What is poverty? A study of the factors affecting the judgement of poverty by future welfare state professionals

Page created by Carrie Bowers
 
CONTINUE READING
Draft – do not quote

What is poverty? A study of the factors affecting the judgement
of poverty by future welfare state professionals

– Paper on methods
                                                                            Merete Monrad & Morten Ejrnæs

Introduction
In Denmark, an important discussion regards what dimensions should be included in a concept of poverty.
In this context it is interesting to analyze, what welfare state professionals emphasize when judging
whether a person or family is poor or not. This question is crucial in terms of how professionals respond to
poverty and the study thus seeks to inform public debate, social policy, social work practice and the
education of welfare state professionals. The objective of the study is to examine what factors impact the
perception and judgement of poverty. In the paper, we use factorial survey (horizontal vignette
methodology) to examine, what factors students in a range of different welfare state professions take into
account when judging a person or family to be poor or not. We analyze what factors students take into
account when making their judgements and discuss the significance of the social position and professional
affiliation of the students for their judgements. The study includes students from the fields of nutrition and
health, nursery teaching and social work. Students rather than professionals in practice are included in the
study since their judgements are important to examine in order to evaluate and improve teaching on
perceiving, understanding and addressing poverty. Furthermore, students are the welfare state
professionals of tomorrow who during their future practice may encounter and have to respond to the
social problem of poverty. The study uses a factorial survey that is a form of survey experiment. This
methodology offers some unique possibilities for examining the factors impacting on people's judgements.
The use of a factorial survey entails a systematic variation of the content of a vignette describing a person
that may or may not be perceived as poor. Since the description of the person is varied with regards to
income, deprivation, duration of deprivation, gender, age, position and ethnicity it is possible to analyze the
extent to which each of these characteristics are significant for social work students’ judgements of
poverty. The article thus examines the associations between the social problem of poverty and the
judgements regarding this social problem made by future welfare state professionals.

The questions we seek to address in this study are the following: 1) What characteristics of a person and his
or her social situation enter into the judgement of him or her as poor or not poor? 2) What weights are
given to different characteristics in making such judgements? 3) How much consensus is there about the
characteristics entering into the judgements and the weights given to each characteristic?

This paper will describe and discuss the methodological design of the study. At the conference, preliminary
results will be presented.

Social definitions of poverty
In a now classical piece, Fuller and Myers (1941) emphasized that social problems contain both objective
conditions and subjective interpretations. In this study, we focus on the subjective side of the social
problem of poverty by studying social definitions of poverty. Related to an overall interest in how future
welfare state professionals respond to the social problem of poverty, the ways in which these future
professionals interpret poverty is important, since such interpretations may affect the way the social
                                                       1
Draft – do not quote

problems faced by people living in poverty are interpreted (e.g. what causes are ascribed to these social
problems) and the actions taken to address poverty and other social problems faced by people living in
poverty. We approach the social definitions of poverty by examining patterns in judgements of when a
person is poor thereby identifying what factors in a person’s life situation are significant for the person to
be defined as poor. Thus, we seek to clarify what factors future professionals regard as crucial for a person
to be poor.

In recent years, several studies have addressed the objective conditions of poverty in Denmark (e.g. Ejrnæs,
Hansen, Hansen, Hussain and Larsen 2011; Hansen and Hussain 2009), and this study is an attempt to
supplement this research by examining the social definitions of poverty among future welfare state
professionals who are likely to encounter and have to respond to poverty in their professional work.

The study examines attitudes or judgements of poverty, not actions taken to address poverty. The
difference between attitudes and actions is important to keep in mind, at the same time as attitudes and
actions are often found to be moderately correlated (Hill 1981). However, in this study we do not aim at
studying what should be done to address poverty, rather we study what is seen as poverty. The perception
or implicit definition of poverty held by professionals must be considered as a foundation for actions
towards poverty. If something is not perceived as poverty, it is unlikely to be addressed as such, and
therefore the perception of poverty is important for social work addressing poverty. Further, we approach
the attitudes of professionals as a part of their professionalism because components of attitudes such as
knowledge and experience, emotions and values are important parts of professionalism (Ejrnæs & Monrad
2010). In this study we do not examine the different components of attitudes, rather we study professional
assessments where future professionals apply their knowledge and experience, emotions and values to
concrete situations described in vignettes in order to determine whether the person described is poor or
not.

