Marketing research 191
they are sampled using a system called the
probability proportionate to size method. Thus,
if a county has five times the population of the
other counties, then in the sampling process it
should be allocated five times the chance of
being selected. The first stage thus results in a
number of counties etc. drawn from the pop-
ulation of counties. Then the research will select
from these areas a number of, for example,
cities, and again they will be selected using the
probability proportionate to size method. These
stages may be repeated until the research
arrives at the desired final sample. The method
has the advantage that the process delivers a
sample chosen at random, but concentrated in
certain geographical areas, useful when the
costs of travel and communication can be high.
It also means that probability sampling may be
used when, at the macro level, there is no
sampling frame. When the final stage has been
completed, and the research has arrived at the
micro level, sampling frames will be available –
city maps, electoral rolls etc.
Non-probability sampling techniques
The researcher does not know the chances of a
unit’s selection if non-probability sampling
techniques are employed. Therefore, the ability
to generalize about a population, using the
laws of probability, is much reduced and it is
not possible to calculate the degree of con-
fidence in the results. The sample is chosen at
the convenience of the consultant or to fulfil the
demands of some predetermined purpose.
Convenience sampling
Here the sample is chosen for the convenience
of the research worker. A street interviewer
who needs to sample 50 people, for example,
might question the first 50 people who walk
past the street corner where the interviewer is
standing. It is a quick method and carries the
minimum cost. It is a method useful in explora-
tory research.
Judgement sampling
This makes an attempt to ensure a more
representative sample than that gathered using
convenience techniques. Research consultants
use their expertise, or consult an expert, to
evaluate populations and to make recommen-
dations as to which particular units should be
sampled. With small populations, accurate
assessments and guidance as to a unit’s selec-
tion, judgement sampling can render samples
with less variable error than might result with a
sample chosen using a simple random tech-
nique, though this cannot be conclusively
proved.Purposive sampling
This does not usually aim for representative-
ness. Here the choice of the sample is made
such that it should meet certain preconditions
deemed appropriate to the fulfilment of the
objects of the research. Thus, a project might
stipulate that the top 50 Professors of French be
interviewed as part of the project, so there is no
true ‘sampling’, merely the need to contact
those units the research has already
delineated.Quota sampling
This attempts to reflect the characteristics of the
population in the chosen sample, and in the
same proportions. From national statistics,
researchers gather the percentages for such
‘stratifiers’ as age groupings, income levels etc.
and use them to construct ‘cells’. This results in
statements such as ‘23 per cent of the popula-
tion is female, aged between 30 and 40 and
earning £12 000–15 000 per annum’. The sample
would then be collected, and 23 per cent of it
would have to fulfil those demands. Quota
controls must be available, easy to use and
current. Quota ‘stratifiers’ shouldn’t be used
merely because they are available – they must
be relevant to the project. This method may be
cheaper to operate than a probability-based