Professional Documents
Culture Documents
i. Commercial Surveys 9
ii. Audits 11
iii. Panels 12
3. Survey Research 15
The design of the research project specifies both the data that are needed and how
they are to be obtained. The first step in the data-collection process is to look for
1
Chapter 6: Primary and Secondary Data Sources
secondary data. These are data that were developed for some purpose other than
for helping to solve the problem at hand. The data that are still needed after that
search is completed will have to be developed specifically for the research project
and are known as primary data.
The secondary data that are available are relatively quick and inexpensive to obtain,
especially now that computerized bibliographic search services and databases are
available. The various sources of the secondary data and how they can be obtained
and used are described ahead.
Most secondary data are generated by specialized firms and are sold to marketers
to help them deal with a category of problems. Nielsen’s television ratings, which
marketers use in making advertising decisions, is the best-known example. Many of
these services, broadly categorized as audits, commercial surveys, and panels,
allow some degree of customization and thus fall between secondary and primary
data. These sources are treated in detail ahead.
Secondary Data
In April 1991, Buick and its advertising agency, McMann-Erickson Worldwide,
launched its new Roadmaster station wagon with a revolutionary new advertising
approach. A major component of the advertising for Roadster is a print campaign
with ads appearing in Time, Newsweek, U.S. News. & World Report, People, Sports
Illustrated, Entertainment Weekly, and Money. However, not all subscribers will see
these ads. In fact, only 4,940 of the more than 40,000 ZIP codes in the United
States will receive the ads. Subscribers is these ZIP codes will not only have a
chance to see the ads, their magazines' will come with a personal addressed card
inviting them to send for more information on the Roadmaster.
The target households, which are located mainly in affluent suburbs in the Northeast
and Midwest, represent less than 20 percent of US households. However, these
households buy over 50 percent of all large Station wagons. Buick was able to
select the appropriate ZIP codes by using McMann-Erickson McMapping database.
McMapping is based on data from several syndicated sources as well as the U.S.
Census. It describes ZIP codes (and larger areas) in terms of standard
demographics, values, primary lifestyle and media use. It works by matching the
characteristics of the firm's target market with the characteristics of ZIP code
residents.
2
Chapter 6: Primary and Secondary Data Sources
McMapping does more than simply allow precise targeting of ads and efficient
media buys. It also helps develop effective commercials. For example, a traditional
system might describe the typical buyer of a specific pickup truck as a mate between
25 and 54 with a household income of $30,000. The McMapping profile might add
such information as he lives alone, owns a dog, likes sports which he often watches
on cable with friends in a bar, and has a very macho self image. Obviously, this
added information would be invaluable in developing effective ad copy.
In a few short years the increase in the number of commercially available databases
and of computers on which to access them have brought about dramatic changes in
the utilization of secondary data. In this chapter we describe this development and
discuss the traditional sources of secondary data.
Availability
For some marketing problems, no secondary data are available. For example,
suppose J.C. Penney’s management was interested in obtaining consumer
evaluations of the physical layout of the company's current catalog as a guide for
developing next year’s catalog. It is unlikely that such information is available from
secondary sources. It is probable that no other organization that had collected such
data would be willing to make it available. Sears may have performed such a study
to guide in the development of their catalogs; it is, however, unlikely that a
competitor would supply it to Penney’s. In this case, the company would have to
conduct interviews of consumers to obtain the desired information. Secondary data
on the spending patterns, media preferences, and lifestyles of some population
3
Chapter 6: Primary and Secondary Data Sources
segments are very limited. For example, there is a shortage of data on African-
Americans, Hispanics, and Asian - Americans.
Relevance
Relevance refers to the extent to which the data fit the information needs of the
research problem. Even when data are available that cover the same general topic
as that required by the research problem, they may not fit the requirements of the
particular problem.
Four general problems reduce the relevance of data that would otherwise be useful.
First, there is often a difference in the units of measurement. For example, many
retail decisions require detailed information on the characteristics of the population
within the "trade area." However, available demographic statistics may be for
counties, cities, census tracts, or ZIP code areas that do not match the trade area of
the retail outlet.
