................9
4. Identification Strategy ........................................................................................................12
5. Results ................................................................................................................................15
5.1 Defense Team Results ..............................................................................................15
5.2 Attack Team Results .................................................................................................19
6. Discussion, Limitations, and Conclusion...........................................................................21
6.1 Discussion ...............................................................................................................21
6.2 Limitations ..............................................................................................................22
6.3 Conclusion ...............................................................................................................23
7. References .........................................................................................................................24
8. Appendices ........................................................................................................................26

2
 1. Introduction:

1.1 Overview, Relevance, and Contribution

Businesses increasingly rely on teams. Drew et al. (1996) survey 100 U.K. firms and provide
evidence of team-based organizations becoming the norm in modern companies. Current business
activities often rely on group efforts. For example, product launching in a specific region or quality
management for an R&D prototype. Hamilton et al. (2003) identify the effect of teams on
productivity. Their findings point to an increase in productivity by 14% in a garment plant after
team adoption. Workers are able to utilize collaborative skills that would be wasted if no team
work happens, like intrateam bargaining or mutual team learning. Given the importance of teams
in the workplace, absenteeism may potentially affect firm productivity, leading me to the research
question: ‘Does absenteeism in the workplace influence group performance?’.

Absenteeism in the workplace refers to the absence of a worker in his/her job for a prolonged
period of time due to reasons different than personal time or holidays, that is, unplanned leaves.
The concept of absenteeism is a key aspect of an operational business since it heavily impacts the
production function. The economic consequences of absenteeism should not be underestimated.
Stewart et. al. (2003) reveal that the productivity loss of absenteeism costs the United States around
$226 billion or $1635 per worker every year. Hoey et al. (2023) expand on the influential role
absenteeism plays in the workplace using a sports setup. My paper aims to fill in the gaps that
Hoey et al. (2023) started covering and broaden the understanding of the implications of absences.
The research's ambition is to provide measurable effects of sudden nonparticipation of employees
in setups where collaboration and synergies among individuals help consummate the objective and
expand the knowledge on production complementarities within teams.

Existing research suggests absenteeism may be an important factor in group performance at work.
Bartel et al. (2014) use panel data from hospital nurses’ experience level to prove how exits of
experienced nurses and the consequent replacement by a temporary nurse harm nurse team
performance, suggesting workers’ productivity is dependent on peers. Finding equally productive
3
replacement workers after a worker departure remains a challenge for employers, potentially
harming team workflow (Herrmann and Rockoff, 2012).

1.2 Hypothesis and Research Structure:

The hypothesis tested on this paper is:

Hypothesis: ‘The absence of coworkers has an impact on group performance and outcomes’

To test this hypothesis I will make use of the NFL data from the regular season of 2022 and run
OLS regressions on the performance outcome variable in the form of points and yards
conceded/achieved on the number of injuries a team sustains throughout the season, after
controlling for variables of interest. Moreover, different roles of workers may have different
impacts on the final team outcome. To account for this, I run the OLS models on two subsets of
data to distinguish between defense and offense players. I take this approach because there are
parallels between the sport and a traditional work environment. The context of the NFL allows
these similarities as the value output metrics for performance are easily measured, the independent
variable for injuries is identifiable, and being injured has direct consequences on the players’ pool,
and lastly, because it is a team activity where the outcome depends on the complementarities of
production from peers. Additionally, absences in this study refer to unplanned leaves of employees,
the same way injuries in sports have a quasi-random variation property causing exogenous sudden
leaves. By means of the OLS models, I obtain point estimates that hint at the extent of injured
players causing an effect on the group outcome. Yet, obstacles and limitations may unveil when
exercising these models. Limitations are discussed in Section 6.

I divide this study in different sections; Section 2 is a literature review where I discuss previous
research related to this paper’s topic, Section 3 describes the data and settings for this research,
Section 4 explains the theoretical background and methodology of the experiment, Section 5
presents the results, and Section 6 wraps up the results and concludes the paper.

4
2. Literature Review:

Returning to this paper's motivation, teamwork can benefit individual and group performance. Mas
and Moretti (2009) explain how production functions of workers are codependent with one
another. They use data from a large grocery chain to provide evidence of low productivity workers
benefiting from the presence of high productivity workers, leading to an optimal production
function of the group. Falk & Ichino (2006) show how peer effects favorably influence an
individual's performance in simple tasks like putting letters in an envelope, implying a positive
relationship between the performance of two individuals. Bringing these findings to a larger scale
indicate that a grouped team with the optimal mix of workers should yield better productivity
relative to individual working cells. Similarly, Mailranta et al. (2009) provide evidence that inter-
company labor mobility has positive spillover effects that boost productivity and profitability,
mainly when job reallocating from a technical to a non-technical role. However, this is still a
debated topic for two reasons. Dahl et. al. (2014) evaluate that the peer effects are more substantial
as more individuals become part of a group as a causal chain makes the direct influence of the
original peer amplify by intervening coworkers; the so-called 'snowball effect'. Alternatively, they
also become aware of the decaying influence of the original peer over time. Yet, peer effects are
affecting individual performance and their effects are not limited to one single individual, hence it
can be argued that peer effects do have an impact on team production output and performance.

Knowing that teamwork is desirable for employers, the effects of absenteeism in the work
environment are profoundly significant. Most of the contemporary papers about absenteeism
discuss the various sources of absenteeism. Some researchers link absenteeism to the consumption
of demerit goods like alcohol (Bacharach et al., 2010) and source absenteeism from the individual
decision to binge drink. Others, like Halpern et al. (2001), argue that micro-absences from smoke
breaks during working hours also lead to impoverished productivity. Further explanations about
the sources for absenteeism are self-reported illnesses (Jacobson et al., 1996), peer attendance and
organizational commitment (Hassan et al., 2014), or changes in government policies regarding
paid leaves (De Paola et al., 2014).

