Modelling Business Cycle Movements in the UK Economy

Abstract

This paper models the phases of the UK business cycle using GDP data with a time-varying transition probabilities (TVTP) Markov-switching regime model and exogenous leading indicator variables. Single indicators in linear models are compared with the TVTP framework, with logistic and exponential functions used in the latter. The Markov-switching models capture the major recessions of the sample, but the use of leading indicators through the TVTP framework can improve this regime recognition. Finally, a forecast comparison shows that the TVTP models perform relatively well in predicting during the 1990s, particularly when nominal interest rates are used to generate the regime-switching probabilities.

Forecasting UK industrial production over the business cycle

“…Search algorithms were employed to derive the linear univariate AR and single leading indicator models; details of these can be found in Simpson et al (2001). Suffice it to say here that autoregressive orders up to 4 are considered in equation (1), while the maximum leading indicator lag considered in equation (2) is 8.…”

“…Thus, the probability of remaining in a regime is conditional on the lagged value of the leading indicator, x t j , as well as the lagged regime, S t 1 . In Simpson et al (2001) we also apply an exponential form for the TVTP in the context of UK GDP. However, when this exponential function was applied to IOP there were many estimation problems so those results are not reported.…”

“…, S t rC1 D s t rC1 jY t , X t ; O Â] is the one-step-ahead predicted probability distribution of states conditional on Y t and X t which denote the set of information available on the dependent variable and the leading indicator respectively through to date t. This is also conditional on the vector of estimated parameters for the model, denoted by O Â. An algorithm (detailed in Simpson et al, 2001) is employed to select the Hamilton fixed transition probability (FTP) and the single indicator TVTP models. Essentially, SIC is used to select the autoregressive lag r of the FTP model.…”

“…Nevertheless, we consider both linear and non-linear specifications, and also univariate models as standards of comparison for the models employing leading indicator variables. In our complementary paper (Simpson et al, 2001) we focus on UK gross domestic product (GDP) and find evidence that the Markov-switching models and the use of leading indicators (especially interest rates) deliver improved forecast accuracy. Here we examine forecasts of industrial production.…”

Sincronización cíclica del sector manufacturero de méxico y estados unidos desde una perspectiva no lineal autorregresiva con transición suave

“…De acuerdo con sus características lineales, tanto con los modelos AR como con los ARMA, se pueden realizar estimaciones considerando una relación lineal de la variable dependiente y sus predictoras; por tal motivo, no poseen la capacidad de captar las variaciones producidas en la variable dependiente por el comportamiento asimétrico de las fluctuaciones cícli-cas (Neftçi, 1984;Hamilton, 1989;Teräsvirta, 1994;Simpson, Osborn & Sensier, 2001;Mourelle, 2010;Zepeda-Mercado, 2012).…”