Shadow "banks" are unidentified entities operating in the financial sphere, in different countries. Therefore their lending and borrowing activities must be based on some rationally determined interest rate. We make the hypothesis that it is determined by a Reinforced Learning process and try to model this behaviour alongside a classic CCLM/DSGE model. This is a follow-up from the presentation made at EcoMod July 2012 Seville "A Markov random model of Interbank Dynamics with Filtering and Learning". We go farther in the implementation of the model. Starting from a DSGE equilibrium, forward simulations are made using Sequential Monte Carlo and interacting particle systems technique. Loopbacks are made with RL equations implemented in a dozen of heterogenous agents as "shadow bankers", each with own set of utility preferences and learning capacities. Tentatively, it appears that in a interbank liquidity shortage environment, the "shadow entities" will split up into 2 categories according to risk aversion and learning speed and their own size. I will upload a first version of the paper in February.
