Annex B: DWP SCBA model parameter choices and cost methodology
This annex summarises parameter decisions made during this evaluation. It also sets out the cost methodology for calculating the cost per job start and the cost per job outcome. Table B1 sets out the parameters and assumptions that the DWP SCBA model allows the user to choose from. Most of these were described in Fujiwara's work discussed earlier.
Table B1. Parameters and assumptions that the DWP SCBA model allows for
- Distributional analysis
- Social cost of Exchequer finance
- Substitution effect
- Applicable substitution effect
- Real or nominal values
- Travel costs
- Hours worked per day - for alternative travel costs method
- Number of participants
Distributional analysis captures the idea that participants who have relatively low incomes are assumed to value each additional pound more highly than those on higher incomes, i.e., higher welfare is achieved via distribution from public finance (modelled as median income) to low-income groups. This is a standard approach, detailed in the Green Book (2020) and described in Fujiwara (2010). This is built into the DWP SCBA model and is included in the central scenario.
Social cost of public finance captures the loss of welfare due to funding employment programmes through taxation. According to Fujiwara (2010) there is some uncertainty over the size of this effect and little evidence from the UK. Because of the very weak evidence on the size of this impact (which could be zero in practice), and the precedent of Scottish treatment of public finance costs, this measure is not used for the main analysis or for the sensitivity checks.
The substitution effect refers to the possibility of newly employed people displacing existing workers. However, since Fair Start Scotland is highly targeted, it is likely that this effect is negligible, with increases in employment representing a fall in the natural rate of unemployment. The substitution effect can sub in for the counterfactual – some of the participants could have found employment (or derived benefits) even in the absence of the programme. Therefore, the 20% discount is used to account for the counterfactual. This parameter is varied in the sensitivity checks to account for better and worse performance of Fair Start Scotland.
Travel costs are incurred by participants who become employed or increase the number of hours in which they work. The DWP SCBA model gives two options for including travel costs i) "linear" travel costs and ii) 8 hourly travel costs. The linear cost captures the increase in transport costs as the income of the claimant goes up from zero. The 8 hourly cost captures extra travel costs if the number of increased hours is enough to warrant a new full day (i.e., it assumes that the claimant already has travel costs). Therefore, for individuals moving from unemployment to employment (starting with an income of zero), the correct choice is to apply the "linear" costs, so this is the default assumption. Just like travel costs, this is likely to be an extra cost incurred by parents who find a job. This is built into the DWP Model for parents and lone parents, based on the number of children.
Moving from unemployment to employment is likely to lead to improvements in individuals' health. This will result in two benefits: i) savings for the NHS and ii) improvement in individual's quality of life (captured by quality-adjusted life years (QALYs)). For the former, NHS savings is predicted to be larger for Employment and Support Allowance (ESA) participants who find work, as they incur greater initial public health costs. For the later, Public Health England (2017) estimates improvement in quality of life to be around 0.0675 QALY gain per person moving from unemployment to employment, assuming the benefit is sustained for one year.
When it comes to the methodology for calculating cost per job start across different characteristics, it is important to keep in mind that there are three types of costs; outcome spend, service fees, and non-supplier cost. The outcome spend is broken down by year, Lot, and participant group; the service fee is broken down by year and Lot; and the non-supplier cost is broken down only by year.
Therefore, some assumptions and calculations are needed in order to apply the appropriate service fee and non-supplier cost across Lots and participant groups. The costs incurred between 2018 and 2020 are adjusted for inflation to reflect real 2021 value.
In order to calculate the cost per job start for participant group, it is necessary to split the service fees and the non-supplier cost across Core, Advanced, and Intense participants. The way the non-supplier cost is split is in accordance with the share of participant group who recorded a job start (e.g., 60% of participants who recorded a job start were from the Core group, therefore 60% of the non-supplier cost went to the Core group cost).
However, the way that the service fee is split is in accordance with the share of participant group who recorded starting on the service as opposed to recording a job start (e.g., 40% of participants who started on the service were from the Core group, therefore 40% of the service fee went to the Core group cost). This is because the service fee is likely to be more impacted by the participant group that suppliers worked with rather than the group of participants that ended up finding a job.
As for calculating the cost per job start across Lot, the cost breakdown is available for outcome spend and for the service fee. However, the breakdown across Lot for the non-supplier cost is not available. Therefore, the non-supplier cost is split equally across the 9 Lots.
The calculation for the cost per job start also includes forecasted costs and forecasted number of jobs starts. These are added to the measure of cost per job start for the 2020 cohort. This is because at the time of this evaluation, it was too early to observe the complete outcomes for the 2020 cohort.