New Deal for Partners:

characteristics and labour market

transitions of eligible couples

August 2002

Commissioned by the
Department for Work and Pensions

Dorothe Bonjour and Richard Dorsett

Disclaimer

The views expressed in this report are those of the authors and do not
necessarily reflect those of the Department for Work and Pensions.

i
 Contents

Contents

Disclaimer i
Contents iii
List of tables v
List of charts vii
Acknowledgements ix
Abbreviations and acronyms xi
Executive Summary xiii

Chapter 1 Introduction 1

Chapter 2 Summary and conclusions 5

Chapter 3 Defining the sample 7

Chapter 4 Characteristics of couples on benefits 11

4.1 Worklessness and benefits 11
4.1.1 Type of worklessness 11
4.1.2 Worklessness and benefits 13
4.2 Unemployment, and job search 17
4.2.1 Looking for work 17
4.2.2 Work experience 18
4.2.3 Characteristics of last job 19
4.3 Personal and household characteristics 20
4.3.1 Basic demographics 20
4.3.2 Region 22
4.3.3 Ethnicity and country of origin 22
4.3.4 Qualifications and education 23
4.3.5 Disability and health 24
4.4 Similarities within couples 25
4.4.1 Age 25
4.4.2 Ethnicity and country of origin 26
4.4.3 Qualifications and education 27
4.4.4 Disability and health 28
4.4.5 Type of worklessness 29
4.4.6 Duration of unemployment 29
4.4.7 Work experience 30
4.4.8 Length of time since last job 30
4.4.9 Manual vs. non-manual employment 31

Chapter 5 Changes over time 33

5.1 Changes in the unemployment/inactivity 33
mix, 1996-2000
5.2 Changes over the observation year 36
5.2.1 The changing economic status of men and 36
women over the observation year

iii
NDP: Characteristics and labour market transitions

5.2.2 Changes in economic status for the full 38

sample
5.2.3 The changing joint economic status of 40
partners over the observation year
5.3 Transitions between individual economic 41
states
5.3.1 The proportion of the sample with experience 42
of each employment status
5.3.2 Transitions between economic states 43
5.3.3 The proportion of the sample with experience 45
of each joint employment status
5.3.4 Transitions between joint economic states 46
5.4 Changes and transitions in benefits 49
claimed
5.4.1 The changing types of benefits claimed over 49
the observation year
5.4.2 The proportion of the sample with experience 50
of the different benefit types
5.4.3 Transitions between types of benefits 51
claimed

Chapter 6 Modelling transitions 55

6.1 The models to be estimated 55
6.1.1 Couple-level analyses 55
6.1.2 Individual-level analyses 56
6.2 Couples’ exits from worklessness 56
6.3 Individuals’ transitions between 63
employment states
6.3.1 Transitions away from unemployment 63
6.3.2 Transitions away from type 1 inactivity 69
6.3.3 Transitions away from type 2 inactivity 74
6.3.4 The relationship between partners’ 78
employment statuses

Appendix 1 Methodology 81

Appendix 2 Results for couples 85

Appendix 3 Results for individuals 89

Bibliography 95

iv
 List of tables

List of tables

3.1 Structure of the sample 8

4.1 Types of worklessness by gender 11

4.2 Aggregated type of worklessness – Distance to the labour market 12
4.3 Types of Benefits claimed 14
4.4 Combination of benefits claimed by couples 15
4.5a Male benefits and distance to labour market 16
4.5b Female benefits and distance to labour market 17
4.6 The type of work sought 17
4.7 How long been looking for work 18
4.8 Whether ever worked 18
4.9 Socio-economic class in last job 19
4.10 Industry of last job 19
4.11 Occupation in last job 20
4.12 Household demographics 21
4.13 Region 22
4.14 Ethnicity and country of origin 23
4.15 Qualifications and education 24
4.16 Disability 24
4.17 Health 25
4.18 Dependence between age of partners 26
4.19 Dependence between ethnic group and country of origin 27
4.20 Dependence between qualifications of partners 27
4.21 Dependence between age completed full-time education 28
4.22 Dependence between disability of partners 28
4.23 Dependence between health problems (lasting more than a year) 29
of partners
4.24 Dependence between economic status of partners 29
4.25 Dependence between duration of unemployment of partners 30
4.26 Dependence between work experience of partners 30
4.27 Dependence between how long since last worked 31
4.28 Dependence between whether last job was manual 31

5.1 The changing economic status of men and women (balanced 36

panel)
5.2 The changing economic status of men and women (unbalanced 40
panel)
5.3 Changes in joint economic status of partners 41
5.4 Proportion of sample with experience of each employment status 42
5.5a Transitions between economic states - Men 43
5.5b Transitions between economic states – Women 45
5.6 Proportion of sample with experience of each joint employment 46
status
5.7 Transitions between joint economic states 47
5.8 The changing type of benefits claimed for men and women 49
(balanced panel)

v
NDP: Characteristics and labour market transitions

5.9 The couple’s changing benefit claiming status (balanced panel) 50

5.10 Proportion of sample with experience of different benefit claiming 50
patterns
5.11 Proportion of couples with experience of different benefit claiming 51
patterns
5.12a Transitions between types of benefits claimed - Men 52
5.12b Transitions between types of benefits claimed – Women 52
5.13 Transitions between couples’ benefits claiming states 53

6.1 Modelling exits from worklessness – selected results 59

6.2 Observed and predicted probabilities of remaining workless 62
6.3 Effect of individual characteristics on unemployment exit 67
6.4 Effect of individual characteristics on type 1 inactivity exits 72
6.5 Effect of individual characteristics on type 2 inactivity exits 77

Appendix Table A1 Modelling exits from worklessness 85

Appendix Table A2 Modelling exits at the individual level: Men 89
Appendix Table A3 Modelling exits at the individual level: Women 92

vi
 List of charts

List of charts

5.1 The relative proportion of unemployment and inactivity over time for 34
men in workless couples
5.2 The relative proportion of unemployment and inactivity over time for 34
women in workless couples
5.3 The numbers in unemployment and inactivity over time for men in 35
workless couples
5.4 The numbers in unemployment and inactivity over time for women 35
in workless couples
5.5 The changing economic status of men 38
5.6 The changing economic status of women 38

6.1 Empirical survivor functions by initial state 58

6.2 Predicted exits from worklessness by initial state 61
6.3 Empirical survivor functions – unemployed men 64
6.4 Empirical survivor functions - unemployed women 64
6.5 Predicted exits from unemployment – Men 66
6.6 Predicted exits from unemployment – Women 66
6.7 Empirical survivor functions - type 1 inactive men 69
6.8 Empirical survivor functions - type 1 inactive women 70
6.9 Predicted exits from type 1 inactivity – Men 71
6.10 Predicted exits from type 1 inactivity – Women 71
6.11 Empirical survivor functions - type 2 inactive men 74
6.12 Empirical survivor functions - type 2 inactive women 75
6.13 Predicted exits from type 2 inactivity – Men 76
6.14 Predicted exits from type 2 inactivity – Women 76

vii
NDP: Characteristics and labour market transitions
 Acknowledgements

Acknowledgements
Material from the Labour Force Survey is Crown Copyright; has been made
available by the Office for National Statistics through the Data Archive and has
been used by permission. Neither the ONS nor the Data Archive bear any
responsibility for the analysis or interpretation of the data reported here. The
authors gratefully acknowledge the advice and assistance of James Calverley,
Jenny Crook, Rachel O’Brien and Dilip Shah at the Department for Work and
Pensions.

Dorothe Bonjour is a Research Fellow and Richard Dorsett is a Principal

Research Fellow, both at PSI.

ix
NDP: Characteristics and labour market transitions
 Abbreviations and acronyms

Abbreviations and acronyms

JSA Jobseeker’s Allowance
IS Income Support
IB Incapacity Benefit
SDA Severe Disablement Allowance
ICA Invalid Care Allowance
LFS Labour Force Survey
NDP New Deal for Partners

xi
NDP: Characteristics and labour market transitions
 Executive Summary

Executive Summary
This report uses longitudinal survey data on 5,600 workless couples observed
over the period 1996-2000 to examine the characteristics of those claiming JSA,
IS, IB, ICA or SDA and therefore eligible for New Deal for Partners. As with
earlier reports on workless couples more generally (ESR79, ESR98), transitions
between employment, unemployment and types of inactivity are a particular
focus and these are also examined using econometric techniques. Three types
of worklessness were considered: unemployment, type 1 inactivity (inactive but
with some desire to work) and type 2 inactivity (inactive and with no desire to
work). These categories can be interpreted as a measure of distance to the
labour market: unemployment is closest, type 2 inactivity furthest away.

Characteristics of couples on benefit

When first observed, about a third of male worklessness was explained by

unemployment, a quarter by type 1 inactivity and two-fifths by type 2 inactivity.
For women, only a tenth were unemployed, a quarter were type 1 inactive, and
three-fifths were type 2 inactive. Male inactivity was dominated by health and
disability. For women, family considerations were more important, especially for
type 2 inactivity.

Men were twice as likely to be seeking full-time rather than part-time work. Only
four per cent of men and 16 per cent of women had never worked. Forty per cent
of men and 24 per cent of women had worked within the last two years. The
proportion of women without work for more than 10 years (at 36 per cent) was
more than twice that for men.

Men were 42 years of age on average compared to 39 years for women. Three-
quarters of couples were married and most lived in rented accommodation. The
sample was predominantly white (90 per cent) and most were originally from the
UK. More than half the women had no qualifications compared to 43 per cent for
men. About two-thirds of the men and half of the women reported a disability of
some form. Similar proportions reported a long-term health problem.

Strong evidence was observed of the similarity of partners with respect to a

number of characteristics: age, ethnicity, country of origin, qualifications and
education, disability and health, type of worklessness, duration of unemployment,
work experience, length of time since last job, type of job and benefits claimed.

In 70 per cent of the couples only the male partner was claiming benefits, in 10
per cent the female partner was the sole claimant and in the remaining 20 per
cent both partners were claiming benefits. For male claimants, the most popular
benefit was JSA. For female claimants the most popular benefit was Incapacity
Benefit. Over 80 per cent of couples claimed only a single type of benefit.

xiii
NDP: Characteristics and labour market transitions

Changes over time

There was a decline in the number of workless couples over the period 1996-
2000. For men, type 2 inactivity accounted for a growing proportion of
worklessness. For women, there was more stability although an increase in the
proportion of type 2 inactivity for other than family reasons was observed.

Over the course of a year, 17 per cent of couples moved from the position of
worklessness to having at least one partner in work. In 3 per cent of cases they
became dual-earner couples, in 10 per cent the man only found work and in
about 5 per cent the woman only.

Of men in employment, 85 per cent were still employed when next observed.
Hence, this appears to be quite a stable economic status. The majority of
leavers became unemployed. Over three-quarters of unemployed men were still
unemployed when next observed, most leavers finding work. More left to type 1
than type 2 inactivity. Inactivity as a whole was characterised by few exits,
although there was some movement between the two types.

For women, employment and type 2 inactivity were even more stable. There was
more movement between unemployment and type 1 inactivity. Those exiting
employment were more likely to become inactive than unemployed. The reason
for female inactivity remained constant for those switching between type 1 and
type 2 inactivity.

Three-quarters of dual-earner couples retained this status when next observed.

In nearly all cases, they were not workless. This was also true for single-earner
couples in most cases. For dual unemployed couples, the chances of being non-
workless when next observed were 23 per cent. For a mix of unemployment and
inactivity, the chances were lower still. Where both partners were inactive, the
chances of being non-workless when next observed were negligible. Partners’
transitions seem to be related - where one partner is further from the labour
market, the other partner appears drawn in the same direction.

All couples were claiming benefits when first observed. After a year, sixteen per
cent were no longer claiming any benefits. A quarter of men were claiming no
benefit by the end of the year. This was due largely to a reduction in the
proportion receiving JSA but also in those claiming IB. For women, there was no
real change.

Overall, benefit claims were relatively stable for men: more than three-quarters of
men were observed claiming the same benefits as in the period before. A fifth of
those claiming JSA were claiming no benefits when next observed. For women,
a similar picture of stability was seen. However, the proportion moving from JSA
to no benefits was twice that for men.

xiv
 Executive Summary

Modelling transitions

The further a couple is from the labour market the more likely it is to remain
workless. For couples where at least one partner was unemployed, by the end of
the observation period 57 per cent remained workless. The corresponding
proportions for other couples ranged from 87 per cent (both type 1 inactive) to 94
per cent (both type 2 inactive).

The longer the workless spell, the less likely the couple is to find work. This
probability is also reduced where either partner has a long term health problem
affecting work or the male partner is claiming either IS and IB or another
combination of benefits. Having qualifications (especially NVQ4 equivalent or
higher) is associated with increased probability of exit.

Over the one year period, more than half the unemployed men changed status,
most finding work. For women, only about a quarter remained unemployed.
Controlling for other characteristics, women are likely to exit unemployment more
quickly than men but not necessarily by finding work.

For type 1 inactivity, most exits are to type 2 inactivity. For both men and
women, exits are concentrated at the beginning of the inactivity spell. The
majority of these exits are to type 2 inactivity. However, within a year and a half,
about a fifth had left type 1 inactivity for unemployment or work.

There were relatively few exits, for either men or women, from type 2 inactivity.
This indicates the rigidity of this type of worklessness. The vast majority of
observed exits were to type 1 inactivity. There were few transitions into work or
unemployment but women were more likely to move from type 2 inactivity straight
into employment.

An inter-relationship between the partners’ statuses was evident. For both men
and women, their partner finding work is mostly associated with moving closer to
the labour market, although not necessarily entering work.

Summary and conclusions

The tendency for exits from worklessness to be concentrated in the early stages
of the spell suggests that it is in these early stages that policy might be most
effective. While the task of helping inactive people find work is certainly difficult,
the movement within types of inactivity provides some grounds for optimism.
Simply helping people take the first step towards finding work may be a
significant achievement with a possible longer-term payoff. However, it is
bridging the gap between unemployment and type 1 inactivity that is likely to
pose the greatest challenge. To be effective, measures should acknowledge
how characteristics affect transitions. The qualitative difference between men
and women in their employment transitions is very apparent.

xv
NDP: Characteristics and labour market transitions
 Introduction

1. Introduction
Unemployment is now at its lowest level for more than twenty years. This,
combined with the fact that inflationary pressures remain weak, suggests that the
UK labour market is currently in good shape. However, while these
macroeconomic indicators are undoubtedly positive, there are microeconomic
considerations that are a cause for concern. In particular, the aggregate
unemployment rates mask important distributional changes in unemployment and
employment. Perhaps the most significant change to the labour market over the
last thirty years has been the growing polarisation of work. That is, there has
been an increased tendency for households to be either ‘work-rich’ (all adults in
work) or ‘work-poor’ (no adults in work) with the intermediate status of a mix
between working and non-working adults becoming increasingly rare. As a
dramatic illustration of this, Gregg et al. (1999) show that the proportion of
households with nobody in work almost tripled from a level of 6.5 per cent in
1975 to 17.9 per cent in 1998. Going back further, the rates are even lower.
Over the same period, the proportion of households where all adults are in work
has risen from 56 per cent to 63 per cent. In 1996 the UK had the fourth highest
rate of workless households out of all OECD countries. The level of polarisation
was higher than in any other OECD country.

There are reasons why this is an important development. From the

macroeconomic perspective, received wisdom suggests there is a relationship
between the extent to which unemployment is concentrated among certain
groups and the extent to which it is effective in reducing wage pressure (Layard
et al, 1991). Hence, an even spread of unemployment maximises its inflation-
quelling efficiency. There are also other concerns at the individual and
household level. Since earnings are the main generator of wealth, households
without work are more likely to be poor. To illustrate this, in 1996 some 70 per
cent of workless households had less than half mean household income. The
corresponding figure for workless households with children was 90 per cent
(Dickens et al., 2000). In 1996, the UK had by far the highest proportion of
children growing up in workless households of any OECD country.

As well as the poverty implications, there are also wider ramifications. Lack of
employment can result in social exclusion as individuals and households become
increasingly distanced from mainstream activities and unable to afford to
participate in outside leisure activities. Being reliant on benefits can result in a
culture of dependency for adults, and children in such households may grow up
lacking an employed role model. Hence, there may be some concern that
children growing up in workless families may themselves have labour market
disadvantages by the time they reach working age. In support of this view,
Johnson and Reed (1996) show that while one in ten men aged 33 had been
unemployed for more than a year in the period 1981-91, this rose to 19 per cent
when considering those men who, at age 16, had unemployed fathers. Using the
same data, Machin (1998) finds that inter-generational mobility is also limited in

1
NDP: Characteristics and labour market transitions

terms of earnings. Furthermore, unemployment can have scarring effects.

Gregg (2001) shows that British men experiencing unemployment when young
are likely to endure long-term labour market disadvantage as a result.
Arulampalam (2001) and Gregory and Jukes (2001) show that this scarring effect
is also evident when considering wages: unemployment imposes a penalty on
future earnings.

An obvious question to ask is why this increase in worklessness arose. One

possibility is that it simply reflects a demographic change. The increased
prevalence of single adult households will, by definition, increase polarisation.
However, this only accounts for a fraction of the trend that has been seen.
Dickens et al. (2000) show that only a third of the observed polarisation can be
explained by changing household composition. The bulk of observed
polarisation is accounted for by different underlying factors. This is clear when
considering a single type of household. For couples, 10.4 per cent of couples
without children and 7.5 per cent of those with children were workless in 1996.
This represents a huge rise on the corresponding proportions in 1968: 2.7 and
1.6 per cent respectively (Gregg et al., 1999).

Some further insight is possible. Two notable labour market trends over the past
twenty or so years have been the increase in female participation and the
increase in male inactivity. If these transitions do not take place in tandem within
the household, increased polarisation must result. Further examination of the
trends shows that, despite the overall rise in women’s employment, there has
been little change for those partnered with jobless men (Desai et al., 1999).
Almost all the increase has been among those with working partners. The main
increases in male inactivity, on the other hand, have been among those aged
over 50. Taken together, these trends are unlikely to both be found within a
single household. Consequently, polarisation has resulted.

An interesting point noted by Gregg et al. (1999) is that while the proportion of
workless households rose rapidly in each of the last three recessions, these
proportions did not revert to their earlier levels during the subsequent recoveries.
In effect, the proportions were ratcheted up with each successive recession.
This suggests that the recovery takes a longer time to permeate workless
couples. There is evidence that this is finally happening. Cooper-Green (2001)
shows that the rate of worklessness has fallen from 18.9 to 16.3 per cent over
the five years to Spring 2001. While this is encouraging, these levels are still
high in a historical context.

All these points suggest the urgency of addressing the problems of worklessness
both from the viewpoint of alleviating existing poverty and preventing longer-term
problems from arising. Furthermore, the importance of encouraging inactive
workers to engage with the labour market is evident. Labour market policy in the
UK is beginning to acknowledge this. A number of programmes are in place that
aim to encourage individuals to move towards employment. Key examples

2
 Introduction

include the New Deal for Lone Parents, the New Deal for Disabled People and
the New Deal for Partners. Furthermore, the New Deal 50 plus is aimed at older
people, not necessarily in receipt of Jobseeker’s Allowance. As another
example, from 2004 work-focused interviews for partners will be available under
Jobcentre Plus. Hitherto, these client groups have not been the focus of
employment efforts. In contrast to the New Deal for Young People and the New
Deal 25 plus which are focused on those in receipt of JSA, the inactivity-focused
New Deals are voluntary.

There is a general acceptance that individuals who are inactive may face
considerable obstacles to employment. First, they may have a specific
characteristic that reduces their employment options (poor health or disability, for
example) or may be trapped by circumstances (as a lone parent, perhaps).
Second, inactivity tends to be long-term in nature. Consequently, those who
have been workless for some time may have outdated skills and be unattractive
to employers who interpret their long spell of inactivity as signalling a more
fundamental problem. In 60 per cent of workless couples, no adult has been in
work within the last three years (Dickens et al., 2000). Machin and Manning
(1999) show that the extent to which the probability of exiting employment falls
with the duration of worklessness is more severe in the UK than anywhere else in
Europe. In view of this, programmes aimed at helping the inactive into work face
the most exacting challenges of any labour market policies. Rather than
measuring the success of such programmes in terms of the number of jobs
secured for participants, it is acknowledged that simply helping individuals move
towards finding a job constitutes a success, albeit qualified. The more tangible
benefits of participation may not arise for some time after the programme itself.
The immediate contributions are likely to be less concrete and include such
things as increased employability, improved confidence and a more positive
attitude towards work.

Clearly, if programmes are to be effective they must be based on a good

understanding of the client group to which they relate. In this report, attention
focuses on those couples eligible for the New Deal for Partners (NDP). NDP is a
voluntary programme which aims to help partners enter or move closer to the
labour market. Those couples who are in receipt of at least one qualifying
benefit1 are eligible to participate. The help on offer includes assistance with job
search, information about in-work benefits and financial assistance towards
childcare and travel costs for partners taking part in labour market programmes,
training and education and access to the Adviser Discretionary Fund. In addition
to understanding the characteristics of these couples, a key interest is in their
transitions between employment states. Using a dataset of eligible couples
drawn from the Labour Force Survey and interviewed five times over the period
of a year, such transitions can be observed. As noted above, moves from
inactivity to employment are rare in the short-term. Consequently, four

1
These comprise Jobseeker’s Allowance, Incapacity Benefit, Income Support, Severe
Disablement Allowance, Disability Living Allowance and Invalid Care Allowance.

