Sunteți pe pagina 1din 14

Analele Universitii Constantin Brncui din Trgu Jiu, Seria Economie, Nr.

1/2008

RISCUL ANGAJRII - O

RISK OF EMPLOYMENT A

ABORDARE PRIN PRISMA

LOGISTIC REGRESSION

REGRESIEI LOGISTICE

APPROACH

Daniela-Emanuela Dncic,
Lect.univ.drd.

Daniela-Emanuela Dncic,
Lect.univ.drd.

Ana-Gabriela Babucea, Prof. univ.dr.

Ana-Gabriela Babucea, Prof. univ.dr.


Faculty of Economics

Facultatea de tiine Economice

Constantin Brncui University of


Trgu-Jiu

Universitatea Constantin Brncui din


Trgu-Jiu

Abstract

Abstract
The aim of this paper is to determine
in which way gender, age and educational
level influence the risk of a person to find a
job or to be re-employed. The empirical
investigation was made only for a Romanian
county, because the great volume of data
made it impossible to achieve them for the
entire country.

The aim of this paper is to determine


in which way gender, age and educational
level influence the risk of a person to find a
job or to be re-employed. The empirical
investigation was made only for a Romanian
county, because the great volume of data
made it impossible to achieve them for the
entire country.

Annals of the Constantin Brncui University of Trgu Jiu, Economy Series, No. 1/2008

123

Analele Universitii Constantin Brncui din Trgu Jiu, Seria Economie, Nr. 1/2008

1.

1.

Introducere

Scopul acestui studiu este acela de


a determina in ce masura genul, varsta si
nivelul educational influenteaza riscul
unei persoane de a isi gasi un loc de
munca sau de a se reangaja. Ca
metodologie
am
folosit
regresia
logistica. Genul, varsta si nivelul
educational sunt variabile independente
(factoriale) in model, iar variabila
dependenta a fost denumita in studiu
statut , variabila calitativa binara ce ia
valorile 1 pentru o persoana devenita
angajat si 0 pentru o persoana ramasa
neangajata
la
sfarsitul
perioadei
analizate. Esantionul contine 80961
inregistrari cu informatii referitoare la
data de intrare in somaj, data de iesire
din somaj, sex, varsta, nivel educational
si motivul iesirii din somaj pentru fiecare
persoana
inregistrata
la
Agentia
Nationala pentru Ocuparea Fortei de
Munca Bucuresti, in perioada 1 Ianuarie
2002 31 August 2006. Richard
Berthoud
(2003)
in
capitolul
Disadvantaging
characteristics
al
monografiei sale intitulate Multiple
disadvantage in employment analizeaza
influenta
variabilelor
independente
varsta, status familial, nivel educational,
stare de sanatate, apartenenta etnica
asupra probabilitatii de a isi gasi un loc
de munca sau de a se reangaja al
subiectilor din Marea Britanie. Alba
Ramirez (1998) in lucrarea sa ReEmployment Probabilities of Young
Workers
in
Spain
investigheaza
influenta genului, mediului urban/rural si
a
nivelului
educational
asupra
proabilitatii de a se reangaja a tinerilor
lucratori din Spania. Studii similare au
fost efectuate de catre Meghir C. si
Ioannides Y. (1989) pentru Grecia,
Tansel Aysit (2001) pentru Turcia sau
Tunali I si R. Asaad (1992) pentru Egipt.
Aspectele legate de somaj, angajare, si
alte probleme generate de dinamismul
pietei muncii in Romania au fost

Introduction

The purpose of this survey was to


determine the risk of the persons from the
database studied to be employed or unemployed
at the end of the period subject to analysis, risk
estimated according to independent variables
gender, age and educational level. As a
methodology I used the logistic regression.
Gender, age and educational level are
independent (factorial) variables in the model,
and the dependent variable has been called
status in the survey, the binary qualitative
variable that takes values 1 for a person become
employed and 0 for a person unemployed at the
end of the analysed period. The sample has
80961 records with information concerning the
entrance date into unemployment, the date of
unemployment end, sex, age, educational level
and the reason for unemployment leaving for
each person registered at the National Agency
for Employment Bucharest during January 1,
2002 - August 31, 2006.
Richard Berthoud (2003) in the chapter
Disadvantaging characteristics of his
monograph called Multiple disadvantage in
employment analyzes the influence of
independent variables age, family status,
educational level, health and ethnic background
on the probability of subjects from Great Britain
to find a job or to be re-employed. Alba
Ramirez (1998) in his paper Re-Employment
Probabilities of Young Workers in Spain
inquires into the influence of gender,
urban/rural area and educational level on the
probability of young workers from Spain to
become employed. Similar surveys were made
by Meghir C. and Ioannides Y. (1989) for
Greece, Tansel Aysit (2004) for Turkey or
Tunali I and R. Asaad (1992) for Egypt.
Aspects related to unemployment, employment
and other issues generated by the dynamism of
the labour market in Romania were
insufficiently investigated and there are no
surveys of this kind for our country.