5
While many researchers discuss this topic and its origin in different setups, this research aspires to
shed light on the implications of labor force voids in team performance and outcomes. Mollerman
& Slomp (1999) investigate the impact of the distribution of workforce flexibility on team
performance. They measure performance in various ways, but the total cumulative production time
is the most related to this thesis. Findings from their research suggest that absenteeism has a strong
negative effect on cumulative production time, regardless of the distribution of flexibility, and that
the workload of the bottleneck worker is greater relative to the other workers.

The most related paper to the current study is Hoey et al. (2023). In their paper, they propose two
channels for absenteeism influencing team productivity. A direct channel causes a direct loss of
production as the employer fails to find a replacement for the absentee, and an indirect channel
causes coworkers to produce less because their productivity depends on the absentee. They also
use the sports context to analyze absenteeism effects by measuring an ex-ante effect of injuries on
individual and team performance. They find that when absences occur, the group experiences
disarray in the coworker network and a decrease in coworker skills, harming group productivity.
Berman et al. (2002) use the sports setup to build a bridge between shared team experience and
favorable team performance. They find that share team experience extends to tacit knowledge, and
consequently becomes a resource for comparative advantage, but has diminishing returns even
turning the effect on team performance negative at some point.

3. Setting, Background, and Data:

In this section, I discuss why the selected setting for studying absenteeism in the workplace is
valid and adequate. I briefly review the NFL structure and describe the dataset I use for this
research.

3.1 Setting:

In this paper, I use the NFL workplace setting during season 2022 to explore further the effect of
voids in the labor force on team productivity, using players' injuries as a proxy for absenteeism.
The NFL setting is adequate to monitor the production function of a team because it has both team
size constraints (i.e., the team units cannot be larger than what the rules specify) and time
6
constraints. The game is played in 4 x 15-minute quarters for a combined playing time of 60
minutes. The highest score at the end of 60 minutes wins. This allows for testing productivity as
the outcome spreads over a predefined period. Ties are rare in American Football, and overtime
periods occur if necessary to determine a winner. Overtime periods are not frequent, so I exclude
them from this analysis.

Highly inspired by Hoey et al. (2023), I address endogenous staffing changes in football teams
during the 2022 season by exploiting quasi-random coworker injuries in NFL team members as a
source of temporary. These 'replacement' players are expected to have lower productivity than the
injured players, for they would be part of the original line-up for all games if they could generate
a better output for the team. Furthermore, I find compatibilities between a company's labor
economics and a football team. As a business owner, an efficient production function that properly
executes operations is a key enabler to reaching the ultimate goal, profit maximization. In other
words, in a setting where the size of the workforce cannot increase when a worker is absent,
coworkers may have to work additional time to make up for the productivity loss. This can lead to
sub-optimal individual and team performance in producing outputs. However, not all coworkers
can compensate for the absence of a peer, for not all of them have the skills to carry out the activity
of the absent peer.

Similarly, in the NFL setup, the team’s goal is to beat their rivals to secure a high standing by the
end of the season. A team needs to maximize points compared to their opponent to win the game.
If a player gets injured, productivity loss in terms of the (in)ability of the team to not concede (or
score) points might occur and the path to victory becomes hazy. Moreover, the nature of the game
requires some degree of specialty of the players. Defensive players may have different
characteristics than offensive players. This means that not all players can compensate for the
absence of a team member, since not all of them have similar skills that can substitute absentee’s
tasks.

7
3.2 Background Knowledge of American Football:

The objective of an NFL game is for a team to score more points than the opposing team. Teams
are formed by 46 players in the NFL, with 11 players taking the field at any time. The field is 100
yards long by 53 yards wide, with two 10-yard endzones at each end. The game starts with a
kickoff; the team with possession of the ball is the offense, and the team without the ball is the
defense. The distinction between these two sub-teams is crucial for this paper. The offense's job is
to move the ball up the field and score points. This can be done by rushing forward with the ball
or passing it up the field for a teammate to catch. The offense is given four chances (or four downs)
to make at least 10 yards. If the offense manages to move the ball 10 yards or more, they will retain
possession of the ball while given another four downs to make an additional 10 yards.

The defense's job is to stop the offense from moving the ball forward by tackling, stopping them
from moving forward, or forcing them off the field. If the offense fails to move the ball 10 yards
within four downs, the ball is given to the defending team at that point. The defending team will
then bring on their offensive players and try and move the ball in the opposite direction so that
they can score.

The teams will usually have three different units of 11 players that come on the field at different
times. They include: i) the offense, ii) the defense, and iii) special teams. Special teams are
specialist players that come on the field when there is a kick involved. Within the special teams is
a mix of offensive and defensive players mixed with either a punter or kicker for offense or a punt
returner for defense. For the scope of this study, I will exclude the special team since their
participation needs measurement mechanisms in the dataset.

There are four different ways of scoring: 1. Touchdown, 2. Extra points (Conversion), 3. Field
Goal, 4. Safety. The scope of this research applies total points scored/conceded, only
distinguishing the source of points for touchdowns.

8
3.3 Dataset

The datasets of this study come from the same source, [Link]. I use this
website's data alone to dodge potential mismatches across datasets. Data pertains to the NFL 2022
regular season, excluding Wild Card games, Divisional playoffs, and the well-known Super Bowl
game. There are 18 weeks in the regular season, where all teams play 17 games and have one bye
week. The NFL consists of 2 conferences, the AFC and the NFC, with 16 teams each, adding up
to 32 teams. For the scope of this paper, the data sample represents 25% of the NFL teams,
specifically from the four teams in the NFC North Division – Chicago Bears, Detroit Lions, Green
Bay Packers, and Minnesota Vikings – and the four teams in the NFC East Division – Dallas
Cowboys, New York Giants, Philadelphia Eagles, and Washington Commanders1.

