American Bar Association Antitrust Law Journal

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

A.C.

NIELSEN MARKET SHARE DATA


Author(s): Lloyd E. Oliver, John Pisarkiewicz and Jr.
Source: Antitrust Law Journal, Vol. 47, No. 3, TWENTYSIXTH ANNUAL MEETING (August
7-9, 1978), pp. 1067-1076
Published by: American Bar Association
Stable URL: https://www.jstor.org/stable/40840032
Accessed: 06-04-2020 15:08 UTC

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

American Bar Association is collaborating with JSTOR to digitize, preserve and extend access
to Antitrust Law Journal

This content downloaded from 103.77.138.95 on Mon, 06 Apr 2020 15:08:10 UTC
All use subject to https://about.jstor.org/terms
A.C. NIELSEN MARKET SHARE DATA
Lloyd E. Oliver

Economist, Glas sman- Oliver Economic


Consultants, Inc., Washington, D.C.

John Pisarkiewicz, Jr.


Economist, Glassman- Oliver Economic
Consultants, Inc., Washington, D.C.

With the enactment of legislation conferring greater authority o


Federal Trade Commission to seek injunctive relief in merger case
passage of the Hart-Scott-Rodino Antitrust Improvements Act wit
implicit (and justifiable) recognition of injunctions as the preferred
of enforcement of the merger laws and the increased incidence of
merger suits due to the recent spate of unfriendly take-over attempts,
has been great interest in readily available individual company marke
information published by private organizations which sell these d
profit. Economic researchers, of course, long have sought individual com
market share data as critical structural variables in industrial organ
analyses.
The A.C. Nielsen Company is a publicly held firm which has been provid-
ing this type of service to the business community for over forty years.1
It now has operations in twenty-three countries and annual revenues of
about $300 million. The company is presently organized into five basic
groups - Marketing Research Services, Media Research Services, Clearing
House, Neodata and Petroleum Information Services. The first is the subject
of this paper, but the company itself probably is known best for its television
program ratings which are produced by Media Research Services. The Clear-
ing House processes coupons for retailers and manufacturers; Neodata
collects information on magazine circulation and provides other statistical
information for publishers; and Petroleum Information provides statistical
services for the oil and gas industry.
The Market Research Service group sells information on the movement
of branded package goods through grocery stores, drug stores and mass

lrThe headquarters of Nielsen is located at Nielsen Plaza, Northbrook, Illinois. They have
a number of offices located throughout the country.

1067

This content downloaded from 103.77.138.95 on Mon, 06 Apr 2020 15:08:10 UTC
All use subject to https://about.jstor.org/terms
1068 A. C. Nielsen Company

merchandising outlets. It thus is able to provide


share information at the retail level in unit and dollar terms on a wide

array of food and pharmaceutical consumer products. The heart of its d


collection system is still the Food and Drug Indexes, which are two of t
oldest Nielsen services. The Drug Index was first initiated in 1933 and t
Food Index was begun in 1934. A more recent service, the Mass Merchandiser
Index, really was designed to supplement the Food and Drug Indexes in
order to expand coverage of health and beauty products, although it h
enabled the company to measure other products such as automotive supplies.

The work of the Market Research Service organization is carried out


by four departments whose basic functions are suggested by their title
Field, Production, Statistical Research and Client Services. Professional st
isticians in the Statistical Research Department develop the universe cont
information and work out the sample design. The more than 500 full-ti
field auditors who collect the basic information at the store level receive
six months of training at Nielsen's training school (which has models o
a "working" grocery store and a drug store) and another six months of
on-the-job training. Another task for the field personnel is to secure th
cooperation of the stores designated for inclusion in the sample by the Statis
tical Research Department and also to pay these stores for their cooperation.
The company claims to have a high rate of cooperation from the designat
stores. Nielsen has continuing cooperative agreements with seventy-three of
the ninety-two largest grocery chains as well as with many important region
chain organizations and does not feel that any significant omissions exis
in the Food Index. For the Drug Index, however, two West Coast chains
Thrifty and Sav-On, do not cooperate, and no K-Mart stores are in the
Mass Merchandiser Index.

