A cross-province comparison of Okun's coefficient for Canada

Abstract

This study estimates Okun's coefficients for ten Canadian provinces using real GDP and unemployment rate data across the provinces. An average estimated Okun's coefficient of −1.58 is obtained under the Hodrick-Prescott detrending method and −1.32 under the quadratic detrending method. There is relative stability of the coefficients across the two detrending methods. Generally, the cost of unemployment in terms of the loss in real GDP is higher in the bigger and more industrialized provinces ranging from −2.14 for Ontario to less than −1 for the Maritime provinces.

Acknowledgements

We thank Prof. David E. A. Giles for his helpful comments and suggestions.

Notes

The GNP gap is the difference between GNP and potential GNP while the unemployment rate gap is the difference between the unemployment rate and natural unemployment rate.

Okun's approach is a very novel way of getting around the problem of predicting potential GNP.

Prachowny used two main US data sets that have been used by several other authors. These are the Gordon data set (1947:1 – 1986:2) and the Adams and Coe data set (1965:1 – 1988:4). The main difference between the two data sets is that Gordon's output gap refers to GNP, while the Adams and Coe measure is for the nonfarm business sector (i.e. about 80% of GNP).

Some are actually as simple as just drawing a line linking the peaks of the series.

Moosa (1997) used annual 1995 GDP (measured in 1985 prices) data covering the period 1960–1995 to estimate Okun's coefficient values for the G7 countries. No unit root test results were reported.

Both Moosa (1997) and Harris and Silverstone (2001) obtained the estimated Okun's coefficients using the reversed regression (regressing unemployment on real output) thus this estimated Okun's coefficient is usually inverted to obtain the true coefficient.

More detailed information on the various Canadian provinces can be found at government of Canada site (http://canada.gc.ca/).

This is the smoothing parameter recommended for annual data series by Hodrick and Prescott.

See Dickey and Fuller (1979).

See Kwiatkowski et al. (1992).