Duration of Regional Unemployment
Spells in Slovenia
Darja Boršic
Alenka Kavkler
The paper begins with an overview of the unemployment rate in Slove-
nia and focuses on duration of unemployment and regional charac-
teristics of the unemployment rates. It is shown that the dispersion of
regional unemployment rate is gradually decreasing and is also slightly
below European average on nuts 3 level. The analysis of the duration
of regional unemployment spells is based on the data obtained from the
Employment Office of the Republic of Slovenia, which consists of the
unemployment spells between January 1st, 2002 and November 18th,
2005 with more than 450,000 entries. The Kaplan-Meier estimates of
the survival function are presented and the effects of region on the du-
ration of unemployment spells are discussed.
Key Words: unemployment, regions, survival analysis, Kaplan-Meier
estimator, Slovenia
jEL Classification: P33, P34
Introduction
In the previous system the regions within transition countries had been
more or less equally developed and had registered similar volume of eco-
nomic activity. The structural change witnessed by these countries due
to adopting the market economies in many cases caused substantial re-
gional differences.
According to Huber (2007) the regional consequences of transition
raise a number of issues: identifying the causes of regional differences,
labour market mechanisms regarding wage flexibility and migration,
new firm creation and economic policy response to regional disparities.
Answering these questions could help in coping with characteristics of
national labour markets, and thus, improve the efficiency of economic
policy.
Dr Darja Boršic is an Assistant Professor at the Faculty of Economics
and Business, University of Maribor, Slovenia.
Dr Alenka Kavkler is a Teaching Assistant at the Faculty of Economics
and Business, University of Maribor, Slovenia.
Managing Global Transitions 7 (2): 123-146
Analyses of regional unemployment in transition countries include
Römisch and Ward (2005) and Landesmann and Römisch (2006). Re-
cent review papers on the topic of regional unemployment include El-
horst 2003, Huber 2007, Ferragina and Pastore 2008. Empirical estimates
of regional unemployment and its persistence in transition economies
are presented in Bornhorst and Commander (2006). Western European
countries are empirically tested in Bayer and Jüssen (2006), while Taylor
and Wren (1997) evaluate uk regional policy in the mid-1990s. Feragina
and Pastore (2008) address two cases for explaining regional disparities
in transition countries: (1) regions with high unemployment deal with
low job creation rates due to similar labour market trends after the tran-
sition on the national level, (2) on the other hand the regional differences
are an outcome of diverse speed of restructuring.
In the last couple of decades different survival analyses and duration
techniques have gained popularity in the social sciences to model the
length of unemployment spells and strike duration. Moffitt (1999) ap-
plied the usual econometric techniques in labour economics, includ-
ing the proportional hazard methods and the duration models. Exam-
ples of duration model applications in labour markets can be found in
Green and Riddell (1995), D'Agostino and Mealli (2000), and Arranz and
Romero (2003). They explain the effects of different determinants of un-
employment duration for Canada, nine EU15 members and Spain, re-
spectively. Newell and Pastore (2006) apply duration models to estimate
the effect of regional unemployment variation on aggregate economic
restructuring in Poland. Factors of unemployment duration in Ukraine
are discussed in Kupets (2006) by using the Cox proportional hazard
model with two competing risks. The author concludes that age, mari-
tal status, income during unemployment and local demand constraints
significantly affect the duration of unemployment.
Analytical papers investigating unemployment in Slovenia are rather
rare. Slovenian unemployment is analysed by Vodopivec (1995a; 2004),
Domadenik and Pastore (2006), Orazem, Vodopivec, and Wu (2005)
and Kavkler and Boršic (2006; 2007). Vodopivec (1995a) uses a duration
model in presenting effects of unemployment insurance on the unem-
ployment duration. While Domadenik and Pastore (2006) test the im-
pact of education and training systems on the participation of young
people in the labour markets of Slovenia and Poland. Kavkler and Boršic
(2006; 2007) estimate the effects of age, gender and education on dura-
tion of unemployment by duration data techniques.
All the above listed papers are devoted to the aggregate labour mar-
ket in Slovenia. To the best of our knowledge, there is no analytical pa-
per dealing with regional issues of unemployment in Slovenia. Thus, this
paper attempts to estimate the effect of regional disparities on the dura-
tion of unemployment in Slovenia by duration data techniques, namely
the Kaplan-Meier estimator. Although Slovenia is in general regarded as
relatively equally developed throughout the geographical area, regional
differences do exist. There are disparities of regional activity rates and
regional gdp. The pace of restructuring after the transformation of the
economic system was not equally distributed throughout the country.
Consequently, regional unemployment rates differ significantly. That is
why we expect the results of our empirical analysis to confirm the sig-
nificant effect of regional differences on the duration of unemployment
in Slovenia. Given that the analysis is based on a huge database consist-
ing of more than 450,000 entries, this paper sheds light on additional
information about regional differences and notably contributes to the
existing studies about the Slovenian labour market. Namely, analysing
regional labour market issues can give information about labour mar-
ket flexibility, which is of prime importance for an effective functioning
of monetary union. Additionally, such analysis also provides vital infor-
mation for establishing an appropriate structural funds policy. Thus, the
message of this paper can be important not only to national but also to
eu authorities.
The paper is structured as follows. The introduction is followed by
an overview of the characteristics of total unemployment rate in Slove-
nia presenting important stylized facts about the labour market against
which regional differences evolved. Then, the regional unemployment
rate is discussed and compared to European regional developments.
Next, a description and preliminary analysis of the dataset is presented. It
is followed by a brief presentation of duration models and survival analy-
sis. Results of Kaplan-Meier estimates are discussed. The paper concludes
with a short summary of the main findings.
