Heuristic (rule-based) based equity trading strategies have existed for well over 50 years. Ben Graham’s value investing approach was by itself to some extent heuristic-based (low PE ratio, low debt to current assets ratio, positive EPS growth etc.). The advent of powerful computation tools and the availability of data has only accelerated research on these criteria-based trading models. Large hedge funds like Renaissance Technologies generate several terabytes of data each day and run advanced quantitative models to select stocks.
One such approach is the systematic sector-rotation strategy published by two Italian academics – Paolo Sassetti and Massimiliano Tani. The paper was originally published in 2003 and has been quoted in various subsequent research on this subject. At its heart, the ideas comprised in it are quite simple.
- At different points in the economic cycle, different sectors tend to perform well. Financials and consumer discretionary perform well during the early stages of the business cycle as the availability of cheap credit gets firms and consumers borrowing. As the economic cycle improves further and more businesses start making capital investments, capital goods, industrial and technology companies tend to do well. As the economy starts contracting again, consumer staples and healthcare stocks tend to outperform other stocks. This trend has been quite well documented notably by Fidelity research.
- The sectors that have performed well in the recent past, will continue to perform well in the ‘recent’ future. This is at the crux of momentum-based strategies.
Sassetti and Tani ran a backtesting algorithm to test their hypothesis that a systematic-sector rotation strategy based on momentum indicators will beat the benchmark over the long run. Lo and behold, they found evidence of significant outperformance. They tested their hypothesis using three different approaches. This article will only cover the first approach called Rate of Change. The other two approaches are available for you to read on their paper right here.
Approach 1 – Rate of Change
Here is how this works.
- You wake up one fine morning and decide that you want to find the best ‘sectors’ to invest in.
- You go into BSE’s website and find out that the BSE classifies all the stocks listed into 19 different sectors. These include healthcare, telecom, utilities, energy etc. The composition of each of these sectoral indices is those companies that fall into one of these sectors/categories. The BSE Basic Materials Sectoral Index for instance has cement companies, chemical companies, fertilizer makers, some steel firms amongst others. There are about 140 of them in total that fall into this category.
- You download the historical pricing (index) data for each of these 19 different sectoral indices from BSE’s website from May 2011 to October 2020 and put them all in an excel spreadsheet. This is what the first few rows of the spreadsheet look like.
The first column has the dates while the rest of the columns have different sector index levels(prices. The top row has the sector codes (SI900 refers to BSE Auto, SIBANK refers to BSE Bankex etc.)
4. You calculate the 30-day returns on each of these sectors and find the top sectors which have the highest returns over the past 30 days.
5. You pick the sector with the highest 30 days return and invest in it. Then you pick the sectors that are top 2 on that list and invest in them. Then you pick the sectors that are top 3 on that list and invest in them. And so on till you pick all the sectors on that list and invest in them. The key question is, what is the optimal number of sectors to choose from the list to invest in? Given that momentum-based trading strategies suggest persistence of stock/sector returns, you expect that you’d only have to pick a few of the top sectors with the highest 30-day historic returns in order to beat the return on the benchmark (in this case, the benchmark return is the return you could obtain from investing in all the sectors)
6. You sell these holdings 30-days later (a 30-days holding period) and start from step 4 again. You do this for 9 years from 2011 to 2020.
7. You repeat steps 4, 5 & 6 by calculating 60-days returns (with 60 days holding period) and 90-days returns (with 90 days holding period) to figure out the ideal time-period to calculate historic returns.
This is more or less the approach that Sasseti and Tani followed in part-1 of their paper to validate this trading strategy. Here are their results.
The way to interpret the above table is this – If the model had picked the top 10 funds based on 30-day historic returns, invested in these and held on to these funds for 30 days before selling them and doing the next trade, the overall return over a period of 5 years (1998-2003) would have been 52% against the benchmark return of 37%. Kapito?
Their model made the highest returns (215%) when the 60-day returns window was used and the top 3-funds/sectors were chosen for investment for a period of 30 days. Note – They used a minimum holding period of 30 days and their asset base was 42 different sectoral funds under management with Fidelity. However, the model that we discussed above uses 30-, 60- and 90-days holding period for 30, 60 and 90-days historic return windows respectively. The asset-base on our model is the sectoral indices on BSE. So, results are not strictly comparable
Here are our results on a similar table as produced above.
Returns computed on this table are average arithmetic mean yearly return based on all trades.
Our model also made the highest returns when the 60-day historic returns window was used, but we only needed to pick the top 2 sectors. The benchmark return is about 7% during this period as you can see at the bottom of the table and the best return highlighted in yellow (60 days, 2 sectors) is more than double the benchmark returns. Essentially, what it is saying to us is that if we had computed the 60-day historic returns, picked the top 2 sectors from that list, invested in them for a period of 60 days and repeated this process over & over from 2011 to 2020, we would have made 14.42% in average annual returns over the last decade – way over the 7% that all the BSE sectors combined delivered.
The table below shows the % of profitable trades with each strategy. The distribution on this table is slightly different from the one above. This table suggests that over 69% of trades were profitable (the highest) when we used a 60-day window and picked the top 4-sectors. But once again, the number of sectors required to choose is less than a quarter of the overall number of sectors (19).
Finally, the table below shows return per unit risk (return/standard deviation). Higher ratios indicate better strategy performance.
This table once again drives home the same point – The fewer the sectors chosen, the better the performance.
So, what does this tell us about where to invest in TODAY?
Well, we cannot rely on this single approach to guide our asset allocation process entirely. Past returns and trading models offer no guarantee that the future return profile will be similar. But with the data we have collected, the top 3 sectors in India that are flashing green TODAY based on the results of this model are BSE Basic Materials, BSE Industrials and BSE Power.
We are working on extending this model to test, from an Indian context, the other two approaches that Sasseti and Tani have published in their paper. A powerful heuristic-based trading model should incorporate several signals based on a variety of data points & approaches. The signals from this model should be incorporated with other signals generated from structured and unstructured data. Is the model singing a similar song to what the financial newspapers are singing? Is the retail public in on it? What does the Reddit population say about the sectors in vogue? How does that correlate with fundamentals data, in terms of the EPS growth rate of the industry? At this time, we have more questions than we have answers for.