Microsoft Word - 2008_01_02.doc
ORIGINAL PAPER
Job Stability and Gender Perspectives:
Application of a Logistic Regression Model
J. Antonio Ariza Montes, Mariano Carbonero Ruz,
Mª Del Carmen López Martín, Belén Gutiérrez Villar
Faculty of Business – ETEA – affiliated to the University Of Cordoba
KEY WORDS: Gender discrimination, Temporary employment, Logistic regression models
UDC: 331.564:316.647.82 JEL: B41, C51, J16
ABSTRACT ‐ Several changes have been taking place in the labour market since the 1970s that have
created the right climate to spur organisations and workers to demand greater flexibility in employment. In
this context, temporary employment has been the focus of many research papers and temporary contracts
have been used as a tool to achieve labour flexibility. In order to understand the situation in Andalusia
(Spain), this paper aims to identify the decisive factors in permanent employment. To this end, starting hy‐
potheses will be defined about the decisive factors in permanent employment and the positive or negative
significance of their influence; the starting hypotheses will then be tested empirically using a logistic regres‐
sion model on a sample population of wage earners in Andalusia. In the second stage, given that the ratio of
temporary contracts is much higher among women, the variable ‘gender’ is likely to be decisive in the con‐
struction of the regression model, therefore the decisive factors for permanent employment in Andalusia will
be evaluated separately for men and women, in order to calibrate the impact of gender on job stability. Fi‐
nally, based on the estimated probabilities of having a permanent job depending on gender, the degree of la‐
bour discrimination faced by women in the Andalusian labour market will be analysed.
Introduction
Many researchers have examined the decline of standard labour relations, in other words,
the model of employment in which the worker has a clearly defined employer, works full time
throughout the financial year under the supervision and premises defined by the employer, enjoys
certain legal and social benefits, and finally, expects this working situation to continue indefinitely
(Muckenberger, 1989; Schellenberg and Clark, 1996; Vosko, 1997).
The changes that have been taking place in the labour market since the 1970s have created
the right climate in many countries to spur organisations and workers to demand greater flexibility
in employment. Researchers in this area define nonpermanent employment as contingent work, in
other words, jobs in which the worker does not have an indefinite contract, either implicit or ex‐
plicit, or in which the minimum limit of two hours’ work could vary at any time (Connelly and
Gallagher, 2004).
Temporary employment is possibly the most common form of contingent work in the labour
market, which is why it has been the focus of many research papers in the field of labour relations.
Temporary contracts have been used as a tool to achieve labour flexibility, which is why they are
more commonly found in countries with strict labour relations ‐ such as Spain, France or Italy ‐ in
comparison with labour markets that are less tightly regulated, such as the United Kingdom or the
United States (Brown and Sessions, 2003; Brown and Sessions, 2005).
In order to understand the current labour market situation in Spain, we have to go back
around three decades. At the end of the 70s and early 80s, the situation was extremely worrying,
with unemployment rates reaching over 20% of the working population. This sparked the need to
implement flexibility measures when hiring workers in order to promote job creation. As a result
2007 ‐ 60 • Economic Analysis®
of the new legislative framework provided by the labour reforms introduced in the 1980s, the
presence of temporary contracts increased from 8% of all contracts in 1984 to 33% in 1994, and in a
short space of time Spain had one of the highest rates of temporary employment in Europe.
Subsequent reforms introduced between 1994 and 2002 attempted to change this trend, by
applying measures aimed at increasing job stability, mainly by restricting the facilities granted to
employers previously in terms of temporary contracts and reducing certain methods of dismissal.
In spite of these policies, currently, the ratio of temporary contracts in Spain is still one of the high‐
est in the European Union1. In an attempt to remedy this situation, on the 9th June 2006, a Royal
Decree Reform Act was passed for the labour market, which chiefly aimed to reduce temporary
contracts and promote stable employment. To this end, certain measures were established that pro‐
mote the application of permanent contracts and reduce certain business contributions. The effects
of these latest reforms will only become clear over the next few years.
Certain authors specify that there is no rational reason to prefer a temporary contract, since a
permanent contract offers all the advantages of a temporary one with the added bonus that the
worker is able to prolong the working relationship for as long as she or he wishes (Brown and Ses‐
sions, 2005). However, this argument is not shared by the scientific community as a whole. García
and Rebollo (2005) suggest two basic points of view in relation to temporary contracts. On the one
hand, they can be viewed as a previous step before moving onto other more stable situations
(Booth, Francesconi and Frank, 2002; Varejao and Portugal, 2002), especially among younger job‐
seekers, who are looking to gain experience. On the other hand, the temporary link with a com‐
pany does not always have to be imposed and in some cases it can be the free choice of the indi‐
vidual. However, the latter alternative is fairly improbable in the case of Spain, given that tempo‐
rary contracts are somewhat detrimental from an economic point of view in this country (Jimeno
and Toharia, 1993; De La Rica, 2004; García and Rebollo, 2005).
Whatever the case may be, the segregation between employees with stable contracts and
those with temporary contracts creates a significant dualism in the Spanish labour market, as
pointed out by several authors (Bentolila and Dolado, 1994; Saint Paul, 1996; Toharia, 2002). As if
they were two different models, according to García and Rebollo (2005), there are two clearly sec‐
tors in the labour market. On the one hand, a primary sector characterised by workers with per‐
manent contracts who have a strong influence on collective bargaining processes. This group en‐
joys job stability, high salaries and opportunities for internal promotion. And on the other hand,
there is a secondary sector, made up of workers with temporary contracts, limited salaries, low
internal mobility and who are excluded from bargaining processes.