The collection of brand movement information works much the same


way for each product included in the basic Indexes and illustrates the
uniqueness of the Nielsen system. The company measures retail movement
during a bimonthly period. A bimonthly period is used (although data can
be produced on a monthly basis, if requested) since the company considers
it efficient to use its work force basically to collect food information one
month and drug information the next month. The quantity of the product
on hand is counted, whether on the shelf or in inventory in the store, on
the day of the store audit. From that figure is subtracted the total inventory
from the previous audit and this result is added to the quantity of product
delivered to the store during the past two months to arrive at total unit
movement. Special circumstances such as store displays and promotional
activity are recorded, and this information is supplied to subscribers. Items
which have promotional deals attached by the manufacturer {e.g., "twenty-
five cents off') are shown separately, but no adjustments are made to any

This content downloaded from 103.77.138.95 on Mon, 06 Apr 2020 15:08:10 UTC
All use subject to https://about.jstor.org/terms
Lloyd E. Oliver/John Pisarkiewicz, Jr. 1069

subsequent data to reflect this promotional activity. Furthermore, a limited


amount of information on the extent of couponing is available, if requested.
The unit movement data obtained are converted to dollar sales by using
the price in effect on the date of the audit (i.e., at the conclusion of the
bimonthly period). Dollar sales are not adjusted if there were a different
price in effect during the period (if that can be ascertained), but, again,
this fact is noted to subscribers.

The actual date of the audit to measure the bimonthly movement extends
over a four-week period. For example, data for what normally is referred
to as the "January- February" reporting period typically would be gathered
during the last two weeks of February and the first two weeks of March.
Each store, while visited monthly, is on a two-month schedule for individual
products. Consequently, some audits are reporting movement from December
15 to February 15 while others are covering the period January 15 to March
15. On average, the period covered is from late December to late February.
This difference in the time of the audit provides for variation in prices
recorded and promotional efforts by retailers.

This "point-of-sale" measurement technique is a distinguishing charac-


teristic of the Nielsen system. The company feels that this approach is supe-
rior to alternative methods since it best enables a manufacturer to assess
his marketing program. In* the literature it makes available for potential
clients,2 Nielsen lists the advantages of this technique over measurement
of warehouse withdrawals, the device employed by Selling- Areas Marketing,
Inc. (SAMI). As Nielsen observes, warehouse withdrawals really do not mea-
sure final sales but only the interim movement from one point in the distribu-
tion system to another. Furthermore, warehouse records do not account for
goods shipped directly from manufacturers to retailers, which can result
in significant distortions for some products such as frozen pizza. The records
of some warehouses also are not kept with consistent accuracy, particularly
if they are unwilling or unable to provide computerized information. Finally,
in-store measurement offers the capability of recording additional informa-
tion on out-of-stock situations and promotional and display activities which
lend greater (but sometimes subjective) understanding to actual movement
conditions at the retail level.

Nielsen claims that obtaining this type of information from a house-


hold panel by diary methods would not be effective because it would be
economically unfeasible to get the cooperation of a representative sample
of families; the records kept in diaries are often inadequate and inaccurate;
and it would require a household sample much larger than Nielsen's to
obtain the same degree of precision.

2A.C. Nielsen Co., Nielsen Retail Services Index Services 20 (1975).

This content downloaded from 103.77.138.95 on Mon, 06 Apr 2020 15:08:10 UTC
All use subject to https://about.jstor.org/terms
1070 A. C. Nielsen Company

For the Food Index, movement informatio


about 1,300 stores (which can include multipl
and it is the data base which Nielsen uses to m
stores in the forty-eight contiguous states. These
refers to as its "sample," are drawn from a list
which attempts to include all food stores. This
on U.S. Census data, outside sources such as r
canvasses of major markets by their own field ag
A store remains in the sample until an in-hous
that it should be eliminated or replaced.
Nielsen describes its statistical estimation metho
based on a stratified, disproportionate random
disproportionality are employed as techniq
sentativeness of the sample. The strata vary fo
are by geographic area and by store type and s
is thought to be necessary to account for diffe
chain organizations with varying merchandising
of manufacturers and brands and in food taste
forty-eight states are divided into ten regions, o
are New England, Metropolitan New York,
Chicago, West Central, Southeast, Southwest,
Pacific. The counties within each region are fu
"A," "B," "C," or "D." The county designation
urbanization. An "A" county is highly urbaniz
tially is rural.
Due to perceived differences among store ty
multiple stratification is employed, and separ
ployed for each category. Table I4 lists these c
volume and the sampling rate applied to grocery
Table I

Average Volume
per Store Sampling

Category (Thousands of $) Rate

Super Market Chain $3,700 1 :40


Large Independent 1,600 1:67
Convenience Chain 233 1 :220
Medium Independent 222 1:220
Small Independent 45 1 :400

3Address by Thomas D. Cole, Vice President and Man


A.C. Nielsen Co., to the Triangle Advertising Federat
(April 12, 1978) (mimeo).
Hd.

