The C.D. Howe Business Cycle Council defines a downturn as a recession if there is, broadly speaking, a pronounced, persistent, and pervasive decline in aggregate economic activity. In other words, to identify a recession three dimensions need to be considered simultaneously: amplitude, duration, and scope – or how widespread a downturn is.
The Council does not follow hard-and-fast rules on these dimensions, instead relying on several guidelines in making our recession determinations. First, a decline in aggregate economic activity lasting at least one quarter is a necessary minimum for a recession. If whatever is provoking the slowdown in economic activity does not lead to at least one quarter of outright decline, it has not produced the sequence of cutbacks one expects to see when studying business cycle downturns.
Second, in terms of amplitude, a 0.1 percent quarterly decline in aggregate economic activity is a necessary but not sufficient condition. For instance, a transitory 0.1 percent quarterly dip in a year otherwise characterized by robust growth is clearly not a cyclical event, whereas a 0.1 percent decline spanning multiple quarters would provide a much clearer claim to an economy in recession. Typically, a quarterly drop in the economy, particularly if it is of low amplitude, would need to be validated by accompanying weakness in contiguous quarters, but not necessarily by outright decline. This specification implies a net contraction, or at least stagnation, in the economy over a two-quarter period. Most analysts would agree that such a period is a recession, because the suppression of the economy’s natural buoyancy over such a time span signifies the presence of true recessionary forces rather than data gremlins or one-off artificial factors.
Third, determining the minimum scope of a recession similarly requires some judgement. We use two different diffusion indices, generated internally, to gauge the breadth of a particular downturn. The first index is an unweighted measure built as follows: expanding industries over a given period receive a score of 100, those falling receive a score of zero, and those with unchanged output receive a score of 50. A simple average is then calculated using the scores for each industry. A score above 50 indicates more expanding than contracting industries and a score below 50 the opposite.
The unweighted version of diffusion provides a pure measure of breadth of changes in economic output, as it abstracts completely from both the depth and amplitude of economic growth. However, there are statistical techniques that can yield a truer measure of the underlying trend in the economy by better taking sectoral co-movements into account. One of those statistical techniques is principal components analysis, which serves to remove idiosyncratic shocks at the sector level in order to properly identify true business cycle shocks that reverberated across multiple sectors. Similar scoring as in the unweighted index is used with 50 as the dividing point.
Finally, the practical definition of aggregate economic activity changes over time to allow analysts to use the best available data. For recent decades, we begin with quarterly GDP and employment data as the primary means of identifying probable recessions, before turning to more granular monthly data – monthly GDP starting in 1961 or industrial production prior to 1961, along with monthly employment – to isolate the precise peaks and troughs. Before 1980, employment was often less sensitive to declines in output; after 1980, employment moves more closely with the business cycle. With expenditure GDP data now extending back to 1961, we are able to use our preferred GDP measure – the average of expenditure and industry GDP – over a longer historical period.
These criteria provide a useful framework for identifying and dating recessions, but fall somewhat short of providing a mechanical rule. It is impractical to establish preset conditions with respect to amplitude, duration, and scope because these considerations need to be judged simultaneously and because the data and statistical tools available for vetting the economy change over time. Thus, our decision-making is the result of a careful balancing of these different considerations.
While not a requirement for determining whether the economy is in recession or not, it is possible to take the dating of recessions one step further and provide additional information on severity. We do this by grouping recessions into five categories with 1 being the mildest and 5 being the most severe. Category 1 recessions are characterized by only a short, mild, but broad drop in real GDP and no decline in quarterly employment; in Category 2 recessions, there is a decline in real GDP similar to that in Category 1, but enough to elicit a drop in employment, often because the contraction is quite severe in some sectors of the economy; in Category 3 recessions, there is a longer, and usually more marked, broad decline in real GDP, often followed by a drop in employment; in Category 4 recessions, there is a substantial decline across many sectors in both real GDP and employment, usually for a period of about a year or longer; and Category 5 recessions involve extremely rapid and widespread contractions in both GDP and employment over an extended period of time.
Cross, Philip. 2004. “A Diffusion Index for GDP.” Canadian Economic Observer vol. 17, no.5.
Cross, Philip, and Philippe Bergevin. 2012. Turning Points: Business Cycles in Canada since 1926. C.D. Howe Institute Commentary No. 366. October.
Kronick, Jeremy. 2016. "Taking the Economic Pulse: An Improved Tool to Help Track Economic Cycles in Canada." C.D. Howe Institute E-Brief. July.