Table of Contents
In investment practice, industry rotation has long been a key basis for capital allocation and stock selection. As the economic cycle progresses, market capital often shifts gradually from leading industries to lagging ones, forming a structural sector rotation. For instance, during the early stage of economic recovery, cyclical sectors such as shipping and steel often react first. As the economy continues to expand and corporate capital expenditures increase, growth-oriented sectors like semiconductors and technology stocks may take the lead.
Although this type of rotation logic is widely referenced in macroeconomic analysis, accurately identifying its starting and ending points still requires the support of quantitative tools and empirical data. This article focuses on historical data from various industry indices and utilizes statistical tools such as Transfer Entropy and rolling correlation coefficients to examine whether stable lead-lag relationships exist between different sectors. It also explores the potential role of economic indicators, such as the business cycle signal lights, in these inter-sector dynamics.
Taiwan’s economic structure is heavily reliant on technology manufacturing and export-oriented industries, with the semiconductor sector at its core. According to data from the Ministry of Economic Affairs, Taiwan holds a leading global market share in semiconductors, encompassing a complete supply chain from upstream IC design, to midstream wafer fabrication, and downstream packaging and testing. The semiconductor industry serves as a major driver of Taiwan’s GDP and exports, making its business cycle and capital flows a key indicator not only for the domestic stock market but also for the broader economy.
However, across different phases of the economic cycle, market capital does not always concentrate in a single sector. Instead, it rotates among various segments depending on prevailing macroeconomic conditions. This rotation may be influenced by multiple factors, such as fluctuations in global trade, inflationary pressure, interest rate policies, and capital expenditure cycles. Therefore, understanding the relative dynamics between the semiconductor sector and other industries can help investors make more forward-looking decisions in sector allocation.
In light of this, the semiconductor industry is selected as the core observation target in this study. We further examine whether it exhibits clear lead-lag or rotational relationships with other sectors—such as finance and insurance, shipping, and biotechnology—and attempt to construct a practical framework for predicting sector rotation.
For the purpose of conducting empirical analysis on industry rotation, this study utilizes historical industry index data provided by TQuant Lab. The platform covers major sector classifications in the Taiwan stock market, including semiconductors, electronics, finance & insurance, shipping, and biotechnology. It offers long-term, daily historical data, making it well-suited for analyzing rotation behavior over time.
Since industry indices are price-based and typically show a gradual upward trend over the long run, they do not fluctuate within a fixed range. As a result, directly analyzing raw price data makes it difficult to detect relative strength changes across sectors. To address this issue, the study applies the Relative Strength Index (RSI) to each industry index. RSI, a momentum-based indicator, reflects the proportion of gains versus losses over a given time window, allowing us to identify periods of relative strength or weakness among sectors.
After computing the RSI time series for each sector, the following three analyses are performed:
Through these methods, the study aims to quantify and verify whether time-shifted rotational logic exists among industries, thereby providing practical insights for investment strategy decisions. The analysis uses data from early 2012 to the end of 2019 for model development, while the period from early 2020 onward is reserved for strategy backtesting to avoid overfitting.
We can clearly observe that complex rotation patterns do exist among the various industries, and the RSI values appear to exhibit a cyclical lead-lag relationship. However, such visual observations alone are insufficient to confirm whether these rotational dynamics are statistically valid or meaningful. Therefore, we proceed with the aforementioned statistical analysis methods to further investigate and identify reliable patterns of industry rotation.
This chart uses a 60-day rolling correlation coefficient to examine changes in the relative co-movement among industries. From a time series perspective, the electronics sector shows a consistently high level of synchronization with semiconductors (red line), indicating a structurally linked relationship. In contrast, the correlation between the shipping sector and semiconductors exhibits cyclical fluctuations, suggesting a possible time-lagged rotation pattern between the two. The financial and biotechnology sectors display more inverse or unstable correlation patterns, supporting their classification as asynchronous or defensive assets.
In summary, the performance of the shipping, financial, and biotechnology sectors may reflect potential rotational relationships with the semiconductor sector, offering practical insights for sector rotation analysis.
The chart illustrates the changes in Pearson rolling correlation coefficients between each industry and the semiconductor index within a ±60-day lag window, aiming to uncover potential lead-lag relationships across sectors.
Notably, the correlation between the shipping sector index and the semiconductor index peaks at approximately +30 days (purple line). This suggests that when the shipping sector demonstrates relative strength, the semiconductor sector tends to follow with a noticeable upward movement about 30 days later. This structural delay supports the sector rotation hypothesis of “Shipping → Semiconductors.”
