• No results found

In sickness and in health : A study of work absence

N/A
N/A
Protected

Academic year: 2021

Share "In sickness and in health : A study of work absence"

Copied!
29
0
0

Loading.... (view fulltext now)

Full text

(1)

J

Ö N K Ö P I N G

I

N T E R N A T I O N A L

B

U S I N E S S

S

C H O O L

JÖNKÖPING UNIVE RSITY

In sickness and in health

A s t u d y o f w o r k a b s e n c e

Paper within Economics

Author: Per Leander and Fredrik Ljung Tutor: Johan Klaesson, Johanna Palmberg, Hyunjoo

(2)

Bachelor Thesis within economics

Title:

In sickness and in health – a study of work absence

Authors:

Per Leander and Fredrik Ljung

Tutors:

Johan Klaesson, Johanna Palmberg, Hyunjoo Kim

Date:

2007-12-23

Abstract

The average number of sick days per year and employee differs widely among countries from around 4 days in the US to almost 30 in Slovak Republic, this pa-per analyze the impact of social generosity on absence due to sickness.

The authors determines some variables that affect the absent level, and also how, or by how much those variables affect the average number of sick days in a country.

The analysis is made by using a cross sectional regression and a sample of 18 dif-ferent countries. The results show that two of the original ten independent vari-ables; the generosity index and total expenditure on health variables, have a sig-nificant impact on the average number of sick days in a country.

The results show that there is a positive effect between the Generosity index variable and the average number of sick days, and a negative relationship be-tween the total expenditure on GDP and the average number of sick days.

(3)

Kandidatuppsats inom nationalekonomi

Titel:

In sickness and in health – a study of work absence

Författare:

Per Leander and Fredrik Ljung

Handledare:

Johan Klaesson, Johanna Palmberg, Hyunjoo Kim

Datum:

2007-12-23

Sammanfattning

Det genomsnittliga antal sjukdagar per år och anställd varierar kraftigt mellan olika länder från drygt 4 i USA till nära 30 i Slovakien, den här uppsatsen analyserar hur den allmänna generositeten i sjukförsäkringsystemet påverkar antalet sjukdagar i ett land.

I analysen använder författarna tvärsnittsdata från 18 olika länder. Datan analyseras i en regressionsmodell och av de 10 ursprungliga variablarna visar sig två vara signifikanta, generositetsindexet samt hälsovårdkostnader som procent av BNP.

Författarna kommer fram till att generositetsnivån i sjukförsäkringsystemet har en positiv inverkan på antal genomsnittliga sjukdagar i ett land samt att ökade anslag till hälsovård minskar antal sjukdagar.

(4)

Table of Contents

Tables and figures ... 4

1

Introduction ... 4

1.1 Purpose ...6 1.2 Hypothesis...6 1.3 Delimitation...6 1.4 Method ...6 1.5 Outline ...6 1.6 Earlier studies...6

1.7 General health and absenteeism...8

2

Theory ... 11

2.1 Labor – leisure choice ...11

2.2 Asymmetric information ...12 2.3 The model ...13

3

Empirical framework... 14

3.1 Descriptive statistics...14 3.1.1 Dependent variable ...14 3.1.2 Independent variables ...15 3.1.3 Construction of an index...17 3.2 Correlation matrix ...19 3.3 Regression analysis ...19 3.3.1 Regression output ...20

4

The Swedish experiment... 21

5

Analysis ... 22

5.1 Interpreting the regression results ...22

5.1.1 Percentage of GDP spent on health care ...23

5.1.2 The generosity index ...23

6

Conclusion... 24

6.1 Author’s remarks ...24

7

Future studies ... 25

(5)

Tables and figures

Equation 1_______________________________________________________________________ 13 Equation 2_______________________________________________________________________ 22 Table 1 Sick days_________________________________________________________________ 15 Table 2 Generosity index _________________________________________________________ 16 Table 3 Descriptive statistics ______________________________________________________ 17 Table 4 Index ____________________________________________________________________ 18 Table 5 Correlation matrix ________________________________________________________ 19 Table 6 Regression output with Sick days as dependent variable ____________________ 20 Table 7 Sweden descriptive statistics ______________________________________________ 21 Table 8 Sweden Regression ______________________________________________________ 22 Figure 1 Life expectancy __________________________________________________________ 9 Figure 2 Infant Mortality ___________________________________________________________ 9 Figure 3 Average sick-days per year_______________________________________________ 10

(6)

A widely discussed matter in Swedish politics today is the proposed changes to the health insurance system. In the budget proposal of 2007, the newly elected Swedish government suggested changes to this system with the aim to provide an incentive for people who are absent from work due to illness to go back to work.

The proposal implies some changes to the current system. The main point is that one should benefit from returning to work after being absent due to illness. Therefore, the government intends to decrease the amount compensated to peo-ple who are absent due to sickness. This is believed to drive peopeo-ple to return to work as soon as they feel capable and reducing the risk for people taking advan-tage of the system by not returning to work when they are.

The discussion following this proposal has been whether lowering the compen-sation would indeed decrease the amount of people who are absent from work. One of the opponents to the current government, the social democrats, believe that a high rate of compensation helps people who have been unemployed or sick during a long period. By giving them a sense of security during the time of absence they would have an opportunity to focus on finding a new job.

The number of sick days varies widely among comparable industrialized coun-tries and later years the discussion about what causes the difference also suggests widely diverse reasons. The variables include working environment, social wel-fare measured by public health as well as macro economical conditions such as employment level, labor force participation rates and size of the black labor mar-ket. Two of the most common arguments in recent studies are the conditions in the workplace and the public health.

