Average retail food prices: a brief history of methods Essay

The Bureau of Labor Statistics publishes average retail food prices
on a monthly basis in a news release, Consumer Prices: Energy and Food.
Data are published for the United States and for four major geographic
regions–Northeast, North Central, South, and West. The report present
average prices for 94 food items that are calculated from data used in
compiling the Consumer Price Index (CPI). All of the major CPI
“food at home” categories–cereals and bakery products; meats,
poulty, fish, and eggs; dairy products; fruits and vegetables; and other
foods at home–are represented in the list of average food prices. Each
report also contains data for the two preceding months.



Average retail food prices are among the oldest data series
published by BLS. The first report, issued in 1904, contained average
monthly retail prices for about 30 foods for the years 1890-1903. Input
data for the report were obtained retroactively from account books and
records of about 800 firms in 171 cities.



Prior to 1964, retail food prices were weighted averages of prices
collected for use in compiling the CPI. From December 1963 through June
1978, average food prices were estimated from the movement of the CPI.
Each year, usually in January, special benchmark prices were calculated
for narrowly defined classes of food products. These benchmark prices
were adjusted in succeeding months by price changes reflected in the
appropriate CPI series. Because the CPI series pertained to more
broadly defined product categories than did the benchmark average food
prices, a new set of benchmark prices was computed annually to prevent
estimated prices from deviating widely from a true average of collected
prices.


The Bureau adopted this estimation technique for average prices as
a result of changes made in the specification pricing procedures during
a revision of the CPI, completed in December 1963. As a part of that
revision, the specifications used in collecting CPI prices were
broadened to encompass a wider sample of goods and services. While this
procedure improved the item sample for the CPI, it made calculation of
the average good prices difficult because of the greater heterogeneity of foods being priced within a specification. The “benchmark and
estimation” technique for calculating average food prices was then
developed to meet the continuing needs of users of such information.



Because of the major methodological changes introduced in the 1978
revision of the CPI, a completely different approach had to be developed
for calculating average food prices. The demanding schedule for the
completion of the 1978 CPI revision made it impossible to revise the
average food price program in time to coincide with the release of the
revised CPI. Therefore, average retail food prices are not available
from July 1978 through December 1979. Data based on the revised CPI
sample are available beginning in January 1980, but average prices in
the current series are not comparable to estimates published through
June 1978.



Development of the new average food price program for 1980
presented the BLS staff with a number of difficulties. Because of the
substantial change in price collection methodology employed in the
revised CPI, a greater variety of food items (as well as nonfood goods
and services) have been selected for pricing. For the pre-1978 CPI, BLS
field representatives had priced items that conformed to detailed
specifications which were basically the same for every store across the
country. Thus, a large number of prices were obtained for each of the
almost 100 food items. For an item such as cookies, for example, about
1,100 prices were collected nationally each month. The prices were for
almost identical types of cookies varying only by brand and package
size. Therefore, an adequate number of observations were available to
calculate an average price for a specific type of cookies in individual
cities as well as nationally.



In the revised CPI, collection methodology was changed to allow for
almost the full range of goods and services to be sample. Under this
procedure, the selection of each item is keyed to the sales experience
of the store in which it is priced. The field representative works from
a list of general categories in selecting the item to be priced. This
procedure gives each variety, brand, size, and so forth, a chance of
selection proportional to its importance in total sales for the general
category in the particular store. Once selected, the same item
continues to be priced over time. This procedure results in a
considerably larger range of goods and services being selected for the
food item sample.


For calculating the CPI, the revised procedure produces an index
which is much more representative of the goods and services purchased by
consumers. Fewer prices are obtained, however, for any specific item
because data collection is spread over a much broader range of food
products. For example, about 570 prices are presently being collected
nationally for cookies. These prices are representative of virtually
all kinds of cookies available in the marketplace, including packaged
cookies, cookies sold loose in bakeries, dietetic cookies, and all of
the various combinations of ingredients. Therefore, there are
relatively few observations for any one type of cookie, compared to the
1,100 prices that were obtained for a specific type of cookie prior to
1978. Because of the smaller number of quotations obtained for nearly
comparable food items, published average prices currently are available
only at the national and regional level.



The number of prices available to calculate average prices for any
food category in the CPI is dependent upon two factors: 1) the number of
price quotations assigned to the product stratum (which assignment is
designed for maximum accuracy of the CPI); and 2) the homogeneity of a
specific item with respect to ingredient composition, package size, and
packaging. Thus, for an item such as white pan bread, which has a large
number of price quotations assigned to its stratum and which is a
relatively homogeneous product, about 930 prices are obtained
nationally, of which about 60 percent are used to calculate the U.S.
average price. Generally, for the purpose of average price calculation,
very few items have usable sample sizes which approach that for white
pan bread.



In developing post-1980 calculation procedures for average food
prices, several procedures were considered, including the use of the
benchmark and estimation procedure used in the earlier series. It was
decided, however, to adopt a methodology in which actual weighted
average prices would be calculated each month. In determining the items
for which to develop average prices, BLS identifies the narrowest
possible specification for which a usable sample can be obtained and an
average price calculated. If the specification is judged narrow enough
to be useful, an average price is published. For example, average
prices are calculated for freeze-dried instant coffee in jars ranging in
size from 6.1 to 14 ounces. The specification was narrowed to this
range because the per-ounce price of freeze-dried instant coffee varies
widely from small jars (6 ounces or less) to large jars (more than 14
ounces). Therefore, prices for jars outside the 6.1- to 14-ounce size
range are excluded to eliminate price extremes which would not yield
realistic average prices.



The first step in calculating an average food price is the
computation of an “effective price.” This procedure involves
converting a reported price to a price per standard unit of measure
(weight, volume, or count). The published average prices are weighted
averages of the individual effective prices. The weight of each
observation reflects the relative share of expenditures which the
individual observations were selected to represent in the CPI. (See
“Consumer Price Index,” BLS Handbook of Methods, Volume II,
Bulletin 2134–2, for a detailed methodological description.)



Users of average retail food prices should be aware that these data
are best suited to measure price levels in a particular month. The
estimates are not designed to track price changes over time, nor are
they intended for use in making interarea comparisons. Ongoing updates
of the item and outlet samples will cause movement of average prices
over time to differ from the movement of an index for the same item,
because the index reflects only price change for the same product in the
same retail outlet. In calculating average prices, individual quotes
that meet the item and geographic definitions are included, regardless
of whether they are used for index calculation. Differences in prices
among geographic areas may not represent true differentials because of
variations in brand, quality, and size of the sample. Of course, such
differences will vary considerably depending on the item being observed.
For an item such as boneless round steak, for which U.S. Department of
Agriculture grades are used to define the quality of the cut of meat,
comparison of prices among regions is likely to be more informative than
for items such as fresh pork sausage, ice cream, canned tomatoes, and
smoked ham, for which differences in brand and quality can be quite
substantial.