1Estimation of wage differential for the Czech Republic: Hedonic wage [615619]

1Estimation of wage differential for the Czech Republic: Hedonic wage
model testing on three datasets

Milan Š časný1, Jan Urban
Charles University of Prague, Environment Centre

Abstract

The aim of the paper is to test hedonic wa ge model on data from the Czech labour
market in order to obtain the Value of a Statis tical Life from wage differential. In order to
support an empirical basis fo r using the willingness-to-accept concept as a reliable
method, we also examine the statistical relationship betw een changes in occupational
rate of fatal risks and yearly average wages fo r economic sectors. We confirm statistically
significant effect of objective fatal risk rate on employee’s wage for the Czech Republic.
Using individual data from the national representative survey conducted in October 2006,
we estimate the VSL about 6 million €, while we display higher VSL for the employees
exposed to higher occupational risks, partic ularly for exposed males. We do not confirm
significant effect of either objective fatal ri sks or subjective perception of occupational
risk on wage for individual data coming from the 2000 survey. Wage differential is
obtained from statistical averages for economic sectors during 2003-2005 that also gives
us empirical basis for using the willingness- to-pay concept as a reliable method for
valuing a life. The VSL derived from macro da ta, i.e. statistical averages, then ranges
between 3.2 to 3.6 million €. The VSL’s derived from hedonic wage models are pretty much comparable with the VSL obtained fr om a CV study on willingness to accept
compensation paid through higher wage for in creased risk rate by 50% as well as they
are in line with empirical evidence provided by the literature. The VSL’s derived from the
Czech labor market are, however, proved to be larger than the VSL’s obtained outside of
labor market that are about 0.2 to 1.0 million €.

K
EYWORDS : hedonic wage model; health risks; fata lity; value of a statistical life; Czech
Republic

JEL CLASSIFICATION : J17, J28, J31

1 Corresponding author: Milan Š časný, Charles University Environment Center, U K říže 8, 158 00, Prague 5,
Czech Republic, e-mail: [anonimizat] .

2Introduction

Policy intervention involves – directly or indirectly – many health impacts. Except the effects on morbidity and human impairment , there are the effects on mortality and
related risks that need to be considered in cost-benefit analysis. Welfare measures of
such changes might therefore emerge for relevant economic analysis.

Despite the very specific charac ter of the product, i.e. a premature death, an attempt to
put monetary value on mortality has a long tradition in public policy analysis. For
instance, Landefeld et Seskin (1982) trace this idea more than 300 years back in impact
assessment of public programmes combating ai rborne pollution in Great Britain. More
recently, the value of statistical life (hereinafter VSL) is used in current policy practise as
in the US as in EU. US EPA uses as a base VSL of $6.3 million (1999 dollars) in its policy
recommendations on groundwater regulations (US EPA, 2000), The Department of Transportation and other US governmental ag encies use similar estimates in evaluating
regulatory effects (Adler and Posner, 2000; cit. in Jennings and Kinderman, 2003). The
European Commission, in its CBA guidelines, e.g. for CAFE Programme (EC, 2006) or for
the external costs quantifica tion by the ExternE method (EC, 2005), recommends to use
a unique value of a statistical life as hi gh as 1 million Euro, with 50% premium for
cancer.

Such value of mortality can be derived by applying three possible methods. The oldest
approach uses observed economic values an d bases on a macroeconomic vision of the
role of the individual as an agent contributing to the activity of the system (OECD, 2002).
The value of preventing a fatality, at a given time, is equal to the future productive loss
evaluated as the discounted sum of the earnings that the individual would have otherwise
earned. This is known as the loss of prod uction method or more familiar as ‘human
capital approach’. Altern atively, one can consider that an y individual is a consumer and
premature death results in the loss of these consumption possibilities. Then, households’
final consumption is used to value life. Both of these methods display many
disadvantages as summarized in e.g. OECD (2002); by considering market production,
the value of life is zero for people outside the labour market such as pensioners and disabled people, due to discounting, the value of child’s life is sma ller then of economic
active person. Moreover, the values vary with age and are very sensitive on magnitude of
discount rate being chosen. The biggest problem, however, lies in inconsistency with the
fundamental principles of welfare econom ics by not taking into account agents’
preferences
2.

The second approach to derive monetary value of life – mostly used in medical practise –
might be based on medical treatment costs pe r Quality Adjusted Life Year (QALY). The
QALY is kind of quality of life measure and presents a weight expressed over the range (0,1), with 1 indicating ‘full he alth’ or health in normal expe cted quality, and 0 indicating
a state in that a person is indifferent to de ath and life (the negatives can even indicate a
preference to die rather than to live). The weights are mostly stated by medical experts
and physicians. The costs of relevant medica l treatment or intervention are identified.
Then, the costs per QALY are used in costs-effe ctiveness analysis or even as an arbitrary
threshold of cost-effectiveness set out by soci al planner. For instance, the costs per QALY
of 20,000£ is identified as a threshold (Tow se et Pritchard, 2002; Devlin et Parkin,
2003), while 30,000£ needs a special reasoning (Rawlins et Culyer 2004) for NICE in the
UK. The costs of 60,000€ was found as a thresh old for Sweden in the studies by Kolbet et
al. (2004). Similarly, as the first approach, the QALY approach is not consistent with the
fundaments of welfare economics.

2 By only considering the productive aspect of the individual, this method used to underestimate the value of
life compared with estimates derived from WTP approaches (Le Net, 1994).

3Only the methods that are consistent with welfare economics focu s on welfare change
due to change in undesirable health outcome. These methods focus specifically on trade-
offs between risk and money (or income). Then, in policy assessment, the Value of
Statistical Life is a reference point for any mortality benefits.

The VSL can be experimentally derived using several methods. The researcher might use
evidence on market choices that involve implicit tradeoffs between risk and money.
Particularly, one can observe the compensating wage differentials that workers must be
paid to take riskier jobs (see Viscusi and Al dy, 2003, for a recent literature review). In
order to estimate the VSL, one can even ex tend the hedonic wage model by relating
housing prices and wages to climate (Moore, 1998; Maddison and Bigano, 2003) or air
quality (Portney, 1981). Most of the empirica l estimates of VSL have relied just on the
hedonic wage approach.
Next group of approaches – using aver ting behaviour method – might examine
behaviours where people weigh costs against risks (Blomquist, 2004; review of US studies in Viscusi and Aldy, 2003). Costs for safer car equipment such as seat-belts,
child-seat or airbags or expenditures on sport- or mo torbike-helmets or smoke-detectors
might be the examples.

The VSL can also be obtained from stated preference through contingent valuation or
contingent choice experiments. Based on original CV questionnaire developed by
Krupnick et al. (2002), the VSL is obtained from willingness-to-pay for risk reduction of dying in wide group of countries; the VSL is estimated in such way for the US and
Canada (Krupnick et al., 2002), for Italy, France and UK (in NewExt project; see Alberini
et al., 2005), while Krupnick et al. (2006) compare these results with VSL derived for
Japan. Alberini et al. (2005) and Alberini et al. (2006) obtained the VSL from WTP for
avoiding risk of dying from respiratory and cardiovascular diseases for Italy and for the
Czech Republic, while Giergiczny (2006) tests this approach in Poland. Alternatively, the
Value of Life Year Lost was obtained from WTP for prolongation of life expectancy by
DEFRA study (Chilton et al. 2004), and more recently the values for 8 European countries
are provided by NEEDS project (Desaigues et al. 2006).
Using stated preference approach in choice experiments, the VSL was derived by e.g.
Tsuge et al. (2005) who experiment with various characteristics such as risk type
(cancer, heart attack, accident) and latency or Itaoke et al (2006) who treat labelling
effect for mortality risk reduction from electric power sector.

Most recent and comprehensive review of the VSL estimates provides Kochi, Hubbell, and
Kamer (2006). They compare data coming from 40 selected studies published between
1974 and 2002, containing overall 197 VSL estimates (although there are increasing numbers of CV or HW studies in countries with lower income such as Taiwan, Korea,
India or CEEC countries, they exclude these from their analysis). Their estimate of
composite distribution of empirical Bayes adju sted VSL yields a mean of $5.4 million and
a standard deviation of $2.4 million.

Despite the real policy demand for having su ch values, according to our best knowledge,
there are only few empirical studies that estimate the VSL or VOLY performed in
transition countries in Central and Easter n Europe. Regarding the implementation of
hedonic wage model, there is only one such study conducted in Poland (Giergiczny,
2006) who found a sample mean of VSL as hi gh as 2.26 million €, with a 95 percent CI
from 2.05 to 2.44 million € (2005).

