Occasional Paper Series ◆ 34

Financial capability
and wellbeing:
Evidence from the BHPS

Mark Taylor
Stephen Jenkins
Amanda Sacker

Institute for Social and Economic Research

University of Essex
Wivenhoe Park
Essex, UK

May 2009
FSA OCCASIONAL PAPERS IN FINANCIAL REGULATION

Foreword

We are committed to encouraging debate among academics, practitioners and policy-makers in all aspects of
financial regulation. To facilitate this, we are publishing a series of occasional papers in financial regulation,
extending across economics and other disciplines.

These papers cover topics such as the rationale for regulation, the costs and benefits of various aspects of
regulation, and the structure and development of the financial services industry. Since their main purpose is
to stimulate interest and debate, we welcome the opportunity to publish controversial and challenging
material, including papers that may have been presented or published elsewhere.

The main factor in accepting papers, which will be independently refereed, is that they should make
substantial contributions to knowledge and understanding in the area of financial regulation. We encourage
contributions from external authors as well as from within the FSA. In either case, the papers will express
the views of the author and not necessarily those of the FSA.

If you want to contribute to this series, please contact Maria-José Barbero or Peter Andrews at:

The Financial Services Authority,

25 The North Colonnade,
Canary Wharf, London, E14 5HS.

Telephone: +44 (0)20 7066 5808 or 3104

Email: maria-jose.barbero@fsa.gov.uk or peter.andrews@fsa.gov.uk

FSA Occasional Papers are available on our website: www.fsa.gov.uk. We welcome comments on these papers;
please address them to the contacts listed above.
 Table of Contents
1. Summary .................................................................................................................2
1.1 Introduction ...................................................................................................2
1.2 The data..........................................................................................................2
1.3 Summarising BHPS variables relevant to financial capability.......................3
1.4 Constructing an index of financial capability ...............................................3
1.5 Relationships between financial capability and other characteristics .........4
1.6 Relationships between financial capability and psychological wellbeing ....4
1.7 Estimating the effect of financial capability on psychological wellbeing....5
1.8 Summary and conclusions ..............................................................................5
2. Introduction ...........................................................................................................7
3. The data................................................................................................................10
3.1 Measures of financial capability ..................................................................10
3.2 Measures of psychological wellbeing...........................................................11
4. Summarising BHPS variables relevant to the concept of financial capability....14
4.1 Measures of perceived financial wellbeing ..................................................14
4.2 Savings behaviour ........................................................................................17
4.3 Housing payment problems..........................................................................18
4.4 Material wellbeing........................................................................................19
4.5 Summary .......................................................................................................20
5. Constructing an index of financial capability .....................................................21
5.1 Correlations between measures....................................................................21
5.2 Constructing indices of financial capability................................................23
5.3 Validity checks .............................................................................................27
5.4 Adjusting for income....................................................................................32
5.5 Changes in individual financial capability from one year to the next .......36
6. Relationships between financial capability and other characteristics ...............41
6.1 Gender ..........................................................................................................41
6.2 Age ...............................................................................................................42
6.3 Marital status ...............................................................................................43
6.4 Number of children.......................................................................................45
6.5 Household type ............................................................................................47
6.6 Health status ................................................................................................49
6.7 Education levels ...........................................................................................50
6.8 Housing tenure.............................................................................................51
6.9 Current house value .....................................................................................53
6.10 Labour market status ...................................................................................54
6.11 Job type........................................................................................................56
6.12 Income..........................................................................................................57
6.13 Summary .......................................................................................................59
7. Relationships between financial incapability and psychological wellbeing.......61
7.1 Psychological wellbeing in the BHPS ..........................................................61
7.2 Changes in individual psychological wellbeing from one year to the next 65
7.3 Financial capability and psychological wellbeing .......................................69
8 Estimating the effect of financial capability on psychological wellbeing..........77
10 Estimation procedures..................................................................................77
10 Estimation results ........................................................................................79
10 Extensions to the analysis ...........................................................................87
9 Summary and conclusions ....................................................................................91
10 References ........................................................................................................93
11. Appendix...........................................................................................................94

1
1. Summary

1.1 Introduction
This report presents the analysis from the project ‘Financial capability and
wellbeing: Evidence from the BHPS.’ The project focuses on the potential
relationships between people’s ability to manage and take control of their finances
(their ‘financial capability’) and their psychological wellbeing. One of the Financial
Services Authority’s (FSA’s) four statutory objectives, set by the Financial Services
and Markets Act (FSMA) 2000, is to promote public understanding of the financial
system. Under this remit, the FSA leads the UK’s National Strategy for Financial
Capability, which brings together industry, government and the third sector to
deliver financial capability programmes which create confident, capable consumers,
doing so in ways that people will most understand and in places most useful to
them. These programmes target different groups, such as expectant parents,
employees, young people and hard-to-reach groups, with tailored initiatives
designed to improve financial skills. This research explores how levels of financial
capability might relate to and affect psychological wellbeing.

This insight contributes towards a deeper understanding of how financial capability

affects individuals, helping the FSA’s financial capability work to be targeted more
appropriately. Part of this is identifying new ways to work with trusted
intermediaries and stakeholders with an interest in psychological wellbeing,
resulting in a wider reach of financial capability programmes. In addition, the
findings will help to inform the evaluation of policies and programmes against the
statutory objective of promoting public understanding, and the FSA’s strategic aim
of creating more capable and confident consumers.

It is important to note that financial capability is not correlated with income

(Atkinson et al 2006); people across society require financial management skills to
be in control of their money, regardless of how much money they have. Although
financial management is important at any time, in the current economic downturn,
reaching a wide swathe of the population is even more vital, as an increasing
number of people find themselves in difficult financial situations. This research
makes a valuable contribution towards the development of effective policy and
practice aimed at improving the UK’s financial management in this difficult time
and beyond.

We use data from the British Household Panel Survey (BHPS) to construct indices
of financial capability, based on the hypothesis is that there is some underlying
factor (financial capability) which is better captured by reviewing a range of
indicators of a person’s current financial situation than by any of the specific
items of information. We describe how financial capability varies according to
individual and household characteristics. Then we examine in detail the
relationship between financial capability and psychological wellbeing, using
multivariate statistical models.

1.2 The data

This project uses data from the first 16 waves of the BHPS, covering the period
1991–2006. To assess financial capability, we focus on financial variables available
in all 16 waves of the BHPS. These are: the respondent’s current financial situation;
2
change in financial situation in the last year; expected change in financial situation
in the coming year; whether respondent saves; the amount saved per month;
whether the household has problems keeping up with housing payments; whether
such problems have required borrowing; whether such problems have required
cutbacks; whether the household has been more than two months in housing arrears
in the last 12 months; and the number of consumer durables to which the household
has access. The measures of psychological wellbeing that we focus on are the
General Health Questionnaire (GHQ) score, reported life satisfaction, and having a
health problem associated with anxiety or depression.

1.3 Summarising BHPS variables relevant to financial capability

We introduce and describe the variables available at all waves of the BHPS that are
relevant to the concept of financial capability. Some of these are hard measures of
financial wellbeing (such as the ability to keep up with housing payments) while
others are perceptions of the individual respondent (such as perceived current
financial situation). In all cases, the source of information is the respondent. We
distinguish between four main groups of variables related to financial capability:
measures of perceived financial wellbeing; saving behaviour; housing payment
problems; and material wellbeing. On average, the proportion of individuals
reporting that they are ‘living comfortably’ or ‘doing alright’ has been increasing
since 1991, while the fraction reporting financial difficulties has fallen significantly
(the data period ends pre-credit crunch). The proportion of respondents reporting
being worse off financially than one year ago and less optimistic about the future
has been falling since 1991. These perceptions are reflected in other measures of
financial wellbeing, with respondents on average saving more and having access to
more consumer durables over time, and fewer respondents living in households with
housing payment problems over time.

1.4 Constructing an index of financial capability

We examine the degrees of association between the various indicators of financial
capability that are available at all BHPS waves. Analysis of average inter-item and
item-rest correlations indicate that a reliable and consistent index can be
constructed from the following variables:
• perceived current financial situation;
• reporting that financial situation has worsened since last year;
• whether saves;
• has housing payment problems;
• problems required borrowing;
• problems required cutbacks; and
• been at least two months in housing arrears in last 12 months.

We call the resulting index the index of financial incapability. As an alternative

approach, and to check the validity of the index, we add together the number of
financial problems individuals currently face, using information on whether the
respondent:
• is finding their financial situation quite or very difficult;
• is reporting that financial situation has worsened since last year;
• is not currently saving;
• has housing payment problems;
• has had to borrow to meet payments;

3
 • has had to cut back to meet payments; and
• has been at least two months in housing arrears in the last 12 months.

The number of financial problems takes a value between 0 (none of the listed
problems) to 7 (all of the listed problems).

We find that the two summary measures of financial capability are very highly
correlated. Furthermore, we find that the two summary measures are relatively
highly correlated with other financial variables available at intermittent waves of
the BHPS. This suggests that the summary measures are valid and consistent
indicators of financial capability.

We also construct a version of the index of financial incapability that is adjusted for
income and examine how individuals’ financial capability varies over the BHPS
sample period. On average, people’s financial capability improved but at a declining
rate between 1991 and 2006. However, at the individual level, financial capability
fluctuates considerably between one year and the next, presumably in response to
other events in people’s lives.

1.5 Relationships between financial capability and other

characteristics
We introduce the individual and household variables with which we describe
patterns of financial capability. Our indices of financial incapability are significantly
associated with gender, age, marital status, number of children, health, employment
status, job type, housing tenure and income, and also with changes in marital
status, the number of children, health, employment status, housing tenure and
income. In particular we find that people with the highest financial incapability
tend to be young (aged less than 35), divorced or separated, have more than one or
two dependent children, are single, non-elderly, lone parents, in fair or poor health,
live in rented accommodation and are unemployed or economically inactive but
would like a job. In contrast, people with the lowest levels of financial incapability
are, on average, older (aged 55 or above), married or widowed with no dependent
children, in good health, home owners and working in a full-time permanent job.
Taking advantage of the panel nature of the data reveals that getting married,
improvements in health, becoming a home owner and entering work are associated
with an increase in financial capability, while death of a spouse, marital dissolution,
an additional child, deterioration in health and unemployment are associated with a
decrease in financial capability. These findings are consistent with the Financial
Services Authority’s Baseline Survey.

1.6 Relationships between financial capability and psychological

wellbeing
Average psychological wellbeing fell between 1991 and 2006 using all three
measures of wellbeing. Average GHQ scores increased by 5% indicating higher
mental stress; the proportion of the sample experiencing anxiety or depression also
increased; and average life satisfaction scores fell by 1%. We also find that these
measures of psychological wellbeing are highly correlated.

However, at the individual level there is considerable change in reported

psychological wellbeing between one year and the next. In addition, individuals

4
whose financial capability varies a lot from year to year also have high year-on-year
variability in psychological wellbeing.

There is a strong association between financial capability and psychological

wellbeing and also between changes in financial capability and changes in
psychological wellbeing. We find that greater financial incapability is associated
with greater mental stress, lower reported life satisfaction, and a greater likelihood
of reporting health problems associated with anxiety or depression.

1.7 Estimating the effect of financial capability on psychological

wellbeing
We estimate the impact of financial capability on psychological wellbeing using
fixed effects panel data models. These allow us to estimate the effects of financial
capability while also taking into account both observable characteristics and time-
invariant unobserved characteristics of individuals (such as personality traits) that
may be related both to an individual’s level of financial capability and their
psychological wellbeing.

Controlling for a range of observable individual and household characteristics, and time-
invariant unobserved effects, people’s financial capability is strongly related to their
psychological wellbeing. For example, moving an individual from relatively low levels of
financial capability (the 90th percentile of the distribution of the index of financial
incapability) to average financial capability levels (the 50th percentile – and therefore
improving their level of financial capability) reduces their GHQ score by about 0.65 GHQ
points (or almost 6%), increases their reported life satisfaction by 0.12 (or 2.4%), and
reduces the probability of an individual suffering a health problem related to anxiety or
depression by 15%. These results are consistent with the hypothesis that changes in
financial capability lead to changes in psychological wellbeing.

Additional analysis shows that the relationship between financial incapability and
psychological wellbeing varies over the distribution of financial incapability: it is
strongest at the bottom of the financial incapability distribution. This implies that
increasing financial capability will improve the psychological wellbeing of most
people, although focusing on those with the highest levels of financial incapability
may have less effect. The impact of financial capability on psychological wellbeing
also differs across different population groups. In particular, financial incapability
compounds the already psychologically harmful effects of unemployment or divorce,
while being in good health or retirement reduces the psychologically damaging
impacts of financial incapability.

1.8 Summary and conclusions

This project has investigated in detail the relationships between financial capability
and psychological wellbeing using data from the first sixteen waves of the BHPS, in
order to contribute towards the FSA’s financial capability policy and programmes.
There is evidence of strong association between both financial capability and
psychological wellbeing, and between changes in financial capability and changes in
psychological wellbeing. Higher financial incapability is associated with higher
mental stress, lower reported life satisfaction, and health problems associated with
anxiety or depression. Estimates from fixed effects panel data models indicate that,
even after controlling for a range of observable individual and household

5
characteristics and time-invariant unobserved effects, people’s financial capability is
a strong predictor of their psychological wellbeing. Moving an individual with
relatively low levels of financial capability to an average level of capability improves
their psychological wellbeing by about 6% (compared to an 8% deterioration in
wellbeing associated with being divorced, and a 10% deterioration from being
unemployed). However, the relationship between financial incapability and
psychological wellbeing varies over the distribution, and in particular is strongest at
the bottom of the financial incapability distribution. Financial incapability
compounds the already psychologically harmful effects of unemployment or divorce,
while being in good health or retirement reduces the psychologically damaging
impacts of financial incapability.

A number of further questions emerge from these analyses. The first is the extent to
which financial capability is related to favourable economic circumstances or by
financial management skills. We have modelled psychological wellbeing as a
function of financial capability (and found a relationship), but have not modelled
the determinants of financial capability itself. If financial capability at the
individual level is highly variable from one year to the next in an unpredictable way,
then this makes it harder to design policies to improve it. The second is the
complex relationship between an individual’s income, their financial management
skills and their savings behaviour. Our analysis touches on this, but it deserves
further attention. Finally, and crucially for our results, is the extent to which people
experience shocks or events that we do not observe that might affect both their
financial capability and psychological wellbeing and confound the effects we found
using statistical models.

6
2. Introduction

This report presents the analysis from the project ‘Financial capability and wellbeing:
Evidence from the BHPS’. The project focuses on the potential relationships between
people’s ability to manage and take control of their finances (their ‘financial
capability’) and their psychological wellbeing. The motivation for this research is to
investigate the extent to which financial capability predicts psychological wellbeing.
A key theme of this research project is the need to develop suitable measures of both
financial capability and wellbeing, together with an understanding of the links
between the two. One of the FSA’s four statutory objectives, set by the Financial
Services and Markets Act (FSMA) 2000, is to promote public understanding of the
financial system. Under this remit the FSA leads the UK’s National Strategy for
Financial Capability, which brings together industry, government and the third sector
to deliver financial capability programmes which create confident, capable consumers,
doing so in ways that people will most understand and in places most useful to them.
These programmes target different groups, such as expectant parents, employees,
young people and hard-to-reach groups, with tailored initiatives designed to improve
financial skills. This research explores how levels of financial capability might relate
to and affect psychological wellbeing.

This insight contributes towards a deeper understanding of how financial capability

affects individuals, helping the FSA’s financial capability work to be targeted more
appropriately. Part of this is identifying new ways to work with trusted
intermediaries and stakeholders with an interest in psychological wellbeing,
resulting in a wider reach of financial capability programmes. In addition, the
findings will help to inform the evaluation of policies and programmes against the
statutory objective of promoting public understanding, and the FSA’s strategic aim
of creating more capable and confident consumers.

It is important to note that financial capability is not correlated with income

(Atkinson et al 2006); people across society require financial management skills to
be in control of their money, regardless of how much money they have. Although
financial management is important at any time, in the current economic downturn,
reaching a wide swathe of the population is even more vital, as an increasing
number of people find themselves in difficult financial situations. This research
makes a valuable contribution towards the development of effective policy and
practice aimed at improving the UK’s financial management in this difficult time
and beyond.

Financial capability may be defined in several ways. The government defines

financial capability as:

“… a broad concept, encompassing the people’s knowledge and skills to understand

their own financial circumstances, along with the motivation to take action.
Financially capable consumers plan ahead, find and use information, know when to
seek advice and can understand and act on this advice, leading to greater
participation in the financial services market” (HM Treasury 2007).

From this it is clear that financial capability should capture a range of skills,
behaviour and knowledge. Atkinson et al (2006) identify five separate strands that
7
contribute to the concept: making ends meet, keeping track, planning ahead,
choosing products and staying informed. The problem as researchers is how to
operationalise this concept using available survey data. Melhuish and Malin (2008)
created a measure of financial capability using survey responses to questions asking
how well individuals are managing financially, how well they manage mortgage or
rent payments, the number of unpaid bills, and the number of items which they
cannot afford. NIACE (2007) stress the importance of defining financial capability in
terms of relating the skills needed to earn income with those needed to manage
savings and consumption. The starting point of this research is to extend existing
knowledge about potential ways of measuring financial capability using responses to
survey questions in the British Household Panel Survey (BHPS).

The analysis follows four distinct steps:

1. identify variables available in BHPS data relevant to the concept of financial

capability;
2. test for the possibility that all, or some subset of, the identified variables
might be combined into a single index measuring a single and common
factor (‘financial capability’);
3. establish how this index is distributed across different groups in the
population and how this changes over the BHPS sample period;
4. examine the relationship between the index of financial capability and
psychological wellbeing and, using suitable multivariate analysis and panel
data models, explore the existence of a causal relationship.

This report summarises the results from each step. The first half of the report
focuses on Steps 1 and 2 – identifying variables in the BHPS data that are relevant
to the concept of financial capability, and then testing which of these might be
combined into a single index that measures financial capability. The latter provides
an indication of the relative importance of relevant variables as contributors to the
underlying concept of financial capability. We might conclude that some variables
do not contribute to that concept at all, having failed the test of being ‘linked and
mutually reinforcing’. We test for the possibility that some or all of the variables
might be combined into a single index. At its simplest, this index might be a
straightforward count across a number of variables, or instead an index created
using inter-item correlations. The hypothesis is that there is some underlying factor
(financial capability) which is better captured by reviewing a range of indicators of
a person’s current financial situation than by any of the specific items of
information. If so, we might expect financial capability to be more stable than any
of the single indicators – people could make short term moves in and out of housing
payment problems (or saving, or being unable to afford certain items, and so on)
without having much effect on their overall level of financial capability.

The second half of the report focuses on Steps 3 and 4 – investigating the relationship
between our measure of financial capability and a range of individual and household
characteristics, including psychological wellbeing. We summarise how the index is
distributed across different groups in the population, and over time. We provide
summaries of the index by a range of individual and household characteristics including
age, gender, marital status, number of children, health status, employment status and
housing tenure. This highlights whether high financial capability is associated with
particular subgroups in the population. We then examine the relationship between

8
financial incapability and psychological health, and whether the index of financial
capability has any power in predicting psychological wellbeing. We do this by first
describing the relationships between our measures of financial capability and
psychological wellbeing, before estimating multivariate statistical models that help to
control for potentially confounding and mediating factors that affect both financial
capability and psychological wellbeing. Our results suggest that people’s financial
capability is a strong predictor of their psychological wellbeing, although the size of
the effect depends on their initial level of financial capability and on a number of other
individual and household characteristics.

The report is divided into three main sections. Section 2 introduces the data set
used in the project (the BHPS), the variables that may be related to the concept of
financial capability, and measures of psychological wellbeing. Section 3 summarises
variables related to the concept of financial capability, and describes patterns in
responses over time. Section 4 examines how the measures of financial capability
are associated with each other, and investigates the possibility of creating an
overall index of financial capability. Section 5 summarises relationships between
financial capability and a range of individual and household characteristics, while
Section 6 provides an initial analysis of how psychological wellbeing relates to
financial capability. Section 7 investigates the strength of this relationship when
controlling for potentially confounding and mediating factors through multivariate
analysis. Section 8 summarises and draws some conclusions.

9
3. The data

In this section we introduce the data, the variables that may be relevant to the
concept of financial capability and available measures of psychological wellbeing.
This project uses individual-level data from the first sixteen waves of the British
Household Panel Survey (BHPS), covering the years 1991–1996. Every year the BHPS
follows and interviews the same adults (aged 16 and above), collecting information
about their incomes, labour market status, housing tenure and conditions,
household composition, education, health and many other aspects of people’s lives.
The BHPS is unique among British surveys in having annual snapshots on the details
of people’s lives over a relatively long time period. Changes in people’s lives can be
identified over a 15-year period.

3.1 Measures of financial capability

There is a range of variables within the BHPS that capture different dimensions of
financial capability, and for each the source of information is the respondent. These
variables, together with their availability in the BHPS, are described in Table 1 below.

Some of these measures relate specifically to individual adults (e.g. How well would
you say you yourself are managing financially these days? Would you say you are
living comfortably, doing alright, just about getting by, finding it quite difficult, or
finding it very difficult?), while others refer to the household context (e.g. Many
people these days are finding it difficult to keep up with their housing payments. In
the last 12 months would you say you have had any difficulties paying for your
accommodation?). In all of the following, the unit of analysis is the individual
adult, though sometimes the personal measure refers to the household context – we
have allocated the household level variable to each individual adult living within
that household.

Also, a number of variables of interest are not available at every BHPS wave. This
raises potential problems for constructing a consistent measure of financial
capability that is available each year. Initially, therefore, we focus on variables that
are available at all BHPS waves (the first 10 variables, in Panel A of Table 1), and
then examine how any resulting index correlates with other relevant variables
collected intermittently over the sample period (the following 10 variables, in Panel
B of Table 1). The latter is carried out to help validate the reliability and robustness
of the index.

10
 Table 1: Financial capability: Relevant BHPS variables Waves available
in BHPS
PANEL A
Many people these days are finding it difficult to keep up with their housing payments. In the last All
12 months would you say you have had any difficulties paying for your accommodation?
Did you have to borrow in order to meet housing payments? All
Did you have to make cutbacks in order to meet housing payments? All
In the last 12 months have you ever found yourself more than two months behind with your All
rent/mortgage?
How well would you say you yourself are managing financially these days? Would you say you are All
living comfortably, doing alright, just about getting by, finding it quite difficult, or finding it
very difficult?
Would you say that you yourself are better off, worse off or about the same financially than you All
were a year ago?
Looking ahead, how do you think you yourself will be financially a year from now, will you be All
better than now, worse than now, or about the same?
Do you save any amount of your income, for example by putting something away now and then in All
a bank, building society, or Post Office account, other than to meet regular bills?
About how much on average do you manage to save a month? All
Access to consumer durables (colour TV, VCR, washing machine, dishwasher, microwave, home All
computer, compact disc player).
PANEL B
Do you or anyone in your household have to make repayments on hire purchases or loans? Please 5 onwards
do not include mortgage loans but do include DSS social fund loans.
To what extent is the repayment of such debts and the interest a financial burden on your 5 onwards
household? Would you say it is a heavy burden, somewhat of a burden, not a problem?
Townsend/Breadline Britain-type indicators (keep home adequately warm; pay for annual holiday; 6 onwards
replace furniture; buy new clothes; eat meat on alternate days; feed visitors once a month; would
like to keep home warm; would like to pay for annual holiday; would like to replace furniture;
would like to buy new clothes; would like meat on alternate days; can’t afford visitors once a
month).
I would like to ask you now about any other financial commitments you may have apart from 5, 10, 15
mortgages and housing related loans. Do you currently owe any money on the things listed on
this card: Hire purchase agreements, personal loans, credit cards, mail order purchase, DSS social
fund loan, loans from an individual?
About how much in total do you owe? 5, 10, 15
Do you currently have any money in any of the investments shown on this card? National Savings 5, 10, 15
Certificates, Premium bonds, Unit trusts, Personal Equity Plans, Shares, National Savings/Building
Society/Insurance bonds?
Thinking of all your investments, about how much do you have invested in total? 5, 10, 15
Would you say your savings are mainly long-term savings for the future or mainly short-term 10 onwards
savings for things you need now and for unexpected events?
Do you save on a regular basis or just from time to time when you can? 10 onwards
Thinking first about your savings accounts, TESSA or ISA, about you much do you currently have 10, 15
in total in these accounts?

3.2 Measures of psychological wellbeing

The BHPS has collected five different measures of psychological wellbeing to date.
These are:
- the General Health Questionnaire (GHQ), asked at all waves;
- whether a respondent suffers from a health problem related to anxiety or
depression (asked at all waves as part of a battery of questions about current
health problems);
- the SF-36 at waves 9 (1999) and 14 (2004);

11
 - life satisfaction scores collected at waves 6-10 (1996-2000) and waves 12-16
(2002-2006); and
- the CASP-19 scale at waves 11 (2001) and 16 (2006).