An Overview of Aggregate Unemployment Rate in Slovenia
The low rate of registered unemployment in Slovenia prior to transi-
tion (below 2%) is not difficult to explain, since the old economic system
provided assurance in the labour market by striving to achieve full em-
ployment and equal wealth distribution. Consequently, severe regulation
of labour market was necessary in order to provide jobs for practically
everyone. The unlimited assurance of employees was even a constitu-
tionally guaranteed right in former Yugoslavia (Vodopivec 1995b). After
the economic transformation the registered rate of unemployment rose
to more than 14.4% in 1993. It remained above 13.9% until 1998, when it
gradually started to fall and had reached 7.7% in 2007.1
Significant changes occurred in labour demand during the transition,
which caused the high increase in the unemployment rate at the begin-
ning of the nineties. Young people suddenly had no assurance of getting
a job after completing education. Since the high level of employment in
the former system was artificial, they were also unlikely to get a job eas-
ily. In order to produce more efficiently, most of the companies started
with massive lay-offs. Many older workers became unemployed and were
also very unlikely to get a new job. The relative advantage of high edu-
cated workers increased. Higher educated employees were changing jobs
more easily; they were less likely to become unemployed and had better
chances of re-employment, if needed.
According to Vodopivec (1995b), women represented a higher share
in the two highest vocational classes (managers and leading clerks) in
comparison to men in the mid-1980s. Consequently, women in Slovenia
had no disadvantages in the labour market at the beginning of transition,
which was not the case in other transition economies.
According to Kajzer (1998) the relatively high level of unemployment
in the nineties is mostly a consequence of a combination of the following:
•	increasing disequilibrium in the labour market was prolonging the
long term unemployment;
•	the effect of increasing investment activity on employment was neu-
tralised by the structure of investment and increasing structural dis-
crepancies;
•	the effect of economic growth on employment was decreased by
wage growth, hidden unemployment among the employed in the
former system and employers' caution in employment.
The main characteristics are low level of employment and high level
of unemployment among the low educated, extremely low employment
rate of older people, relatively high rate of unemployment among young
people and non-intensive human resource management in enterprises
(Kajzer 2005).
Table 1 presents the duration of unemployment in Slovenia in the last
seven years. It can be seen that the highest share of unemployment is
table 1 Duration of unemployment in Slovenia in % of total number of unemployed
Duration	2000	2001	2002	2003	2004	2005	2006	2007
Up to 3 months	17.9	20.2	20.2	23.6	25.5	22.7	22.4	21.7
From 3 till 6 months	10.0	13.2	13.0	14.0	14.3	13.9	11.3	12.6
From 6 till 9 months	5.5	6.2	7.0	8.4	8.2	8.8	8.1	7.8
From 9 till 12 months	5.2	5.8	7.5	8.0	7.5	8.3	7.5	7.2
From 1 till 2 years	15.3	14.2	18.5	19.1	19.4	18.3	20.4	18.2
From 2 till 3 years	12.1	9.0	8.1	9.7	9.1	9.5	9.7	10.3
From 3 till 5 years	16.4	13.5	10.2	7.3	7.7	8.8	9.7	10.1
From 5 till 8 years	10.5	9.3	7.5	4.4	3.6	4.5	5.3	6.3
More than 8 years	7.2	8.6	7-9	5-4	4-7	5.2	5-5	5-9
Total	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0
Total in persons	104,583 104,316 99,607 95,993 90,728 92,575 78,303 68,411
note Based on data available on the Employment Office of the Republic of Slovenia
web site (http://www.ess.gov.si/).
represented by the spells of less than 3 months. This is mainly due to the
fact that all unemployed receive a financial compensation as a certain
share of their previous wage in the first three months of unemployment.
The duration of up to 3 months is followed by the duration of 1 to 2 years,
which represents the long-term unemployment. It has increased in the
last seven years from 15.3% in 2000 to 20.4% in 2006. On the other hand,
very long-term unemployment, measured as duration of unemployment
above 2 years, amounted to 30.2% of total unemployed workers in 2006,
which is 16 percentage points less than in 2000.
Relatively generous unemployment benefits throughout the observed
period makes people less motivated for searching for a new job (Kajzer
2005). This can be also confirmed by observing the difference among
no2 unemployment rate and registered unemployment rate. In Slove-
nia, there is a large discrepancy among the two rates of unemployment.
Table 2 shows significant differences among the rates, although it is grad-
ual in the recent years. This is mainly due to above mentioned financial
aid, relatively high level of informal work and a high share of long term
unemployed who become passive and do not meet the second criterion
of actively seeking for a job in the Labour Force Survey (Kajzer 2005).
Consequently, registered rates imply that Slovenia has high unemploy-
ment, while no rates show that it has low unemployment. Taking into
account international criteria, the ilo rates should be considered. We
table 2 Comparison of the registered and no unemployment rates for Slovenia
Unemployment rate	2000	2001	2002	2003	2004	2005	2006	2007
Registered	12.2	11.6	11.6	11.3	10.6	10.2	9.4	7.7
ilo	6.7	6.2	6.3	6.7	6.3	6.5	6	4.9
Difference (% points)	5-5	5-4	5-3	4.6	4-3	3-7	3-4	2.8
note Based on data available on the Employment Office of the Republic of Slovenia
website (registered rates, see http://www.ess.gov.si/) and on the Eurostat website (ilo
rates, see http://ec.europa.eu/eurostat).
present the comparison of Slovenia in the eu according to ilo rates in
the rest of this section. While the analytical part of the paper below is
based on registered unemployment data, since the database required for
the duration analysis is not available in the Labour Force Survey but was
provided by the Slovenian Employment office.
