Measuring inequalities in health

1
Measuring inequalities in health

• Adam Wagstaff
• Abdo Yazbeck

2
Todays menu

• Concentration curves and indices (AW)
• Combining levels and inequalities into a single
  achievement index (AY)
• Benefit incidence analysis (AY)
• Inequalities in financial burden of health care
  payments (AW)

3
Concentration curves and indices

• Adam Wagstaff

4
Which country is less equal?
5
U5MR concentration curves
CI 2 x area between 450 line and
concentration curve CI lt 0 when variableis
higher amongst poor
6
Setting data up for CC chart
7
Computing CI grouped data
8
Computing CI micro-data

• Can use where variable of interest (y) defined
  and measured at individual levelnot case with
  U5MR
• Use convenient covariance result
• Compute mean of ycall it m
• Generate individuals fractional rank in SES
  distributioncall it R
• Then compute CI 2 cov(y,R) / m
• If data are weighted,
• generate a weighted fractional frank, and
• compute a weighted covariance

9
Computing std errors for CIs

• Grouped data case
• Are variances of group means known? If they are,
  can get a more precise estimate
• Use formulae in TN 7compute in Excel
  spreadsheet available from Bank team
• Micro-data case
• Estimate b in regression below using Newey-West
  estimator in Stata equals CI std error is
  robust std error of CI

10
Health care payments

• Adam Wagstaff

11
Different concerns over health care payments

• Health care payments affect HHs ability to
  purchase other things that matter to their well
  beingfood, shelter, etc.
• But whats an equitable distribution?
• One where payments dont absorb more than x of
  incomei.e. arent catastrophic
• One where payments dont push HHs into poverty or
  further into poverty if already there?
• Or one where payments are proportional to ability
  to pay?

12
Rural Chinapayments relative to income
13
Rural Chinapayments relative to 15 threshold
14
Rural Chinapayments relative to poverty line
15
How much catastrophe?
Vietnam case study
18 of Vietnamese population in 1993 had
out-of-pocket expenditures in excess of 25 of
non-food consumption
16
How much catastrophe?

• Incidence of catastrophic costs can be measured
  as proportion (headcount) exceeding threshold
  level zcat Hcat
• Intensity of catastrophic costs can be measured
  as the average excess (or gap) Gcat
• If , in addition, we want to take into account
  that the incidence of catastrophic costs matters
  more for the poor, we can use the rank-weighted
  intensity, defined as
• where CO is the concentration index of the
  overshoot spending.
• Clearly, if excesses concentrated amongst the
  poor, CO will be negative and

17
Catastrophe in Vietnam
18
How much poverty impact?
Income
Pre-payment income
Poverty line
A pre-payment poverty gap
Cum sample
Pre-payment headcount
19
How much poverty impact?
Income
Pre-payment income
Post-payment income
addition to poverty gap from the new poor
Poverty line
C
A pre-payment poverty gap
Post-payment headcount
deepening poverty of pre-payment poor
B
Cum sample
Pre-payment headcount
20
Impoverishment in Vietnam
Out-of-pocket payments for health care pushed
2.6m Vietnamese into poverty in 1998. Increased
headcount by 23 and poverty gap by 25
21
How progressive?

• Regressive OOPs larger (as a of income) at
  lower income levels ? less inequality in OOPs
  than in pre-payment income Cf. progressive

• Lorenz curve shows income inequality
  concentration curve shows OOPs inequality
• Gini is twice area between Lorenz curve 450
  line concentration index is twice area between
  CC and 450 line
• Kakwani index is twice area between CC and Lorenz
  curve, or positive when progressive

22
How regressive are OOPs?
22
Sources Wagstaff, van Doorslaer, et al. (1998),
authors calculations