Two primary datasets are of interest. First, the team game-by-game data, where there is a split
between defensive and offensive plays for every game a team played throughout the regular
season. Here, data pertaining to the points scored/received, yards scored/received, tackles, passes,
and further specifications on the nature of these result variables are found. Although some players
are present exclusively during defensive/offensive plays, others may have data for both types of
plays. That is, even if there is a clear distinction between defensive/offensive teams, some players
are able to play for both teams if the coach requires them to do so. Second, the injury data at the
individual level comes in the form of a weekly report, where all players who have suffered injuries
are displayed. This report categorizes players by the graveness of their injuries and gives the
probability of a player missing the next game due to the injury. The scope of this paper is
absenteeism in the workplace, resembled by player injuries that disallow them to play a game in
the NFL context. Thus, I move forward with the dataset exclusively considering players that
missed next game; that is, those players with probability of missing next game being 100%. I
follow this identification of injuries structure because it’s the structure the official website of the

1
See Appendix 1 for team codes.
9
 NFL ([Link]/injuries) follows2. After data cleaning, I merge these two datasets into a single
dataset to perform the STATA Software analysis. The final number of observations is 271, spread
across the 18 weeks observed during the analysis.

Table 1: Descriptive statistics of main variables of interest:

Obs Mean Std. dev. Min Max % of total

Weeks played 271 9.46 5.30 1 18 -

% injured per team 271 .084 .061 0 .26 -

Number of injured per team 271 2.52 2.18 0 11 -

Offensive players 136 - - - - 50.18

Defensive players 135 - - - - 49.82

Team points 271 23.55 9.95 0 54 -
Opponent points 271 22.69 9.45 -49 -3 -
Team achieved yards 271 349.65 74.74 182 520 -
Opponent achieved yards 271 348.74 81.72 148 570 -

Notes: These descriptive statistics belong to the merged dataset. All injury and individual-level data belongs
to the 8 teams studied: CHI, DET, DAL, GNB, MIN, NYG, PHI, and WAS. Team data belongs to the studied
teams. Opponent team data belongs to studied teams, for they play against each other, and to the following
teams: ARI, ATL, BAL, BUF, CAR, CIN, CLE, HOU, IND, JAX, LAR, NEW, NOR, NYJ, MIA, PIT, SEA,
SFO TAM, and TEN. All data from this table pertains to the regular NFL season 2022.

Table 1 displays the summary statistics of the most relevant variables for this study. The split
between defensive and offensive players is nearly symmetric. The average percent of injured
players for a team during a particular game is 8.4%. Total points scored by a studied team are

2
See Appendix 2 for detailed examples on how the injury report works and how I define an injured player for a
specific week.
10
 marginally higher than total points conceded. Similarly, yards achieved by a studied team are also
slightly above yards achieved by opponent teams. This may hint at a moderate superiority of the
teams subject to this research relative to their opponents3, only 2 of the 8 observed teams finished
the season below the (median) 16th standing. Not all teams have the same level, thus, their
performance varies.

Figure 1: Trend of % of injured players throughout the season:

Notes: These figures belong to the merged dataset. They display the trends of percent injured players over the
18 weeks that the regular season lasts by matching the. Figures to the right show data for offense players and
to the left for defense players, segmented into the two divisions from the NFC, North and East.

Figure 1 helps explain the trend in the percentage of injured players across teams throughout the
season. On the left side, I display the trend data for offense players, and on the right, defense
players. I expect an upward trend in the percentage of injuries because the likelihood of injury

3
Standings of observed teams: PHI (2nd), MIN (4th), DAL (7th), NYG (10th), DET (11th), WAS (16th), GNB (17th),
CHI (32th).
11
increases after each game slightly (Ekstrand et al., 2011). While defense players seem to follow
this trend, offense players have a hazy trend, with teams like PHI carrying 0 percent injuries during
a period of time.

4. Identification Strategy:
To estimate the effect of injury-driven absences in American Football teams and thus quantify the
effect of absenteeism in the workplace, I will use an OLS regression equation of the following
sort:

𝑌𝑗𝑡 = 𝛽0 + 𝛽1 ∗ 𝐼𝑁𝐽𝑗𝑡 + 𝛿𝑗𝑡 + 𝜌𝑗𝑡 + 𝜏𝑗 + 𝜑𝑗𝑡 + 𝜀𝑗𝑡 (1)

The outcome variable 𝑌𝑝𝑗𝑡 serves as a measurement of production output or performance. In this
case, it takes values for both performance indicators – points and yards achieved/conceded – for
team 𝑗 in game 𝑡. Points scored and conceded are not solely derived from offensive and defensive
plays, but also can be the result of a converted kick, a safety or a field goal from one of the two
teams in a game. Conversely, yards obtained and allowed exclusively stem from offensive and
defensive plays. Since the dataset only contains data on the team-game level for offensive and
defensive plays, the preferred output metric is yards in favor and against a team.

The focal explanatory variable for variation in output is 𝐼𝑁𝐽𝑗𝑡 , resembling the injuries a team 𝑗
sustains during a specific game 𝑡 and relative to the injury level, or in a conventional workplace,
the number of absent workers 𝑗 during a specific day 𝑡 relative to the rate of absent coworkers.
During the analysis, injuries were evaluated in 2 dimensions. First, injuries were perceived as the
percent of injured players a team sustained for a particular game. This calculation was done by
simply dividing the number of injured players by the total size of the squad - either attacking or
defending. Second, injuries are represented by the number of injured players of a team 𝑗 during
game 𝑡.