The general health conditions might be a plausible explanation for average sick days but the variation in health status cannot possibly be as large as the variation in absence between the industrialized countries. For example according to OECD health data (2007) the number of average sick days per person in the US is 4.3 days per year and in Sweden 20 days per year. When investigating the health status in the countries one can notice that Sweden has lower infant mortality and longer life expectancy than the US, (OECD health data 2007), thus considered having a healthier population. This being the case; should not the Swedish labor force be healthier and less frequent absent from the work place?

This conclusion suggests that there are other more important factors than public health that determine the number of average sick days in an economy. One of these, the one handled in this paper, is how the economic incentives to work and the opportunity cost not to affect the frequency and duration of absence due to sickness.

(7)

1.1 Purpose

The purpose of the paper is by using economic theory and existing research to analyze how large impact the level of generosity in the social security and labor market has on the aggregate number of sick days in a given set of economies.

1.2 Hypothesis

The null hypothesis is: the generosity in social security does not affect the aver-age number of sick leave days positively.

1.3 Delimitation

The analysis is based on a cross country study and the most obvious delimitation is the lack of data. There are only a limited amount of countries that report the average days of sick leave and this is a problem when doing cross sectional re-gressions.

The presence of sickness spells is also not accounted for, the measurement of the health conditions in the different economies are based on life expectancy and in-fant mortality. These measurements are frequently used to measure the state of the healthcare and the over all health status in countries.

1.4 Method

The analysis will be based on quantitative secondary data and the econometric study will be in the form of cross sectional data regression. The dependant vari-able is the number of sick days in a country per year and per employee. The in-dependent variables are; a generosity index, unemployment, average spending on health care, days before sickness compensation, and 3 different index vari-ables based on compensation, people between 55 and 64 years of age on the la-bor market, infant mortality and life expectancy.

1.5 Outline

In chapter two the theoretical framework will be presented. In the third chapter the empirical framework consisting of descriptive statistics and regressions are presented. In the fourth chapter the results from the regressions will be analyzed and discussed. The fifth chapter consists of a conclusion where the purpose question will be answered and the authors make some own remarks. The last chapter is where suggestions of future studies are presented.

1.6 Earlier studies

The subject of sick leave and the reasons why it is becoming more and more fre-quently researched is probably a result of a large increase of absenteeism in sev-eral countries. Most studies treat the problem not from an economic point of

(8)

view but as a health related issue usually including topics such as physical work-load, stress and general health etc.

Very few reports analyze the personal incentives not to go to work i.e. the eco-nomic compensation of staying at home.

To take the economical approach is interesting for a number of reasons, first of all to see if it is possible to explain some differences using economic analysis in an area classically dominated by health research.

A more important feature is brought up in “Do economic incentives affect work absence?” P Johansson and M Palme (1996). The theory is that since most indus-trialized countries have compulsory health insurance, the government has the power to increase or decrease the work absence. If, of course the employees are affected by the economic incentives.

Further, most studies focus on time series data in single countries with the occa-sional look on other countries. Very few studies compare the absence across dif-ferent countries and even fewer includes analysis about why there are differences between countries.

The first thing to realize is that theoretically it is only interesting to investigate ab-senteeism, as something else than pure health related, if the number of hours worked each year exceeds the number that individuals desire. If the number of hours desired by the worker would agree with what the employers want, the workload would be optimal and the employee would have no incentives of be-ing absent from work.

There is a marginal rate of substitution (MRS) between working one hour and having one our leisure time, this MRS is individual and is probably very different between countries. The basic point is that if a worker accepts an agreement or a contract where the marginal rate of substitution is not in accordance, the worker will have an incentive to absent him/herself, Frick, B. and M.A. Malo (2005). Already in (1952) Buzzard and Shaw showed that there is a positive relationship between the compensation for absence and the level of absenteeism and since then economists have recognized the relation between the marginal rate of sub-stitution and the absence. The conclusions differ though, in “Absenteeism as a Mechanism for Approaching an Optimal Labor Market Equilibrium: An Empirical Study” by L. F. Dunn; Stuart A. Youngblood (1986), the authors argue that there is a strong positive relationship between the absence rate and the difference be-tween workers MRS and wage rate. Dunn and Youngblood concluded that “hours lost to non-medical absence are found to increase with increases in the difference between a workers MRS and the marginal wage rate”.

Other papers like the one by Drago, R. and M. Wooden (1992) supports the la-bor-leisure model but in the opposite form. They argued that high wage workers,

(9)

thus workers with higher opportunity cost of leisure, have a lower propensity of absence than lower wage employees.

In more recent studies, Osterkamp, R and Röhn, O. (2007), emphasizes the con-struction of matrices measuring the overall generosity in the social security sys-tem. The data usually included in these generosity measures are; level of com-pensation, how many days until the compensation stats, for how long can one receive compensation, job security, who certifies sickness (independent doctor or firm related health care) etc. These studies concur with the others and conclude that there is a positive relation between the generosity of the insurance systems and the level of absenteeism in a country. This approach is very thorough but it requires caution because the index is subjectively measured, the authors them-selves fabricate the index, in the Osterkamp, Röhn report, based on data that is naturally hard to measure as it lacks nominal values.

Holmlund B. (2004) also takes into account the relevance of unemployment benefits when analyzing absenteeism. He argues that the rewards for employ-ment increase with the sick leave benefits, meaning that people will move from unemployment to employment if the benefits increase and from employment to unemployment if the benefits decrease or if unemployment benefits increase. Holmlund comes to the conclusion that higher unemployment benefits result in increased unemployment, more interesting his model predicts that higher unem-ployment benefits will increase the general sickness absence. This supports the argument that the absence frequency increase when the risk i.e. loosing ones job and becoming unemployed, decrease.