Another characteristic feature of the Spanish labour market is that the ratio of temporary to
permanent contracts varies a great deal from region to region. Hence, in the fourth quarter of 2006,
regions such as Andalusia, Extremadura and Murcia recorded figures of over 40% temporary em‐
ployment, whereas in Catalonia and La Rioja, three out of every four wage earners had a perma‐
nent contract. This regional variation could be the result of several factors: heterogeneity in the
composition of the workforce, production specialisation, different levels of productivity, labour
costs, etc.
Given the impact of temporary work in the Spanish and Andalusian labour markets (Anda‐
lusia ‐ 46.4% temporary contracts ‐ has the highest ratio of any of Spain’s autonomous regions), this
paper aims to identify the decisive factors in permanent employment. To this end, firstly, a review
will be provided of previous research in this area; starting hypotheses will be defined about the
decisive factors in permanent employment and the positive or negative significance of their influ‐
1 Hence, in Q4 2006, 33.8% of salaried workers had temporary contracts, although these figures conceal a marked gender
inequality, since this percentage drops to 32.0% among men and increases to 36.2% among women
Volume 40 • Spring 2008 • 61
ence; the starting hypotheses will then be tested empirically using a logistic regression model on a
sample population of wage earners in Andalusia taken from the National Institute of Statistics Ac‐
tive Population Survey. In the second stage, given that the ratio of temporary contracts is much
higher among women (36.2% in Spain and 49.2% in Andalusia) than among men (32.0% and 44.6%,
respectively), the variable ‘gender’ is likely to be decisive in the construction of the regression
model2, therefore the decisive factors for permanent employment in Andalusia will be evaluated
separately for men and women, in order to calibrate the impact of gender on job stability. Finally,
based on the estimated probabilities of having a permanent job depending on gender, the degree
of labour discrimination faced by women in the Andalusian labour market will be analysed.
Temporary vs. permanent contracts: empirical background
The pairing temporary/permanent employment has been studied from various angles. Many
authors have focused on analysing the harmful effects of temporary employment: lack of stability
and opportunities for professional development (Farber, 1999; Arulampalam and Booth, 1998),
demotivation and staff turnover (Purcell et al., 1999), dissatisfaction, fewer opportunities for train‐
ing and lower wages (Booth, Francesconi and Frank, 2002; Aronsson et al., 2002, Nollen, 1996;
Segal and Sullivan, 1997)3, less job security, lower health and safety standards (Quinlan, Mayhew
and Bohle, 2001), insecurity and concern about the future (Purcell, 2000; Sengenberger, 1995).
More in keeping with the objectives of this paper, another area of research has focused on de‐
termining the variables associated with temporary or permanent employment. Dolado et al. (2002)
performed regression analysis for Spain, confirming that age is negatively associated with tempo‐
rary employment. These conclusions are coherent with the trinomial logistic model (in which the
reference category, being unemployed, is compared with the situation of temporary and perma‐
nent wage earners) created by Morris and Vekker (2001), who concluded that individuals between
the age of 16 to 24 are less likely to sign a permanent contract and more likely to have temporary
employment. Furthermore, the probit analysis performed by Dolado et al. (2002) confirms that
there is a lower incidence of temporary employment among older wage earners. Similar results
were obtained in studies performed by Nollen (1996), Farber (1997, 1999), Segura et al. (1991) and
Álvarez Aledo (1996).
Training is also negatively associated with temporary employment, as confirmed by Dolado
et al. (2002) in a study comparing wage earners who gained a secondary education and those with
a lower level of education. Brown and Sessions (2005) arrived at similar conclusions using perma‐
nent employment as the point of reference. They applied probit analysis to a sample of close to
9000 wage earners and concluded that every year an individual remains within the education sys‐
tem increases his/her chances of gaining permanent employment. Other authors have obtained
empirical evidence in this respect, including Nollen (1996), Farber (1997, 1999), Álvarez Aledo
(1996) and Segura et al. (1991), confirming that temporary contracts are much more frequent
among uneducated workers in comparison to other educational categories.
2 According to a group of gender researchers, new non‐standard working methods merely contribute to the perpetuation
of gender inequalities, since historically atypical labour relations (outside the regulations of full‐time permanent em‐
ployment) have always been linked (and continue to be so) with women (Fudge, 1991; Armstrong, 1996; Spatler‐Roth
and Hartmann, 1998; Vosko, 2000; Fudge and Vosko, 2001)
3 The study carried out by Brown and Sessions (2003) quantifies the difference between the salaries earned by permanent
and temporary employees, giving a figure of 13%. Some researchers in Spain have obtained similar results (Jimeno and
Toharia, 1993; Alba‐Ramírez, 1994). Furthermore, Brown and Sessions (2003) point out that salary discrimination affects
women more than men regardless of whether the contract is permanent or temporary
2007 ‐ 62 • Economic Analysis®
Marital status is another variable that certain researchers had attempted to link with the type
of contract signed by the wage earner. Segura et al. (1991) claim that temporary contracts are more
common among single people than the married population, in line with the results obtained by
Morris and Vekker (2001) for a sample of women. According to these authors, being married in‐
creases an individual’s chances of having a permanent contract and reduces the chances of having
a temporary contract.
However the link between marital status and contract type does not always manifest itself as
expected, since Laird and Williams (1996) and Golden (1996) found that the proportion of married
women in the labour market is positively linked to levels of temporary employment.
Cranford et al. (2003) state that the role taken on by women, taking care of the everyday and
generational needs of their families, causes them to choose part‐time or temporary employment of
their own free will. Consequently, it can be expected that the presence of children in the home,
especially little children, or another dependent relative, will increase the frequency of temporary
contracts among the female population. This hypothesis was tested by Morris and Vekker (2001)
on a sample of women with children under the age of 18 still living at home, concluding that this
variable reduces the probability of having a permanent contract, increasing the possibilities of
temporary employment.
