Collection of Data

Primary and secondary data, questionnaires and the three ways of interviewing, census and sample surveys, random sampling, and the errors that creep in.

Key ideas

• Primary data are collected first-hand by the investigator; secondary data come from sources that collected them earlier, such as reports.
• A good questionnaire is short and clear, moves from general to specific, and avoids double negatives, leading questions and suggested alternatives.
• Personal interviews give the highest response but cost most; mailed questionnaires are cheapest and reach remote areas; telephone interviews are quick but miss people without phones.
• A pilot survey tries the questionnaire on a small group first to find weak questions and judge cost and time.
• A census covers every unit; a sample survey studies a representative part and is cheaper, quicker and allows more detail.
• In random sampling every unit has an equal chance of selection, as in a lottery.
• Sampling error falls as the sample grows; non-sampling errors (wrong recording, non-response, bias) do not, even in a census.

Remember

Watch out

• Thinking a census has no errors: non-sampling errors can occur in it too.
• Calling a “Don’t you think…?” question neutral: it pushes the answer.
• Mixing up the population (all units) with the sample (the part studied).

Worked example