Join the Union and be Safe; The Effects of Unionisation on
Occupational Safety and Health in the European Union
Abstract: This paper investigates the effect of unionisation on occupational safety and health (OSH),
as measured by the fatal and non-fatal work accidents, after controlling for the country GDP. It uses a
panel sample of 10 European Union countries, for the period 1982-2006. The study takes into account
the time persistence in work injuries and the endogenous nature of the work injuries – union density
relationship, by using GMM regression models to control for endogeneity. In addition, the effect of
union density is decomposed into a temporary and permanent effect. It is shown that increasing union
density is associated with a decrease in the number of both fatal and non-fatal work injuries. These
findings have important implications for the design of OHS and industrial relations policies.
Keywords: Work accidents, Union Density, Mundlak Decomposition, GMM
JEL: J28, J51, J81, C33
1
1. Introduction
Health and safety at the workplace is a significant determinant of the well-being of the
employed population. Although work injuries exhibit downward trends the last decade, they
still cause significant economic, social and emotional costs to the employees’ involved and
negative externalities to their families as well (European Agency for Health and Safety at
Work, 2001). In particular, the economic costs of work-related injuries and diseases, reflected
in reduced productivity and output, are estimated at 4% of the worldwide GDP (Pouliakas
and Theodossiou, 2010), while the non-pecuniary costs and the negative externalities to the
victims’ families are also considered to be significant (Boden, 2005).
An increasing literature focuses on the workplace health practices, institutions and norms
that may affect OSH (occupational safety and health). Labour unions play an important role
towards the improvement of OSH (Pouliakas and Theodossiou, 2010). While theory
postulates that unions contribute in numerous ways towards the enhancement of OSH, the
empirical evidence provide
conflicting findings regarding the role of unionisation on
workplace safety (Morantz, 2013).
The present study revisits this issue by investigating the effect of unionisation on workrelated injury rates using a panel of ten European Union countries during the period 19822006. The degree of unionisation is approximated by the union density index. Feasible
Generalized Least Square (FGLS) regression models are estimated in order to control for
group-wise heteroskedasticity and cross-panel correlation. In addition, as a robustness check
and in order to take into account the time persistence in work injuries and the endogenous
nature of the work injuries – unionisation relationship, GMM (Generalized Method of
Moments) regression techniques are utilised. The regression results show that both fatal and
2
non-fatal work injuries tend to decrease as union density increases, indicating the protective
role of unions on occupational safety and health.
The next section (Section 2) provides a literature review survey on the conflicting
evidence regarding the role of unions on OSH. Section 3 presents the variables utilised in the
analysis and Section 4 provides details on the econometric methodology. Section 5 presents
the empirical results and finally, Section 6 concludes.
2. Literature Review
The literature indicates that labor unions use their political influence and engage actively
through collective bargaining, representation in health and safety committees and the
undertaking of relevant actions, in enhancing workplace safety (Donado, 2007). In addition,
unions can promote OSH educational activities, disseminate OSH information to workers and
contribute to the enforcement and implementation of OSH regulations (Morantz, 2013).
Theory predicts that unions contribute to the improvement of working conditions, obtain
higher compensation benefits for employees who suffered work-related health problems, and
in general, represent effectively the employees’ interests regarding health and safety at the
workplace (Fenn and Ashby, 2004; Freeman, 1994; Hirsch et al., 1997; Morse et al., 2003).
However, empirical studies often fail to support these hypotheses. Boal (2009) in a similar
study for the US coal mine industry, found evidence that the negative relationship between
unionism and workplace fatalities is mainly driven by the “efficient-bargaining” model of
union behavior, namely workers bargain and succeed in enjoying safer working conditions.
However, other possible paths through which unionism might affect workplace injuries are
also suggested in the literature. Since union presence is associated with higher fringe and
compensation benefits receipts (Hirsch et al., 1997) and higher wages, it is possible that
3
union members may be more prone in exchanging higher wages for lower workplace safety
(Donado, 2007).
Many studies find that in industries with strong unionisation presence, work injuries
appear to be lower in comparison to industries with weaker union presence (Donado, 2007).
Siebert and Wei (1994) provide evidence that unionised workers experience lower fatal injury
rates. A more recent study of Litwin (2000) found that the presence of unions in the British
labour market contribute to the reduction of work injuries. Nichols et al. (2007) argued that
unionisation can reduce work injuries through the participation of union members in health
and safety committees at the workplace. The above findings are in line with Reilly et al.
(1991), who show that lower injury rates are observed in plants where union representatives
participate in the occupational safety and health committees compared to the remainder.
Several studies provide empirical evidence in favour of a positive relationship between
unionisation, work-related injuries and absenteeism (Fishback, 1986; Leigh, 1981; Veliziotis,
2010). One explanation for the observed positive relationship is the presence of endogeneity.
