The author has used a 15 year sample to backtest this strategy, which is, "buy the previous year's worst performing bank relative to the group". According to his research, this strategy "

*is a remarkable and effective stock-picking strategy*".

My comments are below:

- BNS has about 47% Int'l exposure vs. 53% Cdn exposure in terms of revenues. Latam accounted for $5.3B of total revenue in 2015, or 22% of total, compared to $5B of total revenue in 2014, or 21% of total. BNS also has significant revenues and income from "Other International" (Europe, Asia, etc), about $4.8B (20%) / $1.3B (19%) in 2015 vs. $4.5B (19%) / $1.3B (18%) in 2014. The author does not provide much if any real detailed analysis in the article about what the geographic distribution means. BNS has obviously lagged the peer group for a reason though, and it is the job of the insightful investor/analyst to try to understand why. Is close to 36% (2015) vs. 35% (2014) of profits coming from Other International and Latam a concern? Yes/no/maybe? And if it isn't, why has BNS underperformed peers? The author seems to be looking past this relative under-performance without asking why

- I analyzed all of the CDN banks on price book basis. BNS is not absolutely cheap, but it is cheaper than it's peer group. It's trading @ 1.37x book. This compares to BMO @ 1.38x book, TD @ 1.6x book, RY @ 1.88x book, NA @ 1.42x book, and CM @ 1.78x book. Cdn Western Bank (the true contrarian pick of all the Cdn banks due to loan exposure in Western Canada) trades at .87x book.

- ROE for all the Cdn banks is significantly higher than US peers. I think this is b/c large US banks have had to maintain higher capitalizations under more stringent regulatory requirements coming out of the 08/09 crisis. I take this to mean that in Canada, higher ROE is more a case of higher leverage allowed, until now. Apparently, this is about to change under new OSFI capital adequacy requirements leading up to 2019. If Cdn banks have to post more regulatory capital, both BVPS will increase and ROE will decrease, see link here

- Is 15 years enough of a sample? It includes the 2000 tech bubble crash (mostly irrelevant to Cdn mortgage loans) and the 2008 crash (minimal impact due to US subprime issues). If the author really wants to analyze Cdn bank performance properly, he needs to expand his sample to include Cdn real estate busts, and for this, he needs to go back to 1979-1982 and possibly earlier

- BNS despite being down 15% in 2016, and down 24% from its peak in 2014, is still up 12% compound since the 09 lows (not including dividends) and is up almost 200% on an absolute basis. And yet, the author arbitrarily suggests that buying BNS on the first trading day of 2016 is bound to work because
*"bank stocks tend to revert to the mean"*. Again, I go back to my point above. 15 years is too short a sample to suggest that banks revert to the mean, because the mean over 15 years is a) misleading, and b) will be completely different than the mean over say, 100 years. If he really wants to write about mean reversion, he should do the correct work up front in order to determine a reasonable estimation of what the mean should be. Another inconsequential point, define "mean"

- If the mean over a a statistically representative time period should actually be book value, and a reasonable reversion statistic is 1 x book, then BNS reverting to the mean would suggest a decline from current of another $15, or -27% from here. The author's time period includes at least two financial bubbles!

Here's BNS monthly since 1988. Monthly trend going back to 1993 suggests a target of $40, which coincidentally, is 1 x book:

__Concluding Thoughts__
I find it absolutely hilarious that financial writers suggest ideas without doing in depth research. It is the job of the curious investor to fill in the holes of any investment thesis with their own research, and never take any suggestions used to fill the information void in the local paper as advice to act on.

__Update:__

**I was curious to see what the historical P/B was using the author's outside of sample years (inclusive of three equity bubbles (the third of which, we're arguably still in the midst of). Funnily enough, P/B x 1 is not out of the question, especially in the context of years in which Canadian housing got hit. The author has made the error of using mean price performance of the group as a statistical proxy for mean reversion. Oh well, back to my accounting hole.**

Ironically one of your points against his lack of fundamental analysis is using technical analysis of the monthly trend line. Personally, I think if you can buy Canadian banks when book is less than 10x ROE you will do ok but its all about what your hurdle rate is.

ReplyDeleteI think my main issue is the lack of definition of mean over a non-representative sample in terms of time. I've suggested two possible definitions, 1) p/b = 1, and 2) lt monthly trend. If book value is an acceptable measure of valuation over time, and 1x book is mean, I'd be surprised if lt monthly trend didn't also encapsulate 1, but then again, what do I know. Nothing really.

ReplyDeleteBut isn't P/B just a function of ROE? If a business has a sustainable ROE of say 30% (about double the CAD banks just to make a point) wouldn't you be willing to pay a much higher P/B multiple? Obviously your margin of safety is lower if the ROE starts to compress than if you can buy at 1x. Its why the US banks who generally now have ROEs and P/Bs below 10% and 1 are so interesting. The margin of safety seems high (low P/B) and ROEs should trend north of 10% over time so book value growth and multiple growth.

ReplyDeleteGreat points and totally agree. Probably explains much of the premium p/b in Canada relative to the US. What happens post 2019 is anyone's guess though

ReplyDelete