iassist quarterly - 3

Problems in the Use of

Business Data on Tape

by E Kay Worrell,

EHrector, Survey Research Center

The Conference Board, New York

Introduction

This will be a description of some types of

business data available on magnetic tape, the

uses made of these data by companies and

business researchers, and some problems that

may be encoimtered. Examples will be taken

from experience with Conference Board

applications and from discussions with business

users of such data among Conference Board

associate companies.

The Conference Board publishes research

reports, economic forecasts and newsletters, and

organizes conferences for the business

community. It is supported by subscription

income from associate companies, conference

fees, and publication sales.

The Survey Research Center provides assistance

to research staff in the collection and processing

of survey data, including maintenance of

research-related mailing lists. We also provide

statistical support, training in microcomputer

software, and assistance in database design and

interface between the mainframe and

miCTOcomputers.

Types of business data

There are three broad types of business data

available on magnetic tape.

1 Directory information, including names, titles

and addresses of executives

2 Financial and other descriptive information

about individual companies

3 Aggregate statistical information on the

economy, the workforce and industrial

production and services.

I will focus on the first two types: directory-

information on executives, and descriptive

information about individual companies.

Two of the most commonly used sources of

data on tape are Dim & Bradstreet Corporation

and Standard & Poor's Corporation. Both

companies were formerly best known for their

printed directories, and tape products were

derived from computerization of these data.

Dun & Bradstreet also offers extensive business

data processing and mailing services. Both

companies are producers of proprietary on-line

databases.

Standard & Poor's and Dun & Bradstreet rely

to some extent on self-reporting by companies,

in addition to publicly available financial reports

and other published sources. The titles of

individual executives are coded in some detail.

Standard & Poor's codes up to four separate

titles for each individual, using a two-part code

consisting of 35 discrete designations of rank

(e.g. assL V.P.) and 25 functions (e.g. finance)

A maximum of 20 unique standard industrial

classification (SIC) codes describing products or

services produced can be provided, though some

tape products may not contain all.

fall 1986



iassist quarterly

An administrative database containing

information on all publicly-held companies is

now available in machine-readable form from

Disclosure Inc. All companies whose stock is

publicly traded in the United States must file

annual and quarterly reports with the Securities

and Exchange Commission (SEC).

The annual forms — the 10-K for U.S. based

companies, 2(>-K for foreign companies — and

the 8-K quarterly forms have been available on

miCTofilm from Disclosure Inc. for more than

ten years. Data are extracted from these, as

well as from annual reports and proceedings of

annual meetings, and are now available on-line

from Disclosure, as well as on magnetic tape

and microdiskene.

Information on the chief executive officer and

directors of a corporation is also included. The

cash compensation awarded to the lop five

officers is provided, along with their names and

titles. Little other information on individuals is

available, however.

A recent entry into the tape product market is

the National Register Publishing Company
(NRPC), publisher of the Directory of

Corporate Affiliations, the Directory of

Advertisers, and the Corporate Bluebook of

Financial Executives. NRPC is an especially

good source of detailed information on

subsidiary units of corporations.

Tape files from NRPC are now available as

parallel products to each of the printed

directories currently produced. Each directory

product is tailored to one specific business

market Like Standard & Poor's and Dun &
Bradstreei, public data are supplemented by data

solicited directly from companies, and all data

are submitted annually to these companies for

verification.

Uses of Business Data on Tape

Three types of information may be of interest

to the purchaser of business databases on tape,

and the type that is of primary importance may
determine the choice of vendor. The types of

information are:

1. classificatory information, such as SIC

code, or even sales or assets of corporations,

2. names, titles and addresses of individuals,

most often required coded by title — both

level and function,

3. detailed financial information on

corporations.

Companies purchase data on tape for a variety

of reasons, but primarily for purposes related to

marketing. In addition, they may use these data

for analysis of financial and other information

for plannmg and comparisons relating to

investment, mergers, acquisitions and

divestitures.

Marketing activities involving use of such data

fall into three categories:.

1 production of labels for direct mail,

2 update and improvement of in-house lists,

3 analysis of information for marketing

comparisons.

There is some overiap among these activities.

Information may be used for mailing purposes

directly from the purchased tape, or may be run

against in-house computer files — a

"merge/purge" run — lo produce a

non-redundant mailing. This merge/purge run

ma> be done by the company's in-house EDP
department or may be subcontracted to a list

broker or service bureau. Dun & Bradstreet

Fall 1986



iasast quarterly - 5

offers this service to purchasers.

Data from purchased tape may be merged with

other machine- readable data, such as the

corporation's own list of clients, for update

purposes or for the addition of analytical

variables which the user company may not

collect and maintain, sub as sales, assets, or SIC

classification.

