Paul Lambe
School of Education and Lifelong Learning
St Luke’s Campus
Heavitree Road
University of Exeter
Exeter
Conceptualising and measuring agency using the British Household Panel Survey
Paper presented at the British Educational Research Association Annual Conference, University of Warwick, 6-9 September 2006
This working paper was produced as part of the Learning Lives Project*. Copyright lies with the author. If you cite or quote, please be sensitive to the fact that this is work in progress. Special thanks go to Ian Alcock (ESRC funded PhD student attached to the Learning Lives Project).
*The Learning Lives: Learning, Identity and Agency project (see learninglives.org) is a collaboration between the Universities of Exeter, Brighton, Leeds and Stirling and is funded by the Economic and Social Research Council as part of their Teaching and Learning Research Programme (TLRP) see
Conceptualising and measuring agency using the British Household Panel Survey
Introduction
The concept of agency is of central concern to the Learning Lives project. The project aims to understand the relationship between adult learning and agency over the life course in terms of how learning impacts on agency and, conversely, how agency impacts on learning. The project approaches learning as one of the ways in which people respond to events in their lives, often to gain more control over their lives. To further these aims it is necessary to critically evaluate whether empirical evidence supports or disputes theoretical connections between agency and adult learning. However, as outlined below, agency is a rather broad and elusive latent concept. This paper takes a preliminary step in this direction as its purpose is to characterise one major aspect of the self’s agency, self-efficacy, using items (questions) from the British Household Panel Study[1] (BHPS). Furthermore, this paper illustrates the application of Latent Class Analysis, a powerful new tool in the analysis of typologies and scale response patterns from categorically scored survey data using the LEM [2]software programme.
Theorisation of Agency
In recent years much valuable theoretical writing has appeared that has increased our understanding of agency (Yoder 2000, see Cote and Levine 2002 for a review of the psychological and sociological literature, see Emirbayer and Mische 1998 for a review of literature theorising agency). Nevertheless, the concept of agency, as the following brief review of literature points up, has maintained an ‘elusive… vagueness’ (Emirbayer and Mische 1998:962). Karen Evans, has identified 12 factors of importance in the analysis of ‘bounded’ agency and control :- sociability, confidence, fulfilled work life and fulfilled personal life, belief that opportunities are open to all, belief that own weaknesses matter, belief in planning not chance, belief that ability not rewarded, active career seeking, unlikely to move, politically active, helping/people career oriented, and negative view of future (Evans 2002:255). John Bynner sees personal agency as comprised of an individual’s disposition (sense of self-efficacy, sense of internal locus of control, motivation, aspiration and self-esteem) and resources (social, cultural, human capital), including community resources and an individual’s membership of, and activity in community/civic organisations which may enhance an individual’s social and human capital and improve that individual’s prospects and indeed, sense of agency (Bynner 2001:23). Yaojun Li, Andrew Pickles and Mike Savage, using BHPS data, have taken a more nuanced view and have broken down the concept of social capital into three distinct dimensions; ‘civic participation, social networks and neighbourhood attachments’, and argue that ‘civic participation is just one out of a range of channels of social capital generation’ and that more ‘informal networks affect people’s attitudes, values, preferences and key aspects of their life-chances’ (Li et al 2003:16). Emirbayer and Mishe comment that the concept of agency has been associated with, inter alia, motivation, will, sense of purpose, intentionality, choice, initiative, and goal seeking (Emirbayer and Mische 1998). Whilst Cote and Levine’s construct, identity capital, cites agency residing in and emanating from tangible personal assets such as memberships of organisations, and intangible personal assets such as internal locus of control, self-esteem, self-efficacy, a sense of purpose in life, the ability to self-actualise and ideological commitment which ‘combine to predict identity capital in terms of formulating a stronger sense of adult identity’ (Cote and Levine 2002:143-149).
Gert Biesta and Mike Tedder have conceptualised agency not as an ‘individual possession but something that is achieved in action’ which they argue calls for ‘an ecological understanding of agency’ (Biesta and Tedder 2006). Like Emirbayer and Mishe, they contend that individuals, influenced by past achievements, understandings and patterns of action, who are motivated to realise a future that is different from their past, can only act out this intention in the present, or plan for the future. Thus in a context in which each individual’s capacity to make ‘practical and normative judgements among alternative possible trajectories of action [is conditioned by the] emerging demands, dilemmas and ambiguities of presently evolving situations’(Emirbayer and Mishe 1998, quoted by Biesta and Tedder 2006). Their emphasis on changing contexts-for-action over time raises fundamental questions about the relationship between agency and learning in the life course. Not least : why individual’s can be agentic in one situation and not another, why an individual’s capacity for agentic behaviour can change over the life course, and what are the catalysts that initiate the learning processes which enable people to reconstruct their agentic orientations. At the same time Biesta and Tedder emphasise that ‘agentic orientations … are never sufficient to understand actual agency … because [direction of one’s life] … also depends on contextual and “ecological” factors and on available resources within a particular ecology’( ibid: p9).
