Measuring Consumption and Poverty in Zambia GSS methodology conference, 27 June 2012

Measuring Consumption
and Poverty in Zambia
GSS methodology conference, 27 June 2012
Presentation summary
• Why a poverty trend needed to be developed
retrospectively in Zambia
• LCMS 2010 fieldwork delays lead to
problems
• Adjustments to 2010 data to ensure
comparability
• Political environment – Lessons?
Zambia consumption surveys 1996-2006
• Living Conditions Monitoring Surveys (LCMS)
1996, 1998, 2004, 2006
• Conducted over 2 – 4 months
• Consumption data collected in a one time HH
interview
• LCMS 2002/3
• Longitudinal survey
• Diary method
• In field for 12-15 months
Zambia consumption surveys 1996-2006
• Differences in survey design:
• Questionnaires
• Sample size and selection
• Dates for data collection
• Differences in analysis methods
• Poverty lines
• Consumption aggregate
• Thus no comparable trend available
Questionnaire differences 1996-2006
• Differences in lists of food items (e.g . ‘other
vegetables’, poultry)
• Significant differences in maize and maize
products
• Use of different recall periods (2 weeks, 4 weeks, 1
month, 12 months)
Questionnaire differences 1996-2006
• Some surveys included quantities and values
of own production, allowing prices to be
generated.
• Others had seperate community price
questionnaires
Differences in sample size and sample
selection
• Sample selection of SEAs based on 2000
census sample frame - consistent over
surveys
• Sample size varies from 6,000 to 20,000
households
• Sample of households per strata within SEAs
selected by supervisors or enumerators
Zambia poverty lines 1996-2006
Zambia poverty lines
food poverty line overall poverty line
1996
20,181
28,979
1998
32,861
47,187
2004
78,223
111,747
2006
65,710
93,872
Decline in the poverty line btw 2004 and 2006 shows that the
methodology of updating the poverty line changed over time. (CPI
Food inflation 2004-2006 = 20%)
8
Developing methodology to establish the
poverty trend, retrospectively
• Reviewed questionnaires and available data
for 1996, 1998, 2004 and 2006
• Developed a new method that could be
applied to data 1996-2006
• New food basket based on consumption
shares on each food item for HHs in 5th and
6th deciles (2006)
Developing methodology to establish
the poverty trend, retrospectively
• Valued food basket for each year using item
specific CPI indices (avoiding differences in
CPI structure over time)
• Derived overall poverty line based on non
food share of HHs in 5th and 6th deciles
• Developed a consistent consumption
aggregate based on best practice and
available data across surveys
Comparison of overall poverty lines ('000
Kwacha), 1996 - 2006
120
100
80
60
40
20
0
1996
1998
Old CSO poverty line
2004
2006
New CSO poverty line
New Poverty trend – Zambia 1996-2006
Poverty headcount (% poor) Zambia, urban and rural
90
80
70
Percentage
60
Rural
50
Zambia
40
Urban
30
20
10
0
1996
1998
2000
2002
2004
2006
LCMS 2010
LCMS 2010
• CSO required that the LCMS 2010 serve two
purposes:
1. To monitor poverty trends 2006 – 2010
2. Develop separate urban and rural poverty
lines.
• Lessons learned from previous surveys Questionnaire had more specific consumption
items listed
LCMS 2010: Continuing poverty trend
• Developed a ‘Narrow’ consumption
aggregate, strictly excluding all items NOT in
the 2006 LCMS
• Included items likely to be under ‘other’
categories in 2006
LCMS 2010 ‘Other’ categories
excluded
• Own production/’receipt from other sources’
of other vegetables, fruits and own poultry
these were ‘accidentally’ crossed out in 2004
and 2006 questionnaires
• ‘Narrow’ consumption aggregate therefore
excludes these items
LCMS 2010: A key difference
• Problem: LCMS field work was delayed by 2
months
• Rainy season (access)
• Lean season
• 2010 data showed a large increase in poverty
when trends (2009) methodology applied
Attempts to deal with ensuring
comparability in analysis
• Adjustment (i): Use Dec./Nov. 2009 prices
to evaluate the food poverty line
• Argument: Fieldwork was 2 months later than the 2006
survey period, Prices in the period January to March
were slightly higher than Nov-Dec .
