Demonstrating A Python-based Cross-platform Novel Tool for Panel Data Analysis In Econometrics

Abstract

Background Panel data is prevalent in the contemporary economic landscape because of the digitalisation of academic and financial resources. However, advanced econometric models in panel data analysis are out of the reach of early-stage and under-funded researchers due to technical and institutional barriers. Panel Data Analysis requires special software, such as STATA and SPSS, which are considered expensive re- sources that are helpful for a competitive advantage in well-funded institutions and for senior researchers. While low-cost alternatives to these tools exist, they are often complex, tedious, and inaccessible. Even the proprietary tools are tedious and cumbersome for complex analyses since they are designed to perform one study at a time. Objectives This thesis aims to develop and demonstrate a free and open-source, cross-platform, novel Python-based tool for panel data analysis that automates the tedious proce- dure to make it efficient for senior researchers and removes any technical and cost barriers for early-stage researchers and institutions with limited resources, thereby empowering them to use Panel Data Analysis to employ advanced econometric models to generate higher quality insights from the panel data. Methodology The tool developed in this thesis can perform panel data analysis in two modes: a) Exploratory and b) Autonomous. In Exploratory mode, the tool can test hundreds of models, exploring all possible permutations of relationships between variables under scrutiny. This mode can identify relevant econometric models to investigate any meaningful relationship between the variables further. In Autonomous mode, the tool can perform descriptive analyses, model diagnostics, and endogeneity testing, and selection between fixed-effects, random-effects, instrumental, or dynamic models, and provide a step-by-step analysis report. Theoretical Framework The theoretical grounding of this thesis can be described in two parts: a) Econo- metric theories that dictate the workflow or the algorithm of the tool for the panel data analysis, and b) Innovation adoption theories that explain the dominance of current tools and conditions required for the acceptance of the novel tool, thereby, also directing the development of the tool. Econometric theories such as panel data modelling, endogeneity correction using instrumental variables, and statistical tests to test the validity of the analysis from the theoretical basis for this tool's functionality. Innovation adoption theories used in this thesis include the Resource-Based View (RBV), the Technology Acceptance Model (TAM), the Uni- fied Theory of Acceptance and Use of Technology (UTAUT), and the Diffusion of Innovation Theory (DIT). Results This tool demonstrates the execution of econometric analysis on two real-world panel datasets in a transparent and replicable manner. This tool automates the process of panel data analysis, independent of the platform, at a low cost, with enhanced usability and reduced barriers to use. Conclusion The thesis contributes a theoretically informed tool with enhanced functionality that democratizes access to advanced economic analysis of panel data while providing a platform to modify and extend the capabilities of open-source software and cross-platform capabilities for widespread institutional and individual adoption in academic discourse.