The study is inspired by Scandinavian welfare theory, where social problems are conceptualized as lack of
satisfaction of a range of needs (Hillgaard and Keiser 1979: 26). Thus, social problems are a state of lack or
absence of welfare. The Finnish sociologist Allardt (1975) has presented a concept of welfare that does not
only encompass material welfare (to have), but also social welfare (to love) and psychological welfare (to
be). Inspired by this theoretical framework, we are interested in examining what deprivations are regarded
crucial for a person to be poor.

Methods
Methodologically, social definitions of poverty could be studied in a range of ways: participant observation
of professional discussions among future welfare state professionals and focus group interviews about
poverty would for instance reveal negotiations of social definitions. Instead of focusing on the negotiation
of poverty in professional groups, we chose to study the factors that enter into judgements of poverty. That
is, when a future welfare state professional is confronted with a person that may or may not be poor, what
in this person’s life situation does he or she emphasize when making the judgement of whether the person
is poor or not? That is, what patterns exist in social judgements of poverty? This type of study may reveal
both factors that are generally emphasized by the future professionals and factors that may theoretically
be thought relevant to definitions of poverty, but are generally overlooked or thought unimportant by the
future professionals. Thereby, the study provides important insights that may be used to qualify discussions
among professionals and qualify educators of welfare state professionals to teach students to better
respond to the social problem of poverty. The method used to examine what factors enter into and affect
the judgement of poverty is a quantitative factorial survey design.
                                                       2
Draft – do not quote

The factorial survey
The factorial survey is a form of survey experiment using vignettes. In the factorial survey, the content of
short vignettes i.e. descriptions of persons that may be judged as poor, is systematically varied in order to
determine the impact of this variation on the judgements made (the responses to the vignettes) (for a
review on factorial surveys, see Wallander 2009). For instance, we may construct descriptions of two
persons that are similar, but for one person it is described that she due to financial reasons has refrained
from inviting guests home, for the other this information is not provided. Comparing these two vignettes
makes it possible to examine how the information regarding this particular deprivation affects the
judgement of whether the person is poor or not. In the actual study, a range of factors are varied at the
same time at random, and the effect of each variation can be examined through quantitative analysis.

Factorial surveys are used to examine the structure of judgements by examining the factors impacting
judgements or attitudes (Wallander 2009). The idea behind a factorial survey is to construct vignettes
where the factors that are thought to affect the judgements made by respondents are varied across the
vignettes and then in the following analysis to disentangle these factors and determine what factors are
significant for the judgements made. Hence, the factorial survey is an indirect method of studying the
underlying criteria shaping judgements and attitudes. The advantage of studying attitudes indirectly is that
respondents may not be aware of what factors impact their judgements, but they are nonetheless able to
make judgements in concrete situations depicted in vignettes (Alexander & Becker 1978). Core aspects of
the factorial survey design of this study are described in the following.

The vignette design
The main part of the factorial survey design is deciding what factors to include in the vignettes. The factors
have to be relevant and realistic in terms of the judgements made otherwise the judgements of
respondents will become too hypothetical to have any relationship to the judgements they make in their
everyday life (and thus loose external validity). We have selected factors based on previous research on
poverty, particularly in terms of income, deprivations and length of deprivations and have furthermore
included some social background variables on the persons described in the vignettes (gender, ethnicity, age
and position). The vignettes are then constructed by combining these factors into descriptions of persons
that may or may not be judged as poor.

In order to be able to separate the effect of each of the factors in the vignettes, the factors should ideally
be varied independently of each other across vignettes. In this study, this is ensured by randomly selecting
the outcome of each factor to be included in each vignette (this is done by computer software). Thus, each
vignette is a combination of a random selection of outcomes on each factor.

