Bloomberg Professional Services
In quantitative trading, information on the point-in-time revenue segments a business participates in is traditionally used to map relationships across the equity landscape which can be used to augment established trading strategies.Â
What these strategies might sometimes overlook is the inherent value in the data as a standalone product. This research report posits and backtests an easy-to-implement long-short investment thesis to show that revenue segmentation data has orthogonal information content on its own.
Strategy and summary results
The strategy applied in this report involves investing in businesses that are highly concentrated in a particular revenue segment and selling firms that participate marginally. On each day, we ranked all the firms in our universe by the percentage of revenue they derived from a particular sector, buying the top quantile and selling the bottom.
According to our research, the strategy generates positive returns across seven segments (see chart 5). Of these, the best performing is Advertising & Marketing. We normalized these returns against common market factors to verify that there is truly orthogonal information content in the data. Accounting for growth, momentum, size and value, four sectors still generated positive annualized returns (chart 7). These were Beverages, Construction Materials, and Electricity & Gas Marketing & Trading. Concentration in these sectors can lead to outperformance.
Company Segment Fundamentals
Bloomberg’s Company Segment Fundamentals – Industry Standardized product details Revenue, Total Balance Sheet Assets, Net Revenue and Operating Income for each reported segment a company is involved in is based on the standardized Bloomberg Industry Classification System. The dataset has history back to 2006 and covers 53,000 firms.
Chart 1 shows an excerpt of the data with select fields for identification (id_bb_global, long_comp_name and fundamentals_ticker), point-in-time history (fiscal_year_period) and values to create a trading signal (sales_rev_turn or is_percentage_of_revenue). Chart 2 visualizes the evolution of is_percentage_of_revenue back to 2016Â
As shown, many firms derive relatively stable revenues from their Advertising & Marketing segments. Converting this field into a trading signal is a straightforward process since it is expressed as a percentage.
Backtest details
Understanding all the revenue segments a company participates in is useful for a few reasons, these include: universe identification for investing in a specific sector or trend, understanding the competitive landscape or isolating growth drivers within a firms business.
These use cases are particularly important for fundamental analysis. But what about more quantitative approaches? To this end, we ran systematic backtests that attempt to unearth uncorrelated market returns that are native to the dataset. The details are as follows:
Thesis: Companies that are more highly concentrated in a sector outperform those that participate but are less concentrated in terms of revenue
Factor: IS_PERCENTAGE_OF_REVENUE from each sector
Time Range: 2016-2023
Universe: Russell 2000
Rebalance: Quarterly
In each calendar quarter, we split the companies in our universe into quantiles based on the IS_PERCENTAGE_OF_REVENUE they derive from the segment in question. If a company has high revenue from the sector, it goes into the top quantile and vice versa. Within each quantile, we are equiweighted and long-only. An example of this, wth five quantiles, is shown in chart 3.
The final output below is a 50:50 long-short based on going long on the quantile with the highest signal and short on the quantile with the lowest signal. In chart 3, the long would be on 05 Very High and the short on 01 Very Low. Bear in mind that some segments have less than 5 quantiles because there isn’t enough diversity of signal to split into 5 evenly sized buckets.
The results for each segment are shown in chart 4. With most segments generating sizable absolute annualized returns (>10%), we can conclude that there is orthogonal information content in this dataset. The goal with this type of analysis is to derive a differentiable signal, the fact that most returns are negative does not matter as much since we can simply flip our thesis and therefore our long-short.
For the purposes of this report, we will focus on the sectors that generated positive returns as we want to better understand the benefits (or lack thereof) of revenue concentration within a segment.
Segments with positive returns
Chart 5 shows the best segments for concentration.. As these industries tend to be capital intensive (high CapEx),, concentration is required to achieve the economies of scale necessary for high performance. Note that a full investigation of this finding is beyond the scope of this report.
Chart 6 shows the cumulative returns of the long-short strategy over time for each of the sectors shown in chart 5. Construction Materials and Electricity & Gas Marketing & Trading show similar returns as the S&P 500 over this period, Advertising & Marketing far outstrips the index.
Factor exposure analysis
On the whole, returns are positive. But how can we be sure that these profiles are a direct result of the signal we chose and not because of high factor exposure to something like momentum or growth? To verify that our strategy cannot be explained by exposure to common market factors, we normalized against momentum, growth, value and size.Â
Chart 7 shows the final results. In sum, the four sectors below all generated positive annualized returns even after accounting for these common market factors.
Conclusion
As demonstrated, point-in-time revenue segmentation data can provide orthogonal information content. This is an additional benefit to the more common use cases of fundamental business analysis and building knowledge graphs. This dataset is available for trial through data.bloomberg.com.