The selected sample must be representative of the population being studied because normally sociologists wish to generalise. It is also important to find a sampling frame (a list of people who may potentially take part in a survey) which is representative of the population being studied. Sociologists prefer to use random sampling methods in order to minimise the possibility of bias. At its most basic, random sampling allows everyone the same chance of being selected. A number of sampling methods are available to sociologists:
- Systematic sampling: Every nth person is chosen from a sampling frame. For example, 50 people out of a group of 500 may be chosen by randomly selecting a number between 1 and 10, e.g. 6. Every tenth name beginning with the randomly selected number ‘6’ is taken from the list, e.g. 6, 16, 26, 36, 46... up to 496. This process will generate fifty names for the sample.
- Stratified sampling: A random sample is taken from particular social categories, e.g. age, gender, race, etc., which make up the population being studied.
- Cluster or multi-stage sampling: Households may be randomly selected from a random sample of streets from a random selection of areas.
Sometimes non-random and unrepresentative sampling methods may be
preferred despite the danger of unreliability:
- Quota sampling is mainly used by market researchers in the street. For example, they may be under instructions to stop and interview a quota of housewives aged between 25 and 40 years of age.
- Snowball, or opportunity, sampling is mainly used in areas in which it is difficult to find a sample because behaviour is seen as deviant by society, e.g. a researcher interested in heroin addiction may find an addict willing to introduce him or her to other addicts.
Positivists approve of the social survey because it is regarded as scientific (variables are controlled via sampling and questionnaire design), reliable (standardised questionnaires can be replicated), objective (samples are randomly selected) and quantifiable. However, some social groups are difficult (and sometimes impossible) to survey, e.g. people who are illiterate and criminals.
Surveys which monitor a group over a period of years are called longitudinal surveys. They supply an in-depth picture of a group or social trends over time. Trust can be built up between a group and the researchers. This may generate more valid data and lessen the possibility of non-response. However, it can be difficult to find samples and research teams committed to long-term research. The sample may drop out, die, move away, etc. This increases the chance of it being unrepresentative. If different researchers are used, it can be difficult to re-establish trust with a group. The sample may become too ‘survey-friendly’. They may consciously or unconsciously tell researchers what they want to hear.