The type of work undertaken is also linked to the type of contract signed by the wage
earner, in the sense that jobs that require a higher degree of skill, and which are therefore more
difficult to fill with qualified staff, must be duly compensated, among other things with greater job
stability, in other words the offer of a permanent contract. The probit model developed by Brown
and Sessions (2005) confirms this link among professionals, and with even greater intensity among
management positions.4 Dolado et al. (2002) examined at this same area and noted that temporary
contracts are less frequent in qualified positions.
Another interesting variable when studying this issue is the sector or industry. The probit
model created by Dolado et al. (2002) reveals that temporary contracts are negatively associated
with employment in the public sector, a similar conclusion to the findings obtained previously by
Segura et al. (1991). Furthermore, the study performed by Dolado et al. (2002) concludes that there
are fewer temporary contracts in the industrial sector and the services industry, and a higher per‐
centage in the construction industry. Breaking it down even further, Segura et al. (1991) found a
higher incidence of temporary employment in construction, agriculture, commerce and financial
institutions.
Research has also been conducted into the influence of nationality and parttime work, al‐
though not to the same extent as for other variables. In relation to the first variable, Morris and
Vekker (2001) demonstrated that non‐US wage earners were less likely to have a permanent con‐
tract and more likely to have a temporary one. In relation to the second variable, the analysis con‐
ducted by Brown and Sessions (2005) concluded that permanent employment is negatively linked
with part‐time work.
And finally, one of the fundamental variables in this research project, which has purposefully
been left until last, is the gender of the wage owners. The increasing integration of women into the
workforce in recent decades is irrefutable, although this situation masks considerable diversity that
exists at different levels and in different forms (Bradley et al., 2000). Certain authors such as Stand‐
ing (1989) believe that there is no such thing as fairness in the labour market, and that women tend
to cluster at the very lowest levels, occupying positions with the lowest salaries, which are the least
4 The authors already confirmed this hypothesis in a previous study together with the positive association between tem‐
porary contracts and smaller companies (25‐100 employees) and the inverse relationship between the variable studied
and the presence of trade union representation (Brown and Sessions,2003)
Volume 40 • Spring 2008 • 63
secure and with the worst professional expectations5. This fact is borne out by official statistics: as
indicated previously, at the end of 2006 the rate of temporary employment among women in Spain
was more than four percentage points higher than among men; furthermore, 11.9% of women
work part‐time in comparison to just 4.1% of men; or to cite another example, one out of every five
women works in unskilled employment (as opposed to 11.9% of men), whereas only five out of
every one hundred women occupy management positions (8.4% among men). Studies that have
analysed the relationship between gender and labour relationships have concluded that women
are more likely to have a temporary contract than men. The probit analysis conducted by Brown
and Sessions (2005) confirms that permanent employment is positively linked with the male gen‐
der. These results are coherent with the findings of Dolado et al. (2002), who confirm that being a
female wage earner in the private sector increases the probability of having a temporary contract.
This effect is even more marked in the case of public employment.
The research of Booth et al. (2001) is particularly relevant for the purposes of this paper, since
these authors performed separate multinomial logistic regression studies for men and women. The
conclusions of this research showed that women who occupy skilled positions (professionals, spe‐
cialists and teachers) are more open to the idea of having a temporary contract than men. Further‐
more, in comparison with the male population, these authors noticed that the probability of
women having a temporary contract is higher in public administration and charity organisations.
Finally, having young children is also a discriminatory factor, since the regression models obtained
confirm that in such a family context the chance that a woman will have a temporary contract is
higher than for men.
Research hypotheses
The above bibliographic review shows which variables are linked, in different markets, with
the type of contract signed by wage earners, as well as the foreseeable influence (positive or nega‐
tive) of each relationship. Taking these previous studies as a point of reference, the main working
hypotheses to be tested in relation to temporary contracts in the Andalusian labour market are as
follows:
o Hypothesis 1. Gender, age, level of education, family structure, nationality and la‐
bour‐related aspects are decisive factors when determining the probability of gaining
permanent employment.
o Hypothesis 2. The above‐mentioned factors have a different impact depending on
gender.
o Hypothesis 3. All things being equal regarding the other variables, women in Anda‐
lusia have fewer chances of gaining permanent employment than men.
Empirical study
Methodology: binary logistic regression model
The methodology used to achieve our research objectives is based on the binary logistic re‐
gression model, a specific type of dichotomous response regression model6. This statistical tech‐
5 Krahn (1995) claims that the probability that women will remain in non‐standard employment (including self em‐
ployed, part‐time and temporary work) is higher in comparison with men
2007 ‐ 64 • Economic Analysis®
nique determines the probability of the occurrence of an event ‐ having a permanent contract in
this case ‐ compared to the probability of the occurrence of the opposite event.
Sample
In order to analyse which factors determine permanent employment in Andalusia, a sample
population was obtained from the microdata file of the Active Population Survey (EPA) for Q4
2006. In order to avoid the problem of oversensitivity of the model, which can occur when an ex‐
cessively high number of observations are considered, 1628 records were selected: 929 Andalusian
men and 699 Andalusian women7.
The sample was then divided in two. The first part, comprising 80% of the observations, was
used to train the model, and the second part, containing the remaining 20% of observations, was
used for model validation. The comparison between the percentages of success in both stages indi‐
cates the general validity of the model and detects possible overtraining. The division of the sam‐
ple into two sub‐samples was conducted in a random manner, although the initial structure re‐
garding contract duration and gender was maintained.
The variables included in the model were taken from the EPAʹs anonymised microdata file.
In some cases, new variables had to be created in order to reflect the most important aspects of
each subject’s family structure, and in other cases, a certain degree of recoding had to be applied in
order to simplify the study or adapt the content of the original data to the objectives pursued in
this paper. The coding and frequency of the variables included in the regression model are shown
in table 1.