3
NDP: Characteristics and labour market transitions

employment states are considered. These comprise employment,

unemployment, inactivity but with a desired to work and inactivity with no desire
to work. These four states can be viewed as discrete points on a continuum
representing ‘distance’ from work.

The structure of the report is as follows. Chapter 2 summarises the main results
and offers some conclusions. In Chapter 3, the data are described. While the
Labour Force Survey is a widely used source of labour market information, the
manipulations required for the purposes of carrying out the analysis in this report
deserve some explanation in order to make subsequent results understandable.
Next, the main results are presented. There are three stages to this. The first
(Chapter 4) is an examination of the characteristics of NDP-eligible couples. In
addition to basic demographics, this section has a particular focus on type of
worklessness. There is also an examination of the extent to which the
characteristics of partners are related. The second set of results (Chapter 5) is
concerned with changes over time. This section exploits the longitudinal nature
of the data. Chapter 6 presents the results of duration models used to analyse
transitions between employment states.

4
 Summary and conclusions

2. Summary and conclusions

This report considers those workless couples eligible for NDP. Descriptive
analysis summarises the individual characteristics of partners within a couple and
shows the tendency for partners to share these characteristics. Making use of
the longitudinal nature of the data, changes over time in employment status and
in benefit receipt are also investigated. The changes in economic status are
investigated more rigorously using econometric techniques. Through duration
analysis, the extent to which these transitions away from worklessness are
associated with particular characteristics is explored. A number of findings
emerge. These mostly confirm the impressions created from the descriptive
analysis but also provide a new interpretation of the results.

Many of the findings are similar to those of Dorsett (2001a, 2001b). In fact, the
couples considered in this report have even less favourable labour market
characteristics than workless couples as a whole (who provided the focus for
these earlier analyses). The tendency for partners to share characteristics,
including obstacles to employment, again suggests that policies that have been
ineffective for one partner may be equally ineffective for the other. Hence, even
within this marginalised sub-group of workless couples, it appears there may be
a polarisation of disadvantage and consequently a core of couples who are very
hard to reach. It is possible to interpret the observed tendency for movements
towards the labour market to be concentrated within couples as being, to some
extent, a manifestation of these shared characteristics.

Another finding shared with the earlier research is that exits from worklessness
are concentrated in the early stages of the non-employment spell. This
reinforces the suggestion that it becomes increasingly difficult to find work as the
length of time without work increases. There are a number of possible reasons
for this but the implication is that policy may be most effective if it tries to help
people early in their worklessness spell.

An important innovation in this report is the concept of distance from the labour
market. Four employment states are considered each one successively ‘further’
from the labour market: working; unemployed; economically inactive but with
some desire to work (‘type 1 inactive’); and, economically inactive with no desire
to work (‘type 2 inactive’). The results demonstrate the robustness of this
characterisation of worklessness and its usefulness in interpreting transitions
between economic states. This lends a greater subtlety to the results. The
richness of the insight that this allows is evident when considering movement
within inactivity. Taken as a whole, inactivity is a rigid employment status with
few exits observed over the one year period for which data were collected.
However, when one allows for the possibility that some types of inactivity are
actually closer to the labour market than others, the impression of inactivity
becomes less one of stagnation and more one of internal transition. This paints

5
NDP: Characteristics and labour market transitions

a more optimistic picture since if policy can operate to influence some movement
towards the labour market, in the long-run this may translate into actual job-entry.

As a general comment, large jumps either towards or away from the labour
market are relatively rare. Moves in both directions tend to be more incremental
in nature. This is particularly true for men for whom most unemployment exits
are accounted for by job entry, and most exits from one type of inactivity are to
the other type of inactivity. For women, the pattern is not so fixed. For example,
a sizeable proportion of female unemployment exits are to type 1 inactivity.
Furthermore, moves from type 2 inactivity to employment, while still rare, are
noticeably more likely than for men. This is likely to be due, at least in part, to
women returning to work after their children reach a particular age. It seems
appropriate to echo the conclusion from the earlier report that policy must be
sensitive to gender differences.

The principal difference between the couples considered in this analysis and
those considered in the earlier research is that the sample in this study is defined
on the basis of benefit receipt. Clearly, it is of policy interest to examine how
labour market transitions vary by benefit type. JSA stands apart from other
benefits as being predominant among those couples and individuals closest to
the labour market. This is unsurprising since JSA is an unemployment rather
than an inactivity benefit. The remaining benefits are largely concentrated
among the inactive, especially type 2 inactivity for women. Consistent with this,
those claiming JSA are most likely to move closer to the labour market and least
likely to move away from it. Conversely, IB is the benefit least associated with a
move towards the labour market.

Finally, the results make clear that a measure of churning is evident between
employment and unemployment on the one hand and the two types of inactivity
on the other. The implication of this is that there is a greater distance between
unemployment and type 1 inactivity than there is between other neighbouring
categories. Seen in this way, the immediate challenge to policy is not
necessarily to seek to achieve a move from inactivity to employment. Rather, it
is to bridge the gap between inactivity and unemployment. It is inactive rather
than unemployed men and women who are most difficult to help into work. The
role for policy in helping unemployed people is to help them overcome the hurdle
of finding and securing employment. With economically inactive people, there is
the additional hurdle of encouraging them to take an interest in the labour market
in the first place – this is likely to be just as significant an obstacle.
Consequently, when assessing the success of a policy intermediate
achievements (increased motivation, for example) should be taken into account.

6
 Defining the sample

3. Defining the sample

All the analysis contained in this report is based on the Labour Force Survey
(LFS). The LFS is a quarterly survey of 60,000 households in the UK with a
focus on those characteristics related to the labour market. It is carried out as a
rotating panel with one-fifth of the respondents being replaced each quarter.
Hence, each (fully-participating) household is interviewed five times over a period
spanning 12 months. All household members at a given address are sampled,
although information on unavailable members of the household is collected by
means of proxy interview. It is the address rather than the household that is the
sampling unit. This means that households leaving or moving to a new address
will not be observed for the full year.2

The longitudinal element of the LFS is important for this analysis and permits
changes over time between economic states to be considered. To do this
required linking records for partners within households, and records for
individuals across the five quarters over which the LFS tracks each household.
In order to maximise the number of observations on the populations of interest, a
number of waves of LFS data were pooled. The resulting dataset spans the
period from the Winter quarter of 1996 to the Winter quarter of 2000. This
observation period pre-dates the introduction of NDP. It comprises those
couples who were observed to be jointly workless when they were first observed
as a couple. This could either be an existing couple being interviewed for the
first time (wave one) or a partnership formed during the time a household was
observed, thus a newly formed couple could be first observed at any wave of the
survey. Since the focus was on working age couples, those couples where one
or both partners were aged 60 years or over at any point were excluded from
further consideration.

The selection of the data set differs slightly from that used in the earlier report on
workless couples (Dorsett, 2001a, 2001b) in three key respects:

• Only couples receiving one or more of the following benefits are included: Job
Seeker’s Allowance (JSA), Income Support (IS), Incapacity Benefit (IB),
Severe Disablement Allowance (SDA) or Invalid Care Allowance (ICA).
These are the benefits that qualify couples for NDP.3
• The dataset starts with the Winter 1996 rather than Winter 1994 quarter. This
later starting point was chosen because it is the first LFS wave which lies fully
in the period after the introduction of JSA in October 1996.
2
In contrast, the British Household Panel Survey (BHPS) tracks movers and those who leave the
household. However, the sample size of the BHPS is too small for the purposes of this report.
3
It has been shown (Jenkins and Laux, 1999) that in the LFS benefits are underreported by
around 20 per cent. This is of course a disadvantage of the LFS as the data selection for this
project is based on benefits. However, there is no other data source at hand which covers
benefits and labour market transitions and has a sufficient sample size.

7
NDP: Characteristics and labour market transitions

• Only couples who were workless when they were first observed were
selected. In the earlier reports, couples were selected if they were observed
workless at any point. This change in the selection criteria should help to get
a more representative mix between the different forms of worklessness.4

In many cases, couples were observed fewer than five times in the final dataset.
Such attrition of the sample may be for a variety of reasons, and the implications
are considered more fully later in the report. However, in addition to the usual
problem of attrition due to non-response to subsequent interviews, there are
problems introduced by the complicated structure of the data. Specifically,
couples only feature in the data while the partnership lasts. The third change in
the sample selection process noted above means that a higher percentage of
couples were observed for all five waves compared to the earlier report and
therefore the attrition problem is less severe.

In Table 3.1 the structure of the sample in terms of response to the five waves is
considered. For each cell, a cross indicates a response to a particular wave.
The first column shows that (by construction) all couples responded at the time of
first being observed. For most couples this time coincides with their first LFS
interview. The exceptions are relationships formed during the time the
household features in the LFS. Such couples can be observed for the first time
at the second, third, fourth or fifth interview.

There were 5,600 couples in the dataset. Of these, over 40 per cent participated
fully. These couples will be referred to as ‘full participants’ later in the text and
the term ‘balanced panel’ will be used to refer to the dataset comprising only full
participants.

Table 3.1: Structure of the sample

Months since first observed jointly non-employed:
0 3 6 9 12 N %
Whether responded in this wave:
X 1195 21.3
X X 811 14.5
X X X 633 11.3
X X X X 601 10.7
X X X X X 2360 42.1
5600

In the remainder of this report, the results concerned with examining the
characteristics of workless couples and changes over the period 1996-2000 are
based on the full sample of 5,600 couples. The descriptive results that consider

4
Unemployed couples might have been overrepresented in the previous report because there is
more movement in and out of unemployment compared to inactivity.

8
 Defining the sample

changes over the observation year and transitions between economic states are
based on the balanced panel, although there is some consideration given to the
likely effects of excluding couples observed for fewer than five waves. The
econometric analysis is based on the unbalanced panel.

9
NDP: Characteristics and labour market transitions
 Characteristics of couples on benefits

4. Characteristics of couples on benefits

In this chapter, the characteristics of all workless couples claiming one of the qualifying
benefits when first observed are reported. To start with, the concept of ‘distance to the
labour market’ is introduced and the relationship between benefits claimed and different
types of worklessness is considered. This is followed by an examination of
unemployment and job search and then a more general consideration of basic individual
and household characteristics. Also of interest is the extent to which partners in a
couple tend to be similar. This is the focus of the last section in this chapter.

4.1 Worklessness and benefits

4.1.1 Type of worklessness

Table 4.1: Type of worklessness by gender

Male Female
ILO unemployed 35.1 10.1
Inactive – seeking, unavailable, student 0.1 0.0
Inactive – seeking, unavailable, looking after family, home 0.2 0.4
Inactive – seeking, unavailable, temporarily sick or injured 0.5 0.1
Inactive – seeking, unavailable, long-term sick or disabled 0.2 0.1
Inactive – seeking, unavailable, other reason 0.6 0.2
Inactive – seeking, unavailable, no reason given 0.1 0.0
Inactive – not seeking, would like work, waiting results of job application 0.1 0.1
Inactive – not seeking, would like work, student 0.2 0.2
Inactive - not seeking, would like work, looking after family, home 2.7 13.6
Inactive - not seeking, would like work, temporarily sick or injured 2.4 0.8
Inactive - not seeking, would like work, long term sick or disabled 17.8 6.7
Inactive - not seeking, would like work, believes no job available 0.6 0.5
Inactive - not seeking, would like work, not started looking 0.2 0.8
Inactive - not seeking, would like work, not looked 0.8 2.1
Inactive - not seeking, would like work, no reason - -
Inactive - not seeking, not like work, waiting results of job application 0.1 0.0
Inactive - not seeking, not like work, student 0.4 0.9
Inactive - not seeking, not like work, looking after family, home 4.8 38.8
Inactive - not seeking, not like work, temporarily sick or injured 1.5 1.1
Inactive - not seeking, not like work, long term sick or disabled 29.4 19.3
Inactive - not seeking, not like work, not need or want job 0.2 0.6
Inactive - not seeking, not like work, retired 0.9 1.0
Inactive - not seeking, not like work, other reason 0.8 2.3
Inactive - not seeking, not like work, no reason given 0.5 0.3
Total 5600 5600

11
NDP: Characteristics and labour market transitions

Table 4.1 shows the profile of non-employment for men and women.5 There are clear
differences. For men, 35 per cent of non-employment was explained by unemployment.
Nearly half was accounted for by long-term sickness or disability. For women, both
unemployment and inactivity due to long-term sickness or disability were less important,
they accounted for ten and 26 per cent respectively. More important were domestic
responsibilities with over half of workless women inactive due to looking after the family
or home.

One possible explanation for the higher number of men than women that were inactive
due to sickness or disability may be the persistence of traditional gender roles whereby
the man’s role is that of breadwinner while the woman’s role is that of homemaker. If
partners of sick or disabled men are looking after the family or home and partners of
sick or disabled women are working then one would expect a higher percentage of sick
or disabled men among workless couples.

Table 4.2: Aggregated type of worklessness – Distance to the labour market

(column percentages)
Aggregate type of worklessness Male Female
Unemployed 35.1 10.1
Inactive, type 1 26.4 25.7
family reasons 13.6
other reasons 12.1
Inactive, type 2 38.5 61.2
family reasons 38.8
other reasons 25.4
Total 5600 5600

Table 4.2 aggregates the information in Table 4.1 to derive a measure of distance from
the labour market. Three categories are considered: unemployment and two types of
inactivity which, for convenience, will be labelled type 1 inactivity and type 2 inactivity in
the remainder of this report. Type 1 inactivity comprises people who are inactive and
who are either seeking work or, if not seeking, would like work. Type 2 inactivity
includes those who are inactive and who are not seeking and do not want work. It
seems justifiable to view these categories as indications of distance from the labour
market. Unemployed people are trying to find work and are therefore closer to the
labour market than type 1 inactive people who show signs of being interested in
working. In turn, type 1 inactivity can be seen as being closer to the labour market than
type 2 inactivity since, in the latter case, there is no interest in working.

There are differences between men and women in how they are distributed across
these categories. A quarter of the men in workless couples were type 1 inactive and
two-thirds of these were suffering from long-term sickness or disability. Two-fifths of the

5
These categories are based on the International Labour Organisation (ILO) definitions.

12
 Characteristics of couples on benefits

men were type 2 inactive and this was due to being long-term sick or disabled in three-
quarters of cases. Inactivity was more pronounced for women and less dominated by
health considerations. In view of this, Table 4.2 draws a distinction between being
inactive for family reasons and being inactive for other reasons. As with men, about a
quarter were type 1 inactive. Slightly more than half of this type of inactivity was due to
family responsibilities. Type 2 inactivity was much higher than for men and within this
category, family responsibilities were considerably more important than other reasons.
It should be noted that long-term sickness or disability accounted for the majority of
other reasons for both type1 and type 2 inactivity and consequently, these other
reasons are comparable to the reasons cited by men for their inactivity.

Introducing these additional sub-divisions for women is not intended to achieve a finer
gradation of the concept of distance from the labour market. For example, it is unclear
whether type 1 inactivity for family reasons or type 1 inactivity for other reasons is closer
to the labour market. The purpose of retaining the distinction is that it is useful in
interpreting some of the findings that follow.

4.1.2 Worklessness and benefits

The remainder of this section considers the extent to which particular benefits
characterise types of worklessness. Further consideration is given to benefits in
chapters 5 and 6.

The sample consists of couples who were claiming one of the following benefits: JSA,
IS, IB, SDA or ICA. In the next three tables, more details about the types of benefits
claimed, the correspondence between partners’ claims, and between claims and type of
worklessness are presented. Throughout this section it is helpful to keep in mind that
some of the benefits are paid out to one person for the couple (JSA, IS) while others are
designed as individual benefits (IB, SDA, ICA).

It is also important to note that benefits in the LFS are self-reported. Jenkins and Laux
(1999) show that this under-counts true numbers of benefit claimants by around 20 per
cent. There are differences in this under-reporting by benefit type. Among the NDP
benefits there was least under-reporting of IB (3 per cent). For the other benefits,
roughly similar levels of under-reporting were observed: JSA - 23 per cent; IS - 27 per
cent; SDA - 26 per cent; and, ICA - 26 per cent (all numbers taken from Jenkins and
Laux, Table 2). Non-reporting may sometimes validly occur among individuals within
couples when, for example, one partner is claiming JSA at the higher rate for a
dependent partner and the other partner reports not receiving JSA. However, in this
scenario, the couple will still be identified as eligible for NDP by virtue of the partner who
does report a claim.

13
NDP: Characteristics and labour market transitions

Table 4.3: Types of Benefits claimed

(column percentages)

Type and combination of benefits claimed Male Female

No benefit 9.9 70.1
JSA only 33.7 4.2
IS only 11.8 6.1
IB only 25.3 10.1
SDA only 1.0 2.7
ICA only 1.0 2.3
JSA and IS 0.7 0.0
JSA and IB 0.1 0.0
JSA and ICA 0.1 -
IS and IB 12.9 2.2
IS and SDA 0.9 1.0
IS and ICA 1.4 0.5
IB and SDA 0.5 0.3
IB and ICA 0.2 0.1
SDA and ICA 0.1 0.1
JSA, IS and IB 0.1 -
IS, IB and SDA 0.4 0.2
IS, IB and SDA 0.1 0.0
IB, SDA and ICA 0.1 0.0
IS, IB, SDA and ICA 0.1 -

Both partners claiming benefits 20.1

Total 5600 5600

Number of different benefits claimed

One benefit 80.6 84.8
Two or more benefits 19.4 15.2
Total 5047 1676
Notes: Combinations not listed are not observed in the data set. Based on self-reported benefits, thus
contains theoretically impossible benefit combinations.

Table 4.3 contains information about the types and combinations of benefits claimed,
whether both partners claim and the number of different benefits claimed. In 70 per
cent of the couples only the male partner was claiming benefits, in 10 per cent the
female partner was the sole claimant and in the remaining 20 per cent both partners
were claiming benefits. For male claimants, the most popular benefit was JSA, followed
by IB, the combination of IS and IB, and IS. For female claimants the most popular
benefit was IB and a higher proportion of females compared to males were claiming
SDA and ICA only. These gender differences can be explained by the fact the some of
the benefits are designed for couples while others are individual-based. Where benefits
are for a couple they seem to be claimed for by the male partner in the majority of

14
 Characteristics of couples on benefits

cases. Among those claiming benefits, over 80 per cent claimed only a single type of
benefit.

Table 4.4: Combination of benefits claimed by couples

(row percentages)
JSA only IS only IB only IS & IB Other Total
Only one is claiming:
Only male partner claiming 40.8 12.3 25.3 16.2 5.5 3924
Only female is claiming 17.4 23.0 27.5 13.7 18.4 553

Both are claiming:

Female claiming:
Male claiming: JSA only IS only IB only IS & IB Other Total
JSA only 39.1 25.4 21.5 2.5 11.6 284
IS only 3.9 30.3 22.5 7.9 35.4 178
IB only 3.8 9.9 57.7 2.4 26.2 423
IS & IB 3.5 20.7 26.4 5.8 43.7 87
Other combination 0.7 18.5 31.1 8.6 41.1 151

Table 4.4 contains information on benefits claimed by couples. The first two rows show
the benefits claimed when only one in the couple was claiming. The emerging picture is
similar to the overall distribution of benefits. The lower part of the table contains the
combination of benefits claimed by couples where both partners were claiming benefits.
The claims of the two partners were clearly related and there is some indication that
couples were claiming similar benefits. It should be kept in mind that all the information
on benefits in the LFS is based on self-reporting.

Tables 4.5a and 4.5b focus on the dependence between benefits claimed and distance
to the labour market of the claimant and his/her partner. Table 4.5a looks at the
distance to the labour market by the type of benefits the male partner is claiming and
Table 4.5b is the equivalent for couples where the female partner is claiming benefits.
To reduce the size of the tables and increase their readability, information on couples
where both partners are claiming is ignored.

15
NDP: Characteristics and labour market transitions

Table 4.5a: Male benefits and distance to labour market *)

(row percentages)
Men’s (own) distance: Unemp Inactive type 1 inactive type 2 Total
JSA only 90.3 6.8 3.0 1601
IS only 1.9 41.2 57.0 481
IB only 2.4 32.9 64.7 994
IS and IB 2.8 47.2 50.0 634
Other combinations 17.8 24.8 57.5 214
Total 39.1 25.1 35.8 3924
Inactive type 1 inactive type 2
Women’s (partner) Unemp All family other All family other Total
distance:
JSA only 16.4 28.7 19.8 8.9 55.0 44.9 10.1 1601
IS only 4.8 22.7 16.0 6.7 72.6 55.3 17.3 481
IB only 4.8 20.6 12.0 8.6 74.6 52.0 22.6 994
IS and IB 4.3 28.7 18.8 9.9 67.1 51.0 16.1 634
Other combinations 5.1 20.6 11.2 9.4 74.3 51.9 22.4 214
Total 9.5 25.4 16.7 8.7 65.1 49.3 15.8 3924
*)
Couples where only male partner is claiming benefits

As would be expected, there is a high association between the type of benefit claimed
and labour market status. Ninety per cent of males who claimed JSA only were
unemployed. Men claiming IB were more likely to be type 2 inactive compared to men
claiming IS or IS and IB. The female partner’s labour market status was less dependent
on the type of benefit her partner was claiming. The most notable association was that
women were much more likely to be unemployed if their partner was claiming JSA.
With regard to inactivity, women were more likely to be type 2 inactive for other than
family reasons if their partner was claiming IB. They were least likely to be type 2
inactive if their partner was claiming JSA.

