+ All Categories
Home > Documents > Regional Unemployment and Individual Satisfaction

Regional Unemployment and Individual Satisfaction

Date post: 15-Nov-2023
Category:
Upload: uni-flensburg
View: 0 times
Download: 0 times
Share this document with a friend
12
Transcript

Bibliografische Information Der Deutschen Bibliothek

Die Deutsche Bibliothek verzeichnet diese Publikation in der Deut-

schen Nationalbibliografie; detaillierte bibliografische Daten sind im

Internet über <http://dnb.ddb.de> abrufbar.

Metropolis-Verlag für Ökonomie, Gesellschaft und Politik GmbH

Bahnhofstr. l6a, D-35Q37 MarburgCopyright: Metropolis-Verlag, Marbury 2004http://www. metropolis-verlag.deAllrights reservedPrint: Rosch-Buch, Scheßlitz (Germany)

ISBN 3-89s18-454-3

Contents

Gerd Grözinger, Anne van AakenIntroduction; Inequality: New Analytical Approaches ............... ..........7

Georg P. MuellerEntropy Measures for Comparative Analyses of Social Justiceand Discr iminat ion . . . . . . . . . . . . . . . . . .15

Dierk HirschelDo high incomes reflect individual performance? ...............................33

Jobannes Schanrze and Marco HarpferIncome Inequality, Redistribution and Vell-Being.Evidence From Longitudinal Data on Life Satisfaction.......................61

Gerd. Grözinger and Wenzel MatiaskeRegional Unemployment and Individual Satisfaction .........................82

[Jvte Blien, Phan thi HongVanOccupational wage differentials ................105

Andrea VleifbrichPolarisation of the Vorld Income Distribution ............121

Thomas ZiesernerMultiple-steady-state growth models explainingtwin-peak empirics?......... ......145

86 Johannes Schwarze and Marco Härpfer

Self employed

Unemployed

Household income(log)

Pre-government house-hold income (log)

Lowest income quintile

Second income quinti le

Third income quinti le

Founh income quintile

Household receivessocial assistance

Pre-GovernmentInequaliry (IPRE)

Percent Redistribution(STATE)

Number ofquestionnaires filled in

-0.03s5

(0.0266)

-0.3687""

(0.022e)

0.2334""

(0.0256)

0.0322".

(0.0056)

-0.0721."

(0.0365)

-0.0233

(0.0268)

-0 .01 15

(0.0212)

0.0041

(0 .0161 )

-0.1 1 8 1. . .

(0.0352)

-U.JJ+U

(0.r067)-0.0323

(0.081 5)

0.0038..

(0.001e)

-0.0103

(0.01e3)

-0.3856"'"

(0.0188)

0.24s2"..

(0.022s)

0.03 1 8..

(0.0046)

-0.0705

(0.0321)

-0.0276

(0.0233)

-0.0014

(0.0185)

0.0193

(0.0148)

-0.0481"""

(0.0308)

-0.4506."

(0.0e40)

-0.1 036

(0.0761)

0.0073""

(0.0022)

-0.0734'

(0.0328)

-0.7054.."

(0.042e)

0.4267"""

(0.0457)

0.0652.".

(0.0105)

-0. 1 048'

(0.0636)

-0 .01 15

(0.046e)

-0.0016

(0.0365)

0.0216

(0.0272)

-0.1925.".

(0.0668)

-0.4710.".

(0.1 828)

-0.1 039

(0.1 368)

0.0090.".

(0.0031)

-0.0212

(0.0284)

-0.5269"""

(0.0263)

0.3 193"-.

(0 .0311 )

0.3869'..

(o.oo66)-0.0494

(0.0440)

-0.0018

(0.0322)

0.0203

(0.02s6)

0.0424""

(0.0203)

-0 .0151

(0.0424)

-0.5015.""

(0.1273)

-0 .1841"

(0.1 024)

Regional Unemploymentand Individual Satisfaction

Gerd Grözinger and Wenzel Matiaske

The negative effects of unemployment on those affected and their

families have been a standard subiect of research in the social sciences

since the 1920s, when in their renowned Marienthal study, Marie Ja-hoda, Paul F. Lazarsfeld, and Hans Zeisel explored the consequences

of joblessness on a region as a whole (Jahoda/Lazarsfeld/Zeisel 1960).

Since the publication of this prominent case study, the various indi-

vidual impacts of unemployment have been thoroughly investigated.

The regional component, however, has only recently become the sub-

iect of new research.

Based on an empirical analysis of the Socio-Economic Panel

(SOEP, a wide-ranging representative longitudinal study of private

households in Germany), we will explore the impact of regional un-

employment on individual satisfaction in Germany at the end of

1990s. The investigation is not limited to unemployed individuals and

their immediate environment, but includes the entire regional popula-

tion.

This article is comprised of four parts. It begins with a short over-

view of the current state of research on the externalities of unem-

ployment seen from the viewpoint of current economic thought. This

overview is followed by an explanation of the underlying data basis

and of the model specification. In the third part we Present our re-

sults, and by way of a conclusion discuss their significance in connec-

tion with existing publications.

Other variables; Fixed time effects (1 4) , fixed" Statistically significant at the 10% level, ""

level.N of cases: 94528/ 11838.Source: GSOEP. 1985-1998.

regional effects (73).at the 5"/" level.

--" at rhe 1To

88 Gerd Grözinger and Venzel Matiaske

1. Externalit ies of Unemployment in the Economic Discussion

According to the standard formulation of economic theory, involun-tary unemploymenr is a state that entails a reduction in income andtherefore involves a loss of uti l i ty for those effected. \(hile this is cer-tainly not false, the focus is now shift ing increasingly to additional,non-financial consequences of unemployment. An aspect that psy-chologists and sociologists have long been investigating - the seriouseffects of unemploymenr beyond the loss of income - is slowly be-coming part of economic theory.

