Global Smallcap Momentum Investing And The Question Of Regional Allocation
The Question
In my global ranking system, I often see that the regional allocation, especially between US and exUS varies widely depending on the current market environment. If momentum of broad regional markets fades or certain stock markets become expensive or cheap as a whole, a concentrated long-only approach without constraints will start to concentrate allocations in a certain country or continent.
I now asked myself: “Is is better to maintain a certain regional allocation and treat regional momentum portfolios as separate components or should I let regional allocations ride freely depending on global single stock momentum.
Fixed Allocation
To do some tests, I downloaded the return data of 5x5 Size-Momentum sorts (cap-weighted) from the Ken French Data Library for Developed Markets, North America, Europe, Japan and Asia ex Japan. For each region, I extracted the Smallcap-HighMomentum corner portfolio.
The Developed Market Portfolio as published by Fama-French ranks regionally and puts all qualifying stocks back together into one portfolio. Depending on the weighting scheme (cap-weighted or equal-weighted), the resulting country allocation highly depends on the number and size of single stocks in each region. Here is the region categorization according to the database:
In comparison to the original Developed Markets Portfolio, I also created a own equal-weight portfolio allocating 1/3 to North America, 1/3 to Europe and 1/3 to Asia (50/50 Japan/exJapan)
Here is the cumulative wealth chart:
The EW portfolio already comes close to the Fama-French Developed construction. Slight deviation to the downside likely originates from the higher share of Asian stocks. Especially Japan is known as an outlier regarding its untypically low Momentum returns.
So equal-weight continental allocation is a good starting point. Let’s now have a naive look into optimized (fixed) regional allocation regarding risk-adjusted returns. If we solve for maximum risk-adjusted returns in the given time frame, we get the following “optimal” global Smallcap-Momentum allocation:
North America 34%, Europe 60%, Japan 0%, Asia ex Japan 6%
Of course, this is highly data-mined. Also you could argue that this is somewhat unfair because the late 90s were particularly rough for Asia investors. So let’s have a look at the same optimization but since 2000. The results are surprising. The optimal allocation since 2000 would have been:
North America 0%, Europe 85%, Japan 10%, Asia ex Japan 5%
Since 2000, it seems like regional differences became much smaller. Despite the drastic allocation without any US exposure, the optimized portfolio return doesn’t deviate very strongly from the EW Developed returns.
To conclude lesson #1:
Fixed long-term regional allocations to smallcap momentum portfolios most likely have to be very crazy and data-mined to make a difference in comparison to a simple equal-weighted continent allocation. Allocating 84% to Europe because of a over-fitted backtest does not add much value but adds a lot of uncertainty and fragility. So if you want a fixed regional allocation, consider going with an equal-weight approach. Other possibilities with a reasonable anchor would be weighting by total market cap (global asset allocation portfolio), by GDP or similar.
Momentum-Timed Allocation
In the next step, we try to find a better way to time the over- and underweighting of certain regions in the portfolio. As mentioned in the beginning, a global single stock momentum ranking over time will accumulate stocks of a region with high momentum as a whole. As a proxy, I analyze the rolling 12-month performance of the regional Smallcap-Momentum portfolios to find the “winner” region in each month.
The leaderboard over time looked as follows:
So, the question now would be if overweighting the prior winner region improved or deteriorated returns. To test this, we will compare the EW Developed portfolio to different momentum-timed portfolios.
In each momentum-timed portfolio, a different minimum allocation parameter (MAP) is applied. MAP ranges from 0 to 20. In each month, MAP % are allocated to each regional portfolio and the rest ((100 - 4*MAP)%) is allocated to the winner region to create an overweight.
Additionally, I created an “ExLoser” Portfolio, which excludes the portfolio with the lowest prior 12-month performance in each month and allocates 1/3 to each of the remaining 3 regions.
In a last step, I combined the two approaches to create an “ExLoser-OverweightWinner” Portfolio (0% Loser, 50% Winner, 25% each to the remaining 2 regions)
Here are the results:
We see that there is a trade-off between risk and return. While the most aggressive strategy (MAP=0), which only allocates to the winner region, outperformed the EW Developed Portfolio by 2.5%, it also has 6.2% higher volatility. Additionally, this approach would cause a lot of turnover, slippage, taxes and trading costs. If we go to higher values for MAP, we quickly lose the benefits.
A more promising approach seems to be the “ExLoser” approach. Kicking out only stocks from a loser region is less costly and already added 1.8% return by increasing volatility by only 1%. The “ExLoser-OverweigthWinner” approach did not add much additional value on paper.
So basically - as for many other investment cases - avoiding the losers is more important than chasing the winners.
PS: Here are the statistics since the year 2000:
Conclusion
So what can we conclude from the findings for our global single stock momentum ranking?
Fixed equal-weight allocation to each region is a good starting point if you want a simple, low-brain-damage solution. Other over-fitted fixed allocation themes are probably not worth it. If your universe is limited to EU + US or even US-only that might expose you to a “Japanification” risk if a single continent might experience a similar path in the future. However, most likely it won’t hurt too much since difference between regional markets became smaller and smaller post 2000.
Letting a global single stock ranking decide freely about regional allocation won’t be catastrophic either (unless severe regional political risks occur). It might even boost your returns, even if allocation reaches extremes (100% in one region), but it will also increase volatility.
A better approach would be excluding the loser region, which should happen more or less automatically if you rank globally. Of course, actively checking for momentum of niche corner portfolios for each region to exclude a loser means additional work (and sources of error).
The ExLoser-OverweightWinner approach (although not adding much juice) might show us a viable alternative. One interpretation of this approach could be that we can do the following:
let regional allocation ride freely without minimum allocations, so that losers can be excluded
but introduce a maximum allocation cap to each region (for example 50%) to avoid excessive concentration and volatility
In this way, we don’t force allocation to losers but we force diversification if the portfolio gets too concentrated. However, even this approach lost much appeal after 2000 vs. an equal-weight allocation.
Of course, the shown models and backtests can only be proxies for factor-based investing in a real-time, concentrated, single stock portfolio with much higher idiosyncratic risk, tax considerations, varying trading costs, etc.
Nonetheless, this small exercise helped me a lot to clarify the regional allocation question and I hope it helped you as well.
Next time I will repeat the investigation with regional Smallcap-HighB/M portfolios.
Stay tuned!