GERS: Murphy’s Flawed Method

This week the Government Expenditure and Revenue Scotland (GERS) figures for 2016/17 were released by the Scottish Government.  Most people will have barely heard of them, not many will care much and even fewer will ever change their mind based on the contents of the report or blogs like this.

Despite this, stats nerds like me like to dissect the details and find it impossible to bite our tongues (or be-still our twitter thumbs) when the figures are misrepresented for political gain.  I imagine everyone reading this blog will be well aware of the various arguments, the excellent blogs from Kevin Hague  and Neil Lovatt, and my own humble efforts at dispelling some of the popular myths [LINK] and [LINK].

Recently those who find GERS inconvenient to their constitutional obsession have found new solace in the blogs of Professor Richard Murphy, a Professor of Practice in International Political Economy at City University London.  Professor Murphy has published several blogs in which he casts doubt not only on the conclusions of the report but the good faith with which it is produced.

There has been considerable debate driven by the contents of these blogs, not least from the aforementioned Chokkablog [LINK] quoting Professor Angus Armstrong and Professor Ronald MacDonald, or indeed (indirectly) the Fraser of Allander Institute [LINK]; people far more qualified than me to respond.

Distorted deficit

However, Professor Murphy’s latest blog (GERS: is this why it always says the Scottish deficit is so large? [LINK]) has prompted me to have a go myself.  In short, the blog appears to argue that GERS may be distorted because it might not include tax revenues created in other parts of the UK for spending that Scotland contributes to.  In other words, Scotland contributes to a share of UK spending on common services in, for example, Bristol.  This spending results in taxation (income tax from employees, corporation tax from profits on procurement contracts) that is accrued to England, not Scotland.  Further, this spending results in fiscal multipliers as it generates further indirect employment and taxation.

The spending Professor Murphy is referring to is ‘non-identifiable spending’, although he doesn’t explicitly refer to it as such, instead quoting the methodology’s section on the distinction between in and for spending.  Indeed it seems that he has only just discovered that non-identifiable spending exists within GERS which, for someone who has written over 6000 words professing his opinion on the legitimacy of the figures, comes as something of a surprise.

I’ve never been one for personal attacks, and I will try to avoid doing so here, but I find Professor Murphy’s entire approach to the topic rather bizarre.  Never more so than in the latest article where he seems to believe that GERS are unrealiable simply because the results sound ‘improbable’ to him and ‘the last thing they should do is trust that from London’.  With a belief in his own infallible ‘intuition’, he then goes looking for reasons to confirm it.

Non-identifiable spending

I have written on non-identifiable spending before [LINK] but at this point it may be wise to summarise what it is.  Essentially this is public spending that is conducted at a UK-wide level which cannot, or should not, be specifically assigned to one component part of the UK.  The classic example being Defence, which is conducted (whether you agree with the government’s policy or not) for the benefit of the entire United Kingdom equally.  Therefore everyone should pay an equal per capita share and GERS assigns the vast majority of Defence spending exactly thus as non-identifiable spending (NB: there are various methodologies for assigning non-identifiable spending, it is not always split on a per capita basis)

It is not necessarily spending in London or other parts of the UK, significant elements of non-identifiable spending take place in Scotland and plenty of it overseas.  It most certainly is not infrastructure spending in London or the South East; the myth spread by Wings Over Scotland that Scotland contributes to Crossrail or the London Sewers has no basis in reality whatsoever [LINK].

Spend and receive

In contrast to common misunderstandings of non-identifiable spending, Professor Murphy doesn’t appear to contest the validity of the charges to Scotland in GERS; rather he focuses on whether GERS contains an allocation of the tax revenue which has resulted from that spending.

As an example, the methodology above, produced in 2008 under an SNP government, explains the logic for assigning defence spending on a per capita basis – something I doubt many would disagree with. Meanwhile revenue, or tax income, is based on geographical burdens rather than contributions via non-identifiable spending.

So whilst Scotland contributes 8.3% of UK defence spending, GERS includes ‘only’ the taxes generated by military bases, personnel deployment and procurement within Scotland.  Professor Murphy argues that this is incorrect, that the “accounting basis is flawed and that it is very likely to seriously overstate the Scottish deficit as a result”.

There are two parts to that argument: first that the accounting basis is ‘flawed’, second that ‘correcting’ it would result in a massive reduction in Scotland’s notional deficit.

Now and later

I will let you make your own judgement on the first part, or rather wait for people more qualified than me to weigh in with their opinion.  As an amateur, I’d say this is a perfect example of what we’ve always said about GERS and indeed what the SNP mean when they say the figures are only representative ‘of the current constitutional arrangement’.  GERS shows Scotland’s position just now, so you tell us what would change with independence.  If you think ‘repatriating’ this non-identifiable spending would result in additional revenue for Scotland, thereby correcting some of the deficit gap, then fine.  Show us how.

Leaving that aside, however, the second part is perhaps the more interesting.  From where does Professor Murphy get the opinion that failing to include such revenues within GERS means “that is is very likely to seriously overstate the Scottish deficit”.  And what does “seriously overstate” mean, in the context of a £13.3bn, 8.3%GDP deficit?  Professor Murphy confesses to not knowing the figures with any accuracy but contests that “the sum in question is unlikely to exceed £10 billion” and “a significant sum is spent for but not in Scotland”.  Perhaps I can help shed more light on that.

Alongside the spreadsheets containing the summary charts and tables published in the GERS report itself, the Scottish Government very helpfully provide the database that is used to compute the figures [LINK].

