DAD A software for Distributive Analysis / Analyse Distributive By Abdelkrim Araar and Jean-Yves Duclos

1
DADA software for Distributive Analysis /
Analyse DistributiveByAbdelkrim Araarand
Jean-Yves Duclos
2
What is Distributive Analysis?

• Distributive analysis is concerned with the
  distribution and redistribution of well-being,
  usually captured by living standards at the
  household level.
• The distribution of living standards depends
  dynamically on a number of factors, such as
• Average living standards at the level of the
  population
• Living standards relative to the mean
• The structure of the economy and the
  distributional channels of the richness.
• The economic policies in place (redistribution
  policies)
• Economic shocks

3
What is Distributive Analysis?

• Main topics linked to distribution and
  redistribution
• Absolute relative poverty
• Absolute relative inequality
• Polarisation
• Vertical horizontal equity
• Redistribution
• etc..

4
What is Distributive Analysis?

• Example of some relatively recent economic shocks
  in developing countries
• Economic transition from planned to market
  economies
• Application of macro adjustment programs
• Trade liberalisation
• Globalisation
• These shocks can have a significant impact on the
  distribution of living standards at different
  levels (regions, countries, within households).

5
Positioning DAD in distributive analysis

• The main features of the software can be
  summarized as follows
• Free!
• User friendly no need for programming
• Estimates easily a number distributive indices
  and curves that are extensively used in the
  literature about the distributive analysis.
• Estimates accurately the sampling distribution of
  such indices and curves
• by taking into account the sampling design of
  household surveys
• by means of analytical and numerical procedures
• Provides tools for testing the robustness of
  comparisons
• Insists on the power of graphs to provide
  informative pictures of the distribution of
  living standards

6
Basic descriptive tools

• Estimation of means, quantiles, variances
• Non parametric estimation of
• density
• joint density
• non parametric regression between two variables
• regression slopes
• Scatter graphs
• Important and flexible graphical abilities

7
Poverty decomposition

• Static decomposition
• Population subgroups
• FGT index - analytical approach
• Income components
• FGT Index - Shapley approach
• Dynamic decomposition
• Growth and redistribution
• FGT index analytical Shapley approaches
• Sectoral decomposition
• FGT index analytical Shapley approaches
• Transient and chronic
• FGT index analytical approach
• EDE index - analytical approach
• Absolute transition matrix - analytical approach

8
Inequality decomposition

• Static decomposition
• Population subgroups
• S-Gini index - analytical Shapley approaches
• Generalised entropy index - analytical approach
• Income components
• S-Gini Index - analytical Shapley approaches
• Coefficient of variation index analytical
  approach
• Dynamic decomposition
• Difference population subgroups
• S-Gini index- analytical Shapley approaches
• Difference income components
• S-Gini index- analytical Shapley approaches
• Social welfare
• Atkinson index analytical approach

9
Simulations and policy applications

• Impacts of income-component growth on Inequality,
  poverty and social welfare
• Impact of marginal price changes on poverty,
  social welfare and inequality
• Impact of demographic changes on poverty
• Impact of sectoral changes on poverty
• Impact of lump-sum targeting on poverty
• Impact of inequality-neutral targeting on poverty

10
Simulations and policy applications

• Gini income-component elasticity
• Growth elasticity of poverty
• Impact of marginal tax reforms on poverty and
  inequality
• Impact of reforms to poverty alleviation
  programmes, by targeting/allocation effects

11
Estimation of curves for descriptive and
normative purposes

• Lorenz generalized Lorenz curves
• Concentration generalised concentration curves
• Quantile and normalised quantile curves
• Poverty gap cumulative poverty gap (CPG) curves
• FGT curves
• Pro-poor curves
• Bi-polarisation curves
• Deprivation curves
• Consumption dominance (CD) and normalised CD
  curves

12
Checking the robustness of poverty, social
welfare, inequality and equity comparisons

• Estimation of stochastic dominance curves for
• poverty
• social welfare
• inequality (normalised stochastic dominance)
• relative poverty
• indirect tax reforms
• Efficient targeting reforms
• Estimation of  critical  poverty lines for
  absolute and relative poverty
• Estimation of crossing points for Lorenz, CPG and
  concentration curves

13
Estimating sampling distributions

• Data from sample surveys usually display four
  important characteristics
• they are stratified
• they are clustered
• they come with sampling weights (SW), also called
  inverse probability weights
• sample observations provide aggregate information
  (such as household expenditures) on a number of
  statistical units (such as individuals)

14

• Simple Random Sampling

15
Usual sampling procedures

• A country is first divided into geographical or
  administrative zones and areas, called strata.
• Each zone or area thus represents a strata.
• The first random selection takes place within the
  Primary Sampling Units (denoted as PSUs) of each
  stratum.
• Within each stratum, a number of PSUs are
  randomly selected. This random selection of PSUs
  provides clusters of information.
• PSUs are often provinces, departments, villages,
  etc. Within each PSU, there may then be other
  levels of random selection.

