IAOS2018 - Modelling for Improving Measurement: Strategies for Contextualization of Well-Being, W. Okrasa, D. Rozkrut

Modelling for Improving
Measurement:
Strategies for Contextualization of Well-Being
Włodzimierz Okrasa
Dominik Rozkrut
Statistics Poland
Paris, 19 – 21 September 2018
IAOS2018_OECD Conference
Better Statistics for Better Lives
1

demand for models/methods program-/policy-driven
information of converting information request for research
into a knowledge base /users needs
process*:
from data collection to knowledge utilization and decision making
(evidence-based policy)
data  information  knowledge  decision
c o m m u n i c a t i o n
strategy
quantification
and
measurement
*) „Numbers. figures and patterns are first of all of the communication strategies …”
(Porter. 1995) . and as such are ‘socially constructed (Schield. 2013) 3
data/information
producer (statistician)
knowledge producer
(researcher/modelers
)
knowledge user (policy
makers & practitioners)

From modelling to measuring (or measuring for modeling):
Communication chain in statistical process:
between data and decision/policy-making – alternative paradigms
Combining two paradigms – toward a new meta-paradigm?
(a) data ‘first’ (as a policy input)
 Data-based Policy/decision making
 Data-driven (evidence-based policy)*
 Data/evidence-informed
(b) users ‘first’ (needs for…)
 Policy-oriented
 Policy-based Data/information production
 Policy-driven **
*) „ Evidence-influenced politics is suggested as a more informative metaphor. descriptively
and prescriptively. than evidence-based policy” (Prewitt. 2012; p. 4).
**)”Evidence-based policy or policy-based evidence?” (Sanderson 2011) 4

Scientific (statistical) communication – elements of the process
Communication as a way of ‘contextualization’ [in Desrosieres’ sense: statistics (generated in
statistical process) accounts for the peculiarities of social. cultural and political contexts in
which it is practiced - ‘place’ provides the circumstance conducive for their identification
 hence. ‘spatial approach’
Communication entails the following elements (eg. Maggino and Trapani. 2010. Nymand-
Andersen. 2017.)
(i) contents /information (‘message’) to be communicated /transmitted to a receiver;
(ii) communicator / transmitter (‘statisitician)
(iii) receiver / audience (addressees – experts / researchers, policy makers, users of
statistical data/product etc.
(iv) communication channel - “a medium through which a message is directed to and
exchanged with its intended audience”; multiple-channel communication;
(v) code (transmitter’s / receiver’s code):
- the way statistics are reported (and presented);
- the tools used in order to transmit statistics;
(vi) context; (vii) feedback and (viii) noise.
5

Contextualization starts with quantification:
convention and measurement
 Quantification: process of data transormation in numerical form through
measuring characterisitcs of a set of units or items on a scale in form of numbers.
letters or verbally (e.g. Federer. 1991)
• Quantification is a more than measurement (Desrosières . 2001) - it requires
conventions on which the measurement is based
• Quantifcation (a sociological phenomenon. eg.. Espeland and Stevens. 2008; Alonso
and Starr. 1987; Federer. 1991) : production of information and
communication using numbers (precondition for measurement is
classification).
6
 system of communication codes - in the context of spatial analysis
 integration of data from different sources

Communication between ‘two communities’:
researchers and policy makers (Prewitt et al. 2012)
• Translation. a supply-side solution to the use problem - applicable to questions of using
science in policy (eg. social science, medical science).
• Brokering involves filtering, synthesizing, summarizing and disseminating research
findings in user-friendly packages.
• (in contrast to translation strategies) brokering involves two-way communication
• An interaction model. covers a family of ideas directed to systemic changes in the means
and opportunities for relationships between researchers and policy makers
• emphasis on use-inspired research and increased visibility and an insight into what
the use of science means in practice.
The issue of particular importance in this context concerns how data collectors/producers
communicate with data providers?
consequences for measuring well-being

Issues in measuring community well-being
for policy research and evaluation purposes
Interpetation of Community
well-being
[‘concepts s developed by synthesizing
research constructs related to
resident’s perceptions of the
community, … needs fulfillmemnts,
observable community conditions, and
the social and cultural context…” (Sung
and Phillips 2016:2 [in Phillps and
Wong2017: xxix])
Focus on monitoring and evaluation
– possible changes
Changes in community
relevant (well-being)
measures only
Changes in both
community and individual
(residents’) measures
- objective community well-
being (CWB)
A. One-level
cross-section or dynamic
(eg.. scales of ‘local deprivation’)
B. Multilevel w/cross-level
effects
- subjective communiy well-
being (SCWB)
C. One-level w/CWB as a
‘context’ (subjective ‘cohesion’,
’sense of community’*)
)
D. Multilevel
with (possible) reciprocal
influence and interaction
*) For example. the Sense of Community Index developed by Chavis and McMillan (1986) embraces four
components – membership, influence, meeting needs,and a shared emotional connection – as factors of well-
being (Chavis et al.. 2008).

