Trends and random walks in macroeconomics time series: The unit root test considerations

In the time series econometric literature, data generation and stationary are important issues in model selection and estimation method. Difference Stationary and Trend Stationary processes are data generation procedures. In Difference Stationary specification (integrated), the stochastic component follows a random walk process (unit root process) that it yields stationary by differencing, while in trend Stationary specification process, the stochastic component follows a stationary process. The variable variation pattern in the random walk model with trend (random) and the trend stationary model (deterministic trend) is very similar. The revealed facts show the upward trend of Iran's macroeconomic variables over the past few decades, which are very close to the variation pattern of trend and Difference Stationary models. In empirical work, the distinction between these two models is not simple, and misapplying of tests cause incorrect results in the research process. The purpose of this paper is to review again how to perform a unit root test and identify the nature of (deterministic or random) trends of macroeconomic time series of Iran. In the first step, the generalized Dickey Fuller root unit test (ADF) was performed using the Dolado et al. (1990) and Hamilton (1994) approach and then Perron's test (1989) was used to investigate structural break. The results show that 4 variables of 6 time series including nominal GDP, industrial value added, consumer prices and stock money follow the random walk process with positive drift (random trend), but real GDP and stock price index follow the trend stationary process (deterministic process).