1.7 General health and absenteeism

One would think that the most important data when analyzing sick absence would be the health of the workforce. In 1997 the European foundation for the improvement of living and working conditions presented the report - “preventing absenteeism at the workplace” page 12. In The report it says “although there is some cynicism and skepticism about the issue it is absolutely clear that bad health is the main reason for workers to be absent from work”. Further they ar-gue that sickness does not necessary mean that worker stay home from work. The statement that most people in a country that are absent from work are sick is probably true, but it does not explain why the numbers differs so widely.

For example; the two most widely used factors when measuring the general health in a country are life expectancy at birth and infant mortality rate. The ta-bles below are a comparison of our subject countries.

(10)

Figure 1 Life expectancy

Source OECD health data 2007

Figure 2 Infant Mortality Source OECD health data 2007

As one can notice in the graph Switzerland has the highest life expectancy and infant mortality is lowest in Sweden. Though eastern European countries still

(11)

struggle the data does not differ all that much, Life Expectancy (LE) starts at high 70 years to low 80 years and infant mortality from 2 to 7 deaths per a 1000 births. Overall the general health seems to be quite similar in the subject countries. If the general health in the countries is similar and the main reason for absentee-ism is sickness the number of sick days should correspond with the health status i.e. there should be no critical fluctuation between the countries.

The graph below shows that there is not only a bigger difference in the sick days between the countries but also that the general health status sometimes is nega-tively related to the number of sick days, thus suggesting that the absence has lit-tle to do with health.

Sick days per year and employee

0 5 10 15 20 25 30 35 Switze rland Austr alia Spain Swed en Italy Fran ce Cana da Austr ia Belgi um Neth erland s Gree ce Luxe mbo urg Ger many UKFinland Korea Portu gal Denm ark US Czec h Repu blic Slov ak R epubl ic Hung ary Sick Days

Figure 3 Average sick-days per year Source OECD health data 2007

(12)

2 Theory

Most studies on absence due to sickness has mostly been dominated by empirical studies focusing on how individuals react on changes in sick pay or trying to find plausible determinants to increases and decrease in absence.

Steven, G. Allen (1981) presents a theory on absence as a result of an individual deciding to participate in nonworking activities during a scheduled work period, since the utility of absence is greater than the wage he receives during that time. In the paper, Allen presents ways to reduce the absenteeism by raising the cost for being absent from work.

Theoretical framework on absence due to sickness is not as common but do ex-ist. The paper of Barmby, Sessions, and Treble (1994) explain how asymmetric information about a persons health effects absence rates and the cost for firms to monitor individuals health. In the paper they show how people tend to overstate their sickness when being indifferent between work and absence.

Coles, Trebles (1996) analyzed the relation between absenteeism rates relation to a firms technology. They show how a firm’s attitude towards absence has much to do with technology. For example there is low absence and high wages in a firm where the worker complement each other since it is too costly for a firm let-ting them being absent. In their paper they assume that healthy workers have the option of choosing to be absent or not and the firm will have to set a contract which will decrease absenteeism by setting a high wage, in short trying to set an optimal wage.

A similar study was conducted by Chatterij, and Tilly (2002), this study also fo-cused on the optimum wage level but with a twist. They claim that Coles and Treble left out the fact that it is not only healthy people who make the decision of being absent but sick people to. They argue that even if people are sick they can still work, and should therefore be accounted for when setting the optimal wage.

It is clear that most of the theory has been based on how to give people an in-centive to go to work instead of being absent, to increase the utility of going to work instead of being absent due to sickness. The rest of this section will use some of the theory stated above to explain what problems people encounter when making the choice of being absent or not. The theory will be based upon Steven, G. Allen’s theory on the labor-leisure model and Barmby et al theory on asymmetric information of health state but will not include how firms try to solve the problem of absenteeism but instead explain why people sometimes choose to be absent from work.

2.1 Labor – leisure choice

When presented with a job offer, the offer has a specified wage rate and a work schedule. Since there is a cost related to job searching, a worker may accept an offer even though at the contracted working hours the marginal rate of substitu-tion between income and leisure do not equal the wage. Thus, when someone agrees upon working more hours than he or she desires given a certain wage rate, he or she has an incentive to consume leisure, which can be done by being

(13)

absent from work, since it has a higher utility value than wage. (Steven, G. Allen 1981)

When analyzing the choice of being absent from work one has to consider the cost involved. As a worker chooses to be absent from work there is a cost both to the worker and the employer. The worker’s most obvious cost is the loss of wage for not going to work that day.

However in the case of paid sick leave absence, a change in wage will not have any effect on the decision of being absent due to sickness, since there is no longer a substitution effect between absences and wage. There for the worker is therefore indifferent between working and not working, (Steven, G. Allen 1981) However there is still a cost involved which comes more in the form of punish-ment. Since the employer faces the cost of being forced to find a replacement for the absent worker, time which he or she could have been using for more pro-ductive matters. This is a cost which the company wants to be compensated for. Therefore, the worker has to reimburse the employer for the time he lost. The re-imbursement reveals itself as an increased probability of loosing your job, or a decreased probability for getting a promotion. So, the only way to make people go to work when there is paid sick leave absence is to make the punishment steeper. (Steven, G. Allen 1981)

The results of being indifferent between working and not working hold if there is some schedule flexibility. Where there is less flexible work schedules the absence rate is higher and thus where more work schedule flexibility there is less ab-sence. So when schedules become more inflexible the workers opportunity to participate in non-work activities decreases, and by being absent for work this opportunity increases. Thus absence is alternate mean of schedule flexibility. However there is a difference between having a flexible schedule and being ab-sent from work, this difference is the cost it implies for the worker. (Steven, G. Allen 1981)

In firms where there is little flexibility the absence rates can be adjusted by the function of wage and punishment of being absent so that the total compensation package (consisting of wage, hours, flexibility and punishment) to the marginal worker remains unchanged. The choice of adjustment mechanism depends on the relative cost of wage increase and higher absenteeism to the firm. (Steven, G. Allen 1981)

2.2 Asymmetric information

According to barmby, sessions, and treble (1994) there is a relationship between asymmetric information of health and the level of absences from work. Workers are uncertain of their state of health and supply labor on the basis of a utility maximisation decision once they have realised whether they are sick or not. The utility is based on income, leisure and health, and as people get sicker they value leisure more than work (labor supply). Firms are able to control the behaviour by threatening to fire those workers who absence themselves by unacceptable sick-ness. Firms can use an efficiency wage effect through which wages can be used to decrease absenteeism. The optimal response of a firm to an increase in moni-toring costs is to discourage workers absenting themselves with unacceptable sickness by increasing wages.