Community and individual well-being relatonship
– interconnected research tasks: measurement – data – models
 Conceptualization. operationazation and the measurement
what ? - how? – and why?
• there is no theoretical justification for maximizing either happiness or life satisfaction.
because neither correspond to utility - e.g. Gibson (2016);
• acc. to Sen: ‘Happiness is not all that matters. but first of all. it does matter (…). and second.
it can often provide useful evidence on whether or not we are achieving our objectives in
general’ (Sen 2008).
• Capability approach – the ‘functioning‘mode(*)
[following A. Sen: well-being should be
conceived directly in terms of functionings and capabilities instead of resources or utility]
 time-use as a domain(***)
or as a measure of W-B
(*) ‘Functionings is a broad term used to refer to the activities and situations that people spontaneously
recognize to be important (Alkire 2015: 3-4).
**) Capability approach at the community level (?) Eg., community well-being acc. to Wiesman and Brasher
(2008):”combination of …conditions .. essentials [for individuals and their communities] to fluorish and
fulfill their potentials” (in Phillips and Wong (2017:xxx).
(***) For instance, OECD Better Life Index (2015); Canadian Index of W-B; … 9

Remarks on conceptualization of ‘community well-being’
• There are several reasons for focusing on community well-being in both research
and policy considerations. especially in the local development context. Many of them
have been recognized and discussed thoroughly in the literature either as a part of
the process or outcome of such development, challenging the tradition of using GDP
and other economic indicators as measures of social progress (Philliand Wong. 2017.
Kim and Ludwigs. 2017. Lee et al. 2015).
• Efforts to go ‘beyond GDP’ in the evaluation of socio-economic progress were
undertaken several decades ago – for instance social indicators movement see
Land’s and Michalos’ “Fifty Years After the Social Indicators Movement...” (2017).
• Methods of community well-being assessment including subjective aspects of well-
being are becoming standard tools for policy purposes in several countries (eg.
Australia, Canada, the USA and the UK).
• They all have one feature in common. namely. they are based on self-reported
feeling about selected aspects of well-being in connection with community, and
community itself is among the components of the well-being measures.

Two complementary outlooks:
(i) methodological. starting with an overview of the main approaches
(paradigms) to measuring CWB in public statistics and
(ii) analytical. using distributional potentials of the CWB indicators for geographic
targeting of (public) resources – e.g. subjective well-being as welfare
measure (Fischer, 2009) including their effects for changes/reduction in :
- the level of local deprivation (gminas) and contributing to ‘social progress’
(‚social welfare’) in the local context (beyond GDP : Stiglitz et al.. 2010.
OECD.2013/2015) and using dual type measures – objective and subjective – for
CWB (e.g.. Kim and Ludwigs. 2017. Okrasa 2017).
- inequalities between 'localities' (NUTS5 units / gminas) - implications for
spatial cohesion and ‘community cohesion’ (Okrasa 2017. Okrasa and Rozkrut 2018)-
[e.g.. following Forrest and Kearns (2001) and focusing on selected (out
of five) topic areas: reduction in wealth disparities and place attachment and
identity].
[The others are: common values and civic culture - social order and social control - social networks and
Conceptualization and operationalization of community well-being (CW-B)
in the evaluation policy research context

Community well-being – contin.
operationalization and measurement approaches
Formal concept of
the ‘community’ as
a type of set
Interpretation of well-being characteritics / measures:
Attributive
/non-decomposable
Descriptive
Collective
/focus on community
as an entity / unit
aggregative/ ‘holistic’:
community deprivation (Okrasa 2014.
Strubelt 2005); commnity survey
data; also Hunter’s typology;
social indicators
typological /taxonomic - top-
down or normative
conceptualization –based : OECD
2013/2015; national versions (eg.
Atlas Project (Kim and Ludvigs
2017).
Distributive
/community as a
composition of
residents /members of
compositional: sub-population group-
derived composed characterisitcs; eg.
‘Sense of Community’; household
survey -based community data
individual summary item /bottom-
up or data-driven:
TUS /DRM data-based (eg. Krueger
et al. 2009, Okrasa 2017; Okrasa
Mixed

Choice of modelling strategy affects data and measurement
(methodological communication)
– towards a multilevel approach with moderating factor and interaction effect
(„D”)[Pragmatic reason / policy purpose] Putting emphasis on the relationships between
community and individual well-being:
- demand for devices (data and measures) to deal with problem of an optimal
allocation of scare resources across communities (communes/gminas) while accounting for
individual (subjective) well-being of residents.
 research design and the measurement issues:
- ‘nested’ (hierarchical) data structure and/or
- parallel compatible measures of community and individual well-being [range
of measures: OECD 2013/15-Better Life/How’s Life?; Steuer & Marks/LSE- Project. 2008;Philips
and Wong (eds) 2017.]
 multilevel spatial modelling -
interaction-focused model (Subramanian 2010; Pierewan and Tampubolon 2014)
- influence (‘causal’) and moderating factor model (Morgan and Winship 2007; Hong 2015.
Okrasa 2017).
 pre-condition: integration of data from different sources

Types of goods, units of analysis and measures of well-being:
a data-focused typology
Mixed /hybrid measures: TUS – activities: ‘pleasent-
unpleasant (eg. Kahneman and Krueger 2006; Okrasa and Rozkrut
2018)
14
Mixed
Mixed /hybrid measures: community cohesion (eg. Cantle 2008)