The ways in which people in a strained financial situation differ from one another are infinite and thus the
amount of factors that could be included in a judgement of a person as poor or not is enormous. Since it is
impossible to include all the factors that may be important to some respondents when making their
judgements, we have been guided by theories of poverty, welfare and social stratification as well as
previous empirical research on poverty in Denmark when selecting the factors to include. Furthermore, the
random factorial survey design somewhat delimits the kinds of factors that may be included in the study,
since these have to remain realistic when combined at random or almost at random with each other. This
concern made us omit accounts of the vignette person’s living expenses and housing conditions since these
could not realistically be combined randomly with income level (in particular the total cost of living could
not be much larger than the income level for the vignettes to seem realistic). Non-random combinations
could have been defined, but we judged the constraints on the associations between these variables that
                                                      3
Draft – do not quote

would have to be introduced in the study to preserve the realism to be too demanding to work well.
Exceptions from a random design, however, were made, since random combinations of position and age
would result in unrealistic vignettes. The associations between these variables were therefore constrained
so that a person described as a student in the vignettes cannot be 54 or 70 years old.

The vignettes have been developed in two versions: one for a single adult and one for a single adult with a
child. This was necessary in order to include deprivations related to children, since these would be
impractical to include in one random vignette design. For the sake of simplicity, the adult described in the
vignettes is always single and the child is always 12 years old. The age of 12 years was chosen for the child
to be young enough to be dependent on the parent, but to avoid the complexity of defining expenses
related to child care that would be relevant if the child had been younger.

We assumed that it would be possible for each respondent to judge 20 vignettes in a relatively short
amount of time and still remain concentrated about the judgements (an assumption that was confirmed in
a pilot study). To get each person to judge 20 vignettes means that the variation of factors will be evident
to the respondents. However, we deemed this to be relatively unproblematic since we are interested in
examining their professional judgements, not in revealing the hidden biases of respondents (which would
require the manipulation of factors to remain obscure). Further, the variation of a range of factors at the
same time makes the variation of each one less evident. In addition, having each respondent answer 20
vignettes (10 single-adult vignettes and 10 single-parent-with-one-child vignettes) holds the advantage of
making it possible to study more combinations of factors and still be able to analytically tease them apart.

Factors – independent variables contained in the vignettes
The factors in the vignettes fall in three categories: income, deprivations and social background
information. The factors varied in the vignettes function as independent variables along with the
respondent characteristics. The different types of factors included are described in the following.

Income:
When studying what factors affect the judgement of a person as poor or not, the income level of the
person is a central variable. In defining this variable, we wanted to include both income levels above and
below a poverty line. We used a relative poverty definition as 50 % of the median income as our poverty
line and defined three income levels below this line and three levels symmetrically above, resulting in seven
income levels. We made the difference between the income levels of equal size in percent (15 % of the
income at the poverty line).

For the single-adult vignettes, we thus used 50 % of the median income (in Denmark in 2012: 8.788
kr/month after taxes) and defined levels +/- 15 %, +/- 30 % and +/- 45 %, resulting in an income range of
4.833 – 12.743 kr/month after taxes. For the single-parent-with-one-child vignettes, we had to take
account of the increased expenses associated with having a child and simultaneously the economies of
scale within a household. To do this, we used the household equivalent disposable income, which is the
income corrected by economies of scale in families with several family members (Arbejderbevægelsens
Erhvervsråd 2012: 3). Taking a point of departure in a poverty line of 50 % of the median income, the
poverty line of a two-person household in Denmark in 2012 can be set at 13.320 kr/month after taxes
(Arbejderbevægelsens Erhvervsråd 2012: 3). In a similar fashion as with the single adult vignette, we
defined levels of +/- 15 %, +/- 30 % and +/- 45 %, resulting in an income range of 7.326 – 19.314 kr/month
after taxes.

Deprivations:
                                                      4
Draft – do not quote

In deciding what deprivations to include in the vignettes, we took our starting point in extant research on
poverty in Denmark. Ejrnæs, Hansen, Hansen, Hussain and Larsen (2011: 34) describe five types of
deprivations that are all included in this study: daily necessities, health, leisure, social relations and material
comforts. Ejrnæs et al. (2011) examine a range of deprivations of each of these types, we have only
included the ones most widespread among the recipients of the lowest levels of social benefits in Denmark
(that were studied by Ejrnæs et al. 2011). In the vignettes with a single parent with one child we have
furthermore included deprivations specifically related to the child. These deprivations are taken from
Hansen and Hussain (2009) in an attempt to cover different dimension of deprivations among children. The
age of the child in the vignette is set to 12 years (for the sake of simplicity it does not vary across vignettes),
and the deprivations have been chosen to be relevant to a child of this age.