A second factor that can reduce the relevance of secondary data is the necessity in
some applications to use surrogate data. Surrogate data are a substitute for more
desirable data. This was discussed earlier as surrogate information error. Had
Buick had access only to data on new car purchases by ZIP code, it would have
been much less relevant than data on purchases of new station wagons.
A third general problem that can reduce the relevance of secondary data is the
definition of classes. Social class, age, income, firm size, and similar category-type
breakdowns found in secondary data frequently do not coincide with the exact
requirements of the research problem. For example, Gallup and other public
opinion polls frequently collect data on alcohol consumption and attitudes toward
alcohol as part of their periodic surveys. Bacardi Imports would like to use this
readily available data. Unfortunately, Gallup and most other polls define adults as
individuals 18 and over while Bacardi is interested in adults 21 and over. The
different definitions of classes is one reason Gallup estimates that 56 percent of
adults “ever consume” alcoholic beverages compared to the 70 percent indicated by
Bacardi's surveys.
Accuracy
Accuracy is the third major concern of the user of secondary data. When using
secondary data, the original source should be consulted if possible. This is
important for two reasons. First, the original report is generally more complete than
a second or third report. It often contains warnings, shortcomings, and
methodological details not reported by the second or third source.
4
Chapter 6: Primary and Secondary Data Sources
Sufficiency
Secondary data may be available, relevant, and accurate, but still may not be
sufficient to meet all the data requirements for the problem being researched. For
example, a database that contained accurate, current demographic information on
the purchases of various brands and types of automobiles could still be insufficient
in terms of providing information to assist in developing new products or
advertisements.
Accounting Records
The basis for accounting records concerned with sales is the sales invoice. The
usual sales invoice has a sizable amount of information on it, which generally
includes name of customer, location of customer, items ordered, quantities ordered,
quantities shipped, dollar extensions, back orders, discounts allowed, date.
5
Chapter 6: Primary and Secondary Data Sources
Sales personnel often lack the motivation and/or the means to communicate key
information to marketing managers. To obtain the valuable data available from most
sales forces, several elements are necessary: (1) a clear, concise statement,
repeated frequently, of the types of information desired; (2) a systematic, simple
process for reporting the information; (3) financial and other rewards for reporting
information; and (4) concrete examples of the actual use of the data.
Miscellaneous Reports
Miscellaneous reports represent the third internal data source. Previous marketing
research studies, special audits, and reports purchased from outside for prior
problems may have relevance for current problems. As a firm becomes more
diversified, the more likely it is to conduct studies that may have relevance to
problems in other areas of the firm. For example, P&G sells a variety of distinct
products to identical or similar target markets. An analysis of the media habits
conducted for one product could be very useful for a different product that appeals
to the same target market. Again, this requires an efficient marketing information
system to ensure that those who need them can find the relevant reports.
Internal Experts
One of' the most overlooked sources of internal secondary data is internal experts.
An internal expert is anyone employed by the firm who has special knowledge.
Reuse data, stretch it out to the max and reserve your budget for truly new,
necessary primary studies. That's why we developed our "research library."
Everything we have ever done is in there, including subsequent actions and results.
And, it is organized for easy access. Now it is company policy that any research
request has to include proof that the library has already been searched and found
lacking-before any new can be conducted! related to the question at hand. While
6
Chapter 6: Primary and Secondary Data Sources
this knowledge is stored in individuals' minds rather than on paper or computer disk,
it can be as valid and valuable as more formal sources.
Had the marketing manager quickly asked the most obvious internal experts-
members of the sales force-to explain the sales decline, work on a competitive new
product could have begun almost a year earlier. In addition to the sales force,
companies have discovered that marketing research personnel, technical
representatives, advertising agency personnel, product managers, and public
relations personnel often have expert knowledge of relevance to marketing
problems.
Databases
Associations
Government Agencies
Federal, state, and local government agencies produce a massive amount of data
that are of relevance to marketers. In this section, the nature of the data produced
by the federal government is briefly described. However, the researcher should not
overlook state and local government data. There are also a number of specialized
analytic and research agencies, numerous administrative and regulatory agencies,
and special committees and reports of the judicial and legislative branches of the
government.
