In terms of our methodology, we looked at impact on spending of welfare policies scored by the OBR at end of forecast window for each fiscal using the Policy Measures database at obr.uk/data/ & converted these into real terms using GDP deflator, so they're comparable.
The fact I had to look at 400 rows for welfare changes in last 30 fiscal events (i.e. >10 per event) shows the "absolute lack of stability" for recipients @RattusMalumus describes here x.com/RattusMalumu..., with 20 of the 30 resulting in less money being spent on welfare.