Exchange-traded funds (ETFs) have made building a diversified investment portfolio easy, liquid, and cost-effective. Most equate ETFs with broad index investing, and the all-in-one asset allocation ETFs are an elegant solution for that. They are designed intentionally to make diversification work for you. By its nature, dividend investing will be less diversified because you are concentrating on dividend-payers and eschewing stocks that pay low or no dividends. In the last post, I discussed the challenges of picking dividend stocks. Using a dividend ETF that applies a solid methodology at a reasonable cost seems like an elegant solution to those issues. This post will do a deep dive into the best Canadian dividend ETFs to see how they measure up.
Acknowledgements & Data Source
This post will go well beyond a simple listing of common ETFs and their fees. There will be an analysis of ETF characteristics (factor weighting) known to outperform in other broad-market datasets. There will be an explanation of the more advanced analysis concepts used. How the ETF operationalizes its strategy also matters. There can be hidden costs to the buying and selling of stocks, which make up these funds. Those frictions, along with the fees, can lead to tracking error relative to what you would expect.
If the premium from factor weighting does not offset the drag from fees and implementation costs, it may be detrimental overall. The juice may not be worth the proverbial squeeze. While historical performance does not predict the future, it can provide insight into how effectively the strategy captured these premiums (net of costs) in the past.
This data and analysis are difficult to obtain for Canadian ETFs. So, I enlisted the help of Andrew Jones, CFA. He is a chartered financial analyst. Outside of his regular job as a flat-fee financial coach at Cardin Partners, he is an enthusiast of analyzing factor ETFs and educating others about them. His site, Verified Beta, includes an up-to-date list of both US and Canadian ETFs with meaningful factor exposure. The AQR factor database was used for today’s analysis — it is the best publicly available database for deep factor analysis of Canadian funds.
Disclaimer: This article is for information and entertainment purposes only. None of this is meant to be investment advice. You must do your own due diligence and make the best investing decisions for your specific situation. Seek professional advice as required.
Dividends, Factors, & Broad Market ETFs
Dividend ETFs
Dividend-focused ETFs cover a more concentrated portion of the broader market. Even though that deviates from the principle of diversification, it is attractive to some investors for psychological reasons. The tax treatment of dividends varies for personal and incorporated dividend investors. It may be a boon or a drag, depending on the situation. There is potential for a tax advantage with Canadian dividend payers while foreign dividends face higher taxes. So, this post will focus on Canadian dividend-paying ETFs.
There are a variety of dividend-paying ETFs on the market. Even ETFs that are not labelled as “dividend ETFs” may pay high dividends – often at a lower cost. My previous post about the evidence for outperformance of dividend stocks stated that dividends are actually a surrogate marker for other “factors” that are more predictive of performance, using the Fama-French five-factor model. Those predictive characteristics are value pricing (HML – high minus low), conservative investment management (CMA – conservative minus aggressive), and profitability (RMW – robust minus weak).
Fama-French Factor ETFs
You could choose factor ETFs, also commonly known as “Smart Beta” or “Quant ETFs” that target those factors directly rather than indirectly through dividends. The US market has multiple options for doing that, including ETFs from AQR, Avantis, Dimensional Fund Advisors, and others. These funds target combinations of factors from the Fama-French model. The Canadian market is more limited. There are few good Canadian factor ETFs targeting combinations of the Fama-French factors.
Other Factor ETFs
BlackRock & BMO ETFs have some offerings that target other factors. BMO’s low Volatility ETF (ZLB) was previously discussed on the blog. Like dividends, low volatility is a surrogate for CMA and RMW factors from the Fama-French five-factor model. One option is to use ZLB for some factor exposure and it has historically outperformed compared to the S&P/TSX60. BMO ETFs have also recently opened a Quality Factor ETF. BlackRock also has its versions of factor ETFs. As explored in this article, dividend ETFs are a good way to capture Value and Quality factors.
Broad Market ETFs as a Comparator
One challenge in finding data on factor weighting for Canadian ETFs is that most databases use the US market as the comparator. For example, a Canadian ETF with a high concentration of the profitability factor relative to the broad US market is what most online portfolio analyzers would report. However, that is not how Canadian equity factors are determined. They are composed of the universe of stocks in the Canadian stock market.
