The secret political battle over the States’ tax modelling shows no signs of dying down as both sides dig in for a war of words ahead of next week’s GST showdown.
An explosive email chain seen by Express has exposed a fierce, behind-the-scenes battle between backbench deputies and civil servants – as well as Policy and Resources (P&R) – over the numbers underpinning the latest tax proposals – as well as the States’ £8k GST calculator.
Tax rebel Deputy Rob Curgenven demanded answers from the States’ treasury team after an independent analysis of the calculator appeared to show it was under- and over-reporting the effect of the tax changes for some local residents.
Deputy Curgenven also repeated his call for civil servants to share their workings, after consistent claims from P&R that the models contained sensitive data which made it illegal to share them.
A senior government economist immediately hit back, firing off a defensive rebuttal that protected the civil service’s calculations and justified the lack of public transparency.
However, that official defence was in turn attacked in a scathing, point-by-point counter-rebuttal by Mr. Collings.
The software expert warned that the civil service’s technical excuses were fundamentally flawed, adding severe weight to doubts over whether officials actually understand their own numbers.
The leaked correspondence reveals that the walls of secrecy around the £8,000 tool have completely broken down as backbench politicians and the public demand total transparency over the data.
Here’s the leaked email chain in full…
Deputy Curgenven questions the numbers again
Dear Members,
I found a few copies of this email (copy & pasted below), for reasons unknown, in my spam box. Could P&R please confirm whether it is accurate – before the debate?
Thanks.
Pick a Number: a look at the GST proposals
There is a lot to like about the 2026 Tax Reform Package calculator. Genuinely. It loads quickly. It works offline. It doesn’t phone home with your salary details, and for the first time in history it might be a government IT project which hasn’t gone millions over budget. And if you are a single employed person with no mortgage, no savings, no partner and no car, it will tell you exactly what the policy letter says it should. Congratulations to that person, whoever they are.
For everyone else, the experience is more of an adventure.
We spent an evening doing something P&R presumably hoped nobody would bother to do: we sat their calculator down next to their policy letter and asked both of them the same questions. Not our numbers against theirs — theirs against theirs. The policy letter of 8 June and the calculator of the same summer, written by the same organisation, describing the same tax package. You would expect them to agree.
Reader, they do not always agree.
What P&R has, in effect, is two prospectuses: one package being sold to the deputies who will vote on it, and a rather different one being sold to the public who will live with it.
Take a household earning £48,000. The package being sold to deputies — Appendix 8 of the policy letter — leaves them about £450 a year better off. The package being sold to the public leaves them £24 better off. A £44,000 single pensioner is £703 better off in the deputies’ version and £218 *worse* off in the public’s. A £66,000 pensioner couple gets it the other way round — £396 down on the deputies’ desks, a cheerful £109 up on the public’s screens. Same States, same month, and up to £900 a year between the two sales pitches. Which one you’re quoted turns on an optional savings box most people will scroll straight past, and nothing on the results page thinks you need to know that.
It gets better. The public’s package contains a policy the deputies have never been shown: a transferable tax allowance for unmarried couples. In the version before the States, the transfer of unused allowance is strictly for married couples and civil partners. In the version the public is testing their finances against, any two adults sharing a household qualify — £15,800 of allowance, no ceremony required. The calculator does ask whether you’re married, out of politeness apparently, since the answer changes nothing. Romantic, progressive even, and worth several hundred pounds a year to Guernsey’s cohabiting couples. It’s just not in the package anyone gets to vote on.
Homeowners do well too, because the public’s package includes something the deputies’ package doesn’t: a mortgage-interest GST rebate scheme. The version before the States offers relief on mortgage interest capped at £3,500, and there it stops. The public’s version keeps going — for every pound of interest you pay, it quietly hands back three pence of your GST, with no cap at all. That’s worth £600 a year to a homeowner with £20,000 of interest, which in our tests was enough to wipe out most of their GST bill entirely. It’s a lovely scheme. Someone should show it to the deputies.
Not everyone is so lucky. An employed person earning £10,000 in wages with £30,000 of income from investments is told they owe £2,454.63 a year in contributions. In the package before the States they owe nothing: wages below the allowance attract no contribution, and the investment income of working people is out of scope. The public’s package takes a different view — it quietly extends contributions to the unearned income of the employed, a brand-new charge no deputy has been asked to approve. The couples and the homeowners got their unannounced giveaways; this household gets the unannounced tax rise, and walks away from P&R’s own tool believing the reforms cost them two and a half thousand pounds a year that, in the version being voted on, they don’t.
