You are on page 1of 20

Journal of Comparative Economics 34 (2006) 228247

www.elsevier.com/locate/jce

Determinants of unemployment duration in Ukraine


Olga Kupets a,b
a Labor Group, Economics Research and Outreach Center, National University Kyiv-Mohyla Academy,

Voloshskaya Street, bld. 10, office 214, 04070 Kiev, Ukraine


b IZA, Bonn, Germany
Received 2 February 2006
Available online 29 March 2006

Kupets, OlgaDeterminants of unemployment duration in Ukraine


There are few studies of unemployment duration in transition economies, including members of the
CIS. This paper presents the first evidence of the determinants of unemployment duration in Ukraine. We
examine the effects of various individual characteristics and local demand conditions on the hazards to
employment or inactivity using multiple unemployment spell data from the Ukrainian Longitudinal Monitoring Survey (ULMS) for the years 19982002 and estimating the Cox proportional hazards model with
two competing risks. The main estimated results are generally similar to those obtained in developed and
other transition countries. The individuals age, marital status, level of education, income while unemployed
(including income from casual work activities and subsistence farming), and local demand constraints are
significantly related to the total time in unemployment. The estimates of the baseline hazard to employment
suggest marked negative duration dependence after 14 months in unemployment. Journal of Comparative
Economics 34 (2) (2006) 228247. Labor Group, Economics Research and Outreach Center, National University Kyiv-Mohyla Academy, Voloshskaya Street, bld. 10, office 214, 04070 Kiev, Ukraine; IZA, Bonn,
Germany.
2006 Association for Comparative Economic Studies. Published by Elsevier Inc. All rights reserved.
JEL classification: J64; J68
Keywords: Duration analysis; Unemployment benefits; Labor markets in transition; Ukraine

E-mail address: kupets@eerc.kiev.ua.


0147-5967/$ see front matter 2006 Association for Comparative Economic Studies. Published by Elsevier Inc. All
rights reserved.
doi:10.1016/j.jce.2006.02.006

O. Kupets / Journal of Comparative Economics 34 (2006) 228247

229

1. Introduction
Ukraine is one of the transition countries that have been lagging behind in reforms in view
of considerable adjustment costs and social unrest usually associated with radical reforms. However, delay in reforms created an environment of rent-seeking, state capture, and freezing of
transition (Havrylyshyn, 2005). Prolonged recession coupled with postponed enterprise restructuring has brought about a sclerotic labor market in which losing a job might be relatively rare,
but once lost, finding a new job has been increasingly difficult. The weak demand for workers
and competition with those still employed but looking for another job have combined to make it
difficult for the displaced workers and new entrants to find jobs. The immediate result has been
a build-up of a stagnant pool of unemployment and a surge of long-term unemployment at the
levels similar to those of the less dynamic OECD countries. In 2003, for example, the fraction
of the unemployed who have been looking for a job for more than a year (the measure of the
incidence of long-term unemployment) amounted to 50.3% in Ukraine and to 42.3% on average
in the OECD European countries (see Table 1 for Ukraine).
Despite the extensive literature that examines the causes and consequences of long-term unemployment in developed economies (e.g. Machin and Manning, 1999; OECD, 1993, 2002), the
determinants of exits from unemployment and the impact of unemployment benefits on unemployment dynamics in these countries (Devine and Kiefer, 1991; Atkinson and Micklewright,
1991), relatively little has been written about unemployment duration and its determinants in
transition countries with lagged reforms. Although there are many potential reasons for the emergence of long-term unemployment including demand and supply shocks, institutional features
and outside labor markets, the Western literature focuses primarily on various characteristics of
labor market institutions as the main factor behind long-term unemployment in Europe. These
include strict employment protection legislation, powerful trade unions, wide use of permanent
contracts, generous unemployment and welfare benefits, high labor taxes and minimum wages.
However, the functioning of the labor market in a transition economy is less likely to be driven
by the same institutional factors as in developed countries. According to Ham et al. (1998),
the level of unemployment compensation has a moderate negative effect on the duration of unemployment in the Czech and Slovak republics, compared with the corresponding estimates in
western countries, while the principal factors underlying the differences in exit rates from unemployment include growth rate of the new service sector, speed of privatization and restructuring,
amount of foreign direct investment, enforcement of labor legislation, and alternatives for the
working-age population. A similar argument can be used to explain differences between Russia and Ukraine in terms of unemployment dynamics and long-term unemployment. Brown and
Earle (2006) show that increases in job reallocation and in the productivity-enhancing consequences of the labor reallocation process appear to have been somewhat slower in Ukraine than
in Russia. They conclude that a more aggressive reform strategy implemented in Russia has produced greater job reallocation, faster job creation, and less net employment decline. This, in turn,
has resulted in higher intensity of flows into and out of unemployment and shorter unemployment
duration.
Therefore, low outflows from unemployment and long unemployment spells in many transition countries could be blamed on unsuccessful transition reforms leading to insufficient job
creation and job reallocation rather than on labor market rigidities and generous unemployment
compensation. On the other hand, we should not forget that supply-side determinants can be
also important in a transition context (Boeri, 2001). Some categories of the unemployed with
unattractive work-related characteristics may encounter much greater difficulty in finding regu-

230

O. Kupets / Journal of Comparative Economics 34 (2006) 228247

lar jobs than their more competitive counterparts. Also, depreciation of human capital, erosion of
work habits, discouragement, loss of motivation, and other consequences of long-lasting spells
of joblessness (so-called duration dependence effect) may result in ever-declining chances of
reemployment (Machin and Manning, 1999).
In this paper, we examine the factors which may affect the probability of reemployment or
withdrawing from the labor force after the period of unemployment in Ukraine. The only study
directly related to our research has been done by Stetsenko (2003). The author examines the
determinants of duration of the registered unemployment in Kiev using the registered data from
the city employment center over 20012003 and employing the Cox proportional hazards and
the piece-wise constant exponential models. The author finds significant positive effect for the
level of unemployment benefits on the duration of registered unemployment. The other findings are that younger workers and males are more likely to leave the register to both competing
destinations (to job and for other reasons out of the register); that married females tend to have
significantly lower probability of transition to employment; that individuals with less than general
secondary education have higher probability of transition to employment relative to individuals
with higher level of education; that having profession increases chances to get a job; and that
unobserved heterogeneity is insignificant.
Our paper makes several contributions. It provides evidence on the duration and demographic structure of general unemployment (not only registered unemployment) in Ukraine. We
use individual-level data from the first wave of the Ukrainian Longitudinal Monitoring Survey
(ULMS), a nationally representative survey of individuals aged from 15 to 72. Therefore, our
results refer to Ukraine as a whole and not only to its capital city, which is often considered
an outlier in terms of the labor market conditions. We use the sample of unemployment spells
that started between January 1998 and December 2002. Thus, we analyze distribution of unemployment spells over the period which covers the years before and after 2000the year of
economic reversal in Ukraineas well as the years before and after the 2001 reform of the unemployment benefit system into unemployment insurance system. Finally, we test our hypothesis
about the disincentive effect of income from casual activities and subsistence farming during a
non-employment period and the negligible effect of unemployment benefits with respect to exits to employment together with some basic hypotheses suggested by a job search model.1 The
issue of casual work activities and subsistence farming is very important in a transition economy like Ukraine given the high share of individuals involved in various informal activities and
weak monitoring capacity. Although occurrence of casual work activities during unemployment
is potentially endogenous, the estimated effect on the conditional probability of exit from unemployment may inform policymakers about important policy direction.
Our findings confirm broadly the results of studies for developed and transition countries.
Married, younger and educated individuals living in large cities are more likely to leave unemployment to employment. The higher the regional unemployment rate is at the start of unemployment the lower the probability of reemployment (controlling for oblast and calendar time
dummies). Those who have alternative sources of subsistence during unemployment including
income from casual work activities or subsistence farming, household income, or pensions tend
to stay in unemployment before exiting to a job significantly longer. The effect of unemployment
benefits with respect to exits to a regular job is found to be insignificant in the total sample of
unemployed (i.e. with and without income from casual activities or subsistence farming), while
1 Discussion of the definition of unemployment adopted in our study is offered in Section 4 on Data.