It seems that while increased unionisation affects workplace injuries, at the same time riskier
industries are more unionised, since higher accidents rates may motivate workers to organise
in unions in order to protect themselves from the hazardous working conditions (Fenn and
Ashby, 2004; Nichols et al., 2007). Indeed, the evidence indicates that workers who face
increased hazards at work exhibit increased participation in union activities, as a response to
the hazardous job conditions (Hirsch and Berger, 2001; Robinson, 1990).
Furthermore, many studies argue that the report rate of work injuries is higher in
workplaces with strong union presentation and such an effect can bias upwards the effects of
unionisation on injury rates (Donado, 2007; Fenn and Ashby, 2004; Morse et al., 2003;
Nichols et al., 2007). Fenn and Ashby (2004) provide corroborating evidence indicating that
the higher is the proportion of unionised workers in an establishment, the higher would be the
4
risk of reported injuries or illnesses. They argue that highly unionised enterprises show high
reporting of injuries at work, since they enjoy higher compensation due to the union
representation. Towards this end, Morantz (2013) finds that unionisation is associated with a
large decrease in traumatic injuries and overall fatalities in the coal mine industry for the
USA. However, she also reports higher total and non-traumatic injury rates, which is
attributed to the higher reporting rates in the unionized industries. In addition to the above,
Donado (2007) attributes the observed increased work injury rates in unionised sectors to
behavioral issues to the workers involved, since they tend to underestimate the true job
hazards, and thus experience higher work accident rates. All in all, it seems that up-to-date
literature has not provided conclusive evidence on the role of unions upon OSH, thus further
research is needed to shed more light on this issue.
3. The Dataset
The data used in this study is a panel of 10 European Union countries (Austria, Denmark,
Finland, France, Ireland, Italy, Portugal, Spain, Sweden, UK) for the time period 1982-2006.
Data on fatal and non-fatal work accidents and employee population are drawn from the
International Labour Organization (ILO) database (LABORSTA). The ILO database contains
a great variety of information regarding labour market characteristics for numerous countries
and years. The variables to be explained are therefore1:
-
Total fatal injury rates per 100,000 employees
-
Total non-fatal injury rates per 100,000 employees
-
The independent variable of interest is the Union densityand
1
Data on fatal and non-fatal injuries disaggregated by industrial sector are also available at ILO database.
Unfortunately, the increased missing values on the employee population per industry and union density, did not
allow exploitation of this information. Data can be retrieved from: http://laborsta.ilo.org/
5
-
GDP per capita (in PPP) which enters as a control variable to capture the level of
economic performance of the country.
Since numerous studies (Boone and van Ours, 2002; Davies et al., 2009; Catalano, 1979;
Sasaki, 2010; Steele, 1974; Ussif, 2004) show that occupational accidents and injuries rates
are affected by the macroeconomic conditions and the business cycle, a proxy for these
conditions (real GDP per capita) is included in the regressions. The data on real GDP per
capita are derived from the European Database ‘Health for All’ of the World Health
Organization (WHO). Data on trade union density (the ratio of wage and salary earners that
are trade union members divided by the total number of wage and salary earners) are drawn
from the Annual Labour Force Statistics of the OECD database.
As Pouliakas and Theodossiou (2013) argue, work-related accidents are attributed to both
the individual employee characteristics and the overall economic and socio-cultural climate.
For example, working hours are among the most commonly met determinants of work-related
injuries (Souza et al., 2014). Towards this end, information on country unemployment rates,
the average annual hours worked per worker and the annual growth rate of unit labour cost
are drown from the OECD Statistics database. Average years of schooling for individuals of
15 years and older are drawn from the Barro-Lee database. The values are available every
five years therefore the in-between values are calculated by linear interpolation. To examine
the sensitivity of the results a number of covariates are introduced into the models, as
follows:
-
Country Unemployment Rates
-
Average Years of Schooling
-
Hours of Work
6
-
Unit Labour Cost (Annual Growth Rate)
Finally, the instrumental variable used to control for the endogenous relationship
between work injuries and union density, is the “days lost due to strikes and lockouts per
100,000 employees” and the information is drawn from the International Labour
Organization (ILO) database (LABORSTA). The summary statistics of the variables included
in the regressions are shown in Table 1.
