FT article on withholding tax and switching to synthetic ETFs

[Lauretta]

Just read this article in the Financial times saying that BlackRock have had a very big outflow from their S&P 500 ETF in Europe & are now creating a synthetic ETF because it is more advantageous as it seems to avoid withholding taxes.
Is anyone using synthetic ETFs for tax reasons or are you considering using them?
BlackRock S&P 500 fund is Europe’s biggest equity ETF loser in third quarter - https://on.ft.com/33I7ujn via @FT

[krasnall]

I think there are a lot of myths and misunderstandings around synthetic ETFs. In Europe, UCITS swap based ETF limit the counter party risk to 10% of the fund's value. What you gain is a slightly better return, mostly due to lower withholding taxes paid by the ETF (level 1).

For example, Lyxor's swap based S&P500 ETF has an average tracking difference (TD) of -0.43%, while Vanguard's physical replicating ETF has an average TD of -0.29% (iShares is a bit worse with -0.22% for Acc. and -0.06% for Dist.). See comparison.

This makes sense since they all track a net S&P500 index, which deducts 30% withholding tax on dividends for index return calculations. ETFs based in Ireland pay only 15% of withholding tax, while swap based ETFs might reduce it even further. That's why all of them have a better return than the index even including management and trading fees. And clearly synthetic ETFs have the largest tax savings.

Personally, I'm not a big fan of this 0.15-0.20% gain for a 10% counter party risk. The real risk might be actually close to 0%, since Lyxor claims it resets the swap gain/loss daily. Still, it doesn't "feel right" to hold some random stocks (mostly mega caps from the US and Europe) instead of the real index, even if it's a rational choice. That being said, I do hold shares of Lyxor swap-based ETF due to the fact that it's listed on Warsaw Stock Exchange and my main bank doesn't offer access to other exchanges - it does however offer me some nice benefits for using their brokerage service. So 0.15% extra from the ETF plus some insurance and travel benefits from my bank is enough to convince me to keep a part of my portfolio in synthetic.

Anyway, the advantages that swap-based ETFs offer seem be only realized with ETFs tracking the S&P500 index. For MSCI World and most others there is basically no difference in real returns between synthetic and physical and the major difference in tracking comes from lower or higher management fees.

[Lauretta]

I think there are a lot of myths and misunderstandings around synthetic ETFs. In Europe, UCITS swap based ETF limit the counter party risk to 10% of the fund's value. What you gain is a slightly better return, mostly due to lower withholding taxes paid by the ETF (level 1).

For example, Lyxor's swap based S&P500 ETF has an average tracking difference (TD) of -0.43%, while Vanguard's physical replicating ETF has an average TD of -0.29% (iShares is a bit worse with -0.22% for Acc. and -0.06% for Dist.). See comparison.

This makes sense since they all track a net S&P500 index, which deducts 30% withholding tax on dividends for index return calculations. ETFs based in Ireland pay only 15% of withholding tax, while swap based ETFs might reduce it even further. That's why all of them have a better return than the index even including management and trading fees. And clearly synthetic ETFs have the largest tax savings.

Personally, I'm not a big fan of this 0.15-0.20% gain for a 10% counter party risk. The real risk might be actually close to 0%, since Lyxor claims it resets the swap gain/loss daily. Still, it doesn't "feel right" to hold some random stocks (mostly mega caps from the US and Europe) instead of the real index, even if it's a rational choice. That being said, I do hold shares of Lyxor swap-based ETF due to the fact that it's listed on Warsaw Stock Exchange and my main bank doesn't offer access to other exchanges - it does however offer me some nice benefits for using their brokerage service. So 0.15% extra from the ETF plus some insurance and travel benefits from my bank is enough to convince me to keep a part of my portfolio in synthetic.

Anyway, the advantages that swap-based ETFs offer seem be only realized with ETFs tracking the S&P500 index. For MSCI World and most others there is basically no difference in real returns between synthetic and physical and the major difference in tracking comes from lower or higher management fees.