7
Chapter 6: Primary and Secondary Data Sources
These sources produce five broad types of data of interest to marketers. There are
data on (1) population, housing, and income; (2) agricultural, industrial, and
commercial product sales of manufacturers, wholesalers, retailers, and service
organisations; (3) financial and other characteristics of firms; (4) employment; and
(5) miscellaneous reports.
Syndicated Services
A wide array of data on both consumer and industrial markets is collected and sold
by commercial organizations.
Directories
External Experts
External experts are individuals outside your organization whose job provides them
with expertise on your industry or activity. State and government officials associated
with the industry, trade association officials, editors and writers for trade and
publications, financial analysts focusing on the industry, government and university
researchers, and distributors often have expert knowledge relevant to marketing
problems.
8
Chapter 6: Primary and Secondary Data Sources
Commercial surveys are conducted by research organizations and fall into three
categories: periodic, panel, and shared. Periodic surveys measure the same
attitudes,, knowledge, and/or behaviors using different samples at regular points in
time. Panel surveys generally measure differing attitudes, knowledge, and/or
behaviors using the same basic set of respondents at either regular or unique time
intervals. Finally, shared surveys are surveys that are administered by a research
firm and are composed of questions submitted by multiple clients.
Periodic Surveys
Periodic surveys are conducted at regular intervals, ranging from weekly to annually.
They use a new sample of respondents (individuals, households, or stores) for each
survey, focusing on the same topic and allowing the analysis of trends over time,
though changes in individual respondents cannot be traced. These surveys cover
topics ranging from values to media usage and food preparation.
Periodic surveys are conducted by mail, personal interview, and telephone. They
are subject to all of the problems of questionnaire design, sampling, and survey
method that affect custom surveys. In addition, when periodic surveys are
conducted at known intervals, they may affect the behavior being measured. For
example, periodic surveys are used to measure television viewing. Telecasters
have responded by scheduling specials and particularly popular shows to coincide
with these surveys
Panel Surveys
Panel surveys, sometimes called interval panels, are conducted among a group of
respondents who have agreed to respond to a number of mails, telephone, or,
occasionally, personal interviews over time. The interviews may cover virtually any
topic and need not occur on a regular basis. In contrast, a Panel, a continuous
panel or panel data refers to a group of individuals who agree to report specified
behaviors over time.
In an interval panel, the research firm initially gathers detailed data on each
respondent, including demographics and attitudinal and product-ownership items.
Because the researchers need not collect this basic demographic data again, they
can now obtain more relevant information from each respondent. These basic data
also allow researchers to select very specific samples, For example, a researcher
can select only those families within a panel that have one or more daughters
between the ages of 12 and 16, or that own a dog, or that wear contact lens. This
ability to select allows a tremendous swings over a random survey procedure if a
study is to be made for a product for teenage girls, dog owners, or contact lens
wearers, and so on.
It is possible to survey the same interval panel members several times to monitor
changes in their attitudes and purchase behavior in response to changes in the
9
Chapter 6: Primary and Secondary Data Sources
firm's or a competitor's marketing mix. However, interval panels are used more often
for cross-sectional (one-time) surveys. A major advantage is the high response rate
obtained by most interval panels. Return rates in the range of 70 to 90 percent are
often obtained. In addition, the firm does not have to generate a sampling frame, a
process that is both time consuming and costly. Finally, since panel members are
convinced of the legitimacy of the firm maintaining the panel, they may supply more
detailed and accurate data to both neutral and sensitive questions.
Data are normally collected by mail, but telephone, personal, and even focus groups
can be used. Clients can survey the entire panel, a stratified random sample of the
larger panel, or a specific type, size, or location category. Panel surveys obtain very
high response rates. However, the response rate when individuals are initially
asked to join a panel may be quite low. Thus, panels do not eliminate nonresponse
error. This issue is discussed in depth in the section on continuous panels.
Shared Surveys
Shared surveys offer the client several advantages. First, since several clients
share the fixed cost of sample design and most of the variable surveying costs, the
cost per question is generally quite low.
Since these data are collected frequently, responses can be obtained very quickly.
This feature is helpful for measuring consumers’ responses to competetive moves,
adverse publicity and environmental changes.
Audits
Audits involve the physical inspection of inventories, sales receipts, shelf facings,
prices, and other aspects of the marketing mix to determine sales, market share,
relative price, distribution, or other relevant information.
