Wednesday, 6 September 2017

Another Example of Murphy's Flaw

A quick couple of observations on the specifics of Professor Richard Murphy's latest foray into the GERS debate.

Firstly, some context.

As covered in previous blogs (e.g. here), Murphy is on record as asserting that the GERS figures are “untrustworthy”, “rigged by Westminster”, “literally made up” and “nonsense”.  Incredibly in his latest foray (here) he’s even suggested that those using the GERS methodology are “risking the allegation of professional misconduct”. So that includes: the Scottish Government’s Chief Economic Adviser; The Scottish Government economists in St Andrews House; The Office for National Statistics (who judge the methodology to pass the standards required to qualify as National Statistics); presumably also those Nobel Laureats who relied on GERS when they sat on the SNP’s Fiscal Commission Working Group.

He's been shown to have made these allegations based on a fundamentally flawed understanding of the facts (e.g. here) and frankly the above should be enough to disqualify Murphy from any civil debate on this subject. But some seem determined to take him seriously, so let's attempt to ignore his previous form and focus on the specifics of his latest contribution (here).

The thrust of his latest argument is based on the (widely understood1) fact that money spent for Scotland (allocated to our spend in GERS) isn't always spent in Scotland. Murphy argues that because we don't allocate any of the tax revenue that may be indirectly generated by that spend back to Scotland, there's a flaw in the GERS methodology. He boldly concludes (without any supporting analysis or evidence) that this "flaw" means GERS "is very likely to seriously overstate the Scottish deficit as a result".

This blog will explain why he's simply wrong, again.

There are two big problems with this argument

1. He appears to still not understand what the GERS report actually is

GERS provides historical actual information - it describes what has actually happened under current constitutional arrangements. It's perfectly fine to ask the question "if some of the spend that's allocated to Scotland but not spent here was spent in Scotland, what might the impact on taxes raised be?" - but that's not the job of GERS.

For example, the Common Weal White Paper project assumes £50m additional tax revenue would be generated in Scotland as the net result of reducing the defence budget but spending more of it in Scotland, and that £719m of new taxes would come from "relocated government activity". This blog will show why those assumptions aren't realistic, but at this point let's simply observe that this is how this effect is factored in to the independence debate: by modeling alternative scenarios.

So to be clear: nobody suggests that the historical actual figures in GERS can or should represent an alternative hypothetical future - that's not what GERS is for. The SNP's own White Paper correctly described GERS as “a useful indication of the relative strength of Scotland’s public finances as part of the UK and a starting point for discussions of Scotland’s fiscal position following independence”.

Suggesting GERS is flawed for not showing what would happen if money that isn't spent in Scotland was (and vice versa) rather misses the point of what GERS is.

2. He fails to grasp the concept of materiality

We can embark on a thought experiment here, which shouldn't be mistaken for some kind of tacit acceptance that the GERS report is currently flawed. The GERS figures are precisely what they claim to be, no more and no less. But as a thought experiment, we could consider what would happen if we were to decide to change what GERS shows. We could say that instead of accounting for tax revenues based on where they're raised, when those tax revenues are somehow related to Government expenditure we could try and allocate those revenues on the same basis as the related expenditure has been allocated in GERS.

First of all, this would require some heroic assumptions.

To illustrate the general point with a specific example. Consider a Scottish pensioner taking a trip to London and spending some of their money in shops on Regent Street. They'll generate VAT in England and contribute revenues to businesses who employ people and pay taxes in England. To follow Murphy's logic we'd need to allocate a proportion of that VAT and other taxes back to Scotland because the money to generate those taxes was a cost to Scotland - if Scotland didn't pay the pension, that money wouldn't get spent in London. The absurdity of this argument is obvious - it's technically correct but it would be impossible to robustly calculate, it's an effect which happens in both directions anyway and - and this is the key point - it's most certainly not material to the figures we're dealing with.

Murphy is on record as suggesting that using statistically significant sample data to attribute some of the revenue lines in GERS (such as VAT, presented in GERS with explicit confidence intervals)  makes the report "nonsense". What would he call figures which made finger-in-the-air fiscal multiplier assumptions and had a go at guessing, for example, how much tax was generated from spending in Edinburgh during the festival where the money spent ultimately came from English social welfare spending and so shouldn't be allocated to Scotland?

OK so these second and third-order effects are amusingly daft examples, but they serve to illustrate the problems that arise if we head down this path. Let's carry on anyway and focus just on the directly identifiable examples in GERS and assess the materiality of the issue: how much expenditure is attributed to Scotland which generates taxes in rUK and how much is attributed to rUK which generates taxes in Scotland?

First we should consider what our materiality threshold should be. What is a significant issue in the context of figures explicitly stated as being accurate (with 95% confidence) to +/-£729m. Does this affect make a jot of difference when we're looking at a deficit of £13.3 billion, a deficit gap with rUK of over £10 billion?

I'd suggest we need to be talking about a change of £0.5bn or more and for that change to be systematically in one direction or another to become material in the context of the independence debate.

Assuming we ignore the "travelling consumer" argument above and focus on the possibly material issues, this isn't too difficult a scoping exercise. By digging about a bit - as the ever diligent Fraser Whyte has done in this excellent blog - we can find out roughly how much is allocated as spent "for" Scotland and then work out whether materially more or less than that is actually spent "in" Scotland.