According to the Labour Force Survey the unemployment rate in
Slovenia is below the European average. The light gray columns for indi-
vidual countries in figure 1 present total ilo unemployment rates, while
the gray columns denote the long term unemployment rate, measured as
a share of long term unemployed3 in the active population of individual
countries. Both unemployment rates for Slovenia are below EU15 and
EU25 in 2006. The total unemployment rate is 6.0%, which is 1.4 and 1.9
percentage points below the EU15 and EU25 averages, respectively.
While the long term unemployment in Slovenia is 2.9%, which is 0.2
percentage point below EU15 average and 0.7 percentage point below the
EU25 average. The highest total unemployment rates were recorded for
Poland (13.8%), Slovakia (13.4%) and France (9.5%), while the lowest
rate is 3.9% for Denmark and Netherlands. As far as the long term un-
employment rate is concerned, the highest rate was noted for Slovakia
(10.8%), Poland (7.8%) and Germany (4.6%). On the other hand, the
lowest long term unemployment rates are recorded for Denmark (0.8%),
Cyprus (0.9%) and Sweden (1.1%).
Regional Unemployment in Slovenia
According to the Statistical Office of the Republic of Slovenia there
are twelve statistical regions in Slovenia: Pomurska, Podravska, Ko-
roška, Savinjska, Zasavska, Spodnjeposavska, Jugovzhodna Slovenija,
Osrednjeslovenska, Gorenjska, Notranjsko-kraška, Goriška and Obalno-
kraška. Registered unemployment rates in these regions clearly indicate
the East-West distribution. Eastern regions, such as Pomurska and Po-
EU25	3.6	7.9
EU15	3.1	7.4
BE	4.2	8.2
cz	3.9	7.1
DK	0.8	3.9
DE	4.7	8.4
EE	2.8	5.9
IE	1.4	4.4
GR	4.8	8.9
ES	1.9	8.5
fr	4	9.5
it	3.4	6.8
cy 0.9	4.6
Lv	2.5	6.8
lt	2.5	5.6
lu	1.4	4.7
hu	3.4	7.5
mt	2.9	7.3
NL	1.7	3.9
at	1.3	4.7
pl	7.8	13.8
pt	3.8	7.7
si	2.9	6
sk	10.2	13.4
fi	1.9	7.7
se	1.1	7.1
uk	1.2	5.3
figure 1 ilo unemployment rates in eu25, 2006, in % (light gray - harmonized
unemployment rates, yearly averages; gray - long-term unemployment
in % of active population)
dravska, have the highest unemployment rates while Western regions,
such as Gorenjska, Goriška and Obalno-kraška, have the lowest unem-
ployment rates. The order of regions is the same regardless of which
unemployment rate is taken into account. The difference is only in the
levels as noted above for the aggregate data (table 2).
The reasons for the East-West distribution of unemployment rates
can be mainly attributed to different patterns of economic activity in the
former system as well as different pace of restructuring after the eco-
nomic transformation. Namely, in the former system eastern regions
were characterized by a large share of manufacturing activities specifi-
cally specialized in labour intensive industries, such as textile industry,
automobile industry, etc. When restructuring took place many manu-
facturing companies were abolished, leaving behind huge numbers of
unemployed workers, who had difficulties in finding new jobs due to at
least two reasons: (1) required qualification in other sectors and (2) the
problem of overall lower labour demand. Despite the fact that many new
small firms arose, their labour demand was/is much lower compared to
the large socialist companies in the previous system. Moreover, investi-
gating patterns of regional specialization and concentration of manufac-
turing, Traistaru, Nijkamp and Resmini (2002) found that in Slovenia
regional specialization had not changed significantly in the period 1990-
1999. Thus, the regional distribution of sectoral activity has not changed
much in comparison to the former system. Consequently, the high
level of unemployment persists in regions with a high share of manu-
facturing.
Differences among regions can be presented by the coefficient of vari-
ation. This is one of the structural indicators by which the eu follows the
implementation of the Lisbon strategy in employment, innovation and
research, economic reforms, social cohesion and environment (Pecar
2005). According to Eurostat's methodology, the coefficient of variation
of regional unemployment rates is based on the weighted variance of un-
employment rates, which is defined as:
where xi denotes unemployed persons in region i, yi represents active
population in region i, X and y stand for the averages of xi and yi, and x/y
is the unemployment rate at national level. The coefficient of variation of
unemployment rates is the square root of the variance stated in Equation
1 divided by the unemployment rate at national level. It gives a measure
of the regional spread of unemployment rates.
In the period 2000-2002 the coefficient of variation was increasing and
it reached the highest point of 35.1% in 2002. Afterwards it was decreas-
ing and it reached the lowest value of 30.8 in the first half of 2006. This
development of coefficient of variation implies that the regional differ-
ences in the unemployment rate in Slovenia have been decreasing during
the last four years (figure 2).
Also the regional level of unemployment and its regional dispersion
can be compared at the European level. Eurostat (2006) divides statisti-
cal units according to the Nomenclature of Territorial Units for Statistics
(1)
figure 2 Coefficient of variation of regional unemployment in Slovenia, 2000-2006
(adapted from Pecar 2006)
(nuts) adopted in July 2003 and revised in 2005. The nuts regulation
comprises three main levels of statistical regions according to popula-
tion size nuts 1 (3 million-7 million), nuts 2 (800,000-3 million) and
nuts 3 (150,000-800,000). Below nuts 3 level, there are so called Local
Administrative Units (lau), which represent districts and municipalities
and are not subject to nuts regulation. Thus, nuts classification divides
Europe into 89 regions on nuts 1 level, 254 regions on nuts 2 level and
1214 regions on nuts 3 level (Eurostat 2007).
