Empirical analysis of business cycles using the unobserved components model
thesisposted on 2022-03-29, 03:31 authored by Andrew E. Evans
Timely estimates of the current state of the business cycle in a number of economic indicators are essential in the setting of the stance of macroeconomic policy. There are several methods for extracting estimates of the underlying trends and cycles but results vary considerably by method, and it is sometimes uncertain whether the estimated components actually represent real underlying phenomena. The central challenge of each of the three articles forming the main body of this thesis is finding a meaningful statistical characterisation of the business cycle component in various aggregate macroeconomic variables using an unobserved components model.The first article demonstrates how a common cyclical component can be extracted from a set of related labour market statistics to make new estimates of Okun coefficients and participation coefficients by age and gender. The estimated coefficients were generally higher than those typically reported in the literature and reveal the characteristic cyclical response of different groups of workers to an economic downturn. The second article sheds new light on previously disparate empirical findings of procyclical or countercyclical productivity, against a theoretical framework in which it is usually held to be procyclical. An extension of the unobserved components model which allowed direct estimates of phase-shift amongst similar cycles was used to explain the apparent countercyclical and lagging behaviour of average labour productivity relative to the output cycle. The third article develops a non-linear version of the model which incorporates the possibility of asymmetric responses to an exogenous recession indicator. Two related questions are considered in the article, namely, whether a non-linear model with asymmetric features provide a better fit to macroeconomic data than a symmetric model, and whether a recession is a real macroeconomic phenomenon, more than simply the mirror image of the expansionary phase of the cycle. Evidence is found in favour of asymmetry. One important aspect of business cycle research is revealing the characteristic comovement of the cyclical components of several economic aggregate variables in a multivariate framework. The three articles in this thesis have demonstrated how such a framework can be used in situations where economic policy making will benefit from a better understanding of the cyclical behaviour of key macroeconomic aggregate variables and the interaction between them.