This article responds to Dan Neidle’s response to our critique of his claim that Scotland’s 48% top rate has reduced public revenue. Readers may find it helpful to read both Neidle’s original analysis and our initial critique first.

Over the weekend we published a detailed critique of Dan Neidle’s headline-grabbing claim that Scotland’s 48% income tax has actually lost rather than raised public revenue. Dan has now responded to our critique, describing it as the most thoughtful challenge to his work so far. His response is detailed, constructive and, in several respects, persuasive [1].

Many of his points are fair and well made. In particular, Neidle is right that the weakness in Scottish top incomes over the past two years appears unusual when compared with the historic data. We also agree that the lowering of the top rate threshold in 2023-24 cannot explain the further fall in 2024-25, when the threshold remained unchanged. And his interpretation of our alternative HMRC total income series – that income switching from salary into dividends could reduce Scottish non-savings, non-dividend (NSND) income without reducing total income – is a plausible possibility.

These are fair points, and they strengthen the argument that tax related behavioural change may have contributed to the recent divergence between Scotland and the rest of the UK (rUK). In the absence of an obvious alternative explanation, Neidle concludes that his behavioural change hypothesis is still the most plausible explanation for Scotland’s weaker top income growth relative to rUK – and consequently for his estimated £22 million revenue loss.

Importantly however, we never claimed that behavioural change hasn’t played any role. On the contrary, we said this was a “near certainty”. In practice, Neidle’s response subtly attempts to reframe the debate. Our argument wasn't that tax had no effect, or that the recent fall should be dismissed as random volatility. Our contention was that his calculation attributes all of the observed divergence to tax, even though the historic data show that other non-tax factors have consistently driven divergence between the growth of top incomes in Scotland and the rest of the UK.

As we will see, this distinction is crucial to the credibility of the claim that Scotland’s top rate tax changes have actively lost revenue.

Neidle’s original analysis begins with a real and interesting observation. In 2024-25, the average income of Scottish top rate taxpayers fell by around 8%, while it rose by around 5% across the rUK. His hypothesis is that this 13 percentage point divergence largely reflects tax related behavioural change by high earners in response to Scotland’s higher tax rates.

That is a reasonable argument. It may even be the most plausible explanation for a substantial part of the fall, though we just don’t know. But crucially, it isn't sufficient to justify the claim that the 48% tax rate has lost public revenue – never mind on the scale of £22 million.

This is because Neidle’s analysis doesn’t merely assume that tax played some role. It assumes, for the purpose of the £22 million loss estimate, that tax explains the entire gap. This is the most important issue our original analysis sought to highlight, but Neidle’s response doesn't fully address it. 

Our historic charts showed that Scottish top income growth has repeatedly diverged from the UK’s, typically being significantly more volatile in both directions. This includes in earlier periods when Scottish tax divergence couldn't plausibly explain these differences.

The implication is not that the latest divergence must be random volatility, as Neidle suggests. It is that rUK growth is not a reliable estimate of what Scottish top incomes would otherwise have done in any single year. Neidle’s response to our critique largely recasts this as a narrower claim that the 2024-25 result might simply be “one of those swings”. 

He then tests whether a divergence of this scale is unusual from a statistical standpoint. That is a useful exercise, but it doesn’t answer the central challenge we sought to pose: how much of the gap can really be attributed to tax, rather than to the other forces that have repeatedly driven divergence between Scotland and the rUK in the past?

Neidle’s new statistical tests suggest that the 2024-25 divergence in top income growth between Scotland and the rUK was unusual relative to the historic figures. Depending on the test used, he estimates it as roughly a 5% to 7% lower tail event. However, the 13 percentage point fall lies only just beyond the simple two standard deviation benchmark implied by Neidle’s own figures. This certainly makes tax related behavioural change a plausible contributor to the divergence. But it doesn't by itself show that tax caused all of the divergence. This effectively makes no allowance for other forces that may be driving Scotland-rUK divergence.

In reality, an unusual outcome can reflect multiple forces interacting at once. Tax may have played a significant role, but it's also possible that shifts in sectoral conditions, bonuses and self-employment income, or relative economic performance could also have contributed.

The historic volatility shown by our charts matters because it demonstrates that these other forces exist ,and have previously contributed significantly to Scotland-rUK divergence. Yet the £22 million calculation allocates none of the most recent divergence to them – it assumes them away entirely. Although Neidle acknowledges in the accompanying text that non-tax factors could in fact matter, the £22 million estimate that dominated the headlines makes no allowance for them (and neither does the £15 - £30 million range). The entire shortfall from the rUK counterfactual is treated as tax related. [2]

This is why his statistical analysis doesn't settle the argument. Showing that the latest divergence is unusually large may strengthen the view that tax contributed in some way. It doesn't justify assigning 100% of the gap to tax. As we will see, this really matters. 