(14)

Workers are risk neutral utility maximisers with a certain amount of time which they allocate between leisure and work. The utility function is an increasing func-tion of income and leisure, with the worker putting weight on each variable (i.e. income and leisure) depending on his/her general level of health. As the worker becomes sicker he/she values leisure time more. (Barmby, Sessions, and Treble 1994)

Potential workers sign a contract specifying the wage rate for a certain amount of labor supply. The quality of the work is not taken into consideration as well as the productivity instead for simplicity, how you manage your job is only depend-ent on your attendance. There are no costs tied to firing a worker and not pro-viding contracted job hours for the employer. Also, before a worker starts he re-alises his state of health and makes a utility maximising decision regarding his absence rate. Furthermore, there is no re-contracting and therefore the absence rate decision is only based on outside options available. (Barmby, Sessions, and Treble 1994)

The outside options available are: Workers how are absent with acceptable sick-ness are entitled to sick pay. The sick pay is set exogenously and is payable to all that are sicker than some set exogenous minimum acceptance level of sickness. (Barmby, Sessions, and Treble 1994)

Considering the outside option, firms will pay sick pay to all that are sicker than or as sick as the minimum exogenous acceptance level of sickness. And since workers prefer absence for all cases when they are sicker than, or as sick as the minimum exogenous acceptance level of sickness, or when they are indifferent between working and not working, and given the asymmetry of information of health, people have an incentive to overstate their true sickness and will there-fore do so. This kind of behaviour may be very costly for a firm and firms may perhaps start increasing monitoring. (Barmby, Sessions, and Treble 1994)

2.3 The model

To determine which variables affect the level of absents from work due to sick-ness, this paper will use ten explanatory variables. Therefore, a multiple regres-sion model will be used and since the paper intends to analyse the change in the dependent variables to the change in the independent variables, i.e. the elasticity, a log-linear regression model will be used.

Thus the model is constructed as follows:

Equation 1 E e M m L l H h I i I i I i C c G g Un un s S ln ln ln ln 60 ln 60 10 ln 10 3 ln 3 ln ln ln ln β β β β β β β β β β β + + + + + + + + + + =

Where S is the dependent variable sick days, Un is the unemployment variable,

Gis the generosity index variable, Cis the days before compensation variable,

3

I is the index 3 variable, I10is the Index 10 variable, I60is the Index 60-120 variable, His the total health on expenditure variable, Lis the life expectancy variable, M is the infant mortality variable, and Eis the employees between 54-65 variable.

(15)

3 Empirical

framework

3.1 Descriptive statistics

When estimating the hypothesis, one has to start off by deciding which variables to include. The intended model was to consist of one dependent variable and ten independent variables. The independent variables were chosen based on earlier studies and our own assumptions of what could have impact on the dependent variable.

3.1.1 Dependent variable

The number of sick days per year and employee (the dependent variable) is coming from the same source, OECD Health Data 2007. The reason for this is simply that the OECD is the only institution reporting the specific data in the form that is used in this project. Only using one source for this data is associated with some limitations, mainly that there are an imperfect number of countries that measure the average number of sick days thus the sample tends to be too small for a significant cross sectional data analyze. Another drawback in the data is the year of the statistical measures, the data is derived from the OECD report from 2007 but not all countries are making measurements each year. To enlarge the sample as much as possible without taking all the reliability the sample data reaches from last year (2005) to ten years before (1995).

Country Year Average Sick Days Per year and employee.

Australia 2004 6.9 Austria 2005 11.5 Belgium 1995 7.1 Canada 2005 7.8 Czech Republic 2005 22.4 Denmark 2005 10.2 Finland 2005 8.6 France 2005 8.4 Germany 2005 13.8

(16)

Greece 2002 5.3 Hungary 2005 13.4 Italy 1999 5.7 Korea 2005 2.7 Luxemburg 2005 11.1 Netherlands 2005 12.0 Portugal 1996 6.3 Slovak Republic 2003 29.2 Spain 2003 18.6 Sweden 2005 20 Switzerland 2002 10.6 UK 2005 6.6 US 2005 4.3

Table 1 Sick days OECD health data 2007

When doing a standard analysis of the data set one can see that the standard viation (SD) from the mean is 6, 50 (table 3). Two countries fall outside one de-viation from the mean. This implies that about 73% (16 countries) has a dede-viation less or equal to one SD from the mean. Further only one country (Slovak repub-lic) lies outside two SD from the mean, thus 95, 5% lies within two SD from the mean. No value can be found outside 2 SD from the mean and the data is nor-mally distributed according to the central limit theorem.

3.1.2 Independent variables

The ten independent variables were gathered from different sources. For the EU countries the index data is derived from “Security programs throughout Europe” (2006). Most European countries have public health insurance which makes the data derived from the report good quality and since it is published by a govern-ment source (Social security administration). Unemploygovern-ment numbers for the European countries are collected from Eurostat (2006 values), the non European countries unemployment statistics are from CIA world fact book (2006 values). The average days waiting for compensation is also derived from the “Security programs throughout Europe” (2006). The average spending of GDP, life expec-tancy and infant mortality are all from OECD health data (2007) and the values are from 2006.

