In this blog post, I will share with you a project I did with my team, where we tried to answer whether the R&D to market cap ratio, which measures how much a company invests in Research and Development per dollar of its market value, can be a signal of how successful a company’s R&D efforts are.
Our hypothesis was that when a company’s R&D projects are going well, it is more likely to disclose its high R&D expenses. On the other hand, when a company’s R&D projects are not going well, it may try to hide its R&D expenses in other categories, such as general and administrative expenses, to avoid revealing its failure and losing its market value.
Therefore, we expected to see a positive relationship between the R&D to market cap ratio and the risk-adjusted returns of the companies. We also wanted to see if we can exploit this relationship to create a profitable investment strategy.
In all periods, we created six portfolios. These portfolios were a Non R&D portfolio, and five
sorted R&D portfolios. The portfolios were reconstituted in April of every year with rebalancing
every month.
Although equally weighted would be ideal, to verify if our results were not being biased because
of negative autocorrelations in stocks causing higher returns we also tried value-weighted
portfolios.
Finally, to make sure that big stocks were not dominating the portfolios in value-weighted
rebalancing, we also tried value weighting after removing the top 1000 firms based on the
market cap during reconstitutions.
All returns are monthly and in percentage points
Equally weighted portfolios (Excess returns)
Period | Non R&D | L | 2 | 3 | 4 | H |
---|---|---|---|---|---|---|
1981-2012 | 0.81 | 0.88 | 1.17 | 1.81 | 1.86 | 1.83 |
1981-1999 | 0.51 | 0.46 | 0.61 | 0.85 | 0.98 | 0.98 |
2000-2012 | 1.05 | 0.78 | 0.82 | 1.45 | 1.18 | 1.44 |
Value weighted portfolios (Excess returns)
Period | Non R&D | L | 2 | 3 | 4 | H |
---|---|---|---|---|---|---|
1981-2012 | 0.54 | 0.55 | 0.64 | 0.92 | 0.92 | 0.87 |
1981-1999 | 0.57 | 0.58 | 0.82 | 0.74 | 0.82 | 0.77 |
2000-2012 | 0.46 | 0.21 | 0.06 | 0.61 | 0.87 | 0.36 |
Value weighted portfolios after removing top 1000 firms (Excess returns)
Period | Non R&D | L | 2 | 3 | 4 | H |
---|---|---|---|---|---|---|
1981-2012 | 0.74 | 0.44 | 0.64 | 0.88 | 1.07 | 1.29 |
1981-1999 | 0.37 | 0.07 | 0.24 | 0.41 | 0.47 | 0.47 |
2000-2012 | 1.13 | 0.70 | 0.58 | 0.92 | 0.90 | 1.47 |
Value weighted extended period (Excess returns)
Period | Non R&D | L | 2 | 3 | 4 | H |
---|---|---|---|---|---|---|
1981-2021 | 0.75 | 0.91 | 1.36 | 1.90 | 1.88 | 1.78 |
The results of the analysis show that the equally weighted returns have a monotonically increasing trend. The returns for the period of 2000-2012 are higher compared to the other periods. On the other hand, the value-weighted returns exhibit an upward trend, but it is not as clear and consistent as the equal weighting. However, when the top 1000 firms are excluded, the upward trend becomes more prominent. The uptrend also remains consistent when the analysis is extended to the period of 1981-2021
Equally weighted portfolios (High-Low)
Period | Three factor alpha | Three factor t-stat | CAPM alpha | CAPM t-stat | Annualized Sharpe Ratio |
---|---|---|---|---|---|
1981-2012 | 0.84 | 3.07 | 0.89 | 3.32 | 0.77 |
1981-1999 | 0.52 | 2.42 | 0.55 | 2.61 | 0.72 |
2000-2012 | 0.51 | 1.61 | 0.63 | 2.05 | 0.72 |
Value weighted portfolios (High-Low)
Period | Three factor alpha | Three factor t-stat | CAPM alpha | CAPM t-stat | Annualized Sharpe Ratio |
---|---|---|---|---|---|
1981-2012 | 0.19 | 0.63 | 0.24 | 0.82 | 0.23 |
1981-1999 | 0.14 | 0.46 | 0.17 | 0.56 | 0.17 |
2000-2012 | 0.05 | 0.12 | 0.11 | 0.28 | 0.12 |
Value weighted excluding the top 1000 portfolios (High-Low)
Period | Three factor alpha | Three factor t-stat | CAPM alpha | CAPM t-stat | Annualized Sharpe Ratio |
---|---|---|---|---|---|
1981-2012 | 0.72 | 2.86 | 0.80 | 3.20 | 0.74 |
1981-1999 | 0.44 | 1.95 | 0.44 | 1.97 | 0.53 |
2000-2012 | 0.50 | 1.50 | 0.73 | 2.22 | 0.78 |
Value weighted portfolios (High-Low) extended period
Period | Three factor alpha | Three factor t-stat | CAPM alpha | CAPM t-stat | Annualized Sharpe Ratio |
---|---|---|---|---|---|
1981-2021 | 0.78 | 3.15 | 0.78 | 3.17 | 0.66 |
The results show that the equally weighted portfolios and value-weighted portfolios excluding the top 1000 firms generated significant three-factor and CAPM alphas. However, when all firms were taken into account, the value-weighted strategy did not produce any significant non-zero alphas. This suggests that the value-weighted strategy may not be as effective in generating positive alpha returns when considering the entire population of firms.
It should be relatively easy to use the above findings to develop a trading strategy. Our results note that equally weighted portfolios of high R&D stocks perform the best, but they can be difficult to implement in practice due to the constant rebalancing required. Thus, we would suggest using value weighting but with the top 1000 stocks excluded since that also has significant non-zero alphas in the Long-short portfolios formed.
In future research, we would like to improve our estimation of missing returns in the data, as we currently assume them to be zero. We believe this could potentially explain the non-monotonic and inconsistent results observed in the value-weighted backtests. Additionally, before implementing such a strategy, it is important to consider trading costs which we assume to be zero in our backtests.