The most important part of Neidle’s response appears in a new footnote. He accepts that:

Here he is referring to the table we included in our original analysis showing how the £22 million revenue loss is highly sensitive to even small changes to the key assumption that Scottish top incomes would otherwise have grown in line with the rUK, absent tax changes. In the model, revenue loss disappears if Scottish top incomes grew only around 1.2 percentage points more slowly than those in rUK. If the gap were 2 percentage points, the model would instead deliver a revenue gain of £14 million. 

In other words: a relatively small adjustment to the assumed counterfactual is enough to reverse the apparent revenue loss.

This point is more significant than it may first appear. The headline 2024-25 divergence was around 13 percentage points – average Scottish top incomes fell by around 8%, while those in rUK rose by around 5%.

Neidle accepts our finding that the estimated £22 million loss disappears if Scottish incumbent top incomes would otherwise have grown just 1.2 percentage points more slowly than his rUK benchmark. Compared illustratively with the headline 13 percentage point divergence, this is equivalent to around 9% of the observed gap in top income growth between Scotland and rUK (1.2 divided by 13). On this simple and crude comparison, tax could account for roughly 91% of the observed divergence and the alleged revenue loss would still disappear. This is a crucial point. [3]  

Our position does not require tax to have had no behavioural effect. It doesn't even require tax to have been a minor contributor to the gap. Tax could have accounted for the overwhelming majority of the recent divergence and the Scottish Government would still have raised rather than lost revenue overall. Once even a small allowance is made for other non-tax forces that we know have historically caused Scottish and rUK top incomes to diverge, the alleged revenue loss vanishes.  

The chart below presents the same sensitivity using the tax bases implied by Neidle’s revenue figures, rather than extrapolating from the simpler comparison outlined in the table above. It shows how the estimated revenue impact changes depending on what share of the divergence in the tax base is attributed to tax related behaviour. On this basis, the calculation reaches break-even when tax explains around 85.5% of that divergence. [4] In other words: if non-tax factors account for more than around 14.5% of the divergence, the model implies a revenue gain rather than a loss. Crucially, the headline £22 million only materialises when the entire divergence is attributed to tax related behaviour. 

This doesn't tell us what the actual split between tax and non-tax causes was – this remains essentially unknown. Instead it shows that the £22 million loss depends on assuming that non-tax factors played no role in 2024-25 whatsoever. 

In practice, Scottish and rUK top incomes have never moved in perfect annual lockstep, meaning that factors beyond tax have repeatedly driven divergence between them. How credible is it then to assume that non-tax factors explained none of the 2024-25 gap? This is possible, but it is an exceptionally strong assumption. 

Crucially, the absence of an estimate of how significant a role non-tax factors played doesn't mean their contribution was zero – particularly when the charts show these other factors have existed every year historically. Yet this in essence is what the £22 million calculation assumes.

Neidle offers several reasons why he thinks tax related behavioural change is the most plausible explanation for the recent divergence.

Firstly, he argues that the 2024-25 Scotland-rUK gap is statistically unusual. It falls just outside a two standard deviation range based on the historic series (but only marginally), with Neidle estimating that it would occur in roughly one year out of twenty. This makes the result worthy of attention, but it doesn't identify its cause – or justify making no allowance for the non-tax influences that have repeatedly driven the historic volatility. It shows only that a divergence of this scale is relatively uncommon.

Secondly, he notes that the two largest falls occur in the two most recent years, coinciding with Scotland raising its top rate first to 47% and then to 48%. The timing is clearly suggestive, but 2023-24 was also when the top rate threshold was reduced from £150,000 to £125,140, changing the population being assessed. Neidle assumes that this these threshold effects cancel out in the Scotland to rUK ratio, citing a standard Pareto distribution assumption. However, as we explained in our original critique, the lowering of the threshold brought proportionally more taxpayers into the threshold in Scotland than the rUK, mechanically pulling down Scotland’s average income relative to rUK. As such, it isn't possible to untangle these effects from true divergence in 2023-24, meaning only the fall in 2024-25 provides a clean comparison with consistent thresholds.

Neidle also makes a reasonable point that the Scotland-rUK ratio showed no sustained downward trend before the latest tax rises. Again this is true, and strengthens the case that tax may have contributed to the divergence. However, the absence of a downward trend doesn't mean there wasn't significant annual volatility. Scottish top income growth still diverged from the rUK materially from year to year, as shown in the charts presented in our initial critique. This matters because Neidle’s revenue estimate doesn't depend on the long run trend, it relies on the much stronger assumption that Scottish incomes would have matched rUK growth in 2024-25, absent the increases to the top tax rate. The timing of the recent declines is therefore suggestive of a tax effect, but it doesn't mean that the contribution of all other non-tax forces was zero (or sufficiently close to zero to sustain the “revenue loss” claim).