According to the size of Slovenia, the whole country belongs to nuts
2 level, while each of the 12 statistical regions described above belongs
to nuts 3 level. Thus, Slovenia as a whole is one of the 254 regions and
Slovene statistical regions are 12 among 1214 European regions.
At nuts 2 level the lowest rates of unemployment in 2005 were
recorded in Herefordshire, Worcestershire and Warwickshire in uk
(2.6%), Provincia Autonoma Bolzano/Bozen in Italy (2.7%) and North
Yorkshire in uk (2.9%). While the highest regional unemployment rates
were attained in Vyhodne Slovensko in Slovakia (23.1%), Dolnoslaskie
(22.8%) and Zahodnieopomorskie (22.7%) in Poland (Mlady 2006). As
noted above, at the same level Slovenia has reached the unemployment
rate of 6.5% (sors 2006).
Table 3 presents regional unemployment rates at nuts 3 level for 2005.
The first part are the lowest rates, below 3%, in the second part are the
12 statistical regions in Slovenia and the third part of the table shows the
highest unemployment rates in Europe, above 25%. It can be noted that
the unemployment rates below 3% were recorded for 6 regions in uk,
4 regions in the Netherlands, 3 regions in Italy and 2 regions in Austria.
On the other hand, the highest unemployment rates were registered for 8
German regions, 3 Polish regions and 1 Greek region. Slovenian regional
table 3 European Regional Unemployment Rates, nuts 3 level, 2005
Region (country)	Rate
EU25 minimum	
Overig Zeeland (nl)	2.4
Oxfordshire (uk)	2.4
Bolzano-Bozen (it)	2.7
Bologna (it)	2.7
North and North East Somerset, South Gloucestershire (uk)	2.7
Salzburg und Umgebung (at)	2.8
Innsbruck (at)	2.8
Warwickshire (uk)	2.8
Kop van Noord-Holland (nl)	2.9
Midden-Noord-Brabant (nl)	2.9
Devon cc (uk)	2.9
Oost-Zuid-Holland (nl), North Yorkshire (uk), Surrey (uk)	3-0
Slovenia	
Goriška	4.2
Gorenjska	4.7
Obalno-kraška	4.8
Osrednjeslovenska	4.9
Notranj sko -kraška	5.1
Jugovzhodna Slovenija	5.6
Koroška	6.8
Spodnjeposavska	7.4
Savinjska	8.1
Podravska	8.7
Zasavska	8.8
Pomurska	11.0
Continued on the next page
unemployment rates in 2005 range from 4.2% in Goriška to 11.0% in
Pomurska region.
At nuts 3 level Eurostat has information about 21 countries. Their
dispersion of regional unemployment rates, measured by the coefficient
of variation, is shown in figure 3. In 2005 the estimated coefficient of
variation for Slovenia was 30.9%, which is 0.9 percentage point below the
Duration of Regional Unemployment Spells in Slovenia 133
table 3 Continued from the previous page
Region (country)	Rate
EU25 maximum
Nordvorpommern (de)	25.0
Ostvorpommern (de)	25.7
Uckermark (de)	25.9
Görlitz, Kreisfreie Stadt (de)	26.0
Sangerhausen (de)	26.0
Mansfelder Land (de)	26.1
Jeleniogorsko-walbrzyski (pl)	26.5
Slupski (pl)	27.2
Elcki (pl)	27.3
Demmin (de)	27.9
Kastoria(GR)	28.5
Uecker-Randow (de)	29.8
average of 31.8%. Below the average are also differences among regions in
Sweden (15.6%), Ireland (16.6%) and Denmark (20.3%), while the most
evident are the differences among regional unemployment rates in Italy
(62.5%), Czech Republic (46.6%) and Germany (45.3%).
In general, the European regional dispersion of unemployment rate
has decreased at nuts 3 level in the last three years. Considerable falls in
the regional difference were recorded for Italy (19.4 percentage points),
Hungary and Germany (both 6.8 percentage points). On the other hand,
there were also some gaps expanded, as for Slovakia, Estonia and Lithua-
nia with an increase in coefficient of variation by 6.8, 6.1 and 3.7 percent-
age points, respectively.
Regional Comparison of the Length of Unemployment Spells
in Slovenia
The data for our empirical investigation were obtained from the Em-
ployment Office of the Republic of Slovenia. The database consists of the
unemployment spells completed between January 1st, 2002 and Novem-
ber 18th, 2005 and all of the ongoing spells on November 18th, 2005. For
each of the unemployment spells, the start and end date and the factors
sex, age, level of education and statistical region were made available to
us. Since the Employment Office of the Republic of Slovenia is not al-
cz	46.6
dk	20.3
de	45.3
EE	33.8
IE	16.6
gr	29.9
es	33.2
FR	35.7
it	62.5
LV	23.4
LT	20.7
hu	29.9
nl	25.5
AT	40.8
PL	21.9
PT	29.9
si	30.9
sk	42.3
FI	29.2
se	15.6
uk	34.2
figure 3 Dispersion of regional unemployment rates in eu (nuts level 3) in 2005,
in %
lowed to disclose personal data about the unemployed, only a personal
identifying number was added to enable identification of repeated spells.
455,581 unemployment spells are included in our database with the max-
imal length of 13,547 days.