4. Econometric Methodology
The following Random Effects models are utilised:
Fatal Work Accidentsi ,t  b1  GDPi ,t  b2  Union Densityi ,t  St  ai   i ,t 
(1)
Non  Fatal Work Accidentsi ,t  b1  GDPi ,t  b2  Union Densityi ,t  St  ai   i ,t 
(2)
The subscripts i and t denote the country and the year period respectively, ai is the
country-specific random-effects intercepts and St is the year dummies variables. It is
assumed that VAR   i ,t    2 , VAR  i    2 and the ai ,  i ,t are independent. Equations
(1) and (2) are estimated with FGLS. The advantage of using FGLS is that this methodology
takes into account the existence of both groupwise heteroskedasticity and contemporaneous
correlation. It is appropriate to assume that the common European legislations on health and
safety at work will affect work injuries across the countries under examination. As
Wooldridge (2002) and Greene (2002) argue, if there is cross-panel correlation then the
FGLS estimator is more efficient.
7
In addition, the Mundlak (1978) methodology is utilised as follows:


Fatal Work Accidentsi ,t  b1  GDPi ,t  b2  Union Densityi ,t  Union Density i 
 b3  Union Density i  St  ai   i ,t 
(1)


Non  Fatal Work Accidentsi ,t  b1  GDPi ,t  b2  Union Densityi ,t  Union Density i 
 b3  Union Density i  St  ai   i ,t 
(2)
The Mundlak (1978) methodology offers a simple and straightforward way to distinguish
between the permanent and the temporary effect of union density upon work-related fatalities
and injuries. In particular, Union Density i is the mean union density for each country and
it is the permanent effect. The term
Union Density
i ,t

 Union Density i captures the
temporary effect, namely the deviations from the country average union density. Both FGLS
models and FGLS models with Mundlak decomposition are re-estimated with the inclusion of
more independent regressors, in order to check the sensitivity of the results.
Two methodological shortcomings are addressed. First the dynamic nature of work-related
injuries is taken into account. This is an implication of the fact that the experience of a work
related health problem at present may be affected by work related health condition suffered in
the past. For example, musculoskeletal disorders are found to be the most frequently
experienced work-related health problem in the European Union countries. After the onset of
a musculoskeletal disorder (for example, low back pain) due to working conditions there is a
high probability of recurrence of this problem (De Beeck and Hermans, 2000). Second, the
endogenous nature of the relationship between trade union density and work injuries should
8
be controlled for in order to derive unbiased estimates. Both the dynamic relationship and
endogeneity issues can be accounted for by the use of GMM models.
The Arellano and Bond (1991) first differenced GMM estimator estimates equations (1)
and (2) after introducing lagged values of the dependent variable and first differencing the
variables in order to eliminate unobserved heterogeneity. The first differenced equation uses
as instruments the past values of union density. However, past union density should continue
to affect the growth rate of work injuries at subsequent years through the union’s actions
regarding occupational health and safety policies and framework.
Arellano and Bover (1995) argue that the system GMM estimator is improved in terms of
efficiency in comparison to the differenced GMM estimator, since the lagged variables in
levels may be weak instruments for the differenced equation. In line with the above, the main
model utilised in this study is the two-step system GMM estimator (the equations (1) and (2)
respectively) to obtain the following system of two equations (one in differences, as in the
differenced GMM estimator and one in levels) for fatal and non-fatal work injuries ( 
denotes the difference operator)2:
 Fatal Work Accidentsi ,t 
 Fatal Work Accidentsi ,t 1 
 GDPi ,t 
 Fatal Work Accidents   b0  b1   Fatal Work Accidents   b2   GDP  
i ,t 
i ,t 1 
i ,t 



 Union Densityi ,t    i ,t 
b3  


 Union Densityi ,t    i ,t 
(3)
2
The system of equations also includes control for the time trend variable. Alternative estimation models with
year dummies were also utilised but the models presented indicated a better fit, due to the strong collinearity of
the year dummies with the independent variables.
9
 Non  Fatal Work Accidentsi ,t 
 Non  Fatal Work Accidentsi ,t 1 
 Non  Fatal Work Accidents   b0  b1   Non  Fatal Work Accidents  
i ,t 
i ,t 1 


 GDPi ,t 
 Union Densityi ,t    i ,t 
+b2  
 b3  



 GDPi ,t 
 Union Densityi ,t    i ,t 
(4)
The above systems of equations for fatal work injuries (3) and non-fatal work injuries (4)
respectively, are estimated using the two-step system GMM estimator. This estimator is
considered to be more efficient than the differenced GMM estimator, since it uses as
instruments the lagged differences of union density for the equation in levels and the lagged
levels of union density for the equation in differences. The lagged values of Union Densityt 2
and so on, are assumed to be valid instruments for the first differenced equation.
Donado (2007) also utilised GMM regression models to examine the endogenous nature of
work injuries – unionisation, by using as identifying restrictions the past values of union
indicators on the assumption that union OSH actions exert an effect on work injuries at
subsequent years. In order to improve the performance of our models, the variable “days lost
due to strikes and lockouts per 100,000 employees” is also included as an instrumental
variable. This indicator of union activities related to strikes and lockouts is found consistently
to affect trade union density in existing research (Lesch, 204). However, there is no a priori
reason to expect that strike activity is related to work-related injuries.
