The general population census is an account of every member of the population. We have very early examples from the Babylonians and the Chinese, who created accounts of their people for tax and military reasons. The Egyptians also needed knowledge of available manpower to plan the building of the pyramids (ONS 2001). From 5 bc, a census was held every five years across the Roman Empire. In order to eliminate errors due to mobility, every man in the Empire was obliged to return to his place of origin.
In the British Isles, a detailed inventory of land and property was completed in 1086, known as the Domesday Book, and this was the result of several years of work. There are other examples of early census studies: 1666 (Quebec), 1703 (Iceland), 1749 (Sweden). In the 1700s, the census was resisted in the USA and Britain for various reasons: one was a fear that a census ‘might incur the
wrath of God’; another was that foreign enemies would be able to detect weaknesses or ‘individual liberty would be impaired’. Despite these objections, regular census-taking began in 1790 in the
USA, and in 1801 in Britain.
Because the census is an account of every member of the population, it relies on cooperation from the public. Resistance is usually overcome by explaining that statistics will mask individual identities. Refusal to complete the UK 2001 census was met by an offer of prosecution and a fine
of up to £1000. Indeed, 38 people were prosecuted and fined in the 2001 UK census; one person was imprisoned for refusing to pay the fine.
The disadvantage of the census is that a great deal of skilled manpower is required. Leading members of the ‘parish’ were used to take the UK 1801 census and, in 1841, some 35,000 male enumerators were used. The US 1880 census used 31,382 interviewers and the 1960 US census required 160,000 enumerators. For the 2011 census in the UK, 37,000 people were employed and
3 million reminder letters were posted. A large budget is needed and it can take a long time to complete: in the 2001 UK census, the programme plan was for 13 years (1993–2006) and the cost was estimated to be £254 million. The cost of the 2011 census was estimated to reach £480 million, the increase down to inflation and an extra 3.5 million people to count, mainly immigrants.
Compiled by Nigel Bradley 2012.
Sources:
HMSO (2010) The 2001 Census of Population, www.ons.gov.uk/ons/guide-method/census/census-2001/index.html Mouncey, P. (2011) 2011 Census: a CGG Seminar. IJMRS 53(5), pp. 569–570.
Office for National Statistics (2001) 200 Years of the Census. London: ONS, http://www.ons.gov.uk/ons/guide-method/ census/2011/census-history/200-years-of-the-census/index.html
Questions
1 Why do most nations carry out a census regularly?
2 Who benefits from census results?
3 Besides prosecution in a law court, what steps could be taken to encourage participation?
© istockphoto.com/Ian Jeff ery
Part 2 Data collection
150
Sampling does not always select ‘people’
to be questioned; sometimes, ‘situations’
or ‘locations’ are sampled. In observational
research, ethnography, and action research, it is
probable that neither situations nor people can
be predicted, that new situations will introduce
themselves, and that people will enter and exit
from the research project. Additionally, some
research naturally moves to new locations. We
may take a sample of speech to illustrate points
made in a qualitative study; in secondary data
analysis, we may sample certain documents, but
not all. Therefore sampling must relate back to
the purpose of the study. We may be sampling:
• People (as individuals or as groups)
• Time (in terms of minutes or days)
• Places (public or private)
• Behaviour (in terms of individual events or states)
• Items (documents in archives).
Time sampling can off er some useful data, but the disadvantages can be that the period
sampled is not ‘representative’ of the full behaviours. Sampling called ‘behaviour’ above can be
usefully classifi ed ad lib sampling, focal sampling, all-occurrence sampling, and scan sampling.
We can thank Altmann (1974) for these distinctions.
1. Ad libitum sampling A record is made of as much information as possible. This attempts
to monitor all activities. Such observations will always be biased by the behaviours,
individuals, or situations that most attract the observer’s attention. It is costly and time-
consuming; it can be used as a qualitative phase or to plan a study.
2. Focal sampling All occurrences of specified actions of one individual are recorded
during a certain time period, often 60 minutes. The advantage is that unbiased data can
answer numerous questions.
3. All-occurrence sampling The observer focuses on a particular behaviour, rather than a
particular individual. For example, we might count the number of requests for information
coming to a helpdesk in a supermarket. This can give a quantitative measure of the rate of
occurrence of behaviour.
