Executive Summary

I analyzed the relationship between R&D expense as percentage of sales and stock returns. I initially conducted an unconditional analysis on the Russell 1000, after which I conducted an analysis conditioned on a sector wise basis. I segmented the sample into four “fractiles”, stocks in the first fractile spent the most on R&D whilst the constituents of fractile 4 spent the least. The unconditional analysis failed to produce conclusive results (transaction costs were not accounted for). Furthermore, the volatility of returns when deploying the R&D factor is higher than other. Contrary to lore a robust R&D program does not decouple a firm from the macro economy, thus constituents of fractile 1 were as susceptible to contractions as the broader index.

The results of the conditional analysis were more encouraging. The R&D factor (spending) is not a super-dominant strategy nor does it stochastically dominate other well-known factors. However, it was a dominant strategy therefore a viable investment factor in specific sectors. This article will establish the groundwork for the analysis and provide you the reader with an eagle’s view on the feasibility of investing based on the R&D factor. Subsequent articles will cover the following sectors:

Energy (NYSEARCA: VDE), Healthcare (NYSEARCA: PJP) (NYSEARCA: XPH) (NYSEARCA: TECZ) Material and Processing (NYSEARCA: JHMT), Technology (NYSEARCA: XLK) (NYSEARCA: VGT) (NYSEARCA: XITK) (NYSEARCA: ROM), Producer Durables (NYSEARCA: XLY), Consumer Staples (NYSEARCA: XLP), and Utilities (NYSEARCA: XLU).

If the list contains a sector that you are intrigued by or invest in please select the follow button below my profile. Please also browse my author page in case the sector that piqued your interest has already been published.


I used data as far back as available, which in this case is 1996. I defined the R&D factor (independent variable) as the last twelve-month trailing R&D expense divided by sales over the same period. I opted to remove companies who did not disclose their R&D expenditure from the sample. The cumulative return statistics are derived for High Minus Low (F1-FN) portfolios. Where a F1 portfolio comprised of the highest R&D spenders during the period. In the event of a tie the firms would be moved to the lower quartile.

I used both single factor and multiple factor regressions on a rolling basis to assess the predictive power of R&D. The Y variable in all tests were the universe’s excess returns over the risk-free rate (90-day T-bill). To correct for different sector exposures, the independent variables were layered on top of sector groupings.

Methodology Summary


Market Weight


Russell 1000

Balancing and Reweighting Frequency

3 Month

Reporting Lag

45 Days

Regression Type

Rolling Multifactor

I used a 45-day lag to avoid look ahead bias. Given that firms have forty-five days to file earnings in the US, a 45 day lag would circumvent the impact of press releases that precede SEC filings and it would ensure that firms in the entire universe have reported their earnings. On aside some well-respected data sources omit impact of look ahead bias on their factors.

Key assumptions

The assumptions I used are as follows:

The acquisition of smaller firms is not a material substitute for internal R&D activity. Investments in R&D can, at a minimum, result in the creation of (new) economic advantage that matches the rate of rate of decay of a firm’s current economic advantage. Rate of decay of an accumulation of economic advantage can be estimated by managers. Managers prefer to report smoother earnings over time. Capital markets are constrained; managers can deploy R&D dollars more efficiently than investors can gain that exposure. R&D spend conveys an economic advantage that are independent of economic regimes (stable risk premia).

The strength of some my assumptions, especially the first and last, are hard to test but are necessary given time constraints. Note that assumption six allows one to conduct an unconditional test of my hypothesis

The second is necessary to test the hypothesis across a sample that spans over time and across industries. Note, that if the second and third assumptions are true, then the time taken to develop a product is not relevant to my analysis-I do not need to look at R&D spend over different time horizons for different sectors. Managers will adjust the spending on R&D so that it matches the rate at which new technology enters the market. The second assumption becomes tenuous if the firm only has one product. If this is the case, then R&D will not reflect the rate of decay of competitive advantage but rather the development cycle of new products. Given that all firms are publicly listed and follow US GAAP (R&D is mostly expensed than capitalized) I believe that the second is relatively weak.

Analysis by Sector

After evaluating R&D as a possible factor of returns with respect to the overall market (Russell 1000), I turned my gaze to different sectors and tested my hypothesis on a conditional basis. To this end I analyzed seven industries in total. The test was simply constructed; I analyzed the spread between fractile 1 and fractile 4 across different sectors.

Materials and Processing

I ran a Factset simulation on the Russell 1000 Materials and Processing index from 1996-2015. The detailed results are shown in the chart below. The analysis was done by breaking the R&D expense/sales into four fractiles, where the highest R&D/sales are at fractile 1, and the lowest at fractile 4. The Russell 1000 Materials and Processing includes companies that manufacture chemicals, construction materials, plastic, and related packaging products. For fractile 1, the results yield an insignificant positive alpha (t-stat=0.443), and a significant positive beta equals 0.8 (t-stat = 11.664).

Materials and Processing Fractiles’ Performance and Statistics

I then compared the fractiles’ performance to the benchmark and to each other. Note that all the fractiles have accumulated negative cumulative excess returns. The fractiles’ performance in order is the following: Fractile 4, Fractile 2, Fractile 1, and lastly Fractile 3. Thus, I believe that R&D expense/sales in an inconclusive factor in determining excess returns for companies in the materials and processing sector. In fact, a trading strategy that shorts fractile 4 and longs fractile 1 would have accumulated severe losses. Similar to the results in the energy sector, I believe that materials and processing industry is a mature industries that hasn’t experienced much innovation in the last two decades. For example, most of the products that we use today are products that the chemical industry pioneered in the first half of the 20th century, such as nylon and polyethylene. If you reward R&D spend in this sector, your portfolio too may be in secular decline.

Disclosure: I/we have no positions in any stocks mentioned, and no plans to initiate any positions within the next 72 hours.

I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.

Source link