Annals of the Constantin Brncui University of Trgu Jiu, Economy Series, No. 1/2008

124

Analele Universitii Constantin Brncui din Trgu Jiu, Seria Economie, Nr. 1/2008

insuficient investigate, neexistand studii


de acest tip pentru tara noastra.

2. Descrierea bazei de date

2. Database description

Of 80961 persons registered in the


database of Gorj County as unemployed, during
1.01.2002-31.08.2006, 19369 persons found a
job, until August 31, 2006; the reason for their
unemployment leaving was filled in the
database with employed. The average
duration of unemployment until finding a job is
of 6 months, the median of 2 months, the
maximum value of 57 months and the
minimum value of 0 months. Of these 6390
persons (33%) are women and 12979 (67%) are
men. If the percentage difference between the
male unemployment and the female
unemployment registered in the database is of
17.8%, the percentage difference between the
number of men who become employed and the
number of women in the same situation is of
34%, which shows that although there are more
unemployed men, and on the average they stay
unemployed about one more week than women,
however they are preferred by employers. As
concerns the distribution on age groups, most of
them (30.1%) belong to the 25-34 year-old age
group, followed by the 15-24 year-old age
group with 25.7%, 35-44 group with 25.4%, 4554 group with 17.5% and 55-64 year-old age
group with 1.3%. 0.4% of the employed are
without education, 44.7% of the employed have
an educational level of up to 10 grades, 37.4%
are high school graduates, 8.6% of the
employed graduated from vocational or
foremen schools and 8.9% of the employed are
university education graduates.

Din
80961
persoane
inregistrate in baza de date a judetului
Gorj ca someri, in perioada 1.01.200231.08.2006, 19369 persoane si-au gasit
loc de munca, pana la data de 31 august
2006; pentru acestia, in baza de date, la
motivul iesirii din somaj figureaza
angajat. Durata medie a somajului
pana la gasirea unui loc de munca este de
6 luni, mediana de 2 luni, valoarea
maxima 57 luni si valoarea minima 0
luni. Dintre acestia 6390 persoane (33%)
sunt femei si 12979 (67%) sunt barbati.
Daca
diferenta
procentuala
dintre
numarul somerilor barbati si numarul
somerilor femei inregistrati in baza de
date este de 17.8%, diferenta procentuala
dintre numarul barbatilor care s-au
angajat si numarul femeilor in aceeasi
situatie este de 34%, ceea ce arata ca
desi sunt mai multi barbati someri, si
stau in somaj in medie mai mult cu
aproximativ o luna decat femeile, totusi
acestia sunt preferati de catre angajatori.
In ceea ce priveste distributia pe grupe
de varsta, cei mai multi dintre acestia
(30.1%) apartin grupei de varsta 25-34
ani, urmata de grupa de varsta 15-24 ani
cu 25.7%, grupa 35-44 cu 25.4%, grupa
45-54 cu 17.5% si grupa 55-64 ani cu
1.3%. 0.4% dintre cei angajati sunt fara
studii, 44.7% dintre cei angajati au un
nivel educational de pana la 10 clase,
37.4% sunt absolventi de liceu, 8.6% 3. Methodology
dintre cei angajati sunt absolventi de
scoala profesionala sau de maistrii si
As methodology, in order to determine
8.9% dintre cei angajati sunt absolventi
the risk of the persons from the studied database
de invatamant superior.
to be employed or unemployed at the end of the

Annals of the Constantin Brncui University of Trgu Jiu, Economy Series, No. 1/2008

125

Analele Universitii Constantin Brncui din Trgu Jiu, Seria Economie, Nr. 1/2008

period subject to analysis, risk estimated


according to the independent variables gender,
age and educational level, I used the logistic
regression.

