Summer School on Capability and Multidimensional Poverty

1
Summer School on Capability and Multidimensional
Poverty

• 28 August-9 September, 2008
• New Delhi, India

OPHIOxford Poverty Human Development
InitiativeUniversity of Oxford www.ophi.org.uk
HDCP-IRC The Human Development, Capability and
International Research Centre Instituto
Universitario di Studi Superiori www.iusspavia.it
2
Capability production functions

• Enrica Chiappero-Martinetti
• Emma Samman
• 5th September, 2008

3
Capability Approach

• Instrumental role of monetary resources in
  well-being
• Clear distinction between means and ends
• Other factors other than lowness of income affect
  capability deprivation
• The instrumental role between low income and low
  capability is parametrically variable between
  different communities, households and
  individuals (Nussbaum 1999)

4
Methodological requirements

• Many dimensions
• Human diversity and plurality of context
• Correlation among variables, dimensions and
  context of analysis
• Aggregation across attributes, across units of
  analysis (i.e. individuals, households, groups)
  for determining an overall index

5
Most common measurement strategies

• For dealing with issues (i) to (iii)
• bringing multidimensionality back to a univariate
  approach
• making use of multivariate techniques
• For aggregation and comparison purposes (iv)
• multidimensional decomposable poverty indexes
• extension of dominance stochastic conditions
  within a multidimensional framework

6
Linking means and ends WB production process

• To conceptualize the well-being process as a sort
  of
• production function transforming inputs
  (resources, public and private goods and
  commodities) into outputs (achieved
  functionings). Sen 1985,p.11
• xi vector of commodities of individual i,
• b vector of functionings,
• c function that maps commodities into
  characteristics
• fi personal utilization or conversion function

7

• Kuklys (2005)
• groups conversion factors in three different
  categories (non-monetary constraints) personal
  (zi), social (zs) and environmental factors (ze)
• NB distinction between
• conversion function similarly to a production
  function, transforms inputs (resources, public
  and private goods and commodities) into outputs
  (achieved functionings)
• conversion factors internal factors, external,
  social and environmental circumstances
• conversion rate the ability to transform means
  into achievement (technology)

8
Some recent literature

• Ruggeri Laderchi (2008) RL
• ECM,Salardi (2007,2008) ECM/PS
• Ramos, Silber (2005) - RS
• Ramos (2008) -R
• Binder, Broekel (2008) - BB

9
1. RL (2008)

• A simple representation on how resources are
  mapped into functionings
• bi fi (c (xi (yh, ti, to, dh, zh, l) ?i
• yh household resources
• ti set of individual characteristics
• to set of characteristics of other household
  members
• dh set of demographic characteristics of the
  household
• zh set of public goods available to the household
• l a location variable capturing other
  area-specific influences
• xi basket of goods and services whose
  characteristics (c) are combined according an
  utilisation function (f)
• ?i error term

10
Ruggeri Laderchi (2008) II

• A crucial level bline such that
• bi 1 if bi gt bline and bi 0 if bi bline
• Estimation of the relationship between resources
  and basic capabilities with limited dependent
  variable techniques
• Testing for the existence, size and significance
  of differences between individuals with ? ti , to
  and l
• Comparison between capability based and
  consumption based approaches
• Some relevant information for policy purposes
• Empirical analysis
• Household survey, Peru, 1994 (LSMS-type)
• Health (self-reported morbidity, for the overall
  population) and education (functional literacy4
  years schooling for children 12-15)
• Three models estimated
• Probit (sick, non-sick)
• Tobit (n of days of illness for those who are
  sick)
• Fixed effects associated with different
  departments

11
Ruggeri Laderchi (2008) III

• Main results
• 1. Morbidity
• Monetary resources strongly and positively
  correlated with both health indicators
• Age affects duration but not morbidity
• Years of schooling play a significant role and
  the effect persist even controlling for
  individuals and parental background ? education
  improves the technology with which resources are
  transformed into health
• Ethnicity and female head of hh do not play
  independent roles in determining whether people
  are sick or not
• access and quality of health (time to reach and
  to be attended in health facility) surprisingly
  do not play any significant role
• Effects of education and public water on
  morbidity (tab.6) both edu and water provision
  have greatest effect for the poorest
• Policy implication investing in the edu of the
  poorest would provide returns in terms of
  self-reported morbidity almost three times as
  high as those for the richest and community
  level access to public water reduces

