Chapter 15.
International Price Indexes
Sampling
The objective of the International Price Program sample
design is to provide an unbiased measure of price change
in each published index. A multistage design is used to
select up-to-date specific import and export items that
can be priced over time. The first stage selects
establishments independently within each broad product
category (stratum); an establishment can be selected in
more than one category. The second stage selects detailed
product categories within each establishment/stratum pair
and is designed to support all of the IPP classification
systems (described in the Publications section of this
chapter). Each pair can be sampled multiple times,
allowing for more quotes to be selected from detailed
product categories with larger proportions of the stratum
value traded by the establishment. Subsequent stages of
subselection are conducted until a unique item that can
be priced over time is achieved. The first two stages are
completed at BLS using the appropriate sampling frame,
and the final stages take place during the interview with
the respondent.
The sample design is responsive to the constraints of
both cost and respondent burden. Budget constraints
determine the maximum number of establishments selected
in a sample, and respondent burden limits2 control the number of items
priced by each establishment.
Unfortunately, the majority of detailed product
categories sampled in the second stage do not contain
items that are traded frequently enough to be
consistently priced over time. IPP receives initial price
data for approximately 40 percent of the sampled
establishment/detailed categories.
The earliest IPP samples were based on judgmental
selection of establishments and items whose price
movement was considered representative of the
respondent's other items in the same detailed product
category. IPP began using a probability sampling
technique3 for the
first two stages in 1976 and for the subsequent stages in
1982. In order to maximize productivity, efforts are made
to ensure that frequent traders make up 95 to 99 percent
of each sample.
Starting in 1989, IPP divided the import and export
merchandise universes into halves. Samples for one import
half and one export half are fielded each year, so both
universes are fully re-sampled every 2 years. The sampled
products are priced for approximately 5 years until they
are replaced by a fresh sample of the same half-universe.
Generally, each index is composed of two samples.
As of 1991 a selected item can be represented in the
indexes up to 10 times per sample, an increase from a
maximum of two times for older samples. This allows items
with larger shares of the detailed product category to
have more weight in the indexes and reduces the number of
items for which the respondent needs to provide monthly
prices.
Footnotes 2Burden limits are usually determined by the
establishment's relative size, diversity, and consistency
of trade in the stratum. These assignments, however, can
be negotiated with the respondent. 3The technique used is commonly known as a
systematic probability proportionate to size design.
Next: Pricing
|