Our focus is on the relationships between financial capability and psychological

wellbeing, with particular interest on the dynamics of any relationship and whether
the index of financial capability has any power in predicting psychological
wellbeing. To be able to address these issues requires repeated observations of the
same measures for the same individuals over time. For this reason, our analysis of
psychological wellbeing concentrates on the GHQ measure (available at all waves),
whether a respondent suffers from a health problem related to anxiety or depression
(available at all waves), and overall life satisfaction scores (collected at waves 6-10
and waves 12-16).

The GHQ is one of the most widely applied self-completion assessment measure of minor
psychiatric morbidity in the UK (Goldberg & Williams 1988; McCabe et al 1996). It is a
reliable indicator of psychological distress (Argyle 1989), widely used in medical literature
(Goldberg 1972, 1978). The 12-item GHQ score has been used in all waves of the BHPS.
The items take the form of responses to the following questions:

“Have you recently:

1. Been able to concentrate on whatever you are doing?*
2. Lost much sleep over worry?
3. Felt that you are playing a useful part in things?*
4. Felt capable of making decisions about things?*
5. Felt constantly under strain?
6. Felt you couldn’t overcome your difficulties?
7. Been able to enjoy your normal day-to-day activities?*
8. Been able to face up to your problems?*
9. Been feeling unhappy and depressed?
10. Been losing confidence in yourself?
11. Been thinking of yourself as a worthless person?
12. Been feeling reasonably happy all things considered?*”

Answers are coded on a four-point scale running from ‘Not at all/Much less than
usual’ (coded 0) to ‘Much more than usual/Better than usual’ (coded 3 - asterisked
questions are coded in reverse), and added together provide a total GHQ score of
mental distress ranging from 0 to 36. High scores correspond to low feelings of
wellbeing (high stress) and vice-versa. This is sometimes known as a Likert scale.
The GHQ in the BHPS has been shown to be robust to retest effects making it a
1
suitable longitudinal instrument (Pevalin, 2000). This is our primary measure of
psychological wellbeing.

Every wave, BHPS respondents were also shown a card with various health
conditions and asked whether they had any of the health problems listed on it. The

1
More recently, several papers have been published on alternative scoring schemes for the GHQ to
measure positive wellbeing rather than mental distress (Huppert & Whittington, 2003; Hu et al, 2007). An
alternative ‘caseness’ measure could be used, which takes a value of 0-12 and indicates the number of
items with which an individual ‘strongly agrees’ with each statement. We choose to use the 36-point Likert
measure because it is more appropriate to view it as a continuous measure of wellbeing. However all
results presented in this report are robust to using the alternative ‘caseness’ measure.
12
condition most related to psychological wellbeing relates to health problems
associated with anxiety and depression.

Our third measure of psychological wellbeing relates to overall life satisfaction. In

particular, at waves 6–10 and waves 12–16 respondents were asked “How
dissatisfied or satisfied are you with.........your life overall?” using a seven point
scale where one equates to not satisfied at all and seven to completely satisfied.

In subsequent sections we summarise responses over time to financial questions

asked in the BHPS, and the relationships between them to investigate the potential
for constructing an index of financial capability. An analysis of these data over the
BHPS sample period allows us to examine how patterns of financial capability have
evolved over time. We then examine how movements in an individual’s financial
capability relate to changes in their wellbeing.

13
4. Summarising BHPS variables relevant to the concept of
financial capability

In this section we introduce, describe and summarise the variables available at every
wave of the BHPS that are relevant to the concept of financial capability. Some of
these are hard measures (such as the ability to keep up with housing payments) while
others relate to the perceptions of the individual adults (such as perceived current
financial situation). Here we treat the BHPS data as a series of cross-sections and do
not make use of the panel nature of the data – we do that in later stages of the
analysis. We distinguish between four main groups of variables related to financial
capability – measures of perceived financial wellbeing; saving behaviour; housing
payment problems; and material wellbeing. We describe responses to such questions
in detail, and examine how patterns in responses have changed over the sample
period. We use the Pearson chi-squared statistic to test the null hypothesis that the
responses to each survey question are independent over time.2 The value of the chi-
squared statistics cannot be compared across tables, although the reported level of
statistical significance indicates whether the null hypothesis of no association can be
rejected. In all tables the data have been weighted to take account of potential non-
random attrition and non-random response (using weighting variable wXRWGHT). In
the analysis we include all adult (aged 16 and above) respondents, irrespective of
age, and focus on adults who provide non-missing responses to the variables of
interest. This yields a sample size of 16,598 adults contributing 124,940 person-year
observations. We provide sample sizes by wave and gender in the Appendix and do
not show them in each table for brevity.

4.1 Measures of perceived financial wellbeing

At each date of interview, respondents are asked ‘How well would you say you
yourself are managing financially these days? Would you say you are living
comfortably, doing alright, just about getting by, finding it quite difficult, or
finding it very difficult?’ This relates to what Atkinson et al (2006) identify as the
‘keeping track’ and ‘making ends meet’ strands of the concept of financial capability.
Table 2 summarises responses to this question over the sixteen available waves.

This table indicates that on average over the sample period, almost two thirds of
BHPS respondents report either living comfortably or doing alright, and that this
proportion has increased significantly. For example, in 1991, 54.2% of respondents
reported either living comfortably or doing alright, while this had increased to
71.4% in 2006. The proportion reporting finding it quite or very difficult has fallen
correspondingly from 13.5% in 1991 to 6.6% in 2006. Most of these changes
occurred during the 1990s, with little systematic movement since 1999.

2
These statistics take into account the clustering of individuals within households.
14
 Table 2: Perceived current financial situation by year: BHPS 1991–2006
Year Living Doing Just Finding it Finding it very
comfortably alright getting quite difficult difficult
by
1991 0.265 0.277 0.323 0.087 0.048
1992 0.254 0.290 0.323 0.085 0.048
1993 0.256 0.303 0.321 0.084 0.036
1994 0.270 0.314 0.311 0.072 0.033
1995 0.265 0.329 0.308 0.070 0.029
1996 0.282 0.351 0.282 0.058 0.027
1997 0.313 0.351 0.258 0.053 0.026
1998 0.328 0.358 0.245 0.052 0.017
1999 0.315 0.368 0.248 0.049 0.020
2000 0.300 0.378 0.255 0.050 0.018
2001 0.327 0.391 0.222 0.044 0.017
2002 0.324 0.399 0.221 0.041 0.015
2003 0.337 0.399 0.216 0.035 0.013
2004 0.329 0.397 0.215 0.042 0.016
2005 0.304 0.408 0.229 0.042 0.018
2006 0.320 0.394 0.220 0.046 0.020
Total 0.298 0.355 0.264 0.058 0.026
Notes: Weighted using cross-sectional respondent weights. Table reads, for example, that in 1991 26.5% of
respondents reported that they were living comfortably. Pearson χ2 = 42.3 P=0.0000. ‘Total’ shows
data pooled from waves 1 to 16.

As well as being asked about their current financial situation, BHPS respondents are asked
to evaluate the perceived change in their finances over the previous year. In particular,
they are asked ‘Would you say that you yourself are better off, worse off, or about the same
financially than you were a year ago?’ Again, this relates to the ‘keeping track’ strand of
financial capability. Table 3 summarises responses to this question.

15
 Table 3: Change in financial situation since one year ago: BHPS 1991–2006
Year Better off About the Worse off
same
1991 0.234 0.478 0.288
1992 0.214 0.480 0.306
1993 0.251 0.428 0.321
1994 0.248 0.440 0.312
1995 0.269 0.451 0.280
1996 0.286 0.472 0.242
1997 0.309 0.468 0.224
1998 0.300 0.479 0.221
1999 0.294 0.499 0.208
2000 0.305 0.482 0.213
2001 0.311 0.491 0.198
2002 0.285 0.509 0.207
2003 0.279 0.504 0.217
2004 0.272 0.520 0.208
2005 0.262 0.508 0.230
2006 0.262 0.517 0.222
Total 0.274 0.482 0.245
Notes: Weighted using cross-sectional weights. Table reads, for
example, that in 1991 23.4% of respondents reported that they
were better off financially than last year. Pearson χ2 = 37.1
P=0.0000. ‘Total’ shows data pooled from waves 1 to 16.

This table indicates that the proportion reporting being better off than one year ago
increased consistently throughout the 1990s, from 23.4% in 1991 to 30.5% in 2000.
However, this has since fallen to 26.2% in 2006. There was a corresponding fall
(and subsequent increase) in the proportion reporting being worse off that one year
ago, while approximately half of all respondents report their financial situation as
being about the same.

The final question on respondents’ perceived financial wellbeing relates to the expected
change in their financial situation of the coming year. In particular, respondents are
asked ‘Looking ahead, how do you think you yourself will be financially a year from
now, will you be better than now, worse than now, or about the same?’ In contrast to
the previous questions, this variable may relate to the ‘planning ahead’ strand of
financial capability. Table 4 summarises responses to this question.

This table indicates that there has been little change in the proportion of
respondents who expect to be better off financially (which has averaged 27.7% over
the sample period). However there has been an increase in the proportion reporting
that their financial situation in one year’s time will be about the same as now (from
55.2% in 1991 to 62.1% in 2006), and a fall in that reporting that their financial
situation will be worse (from 16.1% in 1991 to 10.8% in 2006).

We construct three summary variables from the subjective measures of financial

wellbeing in order to simplify the construction of an index.3 The first is a variable
(‘financial situation’) which takes the value 1 if the individual reports living
3
We have experimented with several different combinations and definitions of these subjective
variables, but these proved to have the highest correlations with the other financial variables.
16
comfortably, 2 if doing alright, through to 5 if the individual reports finding it very
difficult. Therefore high values of this variable correspond to low financial
capability. The second is a variable (‘situation worsened’) which takes the value 1 if
the individual reports that he/she is worse off financially today than one year ago,
and zero otherwise. The third is a variable (‘expect to worsen’) which takes the value
1 if the individual expects his or her financial situation to worsen in the coming 12
months. These contribute to what Atkinson et al (2006) identifies as the ‘keeping
track’, ‘making ends meet’ and ‘planning ahead’ strands of financial capability.

Table 4: Expected change in financial situation over coming year: BHPS 1991–
2006
Year Better off About the Worse off
same
1991 0.287 0.552 0.161
1992 0.239 0.543 0.219
1993 0.255 0.536 0.209
1994 0.259 0.575 0.166
1995 0.273 0.589 0.138
1996 0.286 0.594 0.120
1997 0.285 0.615 0.099
1998 0.298 0.607 0.095
1999 0.295 0.609 0.096
2000 0.292 0.621 0.087
2001 0.272 0.638 0.090
2002 0.277 0.638 0.085
2003 0.278 0.635 0.088
2004 0.286 0.618 0.097
2005 0.285 0.603 0.112
2006 0.271 0.621 0.108
Total 0.277 0.599 0.124
Notes: Weighted using cross-sectional weights. Table reads, for example, that in
1991 28.7% of respondents reported that they expected to be better off financially
in a year from now. Pearson χ2=60.8 P=0.0000. ‘Total’ shows data pooled from
waves 1 to 16.

4.2 Savings behaviour

Two questions related to savings behaviour were asked at all available waves of the
BHPS. The first is related to whether or not respondents are able to save some of
their income, while the second relates to the average amount saved per month. In
particular, respondents are asked ‘Do you save any amount of your income for
example by putting something away now and then in a bank, building society, or
Post Office account other than to meet regular bills?’, and ‘About how much on
average do you manage to save a month?’ These variables contribute to the
‘planning ahead’ and ‘making ends meet’ strands of financial capability.

Table 5 summarises responses to these questions, reporting whether or not

respondents report saving, the amount saved averaged across the sample as a whole
(where non-savers are given a value of 0) and the amount conditional on saving.
The amount saved has been deflated to January 2006 prices to allow a more direct
comparison over time.

17
 Table 5: Saving behaviour: BHPS 1991–2006
Saves Amount Amount saved
saved conditional on saving
Year Yes No (per month) (per month)
1991 0.386 0.614 59.47 154.23
1992 0.369 0.631 57.92 157.04
1993 0.388 0.612 61.61 158.71
1994 0.384 0.616 63.15 164.40
1995 0.384 0.616 64.68 168.29
1996 0.390 0.610 65.66 168.46
1997 0.407 0.593 67.02 164.64
1998 0.417 0.583 77.92 187.03
1999 0.388 0.612 67.19 173.40
2000 0.403 0.597 68.60 170.29
2001 0.398 0.602 73.70 185.39
2002 0.396 0.604 72.29 182.58
2003 0.390 0.610 74.87 192.01
2004 0.387 0.613 76.75 198.10
2005 0.399 0.601 75.51 189.33
2006 0.381 0.619 75.12 197.33
Total 0.392 0.608 68.58 175.14
Notes: Weighted using cross-sectional weights. Table reads, for example, that in 1991
38.6% of respondents saved from their income; that on average respondents saved £59.47
per month, while those that were able to save on average saved £154.23 per month.
Amounts saved in Jan 2006 prices. Pearson χ2=4.4 P=0.0000. ‘Total’ shows data pooled from
waves 1 to 16.

This table indicates little change in the proportion of respondents who report being
able to save from their income. There is some evidence of an initial increase in the
proportion saving, from 38.6% in 1991 to 41.7% in 1998, but this proportion has
since declined (if not continuously) to 38.1% in 2006. In terms of amounts saved,
there is evidence of a reasonably consistent increase over time, from £59 in 1991 to
£75 in 2006. If we focus only on those that are saving at any particular year, this
increase is more pronounced – increasing from £154 in 1991 to £197 in 2006.

4.3 Housing payment problems

There are four questions asked at each BHPS wave that relate to difficulties in
meeting housing payments. These are asked of only one individual per household
(normally the head of household – the individual mainly responsible for paying for
housing), but for the purposes of this analysis we have allocated the response to all
adult household members.

Households are asked ‘Many people these days are finding it difficult to keep up
with their housing payments. In the last 12 months would you say you have had any
difficulties paying for your accommodation?’ Households who say yes are
subsequently asked ‘Did you have to borrow in order to meet housing payments?’,
‘Did you have to make cutbacks in order to meet housing payments?’ and ‘In the last
twelve months have you ever found yourself more than two months behind with
your rent/mortgage?’ These clearly relate to the ‘making ends meet’ strand of
financial capability. Responses to these questions are summarised in Table 6.

18
 Table 6: Housing payment problems: BHPS 1991–2006
Housing payment Required Required Been 2+ months
problems borrowing cutbacks in arrears
Year Yes Yes Yes Yes
1991 0.133 0.027 0.112 0.035
1992 0.123 0.023 0.103 0.028
1993 0.108 0.021 0.092 0.025
1994 0.090 0.013 0.071 0.020
1995 0.076 0.013 0.061 0.016
1996 0.064 0.011 0.052 0.011
1997 0.065 0.014 0.052 0.012
1998 0.059 0.012 0.048 0.009
1999 0.053 0.010 0.040 0.008
2000 0.060 0.013 0.045 0.008
2001 0.047 0.011 0.035 0.009
2002 0.046 0.012 0.037 0.008
2003 0.046 0.013 0.036 0.007
2004 0.040 0.013 0.029 0.011
2005 0.051 0.012 0.041 0.011
2006 0.041 0.016 0.040 0.009
Total 0.071 0.015 0.057 0.014
2
Pearson χ (p-value) 51.67 (0.0000) 6.4 (0.0000) 47.5 (0.0000) 18.6 (0.0000)
Notes: Weighted using cross-sectional weights. Table reads, for example, that in 1991 13.3% reported
having problems meeting housing payments, 2.7% had to borrow to meet payments, 11.2% had to make
cutbacks to meet payments, while 3.5% were at least 2 months in arrears with their payments in the last
12 months. ‘Total’ shows data pooled from waves 1 to 16.

The table shows that in general meeting housing payments became less of a problem
over the sample period. For example, the proportion of respondents in households
reporting having problems meeting their housing payments fell from 13.3% in 1991
to less than 5% in 2006 (although the minimum was 4% in 2004). There were
similar falls in the proportions reporting having to borrow or make cutbacks in order
to meet their housing payment problems, from 2.7% to 1.6% and from 11.2% to 4%
respectively. The proportion of respondents living in households that were two or
more months in housing arrears at anytime in the last 12 months fell from 3.5% in
1991 to less than 1% in 2006.

4.4 Material wellbeing

At each date of interview, respondents are asked a series of questions relating to
whether they, in their current accommodation, have access to a number of different
consumer durables – a colour television, a video cassette recorder (VCR), washing
machine, dishwasher, microwave oven, home personal computer (PC) and a compact
disc (CD) player. Rather than examine the extent to which respondents had access
to each consumer durable, we have combined these indicators into a summary
measure that simply counts the number of consumer durables to which an individual
has access. This variable therefore takes a value between 0 and 7. Responses to this
set of questions provide an insight into the respondents’ standard of living, and may
contribute to the ‘making ends meet’ strand of financial capability. Responses are
summarised in Table 7.

19
 Table 7: Number of consumable durables: BHPS 1991–2006
Year <3 3 4 5 6 7
1991 0.188 0.165 0.262 0.220 0.127 0.038
1992 0.167 0.141 0.258 0.238 0.139 0.057
1993 0.142 0.127 0.244 0.248 0.167 0.072
1994 0.125 0.113 0.226 0.260 0.183 0.093
1995 0.114 0.096 0.211 0.273 0.200 0.106
1996 0.090 0.093 0.198 0.275 0.227 0.117
1997 0.080 0.077 0.186 0.279 0.242 0.136
1998 0.065 0.074 0.169 0.279 0.254 0.159
1999 0.063 0.059 0.155 0.264 0.276 0.183
2000 0.047 0.060 0.140 0.246 0.285 0.222
2001 0.040 0.052 0.131 0.236 0.300 0.241
2002 0.032 0.045 0.112 0.218 0.309 0.284
2003 0.028 0.037 0.101 0.203 0.316 0.315
2004 0.025 0.035 0.091 0.185 0.323 0.341
2005 0.025 0.029 0.078 0.174 0.341 0.353
2006 0.022 0.029 0.081 0.164 0.345 0.359
Total 0.080 0.079 0.168 0.237 0.249 0.187
Notes: Weighted using cross-sectional weights. Table reads, for example, that in 1991 18.8% of
respondents lived in a household with access to less than 3 consumer durables. Pearson χ2=121.8
P=0.000. ‘Total’ shows data pooled from waves 1 to 16.

This table reflects the general increase in living standards over the sample period,
with a significant increase in the number of consumer durables to which
respondents had access. The proportion with access to fewer than three of the listed
consumer durables has fallen from 18.8% in 1991 to just 2.2% in 2006, while that
with access to all seven has increased from just 3.8% in 1991 to 36% in 2006.

4.5 Summary
In this section we have summarised variables that are available at all BHPS waves
and that may be related to the concept of financial capability. All contribute to the
different strands of financial capability identified by Atkinson et al (2006). They
include measures of perceived financial wellbeing, savings behaviour, problems
meeting housing payments and material wellbeing. On average, the proportion of
individuals reporting ‘living comfortably’ or ‘doing alright’ has been increasing since
1991, while the fraction reporting financial difficulties has fallen significantly. The
proportion of respondents reporting being worse off financially than one year ago
and less optimistic about the future has been falling since 1991. These perceptions
are reflected in other measures of financial wellbeing, with respondents on average
saving more and having access to more consumer durables, and fewer respondents
living in households with housing payment problems. In subsequent sections, we
use responses to these variables to construct an index of financial capability and
then examine correlations between this index and other financial variables that are
available at intermittent waves of the BHPS.

20
5. Constructing an index of financial capability
Having described variables available at every BHPS wave that may be related to financial
capability, we now turn our attention to the degrees of association between these variables.
The ultimate aim is to examine the possibility of constructing an index of financial capability.
This involves experimenting with a number of different ways of combining information
collected in responses to BHPS survey questions on financial wellbeing described previously. A
necessary first stage in this process is to examine the degree of correlation between responses
to each question. A simple way of constructing an index would then be to simply sum
variables with a high degree of correlation to provide a straightforward measure of financial
capability (e.g. Taylor et al 2004). Another popular way of constructing an index is to employ
factor analysis (or principal component analysis) which uses correlations between variables to
determine the underlying factor (in this case financial capability) represented by the variables
(e.g. Taylor et al 2004; Capellari and Jenkins 2007). This method allows us to construct a
factor score for each individual that measures the particular combination and weighting of
variables used. We adopt both procedures.

5.1 Correlations between measures

As a first step in developing an index, we present a correlation matrix which
illustrates the degree of association between the available variables, shown in Table
8 below. Here we have pooled all 16 waves of BHPS data, as our interest is in
constructing an index of financial capability that can be applied across the whole
sample period (rather than examining changes in associations over time). The
statistic reported is the Spearman rank correlation coefficient, which is a measure of
association taking a value between –1 (indicating perfect negative correlation) and
+1 (indicating perfect positive correlation).4 A value of zero indicates no correlation
between the relevant variables. This table can be used to examine the degree of
association between variables, allowing us to identify variables that are likely to be
capturing a common underlying factor (financial capability). Variables that have the
closest association (with rank correlation coefficients of 0.3 and above) are
highlighted in bold. Those with correlation coefficients between 0.1 and 0.29 are in
normal print, while those with the weakest association are in grey. By construction,
the matrix is symmetrical around the lead diagonal.

This table shows that the strongest correlations (of above 0.3) are found between
an individual’s perceived current financial situation and their savings behaviour, and
between an individual’s perceived current financial situation and reporting that
their situation has worsened over the previous 12 months. This suggests that people
reporting finding it difficult to get by financially are also more likely to report a
worsening financial situation, and are less likely to save. (We’ve standardised the
correlations with the saving behaviour and consumer durables variables so that the
positive correlations here indicate that individuals in a difficult financial situation
are less likely to save and have access to fewer consumer durables.) Other strong
correlations are found between the housing payment variables, which are to be
expected given the structure of these questions.

4
We use the Spearman rank correlation coefficient rather than the more common Pearson’s
correlation coefficient because the former is non-parametric and less likely to be distorted when the
normality assumption does not hold.
21
 Table 8: Correlations between financial variables: BHPS 1991–2006

Financial Worsened Expect to Saves (-) Amount Housing Required Required Arrears Durables
situation worsen saved (-) payment borrowing cutbacks (-)
problems
Financial situation 1.00 0.35 0.10 0.29 0.33 0.26 0.13 0.24 0.13 0.16
Situation worsened 1.00 0.22 0.14 0.15 0.14 0.08 0.14 0.06 0.05
Expect to worsen 1.00 0.02 0.02 0.02 0.00 0.02 0.00 0.07
Saves (-) 1.00 - 0.12 0.07 0.11 0.07 0.14
Amount saved (-) 1.00 0.13 0.07 0.11 0.07 0.16
Housing payment problems 1.00 0.46 0.89 0.44 0.08
Required borrowing 1.00 0.42 0.27 0.04
Required cutbacks 1.00 0.41 0.08
Arrears 1.00 0.05
Number of durables (-) 1.00
Mean 0.22 0.15 0.05 0.12 0.13 0.28 0.17 0.27 0.17 0.09
Notes: Figures reported are Spearman rank correlation coefficients. See text for how variables are constructed and defined.

22
The final row of the table shows the average correlation between each variable and the
others. This indicates that the variables most highly correlated with the others are
perceived current financial situation, having housing payment problems and housing
payment problems required cutbacks. It is clear that expecting one’s financial position
to worsen over the coming year has little correlation with the other variables, and for
this reason we discard it from the remainder of the analysis. This lack of correlation is
explained by the fact that individual’s expectations about changes in their financial
situation can be independent of their current financial situation. We now use the
remaining variables to construct an index of financial capability.

5.2 Constructing indices of financial capability

We adopt two approaches to constructing an index of financial capability, based on
the correlations presented in Table 8. The first approach uses factor analysis. The
second approach sums the variables with a relatively high degree of correlation to
provide a straightforward and easily interpretable measure of financial capability.
The latter is a commonly used procedure in the deprivation and hardship literature,
and often appears to work at least as well as much more complicated methodologies
(Skrondal and Rabe-Hesketh 2004). We describe the procedure used in constructing
each of the indices in detail below.

Identifying the common characteristic

Our aim is to construct an index of financial capability that can be traced over time.
The individual variables can be interpreted as reflecting a common, underlying
characteristic (‘financial capability’) if there is a consistent tendency for an individual
who scores highly on one also to score highly on each of the other variables. We test
the internal consistency of such summary measures using Cronbach’s alpha which is
calculated on the basis of the number of contributing variables and the correlations
between them. Alpha takes a value between 0 and 1, with one indicating perfect
internal consistency. The literature suggests that a good summary indicator should have
a value of alpha of at least 0.7 (Nunnally and Bernstein 1994). Before constructing an
index, we examine the inter-item correlations, which we present in Table 9 below.
Because some of the variables have different scales (e.g. perceived current financial
situation, amount saved, number of consumer durables), we have standardised all the
variables to have mean zero and variance one.

