Sample Design

Workshop on Price Index Compilation
February 23-27, 2015
Sample Design, Selection, and
Gefinor Rotana Hotel, Beirut, Lebanon
Sample Design and Selection
Lecture outline
Common problems in price survey sampling
Where to start
Sample design
Sample selection and recruitment
Sample Design and Selection
Use of statistically sound sampling
techniques vs. judgmental samples
Approach dependent on availability of data
Need to maintain sample size
Sample Design and Selection
Common Problems in Price Sampling
Without probability sampling methods it is
difficult to measure accuracy and optimize
Attrition of sampled establishments and
products at the same time as new goods
Lack of up-to-date sampling frame
Voluntary nature of most price surveys
Purposive sampling increases chance of bias
Sample Design and Selection
Where to Start
Objectives of the surveys
 measure of inflation, use as deflator
 monthly or quarterly estimates
 level of detail for publication
 important commodities, groupings,
Sample Design and Selection
 National, regional, urban only, selected
 Economic activities (industry, services, etc.)
 Income / population groups (e.g., low,
middle, high income households; family size;
Sample Design and Selection
Data to be collected
Purchaser, producer, basic price?
Type of price (transaction, list, average, cost
Time of recording (cash or accrual)
Detailed specification of the product which
includes all the terms involved in the purchase
Temporary non-availability of some prices
Point in time price or average over the period
Sample Design and Selection
Sampling frames
 Business register (complete and up-todate?)
 Household lists
 Stratification variables included (ISIC, output
or sales, location; income, household
expenditures, etc.)
Sample Design and Selection
Level of accuracy and sample size
 Probability sampling enables us to use
population variance to determine sample size
 Non-probability sampling must be applied
carefully to ensure adequate representation
Sample Design and Selection
Available resources such as staff involved and
method of collection are constraints
Legislative issues such as voluntary or
mandatory reporting and laws governing
confidentiality or privacy affect our procedures
Sample Design and Selection
Sample Design
Probability vs. non-probability
Simple random sampling
Stratified random sampling
Probability proportionate to size (PPS)
Cutoff samples
Sample Design and Selection
Sample frame selection
 business register, business census, industry
 population census, housing lists, household surveys
Sample structure and stratification
 PPI, XMPI—region, industry, establishment,
 CPI—area, item, outlet, variety
Sample Design and Selection
Sample allocation—what will resources allow?
 XMPI, PPI by product and industry; CPI by
area and item
 specific products, industries, items that
should be published
 cutoff sampling may be best approach
Sample Design and Selection
Sample Selection and Recruitment
Specific issues in the CPI sample design
 selection of geographic areas (cities and rural
 selection of outlets and/or markets for pricing
 selection of items to include in the basket
 number of price observations per item
 frequency of price collection per item (often
determined by price variability)
 selection of individual varieties to represent items
Sample Design and Selection
Specific issues in the XMPI, PPI sample design
 selection of industrial coverage (mining,
manufacturing, energy, agriculture, transport,
 selection of representative industries
 selection of representative establishments
 selection of representative
 frequency of price collection per products
 selection of individual transactions to represent
Sample Design and Selection
Recruitment of establishments/outlets via
personal visit
Selecting transactions / varieties for pricing
•  identify important price determining
•  record the detailed specifications on collection
Thank you