Note the database is only available up to 2015/16, as the frontcover explains:

“This database is based on the CRA 2016, which provides detailed spending figures for the five years to 2015-16. As detailed spending figures are not yet available for 2016-17, spending figures in this database are similarly only available for the years to 2015-16.”

The figures below are computed by taking the average percentage of total spend in the HMT function group that was identified as non-identifiable for the preceding five years and applying this average to the total spend for 2016/17.  See Note 1 below on variance from the average, suffice to say Accounting Adjustments are the only likely significant variable.

Using the data contained therein, we can estimate the non-identifiable spend for the latest GERS (2016/17) (ID = identifiable, non-ID = non-identifiable).

 

Professor Murphy wasn’t far off in his assumption of the total, just exceeding his ten billion ceiling at £10,494m.

Of more interest is what makes up this sum.  In descending order, some items to note:

I’m going to exclude Accounting Adjustments (AA) from this discussion for the reasons provided above .  Total non-identifiable spending not shown in the table is ~£815m and follows much the same logic as those described below.

Debt and defence

The first thing to note is that the vast majority of non-identifiable spending comes from two pots – public sector debt interest at £3,226m and Defence at £3,050m.  Together, these make up over 70% of non-identifiable spending (with AA removed) and are both assigned on a strictly per capita share to Scotland.

Public sector debt interest – paid to investors, banks and pension companies across the UK (not to mention 30% of debt being foreign owned) – would appear to be irrelevant to this discussion.

Defence is the second largest sum at £3,050m.  For this Scotland benefits from the protection of nearly £36bn of Defence spending, 7.5% of UK-based military personnel  [LINK] and 8.04% of MoD regional spending – just £10 per capita below the UK average [LINK].  Should Scotland be assigned a proportion of the fiscal multiplier from Sandhurst?  In that case, shouldn’t Wales be assigned a proportion of the fiscal multiplier from Faslane, for which Scotland contributes just 8.3% of the cost and receives 100% of the benefit to the regional economy?

I struggle to see the value in such petty tit-for-tat.

International services also provide a significant chunk of non-identifiable spending, at £793m.  This consists almost entirely of foreign economic aid, with no revenue implications, alongside around £150m for the administration of the Foreign and Commonwealth Office (FCO).

Social protection at £477m is largely pensions for those who now reside overseas.  Clearly there is no ‘lost revenue’ to Scotland there.

And the last significant item is Public and Common Services (£407m) – which essentially refers to the administrative organs of central government and is dominated by ~£250m for the services of HMRC. Professor Murphy chose to specifically reference such civil service as an example of misallocated revenue: “Take an example of spending on the civil service in London charged to Scotland in GERS. The cost is in GERS. But where is the revenue? That’s in south east England.”

Again, however, it is worth pointing out there is significant civil service employment in Scotland.  10.3% of civil service employees are based here, significantly higher than our 8.3% population share.  The argument has been made that higher paying, and therefore higher taxed, roles are based in London.  This is true but with the Scottish civil service median income at 92% of the UK average, the difference would not appear to be significant [LINK].

The majority of the non-identifiable cost for public services assigned to Scotland is for HMRC, for which ~£250m is included in GERS – an 8.3% population share of the total cost.  Yet 12.95% of HMRC employment is based in Scotland, well above that population share.  Similarly 11.44% of Department of Work and Pensions staff are based in Scotland.  So rather than losing out on revenue accrued from Scottish non-identifiable spending, it could be argued Scotland disproportionately benefits from this arrangement.

What missing revenue?

So again I come back to the question of how Professor Murphy has reached the conclusion that attribution of revenues accrued from Scottish non-identifiable spend would remove an assumed “serious overstatement (of) the Scottish deficit”.

As a reminder, GERS 2016/17 shows Scotland with a notional fiscal deficit of £13,300m.  Scotland’s non-identifiable contribution to the UK civil service is about £400m a year.  We know that a significant proportion of this, if not all of it, is actually spent in Scotland but let’s pretend we aren’t aware of that and assume it is all spent in London instead.  Even utilising the largest fiscal multiplier Professor Murphy’s own blog suggests (at a factor of 2.5 [LINK]) for that entire sum, we’d be estimating a £1,000m increase to GDP/GNI.  With revenue rates around 37% of GDP that would be an estimated increase in tax take of £370m.  Or a reduction to the deficit of just 0.23%.

And that’s ignoring the significant portion of non-identifiable spending which already takes place in Scotland, for which you’d need to provide an additional, and not insignificant, estimate.  Not to mention non-identifiable spending in Scotland for which we actually contribute a lower proportion than is actually spent here – a good example being nuclear decommissioning for which we contribute an 8.3% per capita share but see 15.6% of total UK spend take place here.

As an aside, I do find it curious that a man who has derided the conclusions of GERS based on their reliance on mere estimates has now decided that the answer to his confusion over their “improbable” conclusions is that more estimates are required.  It is seemingly implausible for the Treasury to accurately estimate Scottish income tax revenues but their refusal to provide an estimate of the revenue accrued from Scottish non-identifiable spending on civil service employment across the rest of the UK, along with the resulting fiscal multipliers, well that’s just unforgivable.  Clear evidence that Whitehall wishes to starve Scotland of data.

So where is the material change here?  Where is the missing revenue that has created this “serious overstatement”?  I struggle to believe that anyone who has studied the inputs to GERS, reviewed the methodology and understands what constitutes the non-identifiable spending would go on to claim that revenue accrued could go on to fill a £13bn deficit.

Again, I’m forced to conclude that Professor Murphy has reached his conclusion first and then gone looking for reasons to justify his argument.  I suspect he’ll soon find that he needs to keep looking.

This is an abridged version of the analysis frst published in full (with notes) on the author’s site