The growing interest in social protection in Africa over the past two decades has led to a renewal of academic research and institutional literature, rang-ing from technical and evaluation approaches to political economy studies. The latter have the analyt-ical singularity of linking the outcomes of social pro-tection policies to their modalities of political inser-tion and appropriation rather than to their original conception and the manner in which they are im-plemented. As such, this report is an original contribution to the analysis of public policies in countries under foreign aid regimes. Considering the ‘political construction of public policies’ as a determinant of their success, we present here an empirical analysis of the elabo-ration of Madagascar’s new social protection policy. The study of the relationships between stakeholders reveals the coalitions of actors involved and their role in the ongoing changes in orientation. The empirical strategy we have chosen combines and applies the policy network and advocacy coali-tion framework (ACF) approaches by testing them with the tools of social network analysis. It is in line with the research on developed or emerging coun-tries that is rare or non-existent in low-income countries. The inter-organisational network data is drawn from a sociometric and qualitative survey carried out in 2018 and 2019 among the member organisations of the Groupe de travail sur la protec-tion sociale – GTPS (Social Protection Working Group). Under the auspices of the Ministry of Popu-lation, this group is responsible for drafting social protection policy in Madagascar. Joining the ACF and Policy Network methodological approaches, two complementary steps support our original empirical strategy. The first step deals with a structural analysis of social protection networks, using three cumulative criteria to identify coalitions of political actors. Foremost, a coalition necessary brings together structural equivalent actors within the network of collaborations (we applied one of the most relevant blockmodeling algorithm). After-ward, the coalition's subnetwork has higher within-clique density than between-clique density on col-laboration, sharing information and agreement ties. Finally, the coalition's subnetwork has higher be-tween-clique density than within-clique density on disagreement ties. The second step explores the resource circulation within the network and the cognitive consistency of each political coalition (closeness of values between actors). This then makes it possible to identify the coalition of power, with a strong capacity for mobili-sation and influence, that is at the heart of the new social protection policy. Our results show that Madagascar's approach direct-ly reflects the paradigm shift that took place in the international political arena at the turn of the 2000s. The five relational spaces under study reflect the singular way in which this has been translated in the Malagasy institutional and political context. That of a fragile, liquefied state, with a chronic inability to resolve the redistributive conflict, particularly in a phase of economic growth. Social protection policy is dominated by a ‘pro-vulnerable’ or, in other words, a ‘pro-cash’ coalition, which is much more decisive than the second, ‘pro-rights’ coalition. Composed mainly of actors from the relief sector, the leading coalition has a view of economic security issues based on the understand-ing of individual risks and market integration. Its organization is based on the centrality of UNICEF and includes the two ministries historically in charge of social protection in the country: the Ministry of Health and the Ministry of Social Protection. The over-determining role of international donors is confirmed on analysis. In a position to control the content of social protection policy statements and of the related policy tools, negotiations with national public actors remain limited. In this configuration, where the failure of politics is reflected even in the marginalization of civil society actors, the external global offer tends to be hegemonic. However, this conclusion calls for some nuance. Although they do not occupy central positions, gov-ernment institutions (ministries and agencies) often act as brokers. They build bridges between the sep-arate worlds of social protection. Even if they do not govern social protection policy, the state and its administration disseminate its principles and ideas. This role as an interface between the central inter-national organizations and the population, which is characteristic of a country under foreign aid regime, places the government institutions in the position of a “development-broker”. This encourages the re-production of resource accumulation strategies. Due to a lack of a dense internal social and political construction, social protection policy can only count on the accuracy and relevance of a comprehensive offer of protection and its financing through aid. From this point of view, the development of a new Malagasy social contract that would create solidarity is not on the agenda.