Splitting the world: creating tax-advantaged global stocks portfolio using funds in 3a account

I was thinking about updating this post, but decided not to. Too many variables that affect total investment costs. Instead I decided to outline what I think is a right way to calculate the total investment costs for an ETF.

As a Swiss investor, one is only taxed on dividends, and namely up to 3 times. Let’s consider the whole process.

All companies included in a fund are paying out dividends, which all together make gross dividend yield (in %) D_G. The gross dividend yield of various baskets of stocks, for example built according to MSCI indices, can be found in current index factsheets.

When the dividends are being paid to the fund, Level 1 withholding taxes are being levied by the countries where the paying companies are domiciled. Let’s call the relative amount of this tax t_L1. An estimation of this tax for funds of different domiciles investing in various stock baskets can be found in the previous posts. At this stage, we are left with the amount of money equal to

D_G*(1 - t_L1)

At the next step, fund’s management takes management fees TER from the dividends. We are left with

D_G*(1 - t_L1) - TER

Then, this amount is distributed to the investor. Depending on the fund’s and investor’s domicile, there is another withholding tax on the distributed amount, called Level 2 withholding tax t_L2. So a fund investor receives

(D_G*(1 - t_L1) - TER)*(1 - t_L2)

As for a Swiss investor dividends count as income, it is taxed at the personal marginal tax rate t_m. Finally we are left with the net dividend yield D_N (in %).

D_N = (D_G*(1 - t_L1) - TER)*(1 - t_L2)*(1 - t_m)

The difference between D_G and D_N is the money loss (in %) associated with the investment and therefore it is the total cost of investment:

TCI = D_G - D_N

As you can see, there are 5 parameters that affect TCI, so if you compare different investment vehicles, calculate it for your situation.

However I explored certain examples of interest.


It is more advantageous to invest in US stocks via a US ETF even if you don’t get any reimbursements of level 2 withholding tax. For a simple reason: they have lower TER. But the advantage is slightly lower than the difference of TER!


With full (well, for Swiss residents) 15% Level 2 withholding tax an investment via VT costs you basically the same as VWRL. But with > 15% Level 2 withholding tax (no DTA, no or insufficient tax reimbursement) VT is more expensive than VWRL!


Something that specifically interests me: despite higher TER of IE ETF, it is a more advantageous vehicle for investments in MSCI Emerging Markets if your personal Level 2 withholding tax is more than 5%. And even if it is not the case, investment with an IE ETF is only slightly (ca. 0.09% p.a., i.e. TER difference) more expensive.


For European stocks, IE ETFs are already better than US ETFs despite a slightly higher TER, and if you pay Level 2 withholding tax, the difference increases.

16 Likes

It should be clear from the posts above that, when building a globally diversified stocks portfolio using 3a and taxable accounts, it is advantageous to hold US stocks as US ETFs at Interactive Brokers and Developed Markets ex US as CSIF funds in 3a account (finpension). The question now is: how advantageous it is and if it is worth extra efforts? I will use parameters which are pessimistic for 3a investment (marked red), so an actual advantage should be higher.

Note that I refer to MSCI classification and indices and I won’t talk about MSCI Emerging Markets. MSCI EM is a nicely segregated geographic segment of the global stocks market (10-11%) and, based on a comparison presented above, I decided to invest in this geographic segment via an Irish MSCI Emerging Markets IMI ETF.

So, let’s consider a simple portfolio 1: an MSCI World ETF in a taxable account (48967 CHF) and CSIF MSCI World ex CH Index Fund in 3a account (20000 CHF). To make comparison easier, let’s assume that both of them are equivalent to the following combination of ETFs or index funds:

Note that our “synthetic” MSCI World ETF has TER of 0.055% p.a. and a preferable treatment of withholding taxes for each geographic segment. For CSIF funds I assumed L1 withholding taxes equal to that of CH based fund in a taxable account, unless it is known that they have a preferential treatment. This is what I mean with “parameters which are pessimistic for 3a investment”.

The total investment cost for the portfolio 1 is 397.93 CHF p.a.

