Earlier this week someone asked me whether the forecast GST figures were really based on evidence.
I replied that forecasts are a bit like Guernsey weather. I pay attention to them, but I still glance out the window before deciding whether to take a coat.
In other words, the estimates presented in the tax policy letter may not be wrong. But that doesn’t mean I’m prepared to accept repeated assurances that the modelling is ‘robust’, without understanding what sits underneath it.
So that’s what I’ve been doing.
I’ve been working through the figures, comparing public statements with written responses, examining the data, talking to those who crunch the numbers, and testing how Policy & Resources has reached its estimates, rather than simply accepting them as fact.
And the more I look, the more I worry.
Take household spending.
To estimate how much Guernsey spends, P&R has relied on household expenditure data from May 2018 to April 2019.
I think we could all agree that the world has changed dramatically since then. A global pandemic, sustained inflation, increased military conflict, geopolitical instability, the explosion of generative artificial intelligence, and another presidential win for Donald Trump.
Closer to home, Guernsey faces some of the highest electricity prices in the world, a housing crisis, a cost-of-living crisis, and wages that have barely shifted in years. None of which show any sign of easing.
If anything, GST simply adds another bill to the pile.
I can’t help thinking we’re living through one of the strangest periods in modern history. The world has become one enormous head-scratcher. Every week seems to bring something that would have sounded absurd only a few years ago.
Which brings us to the most obvious question of all – are we really spending the same way we did seven years ago?
This matters, and matters a lot, because this assumption does a lot of the heavy lifting in the ‘household’ revenue estimate. In other words, the committee’s estimate of how much GST it expects to collect from ‘households’ hinges almost entirely on the answer.
What concerns me is that the policy letter appears to assume spending habits haven’t changed in any material way. In seven years. In this climate. No modelling appears to have been done. If this single assumption is off, the revenue forecast will be off. If it is significantly wrong, the revenue forecast will be significantly wrong too.
It feels a bit like using yesterday’s weather forecast to decide whether to pack an umbrella for next year’s outdoor birthday party.
Standard practice for a forecasting exercise of this importance would be to revisit that baseline, update it, or, at the very least, explain why no adjustment is needed.
This does not appear to have been done.
Instead, data from 2018/19 has been carried forward with only limited consideration – and certainly no modelling – of how people’s spending habits have changed.
The committee also says GST will increase inflation by 1.9%. Many rental agreements have inflation-linked increases baked in. Higher rents mean less disposable income. Less disposable income means less spent on goods and services. Less spending means less GST collected. Yet, once again, there appears to have been no adjustment to the revenue forecasts to reflect this.
It is also stated that ‘no explicit allowance has been made for either a reduction or an increase in consumption following the introduction of a GST at 3%’. In other words, behavioural change has (again) not been explicitly modelled. That leaves a central driver of revenue – how people respond – effectively untested in the numbers.
Nor does this issue appear confined only to ‘household’ revenue.
Despite GST representing one of the most significant tax changes in decades, there is no clear evidence of structured modelling of how this will affect local businesses or the economy. Instead, the approach relies heavily on ‘discussions’ with businesses, rather than systematic analysis of pricing, investment, demand impacts, and so on and so forth.
The tools underpinning the forecasts also raise serious concerns. One model is based on an Excel workbook developed in 2017/18 using 2019 data. It is clunky, slow and difficult to update. A second model, built in 2018/19, relies on similarly dated inputs from 2020. Both appear to be operating well beyond the environment they were designed for. The latter in particular struggles, taking up to half an hour to tackle a single simple edit.
There is also the question of data quality. The States’ Treasury lead recently described the Revenue Service as being in ‘special measures’, citing significant backlogs and reliance on external support. Yet GST modelling is heavily dependent on data from the same system. This naturally raises further concerns about how reliable the long-term projections are.
Even the pensions information in the policy letter appears to rely on a report from 2020.
All together, then, this does not look or feel like an up-to-date or properly stress-tested forecasting framework. It can hardly be called ‘robust’. It is two outmoded models, heavily dependent on legacy data and untested behavioural assumptions, presented with a level of confidence that the underlying evidence simply cannot justify.
What is also striking, to me at least, is the apparent double standards.
Those daring to question GST are met with the same old, ‘show your own workings’. But the same expectation does not seem to apply to the Policy & Resources Committee. Key assumptions remain inaccessible, questions are either ignored or conflated, limiting meaningful external scrutiny or challenge.
So far, only three members not belonging to P&R have reviewed the supporting material behind the policy letter. I am one of them. And this article comes off the back of my review.
The rest of the Assembly is therefore being asked to accept conclusions without seeing the evidence base in full. That is not a strong position for the Assembly to find itself in less than two weeks away from voting on major tax reform.
Which leads me to my next point.
I do not accept the argument that the underlying data or modelling is ‘too sensitive’ to be published. I have seen an extract of the workings, along with Deputy Camp and Deputy Collins, and what we were shown was anonymised and aggregated data. There was no personal or identifiable information to be seen.
There is no conceivable way I could identify you or your household from any of the data, which is essentially a best guess of what you might spend in a year. Heck, I can’t even identify my own, my wife’s, or our household spend.
On this basis, Deputies Camp, Collins and I agree that the argument this information cannot be shared on the basis it is ‘too sensitive’ has not been convincingly made out.
If the modelling is indeed ‘robust’, publishing the workings should not be a risk. It is standard practice in most comparable jurisdictions. And the fact this has not been done here raises yet another red flag among many.
Deputies are elected to represent the people and to scrutinise proposals on behalf of the people, not to blindly accept the assertions of a committee or its individual members. Scrutiny depends on access to the evidence, and clear and simple explanations, not just the conclusions drawn from who knows what.
The easiest and quickest way to rebut any of the conclusions or claims in this article would be, as a member of P&R recently suggested, to ‘show your own workings’.
Because people are becoming increasingly fed up of statements with nothing to back them up.