As a prelude to regression analysis, the degree of association was measured between each
pair of independent variables, chosen by a coefficient in each case, depending on the nature of the
6 The most common way of presenting this model is:
The first member of this equation is called the odds ratio (OR) and represents the relative probability of the researched
event as opposed to non‐occurrence. Hence, an OR with a value of x would indicate that it is x times more likely that an
individual will have a permanent contract as opposed to a temporary one. From this point of view, the interpretation of
the model coefficients is simple. Once the variable Xj has been chosen along with the other independent variables, in the
event of any unitary increase in this predictor, relationship between the odds ratios will be calculated as follows:
Hence, if βj>0, the increase in the odds ratio will be greater than 1, which indicates an increase in the probability of oc‐
currence of the researched event. Just as in any statistical model, logistic regression can test several hypotheses in order
to determine its validity both in global and individual terms. Global analysis evaluates as a whole the possibility that the
investigated phenomenon really can be modelled using the chosen type of expression, whereas individual analysis ‐ in
which validity is subject to the global analysis performed ‐ examines the appropriateness of including each of the vari‐
ables in the model considered. In this paper, the Hosmer‐Lemeshow goodness‐of‐fit index was used as a global meas‐
urement, and the Wald test was applied for individual analyses.
7 These amounts were chosen by configuring the gender composition of the EPA sample on the one hand and the num‐
ber of independent variables of each of the estimated models on the other, taking a ratio of about 85 observations per
variable in the case of women and 115 in the case of men, which are sufficient values to be able to maintain the freedom
of the model at a reasonable level. The selection of these observations was made by sampling the original EPA file , re‐
stricting the sample to Andalusian individuals over the age of 16 engaged in salaried employment during the reference
period, and by sampling proportionally by gender and contract duration
Volume 40 • Spring 2008 • 65
variables (nominal or ordinal). The maximum value has been found to be between Occup1 and
Leduc (0.669). These values do not seem to indicate significant evidence of colinearity.
Table 1 ‐ Explanatory variables: coding and frequency
Variable and coding Name Frequency
Value 0 Value 1 Value 2 Value 3 Value 4
AGE (0: 16‐25; 1: 26‐40; 2:41‐55; 3: 56 or over) Age5 419 704 448 57 ‐‐‐‐
GENDER (1: male; 0: female) Gend1 699 929 ‐‐‐‐ ‐‐‐‐ ‐‐‐‐
NATIONALITY (1: Spanish; 0: foreign) Nat1 56 1572 ‐‐‐‐ ‐‐‐‐ ‐‐‐‐
LIVES WITH PARTNER (1: Yes; 0: No) Livp 600 1028 ‐‐‐‐ ‐‐‐‐ ‐‐‐‐
LEVEL OF EDUCATION (0: primary educa‐
tion; 1:secondary education; 2: higher educa‐
tion)
Leduc 255 993 380 ‐‐‐‐ ‐‐‐‐
OCCUPATION (1: unskilled; 2: skilled, opera‐
tional work; 3: skilled, administrative work; 4 :
highly skilled)
Occup1 ‐‐‐‐ 302 717 350 259
SECTOR/INDUSTRY (0: agriculture; 1: indus‐
try; 2: construction; 3: services)
Sect1 90 192 213 1133 ‐‐‐‐
NUMBER OF CHILDREN IN THE HOME (0:
none; 1: one; 2: two; 3: three; 4: four)
Chil 998 337 263 29 1
WORKING DAY (1: full‐time; 0: part‐time) Fpar1 182 1446 ‐‐‐‐ ‐‐‐‐ ‐‐‐‐
Source: Authors’ own calculations.
NB: the variable Chil indicates the number of children under the age of 15 living with the person interviewed.
Results analysis
The analysis was conducted in two stages: in the first stage, considering both sub‐samples
(men and women) as a single sample, the aim was to analyse which variables or factors determine
job stability among Andalusian wage earners, with particular emphasis on gender, since if this
factor is significant, the separate estimation of probabilities for men and women is justified, which
is the second objective of this research paper.
General model
Table 2 shows the results of the estimation using logistic regression of the factors that deter‐
mine the chances of gaining permanent employment in Andalusia. Firstly, it should be pointed out
that the comparison statistic applied to evaluate the validity of the model as a whole indicates that
there are sufficient reasons to accept the validity of the model8, in other words, to confirm that the
fact of whether a wage‐earner has a permanent contract or not can be satisfactorily explained by
the set of variables considered in this paper. This general model shows that the regression coeffi‐
cients for age, the level of skill required for the occupation, working full time and the number of
children under 15 living in the home, provide positive results, thereby contributing to permanent
employment. Hence, the older the individual, the higher the skill level required for the job per‐
formed, the number of children living in the home and the fact that the individual works full time,
8 The Hosmer‐Lemeshow test, used for this purpose, presents the following results: Chi‐Squared: 9.230; Sig. 0.323
2007 ‐ 66 • Economic Analysis®
the greater the probabilities of having a permanent job. However, the coefficients for the variables
gender, nationality and the agriculture and construction industries are negative, which suggests
that being a woman, foreign, or working in these two economic sectors decreases the individual’s
possibilities of finding permanent employment, increasing the probability of having a temporary
contract9.