In short the independent variables are the following:

(17)

• Generosity index (O&R). The O&R generosity index in the data set refers to the index computed by Rigmar Osterkamp and Oliver Röhn in the Os-terkamp, R and Röhn, O. (2007) “Possible Explanations for Differences of Sick-leave Days Across Countries” CESifo Economic Studies, Vol. 53, 1/2007, 97–114 report. This index is constructed by measuring seven dif-ferent variables on a 0-1 scale where 1 is the most generous and 0 the least thus ending up with a generosity index between 0 and 7. The seven variables accounted for are:

! Is there a waiting period (and of how many days), after which sick-leave pay starts?

! Is there the possibility of self-certification for being sick (and for how many days)?

! The official sickness certificate is issued by whom—by the patient’s own doctor or by an independent examining doctor who works on behalf of the employer or the sickness fund?

! In case of sickness absence, how long does the employer continue to pay the salary, and is there any reduction? (Two variables)

! In case of sickness absence, how long does the sickness fund continue to pay the salary, and is there any reduction? (Two variables)

Country Generosity Index

(O&R) Australia 4.10 Austria 5.46 Belgium 4.38 Canada 3.52 Czech Republic 5.15 Denmark 5.40 Finland 2.60 France 5.24 Germany 6.11 Hungary 4.75 Netherlands 3.40 Portugal 4.75 Slovak Republic 5.00 Spain 4.75 Sweden 6.73 Switzerland 5.09 UK 3.87 US 2.70

(18)

Osterkamp and Röhn (2007)

• Days before compensation. This is the number of “waiting” days before the sick leave compensation begins.

• Index (3), (10), (60-120). These are computed indexes that measure the monetary compensation during different absence periods. The indexes are computed by using a hypothetical salary of 100 units of money per day, then calculate how many units are still received the first 3, 10 days and the 60th to the 120 th day of absence.

• Total Expenditure. On health care as % of GDP. This variable represents the money spent on health care in each country as a percentage of GDP; both private and public expenditure is included.

• Life expectancy and infant mortality. The life expectancy in years and in-fant mortality per 1000 births.

• Employees 55-64 years of age (% of tot) the employees between 55 and 64 years of age as a percentage of the total number of employees.

Variables Mean Min Max Standard Deviation

Sick Days 12.09 4.30 29.20 6.50

Unemployment 2006 (%) 6.87 3.30 13.40 2.51 Generosity Index (O&R) 4.61 2.60 6.73 1.10 Days Pefore Comp. 1.11 0.00 7.00 1.90 Index (3d) 51.96 0.00 100.00 38.57 Index (10d) 69.17 35.00 100.00 21.10 Index (60-120d) 69.61 50.00 90.00 10.92 Tot exp on health care as % of GDP 9.25 6.90 15.30 2.00

Life Expectancy 78.66 72.80 81.30 2.32 Infant mortality 4.49 2.40 7.20 1.28 Employees 55-64 years of age (% of tot) 12.45 8.10 19.20 2.83

Table 3 Descriptive statistics

3.1.3 Construction of an index

The test hypothesis, and the nature of the data, demands a unifying preparation of the numbers. To be able to use different types of data in the regression an in-dex variable that reflects sick benefit generosity have to be constructed. The con-struction of an index will always include some level of subjectivity, since we the authors have to create it.

When looking at the non EU countries the programs are different, most non EU countries in this study have very different policies. The major difference is the

(19)

lack of unified health insurance policy. These countries mostly have private in-surances or a mixture between private and public. There lies a problem; since different policy makers (insurance companies) provide unjust benefits the data does not fit the model. Other countries like Korea becomes problematic as there are few or no publications regarding the health insurance and sick benefits. Given this, some of the countries had to be removed in the process of creating the indexes. However the remaining countries should provide enough informa-tion to be able to include the index in the correlainforma-tion.

Both the short term and long term indexes will be created through aggregating the two measures of economic generosity in the sick benefit insurance, the level of compensation and the waiting time before compensation starts. The short term index will be based on data from the first three and the first ten working days of absence and the long term will be from the 60th

to the 120th

day. All indexes will be calculated through the usage of a hypothetical salary and the percentage of compensation will be the index number for the regression. The long term index is from day 60 which by Swedish standards is the definition of long term ab-sence.

Unfortunately some countries including the UK and US have insufficient data and are excluded from the index creation part of the analasys.

After adjusting the data set and constructing the Index variables the statistics now include 16 countries.

Country Average Sick Days Comp. (%) Days Pefore Comp. Index (3d) Index (10d) Index (60-120d)

Austria 11.50 100 0 100 100 70.0 Belgium 7.10 60 0 60 60 55.0 Czech Republic 22.40 25-69 0 25 69 69.0 Denmark 10.20 100 0 90 90 90.0 Finland 8.60 100/70 0 100 100 70.0 France 8.40 50 3 0 35 50.0 Germany 13.80 100 0 100 100 70.0 Hungary 13.40 70 0 70 70 70.0 Netherlands 12.00 70 0 70 70 70.0 Portugal 6.30 65 3 0 46 67.5 Slovak Republic 29.20 25-55 0 25 46 55.0 Spain 18.60 60 3 0 42 75.0 Sweden 20.00 80 1 53 72 80.0 Italy 5.70 50-66 3 0 35 66.6 Luxemburg 11.10 100 0 100 100 100.0 Greece 5.30 50-70 3 0 35 50.0 Table 4 Index

(20)

3.2 Correlation matrix

In order to identify whether the independent variables had any significant impact on the dependent variable, the independent variables where tested in a correla-tion matrix. The results from the correlacorrela-tion presented in table 5. There is a mul-ticollinearity problem between our index variables (index 10 and index 3); due to this the index 10 variable will be dropped from the regression analysis.