Now we consider an “advanced” portfolio 2. It is correctly weighted geographically, and both taxable and 3a account have exactly the same value. However now the taxable account holds only a US ETF on US stocks, all other geographic segment are held in 3a account as CSIF index funds. The 3a account is therefore a synthetic “MSCI World ex CH ex US” fund.

The total investment cost for the portfolio 2 is 346.31 CHF p.a. The advantage of the portfolio 2 is 51.62 CHF p.a., 0.075% p.a. with respect to the total value of both portfolios or 0.258% with respect to the value of the 3a account.

And this is how much we save (negative numbers) per geographic segment with portfolio 2:

3a-adv-3

Now the question is: is it worth to create a synthetic MSCI World ex CH ex US fund in a 3a account using CSIF index funds? For me it is clearly worth. Even this conservative estimation shows an advantage of such approach. Real savings should be higher.

But there is another important non-financial consideration. Namely that holding ex US stocks as a US ETF means that one pays US taxes that US is not entitled to receive: withholding taxes on dividends distributed by companies domiciled outside of US. We are always considering that we are getting reimbursed, at least partially, withholding taxes paid to US, but this is not exactly what happens. US does not reimburse our withholding taxes, instead we are paying less taxes to Switzerland. So we are giving money to US and taking it from Switzerland. As a good Swiss citizen, I would like to avoid it. So I invest into Developed Markets ex US via CSIF index funds. This need some calculations, but nothing complicated.

11 Likes

I am doing the same. I only hold VTI at IBKR and cover the rest with my 2nd/3rd pillar investments.

Not only because of withholding taxes, also because of income taxes. ExUS has a 1-2% higher dividend yield, so you you’ll reduce your income taxes by splitting it up that way.

5 Likes

How do you handle the differences in sizes of your pots vs. target allocations?

E.g. If your 3rd pillar is 10%, or 60% of your total investments?

1 Like

Your 3rd pillar invested in ACWI IMI ex-US should represent 45.5% of your invested wealth.
It will rapidely become unbalanced so you will need to balance it with other ETF such as

VT US9220427424 Vanguard Total World Market 72,4%
VERX IE00BKX55S42 Vanguard FTSE Developed Europe ex UK 13,6%
VWO US9220428588 Vanguard FTSE EMERGING MARKETS 9,6%
AVUV US0250728773 AVANTIS US SMALL CAP VALUE 12,0%
3 Likes

Be creative and find your way!

First, you can always use VT ETF or MSCI World ex CH Index Fund to invest “unbalanced” funds.

If you have “too much” in 3a, that is relatively easy. 3a: developed ex US + MSCI World, taxable: US, MSCI Emerging Markets doesn’t really matter where. That would be the most optimal allocation. A system of 2 linear equations to solve, I would say :joy:

In extreme cases, when 3a account is much larger than taxable, one can have Emerging Markets only in taxable and MSCI World ( + Emerging Markets if needed) in 3a.

If you have a bit too little 3a, you either complement US stocks with Developed Markets ex US ETF, or buy VT. If the difference is quite big, one can for example buy MSCI Europe ETF in addition to US (+ Emerging Markets) in taxable account.

If it’s 10% of your total assets, you’ll probably end up with VTI and VXUS (or VEA + VWO) at IBKR as well. If it’s 60% of your total assets, you’ll include World ex CH or US there. You just have to account for contributions as they are limited for the 3rd pillar. I rebalance 1-2x/year.

(crosslinking from 3a solution from Finpension - #526 by Dr.PI; seems better fitting here)

Alright so here’s what I came up with
(I have 2+3 accounts at Finpension+VIAC; not moving them around yet)

Goal

Keep Dev ex US in 3a pillar, US + EM at IBKR.

