روند کنونی در تجزیه و تحلیل رشد بهره وری کانادا

For more than a decade, debates over the impact of new information technologies on trend productivity growth rates have played a key role in the formulation of monetary policy in many countries, including the United States and Canada. However, the question of whether the trend growth rate of aggregate productivity has changed significantly is rarely examined formally. This paper examines the latest aggregate labour productivity data for Canada using a new testing approach specifically designed to detect recent changes in trends. In addition to showing the strength of the evidence for changes in long-run trends, it considers the effect that data revision and changing sample period has had on inference about structural changes. In an appendix, it investigates how large such changes must be before they can be detected and to what degree detection tends to lag the structural change. Evidence of a decline in the trend rate of labour productivity growth in Canada since 1990 is mixed. In particular, conclusions vary considerably from year to year as data are revised and as the accumulation of observations after purported breaks changes initial perceptions. The instability of test results suggest that policymakers need to use extreme caution in interpreting claims of changes in labour productivity trends and highlight the uncertainty that they face.

Trends in productivity growth play a key role in the formulation of public policy. They are important factors in determining long-run economic growth and therefore play central roles in the management of public pension systems and government debt. They are an essential component in defining measures of economic slack and have therefore played a key role in the formulation of monetary policy, particularly in the United States and Canada.2 International differences in such trends in turn have profound influences on the balance of world saving and investment. Not surprisingly, the possibility of a persistent change in aggregate productivity growth casts a long shadow over many of the most important economic policy debates. For all these reasons, great effort is devoted to accurate productivity measurement and to the analysis of sources of productivity growth.3 Surprisingly, however, the question of whether the trend growth rate of aggregate productivity has changed significantly is rarely examined formally. 4 The answer to this question is frequently ambiguous, even in the much-studied context of recent trends in U.S. productivity growth. For example, Gordon (2003) concludes Productivity growth experienced a second acceleration in 2000–03 following the initial productivity revival of 1995–2000. [p. 272] a view shared by Bailey (2003). 5 Blinder and Yellen (2001) present evidence for a change at the end of 1995, and Hansen (2001) finds that formal statistical analysis supports this timing. 6 However, Fair (2004) argues that late 1995 is simply a cyclical trough in productivity growth and that cyclically adjusted productivity growth shows very little improvement in the 1995–1999 period. Maury and Pluyaud (2004) find weak evidence of trend break in US hourly productivity in 1995Q3, but none after the early 1980s when using per capita GDP. We argue that at least two potential problems complicate the analysis in this literature. First, productivity data are revised over time, with the revisions often causing non-trivial changes in measured growth rates. None of the above papers investigate this problem in a systematic way. Second, very few papers perform statistical tests for changes in productivity growth trends, and some of those that do (Hansen, 2001 and Maury and Pluyaud, 2004) use methods that are known to be unreliable close to the end of sample. This paper examines the latest aggregate labour productivity data for Canada using a new testing approach specifically designed to detect recent changes in trends. Using real-time data, it then considers the impact of data revision on the detection of trend breaks. In an appendix, we also consider how large changes must be before they can be detected and to what degree detection tends to lag the structural change.

The results presented above support two main conclusions. The first is that there appears to be evidence of a slowdown in productivity trend growth in Canada that took place around 1997. There is also recent evidence of a further slowdown that coincides with the 2008 recession. The second conclusion, which may be of more general applicability, is that such statistical evidence of structural changes is frequently fragile. We have shown that data revision may overturn strong statistical evidence of structural changes, or create it where none has existed before. Furthermore, this does not appear to be the result of the refinement of preliminary estimates in the first quarter or two after their release; rather, data revisions may lead to re-evaluation of events several years after the fact. We have also seen that the data revision is not the sole factor responsible for the fragility of statistical test results. The accumulation of observations can also be very informative about the validity of perceived changes in productivity trends. Even in the absence of data revision, therefore, judgements about productivity growth trends may be revised for many years after the fact. The instability of test results suggest that policymakers need to use extreme caution in interpreting claims of changes in productivity trends and highlight the uncertainty that they face. Improving the reliability of existing statistical methods remains an important task for future research.