The self-employed get a windfall of their own. The deputies’ package includes a 2.5% charge on self-employed earnings up to £300,000 — it’s in the policy letter, and the States will be voting on it. The public’s package has quietly abolished it. That’s worth £2,586 a year to someone earning £300,000, and it has been announced precisely nowhere. We’d suggest the successfully self-employed not spend it just yet.
And for connoisseurs of small print: the policy letter in front of deputies promises the new vehicle tax runs from £25 to a maximum of £280; the figure for the public goes up to £300. The Essential Costs Relief figures are provided to the public — £310, £520, a cliff edge at £32,400 — but appear nowhere in the policy letter to deputies. That £32,400 does ring a bell, though: it’s a number from the 2024 package, the one the States rejected. The browser tab, for what it’s worth, still reads “GST-Plus Tax Calculator” — the last package’s name. It’s nice to keep souvenirs.
Which leaves the interesting question, and it isn’t a technical one. Deputies are being asked to vote on one version of the biggest change to Guernsey’s tax system in a generation, while islanders form their views — and write to those same deputies — on the strength of another. Either the committee cannot keep its own package straight from one audience to the next, or the differences suit somebody. We honestly don’t know which. We’re not sure which would be more reassuring.
But it does seem the sort of thing you’d want sorted out before asking an island to trust the maths.
Deputy Rob Curgenven
States economist fires back
Good afternoon,
Thank you for raising this. I can see where the confusion may have arisen, but the Policy Letter and the calculator present different things.
Without full details of exactly what parameters were put into the calculator I cannot be specific, but the differences referenced here will likely reflect the difference between the impact on an individual, according to the specific set of parameters entered, and the average impact of all the households in our data set (or in a specific group within it) presented in the Policy Letter, which will have a wide variety of variables.
The data in the Policy Letter, because it aggregates data for households of all manner of shapes and sizes, also uses a process called “income equivalisation”, which is a set of recognised standard adjustments used in analysis to reflect the number of people who are being supported on an income. Think of it as adjusting for the fact that you can achieve a better standard of living as a single person on their own with £50,000 than if you are a couple with a total of £50,000 of income, but not twice as good. The data and analysis unit include a good explanation in their Household income publication here: CHttpHandler.ashx.
Applying equivalisation adjustments is not necessary (and is actively confusing) when you are looking at the outcome for a single household in isolation for presentation to the general public, so this is not something the calculator does.
Regarding some of the other issues raised in this and other posts in regards to the calculator:
The calculator is designed to give an indication of the impact for the majority of households with common circumstances. It is less accurate when dealing with marginal and unusual cases because it needed to be kept simple to function as an online form and to avoid having to ask for too much detail from users. The treatment of unearned income is an area which is difficult. This is clearly stated in the caveats and health warning for those using the calculator.
In the case below referencing an individual with employed income of £10k and unearned income of £30k, it would be a very niche and marginal case with some complex considerations, in part because the current social security contributions set-up is complicated.
In the current set up this individual would be classified as employed because their income falls above the lower earnings limit of £9,984 and therefore not liable for contributions on their unearned income.
In the calculator, the application of an allowance in the proposed structure means that they would no longer qualify as employed because their earned income is below the £11,122 allowance. Since their total income is above the lower income threshold (£24,960) for non-employed contributions they are identified by the calculator as non-employed and their estimate based on them being charged contributions at 8.5% on all their income over £11,122.
This is an artefact of the fact that their employed income falls between the current lower earnings limit (£9,984) and the allowance (£11,122). There have been policy discussions with the Committee for Employment & Social Security about protecting the employment classification of a very small cohort of people whose earned income falls within this gap, but this is a level of detail it is not practical to apply to a simple online calculator.
The calculator does not request information on whether you’re married, but it does make some assumptions about whether or not members of a two adult household might be able to transfer any unused allowances between them.
Mortgage interest relief is capped at £3,500 (£7,000 for a couple) in the analysis. This is incorporated in the calculator. The calculator also incorporates the fact that mortgage repayments (interest or capital) are not subject to GST so there are two different ways in which a mortgage impacts the outcome.
The Essential Cost Relief Payment was discussed and agreed in February. It is referenced in this Policy Letter, but it does not go into detail again when no decision is required.
The 2.5% charge on self-employed earnings between the Upper earnings limit and £300,000 is still part of the policy and the primary modelling. It is missing from the calculator, and we are looking to address this. It would impact a very small number of high-income self-employed people.