In addition, the Sargan test is also used in order to check for the validity of the overidentifying restrictions. The AR(1) and AR(2) tests are also presented since there should not
be any evidence of second-order serial correlation while first-order serial correlation is
expected in the model of first differences (Arellano and Bond, 1991). The GMM
methodology is recommended for "small T, large N" panels (Roodman, 2009). However, the
10
dataset used in this study provides a small number of countries. Hence, the GMM estimation
is used as a robustness check for the FGLS specification which is the main specification in
this study.
5. Regression Results
Table 1 presents descriptive statistics for the variables of interest at country level. Fatal
injury rates per 100,000 workers vary greatly across countries. Sweden exhibits the lowest
rate with only 0.28 work related fatalities per 100,000 workers. Spain faces a high incidence
of work-related fatalities (8.38 per 100,000 workers) and of non-fatal work injuries (5,026.84
per 100,000 workers). The lowest incidence for non-fatal work injuries is observed for the
UK (632.91 non-fatal work injuries per 100,000 workers). Union density also varies across
countries with the lowest participation observed for France (10%) and the highest for Sweden
(81%).
Figures 1-3 show the mean fatal and non fatal work injury rates and union density rates for
the countries included in the sample. A downward trend is observed for all three series. The
mean rate of work-related fatalities exhibits a greater volatility throughout the period 19822005, and it has a similar pattern to the mean of non-fatal work injuries. Interestingly, both
series exhibit a sharp increase in the middle to late 1980’s and they drop afterwards. The
mean union density rates are declining from 1982 up to 2005, with a sharp increase in the
1990-1995 period.
Fatal work injuries
The results for the FGLS models and the system GMM estimator for fatal and non-fatal
work injuries are presented in Tables 2 and 3 respectively. Table 2 presents the regression
results for fatal work injury rates for 100,000 workers in the period 1982-2005. In the first
11
column of Table 2 (Model 1), the results from the FGLS effects model are presented, and the
second column (Table 2, Model 2) presents the FGLS Mundlak decomposition model results.
The third column (Table 2, Model 3) presents the system-GMM results when endogeneity
and time dependence is taken into account. In this first set of models only GDP per capita is
included as a regressor apart from union density.
The findings indicate that increasing union density is associated with a decreasing
incidence of fatal work-related injuries. The results are in accordance with the GMM results
(which are used to evaluate the robustness of the estimates). The results do not change when
the union density effect is decomposed into a temporary and the permanent effect. Both the
temporary and the permanent effect of union density on OSH are statistically significant. All
in all, an increase in union density is associated with a lower rate of fatal work injuries. This
implies that increasing union density helps unions to achieve better outcomes on occupational
health and safety conditions.
The findings regarding the impact of union density upon work-related fatalities remain
unaltered when more covariates are introduced in the models to evaluate the sensitivity of the
model to variations in specification. Models 4 and 5 of Table 2, extend the regression Models
1 and 2 by including more regressors in the models. The impact of increasing union density
upon work-related fatalities remains strong but the temporary effect of union density
becomes statistically insignificant in this case (Model 5), whereas a permanent negative
effect of union density upon work-related fatalities is retained. This last finding is in line with
the conventional wisdom since it is expected that the action of unions towards the decrease of
work-related fatalities takes some time to manifest itself.
Regarding the effects of the control variables, the effect of per capita GDP upon fatal work
injuries is not straightforward. The sign of the relationship between per capita GDP and
work-related fatalities reverses when more covariates are added to the models. When adding
12
more covariates to the models (Models 4 and 5) economic upturns are associated with an
increase in work-related fatalities. In addition, during increasing unemployment work-related
fatalities increase. The remaining covariates hold the expected signs. An increase in average
schooling is associated with a decrease in work-related fatalities. Similarly, an increase in
hours of work and the labour cost is associated with an increase in work-related fatalities.
Non-fatal work injuries
The findings are similar when one considers the effect of union density upon non-fatal
work injuries (Table 3, Models 1-5). Increasing union density is associated with reduced nonfatal work injury rates (Model 1). This finding is also supported by the GMM estimations
(Model 3). The Mundlak decomposition reveals only a permanent effect of union density
upon non-fatal work-related injuries, while the temporary effect is insignificant (Model 2). In
the augmented models, the effect of union density upon work-related injuries remains strong
and negative (Model 4). The inclusion of more covariates in the Mundlak decomposition
models cause the temporary effect to become significant whereas the permanent effect
becomes statistically insignificant (Model 5).