Table 9: Standardised inter-item correlations: BHPS 1991–2006

Variable Item-rest Average inter- Alpha
correlation item correlation
Financial situation 0.451 0.180 0.638
Situation worsened 0.240 0.211 0.667
Saves (-) 0.309 0.200 0.683
Amount saved (-) 0.232 0.212 0.681
Housing payment problems 0.606 0.160 0.603
Required borrowing 0.351 0.194 0.659
Required cutbacks 0.570 0.164 0.611
Arrears 0.342 0.196 0.660
Number of durables (-) 0.151 0.224 0.698
Total 0.193 0.683

23
The item-rest correlation shows the correlation between each variable and the index
that is formed by all the other items, while the average inter-item correlation shows
the inter-item correlations excluding the relevant variable, and therefore indicates
whether or not excluding the relevant variable would increase the average inter-item
correlation. The last column of the table presents Cronbach’s alpha for the index
formed by excluding the relevant variable, and therefore indicates whether the
internal consistency of the index would be improved by excluding the relevant
variable. The results presented in Table 9 indicate that both the amount saved and
the number of durables appear to be least well correlated with the other variables.
They have the lowest item-rest correlation (indicating they are least well correlated
with an index formed by all other items), and the average inter-item correlation and
alpha would both increase if they were removed. This may be because both these
variables reflect income levels as much as financial capability. Therefore it appears
that it is the act of saving itself that is a more important indicator of financial
5
management than the amount saved.

This leaves us with the following variables from which to construct an index:

• perceived current financial situation;

• reporting that the financial situation has worsened since last year;
• whether saves;
• has housing payment problems;6
• problems required borrowing;
• problems required cutbacks; and
• been at least two months in housing arrears in last 12 months.

The internal consistency of such a summary measure yields a Cronbach’s alpha of

0.71 and an average inter-item correlation of 0.26, which suggests it is a good
summary indicator and that the individual variables all contribute to the underlying
financial capability component in the same way. Wave-specific estimates show
Cronbach’s alphas that vary between 0.68 and 0.74, and average inter-item
correlations that vary between 0.23 and 0.29, suggesting that the index has internal
consistency across time. The distribution of the underlying factor score is
summarised in Table 10 and Figure 1. Because this factor is essentially measuring
financial incapability, we call it an index of financial incapability. Higher values of
this index are associated with higher financial difficulty (lower financial capability),
and vice versa.

Figure 1 shows that although there is a long right hand tail to the distribution of
the index, the majority of observations actually lie between –0.537 and zero.
Therefore, consistent with the Financial Services Authority’s Baseline Survey, most

5
We have also experimented with using savings as a proportion of income. However this too is less
well correlated with the underlying factor of financial capability than the act of savings.
6
We have experimented with a number of different combinations of the housing payment problems
variables, including creating a single variable measuring the scale of the problems and including the
separate variables independently of the others. The current specification appears to provide the most
consistent index.

24
people are financially capable but those that are not can suffer extreme difficulties.
Table 10 indicates that the index has a mean of zero and a standard deviation of
0.601 and varies between –0.537 (indicating no financial difficulty) and 4.1
(indicating high financial difficulty).

Figure 1: Distribution of the index of financial incapability: BHPS 1991–2006

20
15
Percent
105
0

-1 0 1 2 3 4
Index of financial incapability

Table 10: Index of financial incapability: BHPS 1991–2006

Mean Std Dev Min Max
Financial difficulty index 0.000 0.601 –0.537 4.100
Notes: Index constructed using factor analysis from: Current financial
situation; Financial situation worsened since last year; Whether saves; Has
housing payment problems; Problems required borrowing; Problems required
cutbacks; and Been at least two months in arrears in last 12 months.

An alternative approach
As an alternative approach, and to check the validity of the index constructed
above, we have constructed a summary measure by simply adding together the
indicators of financial incapability that individuals currently face. Such ‘sum-score’
indices are commonly used in the deprivation and hardship literature. To do this we
have again focused on those variables with high average inter-item correlations:
perceived current financial situation, reporting that the situation worsened, whether
saves, housing payment problems, whether problems required borrowing, whether
required cutbacks, and whether been in housing arrears. First we have used
perceived current financial situation to define as having low financial capability
individuals who are finding it quite difficult or very difficult. We also define as
having low financial capability those who are not currently saving. Then we

25
construct an index by adding together whether the individual: is finding it quite or
very difficult, reports a worsened financial situation, is not currently saving, has
housing payment problems, has had to borrow to meet payments, has had to
cutback to meet payments, and has been in two or months arrears. This index takes
a value between 0 (has none of the listed problems) to 7 (has all of the listed
problems). Table 11 and Figure 2 summarise the distribution of this index which, for
simplicity, we call the number of financial problems.

Table 11: Number of financial problems: BHPS 1991–2006

0 1 2 3 4 or more Mean
Number of financial problems 0.313 0.448 0.146 0.053 0.040 1.09
Notes: Table reads, for example, that 31.3% had no financial problems. Number of financial problems
is sum of whether individual: is finding it quite or very difficult, has a worsened financial situation,
is not currently saving, has housing payment problems, has had to borrow to meet payments, has had
to cutback to meet payments, has been in two or months arrears, and has access to fewer than three
consumer durables.

This table shows that on average over the sample period, individuals suffered from
1.09 financial problems each year. As with the index of financial incapability, the
distribution of the number of financial problems has a long right hand tail (Figure 2).
Again, most people suffer few problems, but those that do can suffer from extreme
difficulty. More than three quarters of observations had at most one financial
problem, while 15% had two. Only 4% suffered from four or more financial problems.

Figure 2: Distribution of the number of financial problems: BHPS 1991–2006

40
30
Percent
20 10
0

0 2 4 6 8
Number of financial problems

26
5.3 Validity checks
Before taking these two measures onto the next stage of the analysis, we carry out
some validity checks. These take two forms. Firstly we examine the degree of correlation
between our index of financial incapability and the number of financial problems.
Secondly, we examine how each of these measures is correlated with other measures of
financial wellbeing collected intermittently over the BHPS sample period.

Correlations between measures

The first validity check is to ensure that the two measures exhibit high degrees of
association. Table 12 indicates that the mean index of financial incapability
increases monotonically with the number of financial problems. Individuals with no
financial problems have an average index of financial incapability of –0.44. This
increases consistently, such that those with six or seven financial problems have a
mean index of financial incapability exceeding 3. The two constructed measures
have a Spearman rank correlation coefficient of 0.94. This indicates that there is a
very high degree of association between these two indicators of financial capability.

Table 12: Association between number of financial problems and index of

financial incapability: BHPS 1991–2006
Number of financial problems Mean financial incapability
0 –0.443
1 –0.095
2 0.307
3 0.792
4 1.746
5 2.160
6 3.033
7 4.044
Spearman’s rank correlation coefficient 0.943
Notes: Index of financial difficulty constructed from: Current financial situation;
Financial situation worsened since last year; Whether saves; Has housing payment
problems; Problems required borrowing; Problems required cutbacks; and Been at least
two months in arrears in last 12 months. Number of financial problems is sum of
whether individual: is finding it quite or very difficult, has a worsened financial
situation, is not currently saving, has housing payment problems, has had to borrow to
meet payments, has had to cutback to meet payments, has been in two or months
arrears, and has access to fewer than three consumer durables.

As a further check, we have estimated Ordinary Least Squares regressions, with the
number of financial problems as the dependent variable and the index of financial
incapability as the explanatory variable. The estimates from such a regression (not
shown) indicate that the index of financial incapability explains 88% of the total
variance in the number of financial problems. This relationship is highlighted
graphically in Figure 3, which plots the two measures together with a superimposed
fitted regression line. The fitted line does not pass through the centre of the dots,
indicating that the dots are denser at lower values of the index of financial
incapability. Again therefore, there is evidence of a high degree of correlation
between the two measures.

27
As a further validity and robustness check, we examine correlations between our
summary measures of financial capability and the financial variables available
intermittently across BHPS waves. Such variables were not considered in
constructing the indices because they are not available at all survey waves, and
therefore reduce both the time coverage of the index and the number of
observations for which it can be calculated. Before presenting correlations, Table 13
describes the variables concerned.

Figure 3: Plot of index of financial incapability and number of

financial problems: BHPS 1991–2006
4
Index of financial difficulty
0 1 -1 2 3

0 2 4 6 8
Number financial problems

Index of financial difficulty Fitted values

Source: BHPS 1991-2006

28
 Table 13: Variables available at intermittent BHPS waves 1991–2006
Variable name Description BHPS
Availability
Repayments Individual or household has to make repayments on hire purchases or 5 onwards
loans (excluding mortgages).
Repayments burden Is the repayment of such debts and the interest a heavy burden, 5 onwards
somewhat of a burden, not a problem?
Lifestyle The number of the following which the household is able to do: Keep 6 onwards
home adequately warm; pay for annual holiday; replace furniture; buy
new clothes; eat meat on alternate days; feed visitors once a month.
Financial commitments Number of the following financial commitments: Hire purchase 5, 10, 15
agreements, personal loans, credit cards, mail order purchase, DSS
social fund loan, loans from an individual, something else.
Amount of debt The amount owed on the above. 5, 10, 15
Number of investments Which of the following investments individuals have money in: 5, 10, 15
National Savings Certificates, Premium bonds, Unit trusts, Personal
Equity Plans, Shares, National Savings/Building Society/Insurance
bonds, other.
Amount invested How much money invested in the above? 5, 10, 15
Long-term saver Are savings mainly long-term savings for the future? 10 onwards
Regular saver Does respondent save on a regular basis? 10 onwards
Amount in savings accounts How much respondent has in total in savings accounts, TESSAs or 10, 15
ISAs.

These variables capture aspects of individuals’ credit, savings and debt, ranging
from the burdens of debt repayments and financial commitments, to lifestyle
information, investments and savings behaviour. A priori, we would expect any
measure of financial capability to be correlated with at least some of these
variables, as they will also contribute to the different strands identified by Atkinson
et al (2006). As a precursor to examining correlations between these variables and
our two summary measures, in Table 14 we present a correlation matrix which
illustrates the degree of association between the variables (this is symmetrical
about the lead diagonal). Again we have pooled all waves of relevant data (the
number of which vary according to the availability of the variables). Again, the
statistic reported is the Spearman rank correlation coefficient, with variables having
the closest association (with rank correlation coefficients of 0.3 and above)
highlighted in bold. Those with correlation coefficients between 0.1 and 0.29 are in
normal print, while those with the weakest association are in grey. (Again, we’ve
standardised the correlations with lifestyle, investments and saving variables so that
the positive correlations here indicate that individuals in a difficult financial
situation are less likely to have investments and to save).

The table shows that the strongest correlations (of above 0.3) are found between
repaying a loan and the number of financial commitments and amount of debt,
between the number of investments, amount invested and amount in savings
accounts, and between saving regularly, saving long-term and amount in savings
accounts. These results accord with intuition – individuals repaying loans are likely
to have more financial commitments and debt, while those with a larger number of
investments are likely to have more invested. Similarly, individuals who save
regularly and on a long-term basis are likely to have more money in savings
accounts. The average correlations shown in the final row indicate that the amount

29
of money in savings accounts is most highly correlated with the other variables
(average correlation of 0.25). The lifestyle variable (capturing the number of things
the household is able to do) has the weakest correlations with the other variables.

Given the relatively low correlations between many of these variables, we might
expect our two summary measures of financial capability to also be relatively poorly
correlated with these variables. We examine this in Table 15, again presenting
Spearman rank correlation coefficients. These, however, indicate relatively strong
correlations between our measures and the other variables. In particular, our index
of financial incapability and the number of financial problems exhibit relatively high
correlations with the lifestyle measure, being a long-term saver, being a regular
saver and the amount held in savings accounts. Relatively weak correlations emerge
with repaying loans, the number of financial commitments and the amount of debt.

The average Spearman rank correlation coefficients (of 0.237 with the index of
financial incapability and 0.226 with the number of financial problems) are greater
than all but one of the average inter-variable correlations. This indicates that our
summary measures are more highly correlated with these variables than the variables
are correlated between themselves, and gives us confidence that the summary
measures are valid, and consistent, indicators of financial capability.

30
 Table 14: Correlations between financial variables intermittently available: BHPS 1991–2006

Repayments Repayments Lifestyle Financial Amount N Amount LT saver Regular Amount

burden (-) comms debt investments invested (-) saver savings
(-) (-) (-) (-)
Repayments 1.00 0.24 0.03 0.50 0.49 0.11 0.08 0.01 0.04 0.14
Repayments burden 1.00 0.13 0.16 0.14 0.07 0.06 0.03 0.06 0.12
Lifestyle (-) 1.00 0.03 0.02 0.04 0.04 0.08 0.16 0.20
Financial commitments 1.00 0.91 0.07 0.04 0.02 0.02 0.14
Amount of debt 1.00 0.06 0.02 0.01 0.05 0.12
Number of investments 1.00 - 0.16 0.14 0.43
(-)
Amount invested (-) 1.00 0.17 0.14 0.43
Long-term saver (-) 1.00 0.47 0.28
Regular saver (-) 1.00 0.38
Amount in savings 1.00
accounts (-)
Mean 0.18 0.11 0.08 0.21 0.20 0.14 0.12 0.14 0.16 0.25
Notes: Figures reported are Spearman rank correlation coefficients. See text for how variables are constructed and defined.

31
 Table 15: Correlations between measures of financial capability and financial
variables available intermittently: BHPS 1991–2006
Correlation with
Variable Index of Number of financial
financial problems
incapability
Repayments 0.040 0.018
Repayments burden 0.161 0.155
Lifestyle (-) 0.276 0.237
Financial commitments 0.083 0.054
Amount of debt 0.069 0.044
Number of investments (-) 0.211 0.177
Amount invested (-) 0.194 0.167
Long-term saver (-) 0.353 0.385
Regular saver (-) 0.578 0.650
Amount in savings accounts (-) 0.401 0.374
Mean 0.237 0.226
Notes: Figures reported are Spearman rank correlation coefficients. See text for how variables are
constructed and defined.

5.4 Adjusting for income

Of course, financial difficulty is strongly related to income and it can be argued that
any measure of financial capability should be adjusted for income. Financial
capability should capture how capable people are at managing their finances
independent of their income levels. Here we investigate the relationship between
our index of financial incapability and income, defined as real equivalised gross
household income (in the month prior to interview), deflated to January 2006
prices. Our index of financial incapability yields a Spearman rank correlation
coefficient with income of –0.34, suggesting that financial incapability falls as
income increases. Figure 4 provides smoothed plots to highlight the relationships
between income and the index of financial incapability and the number of financial
problems. This indicates that the relationship is stronger (the slopes are steeper) at
lower income levels – financial capability increases with income at a faster rate for
those with higher levels of financial incapability than for those with lower levels.
The lines are relatively flat at higher income levels.

To create an income adjusted measure of financial incapability, we follow the

procedure adopted in Melhuish and Malin (2008) and regress the index of financial
incapability on real equivalised monthly household income (in January 2006 prices)
and use the residuals as our income-adjusted index of financial incapability. The
results from this Ordinary Least Squares (OLS) regression are shown in Table 16. The
residuals from this regression can be interpreted as the part of financial incapability
that is not explained by income, which we call our income-adjusted index of
financial incapability. The relatively small (if statistically significant) coefficients on
the quadratic and cubic terms suggest that the non-linearities in the relationship
between income and financial incapability are small. This is highlighted in Figure 5,
which plots the index of financial incapability, the income-adjusted index, and the
estimated regression line. The closeness of the estimated line to the income-

32
unadjusted index indicates that the income-adjusted and income-unadjusted indices
will only differ at low and very high equivalised household income (below £1,000
and above £6,000 per month). Given that over 80% of income observations lie
within this range, we expect the income-adjusted and the income-unadjusted
indices to provide very similar results. This figure also confirms that income-
adjusted index is unrelated to income.

Figure 4: Relationship between financial incapability and income: BHPS 1991-

2006
2 1.5
Financial incapability
.5 0 1
-.5

0 5 10
Equivalised monthly gross household income (£1000)

Index of financial incapability N financial problems

Source: BHPS 1991-2006

Table 16: OLS Regression of household income on index of financial

incapability: BHPS 1991–2006
Variable Coefficient t-statistic
Real equiv. month household income (£1,000s) –0.1553 33.32
Real equiv. month household income2 (£1,000s) 0.0091 15.24
3
Real equiv. month household income (£1,000s) –0.0001 10.91
Constant 0.2923 34.87
R2 0.0647
N individuals 16598
N observations 124940
Notes: Estimates from ordinary least squares regression where dependent variable is index of
financial incapability. Standard errors adjusted for clustering on individuals.

It is important to note that, according to Figure 5, financial incapability falls much

faster at the lower end of the income scale. For example, an additional £1,000 per
month in household income reduces financial incapability by more for an individual
with a household income of less than £3,000 per month than for one with an
income of more than £5,000 per month. Therefore increasing incomes of those at

33
the bottom of the income distribution will have relatively larger effects on financial
capability than increasing incomes of those at the top of the distribution.

Figure 5: Relationships between financial incapability and income: BHPS 1991-

2006
.5
Financial incapability
-.5 -1 0

0 2 4 6 8 10
Equivalised monthly gross household income (£1000)

Income unadjusted Income adjusted

Regression estimates
Source: BHPS 1991-2006

Table 17 and Figure 6 describe the distribution of the income-adjusted index of

financial incapability. Table 17 shows that the income-adjusted index has a mean of
zero and a standard deviation of 0.582 and varies between –1.978 (indicating no
financial incapability) and 4.4 (indicating high financial incapability). Figure 6
indicates that the income-adjusted index has a long right hand tail (although this is
less pronounced than with the income-unadjusted index) and that the majority of
observations have values between –1 and zero. The clustering of observations at low
levels of financial incapability indicate that most people manage their finances
relatively well, and the long right hand tail indicates that those that have problems
can suffer from extreme difficulty. Our income-adjusted index of financial
incapability has a Spearman rank correlation coefficient of 0.88 with the unadjusted
index, and exhibits an almost identical relationship with the number of financial
problems (not shown).

Table 17: Income-adjusted index of financial incapability: BHPS 1991–2006

Mean Std Dev Min Max
Income-adjusted financial incapability 0.000 0.582 –1.978 4.400
index
Notes: Index constructed using factor analysis from: Current financial situation; Financial situation
worsened since last year; Whether saves; Has housing payment problems; Problems required borrowing;
Problems required cutbacks; and Been at least two months in arrears in last 12 months.

34
Figure 7 below plots the evolution over the 16 years of available BHPS data of the
means of both the income-adjusted and income-unadjusted index of financial
incapability, and the number of financial problems. This shows, first, that all three
measures indicate a decline in average financial incapability from the early 1990s
until 2004, after which there is some evidence of an increase. Second, as expected
given the relationships plotted in Figure 5, the averages in the income-adjusted and
income-unadjusted index are almost identical over time. As we would expect, the
income-adjusted index shows less variation over time, but the differences are small.

Figure 6: Distribution of the income-adjusted index of financial incapability:

BHPS 1991–2006
20
15
Percent
105
0

-2 0 2 4
Income-adjusted index of financial incapability

35
Figure 7: Plot of Index of financial difficulty and number of financial problems:
BHPS 1991–2006
1.5
Financial incapability/problems
.5 0 1

1990 1995 2000 2005

Year

Income adjusted Income unadjusted

N financial problems
Source: BHPS 1991-2006

5.5 Changes in individual financial capability from one year to the

next
Until now, we have analysed the indices of financial incapability from a cross-
sectional perspective. We have not taken advantage of the panel nature of the data
to examine how financial capability changes from one year to the next for each
individual. Table 18 presents our first look at this. In this table we summarise
individuals’ mean financial incapability over two consecutive years, as well as the
average change.

The table indicates that on average over the sample period, people’s financial
incapability fell between one year (“t–1”) and the next (“t”). The mean changes in
the indices were negative, showing that financial capability was improving. For
example, the mean income-adjusted index fell from –0.020 in year t–1 to –0.032 in
year t, while the mean number of financial problems fell from 1.077 to 1.053. This is
consistent with Figure 7 which shows a downward trend in financial incapability
over time. The table also presents average within-individual variances in the indices,
which are very large relative to the means. This indicates a great deal of change in
the indices at the individual level – financial incapability changes considerably
between one year and the next. This is reinforced by a Spearman rank correlation
coefficient between current index of financial incapability and index of financial
incapability one year ago of 0.5. The implication of this is that financial capability
is not a relatively stable characteristic but instead fluctuates considerably at the
individual level presumably in response to other (possibly expected and unexpected)
events that individuals experience. What factors determine this longitudinal flux is
an interesting avenue for further research.
36
Table 18: Within-individual year-on-year changes in financial incapability: BHPS
1991–2006
Financial incapability index Means
t–1 t Change Within-individual
variance
Income-adjusted –0.020 –0.032 –0.012 0.365
Income-unadjusted –0.013 –0.028 –0.015 0.365
Number financial problems 1.077 1.053 –0.024 0.776
N 95935
Notes: Weighted using cross-sectional weights. Table reads, for example, that on average
individuals had an income-adjusted index of financial incapability of –0.02 in year t–1 and of –
0.032 in year t, indicating an average improvement in financial capability of 0.012.

Figure 8 plots the distribution of within-individual year-on-year changes in the

income-unadjusted index of financial incapability. This shows that over 30% had no
change in financial incapability from one year to the next. While this is clearly the
modal value, the figure suggests that in almost 70% of cases, individuals’ financial
capability changed from one year to the next. Figures 9 and 10 reveal a similar
pattern when looking at year-on-year changes at the individual level in the income-
adjusted index and in the number of financial problems.

Figure 8: Within-individual year-on-year changes in income-unadjusted index of

financial incapability: BHPS 1991–2006
40
30
Percent
2010
0

-4 -2 0 2 4
Within-individual year-on-year change in income-adjusted index

37
Figure 9: Within-individual year-on-year changes in income-unadjusted index of
financial incapability: BHPS 1991–2006

30
20
Percent
10
0

-4 -2 0 2 4
Within-individual year-on-year change in income-adjusted index

Figure 10: Within-individual year-on-year changes in the number of financial

problems: BHPS 1991–2006
50
40
30
Percent
20
10
0

-10 -5 0 5 10
Within-individual year-on-year change in number of financial problems

These plots are consistent with the Financial Services Authority’s Baseline Survey,
which finds that almost one-third of people experience large unexpected falls in

38
income over a three year period while one in five experiences a large unexpected
expense. Such shocks will have implications for their financial capability.

Figure 11 plots how these within-individual average changes in financial

incapability changed between 1991 and 2006. Note, first, that there are large
average within-individual changes in these indices over time, relative to the average
values of the means of the indices. Second, the average rate of improvement in
financial capability has, on average, declined over the period. In the early 1990s,
the average year-on-year change was more negative (indicating larger improvements
in financial capability) than in more recent years. In fact, since 2004 there is
evidence that the improvement has reversed, and financial incapability has started
to increase (with average within-individual changes above zero).

The advantage of using a categorical (rather than continuous) measure of financial

incapability is that it allows a more direct assessment of year-on-year change. We
take advantage of this in Table 19, and summarise individual-level changes in the
number of financial problems faced in two consecutive years. If there was no change
in financial incapability, then all individuals would lie on the leading diagonal –
they would have the same number of financial problems each year. Therefore the
degree of change can be assessed by the proportion of individuals that lie off the
leading diagonal – those that experience either an improvement or deterioration in
the number of financial problems they face.

Figure 11: Mean within-individual year-on-year changes in financial

incapability: BHPS 1991–2006
.05
Change in financial incapability
-.05 -.1 0

1991 1994 1997 2000 2003 2006

Year

Income adjusted Income unadjusted

N financial problems
Source: BHPS 1991-2006

This table indicates that there is much year-on-year fluctuation in financial

incapability. Although 60% of those with zero or one financial problem in one year
also have zero or one financial problems in the subsequent year, the vast majority of
39
 those with two or more financial problems experience a change in the number they
have in the following year. For example, of those with two financial problems in one
year, only 29.8% have two financial problems in the subsequent year. The majority
(58%) have fewer than two, while 13% have more than two. Even more change is
evident among those with more financial problems. Of course, such downward
mobility is good, indicating that individuals are improving their position on
average.