Sickdays Unemp GenIndex DaybeforeComp Index3 Index10 Index60-120

Totexphealth LifeExp InfantMort Employ5465 Sickdays 1 .561(*) .501(*) -.249 -.179 -.179 -.097 -.579(*) -.424 .061 -.123 Unemp .561(*) 1 .261 -.035 -.326 -.455 -.661(*) -.357 -.396 .058 -.377 GenIndex .501(*) .261 1 -.027 -.073 .005 .090 -.153 .005 -.383 .065 DaybeforeComp -.249 -.035 -.027 1 -.632(**) -.505 .008 -.042 .370 -.085 .011 Index3 -.179 -.326 -.073 -.632(**) 1 .926(**) .441 -.009 -.085 -.027 .221 Index10 -.179 -.455 .005 -.505 .926(**) 1 .540(*) -.055 .018 -.190 .356 Index60120 -.097 -.661(*) .090 .008 .441 .540(*) 1 -.234 .137 -.142 .624(*) Totexphealth -.579(*) -.357 -.153 -.042 -.009 -.055 -.234 1 .341 .153 .039 LifeExp -.424 -.396 .005 .370 -.085 .018 .137 .341 1 -.529(*) .316 InfantMort .061 .058 -.383 -.085 -.027 -.190 -.142 .153 -.529(*) 1 -.436 Employ5465 -.123 -.377 .065 .011 .221 .356 .624(*) .039 .316 -.436 1

Table 5 Correlation matrix (*) = significant at 0.05 level (**) = significant at 0.01 level

3.3

Regression analysis

After dropping the Index 10 variable due to multicollinearity we also had to drop the other two Index variables along with the variable for days waiting before compensation starts. These variables were taken from the regression data be-cause the regression equations measures how change in the independent

(21)

vari-ables alter the dependent variable i.e. we needed to analyze the logged varivari-ables. Since our Index variables and the waiting variable included values of 0 and therefore impossible to log they were dropped from the analysis.

After removing these variables the regression included the logged variables of; Sick days, Unemployment, Expenditure on health as percentage of GDP, Life ex-pectancy, Infant mortality and the employment of people between 54 and 65 years of age.

3.3.1 Regression output

Adjusted R-squared = 0.515

Unstandardized Coefficients T statistics

B B

Constant 4.951 .269

Unemployment .124 .419

Generosity Index 1.032 2.778

Total expendirure on health -1.465 -2.700

Life expectancy -.356 -.087

Infant mortality .220 .461

Employed 54-65 years of age .063 .130

Table 6 Regression output with Sick days as dependent variable Dependent Variable: Sick days

The significant independent variables are the generosity index and the the % of GDP spent on health care. The remaining independent variables are not signifi-cant at the 5% level due to weak p-values.

The coefficient for the generosity variable is 1,032 and the coefficient for the ex-penditure on health is -1,465.

(22)

4 The Swedish experiment

The unemployment rate was proven to have significant correlation with the aver-age number of sick days but did not show significance in the regression analysis. To further analyze the impact of the unemployment rate in a single country sce-nario we choose to do a regression analysis with the Swedish unemployment rate and the Swedish absence numbers. This will be a time series analysis from 1990 to 1995. The table below shows the sick day and unemployment data used in the analysis.

Year Sick days Unemployment % LN SD LN UE

1990 24.1 1.80 3.18 0.59 1991 22.5 3.30 3.11 1.19 1992 20.0 5.80 3.00 1.76 1993 18.0 9.50 2.89 2.25 1994 17.0 9.80 2.83 2.28 1995 16.0 9.20 2.77 2.22 1996 15.0 10.00 2.71 2.30 1997 15.0 10.20 2.71 2.32 1998 16.0 8.50 2.77 2.14 1999 19.0 7.20 2.94 1.97 2000 22.0 5.90 3.09 1.77 2001 25.0 5.10 3.22 1.63 2002 27.0 5.20 3.30 1.65 2003 26.0 5.80 3.26 1.76 2004 24.0 6.60 3.18 1.89 2005 20.0 7.80 3.00 2.05

Table 7 Sweden descriptive statistics

(23)

Regression output:

Adjusted R Square = 0.4614

Coefficients t Stat P-value

Intercept 3.57 22.59 0.000

LN UE -0.31 -3.72 0.002

Table 8 Sweden Regression

Dependent variable = Sick days.

In this regression the unemployment variable is significant at the 0.5% level and had a negative impact on the average sick days. The result indicates that the ab-sence frequency decrease as the level of unemployment increase. Since people tend to be more protective of their jobs when the unemployment is high this re-sult it not very surprising. More unexpected is the rere-sult from the first regression that included more countries where the unemployment data was insignificant.

5 Analysis

By using cross sectional data this paper intended to analyze how changes in the compensation level granted to a person on sick leave would affect the aggregate number of sick days in an economy. The intended analysis was to use the num-ber of sick days as a dependent variable and the compensation level as the only independent variable. A problem that revealed itself early in the study was the lack of data available especially when it came to the compensation level in each country. Since there were too few variables in order for the study to actually give any results, it was clear that more variables had to be accounted for. The model ended up using 10 independent variables and still using the initial variable sick days as the dependent variable.

5.1 Interpreting the regression results

The linear equation based on equation 6 in theoretical framework, only keeping the significant variables:

Equation 2 U x B x B A Y = + 1 1+ 2 2+

Where Y is in average number of sick days per year and employee in a given economy, B1is the regression coefficient for the index variable, B2is the regres-sion coefficient for the percentage of GDP spent on healthcare variable and

2 / 1 x

x is the change in the independent variables respectively. U= residual vari-able.