Individual (Subjective) well-being:
TUS data-based measures
 Social indicators approach – attempts to exploit TUS data (Th. Juster; and others. e.g.
F. Andrews 80s.) :
- survey research (day reconstruction techniques- e.g. day and week-recall data
-TUS_2013 ; N=23 )
 Psychometric measures
 Econometric research and econometric/psychometric combined approaches - Krueger
and Khaneman et al. (2008) – indicator of emotion / negative /positive affects
associated with activities / ‘time of unpleasant state’ - U-index :
Ui = Σj Iijhij / Σjhij (in TUS2013: I = -1. 0. +1)
and U = Σi(Σj Iijhij / Σjhij ) / N for N-persons / group in population
(used also in poverty research) 15

 Spatial integration of data – geo-referenced data (coordinates X. Y) and statistical
matching techniques (e.g.. D'Orazio et al. 2006) allow to solve several methodological
problems, for instance:
 integration of objective and subjective measures of well-being such as individual
and community well-being - a dual system of indicators (beyond GDP)
 multidisciplinary integration of scientific knowledge – spatially integrated social
research (Goodchild et al. 2004) a framework for integration of social and
behavioral disciplines (Stimson 2014; Okrasa 2012;)
Spatial integration in social research:
– key role of geographical concepts and spatial data: place, locality.
distance/proximity, distribution, neighborhood. region;
– exploration and analysis of spatial patterns of social phenomena.
Integration of data from different sources
- methodological motivation: spatial contextualization
16

Spatial integration – possible research situations
17

-enabling analysis of relationship between individual (subjective) and community
well-being due to construction of multilevel multi-source database:
• Local Data Bank (LDB)
• Time Use Survey data (TUS /GUS 2013)
• Social Diagnosis (SD)
[Examplification] Spatial integration:
‘place’ / locality (NTS5=gmina) as an integrator / merging element
18

A. The nonresponse (missing observations) problem and
identification of missing data generating mechanism standard
procedures (omitted as not specific to data integration context – eg..
Särndal and Lundström. 2005) – see D’Orazio et al.. op cit.
B. Problem of missing joint information on random variables X. Y. Z
with density f(x. y. z).
- units in data set A (sample survey A) have Z missing:
(xa
A
. ya
A
) = (xa1
A
.…. xaP
A
. ya1
A
.….yaQ
A
).
a = 1. …. nA
- units in data set B (sample survey B) have Y missing:
(xb
B
. zb
B
) = (xb1
B
.…. xbP
B
. zb1
B
.….zbQ
B
).
b = 1. …. nB
[Data set A U B is unique and has joint distribution f(x.y.z) [strong
version requires that the datasets are from the same time of research]
Statistical matching problem
19

Statistical matching problem – contin.
(D’Orazio et al.. 2006. p.5)
Sample
Y1 … Yq … YQ X1 … Xp … XP Z1 … Zr … ZR
A
y11
… y1q … y1Q x11 … x1p … x1P
… … … … … … … … … … .
ynA1 … ynAq … ynAQ xnA1 … xnAq … xnAQ
B
x11 … x1p … x1P z11 … z1p … z1P
… … … … … … … … … …
xnB1 … xnBq … xnBQ znB1 … znBq … znBQ
Unobserved variables
Unobserved variables
20

Multisource analytical database/MADb:
Local Data Bank (LDB-2014)
Time Use Survey (TUS-2013)
Social Diagnosis 2013.
L1
L2

 Multi-source database:
(a) commune/gmina level data: Regional / Local Data Base (CSO – public
file 2004. 2008. 2010 and 2012. 2014. and 2016); NUTS5/LAU2;
(N= 2 478);
 Measuring area deprivation at the commune level
 Multidimensional Index of Local Deprivation (MILD)
‘Confirmatory’ Factor Analysis / PCA (single-factor selection):
Eleven (pre-selected) domains of local deprivation - each characterized
by a number of original items: ecology – finance – economy –
infrastructure – communal utilities – culture – housing – social
assistance – labour market – education – health [65 items]
DATA and MEASURES for selected models:
Local Deprivation and Subjective Well-Being (SW-B)
22

Measures employed in the analysis:
–CWB: aggregative/holistic – objective : MILD
–SCWB: individual summaries – (quasi) subjective :
TUS database
–SCWB: compositional – subjective: Social Diagnosis:
3 scales concerning satisfaction from selected
aspects of life in the community:

Model
Women
Std. Coeff.
t Signific.
Men
Std. Coeff.
t Signific.
BetaBeta
Constant 3.842 0.000 1.819 0.069
Job-time. major and additional 0.331 22.809 0.000 0.398 21.533 0.000
Monthly income -0.002 -0.151 0.880 -0.025 -1.365 0.172
Risk due to local labor market
deprivation
0.047 3.028 0.002 0.063 3.231 0.001
Subsidies to commune (gmina)
received bigger than expected (as
proportional to local deprivation)
0.046 2.956 0.003 0.034 1.749 0.080
2 2
Table 1. Approximation of ‘life satisfaction equation’ (eg. Clark. 2018)
using TUS2013-data (U-index for ‘life satisfaction’) and BDL-data (MILD2016)
Opposite directions of influence of income and time of work on well-being , acc.to U-index: while greater income is positive for
individual well-being, the increased amount of time spent on work is negative (U- index increases) - question arises about the
poin t of balance (tradeoff between the two factors of well-being). (See Kahneman and Deaton 2010 for comparison of income effect)

Satisfaction from life and main source of income
(CORA-outcome for 400 HHlds surveyd in South-East
Poland/Subcarpathian region
In addition to income
effect hypothesis should
be considered the way
income is earned, as a
control variable.