Table 19: Year-on-year changes in number of financial problems: BHPS 1991–

2006
N problems N financial problems at t
at t–1 0 1 2 3 4 5 6 7 N
0 61.5 29.5 6.9 1.5 0.3 0.2 0 0 31224
1 21.7 60.0 13.3 3.2 0.9 0.6 0.2 0 42505
2 13.4 44.5 29.8 7.9 2.2 1.7 0.4 0.1 13803
3 10.2 35.0 22.4 21.1 4.9 4.7 1.5 0.3 4991
4 8.2 31.8 17.0 15.3 13.7 9.3 4.1 0.6 1812
5 5.7 22.0 17.5 17.4 12.3 15.9 7.2 2.0 1447
6 3.7 13.1 13.7 19.9 16.1 18.4 10.5 4.5 600
7 2.4 16.6 9.1 16.5 9.2 14.9 14.6 16.7 138
N 31282 43380 13596 4679 1622 1309 521 131 96520
Notes: Row percentages. Table reads, for example, that 61.5% of individuals with no financial problems
at t–1 also had no financial problems at the year t interview, while 29.5% had one financial problem at
the year t interview.

Having created these indices of financial incapability, together with the number of
financial problems, we now turn to describing their relationships with a range of
individual and household characteristics. We take all three measures of financial
hardship forwards – the income-adjusted index, the income-unadjusted index and
the number of financial problems – to highlight the differences and similarities that
controlling for income makes in these bivariate relationships.

40
6. Relationships between financial capability and other
characteristics

In this section we introduce the individual and household variables collected in the
BHPS with which we describe patterns of financial capability. To maximise sample sizes
and to simplify the analysis, we again focus on variables collected at all BHPS waves.
For the purposes of this section, we treat the data as a series of separate cross-sections
and for the time being do not make use of the panel nature of the data. We summarise
how the constructed index is distributed across different groups in the population, and
over time. We provide summaries of indices by a range of individual and household
characteristics including age, gender, marital status, number and ages of children,
health status, employment status, job type, housing tenure and income (note that it is
not possible to summarise by ethnicity because of small sample sizes within the BHPS).
This is important, because our analysis of the relationship between financial capability
and psychological health needs to take account of such characteristics that may
potentially confound or mediate the effects. (For example, it is very plausible that
mentally healthy people are both more likely to get married, to be in employment and
to be able to manage their finances effectively.) We need to ensure that factors that
may affect both financial capability and psychological health are controlled for, and in
this section we explore relationships between financial capability and a range of
individual and household characteristics in detail.

As before, in all tables the data have been weighted to take account of potential
non-random attrition and non-random response (using weighting variable
wXRWGHT), and we include all adult (aged 16 and above) respondents, irrespective
of age, and focus on adults who provide non-missing responses to the variables of
interest. Because of missing values on some of the variables, the sample sizes are
slightly reduced to 16,348 adults contributing 122,231 person-year observations. In
each table, the ‘Average’ column shows the relationship using data pooled from all
16 waves of data.

6.1 Gender
Table 20 summarises mean financial capability by gender. This shows that, adjusting
for income, the index of financial incapability does not differ significantly by
gender. The averages for men over the sample period are consistently above those
for women, indicating higher financial incapability, but these differences are small
and not statistically significant. The income-unadjusted index and the number of
financial problems measure, however, suggest that women have higher financial
incapability than men (0.002 compared with –0.015, and 1.113 compared with
1.071), and that these differences are statistically significant. Furthermore, they
persist over the sample period. From this we conclude that women on average have
greater financial incapability than men, but that this difference can be explained by
differences in incomes between men and women.

41
 Table 20: Mean financial incapability by gender: BHPS 1991–2006
Year Average
1991 1996 2001 2006
Income-adjusted
Male 0.112 –0.028 –0.079 –0.032 –0.010
Female 0.098 –0.029 –0.081 –0.038 –0.017
Income-unadjusted
Male 0.143 –0.030 –0.091 –0.060 –0.015 *
Female 0.152 –0.004 –0.068 –0.045 0.002
N. financial problems
Male 1.327 1.044 0.929 1.009 1.071 *
Female 1.362 1.099 0.990 1.032 1.113
N 8514 8012 7542 6971 122,231
Notes: Weighted using cross-sectional weights. Table reads, for example, that in 1991 men
in the BHPS sample had a mean income-adjusted financial capability score of 0.112,
compared to 0.098 for women. ‘Average’ shows data pooled from waves 1 to 16. *
indicates that the average scores by gender over the sample period are significantly
different at the 5% level.

6.2 Age
Table 21 summarises mean financial capability by age category, together with the
Spearman rank correlation coefficient measuring the association between age as a
continuous variable and the constructed indices. The table indicates a statistically
significant association between age and financial capability using all three
indices. In particular, we find that on average financial capability increases with
age (that is, the indices get smaller). For example, the mean income-adjusted
index of financial incapability for people below 25 years of age is 0.051, compared
to –0.158 for people aged 65 and above. This pattern emerges consistently over
the sample period. A similar picture emerges using the income-unadjusted index,
although the relationship is less pronounced suggesting that adjusting for income
enhances differences in financial capability across age groups. Given that the
population mean for these indices are zero (see Tables 11 and 17), this indicates
that people aged below 45 have above average financial incapability (below
average financial capability), while those aged 55 and above have below average
financial incapability (above average financial capability). These results are
consistent with those from the Financial Services Authority’s Baseline Survey,
which found that younger people (particularly aged under 30) had the most
problems managing their finances.

The differences in the number of financial problems across age groups are small, and
on average over the period again suggest that financial incapability is more
pronounced among younger age groups. However this pattern is less consistent
than with the indices of financial incapability, with some suggestion of a non-linear
relationship (that is, the number of financial problems is greatest for the youngest
and oldest age groups).

42
 Table 21: Mean financial incapability by age: BHPS 1991–2006
Year Average
1991 1996 2001 2006
Income-adjusted
Under 25 0.175 0.008 –0.034 0.057 0.051 *
25–34 0.257 0.023 0.010 0.029 0.076
35–44 0.144 0.005 –0.037 0.034 0.034
45–54 0.133 0.026 –0.086 –0.035 0.008
55–64 –0.008 –0.074 –0.104 –0.076 –0.063
65 and above –0.092 –0.148 –0.198 –0.162 –0.158
Spearman correlation –0.117 –0.094 –0.124 –0.100 –0.118
Income-unadjusted
Under 25 0.213 0.032 –0.027 0.064 0.067 *
25–34 0.267 –0.013 –0.032 –0.034 0.038
35–44 0.156 –0.009 –0.066 –0.021 0.012
45–54 0.127 –0.004 –0.123 –0.090 –0.026
55–64 0.054 –0.050 –0.090 –0.095 –0.047
65 and above 0.036 –0.048 –0.115 –0.101 –0.063
Spearman correlation –0.033 –0.010 –0.029 –0.016 –0.024
N. financial problems
Under 25 1.408 1.116 1.020 1.192 1.178 *
25–34 1.477 1.024 0.976 0.987 1.110
35–44 1.346 1.057 0.968 1.046 1.095
45–54 1.319 1.091 0.872 0.924 1.052
55–64 1.251 1.060 0.968 0.967 1.065
65 and above 1.236 1.099 0.978 1.030 1.072
Spearman correlation –0.011 0.017 0.003 0.014 0.006
N 8514 8012 7542 6971 122,231
Notes: Weighted using cross-sectional weights. Table reads, for example, that in 1991
those aged under 25 in the BHPS sample had a mean income-adjusted financial capability
score of 0.051, compared to –0.158 for those aged 65 and above. ‘Average’ shows data
pooled from waves 1 to 16. * indicates that the average scores by age category over the
sample period are significantly different at the 5% level.

6.3 Marital status

Table 22 summarises the indices of financial incapability by marital status, and
shows that financial incapability differs significantly by marital status. Focusing
initially on the income-adjusted index, this indicates that on average widowed
individuals have the lowest financial incapability score of –0.137 (indicating above
average financial capability). The divorced or separated have the highest financial
incapability (0.154) indicating below average financial capability. This pattern is
consistent over the time period, although there is evidence that in more recent
years financial incapability has increased among the cohabiting relative to other
marital status groups. The income-unadjusted index shows a different pattern, in
that the married have the lowest financial incapability (and are on average the most
financially capable) while the divorced or separated have the highest average

43
financial incapability. The differences in the income-adjusted and unadjusted
indices suggest that the widowed are particularly good at managing their finances
(their average incapability index falls when controlling for income). The divorced or
separated also suffer the largest average number of financial problems over the
period (at 1.49), while the married suffer the fewest at 1.04. Therefore on average
the divorced suffer 44% more financial problems than married people.

Table 22: Mean financial incapability by marital status: BHPS 1991–2006

Year Average
1991 1996 2001 2006
Income-adjusted
Married 0.075 –0.050 –0.106 –0.069 –0.040 *
Cohabiting 0.362 0.064 0.008 0.111 0.102
Widowed –0.058 –0.126 –0.185 –0.156 –0.137
Divorced/separated 0.401 0.132 0.048 0.045 0.154
Single never married 0.095 –0.015 –0.054 –0.011 0.006
Income-unadjusted
Married 0.108 –0.051 –0.113 –0.096 –0.043 *
Cohabiting 0.359 0.001 –0.033 0.050 0.049
Widowed 0.078 –0.013 –0.100 –0.096 –0.034
Divorced/separated 0.482 0.196 0.090 0.066 0.207
Single never married 0.129 0.002 –0.055 –0.020 0.015
N. financial problems
Married 1.281 1.015 0.913 0.952 1.036 *
Cohabiting 1.656 1.100 1.033 1.175 1.160
Widowed 1.313 1.138 0.982 1.002 1.102
Divorced/separated 1.964 1.510 1.244 1.295 1.490
Single never married 1.269 1.057 0.953 1.025 1.085
N 8514 8012 7542 6971 122,231
Notes: Weighted using cross-sectional weights. Table reads, for example, that in 1991 the
widowed in the BHPS sample had a mean income-adjusted financial capability score of –0.137,
compared to 0.154 for the divorced or separated. ‘Average’ shows data pooled from waves 1 to
16. * indicates that the average scores by marital status over the sample period are significantly
different at the 5% level.

However Table 22 focuses on levels of financial incapability rather than change. The
advantage of panel data is that we can examine how financial incapability changes
over time and how this is associated with other life events. In Table 23 we examine
how changes in financial incapability between two consecutive years are associated
with changes in marital status over the same period. This table shows quite clearly
that getting married is associated with a relative improvement in financial
capability – on average people who get married experience a reduction in their
financial incapability scores, and this reduction is larger (more than double) than
the average year-on-year reduction experienced by the sample as a whole. For
example, using the income-adjusted measure we see that individuals who married
had a financial incapability score of 0.059 in the year before they were married, and
of 0.034 in the year after marriage. This is a fall in financial incapability of 0.029,
compared to a sample average fall of 0.012. In contrast, it is clear that those who
suffer the death of a partner or who divorce or separate experience increases in
their financial incapability. The average changes in the indices for such individuals

44
are positive and are especially large for those who experience a marital dissolution.
For example, those who divorce experience an increase in their income-adjusted
financial incapability index from 0.133 to 0.249. This represents an increase the
index of 0.116, compared to an average fall of –0.012. Clearly spousal bereavement
and marital dissolution are associated with large increases in financial incapability.

Table 23: Mean changes in financial incapability by changes in marital status:

BHPS 1991–2006
Means of financial incapability
indices
t–1 t Change N
Sample average 95935
Income-adjusted –0.020 –0.032 –0.012
Income-unadjusted –0.013 –0.028 –0.015
Number financial problems 1.077 1.053 –0.024
Got Married 1872
Income-adjusted 0.059 0.034 –0.029
Income-unadjusted 0.002 –0.042 –0.046
Number financial problems 1.049 0.996 –0.052
Became Widow 512
Income-adjusted –0.080 –0.078 0.001
Income-unadjusted 0.009 0.020 0.012
Number financial problems 1.176 1.250 0.074
Became Divorced/separated 843
Income-adjusted 0.133 0.249 0.116
Income-unadjusted 0.146 0.291 0.145
Number financial problems 1.360 1.708 0.349
Notes: Table reads, for example, that individuals who got married between two consecutive years on
average experienced a fall in their income-adjusted financial incapability from 0.059 before the
marriage to 0.034 post-marriage.

6.4 Number of children

Table 24 shows that average financial incapability scores vary significantly with the
number of children. In particular, we find that those with no children have the
lowest average financial incapability while those with four or more children have the
highest, and this pattern is evident using all three measures. For example,
individuals with no children have an average income-adjusted index of financial
incapability score of –0.046 (and suffer from 1.04 financial problems), compared to
0.147 (and 1.583 financial problems) for those with four or more children. Although
consistent over the time period, this relationship is non-monotonic. Adjusting for
income reduces the differences in averages by number of children, suggesting that
these are partly explained by differences in (equivalised) household income.

This focuses on levels of financial incapability, rather than change. Instead, Table
25 focuses on the change in financial incapability associated with the birth of an
additional child. This suggests two things. Firstly that those about to have an
additional child in the following year are already relatively high in the financial
incapability distribution – they have above average levels of financial incapability at

45
t–1 irrespective of the index used. Secondly, it suggests that the birth of a child is
associated with increases in financial incapability. For example, the income-adjusted
index of financial incapability for individuals who experience an additional child in
the household increases from 0.112 to 0.160. This represents an increase in
incapability of 0.048, compared to a fall of 0.012 for the sample as a whole. Those
with an additional child face an increase in the number of financial problems from
1.185 to 1.405 (or 19%). Financial incapability is positively related to family
formation.

Table 24: Mean financial incapability by number of children: BHPS 1991–2006

Year Average
1991 1996 2001 2006
Income-adjusted
0 0.046 –0.054 –0.107 –0.071 –0.046 *
1 0.253 0.105 –0.014 0.086 0.097
2 0.216 –0.029 0.019 0.059 0.047
3 0.320 0.051 –0.048 –0.016 0.086
4 or more 0.424 0.141 –0.103 0.305 0.147
Income-unadjusted
0 0.083 –0.046 –0.108 –0.088 –0.041 *
1 0.288 0.101 –0.032 0.059 0.087
2 0.273 –0.000 0.028 0.042 0.066
3 0.408 0.123 0.012 0.008 0.150
4 or more 0.563 0.276 0.022 0.353 0.264
N. financial problems
0 1.240 1.025 0.915 0.970 1.037 *
1 1.584 1.271 1.029 1.184 1.247
2 1.542 1.072 1.135 1.144 1.191
3 1.765 1.352 1.106 1.175 1.359
4 or more 2.084 1.625 1.164 1.557 1.583
N 8514 8012 7542 6971 122,231
Notes: Weighted using cross-sectional weights. Table reads, for example, that in 1991
those with no children in the BHPS sample had a mean income-adjusted financial
capability score of 0.046, compared to 0.424 for those with four or more children.
‘Average’ shows data pooled from waves 1 to 16. * indicates that the average scores by
number of children over the sample period are significantly different at the 5% level.

46
 Table 25: Mean changes in financial incapability by the birth of an additional
child: BHPS 1991–2006
Means of financial incapability
indices
t–1 t Change N
Sample average 95935
Income-adjusted –0.020 –0.032 –0.012
Income-unadjusted –0.013 –0.028 –0.015
Number financial problems 1.077 1.053 –0.024
Has an additional child 3358
Income-adjusted 0.112 0.160 0.048
Income-unadjusted 0.061 0.141 0.080
Number financial problems 1.185 1.405 0.220
Notes: Table reads, for example, that individuals who had an additional child between two
consecutive years on average experienced an increase in their income-adjusted financial incapability
from 0.112 to 0.160.

6.5 Household type

Table 26 looks at the relationship between the types of household in which the
individual lives and their financial incapability in more detail. The results suggest that
average financial incapability differs significantly between household types. Focussing
initially on the income-adjusted measure, we find that the lowest average index scores
(indicating the highest level of financial capability) are found among the single elderly
(–0.166), while couples with no children and couples with non-dependent children also
have below average index scores. The highest average index scores (indicating the
lowest financial capability) are found among lone parents (0.112) and adults living in
unrelated multi-occupant households (0.208). Lone parents face almost 50% more
financial problems than couples with no children (1.422 compared with 0.966). The
Financial Services Authority’s Baseline Survey reports similar findings.

47
 Table 26: Mean financial incapability by household type: BHPS 1991–2006
Year Average
1991 1996 2001 2006
Income-adjusted
Single non-elderly 0.266 0.122 0.019 0.047 0.097 *
Single elderly –0.084 –0.149 –0.210 –0.183 –0.166
Couple no children 0.005 –0.076 –0.135 –0.108 –0.078
Couple dependent children 0.188 –0.010 –0.045 0.039 0.038
Couple non-dependent children 0.045 –0.045 –0.077 –0.077 –0.038
Lone parent 0.294 0.062 0.035 0.095 0.112
2+ unrelated adults 0.155 0.229 0.142 0.024 0.208
Other households 0.229 –0.056 –0.171 0.112 0.004
Income-unadjusted
Single non-elderly 0.276 0.118 0.001 0.005 0.081 *
Single elderly 0.072 –0.019 –0.110 –0.116 –0.048
Couple no children 0.028 –0.096 –0.147 –0.137 –0.092
Couple dependent children 0.232 0.001 –0.050 0.014 0.042
Couple non-dependent children 0.046 –0.068 –0.121 –0.138 –0.069
Lone parent 0.391 0.140 0.090 0.151 0.183
2+ unrelated adults 0.179 0.233 0.105 –0.038 0.184
Other households 0.257 –0.038 –0.120 0.123 0.024
N. financial problems
Single non-elderly 1.566 1.298 1.052 1.106 1.222 *
Single elderly 1.298 1.144 0.977 0.988 1.088
Couple no children 1.156 0.952 0.871 0.899 0.966
Couple dependent children 1.464 1.079 0.998 1.101 1.148
Couple non-dependent children 1.155 0.940 0.867 0.862 0.961
Lone parent 1.773 1.339 1.221 1.371 1.422
2+ unrelated adults 1.477 1.614 1.171 0.966 1.422
Other households 1.365 1.092 0.858 1.215 1.106
N 8514 8012 7542 6971 122,231
Notes: Weighted using cross-sectional weights. Table reads, for example, that in 1991 adults in single
non-elderly households in the BHPS sample had a mean income-adjusted financial capability score of
0.266, compared to –0.084 for those in single elderly households. ‘Average’ shows data pooled from
waves 1 to 16. * indicates that the average scores by household type category over the sample period are
significantly different at the 5% level.

The general pattern remains unchanged when we focus on the income-unadjusted

index, with individuals in couples with no children, with non-dependent children
and single pensioners having the highest financial capability and lone parents and
those living in households with unrelated adults having the lowest financial
capability. (This pattern also emerges when looking at the number of financial
problems.) The average index for lone parents and the single elderly is lower once
we adjust for income – indicating that part of their financial incapability is caused
by relatively low income. In contrast, that for the single non-elderly, adults in
couples with no children, couples with non-dependent children and in households
with unrelated adults increase once we adjust for income, indicating that part of
their financial capability is associated with higher income levels.

48
 6.6 Health status
At each wave of the BHPS, individuals are asked to assess their current health
status. In particular, they are asked “Please think back over the last 12 months
about how your health has been. Compared to people of your own age, would you
say that your health has on the whole been Excellent, Good, Fair, Poor or Very
poor?” For the purposes of this analysis we have collapsed this into being in good
health (reporting excellent or good) and being in poor health (reporting fair, poor
or very poor). Table 27 looks at the relationship between an individual’s health
status and their financial incapability. The results suggest that average financial
incapability differs significantly by health. Focussing initially on the income-
adjusted measure, we find that the lowest average index scores (indicating the
highest level of financial capability) are found among those in good health (–
0.043), while those in fair or poor health have above average scores (0.054). The
general pattern remains unchanged when we focus on the income-unadjusted index,
with individuals in good health reporting consistently lower financial incapability
than those in fair or poor health. (This pattern also emerges when looking at the
number of financial problems, where those in fair or poor health suffer from 32%
more financial problems than those in good or excellent health.) This suggests a
strong correlation between self-assessed health status and financial incapability.

Table 27: Mean financial incapability by health status: BHPS 1991–2006

Year Average
1991 1996 2001 2006
Income-adjusted *
In good health 0.070 –0.067 –0.110 –0.064 –0.043
In fair, poor, very poor 0.200 0.059 –0.014 0.032 0.054
health
Income-unadjusted *
In good health 0.098 –0.071 –0.125 –0.097 –0.052
In fair, poor, very poor 0.284 0.108 0.025 0.056 0.101
health
N. financial problems *
In good health 1.243 0.954 0.867 0.933 0.998
In fair, poor, very poor 1.624 1.346 1.172 1.232 1.314
health
N 8514 8012 7542 6971 121946
Notes: Weighted using cross-sectional weights. Table reads, for example, that in 1991 adults in good
health in the BHPS sample had a mean income-adjusted financial capability score of –0.043, compared to
0.054 for those in fair, poor or very poor health. ‘Average’ shows data pooled from waves 1 to 16. *
indicates that the average scores by household type category over the sample period are significantly
different at the 5% level.

In Table 28 we examine how changes in financial incapability between two

consecutive years are associated with changes in health status over the same
period. This table shows quite clearly that an improvement in health status (moving
from fair or poor health to good or excellent health) is associated with a relative
improvement in financial capability – on average people whose health improves
experience a reduction in their financial incapability scores, and this reduction is
larger (more than double) than the average year-on-year reduction experienced by

49
the sample as a whole. For example, using the income-adjusted measure we see that
individuals who experienced an improvement in their health had a financial
incapability score of 0.033 in the year prior to the improvement and of –0.005 in
the year after the improvement. This is a fall in financial incapability of 0.038,
compared to a sample average fall of 0.012. In contrast, it is clear that those who
suffer deteriorations in their health status experience increases in their financial
incapability. The average changes in the indices for such individuals are positive.
For example, those whose health deteriorates from excellent or good to fair or poor
experience an increase in their income-adjusted financial incapability index from
0.010 to 0.028. This represents an increase the index of 0.019, compared to an
average fall of –0.012. Clearly health and financial incapability are strongly related.

Table 28: Mean changes in financial incapability by changes in health status:

BHPS 1991–2006
Means of financial incapability
indices
t–1 t Change N
Sample average 95935
Income-adjusted –0.020 –0.032 –0.012
Income-unadjusted –0.013 –0.028 –0.015
Number financial problems 1.077 1.053 –0.024
Entered good health 8594
Income-adjusted 0.033 –0.005 –0.038
Income-unadjusted 0.051 0.009 –0.042
Number financial problems 1.200 1.121 –0.078
Left good health 9356
Income-adjusted 0.010 0.028 0.019
Income-unadjusted 0.031 0.048 0.016
Number financial problems 1.158 1.196 0.038
Notes: Table reads that individuals who experienced an in health between two consecutive years on
average experienced fall in their income-adjusted financial incapability from 0.033 to –0.005.

6.7 Education levels

Table 29 presents summaries of the indices of financial incapability by
education, and shows that financial incapability differs significantly by
education levels across all three measures. Both the income-unadjusted index
and the number of financial problems reveal a monotonic relationship with
education. That is, the most highly educated who hold higher or first degrees
have the lowest average income-unadjusted index scores (–0.152 and –0.099)
and the fewest average number of financial problems (0.838 and 0.910), while
the least educated with no qualifications have the highest average income-
unadjusted index score (0.066) and the most financial problems (1.265).
Furthermore, the average index and number of financial problems rise with each
successively lower education level – those with no qualifications suffer 50% more
financial problems than those with a first or higher degree.

50
 Table 29: Mean financial incapability by education level: BHPS 1991–2006
Year Average
1991 1996 2001 2006
Income-adjusted
Higher degree –0.015 –0.070 0.029 –0.067 –0.004 *
First degree 0.098 0.003 –0.038 –0.014 0.017
Other higher qualification 0.023 –0.042 –0.079 –0.030 –0.029
A-Levels or equivalent 0.123 –0.003 –0.081 –0.011 0.003
GCSEs or equivalent 0.124 –0.032 –0.085 –0.029 –0.011
Other qualifications 0.145 –0.040 –0.080 –0.004 –0.006
No qualifications 0.118 –0.028 –0.105 –0.087 –0.022
Income-unadjusted
Higher degree –0.131 –0.210 –0.094 –0.232 –0.152 *
First degree 0.012 –0.107 –0.155 –0.131 –0.099
Other higher qualification 0.011 –0.074 –0.112 –0.071 –0.061
A-Levels or equivalent 0.136 –0.016 –0.101 –0.033 –0.012
GCSEs or equivalent 0.159 –0.021 –0.069 –0.024 0.002
Other qualifications 0.209 0.008 –0.025 0.038 0.047
No qualifications 0.220 0.060 –0.025 –0.016 0.066
N. financial problems
Higher degree 0.777 0.761 0.923 0.661 0.838 *
First degree 1.052 0.884 0.804 0.864 0.910
Other higher qualification 1.066 0.960 0.882 0.981 0.982
A-Levels or equivalent 1.281 1.039 0.885 1.028 1.059
GCSEs or equivalent 1.350 1.033 0.974 1.074 1.081
Other qualifications 1.453 1.124 1.076 1.188 1.188
No qualifications 1.526 1.258 1.104 1.124 1.265
N 8514 8012 7542 6971 122,231
Notes: Weighted using cross-sectional weights. Table reads, for example, that in 1991 adults holding
a higher degree in the BHPS sample had a mean income-adjusted financial capability score of –
0.015, compared to 0.118 for those with no qualifications. ‘Average’ shows data pooled from waves 1
to 16. * indicates that the average scores by education category over the sample period are
significantly different at the 5% level.

