The fundamental issue faced by economists and policymakers—while developing theoretical models, forecasting economic conditions, analyzing and prescribing economic policy—is the selection of appropriate linear or nonlinear representation1. Traditionally, most macroeconomic series such as output are modeled as linear. Such linear representation cannot generate and associate with the asymmetric process. Choosing correct specification for modeling process brings implication for the timing and magnitude of the policy prescription whilst wrong representation may cause inaccurate economic forecast and ineffective policy options. It also reflects the increasing interest being paid to nonlinear theoretical and empirical models in which the conventional linear models being inherently incapable of generating or representing asymmetric time series behavior. Hence, the primary objective of the study is to investigate
whether the data generating process follows a symmetric or asymmetric path in the 11 developing countries. Early and informal measures of asymmetry focussed on the distances from trough to peak and peak to trough in the time series profiles of business cycle indicators. Only recently, the reappraisal in the ‘stylized facts’ of cyclical asymmetry genre has received more rigorous empirical attention. A number of approaches have been used to detect possible asymmetries of the observed economic time series data.

Researchers who based their work on the Markov processtheory include Neftci (1984), Falk (1986), Sichel (1989) and Mills (1995) have suggested little or no evidence of asymmetries for business cycle. In addition, Luukkonen and Terasvirta (1991) and Terasvirta and Anderson (1992) has adopted the Self-Exciting Threshold Autoregressive (SETAR) approach had reported the significant nonlinearities in industrial production indices for several OECD countries. Using similar development, Brock and Sayers (1988) have applied Chaos Theory to detect the asymmetry pattern in business cycle. Limitations of several nonlinearity techniques presented above are the inability to discern different types of asymmetry. To deal with these limitations, Sichel (1993) has intensified the empirical debate by constructing simple robust tests of distinctive types of cyclical asymmetry, which extended the originally approach due to DeLong and Summers (1986)2. Numerous researchers had applied such a technique, see, for example, Sichel (1993), Giles (1996), Speight (1997), Verbrugge (1997), Canova (1998), Stanca (1999), Cook (2000), Bodman (2001) and Olekalns (2001), in terms of GDP, GNP or industrial production, found mixed evidence of asymmetries. Specifically, Cook (2000) has found significant business cycle asymmetry for most of the countries studied while Giles (1996), Speight (1997), Canova (1998) and Olekalns (2001) results do not display any detectable asymmetries at the business cycle frequencies. In addition, Razzak (2001) splits a postwar sample of the US real GDP into the exchange rate regimes of Bretton-Woods and free floating, and discovers a significant deepness only in the latter. Recently, Knuppel (2004) has found strongly significant deepness and steepness in the US GDP.

Business Cycles Asymmetry: An Analysis of Developing Countries,asymmetry, business, economic, linear, series, countries, significant, representation, Sichel, evidence, addition, industrial, Canova, Olekalns, cyclical, policy, developing, deepness, Speight, Terasvirta, ‘stylized, discovers, distances, distinctive, economists, Autoregressive, forecasting, frequencies, fundamental, implication, increasing, BrettonWoods, interest, Limitations