12
I consider the fixed effects (FEs) of some that may influence team performance by including them
as control variables to absorb these effects. 𝛿𝑗𝑡 contains the FE of an observed team 𝑗 during game
𝑡 and 𝜌𝑗𝑡 contains the FE of the opponent team for team 𝑗 during game 𝑡. These effects are included
because not all teams perform in the same way; some teams are better than others, so including
fixed effects for teams and their respective opponents is necessary. 𝜏𝑗 absorbs the FE for the week
in which team 𝑗 plays a given game 𝑡, and is included in the equation since the time factor might
play a role in team performance; teams learn from their own and rivals' capabilities throughout the
season, allowing for strategic performance. Lastly, 𝜑𝑗𝑡 captures the FE of the psychological effect
of playing at home since it may influence game outcomes (Legaz-Arrese et al. 2013). Naturally,
this OLS regression equation could include more possible control variables to avoid Omitted
Variable Bias and bypass endogeneity. The error term 𝜀𝑗𝑡 captures all the unaccounted noise of
equation (1).

A second regression equation (2) is formulated, where I categorize the percent of injured players
into five quantiles and represent them as dummies. Description of these quantiles are found in
Table 2. I categorize percent of injured to have detailed information about injuries without treating
them as a continuous variable.

𝑌𝑗𝑡 = 𝛽0 + 𝛽1 ∗ 𝐼𝑁𝐽𝑗𝑡,1 + 𝛽2 ∗ 𝐼𝑁𝐽𝑗𝑡,2 + 𝛽3 ∗ 𝐼𝑁𝐽𝑗𝑡,3 + 𝛽4 ∗ 𝐼𝑁𝐽𝑗𝑡,4 + 𝛽5 ∗ 𝐼𝑁𝐽𝑗𝑡,5 + 𝜔𝑡𝑗 + 𝜀𝑗𝑡 (2)

Equation (2) explains the same regression as equation (1), but now the explanatory variable percent
of injured players for team 𝑗 during game 𝑡 is categorized into five different quantiles, represented
by five dummies with structure 𝛽𝑛 ∗ 𝐼𝑁𝐽𝑗𝑡,𝑛 . Note that 𝐼𝑁𝐽𝑗𝑡,𝑛 only represents percent of injured
players and not number of injured players and that 𝑌𝑗𝑡 represents performance in the form of points
or yards. The estimates regression tables that use this categorized measure for absences are
benchmarked against the first group of injured players. Let 𝜔𝑡𝑗 capture all the FEs considered in
equation (1) - 𝛿𝑗𝑡 , 𝜌𝑗𝑡 , 𝜏𝑗 , 𝜑𝑗𝑡 – and 𝜀𝑗𝑡 be the error term. In the results section, Tables 3 and 5
display the injury estimates of equation (1) and Tables 4 and 6 of equation (2).

13
 Table 2: Description of categorized percent of injuries per team in quantiles:

DEFENSE PLAYERS OFFENSE PLAYERS

Range of % % Players Players in Range of % Players in Players in

injured in group group injured group group

INJURED 0%-2.3% 15.56 21 0% 25.53 32

GROUP 1

INJURED 2.3%-5.5% 22.96 31 0%-5.3% 19.85 27

GROUP 2

INJURED 5.5%-9.7% 22.96 31 5.3%-9.5% 15.44 21

GROUP 3

INJURED 9.7%-13.9% 17.78 24 9.5%-11.8% 21.32 29

GROUP 4

INJURED 13.9%-25.6% 20.74 28 11.8%-23.8% 19.85 27

GROUP 5

Notes: The bold percentages are crucial for understanding tables in the Results section of this paper. The 5
injured groups resemble the 5 quantiles in which observations are categorized and they aim to find the best
distribution of injury data depending on the frequency and injury value. However, sometimes group size may
differ by many observations, threatening the internal validity of the study as treatment groups have different
sizing.

There are several limitations to this identification strategy. A threat to this study's internal and
external validity could be that the relatively small sample size implies high p-values on the results,
meaning that the models have low explanatory power. Moreover, data might be incomplete as
various control variables are not included in equation (1), for example, individual-player-level data
or time series data potentially explaining the trend a team might be following. Data may also
contain measurement errors corresponding to errors made by the measurer (website) when
retrieving the data from the official final scores. Another threat to internal validity could be the
different measures for outcomes and the sub-sampling split between defense and offense players.
When looking at scored/received points, notice that not all points come from defensive/offensive
plays (e.g., conversions), so the data sample will ignore these points and will not attribute them to
either defense or offense teams.
14
5. Results

In this section, results from the analysis are displayed. Note that the structure of the results
distinguishes between offense and defense players for each team. Tables 3 to 6 show the estimates
of injuries on the team-level performance for both subcategories of teams – defense and offense -
after controlling for the opponent team's performance, the week in which the game takes place,
and whether the game is played at home or away. The performance indicators, brought into a
traditional workplace setup, may refer to 2 routes of measuring outcomes. Points scored/conceded
are interpreted as the end result of the input workers bring, the final product. Yards
obtained/allowed are interpreted as the means towards the end result, the path towards the final
product.

5.1 Defense Team Results:

Table 3 presents regression results for the defense team's performance, measured by points (Model
1 and 3) or yards (Model 2 and 4) conceived to rival teams. Here, the treatment variable injuries
for a given team j in a given match t and is measured by either the percentage of injured players
on team j found on the first row or by the number of injured players on team j in the second row,
during game t. Because players may greatly differ in their skill at the individual level, consider
solely Models 1 and 2, where the percent of injured players is the injury variable. This will allow
a more adequate interpretation of the results.