Here are the figures which are allocated as spent for Scotland - about £10.5bn

With the help of both Fraser's blog and this blog from Fraser of Allander Institute, we can eliminate from that £10.5bn some figures which are irrelevant to this debate;
  • £3.2bn of Public Sector Debt Interest. The issues around what public sector debt interest an independent Scotland would be paying are complicated and fraught, but not relevant here
  • £1.7bn of Accounting Adjustments. These are basically technical accounting adjustments2 , so again irrelevant here
  • £0.8bn of International Services. This is mainly foreign economic aid and expenditure on embassies - money spent overseas, not in rUK
  • £0.5bn of Social Protection - mainly pensions paid to UK citizens resident overseas
So that leaves us with the following that may have material "fiscal multiplier" effects on UK tax revenues:
  • £3.1bn of defence expenditure
  • £1.2bn of "other", which includes
    • £409m of "public and common services" (central government administration, which will include the "spending on civil service in London" that Murphy uses as his example)
    • £294m of "recreation, culture & religion" (the BBC, museums & galleries etc)

Let's deal with defence first of all. The MoD provide information on their regional spend breakdown so we can do some quick back-of-the-envelope calculations to scale the difference between UK tax generating spend "on" and "for" Scotland: the figure is likely to be less than £150m3 more allocated to Scotland than spent in.

The pro-independence Common Weal White paper suggest a fiscal multiplier of 0.9 for defence spending which I presume they're defining as tax generated/spend made - so to be generous we could argue to transfer £135m of tax from rUK to Scotland in the GERS figures if we follow this methodology.

But of course most debates about independence start with an assumption that Scotland would not spend as much on defence as allocated in GERS. The SNP White Paper suggested saving £0.5bn, the current Common Weal "White Paper Project" suggests saving £1.1bn. So the common assumptions used for independence appear to be that we'd actually spend less in Scotland than is currently the case - so our tax revenues would be lower than that currently stated in GERS. Either way we can say with confidence that the fiscal multiplier effect on allocated defence spending is simply not a material issue.

So what about the £1.2bn of other?  Well we have some proxies we can use to see whether it's likely to be biased in one direction or another. As Fraser points out in his blog on this topic: 12.95% of HMRC employment is based in Scotland and 11.44% of Department of Work and Pensions staff are based in Scotland. So in these cases at least the spend "in" Scotland will likely be greater than the 8.2% of spend currently allocated "for" Scotland. Spend on the BBC will be the other way around: more is allocated to Scotland than spent in Scotland. Nuclear decommissioning costs are an example where Scotland will be favoured by the current methodology because, as Fraser Whyte rightly points out, 15.6% of total UK nuclear decommission spend take place in Scotland but we're only allocated our 8.2% per capita share. We could go on, but I'd suggest there's no evidence of a systematic bias in one direction or another in these other "non-identifiable spend areas" (or if there is, I haven't seen it).

That Common Weal paper assumes a fiscal multiplier of 0.6 for "relocated government activity" producing £719m of "new" revenue. They provide no audit-trail for that figure and I'm not surprised - the quick exercise above shows it's completely nonsensical. I suspect they've assumed a load of allocated cost moving in but no costs moving out, but without an audit-trail we can't be sure.

Again: in any debate about independence the question of whether the costs that would replace our share of the UK "scale" functions would be greater or less is a complex one. The White Paper suggested "transitional arrangements" where Scotland would pay rUK for some of these currently centralised administrative functions- so there wouldn't be an immediate shift from "for" to "in" anyway.

Enough. Let's remind ourselves that Professor Murphy asserted this issue is "very likely to seriously overstate the Scottish deficit as a result". It's yet another extraordinary and unsubstantiated assertion on his part. Not only has he not demonstrated the materiality of the issue, he hasn't even demonstrated the direction of the impact.

Yet again I find myself wondering why he is taken seriously in this debate when he makes headline-grabbing claims like "very likely to seriously overstate" with absolutely no analytically quantified foundation. In fact the simple exercise above is enough to show us that in the context of a £13.3bn deficit (+/-£0.7bn) the issue he highlight is not of material significance.

I'm not alone in this view - Fraser of Allander concluded
Changing assumptions about how much spending is allocated ‘for‘ Scotland or spent ‘in’ Scotland in GERS will change the net fiscal position. But any revisions are relatively small.
I'd suggest this blog supports that conclusion and would maybe go a few steps further
  • Any revisions would require assumptions that would make the figures less robust, not more so 
  • It is unclear if the net effect would be to increase or decrease Scotland's GERS deficit
  • Any adjustments would be highly unlikely to exceed £0.5bn in any one direction - hardly material in the context of a total £13.3 billion deficit and a deficit gap of over £10bn between Scotland and the rest of the UK



1. Widely known: Page 8 of GERS

2. Accounting Adjustments (page 8 of GERS)

3. "back-of-the-envelope" in vs for defence spending based on MoD breakdowns
  • Expenditure with Industry & Commerce: a difference of just £36m
    • £2,482m is spent outside the UK
    • £18,745m is spent in the UK - Scotland is allocated 8.23% of that in GERS = £1,543m
    • The amount actually spent in Scotland is £1,507m - so a shortfall of just £36m
  • UK expenditure not "spent with UK industry" = £37,000m - £2,482m -£18,745m = £15,773m, 7.6% of MoD personnel and UK regular forces are employed in Scotland (source here). The difference between 7.6% and 8.2% is 0.6% -- 0.6% of £15,773 = £95m

No comments: