Canadian Regional NAIRU Estimates: A Structural Break Approach

Abstract

Abstract. This paper applies the Bai-Perron method to identify longer-run structural breaks in Canadian regional unemployment rate series in order to estimate natural rates of unemployment for Canada and its regions. The longest samples (Canada, Quebec, Ontario, British Columbia, the Maritimes, and the Prairie region) span the 1946-2011 period while the shortest samples, covering smaller provinces, are for 1966-2011. In all cases the technique reveals significant breaks. On the longer series a jump upward of 1.6%-3.2% is found in the mid-1950s. A second, and larger, increase occurs in the mid-1970s in all but the two western-most provinces and the Prairie region. Further increases are seen in the early 1980s in all provinces west of Ontario as well as in the Maritime region and Nova Scotia. Declines in natural rate estimates are found for several provinces and Canada as a whole in the mid-to-late 1990s.1. IntroductionCanadian regional unemployment rates varied considerably over the period 1946-2011. Some of the variation, indeed the sharpest changes, are undoubtedly due to the business cycle. However, much of it may be accounted for by changes in the level of longer-run equilibrium unemployment rates around which the actual unemployment rate fluctuates; in other words, it is accounted for by changes in the natural rate of unemployment or NAIRU.1 These long-run natural rates are of some interest as indicators of the efficiency of individual regional labor markets and as determinants of regional wellbeing. The complex interplay between these regional NAIRUs and interregional migration is also of much interest. The level of the long-run regional unemployment rate may help determine long-run interregional migration patterns or the business locations of firms. Conversely, these same long-run unemployment rates will in part be determined by interregional labour mobility and its long-run determinants.This paper applies the Bai-Perron method to identify longer-run structural breaks in Canadian regional unemployment rate series in order to estimate natural rates of unemployment for Canada and its regions. The method is described after discussions of alternative approaches and data.2. Measuring the natural rate: Alternative approachesOne branch of the empirical natural rate literature hearkens back to Friedman's (1968) view of a natural rate of unemployment determined by various labor market frictions or imperfections. This imperfections approach estimates unemployment rate equations with proxies for the imperfections and a cyclical measure as control variables. Nickell, Nunziata, and Ochel (2005) is a recent example that generates estimates for Canada as a whole and other OECD countries. Canadian examples which use this approach to generate regional natural rate estimates include Riddell (1980), Miller (1987), Burns (1990), and Johnson and Kneebone (1991). Note that regional context factors that discourage interregional labor mobility will raise regional natural rates. See Cebula and Alexander (2006) for a recent study of the determinants of interregional mobility and Deller (2009) for an examination of regional equilibrium unemployment rates using a wage curve model.An alternative approach infers the value of the natural rate or NAIRU from behavior of a Phillips curve relationship. Examples of this approach include Gordon (1997), Ball and Mankiw (2002), and Ball (2009). These studies use a Hodrick-Prescott filter to separate the effect of the Phillips curve error term from the NAIRU. Laubach (2001) adopts another common approach that starts with a Phillips Curve relationship and then uses a Kalman filter to infer behaviour of the unobserved NAIRU.2 Both the imperfections approach and the Ball-Mankiw Phillips Curve method give natural rate estimates that show considerable variability over time. Indeed, they often resemble the actual rate series with the shorter-term fluctuations smoothed out. …