Statistically Representative Survey Results
Employee surveys should be conducted on a voluntary basis. This means a number of employees may elect not to participate in the survey, which could jeopardise the validity of the survey results. To illustrate, if only 5 out of a total population of 10 employees participated in a survey, the survey results could hardly be described as a true and fair reflection of the opinions of all 10 employees. However, the survey results would from a statistical point of view be sufficiently representative if a sample of 500 out of a population of 1,000 employees were to participate in a survey. It is therefore important that as many employees as possible are motivated to participate in the survey, particularly in the case of smaller business units (see the table below).
To determine whether the number of survey responses from a survey group or business unit are sufficiently representative, it will be necessary to calculate the sampling error, i.e. the extent or margin of error to which a particular sample actually represents the views and opinions of the group being surveyed.
From a psychological research perspective, a confidence level of 95% and sampling error of 5% are deemed acceptable. For example, a survey response of 67% will be representative of the population 95% of the time, with a ±5% margin of error. A sampling error of 5% or less means that the survey results are adequately representative, while survey results with a larger sampling error (e.g. 10% or 20%) may not be adequately representative of the group being surveyed and should be interpreted (and used) with caution.
Bottom line: a sampling error is not the same as calculating the averages or means of the respondents who participated in a survey; a sampling error will tell us whether we can say with confidence that the sample that participated in a survey represents the views of the larger population. See Sample Sizes & Sampling Errors in the analytics dashboard Help file for more on sampling errors.