Our aim is fulfil partially this gap by testin g hedonic wage model for the Czech Republic
during huge restructuralisation of its economy and labour market included. The aim of
our paper is then twofold: (1) to obtain th e VSL from wage differential by applying
hedonic wage model on the Czech employees, (2) to examine the statistical relationship

4between changes in occupational mortalit y rate and in yearly average wages for
economic sectors in order to support an empi rical basis for using the willingness-to-pay
concept as a reliable method for valuing a lif e from individual data, and (3) compare our
estimates with VSL values obtained by using other methods, particularly CV method in
the Czech Republic and at abroad.

II. Valuation of occupational risks

Empirical evidence

The idea that higher risk of occupational mortality may result in higher wage payment to
worker is quite plausible. Adam Smith in hi s well-known book ‘The Wealth of Nations’
(1776; Chapter X, part I) has already noted that “ The wages of labour vary with the ease
or hardship, the cleanliness or dirtiness, the honourable or dishonourableness of the
employment… A journeyman blacksmith, though an artificer, seldom earns so much in
twelve hours as a [labourer] does in eight. His work is not quite so dirty, is less
dangerous…” . The economists therefore have been focusing their effort to find an
evidence for such measurable impact of oc cupational mortality on wages in order to
derive compensating wage differential. Such wage differential has been the most often
used approach to reveal trade-off between mone y and fatality risk in order to obtain so
called value of statistical life.
This is documented in the great review by Viscusi and Aldy (2003) who found more than
50 labour market studies that bring the value of statistical life. Most of these studies are
dominated geographically by US labour market; 30 out of 51 studies. Only six hedonic
wage studies were conducted in developing Asian countries (Hong Kong, India, Taiwan),
and six in Europe, however five of them were conducted in the UK (last one in Austria).
So far, there is no such study conducted in CEEC region except a pioneering Giergiczny’s
2006 study. The VSL ranges between $0.5 to $21 million (2000 dollars) in US, $4 to $74
million in the UK, or $0.2 to $4.1 million in Asia (excluding Japan). Central estimate of
the VSL value provided from meta-analysis by Mrozeck-Taylor (2002) yields $1.6 to $2.7
million, by CSERGE (1999) as high as 6.5 million € and Viscusi and Aldy (2003) provide
mean VSL of 5 million €.

Viscusi and Aldy also documented overall 39 studies of the value of statistical (nonfatal)
injury. Again, these studies are mostly (31) coming from US labour market. They identify
only one being conducted in the UK. Implicit value of a statistical injury used to range
about tens of thousands US 2000 dollars.

Most of hedonic wage studies estimate wage differential on individual data of workers.
There is also a group of empirical studies th at examine relationship between statistical
rate of occupational injuries and wage for in dustries or branches. For instance, Jennings
and Kinderman (2003) use data from the Bureau of Labour Statistics for the period 1992
to 1999 on industry injury and illness rates an d fatality rates to examine the statistical
relationship between changes in occupational mortality rate and in hourly wages in the
USA. Their analysis did not support any statistically significant evidence, and thus
conclude that there is no empirical basis for using the willingness-to-pay concept as a
reliable method for valuing a life.
Econometric model

Econometric estimati on of wage compensating differential from hedonic wage function is
well documented exercise. The wage-risk rela tionship in labour markets is mostly
estimated from following equation:

i i i i i i i X COMP RISK RISK JOB WORKER w εβ β β β ββ ++⋅ + + + +=5 4 3 2 1 0 (1)

5
while WORKER is a vector of personal characteri stic variables including human capital
measures such as education, experience and skills for worker i, JOB is a vector of job
characteristic variables for worker i, RISK might be a vector of variables describing risks
of fatal and non-fatal injuries and occupational illnesses, COMP describes compensations
(pecuniary or in-kind) provided to worker i, and X might include other variables including
interactions of fatality risk and personal char acteristics (gender, age, trade-union status)
to capture heterogeneity in ri sk perception and aversion. εi is the random error capturing
unmeasured factors affecting worker’s wage rate.

The model and estimation of hedonic price eq uation is well documented regular event.
Most hedonic wage studies have estimated the wage equation using linear and semi-
logarithmic specifications. Although, as argued by Rosen (1974), choosing a preferred functional form from these specifications cann ot be determined on theoretic grounds, one
can employ a flexible functional form given by the Box-Cox transformation to identify the
specification with greatest explanator y power (Moore and Viscusi, 1988a).

A more general functional form of hedonic price function proposed by Halvorsen and
Polakowski (1981) follows:

i ig icC
cC
cC
gcg ic c i z z z p εγγα φβλ +⋅⋅ +⋅=∑∑ ∑
== =)( )( 5.0)( )(
11 1

where α’s, and β’s are linear parameters on th e transformed variables, and the
transformation of any variable x is the typical form for Box-Cox models:

= λλ1−x
for λ ≠ 0
x(λ)
= l n ( x ) f o r λ = 0.

Then, the marginal effect of a change in characteristic c for individual i is given by

∑
=−− −+⋅=∂∂C
gigcg ic icc i
iciz z z pzp
11 1 1))( ( γα βγφ λ.

The hedonic wage function as in eq. 1 used to be only estimated by semi-logarithmic
specifications, or with the dependent variable, i.e. wage wi, transformed by the best
lambda, i.e. φ=1 and γ=1.

Then, the marginal effect of fatal risks obtained from the hedonic wage function (1) is
given as

31βλ⋅=∂∂ −
i
iiwRISKw
or
) 2 (3 31
q iw ββλ+⋅=− for a quadratic form of fatal risks.

The Value of Statistical Life is then derived as

6

RwVSLq) 2 (3 31ββλ+⋅=−

where β’s are coefficients estimated for fatal risk variable(s), w is average wage and R is
risk rate, e.g. 1/1,000 or 1/10,000.

Risk data

There is a huge literature on choice of the risk variable and its measurement. Variable
reflecting both worker’s and firm’s (subjective) perception of the risk would be an ideal
measure of on-the-job fatality and injury risk (Viscusi and Aldy, 2 003). In the literature,
measures of risk used to include self-r eported risks based on worker surveys and
objective risks derived from actuarial tables, compensation records, surveys and censuses of death certificates (then compiled by official statistical bodies).

Particularly early papers – as note by Viscusi and Aldy (2003) – included several
qualitative measures of on-the-job risk. Hamermesh (1978), Viscusi (1979, 1980), or
Fairris (1989) estimated the hedonic wage equation with a dichotomous measure of
injury risk based on a worker’s perception of whether his or her job is „dangerous“ by
asking if their job exposed them to physical dangers or unhealthy conditions. Duncan and
Holmlund (1983) use “danger” variable in their hedonic study of male workers in
Sweden. Similarly to this approach, we also control the “danger” by variables indicating
subjective perception of physic al and psychological risks and personal aggression at work
in our hedonic wage models.

Empirical studies use, however, dominantly the measures of objective risk either
occupation-specific (e.g. Thaler and Rosen, 1975; Brown, 1980; or Polish study by
Giergiczny, 2006), or – most recent ones – in dustry-specific ones. An average of at least
several years of observations is used for fatalities which tend to be relatively rare events.

The magnitude of risk premium, and consequently of the VSL, estimated by regression strongly depends on the choice of the risk measure. Moreover, it is just accuracy in
perception of differences in occupational ri sks between professions and branches rather
than absolute magnitude of risks that plays a role for validity of hedonic wage model.
Therefore the choice of risk measure used in econometric model and its perception by
worker requires a special attention of modeller.

How the objective risks are perceived and whether they are considered in real decision
making by worker has remained crucial question for researcher that needs to be
discussed. For instance, Viscusi (1979) co nfirms a correlation between subjective
perception of the risks – indicated by the an swer on question “Is your job dangerous?” –
and objective (statistical) accident data. Viscusi et O’Connor (1984) or Gerking et al.
(1988) report that accident rate data used to be perceived more than they are (about
50% in chemical industry). Slovic et al. (1979) found that people overestimate the likelihood of infrequent causes of death such as death due to botulism, floods or
tornadoes and on the other hand underestima te the probability of death with higher
frequencies, e.g. due to heart attacks and cancer (cit. in Freeman, 2003, p. 404).
McDaniels, Kamlet, and Fischer (1992) examined the relationship between perceived risk
and WTP for increased safety from ten tech nological hazards, five of which are well-
known and five are less well-known. They found that the WTP for well-defined risks is
most influenced by perceived personal exposure, while WTP for less well-known risks is
most influenced by levels of dread and severity.

7Very important note was remarked by Fischhoff et al. (1981) that one must distinguish
between an individual’s perception of the rela tive perception frequency of death in some
population and the individual’s estimate of hi s or her own risk of death. As confirmed by
them or by Hamermesh (1985), the later is often underestimated. We also do confirm
this evidence has been found in our two focus groups conducted with blue-collar workers
in the Czech Republic. It was indeed personal experience with injury at work and skills
that result in underestimation (or perception) of objective risks given by statistical
accident rate data.