However, when adjusting for income the pattern changes completely, and the
monotonic relationship between financial incapability and education disappears.
The highest average income-adjusted index score (indicating low financial
capability) is for those with a first degree (0.017), while the lowest is for those with
other higher qualifications (–0.029). These findings suggest that the large
differences in financial capability between the more educated and the less educated
is related to differences in income levels associated with education attained rather
than the level of education itself. This has important implications, as it suggests
that raising general education levels will not directly improve financial capability
itself, and will only do so through an income effect.

6.8 Housing tenure

There is a statistically significant relationship between housing tenure and all three
measures of financial incapability (Table 30). Home owners without a mortgage have
the greatest financial capability (lowest values on average) while tenants have the

51
lowest financial capability (highest values on average). For example, the average
income-adjusted index for individuals who own their own home with no mortgage is
–0.184, while for private tenants it is 0.124. This pattern emerges consistently over
the sample period and for all three measures. Adjusting for income reduces the
differentials between housing tenure groups, and has a particularly large effect for
local authority tenants indicating that part of their financial incapability stems from
relatively low income. Private tenants suffer from 50% more financial problems than
those who own their home outright, and 35% more than those with a mortgage.
These results are consistent with those found in the Financial Services Authority’s
Baseline Survey, which found that home-owners were most able to make ends meet
while tenants (in social housing in particular) had most problems planning ahead.

Table 30: Mean financial incapability by housing tenure: BHPS 1991–2006

Year Average
1991 1996 2001 2006
Income-adjusted
Own home outright –0.199 –0.194 –0.200 –0.171 –0.184 *
Own home mortgage 0.143 –0.005 –0.064 0.008 0.017
Local authority rent 0.301 0.063 0.010 0.038 0.098
Private rent 0.180 0.115 0.075 0.137 0.124
Income-unadjusted
Own home outright –0.125 –0.153 –0.166 –0.155 –0.146 *
Own home mortgage 0.129 –0.052 –0.118 –0.067 –0.035
Local authority rent 0.429 0.175 0.101 0.116 0.206
Private rent 0.238 0.143 0.100 0.135 0.149
N. financial problems
Own home outright 0.935 0.884 0.865 0.899 0.906 *
Own home mortgage 1.262 0.967 0.862 0.958 1.003
Local authority rent 1.873 1.434 1.283 1.314 1.480
Private rent 1.535 1.398 1.236 1.326 1.358
N 8514 8012 7542 6971 122,231
Notes: Weighted using cross-sectional weights. Table reads, for example, that in 1991 adults
owning their home outright in the BHPS sample had a mean income-adjusted financial capability
score of –0.199, compared to 0.180 for those in privately rented accommodation. ‘Average’
shows data pooled from waves 1 to 16. * indicates that the average scores by housing tenure
category over the sample period are significantly different at the 5% level.

Table 31 introduces some dynamics by focusing on the change in financial

incapability associated with becoming a home-owner. This indicates that those who
become home-owners have lower than average financial incapability both before and
after buying a property. This suggests that those buying their own home have above
average financial capability. However, they also experience a larger than average
reduction in their financial incapability. For example, individuals who become a
home-owner have an average income-adjusted index of financial incapability of –
0.059 (compared to –0.020 for the sample as a whole), and this falls to –0.157 after
the event. Therefore their financial incapability falls by 0.098 compared to a sample
average fall of 0.012.

52
 Table 31: Mean changes in financial incapability by becoming a home-owner:
BHPS 1991–2006
Means of financial incapability
indices
t–1 t Change N
Sample average 95935
Income-adjusted –0.020 –0.032 –0.012
Income-unadjusted –0.013 –0.028 –0.015
Number financial problems 1.077 1.053 –0.024
Became a home-owner 2203
Income-adjusted –0.059 –0.157 –0.098
Income-unadjusted –0.073 –0.168 –0.095
Number financial problems 0.970 0.820 –0.150
Notes: Table reads that individuals who became a home-owner between two consecutive years on
average experienced a fall in their income-adjusted financial incapability from –0.059 to –0.157.

6.9 Current house value

As well as housing tenure, at each year the BHPS asks home-owners to estimate the value
of the house they currently live in. We use this as an approximation to wealth, as it is the
only measure of wealth that is available at all 16 waves of the BHPS. We summarise the
correlations between current house value (deflated to January 2006 prices) and our
measures of financial incapability in Table 32 below. We present correlations both
including non-owners (who are allocated a house value of zero) and excluding them.

Table 32: Correlations between financial incapability and current house value:
BHPS 1991–2006
Year Average
1991 1996 2001 2006
Income-adjusted
All –0.143 –0.064 –0.016 –0.013 –0.066
Home-owners –0.026 0.021 0.090 0.060 0.039
Income-unadjusted
All –0.254 –0.231 –0.191 –0.184 –0.227
Home-owners –0.116 –0.108 –0.077 –0.098 –0.096
N. financial problems
All –0.212 –0.186 –0.142 –0.132 –0.180
Home-owners –0.077 –0.065 –0.024 –0.041 –0.050
Notes: Table shows Spearman rank correlation coefficients. House values deflated to
2006 January prices. ‘Average’ shows data pooled from waves 1 to 16.

The correlations highlight a number of notable patterns. Firstly we find that the
correlations are relatively small, suggesting that financial incapability is only weakly
correlated with wealth, as measured by current house value. Focusing on the
income-unadjusted measures, we find that financial incapability is negatively
correlated with house value, indicating that wealthier individuals have lower
financial incapability (and higher financial capability). This pattern emerges with
both the income-unadjusted index of financial incapability and the number of
financial problems. In addition, the correlation is stronger when tenants are
53
included (and given a house value of zero), suggesting that financial capability is in
fact only very weakly correlated with wealth. When adjusting for income, the
correlations between housing wealth and financial incapability become even smaller,
indicating that much of the correlation was associated with income.

6.10 Labour market status

Table 33 summarises our measures of financial incapability by employment status,
and shows significant differences for all three measures. For the purposes of this
report, we have distinguished between full-time and part-time employees, and also
the self-employed. Furthermore, we have separated the economically inactive into
those who are inactive and would not like a job, and those who are inactive but
would like to work if their circumstances permit it.

Table 33: Mean financial incapability by employment status: BHPS 1991–2006

Year Average
1991 1996 2001 2006
Income-adjusted
Full-time employee 0.059 –0.039 –0.066 –0.022 –0.012 *
Part-time employee 0.104 –0.040 –0.052 –0.023 –0.013
Self-employed 0.232 –0.051 –0.069 –0.043 0.011
Unemployed 0.597 0.290 0.246 0.325 0.355
Inactive not like 0.073 0.052 –0.038 0.017 0.027
job
Inactive like job 0.398 0.132 0.077 0.141 0.201
Retired –0.078 –0.144 –0.187 –0.157 –0.149
Income-unadjusted
Full-time employee 0.027 –0.109 –0.148 –0.120 –0.086 *
Part-time employee 0.144 –0.024 –0.046 –0.040 –0.006
Self-employed 0.241 –0.086 –0.065 –0.066 –0.001
Unemployed 0.718 0.389 0.345 0.418 0.461
Inactive not like 0.171 0.116 0.016 0.065 0.089
job
Inactive like job 0.514 0.236 0.167 0.216 0.299
Retired 0.049 –0.045 –0.102 –0.097 –0.057
N. financial problems
Full-time employee 1.037 0.831 0.772 0.831 0.873 *
Part-time employee 1.311 1.017 0.970 0.991 1.051
Self-employed 1.491 0.896 0.966 0.972 1.090
Unemployed 2.512 1.993 1.868 1.953 2.076
Inactive not like 1.454 1.365 1.183 1.273 1.309
job
Inactive like job 2.059 1.586 1.450 1.540 1.705
Retired 1.274 1.110 1.009 1.039 1.092
N 8437 7908 7417 6874 120482
Notes: Weighted using cross-sectional weights. Table reads, for example, that in 1991 adults
in full-time employment in the BHPS sample had a mean income-adjusted financial capability
score of 0.059, compared to 0.597 for those in unemployment. ‘Average’ shows data pooled
from waves 1 to 16. * indicates that the average scores by employment status over the sample
period are significantly different at the 5% level.

54
The table shows that the highest average financial incapability is found for the
unemployed. The average income-adjusted index score for the unemployed is 0.355,
while the income-unadjusted score is 0.461. Therefore even after adjusting for
income, the unemployed have significantly higher financial incapability than
average. They also suffer from an average of two financial problems. Economically
inactive individuals who would like a job also have above average financial
incapability. The table shows that they have an average income-adjusted score of
0.201 and an income-unadjusted score of 0.299. Again therefore, income cannot
completely explain the relatively low financial capability among this group.

The table indicates that the lowest financial incapability (and highest levels of
financial capability) is found among those in employment and the retired. Those in
full-time employment have an average income-unadjusted score of –0.086,
indicating below average financial incapability, while the retired have an average
income-unadjusted score of –0.057. A similar pattern emerges using the number of
financial problems. Those in full-time employment have on average 0.87 financial
problems, compared with about 1.1 for those in part-time employment, self-
employment and retirement while the unemployed suffer from more than two
financial problems. Therefore the unemployed on average have more than twice the
number of financial problems as those in full-time work. Adjusting for income only
changes this picture slightly for those in employment. However, the average index
falls considerably for the retired when adjusted for income, from –0.057 to –0.149.
Therefore, average financial capability increases when adjusting for income,
suggesting that the retired are able to manage their finances well given their
income level. Our findings are consistent with those from the Financial Services
Authority’s Baseline Survey, which found that the unemployed in particular had
problems in making ends meet.

Table 34 focuses on the dynamic association between employment status changes

and changes in financial incapability. This indicates that entering employment is
associated with significantly larger than average falls in financial incapability, and
this is apparent using all three indices. For example, the average income-adjusted
index falls for individuals who enter work from 0.171 before they enter work to
0.074 afterwards. This represents a fall of 0.098 compared to 0.012 for the sample
as a whole. Those entering work experience a 27% reduction in the number of
financial problems they face (1.145 from 1.559). In contrast, individuals who will
enter unemployment in the following year already had higher than average financial
incapability, and also suffer an increase in financial incapability on becoming
unemployed. The income-adjusted index for those entering unemployment increases
from 0.224 to 0.344 (an increase of 0.121), while the number of financial problems
increases by 32%. Therefore, individuals who enter unemployment have higher than
average financial incapability before entering unemployment, but this increases
even further once unemployed. Furthermore, this increase is not caused by the loss
of income associated with unemployment. The association between changes in
financial incapability and entering retirement depends on the index being used. We
find that using the unadjusted index and the number of financial problems, entering
retirement is associated with an increase in financial incapability. For example, the
income-unadjusted index increases from 0.000 pre-retirement to 0.055 post-

55
retirement. However, the income-adjusted index falls from 0.035 pre-retirement to –
0.004 post-retirement. This indicates that the increases in the unadjusted measures
of financial incapability reflect the fall in income associated with retiring.

Table 34: Mean changes in financial incapability by employment status changes:

BHPS 1991–2006
Means of financial incapability
indices
t–1 t Change N
Sample average 95935
Income-adjusted –0.020 –0.032 –0.012
Income-unadjusted –0.013 –0.028 –0.015
Number financial problems 1.077 1.053 –0.024
Entered work 4350
Income-adjusted 0.171 0.074 –0.098
Income-unadjusted 0.215 0.061 –0.153
Number financial problems 1.559 1.145 –0.414
Entered unemployment 1879
Income-adjusted 0.224 0.344 0.121
Income-unadjusted 0.271 0.445 0.173
Number financial problems 1.578 2.081 0.503
Entered retirement 2102
Income-adjusted 0.035 –0.004 –0.039
Income-unadjusted 0.000 0.055 0.055
Number financial problems 1.104 1.242 0.138
Notes: Table reads that individuals who entered work between two consecutive years on average
experienced a fall in their income-adjusted financial incapability from –0.171 to 0.074.

6.11 Job type

As well as employment status, each year the BHPS collects information on the types
of jobs in which those in work are currently employed. In Table 35 we summarise
financial incapability by whether people are currently employed in permanent jobs,
seasonal or temporary work, or on a fixed-term contract.

56
 Table 35: Mean financial incapability by job type: BHPS 1991–2006
Year Average
1991 1996 2001 2006
Income-adjusted
Permanent job 0.070 –0.054 –0.064 –0.028 –0.017 *
Seasonal/casual job 0.229 0.111 –0.062 0.069 0.089
Fixed term contract 0.311 0.106 –0.045 –0.060 0.114
Income-unadjusted
Permanent job 0.056 –0.103 –0.118 –0.100 –0.068 *
Seasonal/casual job 0.264 0.121 –0.068 0.059 0.099
Fixed term contract 0.295 0.062 –0.094 –0.150 0.061
N. financial problems
Permanent job 1.109 0.849 0.833 0.879 0.917 *
Seasonal/casual job 1.519 1.284 0.948 1.227 1.251
Fixed term contract 1.523 1.111 0.876 0.832 1.168
N 5107 4975 4877 4483 76969
Notes: Weighted using cross-sectional weights. Table reads, for example, that in 1991 adults in
permanent employment in the BHPS sample had a mean income-adjusted financial capability
score of 0.07, compared to 0.311 for those employed on fixed term contracts. ‘Average’ shows
data pooled from waves 1 to 16. * indicates that the average scores by job type category over
the sample period are significantly different at the 5% level.

This indicates that on average over the period, and on all three measures of
financial incapability, those employed in permanent jobs have the lowest financial
incapability scores (and therefore are most able to manage their finances). On
average they suffer 0.9 financial problems, and have an income-unadjusted index
score of –0.068, indicating above average financial capability. When adjusting for
income this increases to –0.017, which suggests that some of their above average
financial capability is due to higher incomes. A similar pattern emerges for those
currently employed on fixed-term contracts – when adjusting for income, the
average index score increases from 0.061 to 0.114. These patterns are evident across
the sample period.

6.12 Income
In Table 36 we summarise the relationships between our three measures of financial
incapability and real monthly equivalised gross household income. This allows us to
establish the impact of adjusting for income on this relationship, and if this has
changed over time.

57
 Table 36: Mean financial incapability by income: BHPS 1991–2006
Year Average
1991 1996 2001 2006
Income-adjusted
Bottom quintile 0.253 0.027 –0.101 –0.017 0.043 *
Second quintile 0.180 –0.047 –0.060 –0.073 –0.026
Middle quintile 0.066 –0.063 –0.089 –0.032 –0.050
Fourth quintile –0.009 –0.069 –0.112 –0.067 –0.054
Highest quintile 0.031 0.008 –0.037 0.013 0.015
Spearman correlation –0.088 0.079 0.112 0.139 0.083
Pearson correlation –0.087 0.021 –0.007 0.034 0.000
Income-unadjusted
Bottom quintile 0.448 0.209 0.068 0.142 0.223 *
Second quintile 0.309 0.055 0.028 0.000 0.072
Middle quintile 0.129 –0.035 –0.072 –0.035 –0.027
Fourth quintile –0.022 –0.133 –0.182 –0.160 –0.120
Highest quintile –0.136 –0.223 –0.265 –0.242 –0.217
Spearman correlation –0.384 –0.350 –0.343 –0.328 –0.339
Pearson correlation –0.245 –0.215 –0.230 –0.229 –0.217
N. financial problems
Bottom quintile 1.992 1.581 1.283 1.401 1.569 *
Second quintile 1.637 1.210 1.166 1.148 1.241
Middle quintile 1.289 0.992 0.954 1.046 1.033
Fourth quintile 1.003 0.815 0.741 0.797 0.854
Highest quintile 0.786 0.676 0.602 0.644 0.686
Spearman correlation –0.358 –0.310 –0.304 –0.292 –0.302
Pearson correlation –0.268 –0.223 –0.228 –0.236 –0.217
N 8514 8012 7542 6971 122,231
Notes: Weighted using cross-sectional weights. Table reads, for example, that in 1991 adults
in the bottom income quintile in the BHPS sample had a mean income-adjusted financial
capability score of 0.253, compared to 0.031 for those in the highest income quintile.
‘Average’ shows data pooled from waves 1 to 16. * indicates that the average scores by
income category over the sample period are significantly different at the 5% level.

Focusing initially on the income-adjusted incapability index, we find (by construction)

an average Pearson correlation coefficient with income of zero, although there are
deviations from zero across time. (The average Spearman rank correlation coefficient
however is non-zero, although still small, which suggests that the normality
distribution assumption may not hold.) The correlations are much larger between
income and financial incapability when using the income-unadjusted measures. The
income-unadjusted index has a Spearman rank correlation coefficient with income of
–0.339, indicating that people with higher incomes have lower financial incapability
levels. Furthermore, the average income-unadjusted index declines monotonically as
income increases. For example, people in the bottom income quintile have an average
income-unadjusted index score of 0.223, compared to –0.027 for those in the middle
income quintile and –0.217 for those in the highest income quintile. This pattern is
evident across the whole period, and is also evident when looking at the number of
financial problems. For example, on average over the period people in the bottom

58
income quintile suffered from more than double the number of financial problems
than those in the highest income quintile (1.569 financial problems compared with
0.686). The Financial Services Authority’s Baseline Survey also found that those with
low incomes struggled to make ends meet, but that to some extent financial
incapability was evident at higher income levels.

Table 37: Mean changes in financial incapability by income changes: BHPS

1991–2006
Means of financial incapability
indices
t–1 t Change N
Sample average 95935
Income-adjusted –0.020 –0.032 –0.012
Income-unadjusted –0.013 –0.028 –0.015
Number financial problems 1.077 1.053 –0.024
Income increase > 10% 33267
Income-adjusted –0.007 –0.000 0.006
Income-unadjusted 0.041 –0.035 –0.076
Number financial problems 1.183 1.019 –0.164
Income fell > 10% 26615
Income-adjusted 0.035 –0.004 –0.039
Income-unadjusted 0.000 0.055 0.055
Number financial problems 1.104 1.242 0.138
Notes: Table reads that individuals who experienced an increase in real monthly equivalised gross
household income exceeding 10% between two consecutive years on average experienced an increase
in their income-adjusted financial incapability from –0.007 to –0.000.

Table 37 focuses on the dynamics of the relationship between financial incapability

and income, by focusing on the changes in financial incapability experienced by
individuals who experienced increases and falls of greater than 10% in their real
monthly equivalised gross household income. The income-unadjusted index and the
number of financial problems reveal the relationships we would expect to find –
substantial increases in household income are associated with falls in financial
incapability while substantial falls in household income are associated with
increases in financial incapability. Those who experience a 10% drop in income
suffer an increase of 12% in the number of financial problems they face, while those
that experience an increase of at least 10% in their income face 14% fewer financial
problems. However a different pattern emerges with the income-adjusted index.
According to this index, individuals who experience substantial increases in their
household income experience an increase in their financial incapability, and vice-
versa. This suggests that such income changes are associated with other factors that
influence an individual’s financial incapability (for example, changes in household
composition or in employment status).

6.13 Summary
In this section we have summarised how financial incapability is related to a range
of individual and household characteristics that are available at all BHPS waves. This
is important, given that such mediating and confounding factors have to be taken

59
into account in establishing any relationship between financial capability and
psychological wellbeing. We find that our measures of financial incapability are
significantly associated with gender, age, marital status, number of children, health,
employment status, job type, housing tenure and income, and also with changes in
marital status, the number of children, health, employment status, housing tenure
and income. In particular, we find that people with the highest financial
incapability tend to be young (aged less than 35), divorced or separated, have more
than one or two dependent children, are single non-elderly, lone parents, in fair or
poor health, live in rented accommodation and are unemployed or are economically
inactive but would like a job. In contrast, people with lowest financial incapability
are on average older (aged 55 or above), married or widowed with no dependent
children, in good health, home owners and working in a full-time permanent job. In
addition, there is evidence that financial incapability is strongly related to
education, but this relationship is much less pronounced when adjusting for income.
These findings are consistent with those from the Financial Services Authority’s
Baseline Survey. As well as associations between states, panel data allow us to
investigate associations between events. Doing this reveals that getting married,
improvements in health, becoming a homeowner and entering work are associated
with increased financial capability, while the death of a spouse, marital dissolution,
an additional child, a deterioration in health and unemployment are associated with
falls in financial capability levels.

These bivariate relationships, while interesting, do not begin to address the

question of what determines financial incapability. For example, we find that the
unemployed on average exhibit lower financial capability than those in employment.
However, we cannot interpret this as suggesting that unemployment reduces
financial management skills, as the lack of financial management skills may have
contributed to individuals losing their jobs. Similarly, although entering
unemployment is associated with falls in financial capability, we cannot say this
relationship is causal because there may be a factor that contributes both to an
individual losing their job and to the fall in financial capability. An interesting and
important avenue for future research would be to investigate these relationships in
more detail and in a multivariate framework which would allow more robust and
interpretable conclusions to be drawn. The subsequent sections of this report focus
on the relationships between our measures of financial incapability and
psychological wellbeing.

60
7. Relationships between financial incapability and
psychological wellbeing
The next stage in the analysis is to investigate the relationships between financial
capability and psychological wellbeing. Of particular interest is the dynamics of any
relationship, and whether the index of financial capability has any power in
predicting psychological wellbeing. As a first step we examine our measures of
psychological wellbeing.

7.1 Psychological wellbeing in the BHPS

We use three measures of psychological wellbeing, each capturing a slightly
different component. Our main measure is the GHQ-12 which we score using the
Likert method, giving a range of 0 (no mental health problems at all) to 36 (serious
mental health problems). As described in Section 2, this measure asks respondents
to rate their level of experiencing each symptom in relation to what is ‘usual’.
Therefore it captures short term changes in psychological health but may
underestimate chronic conditions. For example, if a person is depressed and never
feel as if they play a useful part in things, they may respond ‘same as usual’ to this
despite being depressed.7 However the focus on short-term fluctuations seems
appropriate as our concern is with the impact of financial capability on
psychological health. Also, we have repeated all analysis using the 12-point
‘Caseness’ scale, which may be less sensitive to short-term fluctuations in mental
health, and the results are very similar to those presented here. Furthermore, we use
as a measure of psychological health whether or not a person suffers a health
problem related to anxiety or depression, and this will more clearly capture any
chronic condition. Our third measure of wellbeing is reported life satisfaction,
collected at waves 6–10 and 11–16 using a seven point scale where one equates to
not satisfied at all and seven to completely satisfied.

Table 38 summarises how patterns in GHQ scores, whether suffering from problems
related to anxiety or depression and in reported life satisfaction evolve over the
BHPS sample period. The average GHQ score over the period was 11.09, and there is
some evidence of an increase in scores (and therefore in mental stress levels) over
the period. The average proportion of people reporting a health problem related to
anxiety or depression was 6.8%, and again there is some evidence of an increase
over the period particularly in the decade to 2001. Average life satisfaction scores
were 5.23, and these fell marginally over the period. Therefore this table would
suggest that psychological wellbeing and life satisfaction are getting marginally
worse over the period, despite improvements in financial capability (see Figure 7 –
although of course this does not tell us whether it is the same people whose
financial capability is improving that experience declines in mental wellbeing).
However, these means do not tell us much about the distribution of GHQ scores or
life satisfaction scores. We explore these in more detail in Figures 12 and 13. Figure

7
In fact, this is one of the advantages of using the 36-point Likert scale rather than the more
common 12-point Caseness scale. For the latter, responding ‘same as usual’ would score zero points
while in the former it scores one point.
61
12 plots the frequency distribution of GHQ scores, and clearly shows that the most
common GHQ scores were between 6 and 13. In fact almost two-thirds of all
reported scores lie in this interval, with only 20% of observations having scores
above 13 and 10% above 18. The median GHQ score in the sample is 10.

Table 38: Measures of psychological wellbeing: BHPS 1991–2006

Year GHQ Suffers anxiety/ Life
Score depression satisfaction
1991 10.70 0.051
1992 11.01 0.059
1993 10.99 0.057
1994 11.09 0.058
1995 11.23 0.060
1996 11.20 0.067 5.249
1997 11.14 0.073 5.267
1998 11.07 0.075 5.339
1999 10.95 0.063 5.239
2000 11.30 0.074 5.162
2001 11.15 0.080
2002 11.10 0.080 5.238
2003 11.04 0.076 5.261
2004 11.11 0.074 5.199
2005 11.27 0.075 5.131
2006 11.24 0.079 5.188
Average 11.09 0.068 5.230
N observations 119290 121383 72954
Notes: Weighted using cross-sectional weights. Table reads, for
example, that in 1996 adults in the BHPS sample had a mean GHQ
score of 11.2, 6.7% suffered from health problems relating to anxiety
or depression, while average life satisfaction scores were 5.249.
‘Average’ shows data pooled from waves 1 to 16.