3. Metodologie
Ca
metodologie,
pentru
determinarea riscului persoanelor din
baza de date studiata de a fi angajate sau
neangajate la sfarsitul perioadei supuse
analizei, risc estimat in functie de
variabilele independente gen, varsta si
nivel educational, am folosit regresia
logistica.
Spre deosebire de regresia liniara
multipla, unde se poate prezice, pe baza
mai multor variabile independente, o
variabila dependenta numerica, regresia
logistica ofera posibilitatea prezicerii
unei variabile nominale dihotomice.
Metoda regresiei liniare presupune ca
atat variabilele factoriale cat si variabila
rezultativa sa fie de tip continuu; prin
contrast, regresia logistica permite lucrul
cu alte tipuri de variabile.

Unlike the multiple linear regression


where, based on several independent variables,
a numeric dependent variable can be predicted,
the logistic regression gives the possibility to
predict a dichotomic nominal variable. The
method of linear regression implies that both the
factorial variables and the resultative variable
should be of a continuous type; by contrast, the
logistic regression allows working with other
types of variables.

The logistic regression model describes


the relation between a dichotomic nominal
variable Y, that takes the values 1 (success) and
0
(failure),
and
k
factorial
variables x1 , x 2 , x3 .......x k . Factorial variables
can be quantitative (numeric) or categorical.
Since Y is a binary variable, it has a Bernoulli
type
distribution,
with
the
parameter p = P(Y = 1) , where p is the
probability of success for the given values
Modelul de regresie logistica x1 , x 2 , x3 .......x k of factorial variables. The
descrie relatia dintre o variabila average of Bernoulli type variables is given by:
nominala dihotomica Y, ce ia valorile 1
(succes) si 0 (esec), si k variabile
x1 , x 2 , x3 .......x k . Variabilele
factoriale
E[ y ] = P (Y = 1) = p .
factoriale pot fi cantitative (numerice)
sau categoriale. Deoarece Y este o
(1)
variabila binara, prezinta o distributie de
tip Bernoulli, cu parametrul p = P(Y = 1) ,
unde p este probabilitatea de succes
The logistic regression model is defined
pentru valorile date x1 , x 2 , x3 .......x k ale
variabilelor factoriale. Media unei as follows: assuming that Y1 ........Yn are
independent Bernoulli variables, and be pi the
variabile de tip Bernoulli este data de:
average
of
values Yi ,
E[ y ] = P (Y = 1) = p .
then pi = E[Yi ] = P(Yi = 1) . The value pi can
(1)
be expressed according to the factorial variables
Modelul de regresie logistica se x , x , x .......x thus:
i1
i2
i3
ik
defineste in felul urmator: presupunem
ca Y1 ........Yn sunt variabile independente
Bernoulli, si fie pi media valorilor Yi ,

Annals of the Constantin Brncui University of Trgu Jiu, Economy Series, No. 1/2008

126

Analele Universitii Constantin Brncui din Trgu Jiu, Seria Economie, Nr. 1/2008

atunci pi = E[Yi ] = P(Yi = 1) . Valoarea pi


poate fi exprimata in functie de
variabilele factoriale xi1 , xi 2 , xi 3 .......xik
astfel:
pi =

pi =

1
k

1 + exp( 0 j xij )
j =1

(2)

1
k

1 + exp( 0 j xij )

If we apply the logit transformation to


the relation (2) we obtain a linear type
connection between pi and the factorial
Daca aplicam transformarea logit
relatiei (2) vom obtine o legatura de tip variables, as follows:
linear intre pi si variabilele factoriale,
dupa cum urmeaza:
k
pi
k
log
it
(
p
)
=
log(
)
=

+
j xi , j

i
0
p
1 pi
j =1
log it ( pi ) = log( i ) = 0 + j xi , j
1 pi
j =1
(3)
(3)
j =1

(2)

Ecuatia 3 este cunoscuta si sub


The 3rd equation is also known as a
numele de forma logit a modelului. Logit
( pi ) este logaritmul odds pentru logit shape of the model. Logit ( pi ) is the
succes, pentru valori date ale variabilelor odds logarithm for success, for given values
of the factorial variables xi1 , xi 2 , xi 3 .......xik .
factoriale xi1 , xi 2 , xi 3 .......xik .
Rezultatul regresiei logistice este
tot o ecuatie, care prezice cel mai bine o
variabila
efect
binara
(statut,
angajat/neangajat), pe baza uneia sau
mai multor variabile ce pot fi cantitative
(varsta, nivel educational), sau binare
(gen). In loc de a lucra cu probabilitati,
(care se pot afla intre 0 si 1), regresia
logistica lucreaza cu logaritmul natural
al cotei (odds), care poate lua orice
valoare, pozitiva sau negativa. Ecuatia
regresiei logistice poate fi exprimata
astfel:
ln(oddsY ) = 0 + 1 x1 + 2 x 2 + ....... + n x n
(4)