This new interest is due not only to the persistence of the phe-nomenon of long-term mass unemployment in many countries of the\Testern world. A methodological shift also explains this new focus;in current economic theory, the measurement of satisfaction based onsurveys is increasingly accepted as a valid data basis. In 2001 for in-stance, the Journal of Economic Bebaaior and Organization issued aspecial edition focused on "subjective \üell-Being and EconomicAnalysis." In 2002 Bruno Frey, the most quoted economist of theGerman-speaking world, wrore a book with Aloys Stutzer entit ledHappiness and Economics, mainly based on survey resuhs (Frey/Stutzer 2002). In 2003, an international conference on the subject washeld in Milan: The Paradoxes of Happiness in Economics (http://dipeco.economia.unimib.it lhappiness/program.htm).

Vhile psychologists and sociologists usually collect their own dataon satisfaction, economists tend ro use existing data records, in par-ticular, the SOEP and its American and Brit ish equivalents. The firsttype of new investigations on the externalit ies of joblessness beyondthe loss of income is also based on SOEP data. These studies explorethe extent to which the decrease in l ife satisfaction caused by unem-ployment actually coincides with the decrease in income. Studies car-ried out on the basis of the SOEP have shown that for men of em-ployable age in Germany, rhe income dimension is only a minor rea-son for the sharp decrease in satisfaction (Vinkelmann/\Winkelmann1998). The "felt" impact of unemploymenr is much higher than itsmeasured financial impact. By l imiting the discussion ro rhe decreasein income, the scope of problems involved with the loss of employ-

Regional Unemployment and Individual Satisfaction 89

ment is highly underestimated (see also ('Süinkelmann/\flinkelmann

1998), (Blanchflower/Oswald 2000).

Vhile this approach examines additional impacts on the directly af-fected individuals at the time of their unemployment, a second type of

study expands the investigation along the temporal axis. Again using

SOEP data, these studies investigate to what extent previous spells ofunemployment have a long-term effect on a person even after his or

her successful return to working l ife (Clark/Georgell is/Sanfey 2001).In this group as well, most particularly in men, a strong negative im-

pact on well-being can be observed. This goes beyond the "hysteresis

effect," whereby unemployment results in a loss of competence andself-confidence, making it increasingly diff icult for the person af-

fected to find a new job. Clark et al. also find subjective aftereffects so

profound that the emotional wound caused by dismissal does not heal

for a long time, even after f inding a new job and new source of in-

come.

A third type of research extends the search for externalit ies to the

effects of unemployment beyond the individuals directly affected. For

this purpose, studies were carried out in twelve European countries as

well as in the United States on the "felt" effects of inflation or unem-

ployment (Tella/MacCulloch/Oswald 2001). The authors took into

account individual characteristics as well as macro-economic devel-

opments from the 1970s to the 1990s. The results show that although

both inflation and a rise in unemployment lead to reductions in over-

all l i fe satisfaction, the negative impact of a one percent rise in unem-

ployment is about one and a half t imes the impact of a one-percent

rise in inflation.

Using a different approach, we want to once again verify these re-

sults, that is, the decrease in l ife satisfaction as a result of unemploy-

ment not only in unemployed individuals but also in the entire popu-

lation. Vhile Di Tella et al. used the temporal axis, we have based our

approach on the regional dimension. So far, only a few studies on the

effects of regional unemployment on satisfaction have been published

(Turner 1995), (Clark 1999), (Stutzer/Lalive 2001). \üe wil l discuss in

our findings whether the results for similar studies in the United

States, Great Britain can be replicated in Germany. First of all, how-

ever, we want to present our data basis and our model.

[ . ' s 1 6 ,

90 Gerd Grözinger and.Wenzel Matiaske

2,Data Basis and Model

The socio-Economic paner made availabre by the German Institutefor Economical Research, Dr\ür fDeutsches Institut für'\ü/irtschafts-forschung] forms the data basis o] ou. study. The paner is

" ,.gut".ry

held household survey in German and one of the best d"t" .".oid, fo,social science research; approximatery 2,2eopublications are based onthis data (http:/ /panel.gsoep.delsoeprit/). The individuar data and,/orhousehold data of the SoEp ."n

"rro b. iink.d to data

"" *. ,.gi.""r

srructure compiled by the Federal Institute of Regional ReJearch(Bundesamt für Bauwesen und Raumordnung 2oor).The institute hasdivided Germany into 97 so-called ..planning regions,,, each of whichforms a relatively homogenous settlement area.

The local rares of unemployment in each region are also recorded.fh.: 1..gi:""1unemployment figures refer to the end of tgggwhile theindividual details of the SoEp refer to the beginping of the year 2000.'$fle slightly modify three aspects of the data records. rirst of all, indi-viduals who changed households between 1999 and 2000 were ex_cluded, since the DIV, which carculates the household income, alwayspublishes its data wirh a one-year delay to facilitate -or. .o*jlex co-herence tests. The 2007 data was not yet available at the time of thepresenr study. since the incomes for 1999 were reallocated accordingto number of household members in 2000 by rhe DfV, we requiredsteady population figures for our study. Secondry, households with a'income of under DM 5,OOO per year and per calculated person werenot raken inro accounr due to the lack of frausibility. I' ord.. to ex-clude excessive effects caused by the statistical

"rro-.ly of very high

incomes, possible due to the regional fine structuring, the _"*i_u_income per calculated person was limited to DM 25o,ooo. This left uswith 12,000 cases.