O. Kupets / Journal of Comparative Economics 34 (2006) 228247

231

it is significant and negative in the subsample of standard unemployed without any income
from casual work or farming activities. The baseline employment hazard is non-monotonic: it
increases with duration of unemployment during the first 14 months and decreases afterwards.
Analysis of the determinants of unemployment duration before withdrawing from the labor force
suggests that individuals over the age of 55, those who rely on household income, previously
economically inactive persons, and the residents of rural area or large cities are more likely to
leave unemployment for inactivity.
The paper is set up as follows. Section 2 provides an overview of the unemployment insurance
system and the Public Employment Service in Ukraine and the consequences of their failure to
provide real assistance to the unemployed. Section 3 presents the econometric model used in
the analysis. Section 4 provides the details of the data and variables used. Section 5 offers the
estimation results and Section 6 concludes the paper.
2. Challenge for an unemployed person in Ukraine: unemployment insurance or
alternative sources of subsistence?
The Public Employment Service (PES) and the unemployment compensation system were
established in Ukraine according to the Law on Employment in March 1991. The unemployment
compensation system was relatively liberal in terms of eligibility, entitlement and replacement
ratio until the new Law on Compulsory State Social Unemployment Insurance went into effect
in 2001. In general, the PES is supposed to perform two major functions: to assist unemployed
workers in their job search and to provide income support during a period of unemployment.
However, it is widely believed that it is not very successful at either of these tasks in Ukraine.
First, although firms are obliged to register all vacancies with the local employment center
and to use the center during recruitment, many firms fail to do so, preferring other recruitment
methods. Also, the PES sometimes provides training or retraining for skills that are already in
surplus in the local labor markets, and public works are usually of low skill level (Kupets, 2000).
As a result, the probability of finding a good job with the help of the public employment center
is likely to be small, while the transaction costs of registration and staying on the register may
be relatively high. One such transaction cost of staying on the register is a necessary visit of the
unemployed to the local employment center located in the administrative center of his/her civil
registration (new name of the old system of propiska) at least once a month. Since many people
live far from administrative centers of their registration, the above requirement demands heavy
expenses in terms of time and money in some cases.
Second, the level of unemployment benefits is too low both in nominal and real terms. The
ratio of the average unemployment benefit to the official average wage in the economy is about
2528%, while its ratio to the nationally established level of minimum wage fluctuates between
50 and 70% (Table 1).
Moreover, because of the strict unemployment benefit eligibility criteria and high incidence of
long-term unemployment, the coverage ratio (the ratio of those receiving unemployment benefits
or unemployment assistance to the total number of registered unemployed) has been less than
70% for all years. Although with respect to incentives/disincentives to work the unemployment
insurance system in Ukraine may seem better than in CEE countries, it is certainly worse in terms
of income support and poverty prevention.
A study of the sources of subsistence during the period of unemployment confirms this statement. Only 4.6% of our sample of the unemployed reported that unemployment benefits were

232

O. Kupets / Journal of Comparative Economics 34 (2006) 228247

Table 1
Unemployment dynamics and unemployment benefits in Ukraine

Registered unemployment
Registered unemployment rate, % of workingage able-bodied population
Fraction of registered unemployed receiving
unemployment benefits, %
ILO-defined unemployment
Unemployment rate according to LFS, % of labor force aged from 15 to 70
Incidence of long-term unemployment (fraction of the unemployed who have been looking
for job for more than 1 year), %
UB
Ratio of average UB to minimum wage, %
Ratio of average UB to average wage, %

1998

1999

2000

2001

2002

2003

3.69

4.3

4.22

3.68

3.80

3.6

53.1

52.8

54.3

62

66.7

69.1

11.3

11.9

11.7

11.1

10.1

9.1

37

46.3

50.5

54.8

53.5

50.3

70
25.1

67.3
28

50.3
25.8

72.2
27.4

64.2
28.1

57.7
25.6

Note: Registered unemployment characteristics correspond to the end of years, while characteristics according to the
Labor Force Survey (LFS) are presented for the fourth quarter in 1998 (yearly survey) and on average for 19992003
(quarterly surveys).
Source: Derzhkomstat.

the main source of their support.2 The dominant role in support of jobless is played by household
income, i.e. income of parents, spouses or other relatives (68.3%). Income from various casual
activities or subsistence farming constitutes the second largest group among the main sources
of subsistence (13.9%). It may serve as the only source of subsistence or operate in conjunction
with household income, unemployment benefits, pensions, state transfers, or savings. Markedly,
only 27.5% of those who receive unemployment benefits, along with other sources of subsistence, consider it to be their primary source of income during the period of unemployment. Most
of them rely on household income.
As a consequence of ineffective public employment policy and the unemployment insurance
system, less than half of the actual unemployed (defined according to the ILO unemployment
criteria) bother to register as unemployed in the public employment centers. An analysis of job
search methods among the unemployed in our sample indicates that people rely on the help of
friends and relatives (29.2%), direct contacts with employers (16.4%), job advertisements in the
newspapers or Internet (37.6%) rather than on the assistance of the public employment service
(10.8%).
Following from the above argument, it is unlikely that the unemployment insurance system is
behind the low outflows from general unemployment (as opposed to registered unemployment)
in Ukraine, given how low the benefits are and how few unemployed register to receive them.3
However, another inference based on the weak enforcement of legislation and high payroll taxes
could have more explanatory power. Because of very low unemployment benefits accompanied
2 Information about the main source of subsistence is taken from the answers of respondents for the direct question

about the main source of subsistence during a period of joblessness. Most frequent or noteworthy compositions of the
sources of subsistence are additionally reported in Kupets (2005) but not presented here for brevity.
3 Typically, economists have seen an unemployment benefit system as having a negative effect on unemployment
duration, with high benefits and long entitlement periods causing the unemployed to be less willing to accept jobs.
Extensive discussion of this topic is offered in Atkinson and Micklewright (1991).