The result of the logistic regression is


also an equation, that predicts the best a binary
effect variable (status, employed/unemployed),
based on one or several variables that can be
quantitative (age, educational level), or binary
(gender). Instead of working with probabilities,
(that may range between 0 and 1), the logistic
regression works with the natural logarithm of
the quota (odds), that can take any value,
positive or negative. The equation of the logistic
regression can be expressed as follows:

ln(oddsY) = 0 + 1x1 + 2 x2 + .......+ n xn


(4)

O
prezentare
detaliata
a
metodologiei regresiei logistice si a
A detailed presentation of logistic
problemelor ridicate de folosirea acesteia regression methodology and of the issues raised
a fost realizata de catre Amemiya, T. by its use was performed by Amemiya, T.
(1985),
Balakrishnan,
N.
(1991), (1985), Balakrishnan, N. (1991), Hosmer,

Annals of the Constantin Brncui University of Trgu Jiu, Economy Series, No. 1/2008

127

Analele Universitii Constantin Brncui din Trgu Jiu, Seria Economie, Nr. 1/2008

Hosmer, David W.; Stanley Lemeshow David W.; Stanley Lemeshow (2000), Agresti,
(2000), Agresti, Alan. (2002) si Green, Alan (2002) and Green, William H. (2003).
William H. (2003).
In my survey, the factorial variables are:
In
studiul
meu,
variabilele gender, alternative qualitative variable
factoriale sunt: gen, variabila calitativa (male/female), encoded in the analysis with 1
alternativa
(masculin/feminin), (male) and 0 (female); age, quantitative
codificata in analiza cu 1 (masculin), variable, divided in five intervals, 15-24, 25-34,
respectiv 0 (feminin); varsta, variabila 35-44, 45-54, 55-64, in accordance with the
cantitativa, divizata in cinci intervale, Statistical Yearbook of Romania and
15-24, 25-34, 35-44, 45-54, 55-64, in educational level, qualitative variable in the
conformitate cu Anuarul Statistic al database received from NAE Bucharest, but
Romaniei si nivel educational, variabila changed into numeric variable, in compliance
calitativa in baza de date primita de la with the regulations of Romanian Ministry of
ANOFM Bucuresti, dar transformata in Education. The dependent variable status
variabila numerica, in conformitate cu (persons status at the end of the analysed
reglementarile Ministerului Roman al period), is a dichotomic nominal variable that
Educatiei. Variabila dependenta statut takes the values:
(statutul persoanei la sfarsitul perioadei
analizate), este o variabila nominala
dihotomica, ce ia valorile:
1, if the person found a job
status =
0 , is the person is still unemployed

1, daca persoana si a gasit un loc de munca


statut =
0 , daca persoana este inca neangajata

Ecuatia de regresie este:

The regression equation is:

ln(oddsstatut) = 0 + 1 gen + 2 var sta + 3 niveleducational

(4)
unde 1 si 2 sunt coeficientii de regresie
calculati cu ajutorul programului statistic
SPSS 10.0, si care sunt in realitate
logaritmii naturali ai odd ratio ai fiecarei
variabile,
iar
0 este
constanta,
reprezentand logaritmul natural al odds
statut pentru subiectii care prezinta valori
nule ale tuturor variabilelor factoriale (gen,
varsta, nivel educational in cazul de fata).

Metoda
selectata
pentru
regresia
logistica binara a fost metoda Enter,
variabilele factoriale fiind analizate
simultan.

ln(oddsstatut) = 0 + 1gen+ 2 varsta+ 3niveleducationa


l

(4)

where 1 and 2 are regression coefficients


calculated using the statistics program SPSS
10.0, and which in reality are natural logarithm
of odd ratio and of each variable and 0 is
the constant, representing the natural logarithm
of odds status for subjects who have zero values
of all factorial variables (gender, age,
educational level in this case). The method
selected for the binary logistic regression was
the Enter method, analysing simultaneously the
factorial variables.

In tabelul 1 sunt prezentate


rezultatele testului Omnibus pentru
coeficientii
modelului.
Rezultatele
Table 1 shows the results of the
2
testului si ale ratei de verosimilitate Omnibus test for the model coefficients. The
2LL inregistrate la pasul 1 comparativ cu
results of the test 2 and of the likelihood rate pasul initial 0 ne permit respingerea

Annals of the Constantin Brncui University of Trgu Jiu, Economy Series, No. 1/2008

128

Analele Universitii Constantin Brncui din Trgu Jiu, Seria Economie, Nr. 1/2008

ipotezei nule ( H 0 : i = 0 ) si acceptarea 2LL recorded in step 1 compared to the initial


step 0 allow us to reject the null hypothesis
ipotezei alternative.
( H 0 : i = 0 ) and to accept the alternative
hypothesis.

