As ETF investing becomes mainstream at the international level, differences in withholding tax rates (WHT) on dividends among jurisdictions have become an area of focus for non-resident investors who want to know, Is the WHT differential a significant driver of ETF performance? And how do we compare their performances?
Selecting “the right ETF” extends far beyond its tax on dividends and fees. But, before we examine the holistic set of criteria to choose an ETF, it’s important that we first step back and put the ETF selection into context.
An ETF in a given portfolio is there to track a particular asset class. Therefore, before even selecting the ETF, it is critical to first determine which asset classes will best meet your portfolio’s target performance in a given economic environment. Ultimately, the goal is not to beat the asset class, but to track it. Thus, it requires a holistic view to evaluate ETFs, and then to choose the ETF that is able to do that most effectively and efficiently.
At StashAway, we apply this holistic view, and consider the following criteria when selecting the ETFs that make up the portfolios we manage:
- Low tracking error
- Sufficiently long track record for statistical analysis (minimum requirement of 5 years but we prefer more than 10 years)
- High liquidity and trade-ability
- Ranks well in net returns (adjusted for fees and withholding taxes) against comparable ETFs with similar volatility
- Solidity of the ETF’s fund manager
- Simplicity: As StashAway is providing sophisticated solutions at the portfolio level, we do not wish to incur complexity at the ETF level. In particular, we avoid ETFs that offer leverage (e.g. 3x Gold price) and also those with credit risk exposure to the issuer (e.g. Synthetic ETFs, or ETNs).
To put these evaluation criteria into practice, let’s consider the bellwether example: tracking the S&P 500 index.
Here, we will evaluate the USD-denominated net returns of US-listed ETFs. (Note that for our purposes, we first deduct estimates for fees and dividend taxes.) Simultaneously, we are interested in comparing tracking errors of the ETFs listed in other countries.
First, let’s look at WHT. As can be observed in Table I, the London-listed IUSA LN is the only ETF in the table that does not incur any WHT on dividends. In comparison, a non-resident investor who invests in the US-listed SPY US would incur a WHT of 30% on dividends. The WHT figures for ETFs listed in Ireland and Luxembourg are lower than those in the US, seeing tax rates at 20% and 15%, respectively.
Sources: Bloomberg and Morningstar; WHT rates from Deloitte1 and PwC2.
Please note that these tax figures serve only as a guide to facilitate the comparative study. StashAway is not in the business of providing tax advice and investors should consult their tax accountant for advice before making any investments.
Now, let’s first compare the net returns of SPY US and IUSA LN, as they are the only two ETFs in Table I with long histories (April 2002 to April 2017). As previously mentioned, there is a WHT differential of 30% between the two ETFs; this means the IUSA LN has an advantage of 0.56% per annum over the SPY US. However, the difference in realised net returns between the two ETFs is only 0.05% in the period. So, despite a significant tax differential, the net returns ultimately are very similar to each other.
In another comparison where we have included all of the ETFs in Table I (between October 2010 and April 2017), we fail to observe any discernible relationship between net returns and WHT. For one, the IUSA LN, which is the ETF with the largest WHT advantage, was not the best performing tracker fund in the group. In fact, the ETFs in jurisdictions with higher WHT have greater net returns (e.g. SXR8 LN, XSPX IM and SP5 FP) compared to the IUSA LN. While the XSPX IM outperformed IUSA LN (13.14% compared to 12.97%, respectively), the other Luxembourg listed ETFs, such as the C012 GY, had realised lower net returns of 12.88%. Over the shorter 7-year period, between October 2010 and April 2017, the additional return that IUSA LN realised over the SPY US widened from 0.05% to 0.28%. However, this additional return was only half of the WHT advantage that IUSA LN had over SPY US.
These two comparisons show us that ETFs with vastly different tax structures can achieve almost identical net returns, and that tax advantages are not an indicator of better performance.
Source: Bloomberg, Snapshot as of 30th May 2017.
In Table II, we explore the other ways ETFs tracking the same asset class differ. Here, we find evidence of lower liquidity and trade-ability amongst ETFs that track an index that is based in another time zone. This is due to the bid-ask spread that is a direct result of the time changes. To illustrate, SPY US’s average bid-ask spread is only 0.4bps. The bid-ask spreads widen significantly as soon as we look outside the US at the average bid-ask spreads of non-US listed ETFs also tracking the S&P 500. (These average bid-ask spreads are expressed as a percentage of ETF prices and estimated on May 30, 2017.) Even with larger ETFs, such as the IUSA LN and CSP1 LN, average bid-ask spreads could be as wide as 3.6bps and 3.7bps, respectively. For smaller ETFs such as the C012 GY and SP5 FP, the average bid-ask spread could be as high as 6.9bps and 7.3bps, respectively.
It is no surprise to us that there is a corresponding increase in tracking errors for ETFs tracking an index in another time zone, largely due to the bid-ask spread. In effect, the IUSA LN earns similar net returns as SPY US, but does so with significantly greater uncertainty. Its tracking error of 3.08% per annum (since April 2002) is much larger than the 0.66% with SPY US.
Tracking methodologies have clearly improved in more recent times, and the IUSA LN did reduce its annualised tracking error from 3.08% to 2.64% (since Oct 2010). However, this reduction in annualised tracking error paled in comparison to SPY US, which saw a reduction from 0.66% to 0.25%. In other words, 66.67% of the time, the returns of IUSA LN could deviate from the S&P 500 index by an annualised 2.64%, in any given year, whereas the returns of SPY US could deviate only by an annualised 0.25% in any given year.
Ultimately, ETFs are simply vehicles for tracking an asset class. Our analysis reveals a need to consider performance metrics that are adjusted for both costs and tracking uncertainty. Having said that, we find validity in WHT differential as a significant driver of performance when investing in high yielding assets. Further, we highly recommend adopting a holistic approach to ETF selection, because as illustrated in our comparison between IUSA LN and SPY US, although the two vehicles may achieve the same net returns, the certainty that they will do so varies tremendously.
At StashAway, we devote ourselves to identifying the right mix of asset classes for a given economic regime, because the appropriate selection of asset class mixes is vital for a portfolio to achieve effective diversification over the long term. From there, we use only the securities that are most reliable, cost-effective, and best-performing to represent the asset classes we determined as necessary.
References1 https://www2.deloitte.com/content/dam/Deloitte/global/Documents/Tax/dttl-tax-withholding-tax-rates.pdf2 http://taxsummaries.pwc.com/ID/Withholding-tax-(WHT)-rates