O. Kupets / Journal of Comparative Economics 34 (2006) 228247

233

with still relatively low labor demand, many jobless people leave the formal labor market, preferring to find an informal working activity or to rely on subsistence farming. Others start some
sort of self-employment (usually in a low-productivity sector) just to survive. And some discouraged persons withdraw from the labor market to look for additional sources of income such
as stipends, pensions, child allowance, etc. Hence, prolonged joblessness may force some persons to intensify casual work activities or engage more actively in subsistence agriculture. On
the other hand, those usually unemployed persons who are occasionally engaged in unreported
activities or subsistence farming tend to search for regular jobs less intensively and, therefore,
they are less likely to receive a job offer. For such individuals, alternative income may raise their
reservation wage, lowering the probability of accepting a job offer and thus the probability of
reemployment as a whole. Various kinds of non-labor income during an unemployment spell,
including household income, pensions, other state and private transfers are likely to have the
same effect on the duration of unemployment as casual labor income. Thus, we hypothesize that
the involvement of usually unemployed individuals in informal activities or subsistence farming
in view of labor demand constraints in the formal sector are of much greater importance than
unemployment benefits in explaining the stagnancy of unemployment in Ukraine during the late
period of transition.
3. Econometric model
We analyze duration of unemployment spells in Ukraine using a duration model.4 This model
is preferable to the usual regression model because of its ability to handle time-dependent covariates and right-censoring in the data. The focus in modeling durations of unemployment is
usually on the conditional probability of leaving unemployment, the hazard function. The hazard
model used for this study is the Cox proportional hazards model (Cox, 1972) of the following
general form:


i (t) = 0 (t) exp xi (t) ,
where xi is the set of explanatory variables for individual i, is the vector of parameters to be
estimated, and 0 (t) is the baseline hazard at time t , which is allowed to be nonparametric.5
In this study, most variables are taken as time-invariant (except the year and quarter dummies)
due to the limited data on time-varying characteristics of the unemployment benefit system at
our disposal and the potential endogeneity of certain characteristics, which vary with time in
unemployment (e.g. marital status, number of kids).
One of the key assumptions of hazard models is that all inter-individual heterogeneity is due
to observed factors. However, it is possible that unobserved variables are also a source of heterogeneity. Uncontrolled heterogeneity in duration models can lead to misleading inferences about
duration dependence, and can also bias the estimated effects of the included explanatory variables
(Lancaster, 1990). However, in certain cases, this may not be particularly serious. The empirical
work of Meyer (1990) and of others suggests that failure to model distribution of unobserved
heterogeneity explicitly does not seriously bias results if the baseline hazard is allowed to be
4 See Kiefer (1988) or Lancaster (1990) for more details on duration models and hazard functions.
5 In the general case, explanatory variables may vary with unemployment duration t (classic examples are time-varying

unemployment benefits and the time remaining until their expiration), with calendar time (e.g. local labor market conditions or characteristics of the unemployment insurance system which varies with policy changes), or may remain fixed
over time (as most personal characteristics).

234

O. Kupets / Journal of Comparative Economics 34 (2006) 228247

nonparametric. Furthermore, Narendranathan and Stewart (1993) argue that there is no reason to
expect any distortions imposed by the use of techniques to allow for unobserved heterogeneity
to be less serious than those caused by ignoring unobserved heterogeneity in the first place. For
this reason, we have chosen to restrict attention in this paper to the models without accounting
for unobserved heterogeneity.6
Taking into account that an unemployment spell may end with the individual either starting a
job (exit to employment) or leaving the labor force (exit to inactivity), we estimate an independent
competing risks model.7 This assumes that the hazard rate for exit to any destination is the sum
of the two destination-specific hazard rates. We estimate the two risks separately: spells ending
with transition to inactivity are considered censored when estimating hazard to employment, and
vice versa.
4. Data and variables
Our data are taken from the first wave of the Ukrainian Longitudinal Monitoring Survey
(ULMS-2003), a nationally representative random sample of households consisting of 4056
households and 8641 individuals aged 15 to 72. The ULMS data set is unique in Ukraine, since
it is the richest individual-level data set available, and it allows for the analysis of more than five
years of labor market flows and unemployment duration owing to its retrospective nature.8
We use the inflow sample of unemployed including everyone who started with an unemployment spell between January 1998 and December 2002 and who provided complete responses to
the questions about their period of job search. The spells starting after December 2002 are not
used, and ongoing spells starting before December 2002 are censored at the date of December 31,
2002, in order to avoid the possible effects of changing criteria for employment/unemployment
status when moving from the retrospective part of the questionnaire to the section referring to
the reference week. We have not restricted the age of individuals in the sample according to the
usual age of retirement. We think that the low retirement age (55 for women and 60 for men)
and very low level of pensions lead older Ukrainians to have almost the same work incentives
as those in younger age groups. Moreover, according to the ILO guidelines, pensioners, students
and other individuals mainly engaged in non-economic activities who satisfy the basic criteria
of unemployment should be classified as unemployed.9 Finally, we have not separated men from
women in our analysis because there is no objective reason to expect significant differences in
unemployment dynamics between men and women in a country like Ukraine. The total sample
6 We tested for gamma-distributed heterogeneity in our previous models with various specifications of the baseline
hazard and different sets of explanatory variables. The variance of heterogeneity (gamma-distributed) was found to be
not significantly different from zero. Insignificant unobserved heterogeneity has been also found in Stetsenko (2003) for
Ukraine and in Grogan and van den Berg (2001) and Foley (1997) (for exits to employment) for Russia.
7 We restrict our choice to these two main destination states because retrospective data with a very long recall period,
used in our study, does not allow using more alternative destination states like in many other studies (e.g. Narendranathan
and Stewart, 1993).
8 It should be stressed also that the ULMS is a unique data set in the CIS area because it allows for the hazard analysis
in a continuous-time framework without many simplifying assumptions. Widely used Russian Longitudinal Monitoring Survey (RLMS) and other individual-level panel data sets have serious shortcomings as sources of information on
unemployment durations. According to Kiefer (1988), the two main problems of such data sets are right-censoring (exclusive sampling of the current unemployed for information on unemployment duration) and length-biased sampling
(underrepresentation of short spells). Problems connected with the retrospective ULMS data are discussed below.
9 We control for those who receive any kind of pension (not only for years of service or retirement age but also for
disability and loss of provider) by including a separate dummy for such persons.