For the purposes of the above-mentioned study, individual we[-being is the central dimension of rhe impact caused by regional unem-ployment. This is operationalized in several different wiys. clark etaI. (1999) as well as srurzer and Lalive (2oor) use overall iife satisf""-tion as their target variable. Turner (lgg5), on rhe other hand, con_centrated on health satisfaction in particular.

Regional Unemployment and Individual Satisfaction 91

The SOEP provides us with data on both variables. Overall life sat-isfaction was measured with the question: "Generally speaking, howsatisfied are you currently with your life?" The answer was marked on

an eleven-point scale with 0 signifying the lowest, 10 the highest level

of satisfaction. The perceived state of health was evaluated with thequestion: "How satisfied are you with your health?" This question

was also answered using an eleven-point rating scale.

\ühile overall life satisfaction is the standard variable in studies on"Happiness and Economics," there are two important reasons for the

additional investigation of health satisfaction. For one thing, the ex-

tant literature shows that fear of unemployment may cause psycho-

somatic disorders (Feather 1990), fVitte 1999). In addition, this indi-

cator can be regarded as highly reliable. The data recorded on per-

ceived state of health correlates highly with "hard" data like absentee-

ism, hospital stays, and mortality rate (Idler/Benyamini 1997). The\üflorld Health Organization even announced that in countries where

collecting "hard" data is more difficult, surveys on perceived health

status would be initiated instead (Schwarze/Andersen/Anger 2000).

Health satisfaction is thus a more solid indicator than overall life sar-

isfaction. Based on the correlation of the two variables examined as

described in the existing literature, however, we do expect predicators

to have parallel effects (Praa{Frijters/Ferrer-i-Carbonell 2001). \Ve

have constructed a uniform model based on the basic similarity be-

tween the two indicators and the resulting possibilities of comparison.

The negative effects of regional unemployment on individual well-

being can be attributed to various causes. \üorkers in regional crisis

areas have to be afraid of losing their jobs. Those unsatisfied with

their present employers find it more difficult to change jobs. The so-

cial climate at the workplace may deteriorate, involving cuts in cus-

tomary non-wage benefits such as continuing education plans

(Martin/Düll 2000). In Germany, regional unemployment also has

been found to exert a negative pressure on regional wage levels

(Bellmann/Blien 2001). A high regional unemployment rate can thus

also have a negative impact on those members of a family who are not

actually unemployed themselves. The question whether or not this

explains the presumed negative impact of regional unemployment on

well-being will be discussed in the last part of this article.

92 Gerd Grözinger and Wenzel Matiaske

Our three starring hypotheses therefore can be summarized as fol-lows:

- H l: Regional unemploymenr causes a decrease in overall l i fe satis-faction.

- H 2: Regional unemploymenr causes a decrease in heafth satisfac-t ion.

H 3: The effects of regional unemployment have a srronger impacton health satisfaction than on overall l i fe satisfaction.

Vhile the verif ication of H 1 and H 2 is relatively simple, the thirdhypothesis requires further considerations for operadonalization.\When including the additional variable of regional unemploymenr, astronger effect can be seen as a relatively higher rise of the varianceincrease in health satisfaction as compared to l ife satisfaction (H 3a);alternatively, it can imply a relatively srronger impact of regional un-employment on health satisfaction as compared with l ife satisfaction(H 3b). In the following, the plausibil i ty of both aspecrs wil l be ex-plored.

The other specific aspecrs of the model are based on a long-running discussion on rhe contributory effects on overall l i fe satisfac-tion and on health satisfaction (here collectively referred to as "saris-faction"). This has resulted in a whole series of relevant variables andboth positive or negative effects (Mielck 2OOO), (praaglFrijters/Fer-rer-i-Carbonell ZOOt;, (FreylStutzer 2OO2). They primarily includethe age in years and - due to a knovzn non-linearity of this variable -

the square of the age. Based on rhe already known results, we hy-pothesized that the age would have a negative impact, while the squareof the age would have a positive impact: in other words, ar a more ad-vanced age, rhe decrease in satisfaction should be less marked than isthe case at a younger agc.

The degree of disabil ity has been considered as anorher significantfactor: various medical conditions frequently l inked with a disabil ity;hould result in a decrease of satisfaction. Also relevant in many casess the gender of those affected. Since women generally complain aboutl more fragile state of health and therefore go to the doctor more of-

Regional Unemployment and Individual Satisfaction 93

ten, but have been coded O in this case' the male gender variable of 1

should have a plus sign at least where health satisfaction is concerned.'\üe

also hypothesized that the level of education, measured by the

number of years of schooling, would have a positive impact. A higher

level of education is accompanied by a number of positive effects,

such as physically less demanding jobs, a higher degree of information

if problems occur, and a more health-conscious l ife-style. Another

personal feature included is the family dimension, i 'e. the fact whether

or not an individual is married or l iving in a steady partnership. An ex-

isting partnership (value : 1) was presumed ro increase satisfaction

due to the emotional stabil ity provided.

The available income is yet another variable to be taken into ac-

count. It was determined on household level and then divided by the

number of individuals l iving in that household. Gains in efficiency,

however, mean thar the larger the household the lower the relative

household expenses. There are various methods ro calculare so-called

equivalence scales that take this effect into account (Hauser 2002)'

We decided to use the "old OECD scale" where the first adult is

weighted as 1, each of the following adults is weighted asO.7, and all

under 18s are uniformly weighted as 0.5. This probably is the scale

most frequently used in references on the subject' As a rule, the use of

alternative compurarions has hardly any impact on rhe result of re-

gression equations using the income as one of the explaining variables.

