Social surveys are most frequently conducted by governments and campaigning organisations.

They involve collecting data from large numbers of people.

They describe and explain characteristics and factual information, as well as opinions and attitudes (e.g., census and opinion polls).

Surveys:

• identify and measure characteristics, behaviour and attitudes;
• build up a profile of a person (e.g., types of drug-takers); and
• develop a relationship between variables (e.g., class and political party preferences).

Surveys use SPSS to determine the nature of the relationship between the variables: (a) tests of associations – to indicate how strongly the sample variables are related or correlated; (b) tests of significance – to indicate to what extent it is legitimate to generalise from the sample to the population).

Surveys fit well within the positivist/empiricist tradition of methodology.

A good survey research tests a theory by testing its hypotheses. A hypothesis is a conjecture, deduced from a theory, and must be operationalised.

Answers to survey questions must be capable of categorisation and quantification.

Then the hypotheses are confirmed or falsified.

Survey research must be ‘rigorous’; i.e.,

• standardised – conditions, questions and approach must be the same so that the respondents have the equivalent stimuli (though this is difficult to control);
• replicated – other researchers able to reproduce the same results so giving reliable and valid data (though difficult to achieve in different time and space); and
• representative – reflects the population (though there are different types of people).

• random/probability samples
• selective/non-probability samples

Sampling frame refers to the -population

Simple random sample – e.g., 2% of the population

However, this random sample can be modified to produce more accurate and representative sample. For example, multi-stage cluster sampling; e.g., in a study of poverty in Kyrgyzstan, a researcher would choose 3 out of the 7 oblasts, then select the micro-regions, then select the streets and the households.

To ensure greater representativeness, stratified random sampling is used, where the sample is selected on the basis of their social characteristics such as age, ethnicity, sexuality, gender, income level, type of house, etc. This requires a good understanding and knowledge of society and population.

A more practical procedure can be adopting by using systematic random sampling; e.g., every nth person.

However, samples are required for ‘fit for purpose’, not for statistical representativeness; e.g., research on prostitution, child poverty and youth fashion.

An important method – ‘snowballing’ or snowball sampling: initial contact leads to other contacts. This is important when the nature and size of the population are unknown; e.g., drug-users or bribe-taking in education.

1. Preliminary work

• reading around the topic; i.e., theoretical research and literature review;
• field research so as to develop appropriate questions and issues to address; and
• pilot study of sub-sample in order to test design and questions; to ensure questions are relevant and clear; to produce valid information; and to iron out and correct difficulties.

1. Types of questionnaires

• mail / self-completion questionnaires: has lower cost than face-to-face interviews; deals with sensitive issues where anonymity is essential such as rape and incest; is conducted in respondents own time and can give considered responses; and covers a wide geographical area. However, questions relatively simple; no control on how questions are interpreted; no follow up questions; no control who answers the questions; and low response rate.
• telephone survey: telephone directory – not in the book and class-bias, however cheap.
• face-to-face interview: high response rate and controls the interview situation. However, it is costly.

1. Types of questions

• classification – demographic details;
• factual questions – how many . . . ? where did . . . ?;
• opinion questions; and
• open / closed questions.

1. Coding

Most questionnaires are pre-coded to allow the classification of responses into analysable and meaningful categories; e.g., a question may have five possible answers, numbered from 1 to 5.

2. Attitude Scales

They consist of a set of statements that the researcher has designed, and the respondent is then asked to agree or disagree with the pre-coded answers, or answer them on an attitude continuum.

Here are some points to note when designing a question:

• ensure that the question is not too general;
• use the simplest language;
• avoid prejudicial language;
• avoid ambiguity;
• eliminate vague words and expressions;
• avoid leading questions;
• ensure that respondents have the necessary knowledge;
• consider using filter questions;
• avoid hypothetical questions;
• exercise caution when asking personal questions; and
• recognise the problem of recall.

Nowadays all analysis of survey data is conducted through using statistical analysis software products such as SPSS.

Levels of measurement:

(a) nominal variables – identified by names;

(b) ordinal variables - rank the differences in replies; and

(c) interval scales – measurements of the differences.

The aim of questionnaire analysis is to examine the patterns among the replies to the questions, and to explore the relationships between the variables. Two tests are used for this purpose:

1. test of significance – legitimacy to generalise from sample to population;
2. test of association – how strongly the variables are related.

There are some criticisms of survey method:

• It attempts to show causal relations between variables, a strategy that is simply not applicable to the realm of human action, which is rule-following, not ‘caused’. In addition, two variables may be associated, but this correlation does not mean that one variable causes a change in another.
• It rules out the possibility of understanding the process by which people come adopt particular values and behaviours.
• Researchers have presuppositions leading them to ask particular questions, and such restrictions shape the way in which people can answer questions. Consequently, it becomes inevitable that the theories are ‘proven’. By using standardised questionnaires, people do not have the opportunity to challenge ideas on their own terms. Furthermore, compartmentalising responses into fixed categories can hardly accommodate the myriad of differences in people’s attitudes and the meanings.
• Surveys take no account of change in opinions across time.
• It is suggested that by attention to design, measurement and good pilot work, the above problems can be overcome. However, the central issue in social research for the critics is a hermeneutic one, not a measurement one. Can researchers legitimately understand the ways in which people interpret the world around them, and act within their own social universe? How can survey researchers guarantee that their questions will be interpreted by the respondent in the manner in which they intended when there is no opportunity for a participatory dialogue in order to research understanding? From, this point of view, all quantitative research assumes an ethnographic dimension to its design.
• Even if meaning equivalence can be upheld, another problem arises: attitudes and actions are two different things, or what people say they do is very different from what they actually do.

Nevertheless, in the quest to compartmentalise surveys within a positivist orientation, and to produce a dichotomy between qualitative and quantitative methods of social research, their broad appeal can be easily overlooked.

An important skill in becoming a researcher is the ability to weigh up the practical value and methodological limitations of particular methods.