Academic journal article Journal of Economics and Economic Education Research

The Role of Stationarity in Business and Economic Research

Academic journal article Journal of Economics and Economic Education Research

The Role of Stationarity in Business and Economic Research

Article excerpt

INTRODUCTION

Creating decision models for analysis is a critical requirement of many business and economic applications--forecasting, budgeting, causality, and organizational control systems. According to Arsham (1994), "existing formalisms and methods of inference have not been effective in real-time applications where tradeoffs between decision quality and computational tractability are essential." That is, to effectively perform time series analysis, it is essential to realize the fundamental structure and function of the data generating process (DGP) that generates the observations. Understanding time series would permit for the development of an econometric model that can best represent the data to employ forecasting, monitoring, and/or control (Senter, n.d.).

However, time series data (such as asset and stock prices, unemployment rate, exchange rates, inflation, gross domestic product) are often non-stationary, has unit root, or have means, variances, and covariances that are time variant. The data can possess trends, cycles, and/or random walks (Iordanova, 2007). Consequently, it becomes difficult to identify the systematic patterns within the data (Senter, n.d.). Fortunately, there are time series analysis techniques that can filter the data to reduce the errors.

In this study, I lay emphasis on on the inherent random process, trending behavior, or nonstationarity in the mean of most time series data. Meanwhile, a stationary process reverts around a constant long-term mean and has a constant variance independent of time (Iordanova, 2007). Hence, a mandatory econometric task before performing time series analysis is to establish whether the series is non-stationary, that is, there is a need to determine the most appropriate form of the trend in the time series data (Enders, 2004). To generate consistent and reliable results, the non-stationary data must be stationarized or de-trended prior to analysis.

A crucial characteristic of a high-performing manager is the aptitude to employ a strategic approach in making long-term managerial decisions that will achieve organizational goals. A fundamental quality of long-term managerial decision is its strong dependence on forecasts. According to Arsham (1994), "every decision becomes operational at some point in the future, so it should be based on forecasts of future conditions." Forecasts are critical in any business organization because the impact of the forecasts on actual performance is linked. Variances in forecasts and actual can compel decision to be revised. As such, this warrants the use of a reliable time series data. An organization cannot make a good decision nor create a correct planning strategy without a suitable time series data. A sophisticated time series analysis technique will not be enough given poor data.

Nonstationarity can be detected through various testing procedures. The most popular of which are (1) the Augmented Dickey-Fuller (ADF) and (2) the Phillips-Perron (PP) Unit Root Test. Likewise, non-stationary time series can be stationarized via available detrending procedures such as first differencing and time-trend regression (Enders, 2004). Furthermore, suitable alternative estimation procedure involves checking for cointegration and specifying an error correction model (ECM) (Elder & Kennedy, 2001). However, these alternative procedures are beyond the scope of this study. This study exposit this mandatory initial step in time-series analysis--detect for nonstationarity. This is key because in business and economic research, a well-organized time series data is vital to generating reliable results for analysis, decision-making, and long term planning. Highlighting both the importance and conviction of stationarity tests would assure researchers that they generate results that are not spurious.

Given such limitations and disagreements of various unit root tests to detect nonstationary time series, it is imperative to know their respective powers and shortcomings given the kind of non-stationary properties present. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.