In a preliminary analysis, the descriptive statistics for the 348,281 spells
completed on November 18th, 2005 were calculated. The mean, standard
deviation and the 95% confidence intervals of the mean for the 12 Slove-
nian regions can be found in table 4. Already from the 95% confidence
intervals for the mean one may observe significant differences among
different regions.
One of the visual aids to present such results are the boxplots (figure
4). The boxplot (also called the box-and-whiskers plot) summarizes a
single numeric variable within categories of another variable. Each box
shows the median, the quartiles and the whiskers that extend to the last
point within 1.5 times the interquartile range. The outliers are usually
also given in the boxplots, but we left them out due to the fact that our
distribution has a very long right tail.
The difference between Podravska, Zasavska and Savinjska region on
the one hand and Gorenjska, Goriška and Obalno-kraška region on the
table 4 Descriptive statistics for the duration of unemployment spells, in days
Region	(1)	(2)	(3)	(4)
Total	348,281	479.69	791.11	(477.07, 482.32)
Pomurska	31,122	472.54	815.10	(463.49, 481.60)
Podravska	76,393	565.38	906.26	(558.95, 571.81)
Koroška	13,633	467.75	756.92	(455.05, 480.47)
Savinjska	50,941	509.86	809.46	(502.83, 516.89)
Zasavska	9,873	505.27	791.79	(489.65, 520.89)
Spodnjeposavska	14,542	486.75	813.92	(473.52, 499.98)
Jugovzhodna Slovenija	19,503	510.35	893.39	(497.81, 522.88)
Osrednjeslovenska	60,716	426.60	660.58	(421.34, 431.85)
Gorenjska	33,227	375.98	636.94	(369.13, 382.83)
Notranjsko-kraška	7,947	452.62	714.22	(436.92, 468.33)
Goriška	12,938	395.48	668.45	(383.96, 407.00)
Obalno-kraška	16,736	430.46	779-92	(418.64, 442.28)
Column headings are as follows: (1) N, (2) mean, (3) standard deviation, (4) 95% confi-
dence interval for the mean.
o_500_1000_1500_2000
Unknown
Pomurska
Podravska >—	• -1
Koroška >—	• -1
Savinjska 1—	• -1
Zasavska 1—	• -1
Spodnjeposavska 1—	• -1
Jugovzhodna Slovenija 1—	• -1
Osrednjeslovenska 1—	• -1
Gorenjska 1-	• -1
Notranjsko-kraška 1—	• -1
Goriška >—	• -1
Obalno-kraška >—	• -1
figure 4 Boxplots depicting the duration of unemployment spells (in days)
for different regions
other hand is obvious. The regions of Gorenjska and Goriška are the
most advantageous in the labour market with mean length of unem-
ployment spells of 376 and 395 days, respectively. The unemployed from
Podravska region are in the worst position, as they have to wait for 565
days on average to find a new job. The mean length of unemployment
spells for the unemployed from Savinjska, Zasavska and Jugovzhodna
Slovenija is slightly above 500 days, whereas the mean lengths of other
regions are between 400 and 500 days. The region is thus a crucial factor
when searching for a job. Note that the boxplot in figure 4 displays the
median of the length of the unemployment spells, whereas in table 4 the
mean length of the unemployment spells is given.
To test the null hypotheses that the mean duration of unemployment
spells is the same for every region, we performed the nonparametric
Kruskal-Wallis test. anova is inappropriate in our case, because the dis-
tribution is asymmetrical. The null is strongly rejected, since the p-value
is lower than 10-6.
The above described database allows us to take a look into different
characteristics of duration of unemployment by regions:
1.	According to gender, in Osrednjeslovenska and Notranjsko-kraška
region men and women have similar durations of unemployment.
The biggest difference among them is in Zasavska region, where
women need on average 43 days more than men to find a new job.
Women are in a worse position also in Pomurska, Podravska and
Koroška region. On the other hand, it takes longer for men to get
re-employed in Goriška and Spodnjeposavska region.
2.	Regarding age, in all regions the unemployed aged between 40 and
60 years are in the worst position. It is interesting to note that young
unemployed (18 years or less) have a relatively long duration of un-
employment in Jugovzhodna Slovenija, Spodnjeposavska, Zasavska,
Podravska and Pomurska region.
3.	As for level of education, the length of regional unemployment in
general decreases with higher levels of education with a few excep-
tions. In many regions the longest unemployment was experienced
by those who finished 3-year lower vocational education (Obalno-
kraška, Goriška, Osrednjeslovenska, ...). Also post-secondary vo-
cational education does not prove the above statement, as this level
of education required longer unemployment than lower levels of
education in many regions. The next exception is master's degree
in Obalno-kraška, Gorenjska, Spodnjeposavska, Koroška and Po-
dravska, and doctorate in Jugovzhodna Slovenija and Zasavska.
Survival Analysis and Duration Models
Survival analysis and duration models originate in biostatistics, where
the survival time is the time until death or until relapse of an illness.
During the recent years these techniques have also gained popularity in
the social sciences to model the length of unemployment spells and the
strike duration. One of the unavoidable problems encountered when an-
alyzing the duration data is the so-called censoring. Since the event under
observation (i. e. death or the end of the unemployment spell) has often
not occurred till the end of the study, it is only possible to estimate the
lower bound of the survival time. This kind of censoring is called the
right censoring.
A comprehensive overview of the methods and models used in sur-
vival analysis is given by Therneau and Grambsch (2000) and by Klein
and Moeschberger (1998).
basic notions
Let the random variable T denote the survival time. The distribution
function of T is defined by the equation F(t) = P(T < t) and measures
the probability of survival up to time t. Since T is a continuous random
variable, its density function can be computed as the first derivative of
the distribution function f (t) = F'(t). The survival function S(t) denotes
the probability to survive until time t or longer and is given by
S(t) = P(T > t) = 1 - F(t).	(2)
The limit
IM V P(t<T <t + 6\T>t)
Mt) = lim----(3)
<5^0	0
represents the risk or proneness to death at time t. The function A(t) is
usually called the hazard function or the failure rate and measures the
instantaneous death rate given survival until time t.