This content downloaded from 103.77.138.95 on Mon, 06 Apr 2020 15:08:10 UTC
All use subject to https://about.jstor.org/terms
Lloyd E. Oliver/John Pisarkiewicz, Jr. 1071

Lumping the convenience chain and medium independent into one cate-
gory because of identical sampling rates, the stratification scheme produces
160 cells - ten regions, four county sizes and four store sizes. Within each
cell, stores are selected randomly, but the rate of sampling between cells
varies in accordance with the store size of a cell. Nielsen feels this adjustment
is necessary because, for example, chains and large independents represent
only twenty percent of the total number of stores but account for over eighty
percent of all sales through food stores.
The company uses the term "ratio estimation" to describe their method
of developing national estimates from the sample store audit data. This
technique serves as an internal check on estimates made by expanding the
audit figures on the basis of the number of stores in the country. A ratio
is constructed by dividing the estimated dollar volume of a particular prod-
uct by the all commodity dollar volume of products sold through food
stores - a figure which is available from the Survey of Current Business. The
resulting ratio is checked against previous values of this ratio. If the new
one appears to be out of line with past values, then presumably an effort
is made to check the store audit figure and/or develop an explanation of
why a change is expected {e.g., dramatic rise in coffee prices).

The other two parts of the total Retail Index - the Drug Index and the
Mass Merchandiser Index - are constructed in a similar fashion although
sample sizes and stratification schemes differ. For the Drug Index, 710 stores
constitute the sample which is stratified both geographically in accordance
with the ten Nielsen regions and on the basis of store size. Three size catego-
ries are employed (over $300,000, $300,000-$ 150,000 and under $150,000)
for independent drug stores and a separate category exists for chain stores.
For the Mass Merchandiser Index, 150 stores are audited and only a rela-
tively simple geographic stratification scheme is possible because of the small
sample size. This scheme divides the country into four regions. Also, two
size categories exist based on square footage of the store (60,000 sq. ft. or
more and less than 60,000 sq. ft.).
Within the Market Research Service Group, Nielsen can provide a variety
of other services which are not necessarily directly linked to the Retail Index.
For example, a relatively new service audits a projectable sample of liquor
stores in New York, Boston, Chicago, Los Angeles and San Francisco. Index
service type information is available on thirteen liquor and nine wine catego-
ries. Another service, directly linked to the regular Retail Index, provides
information on either a monthly or bimonthly basis for thirty-eight major
markets. These areas correspond roughly to television market Areas of Domi-
nant Influence (ADIs) which are much larger than the Standard Metropol-
itan Statistical Areas used by the Census. Finally, the Nielsen Early In-
telligence System is an example of a service which is not at all related to
the Retail Index. Available for the past twenty years, this system provides

This content downloaded from 103.77.138.95 on Mon, 06 Apr 2020 15:08:10 UTC
All use subject to https://about.jstor.org/terms
1072 A. C. Nielsen Company

bimonthly information based on food store war


methodology is ironic in light of Nielsen overall
surement technique. In fairness to Nielsen, howe
the information to be a substitute for that available in the Retail Index.

The system monitors a sample of 150 food store warehouses only but doe
provide complete detail on brand, sizes, color, flavor, retail selling price
(i.e., suggested price on the warehouse books), retail distribution and unit
movement for products in approximately 600 product groups. These are
not necessarily the same as those used in the Retail Index. Nielsen thinks
that the major advantage of this service is its ability to provide generalize
marketing planning data to clients quickly. It is particularly useful for moni-
toring new product introductions or for market scanning to determine new
areas of opportunity.

Once all of the information is in place, Nielsen is very flexible with respect
to the data it provides to subscribers. Any individual client can request
data tailored to his specific needs. All brands, flavors and package sizes
can be shown for any product or brands of a product specified by the client.
A firm can indicate, within limits, what to include in the product class
Each company receives reports prepared individually for that particular
company. It should be mentioned, however, that the Nielsen market shar
data seen by the authors have usually conformed to expected product group-
ings. The dates (for the bimonthly periods) and geographic areas covered
ordinarily do not differ, although Nielsen will tailor data to some extent
to the sales areas and time periods desired by individual firms.