For couples where the female partner was the only claimant, JSA was once again
associated with being closer to the labour market than were other benefits. This was
true for both partners. With regards to other benefits, there did not appear to be much
difference in how they correlated with female distance.

16
 Characteristics of couples on benefits

Table 4.5b: Female benefits and distance to labour market *)

(row percentages)
Men’s (partner)
distance: unemp Inactive type 1 inactive type 2 Total
JSA only 57.3 14.6 28.1 96
IS only 16.5 29.9 53.5 127
IB only 23.7 21.7 54.6 152
IS and IB 23.7 42.1 34.2 76
Other combinations 15.7 36.3 48.0 102
Total 26.4 27.9 45.8 553
Inactive type 1 inactive type 2
Women’s (own) Unemp All family other All family other Total
distance:
JSA only 72.9 16.7 5.2 11.5 10.4 5.2 5.2 96
IS only 0.8 24.4 24.4 74.8 11.8 63 127
IB only 26.3 1.3 25 75.5 6.4 69.1 152
IS and IB 1.3 27.6 2.6 25 71 10.5 60.5 76
Other combinations 2.0 25.5 4.9 20.6 72.5 8.8 63.7 102
Total 13.4 24.2 2.5 21.7 62.4 8 54.4 553
*)
Couples where only female partner is claiming benefits.

4.2 Unemployment and job search

4.2.1 Looking for work

Among those looking for work there were gender differences with respect to the type of
work sought and the length of time individuals had been looking for work. Table 4.6
shows that men were nearly three times as likely as women to consider either
employment or self-employment. Women were more likely to be seeking work as an
employee and less likely to consider self-employment. With regard to hours worked,
men were twice as likely to be seeking full-time rather than part-time work. Hardly any
men were seeking part-time work, compared to a third of women.

Table 4.6: The type of work sought

(column percentages)
The type of work sought Male Female
Employee 79.3 92.2
Self-employed 3.7 1.7
No preference 17.0 6.1
Total 2045 605

Full-time 84.5 43.0

Part-time 3.2 34.1
No preference 12.4 22.9
Total 1968 595

17
NDP: Characteristics and labour market transitions

Table 4.7: How long been looking for work

(column percentages)
How long been looking for work Male Female
Not yet started 0.4 0.3
Less than 1 month 6.1 13.7
1 month but less than 3 months 15.1 22.1
3 months but less than 6 months 12.7 14.6
6 months but less than 12 months 16.2 17.6
12 months but less than 18 months 9.2 9.4
18 months but less than 2 years 5.2 3.6
2 years but less than 3 years 7.2 6.5
3 years but less than 4 years 5.0 4.2
4 years but less than 5 years 3.4 0.8
5 years or more 19.5 7.2
Total 2062 615

Table 4.7 shows that the main differences between men and women in how long they
had been looking for work were at the extremes; 36 per cent of women had been
looking for less than 3 months compared to 22 per cent of men, while 20 per cent of
men had been looking for more than 5 years compared to 7 per cent of women.

4.2.2 Work experience

Another interesting point to investigate is the past work experience of couples. Table
4.8 gives details of whether they had ever worked and how long ago.

Table 4.8: Whether ever worked

(column percentages)
Male Female
Whether ever worked:
Yes 96.1 84.1
No 4.0 15.9
Total 5594 5595

How long since last worked:

Less than a year 18.1 9.8
1 to 2 years 21.6 14.4
3 to 5 years 20.4 18.4
6 to 10 years 23.8 21.9
More than 10 years 16.2 35.5
Total 5389 4775

18
 Characteristics of couples on benefits

Very few had no experience of employment. Table 4.8 shows that only four per cent of
men and 16 per cent of women had never worked. There was a difference between the
sexes in how recent this experience was. For men, 40 per cent had worked within the
last two years. For women, the corresponding proportion was 24 per cent. At the other
extreme, the proportion of women without work for more than 10 years (at 36 per cent)
was more than twice that for men. This might be a reflection of the fact that men were
on average closer to the labour market; they were more likely to be unemployed and
less likely to be type 2 inactive than females.

4.2.3 Characteristics of last job

Tables 4.9 to 4.11 consider some of the characteristics of the most recent job. The
largest categories for men in Table 4.9 relate to skilled manual and partly skilled
occupations. Women were most likely to be in the partly skilled and skilled non-manual
categories. Over 60 per cent of men and women were concentrated in one of the top
two categories of their respective gender.

Table 4.9: Socio-economic class in last job

(column percentages)
Socio-economic group in last job Male Female
Professional occupations 1.6 0.3
Intermediate occupations 11.9 13.2
Skilled non-manual occupations 7.6 26.5
Skilled manual occupations 39.8 11.4
Partly skilled occupations 27.0 35.0
Unskilled occupations 11.4 13.6
Members of armed forces 0.6 -
Total 4003 2632

Table 4.10: Industry of last job

(column percentages)
Industry Male Female
Agriculture and fishing 2.1 0.9
Energy and water 2.4 0.3
Manufacturing 27.5 19.8
Construction 18.4 0.9
Distribution, hotels and restaurants 18.6 31.5
Transport and communication 10.0 2.6
Banking, financial and insurance 8.3 9.2
Public administrations, education and health 7.6 27.0
Other services 4.3 7.6
Workplace outside UK 0.8 0.5
Total 3989 2632

19
NDP: Characteristics and labour market transitions

Table 4.10 considers the industry of the last job. The three most popular industries for
men were manufacturing, construction and distribution, hotels and restaurants. In total,
65 per cent of men worked in one of these three industries. Women were even more
concentrated with 78 per cent in the three most popular industries. Two of these –
manufacturing and distribution, hotels and restaurants – were also among the top three
for men. The third main industry for women was public administration, education and
health.

Table 4.11: Occupation in last job

(column percentages)
Male Female
Managers and administrators .5 6.0
Professional occupations 2.5 1.8
Associate professional and technical occupations 2.7 4.1
Clerical, secretarial occupations 5.6 12.6
Craft and related occupations 25.2 6.5
Personal, protective occupations 9.7 23.6
Sales occupations 3.8 14.3
Plant and machine operatives 26.1 12.3
Other occupations 16.1 18.7

Manual occupation 78.7 59.3

Non-manual occupation 20.7 40.7
Armed forces 0.6 -

Total 4003 2632

Table 4.11 presents information on the occupation of the last job. Men had worked
mainly as plant and machine operatives and in craft and related occupations. Women,
on the other hand, had worked mainly in personal and protective and sales occupations.
For both sexes, ‘other occupations’ scored quite highly. A much higher percentage of
men (79 per cent) compared to women (59 per cent) worked in manual occupations.

4.3 Personal and household characteristics

4.3.1 Basic demographics

As shown in Table 4.12, the distribution of ages was quite similar for both partners in
the middle of the age distribution (from age 31 to age 50). Men were relatively
concentrated in the highest age group and women in the three youngest groups (from
age 16 to age 30). This difference is also reflected in the average age which was 42.1
years for men and 39.1 years for women.

20
 Characteristics of couples on benefits

Table 4.12: Household demographics

(column percentages)
Male Female
Age range:
16-17 0.1 1.3
18-24 7.1 11.5
25-30 11.9 14.5
31-40 22.3 23.4
41-50 28.6 27.6
51-60 30.1 21.8
Total 5600 5600

Marital status:
Never married 16.8 17.2
Married, living with husband/wife 76.2 76.2
Married, separated from husband/wife 1.0 0.9
Divorced 5.9 5.4
Widowed 0.1 0.3
Total 5600 5600

Dependent children:
No 40.9
Yes 59.1
Total 5600

Housing tenure type:

Owned outright 10.6
Being bought with mortgage or loan 19.2
Part rent, part mortgage 0.6
Rented 69.3
Rent free / squatting 0.7
Total 5599

The marital statuses of men and women were very similar. By definition, the same
number of men and women were married and living together. This category comprised
three-quarters of the sample, the remainder being cohabitees. Dependent children
were present in three-fifths of all households. Most couples were living in rented
accommodation but about a third owned their property or were buying it on a mortgage.

21
NDP: Characteristics and labour market transitions

4.3.2 Region

The regional breakdown of the sample is represented in Table 4.13. 18 per cent of the
couples in the sample were living in London and the South East.6

Table 4.13: Region

(column percentages)
Region
Tyne and Wear 3.4
Rest of northern region 6.0
South Yorkshire 3.9
West Yorkshire 3.7
Rest of Yorkshire & Humberside 3.0
East Midlands 6.9
East Anglia 2.7
Inner London 4.2
Outer London 4.6
Rest of South East 9.1
South West 4.7
West Midlands (Met County) 5.8
Rest of West Midlands 3.9
Greater Manchester 5.6
Merseyside 3.8
Rest of North West 4.5
Wales 8.0
Strathclyde 5.7
Rest of Scotland 4.4
Northern Ireland 6.5
Total 5600

4.3.3 Ethnicity and country of origin

Just over 90 per cent of the sample was white. This was true for both men and women.
Most were also originally from the UK. Of those with a different country of origin, many
had been in the UK for a long time; 74 per cent of men and 64 per cent of women had
been here for more than 10 years. Considerably fewer men and women arrived in the
UK less than a year ago than was true for workless couples as a whole (Dorsett,
2001a). This may reflect differences in the sample selection process. Here, only
couples on specific benefits were considered. As it takes time to get to know the benefit
system, to submit a benefit claim and to wait for the outcome it is not unexpected that
this sample contains fewer new arrivals to the UK.
6
Note that it is not straightforward to draw a comparison with the proportion of the population living in
London and the South East since the data used in this report have not been weighted to account for the
higher levels of survey non-response typically observed in London and the South East.

22
 Characteristics of couples on benefits

Table 4.14: Ethnicity and country of origin

(column percentages)
Male Female
Ethnic group:
White 90.0 90.2
Black – Caribbean 0.3 0.3
Black – African 0.5 0.4
Black – other 0.1 0.1
Indian 2.3 2.4
Pakistani 3.4 3.5
Bangladeshi 1.5 1.4
Chinese 0.2 0.1
Other 1.8 1.6
Total 5596 5597

Country of origin:
UK, British 87.3 87.5
Irish Republic 1.1 1.0
Other 11.6 11.4
Total 5599 5599

How long resident in UK:

Less than a year 1.8 3.3
1 to 2 years 4.2 5.9
3 to 5 years 9.1 10.0
6 to 10 years 11.3 16.9
11 to 20 years 15.6 24.1
21 to 30 years 24.0 24.1
More than 30 year 34.1 15.8
Total 684 673

4.3.4 Qualifications and education

Table 4.15 shows that men were better qualified than women despite leaving full-time
education at a younger age. More than half the women had no qualifications compared
to 43 per cent for men. The difference in qualifications is particularly notable at the
NVQ 3 level where men were more than three times as represented as women. Slightly
more men (87 per cent) than women (85 per cent) left education at the age of 16 or
under. The apparent contradiction of men leaving education at a younger age yet
appearing more qualified is explained by the fact that the qualifications are given in
terms of their NVQ equivalents. It is likely that vocational qualifications contributed to
the apparent better performance of the men. This is further evidenced by the much
higher percentage of men who had completed an apprenticeship, 19 per cent compared
to only three per cent of women.

23
NDP: Characteristics and labour market transitions

Table 4.15: Qualifications and education

(column percentages)
Male Female
Highest qualification:
NVQ 4 or higher 4.8 4.4
NVQ 3 22.2 6.9
NVQ 2 10.3 16.9
NVQ 1 6.2 8.6
Other 13.5 7.1
None 43.1 56.1
Total 5458 5492

Age completed full-time education:

Under 16 47.7 42.3
16 39.6 42.7
17-18 7.2 10.2
Over 18 5.5 4.9
Total 5533 5465

Whether doing/done apprenticeship:

yes (completed or still doing) 19.4 3.4
no (including begun but discontinued) 80.6 96.6
Total 5567 5581

4.3.5 Disability and health

Table 4.16 shows that there was a high level of disability and ill-health reported among
the sample. This is unsurprising given that eligibility for NDP relies on receipt of either
JSA or another, often health-related benefit. About two-thirds of the men and half the
women reported a disability of some form. Similar proportions reported a long-term
health problem. In the majority of cases, the disability was work-limiting.

Table 4.16: Disability

(column percentages)
Male Female
Whether has a disability:
DDA disabled and work-limiting disabled 56.9 39.2
DDA disabled only 1.4 1.8
Work-limiting disabled only 7.6 6.1
Not disabled 34.1 53.0
Total 4670 4673
Note: DDA disabled relates to individuals who have a long-term disability that substantially limits their
day-to-day activities. Work-limiting disabled relates to individuals who have a long-term disability which
affects the kind or amount of work they can do.

24
 Characteristics of couples on benefits

Table 4.17 presents more detailed results for the health status of the sample. The vast
majority of reported long-term health problems affected the kind and the amount of work
that respondents could undertake.

Table 4.17: Health

(column percentages)
Male Female
Health problem lasting more than a year: 67.3 49.8
If yes:
- Health problem affects kind of work possible: 92.5 86.6
- Health problem affects amount of work possible: 85.4 80.4
Total 5174 5170
Note: Only data from Spring 97 onwards was used in this table as the structure of health questions
changed. As a consequence 419 couples are excluded.

4.4 Similarities within couples

In this section, the extent of the similarity between partners in a couple with respect to
various characteristics is considered. For some of these characteristics (age, for
example) there is no possible causal relationship. That is to say, forming a couple will
not affect the age of either partner. For other characteristics it is plausible that a causal
relationship exists. In other words, the characteristic of one partner may affect the
corresponding characteristic of the other. This section begins by first considering those
individual characteristics which can be regarded as unaffected by the other partner7
before moving on to a consideration of employment-related characteristics, for which
this assumption is invalid. However, in all cases it is important to bear in mind that there
is no attempt to explain any similarities found, merely to report them.

All of the tables in this section follow the same format and tabulate the male
characteristic in question against the corresponding female characteristic. The
percentages in each cell are row percentages and show the proportion of the men with
a given characteristic who are partnered with women with that or other characteristics.
It is possible to compare the observed tabulation with that which would be expected
were the characteristics of the partners independent of each other. This null hypothesis
of independence can then be tested statistically. For completeness, the chi-squared
statistic is presented in each table. However, it would suffice to say that, in all cases,
the null hypothesis of independence is rejected beyond any reasonable doubt.

4.4.1 Age

Table 4.18 shows that there is a clear tendency to partner somebody of a similar age.
In all age categories but one, the entries on the leading diagonal (shown in bold) exceed
7
Although in some cases this assertion may be debatable.

25
NDP: Characteristics and labour market transitions

those elsewhere in the row. The exception to this rule are the seven men aged 16 or 17
who are partnered with older females. Comparing these entries with those in the final
row also shows women to be disproportionately partnered with someone of a similar
age. In most cases the second largest entry in a row is to the left of the leading
diagonal indicating that men were more likely to partner younger than older women.

Table 4.18: Dependence between age of partners

(row percentages)
Female:
Male: 16-17 18-24 25-30 31-40 41-50 51-60 Total
16-17 28.6 71.4 7
18-24 12.2 69.9 13.9 3.5 0.5 395
25-30 2.4 34.9 46.6 14.1 2.0 665
31-40 0.5 9.0 28.9 51.5 9.7 0.4 1251
41-50 0.0 0.9 4.4 29.7 56.5 8.5 1559
51-60 0.1 0.9 4.9 30.1 64.0 1683
Total 1.3 11.5 14.5 23.4 27.6 21.8 5600
Chi-squared (25) = 6990

4.4.2 Ethnicity and country of origin

There were also strong correlations between partners with respect to ethnic group and
country of origin. Whereas nine per cent of non-white men were partnered with white
women, less than one per cent of white men were partnered with non-white women.
The figures for country of origin are slightly less dramatic but are still marked. One
quarter of men originating from outside the UK had partners from the UK, while the
corresponding figure for men from the UK was 97 per cent.

26
 Characteristics of couples on benefits

Table 4.19: Dependence between ethnic group and country of origin

(row percentages)

Ethnic group:
Female:
Male: Non-white White Total
Non-white 91.4 8.6 559
White 0.7 99.3 5037
Total 9.8 90.2 5596
Chi-squared(1) = 4693

Country of origin:
Female:
Male: From outside UK From UK Total
From outside UK 75.0 25.0 713
From UK 3.3 96.7 4886
Total 12.5 87.5 5599
Chi-squared(1) = 2931

4.4.3 Qualifications and education

Table 4.20 displays a clear tendency for partners to have similar levels of highest
qualification. For example, while only 5 per cent of women had a qualification at NVQ 4
level or higher, this rose to 26 per cent among those whose partner had this level of
qualification. At the other end of the scale, 56 per cent of all women had no
qualifications but, among those whose partners had no qualifications, the level was 70
per cent.

Table 4.20: Dependence between qualifications of partners

(row percentages)
Female:
Male: NVQ4 NVQ3 NVQ2 NVQ1 Other None Total
NVQ 4 or higher 26.1 12.5 19.8 8.2 8.6 24.9 257
NVQ 3 6.1 9.8 19.2 8.4 7.3 49.2 1190
NVQ 2 6.2 10.1 30.0 10.8 5.7 37.3 547
NVQ 1 1.2 7.3 27.1 15.6 4.9 43.9 328
Other 3.5 6.9 12.2 7.8 17.8 51.9 721
None 1.6 4.1 12.0 7.1 4.3 70.9 2314
Total 4.5 7.0 16.7 8.4 7.2 56.2 5357
Chi-squared(25) = 875

Unsurprisingly, a corresponding pattern is found when considering the age at which full-
time education was completed. While only 5 per cent of women stayed in education

27
NDP: Characteristics and labour market transitions

beyond the age of 18, for those with partners educated beyond this point the proportion
was 37 per cent. At the other extreme, 63 per cent of women partnered with men who
left school before the age of 16 did likewise. The figure for women in the whole sample
was 42 per cent.

Table 4.21: Dependence between age completed full-time education

(row percentages)
Female:
Male: Under 16 16 17-18 Over 18 Total
Under 16 63.0 28.3 7.0 1.7 2604
16 23.4 63.0 10.8 2.8 2155
17-18 24.9 40.5 22.3 12.3 390
Over 18 18.7 25.4 18.7 37.3 284
Total 42.2 42.8 10.2 4.7 5433
Chi-squared(9) = 1722

4.4.4 Disability and health

Dependence between the statuses of partners was also evident when considering
disability and health. Table 4.22 shows that all levels of disability were more highly
represented among women partnered to men with a similar level of disability than they
were in the sample of women as a whole.

Table 4.22: Dependence between disability of partners

(row percentages)
Female:
Male: Disabled Disabled Work- Not Total
and not limiting disabled
working working disability
Disabled and working 44.3 2.3 5.7 47.7 2655
Disabled, not working 45.5 4.6 1.5 48.5 66
Work-limiting disability 33.3 2.0 13.5 51.3 357
Not disabled 31.7 0.8 5.2 62.3 1590
Total 39.2 1.8 6.1 53.0 4668
Chi-squared(9) = 138

Also self-reported long-term health problems were concentrated within couples albeit
the correlation was somewhat weaker.

28
 Characteristics of couples on benefits

Table 4.23: Dependence between health problems (lasting more than a year) of
partners
(row percentages)
Female:
Male: Problem No problem Total
Long-term health problem 54.7 45.3 3737
No long-term health problem 39.7 60.3 1796
Total 49.9 50.2 5533
Chi-squared(1) = 110

4.4.5 Type of worklessness

Table 4.24 considers the distance to the labour market in order to examine the extent to
which individuals tended to be partnered with those of a similar employment status.
There was clearly a tendency to be partnered with somebody of a similar status.
Whereas ten per cent of women in workless households were unemployed, the
corresponding proportion for those partnered with unemployed men was 20 per cent.
Type 1 inactive women were more likely to be with men with the same type of inactivity
and similarly, type 2 inactive women were more likely to be partnered with type 2
inactive men. For both types of inactivity, women inactive for family reasons were more
likely to be partnered with unemployed men than were women inactive for other
reasons, possibly suggesting that female inactivity due to family reasons is less
permanent and closer to the labour market than inactivity due to other reasons. This
hypothesis will be investigated further when transitions between types of worklessness
are analysed.

Table 4.24: Dependence between economic status of partners

(row percentages)
Female:
Male: Inactive type 1 Inactive type 2
Unemp. family others family others Total
Unemployed 20.1 17.9 10.2 38.7 13.2 1966
Inactive type 1 5.6 20.2 22.2 31.1 20.8 1477
Inactive type 2 4.1 5.2 7.0 44.1 39.7 2157
Total 10.1 13.6 12.1 38.8 25.4 5600
Chi-squared(8) = 1006

4.4.6 Duration of unemployment

There was a strong association of partners’ unemployment duration among workless

couples. More details are revealed in Table 4.25. In jointly unemployed couples, men
were likely to have had longer unemployment spells than their partners. Furthermore,
there is a clear correlation between durations. Comparing the leading diagonal with the

29
NDP: Characteristics and labour market transitions

final row shows that, for all categories of duration, the proportion is higher for those
women partnered with a man in a similar category than among women as a whole. For
example, whereas only 33 per cent of women in jointly unemployed couples had an
unemployment duration of less than three months, for those with partners of the same
duration the proportion rose to 67 per cent.