One of the principal tenets of life course theory is the significance of human agency in life course construction and at the core of human agency is the self, or self-agency
(Elder, 1974, 1995, Gecas 2004:369, Badura 1997:3). Self-agency underpins Bandura’s theory of self-efficacy, indeed ‘efficacy belief [is regarded as]… a major basis of action’ and Bandura defines perceived self-efficacy as ‘not a measure of the skills one has but a belief about what one can do under different sets of conditions with whatever skills one possesses’ (Bandura 1997:37). Gecas defines self-efficacy as referring to the ‘perception of oneself as a causal agent in one’s environment, as having some control over one’s circumstances , and being capable of carrying out actions to produce intended effects’ (Gecas 2004, 370). There is much evidence of the relationship between an individual’s self-efficacy and their functioning and well-being in the domains of academic achievement, occupational achievement, and general physical and mental well-being (see Bandura 1997, Swartzer and Fuchs 1996, ). Indeed, Gecas argues that those with ‘high self-efficacy, especially in such domains as education, inter-personal relations, and occupational contexts, are more likely to be the architects of their own lives and to see themselves as such’, and that the converse is more likely for those with low self-efficacy (Gecas 2004:370). Furthermore, in relation to adult learning recent studies suggest that participation in adult learning has a positive effect upon self-efficacy (Schuller et al 2002, Hammond and Feinstein 2005).Clearly, self-efficacy, an individual’s belief in his or her own efficacy and personal control, is an important facet of the latent concept agency and major influence upon life course construction.
This brief review of the theorisation of agency points up that it is crucial to locate learning and agency in a temporal framework in order to examine their dynamic qualities and the changing influences upon them. Furthermore, that the characterisation of a measure of self-efficacy in isolation from contextual effects emanating from various sources including those of social capital generation will allow us to control for and test hypotheses concerning the effects upon this key element of agency of changing contextual influences, thereby moving us towards an ecological understanding of agency. This necessitates longitudinal analysis of panel data such as the BHPS. However our immediate concern is an analysis using items from a single wave of the BHPS to characterise the latent concept self-efficacy and examine its relationship with adult learning.
Methodology
Much of previous research into agency, has employed factor analysis of survey items to reduce many observed variables to only a few latent factors, and respondents’ predicted factor scores then used in post hoc analysis of the relationship between agency and learning ( Evans 2002, Cote 1997, Li et al 2003, Hammond and Feinstein 2005). It is the case that the vast majority of variables in the most widely used social science data sets, including the BHPS are scored as categorical data, either nominal or ordinal. Clearly, when possible, it is preferable to select analytic techniques that conform to the requirements of both the theoretical concept of interest and the nature of the measures of the data available. For these reasons Latent Class Analysis (LCA) is used here. LCA is a statistical method for studying categorically scored variables that does not require data to meet the assumptions of multivariate normality and continuity of measurement, and is eminently suited for life course analysis using categorically scored survey data. It provides a powerful new tool in the analysis of typologies (classes), either as a method for empirically characterising a set of latent types within a set of observed indicators or as a method for testing whether a theoretically posited typology adequately represents the data. LCA can be considered as a ‘qualitative data analogue to factor analysis. It enables researchers to empirically identify discrete latent variables [ e.g agency, religious commitment] from two or more discrete [categorical] observed variables’ ( McCutcheon 1987:7, McCutcheon and Mills 1998, Vermunt, 1997a). Indeed recent developments allow analysts to include variables of mixed type (nominal, ordinal, continuous/and or count variables) in the same analysis. LCA is a flexible methodology that enables researchers: to identify a set of mutually exclusive latent classes, or typologies, from observed categorically scored survey data, and thus empirically characterise typological classifications (multidimensional). It also enables researchers to analyse the scalability of a set of observed categorical items into ordinal measures of a latent variable, and additionally allows us to discern whether the defined latent variable
( e.g. self-efficacy) is invariant over multiple populations (i.e. comparative analysis of scale characteristics and typologies in different populations, social groups. Furthermore, models for the analysis of categorically scored panel data allow us to examine and test formal hypotheses about the nature of change in scale characteristics and typologies identified using identical measures at different times for the same population (passim McCut cheon 1987, see also McCutcheon 1996, Macmillan and Eliason 2004, Vermunt 1997b). Hence, the appropriateness of LCA for examining the relationship between adult learning and self-efficacy over the life course using panel data such as the BHPS.
At BHPS wave 11 (2001) a block of items, CASP-19 (a 19 item Likert scaled index measuring quality of life), was included as a core rotating component ( CASP-19 will be included again at wave 16). CASP-19 was originally developed as a measure of quality of life in early older age and specifically designed to exclude ‘contextual and individual phenomena that might influence it, such as health, social networks, and material circumstances’ and specifically designed to tap into the agency of older people (Hyde et al 2003:187). The CASP19 scale has been used in the English Longitudinal Study of Ageing, the UK Whitehall Study, the US Health and Retirement Study and a shorter version in the Study of Health, Ageing and Retirement in Europe. A raft of studies have validated this measure[3], and its development is outlined in Hyde et al 2003[4]. However the BHPS is the first survey to use the scale on its whole sample of respondents at all ages because of its proven reliability and validity. The items in the CASP-19 scale (see Table 1 below) tap four domains: control, autonomy, self realisation and pleasure. Clearly, as intended by its authors, in the three domains of control, autonomy and self-realisation the items tap in to sense of agency and self-efficacy as theorised by Bandura , and empirically evidenced by Evans, Bynner, Cote and Levine, Emirbayer and Mische. The inclusion of CASP-19 in the BHPS as a core rotating component will no doubt prove to be an extremely useful resource in longitudinal educational research as an instrument to measure sense of agency and its relationship with learning over time, albeit an instrument whose data collection points are at five yearly intervals.
