A NON-LINEAR APPROACH TO JAPANESE BUSINESS CYCLES

A considerable amount of interest has been taken in empirical tests to determine chaos in business cycles after the introduction of the non-linear approach to economics. Amongst the several empirical works on macro-economic time series. Frank. Gencay and Stengos (1988) found the strong non-linearity in the Japanese data. This paper attempts to provide a precise mathematical version of empirical tests on Japanese time series for the presence of low-dimensional deterministic chaos. The "residual diagnostics" which was first presented by Brock(1986) is conducted on the Japanese macro-economic quarterly time series. including the periods of "rapid growth", "two oil crises" and "bubble economy and after bubble depression "(from 1955 to 2000). For the low-dimensional deterministic chaos tests. the correlation integral correlation dimension and BDS statistics are examined Empirical results show that the correlation dimension grew linearly with the embedding dimension and did not saturate at any value. The BDS statistics distribution has a large peak near 0. Based on these results, evidence for chaos in the Japanese quarterly macro-economic time series is weak. and fluctuations are primarily caused by "random noise."