Abstract
This thesis consists of two independent chapters, united by the topic of unemployment. The first chapter investigates the effect of uncertainty on quantiles of unemployment.
Although both linear and non-linear methods have been used in the literature to study this relationship, all of them have focused on the conditional mean and none have so far applied a vulnerable growth approach to unemployment. We propose a rather simple quantile regression model with lagged unemployment and lagged uncertainty to predict unemployment. The measures of uncertainty commonly found in the literature are used to construct six models -three quantile regression models and three linear regression models that serve as a baseline for our analysis. The measures in question are NFCI, VIX and EPU. Finding a strong and positive effect of uncertainty on unemployment, we continue to evaluate abilities of the three linear models in predicting the mean of unemployment and three quantile regression models in accuracy of coverage based on the respective measure of uncertainty used. NFCI is found to have inferior abilities to predict unemployment compared to VIX and EPU that both demonstrate equal accuracy, whether in the context of linear or quantile regression models.
The second chapter studies the relationship between unemployment, productivity and variance of productivity (macroeconomic volatility). It is using the stylized act uncovered by Benigno, Ricci and Surico (2015) in their paper on the long run relationship between unemployment and productivity in the US. They find the positive and significant relationship between unemployment and the variance of productivity, with the magnitude of the effect being much greater than that of a negative effect of productivity on unemployment. This result is in line with their predictions and macroeconomic theory. Using TVP VAR, Rolling Windows and GARCH as filters for the variance of productivity, we uncover opposite effects for the UK. We find a significant and positive effect of productivity and significant negative effect of its variance on unemployment, with magnitude of the former surpassing magnitude of the latter. Based on the results we conclude that an existence of two regimes is a possible explanation of the signs of coefficients of both productivity and macroeconomic volatility. To address this issue we continue our analysis using Markov-Switching models that are commonly used in the context of business cycles and the resulting asymmetry of the effects, identifying two regimes that are associated with recessions and expansions. The effects are preserved in a two-regime scenario. We propose a possible reason behind it being type of unemployment created by automation of low-skilled jobs happening on a large scale in the UK. This type of unemployment is not connected to recessions, but is indeed associated with higher productivity brought by automation. Another important distinction between the two economies that can play a role explaining the results is UK being a small open economy while US is a closed economy.