Thank you for you feedback and for the cool webpage with tracking differences! Didn't know about that site.
You write:

ETFs based in Ireland pay only 15% of withholding tax, while swap based ETFs might reduce it even further.

I understood that withholding tax is zero for swap based ETFs, isn't that the case? (you say that it might be reduced further)

Interestingly I found that for EM the tracking differences are larger and positive; I wonder why? Also, I had used the Amundi ETF in the past for EM and now invest new money in iShares, but Amundi seems to have done better - not sure whether this implies that it will continue to do so, since it's not clear whether synthetic ETFs for EM also have tax advantages so I don't know what caused Amundi's outperformance in the past.

[alex_686]

I would be relaxed about the counter-party risk.

The standard swap contact requires daily mark-to-market posting of collateral.

[krasnall]

I understood that withholding tax is zero for swap based ETFs, isn't that the case? (you say that it might be reduced further)

It might be 0. I didn't dig deep enough to know exactly. I tried checking the annual report for Lyxor funds, but strangely withholding taxes are nowhere to be found.

Interestingly I found that for EM the tracking differences are larger and positive; I wonder why?

Please note that positive tracking difference means the fund had a worse return then the index, and vice versa.

[alex_686]

Interestingly I found that for EM the tracking differences are larger and positive; I wonder why?

Please note that positive tracking difference means the fund had a worse return then the index, and vice versa.

Tracking Error is always positive. A high Tracking Error means it had a different return, not worse. Could have had more days when it beat the index and more days when it did worse.

Now sure, most of the time it means underperformance but that is not what the number means.

[Lauretta]

I understood that withholding tax is zero for swap based ETFs, isn't that the case? (you say that it might be reduced further)

It might be 0. I didn't dig deep enough to know exactly. I tried checking the annual report for Lyxor funds, but strangely withholding taxes are nowhere to be found.

Yes, and I once wrote to Amundi on withholding tax; they usually reply to my emails but they didn't reply to this question. So I assumed they are zero because someone mentioned this to explain the better performance of BNP ETF in an old thread on BH.

Please note that positive tracking difference means the fund had a worse return then the index, and vice versa.

Yes I realised that (when I looked at US ETFs and read your post). For EM they all have positive tracking errors on the webpage you provided, But Amundi's ETF has a lower magnitude so it means it lags the index by a lower error, hence it does a bit better.

[Lauretta]

Interestingly I found that for EM the tracking differences are larger and positive; I wonder why?

Please note that positive tracking difference means the fund had a worse return then the index, and vice versa.

Tracking Error is always positive. A high Tracking Error means it had a different return, not worse. Could have had more days when it beat the index and more days when it did worse.

Now sure, most of the time it means underperformance but that is not what the number means.

I understood that when the Tracking Error is negative, as in the webpage krasnall provided, it means that the ETF actually outperforms the index. Perhaps by lending securities (if it's a physical ETF)?

[krasnall]

Tracking Error is always positive. A high Tracking Error means it had a different return, not worse. Could have had more days when it beat the index and more days when it did worse.

Yes, you are right about the tracking error, however the website I mentioned calculates the tracking difference between the index return and fund return, not TE.

[webkonsola]

Still, it doesn't "feel right" to hold some random stocks (mostly mega caps from the US and Europe) instead of the real index, even if it's a rational choice.

Hi -

Aren't the banks using futures to replicate the the index ? I thought that's the way to replicate through synthetic structure vs sampling which is the random stocks that you may be mentioning?