It is observed how points conceived against rivals should decrease as the percent of injured players
in a team increases (1) while the number of yards allowed to the rival team seems to increase as
the percent of injured players in a team increases (2). More precisely, if 10 percent of defense
players in a team are injured for a game, the team is expected to concede 3.175 fewer points and
18.16 more yards. Such behavior of variables is surprising as one would expect that if a team,
conditioned by injured defense players, is conceding more yards, they would also allow the
opponent team to score more points. This is not the case for these models. Yet, the coefficients are
not significant at any significance level, implying that injuries have no effect on team performance
in these models supposing a threat to their external validity.
15
 Table 3: The effect of absent defense team players on team performance, uncategorized injuries:

(1) (2) (3) (4)

Points conceived Yards conceived Points conceived Yards conceived
vs opponent vs opponent vs opponent vs opponent

Percentage of -31.75 181.6

injured (27.43) (191.9)

Number of -0.629 5.183

injured (0.706) (4.909)

Constant 25.69*** 334.0*** 25.10*** 331.8***

(2.432) (17.01) (2.497) (17.37)

Observations 130 130 130 130

R-squared 0.502 0.468 0.498 0.470
Notes: robust standard errors in parentheses. The percentage of injured refers to the average % number of
players who were injured for a team throughout the season games; The % of injured ranges between 0 - 25.6%.
The number of injured refers to the average number of players injured for a team throughout the season. The
number of injured ranges between 0 – 11. Data belongs to the North and East division teams from the NFC,
during NHL 2022 regular season. P-values: *** p<0.01, ** p<0.05, * p<0.1

Under the premise that replacement players should perform worse than injured players who would
have played if they were not injured, one can presume that the performance of the team will
worsen. Table 4 showcases injuries in a different way than Table 3. Here, injuries are categorized
into 5 quantiles, each containing roughly 20 percent of observations for the subsample, and using
quantile 1 as a benchmark, the reference quantile. Note that when categorizing the percent of
injured players into these 5 groups, the response of team performance becomes more logical to the
percent of absent workers. Model 5 statistically proves that there is a significant increase of 7
points conceded by a team when the percent of absent workers pertains to the group 2 interval [2.2
, 5.5], compared to quantile 1. However, in Model 5, allowed points to the rivals appear to become
less negative as the quantile of percent of injured players for a game increases relative to group 1,
even turning the coefficient positive when reaching the highest injury rate quantile. A possible
interpretation of this would be that the team or business reacts to workforce losses and starts

16
adapting to these absences, thus they suffer less from absenteeism in later periods of time when
the employer or team manager is able to foresee leaves in the labor pool and is able to strategize
upon it. As described in Figure 1, there is a upward trend in the percent of injured as the weeks
pass by and the season approaches the end.

Model 6 follows a similar line of reasoning as Model 5. Models 7 and 8 assess the source of
conceiving yards against opponents and aim to provide more information regarding what defensive
actions have a larger impact on yards lost by a team, but the interpretation of their coefficients is
rather ambiguous. While Model 7 states that 12.2 fewer yards are conceded due to missed
interceptions for injured group 2, Model 8 states that 36.0 additional yards are lost due to missed
tackles, compared to the reference group. Missed interceptions or tackles refer to the yards that the
opponent team is able to obtain by either passing the ball or rushing. This effect of injured group
2 in Model 8 is statistically significant at the 5% level. However, when looking at injured group 3,
the point estimate of absent workers has the same sign and (almost) the same magnitude as the
estimate for group 3 in Model 7. Lost yards stemming from missed interceptions have a negative
effect (14.6 additional yards lost, Model 7), and yards lost stemming from missed tackles have a
negative effect (14.5 additional yards lost, Model 8). It appears that the effect and magnitude of
intercepting and tackling exchange signs depend on the percent of injured players.

17
 Table 4: The effect of absent defense team players on team performance, categorized injuries:
(1) (2) (3) (4)
Points lost vs Yards lost vs Yards conceived due to Yards conceived
opponent opponent interceptions due to tackles

2. Injured type 7.895*** 23.81 -12.20 36.01**

(2.019) (18.96) (29.14) (13.49)

3. Injured type 3.036 29.00 14.55 14.95

(3.277) (37.40) (24.66) (23.94)

4. Injured type 0.888 34.79 9.186 25.60

(3.757) (31.14) (19.89) (21.67)

5. Injured type -2.575 20.64 31.40 -10.76

(2.894) (36.15) (22.70) (22.40)

Constant 20.72*** 327.6*** 207.8*** 119.6***

(1.793) (24.37) (17.63) (12.26)

Observations 130 130 130 130

R-squared 0.557 0.470 0.525 0.508
Notes: Robust standard errors in parentheses. The types of injuries refer to the different quantiles of the % of
players injured in one team. Each quantile contains roughly 20% of observations. Let X be the variable for %
of injured, then quantiles range as follows: Type 2: 0.023 < X < 0.055; Type 3: 0.055 < X < 0.097; Type 4:
0.097 < X < 0.139; Type 5: 0.139 < X < 0.256. Data belongs to the North and East division teams from the
NFC, during NHL 2022 regular season. P-values: *** p<0.01, ** p<0.05, * p<0.1.

