Descriptive Method

Concept

Descriptive research describes things as they are: the present status of a population, the opinions people hold, the conditions that exist, the processes going on or the trends developing. It does not manipulate variables. The main forms are surveys, case studies, developmental or growth studies, correlational studies, comparative studies and observational studies. A survey collects information from a sample or the whole population, usually by questionnaire, interview or schedule; it may be cross-sectional (once) or longitudinal (repeatedly). In a panel study the same respondents are surveyed again and again, which makes change at the individual level visible. Surveys give breadth and are efficient for large populations, but they establish association rather than cause and are open to response bias.

Concept

A questionnaire is a set of written questions answered by the respondent without an interviewer. It reaches a scattered sample cheaply and removes interviewer bias, but it needs literate respondents and a mailed questionnaire often suffers from a low return rate. A schedule contains similar questions but is filled in by an enumerator, so it can be used with illiterate respondents and gives higher response, at higher cost. Questions can be closed (fixed options) or open-ended; they should avoid leading, loaded and double-barrelled wording, and the instrument should be pilot-tested. An interview may be structured (fixed questions), semi-structured or unstructured, and uses probing and rapport. Observation may be participant (the observer joins the group, as in ethnographies) or non-participant, structured or unstructured, and is open to observer bias and the Hawthorne effect.

Concept

Attitude scales convert opinions into scores. The Likert scale (1932) is a summated rating scale in which respondents mark agreement on a five-point range from strongly agree to strongly disagree and the item scores are added. The Thurstone scale (1929) uses statements pre-rated by judges to be equal-appearing intervals, and respondents agree with those they accept. The Guttman scale (scalogram, 1944) uses cumulative, unidimensional items so that agreement with a stronger item implies agreement with the weaker ones. Osgood’s semantic differential (1957) uses bipolar adjective pairs (good–bad) on a seven-point range to measure the meaning of a concept along evaluation, potency and activity. The Bogardus social distance scale (1925) measures willingness to accept groups at different degrees of closeness. Sociometry, developed by J. L. Moreno (1934), studies social acceptance and rejection within a group and displays them in a sociogram.

Concept

A population is the entire group of interest and a sample is the part actually studied; the list from which the sample is drawn is the sampling frame. In probability sampling every unit has a known, non-zero chance of selection, which allows estimation of sampling error. The main kinds are simple random, systematic (every k-th unit after a random start, where k = N/n), stratified (random sampling within strata, proportionate or disproportionate), cluster (whole groups are randomly selected) and multistage. In non-probability sampling selection is not random, so sampling error cannot be calculated: convenience (accidental), purposive (judgemental), quota (fixed numbers from categories but chosen non-randomly) and snowball (participants recruit others). Fuller treatment appears in Topic 3E.