Table 2 ‐ Logistic Regression: factors that determine permanent employment
in Andalusia and the odds ratio confidence intervals
Variables in the model Odds ratios95% C.I. for OR
B S.D. Wald p OR Lower Upper
Age5 0.0595 0.0063 88.9376 0.0000 1.0613 1.0483 1.0745
Gend1(1) ‐0.3840 0.1528 6.3182 0.0120 0.6812 0.5049 0.9189
Nat1(1) ‐0.7761 0.3535 4.8214 0.0281 0.4602 0.2302 0.9200
Occup1 0.4005 0.0790 25.6737 0.0000 1.4925 1.2783 1.7426
Fpar1 1.3679 0.2099 42.4576 0.0000 3.9269 2.6024 5.9257
Chil 0.1674 0.0833 4.0379 0.0445 1.1823 1.0041 1.3920
Sect1 59.1200 0.0000
Sect1(1) ‐1.5949 0.3454 21.3167 0.0000 0.2029 0.1031 0.3994
Sect1(2) ‐0.0489 0.2246 0.0474 0.8276 0.9523 0.6132 1.4788
Sect1(3) ‐1.3271 0.2642 25.2359 0.0000 0.2653 0.1581 0.4452
Constant ‐2.7344 0.3606 57.5015 0.0000
Source: author’s own calculations.
NB: the variable sector/industry is linked to the industrial sector, therefore the coding would be as follows: Sect1(1),
agriculture; Sect1(2), construction; Sect1(3), Services.
In spite of everything, the impact of each of the significant variables on the probability of
gaining permanent employment differs substantially from one variable to another, as indicated in
the confidence intervals analysis for the corresponding odds ratios (see table 2). Hence, each year
that passes in relation to the variable age only slightly increases the odds ratios (approximately 6%
of the value of the previous year), which is a minimal increase in the probability of gaining perma‐
nent employment. The effect is much more intense in relation to the variable gender. In this case,
the OR for women is 68.12% compared with the OR recorded for men, which reveals the difficulty
women have accessing jobs with greater stability10. Furthermore, being foreign is also a major ob‐
stacle to accessing a permanent contract: 46.02% relative probability in comparison to Spanish‐born
citizens. The OR for the variable occupation (1.4925 with a confidence interval of between 1.2783
and 1.7426) suggests that jobs that require a greater degree of skill also offer greater job stability. As
expected, the most intense effect is observed when comparing part‐time workers with people who
work full time, with an increase of almost four points in the OR. A similar occurrence, although
with less intensity, is observed when analysing the influence of the presence of children in the
home: for every child under the age of 15 in the home, the relative probability of having permanent
employment increases by 18%. Finally, the sector or industry in which the individual works also
plays an important role in the type of contracts signed: taking industry, agriculture and construc‐
tion as the points of reference, the ratios are much lower (20.29% in the first case and 26.53% in the
9 All these results are significant at 1%. For this level of significance, the logistic regression model indicates that there is a
series of variables that are neither positively or negatively associated with job stability: having or not having a spouse,
the educational level of the wage earner, and the fact that the individual works in the services industry in comparison
with the industrial sector
10 Even in terms of confidence intervals, even though the upper limit for the ratio evaluated is slightly above 90%, the fact
that the lower boundary of this same ratio is around 50% cannot be ignored In other words, the probability of a woman
having a permanent contract, in comparison with a man, is 91.89% in the best case scenario, dropping to 50.49% in the
worst case scenario
Volume 40 • Spring 2008 • 67
second), which shows that the probability of having stable work in these sectors is much lower
than in the industrial sector. This situation ‐ when compared to the services sector, which has a
ratio of 95% ‐ shows that the chances of having permanent employment in this sector are practi‐
cally the same as in the industrial sector11.
The final step in this first research stage analyses the model’s generalisation capacity, by
comparing the correctly classified observations in both the training and generalisation stages.
Hence, in the training stage, corresponding to the construction of the model, 73.37% of the indi‐
viduals cons idered were classified correctly, knowing their real situation beforehand. There were,
however, considerable differences between the percentages corresponding to permanent (48.03%)
and temporary contracts (87.01%), which suggests that the latter are easier to identify. In the gener‐
alisation stage, once the model coefficients were estimated, individuals were presented for classifi‐
cation with previous information about their real situation. The results obtained were better than
expected in the training stage (77.23% of the global model, 50.88% for permanent contracts and
91.47% for temporary contracts) both in global terms and for each of the groups, which reveals the
general validity of the estimated model.
Gender model
Given that the variable gender proved to be decisive in the construction of the general model,
the second research objective was to compare the regressions corresponding to the male and fe‐
male population, thereby calibrating the impact of gender on job stability. The results of the Hos‐
mer‐Lemeshow test (Chi Squared: 8.930; Sig. 0.348), in relation to the model corresponding to the
male population once again confirmed that the variables considered were sufficient in order to
explain the type of contract, at least in terms of determining the characteristics that affect it.
Table 3 presents the results obtained from the logistic regression model for the male popula‐
tion sample. The main difference from the general situation relates to the presence of children in
the home. This variable, which showed a positive association in the global model, has no influence
when considering just the male population.
Furthermore, analysis of the odds ratios shows that the effect of the significant variables on
permanent employment differs in certain cases from the global model. Firstly, the type of working
day has more of an influence on job stability among men in comparison to the collective as a whole
(4.76 in comparison to 3.92). Secondly, and in the opposite direction, the importance of nationality
as a decisive factor of job stability is considerably lower among men: an odds ratio of 0.2186 com‐
pared to 0.4602 for the general model. For the other variables, differences are insignificant, indicat‐
ing that age, the level of skill required for the job and the economic sector do not display specifi‐
cally different behaviour for the male collective in the determination of the type of contract to
which they have access.
Just as with the global analysis, the modelʹs generalisation capacity for the male population
is adequate, obtaining similar classification percentages to the general model12, with only exception
of a small decrease in the number of correctly classified individuals in the generalisation stage re‐
garding the total number of individuals with a permanent contract13.