The application of vehicle taxes in the model is deliberately simplified into a limited number of indicative bands. Actual liability is dependent on weight and emissions of each specific vehicle and the rates will be confirmed in the budget.
There is also a small inconsistency in the application of the lower limit applied to the income threshold at which pensioners and non-employed people become liable for social security contributions (by £1) between the two halves of the calculator. This is only an issue if you input exactly £24,960 as an income for someone non-employed or above pension age.
I hope this clarifies things
Kind regards
Data critic challenges official line point by point
A point-by-point examination of the email from the States of Guernsey’s Senior Economist in the Treasury, of 9 July 2026, against the policy letter of 8 June 2026, the February 2026 resolutions, and the calculator’s code as served on 7 July 2026. Every figure below is reproducible; the calculator’s engine and a corrected version run side by side, with the workings published, at thepeoplestrustgsy.com/wp-content/GST/guernsey-gst-calculator-audit.html.
What the response concedes, before anything else
Read as a whole, the Treasury email disputes not one figure. It confirms, in writing and to all deputies, that: the 2.5% self-employed charge is missing from the calculator (“It is missing from the calculator, and we are looking to address this”); the £10,000-plus-£30,000 household is charged 8.5% on the whole of its income above £11,122 as “an artefact”; the calculator never asks about marriage but “does make some assumptions” about transferring allowances in every two-adult household; the £24,960 inconsistency exists; and the calculator differs from “the primary modelling”. Those are the findings. What remains in dispute is only how much they matter, and on that the response rests on characterisations that the record does not support. We take its points in order.
1. “Averages versus individuals” and equivalisation
The response attributes the divergence between the policy letter’s figures and the calculator’s to the difference between group averages and a specific household, and to income equivalisation.
Averages are indeed averages, and nobody has suggested otherwise. But three things are wrong with this as an answer. First, the table from which the comparison figures were drawn is headed, in the policy letter’s own words, “Household gross income percentile”, and is broken out by household type, so a single-pensioner column contains only single pensioners ranked on gross income; the equivalisation explanation, whatever its general merits, does not attach to that table. Second, and more important, the response never mentions the actual mechanism identified in the piece it is answering: unless the user fills in one optional savings box, the calculator assumes every remaining pound of income is spent, and therefore charges GST on the whole of it. The policy letter’s modelling does not assume that. That single undisclosed default, not household composition, is what moves the answer by hundreds of pounds for any household that saves, and it moves it in one direction only. Third, the fix costs one sentence: a line on the results page saying “this estimate assumes you spend everything you have not told us you save”. The response defends the methodology; it does not explain why the public is not told the assumption.
It is also worth noting what the response asks deputies to accept: that the figures presented to them and the figures presented to the public differ for good methodological reasons. Even taken at its highest, that is a confirmation of the point, not a rebuttal of it. Two official answers to the same question differ by up to several hundred pounds a year, and nothing on the public tool says so or says why.
2. “Marginal and unusual cases” and the caveats
The response says the calculator “is less accurate when dealing with marginal and unusual cases” and that this “is clearly stated in the caveats and health warning”.
We have extracted every caveat the served calculator contains, and they are quoted in full in the audit. They say the tool provides “an indication”, is “not a formal tax assessment”, simplifies “some of the tax calculations”, makes “assumptions about how GST might impact… prices”, and “assumes your expenditure does not exceed your income”. That is all. The words “unearned income” appear nowhere. No caveat mentions the treatment of investment income, the assumption about transferring allowances, or the treatment of mortgage interest. No caveat discloses that expenditure is assumed equal to income unless the savings box is completed; the one sentence in the vicinity says the opposite, that expenditure will not exceed income. The claim that the difficulties now being relied upon were “clearly stated” to users is contradicted by the text of the tool itself.
Nor are the affected groups marginal. Households with mortgages, couples who are not married, and people with modest wages alongside pension or investment income are ordinary Guernsey households, not edge cases. The one genuinely narrow case, the £24,960 boundary, we have always described as minor.
3. The £10,000 wage with £30,000 of investment income
This is the response’s longest technical passage, and the one that fails most completely. The response confirms that under the current system this person is employed and pays nothing on their unearned income, describes accurately how the calculator instead reclassifies them and charges “8.5% on all their income over £11,122”, calls this “an artefact”, says there “have been policy discussions… about protecting the employment classification” of such people, and concludes that this “level of detail” is “not practical to apply to a simple online calculator”.