Regarding the effects of the GDP which serves as a control variable, the augmented
models (Models 4 and 5), show that an increase in non-fatal work injuries is observed during
economic recessions. The results on the effect of unemployment rates are similar, indicating
that during periods of rising unemployment, non-fatal work injury rates tend to increase. As
expected, increasing average hours of work and increasing labor cost is associated with
increasing non-fatal work injuries.
6. Conclusions
13
The empirical evidence on the effects of unions on occupational health and safety is
ambiguous. Some studies report a negative relationship between union activity and fatal or
non-fatal workplace injuries (Siebert and Wei, 1994, Litwin, 2000) whereas other studies find
a positive relationship (Donado, 2007; Fishback, 1986). The literature suggests that this
ambiguity may be an outcome of the endogeneity bias (Hirsch and Berger, 2001; Robinson,
1990, Fenn and Ashby, 2004).
This study contributes to the literature by providing evidence of the effects unionisation,
upon fatal and non fatal work accidents for ten European Union countries after controlling for
the effects of endogeneity. The level of unionisation is approximated by the union density
rates. The study utilises FGLS models and system-GMM models which take into account the
endogeneity on the work injuries-union density relationship and the dynamic character of
work injuries. The findings are consistent across the different specification models indicating
that union density is conducive to reducing fatal and non-fatal injuries at the workplace. For
fatal work-related injuries, there is a significant permanent effect of union density on OSH,
indicating that unionionisation has a constructive impact on occupation safety in the long-run.
For the non-fatal work-related injuries there is transitory positive effect of unionisation on
OSH. The above difference in the effects is due to the nature of fatal and non-fatal work
injuries. It is consistent with the findings of other studies which have found differences in the
effect of unionisation upon fatal and non-fatal work injuries (Donado, 2007). One might
expect that union OSH activities may act preventively or may deter more serious work
injuries that lead to fatalities. In addition, if a large part of non-fatal work injuries may be
attributed to individual predisposition characteristics (Pouliakas and Theodossiou, 2010) then
the impact of union presence upon them may be more restrictive in this case.
The augmented regression models suggest that during periods of economic upturns, as
suggested by increasing per capita GDP, fatal work injuries increase but non-fatal work
14
injuries decrease. Only a few studies address the cyclicality of workplace injuries and the
majority of them propose that injury rates are procyclical, due to the increased effort and
working hours exerted during periods of economic upturn (Boone et al., 2011; Shea, 1990).
This hypothesis is supported the findings of this study for fatal work injuries only. Other
studies attribute this positive relationship to the hiring of new workers in economic upturns,
who seem to have higher injury risks due to inexperience (Davies and Jones, 2009). The
present study does not provide evidence in favour of the existence of a reporting bias, namely
that workers are more prone in reporting work injuries during periods of economic growth
due to lower concerns of job loss (Boone and van Ours, 2006). However, other effects may
operate and affect the relationship between economic conditions and workplace injuries, such
as behavioral changes by workers (Butler, 2011). For example, during periods of rising
unemployment, workers may experience increased pressure and they might become more
prone in undertaking hazardous job tasks in order to avoid job loss. This hypothesis is
supported by the impact of unemployment exerts upon work injuries, ceteris paribus. Indeed,
increasing unemployment rates are associated with increasing fatal and non-fatal work
injuries. All in all, economic conditions can greatly affect worker behavior and riskundertaking activities at the workplace.
The control covariates hold the expected signs. Investment in human capital is associated
with lower fatal work injuries. The effect of schooling is not significant in the case of nonfatal workplace injuries. Human capital investment is associated with less risky job
environments and increased awareness of OSH behaviours and risks. Similarly, an increase in
average hours of work and an increase in unit labour cost are both associated with an increase
in fatal and non-fatal workplace injuries. Both these effects are intuitive since an increase in
effort is expected to increase workplace injuries. In addition, an increase in labour cost may
15
induce employers to cut down on OSH measures in the workplace in order to reduce labour
costs.
Overall, the results of this study indicate that unionisation is an important factor in
improving occupational health and safety since increased union membership improves OSH
conditions at the workplace through various mechanisms (efficient bargaining, OSH
monitoring and implementation, participation in OSH committees and the like). The
empirical findings have broad implications for workplace safety, since they identify the
relative importance of unions on workplace safety. This is an important policy issue
especially in the current state of affairs where there is a trend of declining union participation
and there are high economic and social costs of workplace injuries.
Acknowledgements
Thanks go to the participants of the ‘Health and Work’ Organised Session (2011) of the
Scottish Economic Society 2011 Annual Conference, Perth, Scotland for helpful comments.
The financial support of the European Commission is gratefully acknowledged
(HEALTHatWORK project) - 7th Framework Programme "THEME [HEALTH-2007-4.2-3]
Grant agreement no: 200716. The authors are grateful to the referees of this journal for
helpful comments’
References
Arellano M. and Bond S.R. (1991). “Some tests of specification for panel data: Monte Carlo
evidence and an application to employment equations”, Review of Economic Studies, 58:
277–297.
