Trends, cycles and measures of persistence in the Norwegian economy

This study analyses empirically the business cycles in Norway using quarterly national accounts from 1967 to 1994. To extract the cyclical component, we need to eliminate the component in the data. However, the cycle will not be invariant to whether we describe the trend as stochastic or deterministic. Testing for unit roots allows us to determine whether a series is best described by a stochastic trend or by a deterministic (linear) trend. By using a test for unit roots that allows for a structural break in the slope or the drift of the linear trend alternative, we can reject the unit root hypothesis for e.g. unemployment and investment, and instead describe them by a linear trend with one structural break. The structural break point is unknown a priori, and for unemployment and investment it is estimated to have occurred in the late 1980s.

To extract cyclical components in the economic variables, we use a variety of stochastic and deterministic trend alternatives. The detrended data (the business cycles), are thereafter analysed both in the time domain and the frequency domain. In the time domain we concentrate on persistence and correlations, whereas in the frequency domain we establish whether the cycles we have estimated have any important periodic components. We then investigate whether the business cycle components are sensitive to the methods of trend extraction used. We find that for some variables, the measures of business cycles are qualitatively independent of the way we have extracted the trend, although quantitatively, the results may differ somewhat. For instance, analysing the correlations between GDP and other economic time series, we find that some variables are persistently procyclical (e.g. consumption, import, investment and productivity) or persistently countercyclical (e.g. unemployment). However, the magnitude of these correlations varies.

For other variables, the business cycles vary considerably with the detrending methods used. For example, traditional exports, real wage and consumer prices show both a procyclical and countercyclical pattern, depending on how we define the trend component in these series. The sensitivity of business cycles to the measurement of the trend, implies that one should be careful not to extract the trend component without first examining the dynamic properties in the data.

Keywords: Unit roots, structural breaks, measures of persistence, trend-cycle decompositions, stylized facts, spectral analysis, time series analysis.

Acknowledgement: Financial support from the Norwegian Research Council is acknowledged.