Store Audits
The simple accounting arithmetic of
The most -, widely-used store audit service is the Nielsen Retail Index. It is based
on audits every 30 or 60 days of a large national sample of food, drug, and mass
merchandise stores. The index provides sales data on all the major packaged
goods product lines carried by these stores-foods, pharmaceuticals, drug sundries,
tobacco, beverages, and the like (but not soft goods or durables). Nielsen contracts
10
Chapter 6: Primary and Secondary Data Sources
with the stores to allow their auditors to conduct the audits and pays for that right by
providing them with their own data plus cash.
The clients receive reports on the sales of their own brand and of competitors'
brands, the resulting market shares, prices, shelf facings, in-store promotional
activity, stock outs, retailer inventory and stock turn, and local advertising. These
data are provided for the entire United States and by region, by size classes of
stores, and by chains versus independents. The data are available to subscribers
on-line via computer as well as in printed reports
Product Audits
Product audits, such as Audits and Surveys' National Total Market Index, are similar
to store audits but focus on products rather than store samples. Whereas product
audits provide information similar to that provided by store audits, product audits
attempt to cover all the types of retail outlets that handle a product category. Thus,
a product audit for automotive wax would include grocery stores, mass
merchandisers, and drugstores (in this way it is similar to the Nielsen store audits).
In addition, it would include automotive supply houses, filling stations, hardware
stores, and other potential outlets for automotive wax.
Similar to store audits are retail distribution audits or surveys. These surveys do not
measure inventory or sales; instead, they are observational studies at the retail
level. Field agents enter stores unannounced and without permission. They
observe and record the brands present, price, shelf facings, and other relevant data
for selected product categories.
Panels
Retail Panels
A number of organizations offer services based on sales data from the checkout
scanner tapes of a sample of supermarkets and other retailers that use electronic
scanning systems. An estimated 99 percent of all packaged products in
supermarkets carry the universal product code (UPC), often referred to as a bar
code, and so are amenable to scanning. UPC codes are rapidly being expanded to
soft goods and hardware; stores such as K mart, Wal-Mart, and Toys 'R' Us have or
are installing scanners in all their outlets.
Scanning data have many applications in marketing research. Safeway Stores, for
example, has a manager of scanner marketing research whose department
conducts studies on such topics as price elasticities, placement of products in the
stores, and the effects of in-store advertising. One such scanner test showed that
the sales of candy bars increased 80 percent when they were put on front-end racks
11
Chapter 6: Primary and Secondary Data Sources
near the checkout stands. Another study indicated that foil-packaged sauce mixes
sold better when they were placed near companion products-spaghetti sauce near
the spaghetti, meat sauce in the refrigerated meat cases, and so on rather than
when they were displayed with other sauces.
Scanner data as compared to store audit data, have the advantages of (1) greater
frequency-weekly instead of bimonthly collection, (2) elimination of breakage and
pilferage losses being counted as sales, and (3) more accurate price information.
They have certain problems however, including (1) only the larger supermarkets
have scanners, and (2) the quality of the scanner data is heavily dependent upon
checkout clerk.
Consumer Panels
Continuous consumer panels allow firms to monitor shifts in individual or specific
household behaviors or attitudes over time. This allows the firm to determine how
its own or competitors' marketing mix changes affect specific consumers or market
segments Consumer panel data are collected either electronically, by UPC scanners
or Diaries.
Diary Panels
A diary panel, as the name implies, is a panel of households who continuously
record in a diary their purchases of selected products. it is used for those product
categories for which purchase is frequent-primarily food, household, and personal
care products.
Electronic Panels
Electronic panels are composed of households whose television viewing behavior is
recorded electronically. Nielsen Media Research is the main organization active in
this area. Until recently, Nielsen used a national sample of homes with TV sets that
were wired to household meters. The meters were connected to a central computer
by telephone line and automatically recorded when the set was turned on and the
station to which it was turned (a separate sample reported individual viewing in
diaries).