Tabelul 1: Rezultatele testului omnibus pentru coeficientii de regresie / Table 1: Results of the
omnibus test for regression coefficients
Step 1

Step
Block
Model

Chi-square
1367.716
1367.716
1367.716

df
3
3
3

In tabelul 2 este prezentat


resultatul
testului
Hosmer
&
Lemenshow.
Testul
Hosmer
&
Lemenshow divide subiectii la nivel de
decile, pe baza probabilitatilor estmate,
la pasul urmator aplicand testul
2 asupra
frecventelor
observate.
Valorile p=0.0000 sunt calculate pe baza
distributiei 2 cu 3 grade de libertate si
indica faptul ca modelul logistic este
valid din punct de vedere statistic, deci
ipoteza nula poate fii respinsa.
Tabelul 2 : Testul Hosmer si Lemeshow
Step
Chidf
Sig.
square
1 427.73
8
.000
2

Sig.
.000
.000
.000

Table 2 shows the results of Hosmer &


Lemenshow test. The Hosmer &
Lemenshow test divides subjects at
deciles level, based on the estimated
probabilities, applying in the next step
the test 2 on the frequencies noticed.
The values p=0.0000 are calculated
according to the distribution 2 with 3
degrees of freedom and indicate that the
logistic model is valid from a statistical
point of view, therefore the null
hypothesis can be rejected.
Table 2: Hosmer and Lemeshow test

Tabelul
3
prezinta
valorile
estimate ale coeficientilor de regresie ale
modelului de regresie logistica binara.
Valorile Sig. egale cu zero ne arata ca
toate cele trei variabile factoriale ale
modelului de regresie sunt semnificative
din punct de vedere statistic, si
influenteaza variabila dependenta statut.
De asemenea, valorile testului Wald ne
arata ca parametrii de regresie i sunt
diferiti de zero. Ipoteza nula este astfel

Step

Chisquare

df

Sig.

427.73
2

.000

Table 3 shows the estimated values


of regression coefficients of the model of
binary logistic regression. Sig. values,
equal to zero, show us that all the three
factorial variables of the regression
model are significant from statistical
point of view and they influence the

Annals of the Constantin Brncui University of Trgu Jiu, Economy Series, No. 1/2008

129

Analele Universitii Constantin Brncui din Trgu Jiu, Seria Economie, Nr. 1/2008

respinsa.

dependent variable of status. Similarly,


the Wald test values show us that the
Nivelurile
estimate
ale regression parameters are different
i
coeficientilor de regresie i sunt notate
from zero. Therefore the null hypothesis
cu B, iar Exp (B) reprezinta odds ratio is rejected.
(OR) pentru fiecare variabila factoriala,

The estimated levels of the


adica e i . Odds ratio este estimarea
riscul unui subiect de a se ramane regression coefficients i are marked
neangajat (0), la o modificare cu o with B and Exp (B) represents odds
unitate a variabilelor factoriale (cand ratio (OR) for each factorial variable,
aceasta este numerica, ca de exemplu which is e i . Odds ratio is the estimation
varsta in cazul nostru).
of a risk of a subject to remain
Vom avea deci ecuatia de regresie unemployed (0), at a change with one unit
of factorial variables (when it is numeric,
logistica:
such as age in our case).
ln(odds statut ) = 1.986 + 0.472( gen) + 0.66(v
Therefore we shall have the
(5)
logistic regression equation:
ln(odds statut ) = 1.986 + 0.472( gen) + 0.66(va
(5)

Tabelul 3: Variabile in ecuatia de regresie logistica / Variables in the logistic regression


equation
B

S.E.

Wald

Step 1

GEN
.472
.017
734.116
NVARSTA
.066
.007
83.356
NIVEL_E
.223
.009
589.486
D
Constant
-1.986
.027
5233.793
a Variable(s) entered on step 1: GEN, NVARSTA, NIVEL_ED.

Aceste
valori
estimate
ale
coeficientilor de regresie ne arata
legatura dintre variabilele factoriale si
variabila dependenta statut, cu cat
creste (sau descreste, daca semnul
coeficientului este negativ) valoarea
determinata log odds
a variabilei
statut=1 la modificarea cu o unitate a
uneia dintre variabilele factoriale,
influenta celorlalte factoriale fiind

df

Sig.

