The Search and Matching Model: Challenges and Solutions

Abstract

This dissertation focuses on explaining the cyclicality of unemployment, job vacancies, job creation and market tightness in the US economy. The framework used to model unemployment and job creation throughout this work, is the search and matching model, created by Mortensen and Pissarides (1994). This dissertation proposes three different mechanisms to improve the performance of a dynamic stochastic general equilibrium model (DSGE) with search unemployment, to align the model’s predictions with the quarterly US data from 1955-2005. The first chapter proposes a New Keynesian model with search and matching frictions in the labor market that can account for the cyclicality and persistence of vacancies, unemployment, job creation, inflation and the real wage, after a monetary shock. Motivated by evidence from psychology, unemployment is modeled as a social norm. The norm is the belief that individuals should exert effort to earn their living and free riders are a burden to society. Households pressure the unemployed to find jobs: the less unemployed workers there are, the more supporters the norm has and therefore the greater the pressure and psychological cost experienced by each unemployed searcher. By altering the value of being unemployed, this procyclical psychological cost hinders the wage from crowding out vacancy creation after a monetary shock. Thus, the model is able to capture the high volatility of vacancies and unemployment observed in the data, accounting for the Shimer puzzle. The paper also departs from the literature by introducing price rigidity in the labor market, inducing additional inertia and persistence in the response of inflation and the real wage after a monetary shock. The model's responses after a monetary shock are in line with the responses obtained from a VAR on US data. In the second chapter I attempt to solve the amplification puzzle, the inability of the standard search and matching model to account for the volatility in vacancies and unemployment, by exploring the connection between RD thus, when agents need to form multi-period forecasts using past data as in Preston (2005) and Eusepi and Preston (2010), the search and matching model's amplification potential is enhanced. The model can replicate moments from quarterly US data from 1955Q1 to 2007Q4. However the more amplification added to the model through higher gain parameter, the further the correlations generated by the model are from the ones obtained from US data. That is because higher amplification induces vacancies to fluctuate around the rational expectations equilibrium less smoothly. Moreover, higher amplification through learning worsens the prediction of the model for the slope of the Beveridge curve.