Compared with other global markets, the Canadian market is tilted toward industries that should favor conservative management and quality factors. So, just buying a broad Canadian market ETF may provide factor exposure relative to other markets outside of Canada. This means using a Canadian market comparator is important. We also need ETFs that have been around for a long time, to ensure a short performance history is not a limiting factor when we compare.
Broad Market ETFs as Default
Since the starting point for investors is usually a low-fee broad-market fund, we will start by analyzing a couple to use as our baseline comparators. I will also use it to explain what some of the more advanced analysis numbers mean. We want to use ETFs with a long track record to maximize the data. So, we chose VCN.TO (inception Aug 2013) and XIU.TO (inception Nov 1999). iShares’ XIU.TO covers the TSX 60, which includes the 60 largest Canadian stocks, while VCN is broader because it caps the allocation to a single stock, resulting in more mid-sized stocks being added and less concentration in the largest names.
We used the AQR database since it has Canadian market data in Canadian dollars. It’s the “Canadian Market”, an all-cap benchmark. The coefficient of determination (R2) for the factor regression for VCN and XIU against AQR’s Canadian Market factor was 0.99 for VCN and 0.95, respectively. That is extremely high—and not surprising.
Correlation Coefficient & Factor Load
The AQR database uses Quality (quality minus junk or QMJ). The quality factor basically subsumes the CMA and RMW factors from the Fama-French Model. The exposure to Market Risk (Rm-Rf), Value (HML), and Size (SMB) factors should be familiar as the other factors in the Fama-French five-factor model.
We quantify the “factor load” that a fund has for a given factor by using its regression coefficient. A regression coefficient quantifies how much of the fund’s return is explained by that factor exposure. Basically, a negative coefficient means that factor weighting is associated with decreased returns. A positive coefficient means that the fund’s exposure to that factor has increased returns. The closer the number is to one, the heavier the influence, and zero means no influence.
You can see in the chart below that VCN’s return was explained by its market exposure (1.00). There were slight differences due to exposure to fewer small companies (-0.13), lower quality (-0.03), and more value pricing (0.02). However, those were all very weak (close to zero). That is what you would expect because the ETFs cover most of the comparator AQR Canadian market. XIU has slightly less exposure to small companies (-0.22) since it limits itself to the largest 60 stocks.
T-Stat (Is it random chance?)
The table also includes a t-statistic, which measures how many standard deviations each coefficient is from zero. Larger t-values arise when the estimate has a smaller standard error or a larger sample size. The t-statistic determines the p-value used to test whether a coefficient is statistically different from zero—that is, whether the observed relationship is likely genuine rather than due to random variation. As a rule of thumb, t-statistics between -2 and +2 correspond roughly to p-values above 0.05 and are generally considered not statistically significant.
When we look at enough variables in a regression, the probability that one of them will have a p<0.05 rises. With 20 variables, you would expect it once by random chance. A t-stat accommodates for that. So, a t-stat outside +/-2 is considered statistically significant in our analysis. I have highlighted significant positive factor weightings in green and significant negative factor weightings in red.
In the preceding chart, the broad market Canadian ETFs had a significant market exposure and underweighting to small-sized companies. While statistically significant (t-stat below -2), the magnitude of impact from over exposure to larger companies seems minor (coefficients of -0.13 and 0.22). However, that is historically a meaningful weighting. A small difference in returns makes a meaningful difference when compounded over years.
Alpha = Better or Worse Return Not Explained by the Factors
Most investors have heard someone talk about “alpha” at some point. Most commonly, it is a manager or someone who sells managed funds talking about the “alpha” generated by their strategy. Alpha is the increased or decreased return of an investment strategy that isn’t explained by the risk exposure.
For example, a skilled manager may achieve a return that exceeds your expectations compared to their benchmark, net of their fees. Of course, the timeframe and benchmark are key, and managers can improve their apparent performance through careful comparator selection. It is well known that it is rare to find an active manager who will generate alpha consistently over the long run. Further, you would have to identify them before that is known because their advantage is arbitraged away as it becomes known.
Most “passive” funds would be expected to have a negative alpha that is similar to their fees and trading costs. Keeping those costs low is how they get the best return. Our baseline ETFs of VCN and XIU have low MERs of 0.06% and 0.18% respectively. Their alphas in the regression were not statistically significant.
Major Canadian Dividend ETFs
There are many ETFs that focus on Canadian dividend-paying stocks. However, for today’s analysis, I will stick to ones that are large with high trading volume. I have also excluded some with corporate class structures, leverage, or options usage. These factors introduce additional costs and risks, and they did not outperform the Big Five. The tickers of the five major Canadian dividend ETFs are CDZ.TO VDY.TO XDIV.TO XEI.TO ZDV.TO . You can click the linked ETF ticker to get their details, but I have summarized the highlights below.
The Dividend Aristocrats ETF has the highest MER, while the Vanguard and other iShares ETFs have the lowest MERs. Fees do drag on performance. However, with dividend investing, there is also an impact from the dividend stock-picking strategy and how effectively fund managers operationalize it. So, we will need to take a deeper dive into the ETF composition and historical performance to gain insights.
All of these funds have been around a long time and have over $1 billion in assets. So, they are unlikely to be closed down due to lack of interest, as often happens with small niche ETFs.
Factor Regression For Canadian Dividend ETFs
Dividend investing may have historically outperformed due to the implicit overweighting of value (HML), conservative investment management (CMA), and profitability factors (RMW) using the Fama-French five-factor model. Those factors explain the returns regardless of whether the stock pays dividends. So, how do the Canadian dividend ETFs stack up when we perform a factor regression (using Quality as a substitute for CMA & RMW)? As expected, they have had a statistically significant overweighting of value and quality. They were also about as underexposed to small-cap stocks as the TSX 60.
The factor exposure explained roughly 90% of historical return variability (R-squared). The other unexplained 10% of return variability was associated with a lower-than-expected performance of 0.23% to 2.38% (alpha). That may reflect fees, other operational costs, management, or the fact that these funds are not shorting overvalued and low-quality stocks explicitly (they are only excluding them). It could also be attributed to the fact that high-dividend-paying stocks are not always the highest-quality stocks. As these funds rebalance towards higher dividend payers, they may sometimes hold somewhat lower-quality stocks for periods. This is generally called ‘time-varying factor exposure’ and is one of the risks of relying on high dividend-paying stocks as a proxy for both value and quality at the same time.
As mentioned, one reason why real-life factor investing in “long-only” funds is more challenging than factor research model portfolios suggest is that it is difficult to achieve heavy factor exposure. The research portfolios went long the top 30% stocks with a factor and shorted the bottom 30% (except SMB, which was a 50/50 split with the median market-cap stock dividing the two groups). A real-life manager would face the increased costs of shorting or only capture the long-only side of the model. Most opt to use a long-only strategy. So, having factor weightings of low magnitude (0.14-0.21 for value and quality) in our dividend ETFs is not surprising, and it is very difficult to get a heavier factor load without shorting. Those factor weightings are statistically significant, but are they clinically significant enough to overcome the other costs of the ETFs?
The Negative Alpha of Canadian Dividend ETFs
Importantly, the negative alpha in the preceding chart does not necessarily mean that they underperformed the broader market. It just means they had a lower return than expected from their factor exposure—a strong, favorable factor exposure could have overcome those issues. However, the reality is that it is very difficult to do that net of fees, implementation costs, and not actively shorting the stocks with the lowest exposure to desirable factors (as was done in the academic research).
Relative Historical Performance of Canadian Dividend ETFs
Below is the historical performance of our Canadian dividend ETFs compared to each other and to our broad Canadian index ETF baseline comparators. It is during the same time period to keep the market regime comparable and is limited by the newest ETF (XDIV inception in 2017). Of the dividend ETFs, XDIV had the best risk-adjusted performance (Sharpe Ratio), even though it trailed VDY and XIU.TO compound return. VDY had a higher CAGR, but a lower risk-adjusted return.
While it is harder to gauge the risk-adjusted return, the chart below shows the relative growth of $1 invested over the period. There is significant overlap, but the top cluster includes VDY, XIU, VCN, and XDIV. The lower cluster has XEI, ZDV, and CDZ.