Set that against the policy letter deputies will vote on. Page 8 states the decision in terms: “In particular unearned income will not be brought into scope for employed and self-employed people, until more work can be undertaken with the Committees for Employment & Social Security and Housing”. Figure A7.1 states the resulting rate on the unearned income of employees under the 2026 package: 0%. The same table shows 8.5% in one place only, the column for the abandoned November 2024 package. Appendix 2 (A2.11) prices the removal of that 2024 charge at about £5m a year and describes who it affects. The calculator, whose browser title still reads “GST-Plus Tax Calculator”, is charging the dead package’s rate. So the correct treatment is not an open policy question awaiting discussions with another committee; the letter records the discussions as the reason the charge was removed from the package, and the treatment it enacts meanwhile is nil.
The response’s own framing confirms the malfunction. If, as it says, the discussions are about “protecting the employment classification” of these people, then the intended answer is that they remain employed and pay nothing, which is what the letter’s table says and what the calculator fails to do. And the calculator is not even a consistent implementation of the contrary view: give the same person £50,000 of wages instead of £10,000 and it charges nothing on the identical £30,000 of investments. No policy reads “investment income is contribution-free unless your wages are low”. A charge that appears and disappears with the size of an unrelated wage is a defect, which is presumably why the response reaches for the word “artefact”.
As to practicality, the correct behaviour is simpler than the incorrect one. The current-system half of the same calculator already implements it, in one condition: a person with any earned income is not charged as non-employed. Copying that condition to the package side is a one-line change. What is described as too detailed for a simple calculator is, in the calculator’s own code, the simpler branch.
Finally, the scale of the error is not the £1 sort. For this household the calculator turns a gain of about £407 a year into a loss of £1,974, a swing of £2,381 that reverses the one thing the tool exists to tell people: whether they are better or worse off. It is also worth deputies noting the direction: the error overstates the cost of the Committee’s own package. We point it out anyway, because the issue is reliability, not advocacy.
4. Marriage and the allowance transfer
The response states that “the calculator does not request information on whether you’re married, but it does make some assumptions about whether or not members of a two adult household might be able to transfer any unused allowances between them”.
On the first half we agree, and the circulating piece was wrong to say the calculator asks; our published audit states, and demonstrated before this exchange, that it never asks. The correction cuts the other way, however. The calculator’s own data structure contains a marital-status field, created, defaulted and passed into the calculation; someone designed it for the question. No screen asks it and no formula reads it.
The second half is a confirmation wearing the clothes of a reassurance. The “assumption” is that every two-adult household, without exception, qualifies for and has arranged the transfer of up to £15,800 of unused allowance. In law the transfer belongs to spouses and civil partners, and a cohabiting couple can obtain it only by a formal, irrevocable election subject to conditions. The calculator asks one household about Income Support, mortgage relief and child allowance, so a fourth yes-or-no question was hardly beyond it; instead it awards the entitlement to everyone and tells no one it has done so. For an unmarried couple with one earner this misstates the outcome by roughly £566 a year, and, again, in the direction of making the package look worse than it is. The response offers no defence of the assumption’s accuracy, no quantification, and, in contrast to the missing 2.5% band, no commitment to address it.
5. Mortgage interest
The response says relief is capped at £3,500, that mortgage repayments are not subject to GST, and that “there are two different ways in which a mortgage impacts the outcome”.
The calculator contains three. The two the response describes are correctly implemented, and we have said so. The third is the finding, and the response does not mention it: on the package side only, the code subtracts the mortgage interest from the GST base a second time, after the full rent-or-mortgage payment, interest included, has already been removed, and it applies no cap to that second subtraction. The form’s own labels prove the double-count: the housing question asks for the full “Annual rent or mortgage payments”, and the interest field is expressly collected “for relief calculation”, not as an additional GST deduction. The behaviour is directly observable in the tool: hold the payment figure constant, increase only the interest figure, and the GST falls by three pence in the pound, without limit. That is worth £600 a year to a household with £20,000 of interest, it appears in no policy document, and it flatters the package. A response that lists the two lawful channels and passes over the third in silence has not engaged with the finding.
6. The Essential Costs Relief Payment
The response says the payment “was discussed and agreed in February” and that the policy letter “does not go into detail again when no decision is required”.
