Background

The Happiness Pulse measures wellbeing in four domains:

● General wellbeing
● BE (emotional wellbeing)
● DO (behavioural wellbeing)
● CONNECT (social wellbeing)

To measure general wellbeing, we use two of the ONS-4 questions on wellbeing. The BE domain comprises the seven questions from the short version Warwick-Edinburgh Mental Well-being scale (SWEMWBS) as these questions are robust, subject to stringent testing and provide the best way of summarising the main elements of emotional wellbeing. We also use their prescribed method for working out the results of the BE domain. The Warwick-Edinburgh Mental Well-being Scale was developed with funding provided by the Scottish Government’s National Programme for Improving Mental Health and Well-being. It was commissioned by NHS Scotland, developed by the University of Warwick and the University of Edinburgh in 2006, and is jointly owned by NHS Scotland, The University of Warwick and the University of Edinburgh. All the questions in both the DO and CONNECT domains are based on indicators used by recognised data agencies, adapted and tested by Happy City. The additional Community Pulse module, accessed via our Advanced package, uses the following data organisations as its source: the Office for National Statistics, the Understanding Society Survey, the European Quality of Life Survey, the Oxwell Survey. A number of the questions are adapted by Centre for Thriving Places. Where we use a question directly from another source we also use the original response scale for that question.

Click here to see a list of the domains, questions and answer scales in the core Happiness Pulse survey.


Calculating wellbeing domain scores for each respondent

The core Happiness Pulse survey results are automatically calculated within the Pulse app as follows:

Step 1: Standardising scores
Questions in the Pulse have different response scales. To make them easier to compare, the scores are standardised by converting them to z-scores: Z-score = [(original response) - (mean response)] / (standard deviation of responses) The means and standard deviations are taken from a nationally representative sample of 2,000 Pulse responses, which was collected by YouGov in October 2018. We aim to update this every four years.

Step 2: Aggregation
To get a score for a whole domain from the responses to individual questions, scores are aggregated by adding together all the Z-scores within a domain, and dividing by the number of questions in the domain.

Step 3: Rescaling
Although Z-scores are comparable, to aid understanding the domain scores are rescaled so that the overall mean for each domain is 6.5 and the range is 0-10.

Step 4: Transformation to percentiles
To aid interpretation of scores, scores are also transformed into a percentile: a 1-100 scale, where each unit on the scale represents a 100th of the range of scores in the nationally representative sample. For instance, a percentage score of 1 would mean that the user is in the lowest 100th of user scores in the UK; 36 would mean the user is in the 36th percentile in the UK; and so on.
To do this, we calculated percentiles for the nationally representative sample, which can be done using a formula, or manually in the following way:
● Order range of domain scores from lowest to highest
● Divide total amount of domain scores (e.g. 722) by 100 to find out how many user domain scores belong to each percentile (e.g. 7.22)
● Record range of domain scores for each percentile (e.g. the first percentile will consist on the lowest 7 user domain scores)

To show users of the Pulse their percentile score for a domain, the Pulse app looks up what percentile their domain score falls into in the nationally representative sample. For example, if a user got a general wellbeing score of 5, this score falls into the 23rd percentile of scores for the nationally representative sample and a score of 23% is displayed. This means they scored higher than 23% of the UK population.

Sample size and representativeness
In order to get meaningful data the Happiness Pulse must be used in a way which ensures optimum outreach so that both the sample size and the representativeness of that sample provide valid and reliable data at a local level, while preserving the anonymity of the respondent. We work with our clients to ensure that this is the case, and provide guidelines on the parameters for comparison and analysis.