Assessing the fit of each item of the ‘Picture-Based Value Survey for Children’ into the theoretical structure of values

Full text

ASSESSING THE FIT OF EACH ITEM OF THE ‘PICTURE-BASED
VALUE SURVEY FOR CHILDREN’ INTO THE THEORETICAL
STRUCTURE OF VALUES
WOLFGANG BILSKY, WWU Münster, Germany
ANNA K. DÖRING, University of Westminster London, UK
PATRICK J.F. GROENEN, Erasmus University Rotterdam, The Netherlands
ABSTRACT
The Picture-Based Value Survey (PBVS-C) is based on Schwartz’s structural theory of values
and measures value preferences of young children. It has been applied in several studies around
the world. Due to some recurring deviations of single items from theoretical expectations we
took a closer look at their fit into the theoretical structure. We re-analysed the data from former
studies by using a strong confirmatory
MDS that enforces theory-based regional restrictions
onto the data (Bilsky, Borg, Janik & Groenen, 2015). To better understand how much each
single item affects the overall model fit, Stress-per-Point (SPP) coefficients were computed for
every item. Since the global MDS model and the external regional restrictions imposed by con-
firmatory MDS are theoretically grounded, SPP scores are diagnostically relevant with respect
to item and construct validity. The results of our re-analyses are presented and discussed with
regard to the observed misfit of items. Consequences for further research are outlined.
INTRODUCTION
The empirical analysis of children’s values does not have a long tradition. This is partly due to
the lack of adequate instruments that do take the still developing literacy of young children into
account. The Picture Based Value Survey for Children (PBVS-C; Döring et al., 2010) closed this
gap only a few years ago. This instrument is based on Schwartz’s (1992) theory on universals in
the content and structure of values.
Schwartz’s theory identifies ten basic values (i.e., value types) which differ with respect to
motivational content: Universalism (UN), Benevolence (BE), Tradition (TR), Conformity (CO),
Security (SE), Power (PO), Achievement (AC), Hedonism (HE), Stimulation (ST), and Self-
Direction (SD). These motivational differences have consequences for the value structure, which
takes the form of a circumplex (Borg & Shye, 1995) in two-dimensional space: According to
Schwartz’s theory, values of the same value type should form sectors with boundaries emanating
from a common origin; furthermore, adjacent sectors are expected to be motivationally more
similar than sectors farther apart. On a more general level, basic values are subsumed under four
higher-order values (HOVs): “Self-Transcendence” (UN, BE) opposed to “Self-Enhancement”
(PO, AC), and “Openness-to-Change” (HE, ST, SD) opposed to “Conservation” (TR, CO, SE;
see Appendix).
The PBVS-C was designed to measure value preferences of children aged 6 to 11. It was
constructed in close analogy to Schwartz’s Portrait Values Questionnaire (PVQ) and comprises
twenty items, each depicting one value-related situation. Every basic value is represented by two
items which are printed on removable stickers. Children are asked to paste them on a standard

response sheet with five graded alternatives from “very important” to “not at all important” in
my life (see Döring et al., 2010).
Until today the PBVS-C has been administered to children in various countries around the
world (Bilsky et al., 2013; Döring et al., 2010, 2015). Exploratory and weak confirmatory Multi-
dimensional Scaling (MDS; Borg & Groenen, 2005) were employed to compare the empirical
value structure with Schwartz’s structural model. Overall, the validity and the cross-cultural ap-
plicability of the instrument were supported. Some items, however, deviated repeatedly from the
theoretical structure. It is these deviations the present paper focuses upon.
METHOD
The present study builds on and extends a former one, in which value data from a Brazilian
sample of young children (Roazzi et al., 2011) were re-analysed by strong confirmatory ordinal MDS
(Bilsky et al., 2015). As in this former study, strong confirmatory MDS is accomplished by imposing
theory based regional restrictions on value data (Borg & Groenen, 2005).
Data
We analyse data that were collected with the PBVS-C from 10 samples of children aged 6 to
12. These samples were from eight different countries, five European and three extra-European1:
Germany (DE; N1=221, N2=294), Bulgaria (BU; N=439), France (FR; N=306), United Kingdom
(UK; N=352), Italy (IT; N=380), Turks living in Germany (gTR; N=117), Brazil (BR; N=182),
New Zealand (NZ; N=104), and United States (USA; N=69).
Facetization of Values
Former facetizations of values (Schwartz & Bilsky, 1987; Schwartz, 1992) considered only
one motivational facet. In this manuscript, we use two independent facets for analysing value
structure (Bilsky et al., 2015). They derive from a recent reframing of Schwartz’s two “basic
value dimensions” (Schwartz, 2006; see also Schwartz et al., 2012). This reframing implies dy-
namic principles which go beyond congruence and conflict among values (cf. Schwartz, 2006,
947f). The first principle differentiates between social and personal interests, the second con-
trasts anxiety and growth related objectives: the prevention of loss and the promotion of gain (see
Table A1 in the appendix). The following mapping sentence results from this distinction:
p considers value X that refers to his/her {social/personal} interests and the {promotion of
gain/prevention of loss} → {very important … not important} as a guiding principle for p’s life.
Strong Confirmatory MDS
The regional restrictions imposed on our data follow from the above mapping sentence
(Bilsky et al., 2015): Thus, every basic value (and the respective PBVS-items) can be character-
ized independently by the two facets “interest” and “objective” (see Figure 1 and Table A1, col-
umns 2, 4 and 5). In accordance with this facetization, our structural hypotheses suggest a split
into four quadrants in two-dimensional space, i.e. a duplex, as indicated in Figure 1.
1We thank our former students for their collaboration and for collecting these data in the context of their final theses:
K. Aryus, L. Drögekamp, V. Glatzel, N. Janowicz, M. Nyagolova, A. Petersen, E.R.Scheefer, T.S.Şimşek, and J.
Sindermann. The Brazilian data were first published in Roazzi et al. (2011).