Table 4.25: Dependence between duration of unemployment of partners

(row percentages)
Female:
Male: <3 m 3-6 6-12 1-2 y 2-5 y >5 y Total
Less than 3 months 67.3 16.4 7.7 4.8 2.9 1.0 104
3 months but less than 6 36.9 29.2 9.2 15.4 6.2 3.1 65
6 months but less than 12 23.7 17.2 43.0 7.5 8.6 0.0 93
1 year but less than 2 17.7 7.8 29.4 31.4 11.8 2.0 51
2 years but less than 5 19.1 6.4 19.1 12.7 31.8 11.1 63
5 years or more 14.1 4.7 15.6 15.6 20.3 29.7 64
Total 33.2 14.3 20.7 12.7 12.3 6.8 440
Chi-squared(25) = 232

4.4.7 Work experience

Table 4.26 shows the dramatic extent to which women without any employment
experience were concentrated among those whose partners had no such experience.
While only 16 per cent of women in workless households had no experience of
employment, the level was 58 per cent among those partnered with men who had never
worked.

Table 4.26: Dependence between work experience of partners

(row percentages)
Female:
Male: Ever worked Never worked Total
Ever worked 85.9 14.1 5369
Never worked 41.8 58.2 220
Total 84.2 15.8 5589
Chi-squared(1) = 308

4.4.8 Length of time since last job

When considering the length of time since last employed, there is once again evidence
of dependence between the partners. However, in this case, there appears to be a
tendency for men’s employment to be a more recent experience than that of their
partners. All the entries to the left of the leading diagonal are smaller than the

30
 Characteristics of couples on benefits

corresponding entries in the final row of the table indicating that women with a spell
since their last period of employment shorter than that of their partner were
disproportionately absent.

Table 4.27: Dependence between how long since last worked

(row percentages)
Female:
Male: <1 year 1-2 yr 3-5 yr 6-10 yr 10+ yr Total
Less than a year 31.0 20.7 15.7 16.0 16.6 830
1 to 2 years 8.2 24.7 22.3 20.0 24.8 998
3 to 5 years 6.1 10.6 27.7 23.5 32.1 956
6 to 10 years 3.2 7.5 16.3 32.2 40.8 1122
More than 10 years 3.0 7.1 8.4 14.3 67.3 765
Total 9.8 14.1 18.5 22.0 35.7 4671
Chi-squared(16) = 1194

4.4.9 Manual vs. non-manual employment

Table 4.28 considers the tendency for manual workers (as were) to be partnered with
other manual workers. Again, there is strong evidence of dependence with partners in a
couple likely to have been either both manual or both non-manual.

Table 4.28: Dependence between whether last job was manual

(row percentages)
Female:
Male: Manual Non-manual Forces Total
Manual 65.2 34.8 - 1691
Non-manual 33.7 66.3 - 517
Armed forces 87.5 12.5 - 16
Total 58.1 42.0 - 2224
Chi-squared(4) = 168

31
NDP: Characteristics and labour market transitions
 Changes over time

5. Changes over time

In this chapter, attention turns to changes in worklessness over time. The
principal focus is on changes over the period for which each household is
observed in the data (a maximum of one year). This is done to assess both the
changing profile of the sample over the year and to focus more closely on
individual transitions between economic states. However, before this is done,
changes over calendar time are considered. The chapter ends with an analysis
of changes in benefit claiming.

5.1 Changes in the unemployment/inactivity mix, 1996-2000

In this section, the focus is on the change over time in the proportion of
worklessness accounted for by unemployment and inactivity. As in previous
chapters, inactivity is further divided into type 1 and 2 inactivity for men and
women and for women the additional dimension of the reason for inactivity is
included. This analysis is based on the full sample and considers only the
quarter at which the workless couple was first observed.

Figure 5.1 considers the economic status of men in workless households. For
each quarter, the percentages of men unemployed, type 1 inactive and type 2
inactive at the point of the couple first being observed are plotted. The trends
show an increase in the proportion of worklessness due to type 2 inactivity and a
fall in the proportion of worklessness due to unemployment over the period winter
1996 to winter 2000. Type 2 inactivity increases from just over 30 to about 45
per cent while unemployment falls from over 40 to about 27 per cent. Type 1
inactivity shows considerable fluctuation but the value at the beginning of the
observation period is close to that at the end.

33
NDP: Characteristics and labour market transitions

Figure 5.1: The relative proportion of unemployment and inactivity over

time for men in workless couples

50
45
40
35
Percentage

30 Unemployed
25 Inactive type 1
20 Inactive type 2
15
10
5
0
/4
/1
/2
/3
/4
/1
/2
/3
/4
/1
/2
/3
/4
/1
/2
/3
/4
96
97
97
97
97
98
98
98
98
99
99
99
99
00
00
00
00
Year and quarter

For women, the patterns are different. Most of the five states observed for
women were relatively stable over time. A small fall in the proportion of
unemployment from slightly above ten to just over five per cent and of type 1
inactivity due to family reasons from 20 to about 13 per cent can be observed.
The only marked trend is the increase in the proportion of workless women being
type 2 inactive due to other than family reasons. The proportion of this group
increased from 20 per cent in Winter 1996 to nearly 35 per cent in Winter 2000.

Figure 5.2: The relative proportion of unemployment and inactivity over

time for women in workless couples

45
40
35 Unemployed
Percentage

30
Inac. 1, family
25
Inac. 1, other
20
15 Inac. 2, family
10 Inac. 2, other
5
0
/4
/1
/2
/3
/4
/1
/2
/3
/4
/1
/2
/3
/4
/1
/2
/3
/4
96
97
97
97
97
98
98
98
98
99
99
99
99
00
00
00
00

Year and quarter

34
 Changes over time

Given these changes in the proportion of different types of worklessness, it is

instructive to examine the absolute numbers of men and women in different types
of worklessness. Figure 5.3 shows this for men. In absolute numbers all types
of worklessness were falling over time. However, unemployment fell at a faster
rate which explains its reduced proportion. The proportionate increase in type 2
inactivity appears to be caused by its relative stability compared to falls in the
numbers of unemployed and type 1 inactive.

Figure 5.3: The numbers in unemployment and inactivity over time for
men in workless couples.

200
180
160
140 Unemployed
Frequency

120
Inactive, type 1
100
80 Inactive, type 2
60
40
20
0
/4

/1

/2

/3

/4

/1

/2

/3

/4

/1

/2

/3

/4

/1

/2

/3

/4
96

97

97

97

97

98

98

98

98

99

99

99

99

00

00

00

00
Year and quarter

Figure 5.4: The numbers in unemployment and inactivity over time for
women in workless couples.

180
160
140 Unemployed
120 Inac. 1, family
Frequency

100 Iinac. 1, other

80 Inac. 2, family
60 Inac. 2, other
40
20
0
96/4 97/1 97/2 97/3 97/4 98/1 98/2 98/3 98/4 99/1 99/2 99/3 99/4 00/1 00/2 00/3 00/4
Year and quarter

Figure 5.4 shows the absolute number of women in different economic states
over the observation period. Because couples were analysed, the total number

35
NDP: Characteristics and labour market transitions

of workless men and women is the same in each period. Consequently, the
noted decline in worklessness among men over the period 1996-2000 is equally
true for women. Inactivity of both types due to other reasons was relatively
stable, while type 1 and 2 inactivity due to family reasons dropped over the
observation period. There was also a drop in the number of unemployed women.
This suggests that, for women, inactivity due to family reasons may be a less
rigid form of inactivity than that due to other reasons.

5.2 Changes over the observation year

The preceding evidence was based on the point in time at which each couple
was first observed. The longitudinal nature of the LFS permits transitions
between economic states to be considered in more detail. Specifically, one can
inspect changes over those periods for which individuals were observed. As
noted earlier, couples were observed at up to five points spanning a maximum
period of one year. In this section, and the remainder of this report, changes
over these five observation points are considered.

5.2.1 The changing economic status of men and women over the
observation year

Table 5.1: The changing economic status of men and women (balanced
panel)
Months since first observed jointly non-employed:
0 3 6 9 12
Men:
Working 0.0 5.9 9.2 11.5 12.6
Unemployed 30.2 25.5 22.4 20.1 18.8
Inactive, type 1 28.3 30.5 30.0 30.2 30.7
Inactive, type 2 41.5 38.1 38.4 38.2 38.0
Base 2360 2360 2347 2342 2330

Women:
Working 0.0 3.1 4.9 6.4 7.3
Unemployed 8.3 7.0 6.1 5.6 5.9
Inactive, type 1 26.4 27.0 26.9 26.3 25.7
family reasons 13.5 14.0 13.6 13.7 12.6
other reasons 12.8 13.0 13.2 12.6 13.0
Inactive, type 2 65.4 63.0 62.1 61.7 61.1
family reasons 39.5 37.1 35.9 35.5 35.2
other reasons 25.9 25.9 26.3 26.2 25.9
Base 2360 2360 2354 2348 2342
Due to rounding percentages might not add up to 100.

36
 Changes over time

The analysis that follows is based on the sample of full participants (the balanced
panel). This has the advantage that the problems of changing sample
composition and differential non-response can be ignored. Table 5.1 shows that
in approximately 13 per cent of cases the man in a workless couple will have
found work8 within the year covered by the survey. This is twice the level for
women. Men’s movement into employment appears (although see below) to be
due mainly to the reduction in unemployment9 which falls from 30 per cent to
below 20 per cent over the year while the level of the two types of inactivity
remains relatively stable. For women, transitions were much fewer in number
and the proportion in each type of worklessness was relatively stable. The seven
per cent increase in employment was accounted for mainly by a drop in type 2
inactivity due to family reasons (four per cent), unemployment (two per cent) and
type 1 inactivity due to family reasons (one per cent).

It is important to be careful when interpreting these changes. For example, the

rise in the proportion of working women appears to be due mainly to a reduction
in type 2 inactivity, more specifically, among those who are type 2 inactive for
family reasons. This would seem to have implications for regarding type 2
inactivity as the most distant status from the labour market. However, it is not
possible to make such assertions since it is possible that there is a cascade
effect whereby those in type 2 inactivity move to type 1 inactivity, those in type 1
inactivity move into unemployment and those in unemployment move into work.
In this scenario, the snapshot proportions of the sample in any given economic
state will provide no information on flows into and out of that status. To illustrate,
if the same number of people move from type 2 to type 1 inactivity as move from
type 1 inactivity to unemployment, no change in type 1 inactivity will be evident.
To investigate such movements requires focusing explicitly on transitions (the
focus of section 5.3).

For clarity, the results from Table 5.1 are also represented in Figures 5.5 and 5.6.
They confirm the picture that inactivity was more prevalent among women in
workless couples and that the composition of states was more stable for women.

8
The category ‘working’ combines employees and self-employed.
9
From this point onwards unemployment is not only ILO unemployment but also includes the
small number of cases reporting to be on a government scheme or working (unpaid) in the family
business.

37
NDP: Characteristics and labour market transitions

Figure 5.5: The changing economic status of men

100%

80%
Inactive type 2
60% Inactive type 1
40% Unemployed
Working
20%

0%
0 3 6 9 12
Months since first observed

Figure 5.6: The changing economic status of women

100%

80% Inac. 2, other

Inac. 2, family
60% Inac. 1, other
40% Inac. 1, family
Unemployed
20% Working
0%
0 3 6 9 12
Months since first observed

5.2.2 Changes in economic status for the full sample

Focusing on the balanced panel means that information is lost through discarding
observations. In fact, more than half the couples in the sample are discarded by
this process. Given this large reduction, it is useful to consider how the observed
transitions differ when working with the full sample. This provides an indication
of the likely bias introduced through the sample reduction. While, as noted
earlier, attrition in this sample may be for a variety of reasons, if the tendency not
to respond were correlated with the outcome of interest (employment status, in
this case) problems of inference may arise. For example, if those more likely to

38
 Changes over time

find employment were also more likely to disappear from the sample, those
individuals remaining would create an overly pessimistic impression of
movements into work.

Table 5.2 shows that the proportions of the different labour market states 12
months after first being observed was identical for the two samples. This is so by
construction as all those observed after twelve months are the ones forming the
balanced panel. Differences occur at earlier points. For men and women there
is only a small difference in the proportion in employment. There was a higher
proportion of unemployed in the unbalanced panel which was compensated by a
smaller proportion of inactives – more or less evenly distributed over all types of
inactivity.

Overall, the effect of restricting attention to the balanced panel is that inactivity as
a proportion of all worklessness is overstated in the earlier quarters of the
observation period. This is true for both men and women, although it is more
marked among men. Hence, being unemployed as opposed to inactive appears
to be associated with failing to respond to subsequent interviews. This is also
reflected in the fact that 45 per cent of couples where the male was inactive
when first observed responded in all five waves but the same was true of only 36
per cent of couples with an unemployed male in the first period. This may
indicate higher non-response among those in work and a greater tendency for
unemployed people rather than inactive people to find work. In view of this, it is
plausible to view the increased prevalence of work revealed by the balanced
panel as being a lower bound on the true proportion, since those entering jobs
may be lost to the sample.

39
NDP: Characteristics and labour market transitions

Table 5.2: The changing economic status of men and women

(unbalanced panel)
Months since first observed jointly non-employed:
0 3 6 9 12
Men:
Working 0.0 5.8 9.7 11.9 12.6
Unemployed 35.1 29.2 24.5 21.4 18.8
Inactive type 1 26.4 28.6 28.9 29.5 30.7
Inactive type 2 38.5 36.5 36.9 37.2 38.0
Base 5600 4403 3563 2930 2330

Women:
Working 0.0 3.2 4.8 6.1 7.3
Unemployed 10.1 8.3 7.2 6.4 5.9
Inactive type 1 25.7 27.1 26.5 26.2 25.7
Family reasons 13.6 14.5 14.0 13.9 12.6
Other reasons 12.1 12.6 12.5 12.2 13.0
Inactive type 2 64.2 61.4 61.6 61.4 61.1
Family reasons 38.8 36.8 36.4 35.6 35.2
Other reasons 25.4 24.7 25.2 25.8 25.9
Base 5600 4405 3575 2940 2342
Due to rounding percentages might not add up to 100.

5.2.3 The changing joint economic status of partners over the observation
year

In Table 5.3, the economic status of both partners is considered simultaneously.

In this way, it is possible to gain an insight into the existence and persistence of
worklessness at the level of the partnership.

Over the five observation points, 17 per cent of couples moved from the position
of worklessness to having at least one partner in work. This is a positive finding
in terms of addressing the problem of worklessness since it does not appear that
moves into employment among men and women were concentrated within the
same household, at least over the period of a year. In fact, only 3 per cent of
workless couples had become dual-earner couples over the period considered.
In about 5 per cent of cases, only the woman found work and in the remaining 10
per cent it was only the man. The most common combination among couples
finding work was for the man to be employed and the women to be type 2
inactive.

40
 Changes over time

Table 5.3: Changes in joint economic status of partners

Months since first observed:
0 3 6 9 12
Partners’ economic status:
male working, female working 0.7 1.5 2.3 2.7
male working, female unemployed 0.7 1.0 1.0 1.0
male working, female inactive 1 1.3 2.1 2.8 2.9
male working, female inactive 2 3.3 4.5 5.3 5.9
male unemployed, female working 1.0 1.2 1.6 1.5
male unemployed, female unemployed 5.3 4.2 3.6 2.8 3.2
male unemployed, female inactive 1 8.6 7.7 6.3 5.4 4.5
male unemployed, female inactive 2 16.3 12.6 11.3 10.3 9.8
male inactive 1, female working 0.9 1.1 1.2 1.5
male inactive 1., female unemployed 1.5 1.2 0.9 0.9 0.7
male inactive 1, female inactive 1 12.5 13.2 12.6 13.0 13.4
male inactive 1, female inactive 2 14.3 15.3 15.5 15.2 15.0
male inactive 2, female working 0.6 1.1 1.3 1.5
male inactive 2, female unemployed 1.4 0.9 0.6 0.8 1.0
male inactive 2, female inactive 1 5.2 4.9 5.9 5.1 5.0
male inactive 2, female inactive 2 34.8 31.8 30.9 31.0 30.6
Total 2360 2360 2343 2332 2317
Due to rounding percentages might not add up to 100.

The combination of an unemployed man partnered with a type 2 inactive woman

was the combination which declined most in size over the period. Couples with
both partners type 2 inactive accounted for the highest proportion among all
couples during the full observation period. This proportion dropped slightly from
35 per cent at the first period to 31 per cent after one year. Two-thirds of couples
were inactive (either type) at the onset of the observation period. Their proportion
fell only slightly over the year. This indicates that inactivity is a relatively stable
state and it seems of special interest to investigate whether differences by the
type of inactivity can be detected.

5.3 Transitions between individual economic states

The preceding section considered changing economic status over time. This is
not the same as considering individual transitions between states since these
summary levels of employment status provide no information on the extent to
which the changes are concentrated among particular individuals. Put another
way, does the growing proportion observed as being employed, for example,
reflect individuals finding work in the early stages remaining in work and having
their numbers boosted in subsequent years by additional long-term job entrants,
or does it reflect a different (but growing) group of people being observed in
employment at each stage? This is an important distinction since the policy

41
NDP: Characteristics and labour market transitions

implications of stable employment are different from those associated with

employment ‘churning’. The answer is not to be found by considering changing
profiles of the sample in terms of economic status but rather by considering
individual transitions.

5.3.1 The proportion of the sample with experience of each employment

status

Table 5.4 provides some indication of the relative importance of the economic
states over the full observation period by showing the proportion of the sample
having a particular employment status in at least one of the five observation
points. These percentages do not sum to 100 since individuals may experience
more than one status.

The number of men experiencing employment was higher than that for women.
This finding is consistent with the observations made in the previous sections
where changes in status over time were considered. Furthermore, for both sexes,
the proportion experiencing employment in at least one of the five waves was
higher than the proportion in employment at wave five indicating at least some
short-term employment. It was much more common for men than women to
experience unemployment. However, it is clear that inactivity was the most
prominent status. Over the observation year, both types of inactivity were
experienced by more than half the men. A similar proportion of women
experienced type 1 inactivity but more than 80 per cent of women were type 2
inactive at some point. This was mainly for family reasons.

Table 5.4: Proportion of sample with experience of each employment

status
%
Men:
Working 15.8
Unemployed 36.1
Inactive, would like to work 52.3
Inactive, not wanting work 57.0

Women:
Working 9.0
Unemployed 14.1
Inactive, would like to work 49.6
Family reasons 27.0
Other reasons 28.0
Inactive, not wanting work 81.8
Family reasons 51.3
Other reasons 38.6

42
 Changes over time

5.3.2 Transitions between economic states

Tables 5.5a and 5.5b consider the transitions between economic states for men
and women, respectively. This makes full use of the longitudinal nature of the
data and summarises changes in employment status for all couples over all time
periods. Since only fully participating couples are considered, four ‘transitions’
are observed for each individual. Some of the ‘transitions’ are actual transitions
from one state to another while the rest are non-transitions: people being
observed in the same state as in the preceding period. The results in Table 5.5a
and the following tables in this chapter simply show the percentage who change
from one state (original) to another (destination). The destination can be the
same as the original status. In this case the amount represents the percentage
remaining stable – this is given in bold and is along the main diagonal in the
table. The entries for each row sum to 100 per cent. Hence, it is possible to see
the level of stability of the different economic states. This represents an advance
on the results presented earlier for changes in employment status over time
since these could not provide any indication of whether the general trend towards
employment among workless households was explained by a cumulative move
into long-term jobs or a growing number of short-term jobs distributed across the
sample.

Table 5.5a: Transitions between economic states – Men

(row percentages)
Destination economic status:
Working Unemployed Inactive type 1 Inactive type 2
Original economic status:
Working 84.6 12.4 1.6 1.5
Unemployed 13.4 76.9 6.7 3.0
Inactive, type 1 1.9 4.8 75.1 18.1
Inactive, type 2 0.8 1.4 16.0 81.8
Due to rounding percentages might not add up to 100.

Table 5.5a shows that:

• 85 per cent of employed men were still employed when next observed.
Hence, this appears to be quite a stable economic status. Of those who left
employment, the majority became unemployed and relatively few became
inactive.
• Over three-quarters of the unemployed remained unemployed in the next
observation period. Of those who left, employment was the main destination,
while type 1 inactivity claimed twice as many leavers as type 2 inactivity.

43
NDP: Characteristics and labour market transitions

• The two types of inactivity were also relatively stable: 75 per cent of those
who were type 1 inactive stayed in that status for the next period while this
number was 82 per cent for type 2 inactivity.
• Movements out of one type of inactivity were predominantly into the other
type of inactivity. Hence, inactivity as a whole was characterised by few exits.

Two remarks are in place. First, the two extreme states are the most stable.
This provides some support for the interpretation of the different types of
worklessness as reflecting distance to the labour market. The fact that most
exits from an initial state were to a state that can be considered neighbouring
lends further support to this interpretation. Second, the majority of transitions
occur between working and unemployment on the one hand and between the
two types of inactivity on the other hand. Thus, the distance between
unemployment and type 1 inactivity seems to be greater than between working
and unemployment or between the two types of inactivity. More simply, inactivity
is considerably further from working than is unemployment.