III. Data and estimation results
Working conditions have been significantly improving since 1990 in the Czech Republic.
While the official statistics of SUIP (State Labour Inspection Office, Prague) recorded
almost 300 cases of fatal injuries and ab out 100,000 cases of non-fatal occupational
injuries yearly in the middle nineties (fro m about 4.7 million of em ployees), there have
been only 137, respectively less than 80 ,000 cases of fatal and non-fatal injuries
recorded recently.

In relative terms, while the statistics recorded 0.6 cases of fatal and almost 230 non-fatal
injuries per 10,000 employees in the middle ni neties, the relative risks have declined at
0.3 of fatal or 180 non-fatal injuries per 10,000 in the year 2005 (SUIP).

We can document several trends: firstly, a re gulatory system of occupational safety has
been enforced with larger stringency, secondly, the Czech economy has been strongly restructuring what’s resulted in higher shar e of services and firms being orientated on
products with higher value added; thirdl y, growing unemployment has lowered the
number of employees what li kely excludes less skilled and more “troublesome” workers
from the labour market. Each of these fact ors might lower the absolute and relative
occupational risks.

We follow an econometric model with transf ormed net wage by Box-Cox and report the
estimation results for best lambda. Econometric model is estimated in maximum
likelihood by SAS programme.
We use three datasets fo r hedonic wage model tests in the Czech Republic:

1. Individual data of Czech employees from a survey “Quality of Occupational Life –
2006” ;

2. Yearly averages for economic sectors provided by the Czech Statistical Office for
the period 2003 to 2005;

3. Individual data of the survey “Working Conditions – 2000” ;

III.1 Individual data from “Quality of Occupational Life – 2006” survey

This data come from the same dataset as the data we used to obtain the VSL from
willingness-to-accept a compensation, i. e. higher real wage (Urban and Š časný;
submitted to EAERE-2007). The survey was conducted jointly by Sociological Institute of
Academy of Sciences – Public Opinion Resear ch Centre, Occupational Safety Research
Institute and Charles University Environment Center in October 2006. The survey and
sampling strategy are described in more de tail in our other paper. There are 2,043
observations in full sample (see Appendix for the descriptive statistics).
We use statistical data by SUIP on objective risks, i.e. reported number of fatal and non-
fatal injuries per 10,000 employees. After identifying the respondent’s occupation (nine

8profession categories used by statistical bodies was used) and branch (17 industries
followed NACE classification), we tell the respondent the risk of fatal and non-fatal
injuries he or she used to be exposed to.
Besides socio-demographics, the respondent reported his/her net monthly wage. Mean
net monthly wage or salary is then about 15,380 CZK or about 520 €.

We use three dichotomous variab les on detecting whether i) she/he is in contact with
machineries while working, ii) she/he travels by car during business travel (as a driver or
a passenger), or iii) she/he used to be in contact with persons who can physically attack
her/him, i.e. situations dominating causes of fatal injuries reported in the official Czech
statistics. We then use these variables as a filter indicating the respondent exposed to
risks. Only 1,373 (67%) of the respondents pass the filter.
Average net monthly wage is larger for the sample of workers exposed to risks
(16,400CZK, or 550 €), net wage gets even higher if we create a sub-sample consisting only of the males exposed to risks (17,900 CZK, or about 600€). Descriptive statistics for
both sub-samples are reported in Appendix.

We regress net wage on fatal and non-fatal oc cupational risks for the full sample and two
sub-samples consisting only of those expo sed to risks, either all or only male.

Estimation results
In all models, non-fatal risks are not significant, while fatal risk and fatal risk square are
significant at almost 99.9% level for full sa mple, and at the 98% level for sub-samples of
the exposed respondents. All coefficients of covariates have right signs and are
significant at 99% level. Net wage is high er if the respondent is managing people
(BOSS), does business trips by car (CARTRAVELLING), is male, has university degree or
A-level, has more children or brings higher share of money to family (BREADWIN).
Variables described by KZAM denote type of profession (following none categories of
Classification of Occupations) and by OKEC denote classes of econ omic branches (NACE
digit-1 level).

The VSL is then calculated from net monthly average wage as high as 5.9 million € (full
sample), or 6.6 and 8.9 million € for the sample of exposed to risks and for the
exposed males. Semi-logarithmic specifications with λ=0 yield similar estimates of the
VSL.

III.2 Yearly averages for industries for 2003-2005

The aim of this analysis is to support our previous estimates by examining the
relationship between branch-specific occupati onal mortality rates and other measures of
“dangerousness” of the job on one side and an average yearly wage for a relevant branch
of the Czech economy. We follow similar lo gic as Jennings and Kinderman (2003) who
examine the relationship between the change s in occupational mortality rates and in
hourly wages in order to provide an empirical basis for reliability of using WTP/WTA
concept for valuing a life.
We assume that statistically significant evid ence for relationship between risks and wages
being found in individual employee’s behaviour could be, in average, also found for the
economic sectors.

Performing this test, we gather statistical data compiled by the Czech Statistical Office for
following measures:

9ƒ fatal injuries, non-fatal injuries with wo rking-disability longer than 3 days, non-
fatal injuries with any working-disability, new cases of job-related illness, days of
sickness due to job-related illnesses,
ƒ financial compensations provided by the firm to employees and expenditures
averting the occupational risks expended by the firm,
ƒ economic variables such as the added va lue, paid wages and salaries and the
number of employees.

We create variables describing the relative risks and average wage by dividing all
risks and paid wages and salaries. Using information on tax regime, we obtain an
average of the net yearly wage per employee in a given sector and year.

All financial data are recalculated in 2005 price level by CPI. We control the effect of
sectors/branches by composing sector dummies . The effect of productivity and
technological change is controlled by sector-specific labour productivity variables and
Year variable. Descriptive statistics see in Appendix.

We collect yearly data for the period 2003 to 2005 for sectors according to the NACE 2-
digit level having in total 3*57 observations.

We then experiment with several sub-sample s: i) only sectors with positive fatal
occupational accidents (N=86), and ii) sectors with positive fatal occupational accidents
without services, i.e. NACE 65+ is dropped out (N=68).
Estimation results

For macro data, i.e. statistical averages fo r sectors, we conclude that statistical
significance of fatal risk rates on wage at 90% level for all sectors. If we consider only
those sectors with positive fatal risk rates, its significance rises at 95% level. We report
stronger statistical relationship for net yearly wage indicating the workers who
considered real wages in their choices at labour market rather than gross wages and
salaries.
The VSL derived from wage compensating differential for all sectors of the Czech
economy amounts 3.6 million € , however, statistically signif icant estimate is found only
for net yearly wages (90% level). If only th e economic sectors with positive fatal rates
were considered, the VSL obtained from th e gross yearly wage gets higher, almost 4
million € , while the VSL derived from the net wage is as high as 3.2 million € . We
estimate then the highest VS L from the data after dropping out services (NACE 65+).
The VSL from the gross wage is 4.5 million € , and from the net wage amounts about
3.6 million € .

We do not find an evidence for quadratic re lationship of fatal risks as we report for
“Quality of Occupational Life” 2006 data. Moreover only some of the models confirm
statistical significance of other non-fatal risks. We find that the non-fatal injuries without
any following working-disabilities increase em ployee’s wage, while the non-fatal injuries
followed by working-disabilities at least fo r 3 days contribute negatively to wage. The
effect of „job-related illness” is not proved to be statistical significant, similarly as
compensation paid by firm to employ ees for suffering due to an injury.
In line with one’s intuition, labour productivi ty is the strongest predictor of the wage in
the sector with positive and si gnificant coefficient. Trend vari able is significant only in
some of our models with intuitively right sign (+). Some of our models support our
hypothesis that the sectors with higher wages and salaries likely invest more sources for
prevention of occupational ri sks (kcprevent). Sector dummies increase robustness of our
models.

10III.3 Working Conditions – 2000 survey

A survey on “working conditions” was cond ucted by STEM/MARK in the Czech Republic
within the European Survey on Working Conditions regularly conducted in 1991, 1996,
and 2000.

The questionnaire contains the questions on physical factors affecting the quality of
working conditions, working time, work orga nisation, social workplace environment and
other measures describing subjective percepti on and objective factors of the workplace.
All questions are related to a full-time job of the respondent ignoring working conditions
of a part-time job.

