That's a tilde, not a minus sign.
OK, thanks.
Not sure why ChatGPT can only approximately model the outcome of the OKI scheme. But I
think the answer might have something to do with this.
The idea with the 0.85% annual levy is that it's supposed to approximate to the tax you would pay, under the ordinary rules, if you invested in risk-free government deposits. So, for example, if government desposits pay 3.5%, and the tax you pay on earnings from government deposits is 25%, then you would expect to pay tax at 25% on income equal to 3.5% of your investment, which works out at 0.875% of the value of your investment.
In other words, theres's no point in putting your money into an ISK account if you intend to invest the account in government deposits; you'll pay pretty much the same amount of tax either way. The scheme is intended to provide a tax incentive to people who are putting their money into higher-yielding (and therefore riskier, and more volatile) investments.
So, you asked ChatGPT to model an ISK investment yielding 7% a year. But the yield on ISK investments, even if it averages 7% a year, won't be 7% every year; it will go up and down.
And here's the thing; you pay 0.85% of your fund value every year, no matter what the investment return is. In years when the investment return is less than 3.5%, you'd actually be better off if you weren't holding your investment through an ISK account; you'd pay less tax. And, in years where you have negative investment returns, you'd pay no tax at all outside the ISK. But, inside the ISK, you're still paying tax, and because you have no earnings to pay it with it's depleting your principal — not only has the value of your investments fallen by (say) 5% over the year, but you now pay a levy of 0.85% so that the total fall in your account value is 5.85%.
What we have here is the opposite of euro cost averaging — an investment strategy where you invest the same amount each month, but you get more shares/units in months when prices are low than you do in months where prices are high, so the average price you paid for the shares/units you hold is lower than the average price of those shares/units over the period. Here, when the price of your units falls, you have to sell some of them. And, once you've sold them, you don't get to benefit from the subsequent rise in the price.
Which means that how the ISK tax incentives work out for you depends not just on the average investment return but on the volatility of that return — the more volatile the return, the less well you do.
I ran a simple simulation (Excel, not Chat GPT) modelling a 1 million investment for 10 years on three different assumptions:
- A: Investment return of 7% each year: final fund value 1.792 million
- B: Investment return alternating between 4% and 10% each year: final fund value 1.785 million
- C: Investment return alternating between -3% and 17% each year: final fund value 1.715 million.
In case B, even though the investment return every year is above the 3.5% risk-free return on which the 0.85% rate for the levy is calculated, because the investment return is somewhat volatile the final fund value is lower than in case A
In case C, the investment return is very volatile and includes years of negative returns, and the final fund value is signficantly lower than in Case A.
So, perhaps in your model ChatGPT's prevarication about the tax is because it doesn't know how volatile the return will be (because you haven't stipulated that) and a precise answer isn't possible without this datum.