Yet, note that the magnitudes of the coefficients for injuries on performance sub-indicators vary
among Models 7 and 8, especially for groups 2, 4, and 5. Taken into a more conventional workplace
environment, such as retail companies or production plants, and given the restricted sample, these
results imply that there is no clear ranking or guideline to argue what type of play in a game/role
of a worker in a company sustains a greater impact on how production is achieved4. Nevertheless,
there is significant evidence that, when the percent of absentees is not too high ( between ~2% and
~5%), the end product, points against, increases substantially (Model 5), and one of ways the end

4
Here I discuss whether a specific defensive play, intercepting or tackling, has greater or lower impact on how team
performance varies.
18
 product is defined, yards allowed due to missed tackles/opponents rushing yards, also increase
accordingly (Model 8). However, I want to make clear that, except for the coefficients for quantile
2 in Models 5 & 8, my coefficients are statistical zeros, meaning that injuries have no effect on
team performance.

5.2 Attack Team Results:

Pursuing a similar analytical path as Tables 3 & 4, Table 5 presents the results of offensive players'
absences, uncategorized, and their effect on team performance. Models 9 and 10 exhibit logical
estimates of injuries on team performance. A 10 percent of injuries on the offensive team should
yield a decrease of 0.163 points per game and a decrease of 16.3 yards per game. Again, estimates
for Models 9-12 are not significant at any level, so there is no evidence of causal effect of attacking
players’ injuries on team performance.
Table 5: The effect of absent offense team players on team performance, uncategorized injuries:

(9) (10) (11) (12)

Points scored vs Yards earned vs Points scored vs Yards earned vs
opponent opponent opponent opponent

Percentage -1.625 -163.0

of Injured (23.54) (252.5)

Number of -0.199 -10.05

Injured (1.270) (13.02)

Constant 23.72*** 363.6*** 23.89*** 365.7***

(1.871) (20.07) (1.905) (19.53)

Observations 132 132 132 132

R-squared 0.456 0.449 0.456 0.451
Notes: robust standard errors in parentheses. The percentage of injured refers to the average % number of
players who were injured for a team throughout the season games; The % of injured ranges between 0 - 25.6%.
The number of injured refers to the average number of players injured for a team throughout the season. The
number of injured ranges between 0 – 11. Data belongs to the North and East division teams from the NFC,
during NHL 2022 regular season. P-values: *** p<0.01, ** p<0.05, * p<0.1.
19
 Table 6: The effect of absent offense team players on team performance, categorized injuries:

(13) (14) (15) (16)

Points scored vs Yards gained vs Yards obtained Yards obtained
opponent opponent due to passing due to rushing

Injured type 2 -5.421 -11.57 18.03 -29.60

(3.753) (33.70) (23.74) (21.10)

Injured type 3 -0.286 12.95 8.94 4.012

(3.956) (43.74) (38.99) (34.11)

Injured type 4 -2.732 -18.99 -9.55 -9.441

(4.757) (48.43) (36.32) (24.44)

Injured type 5 -3.216 -26.42 -0.036 -26.39

(4.898) (46.19) (28.62) (31.40)

Constant -20.72*** -327.6*** 211.29*** 148.6***

(1.793) (24.37) (21.75) (17.44)

Observations 132 132 132 132

R-squared 0.475 0.456 0.558 0.527
Notes: Robust standard errors in parentheses. The types of injuries refer to the different quantiles of the % of
players injured in one team. Each quantile contains roughly 20% of observations. Let X be the variable for %
of injured, then quantiles range as follows: Type 2: 0.000 < X < 0.053; Type 3: 0.053 < X < 0.095; Type 4:
0.095 < X < 0.118; Type 5: 0.118 < X < 0.238. Data belongs to the North and East division teams from the
NFC, during NHL 2022 regular season. P-values: *** p<0.01, ** p<0.05, * p<0.1.

Table 6 displays the results for the offensive team players’ injuries, categorized in the same way
as in Table 4. Injuries do not have an effect on the attacking performance of a team, however, it is
worth mentioning some incongruencies in this table. Again, under the premise of expecting injuries
to have a negative influence performance, it seems that, in some cases, injuries can take a positive
coefficient on production (Model 15, injured group 2) relative to the reference injury group.
Despite that, Models 13, 14, and 16 appear to have a more predictable behavior for injured group
2, with performance indicators dropping as injuries take place. Nonetheless, the noticeable
variation of estimates’ sign and magnitude when benchmarked against group 1, becoming

20
dependent on which bracket of the injured percent is analyzed together with the non-significance
of point estimates, make these models inconclusive.

6. Discussion, Conclusion, and Limitations:

6.1 Discussion
I investigate how absenteeism affects holistic team productivity in the context of the National
Football League. In this setting, I perform three main analyses. First, even if there is evidence of
positive spillover effects, I evaluate the possibility that defensive and offensive players might have
different abilities or specializations that are untransferable. This implies that measuring their
productivity altogether can be misleading, as different roles in the team might have different
sizeable impacts on the final outcome. To solve this, I divide the dataset into defensive and
offensive groups. Second, I categorize injuries into quantiles to create a more realistic scenario
since some teams sustained absences of >15 percent of the players' pool for a game. Third, I
explore two different outcome measures, points scored/conceded and yards achieved/conceded, to
resemble two ways of measuring team performance – the final outcome (points or end product)
and how the final outcome is achieved (yards or means towards end product).

The main results from this investigation point out several interesting implications of absenteeism,
the most important of them being the value of the estimates. Although all 16 models present point
estimates different than zero, hinting at a possible effect of absenteeism in team performance.
Nonetheless, only two estimates are not statistical zeros, meaning no statistically significant effect
of absences in team productivity for most cases. Next, note that nearly all eight models with
categorized injuries present a negative estimates of absences in group performance. This supports
the idea that collective efforts can yield beneficial outcomes and that a disruption of the coworker
network may lead to inferior productivity.