Example: Odds for Sense of Community (SoC) by Main source of income; Type
of Hhld; Level of local deprivation (MILD) [S-E Region/Subcarpathian; N=400)
Results of Multilnomial
Logistic Regression
confirm that pattern of
relationships between
subjective aspect of
Community well-being
/Sense of Comunity
and income soources
differs for various types
of the sources; and also
for types of Hhld; less so
for the level of local
deprivation

Odds for Satisfaction from Life (aspects of) by Main source of
income; Type of Hhld; and Level of local deprivation MILD).
Subcarpatian
Observation: The much bigger chance for high Safisfaction from Life among selfemployed and
also among earners and even pensioners (compared to farmers); also among residents of less-
deprived communes (compared to high local deprivation/MILD).

Odds of experiencing 'non-positive' feeling associated with activities, U -
index, depending on:
(a) the level of local deprivation/MILD (b) the size of the living
place

Cross-level operating
factors
of well-being
Models type I: Effects of micro –
macro influence
29

Individual well-being and community well-being
relationship
A multilevel model (eg.. Subramanian. 2010)
• yij; well-being of i individual in j commune/gmin ;
• x1ij predictor of indywidual (level-1) – such as: age. education. or
satisfaction (e.g. from life in a community. family life . etc.)
• predictor of level-2 / (macro-level): Multideminsonal Index of Local
Deprivation for j-commune/gmina /MILDj
Model for level-1: yij = β0j + βl x1ij + e0ij (1)
where: β0j – refers to x0ij average score on a well-being scale in j-th
commune/gmina (eg.. . ‘less affluent' or ‘low-income’. etc.. for cases < Me. x0ij
=1);
βl – average differentiation of individual well-being associated with
individual material status (x1ij), across all territorial units (communes/gminas;)
e0ij – residual term for the level-1.
30

Model – cont.
• Modeling fixed-effect we include a level-2 predictor – MILD -(index of local
deprivation) along with individual characteristics. including interaction term
between the two levels characterisitcs
β0j = β0 + α1MILD1j + u0j (3)
β1j = β1 + α2MILD1j + u1j (4)
where MILD1j – context variable. predictor of differences between communes
(gminas.) 31
• Treating β0j as random variable : (β0j – β0) + u 0j . where u0j is locally-
specific associated with average value of β0) for a specified group (eg.
less satisfied from a community) and grouping them into fixed and
random part components (e0ij + u0j ) we obtain variance component
model. or random-intercept model:
yij = β0 + βl x1ij + (e0ij + u0j ) (2)

Two-level model can be specified as below:
yij = β0 + βl x1ij + α1w1j + α2w1j x1ij + (u0j + u1j x1ij + e1ij x1ij + e2ij x2ij ) (5)
32
- where w1j
is a 2-level predictor. i.e. the index of local deprivation. MILD1j
.
The following model was calculated using data from Time Use Survey 2013 (22
695 and 24 065 persons surveyed for weekdays and for weekends/holidays.
respectively):
IWB(U-index)ij
= β00
+ β10
educationij
+ β20
ageij
+ α1
MILDj
+ α11
educationij
*MILDj
+ α21
ageij
* MILDj
+ u1j
educationij
+ u2j
age + u0j
+ eij (6)
[It is assumed that] Such a specification of cross-level (between individual and
community/gmina measures of well-being) with interaction effect should ensure
robust estimation (e.g.. Subramanian, op. cit.: 521; Hox et al.. 2018).
 Preliminary results

Model
predictors
Weekdays Weekend /holiday
Beta t Beta t
Constant (.726) ** (6.316) (.333) ** (3.515)
Education -.085 -1.136 -.089 -1.209
Age -.299**
-4.015 -.008 -.105
Multidimensional Index of
Local Deprivation /MILD2014
-.098*
-2.556 -.046 -1.209
Education * MILD2014
.142*
1.900 .145*
1.97
Age * MILD2014
.115 1.497 -.029 -.383
Urban (rural omitted) .011 1.280 .016*
1.966
F (6. 22698) = 174.860** F (6. 24 068) =
23.515**
Table 2. Multilevel regression of individual well-being - U-index – on
individual and commune/gmina level variables with cross-level interaction
terms.
**,* significant at p < 0.01 and p < 0.05. respectively.
33

Models type II - structural approach: causal mediating
mechanisms:
local deprivation as a factor modifying effect of ‘composition’
(socio-demographic. etc.) on well-being according to U-
index.
• Income - indpendent var. / ‘treatment’
• MILD – mediating factor
34

Hypothesis : The level of deprivation of a commune (gmina) affects the influence of
the level of residents‘ material status (income) on their subjective well-being
Y - U-index (individual well-being)
Z – source of influence: HH income (average in a commune/gmina)
M - mediator: level of local deprivation /MILD_2014
M= γ0+ aZ + εM (9)
Y = β0 + bM + cZ + εM (10)
Substituting for M (6 in 7)  reduced-form model:
Y = (…) = β’0 + c’ Z + ε’Y
Estimation of diffrences between coefficients of ifluence c’ – c (with local
deprivation/MILD as a mediator) allows to assess indirect effect (of MILD) in
estimating influence of HH income on individual well-being (U-index).
 structural modeling (e.g. G. Hong. 2015):
35