Table 5.5b presents the results for women with the additional distinction between
inactivity of either type due to family or due to other reasons. Given those
additional statuses it is not surprising that the picture is not as clear-cut as it was
for men:

• It is still true that the two extreme states (employment and type 2 inactivity)
were the most stable. However, there was more movement among the
intermediate states of unemployment and type 1 inactivity. There was also
more movement from one state to a non-neighbouring state.
• Those women who worked were, if anything, more likely than men to still be
working when next observed. However, those exiting employment were more
likely to become inactive than unemployed. In fact, women leaving to
inactivity were fairly evenly split between types of inactivity. This suggests
they may be choosing to leave work since they were not trying to re-enter
employment.
• As noted, unemployment was less stable than for men. Furthermore, while
only a tenth of unemployed men became inactive, the corresponding
proportion for women was a quarter. Type 1 inactivity accounted for twice as
many leavers as type 2.

It is when considering female exits from inactivity that the benefit of subdividing
type of inactivity becomes apparent:

• exits from one type of inactivity were mainly to the other type of inactivity.
• the reason for inactivity is persistent. That is, most of those moving between
types of inactivity did not change the reason (family, other) for their inactivity.

Possibly the most important finding from these results is the degree of movement
between types of inactivity. While inactivity as a whole was very rigid with few

44
 Changes over time

exits, within inactivity there were a considerable number of moves both towards
and away from the labour market that are of potential policy interest. To
investigate these fully requires an econometric approach. However, in the case
of women, the sub-divisions within type of inactivity provide some clues to the
process. The fact that exits from unemployment were mainly to type 1 inactivity
and then moves from type 1 to type 2 inactivity appear not to be prompted by a
change in the reason for inactivity is suggestive of a process whereby those
women who were looking for work became inactive and, with time, their desire to
return to work disappeared. Such a model of labour market disenfranchisement
highlights the potential value in concentrating on those who are type 1 inactive to
prevent them losing all interest in work. It is noteworthy that the chances of
moving from type 1 to type 2 inactivity are twice those of moving in the opposite
direction.

Table 5.5b: Transitions between economic states – Women

(row percentages)
Destination economic status:
Working Unemp. Inactive type 1 Inactive type 2
family other family other
Original economic status:

Working 86.7 4.1 2.1 2.4 2.1 2.7

Unemployed 12.8 62.2 7.7 9.2 4.4 3.6
Inactive 1, family reasons 2.9 5.0 65.5 4.1 19.4 3.1
Inactive 1, other reasons 3.1 4.5 5.2 63.7 4.7 18.7
Inactive 2, family reasons 1.2 1.1 7.8 1.8 85.0 3.1
Inactive 2, other reasons 0.7 0.6 1.4 10.8 3.2 83.5
Due to rounding percentages might not add up to 100.

5.3.3 The proportion of the sample with experience of each joint

employment status

In this and the following section the transition analysis is repeated for both
partners simultaneously. Table 5.6 shows the relative importance of each joint
economic status over the full observation period. Half of all couples had some
experience of being jointly type 2 inactive. This was by far the largest category.
The next most common joint status was the male partner experiencing type 1
inactivity and the female partner type 2 inactivity. Only 4 per cent of couples
experienced a period of being dual-earners. This number is higher than the
percentage of couples jointly in employment at wave 5 (see Table 5.3) indicating
that part of this joint employment was short-term.

45
NDP: Characteristics and labour market transitions

Table 5.6: Proportion of sample with experience of each joint

employment status
percentage with this status at some point
Partners’ economic status:
male working, female working 3.7
male working, female unemployed 2.4
male working, female inactive 1 5.0
male working, female inactive 2 8.8
male unemp., female working 2.8
male unemp., female unemployed 9.1
male unemp., female inactive 1 16.1
male unemp., female inactive 2 23.4
male inactive 1, female working 2.5
male inactive 1, female unemployed 3.2
male inactive 1, female inactive 1 28.3
male inactive 1, female inactive 2 34.1
male inactive 2, female working 2.3
male inactive 2, female unemployed 2.8
male inactive 2, female inactive 1 14.2
male inactive 2, female inactive 1 49.6

5.3.4 Transitions between joint economic states

In Table 5.7, the transitions between economic states at the level of the couple
are considered. However, not all possible transitions were observed and some
cells are therefore left empty. In contrast, an entry of zero means that the
proportion of transitions in this cell was less than 0.5 per cent. The table is read
in the same way as previous transition tables. Each row contains information on
one specific status combination and each cell gives the proportion of transitions
to the states given in the column headings. ‘Non’ transitions, i.e. couples being
observed in the same state in the next period, are given in bold and form the
main diagonal of the table. Before commenting on specific states and transitions
some key patterns are noted:

• as a general rule, the patterns found in the individual level analysis are
confirmed here. Employment and type 2 inactivity were the most stable
states and any combination of these two within a couple would result in over
70 per cent stability from one quarter to the next
• the partner who was in the less stable state was more likely to cause a
transition. For example in the case of a working male and a type 1 inactive
female, in 84 per cent of the cases the status of the male partner was stable
• where there was a transition, it was the woman who changed in over 60 per
cent of cases
• the transitions were more likely to be to a neighbouring state

46
 Changes over time

• transitions between work and unemployment on the one hand and between
the two types of inactivity on the other hand were more likely than other
transitions
• only in a few cases did both partners change their status from one period to
the next. This happened in less than 7 per cent of all transitions.

Table 5.7: Transitions between joint economic states

(row percentages)
Destination economic status: male partner – female partner)

inact. 1 – inact. 1
unemp. – unemp.

inact. 2 – unemp.
unemp. – inact. 1
unemp. – inact. 2

inact. 1 – inact. 2

inact. 2 – inact. 1
inact. 2 – inact. 2
inact. 1– unemp.
unemp. – work
work – unemp.
work – inact. 1
work – inact. 2

inact. 1 – work

inact. 2 – work

observations
work – work

Number of
Original economic
status: male partner
– female partner
Work – work 74 6 5 3 8 1 1 2 1 169
Work – unemployed 19 44 13 11 11 2 87
Work – inactive 1 9 4 57 14 1 3 6 4 2 1 209
Work – inactive 2 1 1 7 74 2 11 1 0 2 444
Unemployed – work 15 1 61 5 1 4 10 1 2 122
Unemp. – unemp. 4 7 3 2 6 55 12 5 1 2 2 1 0 0 1 449
Unemp. – inactive 1 2 1 8 3 3 8 52 14 0 0 5 3 0 1 1 760
Unemp. – inactive 2 1 0 1 10 1 3 9 66 0 0 1 5 0 0 0 4 1413
Inactive 1 – work 3 10 61 3 4 1 15 3 107
Inactive 1 – unemp. 2 1 7 3 11 42 12 5 8 5 5 125
Inactive 1 – inactive 1 0 0 1 0 0 1 3 1 1 1 62 14 0 0 7 8 1515
Inactive 1 – inactive 2 0 0 1 0 1 4 1 0 12 61 0 0 1 19 1761
Inactive 2 – work 3 13 1 71 1 3 7 104
Inactive 2 – unemp. 2 5 2 5 10 6 15 43 8 3 108
Inactive 2 – inactive 1 0 0 0 1 1 1 0 19 4 1 2 47 22 611
Inactive 2 – inactive 2 0 0 1 0 0 1 0 5 10 0 0 4 79 3728
Total 2 1 2 5 1 3 6 11 1 1 13 15 1 1 5 31
Due to rounding percentages might not add up to 100.

It is helpful to put these general rules into a more concrete context. Dual-earner
couples were quite stable. Although, as shown in the previous table, only a small
proportion of couples comprised two earners at any point in the observation year,
nearly three-quarters of those who did attain this status retained it in the next
period. The main cause of change was either partner becoming unemployed.
This accounted for 14 per cent of transitions. Inactivity among women was also
important, with eight per cent moving to some type of inactivity. In nearly all
cases, dual-earner couples in one period had at least one earner in the
subsequent period. Single-earner couples were the next most likely group to be

47
NDP: Characteristics and labour market transitions

non-workless at the subsequent point of observation. In such couples, there was

an earner at the time of next observation in 83 – 89 per cent of cases. In
summary, where the couple was non-workless, there was a very strong
possibility of also being non-workless in the next time period.

For dual unemployed couples, the chances of being in a non-workless household

when next observed were 23 per cent. For a mix of unemployment and inactivity,
the chances were lower and the move away from worklessness was generally
thanks to the unemployed partner. A slightly intriguing exception was the
combination of a type 2 inactive male and an unemployed female. The chances
of such a couple exiting worklessness (22 per cent) were similar to dual
unemployed couples.

Where both partners were inactive, the chances of being non-workless when
next observed were negligible, never more than two per cent for any combination
of type of inactivity. It is instructive to consider inactivity in more detail. There is
some evidence that transitions out of type 1 inactivity were related to partner’s
status. For men, ten per cent of those partnered with working or unemployed
women would become unemployed. This is twice the level for those with inactive
partners. For type 1 inactive women, ten per cent of those with working partners
and five per cent of those with unemployed partners would move into
employment. The level for those with inactive partners was one to two per cent.
A similar pattern was found when considering movements into unemployment.
Seven and nine per cent respectively of those with working and unemployed
partners would themselves become unemployed, compared with only two per
cent of those with inactive partners.

As noted, moves away from worklessness were rare for jointly inactive couples.
However, there were transitions between the two types of inactivity and these
appear to be influenced by the status of the other partner. To illustrate, of those
type 1 inactive men with type 1 inactive partners, 15 per cent moved to type 2.
This compares to 20 per cent for those with type 2 partners. Similarly, 24 per
cent of type 2 men with type 1 partners move to type 1, compared with only 15
per cent of those with type 2 partners. For women, the results are qualitatively
similar, although a general comment is that there were more moves into type 2
inactivity and fewer moves away from it than for men. For a type 1 inactive
woman, 23 per cent of those with a type 1 inactive partner would move to type 2,
compared with 27 per cent for those with a type 2 inactive partner. For a type 2
inactive woman, 14 per cent of those with a type 1 partner would become type 1
inactive, compared with only nine per cent of those with a type 2 inactive partner.

While the description of such transitions between types of inactivity quickly

becomes cumbersome, the important point is that the movements between types
of inactivity appear related to the partner’s type of inactivity. Where one partner
is further away from the labour market, the other partner appears drawn in the
same direction.

48
 Changes over time

5.4 Changes and transitions in benefits claimed

In this section, a similar analysis of changes and transitions is carried out for
benefits. First, the change in benefit status is analysed for each partner
separately and for the couple, followed by transition analyses similar to the ones
carried out on labour market status in the previous section.

5.4.1 The changing types of benefits claimed over the observation year

Table 5.8 shows the changing composition of the sample in terms of benefit
claiming status. The top half of the table contains information for men and the
bottom half for women. There was a marked increase over the year in the
proportion of men not claiming any benefits. This was compensated mainly by a
decrease in JSA claimants, although there was also a notable move away from
IB. Movements in other benefits were much smaller. The situation was more
stable for women. However, most of the women were not claiming benefits to
start with and therefore one would expect the situation to change more for men.
This gender difference reflects the specific design of the benefits. JSA and IS,
which made up the largest proportion of benefits claimed, are benefits based on
a couple rather than the individual whereas IB, SDA and ICA are designed as
individual benefits.

Table 5.8: The changing type of benefits claimed for men and women
(balanced panel)
Months since first observed jointly non-employed:
0 3 6 9 12
Men:
No benefit 9.9 17.4 21.3 22.2 24.5
JSA only 28.6 22.2 17.8 16.4 14.3
IS only 9.4 9.9 10.1 10.9 10.7
IB only 31.1 28.3 27.7 27.3 26.1
IS and IB 14.4 14.8 15.3 15.3 15.5
other combination 6.7 7.5 7.7 7.9 8.9

Women:
No benefit 73.2 72.6 72.5 72.2 71.9
JSA only 2.4 1.6 1.3 1.1 1.3
IS only 3.2 3.8 3.6 3.6 3.7
IB only 11.6 10.6 10.8 10.8 10.6
IS and IB 1.9 2.7 2.3 2.2 2.1
other combination 7.8 8.7 9.5 10.2 10.4
Base 2360 2360 2360 2360 2360
Due to rounding percentages might not add up to 100.

49
NDP: Characteristics and labour market transitions

Table 5.9 contains similar information for the couple. Again, there was a marked
increase over the observation year of couples not claiming any benefits. By
construction, all couples were claiming benefits when first observed. After just
three months, eight per cent of couples were no longer claiming any benefits.
After a year, this proportion had doubled.

Table 5.9: The couple’s changing benefit claiming status (balanced

panel)
Months since first observed jointly non-employed:
0 3 6 9 12

No benefits claimed 0 8.3 12.1 13.7 15.7

Only male claiming 73.2 64.4 60.4 58.5 56.2
Only female claiming 9.9 9.1 9.2 8.5 8.8
Both claiming 17.0 18.3 18.3 19.3 19.3
Due to rounding percentages might not add up to 100.

As with the analysis of the change in labour market states it is interesting to see
to what extent these changes were stable as opposed to short term. The next
two sections can give an indication of this.

5.4.2 The proportion of the sample with experience of the different benefit
types

Table 5.10: Proportion of sample with experience of different benefit

claiming patterns
%
Men
No benefits claimed 33.1
JSA only 31.0
IS only 18.8
IB only 36.9
IS and IB 22.8
Other combinations 13.4

Women
No benefits claimed 80.3
JSA only 3.7
IS only 7.2
IB only 14.2
IS and IB 3.7
Other combinations 13.9

50
 Changes over time

Table 5.10 exhibits the proportion of men (top half) and women (bottom half) who
had experience of different benefit claiming states. It is important to note that a
third of men had at least one period where they did not claim any benefit. This is
higher than the quarter who were not claiming after 12 months (see Table 5.9).
Thus, some of the movement away from claiming benefits must be short-term.
Also the proportion having claimed other benefits is higher than at any one
period, further supporting the suggestion that some of the movements between
benefits were temporary. The difference is smallest for JSA indicating that
movements away from JSA were relatively permanent. IB also had a relatively
small difference while the difference was greater for IS. This might be due to the
fact that claiming IB is a relatively long process – the incapacity to work has to be
assessed – while it is easier and quicker to claim IS. A more formal assessment
of the stability of benefit states will be conducted in the transition tables below.
The patterns observed for women were similar.

Table 5.11 presents similar evidence for couples. Nearly a quarter of the sample
was not claiming any benefits for at least one period. The situation where the
female was the sole claimant in a couple was the least common state with only
15 per cent of couples having experienced it.

Table 5.11: Proportion of couples with experience of different benefit

claiming patterns
%
No benefits claimed 22.3
Only male claiming 77.3
Only female claiming 14.5
Both claiming 27.4

5.4.3 Transitions between types of benefits claimed

In Tables 5.12a and 5.12b, transitions between different benefit claiming states
are reported for men and women respectively. The first thing to note is that the
different benefit combinations were relatively stable: 76 to 88 per cent of men
were observed claiming the same benefits as in the period before.10 If men
moved from not claiming any benefits back on to benefits it was most likely to be
JSA. This is consistent with the observation that most transitions from work back
to worklessness were to unemployment rather than to inactivity. Similarly, most
men moving away from JSA were not claiming benefits in the next period,
presumably working. A much smaller percentage moved from JSA to other
benefits.

10
As the category ‘others’ is made up of a number of different combinations of benefits, there
could have been changes in the exact combination of benefits claimed.

51
NDP: Characteristics and labour market transitions

Table 5.12a: Transitions between types of benefits claimed – Men

(row percentages)
Destination benefits claimed
No benefit JSA only IS only IB only IS and IB Others
Original economic status:

No benefit 84.7 5.5 3.8 2.6 0.6 2.9

JSA only 18.8 75.8 2.8 0.9 0.7 1.1
IS only 7.8 2.7 77.9 2.3 5.4 4.0
IB only 3.3 0.6 1.0 88.4 5.0 1.7
IS and IB 2.6 0.4 4.8 5.4 85.3 1.6
Other combinations 3.7 1.1 3.7 5.4 3.4 82.6
Due to rounding percentages might not add up to 100.

The picture was similar and in some respects even more pronounced for women.
Nearly all women not claiming benefit at one point in time were also not claiming
in the next period. Among those claiming JSA, only 56 per cent were claiming in
the next period as well. However, it is important to bear in mind the distinction
between household and individual benefits. Since men tend to be the claimant
partner for household benefits (JSA, IS), the picture when considering these
benefits may be misleading.

Table 5.12b: Transitions between types of benefits claimed – Women

(row percentages)
Destination benefits claimed
No benefit JSA only IS only IB only IS and IB Others
Original economic status:

No benefit 96.1 0.5 0.9 0.6 0.1 1.8

JSA only 40.7 56.0 3.3 0 0 0
IS only 16.9 1.2 72.7 1.5 3.6 4.2
IB only 4.8 0.3 1.4 89.3 2.2 1.9
IS and IB 5.2 0 5.6 6.6 80.3 2.4
Other combinations 6.7 0 1.6 2.8 0.7 88.2
Due to rounding percentages might not add up to 100.

Finally, Table 5.13 contains transitions in the benefit claiming state of couples.
The stability of the ‘no benefits claimed’ category is quite remarkable given that
this category is made up solely of couples moving away from claiming benefits.
Thus, three-quarters of couples who moved off benefits during the observation
period were also without benefits the next period. There was very little
movement between the male and the female partner being the sole claimant but

52
 Changes over time

considerable movement from female only to both claiming. Couples where only
one partner was claiming were more than three times as likely to move off
benefits compared to couples where both partners were claiming.

Table 5.13: Transitions between couples’ benefits claiming states

(row percentages)
Destination benefits claimed
No benefits Only male Only female Both claiming
Original economic status:
No benefits claimed 77.1 17.3 4.1 1.5
Only male claiming 7.5 88.9 0.4 3.2
Only female claiming 7.4 1.9 80.5 10.3
Both claiming 2.2 6.9 5.0 85.9
Due to rounding percentages might not add up to 100.

53
NDP: Characteristics and labour market transitions
 Modelling transitions

6. Modelling transitions
In this chapter the modelling results are presented. In section 6.1 the models are
described. Section 6.2 gives the results for the analysis of couple-level exits
from worklessness. In Section 6.3, the results for individuals are presented. This
last section also includes a consideration of the extent to which the transitions of
partners within a couple are related.

6.1 The models to be estimated

The advantage of using econometric models rather than simple descriptive
analyses is that the extent to which the characteristics of couples or individuals
are linked to their probability of exiting a given labour market state can be
assessed. This contributes to the understanding of movements between states.

It is appropriate to use duration analysis to model transitions. This allows an

explicit focus on the time it took to exit from a particular state. As a result, not
only the probability of exit but also changes in this probability with increasing
spell length can be shown. Apart from this, the results of the estimations also
show which characteristics were most associated with transitions between states.
Some details of the analytical approach are provided in the methodological
appendix, which also considers the interpretation of the results. The aim of this
section is to set out the models to be estimated.

The descriptive analysis in the previous chapters has revealed the importance of
the initial type of worklessness. These different types were interpreted as
measuring how far away an individual or couple is from the labour market.
Unemployment was seen as being closest to the labour market, followed by type
1 inactivity. The furthest from the labour market are individuals who are type 2
inactive. The models in this chapter also make use of this distance concept. The
distance to the labour market proves to be strongly associated with transitions
between labour market states. Furthermore, couples are analysed separately
from individuals. With this in mind a number of models are required.

6.1.1 Couple-level analyses

For the examination of worklessness at the couple level, the models examining
the time to exit worklessness are considered for the following groups:

• all couples
• couples first observed having at least one unemployed partner
• couples first observed having both partners type 1 inactive

55
NDP: Characteristics and labour market transitions

• couples first observed having one partner type 1 inactive, the other type 2
inactive
• couples first observed having both partners type 2 inactive.

6.1.2 Individual-level analyses

At the individual level there is interest not only in moves into employment but any
moves away from the original status. Thus, the individual-level models not only
take the initial state into account but also consider transitions to all the other
states. These analyses allow for the possibility that those exiting their original
state can do so in a number of ways. For example, an individual originally type 1
inactive can leave this state to become employed, unemployed or type 2 inactive.
Hence, three models for each initial state of worklessness are estimated
separately for men and women. The resulting six models examine:

• Time to exit unemployment for men

• Time to exit unemployment for women
• Time to exit type 1 inactivity for men
• Time to exit type 1 inactivity for women
• Time to exit type 2 inactivity for men
• Time to exit type 2 inactivity for women.

The next section considers results for the five models estimated at the couple
level, followed by a section presenting the results for the six models estimated at
the individual level. The exposition of results focuses on the most influential
characteristics and makes extensive use of graphical tools. Full estimation
results are given in Appendices 2 and 3.

6.2 Couples’ exits from worklessness

In this section, the outcome of interest is whether either of the partners has found
work. This is a couple-level variable since, at this stage, the concern is not with
which partner finds work but rather with whether the couple exits worklessness.
A complication arises from the fact that the independent variables used to model
this exit are largely at the individual level. This means that, although a
characteristic may be associated with success in a couple exiting worklessness,
one cannot state that it is the partner having that characteristic who has actually
found employment. This is unavoidable when considering worklessness at the
level of the couple. However, it should be borne in mind that the descriptive
analysis highlighted the high degree of similarity between partners in a couple.
Hence, where one partner has a particular characteristic, there is a reasonable
chance of the other partner sharing it. The individual-level analysis in the next
section does not suffer from this ambiguity.

56
 Modelling transitions

The concept of distance to the labour market had to be adapted for the analysis
of couples. Four categories were defined:

• At least one partner unemployed

• Both partners type 1 inactive
• One partner type 1 inactive, the other type 2 inactive
• Both partners type 2 inactive.