The February 2026 resolution is public, and what it set, provisionally, was £520 a year for a single adult and £860 for a couple, calculated as the estimated annual GST of an Income Support household under the then package’s 5% rate. The current package charges 3%, and its policy letter promises payments “proportionally lower” than before, without setting them. Scaling the February basis from 5% to 3% gives £312 and £516. The calculator presents £310 and £520. So the figures the public is shown have been agreed nowhere: February agreed different amounts for a different rate, and the current letter agrees none. Two details deserve particular attention. The calculator’s couple payment, £520, is to the pound the February single-adult payment, which is what a hurried adaptation of the old tool would produce. And the calculator imposes an eligibility cliff at £32,400, a figure whose only appearance in the current bundle, at page 74 within the accompanying Sub-Committee report and in the February 2026 resolutions, is as the abandoned package’s higher-rate tax threshold. Far from no decision being required, the response’s own account, agreed in February, proportionally lower now, establishes that the current amounts are undecided, in which case the tool is presenting undecided figures to the public as settled ones.
7. The 2.5% self-employed charge
The response concedes this in terms: part of the policy and the primary modelling, missing from the calculator, being addressed. Three observations. The charge is worth up to £2,586 a year, so for the people it touches the calculator is not slightly out, it is presenting a different tax system. “A very small number of high-income self-employed people” is an argument about who is misled, not whether. And no date is given for the fix, while deputies vote this month and islanders are being urged to consult the tool now. An omission of an entire measure also cannot be reconciled with the earlier suggestion that differences reflect averaging methodology; this one reflects the measure not being in the code.
8. Vehicle tax
Simplification into bands is unobjectionable, and the tool does mark the figures as indicative. But the policy letter states the proposed schedule runs “£25 to £280”, and the calculator charges up to £300. Bands chosen within a published range are a simplification; a figure above the published maximum is a contradiction of it. If the true position is that the rates are unconfirmed until the budget, then the letter’s stated ceiling and the tool’s higher figure cannot both stand, and either way the public is being quoted a number the letter rules out.
9. The £24,960 boundary
Agreed, and agreed to be minor; our audit has said from the outset that it bites only at exactly £24,960. One correction of scale: the input window is £1 wide, but the consequence at that point is a swing of about £526, because one half of the tool charges 3.8% on £13,838 that the other half exempts. Its significance is not the money; it is that the two halves of one calculator apply different rules to the same statutory limit, which is the signature of a tool whose halves were never reconciled, and it now joins the response’s other admissions.
What the response does not address
For completeness, the response is silent on: the third mortgage subtraction described at point 5; the undisclosed spend-everything default described at point 1; the calculator’s browser title still reading “GST-Plus Tax Calculator”, the abandoned package’s name; and the fact that the charge at point 3 and the payment figures at point 6 both match the abandoned 2024 package rather than the one before the States. Individually these are details. Together they describe the same event: the 2024 tool was adapted in a hurry, incompletely, and released under the 2026 package’s name without being checked against the 2026 package.
A note on the document itself
The policy letter, as offered to the public on gov.gg, is not fully searchable. On 68 of its 172 pages, some or all of the paragraphs are pictures of text rather than text, rendered line by line as images, and on those pages the footnotes are pictures too, which matters because the footnotes are where the sources are cited and where figures such as the £1.6m already spent on implementation are disclosed. A search of the public copy for these things returns nothing. The copy lodged with the parliament.gg papers is a different build of the same letter, and it is markedly more searchable than the one given to the public: the case-study tables and several footnotes that are images in the public copy are text in the deputies’ copy. Yet in both, the passage recording that unearned income will not be brought into scope for employed and self-employed people survives only as a picture, invisible to any search. The pattern, in which the same document is rasterised differently by different publishing runs, is characteristic of exporting from machines whose fonts are not licensed for embedding; whatever the cause, the effect is that the public and their deputies were given copies of one letter that answer the same searches differently, and neither answers them completely. The remedy would cost 10 minutes: re-export the letter from its source files on a machine with correctly licensed fonts. Until then, we have prepared a fully searchable copy, produced by optical character recognition, which can be found at https://www.thepeoplestrustgsy.com/wp-content/GST/2026-07-10 TPT-proposition-P2026-49-MS_ocr.pdf.
Conclusion
The Treasury response confirms the calculator’s departures from the policy in four places, mischaracterises the policy in one, the unearned-income treatment that the letter’s own table sets at 0%, describes two mortgage channels while omitting the erroneous third, relies on caveats that do not say what is claimed for them, and cites a February decision whose figures contradict the calculator’s. Nothing in it disputes a single number. The asks are unchanged and are made easier, not harder, by this response: correct the calculator or withdraw it while the States are deciding, disclose the spending assumption on the results page in the meantime, and publish the model so that the figures deputies are being given can be checked in the way the public tool now has been.
Jamie Collings
