Looking at the policy domain of growth and poverty reduction in developing countries there is agreement that public policy is a key determinant. Investment policies like CAADP, have time-delayed and/or indirect effects on the wanted targets. Therefore making a choice in this domain is highly complex problem and political practitioners faced with this challenge form simple mental models that allow them to form a decision, a narrative. We are interested in the transformation of a policy choice, an allocation of budget to different policy instruments, into policy outcomes like poverty reduction. We separate this transformation into two parts: a policy - growth link and a growth - outcome link. The latter is derived by applying a meta-modeling approach to a CGE. The former is estimated applying the policy impact function approach. Using this approach we can extract the assumed underlying technology into a quantitative model. Applying this framework we are able to estimate actor or group specific technologies and can quantify the implicit narratives actors have when making policy decisions. This also allows to measure the inefficiency of policy choices. In a second step it also allows to disentangle the inefficiency into two parts. Is the inefficiency due to missing knowledge or due to wrong incentives? Another interesting application is the comparison of the ``scientific world'' of economic modelers with the ``practitioners world''? Are they fundamentally different? Beyond the theoretical derivation of the framework, it is applied empirically to the case of CAADP in Senegal.