The client can purchase as much of this information as is deemed useful.


Some manufacturers, given their product and their perception of the compe-
tition they face, may desire only Food Index product data or perhaps only
that relevant to a particular geographical area. Others may want national
information from all three Indexes for their product. For example, manufac-
turers of health and beauty aids who distribute their products through a
variety of retail outlets would appear to be the type of client likely to
purchase information gathered on, say, aspirin or hair spray from all three
Indexes. On the other hand, manufacturers of ready-to-eat cereal undoubt
edly have very little interest in the Drug or Mass Merchandiser Indexes.

A.C. Nielsen copyrights its information and sells it to clients but requires
a strict confidentiality agreement as part of its general client contracts. These
agreements contain an express provision that the client agrees it will not
be used in any legal proceeding. Some exceptions to confidentiality are
permitted, such as manufacturers showing it to retailers (presumably to get
shelf space), but these do not include academic research use, published o
unpublished. This clearly creates a problem for economic researchers and

This content downloaded from 103.77.138.95 on Mon, 06 Apr 2020 15:08:10 UTC
All use subject to https://about.jstor.org/terms
Lloyd E. Oliver/John Pisarkiewicz, Jr. 1073

explains why very little Nielsen market share information is in the public
domain, although it will be occasionally seen in trade journals like Advertising
Age. Even relatively old data (e.g., ten years old) is considered confidential
by the company because they still may be salable to a new firm entering
the product line.

As might be expected, Nielsen receives numerous requests and subpoenas


for its market share data. The company has been quite reluctant to have
its information used in any litigation and claims that it has been successful
in a number of instances in keeping it out. Where it has gone into the
record or been turned over to government agencies or other parties, Nielsen
has insisted on assurances of confidential or in camera treatment. The com-
pany argues that its data (1) are "hearsay" and, therefore, not admissible
unless by stipulation; (2) represent "expert opinion" so that company per-
sonnel cannot be forced to produce or testify about the information; and
(3) are not the "best evidence" because the information is not actual universe
or brand data. Nielsen further claims that because their reports are often
tailored to their clients' individual requests, release of the information may
reveal their clients' individual marketing strategies.

The actual situation with respect to litigation should probably be regarded


as somewhat ambiguous. Nielsen market share information has been used
in a number of cases and government investigations. It is frequently admitted
into evidence because it is relied upon by a party. In most of the cases
known to the authors where Nielsen data have been used, the information
usually has played a corroborative role. It often has been admitted pursuant
to stipulation by the parties (typically because the market shares themselves
were not a critical issue), as in the Warner-Lambert Co. case.5 It was put
into evidence for a number of years in the FTC's on-going cereals case,
Kellogg Co.,6 but confidential treatment was permitted only for the most
recent years. However, in Procter & Gamble Co.,7 Nielsen market share infor-
mation was a principal source of market share data and played an important
role in documenting the effect of entry into the Erie, Pennsylvania, area.

An evaluation of Nielsen market share information must begin with an


understanding of what the data are designed to accomplish. Nielsen data
are intended as management tools, especially to enable manufacturers to
assess their marketing programs and determine the penetration of their own
and competitors' products. They are not necessarily designed to incorporate
the kinds of safeguards which might be expected by the courts before sample

588 F.T.C. 503(1976).


b86 F.T.C. 318, 650, 685, 1195 (1975).
763 F.T.C. 1465(1963).

This content downloaded from 103.77.138.95 on Mon, 06 Apr 2020 15:08:10 UTC
All use subject to https://about.jstor.org/terms
1074 A. G. Nielsen Company

evidence is introduced at trial. The company itself c


tation of its data requires an expert evaluation o
coverage, sampling errors and nonsampling errors.