As seen from the regression output we find a negative relationship between the percentage of total GDP spent on healthcare and a positive relationship between generosity and average number of sick days.

(24)

5.1.1 Percentage of GDP spent on health care

The negative relationship in the regression output indicates that there is a posi-tive relationship between the health care spending and the average number of sick days, i.e. if the government increases the spending on health care the aver-age sick days decrease. In this case the relationship is significant suggesting that a 1% increase in health care spending as percentage of GDP would decrease the sick days with about 1.5%. This result is good because it proves that the govern-ment could indeed lower the average sick days by making health care better and more available. This result however has to be discussed because of the complex-ity of the independent variable.

The total percentage of GPD spent on health care reflects both the private and the public spending on health care. What has to be considered in this case is the fact that some countries have very large expenditure on health care as percent of GDP but the actual health care is not very great or is not received by the whole population. To illustrate the point let us consider two countries Sweden and the US. Sweden spends 9, 1% of total GDP on health while the US spends 15, 3% of total GDP on health, (2007) OECD Health data. These numbers are hard to com-pare when adding that 84,6% of the Swedish expenditure is public spending while the US public spending on healthcare contribute to only 45,1% of the total outlays. Further, as concluded earlier in the paper, the general health conditions in the US are worse than in Sweden, meaning that there is no direct relationship between the level of spending and the public health in a country. This combined with the facts that about 16% (2004) (US census bureau) of all Americans are un-insured meaning that if they don’t show up to work there is no compensation at all. This suggests that in the US less people spend more regarding health which in turn says little about the general health conditions. These conditions make in-terpretation of the result difficult.

These implications suggests that there are more to it than what the data tells us, one of the most likely being that there are behavioral differences and there might be a moral hazard problem. The moral hazard would come from the fact that people tend to be less careful when they know that someone else is going to pay their salary when they are not at work. If you, as in the US, have private or no health insurance at all you personally at some point have to pay for your ac-tions.

5.1.2 The generosity index

The level of generosity has a positive relationship with the number of sick days in a country, meaning that if one would increase the generosity in the sickness benefits the average number of sick days will increase, in this case the sick days would increase with about 1.032 % for every 1% increase in the generosity. Since this index is generated through seven different variables it is hard to pin down exactly what has the greater impact on the absence. The conclusion is that the overall generosity in the institutional variables indeed has effect on the sick ab-sence. The constructors of the generosity index, Osterkamp and Röhn, had a lar-ger regression coefficient on this variable suggesting biglar-ger impact on sick leave when changing variables in the index.

The positive relationship between the generosity index and number of sick days is also concluded by Pålsson (2006). Pålsson found that when institutional rules

(25)

in Sweden concerning issuing sickness certificates were less strict more people tended take out sick days, and when these rules became stricter the sickness ab-sence level decrease. Pålsson also found a relationship between the compensa-tion level and the absence level.

Another thing that would be interesting for the discussion is the possibility of “normal” sick leave; people in different countries all have individual preferences and utility functions when choosing between work and leisure. Some people stay home from work when perfectly healthy and others go to work when sick, probably one could find a value of generosity where a sort of equilibrium is ob-tained. This would be where as few as possible had to work while sick but where the difference between the salary and the compensation for being sick is still substantial enough to minimize non-sick related absence.

Just as L. F. Dunn; Stuart A. Youngblood (1986) pointed out the MRS between work and leisure affects the level of absence in the workplace. In our case the generosity represents the utility contribution of staying home and the salary of the one working. If you change the utility of staying at home the MRS change and rational behaving people maximize their utility by spending more time at home though maybe able to work.

6 Conclusion

The author’s conclude that the null hypothesis i.e. “the generosity in social secu-rity does not affect the average number of sick leave days positively” can be re-jected.

The analysis revealed that there is a positive link between the Generosity index and number of sick days in each country. The regression showed that an increase in the generosity index also increased the number of people going on sick leave. Since the coefficient for the generosity variable were greater than one a 1% change in the monetary security on the labor market i.e. the generosity index corresponds to a greater than 1% increase in the average sick days per year and employee.

The authors also found support for the fact that total expenditure on health care has a negative relationship with the number of sick days in a country. That is, an increase in total expenditure on health decreases the number of people on sick leave.

6.1 Author’s remarks

The author’s did find a strong relationship between the spending on healthcare and the average sick days in a country. Since this was not the main targeted issue in the paper it is a conclusion with reservations for errors. As presented earlier in

(26)

the paper there seems to be no relationship between the general health situation in a country and the sick absence, a look on the data also tells us there is a seemingly insignificant relationship between the spending and the general health status. Why the relationship between spending and sickness absence is so strong in the regression is therefore somewhat surprising and can not be explained in depth without further research.

7 Future

studies

The hardest part when writing this paper was the data collection, databases such as Euro-stat and OECD health data provided much data on the subject, however it was far from complete, i.e. the data did not include all of the OECD and EU countries. Therefore, it was very hard to find a sample big enough to be regarded as an accurate quantitative study. Furthermore, when analyzing other papers on this subject it was clear that the data within the papers did not match each other and were therefore not very reliable. A probable explanation to that could be that they were presented with the same problem of not having enough data, and were therefore forced to manipulate it.

Thus, it would be very helpful when writing a paper on absenteeism to able to find reliable and complete data on the subject and we therefore propose further data collection on this subject.