Model / predictors Standardized Coefficients Difference
(c'- c)Beta t-statistics
Dependent Var: U-index for all activities
M I: ILD_economy
Monthly income/ Mi (c’)
ILD_economy on Mi (c)
-.054
.072 *
-.358 **
-1.565
2.070
-11.807
0.304
M II: ILD_social assistance
Monthly income /Mi (c’)
ILD_soc asst. on Mi (c)
-.091 **
-.111 **
-.104 **
-2.824
-3.439
-3.214
0.013
M III: ILD_labor market
Monthly income /Mi
ILD_labor market on Mi (c)
-.089 **
-.061 *
-.154 **
-2.725
-1.850
-4.802
0.065
M IV: : ILD_health
Monthly income /Mi (c’)
ILD_health on Mi (c)
.054
-.070 *
-.178 **
1.638
-2.137
-5.583
0.108
Table 3. Structural (causal) modeling: quality of living environment
-Index of Local Deprivation by domain – as a moderating factor in assessing
influence of respondents' income on subjective well-being
36

Spatial aspects of between-level
relationships
 Two-step spatial analysis:
(1) checking a tendency to clustering among ‘spatial units’
(communes/gminas) with respect to selected measures – subjective
and objective – using Moran’ I (global):
; i ≠ j (6)
where:
xi. xj- values of a measure at each location;
W is the spatial weights matrix.
37

Spatial aspects … cont.
(2) Estimation of the spatial regression model parameters: (notation for
individual observation i):
yi = ρ ∑
n
j=1 Wij yj + ∑
k
r=1 Xirβr + εi (7)
where: yi – the dependent variable for observation i; Xir k – explanatory variables.r = 1.
…. k with associated coefficient βr ; εi is the disturbance term; ρ is parameter of the
strength of the average association between the dependent variable values for
region/observations and the average of them for their neighbours (eg.. LeSage and Pace.
2010. p. 357)
The above specification of the spatial regression model assumes that εi is meant as the
spatially lagged term – versus spatial error formulation - for the dependent variable
(which is correlated with the dependent variable). that is:
εi = ρ Wi.yi + Xi.β + ϵi (8)
These two types of models allow us to examine the impact that one observation has on
other. proximate observations. 38

Table 4. Spatial dependence /spatial regression of SW-B on
commmune’s attributes and compositional characteristics
DIAGNOSTICS FOR HETEROSKEDASTICITY -- RANDOM COEFFICIENTS
TEST DF VALUE PROB
Breusch-Pagan test 7 54.7759 0.00000
DIAGNOSTICS FOR SPATIAL DEPENDENCE -- SPATIAL ERROR DEPENDENCE
TEST DF VALUE PROB
Likelihood Ratio Test 1 36.1346 0.00000
SPATIAL ERROR MODEL - MAXIMUM LIKELIHOOD ESTIMATION
Dependent Variable : U –index Number of Observations: 937; Mean dep var : 0.361195 Number of Variables : 8;
Degrees of Freedom : 929
Lag coeff. (Lambda) : 0.431769 ; R-squared : 0.123681
Variable Coefficient Std.Error z-value Probability
CONSTANT 0.523731 0.042847 12.2233 0.00000
MONTLY INCOME -0.002730 0.001960 -1.40359 0.16044
AGE_avg (%) -0.014313 0.005653 -2.53177 0.01135 *
EDUCATION_HS+ (%) 0.000381 0.000222 1.71849 0.08571 *
NOT WORKING POP. (%) -0.001304 0.000273 -4.77623 0.00000 *
ILD_ECOLOGY 0.000560 0.000462 1.21309 0.22510
ILD_SOCIAL POLICY -0.000415 0.000312 -1.32693 0.18453
SUBSIDIES_pc 1.2323e-005 1.1588e-05 1.06344 0.28758
LAMBDA 0.431769 0.0677941 6.36883 0.00000
39

Table 5. Spatial dependence /spatial regression of SW-B (U-index) on commmune’s
attributes and self-reported satisfaction
DIAGNOSTICS FOR HETEROSKEDASTICITY -- RANDOM COEFFICIENTS
TEST DF VALUE PROB
Breusch-Pagan test 6 54.9862 0.00000
DIAGNOSTICS FOR SPATIAL DEPENDENCE -- SPATIAL ERROR DEPENDENCE
TEST DF VALUE PROB
Likelihood Ratio Test 1 31.9723 0.00000
SPATIAL ERROR MODEL - MAXIMUM LIKELIHOOD ESTIMATION
Dependent Variable : U-index Number of Observations: 852
Mean dependent var 0.361655 Number of Variables: 7 Degrees of Freedom: 845
Lag coeff. (Lambda) : 0.404392; R-squared : 0.107304
Variable Coefficient Std.Error z-value Probability
CONSTANT 0.514258 6.64023e-005 6.45214 0.00000
LOC. DEPR/MILD_14 7.81132e-005 0.001960 1.17636 0.23945
SATISFACTION from:
• FAMILY LIFE_avg 0.035800 0.018318 1.95436 0.05066*
• FINANCIAL SITUAT._avg 0.021283 0.018318 2.0522 0.04015*
• AMOUNT OF TIME-OFF -0.079008 0.016966 -4.65675 0.00000**
• WAY OF LEISURE_avg -0.0159526 0.017004 -0.93814 0.34817
• LIFE (GENERAL)_avg -0.0220795 0.022785 -0.96903 0.33253
LAMBDA 0.404392 0.069845 5.78979 0.00000
40

