iassisl quarterly - 11

The POLL Database:
Roper Center's Online Source

for Public Opinion Research

by Linda Langschied

Rutgers University

College Avenue Campus
New Brunswick, New Jersey 08903

communications and mass media research. Not

only an archival facility, the Roper Center

provides its clients with services such as data

analysis and interpretation, and searches of the

archive. The thousands of scholars who have

made use of the Center's data archives and

services have tappped a rich source of

machine-readable data, which is stored on

magnetic tape.

In 1980, the Roper Center began planning a

new service, an online retrieval system that

would allow researchers, with the use of a

computer and modem, to tap directly a database

that would contain survey questions and

responses. In 1983 this system, POLL, was

constructed and is now available by subscription

to researchers across the U.S. and

internationally. The content of POLL differs

from machine-readable data files in that the

POLL database contains no raw data. Rather,

POLL approximates a bibliographic database,

with each record giving all the information

necessary to form a complete citation. Searchers

retrieve, upon entering a topic, three basic kinds

of information: the texts of poll questions

related to that topic, the responses to each

question, and "study level" information: the

dates the poll was conducted, by whom, the

type of sample, and so on. The following is an

example of a record, which illustrates the

various fields of information available:

the Roper Center and POLL

Established in 1946, the Roper Center contains

the largest archives of public opinion research

data in the worid. This collection includes the

basic data from over 10,000 public opinion

surveys conducted since 1936, which have been

gathered from over 40 major United States

suppliers, and over 70 foreign counuies. The
Roper Center classifies approximately 65% of its

materials as public affairs studies. 20% as

market and consumer surveys, with the

remaining studies focusing primarily on

Question:

R12 The Reagan Administration has proposed

giving $100 million in military, medical and

economic aid to the rebels fighting the

Sandinista government in Nicaragua. Do you

favor or oppose this proposal?

Responses:

Favor 33%

Spring J 987



12 - iasstsl quarterly

Oppose 54%

Not sure 13%
Survey Organization: NBC/Wall Street Journal

Population: National adult

Population siz,e: 1599

Interview method Telephone

Beginning date: APR 28, 1986 Ending date:

APR 29, 1986

Source Document: NBC News/Wall Street

Journal

Date of Source Document: MAY 12, 1986

Subject: LATIN
DIPLOMACY
PRESIDENCY

FLILL QUESTION ID: USNBCWSJ.051286.R12

The need for an online system like POLL that

would provide accurate and inexpensive data to

their clientele was very apparent to the Roper

staff who conducted manual searches of the

archives for clients. Prior to the start of the

online service, the stafT created manual subject

indexes of the reports, or used indexes that

were sometimes supplied by the contributing

survey organizations. To conduct a search for a

client on a particular topic, in a specific time

frame, was very labor intensive: many times this

included having to sit down and read the entire

report Data were then compiled for patrons

through a great deal of cutting, pasting, and

photocopying. It was not uncommon for the

Roper staff to have to pull the same study ten

times within a year and document the same

things ten times. Obviously, this degree of

labor intensiveness made jobs very costly for the

paying patron. For many people, including

academics, with limited financial resources, the

cost was out of reach. In the online system, the

data are simply input once, in a standard

format, and are readily available. Through

POLL, Roper is meeting its goal to provide

easy access to data which are timely and

accurate, while holding down costs, both

internally and to the clients it serves.

Scope of the POLL Database

Researchers using POLL should have a good

understanding of the types of questions

contained in the database. POLL incorporates

all surveys with national samples that come into

the archives, from most of the major polling

organizations, such as Gallup, Roper,

Yankelovich, Harris, and major media like ABC,

CBS, NBC, Los Angeles Times, and many more.

POLL includes onlv national surveys. Although

the Roper Center does archive data files from

international surveys, such as the U.S.

International Agency Surveys, and state-level

studies such as the Minnesota Poll, these are

not added to POLL. Roper Center has,

however, begun planning to create a separate

database of state-level surveys, though initiation

of this database is still several years down the

road.