In contrast, for the financial and insurance sector, the correlation (blue line) reaches its highest point at around -50 days, indicating that strong performance in semiconductors is typically followed by a notable rise in the financial sector approximately 50 days later. This supports the opposite directional hypothesis of “Semiconductors → Finance.”
Meanwhile, other sectors—such as electronics and biotechnology & healthcare—exhibit more symmetrical or flat correlation curves, lacking clear lagged peaks. This implies that their relationship with the semiconductor sector is more synchronous or non-rotational in nature.
Transfer Entropy (TE) is a metric used to measure the directional strength of information transfer between two time series. Unlike traditional correlation, TE captures non-linear and asymmetric relationships. In simple terms, if the TE value from Industry A → Industry B is greater than that from Industry B → Industry A, it can be inferred that changes in A have stronger predictive power over B, indicating a directional flow of information from A to B.
In the following table, we compare TE values under various lag settings to analyze the consistency of directional information flow. Only when a consistent direction is observed across multiple lag periods do we consider the result to be statistically reliable and indicative of a meaningful sector rotation pattern.
Shipping Sector vs. Semiconductor Sector
| Lag Period (Trading Days) | TE: Shipping → Semiconductor | TE: Semiconductor → Shipping | Information Flow Direction |
| 3 | 0.0177 | 0.0155 | Shipping → Semiconductor |
| 5 | 0.0539 | 0.0376 | Shipping → Semiconductor |
| 7 | 0.0866 | 0.0589 | Shipping → Semiconductor |
Finance and Insurance Sector vs. Semiconductor Sector
| Lag Period | TE: Finance → Semiconductor | TE: Semiconductor → Finance | Information Flow Direction |
| 3 | 0.0188 | 0.0164 | Finance → Semiconductor |
| 5 | 0.0469 | 0.0538 | Semiconductor → Finance |
| 7 | 0.078 | 0.0811 | Semiconductor → Finance |
Biotechnology Sector vs. Semiconductor Sector
| Lag Period | TE: Biotech → Semiconductor | TE: Semiconductor → Finance | Information Flow Direction |
| 3 | 0.0139 | 0.0169 | Semiconductor → Biotech |
| 5 | 0.0487 | 0.0333 | Biotech → Semiconductor |
| 7 | 0.0779 | 0.0494 | Biotech → Semiconductor |
Based on the results of the Transfer Entropy (TE) analysis, this study finds that the information flow from the shipping sector to the semiconductor sector remains consistent across different lag periods (k = 3, 5, 7). Furthermore, the TE values increase steadily with longer lags, indicating strong directional stability and temporal consistency in the information transfer. This suggests that the shipping sector may serve as a relatively reliable leading indicator for the semiconductor sector.
In contrast, the finance & insurance and biotechnology & healthcare sectors exhibit unstable information flow directions with respect to the semiconductor sector. For instance, in the short term, the direction of information flow may appear as A → B, but in longer lag periods, it may reverse to B → A. This high sensitivity to parameter settings implies a lack of stability and consistency in directional influence, making it difficult to draw causal inferences or apply the results to predictive modeling.
Therefore, when using inter-sector information transfer as a basis for asset allocation or trading strategies, the relationship from shipping to semiconductors demonstrates relatively higher explanatory power and practical potential. On the other hand, the results for other sectors should be interpreted with caution to avoid overfitting or misleading conclusions.
Integrating the findings from all three analytical methods, we conclude that the shipping sector serves as a leading indicator for the semiconductor sector. This insight provides a foundation for designing forward-looking sector rotation strategies based on inter-industry dynamics.
We welcome all investors to reference this analysis. Moving forward, we will continue introducing how to use the TEJ database to build various indicators and backtest their performance. For readers interested in strategy development and backtesting, we recommend exploring the TQuant Lab plans. With access to a high-quality data platform, you can design and test trading strategies that best suit your individual investment approach.
Disclaimer: This analysis is for informational purposes only and does not constitute any investment advice or recommendation on specific financial products.
With TEJ’s assistance, you can access relevant information about major stock markets in Asia, such as securities market, financials data, enterprise operations, board of directors, sustainability data, etc., providing investors with timely and high-quality content. Additionally, TEJ offers advisory services to help solve problems in theoretical practice and financial management!
From Business Cycle Indicators to Asset Rotation: A Quantitative Strategy to Avoid Bear Markets
Michael Murphy’s Risk Assessment Rules for Investing in High-Tech Stocks
Charles Brandes’ Value Investing Principles : Building a Portfolio with a Margin of Safety