4. Instantaneous or scan sampling A subject’s activities are recorded at predetermined
instances, such as every 45 seconds. It is a ‘sample of states’ and is used to study the
percentage of time spent in a certain activity. If the behaviours of all members of a group
are surveyed within a short period of time, we call it scan sampling.
Social classification
Social classifi cation or grading is a useful indicator of predisposition to consume specifi c products
and services. At the most basic level, we might use age, sex, and terminal education age (TEA).
Chapter 5 Sampling
151
The attempt will be to fi nd divisions in a society that are not likely to change and that can indeed be identifi ed. Classifi cations then become more complex as we incorporate other aspects such as educational attainment, occupation, family stage, social standing, and income. These classifi cations are a combination of objective and subjective measurements. Despite any problems with the validity and reliability of such measures, they are useful because there are correlations with aspects that are less apparent, but important for marketing. For example, clothing, fashion, shopping, leisure, saving, and spending are similar for specifi c groups in society.
In the UK, there are several established systems that are associated with government surveys: socioeconomic groupings (SEG), social class, standard occupational classifi cation (SOC), and other variants. A newer classifi cation, called socioeconomic class (SEC), was released in 1998 and validated on the Labour Force Survey (see Rose and O’Reilly 1997; Rose 1998). This off ers
a division into 14 groups using over 350 occupations: these 14 groups can be collapsed to nine, eight, fi ve, and three categories; the government has adopted the version with eight divisions. It is based on the premise that there are three types of people in the workforce: employers, self-employed workers, and employees. Furthermore, there are two types of relationship: the labour contract and the service relationship. The labour contract is short- term: the contract is easily ended and so there is a low level of job security. Conversely, the service relationship is longer-term, with greater job security and will feature various ‘packages’ such as pension and health schemes; this is a trust relationship. One benefi t is that it can be converted to the ‘old’ social class. The structure is shown in Figure 5.1.
Of particular note in the UK is ‘social grade’—not actually a government classifi cation. This divides the population into groups denoted with letters and numbers: A, B, C1, C2, D, and E. Working defi nitions are shown in Table 5.2. This classifi cation dates back to before the Second World War and was developed in conjunction with early readership media surveys. It is based
on occupation and is used widely by the market research industry in the UK. However, it is specifi c to the UK and does not appear elsewhere in the same form.
Inevitably, the focus of research will be on the target audience as defi ned by the objectives
of the marketing plan. Target audiences are often defi ned in terms of social grade, sex, age, and region. These are clear demographics also used in marketing research because they distinguish behaviour. In turn media planners use market research audience surveys to choose the best
SEC structure
Employers –large –small
Employees
Service Intermediate Labour
Figure 5.1 SEC structure
Part 2 Data collection
152
mix of vehicles to carry the advertising. It is important to say that such audiences may not
be users, or likely users, of the product or service in question. In some cases, the buyer does
not use the product, but decides on its purchase and actually buys it. There are also other
audiences, such as shareholders or associated companies, who will benefi t from knowledge
about a product or service in their own decision-making. Some generic audiences are
shareholders, users, buyers, and gatekeepers.
The challenge for the researcher is to select the target, but also to ensure that the
procedures used are appropriate to the target audience, and that the actual sample selected is
suffi ciently coherent and articulate to be able to voice its knowledge and perceptions. In many
cases, this might mean replicating the recruitment procedure used on audience measurement
surveys. This may mean investigating the screening procedures and instruments used on the
major studies; this way, the sample matches the intended target precisely.
Choices of sample size should be guided by the planned task. In idea generation, the sample
is less important than the usefulness of the ideas being generated. In testing the eff ectiveness
of an advert, a split sample of matched respondents may be required. This way, there must be
similar profi les that ensure that two diff erent sets of results can be compared, that is results
obtained from people who have been exposed to two diff erent treatments (ad executions).
Populations involved in
social research
Marketing researchers deal mainly with consumers of Fast-Moving Consumer Goods (FMCG),
people who are within their target market. Conversely, social researchers question people
Working defi nitions of the UK social grade system
Grade Defi nition Description
A Upper middle class Higher managerial, administrative, or
professional
B Middle class Intermediate managerial, administrative, or
professional
C1 Lower middle class Supervisory or clerical and junior
managerial, administrative, or professional C2 Skilled working class Skilled manual workers
D Working class Semi- and unskilled manual workers
E Lowest levels of
subsistence
State pensioners or widows, casual, or lowest grade workers
Table
5.2
Chapter 5 Sampling
153
from all walks of life: the unemployed, the old, the inarticulate, and the disabled. Our social and welfare services look to provide disadvantaged, distressed, or vulnerable people with support. Social research has no buyers, but audiences may be numerous. Audiences may receive or need services; they may even act as a resource, perhaps as volunteers; they may
be managers of such resources. Respondents may be drug addicts, carers, manual workers, scientists, voters, etc.