The questionnaire also includes question whether respondent thinks that his/her job
brings him/her a danger for safety and health (Q34). If positively answered, i.e. he/she
is exposed to occupational risks, then he/she is asked on 23 kinds of effects that might
be related with this exposure. Using this info rmation we create two variables on exposure
to physical factors and “psychological” risks. Dichotomous variable on occupational risk
due to physical factors ( risk_fyz ) considers respondent’s suffering from health problems
related with hearing, eyes, skin, respiration or cardiovascular system, or his/her pain in
back, head, stomach, muscles, legs or hands, or allergies caused by working. Psychic
causes such as stresses, tiredness, sleep-de ficit, anxiety, being on edge or having
trauma from the job crea tes dichotomous variable risk_psy .

Awareness of physical aggression stemming fr om other people or hi s/her colleagues in
the job creates our third variable risk_aggres .
On the top, we use statistical data on fatal an d non-fatal risk rates for each profession (9
categories) and for each of the five economic sectors aggregated in agriculture, industry,
construction, transport and services.
We use other variables describing exposure to variety of physical factors (7-level Lickert
scale) such as vibrations from handy-equipment and machineries, noise, high
temperature, low temperature, breathing exha lation, vapours from toxic substances and
dust, manipulation with dangerous produc ts or radiation from roentgen or so.

Measures of subjective perception of occupational risks (risk_fyz, risk_psy, or
risk_aggres) as measure of objective fatality ri sk rate and measures of exposure are not
correlated (Pearson coeff. is the highes t for risk_fyz and risk_psy as 0.54).

We use reported average net monthly wage/s alary from the full-time job and normalise
them on full-time, i.e. 42 hours per week. We may only create dichotomous variables for
wage compensation for risk exposure (pripl atek) and extra payments for working during
weekend (vikend) and night (noc). We control the effect of experience (praxe) and if the respondent is a boss (vedouci). We use five binary variables for economic sectors and 9
for profession (KZAM following standardised Classification of Occu pations). Descriptive
statistics is displayed in Appendix.

In our econometric model, we use only da ta for such respondents who are employees
dropping out the respondent running his/he r own business (assuming different behaviour
and thus referring the modeller rather to the averting behaviour model). We also
dropped out those respondents with more than one job. In order to make our sample
more homogenous we also dropped out empl oyees working with less than 40 hours and
more than 70 per week.

Estimation results

11
Firstly, we examine the relationship between subjective perception of risks (risk_fyz,
risk_psy, and risk_aggres) and reported wage. Our strong model that explains net wage
(adjusted R2=0.68), however, does not contain any va riable on subjective perception of
risks. Wage is thus left to be explained by type of profession, econ omic sector and years
of experience. Hedonic wage model that cons ists of our three variables is very weak
(adjusted R2<0.1) and subjective risk perception ex plains the wage variance by 1 to 3%.
Only the effect of risk_fyz is weakly significant, while the effect of the other two is not
statistical significant at all. Moreover, in some cases, our hedonic wage models yield
wrong signs. This could be due to the fact that the respondents do not consider these
factors in their choice or due to the unobser ved heterogeneity labour in productivity we
have not been able to reveal so far (see e.g. Hwang et al. 1992; or Dorsey 1983 a
Dickens 1984 who report wrong si gns for risk coefficients).
We also do not confirm positive relationship of an interaction of perceived risk and paid
compensation with wage; p-value is 0.16 fo r the relationship. Using these data, we do
not confirm any statistical significant effect of fatal rate on wage; p-value of a square of
fatal rate is about 0.18.

IV Conclusions
We confirm statistical significant effect of objective fatal risk rate on employee’s wage.
Based on estimation of hedonic wage function we derive the wage differential from that
the VSL from the Czech labour market was obtained. The VSL is estimated about 6
million € and the VSL is higher for those employees that are exposed to higher occupational risks, particularly for exposed males.
The VSL obtained from statisti cal averages for economic sect ors ranges about 3.2 to 3.6
million €, if gross wages and salaries were used, the VSL would be about 4 million €. Proved statistical significance of risk rates on wages gives us empirical basis for using the
willingness-to-pay concept as a reliable meth od for valuing a life from individual data.

More attention is required to analyse the hedonic wage func tion; particularly a modeller
may pay attention for the role of subjecti ve perception of o ccupational risks and
perception of objective (statist ical) risk rate. Measures indica ting subjective perception of
occupational risks, either the variables describing fatal risk rate did not explain wage in
data coming from the 2000 survey.

Our results – we get from hedonic wage mode ls – are in line with estimates derived by
other studies. Older review by Viscusi (1992) brough the range of VSL between 0.8 to
17.7 million $, more recent estimates of VSL reported in the literature range between 0.2
million $ (Loomis and du Vair, 1993) to 87.6 million $ (Arabsheibani and Marin, 2000). A
comprehensive review of hedonic wage stud ies by Viscusi and Aldy (2003) show the
range between 0.5 to 21 million $ in the US, 4 to 74 million % in the UK, or 0.2 to 4.1
million of 2000$ in Asia. On the top, Kochi et al. (2006) display a mean of the composite
distribution of empirical Bayes adjusted VSL as high as 5.4 million $ and a standard
deviation of 5.4 million $ (bas ed on 197 VSL estimates).

The VSL’s derived from hedonic wage models are pretty much comparable with the VSL
being just obtained from our contingent va luation study on willingness to accept a
compensation paid through higher wage fo r increased risk rate by 50%. Urban and
Ščasný (2006) found a mean VSL derived from WTA as high as 10.7 million €, while
median is 8.4 million €.

The VSL’s derived from the Czech labor market are proved to be much more higher than
the VSL’s obtained outside of labor market.

12
For instance, the value of preventing fatality calculated for the Czech Republic by human
capital method (Š časný, 2005) and considering average macroeconomic labour
productivity in 2004 is as high as 0.4 to 0.5 milllion € for 40 years old man (d.r.=4%, or
d.r.=3% respectively).

Máca (2005) reviewed the costs per QALY for CEEC countries and brings the range
between 370 € to 16,000 €. If we considered 2,900 € per QALY for acute myocardial
infarction as found by Máca for the Czech Re public, we get VSL as high as 0.2 million €
for average life expectancy; the VSL obtained by such a way for QALY for Statins
following pericutaneous coronary intervention (F luvastatin) in Hungary gets 1.2 million €.

The VSL derived by CV method from WTP for mortality risk reduction from cardiovascular
and respiratory diseases in th e Czech Republic is 1.3 million € (mean), or 0.58 million €
(median) (Alberini et al., 2006). The VSL can be obtained also for Poland from WTP for mortality risk reduction by 1 in 10,000 (Giergiczny, 2006); the VSL is 0.77 million € (mean), or 0.44 million € (median), however, Giergiczny’s study did not pass an external
scope test and no VSL value was orig inally reported in the paper.

Median VOLY derived from WTP from life expe ctancy prolongation by 3 months estimated
by the team led by Desaigues (2006) amounts about 8,000 € for NMS pooled data, or
almost 10,000 for the Czech Republic, 8, 000 € for Poland and 3,000 € for Hungary
(country samples are, however, too small – about 150 each – yielding hardly statistical significant estimates). Very rough estimate can be then derived for average life
expectancy of 75 years by neglecting fact or of time; this value would be about 0.75
million € for the Czech republic, or about 0.22 million € for Poland.

We can conclude that this comparison displa ys the VSL about 0.2 to 1.0 million €, while
the VSL derived from labour market is one order of magnitude higher, i.e. about 3 to 9
million €.

This is contradictory to the empirical findin gs given by economic literature that shows
significantly larger estimates generated by the hedonic method than by the CV approach
(Kochi et al., 2006). The fact that two valu ation methods do not necessarily provide the
same outcome is supported on theoretical ground: while the hedonic wage approach is
estimating a local trade-off, the CV approach approximates a movement along a constant
expected utility locus (Viscusi and Evans 1990). In the other words, marginal utility of
changing risks from its optimal level (analysed by hedonic model) can be expected to be
the highest because marginal utility declines with marginal risk ‘located’ more far from
the optimal risk, i.e. probably described in the contingent (hypothetical) scenario.
One caveat should be point out: values base d on willingness to accept approach used to
yield higher values than those derived from willingness to pay. Hanemann (1991) for
instance argues that the differences betwee n the compensating surplus, i.e. minimum
WTA to consent higher occupational risks in our case and the equivalent surplus, i.e.
maximum WTP to prevent increase of the risk s, need not be insignificant as counter-
argued by Randall and Stoll (1980). Empirical evidence suggests that the minimum WTA
can exceed the maximum WTP several times over. Carson (1991) argues that when
individuals are asked to state their minimum WTA, they tend to state their expectation of
the maximum they could hope to extract as compensation rather than their true
minimum WTA (cited in Markandya et al., 2002; p. 425).