Objective
promotion: gain
prevention: loss
Interest
social
1/1
Self-Transcendence:
Universalism
Benevolence
1/2
Conservation:
Security
Conformity
Tradition
personal
2/1
Openness to Change:
Hedonism
Stimulation
Self-Direction
2/2
Self-Enhancement:
Achievement
Power
Figure 1 Classification of the ten basic human values (Schwartz, 1992) according to the two
facets “Interest” and “Objective”
There exists no standard software until today which would allow testing such a structure by
confirmatory MDS. Therefore we used an experimental program written in MatLAB by Groenen
and already used for similar purposes in earlier studies (Borg et al., 2011; Bilsky et al., 2015). In
order to minimize the risk of running into local minima, we used a starting configuration derived
from Schwartz’s original model of value structure (Schwartz, 1992; see Bilsky & Janik, 2010,
for more detailed information).
To answer our central research question which items of the PBVS-C show recurring devia-
tions from the theoretical structure and to what extent, we compute two indices separately for all 10
samples: Stress-1 and Stress per Point (SPP). While Stress-1 informs about the mismatch between
the correlations and the corresponding distances of an MDS solution, SPP indicates how much
every single item contributes to this mismatch. MDS plots complement our computations.
RESULTS
Table 1 summarizes the results of our confirmatory MDS. For a quick overview, SPP-
coefficients that are equal to or bigger than the respective mean Stress per Point are marked by grey
underlay, those that are equal to or bigger than the third quartile by bold print. The contribution of
AC1 to overall stress, for example, is below-average in all samples; the contribution of CO2 and SD1,
in contrast, is above-average in nine of ten samples.
DISCUSSION
A strong confirmatory MDS enforces theory-based regional restrictions onto the resulting MDS
plot. In the present case, these restrictions are specified by the two facets “interest” and “objective”
(Figure 1). They force the PBVS-items into one of four quadrants in two-dimensional space accord-
ing to their motivational content. These quadrants correspond to the four higher-order values (HOVs)
“self-enhancement vs. self-transcendence”, and “openness-to-change vs. conservation”. They charac-
terize value structure as supposed by Schwartz (1992) value theory on the most general level.