Nevertheless, we must pay attention to the two main drawbacks to the presented models; i) lack of
significant coefficients at the 5% level, and ii) inconsistencies in coefficient signs as the percentage
of injured players changes. With that in mind, the only Models that provide significant estimates

21
are Models 5 & 8. Model 5 proves that any team that sustains between 2.3% and 5.5% of defensive
injured players in a given game will allow opponents to score seven additional points, on average,
and Model 8 indicates that a team with these very same characteristics on injuries will concede
36.0 yards to opponents due to missed tackles, on average. I also underline the difference in the
size of coefficients for both defensive and offensive teams. When comparing points
conceded/scored (Models 5 & 13) and yards conceded/obtained (Models 6 & 14), I observe that
injured defensive players have a larger impact on team performance than injured offensive players.
This suggests that the operational roles of employees affect productivity heterogeneously. Another
divergence between defense and offense is the source of yards conceded/achieved. Defensive
players allow fewer yards due to failed tackles, while the offensive counterpart achieves more
yards from passing the ball, insinuating a larger difference in how each player performs.

In line with the findings of Mollerman & Slomps (1999), I find adverse effects of absenteeism in
the production function of a team independent of the distribution of the workers' roles. Negative
impacts on group performance are noticeable regardless of the player's position. While Hoey et al.
(2023) find that a disruption caused by absences in the work environment leads to a marginal
decrease in the team's productivity, even when a replacement covers the vacancy, my estimates
cannot provide enough support for this statement in most cases. Potential mechanisms for this drop
in efficiency are working complementarities that may arise between workers when collaborating
toward an end result, such as spillover or positive peer effects, and the inability of the employer to
find a replacement that fits the job and matches the productivity of the absentee.

6.2 Limitations:

This paper has limitations and threats to both internal and external validity. The research does not
find estimates statistically different form zero, meaning no effect of injuries on team performance
is can be supported and internal validity is threatened. Because of the small number of observations
of my sample I can potentially suffer from low explanatory power, which decreases my ability to
find an effect, should it exist. Additionally, even if the paper’s aim is not to explain all variation
in team performance, models suffer from poor goodness of fit (<0.56 r-squared value). There is

22
still room for improvement in this field, adding more control variables might explain the
performance variation better.

Concerning external validity, I chose the eight selected teams for the analysis fully arbitrarily,
indicating a non-random sample selection. A substantial threat to external validity is my small
number of observations. This implies a risk of incapability to extrapolate my results to other setups
or larger samples. Due to my sample characteristics and the player specific sporting-laboral
conditions, generalization of findings to other non-sporting settings or different contexts is not
possible. The NFL setting is restricted by specific working hours of players, which is an advantage
when measuring productivity, but it is not in line with the modern work environment where
workers might work more or less than what their contract stipulates. Also, the NFL assumes that
replacements must be available for the next game (working day) so workforce size is constant.
Again, this is favorable to measure productivity, but the reality is that companies do not replace
absentees within one day, making the size of the workforce vary Furthermore, the study only
considers a single season of the NFL. Teams may change over time and so their expected
performance, exposing my paper to temporal validity. Replication of this research on the full
population of the NFL context for several seasons could bring cleaner external validity. Because
of the above reasons, it is erroneous to apply these findings to the labor market in general.

6.3 Conclusion

The broader economic implications of this analysis must be treated carefully. While I find two
coefficients causing a significant negative impact of (defensive) injuries on team outcomes, most
of my estimates are not significant and don’t show any significant effects. Still, due to existing
literature on absenteeism in teamwork the environment (Bartel et al., 2014, Herrmann and Rockoff,
2012, and Stewart et al., 2003, among others) and their statistically significant findings, there are
reasons to believe that absenteeism has a negative effect on group performance. This negative
causal relationship can be found in sporting contexts (Hoey et al., 2023) like the one I study in this
paper. Therefore, despite not showing any significant effects with my results, potentially due to

23
the limitations of my data sample or research setting, we can get an idea of what the consequences
of absenteeism in the workplace are; a decrease in group productivity.

Managers should try to overcome the plausible adversities of the negative effects of absenteeism
and think strategically. That is, to have an available, high-quality labor pool where they can source
replacement workers that can sufficiently (or exceedingly) fulfill the absentee's tasks and maintain
(or improve) the level of team productivity. This is food for thought for companies when
strategizing their recruitment processes and might serve as a motivator to boost their agility to
maintain the team dynamic when absences take place. The challenge remains for businesses and
implies performing profile screening adequately to minimize unforeseen absences or storing
unselected applicants efficiently in case they are needed to cover an unexpected vacancy. Finally,
three avenues for further research are i) the analysis of absenteeism in a different non-sporting
context, ii) the investigation of the seasonality of absenteeism, essentially whether the timing of
the absence has any implications for production outcomes, and iii) the effects of an alternative
behavior to absenteeism that is very common in the post-pandemic era, working remotely.

References:

Bacharach, S. B., Bamberger, P., & Biron, M. (2010). Alcohol consumption and workplace
absenteeism: the moderating effect of social support. Journal of Applied Psychology, 95(2), 334.
Bartel, A. P., Beaulieu, N. D., Phibbs, C. S., & Stone, P. W. (2014). Human capital and productivity
in a team environment: evidence from the healthcare sector. American Economic Journal: Applied
Economics, 6(2), 231-259.
Berman, S. L., Down, J., & Hill, C. W. (2002). Tacit knowledge as a source of competitive
advantage in the National Basketball Association. Academy of management Journal, 45(1), 13-31.
Herrmann, M. A., & Rockoff, J. E. (2012). Worker absence and productivity: Evidence from
teaching. Journal of Labor Economics, 30(4), 749-782.
Dahl, G. B., Løken, K. V., & Mogstad, M. (2014). Peer effects in program participation. American
Economic Review, 104(7), 2049-2074.
De Paola, M., Scoppa, V., & Pupo, V. (2014). Absenteeism in the Italian public sector: The effects
of changes in sick leave policy. Journal of Labor Economics, 32(2), 337-360.