The Generosity Index presented by Röhn and Osterkamp could also be analyzed further perhaps by giving more weight to the different variables used in the index , i.e. giving the waiting period before receiving sick days a higher importance then how long an employer has to pay sickness benefits. By dividing these

(27)

per-haps one can find one variable that is more significant than the others and thereby establishing which variable is “more important” to study than the others. Even though the regression did not show any significant results when it came to the level of compensation when absent one could assume that there is a strong link between the average number of sick days and the compensation level. A probable reason for the variable not working in this analysis is the construction of the index. The index made sense to some extent since they included the wait-ing period and the average level of compensation acquired when on sick leave. By manipulating the compensation variable it is very likely that the variable could be used as an explanatory variable.

Another way to approach the problem would be to try to include the behavioral differences between the people in the different countries. Some economies have a more pronounced work ethic than others; also the different strength of labor unions should have impact on the average sick days in a country.

8 References

Allen, S. (1981), “An Empirical Model of Work Attendance”, Review of Economics and Statistics 63, 77-87.

Andrén, D. (2003) “Sickness related absenteeism and economic incentives in Swe-den: A history of reforms.” CESifo DICE Report 3/2003.

Barmby T, Session J, Treble J (1994), “Absenteeism, Efficiency Wages and Shirk-ing”, the Scandinavian Journal of Economics, Vol. 96, No. 4, pp. 561-566.

Buzzard, R.B. and W.J. Shaw (1952): “An Analysis of Absence under a Scheme of Paid Sick Leave”. British Journal of Industrial Medicine, 9, 292-295.

Coles, M. and Treble, J. (1996), “Calculating the Price of Worker Reliability”, La-bour Economics 3, 169-188.

Chatterji, M. and Tilley, C. (2002), “Sickness, Absenteeism, Presenteeism, and Sick Pay”, Oxford Economic Papers 54, 669-687.

(28)

Drago, R. and M. Wooden (1992), “The Determinants of Labor Absence: Economic Factors and Workgroup Norms across Countries”, Industrial and Labor Relations Review, Vol. 45, No. 4, pp.764-778doi:10.2307/2524592

Dunn L.F. And Youngblood S.A. (1986) “Absenteeism as a Mechanism for Ap-proaching an Optimal Labor Market Equilibrium: An Empirical Study”, Reweiv of economics and statistics volume 68 nr 4

Eurostat (2004), Work and health in the EU A statistical portrait 2004 edition Frick, B. and Malo M.A. (2005), ‘‘Labour Market Institutions and Individual Ab-senteeism in the European Union’’, Mimeo, Faculty of Management and Econom-ics, Witten/Herdecke University.

Holmlund, B. and Engström P. (2005), “Worker absenteeism in search equilib-rium”, CESIFO WORKING PAPER NO. 1607

Jans AC, (2003) ”Förtidspensioner och långtidssjukskrivningar under 1990-talet: Regler och konjunkturer styr” Välfärd Nr 1 2003

Johansson, P. and Mårten. Palme. (1996). “Do Economic Incentives Affect Worker Absence?”, Empirical Evidence Using Swedish Data. Journal of Public Economics 59(2): 195-218.

Johansson P. And Palme M. (2002) “Assessing the Effect of Public Policy on Worker Absenteeism” The Journal of Human Resources”, Vol. 37, No. 2, pp. 381-409.

Kangas O. (1999) “Social Policy in Settled and Transitional Countries: A Compari-son of Institutions and Their Consequences” Luxembourg Income Study Working Paper No. 196

OECD Fact book (2007) OECD Health data (2007)

Osterkamp, R and Röhn, O. (2007) “Possible Explanations for Differences of Sick-leave Days across Countries” CESifo Economic Studies, Vol. 53, 1/2007, 97–114 Pålsson M. (2006), ”Vad är det som orsakar hög frånvaro frekvens”, Master thesis Economics Södertörns University

(1997), “PREVENTING ABSENTEEISM AT THE WORKPLACE “, European Founda-tion for the Improvement of Living and Working CondiFounda-tions,

(2006), “Security programs throughout Europe”, SSA Publication No. 13-11801September 2006

Social departementet (2006), utdrag ur budgetpropositionen för 2007, Proposition 2006/07:1, Volym 6 Utgiftsområde 10 Ekonomisk trygghet vid sjukdom och handikapp 3.8 Politikens inriktning, utdrag ur sidorna 38-41 (avsnitt som rör sjukpenningen)

(29)

Vikenmark S. And Andersson A. (2002) ”Arbetsmiljön gör oss sjuka”, VälfärdsBulletinen Nr 1

Figure

Table 1 Sick days_________________________________________________________________ 15 Table 2 Generosity index _________________________________________________________ 16 Table 3 Descriptive statistics _____________________________________________________
Figure 1 Life expectancy  Source OECD health data 2007
Table 1 Sick days  OECD health data 2007
Table 2 Generosity index
+6

References

Related documents

Therefore, it is not possible to use a diagnose specific instrument (e.g. it is not possible to use the Quality of Life after Brain Injury instrument for other injuries

Two papers in this thesis (I and II) were based on the project ‘Women’s health and living conditions’ (WHL). WHL is a multipurpose study of employed women 17 to 64 years old.

Key Words: vertical fiscal externalities, sickness absence, sickness benefits, health care expenditure, fixed budgets, pharmaceuticals, cost containment, dy- namic panel data

To conclude, given the assumptions made, an increase in the federal tax rate or the sickness bene…t reduces the state’s incentive to cure sick individuals and increases its

The studies included cover, how economic incentives matter for the outcome of sickness absence, what determines return to work for individuals with a stress related

The aims of this thesis were: (1) to evaluate the Swedish versions of the Inflammatory Bowel Disease Questionnaire (IBDQ) and the Rating Form of IBD Patient Concerns (RFIPC); (2) to

The effect of parenthood on the income gap in opposite-sex couples is smaller when birthmothers and partners have similar labor market incomes two years prior

Students attending schools with higher levels of teacher-rated school ethos had a lower likelihood of reporting recurring sickness absence compared with those attending schools