Thanks to technological advances, calculations that would have taken computers 7 years in 1991 can be accomplished in less than 1 second today. With such a rapid pace of technological advances, it’s no wonder that tech stocks have provided such amazing returns for investors.
That being said, technology stocks can be difficult to evaluate. For instance, many technology stocks either have negative earnings or trade at sky-high P/E ratios. Evidence suggests metrics conventionally used to evaluate stocks may not be applicable to technology stocks. As such, we performed an extensive analysis to uncover the best metrics to evaluate technology stocks.
In our analysis, we explored the association between over a hundred financial metrics and historical stock returns in the technology sector using a powerful machine learning technique known as bootstrapped regularized lasso regression. In this article, we present the 7 best metrics to evaluate technology stocks that we uncovered during our analyses.
We include this section for readers interested in how the data is derived. For the results of our analysis, skip to the next section.
For this analysis, we used over a decade of historical financial, market and price data. To derive the best metrics to evaluate technology stocks, we used bootstrapped regularized lasso regression to determine the linear association between over a hundred financial metrics and a binary indicator of strong annual returns relative to the overall market. The S&P 500 index was used as a surrogate marker of overall market performance. Furthermore, to normalize the distributions of financial metrics with heavily skewed distributions, we used a log transformation, and all covariates were standardized.
Bootstrapped regularized lasso regression is an excellent technique to gain inference from data. In fact, it is one of the only machine-learning techniques that allows you to gain an understanding of how the model works. Without delving into the mathematics, regularized lasso regression introduces a regularization term (highlighted in yellow in the equation below) to its cost function, which both improves the accuracy of the model and eliminates less important financial metrics from the model.
Furthermore, by bootstrapping regularized lasso regression, we were able to train a thousand models from our original dataset and to set our final model to the average across all trained models. Finally, we restricted our analysis to financial metrics identified as important in >80% of trained models. This process ensured that our findings had increased reliability and better interpretability.
The key findings of our analysis are summarized in table 1 above. We focus on the direction and magnitude of associations between metrics and predicted future annual returns. This is due to the fact that the transformation and standardization of data during the data preprocessing phase of our analysis limits the interpretability of coefficients derived by our models.
An increase in metrics with a positive direction of association is linked to an increased probability of strong future annual stock returns. The opposite is true of stocks with a negative direction of association. For instance, in this case, stocks with a higher price-to-sales ratio are expected to have higher annual returns whereas stocks with higher growth in R&D (research and development) costs are expected to have lower annual returns.
On the other hand, the magnitude of the association describes the importance of a metric determined by our model. As shown in table 1, the price-to-sales ratio is considered to be the most influential metric in determining the expected returns of tech stocks.
The P/S ratio is a measure of a stock’s valuation. Based on our analysis, there is a positive association between the P/S ratio and future annual stock performance. Furthermore, this association was found to be highly important.
But, wait! Doesn’t a higher P/S ratio indicate a stock is more expensive?
At face value, this assertion is correct. And perhaps, there is some truth to this statement; tech stocks with a higher valuation may offer a higher potential for future returns than those at a lower valuation. Nevertheless, we believe that this is only part of the story.
Consider two stocks with similar P/E ratios, but differing P/S ratios:
It follows that Stock A has higher sales than Stock B. Given the stocks have similar earnings, Stock A must-have lower expenses (i.e., a higher profit margin or lower SG&A expense to revenues) in order to achieve similar earnings. As such, when considered alongside other metrics, a high P/S ratio may serve as an indirect indicator of higher profit margins.
It is also important to note that the P/S ratio was considered a more important metric than the P/E ratio. We associate three potential explanations to this finding:
Our findings indicate a significant positive association between the beta of tech stocks and future annual stock returns. Beta is a measure of a stock’s volatility with higher betas correlating to higher volatility. Since higher beta stocks are inherently riskier investments, they typically provide a higher potential for future stock returns.
We detected a moderate positive association between net income growth and future annual returns. More specifically, net income growth occurring within the two quarters was most strongly associated with future stock performance. As such, you should consider recent EPS growth as an important marker of stock growth when evaluating tech stocks.
Similar to net income growth, our algorithms detected a moderate positive association between free cash flow growth and future annual returns. In this case, free cash flow growth occurring in the most recent quarter had the strongest association with future stock performance.
Based on our analysis, tech stocks with higher market caps were more likely to deliver strong future performance relative to smaller peers. This suggests that investing in larger technology companies may have a more favorable risk-reward profile.
A moderate negative association between R&D expense growth and future annual earnings was detected. At first glance, this association may appear counterintuitive. After all, increased R&D expenditures should translate to future competitive advantages.
It is important to note, however, that potential future advantages of R&D come at the cost of short-term earnings, which may ultimately impact the valuation of a stock. Secondly, our algorithms were trained to detect strong future annual performance. Perhaps, it is the case that investments in R&D may take more than a year to be realized. Thirdly, increased R&D expenses may be a surrogate marker for changes in the marketplace. For instance, the entry of new competitors may prompt a company to increase R&D spending.
In any case, our analysis suggests that R&D expense growth is negatively associated with 1-year stock performance adjusting for other factors.
Selling, general and administrative (SG&A) expenses include expenses that are not directly related to the costs of producing a good or service. We detected a moderate negative association between SG&A expenses to revenue.
While SG&A expenses to revenue is certainly not a common measure, it does share similarities to profit metrics such as profit margin. That being said, why is the SG&A expenses to revenue ratio a superior metric as compared to conventional indicators of profit?
We believe this is due to the fact that SG&A expenses to revenue may be more comparable across companies. For instance, we know that providers of luxury goods operate at higher profit margins than bulk merchants. Nevertheless, both business models are viable leading to the conclusion that the profit margin may be a poor marker for future business performance. The SG&A expense to revenue ratio avoids this pitfall making it a superior profitability marker.
In this article, you learned the best 7 metrics to identify technology stocks today. Those 7 metrics are: