What’s the best employee survey scale to use?
There is lots of research and statistical stuff around survey scales so this post is intended to be a practical view of what is the best employee survey scale to use for employee experience surveys in the real world.
When it comes to employee experience surveys, we usually want to gather information about how our employees feel about working for the organisation. The type of survey questions you use and the way you want to analyse your survey data will influence the type of rating scale you use in your employee experience survey.
So, what are the different types of employee survey scale?
There are broadly three types of survey response scales:
- Dichotomous scales
- Semantic differential scales
- Rating scales
Dichotomous scales have two choices that are diametrically opposed to each other. For example:
- ‘Yes’ or ‘No’
- ‘True’ or ‘False’
- ‘Agree’ or ‘Disagree’
Dichotomous scales prevent any kind of ‘sitting on the fence’. They also make analysing the data fairly simple.
Dichotomous scales are good for gathering precise data, but they don’t allow for any kind of nuance in respondents’ answers.
For example, asking if an employee found their last performance review effective does not provide any real insight into why that was and how to improve performance reviews in the future.
Semantic Differential Scales
Semantic Differential scales are designed to measure a person’s subjective perception of the properties of concepts, objects, and events by making use of a set of bipolar scales.
Semantic differential scales usually have dichotomous words at either end of the spectrum and they measure more specific attitudinal responses.
We’ve used semantic differential scales in our own client research to understand what they valued most about our offering.
As with dichotomous scales, semantic differential scales are great for gathering precise data, but they don’t allow for any kind of nuance.
They also work best where a relatively small number of factors are being compared against each other.
Considering that the typical employee survey can have up to 80 questions in it, using a semantic differential scales to compare every question against every other question would result in a hugely lengthy survey.
How people feel is not black or white, we all have differing views, to differing extents, based on our experiences and feelings. This is why most employee surveys use a rating scale.
Rating scales ask people to select one response based on how strongly they feel about something. For example, imagine being asked to say how much you agree with the statement ‘I can see a future for myself in the organisation I work for’.
Whilst you could give a yes or no answer to this question, it is much more useful to understand the extent to which people agree or disagree with the statement in order to understand whether or not we have an issue across the business.
This is where rating scales offer a major advantage – they can be quantified and they allow a degree of variety in how people can respond.
The most commonly used rating scale in survey-based research is the Likert scale, named after its inventor, psychologist Rensis Likert. There are the pros and cons of Likert scales, but it is probably the most common type of rating scale to use in employee surveys.
When using rating scales there are really only two decisions to be made:
- How many response options should we use?
- What descriptions should we use for each response option in the scale?
How many response options should we use?
One of the biggest areas of debate in statistics and research methods is whether or not to use an even number of response options or an odd number. The core of the argument is:
- using an even number of response options does not allow people to give a ‘neutral’ response and it may encourage people to be more extreme in their views (overly positive or negative) as a result.
- using an odd number of response options does provide people with the option to give a neutral response, but it can introduce the potential for ‘central tendency bias‘. This is where people might just respond to every question using the neutral or central part of the scale.
We actually recommend avoiding a 5-point scale. There are two reasons for this. Firstly, it can be difficult to draw conclusions from the data if there is lots of ‘neutral’ in there. If the average response to a question is neutral, then do we do anything about it or not? Secondly, if a question is likely to result in a neutral response then we suggest that the question wording needs improving.
So, if we recommend avoiding a 5-point scale, that suggests we should use a rating scale with an even number of response options, but how many response options should we use?
We have some clients who use a 4-point rating scale as this fits their other processes such as performance appraisal. However, we find that this limits the range of available response options and it can lead to a low level of variation in the data. When exploring the data using average scores, this can lead to the scores for the survey questions looking a little too similar to each other.
We recommend using a rating scale made up of 6 response options. This provides enough variation in the data and it also provides enough scope for people to be a little bit negative or positive rather than being forced to be very negative or very positive.
What descriptions should we use for each response option in the employee survey scale?
The rating scale wording can be anything you want, but there are a few key pieces of advice we would give:
- Keep the wording consistent throughout the survey. When completing surveys, people quickly learn what the rating scale response options are. Responding the questions quickly becomes a habit and people do not read the response scale for each question. There are some surveys, most notably the NHS Staff Survey that use differing scales throughout the survey – they use scales with differing numbers of response options and they switch the ‘positive’ and ‘negative’ parts of the scale from left to right and vice versa in different parts of the survey. Whilst there may be some highly scientific reason for doing this, it just seems to be one extra thing that can cause confusion and lead to inaccurate data.
- Ensure the wording of the scale fits the context of the question wording. Most rating scales use a variation of ‘agree – disagree’, but we have seen some surveys that might use wording such as ‘Excellent – Poor’, ;Effective – Ineffective’ etc. These are still valid, we just advise that you ensure this works when you read the question statement followed by the response descriptions. For example, if would be difficult to respond to the question ‘I see a future for myself at the company’ with a response scale that read ‘Effective – Ineffective’.
So what’s the best employee survey scale to use?
The rating scale you use in your survey will have an impact on how you ask your questions and how you can analyse the data so it pays to think through the different options.
In practice we find that these two scales work best.