[krasnall]

What synthetic ETFs do (at least in Europe) is they hold a basket of "random" securities and a swap agreement. The swap agreement means that the ETF will pay the basket performance to the swap counterparty, and the swap counterparty will pay the index performance to the ETF. That can happen (I guess it usually does) daily. The basket of securities the ETF holds is not really random, it has to hold a diversified portfolio of liquid stocks. For example, here are the top holdings of Lyxor's S&P500 ETF:

AMAZON.COM INC 9.19%
FACEBOOK INC-CLASS A 4.85%
ALPHABET INC-CL C 4.59%
BOEING CO 3.44%
KERING 3.37%
Apple Inc 3.22%
BNP PARIBAS 3.07%
SALESFORCE.COM INC 2.8%
ADOBE INC 2.36%
ALPHABET INC-CL A 2.04%

[alex_686]

The swap agreement means that the ETF will pay the basket performance to the swap counterparty, and the swap counterparty will pay the index performance to the ETF. That can happen (I guess it usually does) daily.

Swap contracts are almost always long term. The shortest I have seen is a 3 years, 5 years is typical. Anything at a year or under would normally be a forward or future.

Cash would be swapped at the end. Maybe quarterly payments would be made. Generally only margin collateral is posted daily.

The positions you posted look odd to me but might work. You can get a pretty decent replication of the S&P with 20 to 40 stocks. There are different mathematical models one can use. I am going to guess sometime of liner regression of significant factors instead of a sampling method.

[assyadh]

Synthetic ETFs do not avoid the IRS witholding tax. Section 871(m) of the internal revenue code made sure of that:

https://www2.deloitte.com/ch/en/pages/f ... -code.html

[webkonsola]

What synthetic ETFs do (at least in Europe) is they hold a basket of "random" securities and a swap agreement. The swap agreement means that the ETF will pay the basket performance to the swap counterparty, and the swap counterparty will pay the index performance to the ETF. That can happen (I guess it usually does) daily.

Essentially the risk would be fairly limited in that case - i.e. to the delta of performance between these two portfolios ('random basket' vs index), between the date of Bank counterpty default (understand this is where the 10% limit plays a role) and reinstating the swap with another Bank ?

[assyadh]

What synthetic ETFs do (at least in Europe) is they hold a basket of "random" securities and a swap agreement. The swap agreement means that the ETF will pay the basket performance to the swap counterparty, and the swap counterparty will pay the index performance to the ETF. That can happen (I guess it usually does) daily.

Essentially the risk would be fairly limited in that case - i.e. to the delta of performance between these two portfolios ('random basket' vs index), between the date of Bank counterpty default (understand this is where the 10% limit plays a role) and reinstating the swap with another Bank ?

Keep in mind asset managers collide with their sister investment banks.

Lyxor deals with sgcib and amundi with bnp cib.

[alex_686]

What synthetic ETFs do (at least in Europe) is they hold a basket of "random" securities and a swap agreement. The swap agreement means that the ETF will pay the basket performance to the swap counterparty, and the swap counterparty will pay the index performance to the ETF. That can happen (I guess it usually does) daily.

Essentially the risk would be fairly limited in that case - i.e. to the delta of performance between these two portfolios ('random basket' vs index), between the date of Bank counterpty default (understand this is where the 10% limit plays a role) and reinstating the swap with another Bank ?

I think I have mentioned before that the maximum risk is less than 250k. Even this can be mitigated via a single party Coedit Default Swap (CDS).

[webkonsola]

I think I have mentioned before that the maximum risk is less than 250k. Even this can be mitigated via a single party Coedit Default Swap (CDS).

Comes with a premium

[Lauretta]

Synthetic ETFs do not avoid the IRS witholding tax. Section 871(m) of the internal revenue code made sure of that:

https://www2.deloitte.com/ch/en/pages/f ... -code.html

According to this article, that section cannot be applied to ETFs
https://www.lyxoretf.fr/pdfDocuments/DT ... F01%2F2019

[vsquid]

I understood that when the Tracking Error is negative, as in the webpage krasnall provided, it means that the ETF actually outperforms the index. Perhaps by lending securities (if it's a physical ETF)?