There are two types of study which are

undertaken by the polling organizations and

entered into POLL. Most organizations which

store their survey results at the Roper Center

conduct omnibus surveys—ongoing surveys

which measure changes in attitudes over a

period of time. In addition, omnibus surveys

cover many different types of political data or

policy issues in one study. The second type of

study is the special, one-time study. Typically,

these studies focus on a specific, major topic:

"As a result of the weapons deal (with Iran and

the Contras in Nicaragua) that is now coming to

light, do you think the U.S. Congress should

cut back on American military aid to Israel, or

do you think Congress shouldn't do that?" A

single search in POLL will find both types of

studv.

Spring 1987



iassisl quarterly - 13

Who Uses POLL?

The main users of POLL are from academia,

the media, business, and even polling

organizations, who may use POLL to design a

study, or to compare their results to those of

other organizations. Academic institutions may

also be members of ISLA (International Survey

Library Organization), the cooperative,

educational arm of the Roper Center. ISLA

members enjoy reduced rates for many of the

Roper Center's services, including the use of

POLL. Annual subscription rales for POLL
vary according to member/non-member status

and whether the institution receives the reduced

ISLA rate.

The Roper Center is trying to alert users,

especially those in academia, of the value of

making POLL available to students for

classroom use, particularly in disciplines such as

sociology, political science, journalism, and any

of the increasing number of academic

departments that utilize public opinion research

as part of student coursework. The Center is

encouraging institutions to take out a POLL
account, and give students direct access to the

database. At present. Roper Center is in the

process of negotiating a subscription to one

institution for the purpose of direct access for

its students. To open such a subscription

requires a great deal of planning on the part of

both Roper and the institution, mainly because

of the internal bookkeeping required by

allowing general access.

The Plan for Building the POLL Database

The stafT at the Roper Center have set entering

the most current data as their highest priority,

with omnibus surveys taking precedence over

special, one-time studies. Typically, when an

omnibus study comes in, unless it is

extraordinarily large, it is entered into the

database within two weeks, so that the

information is very cunent Special studies are

entered as quickly as possible, though these

surveys are more time-consuming to enter than

the omnibus studies. At present, the entry of

omnibus studies is almost complete for all

studies conducted from 1974 to the present

At the same time that current data are being

loaded, Roper Center is retrospectively entering

data from older surveys. Ultimately, the

database will include information going back to

1937. An average of 500 questions a week are

entered into the system: over 79,000 questions

have been entered thus far. No specific target

date for the completion of the retrospective

entry project has been set Predicting how long

it will take to enter all national surveys is

difficult, as the number of collections available

in the eariier years decreases. Still, it will be

several years before the retrospective entry of

national poll data is completed.

How Can POLL Aid Research?

A researcher may need to use POLL only to

satisfy a public opinion research question. For

example, a researcher interested in the question

of whether people think children should be

taught sex education courses might find this

kind of information through a POLL search:

Spring 1987



14 - lassisl quarterly

Qucsiion:

Q14G Do you think that public elementary

schools in this community should or should not

teach sex education in grades 4 through 8? (If

favor sex education in grade 4 through 8, ask.:)

Should this program include discussions about

AIDS (Acquired Immune Deficiency Syndrome),

or not?

Responses:

Favor sex education/Include discussions

of AIDS 67%
Favor sex education/Don't include

discussions of AIDS 4%
Oppose sex education 21%
No opinion 8%
Survey Organization: Gallup Organization

Population: National adult

Population Size: 503

Interview method: Telephone

Beginning date: FEB 9, 1987 Ending date: FEB
25. 1987

Source Document: Gallup Poll

Date of Source Document: MAR 22, 1987

Subject: SEX
EDUCATION
HEALTH
FULL QUESTION ID: USGALLUP.032287.R1

However, if a more detailed analysis than this is

needed, e.g. a breakdown of responses by age,

race, men vs. women, etc., the machine-readable

data files must be tapped. The researcher,

having isolated the correct studies through a

POLL search, must contact the Roper Center to

establish whether the Center has the

corresponding data sets. These data sets contain

the raw data of individual responses to survey

questions, and are stored on magnetic tape.