Whi le income should be expected to have a posi t ive impact on sat-

isfaction, we can take the share of state transfer paymenrs in the in-

come of a household as having a negarive impact. This assumption is

based on the perceived stigmatization l inked to a dependency on srate

benefits. The last personal feature included in the l ist of variables is

the question of personal unemploymenr. In case of an affirmative

answer (again, value : 1), this should naturally have an negative im-

pact on satisfaction.

In addition ro the rare of regional unemployment we have included

other regional characteristics as dummy variables, using general meas-

ures for rhis purpose. The Bundesamt für Raumordnung classified the

regions into seven categories according ro the type of sertlement.

These seven categories are: (1) rural areas with low popuiation den-

sity, (2) rural areas with higher population density, (3) urban areas

94 Cerd Grözinger and Venzel Matiaske

with medium population density without large cenrers, (4) urban ar-eas with medium population density with large centers, (5) urban ar-eas with a higher population density, (6) urban agglomeration areaswith very large cenrers (Z) densely-populated urban agglomerarion ar-eas. By including six dummy variables - with Rural Areas with I owPopulation Density forming the basis - we hope ro record other ma-jor regional characteristics beyond the issue of unemployment. As faras the impact of the regional dummy variables on the variable to beexplained is concerned, we have no assumprions based on theoreticalconsiderations or on other studies.

The use of an econometric model is problematic in rwo respecrs.For one thing, the self-assessment of satisfaction should only be re-garded on an ordinal data-level, even if the 11-point scale is fairly ex-tensive. secondly, our approach is a hierarchical model involving indi-viduals on a firsr level and regions on a second level. The first problemcan be solved with alternatives to the standard OLS estimate, e.g. or-dinal probit or logit models. However, skew distributions in a widelyspread scale, a phenomenon typical in the assessment of satisfaction,result in empry or half-empty cells. Therefore, existing studies such asDi Tella er al. (2001) prefer to use robust OLS estimates in thesecases. In the following mbles, in addition to the mulriple regressionmodel, we wil l show ordinal logit models which use the otherwiseidentical model approach with a target variable divided at the quarti les(Har re l2001 ) .

To solve the second problem we can choose from differenr ap-proaches of hienarchical regression models or multi-level analysis(Snijders/Bosker 1999). The present article additionally reporrs thefindings based on a linear hierarchical two-level model, with the re-gions specified as coincidental effects and the details of those regions(i.e., unemploymenr rate, rype of serrlement) as fixed effects. The ad-vantage of this estimation model is that the degrees of freedom fortesting the impact of the regional variables are correcred. However,the software used for the present investigation neither permits ordinaltarget variables nor does it deliver standard coefficients for the inter-pretation of the variance explained (Pinheiro/Bates 2000).

Regional Unemployment and Individual Satisfaction 95

3. General results

Table 1 shows the results for overall life satisfaction and for health

satisfaction, respectively, based on various model specifications. Due

to the lack of space, only the values for the OLS regression equations

are reported here. Columns I and 4 show the values for the basic

rnodel with the individual variables, but without any regional data.

The model is extended step by step by inclusion of regional unem-

ployment (columns 2 and 5) as well as the regional dummy variables

described above (columns 3 and 6).

The individual variables show the predicted influences: age has a nega-

tive impact while the square of the age has a positive impact; the de-

Table l. Determinants of Satisfaction

Ovemll Life Satisfaction

OLS

Health Satisfaction

OLS

Constant

Age

Ag"'

Degree oidis-abil ity

Sex

Partnership

Education

Equivalised in-come'fransfer

income

[Jnemployment

Regional mte ofunempl.

Regional D2

Regional D3

Regional D4

Regional D5

Regional D6

Regional D7

8.430

-0.067

0.00r

-0.015

-0.007

0.140

0.016

0.000

-0.908

-0.7 I E

8 . 2 1 0 i . *

_0.065 r.r

0 .001 r r r

-0 .0 I 5 r * t

-0.006

0.329 * . r

0 .021 . r

0.000

,0 .918

.0.679

0 . 1 8 7

0.r 20

-0.004

o.222

0.072

0.248

8.540 *rr

-0.064 *rt

0.001 ***

_0.016 *r*

-0.007

0 .319 * * *

0.028 ***

0.000 rrr

_0.872 ***

-0.631 .**

_0.030 ***

0.057

-0.001

-0.066

0 . r l 2

0 .0 r8

0 .1 l 2

8.980 rrr

_0.091 ***

0.001 **r

-0.029 ***

0 .140 r t r

0 .128 t .

0.044 **r

0.000 ***

-0.401 ***

-0.375 ***

0. | 39 r r r

0 . 1 3 1 r r

0"049 rrr

0.000 r*

_0 .43J r * r

_0 .3 I 5 r * *

0.095

0 .186

-0 . I l 7

0 .107

-0.042

0 .1 78

8 .810

-0.090

0.001

-o.029

9.990 *r*

_0.090 *.*

0 .001 r * '

_0.029 rrr

0 . 1 3 9 * * +

0 . 1 2 5 * r

0.053 ***

0.000 r*

_0.408 +**

_0.290 *ri

-0 .017 * * t

0 .021

0 . i l 8

-0. l 53

0.045

-0.073

0.1 02

R, 0 . 1 0 0 * * r 0 . t 0 l 0 . 1 0 5 r r 0 .213 0 .213 * * r o.214 11

Levels ofsignificance: < 0.001 r*t*', < 0.01 '+r', < 0.05 r*', < 0

96 Gerd Grözinger and'Wenzel Matiaske

gree of disabil ity has a negative impact; pa,:tnership, education, andequalized income all have a positive impact, rhe amount of transfer in-come and individual unemployment have a negative impact. All vari-ables are highly significant. The only exception is the sex, which pro-vides a positive value only for health satisfaction while the value forgeneral satisfaction is not significant.

