Structure of Presentation
Setting the context for measuring multi-dimensional
reporting district level socio economic data using household sample surveys has led to an interest in exploring avenues for analyzing multidimensionality of poverty and informing public policy. (a case from Punjab province) steep rise in food, energy and fuel costs, has pushed the government towards using a uni-dimensional (composite) poverty measure using basic needs based capabilities and functionings for targeting beneficiaries of a cash transfer programme as a counter-cyclical intervention. (the case of Benazir Income Support Programme (BISP) poverty scorecard)
categorical). Captures dimensions and evolutions. Establishes causality for capability poor. Minimize Type I and II errors for targeting. (choice of indicators/ union vs. intersection) Direct versus indirect approaches. (Sampling vs. Counting) Setting poverty cut-offs/thresholds. Inter-temporal and cross-sectional comparability.
(aligned to the Multi-indicator Cluster Survey MICS) 23 indicators representing economic, social and housing concerns included. Factor Analysis (FA) technique used to cluster covariant independent variables and to assign weights by degree of variance/dispersion. Overall Score assigned to household and Cluster Analysis used to categorize into poor and non-poor (This step not done in the Punjab exercise).
Example of an Indicator
Northern
Southern Western
29.00
30.00 31.00
39.00
40.00 41.00
49.00
50.00 51.00
59.00
60.00 61.00
69.00
70.00 71.00
79.00
80.00 81.00
For analysis similar tabulations are possible: by district inter temporal by different regional groupings by additional dimensions with quality indicators e.g. Quality of Governance
Poverty Scoring
Indirect approach which is simple, quick and
inexpensive viz. a sample based direct approach. Verifiable indicators are chosen with strong correlation to poverty and can be replaced over time. Poverty scoring can estimate:
The poverty likelihood Poverty rate of a group of households
above the poverty line) Leakage (share incorrectly predicted to be below the poverty line) Exclusion (share correctly predicted to be above the poverty line)
Likelihood Estimates
life indicators are difficult to measure. Assumption of household consumption expenditure per adult equivalent as the basic welfare measure (dependant variable), still in use. Limitations around indicator selection and questionnaire design lead to under-coverage and leakages of beneficiaries.
treating continuous data with a normal distribution. Application to ordinal and discrete data could be problematic. Constructing a single score assumes substitutability across dimensions. This obscures the nature of deprivation faced by a HH. Assigning cardinal scores to categorical/ordinal data is technically problematic.
Conclusion
In order to create an accurate measure for estimating
multidimensional poverty, techniques need to be adopted which allow capturing the qualitative aspects of living standards usually captured by ordinal data.
Alkire and Foster (2007) have proposed one such
methodology.