The above analysis assumes dividends are reinvested, but it ignores taxes. Dividends are taxed annually, which would further drag on the dividend ETF after-tax performance, except in the lowest tax brackets.
Performance of Canadian Dividend ETF Since Inception vs XIU.TO
Whether a factor increases or decreases returns is not static. There are periods of time or market regimes when they shift in and out of favor. That is why I kept a fixed time period above. The following chart shows the historical performance of the dividend ETFs compared to XIU.TO since each ETF’s inception. This is useful for comparing a dividend ETF to the broad market over the longest possible period (dividend ETF vs XIU). However, it is not useful for comparing dividend ETFs to each other, as they span different market periods.
Again, only VDY slightly outperformed XIU over its lifespan, but also had more risk. Only XDIV had a better risk-adjusted return. However, it has been around for only 8 years and has had a lower CAGR than XIU over that period.
Best Canadian Dividend Paying ETFs
Of the dividend ETFs, CDZ has been around the longest, and VDY is the largest. However, all the major dividend ETFs considered today are large and established enough to likely stick around. They all hold liquid underlying stocks.
Factor Weighting
As expected, the Canadian dividend ETFs had significant exposure to the Market, Value, and Quality factors. The exception was XDIV, which did not have significant Quality weighting. Those factor tilts would be expected to have a return premium based on historical factor research. However, those research portfolios use both long and short strategies to factor tilt. So, the factor weighting is lower in these real-life investment funds.
Another important thing to realize, if you are factor investing or using a dividend ETF as a surrogate, is that different factors outperform or underperform during different market periods. They may also influence each other, such as small size accentuating the impact of other factors. For example, we took a deep dive into the period from January 2020 to January 2025. During this time period, Market exposure, small size, value, and momentum had positive impacts on performance. Quality had a slightly negative impact.
During that time period, the dividend ETFs (except CDZ) held “lower beta” stocks, having a lower exposure to the Market (Rm-Rf) factor, which caused underperformance during this period. The value exposure of our dividend ETFs helped. However, the underexposure to smaller companies hurt them (although not statistically significant with the small sample). The impact of Quality weighting was not significant. This highlights the problem of treating “dividends” as a proxy rather than a deliberately designed factor-based investing strategy. While the dividends stocks did have Value and Quality weighting, the lower exposure to market risk and size overshadowed that.
Overcome by Costs
Real-life funds also face costs to manage and implement their strategy. There are explicit trading costs, and a more subtle loss due to frictions such as the bid-ask spread. All of the dividend ETFs had a negative alpha – a drag on performance from these types of variables. Of the dividend ETFs, CDZ had the smallest negative alpha despite its higher MER. Having a negative alpha (unexplained return) would not be a problem if there was enough excess return explained by the other factors. However, dividend ETFs struggled to outperform, net of their costs, and despite having some favorable factor tilts.
Hard to Beat Low-Cost Index ETFs
Even though the broad Canadian Index ETFs, like VCN and XIU, did not have favorable factor tilts relative to the broad Canadian market, they have extremely low fees and operating costs. That makes them tough to beat. Over a comparable time period (2017-2025), the dividend ETFs were unable to do it. Only VDY came close, but still had a lower risk-adjusted return than XIU. Looking over more extended time periods, only VDY eked out a higher CAGR relative to XIU since inception. They also did this with higher volatility (as measured by standard deviation or Sharpe ratio).
While Canadian dividend ETFs may have slightly higher dividend yields (3.75-4.67%), the broad market ETFs also had eligible dividend yields of 2.40-2.46%. Some investors may prefer dividends for psychological reasons, but they would still get some from a VCN or XIU. Today’s analysis did not include taxes. Eligible dividends may have favorable or detrimental tax treatment relative to capital gains, depending on the specific situation. Personally, my psychological benefit and tax situation would have to be pretty compelling for me to focus on a dividend ETF. However, we are all different in that regard, and my hope is that this dividend investing series along with today’s post will help you more carefully evaluate whether this popular strategy really is the best one for you.
Good analysis
Thanks Park! It took us like a month to pull it together, but it is pretty unique (I think).
Mark