Thus, our first two hypotheses have been fully confirmed: regionalunemployment has a significantly negative impact both on overall l i fesatisfaction (H 1) as well as on health satisfaction (H 2). A higher rateof regional unemploymenr reduces not only general satisfaction, butalso the physical well-being of the population.

The strength of the connecrion is also worrh menrioning. Thestandardized coefficients (not reported here) show: the regional vari-able has more or less the same significance as the level of education orthe income. For the explained variance of the total model ir can besaid that an R2of 0.10 for overall l i fe satisfaction and 0.21 for healthsatisfaction is within the usual scope of such studies (FreylStutzer2002). The higher value in the health dimension can largely be attrib-uted to the inclusion of the disabil ity variable, which has a strong im-pact on health.

Vhere the third hypothesis is concerned, both variety H 3a as wellas 3b are disproved. The inclusion of regional unemployment into thebasic model results in a slightly stronger impact on the explanation ofoverall l i fe sarisfaction than on the explanation of health satisfaction.\When regional unemployment is included, the change in R2 is higher(columns 3 to 2 as compared with columns 6 to 5), and the impact ofthis variable is also grearer (columns 3 and 6). The variables to be in-cluded here do not provide any simple explanation. Our assumption -based on our experience regarding the component breakdown of jobexpectations - is that changes in environmental perspectives have ashort-term effect on overall l i fe satisfaction while any psycho-socialdamage becomes apparenr only after a certain delay (Matiaske/Schramm/Schlese 1996).

Finally it becomes evident that the regional dummy variables, justas in the other versions shown below, do not have any significant im-pact here. Minor influences in the original model are absorbed by in-clusion of regional unemployment, The regional dummy variables are

Regional Unemployment and Individual Satisfaction 97

therefore no longer taken into consideration in the following. Instead,

the results from the two other methods, i.e. ordinal regression and the

hierarchical model, are shown additionally, starting with the basicrnodel in Table 2. However, the results of the different models arer)rore or less the same, therefore our interpretation of the findings is

based mainly on the well-known model of multiple regression.

4. Discussion of regional unemployment rates and satisfaction

The impact of regional unemployment on general satisfaction and on

health has so far rarely been the subject of empirical l i terature. The

clominating thread currently focuses more on the l ink between health

rnd inequality (Grözinger 2002).1Vhile some authors - mainly Brit ish

and American - investigating this subject from a social-epidemio-

logical perspective, have observed a correlation between inequality in

income and unemployment, most publications focus on the effects of

l'uble 2. Delerminants of Satisfaction - Basic Model

Overall Life Satisfaction Health Satisfaction

OLSOrdinal

regressionHieruchical

modelOLS

Ordinalregression

Hierehicalmodel

( onslant

Age

Age'

I)cgree ofdis-ab i l i t y

5cx

Partnership

[:ducation

I rqu iva l i sed in -come

Iransl'er income

I Jncmployment

Rcgional rate ofu nemp r.

8.660

-0.065

0.00 |

- 0 .015

-0.008

0.324

0.02'7

0.000

-0.85 |

-0.653

-0.035

-0.084

0.001

-0.014

-0.016

0.371

0.035

0.000

-0.137

-0.655

-0.044

8.631

-0.065

0.001

-0.015

-0.010

0.323

0.03 |

0.000

-0.829

-0.640

-0.034

9 .140

-0.090

0.00 |

-0.029

0 .139

0 . 1 1 7

0.05 I

0.000

-0.362

-0.33 I

-0.021

-0.095

0.001

-0.021

0 .1 08

0.068

0.049

0.000

-0. | 95

-0.265

-0.026

9.094

-0.090

0.001

-0.029

0 .1 38

0.t22

0.054

0.000

-0.345

-0.320

-0.021

0. I 08 ***

I 2095

0 . t 00 * * 0 . 1 2 5 r r

I 1995

0 . 1 8 8 + r r

eve ls o fs ign i f i cance: < 0 .001 r * r * ' , < 0 .01 t * t ' , < 0 .05 r * ' , < 0 .

98 Gerd Grözinger and \üenzel Matiaske

the first dimension without taking inro account the issue of unem-ployment. The only publication referring ro Germany that deals withinequality and satisfacrion and can be compared ro our approach be-cause it uses the regionalized SOEP fails to include unemploymentnumbers (Schwarze und Härper in this volume).

Only a small number of publications examine the impact of re-gional unemployment on well-being with individual data. Blake J.Turner carried out such a study in the U.S. for the middle of the1980s using a fairly modest data record of just above 1,000 interview-ees (Turner 1995). Gathering values of physical and psychologicalwell-being for the health dimension, he observed a slightly negativeimpact of local unemploymenr on the group of permanenrly em-ployed persons, a strongly negative impact on unemployed personsand a strongly positive impact on employed persons who had previ-ously gone through phases of unemploymenr. In the two lattergroups, however, the number of cases was very low, inducing theauthor himself to interpret his data with caution.

Andrew A. Clark had a much larger number of cases ro srarr from(Clark 1999). He carried our an assessmenr for the 1990s for em-ployed individuals using Brit ish panel data - similar to the SOEp - aswell as covering eleven regions. He used an index for psychologicalwell-being assembled from several different quesrions as the variableto be explained. He observed a significantly negative impact of re-gional unemployment on rhose employed, this coincides with our re-sults. However, his explaining variables also include an indicator forhealth satisfaction, which makes rhe inrerpretation of rhis model verydiff icult. one can either examine rhe effect of satisfaction in certainareas on general satisfaction, or the impact that variables outside sub-

;'ective measurements have on any one dimension of satisfaction. Byblending together these rwo methods the problem of mulri-colinearity is neglected.