16
Arellano M. and Bover O. (1995). “Another look at the instrumental-variable estimation of
error-components models”, Journal of Econometrics, 68: 29–51.
Boal W.M. (2009). “The Effect of Unionism on Accidents in U.S. Coal Mining, 1897-1929”,
Industrial Relations, 48(1): 97-120.
Boden L.I. (2005). “Running on Empty: Families, Time and Workplace Injuries”, American
Journal of Public Health, 95(11): 1894-1897.
Boone J. and van Ours J.C. (2006). “Are recessions good for workplace safety?”, Journal of
Health Economics, 25: 1069-1093.
Boone J., van Ours J.C., Wuellrich, J.-P. and Zweimüller J. (2011). “Recessions are Bad for
Workplace Safety”, IZA Discussion Paper Series, IZA DP No. 5688.
Butler M. J. (2011). “Essays in Health and Labor Economics”, U.C. Berkeley Electronic
Theses and Dissertations, Ph.D. Thesis, pp. 119.
Catalano R. (1979). “Health Costs of Economic Expansion: The Case of Manufacturing
Accident Injuries”, American Journal of Public Health, 69: 798-794.
Davies A.R., Jones P. and Nunez I.C. (2009). “The impact of the business cycle on
occupational injuries in the UK”, Social Science and Medicine, 69: 178–182.
De Beeck R.O. and Hermans V. (2000). “Research on work-related low back disorders”,
Research Report No. 204, Institute for Occupational Safety and Health, Luxembourg:
Office for Official Publications of the European Communities.
Donado A. (2007). “Why do unionized workers have more nonfatal occupational injuries?”,
Paper presented in the 22nd Annual EALE Conference, London, UK.
European Agency for Safety and Health at Work (2001). “Preventing Accidents at work”,
Research Report, European Agency for Safety and Health at Work, Luxembourg: Office
for Official Publications of the European Communities.
17
Fenn P. and Ashby S. (2004). “Workplace Risk, Establishment Size and Union Density”,
British Journal of Industrial Relations, 42: 461-480.
Fishback V.P. (1986). “Workplace Safety during the Progressive Era: Fatal Accidents in
Bituminous Coal Mining, 1912-I 923”, Explorations in Economic History, 23: 269-298.
Freeman E.J. (2004). “Union-management solutions for preventing workplace injury of older
workers”, Work, 22: 145-151.
Greene W. (2000). Econometric Analysis. Upper Saddle River, NJ: Prentice–Hall.
Hirsch B.T. and Berger M.C. (2001). “Union Membership Determination and Industry
Characteristics”, Southern Economic Journal, 50: 665-679.
Hirsch B.T., Macpherson D.A. and DuMond J.M. (1997). “Workers’ Compensation
Recipiency in Union and Nonunion Workplaces”, Industrial and Labor Relations Review,
50: 213-36.
Leigh, J. P. (1981). “The Effects of Union Membership on Absence from Work Due to
Illness”. Journal of Labor Research, II(2): 329-336.
Lesch H. (2004). “Trade Union Density in International Comparison.” CESifo Forum, 5: 1218.
Litwin A. S. (2000). “Trade Unions and Industrial Injury in Great Britain”, Working Paper
Series No. dp0468, Centre for Economic Performance, London School of Economics, UK.
Morantz, A. D. (2013). “Coal Mine Safety: Do Unions Make a Difference?”, Industrial and
Labor Relations Review, 66: 88-116.
Morse T., Punnett L., Warren N., Dillon C. and Warren A. (2003). “The Relationship of
Unions
to
Prevalence
and
Claim
Filing
for
Work-Related
Upper-Extremity
Musculoskeletal Disorders”, American Journal of Industrial Medicine, 44: 83-93.
Mundlak Y. (1978). “On the Pooling of Time Series and Cross Section Data”, Econometrica,
46: 69-85.
18
Nichols T., Walters D. and Tasiran A.C. (2007). “Trade Unions, Institutional Mediation and
Industrial Safety: Evidence from the UK”, Journal of Industrial Relations, 49: 211–225.
Pouliakas K. and Theodossiou I. (2013). “The Economics of Health and Safety at Work: An
Interdisciplinary Review of the Theory and Policy”, Journal of Economic Surveys, 27(1):
167-208.
Pouliakas K. and Theodossiou I. (2010). “An Inquiry into the Theory, Causes and
Consequences of Monitoring Indicators of Health and Safety at Work”, IZA Discussion
Papers 4734, Institute for the Study of Labor (IZA).
Reilly B., Paci P. and Holl P. (1995). “Unions, Safety Committees and Workplace Injuries”,
British Journal of Industrial Relations, 33: 275-288.
