The Magnificent Seven (Mag 7): Strategic Evolution and Distance-to-Frontier Dataset Suite (2010-2022)
Description
This dataset provides the cleaned panel data and MATLAB replication code for evaluating the structural convergence and multi-market diversification of major technology conglomerates. The data map the technological footprint of the 'Magnificent Seven' cohorts relative to the broader IT sector baseline. Revision of the MATLAB code: Some bugs are corrected. (9 September 2026 Files Included 1. Readme.pdf 2. mag7_longitudinal_subclass_GH.csv: The base technological subclass panel dataset restricted to Cooperative Patent Classification (CPC) sections G (Physics) and H (Electricity). This file is utilized for the core Hill number calculations and the baseline distance-to-frontier analysis. 3. mag7_longitudinal_subclass_th66.csv: The threshold-filtered panel dataset incorporating patents across all CPC sections. It includes data for the Magnificent Seven firms and a baseline 'Rest of the IT sector' comprised of firms where sections G and H constitute at least 66% of their total patent portfolios within the analysis period. This dataset is used exclusively for the Jensen-Shannon Divergence (JSD) robustness checks. 4. Replication.m: The master MATLAB script that automatically loads the datasets, executes the analytical loops, and generates the output figures. Data Source & Temporal Windows The underlying raw data were obtained from PatentsView, last updated on December 9, 2025. While the raw data files span from 2010 through 2024 to capture absolute historical allocation distributions, the empirical analysis window is intentionally evaluated up to the year 2022 to control for standard right-truncation biases introduced by patent approval time lags.
Files
Steps to reproduce
To replicate the results, place the master MATLAB script and both CSV datasets in the same directory and execute the script. The suite will process the data in the background and generate eight vector PDF figures. In accordance with the journal's submission guidelines, the main text figures and individual appendix figures follow separate numbering conventions: Main Text Figures (Core $q=2$ Specifications) Figure 1: Plots average portfolio diversification against cross-firm strategic heterogeneity using the Inverse Simpson limit on G/H data. Figure 2: Plots effective firms per domain against cross-domain heterogeneity in the transposed matrix space. Appendix A Figures (Hill Diversity $q=1$ Robustness) Figure A.1: Plots the exponential Shannon entropy variant tracking primal space diversification. Figure A.2: Plots the exponential Shannon entropy variant tracking transposed space crowding. Appendix B Figures (Jensen-Shannon Divergence Analysis Suite) Figure B.1 (Baseline): Plots the standard aggregate $\overline{JSD}_t$ using base G/H data without robustness filters. Figure B.2 (Threshold Data): Plots the aggregate $\overline{JSD}_t$ applied to the comprehensive threshold-filtered dataset. Figure B.3 (Top-100 Subclasses): Restricts the threshold dataset to the top 100 most heavily patented technological subclasses. Figure B.4 (Joint Robustness): Appears under the strictest specification, combining the Top-100 restriction, incidence-based binarisation, and equal firm weighting.
Institutions
- The University of TokyoTokyo, Tokyo