O. Kupets / Journal of Comparative Economics 34 (2006) 228247

235

according to the definition of unemployment adopted in our study (see below) comprises 1536
unemployment spells experienced by 1293 individuals. The subsample of unemployment spells
without income from casual work or farming activities according to the standard ILO criteria
(call them standard unemployed) includes 1102 unemployment spells, and the subsample of
spells with any income from casual work or subsistence farming (call them casual workers)
consists of 434 spells.
The unemployment duration measure refers to the length of an unemployment spell, which is
defined as the number of full months between the date of the beginning of job search (month and
year only) to the date of its end (or to December 31, 2002 in the case of right-censored spells).
In order to isolate the net effect of time out of work on the hazard of leaving unemployment, we
introduced a set of control variables based on theoretical grounds and previous empirical findings
in developed and transition countries (see Devine and Kiefer, 1991 for review of some of them).
The choice of variables was constrained by the data available in the retrospective sections of the
ULMS. Controls are included for gender, age, marital status and number of children under 15
years old (and their interaction with gender), education level, previous labor market status, etc.
The values of the characteristics are determined at the starting date of the unemployment spell
to ensure their exogeneity. Additionally, we use six dummy variables representing the categories
of sources of subsistence. These dummy variables reflect the presence or absence of a certain
type of income during a non-employment period. Unfortunately, the ULMS does not include
retrospective information on the level of income received from various sources. Also, there is
no direct information about calendar time and the length of receipt of unemployment benefits
or other alternative income, as well as the remaining time for benefits to lapse. Due to the lack
of this information, we are unable to analyze the effect of specific features of the unemployment benefit system on reemployment probabilities, an analysis which would provide valuable
policy implications. However, given the relatively small variation in the level of unemployment
benefits and length of payment as well as the low coverage of the total unemployed population
(as opposed to only the registered unemployed), a dummy representing benefit receipt should be
sufficient to capture the expected effect on the duration of general unemployment in Ukraine.10
In addition to individual characteristics, we use variables to account for differences in local
labor demand conditions. Differences in the local labor markets are proxied by the oblastlevel quarterly registered unemployment rate at time of starting unemployment (accounting for
between-region differences) and the type of settlement (accounting for within-region differences).11 In our final model we include also oblast fixed effects to take into account possible
omitted regional characteristics which may affect unemployment duration and labor market conditions. Finally, we add calendar time dummies (year and quarter) which are allowed to change
with time in unemployment to control for changes in the macroeconomic environment and possible seasonal effects.
10 Hunt (1995) finds for Germany that the dummy on receipt of UI is significant while the level of benefit receipt is
insignificant. Addison and Portugal (2003) use a dummy on access to unemployment benefits and find it highly significant
in Portugal, but they group individuals by age (seven elements of age regressor) so as to mimic the stepped increases in
benefit entitlement with age.
11 Regional unemployment rate is the most popular measure of the local labor demand conditions (inter alia
Narendranathan and Stewart, 1993 for the UK; Meyer, 1990 for the US; Foley, 1997 for Russia). The alternative measures are local unemployment and vacancy rates for the individuals education group, real value of regional per capita
industrial production, and regional agricultural/industrial employment ratio (Ham et al., 1998) or Herfindahl-Hirschman
Index of employment concentration in the local labor market (Denisova, 2002).

236

O. Kupets / Journal of Comparative Economics 34 (2006) 228247

The definition of all variables is provided in Appendix Table 1. Table 2 provides some descriptive statistics for the variables used in the unemployment duration analysis for the three samples.
The mean duration of censored spells is almost twice as high as completed unemployment spells,
and the maximum duration is 60 months. Distribution of the unemployed with incomplete spells
by actual duration of their unemployment at the end of 2002 shows that the incidence of longterm unemployed was 58%.12
Table 2 shows that the majority of spells are experienced by the unemployed without income
from casual work (72%) implying that the total sample gives a disproportionate weighting to
such job-seekers. As expected, the mean unemployment durations of completed and censored
spells among casual workers are about 23.5 months higher than among standard unemployed, and the corresponding difference in the incidence of long-term unemployment is about
6 percentage points. There are much more previously employed males and married people with
general secondary/vocational or lower level of education in the subsample of casual workers
than in the total sample or in the subsample of standard unemployed. It is worth noting also
that in contrast to the total sample of unemployed or the subsample without income from casual
work, the proportion of unemployment spells experienced by the unemployed with income from
casual work increases with age up to age 55. Not surprisingly, most of those involved in some
kind of casual work or subsistence farming live in villages or very small towns (about 62%).
Although the samples are very similar across geographic regions, Western Ukraine has fewer
casual workers and Southern Ukraine has more.13 Overall, the data suggest that jobless individuals participate in casual work activities because they have fewer employment opportunities,
greater financial pressures, and generally worse job prospects. In other words, they have been
forced to take these unconventional measures just to survive.
Before turning to the discussion of results, several important methodological issues should
be stressed. The main problem is that the definition and measurement of unemployment differs
across sources, making comparisons difficult. Although we follow the ILO guidelines on defining
the unemployed as people without work and currently looking and available for work during a
given period of time (ILO, 2004), the definition of unemployment accepted in our study differs
from the standard ILO definition due to the retrospective nature of the data with a long recall
period.
First, since labor market states are measured in relation to a long reference period such as
several years rather than to a short period such as one week or one day as in most longitudinal
surveys, the definition of the three labor market states employed in our study refers to the usually employed, unemployed or economically inactive rather than to the currently employed,
unemployed or economically inactive individuals.
Second, according to the standard ILO unemployment criteria, individuals who engage in casual work or casual business activities can not be classified as unemployed. In our study, however,
we do not exclude individuals on the basis of their engagement in irregular activities from the
sample of unemployed if:
12 For comparison, the share of unemployed with duration of non-employment of more than 12 months according to
the official LFS data (fourth quarter in 2002) was 58.6% of all unemployed previously employed, and the share of
unemployed with duration of job search of more than 12 months was 52% of all unemployed who were looking for job
during the preceding four weeks.
13 Both Western and Southern parts of Ukraine are considered to be less industrially developed than the Eastern or
Central parts. The West is predominantly agricultural, while the Southern oblasts have relatively diversified economies
with developed service sectors.

O. Kupets / Journal of Comparative Economics 34 (2006) 228247

237

Table 2
Descriptive statistics
Total sample

Mean duration of completed spells, in months


Mean duration of censored spells, in months
Type of unemployment spell by destination state
% Exit to employment
% Exit to inactivity
% Right-censored
Duration group (completed spells)
% <3 months
% 46 months
% 79 months
% 1012 months
% >12 months
Duration group (censored spells)
% <3 months
% 46 months
% 79 months
% 1012 months
% >12 months
% Female
% Married
Age group
% 24
% 2539
% 4054
% 55
Education
% Primary or unfinished secondary
% General secondary or vocational
% Professional secondary or unfinished higher
% Higher
Sources of subsistence
% Unemployment benefits or stipend during training
% Casual work
% Household income
% Pension
% Other state transfers
% Other sources of subsistence
% Previously employed
Geographic location
% West
% Center and North
% East
% South
Type of settlement
% Village or small town
% Town
% Large city
Number of spells (observations)

Unemployed with
some income from
casual work

Unemployed without
any income from
casual work

12.31
24.83

13.57
27.24

11.87
23.72

46.09
11.39
42.51

40.09
12.21
47.70

48.46
11.07
40.47

23.90
16.76
11.21
6.00
42.13

17.62
18.50
10.57
6.17
47.14

26.07
16.16
11.43
5.95
40.40

16.08
11.18
11.18
3.52
58.04
50.07
54.62

14.01
8.21
12.08
3.38
62.32
33.64
59.22

17.04
12.56
10.76
3.59
56.05
56.53
52.81

34.44
33.27
28.45
3.84

24.42
35.25
37.79
2.53

38.38
32.49
24.77
4.36

12.43
52.67
21.81
13.09

12.44
60.60
18.43
8.53

12.43
49.55
23.14
14.88

19.40
28.26
80.60
9.18
11.33
6.05
67.64

20.28
100
64.75
6.91
8.53
4.61
78.34

19.06
0
86.84
10.07
12.43
6.62
63.43

22.46
29.49
32.94
15.10

19.59
29.72
33.64
17.05

23.59
29.40
32.67
14.34

45.44
32.81
21.74

62.21
22.35
15.44

38.84
36.93
24.23

1536

434

1102

Note: Variables are measured at the beginning of an unemployment spell. Definition of variables is offered in Appendix
Table 1. Sources of subsistence refer to all sources of subsistence (not only the main source) reported by the unemployed.
Source: ULMS, authors calculations.