62
 Figure 12: Frequency distribution of GHQ scores: BHPS 1991–2006

40
30
GHQ Score
20
10
0

0 5000 10000
Frequency

Figure 13 plots the frequency distribution of life satisfaction scores, and clearly
shows that the most common satisfaction scores were 5 and 6. In fact 30% of the
sample reported a life satisfaction score of 5, and 33% a score of 6, with only 23%
reporting a life satisfaction score of 4 or less. The median life satisfaction score in
the sample is 5.

63
 Figure 13: Frequency distribution of life satisfaction scores: BHPS 1991–2006
8
6
Life satisfaction
4
2
0

0 5000 10000 15000 20000 25000

Frequency

Table 39 summarises the relationships between the three measures of psychological

wellbeing. It presents mean GHQ and life satisfaction scores by whether or not
people are suffering from anxiety or depression, and then mean GHQ scores by
reported life satisfaction. This table indicates high levels of association between
these measures. For example, people who reported suffering a health problem
related to anxiety or depression had an average GHQ score of 17.6, significantly
larger than the average of 10.62 for people who reported no such health problems.
Therefore mental stress is significantly higher for people reporting problems related
to anxiety or depression. Similarly, people who reported suffering a health problem
related to anxiety or depression also on average reported lower life satisfaction
(4.089 compared with 5.319). A strong monotonic relationship also emerges
between GHQ scores and reported life satisfaction, with a Spearman rank correlation
coefficient of –0.506. This indicates that high mental stress is associated with lower
reported life satisfaction. People reporting life satisfaction scores of one or two had
average GHQ scores exceeding 20, while those reporting life satisfaction scores of
six or seven had average GHQ scores of less than 10.

64
 Table 39: Relationships between measures of psychological wellbeing: BHPS
1991–2006
GHQ Score Life satisfaction
Anxiety/Depression
Yes 17.60 4.089
No 10.62 5.319
Life satisfaction
1 22.42
2 20.28
3 17.13
4 14.04
5 11.10
6 9.03
7 8.20
Correlation -0.506
Notes: Weighted using cross-sectional weights. Table reads, for
example, that individuals who reported health problems related to
anxiety or depression had an average GHQ score of 17.6, compared to
10.6 for those who reported no such health problems, and an average
life satisfaction score of 4.089 compared with 5.319 among those with
no such health problems. Data pooled from waves 1 to 16.

7.2 Changes in individual psychological wellbeing from one year to

the next
So far in this section, we have analysed the measures of psychological wellbeing from a
cross-sectional perspective. That is, we have not taken advantage of the panel nature of
the data to examine how psychological wellbeing changes for the same individuals over
time. We do this in the same way as we did earlier for financial capability, and Table 40
below presents our first look at this. In this table we summarise individuals’ mean
psychological wellbeing over two consecutive years, as well as the average within-
individual year-on-year change and within-individual variance.

Table 40: Within-individual year-on-year changes in psychological wellbeing:

BHPS 1991–2006
Psychological wellbeing Means
t–1 t Change Within-individual
variance
GHQ scores 11.02 11.12 0.10 16.92
Life satisfaction 5.26 5.23 –0.03 0.82
Anxiety/depression 0.07 0.07 0.00 0.04
Notes: Weighted using cross-sectional weights. Table reads, for example, that on average
individuals had a GHQ score of 11.02 in year t–1 and of 11.12 in year t, indicating an average
increase in GHQ of 0.1.

The table indicates that on average people’s psychological wellbeing fell between
one year (t–1) and the next (t). The mean change in GHQ scores was positive
while that in reported life satisfaction was negative, showing that individuals’
psychological wellbeing was deteriorating over the period. For example, the
mean GHQ score increased from 11.02 to 11.12 between two consecutive years,

65
while the mean reported life satisfaction fell from 5.26 to 5.23. On average there
was no year-on-year change in suffering from a health problem related to anxiety
or depression. The table also presents average within-individual variances in the
measures of psychological wellbeing which, in the case of GHQ scores, are large
relative to the means. Therefore there is some within-individual change in
psychological wellbeing – people’s psychological wellbeing changes from year-to-
year. (This is also reflected in the fact that a person’s GHQ scores at t and at t–1
exhibit a Spearman rank correlation coefficient of 0.52.) This may at least partly
reflect the nature of the GHQ, in that it is likely to pick up short-term
fluctuations in people’s psychological health.

Figure 14 plots the distribution of year-on-year changes at the individual level in

GHQ scores. This shows that 25% had no change in psychological wellbeing from
one year to the next and so their psychological wellbeing was stable. While this is
clearly the modal value, this suggests that in 75% of cases, individuals’
psychological wellbeing changed from one year to the next and in some cases these
changes were large. Again, this reflects the fact that the 36-point GHQ is good at
identifying short-term change in mental health. Figure 15 reveals a similar pattern
when looking at year-on-year changes at the individual level in reported life
satisfaction, although the amount of individual level change from one year to
another is smaller (almost 50% experience no change).

Figure 14: Within-individual year-on-year changes in GHQ scores: BHPS 1991–

2006
25
20 15
Percent
10
5
0

-40 -20 0 20 40
Within-individual year-on-year change in GHQ score

66
 Figure 15: Within-individual year-on-year changes in life satisfaction: BHPS
1991–2006

50
4030
Percent
20
10
0

-10 -5 0 5 10
Within-individual year-on-year change in life satisfaction score

Figure 16: Mean within-individual year-on-year changes in GHQ scores: BHPS

1991–2006
.14
.12
Change in GHQ
.1
.08
.06

1990 1995 2000 2005

Year
Source: BHPS 1991-2006

Figure 16 plots the trend in within-individual average change in GHQ scores. This
clearly shows that the rate of year-on-year change in mental stress has fallen over
the sample period. In the early 1990s, for example, individuals were on average
experiencing increases in mental stress of around 0.12 GHQ points per year. This fell

67
 throughout the decade and into the early 2000s, such that in 2005 the year-on-year
average increases in GHQ scores were fewer than 0.08 points. There is evidence that
this has subsequently increased.

The advantage of using a categorical (rather than continuous) measure of financial

incapability is that it allows a more direct assessment of year-on-year change. We
take advantage of this in Tables 41, 42 and 43, in which we summarise individuals’
movements in psychological wellbeing in two consecutive years. If there was no
change in psychological wellbeing, then all individuals would lie on the leading
diagonal of each table – they would remain with the same wellbeing score each
year. Therefore the degree of change can be assessed by the proportion of
individuals that lie off the leading diagonal – those that experience either an
improvement or deterioration in their psychological wellbeing.

Table 41 focuses on average year-on-year transitions in GHQ scores, and reveals a

large degree of change at the individual level. For example, of individuals with a
GHQ score of 9 at t–1 (and who were therefore psychologically quite healthy), 37%
had a lower GHQ score at the subsequent year (and therefore had an improvement in
wellbeing) while 49% had a higher GHQ score (and suffered a deterioration in their
psychological wellbeing). Furthermore, one in ten had a GHQ score of 15 or above. A
similar pattern emerges in the rest of the table – there are relatively large changes
in psychological wellbeing on a year-to-year basis.

Table 41: Year-on-year changes in GHQ scores: BHPS 1991–2006

GHQ scores GHQ scores at t
at t–1 <9 9 10 11 12 13 14 15+ N
<9 60.8 9.7 7.6 5.7 4.2 2.3 1.6 8.1 33686
9 36.9 14.0 12.7 10.3 8.5 4.2 2.8 10.6 8607
10 28.0 12.9 14.4 12.5 10.5 4.9 3.6 13.3 8587
11 20.9 10.9 12.2 14.8 14.1 3.0 4.4 16.7 8216
12 15.3 7.2 9.1 11.8 23.8 7.8 5.5 19.6 8949
13 15.2 6.5 8.8 11.5 14.5 9.2 6.9 27.4 4770
14 13.9 5.5 7.1 9.8 13.9 8.7 7.6 33.4 3849
15+ 12.7 4.5 5.5 6.2 8.4 6.2 6.1 50.4 19856
N 33011 8501 8471 8293 9258 4765 3818 20403 96520
Notes: Row percentages. Table reads, for example, that 60.8% of individuals with a GHQ score of less than 9
at year t–1 also had a GHQ score of less than 9 at year t, while 9.7% had a GHQ score of 9 at year t.

Table 42 reports year-on-year changes in reported life satisfaction, and reveals a

similar pattern. Again there is a great deal of change from year to year. For
example, only 20% of those who reported a life satisfaction score of two in one year
reported the same score in the subsequent year. This increases to 25% of those
reporting a score of three, 33% of those reporting a four, and 47% of those reported
a five. Nevertheless, more than half report a different life satisfaction score from
one year to another. Table 43 reveals that 45% of those who reported suffering from
anxiety or depression in any particular year were no longer reporting the problem at
the subsequent year. From this it is clear that, even on an annual basis, there is a
great deal of change in reported psychological wellbeing. Our aim is to investigate
the extent to which this change is associated with changes in financial capability.

68
 Table 42: Year-on-year changes in reported life satisfaction: BHPS 1991–2006
Life Life satisfaction t
satisfaction
t–1 1 2 3 4 5 6 7 N
1 28.8 15.1 15.2 15.5 11.9 6.7 6.7 570
2 8.1 19.6 26.5 21.2 15.7 6.2 2.7 977
3 3.0 9.9 25.3 30.0 21.4 8.7 1.8 2788
4 1.6 3.2 12.5 32.9 33.6 13.2 3.1 6699
5 0.3 1.2 4.7 15.8 46.8 27.5 3.7 15097
6 0.3 0.4 1.5 6.0 25.9 54.6 11.4 17086
7 0.7 0.4 1.3 3.3 9.3 29.5 55.5 6781
N 546 1008 2922 6903 15271 16995 6353 49998
Notes: Row percentages. Table reads, for example, that 28.8% of individuals with a life satisfaction
score of 1 at year t-1 also had a life satisfaction score of 1 at year t, while 15.1% had a life satisfaction
score of 2 at year t.

Table 43: Year-on-year changes in anxiety/depression: BHPS 1991–2006

Anxiety/depression t
t-1 No Yes N
No 96.5 3.5 90052
Yes 44.9 55.1 6335
N 89775 6612 96387
Notes: Row percentages.

7.3 Financial capability and psychological wellbeing

We now turn to a descriptive analysis of the relationship between financial
capability and psychological wellbeing. This is the first analytical step towards
developing an understanding of the links between financial management and
capability and psychological health and emotional wellbeing. Descriptive analysis
will not explore the effects of mediating factors (such as the role of education,
employment status, age etc). This will be investigated in the next section.

Table 44: Correlations between within-individual variances in measures of

financial incapability and psychological wellbeing: BHPS 1991–2006
Within-individual Within-individual variance in:
Variance in: GHQ scores Life satisfaction Anxiety/depression
Income-adjusted 0.264 0.194 0.161
Income-unadjusted 0.255 0.187 0.159
N. financial problems 0.266 0.193 0.165
Notes: Spearman rank correlation coefficients.

As a first step, and to establish how the year-on-year changes at the individual level
in psychological wellbeing is related to changes in financial incapability between
one year to the next, in Table 44 we present Spearman rank correlation coefficients
between each set of measures. This shows positive correlations between the within-
individual variances in financial incapability and psychological wellbeing, which is

69
largest with GHQ scores. Therefore, at the individual level there is considerable
change in both financial capability and in psychological wellbeing between one year
and the next, and in addition individuals whose financial capability varies a lot from
year to year also have high year-on-year variability in psychological wellbeing.

Financial capability and GHQ

Table 45 presents Spearman rank correlation coefficients to illustrate the degree of
association between our three measures of financial incapability and GHQ scores.
The correlation coefficients are similar, 0.204 with the income-unadjusted index and
0.18 with the income-adjusted index and the number of financial problems.
Therefore higher financial incapability is associated with higher mental stress. There
is evidence that the degree of association fell between 1991 and 2001 (for example,
the correlation between GHQ scores and the income-adjusted measure fell from
0.225 to 0.165 over this period), although it has since strengthened.

Table 45: Relationship between indices of financial incapability and GHQ

scores: BHPS 1991–2006
Year Average
1991 1996 2001 2006
Income-adjusted 0.225 0.196 0.165 0.178 0.179
Income-unadjusted 0.244 0.223 0.188 0.208 0.204
N. financial problems 0.226 0.194 0.154 0.174 0.177
N 8284 7875 7436 6842 120027
Notes: Figures reported are Spearman rank correlation coefficients. See text for how variables are constructed
and defined.

Figure 17 investigates the relationships in more detail, by plotting average financial

incapability by GHQ score. This suggests that financial incapability is constant when
GHQ scores are very low (indicating low mental stress and high psychological
wellbeing). However, financial incapability increases linearly with GHQ scores
between scores of 6 and 27. Therefore, for 90% of observations, higher GHQ scores
(and higher mental stress) are associated with higher financial incapability. This
relationship becomes less smooth when GHQ scores exceed 27, but here sample sizes
are relatively small (see Figure 12). Therefore from this we conclude that there is a
clear positive relationship between levels financial incapability and mental stress.

We now turn to whether changes in GHQ scores are related to changes in financial
capability – that is, do individuals who experience changes in their financial
capability between one year and the next also experience changes in their
psychological wellbeing at the same time? Figure 18 plots average changes in GHQ
by average changes in financial incapability. The plots suggest a positive
relationship between changes in GHQ and changes in incapability: increases in
financial incapability are associated with increases in GHQ scores. This relationship
is almost linear when using the number of financial problems, but is also apparent
when using income-adjusted and income-unadjusted indices. Figures 8 and 9 showed
that the vast majority of year-on-year changes in financial incapability lie in the
range –2 to 2, and within this range the positive relationship is evident. Outside
this range, the sample sizes become small, resulting in more fluctuations. This is
evidence of a positive relationship between changes in financial capability and

70
changes in psychological wellbeing, which we test using multivariate analysis in
subsequent sections.

Figure 17: Mean financial incapability by GHQ scores: BHPS 1991–2006

2.5
Financial incapability/problems
.5 1 01.5 2

0 6 12 18 24 30 36
GHQ score

Income adjusted Number of problems

Income unadjusted
Source: BHPS 1991-2006

71
 Figure 18: Changes in GHQ scores by changes in financial incapability: BHPS
1991–2006

4 2
Change in GHQ
0-2
-4

-4 -2 0 2 4
Change in financial incapability

Income unadjusted Income adjusted

N financial problems
Source: BHPS 1991-2006

Financial capability and life satisfaction

Table 46 summarises our measures of financial capability by life satisfaction scores,
and shows significant differences for all three measures. That is, we find that
average financial capability varies significantly by reported life satisfaction. (Recall
that the life satisfaction question was not asked before wave 6 of the BHPS and
therefore no data is available before then.)

72
 Table 46: Financial incapability and life satisfaction scores: BHPS 1991–2006
Year Average
1996 2000 2006
Income-adjusted
Not satisfied at all 0.405 0.313 0.537 0.376 *
2 0.280 0.354 0.389 0.280
3 0.180 0.138 0.190 0.167
4 0.093 0.071 0.063 0.058
5 –0.036 –0.044 –0.026 –0.049
6 –0.107 –0.122 –0.117 –0.123
Completely satisfied –0.138 –0.198 –0.205 –0.186
Spearman correlation –0.228 –0.217 –0.248 –0.216
Income-unadjusted
Not satisfied at all 0.501 0.408 0.606 0.465 *
2 0.349 0.380 0.432 0.319
3 0.217 0.176 0.207 0.194
4 0.126 0.080 0.068 0.074
5 –0.044 –0.063 –0.058 –0.069
6 –0.122 –0.154 –0.157 –0.150
Completely satisfied –0.091 –0.161 –0.184 –0.147
Spearman correlation –0.248 –0.228 –0.234 –0.220
N financial problems
Not satisfied at all 2.181 1.962 2.294 2.061 *
2 1.899 1.852 1.920 1.771
3 1.563 1.480 1.537 1.497
4 1.337 1.225 1.246 1.235
5 0.999 0.952 0.987 0.955
6 0.852 0.803 0.822 0.821
Completely satisfied 0.969 0.827 0.820 0.874
Spearman correlation –0.180 –0.180 –0.192 –0.169
N 7946 7592 6849 73617
Notes: Weighted using cross-sectional weights. Table reads, for example, that in
1991 adults in the BHPS sample who reported a life satisfaction score of 1 had a
mean income-adjusted financial capability score of 0.405, compared to –0.138 for
those who reported a life satisfaction score of 7. ‘Average’ shows data pooled from
waves 1 to 16. * indicates that the average scores by reported life satisfaction over
the sample period are significantly different at the 5% level.

The table reveals a similar pattern for all three measures of financial capability, with
low life satisfaction associated with higher financial incapability, with Spearman rank
correlation coefficients ranging from –0.169 with number of financial problems to –
0.220 with the income-unadjusted index. Furthermore, the average level of financial
incapability falls monotonically as reported life satisfaction increases and this too is
evident with all three incapability measures and across the sample period. For example,
people reported not being satisfied at all with their life have, on average, an income-
adjusted financial incapability index of 0.376. This falls to 0.058 for those who report a
life satisfaction score of four and to –0.186 for those who report being completely

73
satisfied with their life. Therefore, we conclude that there is a clear relationship
between financial incapability and reported life satisfaction.

Figure 19 examines the relationship between year-on-year changes in financial

incapability and year-on-year changes in reported life satisfaction. This suggests a
negative association – that is increases in financial incapability are associated with
falls in reported life satisfaction. Again, an almost linear relationship is apparent
between changes in the number of financial problems and changes in life
satisfaction, but a similar trend is also evident between the income-adjusted and
income-unadjusted indices. This is particularly apparent for year-on-year changes in
financial incapability lying in the range from -2 and 2 which account for the
majority of the observations. Outside this range, sample sizes are small resulting in
greater fluctuations. From this, we conclude that increases in financial incapability
are associated with falls in reported life-satisfaction.

Figure 19: Changes in life satisfaction scores by changes in financial

incapability: BHPS 1991–2006
.5
Change in life satisfaction
0
-.5

-4 -2 0 2 4
Change in financial incapability

Income unadjusted Income adjusted

N financial problems
Source: BHPS 1991-2006

Financial capability and anxiety/depression

Table 47 summarises financial capability by whether or not people are suffering from
a health problem related to anxiety or depression. Again we find a statistically
significant association, with individuals reporting such a health problem having
higher financial incapability across all three measures. For example, on average
people who reported having a health problem associated with anxiety or depression
had an income-adjusted financial incapability index score of 0.146, compared with –
0.025 among those who reported no such health problem. A similar pattern is
apparent for all three measures of financial incapability. This is further evidence of
a clear relationship between financial incapability and psychological wellbeing.
74
 Table 47: Financial incapability and suffering from anxiety or depression: BHPS
1991–2006
Year Average
1991 1996 2001 2006
Income-adjusted
Suffers anxiety/depression 0.293 0.219 0.078 0.138 0.146 *
Does not suffer 0.095 –0.046 –0.094 –0.050 –0.025
Income-unadjusted
Suffers anxiety/depression 0.379 0.280 0.127 0.160 0.202 *
Does not suffer 0.135 –0.037 –0.097 –0.052 –0.021
N. financial problems
Suffers anxiety/depression 1.785 1.699 1.367 1.436 1.516 *
Does not suffer 1.321 1.029 0.926 0.986 1.062
N 8489 8009 7542 6971 122,132
Notes: Weighted using cross-sectional weights. Table reads, for example, that in 1991 adults in the BHPS
sample who reported suffering from anxiety or depression had a mean income-adjusted financial
capability score of 0.293, compared to 0.095 for those who did not report suffering from anxiety or
depression. ‘Average’ shows data pooled from waves 1 to 16. * indicates that the average scores by
reported life satisfaction over the sample period are significantly different at the 5% level.

To examine the associations between changes in financial capability and changes in

suffering from anxiety or depression, Table 48 summarises average year-on-year
changes in financial capability by whether or not individuals suffer from anxiety or
depression in each year. This shows that individuals who do not suffer from anxiety
or depression in either year experience changes in the financial capability similar to
the sample as a whole, and on average, have lower financial incapability than the
sample average. Those that suffer from anxiety or depression in both years have
higher than average financial incapability, although their financial capability
improves by more than the sample average over the year. Those that start suffering
from anxiety or depression have above average financial incapability, and their
financial incapability increases. Finally, individuals that stop suffering from anxiety
or depression have above average financial incapability but experience an above
average improvement in capability. Therefore, as with the other measures of
psychological wellbeing, we find that starting to suffer from anxiety or depression is
associated with higher, and deteriorating, financial capability while recovering from
such a health problem is associated with improving financial capability.

This descriptive analysis provides evidence of a strong association between both

financial incapability and psychological wellbeing, and also between changes in
financial incapability and changes in psychological wellbeing using all three indices
of incapability and all three measures of wellbeing. We find that higher financial
incapability is associated with higher mental stress, lower reported life satisfaction,
and health problems associated with anxiety or depression. However, as we have
seen, financial capability is strongly associated with a range of other individual and
household characteristics that are also likely to affect people’s psychological
wellbeing, such as marital status, employment status etc. There may be mediating
variables that jointly determine an individual’s financial capability and their
psychological wellbeing. In the next section, we attempt to disentangle these
associations using multivariate analysis.

75
Table 48: Changes in suffering from anxiety or depression and changes in
financial incapability: BHPS 1991–2006
Means of indices
t–1 T Change B
Sample average 95935
Income-adjusted –0.020 –0.032 –0.012
Income-unadjusted –0.013 –0.028 –0.015
Number financial problems 1.077 1.053 –0.024
Not suffered in either period 86921
Income-adjusted –0.026 –0.037 –0.011
Income-unadjusted –0.032 –0.046 –0.014
Number financial problems 1.034 1.012 –0.023
Suffered in both periods 3481
Income-adjusted 0.178 0.153 –0.025
Income-unadjusted 0.234 0.205 –0.029
Number financial problems 1.580 1.522 –0.058
Not in t–1, but suffered in t 3131
Income-adjusted 0.106 0.139 0.032
Income-unadjusted 0.144 0.175 0.031
Number financial problems 1.388 1.451 0.063
Suffered in t–1, but not at t 2854
Income-adjusted 0.151 0.076 –0.074
Income-unadjusted 0.188 0.111 –0.077
Number financial problems 1.469 1.314 –0.156
Notes: Table reads that individuals who did not suffer a health problem associated with
anxiety or depression at year t–1 or year t on average experienced a fall in their income-
adjusted financial incapability from –0.026 to –0.037.

76
8 Estimating the effect of financial capability on
psychological wellbeing

10 Estimation procedures
The final stage of the analysis investigates the relationships between financial
capability and psychological wellbeing in more detail. Of particular interest is
whether the index of financial capability has any power in predicting psychological
wellbeing. There are two problems which need to be addressed in attempting to
answer this question. The first is that there are likely to be both mediating and
confounding factors that are associated both with an individual’s level of financial
capability and his/her level of psychological wellbeing. Descriptive statistics in
previous sections suggest that this is indeed the case. The second is that there are
also likely to be both unobservable factors (such as ability, personality, ambition or
motivation) and unobserved factors (such as an individual’s attitude towards risk)
that are similarly associated with both financial incapability and psychological
wellbeing. Our estimation procedure attempts to deal with both these issues.

We use multivariate panel data models, and fixed effects models in particular.
Multivariate analysis allows us to control for other (observable) characteristics of
individuals and the households that they live in that might be correlated with both
psychological wellbeing and financial capability (such as age, gender, marital status,
employment status, income, housing tenure, family type etc). The BHPS is a particularly
rich source of a wide range of such characteristics, allowing a more reliable coefficient
on the variables of interest to be estimated. We can write the model to be estimated as
the following, where y is our measure of psychological wellbeing, x our measure of
financial capability and z is a vector of other control variables:

yit = xit β + zit δ + ε it [1]

ε it = η i + hit [2]

å is the error term. Estimating [1] using simple Ordinary Least Squares (OLS)
regression ignores any individual-specific characteristics that are included in ε.
These can be separated, as in [2] where ç is a time invariant individual-specific
effect capturing unobservable (or unmeasured) characteristics. If this is correlated
with the observable x and/or z, then estimating [1] using OLS will yield biased
estimates. This is likely to be especially important in the current context, as latent
time-invariant psychological characteristics have been found to systematically
influence reported mental wellbeing scores (De Neve and Cooper 1999), and
estimation methods that do not allow for such time-invariant unobserved traits are
likely to result in biased estimates. For example, if individuals are innately
optimistic (pessimistic) they are more likely to both report being in a good (bad)
financial situation and also more likely to report high (low) psychological wellbeing.