You should check what index they use. Typical SP500 ETF uses SPTR500N which is net total return which means that it assumes a 30% withholding for dividends. However only 15% is withhold for Ireland domiciled funds. This is enough to explain why any SP500 ETF beats the index.

iShare's new swap-based SP500 ETF also lists SPTR500N as its benchmark. It will be interesting to see what performance they will really be getting. Is there any plausible explanation how they could avoid paying tax on dividends? My understanding is that index options etc omit dividends.

[Lauretta]

iShare's new swap-based SP500 ETF also lists SPTR500N as its benchmark. It will be interesting to see what performance they will really be getting. Is there any plausible explanation how they could avoid paying tax on dividends? My understanding is that index options etc omit dividends.

I think the explanation is here:
https://www.lyxoretf.fr/pdfDocuments/DT ... F01%2F2019

I500 is not yet listed by my broker, and according to the LSE website the last trade was on 16 October!!
https://www.londonstockexchange.com/sto ... mpany-page
not sure if there's some mistake, but it doesn't look good... I am also considering Lyxor SP5L; they have a much larger AUM as they've been around longer; they are LU domiciled but since there's no withholding tax I don't think it makes a difference(?)

[TedSwippet]

Synthetic ETFs do not avoid the IRS witholding tax. Section 871(m) of the internal revenue code made sure of that:

https://www2.deloitte.com/ch/en/pages/f ... -code.html

According to this article, that section cannot be applied to ETFs
https://www.lyxoretf.fr/pdfDocuments/DT ... F01%2F2019

That's a really interesting document. Thanks for linking to it. Its implication is that SP5 could well be an efficient alternative to Vanguard's UCITS VUSD.

Note that SP5's TER is 0.15%, but VUSD's is 0.07%. This higher TER narrows SP5's possible advantage, but doesn't destroy it entirely.

[Lauretta]

Synthetic ETFs do not avoid the IRS witholding tax. Section 871(m) of the internal revenue code made sure of that:

https://www2.deloitte.com/ch/en/pages/f ... -code.html

According to this article, that section cannot be applied to ETFs
https://www.lyxoretf.fr/pdfDocuments/DT ... F01%2F2019

That's a really interesting document. Thanks for linking to it. Its implication is that SP5 could well be an efficient alternative to Vanguard's UCITS VUSD.

Note that SP5's TER is 0.15%, but VUSD's is 0.07%. This higher TER narrows SP5's possible advantage, but doesn't destroy it entirely.

Yes I noted that. I500 could be even better since its TER is 0.07% but

I500 is not yet listed by my broker, and according to the LSE website the last trade was on 16 October!!
https://www.londonstockexchange.com/sto ... mpany-page
not sure if there's some mistake, but it doesn't look good...

[assyadh]

Synthetic ETFs do not avoid the IRS witholding tax. Section 871(m) of the internal revenue code made sure of that:

https://www2.deloitte.com/ch/en/pages/f ... -code.html

According to this article, that section cannot be applied to ETFs
https://www.lyxoretf.fr/pdfDocuments/DT ... F01%2F2019

Really, really interesting document. Thanks for correcting me. That finally kind of explains why Lyxor and Amundi keep registrating their ETFs in Luxembourg.

I did some more googling on The Section 871(m) Qualified Index Rules: https://www.sullcrom.com/files/upload/P ... r-2018.pdf

Requirement that Index be Passive, Diverse, and Widely Used.

Apparently, the MSCI World index would not satisfy this:

Whether listing or minimal trading satisfies the “traded” requirement could affect whether
some widely used indices can constitute a qualified index. For example, the MSCI World Index
is a widely used index for which there is a significant amount of futures that trade on foreign
exchanges. The trading on foreign exchanges, however, will not satisfy the “traded” requirement because U.S. stocks comprise most of the index. Futures contracts on the index are
also listed on ICE (which is a U.S. exchange), but there is very little trading of the contracts.
Similarly, the MSCI USA index is a widely used index for which futures contracts are listed on
ICE, but there is a minimum amount of trading of the contracts. Accordingly, both of these
indices may not constitute a qualified index if the “traded” requirement can only be satisfied if
there is regular or substantial trading of the futures contracts. 43Reg. § 1.871-15(l)(1)

[gougou]

Just buy and roll over S&P 500 futures to avoid withholding tax.