Therefore, while the hedonic wage studies may be subject to bias resulting from
measurement errors (Black 2001), and omitted variables (Hwang et al. 1992; Gunderson
and Hyatt 2001), CV studies may suffer from hypothetical bias. Better understandings of
the role of subjective percep tion of occupational risks in valuation can improve the
models tested in this paper. This task need s to be however left for our next research.

13Acknowledgment

This research has been supported by the R&D Project MPSV 1J 039/05-DP1 “The effect of changes in labour market on quality of life” funded by the Czech Ministry of Labour and
Social Affairs“ within the programme “Mod ern society and its transformation”. The
support is gratefully acknowledged. We also grateful for useful comments to Lenka
Svobodová from Occupational Safety Rese arch Institute in Prague and Zdenka
Mansfeldova and Jiri Vinopal from Sociological Institute of Academy of Sciences – Public
Opinion Research Centre with whome we colla borate on the project. Responsibility for
any errors remains with the authors.

References: Adler, M.D. and E.A. Posner (2000): "Implementing Cost-Benefit Analysis When
Preferences Are Distorted," Journal of Legal Studies 29, 1105-1148.
Alberini, A., A. Chiabai, G. Nocella (2006): Valuing the Mortality Effects of Heat-waves.
In: Menne, B., Ebi, K.L. (eds.), Climat e Change and Adaptation Strategie for
Human Health. Springer, Steinhopff Verlag, Darmstadt. ISBN: 3-7985-1591-3. pp.
345-371.
Alberini, A., Š časný, M., Braun Kohlová, M. (2005): The Value of Statistical Life in the
Czech Republic. Poster presented at the 14th Annual Meeting of the European Association of Environmental and Reso urce Economics EAERE-2005, Bremen, 23-
26 June, 2005.
Alberini, A., Š časný, M., Braun Kohlova, M., Melichar, J. (2006): The Value of Statistical
Life in the Czech Republic: Evidence from Contingent Valuation Study. In: Menne,
B., Ebi, K.L. (eds.), Climate Change an d Adaptation Strategie for Human Health.
Springer, Steinhopff Verlag, Darmstadt. ISBN: 3-7985-1591-3. pp. 373-393.
Arabsheibani, R. G., and A. Marin (2000): Stab ility of Estimates of the Compensation for
Damage, Journal of Risk and Uncertainty 20:3, 247-269.
Black, Dan A. “Some Problems in the Identifica tion of the Price of Risk,” Paper presented
at USEPA Workshop, Economic Valuation of Mortality Risk Reduction: Assessing
the State of the Art for Policy Applications, Silver Spring, Maryland, November 6-
7, 2001.
Blomquist GC (2004): Self-protection and Aver ting Behavior, Values of Statistical Lives
and Benefit Cost Analysis of Environmen tal Policy. Review of Economics of the
Household 2:69-110.
Braun Kohlová, M., Š časný, M., Máca, V., Melichar, J. (2005): Environmentální vlivy na
zdraví d ětí (Environmental effects on children's health) – Final re port from Project
IC/5/6/04 granted by Ministry of Environment of the Czech Republic.
Brown, C. (1980): "Equalizin g Differences in the Labor Market," Quarterly Journal of
Economics 94(1), 113-134.
Carson, R.T. (1991): Constructed markets; in : Braden, J., C. Koldstad (eds.), Measuring
the Demand for Environmental Qualit y, Amsterdam, Elsevier, pp. 121-62.
Chilton, S., J. Covey, M. Jones-Lee, G. L oomes, H. Metcalf (2004), Valuation of Health
Benefits associated with Reduction in Air Pollution. Final Report, DEFRA, UK.
CSERGE, IOS-NLH, IVM, CAS, DAE-UoV. Benefits Transfer and the Economic Valuation of
Environmental Damage in th e European Union: With Special Reference to Health.
CSERGE, IOS-NLH, IVM, CAS and DAE-UoV, 1999.
Desaigues, B., Ami, D., Hutchison, M., Rabl, A., Chilton, S., Metcalf, H., Hunt, A., Ortiz,
R., Navrud, S., Kaderjak, P., Szántó, R., Nielsen, J.S., Jeanrenaud, C., Pellegrini,
S., Braun Kohlová, M., Scasny, M., Máca, V., Urban, J., Stoeckel, M.E., Bartczak,
A., Markiewicz, O., Riera, P., Farreras, V. (2006): Final Report on the monetary
valuation of mortality and morbidity risk s from air pollution. Final report. New
Energy Externalities Developments for Sustainability, Research Stream 1b,

14Workpackage 6, New approaches for valuat ion of mortality and morbidity risks
due to pollution. Project funded by the EC within the 6th Framework Programme.
Devlin, N. and D. Parkin (2003): "Does NICE have a cost effectiveness threshold and
what other factors influenc e its decisions? A discrete choice analysis", Health
Economics, 13(5).
Dickens, W.T. (1984): "Differences Between Risk Premiums in Union and Nonunion
Wages and the Case for Occupational Sa fety Regulation," American Economic
Review 74(2), 320-323.
Dorsey S. and N. Walzer (1983): "Workers ' Compensation, Job Hazards, and Wages,"
Industrial and Labor Relations Review 36(4), 642-654.
Duncan, G.J. and B. Holmlund. (1983): "Was Ad am Smith Right After All? Another Test of
the Theory of Compensating Wage Differen tials," Journal of Labor Economics 1(4),
366-379.
European Commission (2005): Externe – Exte rnalities of Energies: Methodology 2005
Update. Edited by Bickel, P. and Friedrich, R. Published by the Directorate-General
for Research, Sustainable Energy Systems of European Commission.
European Commission (2006): Thematic Strate gy on air pollution, Communication from
the Commission to the Council and th e European Parliament. COM(2005) 446
final, Brussels.
Fischhoff, B., S. Lichtenstein, P. Slovic, S.L. Derby, and R.L. Keeney (1981): Acceptable
Risk. New York: Cambridge University Press.
Freeman III, A.M. (2003): The Measurement of Environmental and Ressource Values.
Theory and Methods. 2nd edition (1st ed ition 1993). Resource for the Future,
Washington, DC.
Gerking, S., M. de Haan, and W. Schulze (1 988): "The Marginal Value of Job Safety: A
Contingent Valuation Study," Journal of Risk and Uncertainty 1(2), 185-199.
Giergiczny, M. (2006): Value of Statistical Life – Case of Poland. Paper prepared for the
the 3rd Annual Congress of Association of Environmental and Resource Economics
AERE, Kyoto, 4-7 July, 2006.
Gunderson, Morley and Douglas Hyatt, "Workp lace risks and wages: Canadian Evidence
from Alternative Models," Canadian Journal of Economics 34:2 (2001), 377-395.
Hanemann, M. (1991): Willingness to pay an d willingness to accept: how much can they
differ? American Economic Review, 81 (3), 635-47.
Halvorsen, R. and H. O. Pollakowski (1981): Choice of Functional Form for Hedonic Price
Equations, Journal of Urban Economics, 10, 37-49.
H a l v o r s e n , R . , H . O . P o l a k o w s k i ( 1 9 8 1 ) : C h o i c e o f F u n c t i o n a l F o r m f o r H e d o n i c P r i c e
Equations. Journal of Urban Economics, 10, 37-49.
Hamermesh, D.S. (1978): "Economic Aspects of Job Satisfaction." In O. Ashenfelter and
W. Oates (eds.), Essays in Labor Market Analysis. New York: John Wiley and Sons,
pp. 53-72.
Hamermesh, Daniel S. (1985): Expectations , Life Expectancy, and Economic Behavior.
Quarterly Journal of Econ omics 100 (2): 389-408.
Hwang, H.-S., W.R. Reed, and C. Hubbard (1992): "Compensating Wage Differentials
and Unobserved Productivity," Journal of Political Economy 100(4), 835-858.
Jennings, P.W., A. Kinderman (2003): The Value of a Life: New Evidence of the
Relationship Between Changes in Occupati onal Fatalities And wages of Hourly
Workers, 1992 to 1999. The Journal of Risk and Insurance, 2003, Vol. 70, No. 3,
549-561.
Itaoke, K., A. Saitoi, A. Krupnick, W. Adamowic z, T. Taniguchi (2006): The Effect of Risk
Characteristics on the Willingness to Pa y for Mortality Risk Reductions from
Electric Power Generation. Environmental & Resource Economics (2006) 33: 371-
398.
Kochi, I., B. Hubbell and R. Kramer (2006) , An Empirical Bayes Approach to Combining
and Comparing Estimates of the Value of a Statistical Life for Environmental Policy
Analysis. Environmental and Resource Econ omics, Volume 34, Number 3, pp. 385-
406.

15Kolbet, G., K. Eberhardt and P. Geborek (20 04): "TNF inhibitors in the treatment of
rheumatoid arthritis in clinical practice: costs and outcomes in a follow up study of
patients with RA treated with etanercept or infliximab in southern Sweden",
Annals of Rheumatic Disease, 63. Available at
http://ard.bmjjournals.c om/cgi/reprint/63/1/4
Krupnick, A., Bhattacharya, S., Alberini, A., Ve ronesi, M. (2006): Mortality Risk Valuation
in Six Countries: Testing Benefit-Transf er Approaches. Prepared for The Third
World Congress of Environmental and Reso urce Economics, Kyoto, Japan, July 3-
7, 2006.
Krupnick, Alan, Anna Alberini , Maureen Cropper, Nathalie Simon, Bernie O'Brien, Ron
Goeree, and Martin Heintzelman (2002): "A ge, Health and the Willingness to Pay
for Mortality Risk Reductions: A Contingent Valuation Study of Ontario Residents,"
Journal of Risk Uncertainty, 24, 161-186.
Landenfeld, J., S., Seskin, E., P. (1982), The Economic Value of Life: Linking Theory to
Practise. American Journal of Public Health, 72(6): 555-566.
Le Net (1994): "Le Prix de la Vie Humaine", Rapport pour le Commissariat Général du
Plan, Paris.
Loomis, John B. and Pierre H. duVair (1993): Evaluating the Effect of Alternative Risk
Communication Devices on Willingness to Pay: Results from a Dichotomous Choice
Contingent Valuation Experience, Land Economics 69:3, 87-98.
Maddison, David and Andrea Bigano (2003): "T he Amenity Value of the Italian Climate",
Journal of Environmental Economics and Management, 45: 319-332.
Markandya, A., Harou, P., BellÙ L. G., Cist ulli, V. (2002): Environmental Economis for
Sustainable Growth: A Handbook for Practi tioners. Chelthenam, UK: Edward Elgar.
McDaniels, Timothy L., Mark S. Kamlet, Greg oryW. Fischer. (1992). “Risk Perception and
the Value of Safety,” Risk Analysis 12, 495–503.
Moore, M.J. and W.K. Viscusi. (1988a): "Dou bling the Estimated Value of Life: Results
Using New Occupational Fatality Data," Jour nal of Policy Analysis and Management
7(3), 476-490.
Moore, Thomas Gale (1998): "Health and Amenity Effects of Global Warming", Economic
Inquiry, 471-488.
Mrozek, J.R. and L.O. Taylor (2002): "Wha t Determines the Value of Life? A Meta-
Analysis," Journal of Policy Analysis and Management 21(2), 253-270.
OECD (2002): Valuing Children's Environm ental Risks. Working Party on National
Environmental Policy. Workshop he ld 20/21 November 2002, Paris.
ENV/EPOC/WPNEP(2002)31
Portney, P.R. (1981): "Housing Prices, Health Effects, and Valuing Reductions in Risk of
Death," Journal of Environmental Economics and Management 8, 72- 78.
Randall, A., J.R.Stoll (1980): Consumer Surp lus in commodity space, American Economic
Review, 70 (3), 449-55.
Rawlins, M.D. and A.J. Culyer (2004): "National Institute for Clinical Excellence and its
value judgements", British Medicine Journal (BMJ), 329.
Rosen, S. (1974): "Hedonic Prices and Implicit Markets: Product Differentiation in Pure
Competition," Journal of Political Economy 82, 34-55.
Slovic, Paul, Baruch Fischhoff, and Satan Liechtenstein (1979): Rating the Risks.
Environment 21 (3): 14-20, 36-39.
Smith, A., 1776 (reprinted 1937): The wealth of Nations (New York: Modern Press).
Ščasný, M. (2005), Quantification of Value of a Statistical Life in the Czech republic: a
review of method. Working Paper prepar ed for the Project VaV/320/1/03 on
External cost from electricity and heat production in Czech Republic and the
methods of their internaliz ation, Charles University Environment Center, mimeo
(in Czech).
Tsuge, T., A. Kishimoto, K. Takeuchi (2 005): A Choice Experiment Approach to the
Valuation of Mortality. The Journal of Risk and Uncertainty, 31:1; 73-95, 2005.
Thaler, R. and S. Rosen (1975): "The Value of Saving a Life: Evidence from the Labor
Market." In N.E. Terleckyj (ed.), Hous ehold Production and Consumption. New
York: Columbia University Press, pp. 265-300.

16Towse, A. and C. Pritchard (2002): "Does NICE have a threshold? An external view", in
A. Towse, C. Pritchard and N. Devlin (eds) Cost-Effectiveness Thresholds:
economic and ethical issues. London: King 's Fund/Office for Health Economics.
U.S. EPA (U.S. Environmental Protection Agency) 2000: Guidelines for Preparing
Economic Analyses. EPA 240-R-00-00 3. Washington, DC: U.S. EPA.
Viscusi, W.K. (1978): "Labor Market Valuations of Life and Limb: Empirical Evidence and
Policy Implications," Public Policy 26(3), 359-386.
V i s c u s i , W . K . ( 1 9 7 9 ) : E m p l o y m e n t H a z a r d s : A n I n v e s t i g a t i o n o f M a r k e t P e r f o r m a n c e .
Cambridge, MA: Harvard University Press.
Viscusi, W.K. (1980): "Union, Labor Market Structure, and the Welfare Implications of
the Quality of Work," Journal of Labor Research 1(1), 175-192.
Viscusi, W.K. and C.J. O'Connor. (1984): "Adaptive Responses to Chemical Labeling: Are
Workers Bayesian Decision Makers?" Am erican Economic Review 74(5), 942-956.
Viscusi, W. Kip and William N. Evans (1990): Utility Functions That Depend on Health
Status: Estimates and Economic Implicat ions, American Economic Review 80:3
(1990) 353-374.
Viscusi, W. Kip, Fatal Tradeoffs (1992): Public and Private Responsibilities for Risk, New
York: Oxford University Press.
Viscusi, W.K., Aldy. J.E. (2003): The Value of a Statistical Life: A Critical Review of
Market Estimates Throughout the World. Journal of Risk and Uncertainty 27(1):5-
76.

17Figure 1: Absolute number of occupati onal injuries, Czech Republic, 1993-2005
Absolute number of occupational injuries
fatal injuriesnon-fatal injuries
050100150200250300350
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005fatal injuries
020 00040 00060 00080 000100 000120 000
non-fatal injuries

Figure 2: Fatal and non-fatal injuries per 10,000 employees, Czech Republic, 1993-2005.
Occupational inuries per 10,000 employees
fatal injuries non-fatal injuties
0,000,100,200,300,400,500,600,70
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005fata injuries
050100150200250
non-fatal injuries

Figure 3: Fatal injuries in br anches, Czech Republic, 1993-2005.
Fatal injuries, CZ 1999-2005
050100150200250
1999
2000
2001
2002
2003
2004
2005Others+H+J+O
Halth (N)
Education (M)
Public admin (L)
Services (K)
Transport (I)
Trade (G)
Buildings (F)
Power €
Manufacturing (D)
Mining ©
Agricultural (A)

18Table 1: Descriptive statistics – “Quality of Occupational Life- 2006” survey.

whole sample only exposed to risks only man and exposed to risks
Variable Variable type N Mean Std Dev N Mean Std Dev Min Max N Mean Std Dev Min Max

wage thousands CZK 1 616 15.38 8.32 1 085 16.42 9.20 3 100 694 17.89 9.71 3 100
fatal cases per 10.000 2 037 1.24 0.80 1 367 1.29 0.86 1 9 906 1.41 0.92 1 9
nonfatal cases per 10.000 2 037 14.71 15.10 1 367 16.15 15.81 1 84 906 19.52 16.78 1 84