Table 5.3 categorises typical populations that are central to social research studies; we see that institutions such as public authorities, hospitals, and utilities may be used. In common with B2B research, these organisations have complex hierarchies and decision-making units,
so respondents can span many people, from council members, department heads, and purchasing managers to administrative staff . The stakeholders and players involved in non- profi t organisations—perhaps volunteers or potential donors—are also used. Members of the professions—lawyers, solicitors, doctors, and architects—will be extremely important to certain projects. This is also the case with opinion leaders, such as journalists, politicians, shareholders, and activists in pressure groups. Finally, the largest group of all is the general public. Members of the public are the recipients of most social services. As a consequence, household members, whether they are families or sharers, will all be central to studies. Specifi c members of the public may become the focus, and each of the subgroups will need careful defi nition. For example, let us look at one defi nition of disabled people. The Disability Discrimination Act 1995 defi nes disability as ‘a physical or mental impairment which
has a substantial and long-term adverse eff ect on a person’s ability to carry out normal day-to-day activities’. Long-term is usually seen as 12 months or more in this context. Clear defi nition of
the population of interest must be created at the outset.
Most marketing research examines consumers and how they process information, make decisions, and consume products. In social research, minority groups may be interviewed for the opposite reasons; such individuals are disadvantaged or vulnerable to many forces in society. For research, these groups often require a modifi cation in sampling and collecting data. Such groups may be a minority because of their race, belief, income level, or behaviour.
Typical populations for social research
Population type Examples of typical respondents
Public authorities, hospitals, utilities Council members, department heads,
purchasing managers, administrative staff Non-profi t organisations, charities Volunteers, potential donors
Professionals Lawyers, solicitors, doctors, architects
Opinion leaders Journalists, politicians, shareholders, activists in
pressure groups
Members of the public Families, household members, institutions.
Specifi c groups, e.g. disabled people and carers
Table
5.3
Part 2 Data collection
154
The stages of sampling
Before proceeding with fi eldwork, we need to defi ne our population and the source of the
sample; we need to decide how to take a sample and we need to decide on the sample size.
These are the basic elements involved in selecting a sample.
The fi rst step is to examine the purpose of the study to decide what degree of precision
is required. After this, we must defi ne the population. We must decide a suitable source for
the population members; this is the ‘sampling frame’. Having done that, we determine the
sampling procedure, to be clear how the members are selected or recruited; this may be by
probability or non-probability methods. Sampling may be done in the offi ce by researchers
or in the fi eld by interviewers; these two approaches are called ‘preselected sampling’ and
‘fi eld sampling’, respectively. The sample size is generally agreed before undertaking fi eldwork,
although in a few projects, the sample size may be determined after it has started. This is
explored further later in this chapter. After fi eldwork, any sampling errors will be identifi ed and
corrected at the publication stage.
Sampling frames
The ‘sampling frame’ is an important part of sampling (see Table 5.4). As shown in Figure
5.2, it should mirror the population of interest in summary form. It should include summary
information of key features of all units in the population of interest. It is the basis by which
respondents are selected: people, telephone numbers, or addresses are sampled from a frame.
It might be a tangible list such as a phone directory or it might just be a set of instructions; it
may take the form of geographic maps, to divide sample by region, or even at street level.
Remember the ‘-ing’; it is sometimes written as ‘sample frame’; to be pedantic, this is
incorrect because it is the frame from which sampling happens. If we imagine books on a
bookshelf against a wall, it is the shelf from which we take down volumes; it is the place from
The stages of sampling
1. Examine the objective of the study—purpose
2. Defi ne the people of interest—population
3. Find suitable source for the population members
4. Decide on the sampling type and approach—procedure
5. Decide on the sample size
6. Proceed with the fi eldwork
7. Correct sampling errors ready for reporting—publication
Table
5.4
Chapter 5 Sampling
155
which sampling occurs. Sampling frames must be up to date, complete, aff ordable, and easy to use, in the sense that they can be manipulated and transferred into other media. They should
be easily exported into software such as spreadsheets (e.g. Excel), databases (e.g. Access), or word processing programs (e.g. Word).