Autocorrelation of local deprivation: MILD_2014
- (Moran) scatter plot and map (country). Moran I = 0.20
41

Autocorrelation individual SWB/U-index (all activities)
- (Moran) scatter plot and map (country /1036
communes/gminas) . Moran I = 0.09
42

Individual SWB/U-index (all activities)
and the level of commune (gmina) local
deprivation/MILD_2014.
Moran’s I = 0.103
Lag coeff. (Rho) : 0.48

Subjective well-being by U-index (all activities) and the level
of commune local deprivation in social policy.
Moran’s I = 0.33
44

Subjective well-being by U-index and the level of commune
local deprivation in the domain of labor market.
Moran’s I = 0.22
45

Summary and Conclusion
 Working with existing databases. eg. public files of official statistics has its advantages and
disadvantages. Despite inescapable limitations it involves it makes it possible on the one hand to
address issues not solvable in other way (due to cost and time of data collection) while on the other
hand imposes reversed frame of methodological decision proces starting with models which imply
demand for data(‘modelling for measuring’).
 Since data play vital role in measuring well-being following the requirements of modelling cross-level
relationship – between individual and community – an internal/methodological communication chain
needs to be established in the statistical research process. Several clarifications of the research
situation – in terms of types of units and measures involved – are meant to facilitate the choice of
data appropriate for model(s) conceived to be capable of dealing with the problem under study
(focused on the nature of the relationship between community and individual well-being).
• In view of the lack of hierarchical data structure (individual/household data ‘nested’ in community-
level data) an integration procedure was applied to construct multi-source database through
matching records of communes - esp. multidimensional index of local deprivation - with individual
records of subjective well-being measured with the time-use survey data (using U-index. i.e. relative
time in ‘unpleasant state’ while performing activities with negative feeling about them). The merging
key was ‘place’, its geographical coordinates (X.Y) at the NUTS5-level (gmina). ensuring also
contextualization’ of the relevant (thought not always exposed) factors of the individual/subjective
well-being.

• Although presented results of the empirical exploration using different kinds of models can not be
considered robust (data not generated by design) several interesting conclusions can be drawn from the
analysis of the type illustrated here, proving validity of the adopted strategy encompassing integration of
data required by a model considered appropriate for the research problem.
• In general, negative association between subjective well-being (according to U/’unpleasant’-index) and
local deprivation means that it is diminishing (U-index grows) along with the lower level of commune's
local deprivation (see Tab. 2.) [In other words, overall conditions in less developed gminas constitute in
general more favorable environment for individual (subjective) well-being – certain aspects like social
interaction, interpersonal relations might be of importance].
• Individual well-being increases along with greater household income. However, community deprivation
reinforces significantly the subjective well-being effect of individual income. Also, deprivation in several
domains shows negative association with U-index (such as in ‘local social policy’ and ‘local labour market’).
• Local deprivation (‘community well-being’) shows clear tendency to spatial clusters. Similar tendency (but
with lower intensity) appears for spatial association of individual and community well-being. Such
association is much stronger with deprivation in the domain of local labour market and local social
assistance.
Summary and Conclusion

References
Alkire S. 2015. The Capability Approach and Well-Being Measurement for Public Policy. Oxford Poverty
& Human Development Initiative Working Paper No. 94.
Allin P. 2015. Official Statistics on Personal Well-Being: Some Reflections on the Development and Use
of Subjective Well-Being Measures in the UK . Statisitcvs in Transition new series. Vol.16.No.3
Anselin L. Syabri I. Kho Y. 2010. GeoDa: An Introduction to Spatial Data Analysis. Chapt. A.4 . [in]
Fischer M.M.. Getis A. Handbook of Applied Spatial Analysis: Software Tools. Methods and
Applications. Springer.
Cantle T., 2008. Community Cohesion: A New Framework for Race and Diversity. Palgrave Macmillan:
Basingstoke.
[The 2016] CIW National Index Report: How Are CanadiansReally Doing?
https://uwaterloo.ca/canadian-index-well-being/(Nov. 2017)
Clark A. E. Four Decades of the Economics of Happiness: Where Next? Review of Income and Wealth.
Volume 64. Issue 2. https://doi.org/10.1111/roiw.12369
Chavis D. M. Lee. K.S.. Acosta J.D. 2008. The Sense of Community (SCI) Revised: The Reliability and
Validity of the SCI-2. Paper presented at the 2nd International Community Psychology Conference,
Lisboa, Portugal.
Davidson W.B. Cotter P R. 1986. Measurement of Sense of Community Within the Sphere of City.
Volume16. Issue7. Pages 608-619 https://doi.org/10.1111/j.1559-1816.1986.tb01162.x