238

O. Kupets / Journal of Comparative Economics 34 (2006) 228247

(1) a person answered that he/she didnt have a job (including entrepreneurship, business activities, individual work, work in a family enterprise or on a farm, and freelance work) at some
time period in the past;
(2) a person gave the reason of not having a job and answered that he/she was seeking and
available for work for any time during that period;
(3) there is no overlap between the period of employment and the period of non-employment
according to respondents answers (if there was such overlapping we reclassified a person as
employed); and
(4) a person reported about the income from casual work or business activities, production and
sale of products from own land plot, or subsistence farming for his/her own needs as one of
the sources of subsistence at that time of non-employment.14
Unfortunately, the ULMS does not allow us to capture accurately the extent and the nature of such
irregular, usually short-term, activities within a long period of non-employment. It is impossible
to know for sure whether casual work or business activities in this case are really short-term
and sporadic or whether they are regular; whether persons without a regular job in the formal
sector have chosen these informal activities in light of unattractive formal sector opportunities,
or whether they have been forced to engage in casual work activities or subsistence farming just
to survive. It is also difficult to say whether engagement in such activities results in prolonging
an unemployment spell, or whether the long-term unemployment intensifies the search for any
kind of economic activity including casual activities or subsistence farming. The last issue raises
the problem of the potential endogeneity of casual activities and subsistence farming, which is
extremely difficult to address in duration models.
Finally, although some categories of individuals classified as out of the labor force are conceptually distinct from the unemployed (e.g. disabled or retired in the US), a substantial portion
of those reporting themselves as economically inactive may be reclassified as unemployed, and
vice versa (Poterba and Summers, 1995). Therefore, some allowance for spurious events that
result from classification error should be made when analyzing unemployment duration and dynamics. Furthermore, we might expect that the problem of classification error may become worse
as respondents must recall details of events that occurred a long time ago.15 Our analysis based
on the retrospective data over more than five years is certainly subject to reliability problems
and recall bias (see Paull, 2002, among many others). Nevertheless, we believe that the relatively
low labor market mobility of the majority of Ukrainians, the salience and social undesirability of
unemployment for most individuals, and the careful design of the questionnaire have minimized
this problem.
5. Estimation results
Results of fitting the Cox proportional hazards model in a competing risks framework to the
three samples of unemployment spells are given in Table 3.16 The figures reported are the esti14 Grogan and van den Berg (2001), who analyze determinants of unemployment duration in Russia, also do not exclude

individuals on the basis of informal sector activity from the sample of unemployed.
15 Paull (2002) argues that time in unemployment is less likely to be recalled correctly than periods of employment and
inactivity, and so the spell of unemployment may be reclassified as the spell of inactivity rather than forgotten at all.
16 At the first stage of our empirical work we experimented also with parametric continuous-time models with Weibull,
log-normal and log-logistic specifications of the baseline hazard as well as with discrete-time semiparametric model of

O. Kupets / Journal of Comparative Economics 34 (2006) 228247

239

Table 3
Estimation resultsduration analysis of exits from unemployment
Variable

Female
Married
Female Married
Number of children
Female Children
Age group
2539
4054
55
Education
General secondary or
vocational
Professional secondary
or unfinished higher
Higher
Sources of Subsistence
Unemployment
Benefits
Casual work
Household income
Pension
Other state transfers
Other sources
Previously employed
Regional registered
unemployment rate
Type of settlement
Town

Exits to employment

Exits to inactivity

Total
sample
(1a)

Unemployed
with some
income from
casual work
(2a)

Unemployed
without any
income from
casual work
(3a)

Total
sample
(1b)

Unemployed
with some
income from
casual work
(2b)

Unemployed
without any
income from
casual work
(3b)

0.226
(0.121)
0.344*
(0.136)
0.264
(0.175)
0.141
(0.084)
0.058
(0.109)

0.404
(0.293)
0.113
(0.262)
0.615
(0.378)
0.096
(0.134)
0.266
(0.201)

0.301*
(0.137)
0.562**
(0.175)
0.328
(0.211)
0.274*
(0.115)
0.170
(0.141)

0.194
(0.255)
0.554
(0.334)
0.389
(0.372)
0.617*
(0.283)
0.600*
(0.298)

0.148
(0.595)
0.485
(0.547)
0.904
(0.712)
0.551
(0.362)
0.022
(0.456)

0.024
(0.306)
0.974
(0.523)
0.733
(0.553)
1.265
(0.648)
1.530*
(0.659)

0.325**
(0.118)
0.564**
(0.128)
0.892**
(0.294)

0.320
(0.259)
0.307
(0.273)
0.534
(0.817)

0.438**
(0.139)
0.752**
(0.156)
1.145**
(0.327)

0.134
(0.266)
0.478
(0.254)
2.132**
(0.438)

0.070
(0.128)
0.199
(0.142)
0.469**
(0.153)

0.199
(0.275)
0.254
(0.322)
0.536
(0.389)

0.006
(0.147)
0.169
(0.162)
0.415*
(0.171)

0.376
(0.236)
0.044
(0.255)
0.636
(0.332)

0.319
(0.472)
0.296
(0.557)
1.752
(1.119)

0.782**
(0.297)
0.242
(0.305)
0.642
(0.374)

0.161
(0.108)
0.432**
(0.103)
0.367**
(0.116)
0.847**
(0.191)
0.046
(0.126)
0.229
(0.151)
0.167
(0.095)
0.132*
(0.060)

0.136
(0.228)

0.301*
(0.130)

0.185
(0.449)

0.374
(0.268)

0.086
(0.185)
1.067*
(0.500)
0.082
(0.311)
0.263
(0.347)
0.415*
(0.211)
0.197
(0.132)

0.642**
(0.158)
0.955**
(0.224)
0.100
(0.143)
0.109
(0.178)
0.091
(0.110)
0.128
(0.070)

0.162
(0.209)
0.030
(0.196)
0.683*
(0.270)
0.267
(0.292)
0.015
(0.255)
0.515
(0.478)
0.642**
(0.180)
0.009
(0.138)

0.801
(0.416)
0.967
(0.699)
0.352
(0.531)
0.004
(0.792)
0.203
(0.371)
0.247
(0.263)

0.707
(0.433)
0.148
(0.373)
0.224
(0.311)
0.788
(0.627)
0.783**
(0.226)
0.126
(0.186)

0.007
(0.101)

0.035
(0.230)

0.031
(0.118)

0.616**
(0.223)

0.563
(0.518)

0.631*
(0.264)

0.776
(0.530)
1.125*
(0.554)
2.567**
(0.971)

0.587
(0.349)
0.245
(0.321)
2.262**
(0.553)

(continued on next page)