By integrating the hazard function over the interval [0, t] one obtains
the so-called cumulative hazard function
A(t) = f A(u)du.	(4)
o
In addition to defining basic notions, we shall also derive the relations
between them that will be needed in the following subsections. Obvi-
ously,
... Pit <T<t + S\T>t)
Ait) = lim---=
<5^0	0
P(t < T < t + ö)/P(T > t)
= lim- =
<5^0	ö
1 F(t + 6)-F(t) F (t)	. .
= W)l™—s— = w	(5)
It follows from the definition of the survival function S(t) given by
equation (4) that F'(t) = -S'(t), therefore
Rewriting the last equation in the form A(u)du = -dlogS(u) and inte-
grating from 0 to t yields
- logS(t) = f A(u)du = A(t),	(7)
0
therefore
S(t) = e-A(t).	(8)
We have observed the fact that logS(0) = log1 = 0 since P(T > 0) = 1.
nonparametric methods
The parametric models are often used because of their simplicity. It has
to be emphasized that they impose a complex structure on the data,
which can lead to distortions in the estimated hazard rates. Better mod-
els may be obtained by using nonparametric methods that impose very
few restrictions.
Kaplan-Meier estimator
The derivation of the Kaplan-Meier estimator of the survival curve can
be found in Greene (2003) and in Zeileis (2002). This estimator of the
survival function is also called the product limit estimator for reasons that
will be clear later on.
Given n individuals with p distinct survival times t1 < t2 < ... < tp
and di deaths at ti, assume at first that no censoring occurs. For the time
t from the interval [ts, ts+1) the survival function can be estimated in the
following way:
n - E L dj
S(t) = 1 - F{t) =-1-, ts < t < ts+1.	(9)
n
If the numerator and the denominator of the previous expression are
successively multiplied by factors of the form n - d1 - d2 - ••• - di, i =
1,2,..., s - 1, one obtains
~ , n - d1 n - d1 - d2	n - d1 - d2 - ■■■ - ds	/ s
S(t) =----—........... (10)
n n - d1	n - d1 - ■■■ - ds-1
Let ri = n, i = 2,..., p, denote the number of individuals whose ob-
served survival time is at least t- and let ri. In other words, the number
at risk ri takes into account all individuals alive during the time interval
[ti-1, ti). Under the assumption of no censoring, the equation ri+1 = ri - di
holds, whereas ri+1 = ri - di - ci if censorings occur, with ci equal to the
number of censored observations in the interval [f!--1, ti). The final ver-
sion of the Kaplan-Meier estimator can thus be written as
*»-K).....K)-flR)' (,i)
RESULTS
Recall that the survival function S(t) denotes the probability of survival
time greater than or equal to t. In our case, the probability for the unem-
ployment spell to last until time t or longer is measured. Figure 5 depicts
only the two extreme levels of the factor region with the highest and low-
est probability of survival, namely the survival curves for Podravska and
Gorenjska region, respectively, to make the figure easier to read. The sur-
vival function estimates for other regions lie between the two extremes.
The differences between the survival function estimates are clearly vis-
ible. The estimate of the unemployed from Podravska decreases to 0 at
a much slower rate, indicating that the unemployed from Gorenjska re-
gion have a far better position in the labour market.
To test the null hypothesis that the survival functions are the same
for two or more levels of a given factor, the so-called log rank test with
the x2-distribution under the null can be used. When performed for our
data, the highly significant p-value (lower than 10-16) confirms the re-
sults derived graphically from the Kaplan-Meier estimates of the survival
functions.
Since the differences between the highest and the lowest unemploy-
ment region might be specific to the regions considered, we also per-
formed the Kaplan-Meier analysis for the 4 lowest and the 4 highest un-
employment regions pooled together (figure 6). The log rank test is again
highly significant with a p-value of less than 10-6.
Years
figure 5 Survival function estimates for the unemployed from Podravska
region and the unemployed from Gorenjska region (in bold)
Although having the advantage of being non-parametric and therefore
of not imposing restrictions on the shape of the survival function, the
Kaplan-Meier estimator has a major shortcoming. Namely, it does not
allow testing for the presence of an omitted heterogeneity bias. This can
be done in the penalized Cox proportional hazards models setting, or
equivalently, with the help of the frailty models. These models embrace
the idea that different individuals have different frailties, and that those
who are most frail will die earlier than the others. The notion of frailty is
modelled as a random effects term in survival models. A comprehensive
discussion about frailty models can be found in Therneau and Grambsch
(2000).
We fitted a gamma frailty model (since hazard cannot be negative)
with the help of the survival package of the open source code statistical
software R (see http://www.R-project.org). The random effects variable
is highly significant (p-value is equal to 0.00046), thus the null hypothe-
sis of no omitted heterogeneity bias has to be rejected. Omitted hetero-
geneity suggests the existence of some unobserved reason why regional
unemployment differences might persist.
We can look for reasons of persistent regional disparities in regional
labour market adjustment mechanisms: migration, wage flexibility, in-
vestments and changes in labour force participation. According to Gacs
and Huber (2005) the unemployment rate accommodates to a minor
part of regional asymmetric shocks in first round candidate countries
(including Slovenia), on the other hand employment losses turned out
Years
figure 6 Survival function estimates for the unemployed from the four
highest unemployment regions and the unemployed from
the four lowest unemployment regions (in bold)
to be highly persistent, while participation rates importantly contribute
to the adjustment.
