Adaptive Learning and Labour Market Dynamics

It is well known that the standard search and matching model with Rational Expectations (RE) is unable to generate amplification in unemployment and vacancies. We show that relaxing the RE assumption has the potential to provide a solution to this well known unemployment volatility puzzle. A model in which agents use Recursive Least Square algorithms to update their expectations as new information becomes available is presented. Firms choose vacancies by making infinite horizon forecasts over (un)employment rates, wages and interest rates at each point in time. Firms have much greater incentive for vacancy posting because of overoptimism about the discounted value of expected profits at the time a positive TFP shock hits the economy. The model with adaptive learning is able to match the relative volatility of labour market variables in US data and the properties of forecast errors of unemployment rates obtained from the Survey of Professional Forecasters.