24
Drew, S., & Coulson‐Thomas, C. (1996). Transformation through teamwork: the path to the new
organization?. Management Decision, 34(1), 7-17.
Ekstrand, J., Hägglund, M., & Waldén, M. (2011). Injury incidence and injury patterns in
professional football: the UEFA injury study. British journal of sports medicine, 45(7), 553-558.
Falk, A., & Ichino, A. (2006). Clean evidence on peer effects. Journal of labor economics, 24(1),
39-57.
Forman, S., Gillete, G., Pullis, K., & Palmer, P. 2022 NFL Standings & Team stats, Sports
Reference [Link]
Forman, S., Gillete, G., Pullis, K., & Palmer, P. 2022 NFL Standings & Team stats, Sports
Reference [Link]
Guryan, J., Kroft, K., & Notowidigdo, M. J. (2009). Peer effects in the workplace: Evidence from
random groupings in professional golf tournaments. American Economic Journal: Applied
Economics, 1(4), 34-68.
Halpern, M. T., Shikiar, R., Rentz, A. M., & Khan, Z. M. (2001). Impact of smoking status on
workplace absenteeism and productivity. Tobacco control, 10(3), 233-238.
Hamilton, B. H., Nickerson, J. A., & Owan, H. (2003). Team incentives and worker heterogeneity:
An empirical analysis of the impact of teams on productivity and participation. Journal of political
Economy, 111(3), 465-497.
Hassan, S., Wright, B. E., & Yukl, G. (2014). Does ethical leadership matter in government?
Effects on organizational commitment, absenteeism, and willingness to report ethical
problems. Public Administration Review, 74(3), 333-343.
Hoey, S., Peeters, T., & van Ours, J. C. (2023). The Impact of Absent co-workers on Productivity
in Teams. Labour Economics, 102400.
Jacobson, B. H., Aldana, S. G., Goetzel, R. Z., Vardell, K. D., Adams, T. B., & Pietras, R. J. (1996).
The relationship between perceived stress and self-reported illness-related absenteeism. American
Journal of Health Promotion, 11(1), 54-61.
Legaz-Arrese, A., Moliner-Urdiales, D., & Munguía-Izquierdo, D. (2013). Home advantage and
sports performance: Evidence, causes and psychological implications. Universitas
Psychologica, 12(3), 933-943.
Maliranta, M., Mohnen, P., & Rouvinen, P. (2009). Is inter-firm labor mobility a channel of
knowledge spillovers? Evidence from a linked employer–employee panel. Industrial and
Corporate Change, 18(6), 1161-1191.
Mas, A., & Moretti, E. (2009). Peers at work. American Economic Review, 99(1), 112-145.

25
Molleman, E., & Slomp, J. (1999). Functional flexibility and team performance. International
Journal of Production Research, 37(8), 1837-1858.
Stewart, W. F., Ricci, J. A., Chee, E., Hahn, S. R., & Morganstein, D. (2003). Cost of lost
productive work time among US workers with depression. Jama, 289(23), 3135-3144.

Appendix 1:Ways of scoring points in the NFL in detail:

There are four different ways of scoring: 1. Touchdown: if the ball is carried into the endzone
area, or thrown and caught in the endzone, this is a touchdown and is worth 6 points. 2. Extra
points (Conversion): once a touchdown has been scored, the offense has the option of kicking the
ball through the uprights for an extra point (15 yards from the endzone), or trying to pass or run
the ball into the endzone again for an extra two points (2 yards from the endzone). 3. Field Goal:
At any time, the team with the ball can kick the ball between the posts and over the crossbar. To
do this, they must hand it to a teammate who will hold it on the ground ready for a kicker to make
the kick. A successful kick scores 3 points. 4. Safety: if the defense tackles an offensive player
behind his own goal line, the defending team scores two points.

Appendix 2: Example of injury report:

In the official injury report from the NFL website, [Link]/injuries, each week injured players get
categorized into: Probable, Questionable, Doubtful, and Out. The reading of these is:

Probable 75% chance of playing next game

Questionable 50% chance of playing next game
Doubtful 25% chance of playing next game
Out The player is out for next game

For this analysis, I only use players who fall under the ‘Out’ category and take them as injured.
More information can be found on the cited dataset website:

[Link]
26
Appendix 3: Team codes & standings NFL season 2022:

TEAM CODE AFC/NFC STANDING IN REGULAR SEASON

Chiefs KC AFC 1
Eagles PHI NFC 2
Bills BUF AFC 3
Vikings MIN NFC 4
49ers SFO NFC 5
Bengals CIN AFC 6
Cowboys DAL NFC 7
Ravens BAL AFC 8
Chargers LAC AFC 9
Giants NYG NFC 10
Lions DET NFC 11
Jaguars JAX AFC 12
Dolphins MIA AFC 13
Steelers PIT AFC 14
Seahawks SEA NFC 15
Commanders WAS NFC 16
Packers GNB NFC 17
Patriots NEW AFC 18
Buccaneers TAM NFC 19
Falcons ATL NFC 20
Panthers CAR NFC 21
Browns CLV AFC 22
Saints NOR NFC 23
Jets NYJ AFC 24
Titans TEN AFC 25
Raiders LV AFC 26
Broncos DEN AFC 27
Rams LAR NFC 28
Colts IND AFC 29
Cardinals ARI NFC 30
Texans HOU AFC 31
Bears CHI NFC 32

27