Thirdly, Neidle argues that our alternative HMRC total income series may itself be consistent with income switching. If high earners moved income from salary into dividends, Scottish NSND income could fall while total income remained stable. This is certainly plausible, but 'consistent with' is not the same as demonstrated. It's also possible that the movements reflect other factors entirely unrelated to this. In practice income switching may well be occurring, but we don't have the evidence to show it's happening on the scale needed to explain all the divergence – or sustain the £22 million loss estimate.

The purpose of our alternative HMRC series wasn't to try and replace or challenge the official Scottish tax data. Instead it was simply to test the credibility of the claim that Scottish top income growth could normally be expected to track the rUK closely. On this point, the evidence remains clear: historically growth in Scotland hasn't matched the rUK's, indicating that other factors unrelated to tax have repeatedly played a role.

Finally, Neidle points again to Scotland’s declining share of UK Self Assessment liabilities. But as outlined in our initial response, this is far from conclusive. The data covers all Self Assessment taxpayers, not only those paying the 48% rate – and once the series’ pre-existing downward trend is taken into account, the fall is no longer statistically significant. At best, the Self Assessment data provides corroborative support for some kind of behavioural response. But it is nowhere near strong enough to demonstrate that all of the top income divergence was caused by tax, which is necessary to validate the alleged £22 million loss.

Taken together, the above points make it plausible that tax related behaviour contributed to the recent unusual divergence. But they don’t allow us to identify what proportion of the divergence tax actually caused. Subjecting the historic data to statistical analysis can show us how unstable the Scotland-rUK relationship has been, but it can’t tell us how much of the 2024-25 divergence was attributable to tax behaviour and how much to other factors.

Neidle’s response helpfully moves the debate forward. It provides additional evidence that the recent weakness in Scottish top incomes was unusual and that tax related behaviour may have played a significant role.

Crucially though, this is not the claim that made the headlines. Some behavioural response to higher top rates was always expected and was already baked into the SFC’s forecast. Neidle’s claim was much stronger – that Scotland’s decision to raise the top tax rate reduced revenue, specifically to the tune of £22 million. His response makes a strong case for some kind of behavioural effect, but it doesn't establish that the reforms lost £22 million – or even that there was any loss at all. As outlined above, a small allowance for the non-tax forces that have repeatedly contributed to Scotland-UK divergence historically wipes out any revenue loss.

However, neither does the evidence allow us to conclude that the revenue effect was definitely positive. The honest conclusion is that there is substantial uncertainty not only about the size of the revenue impact, but about whether it was positive or negative. The debate is therefore not between those who believe tax changes had an effect and those who do not. On that, we largely agree. The real question is whether tax can credibly explain virtually all of the recent divergence. Neidle’s £22 million estimate assumes that it does, but in our view the available evidence doesn't credibly support this. 

And at the end of the day, it is Neidle who is making the specific claim that Scotland’s higher top rate of tax reduced revenue by £22 million. The burden is therefore on him to justify the assumptions required to reach that figure. We don't need to prove a particular non-tax contribution to show that the alleged loss is built on shaky foundations, particularly when other factors have played such a significant role historically.

Ultimately, claiming that tax related behaviour probably contributed to an unusual fall is a defensible statement. But saying that tax related behaviour caused the entire shortfall and therefore cost precisely £22 million is not. The reason we are having this debate is because it is the second proposition, not the first, that is driving the public debate.


Footnotes

[1] Dan’s response to our critique can be found in a new section called ‘Couldn’t it just be volatility?’ which was added on 1 August 2026. 

[2] £22 million is presented as Neidle’s central estimate, with a range of £15 to 30 million. But the range appears to reflect alternative assumptions about the incomes of new entrants to the top rate band, as opposed to uncertainty around the central counterfactual that Scottish top incomes would otherwise have grown in line with rUK.

 
[3] The 91% figure is approximate and illustrative rather than a precise estimate. The 1.2 percentage point difference relates to his modelled growth in the incumbent Scottish top-rate tax base, while the 13 percentage point figure is the headline difference in average top income growth between Scotland and rUK. The two measures are therefore related but not identical, hence why the figure is described as "crude". The comparison remains useful for illustrating the scale of the sensitivity, but it doesn't reproduce every element of Neidle’s model, particularly his separate treatment of new entrants to the top band. As such it should be treated as purely illustrative.

[4] The 85.5% figure is derived directly from the tax bases implied by Neidle’s published revenue estimates. Actual revenue of approximately £2.069 billion at 48% implies a tax base of around £4.31 billion, while counterfactual revenue of approximately £2.091 billion at 45% implies a tax base of around £4.65 billion. The resulting shortfall in modelled taxable income is approximately £336 million. Break-even is reached when around £287 million of that shortfall is attributed to tax related behaviour, equivalent to about 85.5%. Figures are approximate because they are reconstructed from rounded published estimates.