Exp(B)

1
1
1

.000
.000
.000

1.603
1.068
1.250

.000

.137

95.0%
C.I.for
EXP(B)
Lower
1.549
1.053
1.227

Upper
1.659
1.083
1.272

These estimated values of the


regression coefficients show us the
relation between factorial variables and
the dependent variable status, the more
it increases (or decreases, if the
coefficient sign is negative) the
determined value log odds of the variable
of status=1 at a change with one unit of
one of the factorial variables; the
influence of the other factorial is

Annals of the Constantin Brncui University of Trgu Jiu, Economy Series, No. 1/2008

130

Analele Universitii Constantin Brncui din Trgu Jiu, Seria Economie, Nr. 1/2008

considerata constanta.

considered to be constant.

Din tabelul 3 abservam ca riscul


angajarii creste cu 1.6 pentru subiectii
barbati inregistrati in baza de date supusa
analizei. In perioada analizata in baza de
date au fost inregistrati 33270 someri
femei si 47691 someri barbati. Dintre
acestia, la sfarsiul perioadei si-au gasit
loc de munca 6390 femei (19.21%), si
12979 barbati (27.21%). Desi exista un
numar mult mai mare de barbati
inregistrati ca someri in baza de date
analizata, totusi numarul celor care
reusesc sa isi gaseasca un loc de munca
este mai mare comparativ cu femeile,
ceea ce arata ca desi sunt mai multi
barbati someri totusi acestia sunt
preferati de catre angajatori.

From table 3 one can notice that


the employment risk increases with 1.6
for the male subjects registered in the
database subject to analysis. In the period
analyzed in the database there were
registered 33270 female unemployed and
47691 male unemployed persons. Among
them, at the end of the spell 6390 women
(19.21%) and 12979 men (27.21%) found
a job. Although there are a much greater
number of men registered as unemployed
in the analysed database, the number of
those who manage to find a job is higher
compared to women, which indicates that
although there are more unemployed
men, however they are preferred by
employers.

Observam de asemenea ca riscul


angajarii creste cu 1.06 la modificarea cu
o unitate (an) a variabilei varsta. Din 25
776 subiecti cu varsta cuprinsa intre 1524 ani si-au gasit loc de munca in
perioada analizata 4 982 subiecti,
reprezentand 19,33%, din 21 138 someri
cu varsta cuprinsa intre 25-34 ani s-au
angajat 5832, reprezentand 27,59%, din
18225 someri cu varsta cuprinsa intre 3544 ani s-au angajat 4920, reprezentand
27%, din 14452 subiecti someri cu varsta
cuprinsa intre 45-54 ani s-au angajat
23,39 % iar din 1370 someri cu varsta
cuprinsa intre 55-64 ani s-au angajat
18,61%. Observam ca sansa cea mai buna
de a isi gasi un loc de munca o au
persoanele cu varsta cuprinsa intre 24-54
ani. Acestia sunt fie absolventi de studii
superioare ce dupa o perioada scurta de
somaj isi gasesc locul de munca dorit fie
persoane cu experienta, ce inta in somaj
voluntar, generat de insatisfactii la
vechiu loc de munca, sau sunt
disponibilizati, dar isi gasesc un loc de
munca avantajati fiind de experienta
acumulata.

We can also notice that the


employment risk increases with 1.06 at
the change with one unit (year) of the age
variable. Of 25 776 subjects aged
between 15-24 years, 4 982 subjects
found a job in the analysed period,
representing 19,33%, of 21 138
unemployed aged between 25-34 years,
5832 became employed, representing
27,59%, of 18225 unemployed aged
between 35-44 years, 4920 became
employed, representing 27%, of 14452
unemployed subjects aged between 45-54
years, 23,39 % became employed and of
1370 unemployed aged between 55-64
years, 18,61% became employed. We can
notice that persons aged between 24-54
years have the best chance to find a job.
These are either university graduates that
after a short duration of unemployment
find the desired job, or experienced
persons that are voluntarily unemployed
because of dissatisfactions at the old job
or are laid off, but they can find a job
since they are favoured because of their
experience.

As for the educational level


In ceea ce priveste variabila nivel variable, the employment risk increases

Annals of the Constantin Brncui University of Trgu Jiu, Economy Series, No. 1/2008

131

Analele Universitii Constantin Brncui din Trgu Jiu, Seria Economie, Nr. 1/2008

educational, riscul angajarii creste cu


1.25 la modificarea cu o unitate a
acesteia. Intr-adevar, cu cat nivelul
educational este mai ridicat, cu atat
probabilitatea angajarii este mai mare
pentru un subiect.

with 1.25 at its change with one unit.


