Presentation at the Third Conference of International Consortium for Social Well-Being Studies, held at Plataran Hotel & Convention Center, Magelang, Indonesia, March 4, 2018
3. Introduction
• How can we measure the well‐being of a whole society?
• The popular methods to do so are:
– (i) To utilize objective indices about a society, such as indices for
economic production and equality (e.g. GDP, GNI, Gini)
– (ii) To utilize simple aggregated values of subjective evaluations
by individuals, who live in the society, of themselves, such as the
percentage of those who feel happy
• Possible Drawbacks
– Method (i) can be criticized for not considering how ordinary
people feel.
– Method (ii) may also be criticized because there can be people
who tend to feel satisfied even in very difficult life situations,
and such people can more likely be ignored by policy makers, if
they only base their policy on simple subjective indices.
3
5. Introduction
• Our method consists of the following steps:
– (1) Estimating Individual Evaluation Function
• Based on real data
• We estimate the individual evaluation criteria with which
each of ordinary people evaluates a society
• Obtain the criteria with which an individual "subjectively"
evaluates a society using "objective" information.
– (2) Aggregating IEF into Collective Preference Order
• Aggregate the individual evaluations into a collective
preference order with respect to multiple statuses of a
society
• Conducted with a particular rule, i.e. in a transparent and
well‐defined way. In this sense, this aggregation is
"objectively" done.
5
8. Concepts
• Figure 1:
the idea of
Individual
Evaluation
Function, IEF
8
Figure 1. Individual Evaluation Function on Status of Society
Status A
Gini=0.30
Growth=0%
Status B
Gini=0.40
Growth=2%
Individual Evaluation Function
(IEF) represents individuals'
subjective evaluation on status
of society
Status of society is
described by
Objective Indices
11. Step 1: Estimating IEF
• Model:
– Each person has IEF, and one prefers the policy which
maximizes the evaluation under perceived restriction.
– Based on this model, we can estimate the parameters of
IEF for different individuals using real data.
– IEF is evaluated with the equality, EQ, and the economic
growth, GR, and also with parameters of normative criteria
– After knowing the estimated set of parameters for each
individual, we can know how each person evaluates a real
or fictional status of a society, using the IEF formula, with
the indices for equality and growth.
11
GREQIEF iiii )1()( 2
12. Step 1: Estimating IEF
• Data
– JHPS: Japan’s nationwide panel survey
• not SWB data
– Use responses in 2011 and 2012
• Item: asked concrete amounts of money for desired
redistribution in “a fictional society” and the perceived
external effect on economic growth
– Using this set of questionnaire items, we can estimate the
parameters with which people evaluates “objective”
indices on a whole society.
12
21. Step 2: Showing CPO
21
Figure 4-1. How Statuses are Ranked (1)
Data Source: JHPS2011-2012; World Bank; Solt (2009; 2016)
Note: Shown are Japan's Gini coefficients (post-transfer Gini) and growth rates
(annual growth in real GNI per capita) for respective years. In the square
brackets with "R" letter are the rank values of the status among ten periods
compared.
2005 [R3]
2006 [R6]
2007 [R5]
2008 [R7]
2009 [R10]
2010 [R4]
2011 [R9]
2012 [R8]
2013 [R2]
2014 [R1]
‐6%
‐4%
‐2%
0%
2%
4%
6%
0.298 0.299 0.300 0.301 0.302 0.303 0.304 0.305
Growth Rate
Gini
In this area, a status with
more equality is more
preferred collectively
22. Step 2: Showing CPO
• Outside the
area of
ordinary
growth, the
pattern turns
different.
22
Figure 4-2. How Statuses are Ranked (2)
Data Source: JHPS2011-2012; World Bank; Solt (2009; 2016)
Note: Shown are Japan's Gini coefficients (post-transfer Gini) and growth rates
(annual growth in real GNI per capita) for respective years. In the square
brackets with "R" letter are the rank values of the status among ten periods
compared.
2005 [R3]
2006 [R6]
2007 [R5]
2008 [R7]
2009 [R10]
2010 [R4]
2011 [R9]
2012 [R8]
2013 [R2]
2014 [R1]
‐6%
‐4%
‐2%
0%
2%
4%
6%
0.298 0.299 0.300 0.301 0.302 0.303 0.304 0.305
Growth Rate
Gini
The status of 2010 beats
that of 2007 etc. because of
its exceptionally higher
growth rate, although 2010
is more unequal
23. Step 2: Showing CPO
• Outside the
area of
ordinary
growth, the
pattern turns
different.
23
Figure 4-3. How Statuses are Ranked (3)
Data Source: JHPS2011-2012; World Bank; Solt (2009; 2016)
Note: Shown are Japan's Gini coefficients (post-transfer Gini) and growth rates
(annual growth in real GNI per capita) for respective years. In the square
brackets with "R" letter are the rank values of the status among ten periods
compared.
2005 [R3]
2006 [R6]
2007 [R5]
2008 [R7]
2009 [R10]
2010 [R4]
2011 [R9]
2012 [R8]
2013 [R2]
2014 [R1]
‐6%
‐4%
‐2%
0%
2%
4%
6%
0.298 0.299 0.300 0.301 0.302 0.303 0.304 0.305
Growth Rate
Gini
The statuses of 2008 and
2009 are beaten by more
unequal statuses because
of exceptionally low growth
rates
33. 33
GR
EQ
Panel (e)
Figure A1. Concepts: Contour of IEF Values, Restriction Line, and Optimal Point
GR
EQ
Panel (b)
GR
EQ
Panel (a)
Restriction Line (θ < 0)
Status Quo
(EQ0,GR0)
Contour of IEF Values
Evaluated High
Evaluated Medium
Evaluated Low
Optimal Point (θ < 0)
GR
EQ
Panel (c)
GR
EQ
Panel (d)
Optimal Point (θ = 0)
Restriction Line (θ = 0)
Optimal Point (θ > 0)
Restriction Line (θ > 0)
34. Method and Data
• We are to estimate how individual characteristics affect the
parameters in the theoretical model
• Estimation method
– Let us assume the responses of EQ and GR, which optimize IEF,
for each individual
– From the above formulas, Theta (θ) is directly estimated
if EQi ≠ EQ0
– We associate Tau (τ) and Theta (θ) with individual characteristic
variable vector, zi, and unobserved disturbances
– Using aforementioned formula, EQ can be written as follow
where
34
)/()( 00 EQEQGRGR iii
ii , τi βz ii , θi βz
iiiEQ , τi βz
2/)1( )21/(1
36. Method and Data
• Estimation method
– Diagram
36
Figure A2. Diagram for Estimated Equations
Explained Variable EQ
Policy Preference
"How much redistribution
should Gov't conduct?"
Parameter Theta (θ)
Perceived Facts
"Does redistribution improve
economic growth?"
Parameter Tau (τ)
Normative Criteria for Equality
"What is the desirable equality,
other factors being the same?"
Explanatory Variables
Individual Characteristics
e.g. Higher Education
38. Method and Data
• Data
– JHPS: Japan’s nationwide panel survey
• not SWB data
– Use responses in 2011 and 2012
• Item: asked concrete amounts of money for desired
redistribution in “a fictional society” and the perceived
external effect on economic growth
– The information on these is utilized to construct each
respondent’s optimal EQ and GR values
38
41. Method and Data
• Variables of main interest: (answered = optimal) EQ and GR
– EQ is based on Gini coefficient after redistribution
• To include information of assured minimum income and to
avoid bias, a continuous income distribution is used; Gini is
calculated from the distribution, by a Monte Carlo method
• EQ is obtained by monotonically decreasing transformation
of the resultant Gini coefficient
– GR is a value which corresponds to the expression in a natural
language, like “worsen dramatically”
• We conducted another small survey to assign each value, i.e.
a concrete amount, for each natural‐language expression of
“growth in growth”
• Other variables: gender, age, marital status, higher education, white
collar job, household income
41
42. 42
Table A1. Estimation Results
Coef. (S.E.) Coef. (S.E.) Coef. (S.E.)
0.013 *
(0.005)
Male
Age/100 0.002 (0.048) 0.522 *
(0.244) -0.005 (0.048)
Married -0.006 (0.017) 0.105 (0.075) -0.007 (0.017)
Higher Education 0.031 *
(0.013) 0.144 **
(0.054) 0.029 *
(0.013)
Female
Age/100 0.080 (0.054) 0.009 (0.335) 0.080 (0.054)
Married -0.008 (0.018) 0.269 **
(0.093) -0.012 (0.018)
Higher Education 0.052 **
(0.019) 0.162 *
(0.074) 0.050 **
(0.019)
Source : JHPS2011-2012
N of Obs. 3084 3084 3084
Note: +:p<0.10, *:p<0.05, **:p<0.01
The value of θ is multiplied by 1000 so that the coefficients are decently displayed.
θ
(Perceived Improvement of
Growth by Unit Equalization)
--- ---
For both genders, White Collar and Household Income (Log) are included but turn out to
be insignificant. Constant, Male Dummy, and Year Dummy (2012) are also included.
R
2
0.011 0.011 0.016
Explanatory Variables
Model 1 Model 2 Model 3
Dep. Var. Dep. Var. Dep. Var.
EQ θ EQ