Robinson J.C. (1990). “Worker Responses to Occupational Risk of Cancer”, The Review of
Economics and Statistics, 72: 536–541.
Roodman D. (2009). “How to do xtabond2: An introduction to difference and system GMM
in Stata”, Stata Journal, 9: 86-136.
Sasaki M. (2010). “Unemployment and workplace safety in a search and matching model”,
Discussion Papers in Economics and Business No. 10-14, Graduate School of Economics
and Osaka School of International Public Policy, Osaka University.
Shea, J. (1990). “Accidents rates, labor effort, and the business cycle”, SSRI Working Paper
No. 9028, University of Wisconsin, Madison.
Siebert W.S. and Wei X. (1994). “Compensating Wage Differentials for Workplace
Accidents: Evidence for Union and Nonunion Workers in the UK”, Journal of Risk and
Uncertainty, 9: 61-76.
Souza K., Cantley L.F., Slade M.D., Eisen E.A., Christiani D. and Cullen M.R. (2014).
“Individual-Level and Plant-Level Predictors of Acute, Traumatic Occupational Injuries
in a Manufacturing Cohort”, Occupational and Environmental Medicine, 71(7): 477-483.
19
Steele G. R. (1974). “Industrial accidents: an economic interpretation”, Applied Economics,
6: 143-155.
Ussif A. (2004). “An international analysis of workplace injuries”, Monthly Labor Review,
127: 41-51.
Veliziotis, M. (2010). “Unionization and Sickness Absence from Work in the UK”, Institute
for Social and Economic Research Working Paper Series No, 2010-15, University of
Essex.
Wooldridge J.M. (2002). Econometric Analysis of Cross Section and Panel Data.
Cambridge, MA: MIT Press.
20
Table 1. Descriptive Statistics
Country
Variables
Fatal Injury Rates per 100,000 employees
Austria
Denmark
France
Ireland
St. Dev.
Mean
St. Dev.
Mean
St. Dev.
Mean
St. Dev.
Mean
St. Dev.
4.95
1.64
2.76
0.89
2.64
0.68
4.02
1.19
3.26
1.63
1910.03
350.69
3090.66
764.81
Non-Fatal Injury Rates per 100,000
employees
Union Density (%)
Finland
Mean
43.11
6.79
75.81
2.45
74.36
4.08
10.40
2.63
49.69
9.99
GDP per Capita
20537.58
7101.47
21558.83
7240.06
18777.38
6393.45
19857.75
5661.74
18889.67
11197.02
Country UR (%)
5.99
1.02
8.47
2.26
8.82
4.13
10.13
1.17
12.16
5.76
Average Years of Schooling
8.39
0.61
9.71
0.15
8.83
0.60
8.22
1.22
10.71
0.25
1816.29
19.22
1575.75
33.48
1770.96
34.73
1597.75
69.32
1861.63
126.23
0.74
2.74
2.88
2.33
2.80
3.65
2.45
2.81
3.61
3.59
Mean
St. Dev.
Mean
St. Dev.
Mean
St. Dev.
Mean
St. Dev.
Mean
St. Dev.
5.87
1.06
6.68
2.78
8.38
2.44
0.28
0.18
1.14
0.49
3356.84
643.63
4844.57
1315.57
5026.84
526.06
1274.43
617.06
632.91
94.33
Average Hours of Work
Annual Growth Rate of Unit Labor Cost (%)
Country
Variables
Fatal Injury Rates per 100,000 employees
Non-Fatal Injury Rates per 100,000
employees
Union Density (%)
Italy
Portugal
Spain
Sweden
UK
38.35
3.83
27.15
8.78
14.06
2.87
80.63
2.11
36.70
7.18
GDP per Capita
19043.54
5973.48
12466.54
4883.09
14977.96
5767.34
20085.46
5641.46
18984.00
6639.99
Country UR (%)
10.34
1.27
6.13
1.50
17.55
4.41
4.57
2.32
8.03
2.59
Average Years of Schooling
8.08
0.72
6.87
0.52
7.84
1.32
10.54
0.61
8.40
0.39
1861.54
17.90
1885.31
78.81
1746.04
35.51
1590.21
46.70
1727.54
32.83
4.73
4.42
8.85
6.73
5.23
3.19
3.35
3.54
3.81
3.19
Average Hours of Work
Annual Growth Rate of Unit Labor Cost (%)
* Information on fatal work injuries per 100,000 employees is available for the period 1982-2005. Information on non-fatal work injuries per 100,000 employees is available for the years 1985-2006.
2
3
4
5
6
Figure 1. Changes in Mean Fatal Work Injury Rates Over Time, 1982-2005.
1980
1985
1990
1995
2000
2005
Year
2000
2500
3000
3500
Figure 2. Changes in Mean Non-Fatal Work Injury Rates Over Time, 1985-2006.