12

13
Ruggeri Laderchi (2008) IV

• 2. Childrens edu achievement
• Probit model
• Hh income strong and significant determinants of
  childrens schooling achievement
• Gender no role stronger effect of father years
  of schooling, lower for the mother ethnicity not
  significant nor the size of hh while it is the hh
  composition
• access and quality of schooling not correlated
  with low achievements little difference
  rural/urban
• Effectiveness of resources and parental edu by
  decile (tab 10)

14
(No Transcript)
15

• Sum up
• individual, household and community
  characteristics, in ? degrees and to varying
  degrees do affect the way resources are
  translated into individual achievement (a
  monetary indicator is not sufficient or
  appropriate)

16
2. ECM/PS (2007, 2008)

• Empirical evidence on conversion
  factors/conversion rates
• Disaggregated results by subgroups of population
• Regression analysis controlled by a range of
  socio-demographic characteristics (Anand, van
  Hees, 2006 Zaidi, Burchardt, 2005 Kuklys, 2005,
  Ruggeri-Laderchi 2008)
• Equivalence scales literature. Zaidi and
  Burchardt (2005), Lelli (2005),
• Latent variable models (factor analysis,
  structural equation models and multiple
  indicators and multiple causes model). Kuklys
  (2005), Addabbo et al (2004) and Krishnakumar
  (2007)
• In most cases evidence of the (ambiguous)
  correlation between income and functionings,
  impact of personal and socio-environmental
  features, measure of cost differences related to
  individual or household conditions.
• In none of these cases, an estimation of the rate
  of conversions is produced

17

18
ECM/PS (2007, 2008)

• Empirical Analysis (2008). Data both macro-data
  and HH multiporpose survey (Italy, 2006)
• Inputs
• A composite index for public resources
• Public services for health (Indicatori socio
  sanitari - ISTAT, 2006)
• Public services for environmental (Ecosistema
  Urbano - Ambiente Italia, 2008)
• Public services for security and public order
  (Compendio Statistico Eventi Criminosi
  Ministero dellInterno, 2003)
• Private resources
• Problems no income information in our dataset
  possible reversal causality between income and
  the selected achievements income instability and
  fluctuations
• Solution to build up a Wealth index via the
  principal component analysis by using variables
  on housing characteristics, durables and
  affordability

19

• output
• Three achievements
• 1. being healthy a subjective indicator for
  health status (very good, good, fair, bad, very
  bad)
• 2. living in a healthy and safe environment
  a subjective indicator of satisfaction for the
  environmental quality, e.g pollution, parking
  public transport, traffic, noise, etc (bad, not
  much, fair, good)
• 3. feeling secure a subjective indicator of
  satisfaction of public security (bad, not much,
  fair, good)

20

• Conversion rates estimated for six subgroup of
  population (by gender and by age young, adult,
  elderly) and controlling by individual
  characteristics gender, age, levels of attained
  education, marital status, occupational status
  and external/geographical characteristics
  (macro-area, regions)
• Econometric strategy an ordered probit model
  (due to the categorical nature of the dependent
  variables) for each functioning, for the overall
  population and for subgroups of population how
  these groups differ in converting resources into
  functioning (more efficient or more critical
  groups)

21
Bij the achieved functioning Wij, the wealth
index, proxy for private resources Gj, the
variable of public services, proxy for public
resources for the individual i, i1n,
living in the geographical area j, j1m. Cijk,
the k internal conversion factors with k1l,
that varies across individuals or the
geographical areas where they live.
22

• Main results
• Pooled estimation positive and statistically
  significant impact of
• both private and public resources for health
• Private resources only for environment
• Public resources only for security
• individual characteristics provide information in
  line with one could expect
• H worst health reported conditions for F,
  elderly, unemployed
• Elower perceived achievement for F, elderly and
  better educated people
• S similar pattern as before

23
Empirical findings (I) Being healthyPooled
estimations
24

• Subgroups estimations statistically significant
  estimations only for adult and elder female and
  male
• F are more efficient to convert public resources
  while M convert better private resources
• Elderly are less able to convert public as well
  as private resources

25
Empirical findings (II) Living in a safe and
healthy environmentPooled estimations
26
Ramos and Silber 2005

• - EFFICIENCY ANALYSIS
• They use distance functions to estimate the
  standard of living (inputs or resources), the
  quality of life (outputs or capabilities) and the
  efficiency in transforming resources into
  functionings
• They compare four different wb approaches,
  measuring
• 1) an index of achievements reached by the
  individual in each dimensions
• 2) An overall index of well-being/ human
  development.