BOSS dummy 2 042 0.25 0.43 1 373 0.30 0.46 0 1 909 0.33 0.47 0 1
BUILDWOK dummy 2 043 0.07 0.25 1 373 0.09 0.29 0 1 909 0.13 0.34 0 1
cartravelling dummy 2 043 0.44 0.50 1 373 0.65 0.48 0 1 909 0.73 0.44 0 1
EXPERIEN years 2 043 7.17 8.15 1 373 7.20 8.36 0 50 909 7.16 8.52 0 50
MALE dummy 2 043 0.56 0.50 1 373 0.66 0.47 0 1 909 1.00 0 1 1
University dummy 2 043 0.14 0.35 1 373 0.15 0.35 0 1 909 0.14 0.35 0 1
A-level dummy 2 043 0.37 0.48 1 373 0.36 0.48 0 1 909 0.31 0.46 0 1
AGE years 2 036 40.38 29.37 1 370 40.46 29.32 18 72 906 40.52 29.67 18 72
BREADWIN dummy 1 467 0.62 0.23 1 001 0.64 0.23 0 1 642 0.68 0.21 0 1
KIDS no of children 2 043 0.69 0.89 1 373 0.72 0.91 0 4 909 0.70 0.91 0 4

kzam1 dummy 2 043 0.09 0.29 1 373 0.12 0.33 0 1 909 0.14 0.34 0 1
kzam2 dummy 2 043 0.04 0.20 1 373 0.04 0.19 0 1 909 0.04 0.19 0 1
kzam3 dummy 2 043 0.17 0.37 1 373 0.16 0.37 0 1 909 0.12 0.32 0 1
kzam4 dummy 2 043 0.13 0.33 1 373 0.07 0.26 0 1 909 0.04 0.19 0 1
kzam5 dummy 2 043 0.21 0.41 1 373 0.20 0.40 0 1 909 0.16 0.36 0 1
kzam6 dummy 2 043 0.04 0.20 1 373 0.05 0.22 0 1 909 0.06 0.25 0 1
kzam7 dummy 2 043 0.14 0.35 1 373 0.17 0.38 0 1 909 0.25 0.43 0 1
kzam8 dummy 2 043 0.09 0.28 1 373 0.11 0.32 0 1 909 0.14 0.35 0 1
kzam9 dummy 2 043 0.09 0.28 1 373 0.07 0.25 0 1 909 0.06 0.24 0 1

okec1 dummy 2 043 0.04 0.20 1 373 0.05 0.22 0 1 909 0.06 0.23 0 1
okec2 dummy 2 043 0.00 0.05 1 373 0.00 0.05 0 1 909 0.00 0.07 0 1
okec3 dummy 2 043 0.01 0.10 1 373 0.01 0.11 0 1 909 0.02 0.14 0 1
okec4 dummy 2 043 0.16 0.37 1 373 0.16 0.37 0 1 909 0.19 0.40 0 1
okec5 dummy 2 043 0.03 0.16 1 373 0.03 0.17 0 1 909 0.04 0.20 0 1
okec6 dummy 2 043 0.10 0.29 1 373 0.12 0.32 0 1 909 0.17 0.38 0 1
okec7 dummy 2 043 0.14 0.35 1 373 0.13 0.33 0 1 909 0.11 0.31 0 1
okec8 dummy 2 043 0.06 0.23 1 373 0.05 0.22 0 1 909 0.03 0.17 0 1
okec9 dummy 2 043 0.08 0.27 1 373 0.09 0.29 0 1 909 0.12 0.32 0 1
okec10 dummy 2 043 0.03 0.18 1 373 0.03 0.18 0 1 909 0.03 0.17 0 1
okec12 dummy 2 043 0.06 0.24 1 373 0.06 0.23 0 1 909 0.05 0.23 0 1
okec13 dummy 2 043 0.08 0.27 1 373 0.06 0.23 0 1 909 0.03 0.18 0 1
okec14 dummy 2 043 0.08 0.27 1 373 0.09 0.28 0 1 909 0.04 0.20 0 1
okec15 dummy 2 043 0.08 0.28 1 373 0.07 0.26 0 1 909 0.06 0.24 0 1
okec16 dummy 2 043 0.01 0.11 1 373 0.01 0.09 0 1 909 0.00 0.06 0 1

19Table 2: KPZ-2006: full sample.

Model with the best lamda Model with λ=0 FULL SAMPLE
Coefficient Liberal p Coefficient Pr > ChiSq
Intercept 1.768 >= <.0001 1.718 <.0001
fatal 0.148 >= 0.0015 0.136 0.001
fatal2 -0.021 >= 0.0016 -0.019 0.001
BOSS 0.166 >= <.0001 0.150 <.0001
Cartravelling 0.082 >= 0.0001 0.073 0.000
MALE 0.185 >= <.0001 0.168 <.0001
University 0.347 >= <.0001 0.312 <.0001
A-level 0.145 >= <.0001 0.132 <.0001
BREADWIN 0.561 >= <.0001 0.509 <.0001
KIDS 0.035 >= 0.0010 0.032 0.001
kzam1 0.562 >= <.0001 0.506 <.0001
kzam2 0.509 >= <.0001 0.461 <.0001
kzam3 0.342 >= <.0001 0.313 <.0001
kzam4 0.311 >= <.0001 0.286 <.0001
kzam5 0.241 >= <.0001 0.221 <.0001
kzam6 0.198 >= 0.0030 0.181 0.003
kzam7 0.305 >= <.0001 0.279 <.0001
kzam8 0.343 >= <.0001 0.313 <.0001
okec1 -0.108 >= 0.0538 -0.096 0.055
okec4 -0.085 >= 0.0070 -0.076 0.007
okec6 -0.042 >= 0.2679 -0.037 0.284
okec7 -0.129 >= <.0001 -0.118 <.0001
okec13 -0.103 >= 0.0156 -0.091 0.017
okec14 -0.065 >= 0.1122 -0.059 0.109
okec15 -0.053 >= 0.1808 -0.048 0.169
N 1 462 1 462
Lambda used 0.038 0.000
LogLikelihood -2212.0* -406.3
Adj R-Sq. 0.49
WTP 0.106 0.097
VSL (mil.K č) 175.6 179.4
VSL (mil.€) 5.91 6.04
* Loglikelihood of lambda estimate.
Note: Exchange rate used 28.34 CZK/€ (2006).

20Table 3: KPZ-2006: sub-samples of exposed all and exposed males.

All but exposed to risks Male exposed to risks
Coefficient Liberal p Estimate Pr > ChiSq Coefficient Liberal p Estimate Pr > ChiSq
Intercept 1.692 >= <.0001 1.657 <.0001 1.700 >= <.0001 1.770 <.0001
fatal 0.157 >= 0.0045 0.145 0.004 0.138 >= 0.0061 0.163 0.005
fatal2 -0.024 >= 0.0019 -0.022 0.001 -0.017 >= 0.0201 -0.020 0.018
BOSS 0.154 >= <.0001 0.141 <.0001 0.098 >= 0.0014 0.115 0.001
BUILDWOK 0.151 >= 0.1031 0.137 0.102 0.148 >= 0.0571 0.179 0.047
Cartravelling 0.109 >= <.0001 0.100 <.0001 0.080 >= 0.0031 0.095 0.003
MALE 0.189 >= <.0001 0.174 <.0001
University 0.367 >= <.0001 0.335 <.0001 0.258 >= <.0001 0.307 <.0001
A-level 0.151 >= <.0001 0.140 <.0001 0.095 >= 0.0031 0.112 0.003
BREADWIN 0.603 >= <.0001 0.554 <.0001 0.385 >= <.0001 0.458 <.0001
KIDS 0.038 >= 0.0028 0.035 0.002 0.046 >= 0.0002 0.053 0.000
kzam1 0.551 >= <.0001 0.502 <.0001 0.526 >= <.0001 0.628 <.0001
kzam2 0.406 >= <.0001 0.372 <.0001 0.468 >= <.0001 0.556 <.0001
kzam3 0.286 >= <.0001 0.264 <.0001 0.308 >= <.0001 0.361 <.0001
kzam4 0.262 >= 0.0003 0.243 0.000 0.251 >= 0.0034 0.291 0.004
kzam5 0.234 >= <.0001 0.216 <.0001 0.299 >= <.0001 0.351 <.0001
kzam6 0.204 >= 0.0131 0.189 0.011 0.134 >= 0.0520 0.155 0.054
kzam7 0.280 >= <.0001 0.258 <.0001 0.282 >= <.0001 0.331 <.0001
kzam8 0.327 >= <.0001 0.301 <.0001 0.320 >= <.0001 0.376 <.0001
okec1 -0.120 >= 0.0820 -0.109 0.080
okec4 -0.052 >= 0.1704 -0.047 0.171
okec6 -0.145 >= 0.0708 -0.130 0.072 -0.123 >= 0.0701 -0.150 0.057
okec7 -0.137 >= 0.0005 -0.128 0.000 -0.071 >= 0.0794 -0.083 0.081
okec13 -0.097 >= 0.0754 -0.087 0.080 -0.087 >= 0.1727 -0.105 0.158
N 997 997 640 640
Lambda used 0.050 0.000 0.050 0.000
LogLikelihood -1603.7* -313.7 -1096.3* -201.0
Adj R-Sq. 0.47 0.37
WTP 0.109 0.101 0.103 0.122
VSL (mil.K č) 195.2 198.0 263.3 261.5
VSL (mil.€) 6.57 6.67 8.87 8.80
* Loglikelihood of lambda estimate.
Note: Exchange rate used 28.34 CZK/€ (2006).

21Table 4: Descriptive statistics – statistical data for sectors, CZ, 2003-2005.

