IFS report- how to make numbers say what you want
Andy Cooke on Political Betting has written a very well argued critique at post 168 of this thread of the IFS report about whether the budget hit rich people or poor people hardest.
His starting comment is
" 168. I’ve had more of a poke around the IFS report. It’s an impressive example of how to present facts to support whatever case you want to make." and his conclusion is "It’s intriguing how they’ve taken an analysis that shows that the Budget is definitely progressive up until 2012 and managed to present it as “regressive”.
Main points from his argument:
* Appendix A of the report shows the Budget effects without their admittedly uncertain estimates of the effects of renewed DLA testing and the Housing Benefit effects, as well as the Tax Credit changes.
"If you pretend that Osborne had no power to change Darling’s pre-announced changes (that they are binding on him and were the baseline), then the blue bars are what you’ll concentrate on. If you want to see the overall effect of the Budget (assuming that Osborne had the power to implement whatever changes he wanted to, comparing the output of the Budget with what the state was before) then the black line marked “Total” is the overall effect of the Budget. Because that is the effect of the Budget. You’ll note that the poorest deciles are best off (1st 2nd and 3rd deciles are no worse off and in most cases better off), the middle deciles are slightly worse off (gentle downward slope from 4th to 8th deciles), a downward jump for the 9th and a big downward jump for the 10th."
(My use of bold text.)
Appendix B contains the details of how the IFS came up with their changes, and particularly some less than rigorous assumptions about Housing Benefit changes which are assumed to have all their incidence on tenants rather than landlords (very unlikely), Disability Living Allowance, and Tax Crfedits. In Andy's words
"Housing Benefit: They use the DWP paper, whose core assumption is “assuming that they would be renting at the same rent level in the same property and with the same household composition. No behavioural changes have been assumed, such as customers moving to a cheaper property or landlords reducing their rents. As a result, when we report ‘losers’ or ‘losing out’, these could be actual losers (seeing their benefit decrease) or notional losers, meaning that they would not see any benefit decrease, but would receive less HB compared to what they would have done under the previous scheme. So, for example, a new LHA customer applying for benefit after measures take effect may ‘notionally lose out’, meaning that they would receive less than under the current arrangements.”
They then (in essence) assign the savings estimated amongst the claimants as if it’s coming out of their pockets.
The other two are done as follows:
Disability Living Allowance
“The Budget policy costings document says that the effect of this reform will be to remove DLA from around 20% of claimants. We randomly remove entitlement to DLA from the appropriate number of claimants in order to match the long run saving from this policy (around £1.4 billion).”
Tax credit reforms
“There is no way to identify those who will be affected by changes to the way in-year awards are calculated so we simply reduce all tax credit awards by the same percentage amount so that the total amount saved from the policy is correct (around £1 billion).”
These aren’t exactly rigorous techniques, so I’d say that the published headlines are not exactly sturdy. Even with all this done, they have to gloss over the fact that the richest decile still bear a far greater share of the impact than the poorer deciles (in table 2.1) and the expenditure tables in Table 3.1 (which they’ve previously argued (when exploraing VAT effects) should be taken as better indicators when indirect taxation is looked at (overcome effects of wealth-rich but income-low students, etc) still have a total effect that’s progressive pretty much throughout.
Andy Cooke says that he is "disappointed" in the IFS. Indeed.