240

O. Kupets / Journal of Comparative Economics 34 (2006) 228247

Table 3 (continued)
Variable

Large city
N (unemployment spells)
Number of failures
Log-likelihood

Exits to employment

Exits to inactivity

Total
sample
(1a)

Unemployed
with some
income from
casual work
(2a)

Unemployed
without any
income from
casual work
(3a)

Total
sample
(1b)

Unemployed
with some
income from
casual work
(2b)

Unemployed
without any
income from
casual work
(3b)

0.243
(0.136)
1536
708
4582.82

0.007
(0.302)
434
174
887.52

0.307
(0.158)
1102
534
3275.51

0.326
(0.243)
1536
175
1026.30

0.072
(0.509)
434
53
232.85

0.392
(0.292)
1102
122
651.78

Notes. Estimation uses the Cox proportional hazards model. Figures reported are the estimated coefficients. Standard
errors are in parentheses. All models include oblast, year and quarter dummies. Exits to inactivity are considered censored
when estimating exits to employment, and vice versa.
* Significance at the 5% level.
** Idem, 1%.

mated coefficients implying that the proportionate impact of each variable on the state-specific
hazard can be calculated by taking the exponent of the corresponding coefficient. The splitsample estimation (columns (2a), (2b) and (3a), (3b)) points to the striking differences between
standard unemployed (without any income from casual work) and those with some income
from casual work. The estimation results for the total sample and for the sample of standard
unemployed confirm broadly the results of studies for developed and transition countries.17
Marriage in the case of males is associated with more rapid job finding after a period of
unemployment. The simplest explanation of this result is that, as household heads, married man
are under greater financial pressure to return to work; they may have higher opportunity costs
for staying unemployed and search more intensively for a new job. Surprisingly, the number of
small children has no significant effect on the duration of unemployment for either females or
males. This finding may be partly attributed to the cheap and well functioning childcare system
emerged under the Soviet era with the aim of promoting female labor force participation. These
findings for Ukraine are consistent with those obtained by Foley for Russia (Foley, 1997).
Age is negatively associated with the probability of reemployment, implying that older workers are at a disadvantage in Ukraines rapidly changing economic environment. Generally, the
difficulties which older workers face in finding work could be attributed to the restrictive hiring
standards of employers (especially in the emerging private sector) due to objective and discriminatory factors, such as obsolete skills, health problems (which from the employers viewpoint
the form suggested by Meyer (1990). We have also estimated the specifications including variables on religion, nationality, health status, the number of dependants younger than 15 or older than 75 in the household, previous employment
status, sector of previous employment, last wage, and last occupation before moving to unemployment, number of previous unemployment spells, and time-changing national unemployment rate, but these factors appear to be not significant.
Our main results are robust to their inclusion. Models analyzing only individuals with one unemployment spell show
no discernible difference from those analyzing individuals who experienced more than one spell, implying that serial
correlation is not a problem.
17 Review of the studies on the determinants of unemployment duration and labor market transitions in the CEE countries can be found in Svejnar (1999). Devine and Kiefer (1991) offer the literature review with detailed discussion for
developed countries.

O. Kupets / Journal of Comparative Economics 34 (2006) 228247

241

are usually associated not only with a lower productivity of workers but also with a heavy burden of sickness benefits), loss of motivation and discouragement. All these factors may in turn
lead to fewer job offers. These results are consistent with job search theory and empirical evidence for many developed and transition countries (e.g. Foley, 1997; Nivorozhkina et al., 2002;
Stetsenko, 2003) but they are in contrast to the findings of many studies for the early period of
transition, which found longer periods of unemployment for young people (e.g. Earle and Pauna,
1996).
Individuals who have completed higher education have significantly higher hazards to employment than individuals with a lower level of education, ceteris paribus. Higher exit rates
among educated people can be explained by their more efficient ability to search for a job due to
better access to information, higher opportunity costs of unemployment, greater flexibility and
wider range of alternatives for future employment. Whereas higher educated persons are able
to compete for jobs that require fewer years of schooling, the reverse is not generally the case.
This issue is extremely important during the economic transition of former centrally-planned
economies. In her study of occupational mobility in Russia, Sabirianova (2002) found that when
the transition period was accompanied by negative demand shocks, more downward unconventional switches occurred on the career ladder (or downward occupational mobility) with greater
losses taking place among those with more education.18 However, our finding of the positive
effect of education on the re-employment probability is in conflict with Stetsenkos (2003) findings regarding the effect of education on exits from registered unemployment. We attribute this
discrepancy to the difference in the composition of vacancies registered at public employment
service offices and those advertised in newspapers and private employment agencies in Ukraine,
with the former heavily represented by vacancies for less educated persons with lower skills
(Kupets, 2000). As has been shown before, our data favor people finding jobs through direct
contacts with employers, the help of relatives or friends, advertisements in newspapers or private employment agencies. These jobs usually attract more highly educated people. Registered
vacancies available for registered unemployed at the public employment centers, on the contrary,
mainly attract less educated people with low skills level.
The estimate of the variable on receipt of unemployment benefits fails to reject our hypothesis
of insignificant effect of unemployment benefits on reemployment probability in the case of the
total sample of unemployed. However, the effect of unemployment benefits is found to be significant and negative if we take only standard unemployed without any income from casual work.
This implies that the existing unemployment benefit system may contribute to longer unemployment spells in some cases but it should not be considered the primary determinant of stagnant
unemployment in Ukraine. The existence of other sources of subsistence during a period of unemployment, including income from casual work activities and subsistence farming, household
income and pension, significantly lowers the probability of reemployment. This effect is consistent with job search theory, with a longer search duration implied by the higher reservation wages
and lower job search intensity caused by alternative sources of subsistence.
The local labor demand variables proxied in our model by the regional unemployment rate
at start of unemployment and the type of settlement have the expected signs. The residents of
regions with higher unemployment rates, all else equal, tend to have longer unemployment spells
before re-employment. The residents of large cities (more than 500 thousand of inhabitants) are
18 Classic examples of such downward occupational mobility in Ukraine for males include the transition from engineer,
technician, and discharged armed forces serviceman to taxi driver, builder, loader or guard. For females, transitions occur
from any profession requiring a higher level of education to street salesperson, babysitter or housekeeper.

