This dissertation makes contributions in topics related to mechanism design and learn-ing in game theoretic environments through three essays. The rst essay deals withthe question of mechanism design in the principal-agent model. The main contribu-tion of this essay is in extending the work by Piketty (1993). It prescribes a mechanismin incomplete informational settings where the principal is able to implement rst-best contracts while extracting the entire surplus. Importantly, the mechanism issuch that the desired outcome can be uniquely obtained when agents play the actionthat survives iterative elimination of dominated strategies. Furthermore, given themechanism, the desired outcome is shown to be a truth-revealing Nash equilibriumwhich is also Pareto-ecient. It is shown that the proposed mechanism also has thefeature that none of the agents prefer any of the other possible Nash Equilibria tothe status quo. It thus gives insights into possible mechanisms in nite agent settingsthat could improve upon the traditional second-best results.In the second essay, a model of sophisticated learning is developed where itassumes that a fraction of the population is sophisticated while the rest are adaptive learners. Sophisticated learners in the model try to maximize their cumulative payoin the entire length of the repeated game and are aware of the way adaptive learnerslearn. Sophisticated learning contrasts other models of learning which typically tendto maximize the payo for the next period by extrapolating the history of play.The sophisticated learning model is estimated on data of experiments on repeatedcoordination games where it provides evidence of such learning behavior.The third essay deals with the optimal pricing policy for a rm in an oligopolythat is uncertain about the demand it faces. The demand facing the oligopoly, whichcan be learned through their pricing policy, changes over time in a Markovian fashion.It also deduces the conditions in which learning (experimentation) is not achievableand outlines the dierent learning policies that are possible in other settings. Themodel combines the monopoly learning literature with that of the literature on pric-ing behavior of rms over business cycles. The model has interesting insights onthe pricing behavior over business cycles. It predicts that prices jump as the beliefof a possible future boom rises over a certain threshold. The model also predictscompetition to be quite vigorous following a boom while rms are predicted not toexperiment with their (pricing) policies for many periods following a bust.

This dissertation makes contributions in topics related to mechanism design and learn-ing in game theoretic environments through three essays. The rst essay deals withthe question of mechanism design in the principal-agent model. The main contribu-tion of this essay is in extending the work by Piketty (1993). It prescribes a mechanismin incomplete informational settings where the principal is able to implement rst-best contracts while extracting the entire surplus. Importantly, the mechanism issuch that the desired outcome can be uniquely obtained when agents play the actionthat survives iterative elimination of dominated strategies. Furthermore, given themechanism, the desired outcome is shown to be a truth-revealing Nash equilibriumwhich is also Pareto-ecient. It is shown that the proposed mechanism also has thefeature that none of the agents prefer any of the other possible Nash Equilibria tothe status quo. It thus gives insights into possible mechanisms in nite agent settingsthat could improve upon the traditional second-best results.In the second essay, a model of sophisticated learning is developed where itassumes that a fraction of the population is sophisticated while the rest are adaptive learners. Sophisticated learners in the model try to maximize their cumulative payoin the entire length of the repeated game and are aware of the way adaptive learnerslearn. Sophisticated learning contrasts other models of learning which typically tendto maximize the payo for the next period by extrapolating the history of play.The sophisticated learning model is estimated on data of experiments on repeatedcoordination games where it provides evidence of such learning behavior.The third essay deals with the optimal pricing policy for a rm in an oligopolythat is uncertain about the demand it faces. The demand facing the oligopoly, whichcan be learned through their pricing policy, changes over time in a Markovian fashion.It also deduces the conditions in which learning (experimentation) is not achievableand outlines the dierent learning policies that are possible in other settings. Themodel combines the monopoly learning literature with that of the literature on pric-ing behavior of rms over business cycles. The model has interesting insights onthe pricing behavior over business cycles. It predicts that prices jump as the beliefof a possible future boom rises over a certain threshold. The model also predictscompetition to be quite vigorous following a boom while rms are predicted not toexperiment with their (pricing) policies for many periods following a bust.

Climate change must deal with two market failures: global warming and learning by doing in renewable energy production. The first-best policy consists of an aggressive renewables subsidy in the near term and a gradually rising and falling carbon tax. Given that global carbon taxes remain elusive, policy makers might have to rely on a second-best subsidy only. With credible commitment the second-best subsidy is higher than the social benefit of learning to cut the transition time and peak warming close to first-best levels at the cost of higher fossil fuel use in the short run (weak Green Paradox). Without commitment the second-best subsidy is set to the social benefit of learning. It generates smaller weak Green Paradox effects, but the transition to the carbon-free takes longer and cumulative carbon emissions are higher. Under first best and second best with pre-commitment peak warming is 2.1-2.3 °C, under second best without commitment 3.5 °C, and without any policy 5.1 °C above pre-industrial levels. Not being able to commit yields a welfare loss of 95% of initial GDP compared to first best. Being able to commit brings this figure down to 7%.

Submitted by Helena Laneuville Teixeira Garcia (laneuvillehelena@gmail.com) on 2017-05-26T19:21:45Z No. of bitstreams: 1 Dissertacao_Final.pdf: 698751 bytes, checksum: a42e995534698e498fe856b2bc63c1d1 (MD5) / Approved for entry into archive by Marcia Bacha (marcia.bacha@fgv.br) on 2017-05-30T13:36:26Z (GMT) No. of bitstreams: 1 Dissertacao_Final.pdf: 698751 bytes, checksum: a42e995534698e498fe856b2bc63c1d1 (MD5) / Made available in DSpace on 2017-05-30T13:36:52Z (GMT). No. of bitstreams: 1 Dissertacao_Final.pdf: 698751 bytes, checksum: a42e995534698e498fe856b2bc63c1d1 (MD5) Previous issue date: 2017-03-24 / This thesis develops a simple model to represent a procurement situation with two main features. The first is that the optimal level of production cannot be fully anticipated when suppliers build their plants due to demand shocks. The second is that producers competing for a supply contract typically have different technologies within an efficient frontier, characterized by a trade-off between the marginal cost of production and the fixed cost per unit of capacity. With this framework in mind, we investigate how the shape of the frontier and the distribution of shocks affect efficient technology choices when the planner knows firms' technologies (first-best) and when she doesn't (second-best). In addition, we characterize how and when a well established real-life mechanism such as a quasi-linear score auction may implement second-best social welfare. We find that, if there is a strict preference over technologies in first-best, a quasi-linear score auction may implement second-best allocations. However, there is a non-neglectable case in which countervailing incentives arise, i.e. firms' allocations may be distorted either upwards or downwards with respect to first-best depending on their technologies. In that case, the planner may optimally choose to hire more than one firm, and there is no quasi-linear score auction that provides the social welfare achieved in second-best.