The deprivations are shown in Table 1 below. The deprivations were introduced with the text “… has due to
financial reasons refrained from…”. In the vignettes 0-4 general deprivations are included (in the pilot we
found that this was more than enough for respondents to take in). In the vignettes with a parent and a
child, 0-4 deprivations related to the child were also included (making these somewhat more complex than
the single adult versions). Based on a pilot study, we chose to let the number of general deprivations in
each vignette be larger than the number of specific deprivations for the child, since respondents in the pilot
study reacted strongly to vignettes where the child suffered the most deprivations and saw the parents in
these vignettes as unfair or immoral. Since we are interested in studying the judgement of when a person
or family is poor and not the judgement of the morality related to the management of poverty (though
these may not be entirely separated), we decided to let the children in the vignettes suffer fewer or the
same number of deprivations as the family in general.

Table 1: Deprivations varied in vignettes

Type of deprivation            Specific deprivations
Daily necessities              Eat fresh fruit and vegetables on a daily basis
                               Buy clothes, shoes and outdoor things
Health                         Visit the dentist
Leisure                        Go on vacation outside the home
                               Do leisure-time activities
Social relations               Visit friends/family (that live more than 20 km from the home)
                               Invite guests home
Material comforts              Make repairs in the home, or replace equipment in the home that has been
                               broken or spent
Child – daily necessities      Buy clothes and shoes for the child
Child – leisure                Let the child do sports or other leisure time activities
Child – social relations       Celebrate the child’s birthday
Child – social relations and   Let the child participate in school trips, participate in club activities etc.
educational activities
Child – material comforts      Buy the child a mobile phone of his or her own

Each deprivation is binary, that is it is either present or not, it is not scaled. The duration of the deprivation
is included as a separate variable with durations of 1, 2 or 3 years (covering all the deprivations mentioned
in the given vignette). This piece of information is of course omitted when the number of deprivations is set
to 0.

Social background information:

                                                        5
Draft – do not quote

In the vignettes, we have varied gender, ethnicity, age and position. Gender and ethnicity are varied
through the name of the person in the vignette, which is either a common traditional Danish feminine or
masculine name or a common feminine or masculine minority name of Middle Eastern (Muslim) origin. The
names that are included in the vignettes are chosen so as to be common and have as clear a cultural and
gendered reference as possible (using a list from the Department of Scandinavian Research, University of
Copenhagen, and an online database of names, their significance and distribution in Danish society based
on numbers from Statistics Denmark). Only 30 % of the names of the persons in the vignettes are Middle
Eastern names, so as to not make the combination of names in the vignette sample of each respondent too
unlikely to encounter in the Danish society. For the sake of simplicity, we chose only to include minority
names with a Middle Eastern cultural frame of reference (rather than also including names of several
different cultural references e.g. Hispanic, Chinese). Since the minority names in the vignettes may be used
in different regions and may be affiliated with different cultures and nationalities, the interpretation of the
names may differ across respondents. Despite such differences in interpretations of the names, we have
sought to ensure that the ethnically Danish names are traditionally Danish and that the minority names will
be interpreted as belonging to a minority in the Danish society and more specifically a minority of Middle
Eastern origin.

Age is varied in four categories in the single adult vignettes (22 years, 38 years, 54 years and 70 years) and
for the sake of plausibility only in two categories in the parent with one child vignettes (38 years and 54
years). Position describes affiliation with the labour market and covers three broad categories: student,
employed and recipient of pension. The significance of receiving pension will differ depending on the age of
the person in the vignette, for vignettes describing a young person the reader will probably assume that the
person receives early retirement pension, while the reader will probably assume the person to be a
recipient of retirement pension in vignettes describing an older person. These variations of position are
included to allow respondents to make their judgements of poverty depend on the social position and life
phase the vignette person is in (even though these categories only allow for very rough differentiations).