However, what is required for our purposes is an equivalent instrument whose data are recorded in the smallest possible units of time, and in the case of the BHPS this would be at each annual wave. Fortunately, at all waves of the BHPS since 1991 onwards a General Health Questionnaire has been included, and along with other BHPS core questions we are able to construct an instrument whose items approximate those used in the domains of control, autonomy and self-realisation of CASP-19. Table 2 below outlines these core BHPS items and their counterparts in CASP-19.
Table 1:Item wording and domains for CASP-19Each item has the response categories coded as follows, Often =3, Not Often=2, Sometimes =1, and Never=0.
CONTROL / My age prevents me from doing the things I would like toI feel that what happens to me is out of my control
I feel free to plan for the future
I feel left out of things
AUTONOMY / I can do the things that I want to do
Family responsibilities prevent me from doing what I want to do
I feel that I can please myself what I can do
My health stops me from doing the things I want to do
Shortage of money stops me from doing the things I want to do
PLEASURE / I look forward to each day
I feel that my life has meaning
I enjoy the things that I do
I enjoy being in the company of others
On balance I look back on my life with a sense of happiness
SELF-REALISATION / I feel full of energy these days
I choose to do things I have never done before
I feel satisfied with the way my life has turned out
I feel that life is full of opportunities
I feel that the future looks good for me
Table 2: BHPS proxy items and their CASP-19 counterparts ( excludes item 1 My age prevents me from doing the things I would like to). The emboldened items in the CASP-19 column form the eight item version. The BHPS items which appear at all waves, KGHQC-L, KHLLT( *** this item does not appear at waves I and n of the BHPS, however values at preceding and consecutive waves allow imputation of missing values), KFISITX, reverse coded where necessary and dichotomised.
BHPS items all waves / Domain / CASP-19 itemsHave you recently been able to face up to problems ?
(Item KGHQF) / CONTROL / I feel that what happens to me is out of my control *
I feel free to plan for the future
Have you recently felt you are playing a useful part in things?
( Item KGHQC) / I feel left out of things*
( Item KQLFD)
Have you recently been losing confidence?
(Item KGHQJ) / AUTONOMY / I can do the things that I want to do *
************************* / Family responsibilities prevent me from doing what I want to do
Have you recently felt capable of making decisions about things?
(Item KGHQD) / I feel that I can please myself what I can do*
Does your health in any way limit your daily activities compared to most people of your age?
( Item KHLLT)*** / My health stops me from doing the things I want to do*
How well would you say you were managing financially?
(Item KFISIT) / Shortage of money stops me from doing the things I want to do
Have you recently been able to enjoy your normal day to day activities
(Item KGHQG) / I look forward to each day
Have you recently felt constantly under strain?
(Item KGHQE) / SELF-REALISATION / I feel full of energy these days
*********************** / I choose to do things I have never done before
Have you recently been feeling reasonably happy all things considered?
(Item KGHQL) / I feel satisfied with the way my life has turned out*
Have you recently been thinking of yourself as a worthless person?
(Item KGHQK) / I feel that life is full of opportunities*
Looking ahead how do you think you will be financially a
year from now?
(Item KFISITX) / I feel that the future looks good for me*
Analysis: Part 1
Our sample comprised 5,718 adults of working age (BHPS wave 11, England only) females aged 16-59 years and males aged 16-64. Firstly, all CASP-19 items in the domains of CONTROL, AUTONOMY and SELF REALISATION were subject to LCA and a more parsimonious seven-item shorter version derived (see Table 2, the emboldened items in the CASP-19 column, recoded 1 for a non -agentic response and 2 for an agentic response) . Latent class analysis of the seven-item CASP scale produced a two class model which assigned cases into two distinct typologies. Class 1 in which there was a very high probability of an agentic response to each of the seven items and class two in which respondents had very low probability of an agentic response to each item. Table 3 below outlines the conditional probabilities of an agentic response to the items in the scale and the Latent Class probabilities, that is, the probability of a respondent belonging in either the agentic or non-agentic group given their responses to the shortened seven-item CASP instrument. Eighty percent of respondents were classified as agentic. Thus 8 in 10 of our sample are typified as having a positive perception of their self-efficacy, and two in ten of the sample as having a much poorer perception of their self-efficacy. Clearly, as the CASP-19 was designed specifically to tap into sense of agency and is a much validated instrument in its entirety and in its shortened versions, these findings come as no surprise Nevertheless, the analysis confirms that our seven- item version is a valid measure of self-efficacy.