An example of the latter was a recent request

from one of our associate companies seeking

advice on matching SIC codes from a purchased

tape file with their in-house client hsL They

wanted to be able to analyze their client list

stratified by industry groups. Two problems

were encountered.

An alphabetic match on company name resulted

in only a 60% match with their own list. Use

of a unique identification number might have

facilitated a more complete match.

Also, the company was specifically interested in

SIC codes in the retail trade series. The dataset

they had purchased included 6 SIC codes per

corporation. Only the 6 SIC codes thai

described their primary- activities, reflected in

the proportion of the annual revenue generated

by those activities, were listed. There was a

distinct possibility that the codes in which they

were interested were not always included for

companies involved in retail trade activities.

Our own most recent use of a purchased tape

was in conjunction with a survey on corporate

benefits. The tape was acquired from Standard

& Poor's, and the SAS System was used to

produce mail labels (in a time-sharing system).

selecting subjects by title within certain industry

group and size parameters. We wanted to mail

a questionnaire to senior human resource

officers. Although a great variety of funcuons

were coded, that particular release of tape

product had a disappointingly low number of

high-ranking officers coded as this funcuon.

Only about 65% of the largest 1000 corporations

had an identifiable senior human resources

officer coded.

A final example is an investigation we
concluded recently on the feasibility of using

another external tape product The Conference

Board has recently begun to identify a second

level of major corporate leadership, the chief

executive officers of large subsidiary companies.

In our exploratory work we used the printed

version of the Directory of Corporate

Affiliations. A complimentary product is now
available on tape. This is the most

comprehensive source we have found of

information on levels of corporate ownership.

The vendor relies heavily on descriptions of

corporate levels provided by individtial

companies, which are polled annually. Because

of the variety of reporting procediu'es used by

the companies, it is difficult to extract reliable

lists of second-level executives which are

consistent from company to company. We have

resoned to checking the armual reports and

organizational charts that the Conference Board

collects annually from cooperating companies.

Many of the companies listed in the National

Register directory appear to be "paper" entities

rather than corporate profit centers. Several

may be headed by the same individual.

Preliminary work suggests that these tapes will

not be useful for our purposes.

Summar) of Problems and Solutions

Tv-pes of problems one encounters in using

commercial!) available business databases on

tape are:

a. the quality of the data

b. the appropriateness of the content of the

Fall 1986



6 - iassist quarterly

data for the intended purposes

c. logistic problems related to

implementation of the intended use, e.g.

matching external data with internally

produced data

d. format and documentation of the tape.

The industrial classification codes assigned vary

from tape source to tape source, as do printed

sources. Often, these codes are assigned by

clerical staff on the basis of reported

descriptions of business activities provided by

the responding company; these may be

incomplete or enoneous. They may also be

inconsistently or incorrectly interpreted by the

coders. The composition of the corporation may
change through merger, acquisition, divestiture,

or change in the revenues generated in specific

sub-units, so that the original "primary

SIC-code" no longer applies.

To facilitate merging information from an

externally purchased list with one's own
in-house company list, it may be advisable to

add a unique identification number to the

companies in one's own lisL The CUSIP
number, used by the SEC, the D-U-N-S
number assigned by Dun & Bradstreet, and the

FORTUNE number used in the Fortune Data

Bank, have all been incorporated into at least

one other database besides that of the generator

of the number. One or more ticker symbols

(stock market codes) may also be included, but

these are not consistent from exchange to

exchange. Sundard & Poor's includes the

CUSIP number; Disclosure includes both the

Fortune and D-U-N-S numbers, as well as the

CUSIP.

quality-checking.

Computer-produced, customized print lists and

labels are also available from these same

suppliers, and are frequently the most

cost-efficient way to purchase the information.

The list brokers use a variety of trade journal

subscription surveys and other resoiu'ces.

In order to update certain specialized lists, we
have purchased sets of labels or print-outs from

specialized vendors. Periodically, in order to

update our research mailing list of banks, we
have purchased a printed listing from R. L.

Polk & Co., whose primary research focus is on

banks, and compared it to our own in-house

list We find the primed copy easier to work

with, as we have limited database management
capability in-house. Our own database,

maintained on a Burroughs system, can not

readily accommodate external files.

We request from Polk's print-out of the 3,000

largest banks, in alphabetic order by state and

city, showing the total deposit income of each.

We then select for our list those banks with

total deposit income of e.g. over $500 million

(or $100 million or $1 billion). These we check

against our own in-house list, especially noting

possible bank mergers or name-changes. These

are easier to identify in listings ordered

geographically.

The most difTicult problems facing those who
would merge external information with internal

information are the problems of subsidiary units,

and, more recently, the even more complicated

problem of joint ventures.n

There are also list service houses which will

merge the data and do some specified checking,

based on text-matching of compan\ names.

This is far less successful than matching on a

unique identification number, which allows the

end user substantial control over

fall 1986