Optimization of portfolio (withholding taxes)

Hi guys,

I live in CH and my equity allocation is 100% VT.

As I understand the European dividend treatment is suboptimal with a US based fund. For example a german company pays dividend to the US fund and the withholding tax cannot be reclaimed.
An IE- or LUX-domiciled wouldnt have this issue as it’s within the EU.
Is this understanding correct?

Then the main question is, could i optimize the portfolio switching to:
60% US allocation with a US fund, e.g. VTI
40% rest the world ex-US with a LUX or IE fund, e.g. IE0006WW1TQ4

Not considering rebalancing efforts, cost, etc. Just focus on WHT treatment.

Thanks & best regards

I think you need to do a bit more math I believe.

Ex-US is not just Europe. You need to check treaty rates for Japan, EM, Asia as well.

Nearest comparison could be to check effective yields for following -:
EXUS ETF (msci world ex US UCITS)

Ex-US is actually still better on average with US domiciled funds.

Certain countries differ of course. But to do an optimization like you would want to do here, you would need to have an etf for every single country (or certain regions), domiciled in that country and then CH must have a double tax agreement with them as well.

That‘s simply unfeasible and your porfolio must have a certain size, as withholding tax claims need be above 50CHF (or was it 100? Cant remember exactly) of tax withheld.

At the end you also introduce lots of trading cost, a big headache in tax declaration and probably not even ending up materially better.

Very much depends on how much of US WHT you can get back.


I didnt realize that even within EU countries there are withholding taxes.

Found a great page by PWC, listing all country relations for WHT: France - Corporate - Withholding taxes

Seems like the US enjoys the same treatment as LUX/IE for France, germany, China etc. And even more favorable for Japan.

Is it? Out of curiosity, I asked an AI model and it’s struggling with this kind of questions. Maybe the questions are too specific, even though I narrowed it down to European markets, only. Or I’m asking the wrong way :sweat_smile:

Comparing ETF on European markets, it claimed 12-14% withholding tax for US domicile, and linked a tax document for 2023 where half of that are mentioned. Eventually, it corrected and concluded.

“It is surprising that a US-domiciled fund, pays lower withholding taxes than an Irish fund, since this contradicts general expectations. Further investigations and direct insight into the specific tax documents of the respective funds are necessary to confirm and understand this observation. “

There’s a big thread on withholding taxes, if someone did further investigation, let’s hear it there. :wink:

To get a definite answer you can comb through annual reports of funds and see exactly how much withholding tax they paid.

Check for example annual report of VT and annual report of VWRL for direct comparison (you’d need to deduct the 15% US withholding to get the right number probably).

Or the new EXUS (when the report is out) with VXUS.

That’s what the model kept telling me :rofl::
“To confirm these figures, I would recommend reviewing the detailed financial statements or annual reports of the respective funds from Vanguard’s official website”
“For detailed and accurate information, please refer to the respective annual reports and financial disclosures. If you need assistance finding these documents, feel free to ask!”


Banker on Wheels has tackled this question by looking at all 3 levels of taxation and says it depends on ETF’s domicile and legal structure for level 1 taxation on crossborder payments from a company to the ETF.

1 Like

Just keep in mind though that EXUS does not include Emerging markets. VXUS does

1 Like

This post is very informative Dr Pi Post

Some numbers might have changed though if tax treaties got updated.

Thanks, I’ve read it before and did something similar, less detailed myself. I was hoping in these modern times, AI can do it for me. Didn’t work out of the box. Maybe there’s some deep minds here that can get more out of their model :smiley:

Either way, my hypothesis is that multiple ETFs, for example US, Europe, Japan and Emerging markets can beat an efficient single All World ETF by at least some 0.05% considering TER and withholding taxes. I’m ignoring Canada and rest of Asia-Pacific for the sake for simplicity.

A combo VEVE + VFEM will have a nominal TER of around 0.13% already, without any domicile play.

You can‘t just ignore huge markets and then talk about such a miniscule difference. The omiting of those markets alone has a wayyy bigger impact either in positive or negative direction.

We haven‘t even talked about available products, those containing small caps or only large. Your own trading costs. How do you rebalance? What weights?

All those will have impacts and there will basically not a definitive way for you to determine if it‘s better or not.

Let alone all this effort for some meager 5bps.

1 Like

Ok, then include them, it’s some 3% each? Maybe we have different understandings of huge. :sweat_smile:

My settings are

  • Include small caps where available at low fees, ignore them otherwise
  • Trading cost like at IB
  • Weight like MSCI or FTSE all World
  • Ignore rebalancing, movements are pretty correlated these days

It’s partly for the fun of it, a little distraction. But then, 0.05% can make a difference. There’s 100s, if not 1000s of posts on this forum about saving a few bucks on phone, bank or streaming services on this forum with lower amounts (which is great)

I was mainly refering to you wanting to beat a world index by 5 basis points. 6% allocation to anything will skew your picture enough to not get an answer for that.

you can’t do that either

I don’t know what your goal is. But it’s basically meaningless to do a comparison like this, as you will not answer your original question statistcally significant.

If that allocation beats a world index, you cannot make the conclusion it’s more favourable going forward.

Only an apples to apples comparison makes sense.

1 Like

I don’t want to beat it, I want to replicate it in the most cost efficient way, considering TER and withholding taxes.

My goal is to challenge my thesis (see above), and as a bonus get an AI model to do so for me/us. It’s not meaningless. It’s fun and might save you 100s of bucks.
I understand there are people here working on AI and/or in the financial industry, so I salute the provocative exchange of ideas :wink:

Don’t know these, but you can get US market for 0.03% TER, Europe for 0.1% etc. Not including WHT, not even considering tracking difference.
Most discussions I’ve read are US vs. IE. I’m looking at a mix of those and others.

1 Like

I’m not an AI specialist by any stretch of mind but I’m afraid you may have too high expectations of the current level of publicly available LLMs (if that’s what you’re using for AI).

It may be what you’ve done but my understanding is that unless your model is specifically trained in the dedicated field and explicitly fed or directed toward pre-vetted documents with accurate knowledge/data in the relevant field, their answers are very generic and often imaginative as they tend to try to invent things to fill the gaps they may have.

It may be worth (and again, it may be what you’ve done in which case, please ignore this) gathering a few sources of relevant and adequately updated data and directing a prompt specifically toward them while ignoring contradicting other data found openly on the internet to try and get better results. You’d have to check that the AI bot you’re using has access to current information and isn’t limited to data previous to a set date (as, for example, the free version of ChatGPT (3.5) that is limited to data from Autumn of 2021 and before).

My appologies if what I’m stating is plain wrong, I’d be happy to be corrected by someone with specialized knowledge on the topic if they feel a need for it.

1 Like

Yes, for this test, I used the enterprise version of ChatGTP 4o. No specific training or documents, just prompting. There’s tons of data on tax agreements and fund data.
Nevertheless, it does quite confidently provide incorrect or incomplete answers. Same trials and tribulations as in the work place or online :sweat_smile:

If anyone cares and @Dr.PI doesn’t mind, we can compare the WHT table linked above here or in a different thread.

1 Like

Of course I don’t mind, please update those data. You can even make a wiki if you care.

What you can try to do is to feed annual report PDFs to AI and ask to extract specific values for each fund.

Basically all you need is the gross amount of dividends received by a fund (iShare calls it “dividend income”) and “non-reclaimable withholding tax”.