27
ISOQUANT all possible combinations of inputs (x)
that allow to produce a given output L(y) The
distance between an inefficient and an efficient
amount of inputs lying onto the isoquant for the
corresponding output level is measured by an
(input) distance function. Din (y,x) OA/OB gt1
(1 if x is on L(y) how much input vector may be
proportionally contracted given an output vector
28

• Production possibility frontier PPF(x) various
  combinations of output (y1,y2) that can be
  produced using an input vector x
• The distance between any output vector inside PPF
  and an efficient combination of output lying onto
  the PPF is measured by an output distance
  function. Dout (y,x) OA/OB lt1 (1 if x is on
  PPF (x) extent to which the output vector may be
  proportionally expanded given an input vector

29
ESTIMATING ACHIEVEMENT FOR GIVEN DIMENSION OF
WELLBEING Let x (x1,,xN) ? RN denote the
vector of the n elements that are considered as
inputs in the production of achievement levels in
the various domains and u(u1, ,uM) ? RM
denote the vector of achievement levels. Then
an individuals endowment of inputs and levels of
achievement are denoted by the pair (xi, ui), i
1, I individuals
30
They estimate a standard of achievement SA using
an input quantity index SA(u, xs, xt) Din(u,
xs)/Din(u, xt) where xs and xt are two different
input vectors and Din is an input distance
function. Similarly, they estimate an output
distance function WB(x, us, ut) Dout(x,
us)/Dout(x, ut)
where us and ut are two achievement vectors and
x is an input vector
31
Problem Production functions compare two inputs
based on their relative prices. But how do we
value e.g., education versus health in our
capability production function? gt reference
point Level of achievements They assume each
individual has the same vector e of achievement
e.g one unit of achievement for each of M
dimensions. L(e) is called reference set and
bounds the input vectors from below. Individuals
with input vectors onto L(e) share the lowest
level of achievement (with an index value 1)
whereas individuals with large input vectors will
have higher levels of achievement (index values gt
1) Overall wb As above assuming that each
individual is endowed with one unit of each input
PPF (e). If individual has a vector of
achievements that places her on PPF(e) gt max WB
(1) while if it is lower the index values will
be lt1.
32

• Empirical analysis
• British Household Panel Survey, 1997
• Input variables (long and broader set of subj/obj
  var)
• Four lists of functionings
• Allardt 1993 Having (material resources, working
  conditions, health and education), Loving
  (contacts w/ local community, family, friends
  etc) and Being (self determination, leisure,
  meaningful work etc)
• Cummin, 1996 material wellbeing, health,
  productivity, intimacy/friendship, safety, the
  community and emotional wellbeing
• Narayan, 2000 material wb, bodily wb, social
  wb, security, freedom of choice action,
  psychological wb
• Sen 1985 list of indicators they feel
  compatible with CA

33
(No Transcript)
34
(No Transcript)
35

• Inequality and poverty analysis in achievements
  and overall wb (CA higher inequality for the
  right environment and health vs ability to
  undertake mental or physical tasks or to
  socialize).
• WB domains weakly correlated with income
  Correlation between dimensions is also quite low
  whatever approach is used
• - They conclude It might therefore not matter
  too much which of these approaches we select
  given that we make sure that aspects of wellbeing
  that have not much to do with resources, such as
  emotional wellbeing, friendship etc. are taken
  into account.

36
Binder, Broekel, 2008

• They also refer to efficiency analysis and use
  BHPS
• Three functioning
• being happy (mental wb in the General Health
  Quest)
• being educated (highest level of edu)
• being healthy (subj reported)
• I as a proxy for commodity vector
• Two steps
• identify the best practice individuals in terms
  of achievements (no other individuals with lower
  or equal I shows a higher level of achiev.)
  gtreference group. Individuals with I and lower
  achievm is inefficient (euclidean distance
  between obs and efficiency frontier).

37

• Problem outliers
• Solution robust non parametric frontier
  techniques. Instead of comparing the efficiency
  of each person with respect the efficiency of all
  other obs, the comparison is made with a random
  subsample of observations.
• 2nd step influence of individual conversion
  factors (gender, age, job and marital status) on
  the efficiency scores.

38
To conclude

• At what extent the production theory/production
  function approach is coherent with the
  theoretical foundation of the CA
• a) Common problem to summarize or collapse a
  large amount of information into only one
  dimension
• b) Production economics as well-consolidated
  theory and tools
• c) Similarities with Gary Beckers household
  production function
• But see Sens (1985) critique about the
  appropriateness of seeing functionings as
  fundamental commodities produces by the
  household
• the former are not just a simple commodity but a
  state of existence of a person and
• many of the functionings are produced outside the
  household (e.g. through public policy)
• rather restrictive assumptions regarding the
  existence of shadow prices and implicit markets
•