Single-Source Data
Single-source data are continuous data derived from the same respondent or
household, covering at least television viewing and UPC product purchase
12
Chapter 6: Primary and Secondary Data Sources
information. In general the data are collected electronically and also contain in-
store data such as price level, coupon use, and so forth. The advantages of such a
system are substantial, as it can produce virtually real-time measures of advertising
effectiveness, the effects of repetition, product changes, and so forth.
Retail Sales
Retail sales data a-re available from both audits and scanner-based retail panels.
Scanner panels provide more current data at shorter time intervals than do audits.
However, audits cover outlets not required with scanners. Scanner data are
particularly useful to both retailers and manufacturers for measuring the aggregate
impact of coupons, in-store promotions, point-of purchase displaces, price discounts
and so forth. Measuring only the sales of the promoted brand would lead a manager
to conclude that the fifth (least) most popular brand should be promoted. However,
an analysis of category sales reveals that sales increases of minor brands on sale
come as a result of the cannibalization of the more popular brands. In contrast, price
reductions on the leading brands appear to increase overall category sales.
Household Purchases
Data on household consumption are available from both diary- and scanner-based
household panels. Household consumption data allows the firm to monitor shifts in
an individual's or market segment's purchasing patterns over time. This allows the
firm to evaluate the effects of both its own and its competitors' marketing activities
on specific market segments. For example, if a competitor introduces a larger
package, the firm can tell what type (demographic and product usage
characteristics) and how many people are switching to the new size.
Household panel data also serve as an important basis for forecasting the sales
level or market share of a new product. A new product will often attract a number of
purchasers simply because it is new. However, its ultimate success depends on
how much of these initial purchasers become repeat purchasers.
Media Usage
Given the billions spent on advertising, it is not surprising that substantial effort is
expended to measure media usage.
Attitudes/Knowledge/Behaviors
Commercial surveys, both periodic and panel-based, are the primary general
sources of data on consumer attitudes, knowledge, and behavior. For example, a
firm desiring to improve or alter its corporate image could engage in a variety of
advertising and public relations programs in different regions of the country. Using
one of the weekly shared-interview services, it could economically determine the
relative impact on each approach over time.
13
Chapter 6: Primary and Secondary Data Sources
SURVEY RESEARCH
THE NATURE OF SURVEY RESEARCH
Survey research is the systematic gathering of information from respondents for the
purpose of understanding and/or predicting some aspect of the behavior of the
population of interest. As the term is typically used, it implies that the information
has been gathered with some version of a questionnaire. The administration of a
questionnaire to an individual or group of individuals is called interview.
TYPES OF INTERVIEWS
Direct interviewing involves asking questions such that the respondent is aware of
the underlying purpose of the survey. Most marketing surveys are relatively direct.
That is, although the name of the sponsoring firm is frequently kept anonymous, the
general area of interest is often obvious to the respondent. Direct questions are
generally easy for the respondent to answer, tend to have the same meaning across
respondents, and have responses that are relatively easy to interpret. However,
occasions may arise when respondents are either unable or unwilling to answer
direct questions. For example, respondents may not be able to verbalize their
subconscious reasons for purchases or they may not want to admit that certain
purchases were made for socially unacceptable reasons. In these cases, some
form of indirect interviewing is required.
Indirect interviewing, often referred to as disguised, involves asking questions such
that the respondent does not know what the objective of the study is. A person who
is asked to describe the “typical person" who rides a motorcycle to work may not be
14
Chapter 6: Primary and Secondary Data Sources
aware that the resulting description is a measure of his own attitudes toward
motorcycles and this use of them.
Both structure and directness represent continuums rather than discrete categories.
However, it is sometimes useful to categorize surveys based on which end of each
continuum they are nearest. This leads to four types of interviews: structure-direct,
structure-indirect, unstructured-direct and unstructured-indirect.
TYPES OF SURVEYS
Personal Interviews
Mall intercept interviews are the predominant type of personal interview. The
popularity of this type of personal interview is the result of its cost advantage over
door-to-door interviewing, the ability to demonstrate products or use equipment that
cannot be easily transported, greater supervision of interviewers, and less elapsed
time required.
Individuals who visit malls are not a representative of the entire population. An
additional problem with intercept interviews at malls where research firms maintain
permanent facilities is "respondent burnout," That is, a significant portion of a given
mall's customers shop at the mall regularly. over time, these regular shoppers will be
randomly. selected into numerous studies. Both their willingness to cooperate and
the nature of their responses will change as the% participate in more and more
studies.
