Transcript Price Survey Design

Third Workshop on
ICP Western Asia
Beirut, 11-15 October 2004
Design of ICP price survey
Sultan Ahmad, Consultant
[email protected]
Based on Keith Blackburn, Australian Bureau of Statistics
1
Aim of this session
Discuss how the actual scope for your
country can be determined:
geographic scope
time scope
Clarify choices needed when designing
the ICP price survey for your country
Identify possible frame sources
2
Scope – defining the target
Target or ‘ideal’ price
“National annual average price”
• Determined by need to link to GDP values
Arithmetic average price of all transactions
• For the whole country
• For the whole year
Necessary to sample
• Products (what products are purchased; product list)
• Outlets (where the purchases are made; outlet sample)
• Transactions (when the purchases are made; time sample)
3
Errors
National annual average price is an estimate
subject to sample error
subject to many non-sample errors
Product and transaction samples are purposive
Therefore not possible to calculate sample error
Outlet sample may be PPS (probability
proportional to size)
If it is, then it is possible to calculate sample error
Product PPPs result of two estimated averages
May calculate sample error of the ratio
4
Probability sample
Random or probability samples preferred
Unbiased
Efficient – have minimum variance
Prior information needed
Allow stratification
Selection using probability proportional to size (PPS)
PPS size measure (for ICP purposes)
Ideally want product sales volumes
In practice may use ‘total sales’ or ‘staff numbers’
5
Purposive sample
Subjective selection by person or agency
CPI – used extensively
• Could use random sampling for products but don’t
ICP – used for product and transaction samples
Random or probability sampling preferred
Purposive sampling used
• When information not available for random sampling
– When the cost of random sampling too high
6
Product sample
Regional ICP product list provides the ‘frame’
Your list prepared at finalisation workshops
provides national sample:
includes ‘representative’ and ‘available’ items
excludes ‘not available’ items – keep to minimum
avoid changes after survey starts (costs/data quality)
May include items not included in CPI
7
Outlet sample
Regional circumstances vary enormously
Geographically
• City states/Single state countries/Multi-state countries
Available CPI infrastructure
• single national systems/separate provincial systems
• highly developed systems/simple limited-scope systems
Outlet survey designs should reflect individual
national circumstances
Remember CPI design may not be adequate
8
Choosing outlets
Must choose both location and type
Location – city/rural poor province/wealthy province
Type – street stall, small shop, department store
Outlet choice more important for ICP
Price levels differ by outlet type – cause ICP bias
CPI more robust – only needs consistent movements
Care to avoid price collectors taking easy options
Pricing in department stores easier
Risk bias if high price goods easier to measure
9
Informal markets
Sampling frames may exclude informal outlets
e.g. Street stalls and itinerant sellers
Weekly markets may be very important and
account for major part of household purchases
Informal markets may require location sample
Price collectors then select specific sellers
Bargaining – no fixed prices
Special care needed by price collectors
With such markets there is risk of serious bias
10
Using CPI frame
Is suitable
national
CPI frame
available?
No
Yes
Is a new frame
required?
Use existing CPI
frame
No
Yes
Supplement CPI
frame
Develop new
ICP frame
11
Time sample
Aim at average prices for the whole year
Need quantity weighted transaction price
Choices - Monthly/Quarterly/Annual
Frequency – determined by variability
• Type of charge (annual, seasonal etc.)/inflation
12
Sample plan
J A S O N D J F M A M J
Food
Cities
Other urban
Rural
Clothing
X
X
X
X
X
X
X
X
X
X
X
X
Cities
X
X
Other urban
Rural
X
X
X
X
13
Multiple CPI outlet frames
Multi-state countries may have many
independent CPIs (eg India, China)
Design of a single national ICP sample will
need input from skilled sampling experts
A purpose-built sample needed
Outlets should be selected on PPS basis
Stratified sample may be appropriate
Use expenditure weights for PPS, if available
14
Alternative-frames
If CPI frame not available or needs
supplementing alternative frames are
needed
National level sampling frames
• May provide urban/rural, area and other
broad level weights
• National frames may not cover outlets
Local level outlet sampling frames
• Usually are available
15
Alternative-frame issues
Is it available for statistical purposes?
Is it available in the required time-frame?
Is it reliable?
Is it current?
Does it have the required coverage?
Does it include size indicators?
16
Sources of national frames
Retail trade census
Survey of retail sales
Value Added Tax (VAT) records
Business registers
Telephone directories (“Yellow Pages”)
17
Sources of local frames
Local government records
Property tax
Other regulatory systems
Local chambers of commerce
Employer organisations
18
Retail trade census or survey
Frames may exist at national level as
well as local level
Useful for probability proportional to
size (PPS) sampling
Usually maintained by the statistics
office
19
Value Added Tax (VAT)
VAT administrative records provide detailed
lists of businesses
Up to date
May not be available for statistical purposes
May support PPS sampling (provide size details)
May exclude some sectors such as food
May exclude informal sector businesses
If available, care needed with exclusions
May require supplementary sources
20
Business registers
Business registers maintained for:
Statistical purposes
Administrative purposes
May contain retail outlet locations/addresses
May include size measures
sales volume
number of employees
Useful for PPS sampling
21
Telephone/business directories
“Yellow Pages” and similar directories
Do not include size measures
May include detailed lists of businesses
May not be comprehensive
• PPS not possible
• simple random sampling can be used
22
Thank you
Any questions?
23