In his approach, Clark also observed an opposite impact on unem-ployed individuals. In this group, regional unemployment has aslightly positive impact on satisfaction. His explanation for this sur-prising result is that the knowledge that many others are in rhe sameawkward situation can help a person psychologically. However, given

Regional Unemployment and Individual Satisfaction 99

rhe problematic model and the restricted data basis, this may be an

,,ver-interpretation of the results.

Using our data record and our model specification, we tested(llark's hypothesis and examined overall l i fe satisfaction and health

srrtisfaction as a reaction to regional unemployment for the group of

rrnemployed persons. However, there are only BO0 cases ieft in the

SOEP. Due to this narrow data basis, now the sex variable loses all

significance for the determination of the dimension of perception of

health, a rather unlikely result. The Brit ish data record with about(,,OOO interviewees per year is even smaller structurally than that pro-

vided by the SOEP and appears to be more extensive only because

scveral survey waves have been combined. All statements about un-

cmployed persons based on that data record have to be viewed with

cxut ion.

' l 'his results in some previously relevant variables losing their signifi-

cance due to the low number of cases examined and sometimes

changes in sign (Table 3). The impact of regional unemployment con-

'lible 3: Determinants of Satisfaction - Unemployed Persons

Overall Life Satisfaction H€alth Satisfaction

OLSOrdinal

regressionHiemhical

modelOLS

Ordinalregression

Hiemhimlmodel

( onstant

Age

Ag"t

l )cgree o fd is -ab i l i t y

)cx

l'afinership

l:ducation

l rqu iva l i sed in -come'l-ransfer

income

Rcgional rate ofunempl.

9.640 *r I

-0.2234r'

-0.014 * **

-0.281 *

0.496 ** |

0 . 051

0.000 ** *

-0.332

-0.040 rr

_0. I 98 **+

0.002 ++*

-0.0 I 8 ** ,

-0.305 *

0.49 I ** .

0.048

0.000 * * |

-0.268

-0.041 t*

_0.223***

0 003***

-0.014 r r*

-0.281 r

0.496 **r

0.05 |

0.000 * * {

-0.332

-0.040rr

I 0.400 r+

-0.236rr

0.002rr

-0.033 *'

0.000

0 .1 65

0 .1 24 * *

0.000.

-0.078

-0 .013

-0.202ri I

0.002 rr r

. 0 . 021* * r

0.033

0.1 93

0.095 ***

0,000.

-0. | 78

-0.014

-0.243 r **

0.002 * **

-0.033 *r{

-0.002

0 . t 7 0

0 .129* * i

0.000.

-0.086

-0 .012

l{ '

N

0 . 1 l 8 * * *

835

0 . 1 1 8 * + 0.192*1

834

0. I 83 r r+

l -cve ls o fs ign i f i cance: < 0 .001 ' r * * r , < 0 .01 r * r r , < 0 .05 ' r ' ! < 0 .1

100 Gerd Grözinger and \(enzel Matiaske

tinues to be negative, however, with significant values available onlyfor overall l i fe satisfaction. clark's staremenr that a person regards hisor her own unemployment - while certainly as a negarive situation -sti l l as less dramatic if many others in that person's environmenr suf-fer the same fare, is not to be found in Germany.

Alois Stutzer and Rafael Lalive, who evaluated a data record of ap-proximately 1,400 individuals from the labor force at the beginning ofthe 1990s in 133 larger communities in switzerland, also concenrratedon the impact of regional unemployment (Stutzer/Lalive 2001).\(henever a dummy variable is included for the dominant standardview that everyone has to look after themselves, they do not find anysignificant l ink with overall l i fe satisfaction. At that t ime, a referen-dum was held in Switzerland on reducing unemployment benefit, andthe authors define a "strong social standard" as communities where anabove average affirmation of this intention was registered.

Regional Unemployment and Individual Satisfaction 101

Since there is no similarly interesting indicator for regional value posi-

rions in Germany, it was diff icult to compare the results with our

,,wn. \f lhat was possible, however, was to follow uP two possible ob-

icctions against a hasty transference of this result. First of all, the

number of unemployed individual is very low in Switzerland, it came

to just 5% at the time of the survey. This factor most probably had an

irnpact on the regional variance. Secondly, the number of cases in-

cluded in the Swiss data record is much lower than the number of

c,rses provided by the SOEP for Germany; this, too, probably con-

tributes to the lack of significance.

To find out whether or not the statements made by Stutzer and La-

live can be transferred, we again used our model to carry out a test'

We limited the sample to Southern Germany (Bavaria and Baden-

Wuerttemberg) in order to simulate conditions regarding the low

number and regional variance of unemployment similar to those in

Switzerland. Lastly, we also l imited our test to employed persons.' l 'his

resulted in 1,800 cases, i.e., rather more than the Swiss model.

The result shows that regardless of the model specification, the re-

gional rate of unemployment has no impact on satisfaction in South-

crn Germany (Table 4). Stutzer and Lalive's assumPtion that the im-

pact of regional unemployment is influenced by social standards ap-

pears questionable in l ight of the test result. In our opinion, the fact

they found regional ur-remployment to be insignificant results from an

inadequate data base. It is not clear at this point whether or not there

is a general unemployment figure serving as a threshold, below which

rhe regional variance in welfare states no longer has any measurable

in-rpact; neither is it certain whether the relatively low rates of unem-

ployment in Switzerland and Southern Germany are below such a

threshold value.