POLL, in this example, has served as a

reference tool, leading the researcher to the data

sets which are rich in information on the his

topic.

Roper generally does have the data sets referred

to in POLL, (though there arc some exceptions,

e.g. the Harris Poll, which is archived at the

University of North Carolina.) The Center

receives the data directly from the contributing

survey organizations, though the lag time for

receipt of the sets varies greatly from one

organization to another. Some organizations

send out six months worth of data sets at a

time, while others send their data each time a

stud\ is conducted. In addition, academic

researchers whose institutions are members of

the ISLA, may find that their institution has

purchased the tapes from Roper, and that they

may access the files on their own campuses.

How is POLL Searched?

Anyone familiar with library card catalogs or

printed periodical indexes understands that they

may hunt for books or articles using set "fields"

of information—author, title, or subject

Researchers who have made the transition from

print catalogs and indexes to online

searching—whether searching a library's online

catalog or the databases of a commercial

information vendor, such as Lockheed

Information Systems' DIALOG. System

Development Corporation's ORBIT, or the

Bibliographic Retrieval Services (BRS)— find the

possibilities for searching are greatly broadened

through the use of the computer. The computer

can enable them to seek out, for example,

individual words imbedded in a title or abstract

of a book or article, or to limit their findings to

a specific year of publicauon. Perhaps most

important, an online search allows the

combination of two or more distinct concepts in

a single search (e.g. learning disabilities and

college students), thus retrieving very specific

search results.

Spring J i



lassisl quarterly - 15

POLL is, of course, a unique database, and

while ihe fields available for searching differ

from the usual bibliographic format, the same

search concepts and strategies still apply.

Following the instructions in the User's Manual

for POLL, which is distributed to subscribers,

the researcher can execute a search with the

knowledge of some basic commands. The

following is a very brief explanation of the

methodology used in searching POLL:

1. The basic command for searching the POLL
database is the "FIND" command, which

may be abbreviated as "FIN".

2. The searcher must indicate what kind of

search is to be executed, by choosing from

the fields, or "indices" available:

SUBJECT: Topic(s) assigned to question

WORD: Actual text of questions and

responses

ORGANIZATION: Organization conducting

the study

DATE: Beginning date of the study

Examples of the general form that a search

might take are: HNd NAME OF INDEX
ENTRY IN INDEX
HN WORD REAGAN
HN SUBJECT ETHICS
HN ORGANIZATION GALLUP
HN DATE 12/18/85

The User's Manual lists, in its appendices, all

subject category codes and definitions, and

organization codes. A cautionar}' note:

subject categories are very broad, and

searchers should not assume thai their topic

constitutes an official subject For example,

a search done on Reagan as "SUBJECT"
will net z,ero hits, while the same search

done m the "WORD" index will retrieve

over 6,000 items.

3. Searchers may use the truncation symbol
"#" to pick up all forms of a word. A
search of the word "librar#" will find items

containing the words "library,"

"librarian."

"libraries," or

4. The searcher must wait for an arrow to

appear on the screen before entering a

search. The prompt, "->" means that the

system is ready for searching.

