I’m not sure how useful this post will be for a big chunk of you. I’ve been doing some research for a project, how much it costs to rebalance an ETF portfolio if you’re an investor living in Italy, and figured I’d share it rather than let it sit in my spreadsheet.
In its never-ending campaign to punish savers (to be fair: savers who don’t buy Italian government bonds), the Italian tax authority has built one of the more convoluted fiscal systems out there. Convoluted here is a polite word for punitive. Gains and losses from ETF investments can’t be offset against each other. And the capital gains rate isn’t small: 26%. That rate only applies if you’re not holding “white-listed” bonds, which is essentially developed-market government debt. Italy couldn’t exempt only its own bonds without it looking obvious, so this is the least damaging version of the rule they could come up with while still keeping demand for their enormous pile of debt.
For this exercise I picked one of the better-known lazy portfolios: the Golden Butterfly.
I first thought about building a model with the help of AI, but I no longer have a Bloomberg terminal, and the plan of “building my own ETF dataset” has been sitting on my mental shelf for ages. Luckily, testfol.io stepped in to save me the trouble, the platform had recently added a feature to calculate post-tax returns…apparently for a brief window, tho. While writing this post I went back to my analysis, only to discover they’d put the tax calculator behind the paywall.
Testfol.io is not perfectly precise, though: it’s built for the American investor, so it distinguishes between short-term and long-term gains instead of the white-listed-versus-everything-else split that actually matters in Italy. It also lets you offset losses against gains, which Italy doesn’t allow.
Both of those quirks mean my simulation is probably a little off. Total losses came in a bit over 10k against 540k in gains, so the distortion is small. Also worth noting: TLT and SHY generated smaller gains than VTI, VRB, and GLD. Testfol.io gives you all the underlying data to adjust for this, and I did most of that work manually in a Google Sheet. The short-term-versus-long-term gain issue is easily fixed by setting both rates at 26%.
The bigger limitation is that testfol.io doesn’t allow to run post-tax Monte Carlo simulations… or if it does, I couldn’t find the button. So I kept things simple: took the POV of an investor who started on August 1, 1996 with one dollar and added $500 a month for 30 years. I wanted something realistic, and this is close to how most people actually invest. It also matters here for a specific reason: new contributions let you rebalance “tax free,” simply by directing fresh money where it’s needed. That makes the whole analysis path-dependent, and I may run different starting points later to see how much that matters.
To sidestep the issue of US-listed funds paying out dividends, I used SIM-style products, which mimic how UCITS ETFs behave (specifically the non-dividend-paying ones, which are the ones you should be using anyway).
Now that you know the setup and its limitations, here’s the punchline: Italian tax friction costs the Golden Butterfly investor almost 80 basis points a year. That’s a lot. If the Golden Butterfly existed as a single ETF, with all the rebalancing happening inside the fund and out of reach of the Italian tax authority, the same investor would earn 80bps more per year, purely because they held one all-in-one product instead of five separate ones. Frankly, that’s crazy.
Here’s how I got there. Every “tax calculator” backtester has the same blind spot: none of them let you liquidate the whole portfolio at the end of the simulation. But we don’t invest to admire the number, we invest to eventually spend it. Comparing a world where you pay tax on every rebalance to a world where you never pay tax at all isn’t a fair comparison. Sure, you could pass the portfolio on and use a step-up in cost basis to soften the blow, but that only works once you’re dead. So let’s stick to the problems you’ll actually face while alive.
Testfol.io reports the unrealized capital gain, so I just took that figure and applied the relevant tax rate, split by instrument, so I could model the 26% and 12.5% rates separately.
Run the numbers: $500 a month for 30 years turns into $179,500 of contributed capital and $583,555.56 at the end. Once you subtract the capital gains tax paid along the way from rebalancing, you’re left with $500,175.78.
Now compare that to a portfolio that only pays capital gains tax once, at the very end. That one finishes at $584,933.58 (I used a blended tax rate of 24.86%, calibrated to match the same effective rate as the rebalance-as-you-go version).
An 84k gap is material, whichever way you want to slice it, as a percentage of contributed capital or of the final balance.
To land on that 80bps-a-year figure, I goal-sought the drag that would take a tax-free-rebalancing scenario down to that same $500,175.78 final number, with only a single 26% capital gains hit applied at the very end.
That’s the tax advantage “asset allocation” ETFs quietly offer Italian investors. Wrapping your asset allocation inside a single product can save you a genuinely meaningful amount of money.
This analysis is hard to generalize, because the size of the benefit depends on a few moving parts:
- the longer you stay invested, the bigger the savings
- the larger your pot is relative to your ongoing contributions, the bigger the savings
- the less correlated your underlying assets are, the bigger the savings
There’s also a subtler edge here that basically no asset manager is exploiting (yet?). Say I run a Golden Butterfly ETF, just five other providers’ ETFs wrapped inside one box. My product already beats your DIY version by that 80bps. But if a provider launches a cheaper ETF tomorrow, say a large-cap stock one, I can swap it inside the box with zero tax event. I can hand you the lowest-cost beta available at any point in time, permanently. You, running the DIY version, can’t do that: you’d have to weigh the fee savings against giving up all the compounding tied up in the capital gains tax you’d trigger by selling. If you’ve been invested for a few years already, that switch basically never pays off.
It’s a few basis points here and there, sure, but it all adds up. Traditional asset managers have no incentive to chase this, they’re playing a relative game, just trying to keep fees in line with competitors. New entrants, on the other hand, could turn this into an actual differentiator: a genuine long-term relationship with their clients, instead of just another fee war.
I ran this analysis in between an apartment move — packing and unpacking boxes on repeat — and some travelling. If you have different ideas or approaches, I’m all ears.