subtitle

...a blog by Richard Flowers

Wednesday, June 18, 2014

Day 4916: You Can Prove Anything With Statistics

Tuesday:

(Warning: Contains Maths!)

I've seen a couple of people linking to this story: "British public wrongly believe rich pay most in tax, new research shows".


It claims that most people want a tax system that is fairer (good) but that they are wrong to think it's fair now (bad!).

Of course, this is a piece in the Grauniad cherry-picking from a report for The Equality Trust cherry-picking data from the Office for National Statistics and before you can say "confirmation bias" it's proved to the Internets that the Evul Condums are Evul.

Except, of course, it's not true.

Obviously, there's the usual exaggeration by some Graun sub-editor in the headline: we've lost the nuance of "as a proportion of their income". Of course the rich pay most in tax. 35% of more is obviously more. The question is do they pay a higher SHARE (we'll not even get into SHOULD they pay a higher share; we'll take that as read).

Then there's the point that the Equality Trust's research is actually a poll into public perceptions, NOT research into the effects of the tax system.

Just because they're a charity doesn't mean that we should not be cautious of this sort of polling – it's very similar to the sort of puff piece that marketing teams place in papers all the time, you know the sort of thing: "90% of housewives say sunny days are nicer says poll for insert name of suntan lotion retailer here". It's about getting their name in print – i.e. advertising.

Another point that's interesting is that the ONS data excludes Capital Gains Tax and Inheritance Tax, which amount to about eight billion quid between them, not a lot in the grand scheme of things – about 2% of all taxes, but more than cigarette duty which the figures do include – and these are taxes that tend to impact the higher earner more. But the report's methodology does say they specifically do mention Capital Gains Tax when asking poll respondents to think about the taxes that they pay. So you're asking people to remember a tax paid mainly by the better-off that you then exclude from your calculations.

In addition, the ONS figures show that most households benefit "in kind" mainly from state-provided schools and health service equivalent to about seven-and-a-half thousand pounds (which again is less for the top earners, I'd guess because of higher take-up of private schools and health insurance by those able to afford it). The Equality Trust report does not include this benefit in the "total of benefits and income" which they use to compute the tax rates. You may say "fair enough" (and I'd probably agree), but it is still a bit dodgy to ignore SOME of the ONS data you claim to be using.

And there's the very serious fact that the figures are based on EARNINGS, rather than WEALTH. Earnings at least demonstrate some work being done. It may not be entirely USEFUL work, but at least it's being done, whereas wealth, particularly when tied up in land, is more often a dead weight or leads to rentiering.

But, if we move past all of that there is still this BIG problem: the central assertion that the public believe that the tax system is fairer – and that they would like it to be fairer still (nicely coinciding with the aims of the Equality Trust, of course, but fair enough) – and that this belief is WRONG.

This rests on the claim that the tax system actually isn't fair. In fact, the report says in these words:

"The UK’s tax distribution is not only less progressive than the public’s perceived and preferred distribution, it is actually regressive when comparing the richest and poorest 10%"

This is, at best, in error and more than likely actually deceitful.

A regressive system would take money from the less well-off (in this example the lowest 10%) and transfer it to the better-off (i.e. here the top-earning 10%).

And that is very clearly NOT what is happening.

Why not? Because the way they have sliced the data is to compare income INCLUDING BENEFITS against taxes paid.


Here's now the report puts it:

10% of households with lowest income:
Total Income and Benefits: £10,253
Total Direct and Indirect Tax: £4,424
Income after taxes: £5,830
Effective tax rate: 43%

10% of households with highest income:
Total Income and Benefits: £101,291
Total Direct and Indirect Tax: £35,627
Income after taxes: £65,664
Effective tax rate: 35%

Cue shock and outrage!

But let's cut that data up a slightly different way:

10% of households with lowest income:
Total Income: £3,835
Total Benefits less Direct and Indirect Tax: £1,994
Income after tax and benefits: £5,830
Effective tax rate: -52% (yes, that's a negative tax rate of MINUS 52%)

10% of households with highest income:
Total Income and Benefits: £101,291
Total Direct and Indirect Tax: £33,424
Income after tax and benefits: £65,664
Effective tax rate: 34%

Clearly there is a net cost to the top 10% and a net benefit to the bottom 10%. Actually, all the lower four deciles (or 40% of households) receive more in benefits than they pay out in taxes.

The combined tax/benefit rates for all deciles are as follows:

Bottom: RECEIVE 52% (receiving more than paying)
2nd: RECEIVE 49%
3rd: RECEIVE 26%
4th: RECEIVE 13%
5th: PAY 9% (paying more tax than receiving benefit)
6th: PAY 15%
7th: PAY 24%
8th: PAY 30%
9th: PAY 33%
Top: PAY 34%

On that basis the tax and benefit system is pretty positively progressive.

Far from being wrong because the system is unfair, the public hugely underestimate how much the least-well-off are helped. (And probably just as vastly underestimate just how little the least-well-off get paid!)

Does this just mean we're playing with numbers matter? Does it just mean you pays your money and takes your choice? As Obi-Wan Kenobi puts it: "it all depends on a certain point of view". I don't think it does.

To ignore the fact that benefits are a part of the government's effect on household incomes is absurd; worse, it distorts the picture entirely. It suggests that the lowest-earning households actually pay more in tax than they earn altogether, which is clearly impossible.

It matters that statistics are used to tell a story that is true. And by abusing the ONS's numbers this report doesn't just discredit their own version – they discredit mine. People will just go (as my title suggests) "ugh, maths means nothing". Far too many CiF commentators are ready to leap to their own prejudices that Evul Condums are Evul and they pick the numbers that support that story.

(And there's enough "maths blindness" in the world as it is, without deliberately reinforcing it.)

As a Liberal I am anyway naturally wary of putting a society that is "more equal" ahead of a society that has more freedom.

I want to lift people out of poverty so that they are free to live their life the way they want; I care less about how much they earn after that, so long as it's a fair return for their work.

I do think that the tax system does need to be a lot simpler and clearer – it would be fairer if it were easier to understand what tax you pay.

And if we are to balance the government's budget then – like Cap'n Clegg – I say that it's right to ask the better-off to be first to contribute. I'm just not sold on raising taxes – particularly not to punitive levels – as a "good" in themselves. There MAY be something in "The Spirit Level" idea that more equal societies are healthier ones, but I find they (too) pick their evidence to agree with their case.

If we deplore – and we do – Tory (and Labour) ministers pushing the lie that "scroungers" and "benefit fraud" are bankrupting the economy, then we must equally deplore the reverse when I'll have to label them broadly "the left" say that the government are ripping off the poor to pay to the rich. We should not endorse wrong figures from either end of the spectrum.

It is GOOD that the public want and endorse a fair, progressive tax system. Lying to them about it will only induce more apathy and KIPpery. Let's not.

PS:
I encourage you to check sources:

The Equality Trust's report [pdf]
and
The Office of National Statistics data