Sampling the vignettes
Randomly combining these different factors in the vignettes results in a very large number of possible
vignettes (the vignette universe). For the single adult vignette, the vignette universe is for instance
1.411.480 possible combinations. Due to the large number of possible vignettes, we cannot expect every
combination to be included in the study, far from it. This is, however, not problematic as long as we draw
the vignettes – and thus the combination of outcomes on the factors in each vignette - randomly. To cover
as large an amount of different combinations of factors as possible, we constructed 130 unique
questionnaires each including 20 randomly drawn vignettes. Thus, the study was designed so that each
respondent from each of the educational institutions included in the study was given a unique combination
of vignettes. Since each questionnaire includes 10 single adult vignettes and 10 single parent with one child
vignettes, the study includes 1300 single adult vignettes with random combinations of outcomes on the
factors varied and 1300 parent with one child vignettes with random combinations. In Table 2 below are
two examples of vignettes from the questionnaire.

                                                       6
Draft – do not quote

Table 2: Examples of vignettes from the questionnaire

Single adult vignette example:
Thomas is 22 years old and single. He is a student and has an income of 4.833 kr. monthly after taxes. He
has for 3 years due to financial reasons refrained from buying clothes, shoes and outdoor things and eating
fresh fruit and vegetables on a daily basis.

Single parent with one child vignette example:
Fatma is 38 years old and is alone with her 12 year old child. She is a student and has an income of 9.324
kr. monthly after taxes. She has for 2 years due to financial reasons refrained from visiting the dentist and
letting the child participate in school trips, participate in club activities etc.

Several statistical advantages pertain to the factorial survey design. Firstly, it is possible for the researcher
to determine the distribution of each variable, and by making these distributions uniform or rectangular
(equal proportion of each outcome) issues of having too few cases with a given outcome to perform a
sound analysis can be avoided. Further, the researcher can assure that there are only minimal correlations
between variables, which improves the estimates and diminishes problems of multicollinearity
(multicollinearity occurs when two variables are highly correlated making it difficult to separate out the
effect of each variable from the other, thus making the estimates imprecise) (Alves & Rossi 1978: 546).
When the random combinations of factors in all vignettes have been drawn, it is possible to check the
correlations between the independent vignette-variables before conducting the study. This is done in order
to ensure that large correlations have not occurred by chance and thus to ensure that it will be possible to
separate the effect of each factor in the analysis. Such a check showed only minor correlations among
vignette-variables in this study.

Dependent variable
Only one question is asked in relation to each vignette and that is a question asking the respondent to
judge the degree to which the person described in the vignettes is poor ranging from extremely poor to
extremely wealthy on a 9-point scale. This is the dependent variable in the analysis where we seek to clarify
how respondent characteristics and the factors varied across the vignettes impact on the judgement of
poverty.

Independent variables regarding respondent characteristics
Several independent variables describing respondent characteristics are included in the study: gender, age,
children, marital status, educational affiliation, year of study, income and sucjective experience of poverty.
The item measuring subjective experience of poverty calls on the respondent to describe him- or herself on
the same scale as the judgements of poverty made in the vignettes. This subjective measure may more
accurately than the actual income level account for the respondents’ experienced financial situation and is
therefore included in the analysis.

Sample
The data was collected from a sample of students of professions that are likely to encounter poverty in
their professional work. The study thus includes bachelor students from the fields of nutrition and health,
nursery teaching and social work. We had intended to also include students from the field of school

                                                        7
Draft – do not quote

teaching, but due to practical hindrances this was not possible (but a later stage of data collection might be
completed and in this we would seek to include school teachers and also nurses).

The data was collected during scheduled lectures on either poverty or vignette methodology. The data
collection was placed in the beginning of each class and followed a standardized protocol. In this way, a
high rate of attendance at the time of data collection was ensured, at the same time as the study carried
out could be used as a point of reference in the following lecture, thus functioning as a pedagogical tool. All
students present were handed a paper-based questionnaire and only very few students did not return
these. The students present in class were not systematically counted on all occasions of data collection, but
since the questionnaire was answered in class while the lecturer was present, students were urged to
participate and it was the impression that the amount of students present who did not respond at all to the
questionnaire was extremely low (less than 5 %).

