Development economics witnessed an ‘institutional turn’ from the 1990s onwards, with an increasing number of studies examining the two-way relationships between institutions and growth or individual incomes. Most of these studies, however, rely on mathematical models and with empirical evidence being analysed via econometrics, thus assuming the possibility that given institutions may be subsumed under quantifiable variables. Many possible channels of causality have been highlighted over time and are in fine inconclusive. The context-dependence of causalities has been underscored in the literature. This paper goes further, building an original conceptual framework that articulates concepts from development economics, evolutionary theories of institutions and cognitive psychology. Firstly, the paper argues that the modelling of causalities is hindered by the very nature of the concept of institutions: these are composite entities and include varieties of ‘forms’ and ‘contents’, which are driven by heterogeneous evolutionary dynamics and combine between themselves according to contexts. The concept of institution thus does not have the properties of semantic precision and stability in time and space that is yet required by modelling. What econometric studies measure are indeed some public attributes (‘forms’) of an institution, but this gives little information on the beliefs that individuals have about a given institution. Building on this first argument , the second argument is that this ex ante causal indeterminacy (on economic outcomes) is compounded by the heterogeneity of institutions: some institutions have the property in individual mental representations to be more resilient than others, notably those governing group memberships: in the context of a limited presence of other institutions (e.g., those created by states) they exhibit a greater social dissemination - but cumulative causation may generate ‘institutional traps’. Understanding these ‘core’ institutions may explain the vicious circles in which some developing regions seem to be caught.< Leer menos