Dynamic factor Markov switching model and its applications in business cycles

We introduce the dynamic factor Markov switching model and Kalman filter technique in business cycle modeling. We investigate how these mechanisms are developed and can be used in national and international business cycle modeling. Kalman filtering is a technique introduced from control engineering where the observed variables are influenced by unobserved factors. In dynamic macroeconomic modeling, the Kalman filter will be applied to estimate or describe the current state of the economy—the process of the common factor based on the observations of aggregate time series variables. We consider economic variables as indicators for inferences about economic conditions, which means that the movement of these variables can be captured by the dynamic process of the common factor. The economic conditions are not constant and are subject to asymmetric shifts between expansions and recessions in business cycle. Thus, the dynamic common factor is subject to Markov switching in regimes, which is an index for the general state of the national and international economy or business cycles. We must point out that the simultaneous movement among time series in the business cycle does not necessarily mean the phases for each economic series are perfectly synchronized. Although the series in the model are driven by the same factor, some series may lead, lag or coincide in terms of the sequence of time for the phases. We hope to find the regularities for the comovements of the vector of these economic time series so that we can use indicators to make inferences about national and international business cycles. This dissertation models the national business cycles of each of the G7 countries, and characterizes a potential international business cycle reflecting economic linkages among the G7 countries and the 29 country-members of the Organization for Economic Cooperation and Development (OECD). The dynamic factor is a latent variable that captures simultaneous cyclical fluctuations in a set of economic aggregates—expansions and recessions. This framework allows business cycle phases to display asymmetries with respect to their duration, amplitude, or intensity. When applied to country-specific macroeconomic variables that display simultaneous movements to national GDP, the dynamic factor model characterizes country-specific business cycles. When applied to measures of production across different countries, it decomposes fluctuations in these series into a common cyclical factor underlying these countries, and movements specific to each individual country. The common factor is a measure of the international business cycle. We find that the model clearly depicts a regularity and comovement of business cycle, both nationally and internationally, which matches closely the ex post NBER business cycle dating and the OECD peak/trough chronology.