References
Fischer J. 2009. Subjective Well-Being as Welfare Measure: Concepts and Methodology. OECD.
Paris. Online at https://mpra.ub.uni-muenchen.de/16619/
Hong G. 2015. Causality in a Social World: Moderation. Mediation and Spill-over. Wiley.
Kahneman D., Deaton A., 2010. High income improves evaluation of life but not emotional well-
being. PNAS https://doi.org/10.1073/pnas.1011492107
Kahneman D., & Krueger. A. B., (2006). Developments in the measurement of subjective well-
being. Journal of Economic Perspectives. 20. 3-24
Krueger A. B., Kahneman D., Schkade D.A., Schwartz N., Stone A., 2009. National Time Accounting:
The Currency of Life, in : A. B. Krueger (ed) Measuring Subjective Well-Being of Nations: National
Account of Time Use and Well-Being. University of Chicago Press.
Kim Y.. Ludwigs K. 2017. Measuring Community Well-Being and Individual Well-Being for Public
Policy: The Case of thr Community Well-Being Atlas. in: R. Phillips. C. Wang. (ed.). Handbook Of
Community Well-Being Reseach. Springer.
Land K. C.. Michalos A. C.. 2017. Fifty Years After the Social Indicators Movement: Has the Promise
Been Fulfilled? An Assessment an Agenda for the Future. Social Indicators Research.
doi:10.1007/s11205-017-1571-y.

References
OECD 2015. Better Life Index. OECD Publishing. Paris.
OECD 2013.OECD Guidelines on Measuring Subjective Well-being. OECD Publishing. Paris.
Okrasa W., 2018. The Measurement Aspects of Community Well-Being Research: A Urban-Rural
Spatial Justice Perspective. SCORUS2018 Conference in Warsaw. Warsaw. June 6-8.
Okrasa W., 2017. Community well-being. Spatial Cohesion and Individual well-being – towards a
multilevel spatially integrated framework. [in] W. Okrasa (Ed.) Quality of Life and Spatial
Cohesion: Development and well-being in the Local Context. Cardinal Stefan Wyszynski
University Press. Warsaw.
Okrasa W., Rozkrut D., 2018. The Time Use Data-based Measures of the well-being Effect of
Community Development. Proceedings of the 2018 Federal Committee on Statistical
Methodology (FCSM) Research Conference. [in press].
Pierewan A C., Tampubolon G.. 2014. Spatial dependence multilevel model of well-being across
regions in Europe. Applied Geography Volume 47. February 2014. Pages 168-176
Phillips R.. and Wong C.. 2017. Handbook of Community Well-Being Reserach. Springer
Stauer N.. Marks N.. 2009. Local well-being. Can We Measure it.?
https://youngfoundation.org/wp-content/uploads/2013
Subramanian S.V.. 2010. Multilevel Modeling [in] Fischer M.M.. Getis A.. Handbook of Applied
Spatial Analysis: Software Tools. Methods and Applications. Springer.

Scale construction based on Survey Modules for the Oxford Quality of Life Index and Dashboard
(OXQOL) - see Anand et al. (2010)
1. Material Aspects of Life: job and living environment (α ≈ 0,75)
[For the following statements relating to your work and environment, please indicate your level of
agreement, or otherwise: I have good opportunities to use my talents and skills at work I am
treated with respect by others at work. In general, employment opportunities for people like me
are good. My current employment is secure. I find it difficult to make friendships which last with
people outside of work. I have adequate opportunity for holidays and recreational activities. The
quality of my housing is good given what I need .The quality of local services provided in the
area where I live is good. There is very little pollution from traffic or other sources where I spend
most of my time. There are parks or fields near where I live. 5 point scale – Strongly Disagree to
Strongly Agree]
1. Social and Intellectual Dimensions (α ≈ 0,75)
For the following statements, please indicate your level of agreement, or otherwise. I am free to follow my
own religious beliefs. I am free to engage in cultural activities if I so choose. I am free to plan my life out the
way I want. I feel able to contribute to society in ways that I value. I feel people can be trusted in general. I
feel able to articulate my political beliefs. I am able to respect value and appreciate other people. I feel safe
walking around my local neighbourhood in the day. I find it easy to imagine the situation of other people (ie.
'to put yourself in others' shoes'). My idea of a good life is based on my own judgement I find it difficult to
express feelings of love, grief, longing, gratitude and anger compared with most people my age I am able to
appreciate and value plants, animals and the world of nature 5 point scale – Strongly Disagree to Strongly
51

Scales of wellbeing – cont.
3. Satisfaction with (Key Aspects of) Life (α= 0,78)
[Given your own abilities and the environment in which you live, how satisfied are you with
the following aspects of your life? Your health, your job (if employed), your sleep, your free
time, your family life, your community and public life, personal projects, your dwelling, your
personal income, your personal prospects. 5 point scale – Totally Satisfied to Totally
Unsatisfied]
 Sense of Community /Belonginess and Identyfication (α ≈ 0,75)
[Sense of Community (SCI), Chavis et al., (2008): Community members and I value the
same things. Being a member of this community makes me feel good. People in this
community have similar needs, priorities, and goals. I can recognize most of the members
of this community. Most community members know me. I care about what other community
members think of me. I have influence over what this community is like. If there is a problem
in this community, members can get it solved. It is very important to me to be a part of this
community. I expect to be a part of this community for a long time. Members of this
52

