One of the main objectives of the advocacy coalition framework is to explain policy change defined as fluctuations in the dominant system of beliefs, i.e. those incorporated into concrete policies. Over time, policy actors can adjust their understandings, beliefs and preferences related to public policy - a process called 'policy learning'. For this reason, the ACF argues that policy learning is an important pathway to policy change, next to policy actors' efforts to transform their coalition's preferences into policies. However, many studies looking at the relation between policy learning and policy change rely on cross-sectional data, which makes it difficult to establish a causal direction between policy learning and policy change. Does policy learning lead to policy change or the other way around? In this paper, we look at 1000 stances of 200 different policy actors in newspaper articles about the French politics of hydraulic fracturing. To address our question, we compare the proportions of policy actors who revised their beliefs/preferences before and after the major policy changes of 2011 and 2017 and conclude that policy learning is mostly a consequence of, rather than a pathway to policy change. Our paper innovates in two respects. Methodologically speaking, our answer to the research question is not dichotomic \textendash we assess `how much' (rather than 'whether') policy learning leads to policy change and the other way around \textendash and we rely on the analysis of longitudinal data. Theoretically speaking, we do not look at policy learning 'by default', to explain policy change: rather, we test this pathway to policy change before concluding that policy change results from other pathways \textendash i.e., changes in the dominant coalition and personnel turnover. This contribution to understanding policy learning-change relations advance frontiers on comparative policy research (over time) through the ACF research program.< Réduire