The insecurity index that has been developed for this report seeks to fill a gap between government datasets that fail to provide a holistic view of insecurity in a changing economy. However, due to its experimental nature, and precisely because it aims fill in gaps in official data, a number of assumptions and approximations have been made during the analysis. Those assumptions are laid out below. That said, it is our firm view that the insecurity index provides an important level of insight that is missing from official government data and previous analysis. The methodology provided should be taken as proposal that could be more robust over time.
The bulk of the data used in this report comes from the Labour Force Survey. These datasets are the raw data used by the ONS to produce summarised employment data. They are accessible via the UK Data Service but require a registered account and data analysis software such as SPSS or STATA to view. In each case the Two Quarter Longitudinal datasets were used. In order to make a like for like comparison with the most up to date data, the latest quarter in each year was analysed.
Each dataset has a different number of respondents and, in general, the number of respondents has decreased over time. To account for these changes each respondent’s weighting was multiplied by the total weighting of all respondents across all datasets, divided by the total weight of all respondents in that year.
A major issue when trying to account for people’s wages is the fact that only around 15% respondents provide their income. This is assumed to be because of the perceived personal nature of that information and respondents’ general resistance to disclose their income.
Where respondents had not provided their pay, the figure was replaced with an average hourly wage for that sex, region, occupation and industry. However, even with the relatively large sample size overall, the sample size for any one group (say, females in London with an elementary occupation in the mining industry) is often not statistically meaningful, and occasionally contained no respondents at all.
In order to include a figure for income in each group, pay was approximated from the annual survey of hours and earnings (ASHE). This data is provided by the ONS but is summarised separately by industry and occupation. In each case, average hourly wage is provided for each region and sex. To estimate the average wage of each combination of occupational group and industry, we distributed pay in each occupational group according to average pay in each industry:
PIO = PI/PT * PO
PIO = average pay in one occupational group and industry
PI = average pay in that Industry
PO = average pay in that occupational group
PT = overall average pay
This process was repeated for each region, year and for males and females. In each case, the product of average pay and number of jobs for an occupational group and industry, divided by the total number of jobs in that industry remained equal to the average pay for that industry. At the same time the product of average pay and number of jobs for an occupation and industry, divided by the total number of jobs in that occupational group remained equal to the average pay for that occupational group.
This method assumes that pay is distributed similarly across occupational groups in each industry. I.e. That managers are paid more than elementary occupations in in the hospitality sector as well as in the education sector. However, it does allow for the fact that managers in education are paid more than in hospitality. It also assumes that the distribution of jobs in different occupations follows the national average in all industries. I.e. that the same percentage of the hospitality workforce are managers as in the financial sector. Of all the assumptions made in this report, this is probably the furthest from accurate. However, it is almost certainly closer to the truth than applying either the average pay for an industry to everyone in that industry, the average pay for an occupation to everyone in that occupational group, or taking the average of the small sample sets created by looking at such specific sections of the LFS respondents.
Where a respondent is listed as unemployed, we used the rate of job seeker’s allowance (JSA) or universal credit for that time period, based on a respondents age, number of dependent children and whether they lived as a couple of single parent.
Living Cost Calculation
The basis of the living cost calculations is derived from the Resolution Foundation’s living wage calculations. The living wage itself is calculated as two proposed hourly wages, one for London and one for the rest of the UK. However, before being boiled down to these figures, the calculations estimate the basic living costs of a household based on whether it has the income of someone single or of a couple and the number of children in different age brackets. This produces a dataset of living costs based on household type.
In order to improve the accuracy of the insecurity index, we took these datasets for each year and matched respondents to the closest available category. I.e The living cost used for a single parent with three children is different to that used for a couple living together with no dependent children.
These datasets are available via the Resolution Foundation’s annual reports, Calculating the Living Wage reports. As the living wage was only introduced in 2016, previous years were extrapolated using historical inflation data.