These categories are useful in that they provide a clear ranking with regard to
distance from the labour market. Furthermore, they split the sample into roughly
similar-sized groups.

Figure 6.1 shows the cumulative movements out of work over the five
observation points. This type of graph is often referred to as a ‘survival curve’
because it represents what proportion survives in the initial state (here,
worklessness) as time progresses. It is a purely descriptive presentation of the
movement away from worklessness and does not incorporate any statistical
adjustment for other factors which may exert an influence. The “step” nature of
this chart reflects the fact that couples are sampled only at quarterly intervals.
Clearly the proportion remaining workless will fall gradually during the first
quarter, not all at once at the end of the quarter, but the information to show this
is not available.

There are a number of noteworthy points which combine to demonstrate the

importance of controlling for the initial type of worklessness:

• Couples where at least one partner was unemployed were much more likely
to exit worklessness. By the end of the observation period 57 per cent of
such couples remained workless.
• The other three categories are relatively close together with the proportion
remaining workless ranging from 87 per cent (both type 1 inactive) to 94 per
cent (both type 2 inactive).
• The further a couple is from the labour market the more likely it is to remain
workless.

57
NDP: Characteristics and labour market transitions

Figure 6.1: Empirical survivor functions by initial state

at least one unemployed both inactive 1
inactive 1 and inactive 2 both inactive 2

.75
probability still workless

.5

.25

0
0 1 2 3 4
Analysis time - quarters

The observed exits presented in Figure 6.1 do not control for any differences in
characteristics of the couples. These can be taken into account by estimating
duration models which assess the impact of characteristics on the probability of
exiting the initial state. Characteristics can be associated with a higher or a lower
probability of exit. Table 6.1 presents the results for the most important
characteristics. A value greater than one indicates that that characteristic is
associated with speedier exit, while a value less than one indicates the reverse.
A more comprehensive guide to the interpretation of the results is provided in
Appendix 1. The results themselves are given in full in Appendix 2.

58
 Modelling transitions

Table 6.1: Modelling exits from worklessness – selected results

All couples At least 1 Both inac. 1 Inact. 1 & 2 Both inac. 2
unempl.
Age group male partner
up to 30 1.268 1.071 2.179 1.753 1.449
(1.46) (0.35) (1.21) (0.85) (0.63)
31 to 40 1.271 1.027 2.967 1.590 2.292
(1.58) (0.14) (1.78) (0.88) (1.92)
41 to 50 1.344 1.346 1.442 1.706 0.750
(2.13)* (1.63) (0.80) (1.10) (0.71)
Highest qualifications of
either partner:
NVQ4+ 1.538 1.407 3.271 2.148 1.497
(2.77)** (1.79) (2.33)* (1.74) (1.06)
NVQ3 1.327 1.287 0.715 1.980 1.250
(2.74)** (2.04)* (0.82) (2.09)* (0.66)
NVQ2 1.157 1.157 0.627 1.648 0.792
(1.35) (1.18) (1.11) (1.20) (0.50)
NVQ1 1.466 1.429 0.768 2.801 1.155
(3.05)** (2.46)* (0.40) (2.71)** (0.28)
Other 1.267 1.328 0.704 1.658 1.244
(1.97)* (1.96) (0.61) (1.34) (0.55)
Long term health problem
Affecting work: either partner 0.708 0.669 0.612 0.912 0.383
(3.40)** (3.60)** (0.98) (0.18) (2.12)*
Affecting work: both partners 0.778 0.940 0.782 1.119 0.626
(1.83) (0.31) (0.57) (0.31) (1.58)

Length of non-employment 0.939 0.937 0.886 0.918 0.972

Spell (9.36)** (5.53)** (4.27)** (4.46)** (1.43)
Length of spell, squared 1.000 1.001 1.001 1.001 1.000
(6.37)** (2.68)** (3.69)** (3.95)** (0.56)
Initial state of worklessness
Both inactive 1 0.555
(2.92)**
Inactive 1 & 2 0.474
(4.20)**
Both inactive 2 0.284
(6.35)**

Couple claiming JSA 1.234 1.554

(0.94) (1.51)
Male partners benefits
IS 0.806 1.183 0.466 0.515 0.350
(0.85) (0.43) (1.45) (1.33) (1.66)
IB 0.954 1.701 0.393 0.740 0.514
(0.22) (1.60) (1.81) (0.71) (1.50)
IS & IB 0.563 1.420 0.326 0.311 0.156
(2.18)* (0.79) (1.83) (2.22)* (2.99)**
Other benefit (combination) 0.370 0.312 0.170 0.123 0.283
(3.10)** (2.29)* (2.55)* (2.43)* (1.99)*
Reference categories for the categorical variables: aged 51 or over; no qualifications; neither has
health problem affecting work; at least one unemployed; not claiming any non-JSA benefit. *
significant at 5 per cent, ** significant at 1 per cent.

59
NDP: Characteristics and labour market transitions

Inspecting the results for workless couples as a whole (first column of numbers in
Table 6.1) reveals a number of factors to be associated with exits:

• having qualifications (especially NVQ4 equivalent or higher) is associated

with increased probability of exit
• either partner having a long term health problem affecting work is associated
with a reduced probability of exit
• couples with longer spells without work are less likely to exit worklessness.
The strength of this effect declines over time
• the initial state of worklessness is important. Compared to the group where at
least one partner is unemployed the probability of exit is increasingly reduced
the further away from the labour market the couple is initially
• male partner claiming either IS and IB or another combination of benefits is
associated with a reduced probability of exit compared to not claiming any
individual benefit

Models estimated separately for the four initial states of worklessness confirm
these results. On the whole, it is the same group of characteristics that prove to
be important. There are, however, differences in the effects of these
characteristics.

• Having a qualification at NVQ 4 or above has the most pronounced effect on

couples where both are type 1.
• Either partner having a health problem that is affecting work seems to be
important for couples where at least one partner is unemployed and couples
at the other extreme where both partners are inactive 2. However, the
variable is insignificant for the other two types of couples.
• The length of the workless spell has a similar effect across all the four groups
of couples. However, it is not significant for couples where both are type 2
inactive.
• Benefits seem to have most effect on couples where at least one person is
type 2 inactive.

Figure 6.2 uses the estimation results to summarise movements away from non-
employment for four ‘average’ couples in the sample who differ only in their type
of initial worklessness. By varying the length of worklessness spell, it is possible
to trace out how quickly couples move from their initial state. Whereas Figure
6.1 simply showed observed exits over the observation period, the curves in
Figure 6.2 take differences in the characteristics of couples into account. The
model can predict these exits over a longer period than the observation period.
Consequently, exits over a three year (12 quarter) period are given.

60
 Modelling transitions

Figure 6.2: Predicted exits from worklessness by initial state

at least one unemployed both inactive 1
inactive 1 & inactive 2 both inactive 2

1
probability of remaining workless

.8

.6

.4

.2
1 2 3 4 5 6 7 8 9 10 11 12
length of spell (quarters)

The results confirms the findings of the descriptive analysis. The closer a couple
is to the labour market the more likely it is to exit worklessness. The curves
depict the cumulative probability of remaining workless depending on the initial
state of worklessness. The probability of remaining workless is by far the lowest
for couples where at least one partner is unemployed. In contrast, this probability
is highest for couples furthest away from the labour market where both partners
are type 2 inactive.

Predictions from estimating the model separately for each type of initial
worklessness are broadly similar. Table 6.2 below summarises the observed
and predicted probabilities of remaining in the initial state.

61
NDP: Characteristics and labour market transitions

Table 6.2: Observed and predicted probabilities of remaining workless

Duration of worklessness (months)
Percentage remaining workless 3 6 9 12 24 36
At least one partner unemployed
Observed 82.5 70.0 61.8 56.7
Predicted from joint model 84.2 71.3 61.1 53.0 31.9 21.1
Predicted from separate model 76.5 59.4 46.8 37.3 16.9 8.8
Both partners type 1 inactive
Observed 95.8 91.8 89.2 87.9
Predicted from joint model 90.5 82.5 75.5 69.4 51.9 41.1
Predicted from separate model 92.4 86.0 80.8 76.3 64.2 57.3
One partner inactive 1 the other 2
Observed 97.1 94.2 91.3 89.3
Predicted from joint model 91.8 84.7 78.5 73.1 57.0 46.6
Predicted from separate model 95.3 91.2 87.6 84.4 74.5 67.9
Both partners type 2 inactive
Observed 97.9 96.9 95.5 94.3
Predicted from joint model 94.9 90.4 86.3 82.6 71.0 62.9
Predicted from separate model 98.1 96.4 94.7 93.1 87.3 82.4

In most cases the models predict lower percentages remaining workless than is
observed. This is mainly due to the fact that the models predict the probability of
remaining workless for a couple just starting a spell whereas the numbers for the
‘observed’ row contain couples with very different initial lengths of worklessness.
As the probability of exit decreases with the length of the spell, the fact that the
numbers are higher in the latter case is not surprising.

Using the separate models, predicted probabilities of remaining workless are

even lower for couples closest to the labour market and higher for couples further
away. Differences between the two models arise for two reasons:

• the estimated coefficients of the models are different (see Table 6.1)
• with separate models, the predictions are based on an average couple of the
specific type. Consequently, characteristics of couples are not fixed in the
comparisons across different types of initial worklessness.

Overall, the modelling confirms the results of the descriptive analysis:

• couples closer to the labour market have a higher probability of exiting

worklessness compared to couples further away from the labour market
• The probability of exit diminishes the longer the spell of worklessness lasts.

62
 Modelling transitions

6.3 Individuals’ transitions between employment states

As already noted, modelling individuals’ transitions required the estimation of six
models. The full results are provided in Appendix 3. In this section, the key
results are summarised graphically and using tables of predicted probabilities.

When considering couples, the outcome variable was whether they had exited
worklessness. With individuals, all transitions are considered, regardless of
whether they involve an exit from worklessness. It is therefore possible to detect
moves towards or away from the labour market relative to the initial state. As the
analysis below shows, this proves to be important; for some initial states,
transitions into work account for only a small fraction of all transitions.

6.3.1 Transitions away from unemployment

This section is concerned with the transitions of those men and women in
workless couples who were unemployed when first observed. This constitutes
the group that is closest to the labour market and therefore one would expect a
higher proportion of their transitions to be into employment than for those
individuals who were first observed as being inactive.

The transitions are depicted graphically in Figures 6.3 and 6.4 for men and
women respectively. Four lines appear in both charts. These indicate the
destination on exiting unemployment. The line showing the steepest decline
relates to total exits and this overall number of exits can be divided between
those who find work, those who become type 1 inactive and those who become
type 2 inactive. Over the one year period, more than 50 per cent of unemployed
men exited unemployment. Most of these transitions were into work (second line
from the bottom) and fewest were to type 2 inactivity. An even higher proportion
of women exited unemployment during the observation period and only about 25
per cent remained unemployed. However, most female exits were to type 1
inactivity, closely followed by moves to work. Another gender difference is that,
for women, a very high percentage of exits occurred in the first quarter while for
men the exits were more evenly distributed over the whole observation period.

63
NDP: Characteristics and labour market transitions

Figure 6.3: Empirical survivor functions – unemployed men

Any exit Exit to employment
Exit to inactivity 1 Exit to inactivity 2

.75

.5

.25

0
0 1 2 3 4
Analysis time - quarters

Figure 6.4: Empirical survivor functions - unemployed women

Any exit Exit to employment
Exit to inactivity 1 Exit to inactivity 2

.75

.5

.25

0
0 1 2 3 4
Analysis time - quarters

64
 Modelling transitions

As with couples, descriptive analysis does not take into account any differences
in characteristics. Of particular importance is the length of the unemployment
spell. By estimating duration models it is possible to ‘align’ an average individual
at the beginning of his or her workless spell and predict how exit rates change as
the spell lengthens. It is also possible to predict the effect certain characteristics
have on exit rates.

These predicted exit rates for average individuals are presented graphically
below. Figure 6.5 considers male exits from unemployment. It is worth spending
some time explaining the format of this graph as all those that follow are
interpreted in the same way. Also the shading used for the different states
remains the same throughout the remainder of this report. The graph shows
three curves which illustrate the transitions to different states:

• the black area illustrates movements into work

• the striped area shows moves into type 1 inactivity
• the grey area shows moves into type 2 inactivity
• the speckled area shows moves into unemployment (in this case, remaining
unemployed)

The size of these areas illustrates the relative importance of the destinations.
Clearly, employment accounts for the largest proportion of predicted exits from
unemployment for men. The top area gives an indication of the stability of the
initial state. The larger this top area, the more stable is the initial state since
fewer exits are predicted.

One important fact has to be kept in mind when interpreting the graphs. The
models underlying these predictions only consider the first transition. Thus, the
area representing the initial state can only reduce in size over time. In reality,
some people will return to the initial state. The interpretations should thus focus
more on relative differences between initial states of worklessness comparing the
speed of exit and the importance of possible destination states.

65
NDP: Characteristics and labour market transitions

Figure 6.5: Predicted exits from unemployment – Men

100%
80% unemployed
60% inactive 2
40% inactive 1
20% working
0%
0

10

12
spell length: quarters

Figure 6.6: Predicted exits from unemployment – Women

100%
80% unemployed
60% inactive 2
40% inactive 1
20% working
0%
0

10

12

spell length: quarters

Most male exits from unemployment are into work. While this is also the main
destination for female exits from unemployment, nearly as many women move
into type 1 inactivity. There is also a higher proportion of women moving to type
2 inactivity. Overall, women are likely to exit unemployment more quickly than
men, but not necessarily by finding work.

Table 6.3 summarises the estimation results by showing how predicted exit rates
vary with individual characteristics. The top line gives the exit probability for a
man and a woman with ‘average’ characteristics (including average spell length).
While the previous graphs plotted exit probabilities for an ‘average’ unemployed
person starting their spell, the first row in Table 6.4 gives the probabilities that an
‘average’ person exits unemployment to one of the other states in the next period
given that they have been unemployed for the average spell length. Hence, the

66
 Modelling transitions

average unemployed man had a 79 per cent chance of still being unemployed in
the next quarter. The corresponding figure for women was 66 per cent.
Each subsequent row of the table is derived by altering a single characteristic
while holding all others at the average values. Thus the second row shows
probabilities for an average individual in the youngest age group with all other
characteristics at the average level. Comparing this to individuals in the other
age groups gives an indication of the influence age has. The results should be
interpreted with some caution in so far as some of these predictions are based
on estimated coefficients that were not statistically significant.

Table 6.3: Effect of individual characteristics on unemployment exit

Men Women
empl. unempl. inac. 1 inac. 2 empl. unempl. inact. 1 inact. 2
‘Average’ individual 13 79 5 3 13 66 13 7

Age group:
up to 30 11 82 4 2 10 70 15 6
31 to 40 12 79 6 2 14 68 12 7
41 to 50 16 76 4 4 20 60 11 10
51 and older 13 77 7 3 20 50 15 15

Youngest child:
no children 12 81 5 2 12 72 9 6
0-1 years 14 77 4 4 16 45 23 16
2-4 years 12 78 7 3 10 66 13 11
5-10 years 15 78 6 2 18 62 16 4
11+ years 12 80 6 2 14 65 14 6

Non-white 14 79 4 3 8 69 11 11
White 13 79 6 3 14 66 14 7

Highest qualification:
NVQ4+ 15 78 4 3 15 58 23 4
NVQ3 13 81 5 2 12 67 15 6
NVQ2 11 75 8 6 15 63 15 7
NVQ1 17 73 5 5 13 70 5 12
Other 16 79 3 2 12 65 17 6
No qualifications 12 80 6 2 11 67 14 7

Left full-time education

before 16 12 79 5 4 12 65 17 6
at 16 13 80 5 2 12 68 14 6
at 17 to 18 16 76 6 2 11 69 12 8
at 18 and over 15 76 6 2 24 53 6 17

No health problem 13 80 5 2 14 66 13 7
Health problem 12 78 7 4 9 68 14 9

Workless for:
0 quarters 22 71 4 2 20 60 12 7
1 year 19 75 4 2 19 62 13 7
2 years 15 78 5 3 17 63 13 7
3 years 13 80 5 3 16 64 13 7

67
NDP: Characteristics and labour market transitions

Claiming benefits
*
no benefits 12 72 12 5 12 64 16 8
JSA 13 80 5 2 17 72 7 4
IS 9 81 9 2 13 64 15 7
IB 14 61 17 8 13 58 16 13
IS & IB 12 73 10 6 - - - -
other combination 4 86 9 2 - - - -
*
For women, the reference category ‘no benefits’ also contains women claiming IS & IB or
another benefit combination.
Figures might not sum to 100 due to rounding.

Several key results are evident:

• The average woman has a 13 percentage point lower probability of remaining

unemployed than the average man. The additional transitions are to
inactivity, especially type 1 inactivity. The same percentage of men and
women enter work.
• Among men, those aged 41 to 50 have the highest probability of entering
work from unemployment. For unemployed women, the highest probability of
entering work is among those aged over 41 years. The likelihood of women
moving to type 2 inactivity increases with age.
• The presence of children does not seem to affect men’s transitions by much.
Unemployed women with children below the age of one year are much more
likely to move to either type of inactivity. They are also more likely to enter
work compared to their counterparts without children. Women with children
aged two to four are least likely to move from unemployment to work.
• There are hardly any differences between white and non-white men. White
women are more likely to move into work and less likely to move to type 2
inactivity compared to non-white women. However, this prediction is based
on an insignificant coefficient.
• There is no clear-cut picture with regard to the level of qualifications. Women
with an NVQ1 or equivalent qualification have a markedly higher probability of
moving to type 2 inactivity and a lower probability of moving to type 1
inactivity.
• For men, the age of leaving full-time education does not influence transition
probabilities much. Women who left full-time education at 18 or over are
much less likely to remain unemployed and more likely to either start work or
move to type 2 inactivity.
• Having a health problem reduces the probability of moving from
unemployment into work for women.
• The picture with regard to the spell length is very clear-cut. The longer the
spell the less likely unemployed men and women are to move into work.
Increasing the spell length from zero to three years reduces this probability by
7 percentage points for men. The probability of remaining unemployed or
moving to inactivity of either type increases with the length of the
unemployment spell.

68
 Modelling transitions

• Men claiming JSA are least likely to move away from the labour market into
inactivity of either type. Men claiming IB have the lowest probability of
remaining unemployed though most of the transitions are into inactivity. The
highest probability of moving into work is predicted among men claiming IS.
Unemployed men on other benefit combinations (which can include JSA)
have by far the highest probability of remaining in their initial state.

6.3.2 Transitions away from type 1 inactivity

In this section, attention turns to type 1 inactivity. Figures 6.7 and 6.8 describe
observed exits from this type of inactivity for men and women respectively.

Figure 6.7: Empirical survivor functions - type 1 inactive men

Any exit Exit to employment
Exit to unemployemnt Exit to inactivity 2

.75

.5

.25

0
0 1 2 3 4
Analysis time - quarters

69
NDP: Characteristics and labour market transitions

Figure 6.8: Empirical survivor functions - type 1 inactive women

Any exit Exit to employment
Exit to unemployemnt Exit to inactivity 2

.75

.5

.25

0
0 1 2 3 4
Analysis time - quarters

For both men and women, most exits are away from the labour market to type 2
inactivity. Fewer exits to unemployment and employment are observed. Most
exits happen at an early stage with fewer at later quarters over the observation
period.

Figures 6.9 and 6.10 show a very similar picture for predicted exit rates. For both
men and women, more exits take place at the beginning of the inactivity spell.
The majority of these exits are to type 2 inactivity. However, within six quarters
about 20 per cent had left type 1 inactivity for unemployment or (slightly less
often) work.

70
 Modelling transitions

Figure 6.9: Predicted exits from type 1 inactivity – Men

100%
80% inactive 1
60% inactive 2
40% unemployed
20% working
0%

10

12
0

8
spell length: quarters

Figure 6.10: Predicted exits from type 1 inactivity – Women

100%
80% inactive 1
60% inactive 2
40% unemployed
20% working

0%
0 1 2 3 4 5 6 7 8 9 10 11 12
Spell length - quarters

Table 6.4 presents transition probabilities by different characteristics. For the

average man, the probability of remaining in the initial state of type 1 inactivity is
one percentage point lower than for unemployed men. For women, 72 per cent
probability remain type 1 inactive. This is six percentage points higher than the
probability of remaining unemployed, indicating that type 1 inactivity is a more
stable state.