Table 1 PBVS-C; ordinal MDS with theory-based Starting Configuration and Regional Restrictions:
Stress per Point (SPP)
PBVS-
Item
DE1
DE2
BU
FR
UK
IT
gTR
BR
NZ
USA
N
221
294
439
306
352
380
117
182
104
69
UN1
0.036
0.020
0.042
0.031
0.049
0.022
0.085
0.045
0.053
0.070
UN2
0.031
0.024
0.012
0.029
0.018
0.051
0.052
0.024
0.044
0.099
BE1
0.017
0.008
0.020
0.018
0.013
0.018
0.012
0.076
0.064
0.072
BE2
0.055
0.052
0.032
0.067
0.049
0.055
0.061
0.092
0.057
0.083
TR1
0.049
0.029
0.017
0.057
0.044
0.044
0.112
0.104
0.056
0.035
TR2
0.054
0.086
0.014
0.062
0.014
0.063
0.075
0.106
0.087
0.049
CO1
0.029
0.005
0.007
0.028
0.012
0.027
0.041
0.101
0.055
0.050
CO2
0.079
0.056
0.057
0.055
0.045
0.070
0.133
0.091
0.099
0.026
SE1
0.064
0.038
0.033
0.018
0.024
0.031
0.048
0.104
0.079
0.031
SE2
0.075
0.086
0.066
0.102
0.064
0.058
0.098
0.039
0.086
0.061
PO1
0.026
0.020
0.027
0.020
0.021
0.015
0.024
0.042
0.058
0.047
PO2
0.038
0.021
0.037
0.037
0.029
0.032
0.068
0.056
0.054
0.078
AC1
0.049
0.021
0.008
0.031
0.020
0.033
0.033
0.045
0.027
0.054
AC2
0.101
0.028
0.045
0.059
0.054
0.069
0.109
0.165
0.057
0.073
HE1
0.035
0.047
0.055
0.042
0.028
0.029
0.016
0.051
0.029
0.060
HE2
0.106
0.020
0.053
0.040
0.038
0.094
0.044
0.088
0.043
0.101
ST1
0.036
0.023
0.030
0.027
0.019
0.029
0.096
0.065
0.055
0.055
ST2
0.046
0.014
0.026
0.057
0.036
0.019
0.046
0.044
0.041
0.075
SD1
0.073
0.074
0.114
0.073
0.054
0.067
0.092
0.112
0.048
0.077
SD2
0.045
0.029
0.064
0.058
0.042
0.021
0.065
0.144
0.059
0.081
Mean
0.052
0.035
0.038
0.046
0.034
0.042
0.065
0.080
0.058
0.064
Stress-1
0.228
0.187
0.195
0.213
0.184
0.206
0.256
0.282
0.240
0.253
3rd Quar-
tile
0.067
0.048
0.053
0.058
0.046
0.059
0.093
0.104
0.060
0.077
Note. Grey underlay: SPP >= Mean; bold figures: SPP >= 3rd Quartile
Stress-1 informs about the overall mismatch between the correlations of the 20 PBVS- items and
the corresponding distances of the MDS solution. This mismatch is smallest for the sample from
Great Britain (GB: 0.184) and highest for that from Brazil (BR: 0.282). Despite the observed range of
stress coefficients, however, only the coefficient of the Brazilian sample challenges the hypotheses of
non-randomness (Spence & Ogilvie, 1973).
When calculated in the context of a confirmatory MDS, Stress per Point indicates how well the
respective item fits the underlying theoretical model. Singular instances of above-average SPPs may
be attributable to random error, to a poor adaptation of the item to the target population, or to cultural
differences. Which of these alternatives comes true can only be answered by systematic replications.

Figure 2 Ordinal MDS of PBVS-Items with Starting-Configuration and Regional Restrictions;
PBVS-C above and PBVS-10 (short version) below
Finding above-average SPPs repeatedly within and/or across different cultures, however, chal-
lenges the validity of the respective item. As can be seen from Table 1, the number of above-average
SPPs differs considerably between the 20 PBVS-items. For the present it appears too early to draw
far-reaching conclusions from these findings. First, the size of the individual samples and the number
of samples per country is still very small despite the internationally growing number of studies using
the PBVS-C. Second, their representativeness is not warranted. The observed pattern of SPPs found
in the present data base should however be traced in future studies. In the case of recurring high
SPP-coefficients in future and more representative studies, revising or substituting the respective
items would be a reasonable consequence for ameliorating the PBVS-C in the long run.
While the revision or replacement of items is not an option for now, we used our knowledge of
SPPs for another purpose: item selection. This is possible because every basic value is represented by
two items in the PBVS-C. As can be seen from Table 1, for most values one of the two items showed
repeatedly below-average SPPs. Therefore we chose these items to construct a short version of the
Picture-Based Value Survey. The respective items of this PBVS-10 are: UN2, BE1, TR1, CO1, SE1,
PO1, AC1, HE1, ST1, and SD2 (see Figure A1 in the appendix). Such a short version could be useful
as part of a more comprehensive survey on different developmental topics or in cohort studies.