A total of 327 respondents completed the questionnaire. In terms of educational affiliation, 37,6 % are
students of nursery teaching, 21,1 % are students of nutrition and health and 41,2 % are students of social
work. The sample is predominantly female, 81 % of respondents are female. The age ranges from 19-57
years with an average of 25 years, 44 % are married/cohabiting and 18 % have children.

The data collection was constrained by the practical concerns of establishing contact to educational
institutions and setting up data collection in relevant classes and therefore the students participating are at
different stages of their education. The students of nutrition and health and most students of nursery
teaching are in their first year of study, while the social work students are between their second and fourth
year of study. Thus, a systematic comparison across educational affiliations is not warranted by the data.
However, since the main goal is to examine the factors impacting the judgements of poverty, we were
willing to accept this shortfall of the data.

The data was mainly collected in Copenhagen. A supplementary data collection including only social
workers was carried out in Aalborg. The data was collected in cooperation between Metropolitan
University College, University College Capital and Aalborg University.

Analysis
In the data there are two levels of analysis present: the vignettes and the factors varied across vignettes
and the individuals (respondents) and their characteristics. This entails that there are both independent
variables pertaining to the respondents (age, gender, income, educational affiliation etc.) and independent
variables pertaining to the vignettes, i.e. the factors that vary across vignettes (deprivations, position,
ethnicity etc.). Hence, the data contains a hierarchical structure with both a level of vignettes and a level of
individuals. Since each respondent has replied to several vignettes, the level of vignettes are nested within
individuals (Hox, Kreft and Hermkens 1991).

Performing regular OLS-estimation on this kind of hierarchical data violates some assumptions of the OLS
regression and may lead to errors in the estimation of parameters and in the levels of significance
(artificially improving the level of significance). Furthermore, ignoring the hierarchical structure would
entail loosing valuable information about the data (Lolle 2003). Therefore, it is necessary to perform
multilevel analysis in order to account for the hierarchical structure of the data.

Preliminary results of such multilevel regression analysis will be presented at the conference.

                                                       8
Draft – do not quote

References
Alexander, Cheryl S. & Henry Jay Becker (1978): The use of vignettes in survey research, Public Opinion
     Quarterly, 42, 93-104.
Allardt, Erik (1975): Att ha att älska att vara. Om välfärd i norden. Lund: Argos Förlag.
Alves, Wayne M. & Peter H. Rossi (1978): Who should get what? Fairness judgements of the distribution of
     earnings, The American Journal of Sociology, 84, 3, 541-64.
Arbejderbevægelsens Erhvervsråd (2012): Fattigdom i Danmark: Antallet af langvarigt fattige er steget med
    80 procent i Danmark. 13. September 2012, Schytz Juul, Jonas & Helene Regitze Lund Wandsøe.
Ejrnæs, Morten, Finn Kenneth Hansen, Henning Hansen, M. Azhar Hussain and Jørgen Elm Larsen (2011):
     På laveste sociale ydelser et år efter - en kvantitativ forløbsanalyse. København: CASA.
Ejrnæs, Morten & Merete Monrad (2010): Enighed, uenighed og udvikling: Pædagogisk faglighed i
     daginstitutioner. København: BUPL.
Fuller, Richard C. & Richard R. Myers (1941): The natural history of a social problem, American Sociological
     Review, 6, 3, 320-9.
Hansen, Finn Kenneth & M. Azhar Hussain (2009): Konsekvenser af de laveste sociale ydelser
    - forsørgelsesgrundlag og afsavn. København: CASA.
Hill, Richard J. (1981): Attitudes and behavior. In Social psychology, sociological perspectives, eds. Morris
      Rosenberg & Ralph H. Turner, p. 347-377. New York: Basic Books, Inc. Publishers.
Hillgaard & Keiser (1979)
Hox, Joop J., Ita G. G. Kreft and Piet L. J. Hermkens (1991): The analysis of factorial surveys, Sociological
     Methods & Research, 19, 4, 493-510.
Lolle, Henrik (2003): Multilevel analyse i SPSS 11.5, Særtryk Fra Institut for Økonomi, Politik og Forvaltning,
      AAU, 1.
Wallander, Lisa (2009): 25 years of factorial surveys in sociology: A review, Social Science Research, 38, 505-
    20.

                                                        9
You can also read