16
Usual sampling procedures

• For instance, within each province, a number of
  villages may be randomly selected, and within
  every selected village, a number of households
  may be randomly selected.
• The final sample observations constitute the Last
  Sampling Units (LSUs).
• Each sample observation may then provide
  aggregate information (such as household
  expenditures) on all individuals or agents found
  within that LSU. These individuals or agents are
  not selected information on all on them appears
  in the sample.
• They therefore do not represent the LSUs in
  statistical terminology.

17
Sampling Design with two levels of random
selection
18
Sampling design and statistical significance
PSUs
19
Example
   
     
20
Initialising the sampling design

• From the main menu one chooses the item Edit-gt
  Set Sample Design.
• Indicate the variables to set the sample design
  and confirm your choice by clicking on the button
  OK.

21
Performing statistical inference

• Estimating confidence intervals and p-values
• Estimations are included directly for FGT,
  S-Gini and Atkinson indices
• Can be computed via the Confidence interval
  application in DAD
• Testing hypothesis
• Can be performed directly for FGT, S-Gini and
  Atkinson indices
• Can be computed via the Confidence interval
  application in DAD

22
DAD and DATA files

• Shows two sheets to load simultaneously two data
  bases
• Can read ASCII files safely through a data wizard
  
• Can support copy/paste to and from sheets of the
  most common software (Excel, Stata,...)
• Offers its own ASCII format for saving data
• Can edit variable information and content
• Can add or delete observations
• Can generate other variables

23
DAD Graphs

• Flexible Graph Options
• For example, one can change easily the
• main title, title of axis and legends
• graph size
• template choice
• color, width and style of curves
• Saving DADs graphs
• One can save DADs graphs in
• DAD Graph Format .dgf
• JPEG, GIF, BMP
• One can also save curves coordinates in ASCII
  format
• Editing curves coordinates in a new data sheet

24
How to learn to use DAD?

• The book entitled POVERTY AND EQUITY
  MEASUREMENT, POLICY AND ESTIMATION WITH DAD
  covers most of the measurement theory implemented
  in DAD. The book is also a comprehensive
  reference for intermediate and advanced study in
  distributive analysis.
• DADs user manual provides tools for fast
  learning of DAD and can lead to rapid use of any
  of DADs applications.
• Exercises Technical notes were written to
  consolidate the learning of DAD.
• Training sessions are regularly organised to
  teach distributive analysis and the use of DAD
  and other software.

25
DADs data files

• With DAD, micro data from household surveys are
  typically required.
• A database used in DAD is then a matrix (a number
  of columns) whose number of lines is the number
  of observations
• DAD can load simultaneously two databases.
• The maximum number of variables for each DAD file
  is currently 20.

26
DADs spreadsheet
27
The structure of a data file

• I- Sampling design
• Strata Specifies the name of the variable (in an
  integer format) that contains the Stratum
  identifiers.
• PSU Specifies the name of the variable (in an
  integer format) that contains the identifiers for
  the Primary Sampling Units.
• SAMPLING WEIGHT Specifies the name of the
  Sampling Weights variable.
• Finite Correction Gives the Finite Population
  Correction variable that is needed when the
  number of PSU is small and sampling was one
  without replacement.

28
The structure of a data file

• II- Basic distributive variables
• VARIABLE OF INTEREST. This is the variable that
  usually captures living standards. It can be for
  the entire household or for individuals (e.g.,
  per capita or per equivalent adult expenditure).
• SIZE VARIABLE. This refers to the ethical or
  physical size of the sampling observation
• GROUP VARIABLE To perform computations at the
  group level (integer variable ex. Rural (1)
  Urban (2) )

29
The structure of a data file

• II- Basic distributive variables
• GROUP NUMBER tells DAD on which value of the
  GROUP VARIABLE to condition the computation of
  some distributive statistics. The value for GROUP
  NUMBER should be an integer. For example, rural
  households might be assigned a value of 1 for
  some variable denoted as region.
• SAMPLING WEIGHT. Sampling weights are the inverse
  of the sampling rate.

30
Importing data files into DAD

• ASCII files
• After preparation of the required variables, one
  can export an ASCII file to be read by DAD.
• To import safely the data, a wizard is used in
  DAD.
• One can also use Copy/Paste to copy data from
  other software sheets (that is more risky
  however)
• A helpful software that can be used to prepare
  DAF files is Stat/Transfer (though a commercial
  software)

31
Launching DADs applications

• From the main menu one can choose the desired
  application applications are organised by main
  themes.
• After choosing the desired application, a second
  widows appears to indicate the number of
  distributions or data files that should be used.

32
DADs application for the FGT index
33
DADs window of results
34
DADs graphs (ex. Lorenz curves)
35
DADs graphs (ex. difference between Lorenz
curves)
36
DADs graphs (ex. FGT curves (a0))
37
DADs graphs (ex. Concentration Lorenz
curves)
38
DADs graphs (ex. density curves)
39
DADs graphs (ex. Non parametric regression)