For illustrative purposes, we conducted two ordinal MDS with regional restrictions and these ten
items. Data were from the two German samples (see Table 1).The results for the standard and for the
short form of the Picture-Based Value Survey are summarized in Figure 2. As can be seen, items can
be partitioned in four quadrants according to Schwartz’s (1992) theory. However, their sequence
partly deviates from Schwartz’s structural model. Whether these deviations are random or systematic
needs additional analyses and further studies.
REFERENCES
Bilsky, W., Borg, I., Janik, M., & Groenen, P. (2015). Children's value structures - Imposing
theory-based regional restrictions onto an ordinal MDS solution. In A. Roazzi, B.C. de
Souza & W. Bilsky (Eds.), Searching for Structure in Complex Social, Cultural and Psy-
chological Phenomena (pp. 23-37). Recife: Editora Universitária – UFPE.
Bilsky, W., Döring, A.K., van Beeck, F., Rose, I., Schmitz, J., Aryus, K., Drögekamp, L., & Sin-
dermann, J. (2013). Investigating children's values and value preferences - testing and ex-
panding the limits. Swiss Journal of Psychology, 72(3), 123-136.
Bilsky, W. & Janik, M. (2010). Investigating value structure: Using theory-based starting con-
figuration in Multidimensional Scaling. Revista de Psicología Social, 25(3), 341-349.
Borg, I. & Groenen, P. (2005). Modern Multidimensional Scaling. Berlin: Springer.
Borg, I., Groenen, P.J.F., Jehn K.A., Bilsky W., & Schwartz, S.H. (2011). Embedding the organ-
izational culture profile into Schwartz's theory of universals in values. Journal of Personnel
Psychology, 10(1), 1-12.
Borg, I. & Shye, S. (1995). Facet Theory. Form and content. Thousand Oaks: Sage.
Döring, A.K., Blauensteiner, A., Aryus, K., Drögekamp, L., & Bilsky, W. (2010). Assessing
values at an early age: the picture-based value survey for children (PBVS-C). Journal of
Personality Assessment, 92(5), 439-448.
Döring, A.K., Schwartz, S.H., Cieciuch, J., Groenen, P.J.F., Glatzel, V., Harasimczuk, J., Jan-
owicz, N., Nyagolova, M., Scheefer, E.R., Allritz, M., Milfont, T.L., & Bilsky, W. (2015).
Cross-cultural evidence of value structures and priorities in childhood. British Journal of
Psychology, 106, 675-699.
Roazzi, A., Döring, A.K., Gomes, Y.A., Souza, B.C. & Bilsky, W. (2011). The emergence of a
value structure at an early age. In Y. Fisher & I.A. Friedman (Eds.), New horizons for Facet
Theory (pp. 99-112). Israel: FTA Publication.
Schwartz, S.H. (1992). Universals in the content and structure of values: theoretical advances
and empirical tests in 20 countries. In M. Zanna (Ed.), Advances in Experimental Social
Psychology, Vol. 25 (pp. 1-65). New York: Academic Press.
Schwartz, S.H. (2005). Human values. European Social Survey Education Net. Retrieved from:
http://essedunet.nsd.uib.no/cms/topics/1/5/, 31.10.2008.
Schwartz, S.H. (2006). Les valeurs de base de la personne: Théorie, mesures et applications. Re-
vue Française de Sociologie, 47(4), 929-968.
Spence, I., & Ogilvie, J. C. (1973). A table of expected stress values for random rankings in
nonmetric multidimensional scaling. Multivariate Behavioral Research, 8, 511–517.

Appendix
Table A1: Human Values (Schwartz, 1992), classified according to “Interest” (social vs. per-
sonal) and “Objective” (promotion: gain vs. prevention: loss); see Schwartz (2006; Bilsky et
al., 2015)
Higher-Order
Values
(HOV)
Basic Values
and
PBVS-Items
Central
Motivational Goal a
Facets
and
Elements
Interest
Objective
Self-Tran-
scendence
Universalism
UN1, UN2
Understanding, appreciation,
tolerance, and protection for
the welfare of all people and
for nature
social promotion:
gain
Benevolence
BE1, BE2
Preserving and enhancing the
welfare of those with whom
one is in frequent personal
contact (the 'in-group')
social promotion:
gain
Conservation
Tradition
TR1, TR2
Respect, commitment, and
acceptance of the customs and
ideas that traditional culture or
religion provide the self
social prevention:
loss
Conformity
CO1, CO2
Restraint of actions, inclina-
tions, and impulses likely to
upset or harm others and vio-
late social expectations or
norms
social prevention:
loss
Security
SE1, SE2
Safety, harmony, and stability
of society, of relationships,
and of self
social prevention:
loss
Self-
Enhancement
Power
PO1, PO2
Social status and prestige,
control or dominance over
people and resources
personal prevention:
loss
Achievement
AC1, AC2
Personal success through
demonstrating competence
according to social standards
personal prevention:
loss
Openness
to
Change
Hedonismb
HE1, HE2
Pleasure and sensuous gratifi-
cation for oneself
personal
promotion:
gain
Stimulation
ST1, ST2
Excitement, novelty, and chal-
lenge in life
personal
promotion:
gain
Self-Direction
SD1, SD2
Independent thought and ac-
tion; choosing, creating, ex-
ploring
personal promotion:
gain
a Central motivational goals (Schwartz, 2005; http://essedunet.nsd.uib.no/cms/topics/1/)
b Hedonism shares elements of both self-enhancement and openness to change, but is closer to
openness in most cases (Schwartz, 2005).

Figure A1 Short Version of the Picture-Based Value Survey for Children (PBVS-10)
Universalism
Benevolence
Tradition
Security
Power
Achievement
Hedonism
Stimulation
Selfdirection
Conformity