71
NDP: Characteristics and labour market transitions

Table 6.4: Effect of individual characteristics on type 1 inactivity exits

Men Women
empl. unempl. inac. 1 inac. 2 empl. unempl. inact. 1 inact. 2
‘Average’ individual 2 4 78 16 3 5 72 20

Age group:
up to 30 2 6 78 13 2 7 73 18
31 to 40 3 7 76 15 3 5 73 19
41 to 50 1 4 79 16 3 3 70 24
51 and older 1 2 79 17 2 3 73 22

Youngest child:
no children 2 6 76 17 3 5 70 22
0-1 years 2 5 78 16 2 4 71 24
2-4 years 2 4 82 12 2 5 73 20
5-10 years 2 3 78 17 3 4 73 20
11+ years 2 4 80 14 3 6 74 16

Non-white 2 6 80 12 1 2 71 26
White 2 4 78 16 3 5 72 20

Highest qualification:
NVQ4+ 2 8 78 11 6 10 69 15
NVQ3 2 5 76 17 4 7 74 16
NVQ2 2 5 79 14 1 4 71 23
NVQ1 2 5 80 14 2 7 74 17
Other 2 4 79 15 4 5 68 23
No qualifications 1 4 78 17 3 4 73 21

Left full-time education

before 16 2 4 78 16 2 5 74 20
at 16 2 5 78 15 3 5 70 22
at 17 to 18 1 4 80 15 3 4 76 17
at 18 and over 1 1 79 18 3 7 66 24

No health problem 2 4 76 18 3 6 72 19
Health problem 2 5 78 16 2 3 73 22

Workless for:
0 quarters 6 7 73 14 8 5 68 19
1 year 5 6 75 14 7 5 69 20
2 years 3 6 76 15 5 5 70 20
3 years 2 5 77 15 4 5 71 20

Claiming benefits
*
no benefits 2 6 78 14 3 5 72 20
JSA 5 23 63 9 5 12 54 29
IS 2 3 81 14 - - - -
IB 1 2 76 20 1 3 73 24
IS & IB 1 2 82 15 - - - -
other combination - - - - 2 1 76 21
*
For men, the reference category ‘no benefits’ also contains other combination and for women it
also contains women claiming IS only, or IS & IB.
Figures might not sum to 100 due to rounding.

72
 Modelling transitions

Some of the main associations between characteristics and transitions are

highlighted below:

• The older type 1 inactive individuals are, the more likely they are to move to
type 2 inactivity. Men in the youngest age group have a 13 per cent
probability of moving to type 2 inactivity while this is 17 per cent for the oldest
group. For women these numbers are 18 and 22 per cent, respectively.
• Women with very young children (aged one year or less) are the most likely to
move away from the labour market while women with children aged 11 or
older are least likely to make a move from type 1 to type 2 inactivity.
• White men are more likely to move to inactivity 2. For women it is the other
way round, non-white women are more likely to move away from the labour
market while white women are slightly more likely to move to unemployment
or work.
• Men with higher qualifications are more likely to move to unemployment.
There is no pattern with regard to movements to employment or type 2
inactivity. Having an NVQ4 or a higher qualification increases the probability
that women move towards the labour market: 16 per cent move to
unemployment or into work. Women with NVQ2 qualifications are less likely
to move towards the labour market (five per cent) and more likely to move
away to type 2 inactivity (23 per cent).
• The age of leaving full-time education does not seem to affect men’s
transition probabilities. Women who left full-time education late (at 18 or
older) are slightly more likely to move towards the labour market but at the
same time also more likely to move to type 2 inactivity.
• Having a long-term health problem affecting work did not appear to be
strongly associated with exits from type 1 inactivity.
• There is again a very clear-cut picture with regard to the length of the spell an
individual has been workless. The longer the spell, the less likely are moves
towards the labour market and the more likely the individual is to remain in
the initial state or move away from the labour market. This pattern is valid for
men and women.
• Men on JSA are most likely to move from type 1 inactivity. They are by far the
most likely to move to unemployment (23 per cent). At the same time men
claiming JSA are least likely to move further away from the labour market. A
similar picture regarding benefits applies to women. 11

11
These results should be interpreted with some care as both the economic status and the
benefit situation are self-reported and there is likely to be some error as JSA claimants should be
unemployed and not inactive.

73
NDP: Characteristics and labour market transitions

6.3.3 Transitions away from type 2 inactivity

The last set of results refers to individuals initially furthest away from the labour
market - type 2 inactive. Figures 6.11 and 6.12 show that there were relatively
few exits, for either men or women, from type 2 inactivity. This indicates the
rigidity of this type of worklessness. The vast majority of observed exits were to
type 1 inactivity. There were hardly any transitions into work or unemployment.

Figure 6.11: Empirical survivor functions - type 2 inactive men

Any exit Exit to employment
Exit to unemployemnt Exit to inactivity 1

.75

.5

.25

0
0 1 2 3 4
Analysis time - quarters

74
 Modelling transitions

Figure 6.12: Empirical survivor functions - type 2 inactive women

Any exit Exit to employment
Exit to unemployemnt Exit to inactivity 1

.75

.5

.25

0
0 1 2 3 4
Analysis time - quarters

This impression is confirmed when making predictions based on the estimated

models. Figures 6.13 and 6.14 show that the proportion remaining in the initial
state is much larger than for unemployed and type 1 inactive individuals,
especially for women. Type 1 inactivity is by far the most common destination
and there are very few movements to unemployment or work. For men, there
seem to be slightly more movements to unemployment than to work while the
opposite is true for women. In fact, while some transitions into work were
predicted for women, there were hardly any for men.

75
NDP: Characteristics and labour market transitions

Figure 6.13: Predicted exits from type 2 inactivity – Men

100%
80% inactive 2
60% inactive 1
40% unemployed
20% working
0%

10

12
0

8
spell length: quarters

Figure 6.14: Predicted exits from type 2 inactivity – Women

100%
80% inactive 2
60% inactive 1
40% unemployed
20% working
0%
0

10

12

spell length: quarters

Table 6.5 presents predicted exit probabilities by characteristics. For men, type 2
inactivity is very stable - 84 per cent remain in this initial state. This compares
with 79 per cent for unemployed men and 78 per cent for type 1 inactive men. A
similar impression of stability emerges when considering women. Eighty-nine
per cent remain type 2 inactive, compared to 65 and 72 per cent for
unemployment and type 1 inactivity respectively.

76
 Modelling transitions

Table 6.5: Effect of individual characteristics on type 2 inactivity exits

Men Women
empl. unempl. inac. 1 inac. 2 empl. unempl. inact. 1 inact. 2
‘Average’ individual 1 1 14 84 1 1 10 89

Age group:
up to 30 1 2 12 86 1 1 10 88
31 to 40 1 3 17 80 1 1 10 87
41 to 50 0 1 17 82 1 1 10 88
51 and older 1 1 1 86 1 0 8 91

Youngest child:
no children 1 2 15 83 1 1 9 89
0-1 years 1 1 14 84 1 1 8 90
2-4 years 1 2 13 84 1 1 10 88
5-10 years 0 2 13 85 0 1 10 88
11+ years 1 1 15 83 2 1 11 86

Non-white 1 4 7 89 1 1 6 92
White 1 1 15 83 1 1 10 88

Highest qualification:
NVQ4+ 2 1 20 77 1 1 6 92
NVQ3 0 2 13 85 1 1 9 89
NVQ2 1 2 18 79 1 1 10 88
NVQ1 1 1 9 89
Other 1 1 16 82 1 1 12 86
No qualifications 0 1 14 85 1 1 10 88

Left full-time education

before 16 0 2 14 83 1 1 10 89
at 16 1 1 15 82 1 1 9 89
at 17 to 18 1 1 11 87 1 1 11 87
at 18 and over 1 1 9 89 1 3 8 88

No health problem 1 2 14 83 1 1 9 89
Health problem 1 1 14 84 1 1 10 88

Type 2 inactive for:

0 quarters 1 2 17 80 3 2 14 81
1 year 1 2 17 81 3 2 13 82
2 years 1 2 16 81 2 2 13 83
3 years 1 2 16 82 2 1 12 84

Claiming benefits
no benefits* 0 2 16 82 1 1 10 88
JSA 7 10 15 68 2 1 10 87
IS 0 1 13 85 1 1 9 89
IB 1 1 14 85 0 0 9 90
IS & IB 0 1 15 84
other combination
*
For women, the reference category ‘no benefits’ also contains women claiming IS & IB or
another benefit combination; for men, it also contains other benefit combinations.
Figures might not sum to 100 due to rounding.

77
NDP: Characteristics and labour market transitions

Since, for both men and women, there are very few moves to employment or
unemployment, much of the interpretation will focus on the relative stability of
type 2 inactivity as opposed to moves to type 1 inactivity with less attention to the
few moves to unemployment and work. The high stability of the initial state
means that there is relatively little variation in the transition probabilities.

• For men, those aged between 31 and 50 years had the highest probability of
moving away from type 2 inactivity. For women, there was little variation by
age, although those in the oldest age group were most likely to remain type 2
inactive.
• The presence and age of children has little effect on exits from type 2
inactivity. This is true for both men and women.
• Men and women from minority ethnic groups have a larger probability of
staying type 2 inactive compared with their white counterparts.
• Having a NVQ4 or higher has a very different effect on men and women. It
increases the probability of men moving away from inactivity 2 while women
with the highest level of qualifications are least likely to move.
• For men, the age they left full-time education is associated with movements
from type 2 inactivity. The older they were when they left full-time education,
the more stable is their type 2 inactivity.
• Again, there is a clear pattern with regard to the length of the spell – the
longer the spell of worklessness the more stable individuals are in their type 2
inactivity.

6.3.4 The relationship between partners’ employment statuses

In the final section of this chapter, attention turns to the question of the
relationship between the employment statuses of the two partners. It is a well-
established empirical finding that the partners of unemployed men in Britain are
less likely to be engaged in paid work than the wives of men who are employed.
This subject is considered by, for example, Davies et al. (1992) who conclude
that a causal relationship exists between male and female employment statuses
within a couple. The analysis that follows allows for the possibility that the
relationship can operate in both directions. That is, the model assumes that the
man in a couple finding employment can influence the probability of his partner
moving closer to the labour market and that the woman finding employment can
influence the chances of the man moving closer. This is a similar approach to
that used in Dorsett (2001b). Fuller details of this approach are provided in
Appendix 1.

All of the individual-level models discussed above took account of this possible
correlation across the partners. The results were mixed in terms of whether this
association was significant:

78
 Modelling transitions

• For unemployed men, the partner finding work was associated with either
remaining unemployed or moving into employment
• For type 1 inactive men, there was no association
• For type 2 inactive men, the partner finding work was associated with a move
into unemployment
• For unemployed women, there was no association
• For type 1 inactive women, the partner finding work was associated with a
move into unemployment
• For type 2 inactive women, the partner finding work was associated with a
move to type 1 inactivity or unemployment.

Overall, these results suggest an inter-relationship between the partners’

statuses. For both men and women, their partner finding work is associated with
moving closer to the labour market, although not necessarily entering work. It
does not necessarily follow that this is a causal relationship, however. It may be
that there are some unobserved characteristics, shared by partners within a
couple, that are associated with the probability of moving closer towards the
labour market. In this case, the result simply captures the tendency for couples
to be polarised with respect to these movements.

79
NDP: Characteristics and Labour Market transitions

80
 Appendix 1: Methodology

Appendix 1 Methodology
Duration models

The econometric results presented in this report are based on two types of
model: binomial logistic regression and multinomial logistic regression. These
are standard techniques which are routinely used when modelling a dependent
variable12 that is categorical in nature. It is well-acknowledged that such models
can be used for the purposes of duration analysis (Allison, 1982). The binomial
model is appropriate when the dependent variable is dichotomous. This will be
the case when considering worklessness versus non-worklessness, for example.
The multinomial model is simply a generalisation of this dichotomous case and
allows the dependent variable to indicate one of a set number of states. For the
purposes of this research, the multinomial model was used to differentiate
between employment, unemployment and the two types of inactivity.

Clearly, individuals are not observed from the start of their worklessness spell.
This means that only transitions for those couples or individuals who have
remained workless up to the point of first appearing in the data can be observed.
Jenkins (1995) shows that this does not pose any problems so long as the length
of the spell is known. This highlights the importance of knowing the length of the
worklessness spell. In the analysis that follows, this is simply taken as the time
since last employment.13 For those individuals with no previous employment,
the spell of worklessness is taken to begin at the time they were aged 16. This
implicitly treats any post-16 education as inactivity and in so doing is consistent
with the ILO definition of economic inactivity. A complication arises when
considering couples. In this case, the length of the worklessness spell is taken to
be the minimum of the man’s period of worklessness and the woman’s period of
worklessness. The implicit assumption is that the couple has existed at least as
long as the period of worklessness. This assumption is required since there is no
information in the dataset on how long the couple has, in fact, existed.

It is worth pointing out one important difference between the dataset used for the
econometric analysis and that used for the descriptive analysis. For simplicity
and clarity, the descriptive analysis only considered transitions for those couples
who were observed five times in the dataset. The duration analysis used in
modelling transitions is robust to attrition. That is, the influences on changes in
economic status can be estimated across all couples, be they fully-participating
or not. However, this relies on the assumption that attrition from the sample is
independent of the probability of change in economic status. The descriptive
analysis suggests that this was unlikely to be the case, and that focusing on the
balanced panel was tantamount to considering a lower bound on the true

12
The dependent variable is the variable that is to be predicted, ie the variable of interest.
13
Note that all spells are given as quarters. This is in line with the LFS which carries out
interviews each quarter.

81
NDP: Characteristics and Labour Market transitions

proportion moving into work. In the econometric analysis, the full sample was
used but, following the approach of Verbeek and Nijman (1992), variables
indicating the panel nature of the data were included in an attempt to control for
attrition.

The interdependence of partners’ employment status

To allow for the influence of one partner’s employment status on that of the other,
the approach of Mallar (1977) was followed. This is a two-stage approach. In
the first stage, two indexes are calculated, one reflecting the underlying
propensity for the man to enter employment, the other a similar index for the
woman. These indexes are based on two separate logistic regression models of
observed employment for men and women, where the regressors in both are all
those characteristics affecting either male or female employment. In the second
stage, male employment participation is estimated, including as regressors those
variables thought to influence male participation together with the index reflecting
the propensity of the female to participate that was calculated in the first stage.
For women’s participation, an analogous model is estimated. The resulting
estimates allow for the possibility that the male partner finding employment can
influence the probability of his partner doing likewise and that the female partner
finding employment can influence the chances of job entry for the man.

Clearly, this approach is suited to the consideration of the case where the
dependent variable is dichotomous. When considering individual transitions, the
destinations on worklessness exit are also important. In this case, the first stage
is identical (estimating the probability of male or female job entry) but the second
stage involves a multinomial rather than a binomial logistic regression.
Consequently, what the results capture is the effect of one partner moving into
work on the other partner moving to one of the other destinations.

Interpreting the estimation results

In the remainder of this appendix, a guide to interpretation of the model results is

provided. This is for the benefit of those who are unfamiliar with such models, or
with modelling terminology more generally.

The advantage of econometric models over purely descriptive accounts is that

they allow the researcher to investigate the extent to which a dependent variable
is simultaneously associated with a number of other variables. These other
variables are often termed ‘independent’ or ‘explanatory’ variables, although one
should be cautious in assuming the relationship is causal in nature. To examine
the relationship, the dependent variable is ‘regressed’ on the independent
variables with the result that a measure of the separate influence of each
independent variable is obtained. An estimate of the statistical significance of the

82
 Appendix 1: Methodology

influence is also obtained. This allows the researcher to reach a view as to what
are the most important correlations.

With a logistic regression model, the measures of influence appear as ‘odds

ratios’. The interpretation of these ratios differs according to whether the
independent variable is categorical (that is, it indicates a category such as
gender or ethnic group) or continuous (that is, it indicates a quantity such as
wages or age). In the categorical case, the ratios represent the extent to which
the independent variable is associated with an increase in the odds of the
dependent variable taking on a particular value relative to some reference
category. To illustrate, if the dependent variable were a binary indicator of
whether an individual was in work, the ratio attached to the independent variable
indicating presence of children in the household would show the extent to which
having a child changed the odds of working relative to not having a child. The
term ‘odds’ is used here exactly as in betting. If an outcome occurs 1 time in ten,
the odds-against are 9 to 1 (i.e. 9), and the odds-on are 1 to 9 (i.e. 1/9). If the
effect is 1 then the odds are unchanged. If the effect is greater than 1, the odds
are increased (become higher), while if the effect is less than 1, the odds are
decreased (become lower).

When the independent variable is continuous, the interpretation of the odds ratio
is slightly different. Now it represents the effect on the odds of a unit increase in
the value of the predictor variable. Thus, considering the same dependent
variable but a continuous independent variable (minimum net hourly pay for
which the respondent will work), the odds ratio represents the marginal impact of
each extra pound of expected earnings on the odds of being in work. The
estimated odds ratio can be small in size for continuous variables, depending on
the scale used. For example, the estimated effect of the length of non-
employment spell would be smaller if measured in weeks that it would if
measured in years.

With a multinomial logistic regression model, the interpretation is slightly

different. Now the estimated effects show the association between an
independent variable and the odds of the dependent variable taking a particular
value relative to it taking another value. To make this more transparent, consider
the case in which the dependent variable is the individual’s economic status
which can be either employed, unemployed or inactive. There may be interest in
the effect on employment status of there being children in the household
compared to the situation in which children are not present. The results will show
this effect as it relates to a reference category of the dependent variable. Taking
inactivity as the reference category, the results can show how the presence of
children increases the odds of being employed rather than inactive, or of being
unemployed rather than inactive.

83
NDP: Characteristics and Labour Market transitions
 Appendix 2: Results for couples

Appendix 2 Results for couples

Table A1: Modelling exits from worklessness
All couples At least 1 Both inac. 1 Inact. 1 & 2 Both inac. 2
unempl.
Age group male partner
up to 30 1.268 1.071 2.179 1.753 1.449
(1.46) (0.35) (1.21) (0.85) (0.63)
31 to 40 1.271 1.027 2.967 1.590 2.292
(1.58) (0.14) (1.78) (0.88) (1.92)
41 to 50 1.344 1.346 1.442 1.706 0.750
(2.13)* (1.63) (0.80) (1.10) (0.71)
Youngest child:
0-1 years 0.996 1.049 1.321 0.904 0.493
(0.03) (0.33) (0.44) (0.17) (1.13)
2-4 years 1.034 0.956 0.805 2.259 0.622
(0.24) (0.28) (0.31) (1.51) (0.98)
5-10 years 1.141 1.248 0.636 1.492 0.343
(0.98) (1.38) (0.73) (0.84) (1.89)
11+ years 1.184 1.066 1.443 1.062 1.702
(1.19) (0.37) (0.64) (0.11) (1.45)

Either partner not white 0.806 0.849 0.642 0.426 0.574

(1.56) (1.09) (0.48) (1.09) (0.92)
Highest qualifications of
either partner:
NVQ4+ 1.538 1.407 3.271 2.148 1.497
(2.77)** (1.79) (2.33)* (1.74) (1.06)
NVQ3 1.327 1.287 0.715 1.980 1.250
(2.74)** (2.04)* (0.82) (2.09)* (0.66)
NVQ2 1.157 1.157 0.627 1.648 0.792
(1.35) (1.18) (1.11) (1.20) (0.50)
NVQ1 1.466 1.429 0.768 2.801 1.155
(3.05)** (2.46)* (0.40) (2.71)** (0.28)
Other 1.267 1.328 0.704 1.658 1.244
(1.97)* (1.96) (0.61) (1.34) (0.55)
SOC male partner
Manager & administrator 1.797 1.518 4.143 0.409 3.933
(2.31)* (1.32) (2.50)* (1.31) (2.90)**
Professional 1.279 1.524 0.515 1.402 0.901
(1.26) (1.85) (0.60) (0.70) (0.15)
Associate professional & 1.075 1.396 1.039 0.500 0.413
technical (0.45) (1.79) (0.05) (0.81) (1.17)
Clerical & secretarial 1.138 1.114 1.082 2.230 1.750
(0.58) (0.42) (0.09) (1.26) (0.67)
Personal, protective services 1.137 1.322 0.681 0.673 1.107
(1.12) (2.03)* (0.80) (0.90) (0.24)
Sales 0.882 0.962 0.680 0.548 2.275
(0.88) (0.23) (0.62) (0.95) (1.89)
Plant & machine operatives 1.797 1.518 4.143 0.409 3.933
(2.31)* (1.32) (2.50)* (1.31) (2.90)**
Other occupations 1.279 1.524 0.515 1.402 0.901
(1.26) (1.85) (0.60) (0.70) (0.15)

85
NDP: Characteristics and Labour Market transitions

Long term health problem

Affecting work: either partner 0.708 0.669 0.612 0.912 0.383
(3.40)** (3.60)** (0.98) (0.18) (2.12)*
Affecting work: both partners 0.778 0.940 0.782 1.119 0.626
(1.83) (0.31) (0.57) (0.31) (1.58)

Length of non-employment 0.939 0.937 0.886 0.918 0.972

spell (9.36)** (5.53)** (4.27)** (4.46)** (1.43)
Length of spell, squared 1.000 1.001 1.001 1.001 1.000
(6.37)** (2.68)** (3.69)** (3.95)** (0.56)
Initial state of worklessness
Both inactive 1 0.555
(2.92)**
Inactive 1 & 2 0.474
(4.20)**
Both inactive 2 0.284
(6.35)**

