Research · Master's thesis
A Formula for Investing in the Semiconductor Industry
An Empirical Study Using Firm Fundamentals and System GMM
Abstract
This thesis investigates whether firm-specific financial metrics can predict stock returns in the semiconductor industry and whether these predictors can form the basis of a systematic investment strategy. Using a panel of fifty global semiconductor firms from 2006 to 2022, the analysis applies a two-step System Generalized Method of Moments estimator to address endogeneity and firm heterogeneity. Return on assets emerges as the most robust and statistically significant predictor of future stock returns, reflecting the central role of asset efficiency in this capital-intensive sector.
Portfolio tests compare buy-and-hold and annually rebalanced strategies against the NASDAQ benchmark. The results show that ROA-based portfolios achieve superior raw performance, particularly when rebalanced annually, although excess returns largely reflect higher market exposure once risk adjustment is applied. Within the framework of the Adaptive Market Hypothesis, the findings suggest that profitability-based strategies may occasionally capture time-varying inefficiencies, but their effectiveness depends on evolving market conditions.
The results overall show that, in a semiconductor-specific dynamic panel that accounts for unobserved firm effects and endogeneity, return on assets is the only robust fundamental predictor of stock returns. This motivates an investment formula that can narrow the set of firms investors need to evaluate when making investment decisions in this sector.
Key findings
- Return on assets is the one fundamental that predicts returns robustly. Estimated with System GMM, its coefficient of 0.429 is significant at the 0.1% level and passes the tests of the instruments and of autocorrelation. Return on equity, return on capital employed, operating margin, leverage, R&D intensity and revenue growth each fail a test or are not significant.
- An equally weighted portfolio of the five firms with the highest return on assets, capped at 30% and rebalanced each year, returned 27.12% a year from April 2018 to April 2024, against 19.20% for the NASDAQ and 26.78% for the S&P Semiconductor index.
- Updating the portfolio matters: the same five firms bought in 2018 and held for six years returned 24.76% a year. With the cap at 50% instead, the rebalanced portfolio returned 19.81%, consistent with very high returns on assets reflecting temporary conditions rather than lasting efficiency.
- Adjusted for risk, the outperformance is explained by market exposure: the portfolio's beta against the NASDAQ is 1.117, and its alpha of 0.0063, about 0.63% a month, is not statistically significant.
Returns, April 2018 to April 2024
| Portfolio | A year | In total |
|---|---|---|
| Return on assets, capped at 30%, rebalanced yearly | 27.12% | 335.35% |
| The same screen, bought once and held | 24.76% | 289.89% |
| Return on assets, capped at 50%, rebalanced yearly | 19.81% | 204.22% |
| S&P Semiconductor index | 26.78% | 334.84% |
| NASDAQ | 19.20% | 186.24% |
About the study
- Sample
- Fifty listed semiconductor firms in North America, Europe and Asia: chip designers, manufacturers, foundries and equipment makers (GICS 4530)
- Data
- About 850 annual reports, 2006 to 2022, from Refinitiv Eikon
- Method
- Two-step System GMM with year effects and Windmeijer-corrected standard errors; the Capital Asset Pricing Model for risk adjustment
- Portfolio test
- April 2018 to April 2024, against the NASDAQ and the S&P Semiconductor index
- Paper
- Master's thesis in economics, 15 credits, spring 2025; 51 pages
An academic paper, published for information. Its portfolios are historical backtests, not recommendations to buy or sell any security, and nothing here is investment advice. Past returns are no guarantee of future returns. Balk Investmentbolag and its co-owners may hold shares in companies named in the paper.