242

O. Kupets / Journal of Comparative Economics 34 (2006) 228247

likely to have higher exits to jobs than those living in the rural area or very small towns (at the
10% confidence level). These effects presumably reflect the poorer opportunities facing people
in depressed areas with relatively low labor market activity and a less diversified economy. In
addition to these variables, oblast-level fixed effects were included in the model. A comparison
of models with and without oblast dummies (on the basis of the likelihood ratio test) has supported the presence of oblast fixed effects which are assumed to be constant over the observation
period.
When we turn to the multivariate analysis of the factors affecting exits from unemployment to
economic inactivity (column (1b)), several primary results emerge. Individuals with more small
children appear to search for a job longer before becoming discouraged or deciding to focus
on non-market activities. As expected, workers older 55 have significantly higher exit rates to
inactivity than prime-age or younger individuals. This age effect captures the stronger effect of
discouragement and loss of motivation among older individuals, higher possibility of retirement
and stopping the job search process, health problems and some other attributes. Persons relying
on household income during unemployment are more likely to leave the labor force than persons
without alternative income support. Significant effect of presence of income from casual work
activities with respect to the exit to employment accompanied with its insignificant impact with
respect to the exit to inactivity probably indicates that various casual work activities and subsistence farming can be considered as simply survival measures taken by those who would prefer
the stability of a formal regular job but with a reasonable remuneration. Previously employed individuals appear to search longer before withdrawing from the labor force than those who came
from inactivity. This finding presumably reflects higher importance of work and more negative
attributes associated with not having work and being idle for those previously employed. Also
they may anticipate their relative advantage in finding a job and are not willing to leave the labor
market. Finally, we observe significantly longer unemployment durations before withdrawing
from the labor force for the residents of small to medium towns compared to the residents of
rural areas or very small towns, and no significant difference in unemployment durations between residents of cities and the latter. One of the possible explanations is that the residents of
rural areas can move to self-employment (primarily in subsidiary agriculture) as a last resort or
withdraw from the labor market in the case of unsuccessful search of paid employment, whereas
residents of small to medium towns stay unemployed longer hoping to find a regular job subject
to the limited number of alternative activities.
Figure 1 presents the baseline hazard functions for employment and inactivity (panels (a)
and (b), correspondingly). Both functions are well behaved in terms of theoretical predictions.
The employment hazard is non-monotonic: it first increases with duration until about 14 months
and then falls gradually approaching zero in the end. This pattern is fairly close to the one found
by Stetsenko (2003) for exits from registered unemployed in Kiev. The inactivity hazard has
two peaks at 14 and 37 months, suggesting that the probability of withdrawing from the labor
force is increasing during the first and third years of unemployment and is decreasing during the
second year and over the last portion of the analysis time scale after the second peak. The spike
in the outflow rates at 14 months is around the time of the expiration of benefits (12 months after
the date of registration with the local employment center) but this explanation does not seem
appropriate given the composition of our sample (with recipients of benefits comprising only
one fifth of the sample). We would suggest that unemployed individuals do have some control
over when they start work or withdraw from the labor force and that there are other factors
inducing changes in their reservation wage. Further research is thus required to ascertain the

O. Kupets / Journal of Comparative Economics 34 (2006) 228247

243

(a)

(b)
Fig. 1. Estimated baseline hazard functions by destination state. (a) Exits to employment Note: Baseline hazard function obtained from Table 3, column (1a). (b) Exits to inactivity. Note: Baseline hazard function obtained from Table 3,
column (1b).

more fundamental factors at work which could explain the pattern of duration dependence and
the functioning of the labor market in Ukraine.

244

O. Kupets / Journal of Comparative Economics 34 (2006) 228247

6. Conclusion
This paper analyzes determinants of individual unemployment durations in Ukraine, using
a sample of individuals entering unemployment during January 1998December 2002 from a
new, rich, nationally representative data set (Ukrainian Longitudinal Monitoring Survey) and
estimating the Cox proportional hazards models and competing risks of exits to employment
and to inactivity. Given the absence of an effective system of public employment services and
unemployment insurance in Ukraine, this study tries to identify other potential determinants of
unemployment duration.
Our analysis has shown that there is huge heterogeneity among unemployed in the sample.
The estimation results report evidence for existence of the disadvantaged groups of unemployed
with respect to the probability of reemployment. They include older, single, less educated individuals, living in small towns or rural areas and relying on household income, pension or income
from casual work or subsistence farming during unemployment period. The negative effect of
casual work activities is so strong that despite the relatively small share of casual workers in the
full sample it has been captured in the general model when a dummy variable on casual work
is used. Recipients of unemployment benefits do not have significantly different unemployment
durations in the total sample but they tend to remain unemployed considerably longer if a possibility of having income from casual work or farming activities during unemployment is excluded
(the subsample of standard unemployed). After controlling for oblast fixed effects and changing macroeconomic environment (calendar time dummies), local demand constraints, measured
by the oblast-level registered unemployment rate at start of unemployment, are found to have a
significant negative effect on the exit probability. As far as duration dependence is concerned,
our results show positive duration dependence of the hazard to employment until 14 months and
negative duration dependence afterwards.
Demand shocks, technological changes in the early 1990s and delayed policy responses have
brought about persistent and stagnant unemployment in Ukraine in the late 1990s. In this study
we show that the possibility of different casual work activities or subsistence farming can be
viewed as one of the potential contributors to stagnant unemployment during the late period of
transition. It should be noted, however, that a reverse-causality interpretation of this phenomenon is also possible. Taking into account insufficient labor demand in the formal sector and
inadequate assistance in retraining and job matching by public employment centers, many unemployed people, especially from disadvantaged groups at the labor market, may be forced to
engage in informal casual work activities or subsistence farming just to survive. This in turn
leads to ever-diminishing chances of their reemployment especially when a certain unemployment duration threshold is passed. Thus, forced long-term unemployment accompanied with
various casual work activities is likely to be a trap for those who for any reason lost a chance to
get a regular job. In this situation, a good choice of policies and reforms in a number of areas
is crucial to alleviate the problem of long-term unemployment and to boost the outflows from
unemployment.
Acknowledgments
The author is grateful to the two anonymous referees, Hartmut Lehmann, Rostislav Kapelyushnikov, Irina Denisova, Jonathan Wadsworth, Michael Beenstock, John Earle, Christian
Belzil, Atanas Christev, Alexander Skiba, Anna Lukyanova, Inna Maltseva for valuable com-

O. Kupets / Journal of Comparative Economics 34 (2006) 228247

245

ments and suggestions. I benefited also from comments by the participants of the IER (Kiev)
International Conference Labor Market Reforms and Economic Growth in Ukraine: Linkages
and Policies in Kiev (March, 2004), of the 8th IZA European Summer School in Labor Economics in Buch (April, 2005), and of the IZA-EBRD International Conference Labor Market
Dynamics, the Role of Institutions and Internal Labor Markets in Transition and Emerging Market Economies in Bologna (May, 2005). Financial assistance from the EERC (Russia) on grant
R02-237, from INTAS (Belgium) on grant YS 2002-249/F7, and from the Economics Research
and Outreach Center (Ukraine) is gratefully acknowledged. Special thanks to the Institute for
the Study of Labor (IZA, Bonn) as the INTAS host institution for support and hospitality. Data
for this study are taken from the first wave of the Ukrainian Longitudinal Monitoring Survey
(ULMS) which has been carried out by the Kiev International Institute of Sociology on behalf of
the international consortium of sponsors led by the Institute for the Study of Labor (IZA, Bonn,
Germany).
Appendix Table 1
Definition of variables
Variable