Finpension (Swisscanto funds)

Fund %
Europe ex CH NT CHF 49
Switzerland Total (II) NT CHF 9
Canada NT 10
Japan NT 20
Pacific ex Japan NT CHF 11

VIAC (CSIF funds) (*)

Fund %
Europe ex CH 32
SMI + SPI Extra (2:1) 6
Canada 7
Japan - Pension Fund 13
Pacific ex Japan 7
World ex CH hedged (**) - Pension Fund Plus 34

(*) These proportions fit with the remainder of my AA outside of 3a (i.e. 65% of total VIAC portfolio should go to Dev ex US), YMMV.
(**) Could probably reassign it to US+EM, to be 100% correct, but keeping it simpler for now.

IBKR

  • Wipe out VEA
  • Substitute VWO for IEMG (to match on MSCI vs FTSE with the 3a)
  • I hold some small cap value tilts with SLYV/AVUV/AVDV

Objections? :slight_smile:

1 Like

I wish I was able to put so much in my 3rd pillar to cover enough of the world. For now I aim to cover small single markets like Japan or Canada to avoid holding separate ETFs for them. I also hold my CH allocation there to minimize the tax drag on dividends (I think Swiss companies pay lots of dividends)

2 Likes

I re-read the thread and some other resources and now think that I should move my “Pacific ex-Japan” (dominated by a high-dividend Australia) and UK portions of my portfolio to 3rd Pillar. Japan has historically paid lower dividends and Switzerland is also not that high on the dividends. Links:

I will therefore consider adding a Japan (and likely Canada) ETFs to my taxable account. I’ll probably go with a Canada-based ETF for the latter.

Does it make sense?

This is less optimal because of reduced (to 0) L1 withholding tax for the pension fund vs. 10% for a US fund and 15% for a European fund.

Yes that is good.

And there is no L1 tax loss one way or another.

Why would a U.S. fund (with Japanese stocks/dividends) have lower withholding tax than a European one?

EDIT: Got it from the table above and the DTA. Many European countries seem to enjoy 10% too though (including Switzerland). Why 15% for a US fund and 25% for a European fund? Shouldn’t those be 10% and 15% respectively?

Different double tax agreements. The rates are wrong though, sorry.

Regarding L1 tax loss for Japan. It’s 10-15% of a rather small dividend payout. Not having to pay income tax on much larger Australian or British dividends will save more (as per your amazing tables). Although, I don’t see a vanilla UK option on VIAC/Finpension…

One might also want to account for the withdrawal fee of the 3a.
For instance, these computations show that if one invests at year n and plans to withdraw the money from 3a at year n + d with d small, well, this is not very sensible:

With a longer horizon:

I would say it’s a wrong reasoning. If you want to account for taxes paid at the exit, you should also consider taxes saved in the first place. As we know, contributing to the 2nd pillar, saving on taxes and withdrawing after 1 year the same amount can easily result in 30+% of return due to different tax rates saved and paid. That’s why it is de facto forbidden.

I didn’t want to mess with these tax rates also because it is a general feature of pension schemes 2 and 3, not related to what kind of assets you have there.

In fact there is another very distantly possible issue discussed: that with 3a you pay taxes on the whole amount withdrawn, i.e. it becomes a capital gains tax. We tried to do some estimations, looks like that for 40+ years the 3a with 100% stocks can grow so much that due to the exit tax you better invest in a taxable account.

3 Likes

Maybe at least caveat your answer in that thread? It’s true only for people with marginal tax rate < top withdrawal rate (but for those it’s usually more obvious that 3a isn’t a very good deal, if you don’t pay a lot of taxes, don’t try to optimize for it :slight_smile: ).

Those of us with B permit, we cannot get those tax advantages (or it’s debatable whether it’s worth it).

So, I think the reasoning could be okay for some.

I am bit confused with the recent discussion. The calculations from Dr Pi were based on specific scenario and I believe was mainly to illustrate a good way to utilise 3a and optimise taxes.

It is a very good summary of what can be considered while making such decisions but it can never be a complete solution.

The reality is always very specific to individual

  • marginal tax rates
  • withdrawal tax rates
  • asset allocation in taxable account vs non-taxable
  • expected returns
1 Like

I see. I was highlighting a way in which the computations should be expanded to take a not so uncommon personal circumstance into account.