77
Panel data models allow us to control for the effects of unobserved variables that
are fixed over time, and that might also be correlated with other explanatory
variables. Furthermore, fixed effects models allow such traits to be arbitrarily
correlated with the observable characteristics. This may be important if, for
example, more optimistic or more motivated people are also more likely to get
married, be in employment or have higher qualification levels. Such models are
estimated by taking deviations from individual-specific means over time in both the
dependent variable and explanatory variables, and therefore removing the effect of
time invariant characteristics, so we estimate:

yit − y i = β ( xit − x i ) + δ ( zit − z i ) + hit [3]

Therefore, a positive value for β would imply that higher values of x are associated
with higher values of y, while a negative â indicates that a higher x is associated
with a lower y. We estimate whether an individual’s level of psychological wellbeing
varies systematically with their financial capability, controlling for changes in a
wide range of personal, household, family, and housing-related characteristics.

Such models are appropriate when the dependent variable (our measure of
psychological wellbeing) is continuous. For the purposes of this report, we assume
that GHQ scores and reported life satisfaction are continuous variables, and
therefore use within group fixed effects models to estimate the impact of financial
incapability on wellbeing. However, such models are less appropriate when the
dependent variable is categorical, as is the case with the binary variable indicating
whether or not the individual has a health problem related to anxiety or depression.
Binary dependent variable models are more appropriate in these circumstances, and
therefore we estimate fixed effects (or so-called conditional) logit models. The
model specification can be written:

Pr( yit = 1 xit , zit ) = F (ηi + βxit + δzit ) [4]

where F(•) is the cumulative logistic distribution. A feature of this approach is that
when y = 0 or y = 1 for all observations for an individual, this individual’s
contribution to the log-likelihood is zero and their data does not contribute to the
estimation. Therefore, estimation of the impact of financial capability on anxiety or
depression is identified solely by individuals whose anxiety/depression status
changes over time.

There are two issues concerning fixed effects models. The first is that they do not
allow for the impact of time-invariant observable characteristics (e.g. ethnicity,
gender etc.) to be estimated. We estimate models with both men and women
combined as well as separate models for each sex to examine whether the effects of
financial capability on wellbeing differs for men and women.

The second issue is that, although fixed effects models allow for time-invariant
unobserved characteristics, and allow these characteristics to be correlated with
observed characteristics (such as personality traits), they do not account for
unobserved shocks that affect both the dependent variable and the explanatory
variables of interest. So, for example, if individuals with particular characteristics
78
experienced an unobserved event that affected both their financial capability and
their psychological wellbeing, the estimated coefficients would be biased. However,
this problem is shared by all other existing estimation methods. We attempt to allow
for any such potential shocks (or changes in an individuals level of optimism or
pessimism) by including in all our model specifications a measure of whether people
expect their financial situation to improve or worsen in the forthcoming year.

10 Estimation results
GHQ
We first focus on the results with GHQ scores as the dependent variable. We
present estimates on variables of interest in Table 49, while a full set of
estimates is available in the Appendix. Model [1] includes the income-adjusted
index of financial incapability as an explanatory variable, Model [2] includes the
income-unadjusted index, while Model [3] includes the number of financial
problems. The table indicates that the financial incapability measures have a
positive and statistically significant impact on GHQ scores, even when
controlling for a large number of potentially confounding and mediating factors
8
as well as time-invariant unobserved effects.

8
The large t-statistics indicate these estimates are very precisely determined. This precision is likely
to be caused by the high degree of year-on-year fluctuation in both the financial incapability indices
and psychological wellbeing, which our analyses have shown to be experienced by the same
individuals. It is also likely to reflect the fact that our measures of psychological health (and the
GHQ score in particular) tends to emphasise recent events (although very similar results were
obtained using the 12-point GHQ score which maybe less sensitive to short-term fluctuations).
Further investigation suggests that the main driver behind the relationship between financial
incapability and psychological wellbeing is an individual’s perceived current financial situation –
individuals who report finding it quite or very difficult have significantly lower levels of
psychological wellbeing. Our estimation procedure accounts for time-invariant unobserved effects
(such as personality traits), and so this is not caused by innately more optimistic individuals
reporting a better financial situation and higher levels of psychological wellbeing. It is possible,
however, that individuals may experience unobserved events that affect both their perceived
financial situation and their wellbeing, or that individuals’ perceptions and psychological health are
strongly influenced by changes in the real-world financial situation.
79
 Table 49: Within-group fixed effects estimates from the GHQ model
Model [1] Model [2] Model [3]
Income-adjusted index 1.074
[38.62]
Income-unadjusted index 1.092
[39.30]
N financial problems 0.637
[43.53]
Real equiv hh income pcm –0.034 0.015 0.015
(x 1000) [3.16] [1.39] [1.43]
Amount saved pcm 0.036 0.043 0.288
(x 1000) [0.45] [0.53] [3.58]
In good health –1.729 –1.729 –1.723
[46.53] [46.52] [46.44]
Widowed 1.774 1.807 1.767
[10.43] [10.62] [10.40]
Divorced/separated 0.920 0.930 0.897
[7.24] [7.32] [7.07]
Unemployed 1.187 1.114 0.955
[13.98] [13.11] [11.22]
Retired –0.374 –0.441 –0.528
[4.61] [5.42] [6.50]
Inactive not like job 0.383 0.332 0.250
[6.26] [5.41] [4.08]
Inactive like job 0.783 0.723 0.615
[9.64] [8.90] [7.57]
R-squared 0.0510 0.0515 0.0548
N observations 114190
N individuals 15974
Notes: Estimates from within-group fixed effects regressions with GHQ score as
the dependent variable. Absolute ratio of coefficient to standard error in
brackets. Table presents estimates for variables of interest; full estimation
results are presented in the Appendix. Reference categories are in fair or poor
health; single never married; in full-time employment.

The estimated coefficient on the income-adjusted index is 1.074 with a t-statistic of

almost 39, while on the income-unadjusted index the coefficient is 1.092 with a t-
statistic that exceeds 39. The size of the effects indicates that a one standard
deviation decrease (equal to about a fall of 0.6 - see Tables 10 and 17) in the
income-adjusted index of financial incapability would reduce GHQ scores by 0.619
GHQ points (or 5.6% at the sample means). Similarly, a one standard deviation
reduction in the income-unadjusted index reduces GHQ scores by 0.65 GHQ points
(or 5.9% at the sample means). To highlight the effect of reducing or increasing the
index of financial incapability by one standard deviation, Figure 20 below plots the
cumulative distribution of the income-adjusted index.

80
 Figure 20: Cumulative distribution function of income-adjusted index of
financial incapability: BHPS 1991–2006

1
.9
.3 .4 .5 .6 .7 .8
Cumulative distribution
.2
.1
0

-2 -1 0 1 2 3 4
Income-adjusted financial incapability
Source: BHPS 1991-2006

This shows that a standard deviation reduction in financial incapability (which

equates to a reduction in the index of about 0.6 – see Table 17) are approximately
equivalent to moving from the 90th percentile of the financial incapability index to
the 50th percentile. Therefore, it moves an individual from having relatively high
levels of financial incapability to having average levels of financial incapability.
Doing so reduces their GHQ score by 0.62 to 0.65 GHQ points (almost 6%). A
reduction of one financial problem also reduces GHQ scores by 0.637 GHQ points (or
5.7% at the sample means). This evidence supports the hypothesis that financial
capability predicts psychological wellbeing and that improving people’s financial
capability leads to improving their psychological wellbeing.9

These effects are relatively large compared to the effects of other variables known to
affect psychological wellbeing. For example, individuals who report themselves to be
in good or very good health have GHQ scores 1.72 points (or 15.5%) lower than
those in fair, poor or very poor health. Widowhood has the single largest impact on
GHQ scores, and is associated with an increase of 1.8 GHQ points (or 16.2% at the
sample means) relative to people who are single never married, while divorce or

9
We have also estimated models where we allow psychological health at time t to be affected by
financial capability at time t–1, again using within-group fixed effects. The results indicate that
those with high levels of financial incapability at t–1 have lower levels of psychological health at t,
although the size and strength of the association is much lower than those presented here. This is to
be expected given the considerable within-individual year-on-year variation in both financial
incapability and psychological health (both show a Spearman rank correlation coefficient of 0.5
between a person’s current value and that reported last year) – if psychological health is directly
affected by financial capability and if an individual’s financial capability fluctuates from year to year,
then we expect a stronger relationship to emerge when both are measured at the same time.
81
separation is associated with an increase of 0.92 GHQ points (or about 8.3% at the
sample means). Being unemployed also has a large impact on psychological
wellbeing, increasing GHQ scores by 1.1 GHQ points relative to people in full-time
work. These characteristics are those most commonly associated with psychological
ill health, and a comparison of the relative sizes of effects suggests that moving an
individual from average levels of financial capability to relatively low levels of
financial capability (or the experience of an additional financial problem) has an
impact almost comparable to that of divorce or separation.

The table also presents the impact of other covariates of interest on GHQ scores. We
find that income has no effect on GHQ scores in Models [2] and [3], while it has a
negative and statistically significant impact in Model [1]. This is to be expected,
given that the income-adjusted index of financial incapability used in Model [1] is
independent of income by design, while the indices used in Models [2] and [3] still
capture some of the effects of income. In Model [1], the estimates suggest that a
£1000 increase in household income reduces mental stress by 0.035 GHQ points.
Controlling for financial incapability and income, the amount saved per month has
no impact on GHQ scores. (However in Model [3] we find that a £1000 increase in
savings per month increases mental stress, which is difficult to explain, although
there may be problems of co-linearity with the number of financial problems.) Being
retired improves people’s mental wellbeing, reducing GHQ scores by between 0.37
and 0.44 points (or 3–4% at sample means). Economic inactivity increases mental
stress, particular if the individual would like to work. In this case economic
inactivity increases GHQ scores by an average of between 0.72 and 0.78 GHQ points.

Table 50: Gender-specific within-group fixed effects estimates from the GHQ
model
Men Women
Income adjusted 1.094 [28.13 1.055 [26.91]
R-squared 0.0560 0.0502

Income unadjusted 1.103 [28.26 1.081 [27.59]

]
R-squared 0.0562 0.0509

N financial problems 0.643 [31.48 0.631 [30.53]

]
R-squared 0.0601 0.0539
N observations 52308 61882
N individuals 7657 8317
Notes: Estimates from within-group fixed effects regressions with GHQ score
as the dependent variable. Absolute ratio of coefficient to standard error in
brackets. Table presents estimates for variables of interest only.

One of the limitations of the within-group fixed effects approach is that we cannot
identify the effect of time-invariant covariates (such as gender) on the outcome of
interest. To identify whether the impacts of financial incapability on GHQ scores
differ by gender, we have estimated separate models for men and women, and
present the coefficients of interest in Table 50. This indicates that for men the
coefficient on the income-adjusted index of financial incapability is 1.094, while for
82
women it is 1.055. In both cases it is highly statistically significant, although these
coefficients are not significantly different from each other. The same is true of the
gender-specific coefficients on the other financial incapability measures, from which
we conclude that the impact of financial incapability on psychological wellbeing
does not differ for men and women.

Life satisfaction
Having established that financial incapability is associated with increases in mental
stress as measured by GHQ scores, we next examine its impact on life satisfaction.
We present estimates on variables of interest in Table 51, while a full set of
estimates is again available in the Appendix. As previously, Model [1] includes the
income-adjusted index of financial incapability as an explanatory variable, model [2]
includes the income-unadjusted index, while Model [3] includes the number of
financial problems.

The table indicates that the financial incapability measures have a negative and
statistically significant impact on reported life satisfaction scores, even when
controlling for a large number of potentially confounding and mediating factors as
well as time-invariant unobserved effects. The estimated coefficient on the income-
adjusted index is –0.226 with a t-statistic of almost 27, while on the income-
unadjusted index the coefficient is –0.231 with a t-statistic that exceeds 27. The
number of financial problems has a coefficient of –0.122 with a t-statistic of 28.
These results are consistent with changes in financial capability leading to changes
in life satisfaction.

Again, we try to quantify the relative sizes of these effects. The coefficients imply
that a one standard deviation fall in the income-adjusted index of financial
incapability increases life satisfaction scores by 0.119 (or 2.3% at the sample
means). Similarly, a one standard deviation fall in the income-unadjusted index
increases reported life satisfaction by 0.125 (or 2.4% at the sample means).
th th
Therefore, moving an individual from the 90 percentile to the 50 percentile in the
financial incapability distribution (or from relatively low levels of financial
capability to average levels) would increase their life satisfaction by between 0.12
and 0.13 (about 2.5%). A reduction of one financial problem also increases life
satisfaction by 0.122 (or 2.4% at the sample means).

These effects are relatively large compared to the effects of other variables. For
example, as in the GHQ models, being in good health, widowhood and
unemployment have the largest impacts on life satisfaction. Good health is
associated with an improvement of about 0.28 (or 5.4% at the sample means)
relative to being in fair or poor health, widowhood and unemployment are
associated with a fall in reported life satisfaction of –0.23 (4.4%) relative to a
person who has never been married and in full-time work respectively, while
divorce or separation is associated with a fall in reported life satisfaction of 0.18
(or about 3.4%). Again, we find that moving an individual from average levels of
financial capability to relatively low levels (or the addition of one more financial
problem) has an effect on life satisfaction that is comparable to divorce or
separation, but which is smaller than the impacts of good health, being widowed
or being in unemployment.

83
As with the GHQ models, and as we expect, household income has an effect only in
the model including the income-adjusted index of financial incapability. In this
specification, a £1000 increase in income increases life satisfaction by 0.012, or
0.2% at the sample means. The amount saved has no significant effect on life
satisfaction, while life satisfaction is higher for the retired. Being retired is a
associated with life satisfaction scores about 0.16 points higher than full-time
employment, or about 3% at the sample means. Being in economic inactivity
reduces life satisfaction but only if a job is desired. In this case life satisfaction is
reduced by –0.10 points (about 2%).

Table 51: Within-group fixed effects estimates from the life satisfaction model
Model [1] Model [2] Model [3]
Income-adjusted index –0.226
[26.91]
Income-unadjusted index –0.231
[27.43]
N financial problems –0.122
[28.14]
Real equiv hh income 0.012 0.003 0.003
(x 1000) [4.18] [0.97] [1.04]
Amount saved pcm 0.025 0.023 –0.015
(x 1000) [1.18] [1.09] [0.68]
In good health 0.281 0.281 0.281
[27.28] [27.28] [27.26]
Widowed –0.230 –0.239 –0.233
[4.60] [4.77] [4.65]
Divorced/separated –0.177 –0.179 –0.175
[4.82] [4.87] [4.75]
Unemployed –0.231 –0.214 –0.191
[8.88] [8.21] [7.32]
Retired 0.157 0.173 0.185
[6.65] [7.30] [7.82]
Inactive not like job 0.008 0.020 0.031
[0.46] [1.12] [1.78]
Inactive like job –0.104 –0.090 –0.075
[4.27] [3.71] [3.07]
R-squared 0.0397 0.0402 0.0408
N observations 73345
N individuals 12640
Notes: Estimates from within-group fixed effects regressions with life
satisfaction score as the dependent variable. Absolute ratio of coefficient to
standard error in brackets. Table presents estimates for variables of interest; full
estimation results are presented in the Appendix. Reference categories are in
fair or poor health; single never married; in full-time employment.

In Table 52 we present the results from estimating separate models for men and women.
This indicates that for men the coefficient on the income-adjusted index of financial
incapability is –0.229, while for women it is –0.223. In both cases it is highly statistically
significant, although tests indicate that these coefficients are not significantly different
from each other. The same is true of the gender-specific coefficients on the other financial
84
incapability measures, from which we conclude that the impact of financial incapability on
life satisfaction does not differ between men and women.

Table 52: Gender-specific within-group fixed effects estimates from the life
satisfaction model
Men Women
Income adjusted –0.229 [18.83] –0.223 [19.24]
R-squared 0.0460 0.0381

Income unadjusted –0.233 [19.14] –0.229 [19.68]

R-squared 0.0464 0.0386

N financial problems –0.123 [19.73] –0.121 [20.07]

R-squared 0.0472 0.0390
N observations 33594 39751
N individuals 5977 6663
Notes: Estimates from within-group fixed effects regressions with life satisfaction
score as the dependent variable. Absolute ratio of coefficient to standard error in
brackets. Table presents estimates for variables of interest only.

Anxiety or depression
Our final measure of psychological wellbeing focuses on whether or not individuals
have a health problem that relates to anxiety or depression. Because this variable is
binary rather than continuous, we estimate the impacts of financial incapability
using a fixed effects (conditional) logit specification. Recall that in this estimation
procedure, the impact of financial capability on suffering anxiety or depression is
identified solely by individuals whose anxiety/depression status changes over time,
and therefore the sample sizes in these tables are much smaller than in previous
ones. The results, shown in Table 53, are consistent with those using both GHQ and
life satisfaction as measures of psychological wellbeing. In particular we find that
higher financial incapability is associated with a higher probability of having a
health problem related to anxiety or depression – in each case the coefficients on
the financial incapability measures are positive and statistically significant.

These coefficients are difficult to interpret, in terms of the size of the effect that
changes financial incapability have on the probability of suffering a health problem
that relates to anxiety or depression. The coefficients on the income-adjusted and
income-unadjusted indices suggest that a one standard deviation reduction in
th th
financial incapability (equivalent to moving an individual from the 90 to the 50
percentile of the distribution – from high levels of incapability to average levels)
reduces the probability of an individual suffering a health problem related to
anxiety or depression by 15% at the sample means.

Coefficients on the other covariates are also generally consistent with those
presented in previous models. For example, we find that good health significantly
reduces the probability of suffering from anxiety or depression (by 70%) relative to
a person in fair or poor health, while widowhood, divorce or separation, or
unemployment significantly increases the probability of suffering anxiety or
depression (by 51%, 35% and 43% respectively relative to never having married or
85
being in full-time work). However, in contrast to the previous models we find that
being retired reduces psychological wellbeing, in that it significantly increases the
probability of suffering anxiety or depression (by 23%). Household income and the
amount saved have no significant impact here.

Table 53: Fixed effects logit estimates from the anxiety or depression model
Model [1] Model [2] Model [3]
Income-adjusted index 0.196
[6.67]
Income-unadjusted index 0.200
[6.81]
N financial problems 0.123
[7.54]
Real equiv hh income 0.004 0.016 0.017
(x 1000) [0.26] [1.08] [1.14]
Amount saved pcm 0.068 0.072 0.129
(x 1000) [0.53] [0.56] [1.07]
In good health –1.220 –1.220 –1.219
[27.73] [27.73] [27.69]
Widowed 0.414 0.419 0.415
[2.16] [2.19] [2.17]
Divorced/separated 0.299 0.299 0.290
[2.03] [2.03] [1.97]
Unemployed 0.355 0.344 0.310
[3.37] [3.27] [2.93]
Retired 0.205 0.196 0.177
[2.03] [1.94] [1.75]
Inactive not like job 0.466 0.459 0.439
[6.09] [6.00] [5.74]
Inactive like job 0.586 0.578 0.555
[6.50] [6.41] [6.14]
Log likelihood –8602.8 –8601.8 –8596.5
N observations 27170
N individuals 2583
Notes: Estimates from fixed effects (conditional) logit regressions with whether
suffers a health problem related to anxiety or depression as the dependent
variable. Absolute ratio of coefficient to standard error in brackets. Table
presents estimates for variables of interest; full estimation results are presented
in the Appendix. Reference categories are being in fair or poor health; single
never married; in full-time employment.

In Table 54 we investigate whether the impact of financial incapability has a

gender-specific impact on the probability of having a health problem related to
anxiety or depression. The results suggest that they do. In particular, we find the
impact of financial incapability to be larger among women than men – the
coefficient for women is significantly larger on the financial incapability indices
than it is for men. For example, in the models estimated for men, the estimated
coefficient is 0.12, compared to 0.24 for women. These indicate moving from
average levels of financial incapability to relatively high levels increases the
probability of anxiety or depression by 17% for women and by 9% for men.

86
 Table 54: Gender-specific fixed effects logit estimates from the anxiety or
depression model
Men Women
Income-adjusted 0.120 [2.27] 0.236 [6.55]
Log likelihood –2578.1 –5957.0

Income-unadjusted 0.124 [2.35] 0.240 [6.66]

Log likelihood –2577.9 –5956.3

N financial problems 0.091 [3.07] 0.140 [7.07]

Log likelihood –2576.0 –5953.5
N observations 8627 18543
N individuals 847 1736
Notes: Estimates from fixed effects (conditional) logit regressions with
whether suffer a health problem related to anxiety or depression as the
dependent variable. Absolute ratio of coefficient to standard error in
brackets. Table presents estimates for variables of interest only.

10 Extensions to the analysis

Our estimates are consistent with financial capability being a good predictor of
psychological wellbeing. As well as the standard models described previously, we
have also estimated some variants that (i) allow for the effect on psychological
wellbeing of financial incapability to be non-linear; and (ii) allow for the effect of
financial incapability to vary according to a range of other observable
characteristics. In these extensions we focus solely on our GHQ score measure of
psychological wellbeing for simplicity and ease of interpretation, although we draw
similar conclusions when using the life satisfaction and having a health problem
related to anxiety or depression measures. Therefore there is no loss of generality.

Non-linear specifications
Table 55 below presents the estimates resulting from entering a quadratic and cubic
index term into the models. We focus only on the coefficients on the index terms
because those on other variables change only marginally in these specifications. We
find that the coefficients on the quadratic and cubic variable are statistically
significant in each specification, suggesting that the impact of financial
incapability on GHQ scores is non-linear. In particular the coefficient on the income-
adjusted index of financial incapability is 1.528, that on the quadratic term is –
0.460 and that on the cubic term is 0.07. This suggests that although GHQ scores
increase with financial incapability, they do so at a declining rate initially, before
increasing again towards the tail of the distribution. The same pattern emerges with
the income-unadjusted index.

87
Table 55: Non-linear effects of financial incapability on GHQ scores
Index of financial incapability N financial
Income Income Problems
adjusted unadjusted
Index 1.528 1.661 0.256
[31.99] [32.86] [4.87]
Index squared –0.460 –0.572 0.151
[8.73] [10.63] [5.77]
Index cubed 0.074 0.097 –0.013
[5.16] [6.95] [4.00]
R-squared 0.0523 0.0532 0.0555
N observations 114190
N individuals 15974
Notes: Estimates from within-group fixed effects regressions with GHQ
score as the dependent variable. Absolute ratio of coefficient to standard
error in brackets. Table presents coefficients on variables of interest.

Figure 21: Estimated effects of financial incapability on GHQ scores:

BHPS 1991–2006
6 8 10 12 14 16 18 20
GHQ Score/percent
4
2
0

-2 -1 0 1 2 3 4 5
Financial incapability/problems

Incomeunadjusted Incomeadjusted
NProblems Distribution of income-adjusted index
Source: BHPS 1991-2006

To highlight this relationship Figure 21 plots these estimates together with the
underlying distribution of the income-adjusted index. The figure shows that the
relationship between financial incapability and GHQ is stronger at the bottom and
at the top of the financial incapability distribution. That is, the effect on GHQ
scores of changes in financial incapability is larger at the bottom of the incapability
distribution. The distribution of the income-adjusted index shows that for most
people an increase in financial capability (a reduction in the index) will improve
their psychological wellbeing (reduce their GHQ scores). However focusing on those
with the highest levels of financial incapability may have less effect. For example,

88
GHQ scores could be reduced by 3 points by decreasing financial incapability from
zero to –1, but only by 2 points by reducing it from four to zero. This may be
related to the fact that those with very high levels of financial incapability are likely
to face a mixture of financial, psychological and other problems in their lives, and
so improving just their financial capability has little overall impact on their mental
wellbeing. This figure also shows that the impact of the number of financial
problems on GHQ scores is almost linear – each additional financial problem suffered
increases mental stress levels by a similar amount across the distribution.

Including interaction terms

We have also investigated the extent to which the impact of financial incapability
varies by other characteristics of individuals, through introducing interaction terms
into the specifications. Table 56 presents the results on the key coefficients of
interest. We have focused on key factors that are known to affect their
psychological wellbeing such as physical health, widowhood, divorce, unemployment
and retirement.