In general, the methodology employed by the


in the preparation of its market share informat
standard statistical techniques. The company
methods to the universe of food stores and drug
of stores. It has taken the further step of tryin
of health and beauty products by developing a
mass merchandising outlets. While the company wil
for its estimates, the techniques of adjustment b
for differences in geographic patterns and store siz
tionality of sampling and of "ratio estimation"
the "expert opinion" of Nielsen and are conside
company. Consequently, they are somewhat diff
greater access to actual specific formulas used by th

Unless these statistical techniques are carefully


can be magnified considerably. The coverage of N
to its closest rival - SAMI - because it does incorpor
(a significant SAMI problem) and does obtain
than the less precise proxy of warehouse withdr
products covered by Nielsen is fairly broad, the dat
to consumer food and drug products purchased i
Merchandiser Index is allowing coverage of som
does not include meats, produce, bakery or dairy
institutional use (restaurants, hospitals, universit
prescription drugs, industrial products or any marke
the company does not pick up direct home delive
which can be important for something like orange
with the lack of coverage of department stores w
the usefulness of Nielsen data on, say, cosmetics
from coverage are vending machines, airports, p
stations which would be significant for products su

Nielsen does measure sales at the retail level (rathe


sale sales) which is an advantage to a researcher b
demand of the final consumer rather than the derived or intermediate
demand for products. It also avoids the necessity of evaluating exports and
imports which can be nagging problems when measuring sales at other levels
because retail sales automatically include imports and exclude exports.

8A very limited number of prescription items are covered, including some shampoos, but
these represent an exception.

This content downloaded from 103.77.138.95 on Mon, 06 Apr 2020 15:08:10 UTC
All use subject to https://about.jstor.org/terms
Lloyd E. Oliver/John Pisarkiewicz, Jr. 1075

The Nielsen Indexes do have some significant problems, particularly in


the measurement of dollar sales. Using the price of the product on the day
of the audit may introduce distortions since it may not estimate correctly
the average of actual transactions prices during the bimonthly period. It
is not at all clear, however, that this imparts a specific bias in one direction
or the other because the store audits do occur over a four-week period and
some price variability is captured by this technique. Further, Nielsen is not
able to capture price reductions via couponing directly to consumers (i.e.,
"cents-off" coupons appearing in newspapers and magazines or mailed to
the consumers' homes) which could be quite significant with respect to cer-
tain heavily promoted products such as coffee.

The major problem for independent researchers is availability of the Niel-


sen reports themselves. Unless the company were willing to make these
reports available to the public, which they are not inclined to do, the cost
of buying them could be prohibitive.9 Table II illustrates the problem:

Table II10

Relative Cost of
A.C. Nielsen Company Services*

COST OF NIELSEN

SERVICE AS A

TYPE OF CLIENT PERCENT OF SALES

Five Large Grocery Manufacturers 0.05%


Five Large Drug Manufacturers 0.05%
Seven Small Manufacturers

(Grocery and Drug Combined) 0.14%

*Retail Index Division

While these figures may seem nominal to a manufacturer with a vital interest
in the relative performance of his product in the marketplace, the cost can
easily run into several hundreds of thousands of dollars.

In view of this, the economic researcher will be limited to that market


share information made available through the press or courtroom proceed-
ings. Because this is likely to deal with only a single industry, it will be
difficult to assemble the kind of information necessary to do a cross-sectional
study examining many industries during the same time period.

9Reportedly, Nielsen reports are more expensive than SAMI data.


10A.C. Nielsen Co., 1977 Annual Report, at 7.

This content downloaded from 103.77.138.95 on Mon, 06 Apr 2020 15:08:10 UTC
All use subject to https://about.jstor.org/terms
1076 A. C. Nielsen Company

For the antitrust attorney, whether governme


tion, if obtainable, can prove extremely help
can provide important clues regarding the fir
tion. Assuming the Nielsen market correspon
from an appropriate application of the econo
and product market definition, then the sh
valuable.

Mr. Edlund: The next methodology we will consider is the consumer


questionnaire methodology. John Cady is professor of business administra-
tion at Harvard. He holds bachelor's, master's and Ph.D. degrees from the
State University of New York. Prior to joining the faculty at Harvard, he
was a member of the marketing faculty at the University of Arizona. In
addition to being an active consultant for state and local agencies, law firms
and private corporations, he has been involved in this area as a witness
and as an advisor in a number of antitrust and regulatory rulemaking
proceedings. He has been a consultant to the Federal Trade Commission's
Bureau of Consumer Protection and to the Bureau of Competition. He is
an author of many publications on marketing, management and advertising
and is going to speak to us particularly with respect to the consumer ques-
tionnaire methodology and the Target Group Index, which uses that meth-
odology among others.

This content downloaded from 103.77.138.95 on Mon, 06 Apr 2020 15:08:10 UTC
All use subject to https://about.jstor.org/terms

You might also like