A simple secirch might look like this:

-» fin word librar#

-RESULT: 93 items

The above example constitutes the simplest kind

of search that might be done in POLL The

system also supports more sophisticated search

strategies, utilizing standard Boolean protocols

(using AND, NOT, and AND NOT) to combine

two or more search terms into a single

statement:

-»fin word reagan and subject diplomacy

-RESULT: 62 items

Or, items may be "nested" within parentheses

to indicate the order in which terms are to be

searched. Ordinarily, the system searches items

in the order in which they were entered, from

left to right, and from lop to bottom. Nesting

terms makes the system search the items within

parentheses first, and then combine the result

with items outside the parentheses. Here is an

example of a fairly complicated nesting strategy:

-»fin word boycott* and ((south and

africa) or aparthied)

Searches may also be entered step-by-step. The

same search might be done like this:

-»fin word south and africa

-RESULT: 17 items

->X)r apartheid

-RESULT: 38 items

-»and boycott

-RESULT: 9 items

Spring 1987



16 - lassist quarterly

These examples just begin lo touch on ihe

possible search techniques that may be used to

search POLL. POLL is designed to be flexible,

and accurate search results may be obtained

through many different avenues. This flexibility

makes POLL a relatively easy database to access

and search. With the aid of the User's Manual ,

and a little practice, researchers using POLL can

achieve proficiency in a fairly short time.

Once a search is completed, the user may view

the records on the screen by issuing a "TYPE"

command. In order obtain hardcopy results of

the search, the searcher may want to print the

results directly, as the search results scroll by on

the screen. In this case, it may be advisable to

first "sort" the search results, eg. by date, by

using POLL'S "SEQUENCE" command, before

entering "TYPE." Otherwise the records will be

printed out in random order. Another

possibility is to download the results directly to

disk. These results can be edited later with any

word-processing software that can handle a

standard ASCII file. In the event that a search

should result in an extremely large number of

items, using these techniques may not be

practical. The file can instead be printed at the

Roper Center. The searcher issues the

command "..POLLPRT", sending the search

result directly to the Roper Center, where it is

printed and then mailed to the searcher. Roper

Center does charge for this service, and rates

vary according to the size of the file.

doing searches on it at patron request, and

charging a fiat fee to cover some of the cost of

connect time and telecommunication charges.

Thus far, we have had two requests from

graduate students in political science for POLL
searches—one on South Africa and apartheid,

and the other on the attitudes of gays on

political elections. The reactions of these two

students were favorable to POLL as a system,

although the student researching the attitudes of

gays found little relevant materia! in POLL.

As the coordinator of online reference services

at Alexander Library, I do much of the

database searching, and am the person most

familiar with POLL. Overall, I find il

extremely helpful in locating polling results— it

is an easy-to-use resource, and it produces

results far faster than paging manually through

indexes. It is also a relatively inexpensive

database, costing $15.00 per hour of connect

time for ISLA members. (In comparison, we

often search databases on the DIALOG system

which cost upwards of $100.00 per hour of

connect time, plus $1.00 or more for each

citation—and there are many databases which

cost much more than this.) One minor

drawback of POLL, financially speaking, is that

it must be accessed directly in Storrs,

Connecticut, rather than through a

telecommunications network like TYMNET or

TELENET, so telephone charges can be rather

high. Still, when one considers the alternative

of manual searching, POLL more than pays for

itself.

Conclusion

Rutgers University, a participating ISLA

institution, has been a subscriber to POLL since

September of 1986. Rutgers handles its POLL
activities through the Alexander Library, the

graduate library for social sciences and

humanities research. The librarians at

Alexander ueat POLL like anv other database.

I would also like to add that one of the chief

strengths of POLL is the support that Roper

Center gives to POLL users. New users are

bound to have problems initially, with

determining their software communications

parameters, with bad phone lines, or with

signing on to the system. At Alexander Library,

wc encountered all of these problems, and each

time the technical staff at Roper analyzed our

situation over the telephone, and coached us

through to a solution. Anyone interested in

Sprmg 1987



lasstsl quarierly — 17

contaciing ihc Roper Center with quesuons

aboul POLL may do so by writing lo Marilyn

Potior, Assistant Director for User Services and

Administration, Roper Center for Public Opinion

Research, P.O. Box 440, Storrs, CT 06268, or

may call at (203) 486-4440.

*A11 background information on POLL was

obtained directly from Marilyn Potter and

Sterling Green, staff members at the Roper

Center.n

Spring 1987