Thursday, February 08, 2007

BMO CM on Mean Reversion of Canadian Banks

BMO Capital Markets, 6 February 2007


For the period 1970 to 2006, we examined the total returns (dividend income and price appreciation) of the Big Six Canadian banks relative to the S&P/TSX Bank Index, to determine if mean reversion was a significant phenomenon when it came to bank performance. Broadly speaking, we conclude that mean reversion did produce incremental returns, although the incidence of success was no better than average.

Our research was centred on the performance of the two extremes: the best-performing bank and worst-performing bank for each year. Performance was deemed to be the total return of the bank in question relative to the S&P/TSX Bank Index in the year immediately following its best or worst title year. To ensure that outliers didn’t skew our analysis, we excluded the two years with the highest and lowest relative returns.

Even the Worst Bank Stock Produced Strong Returns

While the Big Six Canadian banks might look similar on the surface, in reality, there are differences in how the individual banks have performed over the years, with Scotiabank and TD leading the pack and CIBC, BMO and National trailing (Table 1). Having said that, even the weakest bank stock has beaten the S&P/TSX over the past 36 years.

Table 2 shows the marked differences in the returns of the best and worst performing banks over the years. The gap between the two extremes has averaged 33% over the past nine years. In four of these years, the worst-performing bank has become the best-performing bank in the following year. On the surface, this is a compelling argument for reviewing the tendency of bank stocks to mean revert.

Mean Reversion for the Big Five Banks

In examining the returns of the worst-performing bank one year after its poor performance, we found the incidence of outperformance relative to the Bank Index inconclusive. In the 36 years we considered, the worst-performing bank outperformed the bank index in the following year exactly 50% of the time (Chart 1). This meant that, on an historical basis, an investor who bought the worst-performing bank in one year had only a 50% chance of outperforming the group the following year. However, in examining the scale of outperformance for the same data set, we found the results far more conclusive. The average performance for the worst-performing bank one year later was 3% higher than the bank index. This indicated that when banks did recover from their poor performances, they tended to show meaningfully above-average returns.

The other side of the coin was the performance of the best performers. Over 36 years, the best-performing banks underperformed the group, the year after, on 16 occasions (Chart 2). This indicated that banks which performed the best in one year were likely to outperform the index in the following year. However, when the scale of performance was considered, the best performers tended to falter; the average annual performance of the best-performing bank one year later was 0.5% lower than the bank index.

Mean Reversion Inclusive of National Bank

National Bank was excluded from the bank group in our previous analysis. Our logic has historically been that mean reversion works for homogenous groups. National, with its regional focus and smaller market capitalization, tends to be impacted by different variables. For perspective, the average market cap of the Big Five banks is roughly $48 billion. National has a market cap of only $10 billion.

However, it is important to note that when National Bank was included in the results, it negatively skewed the average incremental returns - particularly for those of the worst-performing mean reversion strategy (Chart 3).

With National included, the average long-term returns of buying the worst-performing bank drop to 1.2% from 3.0%, owing to several years of consecutive underperformance relative to the other banks in the early 1980s and early 90s. Having said that, National has closed the performance gap over the past five years as it has moved its ROE higher to match its bank peer group. We would not be surprised if it begins to ‘mean revert’ like the Big Five Banks.

The Investor Perspective

Mean reversion, therefore, appears to be a valid investment strategy and can produce positive returns if adhered to over the long term. Both strategies - buying the worst-performing stock or short-selling the best performing stock - would have produced positive returns since 1970 (Chart 4). Ignoring transaction costs and taxes, simply buying the worst-performing bank each year since 1970 would have produced a CAGR of roughly 2.8% above the bank index. (Note that this is different from the average shown in Chart 1 due to compounding effects).

Chart 4 highlights two interesting trends. First, the incremental returns earned from buying the worst-performing bank are higher than those earned by selling the best-performing bank. We believe this is a case study in behavioural finance, which contends that investors are more enthusiastic of strategies that have them buying weak stocks than strategies that have them selling strong ones. Put another way, banks tend to be 'oversold' in negative times, rather than 'overbought' in good times.

We also find it interesting to see that the trend in both incidence and scale of mean-reversion actually seems to be picking up, particularly for the worst-performing bank. In the 1970s, 80s and 90s, the average incidence of outperformance for the worst performing bank one year later was roughly 45%. That number has increased to 71% since 2000. Similarly, the average performance of the worst-performing stock one year later relative to the bank index between 1970 and 1999 was 1.7%. Since 2000, that number has increased to 11%. The reasons for this are less clear to us. It is difficult to tell whether investors have become more aware (and more proactive) of the mean reversion strategy, or whether bank executives are becoming more effective at correcting their mistakes. If indeed it is bank executives who are taking action more quickly to repair damaged stock prices, one dares question whether corporate time horizons have shrunk over time, making bank managers less willing to pursue long-term strategies that hurt short-term stock prices.

The Timing of Incremental Returns from Mean Reversion

We also thought it interesting to examine the speed with which the revaluation occured. In the 36 years we examined, the worst-performing bank outperformed the market in the first half of the following year roughly 53% of the time (Table 3). Furthermore, the average relative returns experienced by the worst performers in the fi rst half were 2.2% (interesting, given the average relative returns for the full years were roughly 3.0%). Similarly, the best performers underperformed the market one year later roughly 56% of the time, with an average relative underperformance of 1.8%.

This indicated to us that there is a seasonality component to mean reversion, with the majority of the re-balancing occurring in the first half of the calendar year. In other words, if you are going to employ the strategy, it pays to get in early, before other investors catch on or bank managers pull up their socks.


From the analysis above, we could speculate that a reasonable investment strategy today would be to buy BMO and short CIBC. BMO, the worst-performing bank stock of 2006 (Table 4), has held this title seven other times in the past 36 years. In the years after each of these times, BMO has outperformed the bank index by an average of 3.8%. CIBC, the best-performing bank stock of 2006, has been the best of the pack on six different occasions within the past 36 years. The bank’s one-year return following these incidents has averaged 4.4% lower than the bank index.

However, as analysts we believe that fundamental research does have value and that there are complexities to future bank performance that are not captured in the simple analysis of historical relative returns. While we do not deny that the Canadian banks tend, over time, to move as a group, we believe that their platforms, risks, opportunities and management teams can lead to meaningfully different total return performance in the short term.

We continue to prefer CIBC and TD Bank over other bank stocks. Both banks have significant exposure to the domestic retail environment, which we believe will remain healthy in 2007. CIBC will continue to reap the benefits of its loan portfolio restructuring as well as its cost control. TD Bank has an outstanding domestic footprint as well as a solid platform in the U.S., providing it both diversification benefits and the potential for growth.

Our other recommendation is National Bank, which fits squarely into our mean reversion argument. The second-worst performer of calendar 2006, National’s shares have meaningfully underperformed the bank group over the past 12 months (-0.5% vs. the bank group of almost 12%) due to BMO’s recent outperformance. The somewhat more attractive competitive environment in Quebec, combined with National’s solid retail and wholesale franchises, indicates that NA shares should be due for a catch-up.