Couple claiming JSA 1.234 1.554

(0.94) (1.51)
Male partners benefits
IS 0.806 1.183 0.466 0.515 0.350
(0.85) (0.43) (1.45) (1.33) (1.66)
IB 0.954 1.701 0.393 0.740 0.514
(0.22) (1.60) (1.81) (0.71) (1.50)
IS & IB 0.563 1.420 0.326 0.311 0.156
(2.18)* (0.79) (1.83) (2.22)* (2.99)**
Other benefit (combination) 0.370 0.312 0.170 0.123 0.283
(3.10)** (2.29)* (2.55)* (2.43)* (1.99)*
Female partners benefits
IS 1.087 0.894 0.196 1.695 1.087
(0.35) (0.35) (1.24) (0.99) (0.12)
IB 0.934 1.494 0.238 0.460 0.672
(0.41) (1.80) (1.99)* (1.48) (0.92)
IS & IB 0.382 1.330 0.204 0.516
(1.79) (0.28) (1.52) (0.89)
Other benefit (combination) 0.647 0.436 0.844 0.706 0.596
(1.78) (2.12)* (0.26) (0.54) (0.87)
Interview in:
1997 0.718 0.681 0.633 0.855 0.529
(2.98)** (2.91)** (1.00) (0.41) (1.80)
1998 0.830 0.926 0.485 0.901 0.329
(1.65) (0.58) (1.44) (0.27) (2.70)**
1999 0.809 0.894 0.598 0.910 0.543
(1.73) (0.77) (1.04) (0.21) (1.72)
2000 0.718 0.681 0.633 0.855 0.529
(2.98)** (2.91)** (1.00) (0.41) (1.80)
Quarter 2 0.912 0.965 0.451 1.346 0.708
(0.81) (0.26) (1.78) (0.89) (1.02)
Quarter 3 0.955 1.130 0.509 0.816 0.479
(0.40) (0.90) (1.54) (0.56) (2.00)*
Quarter 4 0.888 1.060 0.446 0.642 0.584
(1.02) (0.42) (1.78) (1.12) (1.55)

86
 Appendix 2: Results for couples

couple not observed in all 0.838 0.819 1.205 0.359 1.117

waves (1.93) (1.90) (0.44) (2.65)** (0.34)

LFS interview 3 0.882 0.873 1.736 0.767 0.746

(1.23) (1.13) (1.35) (0.79) (0.85)
LFS interview 4 0.817 0.787 1.073 1.042 0.906
(1.79) (1.77) (0.14) (0.12) (0.28)
LFS interview 5 0.763 0.627 2.062 1.111 1.089
(2.20)* (3.09)** (1.65) (0.28) (0.24)

Observations 11989 3754 1516 2390 4278

Robust z-statistics in parentheses
* significant at 5% level; ** significant at 1% level
Reference categories for the categorical variables: aged 51 or over; no children; no qualifications;
craft & related occupation; neither has health problem affecting work; at least one unemployed;
not claiming any non-JSA benefit; interview in 1996; interview in quarter 1; LFS interview 2

87
NDP: Characteristics and Labour Market transitions
 Appendix 3: Results for individuals

Appendix 3 Results for individuals

Table A2: Modelling exits at the individual level: Men
Initial state unemployed Inactive type 1 Inactive type 2
Exit to: working inactive 1 inactive 2 working unemployed inactive 2 working unemployed inactive 1

Age group:
up to 30 0.812 0.522 0.571 2.512 2.799 0.780 1.056 2.885 0.971
(0.92) (1.95) (1.12) (1.45) (2.39)* (0.89) (0.08) (1.76) (0.12)
31 to 40 0.889 0.885 0.623 3.108 3.170 0.900 1.543 5.005 1.504
(0.54) (0.44) (1.05) (2.01)* (3.04)** (0.51) (0.54) (2.56)* (2.19)*
41 to 50 1.241 0.631 1.251 1.364 1.495 0.935 0.640 2.411 1.447
(0.99) (1.65) (0.54) (0.61) (1.15) (0.47) (0.70) (1.77) (2.80)**
Youngest child:
0-1 years 1.286 0.940 2.144 0.965 0.789 0.894 1.867 0.687 0.958
(1.47) (0.21) (1.82) (0.06) (0.67) (0.45) (0.83) (0.65) (0.18)
2-4 years 1.069 1.502 1.397 1.191 0.519 0.678 1.496 0.904 0.875
(0.36) (1.55) (0.84) (0.30) (1.75) (1.64) (0.66) (0.17) (0.64)
5-10 years 1.292 1.270 0.783 0.880 0.398 0.998 0.651 0.777 0.851
(1.39) (0.91) (0.55) (0.24) (2.36)* (0.01) (0.49) (0.44) (1.03)
11+ years 0.994 1.425 1.121 1.100 0.629 0.777 1.927 0.397 1.023
(0.03) (1.35) (0.28) (0.17) (1.24) (1.41) (0.95) (1.97)* (0.15)

non-white 1.067 0.736 1.373 0.869 1.476 0.723 0.979 2.880 0.438
(0.36) (0.96) (0.83) (0.18) (0.81) (1.21) (0.02) (2.10)* (3.55)**
Highest qual.:
NVQ4 + 1.390 0.683 1.249 2.141 2.485 0.646 9.997 0.493 1.660
(1.23) (0.95) (0.43) (1.12) (1.70) (1.18) (2.74)** (0.54) (1.90)
NVQ3 1.100 0.838 0.758 1.807 1.340 1.024 0.888 1.168 0.956
(0.60) (0.75) (0.70) (1.31) (0.96) (0.16) (0.14) (0.36) (0.33)
NVQ2 1.035 1.475 2.886 1.650 1.302 0.786 5.630 1.886 1.431
(0.19) (1.54) (2.83)** (0.90) (0.65) (1.00) (2.16)* (1.13) (1.88)
NVQ1 1.610 0.942 2.648 1.434 1.318 0.788
(2.27)* (0.17) (2.26)* (0.60) (0.66) (0.90)

89
NDP: Characteristics and labour market transitions

other 1.472 0.508 1.004 1.419 0.975 0.901 4.164 1.095 1.192
(2.11)* (2.13)* (0.01) (0.67) (0.07) (0.62) (2.24)* (0.18) (1.06)
Left education:
before 16 0.920 0.909 1.970 0.951 0.766 1.111 0.487 1.386 0.902
(0.61) (0.50) (2.48)* (0.13) (1.04) (0.69) (1.10) (0.91) (0.84)
17-18 1.304 1.253 1.054 0.620 0.698 0.972 0.752 0.727 0.655
(1.36) (0.72) (0.11) (0.72) (0.82) (0.10) (0.45) (0.49) (1.64)
over 18 1.254 1.236 1.259 0.556 0.237 1.210 0.635 0.610 0.527
(0.99) (0.57) (0.45) (0.88) (2.20)* (0.61) (0.66) (0.50) (2.10)*

long-term 0.897 1.415 1.914 1.014 1.081 0.813 0.438 0.653 1.016
health prob. (0.78) (1.82) (2.56)* (0.03) (0.25) (1.03) (1.38) (0.82) (0.09)

length of spell 0.936 1.013 0.998 0.906 0.966 1.004 0.971 0.985 0.992
(7.95)** (1.41) (0.14) (3.45)** (2.70)** (0.60) (1.11) (0.67) (1.44)
square of spell 1.000 1.000 1.000 1.001 1.000 1.000 1.000 1.000 1.000
length (5.19)** (0.03) (0.46) (2.87)** (2.31)* (0.47) (0.47) (0.10) (0.50)

Benefits
JSA 1.037 0.341 0.418 4.483 5.735 0.760 25.414 7.264 1.140
(0.15) (4.05)** (2.27)* (2.80)** (4.87)** (0.78) (4.57)** (3.93)** (0.31)
IS 0.675 0.661 0.265 1.033 0.507 0.932 0.438 0.433 0.800
(0.68) (0.78) (1.22) (0.06) (1.81) (0.33) (0.61) (1.26) (1.13)
IB 1.422 1.779 1.797 0.807 0.406 1.434 1.142 0.470 0.814
(0.57) (1.09) (0.80) (0.42) (2.35)* (1.98)* (0.16) (1.52) (1.17)
IS & IB 0.966 0.847 1.172 0.505 0.297 0.980 0.966 0.310 0.880
(0.06) (0.27) (0.20) (1.14) (2.95)** (0.10) (0.03) (1.62) (0.72)
other benefit 0.246 0.654 0.296
(1.81) (0.92) (1.41)
Interview in
1998 0.597 0.617 1.235 0.578 0.483 0.842 1.030 0.360 0.921
(3.37)** (2.16)* (0.65) (1.18) (2.54)* (1.09) (0.05) (2.23)* (0.56)
1999 0.901 0.736 1.195 0.636 0.546 0.754 0.332 0.431 1.030
(0.70) (1.37) (0.51) (0.99) (2.14)* (1.70) (1.53) (1.71) (0.19)
2000 0.899 0.759 1.406 0.757 0.443 0.882 0.892 0.472 0.924

90
 Appendix 3: Results for individuals

(0.66) (1.12) (0.96) (0.65) (2.34)* (0.73) (0.16) (1.95) (0.50)

Quarter 2 1.083 1.153 0.822 0.884 1.402 0.925 0.617 0.504 0.859
(0.51) (0.63) (0.61) (0.30) (1.08) (0.49) (0.79) (1.33) (1.06)
Quarter 3 1.143 0.833 0.998 0.908 1.479 1.082 0.398 0.841 0.900
(0.87) (0.74) (0.01) (0.24) (1.26) (0.50) (1.37) (0.39) (0.74)
Quarter 4 0.945 1.139 0.753 0.598 1.641 1.108 0.733 0.910 1.008
(0.35) (0.58) (0.86) (1.08) (1.60) (0.64) (0.59) (0.24) (0.06)

not observed 0.910 1.092 1.203 0.581 0.986 1.069 0.744 1.537 0.850
in all waves (0.77) (0.46) (0.70) (1.48) (0.06) (0.51) (0.50) (1.21) (1.30)

LFS interview
Number 3 0.959 0.947 1.247 1.197 0.765 0.760 0.552 0.779 0.702
(0.31) (0.27) (0.77) (0.50) (1.05) (2.06)* (1.22) (0.61) (2.86)**
Number 4 0.901 0.805 1.244 1.008 0.640 0.500 0.556 1.111 0.568
(0.66) (0.92) (0.71) (0.02) (1.50) (4.22)** (0.88) (0.29) (3.94)**
Number 5 0.824 0.874 0.937 1.285 0.639 0.505 0.329 0.232 0.463
(1.13) (0.52) (0.17) (0.55) (1.39) (3.68)** (1.30) (2.27)* (4.72)**
Partner’s 1.172 1.211 0.930 0.999 1.013 1.003 1.023 1.317 1.064
prob. of emp (2.56)* (1.99)* (0.48) (0.02) (0.20) (0.09) (0.17) (2.30)* (1.72)

Observations 3256 3256 3256 2646 2646 2646 3500 3500 3500
Robust z-statistics in parentheses
* significant at 5% level; ** significant at 1% level
Reference categories for the categorical variables: aged 51 or over; no children; no qualifications; left full-time education at 16; no benefit (for
unemployed and inactive 1), no benefit or other combination (for inactive 2); interview in 1996; interview in quarter 1; LFS interview 2.

91
NDP: Characteristics and labour market transitions

Table A3: Modelling exits at the individual level: Women

Initial state unemployed Inactive type 1 Inactive type 2
Exit to: working inactive 1 inactive 2 working unemployed inactive 2 working unemployed inactive 1

Age group:
up to 30 0.312 0.719 0.219 1.109 2.533 0.823 1.328 4.368 1.307
(2.38)* (0.62) (2.29)* (0.17) (1.84) (0.80) (0.56) (2.09)* (1.46)
31 to 40 0.462 0.593 0.272 1.808 1.796 0.853 1.618 3.663 1.416
(1.46) (0.94) (1.83) (0.94) (1.14) (0.70) (1.08) (1.98)* (2.09)*
41 to 50 0.801 0.581 0.487 1.509 1.317 1.118 1.422 3.325 1.379
(0.45) (1.01) (1.20) (0.81) (0.63) (0.64) (0.81) (2.05)* (2.62)**
Youngest child:
0-1 years 2.230 4.231 4.770 0.642 0.669 1.075 0.418 0.819 0.930
(1.70) (3.46)** (2.84)** (0.82) (0.96) (0.31) (1.80) (0.41) (0.40)
2-4 years 0.911 1.607 2.031 0.856 0.935 0.891 0.549 1.250 1.178
(0.18) (1.07) (1.23) (0.30) (0.17) (0.53) (1.20) (0.46) (0.94)
5-10 years 1.842 2.086 0.727 1.078 0.808 0.863 0.327 1.340 1.175
(1.62) (2.00)* (0.47) (0.17) (0.55) (0.75) (1.94) (0.67) (1.04)
11+ years 1.372 1.740 1.027 1.299 1.202 0.686 1.398 1.438 1.284
(0.80) (1.44) (0.05) (0.63) (0.52) (2.14)* (0.87) (0.83) (1.93)

non-white 0.550 0.776 1.680 0.375 0.321 1.305 0.593 0.460 0.611
(1.19) (0.54) (1.00) (1.13) (2.06)* (1.12) (0.89) (1.71) (2.94)**
Highest qual.:
NVQ4 + 1.642 2.031 0.579 2.665 2.832 0.756 1.358 0.423 0.610
(0.94) (1.14) (0.72) (1.39) (1.65) (0.81) (0.51) (1.12) (1.86)
NVQ3 1.042 1.117 0.814 1.338 1.825 0.743 0.676 0.489 0.969
(0.08) (0.22) (0.29) (0.60) (1.61) (1.30) (0.75) (1.19) (0.18)
NVQ2 1.462 1.172 1.050 0.458 1.197 1.164 1.029 0.939 1.032
(1.14) (0.54) (0.10) (1.97)* (0.60) (0.93) (0.08) (0.19) (0.24)
NVQ1 1.144 0.339 1.592 0.850 1.786 0.778 0.775 0.856 0.907
(0.32) (2.24)* (0.84) (0.31) (1.75) (1.16) (0.49) (0.35) (0.58)
other 1.102 1.294 0.779 1.546 1.479 1.171 0.737 0.854 1.356
(0.21) (0.53) (0.35) (0.91) (0.96) (0.74) (0.57) (0.29) (1.90)

92
 Appendix 3: Results for individuals

Left education:
before 16 1.039 1.350 1.166 0.611 0.951 0.831 0.911 0.755 1.026
(0.12) (0.99) (0.31) (1.27) (0.18) (1.34) (0.33) (0.84) (0.25)
17-18 0.883 0.856 1.540 0.909 0.795 0.702 0.949 0.765 1.195
(0.32) (0.36) (0.81) (0.20) (0.67) (1.83) (0.11) (0.56) (1.19)
over 18 2.712 0.612 4.503 1.305 1.697 1.129 1.462 2.746 0.838
(2.23)* (0.76) (2.58)** (0.43) (1.11) (0.39) (0.60) (1.65) (0.68)

long-term 0.571 1.053 1.326 0.500 0.575 1.116 0.455 0.804 1.134
health prob. (1.80) (0.15) (0.67) (1.95) (1.92) (0.83) (2.44)* (0.62) (1.19)

length of spell 0.971 1.003 1.000 0.940 0.989 1.002 0.955 0.970 0.989
(2.01)* (0.20) (0.02) (4.64)** (0.96) (0.35) (3.44)** (2.77)** (2.49)*
square of spell 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000
length (0.07) (0.60) (0.26) (3.76)** (1.24) (0.58) (2.18)* (2.17)* (0.65)

Benefits
JSA 1.300 0.361 0.467 2.643 3.481 2.016 2.455 1.256 1.052
(0.98) (2.94)** (1.91) (1.40) (2.31)* (1.60) (1.18) (0.26) (0.08)
IS 1.094 0.997 0.854 1.745 0.691 0.905
(0.13) (0.01) (0.21) (1.07) (0.67) (0.51)
IB 1.296 1.104 1.933 0.229 0.490 1.184 0.109 0.290 0.947
(0.21) (0.09) (0.61) (1.75) (1.27) (0.87) (2.13)* (1.90) (0.38)
Other benefits 0.801 0.132 0.999
(0.31) (2.10)* (0.01)
Interview in
1998 0.668 0.949 3.066 0.562 1.214 1.552 1.000 0.906 0.817
(1.28) (0.18) (2.43)* (1.57) (0.70) (2.89)** (0.00) (0.29) (1.75)
1999 0.859 0.875 3.767 0.552 0.947 1.372 0.702 0.826 0.912
(0.44) (0.39) (2.73)** (1.58) (0.19) (1.95) (0.97) (0.58) (0.80)
2000 1.108 1.142 1.720 0.491 0.672 1.291 0.827 0.756 0.755
(0.29) (0.33) (0.94) (1.80) (1.18) (1.49) (0.51) (0.78) (2.27)*

93
NDP: Characteristics and labour market transitions

Quarter 2 0.761 0.818 1.374 0.502 0.697 0.841 1.049 1.207 0.932
(0.83) (0.65) (0.63) (1.67) (1.25) (1.17) (0.14) (0.54) (0.61)
Quarter 3 0.914 0.959 2.912 0.756 0.704 0.819 1.066 1.374 0.985
(0.29) (0.13) (2.30)* (0.76) (1.23) (1.34) (0.18) (0.94) (0.13)
Quarter 4 0.693 0.866 1.725 0.754 0.658 0.782 0.988 0.869 0.855
(1.09) (0.44) (1.11) (0.75) (1.46) (1.65) (0.03) (0.37) (1.34)

not observed
in all waves 0.966 0.743 1.294 0.996 1.156 0.889 0.912 1.874 1.116
(0.13) (1.10) (0.71) (0.01) (0.63) (0.93) (0.29) (2.39)* (1.16)
LFS interview
Number 3 0.650 0.970 1.147 0.879 0.732 0.501 0.416 0.803 0.694
(1.50) (0.11) (0.41) (0.41) (1.20) (4.98)** (2.45)* (0.75) (3.56)**
Number 4 0.663 0.727 0.575 0.470 0.777 0.544 0.620 0.553 0.615
(1.20) (0.91) (1.09) (1.67) (0.91) (3.99)** (1.44) (1.70) (4.14)**
Number 5 0.489 0.677 0.774 0.598 0.731 0.434 0.292 0.279 0.542
(1.68) (0.96) (0.51) (1.18) (0.99) (4.74)** (2.64)** (2.78)** (4.65)**

Partner’s 0.969 0.930 1.101 1.207 1.279 1.025 1.210 1.447 1.006
prob. of emp (0.41) (0.95) (0.85) (1.93) (3.82)** (0.75) (2.43)* (4.06)** (0.23)

Observations 755 755 755 2321 2321 2321 7209 7209 7209
Robust z-statistics in parentheses
* significant at 5% level; ** significant at 1% level
Reference categories for the categorical variables: aged 51 or over; no children; no qualifications; left full-time education at 16; claiming no benefit
or IS & IB (for inactive 1), claiming no benefit, IS & IB or other combination (for unemployed and inactive 2); interview in 1996; interview in quarter
1; LFS interview 2.

94
 Bibliography

Bibliography

Allison, P. (1982) Discrete-time methods for the analysis of event histories in

Leinhardt, S. (ed.) Sociological Methodology Jossey-Bass Publishers, San
Francisco.
Arulampalam, W. (2001) Is unemployment really scarring? Effects of
unemployment experiences on wages Economic Journal 111(475): F585-
F606.
Cooper-Green, E. (2001) The relationship between work-rich and workless
households Labour Market Trends 109(12): 547-555.
Davies, R., Elias, P. and Penn, R. (1992) The relationship between a husband’s
unemployment and his wife’s participation in the labour force Oxford Bulletin
of Economics and Statistics 54(2).
Desai, T., Gregg, P., Steer, J. and Wadsworth, J. (1999) Gender and the labour
market in Gregg, P. and Wadsworth, J. (eds.) The state of working Britain
Manchester, Manchester University Press.
Dickens, R., Gregg, P. and Wadsworth, J. (2000) New Labour and the labour
market, Oxford Review of Economic Policy 16(1): 95-113.
Dorsett, R. (2001a) Workless couples: characteristics and labour market
transitions Employment Service Research and Development report ESR79.
Dorsett, R. (2001b) Workless couples: modelling labour market transitions
Employment Service Research and Development report ESR98.
Gregg, P. (2001) The impact of youth unemployment on adult unemployment in
the NCDS Economic Journal 111(475): F626-F653.
Gregg, P., Hansen, K. and Wadsworth, J. (1999) The rise of the workless
household in Gregg, P. and Wadsworth, J. (eds.) The state of working
Britain Manchester, Manchester University Press.
Gregory, M. and Jukes, R. (2001) Unemployment and subsequent earnings:
estimating scarring among British men 1984-94 Economic Journal 111(475):
F607-F625.
Jenkins, J. and Laux, R. (1999) Evaluation of new benefits data from the Labour
Force Survey Labour Market Trends 107(9). 505-515.
Jenkins, S. (1995) Easy estimation methods for discrete-time duration models
Oxford Bulletin of Economics and Statistics 57(1).
Johnson, P. and Reed, H. (1996) Intergenerational mobility among the rich and
the poor: results from the National Child Development Survey Oxford
Review of Economic Policy 7: 127-42.
Layard, R., Nickell, S. and Jackman, R. (1991) Unemployment: macroeconomic
performance and the labour market Oxford University Press.
Machin, Stephen (1998) Childhood disadvantage and intergenerational
transmissions of economic status in Atkinson, A and Hills, J (eds.)
Exclusion, employment and opportunity, LSE, CASE paper 4.
Machin, S. and Manning, A. (1999) The causes and consequences of long-term
unemployment in Europe in Ashenfelter, O. and Card, D. Handbook of
labour economics vol. 3, Elsevier Science.

95
NDP: Characteristics and Labour Market transitions

Mallar, C. (1977) The estimation of simultaneous probability models

Econometrica 45(7).
Verbeek, M. and Nijman, T. (1992) Testing for selectivity bias in panel data
models International Economic Review 33(3).

96