Variable FULL SAMPLE only POSITIVE MANUAL
Wage 1000 CZK per empl. 325.16 316.16 301.98
Net_wage 1000 CZK per empl. 262.73 256.34 246.45

r_fatal cases per 1000 0.03 0.06 0.07
r_injur3 cases per 1000 17.39 21.29 24.61
r_injury cases per 1000 8.39 10.22 12.24
r_longill cases per 1000 0.41 0.32 0.39
r_sickdays cases per 1000 35.56 56.59 70.17
r_longrisk cases per 1000 0.05 0.07 0.08

Kcpain 1000 CZK per empl. 0.14 0.09 0.10
Kcprevent 1000 CZK per empl. 0.18 0.17 0.18
L_productivity 1000 CZK per empl. 705.24 616.51 655.76
Labour number of empl.s 71 040 105 225 88 774

okec_agri dummy 0.05 0.07 0.09
okec_ind dummy 0.47 0.44 0.56
okec_ener dummy 0.04 0.06 0.07
okec_stav dummy 0.02 0.03 0.04
okec_serv dummy 0.35 0.33 0.15
okec_tran dummy 0.07 0.07 0.09
Dchemie dummy 0.02 0.03
Dplast dummy 0.06 0.07
Dkovy dummy 0.07 0.09
Dstroje dummy 0.02 0.03

22Table 5: Estimation results – statistica l data for sectors 2003-2005: full sample.

MODEL 1a MODEL 2a MODEL 3a
Coeff. p-value Coeff. p-value Coeff. p-value

intercept 10.777 >= <.0001 9.214 >= <.0001 2.077 >= <.0001
Fatal 0.295 0.1987 0.261 0.1627 0.001 0.0980
kcprevent 0.953 >= <.0001
LP 0.000 >= <.0001 0.000 >= .0007 0.000 >= .0005
Okec_ind -0.122 >= <.0001 -0.081 >= .0004 0.000 >= .0006
Okec_agri -0.213 >= .0006 -0.201 >= .0001 -0.001 >= .0001

Dependent var wage wage net_wage
Lambda used -0.02 -0.05 -0.48
LogLikelyhood λ -1188 -1184 -1801
Adj R-Sq 0.63 0.1888 0.192

VSL(mil.Kč) 91.0 116.8 107.8
VSL(mil. €) 3.06 3.92 3.62
VSL(m€; PPP) 6.50 8.34 7.70
Note: Exchange rate used 29.78 CZK/€ (2005).

Table 6: Estimation results – statistical da ta for sectors 2003-2005: only sectors with
positive fatal rates.

MODEL 4a MODEL 5a MODEL 6a
Coeff. p-value Coeff. p-value Coeff. p-value
intercept 97595.04 >= <.0001 463902.75 >=
<.0001 1884327.06 >= <.0001
fatal 69560.01 0.0350 436430.48 0.0238 1900912.21 0.0219
LP 58.49 >= .0001 352.25 >= .0001 1510.09 >= .0001
okec_ind >= .0006
okec_agri >= .0001
injur3 -2.56 0.1714
injury 2.64 0.11
Dplast 8581.97 0.34
Dkovy 23829.16 0.0087 120773.82 >= .0214 516612.95 0.0219
Dstroje 22663.10 0.12

Dependent var wage wage net_wage
Lambda used 0.96 1.105 1.245
LogLikelyhood λ -702.8 -704.2 -688.6
Adj R-Sq 0.4965 0.4795 0.4798

VSL(mil.Kč) 113.9 119.7 96.4
VSL(mil. €) 3.82 4.02 3.24
VSL(m€; PPP) 8.13 8.55 6.89
Note: Exchange rate used 29.78 CZK/€ (2005).

23Table 7: Estimation results – statistical da ta for sectors 2003-2005: only sectors with
positive and services excluded.

Model 7a Model 7a
kod dat Coeff. p-value Coeff. p-value

intercept 3.249 >= <.0001 3.6212109 >= <.0001
Fatal 0.014 0.068 0.021 0.065
LP 0.000 >= .0001 0.0000125 >= .0001
injur3 0.000 >= .0001 -0.0004094 >= .0001
Injury 0.000 >= .0493 0.0001329 >= .0461
Plast 0.004 0.086 0.0052480 0.091
Dkovy 0.006 0.003 0.0091681 0.003
Dstroje 0.005 0.122 0.0069761 0.124

Dependent var Wage net_wage
Lambda used -0.3 -0.265
LogLikelyhood λ -907 -886.8
Adj R-Sq. 0.4423 0.44444

VSL(mil.Kč) 134.1 106.1
VSL(m€) 4.50 3.56
VSL(m€; PPP) 9.58 7.58
Note: Exchange rate used 29.78 CZK/€ (2005).

24Table 8: Descriptive statistics – Working conditions 2000, CZ 2000.

Variable Description N Mean Std Dev Minimum Maximum

Net wage Net monthly wage per month 786 9 728 7 155 2 125 119 000
priplatek dummy=1 if compensation 892 0,12 0,33 0 1
noc dummy=1 if compensation for
working during night 891 6,97 11,64 0 90
vikend dummy=1 if compensation for
working during weekend 892 1,89 2,29 0 9,33
praxe Working practise 890 8,32 2,50 0 12,58
vedouci Boss 892 0,17 0,38 0 1

RISK dummy=1 if subjectively
perceived occupational risks 892 0,42 0,49 0 1
RISK_fyz dummy=1 if perceived physical
risks 892 0,36 0,48 0 1
RISK_psy dummy=1 if perceived
psychological risk 892 0,21 0,41 0 1
RISK_fn dummy=1 if perceived risk due
to physical aggression 892 0,07 0,25 0 1

okec_zem dummy for sector (agriculture) 892 0,06 0,23 0 1
okec_prum dummy for sector (industry) 892 0,31 0,46 0 1
okec_stav dummy for sector (construction 892 0,06 0,24 0 1
okec_dopr dummy for sector (transport) 892 0,07 0,26 0 1
okec_slu dummy for sector (services) 892 0,50 0,50 0 1
Dummies for profession
kzam1 Managers 892 0,06 0,23 0 1
kzam2 research and intellectuals 892 0,06 0,23 0 1
kzam3 Technicians 892 0,18 0,38 0 1
kzam4 lower administration 892 0,12 0,32 0 1
kzam5 service and sellers 892 0,18 0,38 0 1
kzam6 blue-collars in agri/forestry 892 0,02 0,12 0 1
kzam7 blue-collars in industry 892 0,19 0,40 0 1
kzam8 machinery service, drivers 892 0,11 0,31 0 1
kzam9 non-professional blue-collars 892 0,08 0,27 0 1

zena dummy=1 if female 892 0,51 0,50 0 1
vek Age 886 39,84 11,21 15 76
zenaty dummy=1 if married 892 0,56 0,50 0 1
eaosob Number of economic-active
person in household 892 1,74 0,79 0 6
deti Number of children in
household 892 0,65 0,87 0 4
osob Number of person in
household 892 2,94 1,19 0 7
zivitel dummy=1 if brings large part
of money to family 892 0,61 0,49 0 1
obec1 dummy=1 if small village 892 0,25 0,44 0 1
obec2 dummy=1 if small town 892 0,29 0,45 0 1
obec3 dummy=1 if larger city 892 0,24 0,43 0 1
obec4 dummy=1 if city with more
than 100,000 inhabitants 892 0,22 0,41 0 1

25Table 9: Estimation results – Working Conditions 2000, CZ 2000.

Coeff. liberal p-value
Intercept 5.761 >= <.0001
Fatal-sq 0.003 >= 0.1800
VIKEND 0.003 >= 0.0533
PRAXE 0.010 >= <.0001
ZIVITEL -0.046 >= 0.0604
ZENA_BI -0.065 >= <.0001
Age -0.001 >= 0.0004
Ecoactive person in household 0.015 >= 0.0076
Person in household -0.026 >= <.0001
okec_industry -0.034 >= 0.0028
okec_services -0.025 >= 0.0255
KZAM2 0.038 >= 0.0659
KZAM3 -0.108 >= <.0001
KZAM4 -0.116 >= <.0001
KZAM5 -0.204 >= <.0001
KZAM6 -0.212 >= <.0001
KZAM7 -0.165 >= <.0001
KZAM8 -0.163 >= <.0001
KZAM9 -0.220 >= <.0001
City1 -0.073 >= <.0001
City2 -0.068 >= <.0001
City3 -0.059 >= <.0001
Riskwomen 0.024 >= 0.0496
riskzivitel -0.047 >= <.0001
riskbenefit 0.042 >= 0.0040
pecovatel 0.039 >= <.0001

Dependent var Net_wage
Lambda used -0.12
LogLikelyhood λ -9098
Adj R-Sq 0.19