Definition

Duration of unemployment
(in months)
Observation period
Female
Married

The length of time elapsed between the dates of inflow into and outflow from
unemployment (or censoring date defined as December 31, 2002)
January 1998December 2002
= 1 if Female
= 1 if legally married or cohabiting (i.e. in non-registered marriage), = 0 otherwise (never married, divorced, widowed or separated)
=Integer number from 0 to 4, number of small children aged 15 or less
Three dummy variables for the corresponding age group: from 25 to 39, from 40
to 54, and 55 or older; reference age group is full 24 years or under
Three dummy variables for the corresponding level of education: general secondary or vocational (diploma of high school or PTU); professional secondary
or unfinished higher (diploma of college or at least 3 years of study at the institute/university); higher (diploma of institute/university, any degree); reference
education group is primary or unfinished secondary
= 1 if received unemployment benefits or training allowance during an unemployment spell
= 1 if received income from casual work, production and sale of products from
own land plot, from casual business activities or engaged in subsistence farming
for own needs
= 1 if lived on income of spouse or parents or support from relatives during a
period of unemployment
= 1 if lived on pension during a period of unemployment
= 1 if lived stipend or study loan, child allowance, alimony, social benefits, subsistence allowance, or support by state or municipal institution
= 1 if lived on income from sale of property or rent, dividends, loans or savings
= 1 if employed prior to the start of unemployment, = 0 if previously inactive for
more than 1 month
Registered oblast-level unemployment rate (24 oblasts, Kyiv City and Crimean
Republic) corresponding to the starting quarter of an unemployment spell
Two dummy variables for the corresponding type of settlement where an individual lived at the beginning of an unemployment spell: town (from 20 to 500
thds. inhabitants), large city (more than 500 thds. inhabitants); reference group is
village or very small town (up to 20 thds. inhabitants)
(continued on next page)

Number of children
Age
Education

Unemployment benefits
Casual work

Household income
Pension
Other state transfers
Other sources of subsistence
Previously employed
Regional registered unemployment rate
Type of settlement

246

O. Kupets / Journal of Comparative Economics 34 (2006) 228247

Appendix Table 1 (continued)


Variable

Definition

Macroregion for geographic


location in Table 2

West stands for Chernivetska, Ivano-Frankivska, Khmelnytska, Lvivska,


Rivnenska, Ternopilska, Volynska, Zakarpatska oblasts, Center and North
consists of Kiev City, Vinnytska, Zhytomyrska, Kyivska, Kirovohradska,
Poltavska, Sumska, Cherkaska and Chernihivska oblasts, East includes
Dnipropetrovska, Donetska, Zaporizka, Luhanska and Kharkivska oblasts,
and South consists of Crimean AR and Sevastopol, Mykolayivska, Odeska and
Khersonska oblasts

References
Addison, John T., Portugal, Pedro, 2003. Unemployment duration: Competing and defective risks. Journal of Human
Resources 38, 156191.
Atkinson, Anthony, Micklewright, John, 1991. Unemployment compensation and labor market transitions: A critical
review. Journal of Economic Literature 29, 16791727.
Boeri, Tito, 2001. Transition with labour supply. Discussion paper No. 257. IZA, Bonn.
Brown, David, Earle, John S., 2006. Job reallocation and productivity growth in the Ukrainian transition. In: Handrich,
Lars, Burakovsky, Ihor, Betliy, Olexandra, (Eds.), Labor Market Reforms and Economic Growth in Ukraine: Linkages
and Policies. Alpha-Print Publishing House, Kiev (in Ukrainian and English).
Cox, D.R., 1972. Regression models and life-tables. Journal of the Royal Statistical Society B 34, 187220.
Denisova, Irina, 2002. Staying longer in unemployment registry in Russia: Lack of education, bad luck or something
else? Working paper No. 17. Center for Economic and Financial Research, Moscow.
Devine, Theresa J., Kiefer, Nicholas M., 1991. Empirical Labor Economics: The Search Approach. Oxford Univ. Press,
Oxford.
Earle, John S., Pauna, Catalin, 1996. Incidence and duration of unemployment in Romania. European Economic Review 40, 829837.
Foley, Mark C., 1997. Determinants of unemployment duration in Russia. Discussion paper No. 779. Economic Growth
Center, Yale University.
Grogan, Louise, van den Berg, Gerard J., 2001. The duration of unemployment in Russia. Journal of Population Economics 14, 549568.
Ham, John C., Svejnar, Jan, Terrell, Katherine, 1998. Unemployment and social safety net during transitions to a market
economy: Evidence from the Czech and Slovak Republics. American Economic Review 88, 11171142.
Havrylyshyn, Oleh, 2005. Divergent Paths in Post-Communist Transformation: Capitalism for All or Capitalism for the
Few? Palgrave Macmillan.
Hunt, Jennifer, 1995. The effect of unemployment compensation on unemployment duration in Germany. Journal of
Labour Economics 13, 88120.
International Labour Organisation (ILO), 2004. Sources and Methods: Labour Statistics, vol. 3: Economically Active
Population, Employment, Unemployment and Hours of Work (household surveys). ILO, Geneva.
Kiefer, Nicholas M., 1988. Economic duration data and hazard functions. Journal of Economic Literature 26, 646679.
Kupets, Olga, 2000. The impact of active labor market policies on the outflows from unemployment to regular jobs in
Ukraine. Master of Arts Thesis. EERC, Kiev.
Kupets, Olga, 2005. Determinants of unemployment duration in Ukraine. Working paper No. 05-01. EERC, Moscow.
Lancaster, Tony, 1990. The Econometric Analysis of Transition Data. Cambridge Univ. Press, Cambridge.
Machin, Stephen, Manning, Alan, 1999. The causes and consequences of long-term unemployment in Europe. In: Ashenfelter, Orley, Card, David (Eds.), Handbook for Labor Economics, vol. 3C. Elsevier, North-Holland, pp. 30853139.
Meyer, Bruce D., 1990. Unemployment insurance and unemployment spells. Econometrica 58, 757782.
Narendranathan, Wiji, Stewart, Mark B., 1993. Modeling the probability models of leaving unemployment: Competing
risks models with flexible baseline hazards. Journal of the Royal Statistical Society C 42, 6383.
Nivorozhkina, Lyudmila, Nivorozhkin, Evgenii, Shukhmin, Andrei, 2002. Modeling labour market behaviour of the population of a large industrial city: Duration of registered unemployment. Working paper No. 01-08. EERC, Moscow.
Organisation for Economic Cooperation and Development (OECD), 1993. Long-term unemployment: Selected causes
and remedies. In: OECD Employment Outlook. OECD, Paris, pp. 83117.
Organisation for Economic Cooperation and Development (OECD), 2002. The ins and outs of long-term unemployment.
In: OECD Employment Outlook. OECD, Paris, pp. 187243.

O. Kupets / Journal of Comparative Economics 34 (2006) 228247

247

Paull, Gillian, 2002. Biases in the reporting of labour market dynamics. Working paper No. 02/10. Institute for Fiscal
Studies, London.
Poterba, James M., Summers, Lawrence H., 1995. Unemployment benefits and labour market transitions: A multinomial
logit model with errors in classification. Review of Economics and Statistics 77, 207216.
Sabirianova, Klara, 2002. The great human capital reallocation: A study of occupational mobility in transitional Russia.
Journal of Comparative Economics 30, 191217.
Stetsenko, Serhiy, 2003. On the duration and the determinants of Ukrainian registered unemployment: A case study of
Kyiv. Master of Arts Thesis. EERC, Kiev.
Svejnar, Jan, 1999. Labor markets in the transitional Central and Eastern European economies. In: Ashenfelter, Orley,
Card, David (Eds.), Handbook for Labor Economics, vol. 3B. Elsevier, North-Holland, pp. 18092858.

You might also like