Indeed, the higher the educational level,
the higher the subjects employment
probability.

Annals of the Constantin Brncui University of Trgu Jiu, Economy Series, No. 1/2008

132

Analele Universitii Constantin Brncui din Trgu Jiu, Seria Economie, Nr. 1/2008

4. Concluzii

4. Conclusions

Scopul studiului este acela de a


determina riscul persoanelor din baza de
date studiata de a fi angajate sau
neangajate la sfarsitul perioadei supuse
analizei, risc estimat in functie de
variabilele independente gen, varsta si
nivel educational. Datele statistice au
fost obtinute de la Agentia Nationala a
Fortelor de Munca Bucuresti, si ofera
informatii privind subiectii inregistrati ca
someri in perioada 1 ianuarie 2002-31
august 2006. Esantionul cuprinde 80961
inregistrari, cu informatii referitoare la
data de intrare in somaj, data de iesire
din somaj, sex, varsta, nivel educational
si motivul iesirii din somaj pentru fiecare
persoana inregistrata. Ca metodologie am
folosit
regresia
logistica
binara.
Rezultatele analizei arata ca riscul
angajarii creste cu 1.6 pentru barbatii
inregistrati ca someri comparativ cu
femeile cu acelasi statut. In perioada
analizata in baza de date au fost
inregistrati 33270 someri femei si 47691
someri barbati. Dintre acestia, la sfarsiul
perioadei si-au gasit loc de munca 6390
femei (19.21%), si 12979 barbati
(27.21%). Desi exista un numar mult mai
mare de barbati inregistrati ca someri in
baza de date analizata, totusi numarul
celor care reusesc sa isi gaseasca un loc
de munca este mai mare comparativ cu
femeile, ceea ce arata ca desi sunt mai
multi barbati someri totusi acestia sunt
preferati de catre angajatori. Pentru
variabila varsta, riscul angajarii creste cu
1.06 la o modificare cu o unitate (an)
acesteia. Pe grupe de varsta, riscul
angajarii creste cu 1.3 pentru grupa de
varsta 15-24 ani comparativ cu grupa 5564 ani, cu 2.107 pentru grupa 25-34 ani,
comparativ cu grupa mentionata, cu
2.125 pentru grupa 35-44 ani comparativ
cu ultima grupa de varsta, si cu 1.664
pentru grupa 45-54, comparativ cu grupa

The purpose of this survey is to


determine the risk of the persons from
the analyzed database to become
employed or unemployed at the end of
the period subject to analysis, risk
estimated according to the independent
variables gender, age and educational
level. The statistical data were obtained
from
the
National
Agency
for
Employment Bucharest, and they provide
information on the subjects registered as
unemployed during January 1, 2002August 31, 2006. The sample includes
80961
records,
with
information
concerning the date of unemployment
beginning and end, sex, age, educational
level and the reason of unemployment
leaving for each registered person. As
methodology I used the binary logistic
regression. The results of the survey
show that the employment risk increases
with 1.6 for the men registered as
unemployed compared to women with the
same status. In the period analyzed in the
database there were registered 33270
female unemployed and 47691 male
unemployed persons. Among them, at the
end of the spell 6390 women (19.21%)
and 12979 men (27.21%) found a job.
Although there are a much greater
number of men registered as unemployed
in the analysed database, the number of
those who manage to find a job is higher
compared to women, which indicates that
although there are more unemployed
men, however they are preferred by
employers. For the age variable, the
employment risk increases with 1.06 at
its change with one unit (year). By age
groups, the employment risk increases
with 1.3 for the age group of 15-24 years,
compared to the group of 55-64 years,
with 2.107 for the group of 25-34 years,
compared to the specified group, with
2.125 for the group of 35-44 years,

Annals of the Constantin Brncui University of Trgu Jiu, Economy Series, No. 1/2008

133

Analele Universitii Constantin Brncui din Trgu Jiu, Seria Economie, Nr. 1/2008

55-64 ani. In ceea ce priveste variabila


nivel educational, riscul angajarii creste
cu 1.25 la modificarea cu o unitate a
acesteia. Intr-adevar, cu cat nivelul
educational este mai ridicat, cu atat
probabilitatea angajarii este mai mare
pentru un subiect. Cele mai dezavantajate
grupe educationale s-au dovedit a fi
persoanele fara studii, persoanele cu
scoala generala incompleta, scoala
profesionala si invatamant complementar
de ucenici si invatamant special si
persoanele absolvente de liceu teoretic.

compared to the last age group, and with


1.664 for the group of 45-54, compared
to the group of 55-64 years. As for the
educational
level
variable,
the
employment risk increases with 1.25 at
its change with one unit. Indeed, the
higher the educational level, the higher
the subjects employment probability.
The most disadvantaged educational
groups proved to be the persons without
education, the persons with unfinished
secondary school, vocational school and
apprenticeship complementary education
and special education and the theoretical
high school graduates.