11 It should be noted that the logistic regression model explains the relationship between variables, which does not neces‐
sarily have to be causal. For example, the positive association between the presence of children in the home and perma‐
nent employment does not imply that the former is a source of job stability and the relationship might even be present in
the opposite direction
12 Training stage: 74.97% for the male population as the whole; 49.19% for men with a permanent contract and 87.88% for
temporary contracts. Generalisation stage: 76.88%, 45.16% and 92.74%, respectively
13 At any rate, the figures in themselves and the scant differences between the training and generalisation stages indicate
that the evaluated model is sufficiently valid in general and is not predetermined by the specific observations used in its
2007 ‐ 68 • Economic Analysis®
Table 3 ‐ Logistic regression: factors that determine permanent employment in the Andalusian male popula‐
tion and odds ratios confidence intervals
Variables in the model Odds ratios 95% C.I. for OR
B S.D. Wald p OR Lower Upper
Age5 0.0562 0.0085 44.0292 0.0000 1.0578 1.0404 1.0755
Occup1 0.5219 0.1269 16.9292 0.0000 1.6853 1.3143 2.1610
Fpar1 1.5603 0.4362 12.7915 0.0003 4.7600 2.0243 11.1930
Sect1 50.3354 0.0000
Sect1(1) ‐1.3314 0.3939 11.4251 0.0007 0.2641 0.1220 0.5716
Sect1(2) 0.0256 0.2703 0.0090 0.9244 1.0260 0.6040 1.7428
Sect1(3) ‐1.3774 0.2908 22.4338 0.0000 0.2522 0.1427 0.4460
Nat1(1) ‐1.5206 0.5243 8.4102 0.0037 0.2186 0.0782 0.6108
Constant ‐2.8537 0.5500 26.9170 0.0000
Source: authorsʹ own calculations.
NB: the variable sector/industry is linked to the industrial sector, therefore the coding would be as follows: Sect1(1),
agriculture; Sect1(2), construction; Sect1(3), Services.
Focusing now on the group of women studied, the Hosmer‐Lemeshow test confirms the va‐
lidity of the model (Chi squared: 5.475; Sig. 0.706). The variables that are really associated with
permanent contracts are summarised in table 4, which reveals that the main difference as regards
to the situation of men focuses on two variables14: on the one hand, the presence of children in the
home is related in this case with having or not having a permanent job; on the other hand, nation‐
ality, which was a relevant factor when determining job stability among men, does not influence
this aspect among the female population.
As in the previous analyses, in order to determine the effect of each explanatory variable on
the dependent variable, the estimated odds ratios and their associated confidence intervals were
considered (see table 4). Firstly, of particular importance is the fact that the presence of children in
the home is directly associated with the probability of having a permanent contract. Furthermore,
comparison of the odds ratios for the common variables between the logistic regression models for
men and women reveals, in general, fairly similar values; therefore, from variable to variable, there
are no significant differences in the way in which the different circumstances affect the type of con‐
tract. The only exception, apart from the aforementioned presence of children in the home, can be
observed in terms of the sector or industry in which the individual works, since the probability of
having permanent employment in agriculture (always in relative terms in comparison to industry)
is substantially lower among women than men (OR=0.2641 for men and OR=0.0802 for women)
and slightly lower in the services industry (1.03 as opposed to 0.8218).
To bring this section to a close, it should be pointed out that the classification percentages of
the training and generalisation stages are slightly different to those obtained in the other two mod‐
els analysed15. Furthermore, significant improvements were observed in the generalisation stage in
comparison to the training stage, both for permanent and temporary employment. Moreover, sub‐
stantial differences were observed in the number of correctly classified individuals according to
the type of contract in favour of temporary contracts, just as with the model applied to the male
population, which indicates that the profile of a temporary female worker is easier to predict than
construction
14 It should be emphasised that the construction industry was excluded from this analysis, given that the number of
women employed in this sector is very low and that all those present in the sample had a permanent contract, which
would significantly distort the coefficient estimation associated with this level
15 Training stage: 71.92% for the female population as a whole; 54.81% for women with a
permanent contract and 82.27% for temporary contracts. Generalisation stage: 80.15%, 68.00% and 87.21%, respectively
Volume 40 • Spring 2008 • 69
that of a woman with a permanent job. Just as with the other two logistic regression models, these
results confirm that the model evaluated is sufficiently valid in general and is not predetermined
by the specific observations used in its construction.
Table 4 ‐ Logistic regression: factors that determine permanent employment among the
Andalusian female population and odds ratios confidence intervals
Variables in the model Odds ratios 95% C.I. for OR
B S.D. Wald p OR Lower Upper
Age5 0.0666 0.0098 46.5539 0.0000 1.0689 1.0486 1.0896
Occup1 0.2817 0.1063 7.0196 0.0081 1.3254 1.0761 1.6325
Fpar1 1.4364 0.2521 32.4603 0.0000 4.2055 2.5658 6.8932
Chil 0.3439 0.1313 6.8585 0.0088 1.4104 1.0904 1.8244
Sect1 11.3621 0.0034
Sect1(1) ‐2.5234 0.8056 9.8109 0.0017 0.0802 0.0165 0.3889
Sect1(2) ‐0.1963 0.4201 0.2183 0.6403 0.8218 0.3607 1.8721
Constant ‐3.2485 0.5708 32.3951 0.0000
Source: authors’ own calculations.
NB: the variable sector/industry is linked to the industrial sector, therefore the coding would be as follows: Sect1(1),
agriculture; Sect1(2), construction; Sect1(3), Services.
Gender discrimination in the Andalusian labour market
The previous analyses evaluated the influence of the variables studied in this research paper
on the type of contract signed by wage earners in Andalusia, placing specific emphasis on the vari‐
able gender. This last section makes a graphic comparison between the situation of men and
women. For this purpose, for each of the individuals surveyed, regardless of their gender, the
probability of having a permanent contract was calculated according to the logistic regression
models obtained (see graph 1).