Irinally, let us once more return to the question of whether regional

unemployment has a negative impact because of the individual de-

crease in income, career chances, further education oPtions' and mo-

bil ity, or if i t causes a more general concern beyond such personal

worries. \fle chose civil servants for our test group' since the effects of

regional unemployment do not directly affect them. The result of the

rest is unequivocal (Table 5). The reaction to the regional job situa-

tion is as negative in civil servants as it is in other grouPs' in the di-

Table 4: Determinants of Satisfaction - Employed Persons in Southern GermanyOverall Life Satisfaction Health Satisfaction

OLSOrdinal

regressionHiemhiel

modelOLS

Ordinalregression

Hierehislmodel

Constant

Age

Ag.t

Degree ofdisabil-ity

Sex

Partnership

Education

Equivalised in-come

Transfer income

Unemployment

Regional rate ofunempl.

8.760r**

-0.099 ***

0.00 I ***

_0.02 I ***

0.086

0.228**

0.050**

0.000.

_1 .360* * *

-0.475 **

-0.006

-0. I 26 **

0.00 I **

-0.024*1

0 . I 59 .

0 .330* *

0.059 **

0.000 *r

- 1 .850* *

-0.4t7 *

-0.012

9.627 *++

-0.100***

0.001 ***

_0.02 I ***

0.084

0.224**

0.053 *r

0.000.

- I ,325 r*N

-0.4991*

0 .014

9 .860 i r 1

_0. I 49 +*r

0 ,001* * *

_0.037*r*

0.336*+*

0,094

0.071 *r*

0.000

_1 .250 r r *

-0.375.

0 .012

_0. I 53 ***

0.00 | ***

-0,029 ***

0.327 ***

0 . l 2 l

0.069 **i

0.000

-1.2 '14**4

-0.243

-0.010

9.487 * * {

-0 . I 49 r * '

0 .001 * * *

_0.036 ** r

0 .J33 * * *

0 .085

0.075 r r *

0.000

- l . l 3 9 + *

-0 .416r

0 .058

R,

N

0.099 | **

1844

0.1 03 0 . 1 6 4 N

I 836

0 .140* * *

Leve ls o fs ign i f i cance: < 0 .001 ' * * * ' , < 0 .01 ' * * r ! < 0 .05 ' * r , < 0

142 Gerd Grözinger and \f lenzel Matiaske

mension of general satisfaction as well as i ' the dimension c,f healthsatisfaction. The psychological impact of unemploymenr seems ro gofar beyond the individual threat.

5. Summary

It has been shown on rhe basis of the SOEP that the regional rate ofunemployment in Germany has a negative impact on general satisfac-tion and on health satisfaction. The negative impact manifests itself inboth employed and unemployed persons. Thus, a new and importantexternality of unemploymenr has been identif ied.

'We cannot rule out the fact thar this negative externality occurs

only when a cerrain threshold value of high unemployment is given.When the impact is measured, it equally distresses the entire popula-tion. The impact is not due only to individual remore effects of a dif-

Table 5; Determinants of Satisfaction , Civil Servants

Overall Life Satisfaction Flealth Satisfaction

OLSOrdinal

regressionHieruchical

modelOLS

Ordinalregression

Hieruchiolmodel

Constant

Ags

Ag"t

Degree ofdis-abil ity

Sex

Partnership

Education

Equivalised in-come

Transfer income

Regional rate ofunempl .

7.990+*{

_0.084 +*r

0.00 | r r

_0.0 | | ** i

0.057

0.299+rr

0.05 I *+r

0.000 *r *

- 0 .718*

-0.025 **

_0.1 1 5 **+

0.001 *r*

-0.006

0.078

0.283 '

0.05 | *+

0.000 *r*

-0.37 l

-0.046 ***

9.046 ***

_0 084* * *

0 .001* *

-0.01 I ***

0.049

0.296 r **

0.050*r*

0.000 r **

-0.7 l6 *

-0.025 **

9.690 ***

_0 . l 2 I r + *

0 .001++

_0.02'7 ***

0.036

0.054

0.065r+r

0.000

-0. I 73

-0.036*1r

_0 .1 l 4 * * r

0 .00 t *

_0.021 **r

0 .017

-0.094

0.063 ***

0.000

-0.234

-0.037 **r

9.7 49 ***

-0 .121 * * r

0.001 * *

-0.026 f**

0.027

0.048

0.063 ** t

0.000

-0.125

_0.037 r* |

RT

N

0 . 0 5 8 . t *

I 659

0.059 * * r 0 .104* *

I 660

0 . I l 0 * r +

Leve ls o fs ign i f i cance: < 0 .001 ' * * * ' , < 0 .01 ' i * t ! < 0 .05 ' * , . < 0 .

Regional Unemployment and Individual Satisfaction 103

licult job market, since it is also manifest in civil servants, who are l ir-

t lc affected by the employment situation.

6. Bibliography

Ilellmann, LutzlBlien, uwe. (2001):'wage curve Analyses of Establishment

Data from 'lüesrern

Germany. ln Indusftial €t labor relations review

5 4 , 4

lllanchflower, David G./Oswald, AndrewJ. (2000): .Well-Being

Over time In

Britain and USA. Rep./Working Paper 7487, NBER

ßundesamt für Bauwesen und Raumordnung. (2001): Aktuelle Dhten zur

Entwicklung der Städte, Kreise und Gemeinden' Bonn

Clark, Andrew A. (1999): [Jnemployment as a Social Norm: PsychologicaL

Eaidence from Panel data. Rep./Wor&fug, CNRS and LEO-CRESEP'

Orl6ans

Clark, Andrew E./Georgell is, Yannis/Sanfey, Peter. (2001): Scarring: The

Psychological Impact of Past Unemployment' Ini Economica, 68:221-

241

Feather, Norman T. (1990): Tbe Psychological Impact of Unemployment'

