## steady state predictions firm dynamics steady state for a mass performance differences (some firms are more profitable than others) di managerial practice aren't easy to observe but (org capital improves,) -> management has causal impact on profitability leaders and individuals matter (x=1 vs 0); manager is at the center of the model in terms of modeling (mechanical; how you write apply paper, not relevant with the real words) - no surprising (luck in ) - solve for (single agent model stacked together) -> incorporate strategic - useful for empirical; unify lots of theoretical works - broad knowledge of literature ---- intuitive game theory building routines: learning, cooperation, ## chassang (2010) - exploration and exploitation phase - value of collaboration grows and parties build a common understanding of their environment ### simplified - symmetric information - principal and agent interact repeatedly - choose routine (d=0) or experiment (d=1) - if d=1: agent chooses private effort e in 0 or 1 - p is probability of improving sate (pde) - rule out money btw employer and employee - Levin, nothing interesting happens on path - only exert effort if the agent thinks gurantee is better - firing must happen on path - if you tell her you'll never fire her, then she'll never work hard - fire or not given agent's past is computationally hard - IF FIRING HAPPENS, IT SHOULD HAPPEN IMMEDIATELY - effort is not observable (only know whether probability is high or not - tradeoff ; cost of doing vs ) - - ⭐️trying to build equilirium where agent will exert effort⭐️ () - even ; cost of having to fire a hard working person - - the higher value of relation (AS THE COMPANY MATURES), the less incentive you have to experiment (Chassang) VS tomorrow paper: the opposite - persistence to performance difference (PPD) - if you have money, you'd never fire agent -> loose the intuition - 💰do i put money(so powerful; model become stationary) or not in the money (no firing) -> DON'T HAVE STH IN BETWEEN (wealth constraints) - if R_t = b, not to experiment, --- - implementability of experimentation policies sum across domain, because of the inequality, there will be interdependancy, breakthrough in domain 1, our relation; pool incentive accros domain, no loss in pooling from every domain (defect, punish in every domain) project never expire, average of the project should be sufficiently higher (deteriorate; ) explore until we find "v not", don't know what you don't know - become confident, - exploitation phase is harder to sustain as outcome if fixed (clear winner and loser) WHEREAS exploration phase people all have they will be the winner - you can't characterize it - what i can do today -> what i can do tmr --- charlie - just as you want me to be happy, i want you to be happy, and that was motivation behind trying to add operational piece i wish to have