Graph 1. Probability of having permanent employment according to a gender
Source: authors’ own.
2007 ‐ 70 • Economic Analysis®
If the employability of men was similar to that of women, each and every one of the points
should be located on the diagonal, where both probabilities are equal. When this does not occur,
the location of each point indicates which gender offers greater opportunities for permanent em‐
ployment: men when the point is lower than the diagonal or women when it is higher. This graph
clearly shows that when the other variables are constant, being a man offers greater expectations
for stable employment than being a woman. Furthermore even in cases in which being a woman is
a competitive advantage in terms of job stability, the probability of having permanent employment
is very high for both sexes (top right‐hand corner of the graph), so in actual fact this advantage is
relative.
In order to quantify the content of the graph, the observations were classified into two crite‐
ria: the real gender of each individual and that which, in terms of probability, would offer greater
prospects of having permanent employment. For this second criterion, three possibilities were
taken into account: man, woman, or irrelevant. Here, the term irrelevant refers to situations in
which the difference in the probability of accessing permanent employment is less than 0.05. The
results shown in table 5 confirm that the possibilities of finding permanent employment are in
general much higher for men. In fact, only 49 (7.1%) of the 688 women interviewed and 26 (3.5%)
of the 727 men interviewed would have a better chance of finding permanent employment as
women. However, 412 of the women interviewed, representing 59.9% of the total, would have a
much greater chance as men.
Table 5 ‐ Probabilistic comparison of employability in terms of the real gender of the individual interviewed
and the preferable gender to access permanent employment
Real Preferable
Irrelevant Male Female Total
Male 307 394 26 727
Female 227 412 49 688
Total 534 806 75 1415
Source: authors’ own.
Discussion
The research aims of this paper are: on the one hand, to determine which factors affect the
probability of having a permanent job contract among wage earners in Andalusia, which had the
highest rate of temporary employment of any autonomous region in Spain at the end of 2006; and
on the other hand, to evaluate the impact of gender on job stability.
In order to achieve these objectives, firstly a logistic regression model was applied, indicating
that the probability that a wage earner has of getting a permanent contract increases with age, the
level of skill required for the position, full‐time employment and the number of children living at
home. It was also confirmed that being a woman, being foreign and working in agriculture or con‐
struction reduces the possibilities of finding permanent employment. Our results practically con‐
firm the approach of the first hypothesis: gender, age, some characteristics of family structure, na‐
tionality and labour‐related aspects are as being decisive factors in determining the probabilities
for permanent employment. Our conclusions are similar to previous results in the field back‐
ground.
Given that gender is a decisive variable in the logistic regression model, the model coeffi‐
cients were then estimated for the male and female populations. The most significant results were:
Volume 40 • Spring 2008 • 71
on the one hand, that having or not having children living at home does not alter the possibilities
of finding permanent employment among Andalusian men, whereas it does influence job stability
for women; and, on the other hand, that being a Spanish national ‐ a factor that contributed to
permanent employment among the male population ‐ does not show any influence among the fe‐
male population. This result confirms our second hypothesis.
Finally, from the perspective of gender, a graphic estimation was generated of the probability
of each individual interviewed having a permanent job contract, regardless of gender and main‐
taining all the other variables constant. The evidence obtained confirms the expectations in the
hypothesis 3: being a man in Andalusia offers greater prospects for permanent employment; hence
any legislative measures taken to promote equality among men and women will be welcome in
order to redress this situation.
There are two limitations for this study. Firstly, we focus on only one region; therefore our re‐
sults cannot be generalized to other regions where the labour conditions can be different from the
situation in Andalusia. Second ly, the study is based on selfreported cross‐sectional data, which
hinders causal inferences. Nevertheless, these restrictions will be a target for future for research.
Despite these limitations our study yields valuable new information about the decisive factors
identifying permanent employment and its combined associations regarding gender.
Reference
Alba‐Ramírez, A. (1994) “Formal training, temporary contracts, productivity and wages in Spain”, Oxford Bulletin of
Economics and Statistics, 56, 2, 151–70.
Álvarez, C. (1996) El impacto de la contratación temporal sobre el sistema productivo español, Madrid: Consejo Económico y
Social de España.
Armstrong, P. (1996) The Feminization of the Labour Force: Harmonizing Down in a Global Economy. Rethinking Restructuring:
Gender and Change.
Aronsson, G., Gustafsson, K. and Dallner, M. (2002) “Work environment and health in different types of temporary jobs”,
Work. European Journal of Work and Organizational Psychology, 11(2), 151‐75.
Arulampalam, W. and Booth, A. L. (1998) “Training and labour market flexibility: is there a trade‐off?”, The British Journal
of Industrial Relations, December, vol. 36(4) pp. 521–36.
Bentolila, S. and Dolado, J.J. (1994) “Spanish Labour Markets”. Economic Policy, 18, pp. 55‐99.
Booth, A.L., Francesconi, M. and Frank, J. (2001) “Temporary jobs: who gets them, what are they worth, and do they lead
anywhere?” ISER Working Paper No. 00‐13, University of Essex, April.
Booth, A.L., Francesconi, M. and Frank, J. (2002) “Temporary Jobs: Stepping Stones or Dead Ends?” The Economic Journal,
112 (June), 189‐213.
Bradley, H., Erickson, M., Stephenson, C. and Williams, S. (2000) Myths at Work, Cambridge: Polity. Brown, S. and Ses‐
sions, J.G. (2003) “Earnings, education, and 21 fixed‐term contracts”, Scottish Journal of Political Economy, Vol. 50, No. 4,
September.
Brown, S. and Sessions, J.G. (2005) “Employee Attitudes, Earnings and Fixed‐Term Contracts: International Evidence”,
Review of World Economics, Vol. 141 (2).