1985
1990
1995
2000
2005
Year
45
50
Figure 3. Changes in Mean Union Density Rates, 1982-2005.
40
Mean Union Density
21
1980
1985
1990
1995
2000
2005
Year
22
Table 2. The Effect of Unionisation on Fatal Work Injuries, 1982-2005
Dependent
Variable
Independent
Variables
Fatal Injuries
FGLS,
Model 1
Temporary Union Density
-0.0002 *
(-13.59)
-0.056 *
(-27.26)
(0.67)
-
Permanent Union Density
-
GDP per capita
Union Density (%)
Fatal Injuriest-1
-
FGLS,
Mundlak decomposition,
Model 2
-0.0002 *
(-8.86)
-0.032 *
(-4.39)
-0.059 *
(-26.74)
-
System GMM
Model 3
FGLS,
Model 4
0.0001
(0.14)
-0.344 *
(-1.97)
-
0.00004 *
(1.98)
-0.009 *
(-3.19)
(0.67)
-
-
-
0.222
(0.61)
Country UR (%)
-
-
-
Average Years of Schooling
-
-
-
Average Hours of Work
-
-
-
Annual Growth Rate of Unit Labor Cost (%)
-
-
-
9.977 *
(63.20)
9.728 *
(53.01)
-
-
20.296 *
(1.97)
-0.186
(-0.77)
Constant
Trend
Year Dummies
FGLS
Mundlak decomposition,
Model 5
0.00004 **
(1.83)
-0.018
(-1.26)
-0.009 *
(-2.96)
-
-
0.189 *
(11.59)
-1.207 *
(-15.11)
0.003 *
(6.63)
0.103 *
(5.93)
7.323 *
(5.66)
0.191 *
(10.92)
-1.202 *
(-15.21)
0.003 *
(6.59)
0.103 *
(5.76)
7.184 *
(5.45)
-
-
Yes
Yes
No
Yes
Yes
10609.63
(0.000)
10988.83
(0.000)
60.47
(0.00)
23580.15
(0.000)
21.813.16
(0.000)
AR(1)
-
-
0.927
-
-
AR(2)
-
-
0.900
-
-
Sargan test
-
-
0.416
-
-
240
240
230
240
240
Wald chi2
Observations
* indicates statistical significance for p<0.05 and ** indicates significance for p<0.10. Robust standard errors are calculated. T-statistics are reported in parentheses.
Probabilities are reported for AR(1), AR(2) and Sargan test.
23
Table 3. The Effect of Unionisation on Non-Fatal Work Injuries, 1985-2006
Dependent
Variable
Independent
Variables
Non-Fatal Injuries
FGLS,
Model 1
Temporary Union Density
0.0000 *
(3.36)
-35.441 *
(-27.07)
(0.67)
-
Permanent Union Density
-
GDP per capita
Union Density (%)
Fatal Injuriest-1
-
FGLS,
Mundlak decomposition,
Model 2
0.0000 *
(6.00)
11.025
(0.87)
-36.621 *
(-23.60)
-
System GMM
Model 3
FGLS,
Model 4
-0.047 *
(-2.63)
-0.030 **
(-1.63)
-
-0.0000 *
(-3.63)
-9.849 **
(-1.77)
(0.67)
-
-
-
-0.365
(-0.55)
Country UR (%)
-
-
-
Average Years of Schooling
-
-
-
Average Hours of Work
-
-
-
Annual Growth Rate of Unit Labor Cost (%)
-
-
-
5222.050 *
(62.79)
5084.564 *
(53.62)
-
-
2.314 **
(1.86)
-0.008
(-1.61)
Constant
Trend
Year Dummies
FGLS
Mundlak decomposition,
Model 5
-0.00001 *
(-3.77)
-56.849 *
(-2.78)
-2.152
(-0.40)
-
-
48.874 *
(2.62)
35.681
(0.34)
6.930 *
(7.18)
67.431 *
(2.96)
-9590.587 *
(-4.48)
70.880 *
(3.60)
140.926
(1.25)
8.976 *
(8.76)
89.842 *
(3.65)
-14383.89 *
(-6.24)
-
-
Yes
Yes
No
Yes
Yes
5715.53
(0.000)
16330.42
(0.000)
27.30
(0.00)
2580.32
(0.000)
1797.39
(0.000)
AR(1)
-
-
0.435
-
-
AR(2)
-
-
0.291
-
-
Sargan test
-
-
0.115
-
-
154
154
147
154
154
Wald chi2
Observations
* indicates statistical significance for p<0.05 and ** indicates significance for p<0.10. Robust standard errors are calculated. T-statistics are reported in parentheses.
Probabilities are reported for AR(1), AR(2) and Sargan test.