[hushroom]

For MSCI World and most others there is basically no difference in real returns between synthetic and physical

Apparently, the MSCI World index would not satisfy this:

This Invesco article from June 2020 claims synthetic replication is advantageous for the MSCI World index as well (emphasis mine).

Under US tax law, certain types of derivatives are subject to an equivalent withholding tax rate if they pass through “dividend-equivalent” payments. Specifically, when a US bank writes a swap on an index with a non-US counterparty, it is generally required to withhold US dividend tax at the same rate as would apply to a physical investment by that same counterparty.
This would seem to level the playing field between physical and synthetic replication strategies, but the same rule that imposes this tax treatment on derivatives also specifies certain exemptions, and the MSCI World Index meets the criteria for these exemptions. Namely, section 871(m) of the HIRE Act explicitly excludes swaps written on indices with deep and liquid futures markets from the requirement to pay dividend withholding taxes.
This means that, while a European-domiciled physically replicating ETF will generally be able to achieve a maximum of 85% of the dividend yield of the US holdings in their MSCI World portfolio, a synthetically replicating ETF can achieve up to 100% of the full gross dividend amount. With an average dividend yield for US large-cap stocks of approximately 2%, this exemption means synthetically replicating funds can potentially achieve up to 30 basis points of additional performance on US exposures, which equates to around 19 basis points on the MSCI World Index

[dulac]

This is a fascinating find.

Let's look at Lyxor SP5 vs Vanguard VUSA:
https://www.trackinsight.com/en/compare/1242,1283
https://pasteboard.co/Jx7oeYV.png

Any ideas why is the performance of the lyxor synthetic sp500 below the perf of VUSA?
Considering it's supposed to save on 15% of dividend witholding tax, shouldn't it beat it?

https://www.trackinsight.com/en/compare ... 7698,12058
https://pasteboard.co/Jx7tolO.png

[krasnall]

Note that in the table there you have results until 21 and 22 OCT - different dates, different results.

If you compare official NAV from the fund providers:

Code: Select all

Fund		2016	2017	2018	2019
Lyxor S&P500	11.79%	21.69%	-4.55%	31.31%
Vanguard S&P500	11.51%	21.37%	-4.76%	31.01%

You can also compare it a bit less reliably (because the results are in EUR) but in an easier format on Tracking Differences

[hushroom]

In Europe, UCITS swap based ETF limit the counter party risk to 10% of the fund's value.

Any source for this? If the CP risk is only 10%, it seems a worthwhile tradeoff for the increased performance, especially since there is some CP risk anyway in physically replicating funds that lend shares.

[krasnall]

Any source for this? If the CP risk is only 10%, it seems a worthwhile tradeoff for the increased performance, especially since there is some CP risk anyway in physically replicating funds that lend shares.

Honestly I cannot provide a direct regulation. I can only point you to the Lyxor guide:

Under UCITS regulations, an ETF can only ever put 10% of its value at the risk of a counterparty failure. In practice this means the ETF’s assets must cover at least 90% of its value. At Lyxor, we go much further than the regulation. Our aim is to ensure our synthetic funds hold enough assets to cover their value in full. This means managing the assets daily, and topping them up if they ever drop below their target

On their S&P500 ETF website they also give you a daily counterparty risk level indication, which actually says "N/A" and explains:

From 1st March 2017, the European Market Infrastructure Regulation («EMIR») requires certain EU counterparties, including undertakings for collective investment in transferable securities («UCITS»), to put in place risk mitigation procedures prior to entering into OTC derivatives trades. This regulation obliges both parties to a swap transaction to exchange collateral in order to reduce any counterparty exposure to zero allowing for a minimum transfer amount of €500,000.

I also found a study where you can read more about the counterparty, but also liquidity risks.