Internal migration in Slovenia is low, has fallen during the transition,
and it is not effective in reducing regional disparities (Huber 2007; Fidr-
muc 2004). Such characteristics of migration despite substantial regional
differences are in contrast to economic theory, according to which the
migration should increase rather than fall. One of the major impedi-
ments for migration could be real estate market: house ownership and
high land prices contribute to lower migration, while increased construc-
tion accelerates migration to a region (Huber 2007).
As far as wage flexibility is concerned, there is no straightforward out-
come of existing studies for transition countries as well as for Slovenia.
In general, Slovenia is known to have a high degree of labour market
rigidity (Ferragina and Pastore 2008). Büttner (2007) does find a cor-
rectly signed (negative) and significant effect of regional unemployment
rate on wage level in Slovenia after controlling for industry composition
of employment. Although his findings show that the negative impacts of
unemployment rates on wage level are similar to those of old eu mem-
bers, this does not automatically imply a high level of wage flexibility,
since eu countries are known for a relatively high level of wage rigidities
to regional unemployment rates.
For transition countries Bornhorst and Commander (2006) show that
a substantial fall in labour demand results in very slow employment re-
covery. Capital mobility has a limited role in diminishing regional dis-
parities. Since Slovenia is a small country, close to eu markets, there is a
dispersed structure of fdi across regions (Huber 2007) indicating there
is no capital mobility effect in decreasing regional differences. In other
transition countries stylized facts show that border regions and regions
with the capital city are in better position in comparison to other much
poorer regions. Due to the smallness of the country these effects are in-
significant for Slovenia.
Conclusions
The highest unemployment rates are registered in Pomurska and Po-
dravska. While the unemployment rate used to be the highest in Po-
dravska, it is persistently decreasing in the last couple of years. Above
average unemployment rates were also recorded in Zasavska, Spod-
njeposavska and Savinjska regions. Notranjsko-kraška, Jugovzhodna
Slovenija, Gorenjska, Obalno-kraška, Osrednjeslovenska and Goriška are
the regions with the unemployment rate persistently below Slovenian av-
erage. The lowest unemployment rate was recorded in Goriška: 6.5% in
2005. Slovenian regional unemployment rates belong to the nuts 3 level
of European statistical regions and are below the eu average. Also the
differences among regions are below European average and are gradually
diminishing.
Survival analysis of the duration of unemployment spells based on
a comprehensive dataset, with more than 450,000 observations in the
period from January 2002 and November 2005 yielded the following
results. Regarding the region of the unemployed, the probability of re-
employment is the lowest for the unemployed from Podravska, Sav-
injska, Zasavska and Jugovzhodna Slovenija. The unemployed from
Gorenjska region have a far better position in the labour market. The
probability of remaining unemployed for the latter region is slightly
lower than for the unemployed from Goriška region. The two extreme
regions with the lowest and the highest mean length of unemployment
are Gorenjska and Podravska region, where the unemployed have to wait
for 376 and 565 days on average to find a new job, respectively. The dif-
ferences between the Kaplar-Meier survival function estimates are highly
significant. The probability of re-employment for the unemployed from
Gorenjska is the highest among all regions, while being the lowest for the
unemployed from Podravska.
The analysis has proven that the characteristics of duration of regional
unemployment are specific. The reasons for significant regional dispar-
ities can be found among low internal migration and high level of wage
rigidity. The results can help to identify potential target groups of un-
employed in different regions in order to improve the efficiency of an
active employment policy. Furthermore, one of the prime policy objec-
tives should be to enhance migration by, for example, providing a more
attractive real estate market in the regions with high employment rates.
Slovenia has lately experienced a lack in using eu structural funds. There
is an obvious need for a more effective structural funds policy. Thus, the
results of this study can help in identifying regions which fulfil the re-
quirements and have high needs for structural funds usage. Despite high
wage rigidity to unemployment rates, the wage levels differ across re-
gions, resulting in different levels of unemployment benefits, which also
contribute to different regional incentives for the unemployed to find
new jobs.
Acknowledgments
This research was supported by a grant from the Austrian Science and
Liaison Offices Ljubljana and Sofia. The paper reflects only the author's
views. The Austrian Science and Liaison Offices are not liable for any use
that may be made of the information contained therein.
We are grateful to two anonymous referees for helpful suggestions.
Notes
1	Based on data provided by Employment Service of Slovenia (http://www
.ess.gov.si/slo/Dejavnost/StatisticniPodatki/Kazalci/GibanjeRegBP.htm).
2	ilo unemployment rate streams from the Labour Force Survey conducted
by the Statistical Office of the Republic of Slovenia according to Interna-
tional Labour Organisation (ilo) instructions. It is internationally com-
parable. ilo unemployed are those who meet the following criteria: are
not working for payment, are not employed or self-employed (1), actively
seek for employment (2) and are willing to accept work immediately or
within two weeks (3) (Kajzer 2005). On the other hand, registered unem-
ployed are those who are listed in the register at the Employment Office of
the Republic of Slovenia.
3	According to Eurostat long term unemployed are those who are unem-
ployed for 12 months or more.
References
Arranz, J. M., and J. M. Romero. 2003. An extra-time duration model
with application to unemployment duration under benefits in Spain.
Documento de Trabajo E2003/38, Fundacion Centro de Estudios An-
daluces.
Bayer, C., and F. Jüßen. 2006. Convergence in West German regional un-
employment rates. Http://i29.3.20.4i/eps/urb/papers/04ii/04ii007.pdf.