Adjustments to deal with comparability
in analysis
• Adjustment (ii): Reduce the food share of
the overall poverty line
•
Argument: Households have a lower non-food consumption in the
period January to April (lean period) than from Nov. to Dec.
•
LCMS 2002/03 show that the non-food share in the period Jan. to April
is indeed lower than in the period Nov.-Dec.
•
The non-food share of the total poverty line in 2010 was lowered by 3
percentage points, from 41.5% to 38.5% (was reviewed with 1.5% and
3% reductions)
•
However the LCMS 2002/03 report shows lower poverty in Q1
Adjustments to deal with comparability
in analysis
• Adjustment (iii) Take out the price
increase for imported rice from the pricing
of the poverty line
• Argument: Between 2006 and 2010 the national median
price for local rice increased roughly by the factor 2.14,
for imported rice by the factor 4.17.
Adjustments to deal with comparability
in analysis
• Adjustment (iv): Use an extended
consumption aggregate
• Consultants requested to create an ‘extended’
consumption aggregate, to include own-produced and
items received of other vegetables, fruits and poultry
• Comparing item specific budget shares over time
reveals that own produced ‘other vegetables’ are
extremely important consumption items.
Adjustments to deal with comparability
Adj. (iv) – extended cons. agg.
• Share (of total consumption) of other vegetables
was 1.7% in 2006 and 2% in 2010 – under the
narrow consumption aggregate (inc. only
purchases)
• Share of ‘other vegetables’ in ‘extended’
consumption aggregate (2010) is 11%
• This indicates that the narrow 2010 consumption
aggregate is actually more comparable to the 2006
aggregate.
Adjustment (iv): outcome
• When ‘Extended’ aggregate is used to generate
poverty headcount, results ……………………
• Poorer quintiles gained against the overall trend of a
decline in real consumption (own produced
vegetables, fruits and poultry are of much greater
relevance to poorer quintiles)
• Significant discomfort with this adjustment
Further revisions made by CSO
• Spatial price differences removed
• Remittances sent added to HH consumption
aggregate
• Consumption aggregate for 2010 includes all
food items (e.g. ‘other own production’)
Effects of adjustments on poverty
headcount
Changes in poverty line methodology
Prices used
Imported
Products
Poverty Lines
Headcount Estimates
Food share
Food
Poverty
Line (K)
Overall
Poverty
Line (K)
Base
Extended Cons.
Consumption
Aggregate
Aggregate
Constant (58.5%)
103,611
177,123
70.2%
67.2%
Base scenario poverty line
Jan-March
2010
local and
imported rice
Combinations of changes
Nov-Dec
2009
local rice only
Constant (58.5%)
98,505
168,394
68.4%
65.5%
Nov-Dec
2009
local rice only
Ad-hoc (60% =
+1.5%)
98,505
164,175
67.6%
64.3%
Nov-Dec
2009
local rice only
Ad-hoc update
(61.5% = +3%)
98,505
160,171
66.7%
62.8%
96,366
146,009
Final method: All food consumed and
remittances sent inc. in consumption agg.
provincial price deflator removed
60.5%
Political situation
• CSO senior management change half way
through survey (temporary promotion)
• GRZ preparing for election and CSO under
pressure to show reduction in poverty
• Clear steer for ST that poverty must have fallen
(argument anecdotal)
• President announces fall in poverty before final
numbers agreed
Lesson learning
• Survey had on-going evaluation – Paper
available
• Summary of lessons learned in brief paper
• Political lessons
• Management lessons
• Technical lessons
Thank you!