15
Chapter 6: Primary and Secondary Data Sources
Intercept interviews are not limited to shopping malls. Increasingly, interest
interviews are conducted at locations relevant to the population of interest. An
emerging type of personal interviewing is the in-store intercept. In-store intercept
interviews involve interviewing individuals inside retail outlets, generally immediately
after they have purchased the product category in question. One version of this
approach, the purchase intercepts technique (PIT).
Telephone interviews
Telephone interviews involve the presentation of the questionnaire by telephone.
Computer-assisted telephone interviewing (CATI) dominates large-scale telephone
interviews. A stand-alone CATI system involves programming a survey directly in
one or more personal computers. The telephone interviewer then reads the
questions front a television-type screen and records the answers directly on the
terminal keyboard or directly on the screen with a light pen. The flexibility associated
with the computer provides a number of advantages. often exact set of questions a
respondent is to receive will depend on answers to earlier questions. For example,
individuals who have a child under age three might receive one set of questions
concerning food purchases whereas other individuals would receive a different set.
The computer, 4 in effect, allows the creation of an "individualized” questionnaire for
each respondent based on answers to prior questions.
A second advantage is the ability of the computer to present different versions of the
same question automatically. For example, when asking people to answer
questions that have several stated alternatives, it is desirable to rotate the order in
which the alternatives are presented. This is easy with a CAT1 system.
Another advantage of CATI system- is the ease and speed with which a bad
question can be changed or a new question added. CATI systems also edit data as
they are entered. That is, the computer can be programmed to highlight
inconsistent answers across questions, to refuse answers outside a defined range,
to ensure that constant sum question responses total properly, and so forth.
Finally, data can easily be analyzed and interim reports issued. Interim reports may
allow one to stop a survey if the "answer" becomes clear before the scheduled
number of interviews has been completed. Final reports can also be produced
rapidly.
Mail Interviews
Mail interviews may be delivered in any of several ways. Generally, they are mailed
to the respondent and the completed questionnaire is returned by mail to the
researcher. However, the forms can be left and/or picked up by company personnel.
They can be also distributed by means of magazine and newspaper inserts or they
can be attached to products. The warranty card attached to most Consumer
products as a useful source of survey data for many manufacturers.
A of criteria are relevant for judging which type of survey to use in a particular
situation. These criteria are
16
Chapter 6: Primary and Secondary Data Sources
(1) complexity,
(2) required amount of data,
(3) desired accuracy,
(4) sample control,
(5) time requirements,
(6) acceptable level of Nonresponse and
(7) cost.
A respondent seeing a questionnaire of this type for the first time can easily
become confused or discouraged. Thus, computer, personal, and telephone
interviews are better suited to collect this type of information than are mail
interviews.
Other aspects of complexity also tend to favor the use of personal or computer
interviews. Visual cues are necessary for many projective techniques, such as the
picture response. Multiple-choice questions often require a visual presentation of
the alternatives because the respondent cannot remember More than a few when
they are presented orally. However, most attitude scales can be administered via
the phone.
The telephone, and often mail, are inappropriate for studies that require the
respondent to react to the actual product, advertising copy, package design, or other
physical characteristics. Techniques that require relatively complex instructions are
best administered by means of personal inter-views. Similarly if the response
required by the technique is -extensive, such as with many conjoint analysis studies,
personal inter-views are better, with computers second.
Amount of Data
Personal interviews can, in general, be longer than other types. Social motives play
an important role in personal interviews. It would be "impolite" to terminate an
interview with someone in a face-to-face situation.
17
Chapter 6: Primary and Secondary Data Sources
The accuracy of data obtained by surveys can be affected b a number of factors,
such as interviewer effects, sampling effects, and effects caused by questionnaire
design. In this section, we are concerned with errors induced by the survey method
itself, particularly responses to sensitive questions and interviewer effects.
Sensitive Questions
Interviewer Effects
Questionnaire designs that minimize interviewer freedom also reduce the potential
for interviewer bias. The most effective approach involves the skillful selection,
training, control of interviewers. However, after the most cost effective design
principles have been applied, some interviewer bias is apt to remain. This should be
estimated subjectively or, preferably, statistically.
