Who drives policy change? A Novel Machine Learning Approach for Environmental Policy Analysis

Who affects political decision-making and policy development is a central puzzle for political scientists to solve. Politicians and interest groups are known to affect political decisions, but research on the conditions and to what extent these actors interact and negotiate agreements regarding policy stability and change is still very limited, so are conclusive studies on the independent influence of different types of actors in the policy process. In short, the purpose of the paper is to further contemporary theory and empirical approach on who initiates, drives, and determines when and how policies change. Recognizing the pressing need for policy innovations to meet current and future societal challenges, along with the gaps in current literature, the paper sets out to advance theories of the policy process and of democratic decision-making by better understanding and explaining policy change. While countries around the world prepare to undertake a major push towards a green transition to meet the goals of international agreements and national targets, sustainability transitions also spur critical issues, not the least concerning the extraction of natural resources, that give rise to intractable conflicts among actors and interests that need to be handled politically. To explain policy change, the paper combines state-of-the-art research within political science with novel methodological techniques from data science research and predictive algorithms. This innovative approach applies AI and machine learning methods on political science problems which allows for big data to be processed, analyzed, and interpreted. This is an important methodological contribution and has the potential to revolutionize the field of policy analysis and social science more generally. We apply these methods to analyze policy development over time within Swedish mining policy, a contested policy area permeated with trade-offs between conflicting political goals and societal values as well as mobilization among political actors and organized interests. We apply AI, machine learning, data mining, statistics, and visualization techniques to collect, process, analyze, visualize, and interpret results. Because decision-makers and organized interest groups can use different fora to change policy coupled with the fact that changes are often incremental, the empirical material is large. Our data collection focuses on the most important sources of policy positions, namely official documents, written statements of political decision-makers, agencies, NGOs, and industry representatives. We study mineral policy between the years of 1990 and 2023, this time frame ensures that periods of stability and of change are covered. Our approach combines inductive and deductive topic and tonality modeling. As policy innovations are critical for a sustainable societal transition, there is a pressing need for scholarship to explain who initiates, drives, and determines when and how policies change. While the study is situated in the natural resource governance sector, the findings will be applicable to policy dynamics in other adversarial policy areas, such as climate policy. Hence the paper contributes novel empirical insights on policy process dynamics, develops new machine learning methods, generates theoretical knowledge about drivers and obstacles to disputed policy change, and more importantly, answers the age-old question; who affects political decision-making?