Connelly, C. E. and Gallagher, D.C. (2004) “Emerging trends in contingent work research”, Journal of Management, 30,
956‐83.
Cranford, C.J., Vosko, L.F. and Zukewich, N. (2003) “The Gender of Precarious Employment in Canada”, Industrial Rela‐
tions, 58(3), 454‐82.
De La Rica, S. (2004) “Wage Gaps Between Workers with Indefinite and Fixed‐Term Contracts: The Impact of Firm and
Occupational Segregation”, Forthcoming in Moneda y Crédito.
Delsen, L. (1995) Atypical Employment: An International Perspective—Causes, Consequences, and Policy. Groningen: Wolters‐
Noordhoff.
Dolado, J.J., García‐Serrano, C. and Jimeno, J.F. (2002). “Drawing Lessons from the Boom of Temporary Jobs in Spain”,
The Economic Journal, 112 (June), 270‐95.
2007 ‐ 72 • Economic Analysis®
Farber, H. S. (1997). “The changing face of job loss in the United States, 1981–1995”, Brookings Papers: Microeconomics, 55–
128.
Farber, H.S. (1999) “Alternative and part‐time employment arrangements as a response to job loss”. Journal of Labour
Economics, 17, 4, 142–69.
Fudge, J. (1991) Labour Law’s Little Sister: The Employment Standards Act and the Feminization of Labour. Ottawa: Canadian
Centre for Policy Alternatives.
Fudge, J. and Vosko, L.F. (2001) “By Whose Standards? Re‐Regulating the Canadian Labour Market”, Economic and Indus‐
trial Democracy, Vol. 22, No. 3, 327.
García, J.I. and Rebollo, Y. (2005) The use of permanent and temporary jobs across Spanish regions: Do unit Labour cost differen‐
tials offer an explanation? Sevilla: Fundación Centro de Estudios Andaluces, Junta de Andalucía.
Golden, L. (1996) “The Expansion of Temporary Help Employment in the U.S., 1982‐ 1992. A Test of Alternative Eco‐
nomic Explanations”, Applied Economics, 28, September, 1127‐41.
Jimeno, J.F. and Toharia, L. (1993) “The effects of Fixed‐term employment on wages: theory and evidence from Spain”,
Investigaciones Economicas, 16, 3, pp. 475–94.
Krahn, H. (1995), “Non‐standard Work on the Rise”, Perspectives on Labour and Income, Winter, 35. Laird, K. and Williams,
N. (1996) “Employment Growth in the Temporary Help Supply Industry”. Journal of Labour Research, 17, 663‐81.
Morris, M. and Vekker, A. (2001) “An Alternative Look at Temporary Workers, Their Choices, and the Growth in Tempo‐
rary Employment”, Journal of Labour Research, 22(2), 373‐90.
Muckenberger, U. (1989) “Non‐standard Forms of Employment in the Federal Republic of Germany: The Role and Effec‐
tiveness of the State” in Precarious Employment in Labour Market Regulation: The Growth of Atypical Employment in Western
Europe. G.
Rodgers and S. Rodgers, eds. Geneva: International Institute for Labour Studies, 227.
Nollen, S.D. (1996), “N egative aspects of temporary employment”, Journal of Labour Research, XVII, 4, pp. 567–82.
Purcell, K., Hogarth, T. and Simm, C. (1999) Whose Flexibility? The Costs and Benefits of Non‐Standard Working Arrangements
and Contractual Relations, York: Joseph Rowntree Foundation, York Publishing Services.
Purcell, K. (2000) “Gendered employment insecurity?” In Heery, E. and Salmon, J. (eds) The Insecure Workforce, pp. 112–
139. Andover: Routledge.
Quinlan, M., Mayhew, C. and Bole, P. (2001) “The global expansion of precarious employment, work disorganization,
and consequences for occupational health: A review of recent research”, International Journal of Health Services, 31, 335–
414.
Saint Paul, G. (1996) “On the Political Economy of Labour Market Flexibility”, NBER Macroeconomics Annual, 8, pp. 151‐
196.
Schellenberg, G. and Clark, C. (1996) Temporary Employment in Canada: Profiles, Patterns and Policy Considerations, Ottawa:
Canadian Council on Social Development.
Segal, L.M. and Sullivan, D.G. (1997) “The Growth of Temporary Services Work”, Journal of Economic Perspectives, 11, 117‐
36.
Segura, J., Durán, F., Toharia, L. and Bentolila, S. (1991) Análisis de la contratación temporal en España, Madrid: Ministerio
de Trabajo y Seguridad Social.
Sengenberger (1995) “Employment protection and employment security”, in Sengenberger and Campbell. (eds) Creating
Economic Opportunities: The Role of Labour Standards in Industrial Restructuring. Geneva: Institute for Labour Studies.
Spalter‐Roth, R. and Hartmann, H. (1998) Gauging the Consequences for Gender Relations, Pay Equity and the Public Purse.
Contingent Work: Employment Relations in Transition. K. Barker and K. Christensen, eds. Ithaca, N.Y.: Cornell University
Press, 69.
Standing, G. (1989) “Global feminization through flexible labour”, World Development, 17 July. Toharia, L. (2002) “El
modelo español de contratación temporal”, Temas laborales, 64, pp. 117‐139.
Varejao, J. and Portugal, P. (2002). Why do Firms Used Fixed‐Term Contracts? Mimeo.
Vosko, L.F. (1997) “Legitimizing the Triangular Employment Relationship: Emerging International Labour Standards
from a Comparative Perspective”, Comparative Labour Law and Policy Journal, Vol. 19, fall, 43.
Vosko, L.F. (2000). Temporary Work: The Gendered Rise of a Precarious Employment Relationship. Toronto: University of To‐
ronto Press.