One final problem that arises with the use of telephone and personal interviewer
cheating. That is for various reasons, interviewers may falsify all or parts of an
interview. This is a severe enough problem that most commercial survey
researchers engage in a . process called validation or verification.
Validation involves reinterviewing a sample of the Population that completed the
initial interview. In this reinterview, verification is sought that the interview took
place and was conducted properly and completely.
Sample Control
Each of the four interview, techniques allows substantially different levels of control
over who is interviewed. Personal interviews offer the most potential for control over
the sample. An explicit list of individuals or households is not required. Although
such lists are desirable, various, forms of areas sampling can help the researcher to
18
Chapter 6: Primary and Secondary Data Sources
overcome most of the problems caused by the absence of a complete sampling
frame. In addition, the researcher can control who is interviewed within the sampling
unit and how much assistance from other members of the unit is permitted.
Lists of specialized groups are more readily available However even with a good
mailing list, the researcher maintains only limited control over who at the mailing
address completes the questionnaire. Different family members frequently provide
divergent answers to the same question. Although researchers can address the
questionnaire to a specific household member, they cannot be sure who completes
the questionnaire.
19
Chapter 6: Primary and Secondary Data Sources
To ensure more representative samples, researchers generally utilize some form of
random digit dialing. This technique requires that at least some of the digits of each
sample phone number be generated randomly.
A primary problem with pure random digit dialing is that only about 20 percent of all
numbers within working prefixes are actually connected to home phones. A variety
of techniques have been developed to minimize this problem. The most popular
technique, Plus-one or add-a-digit, simply requires the researcher to select a
sample from an existing directory and add one to each number thus selected.
Although the technique is more expensive than a sample selected directly from a
directory, and it has a higher refusal rate,' it produces a high contact rate and a fairly
representative sample.
Time Requirements
Telephone surveys generally require the least total time for completion. In addition,
it is generally easier to hire, train, control, and coordinate telephone interviewers.
Therefore the number of interviewers can often be expanded until any time
constraint is satisfied. the number of personal and computer interviewers can also
be increased to reduce the total time required. However, problems with training,
coordinating, and control tend to make this uneconomical after a certain point.
Because “at home’ interviewers must travel between interviews and often set up
appointments such interviews take substantially more time than telephone
interviews, however mall intercept interviews can be done fairly rapidly.
Non response error is a potential problem for telephone, personal, and computer
interviews. Not-at-homes and refusals are the major factors that reduce response
rates. The major focus in reducing non response in telephone and personal
interview situations has centered on contacting the potential respondent. This was
based on the belief that the social motives that are present in a face-to-face or
verbal interaction, situation operate to minimize refusals. However, refusal rates are
increasing for both personal and telephone interviews. Therefore, researchers
Must focus attention on gaining cooperation from, as well as making contact with,
potential respondents.
Contacting Respondents
The minimum number of calls in most consumer surveys should be three, and
Callbacks should generally be made at varying times of the day and on different
days of the week. There is, as one might suspect, a definite relationship between
20
Chapter 6: Primary and Secondary Data Sources
both the day of the week and the time of day and the completion rate of telephone
and personal interviews.
Commercial survey research firms vary widely in the number of times they allow a
phone to ring before dialing the next number. Some allow only three rings, whereas
others 90 as high as ten. One study indicates that five rings may be optimal.
Motivating Respondents
Refusals are a problem in telephone and personal surveys. Most refusals occur
immediately after the introductory remarks of the interviewer. After they begin, very
few interviews are terminated prior to completion.
Likewise the length of the interview has a significant impact. Gender of the
interviewer does not appear to affect the refusal rate, but characteristics of the
interviewer’s voice do. Prior notification by letter lowers the refusal rate for
telephone surveys. Likewise, prior notification by telephone increases the
cooperation rate to an at-home personal interview survey. The sponsor of the survey
affects telephone response rates with the rate being higher for university and charity
sponsors than for commercial sponsors.
The promise of a large monetary incentive ($10) was effective in generating a high
response rate to a telephone survey that required respondents to agree to watch a
specific television program.
