Minimum Income Guarantee: medium-term social security modelling

Footnotes

1 https://www.gov.scot/groups/minimum-income-guarantee-steering-group/

2 While our modelling of the mid-MIG is based on the assumption of 100% benefit take-up, our projection of the poverty line is based on partial take-up to reflect the current state of affairs. In theory, 100% take-up could also raise the median, although under the current system we would expect the median household to derive most if not all of its income from earnings.

3 Workers who report wages lower than the statutory Minimum Wage (the National Living Wage) are also brought up to the RLW, even though some proportion of these workers have likely mis-reported their wages. In cases where working hours are missing, an hourly wage cannot be calculated and therefore the earnings of these workers are not altered.

4 There would be no effect on Free School Meals or School Clothing Grant because these policies are either universal or include earnings thresholds. Households newly eligible for UC as a result of the mid-MIG would have earnings above these thresholds.

5 It will be partly reflected in our use of data on people rather than households, which will result in couples being weighted more heavily than single adults.

6 DWP, People on Universal Credit, Stat-Xplore.

7 At first glance, limiting claims that would have otherwise lasted for more than a year may appear to have no effect on the annual cost, since these claims would still last for a year. However, as noted above, the effect will be felt in future years, in proportion to the counterfactual duration beyond the limit. Including these impacts in the present year is therefore in line with our assumption of a steady state.

8 In reality this assumption will be unlikely to hold precisely, for example because a higher proportion of claimants will be in work, but there is no way to precisely predict any change in the distribution of claim durations. At present, claimants in work tend to have shorter claims, but this may not hold under the mid-MIG as it extends further up the earnings distribution.

9 ‘5 years and over’ assumed to have an upper end of 6 years, with a midpoint of 5.5 years (66 months).

10 While our analysis of the mid-MIG is based on cross-sectional data – i.e. the FRS – it is not affected by survivor bias. When it comes to estimating the total cost or average impacts across the year without a limit on duration, the under-representation of shorter claims at any point in time is a valid reflection of the reality that we are attempting to capture, assuming that the size and characteristics of the caseload are relatively stable over time net of on- and off-flows. It is only when we are interested in the distribution of claim durations that survivor bias becomes an issue.

11 This assumption is only reasonable because we have adjusted for survivor bias. The nature of this bias is not only that claims which end up being longer are more likely to be observed; it is also that claims that are further along are more likely to be observed, because they are longer by that point. Thus, on average, we are likely to observe claims beyond their midpoints. Inverse probability weighting should help address both issues by approximating a sample in which all claims are observed at a random point in time within each claim.

12 If an individual who was receiving UC at a point in time in Year 1 is still receiving it at a point in time in Year 2, we cannot ascertain whether or not UC receipt was continuous between two time periods. Conversely, if the individual is no longer receiving UC in Year 2, we cannot ascertain whether the claim finished shortly after the point of measurement in Year 1 or shortly before the point of measurement in Year 2, leaving a wide range.

13 The effect of censoring may be partly offset by the effect of interpolation, i.e. our assumption of continuity between two consecutive years in which UC is reported. In reality some of these cases will represent multiple, shorter claims. On the other hand, some of the claims that we have assumed to be for a year or less will exceed a year; and in any case we would expect the effect of censoring to be more significant.

14 This calculation may underestimate the frequency of repeated claims, since it is based on inter-wave interpolation. However, as above, we must also consider our assumption that each time UC is reported, it represents a full year of UC receipt. Some people with repeated claims may spend less than a year on UC in each claim; indeed we would expect there to be a negative correlation between claim frequency and claim duration. Depending on how the one-year limit was applied, these individuals may continue to be eligible for the mid-MIG for at least part of their subsequent claims.