Bornhorst, F., and S. Commander. 2006. Regional unemployment and its
persistence in transition countries. The Economics of Transition i4 (2):
269-88.
Büttner, T. 2007. Unemployment disparities and regional wage flexibility:
Comparing eu members and Eu-accession countries. Empirica 34 (4):
287-97.
D'Agostino, A., and F. Mealli. 2000. Modelling short unemployment in
Europe. Institute for Social & Economic Research Working Paper 06.
Domadenik, P., and F. Pastore. 2006. Influence of education and train-
ing systems on participation of young people in labour market of cee
economies: A comparison of Poland and Slovenia. International Jour-
nal of Entrepreneurship & Small Business 3 (i): 640-66.
Elhorst, J. P. 2003. The mystery of regional unemployment differentials:
Theoretical and empirical explanations. Journal of Economic Surveys
i7 (5): 709-48.
Eurostat. 2006. Unemployment in EU25. Eurostat News Release stat/06/
i50.
-. 2007. Nomenclature of territorial units for statistics - nuts: Sta-
tistical regions of Europe. Http://ec.europa.eu/comm/eurostat/
ramon/nuts/home_regions_en.html.
Ferragina, A. M., and F. Pastore. 2008. Mind the gap: Unemployment in
the new eu regions. Journal of Economic Surveys 22 (i): 73-u3.
Fidrmuc, J. 2004. Migration and regional adjustment to asymmetric
shocks in transition economies. Journal of Comparative Economics 32
(2): 230-47.
Gacs, V., and P. Huber. 2005. Quantity adjustments in the regional labour
markets of eu candidate countries. Papers in Regional Science 4 (84):
542-58.
Green, D., and W. Craig Riddell. i995. Qualifying for unemployment in-
surance: An empirical analysis of Canada. Http://www.hrsdc.gc.ca/
eng/cs/sp/hrsdc/edd/reports/i995-000322/repi7.pdf.
Greene, W. H. 2003. Econometric analysis. New York: Prentice-Hall.
Huber, P. 2007. Regional labour market developments in transition: A sur-
vey of the empirical literature. European Journal of Comparative Eco-
nomics 4 (2): 263-98.
Kajzer, A. i998. Uporabnost sodobnih teorij brezposelnosti pri obravnavi
brezposelnosti v Sloveniji. ib revija, no. 8-10:46-52.
-. 2005. Pojem fleksibilnosti trga dela in stanje na trgu dela v
Sloveniji. Delovni zvezki 14, umar.
Kavkler, A., and D. Boršic. 2006. The main characteristics of unemployed
in Slovenia. Naše gospodarstvo 52 (3-4): 3-15.
-. 2007. Duration data analysis: Features of unemployment spells in
Slovenia. Paper presented at Warsaw International Economic Meeting,
Warsaw.
Klein, J. P., and M. L. Moeschberger. 1998. Survival analysis: Techniques for
censored and truncated data. New York: Springer.
Kupets, O. 2006. Determinants of unemployment duration in Ukraine.
Journal of Comparative Economics 34 (2): 228-47.
Landesmann, M., and R. Römisch. 2006. Economic growth, regional dis-
parities and employment in the EU-27. wiiw Research Reports 333.
Mlady, M. 2006. Regional unemployment in the European Union, Bul-
garia and Romania. Eurostat: Statistics in Focus, no. 1.
Moffitt, R. A. 1999. New developments in econometric methods for labour
market analysis. In Handbook of labour economics, ed. O. Ashenfelter
and D. Card, 1367-97. Amsterdam: North Holland.
Newell, A., and F. Pastore. 2006. Regional unemployment and industrial
restructuring in Poland. Eastern European Economics 44 (3): 5-28.
Orazem, P., M. Vodopivec, and R. Wu 2005. Worker displacement during
the transition: experience from Slovenia. The Economics of Transition
13 (2): 311-40.
Pecar, J. 2005. Regije 2005: Izbrani socio-ekonomski kazalniki po regijah.
Delovni zvezki 9, umar.
-. 2006. Regije 2006: Izbrani socio-ekonomski kazalniki po regijah.
Delovni zvezki 15, umar.
Römisch, R., and T. Ward. 2005. Regional employment paterns and
prospective developments in the new eu member states. wiiw Re-
search Reports 319.
sors. 2006. Labour force survey results: Slovenia 2005. Rapid Reports
150/2006, Statistical Office of the Republic of Slovenia.
Taylor, J., and C. Wren. 1997. uk regional policy: An evaluation. Regional
Studies 31 (9): 835-48.
Therneau, T. M., and P. M. Grambsch. 2000. Modelling survival data: Ex-
tending the Cox model. New York: Springer.
Traistaru, I., P. Nijkamp, and L. Resmini. 2002. The emerging economic
geography in eu accession countries. Adelshot: Ashgate.
Vodopivec, M. 2004. Labor market developments in the 1990s. In Slovenia:
Yugoslavia to the European Union, ed. M. Mrak, M. Rojec and C. Silva-
Jauregui, 292—314. Washington: The World Bank.
Vodopivec, M. 1995a. Unemployment insurance and duration of unem-
ployment: Evidence from Slovenia's transition. World Bank Policy Re-
search Working Paper 1552.
-. 1995b. Trg dela. In Strategija gospodarskega razvoja Slovenije:
Splošni pogoji za gospodarski razvoj, 145-70. Ljubljana: umar.
Zeileis, A. 2002. Slides for the lecture Biostatistics. Http://www.ci.tuwien
.ac.at/~zeileis/teaching/Biostatistics/.