Attempts to gain cooperation for long or complicated interviews occasionally use the
foot-in-the-door technique. This technique involves two stages. First, respondents
are asked to complete a relatively short, simple questionnaire. Then, at a later time,
they are asked to complete a more complex questionnaire on the same topic. This
technique generally produces at least a small gain in the response rate. However,
given the added expense this involves in telephone and personal interviews,
concentrating on Ion techniques and calbacks may provide a higher payoff.
Refusal conversion or persuasion has been found to increase the overall response
rate by an average of 7 percent. This involves not accepting a no response to a
request for cooperation without malting an additional plea. The additional plea can
stress the importance of the respondent's opinions or the brevity of the
questionnaire. It may also involve offering to re contact the individual at a more
convenient time finally, the time of day that contact is made appears to influence the
refusal rate. Paradoxically, while evening is the optimal time to find respondents at
home, it also generates the highest level of refusals.
Predicting Response
Most mail surveys produce similar response patterns. However, the speed of
response and ultimate percentage responding can vary widely.Researchers can
conduct small scale preliminary mailing to a subsample of their target respondent.. If
a pilot study is not practical perhaps because of time pressure the observed
21
Chapter 6: Primary and Secondary Data Sources
response pattern to earlier similar surveys among similar respondents using similar
respondents using similar response inducements can be used.
Attempts to increase the response rate to mail surveys focus on increasing the
potential respondents motivation to reply. Two complementary approaches are
frequently used. The first is to increase the motivation as such as possible in the
initial contacts with respondents. The second approach is to remind the respondents
through repeated mailings or other contacts. The initial response rate to a mail
survey is strongly influenced by the respondents interest in the subject matter of the
survey. Interest level can be a serious source of non-response bias in the survey
results.
Pre-notification, such as a letter or telephone call that informs the respondents that
they will receive a questionnaire shortly and requests cooperation, is a cost effective
means of increasing response rates. In the absence of monetary inducements, a
number of studies have found pre-notification to double the response rate obtained
without pre-notification. This technique works best with the general public, but it is
also effective in industrial surveys. Evidence suggested that a preliminary letter or
card is more effective than a preliminary phone call. The type of postage has a
moderate impact on the response rate. First-class, hand-stamped outgoing and
return envelopes produce higher response rates than do metered, second-class, or
business reply envelopes. This impact is greatest for return envelopes, where it is
clearly a cost-effective technique.
Lottery incentives have been found to have mixed results. Other types of incentives
are generally more effective.
The effect of gift incentives such as pens or key rings is generally positive but very
moderate. Like cash incentives, gift incentives lose most or all of their effectivess
when they are promised rather than provided with the questionnaire.
The degree of personalization and the related variables respondent anonymity and
assurances of confidentiality produce variable effects on both response rates and
accuracy. Personalization appears generally to increase response rates on non-
sensitive issues, whereas assurances of anonymity or confidentiality are most
effective on questionnaires dealing with personally important or sensitive issues.
However, these effects are generally small.
22
Chapter 6: Primary and Secondary Data Sources
The identity of the survey sponsor influences the response rate, with commercial
sponsors generally receiving a lower response rate than noncommercial sponsors.
The type of appeal used in the cover letter can take a number of approaches, such
as egoistic (your opinion is important), altruistic (please help us), social utility (your
opinion can help the community), or negative (if the questionnaire is not returned by
a certain date, a telephone call or personal follow-up will result). Evidence indicates
that the "best" appeal depends on the nature of the sponsor and purpose of the
study, though negative appeals appear to be dysfunctional.
Subjective Estimates
When it is no longer practical to increase the response rate, the researcher can
estimate subjectively,, the nature and effect of the non-respondents. That is, the
researcher, based on experience and the nature of the survey makes a subjective
research evaluation of the probable effects of the non-response error. For example,
the fact that those most interested in a product are most likely to I return a mail
questionnaire gives the researcher some confidence that non-respondents are less
interested in the topic than respondents.
Imputation Estimates
Trend Analysis
Trend analysis is similar to the imputation technique, except that the attributes of the
non-respondents are assumed to be similar to a projection of the trend shown
between early and late respondents. However, trend analysis should only be used
when there are logical reasons to believe the trend will extend to the non-
respondents.
24