LOCAL DEPRIVATION – DIMENSIONS AND INDICATORS
Factor Analysis – items included in the first factor
The label of the variable
E c o l o g y
1. Plants generating waste – total number [per 1000 people]
2. Emission of dust contamination and gas pollution – total [per 100 hectares of the total area of gmina]
3. Sewage not processed discharged into water or soil, total (dam3) [per 1000 people]
4. Waste generated during the year – total [thousand tons per year], [per 1000 people]
5. Sewage cleaned, discharged – total [dam3/year], [% of total sewage capacity]
6 .Expenditure for public utilities and environmental protection – total [PLN], [per capita]
F i n a n c e
7. Income – total [PLN], [per capita]
8. Expenditure per resident – total [PLN]
9. Total budget expenditure – total asset-related expenditure [PLN], [per capita]
10. Total budget expenditure by budgetary units [PLN], [per capita]
11. Total budget expenditure for materials and services [PLN], [per capita]
12. Income from property tax [PLN], [per capita]
13. Income from asset [PLN], [per capita]
14. Total budget expenditure for salaries [PLN], [per capita]
E c o n o m y
15. Publicly owned enterprises – total [economic units], [per 1000 people]
16. Private sector – number of economic units, firms, in total, [per 1000 people]
17. Stores by sector of ownership, in total [per 1000 people]
18. Private sector – associations and social organizations, [% of private sector units]
19. Public sector – state and local self-government (budgetary) units, in total, [% of public units]
20. Hotel and tourist objects – accommodated [number of people], [per 1000 people]
21. Public sector – commercial units, [% of all public units] 53

I n f r a s t r u c t u r e
22. Expenditure for transport and communication – total [PLN], [per capita]
23. Expenditure for transport and communication as asset-related expenditure, in total [PLN], [per capita]
24. Expenditure for transport and communication – asset-related investment expenditure [PLN], [per capita]
25. Expenditure for transport and communication – public roads and paid motorways [PLN], [per capita]
M u n i c i p a l u t i l i t i e s
26. Dwelling amenities – flush toilet, [% of dwellings]
27. Dwelling amenities – bathroom [% of dwellings]
28. Dwelling amenities – central heating, [% of dwellings]
29. Users of the amenities as proportion of general population – sewer [%]
30. Water supply – population using the water supply network in cities [number of people], [per 1000 people]
31. Electricity in urban households – consumers of electricity at low voltage [% of dwellings]
32. Dwelling amenities – water supply [% of dwellings]
33. Dwelling amenities – gas network [% of dwellings]
34. Gas network – population using gas network [number of people], [per 1000 people]
35. Gas network – gas consumers heating homes with gas [households], [% of dwellings]
36. Water industry – water supply network [km], [per 1000 dwellings]
C u l t u r e
37. Expenditure for culture and conservation of national heritage [PLN], per inhabitant
38. Expenditure for culture and conservation of national heritage – cultural houses and centers, social rooms and clubs [PLN pc],
39. Expenditure for culture and conservation of national heritage – libraries [PLN], [per capita]
40. Libraries – libraries and affiliated units [per 1000 people]
41. Libraries – library stuff [number of people], [per 1000 people]
H o u s i n g
42. Dwelling units delivered, in total – living area [m2
], [per 1000 people]
43. New housing buildings delivered, total – living area [m2
], [per 1000 people]
44. New housing buildings delivered, total – number of buildings, [per 1000 people]
45. Dwelling units delivered, in total – dwellings [per 1000 people]
FA – items included in the first factor (cont.)
54

FA – items included in the first factor (cont.)
S o c i a l w e l f a r e
47. Social welfare expenditure and other needs within the social policy area – total [PLN], [per capita]
48. Social welfare expenditure and other needs within the social policy area – benefits to individuals [PLN], [per capita]
49. Social welfare expenditure and other needs within the social policy area – benefits and in kind assistance, and social security contributions
[PLN], [per capita]
L a b o u r m a r k e t
50. The rate of unemployment, as percentage of the working-age population – total [%]
51. Registered unemployed persons by sex – total [persons], [per 1000 people]
52. Employed persons by sex – total [persons], [per 1000 people]
53. Dependency ratio – persons in retired-age per 100 persons in the working-age
54. Dependency ratio – persons in non-working age per 100 persons in the working-age
E d u c a t i o n
55. Children in kindergarten (kindergartens, kindergartens units in primary schools, teams of kindergarten upbringing and kindergarten points),
[% of children aged 3-6 years]
56. Enrollment Ratio (primary and middle education) gross enrollment ratio – middle schools [%]
57. Expenditure for education and upbringing – vocational schools [PLN], [per 1 child aged 17-19]
58. Day-care centers – children attending during the year (including affiliated units) [persons], [% of children aged 0-3 years]
59. Day-care centers – children (including affiliated units), [% of children aged 0-3 years]
60. Expenditure for education and upbringing – secondary school [PLN], [per 1 child aged 17-19]
61. Expenditure for education and upbringing – kindergartens [PLN], [per 1 child aged 3-6]
62. Expenditure for education and upbringing – middle school [PLN], [per 1 child aged 13-16 ]
H e a l t h
63. Health care institutions – medical practices in urban areas [persons], [per 1000 people]
64. General hospitals – bed in total, [per 1000 people]
65. Health-related expenditure – total [PLN], [per capita]
55

IAOS2018 - Modelling for Improving Measurement: Strategies for Contextualization of Well-Being, W. Okrasa, D. Rozkrut

More Related Content

Similar to IAOS2018 - Modelling for Improving Measurement: Strategies for Contextualization of Well-Being, W. Okrasa, D. Rozkrut (20)

More from StatsCommunications (20)

Recently uploaded (20)

IAOS2018 - Modelling for Improving Measurement: Strategies for Contextualization of Well-Being, W. Okrasa, D. Rozkrut