The identifi cation of a useful sampling frame can be time-consuming, and where one needs
to be created, this can be a project in itself (see Figure 5.3). Any source used must be checked for duplicates. In the case of preselected samples, the duplicates must be removed before
fi eldwork takes place in a process known as ‘de-duplication’. Where sampling takes place in the
fi eld, duplicate interviews must be avoided by careful record keeping.
Distribution channels
Wholesaler
Customer Manufacturer
Retailer
Figure 5.3 Sampling frames: sources of sample
Population Sampling frame
End sample Sample selected
Non-respondents
Figure 5.2 From the population to the end sample
Part 2 Data collection
156
Poor sampling frames are ‘old, incomplete, and inappropriate’. Typically, bad frames
are made up of databases with individuals who have volunteered themselves or have been
selected based on criteria that do not match the purpose of the project in question.
Table 5.5 lists diff erent sources available to the researcher, and a few of these sources will be
described. If several are merged together, then a very powerful sampling frame can be created;
this merging process can make (for example) sampling using the Postcode Address File (PAF)
feasible for the telephone. Some providers of these services may off er their product already
merged with other databases.
The Electoral Register
Until 1990, the main source of general population samples in the UK was the Electoral Register
(ER). This lists all those eligible to vote, so is useful for sampling because it is expected to be
Sample sources and suitability to diff erent methods
Source suitability
for surveys using . . .
Telephone interviews
Personal Post Online
The Electoral Register Yes Yes
Postcode Address File Yes Yes
Random digit dialling Yes
Telephone directories Yes Yes Yes
Announcements Yes Yes Yes Yes
Email directories Yes
Subscriber/members
records
Yes Yes Yes Yes
Customer records Yes Yes Yes Yes
Interest group members Yes
Registration forms Yes
Snowballing Yes
Invitations (e.g. banners) Yes
Hypertext links Yes
Printed directories Yes Yes
Pop-up surveys Yes
Harvested addresses Yes
Website directories Yes
Table
5.5
Chapter 5 Sampling
157
complete. The Register includes all British, Northern Irish, Commonwealth, and Irish Republic citizens who are aged 18 or over, or who will become
18 during the life of the Register. The Register is compiled in October of each year and comes into eff ect in February, remaining in eff ect for 12 months.
From a practical angle, it is not essential to visit local town halls to have access to the electoral roll; records can
be searched by directory services such
as http://www.192.com. Of particular interest is that searches on this website can be made on relationships—
particularly interesting if de-duplication is necessary.
This Register is made up of individuals, so is useful for sampling people, rather than addresses. It is possible to use the ER to sample households, but it is worth noting that,
if no registration form is returned, the same information as the previous year is often left unamended. This can create inaccuracy. If it is used to sample addresses, then selection must
be done carefully, for example, by using only the fi rst entry for any one address (this is known
as ‘fi rsting’). Alternatively, weighting should take place after collection; this is described further
in Chapter 9 on analysis. A disadvantage is that there are reasons why people fail to register. This happened during the introduction of Poll Tax and there are cases where people prefer not
to register (and lose their vote) simply to avoid the arrival of junk mail (because this register can
be purchased by direct marketers).
From Autumn 2002, an ‘opt-out’ facility was off ered; this means that people listed can opt out of making their names available for mailing. In 2003, 21 per cent of people took this option; the percentage rose to 26 per cent in 2004 and then to 32 per cent in 2005. Clearly, the rise is disturbing for sampling and it is important to decide who has opted out, and whether the reason is relevant to the specifi c subject under study.
In April 1990, the Poll Tax was introduced. This meant that a tax was due from each person aged 18 years or over, and one way the government could identify people was through the
ER. Here is an extract from one of many leafl ets produced during the movement against the
‘community charge’ or Poll Tax.
The most important thing to do now is to ensure that your name and address are not added to any local or national government lists, e.g. the Electoral Register. In the event of a census being carried out, refuse to give any information. If
possible, remove snoopers from your area by force. Sabotage, industrial action and refusal by those asked to administer the system are also important possible forms of resistance.
As a result of such resistance, up to 18 million people refused to pay Poll Tax, and one way
to avoid this was not to register for voting. This left the Register incomplete, and the Postcode Address File became the sample source of choice.
‘ ’
Part 2 Data collection
158