NewYork

Frey, Bruno S./Stutzer, Alois. (2002) : Happiness and economics : bow the

economy and institutions affect well-being. Princeton, NJ fet al']

Grözinger, Gerd. (2002). Makroökonomik der Seele? Über den Zusammen-

häng uon Ungleichheit, Sozialkapital und Gesundheit ' In: Öhonomik

und. sozialwissenscbaft. Ansichten eines in Bewegung Seratenen verbah-

zllsses,'Walter Ötsch, Stephan Panther (eds.). Marburg, 119-153

Harrel, F. E.Jr. (2001): Regression ModelingStrategies' NewYork [et al ']

Hauser, Richard. (2002). Zum Einfluss von Aquivalenzskalen auf Ergebnisse

zur personellen Einkommensverteilung und zur relativen Einkom-

mensarmur. In. zur Theorie, Empirie und Politik der Einhommensver-

teilung. Festscbrift fi.tr Gerold Bltimle, Lukas Menkhoff, Friedrich L.

Sel l (eds.) : Ber l in , 175-189

Idler, Ellen L/Benyamini, Yael. (1997): Self-related health and mortality: a

review of twenry-seven community studies. In Journal of Heabh and

Social B ehauior 38, March: 21-37

Jahoda, Marie/Lazarsfeld, Paul F./zeisel, Hans. (1960): Die Arbeitslosen pon

Marienthal.Bonn

104 Gerd Grözinger and Wenzel Matiaske

Martin, Albert/Düll, Herbert. (2ooo). Betriebriche veiterbildune und Ar-bei tsmarkts i ruar ion: eual i f iz ierungsmaßnahmen im Lichte organisa_tionstheoretischer Ansätze und empirischer Datenanaly r". rn, i*orrr-sche organisations- und Entscheidungsforschung, Matiaske venzel,Thomas Mellewigt, Friedrich A. Stein (eds.): Heidelberg, 8l_123

Matiaske, \ü/enzellSchramm, Florian/schlese, Michael. (1996\: Berufl icheErwartungen und ihre Konsequenzen. rn: Zeitschrtrt fi;r personarfor-schung (Sonderband: Sechs Jahre danach - personalarbrit in drn nru*B undesländern: 25-51

Mielck, Andreas. (2ooo): Soziale Ungleichheit und Gesundheit. EmpirischeErgebnisse, Erklärungsansätze, I nterventionsmaglichkeiten. Bern

Pinheiro, J. c.,/Bates' D. M. (2ooo): Mixed-Effects Mod.ers in S and s-pLus.New York [et al.]

Praag, B. M. S. vanlFrijters, P./Ferrer-i-Carbonell, A. (2001): The Anatomyof Subjectioe Well-being. Rep./Working paper 265, Dr\V, Berlin

Schwarze, Johannes/Andersen, Hanfried H./Anger, Silke. (2000): Self_ratedHealtb And Changes In Self-rated Health As predictors of uortat;ty -First Evidence From German paner Dau. Rep./worhing paper 203,DI\X/, Berlin

Snijders, R./Bosker, R. ( 1 999): Multileael Analysis. LondonStutzer, Alois/Lalive, Rafael. (200r): The Role of Sociar work Norms in Job

Searching and subjectizte tilell-Being. Rep./working paper 300, r2A,Bonn

Tella, Rafael Di/MacCulloch, Robert J./Oswald, Andrew J. (2001): prefer_ences over Inflation and unemploymenr: Evidence from Surveys ofHappiness. In: American Economic Revieta 91,1:335_341

Turner, J. Blake. (1995): Economic context and the Health Effects of un-employment. In: Journal of Health and Social Behapior J6, September:2 r 3 - 2 2 9

'winkelmann, Lil iana/vinkelmann, Rainer. (r99g): vhy Are The Unem-

ployed So Unhappy? Evidence From panel Data. In: Economica 65: t_1 5

'üfitte, Hans De. (1999): Job Insecurity and psychological Vell_Being: Re_view of the Literature and Explorarion of some unresolved Issues. In:European Journal Of Worh and Organizational psychologt g, 2: 155_1 7 7

Occupational wage differentials

Urue Blien, Phan tbi Hong Vanl

1. Introduction

,'\ccording to human capital theory, people who invest a specific,rnount of t ime and money in education and training should receive al,rir ly identical level of wages in return. Therefore, the wages of wor-I 'crs who have completed a three-year course of vocational training inr l re German so-cal led "dual system" represenr a test of the theory.' l 'he "dual system" provides a specific forr,-r of training in which,rp;rrentices are trained in an employing firm and at a vocationalr , : hoo l .

f)escriptive analyses show that average wages in different occupa-r ions d i f fer up to 100 percent one year af ter f in ish ing thc t ra in ing. In( icrmany occupational structures in the labour market are more im-

i)()rtant than in other countries. This corresponds to a long tradirion.rnd to the institutional structures shaping the labour market as it isshown below.

Instead of using descriptive statistics we apply a cross-classified,,rult i leveI model to explain occupational wage differentials. This

Institute for Employment Research of the Federai Employment Services1 I AB), Nuernberg/Universität Kaiserslautern (Email: [email protected])L,rthar Dirnfeldner and Karen Scott-Leuteritz are thanked for very valuable,jupport. Elke Jahn, Gerd Grözinger, the participants of a session of the 2003( longress of the European Society of Population Economics (ESPE) in New\',rrk and the attendants of the 2003 GAPE meeting in lVittenberg are thankedl,rr valuable contributions in the discussion of the paper. All responsibil i ty forrhc analysis remains with the authors.


Recommended