Annals of the Constantin Brncui University of Trgu Jiu, Economy Series, No. 1/2008

134

Analele Universitii Constantin Brncui din Trgu Jiu, Seria Economie, Nr. 1/2008

Bibliografie:

References:

[1] Agresti, A. (2002), Categorical Data


Analysis. New York: Wiley-Interscience
[2] Aysit T., (2004), Determinants of
unemployment duration for men and
women in Turkey, Discussion Paper
2004/6,
http://www.tek.org.tr/dosyalar/ATANSEL-UNEMP.pdf
[3] Amemiya, T. (1985), Advanced
Econometrics. Harvard University Press
[4] Balakrishnan, N. (1991), Handbook of
the Logistic Distribution. Marcel Dekker,
Inc
[5] Berthoud R. (2003), Multiple
disadvantage in employment, Published
by Joseph Rowntree Foundation
[6]
Green,
William
H.
(2003),
Econometric Analysis, fifth edition.
Prentice Hall
[7] Ham, J. C. and Rea, A. S. (1987),
Unemployment Insurance and Male
Unemployment Duration in Canada,
Journal of Labor Economics, 5, 325-353.
[8]
Hosmer,
David
W.,
Stanley
Lemeshow (2000). Applied Logistic
Regression,
2nd
ed.
New
York
Chichester, Wiley
[9] Meghir C., Ioannides Y, Pissarides,
C. (1989), Female participation and male
unemployment duration in Greece:
Evidence from the labour force survey,
European Economic Review, Elsevier,
vol. 33(2-3), pages 395-406.
[10] Ramirez A (1998), Re-employment
probabilities of young workers in Spain,
Investigaciones Economicas, Vol 22,
Issue 2, pages 201-224.
[11] Tunali I and R. Asaad (1992),
Market
Structure
and
Spells
of
Employment
and
Unemployment:

Agresti, A. (2002), Categorical Data


Analysis. New York: Wiley-Interscience
Aysit T., (2004), Determinants of
unemployment duration for men and
women in Turkey, Discussion Paper
2004/6,
http://www.tek.org.tr/dosyalar/ATANSEL-UNEMP.pdf
Amemiya,
T.
(1985),
Advanced
Econometrics. Harvard University Press
Balakrishnan, N. (1991), Handbook of
the Logistic Distribution. Marcel Dekker,
Inc
Berthoud
R.
(2003),
Multiple
disadvantage in employment, Published
by Joseph Rowntree Foundation
Green, William H. (2003), Econometric
Analysis, fifth edition. Prentice Hall
Ham, J. C. and Rea, A. S. (1987),
Unemployment Insurance and Male
Unemployment Duration in Canada,
Journal of Labor Economics, 5, 325-353.
Hosmer, David W., Stanley Lemeshow
(2000). Applied Logistic Regression, 2nd
ed. New York
Chichester, Wiley
Meghir C., Ioannides Y, Pissarides, C.
(1989), Female participation and male
unemployment duration in Greece:
Evidence from the labour force survey,
European Economic Review, Elsevier,
vol. 33(2-3), pages 395-406.
Ramirez A (1998), Re-employment
probabilities of young workers in Spain,
Investigaciones Economicas, Vol 22,
Issue 2, pages 201-224.
Tunali I and R. Asaad (1992), Market
Structure and Spells of Employment and
Unemployment: Evidence from the
Construction Sector in Egypt, Journal of

Annals of the Constantin Brncui University of Trgu Jiu, Economy Series, No. 1/2008

135

Analele Universitii Constantin Brncui din Trgu Jiu, Seria Economie, Nr. 1/2008

Evidence from the Construction Sector in Applied Econometrics, 7, 339-367


Egypt, Journal of Applied Econometrics,
7, 339-367

Annals of the Constantin Brncui University of Trgu Jiu, Economy Series, No. 1/2008

136