The concept of reliability of a research instrument

We use the word ‘reliable’ very often in our lives. When we say that a person is reliable, what do we mean? We infer that s/he is dependable, consistent, predictable, stable and honest.

The concept of reliability in relation to a research instrument has a similar meaning: if a research tool is consistent and stable, hence predictable and accurate, it is said to be reliable. The greater the degree of consistency and stability in an instrument, the greater its reliability. Therefore, ‘a scale or test is reliable to the extent that repeat measurements made by it under constant conditions will give the same result’ (Moser & Kalton 1989: 353).

The concept of reliability can be looked at from two sides:

1. How reliable is an instrument?
2. How unreliable is it?

The first question focuses on the ability of an instrument to produce consistent measure­ments. When you collect the same set of information more than once using the same instru­ment and get the same or similar results under the same or similar conditions, an instrument is considered to be reliable. The second question focuses on the degree of inconsistency in the measurements made by an instrument — that is, the extent of difference in the measurements when you collect the same set of information more than once, using the same instrument under the same or similar conditions. Hence, the degree of inconsistency in the different measurements is an indication of the extent of its inaccuracy. This ‘error’ is a reflection of an instrument’s unreliability. Therefore, reliability is the degree of accuracy or precision in the measurements made by a research instrument. The lower the degree of ‘error’ in an instrument, the higher the reliability.

Let us take an example. Suppose you develop a questionnaire to ascertain the prevalence of domestic violence in a community. You administer this questionnaire and find that domestic violence is prevalent in, say, 5 per cent of households. If you follow this with another survey using the same questionnaire on the same population under the same conditions, and discover that the prevalence of domestic violence is, say, 15 per cent, the questionnaire has not given a comparable result, which may mean it is unreliable. The less the difference between the two sets of results, the higher the reliability of the instrument.

Source: Kumar Ranjit (2012), Research methodology: a step-by-step guide for beginners, SAGE Publications Ltd; Third edition.