Table 56: Within-group fixed effects estimates from the GHQ model: The impact
of interaction terms
Index of financial incapability N financial
Income Income unadjusted Problems
adjusted
Index 1.238 [23.67] 1.234 [24.00] 0.763 [27.85]
In good health –1.722 [46.31] –1.718 [46.17] –1.440 [29.68]
Widowed 1.744 [10.24] 1.756 [10.32] 1.610 [8.49]
Divorced/separated 0.824 [6.44] 0.812 [6.33] 0.317 [2.23]
Unemployed 1.056 [11.75] 0.952 [10.57] 0.219 [1.65]
Retired –0.406 [4.95] –0.443 [5.45] –0.091 [0.93]
Inactive not like job 0.391 [6.37] 0.341 [5.53] 0.266 [3.40]
Inactive like job 0.784 [9.46] 0.718 [8.49] 0.422 [3.68]
Interactions:
Index*
In good health –0.314 [6.10] –0.309 [6.15] –0.239 [9.01]
Widowed 0.239 [1.59] 0.221 [1.49] 0.086 [1.23]
Divorced/separated 0.465 [5.83] 0.508 [6.51] 0.375 [8.87]
Unemployed 0.417 [4.47] 0.408 [4.46] 0.368 [7.08]
Retired –0.379 [3.40] –0.334 [3.01] –0.371 [7.58]
Inactive not like job –0.094 [1.32] –0.051 [0.72] –0.010 [0.27]
Inactive like job –0.028 [0.32] 0.006 [0.07] 0.105 [2.15]
Constant 9.224 [6.11] 9.047 [6.00] 8.078 [5.37]
R-squared 0.0521 0.0527 0.0578
N observations 114190
N individuals 15974
Notes: Estimates from within-group fixed effects regressions with GHQ score as the dependent variable.
Absolute ratio of coefficient to standard error in brackets. Table presents estimates for variables of
interest only. Reference categories and in fair or poor health, single never married, in full-time
employment.

89
The estimated coefficients on the financial incapability indices show that mental
stress increases significantly with financial incapability, consistent with our earlier
estimates. However the average sizes of these effects are larger (compared with
those presented in Table 49). The estimates suggest that a standard deviation
increase in the income-adjusted index of financial incapability (moving from
average levels to relatively high levels) increases mental stress levels by about 0.7
GHQ points (or 6.3% at the sample means). However, several of the interaction
terms also have statistically significant effects on GHQ scores. For example, being in
good health reduces the impact of financial incapability, the coefficient on the
financial incapability and good health interaction term is negative and statistically
significant. This indicates that the impact of financial incapability on GHQ scores is
lower for individuals who are in good health. The estimates from the income-
adjusted index indicate that a one standard deviation increase in financial
th th
incapability (again in this case equivalent to moving from the 50 to the 90
percentile in the financial incapability distribution – from average levels to
relatively high levels) for someone who is in good health increases their GHQ scores
by 0.53 points (or 4.8%) relative to an otherwise identical person who was in good
health but whose financial incapability level did not change. The interaction term
on retirement has a similar effect – being retired reduces the psychologically
damaging impacts of financial incapability.

Unemployment and being divorced increase the effect of financial incapability, the
coefficients on the financial incapability and divorce and on the financial
incapability and unemployed interaction terms are positive and statistically
significant. This indicates that the impact of financial incapability on GHQ scores is
greater for individuals who are divorced or separated or in unemployment. The
estimates from the income-adjusted index indicate that a one standard deviation
increase in financial incapability for someone who is divorced or separated or
unemployed increases their GHQ scores by about 0.98 points (or 8.8%) relative to an
otherwise identical person who was divorced or unemployed but whose financial
incapability level did not change. This pattern emerges when using the income-
unadjusted index of financial incapability. Therefore financial incapability
compounds the already psychologically harmful effects of job loss or marital
dissolution, while being in good health or retired reduces the psychologically
damaging impacts of financial incapability.

The results when using the number of financial problems follow similar patterns. The
coefficients from this model indicate that increasing the number of financial problems
that an individual suffers by one results in increasing that individual’s GHQ score by
0.76 points. This increase is larger if that individual is also divorced or unemployed,
to the extent that one additional financial problem in these circumstances raises GHQ
scores by 1.1 points (or 10% at the sample means). Being in good health and
retirement reduces the effect of financial incapability on mental stress.

90
9 Summary and conclusions
The aim of this project is to investigate the relationships between financial
capability and psychological wellbeing in Britain, and in particular to establish
whether low psychological wellbeing is predicted by low financial capability, all else
equal. This hypothesised link between a low level of financial capability and poor
psychological wellbeing is the central issue.

We create an index of financial incapability, which we adjust for income, using

variables measuring perceived current financial situation; reporting that the
situation worsened since last year; whether respondent saves; whether household
has housing payment problems; whether such problems have required borrowing;
whether they have required cutbacks; and whether the household has been two or
more months in housing arrears in the previous 12 months. As an alternative, and as
part of the validity checking process, we have constructed a simpler measure that
counts the number of financial problems each individual is facing.

People with the highest financial incapability tend to be young (aged less than 35),
divorced or separated, have more than one or two dependent children, are single
non-elderly, lone parents, in fair or poor health, live in rented accommodation and
are unemployed or economically inactive but would like a job. In contrast people
with the lowest levels of financial incapability are on average older (aged 55 or
above), married or widowed with no dependent children, in good health, home
owners and working in a full-time permanent job. These findings are consistent with
those from the Financial Services Authority’s Baseline Survey. Taking advantage of
the panel nature of the data reveals that getting married, improvements in health,
becoming a home owner and entering work are associated with increasing financial
capability, while death of a spouse, marital dissolution, an additional child,
deterioration in health and unemployment are associated with falls in financial
capability. Furthermore, and most importantly within the context of this report,
higher financial incapability is associated with higher mental stress, lower reported
life satisfaction, and health problems associated with anxiety or depression.

Estimates from panel data models indicate that, even controlling for a range of
observable individual and household characteristics, and time-invariant unobserved
effects, people’s psychological wellbeing is strongly related to their financial
capability. This is consistent with financial capability being a good predictor of
psychological wellbeing. For example, moving an individual from very high relative
levels of financial incapability to average levels reduces their GHQ score by about
0.65 GHQ points (or almost 6%), increases their reported life satisfaction by 0.12
(or 2.4%), and reduces the probability of an individual suffering a health problem
related to anxiety or depression by 15%. The relationship between financial
incapability and psychological wellbeing varies over the distribution, and in
particular is strongest at the bottom of the financial incapability distribution. This
implies that increasing financial capability will improve the psychological wellbeing
of most people but may have less effect among those with the highest levels of
financial incapability. The impact of financial capability on psychological wellbeing
differs across different population groups. For example, being in good health or

91
being retired reduces the impact of financial incapability on psychological
wellbeing, while unemployment and being divorced increase the effect of financial
incapability on wellbeing.

The results from our analysis lead us to conclude that financial capability has a relative
large and statistically significant impact on psychological wellbeing. This would indeed
suggest that improving people’s financial management skills would have substantial
effects on stress-related illnesses and outcomes associated with such problems. The FSA is
committed to identifying and promoting financial capability programmes which create
confident, capable consumers, and doing so in ways that people will most understand, in
places most useful to them. The indication from this research that financial capability
has a relatively large and statistically significant impact on psychological wellbeing will
be valuable in the formation and implementation of the FSA’s financial capability policy.
The findings will contribute to the targeting of financial capability policy, to the
formation of effective relationships with stakeholders and trusted intermediaries and to
the successful measurement and evaluation of the FSA’s work. Improving the
engagement and effectiveness of financial capability is important at any time, but it is
particularly crucial in the current economic downturn, when an increasing number of the
population are experiencing anxiety and stress about managing their finances.

However, a number of further questions emerge from these analyses. The first is the
extent to which financial capability is related to favourable economic circumstances
or to financial management skills. We have modelled psychological wellbeing as a
function of financial capability (and found a relationship), but have not modelled
the determinants of financial capability itself (although we have discussed bivariate
relationships with other observable characteristics and found strong associations).
Our results suggest that a person’s financial capability varies considerably between
one year and the next, and this also has implications. If, for example, financial
capability at the individual level is highly variable from one year to the next in an
unpredictable way, then this makes it harder to design policies to improve it.

The second is the complex relationship between an individual’s income, their

financial management skills and their savings behaviour. For example, higher
income does not necessarily reflect higher financial capability (and our analysis
indicates this relationship is not linear) as it implies larger financial
responsibilities. This is partly reflected in the relationship between savings
behaviour and financial capability – the incidence of saving is a more important
indicator of financial capability than the amount saved. Our analysis touches on
this, but it deserves further attention.

Finally, and crucially for our results, is the extent to which people experience shocks
or events that we do not observe that might affect both their financial capability
and psychological wellbeing. Our estimation procedure allows for time invariant
unobserved or unobservable characteristics of individuals, such as personality traits,
which may affect both financial capability and psychological wellbeing. However, if
there are particular events that people experience, but that we are unable to
capture in our data, that reduce their financial capability levels and also reduce
their psychological wellbeing, then these may confound the effects we found using
statistical models.

92
10 References
Argyle, M. (1989), The psychology of Happiness, Routledge: London.
Atkinson, A., McKay, S., Kempson, E., and Collard, S. (2006). ‘Levels of Financial
Capability in the UK. Results of a baseline survey.’ Bristol: University of
Bristol.
Capellari, L. and Jenkins, S.P. (2007), ‘Summarising multiple deprivation indicators’,
in Inequality and Poverty Re-examined, S.P. Jenkins and J. Micklewright
(eds), Oxford University Press, Oxford.
De Neve, K.M. and Cooper, H. (1999), ‘The happy personality: A meta analysis of 137
personality traits of subjective wellbeing’, Psychological Bulletin, 125: pp197-
229.
Goldberg, D.P. (1972), The Detection of Psychiatric Illness by Questionnaire, Oxford
University Press: Oxford.
Goldberg, D.P. (1978), Manual of the General Health Questionnaire, NFER: Windsor.
Goldberg, D.P. and Williams, P. (1988). A User’s guide to the General Health
Questionnaire. NFER-Nelson: Windsor.
HM Treasury (2007), Financial capability: the Government’s long-term approach,
HMSO.
Hu, Y., Stewart-Brown, S., Twigg, L. and Weich, S. (2007). ‘Can the 12-item General
Health Questionnaire be used to measure positive mental health?’,
Psychological Medicine, 37: 1005-1013.
Huppert FA, Whittington JE (2003), ‘Evidence for the independence of positive and
negative wellbeing: implications for quality of life assessment’, British
Journal of Health Psychology 8:107-122.
McCabe, C.J., Thomas, K.J., Brazier, J.E. & Coleman, P. (1996), ‘Measuring the
mental health status of a population: A comparison of the GHQ-12 and the
SF-36’, British Journal of Psychiatry 169, 516-521.
Melhuish, E. and Malin, A. (2008), An investigation of the relationship between
financial capability and psychological wellbeing in mothers of young children,
FSA Occasional Paper 30.
National Institute of Adult Continuing Education (2007), Financial capability: the
Government’s long-term approach. A NIACE response to the HM Treasury paper,
http://www.niace.org.uk/organisation/advocacy/Treasury/financial-
capability.htm.
Nunnally, J.C. and Bernstein, I.H. (1994), Psychometric theory, 3rd Edn. New York:
McGraw-Hill.
Pevalin, D.J. (2000), ‘Multiple applications of the GHQ-12 in a general population
sample: an investigation of long-term retest effects’, Social Psychiatry and
Psychiatric Epidemiology 35, 508-512.
Skrondal, S. and Rabe-Hesketh, S. (2004), Generalized latent variable modelling:
multilevel, longitudinal and structural equation models, Chapman & Hall/CRC.
Taylor, M.P., Berthoud, R and Jenkins, S.P. (2004), Low income and multiple
disadvantage 1991-2001: Analysis of the British Household Panel Survey, A
report for the Social Exclusion Unit in the Breaking the Cycle Series, ODPM:
London.
Townsend, P (1979), Poverty in the United Kingdom, Penguin.

93
11. Appendix

Sample sizes by wave and gender: BHPS 1991–2006

Year Men Women Total
1991 3938 4599 8537
1992 3758 4450 8208
1993 3623 4228 7851
1994 3684 4326 8010
1995 3529 4153 7682
1996 3734 4373 8107
1997 3816 4390 8206
1998 3698 4355 8053
1999 3686 4324 8010
2000 3597 4308 7905
2001 3585 4183 7768
2002 3480 4063 7543
2003 3429 4063 7492
2004 3335 3911 7246
2005 3262 3889 7151
2006 3291 3880 7171
Total 57445 67495 124940
Notes: Unweighted sample sizes for adults with non-missing
information on relevant variables at each wave. Total row sums
all waves. BHPS 1991–2006.

94
 Full estimates from the GHQ model
Index of financial incapability N financial
Income adjusted Income unadjusted Problems
Index 1.074 [38.62] 1.092 [39.30] 0.637 [43.53]
Age 0.073 [1.73] 0.076 [1.81] 0.073 [1.74]
Age squared/100 –0.017 [1.45] –0.019 [1.63] –0.012 [1.04]
In good health –1.729 [46.53] –1.729 [46.52] –1.723 [46.44]
Expect improvement –0.201 [5.76] –0.206 [5.91] –0.232 [6.66]
Expect worsen 0.488 [10.71] 0.484 [10.62] 0.429 [9.42]
Real equiv. hh income –0.034 [3.16] 0.015 [1.24] 0.015 [1.43]
Amount saved pcm 0.036 [0.45] 0.043 [0.53] 0.288 [3.58]
Marital status
Married 0.520 [4.54] 0.532 [4.64] 0.507 [4.44]
Cohabiting 0.282 [2.63] 0.291 [2.71] 0.265 [2.48]
Widowed 1.775 [10.43] 1.807 [10.62] 1.767 [10.40]
Divorced/separated 0.920 [7.24] 0.930 [7.32] 0.897 [7.07]
Number of children
One child 0.035 [0.43] 0.019 [0.23] 0.004 [0.05]
Two children –0.012 [0.13] –0.046 [0.52] –0.055 [0.62]
Three children 0.007 [0.06] –0.041 [0.35] –0.039 [0.33]
Four or more children 0.056 [0.27] –0.004 [0.02] –0.020 [0.10]
Household type
Single elderly –0.536 [3.63] –0.525 [3.56] –0.516 [3.50]
Couple no children –0.438 [3.75] –0.400 [3.43] –0.379 [1.38]
Couple dep child –0.235 [2.00] –0.204 [1.73] –0.173 [1.47]
Couple non-dep child –0.041 [0.38] 0.003 [0.03] 0.014 [0.13]
Lone parent 0.202 [1.92] 0.203 [1.93] 0.223 [2.13]
2+ unrelated adults –0.115 [0.78] –0.100 [0.67] –0.084 [0.58]
Other household type –0.261 [1.55] –0.226 [1.35] –0.187 [1.12]
Highest qualification
Higher degree 0.394 [1.36] 0.418 [1.44] 0.398 [1.38]
First degree 0.436 [2.26] 0.450 [2.33] 0.423 [2.20]
Other higher qual. 0.005 [0.03] 0.004 [0.03] 0.003 [0.02]
A-Levels or equiv 0.018 [0.11] 0.006 [0.04] –0.010 [0.06]
GCSE or equivalent 0.112 [0.72] 0.116 [0.75] 0.136 [0.87]
Other qualification 0.235 [1.15] 0.234 [1.15] 0.222 [1.09]
Housing tenure
Own outright –0.126 [2.06] –0.139 [2.28] –0.153 [2.51]
Local authority tenant –0.128 [1.53] –0.153 [1.83] –0.145 [1.74]
Private tenant –0.088 [1.16] –0.112 [1.47] –0.116 [1.52]
House value/£100000 –0.034 [3.16] –0.035 [2.02] –0.039 [2.30]
Labour market status
Part-time employee –0.011 [0.18] –0.041 [0.67] –0.064 [1.05]
Self-employed –0.022 [0.27] –0.064 [0.78] –0.086 [1.05]
Unemployed 1.187 [13.98] 1.114 [13.11] 0.955 [11.21]
Retired –0.374 [4.61] –0.441 [5.42] –0.528 [6.50]
Inactive not like job 0.383 [6.26] 0.332 [5.41] 0.250 [4.08]
Inactive like job 0.783 [9.64] 0.723 [8.90] 0.615 [7.57]
Seasonal/casual job –0.206 [2.54] –0.209 [2.59] –0.217 [2.69]
Fixed term contract –0.429 [4.41] –0.425 [4.37] –0.426 [4.39]
Constant 9.243 [6.12] 9.062 [6.00] 8.434 [5.60]
Notes: Estimates from within-group fixed effects regressions with GHQ score as the dependent variable.
Models also include region and time dummies to capture macro-economic effects. Absolute ratio of
coefficient to standard error in brackets.

95
 Full estimates from the life satisfaction model
Index of financial incapability N financial
Income adjusted Income unadjusted Problems
Index –0.226 [26.91] –0.231 [27.43] –0.122 [28.14]
Age –0.005 [0.41] –0.006 [0.48] –0.006 [0.47]
Age squared/100 –0.013 [3.34] –0.012 [3.25] –0.013 [3.43]
In good health 0.281 [27.28] 0.281 [27.28] 0.281 [27.26]
Expect improvement 0.025 [2.56] 0.026 [2.71] 0.029 [3.05]
Expect worsen –0.039 [2.94] –0.038 [2.86] –0.032 [2.37]
Real equiv. hh income 0.012 [4.18] 0.003 [0.97] 0.003 [1.04]
Amount saved pcm 0.025 [1.18] 0.023 [1.09] –0.015 [0.68]
Marital status
Married –0.006 [0.19] –0.008 [0.24] –0.006 [0.17]
Cohabiting 0.019 [0.62] 0.018 [0.58] 0.020 [0.65]
Widowed –0.230 [4.60] –0.239 [4.77] –0.233 [4.65]
Divorced/separated –0.177 [4.82] –0.179 [4.87] –0.175 [4.75]
Number of children
One child 0.003 [0.12] 0.006 [0.27] 0.007 [0.30]
Two children –0.019 [0.71] –0.011 [0.40] –0.011 [0.44]
Three children –0.016 [0.45] –0.005 [0.14] –0.006 [0.16]
Four or more children –0.113 [1.80] –0.100 [1.60] –0.100 [1.60]
Household type
Single elderly 0.036 [0.82] 0.035 [0.78] 0.035 [0.80]
Couple no children 0.090 [2.68] 0.080 [2.39] 0.079 [2.36]
Couple dep child –0.003 [0.07] –0.010 [0.30] –0.012 [0.36]
Couple non-dep child –0.041 [1.30] –0.051 [1.64] –0.051 [1.64]
Lone parent –0.135 [4.44] –0.135 [4.46] –0.137 [4.52]
2+ unrelated adults –0.058 [1.34] –0.061 [1.43] –0.065 [1.53]
Other household type –0.003 [0.07] –0.012 [0.26] –0.016 [0.33]
Highest qualification
Higher degree –0.022 [0.24] –0.030 [0.33] –0.028 [0.32]
First degree –0.014 [0.24] –0.017 [0.30] –0.014 [0.24]
Other higher qual. 0.089 [1.97] 0.090 [1.97] 0.088 [1.93]
A-Levels or equiv 0.090 [1.85] 0.093 [1.92] 0.094 [1.93]
GCSE or equivalent 0.039 [0.82] 0.038 [0.79] 0.033 [0.69]
Other qualification 0.016 [0.26] 0.017 [0.27] 0.016 [0.26]
Housing tenure
Own outright 0.049 [2.82] 0.051 [2.97] 0.054 [3.15]
Local authority tenant 0.035 [1.38] 0.040 [1.60] 0.036 [1.42]
Private tenant 0.013 [0.57] 0.018 [0.83] 0.016 [0.70]
House value/£100000 0.004 [1.02] 0.004 [1.05] 0.005 [1.18]
Labour market status
Part-time employee 0.035 [2.01] 0.041 [2.41] 0.043 [2.53]
Self-employed 0.034 [1.43] 0.044 [1.88] 0.046 [1.98]
Unemployed –0.231 [8.88] –0.214 [8.21] –0.191 [7.32]
Retired 0.157 [6.65] 0.173 [7.30] 0.185 [7.82]
Inactive not like job 0.008 [0.46] 0.020 [1.12] 0.031 [1.78]
Inactive like job –0.104 [4.27] –0.090 [3.71] –0.075 [3.07]
Seasonal/casual job –0.015 [0.64] –0.014 [0.60] –0.013 [0.55]
Fixed term contract 0.027 [0.97] 0.027 [0.96] 0.028 [0.99]
Constant 5.376 [9.26] 5.420 [9.34] 5.569 [9.60]
Notes: Estimates from within-group fixed effects regressions with life satisfaction score as the dependent
variable. All models also include region and time dummies to capture macro-economic effects. Absolute
ratio of coefficient to standard error in brackets.

96
 Full estimates from the anxiety or depression model
Index of financial incapability N financial
Income adjusted Income unadjusted Problems
Index 0.196 [6.67] 0.200 [6.81] 0.123 [7.54]
Age 0.235 [4.47] 0.235 [4.47] 0.235 [4.47]
Age squared/100 –0.115 [7.76] –0.115 [7.77] –0.114 [7.68]
In good health –1.220 [27.73] –1.220 [27.73] –1.219 [27.69]
Expect improvement 0.039 [0.82] 0.038 [0.81] 0.030 [0.63]
Expect worsen 0.076 [1.37] 0.075 [1.36] 0.056 [1.01]
Real equiv. hh income 0.004 [0.26] 0.016 [1.08] 0.017 [1.14]
Amount saved pcm 0.069 [0.53] 0.072 [0.56] 0.129 [1.07]
Marital status
Married 0.152 [0.94] 0.150 [0.93] 0.147 [0.91]
Cohabiting 0.425 [2.82] 0.424 [2.82] 0.419 [2.79]
Widowed 0.414 [2.16] 0.419 [2.19] 0.415 [2.17]
Divorced/separated 0.299 [2.03] 0.299 [2.03] 0.290 [1.97]
Number of children
One child 0.156 [1.49] 0.155 [1.48] 0.153 [1.46]
Two children 0.044 [0.38] 0.041 [0.35] 0.041 [0.35]
Three children 0.165 [1.07] 0.159 [1.03] 0.162 [1.05]
Four or more children 0.087 [0.35] 0.082 [0.33] 0.083 [0.33]
Household type
Single elderly –0.210 [1.35] –0.209 [1.34] –0.213 [1.37]
Couple no children –0.490 [3.30] –0.482 [3.24] –0.482 [3.25]
Couple dep child –0.590 [3.73] –0.582 [3.68] –0.583 [3.69]
Couple non-dep child –0.412 [2.85] –0.401 [2.78] –0.407 [2.82]
Lone parent –0.158 [1.27] –0.157 [1.26] –0.160 [1.29]
2+ unrelated adults –0.063 [0.30] –0.057 [0.27] –0.056 [0.27]
Other household type –0.490 [2.30] –0.484 [2.27] –0.481 [2.25]
Highest qualification
Higher degree –0.286 [0.66] –0.282 [0.65] –0.291 [0.67]
First degree –0.072 [0.26] –0.069 [0.25] –0.069 [0.25]
Other higher qual. –0.112 [0.56] –0.113 [0.57] –0.109 [0.55]
A-Levels or equiv 0.015 [0.06] –0.012 [0.06] 0.015 [0.07]
GCSE or equivalent –0.124 [0.57] –0.124 [0.57] –0.119 [0.55]
Other qualification 0.160 [0.59] 0.160 [0.59] 0.154 [0.57]
Housing tenure
Own outright 0.001 [0.01] –0.001 [0.01] –0.005 [0.06]
Local authority tenant 0.144 [1.41] 0.141 [1.38] 0.140 [1.37]
Private tenant –0.025 [0.25] –0.028 [0.28] –0.030 [0.30]
House value/£100000 –0.022 [0.70] –0.022 [0.71] –0.023 [0.75]
Labour market status
Part-time employee 0.205 [2.61] 0.201 [2.55] 0.196 [2.48]
Self-employed –0.013 [0.11] –0.018 [0.15] –0.024 [0.20]
Unemployed 0.355 [3.37] 0.344 [3.27] 0.309 [2.93]
Retired 0.205 [2.03] 0.196 [1.94] 0.177 [1.75]
Inactive not like job 0.466 [6.09] 0.459 [6.00] 0.439 [5.74]
Inactive like job 0.586 [6.50] 0.578 [6.41] 0.555 [6.14]
Seasonal/casual job –0.030 [0.27] –0.032 [0.28] –0.037 [0.32]
Fixed term contract –0.231 [1.63] –0.231 [1.63] –0.229 [1.61]
Notes: Estimates from fixed effects (conditional) logit regressions with whether suffers a health problem
related to anxiety or depression as the dependent variable. All models also include region and time
dummies to capture macro-economic effects. Absolute ratio of coefficient to standard error in brackets.

97
The Financial Services Authority
25 The North Colonnade Canary Wharf London E14 5HS
Telephone: +44 (0)20 7066 1000 Fax: +44 (0)20 7066 1099
Website: http://www.fsa.gov.uk
Registered as a Limited Company in England and Wales No. 1920623.
Registered Office as above.