Section 2: Feature article

View the most recent version.

Are monthly changes in the economy becoming more volatile?

by Philip Cross ¹ 

Over the summer of 2008, analysts were surprised by the monthly movements in GDP and employment. In particular, real GDP jumped 0.7% in July, when the consensus forecast called for only 0.2% growth. The record-breaking hike of 107,000 jobs in September mocked the consensus forecast of a 10,000 increase. This surge was soon followed by a 71,000 drop in November. As one analyst noted about the November retreat in jobs, the “employment report blows expectations (yet again) out of the water.” ²  One journalist commented that the report ‘’was just the latest in a string of monthly job numbers that made monkeys out of the country’s best economists.’’ ³ 

Last summer’s divergence between expectations and the outcome for some monthly data may not be just an isolated event. It could reflect structural shifts in the Canadian economy toward industries whose monthly output is inherently more unstable. These industries include the primary sector, utilities and construction, all of which have pronounced volatility because much of the work is done outdoors or is more prone to mishaps or maintenance than most services. For example, energy led the upsurge in output in July as supply disruptions were resolved, while farming drove a gain in hours worked in September.

The volatility of some recent monthly movements in the economy is all the more surprising given the stability of annual growth since 2002. From 2003 to 2007, real GDP rose by 3.1%, 2.9%, 3.1% and 2.7%, the steadiest 4-year period on record since 1961. Hours worked showed a similar stability over that period.

Given the greater stability of the macro-economy on an annual basis (before the turmoil in 2008), this article looks at whether monthly changes are becoming less or more variable. The paper then documents whether the economy is shifting to industries that are more volatile on a monthly basis, and whether the gyrations within these industries have grown over time. It concludes by looking at the variability of other data, notably jobs, housing starts and manufacturing sales.

GDP

Comparing the standard deviation of monthly growth in real GDP before and after 2002 shows no change in volatility, despite the benign pattern of annual growth after 2002. ⁴  A small drop in volatility in the goods sector was offset by a slight increase for services, although the latter remains much less irregular than goods. There were marked increases in volatility accompanying specific disruptions to the global economy both before and after 2002, notably the Asian crisis in 1998, the 9/11 attacks in 2001, the SARS outbreak and Ontario’s blackout in 2003 and the global credit crisis in 2008 (Figure 1).

On top of these specific events, the boom in natural resources and construction starting in 2003 reversed the long-term downward trend in their importance in GDP. From 1981 to 2002, constant dollar GDP in construction and primary industries rose by 39.4%, trailing total output, which expanded 77.2% over the same period. From 2002 to 2007, however, this was reversed as resources and construction increased output 16.6% versus 13.9% for the rest of GDP.

The greater variability of output in resources and construction implies that their relative decline as a share of the economy helped stabilize overall GDP from 1981 to 2002. Since then, their faster growth has had the effect of raising the volatility of total GDP. The increase has been mitigated by lower instability in construction.

The standard deviation of monthly GDP growth remained above average in all the primary industries and construction. Compared with a standard deviation of just 0.34 for total GDP growth from January 2003 to August 2008, the standard deviations ranged from 0.42 for construction to 4.22 for non-metallic minerals. Metal ores, oil and gas, and farming and fishing also had heightened volatility, with standard deviations of over 1.0.

Interestingly, most primary industries have shown little change in volatility over time. The one exception was non-metallic minerals, where the rapid growth of diamond mining heightened its monthly variability due to the unevenness of ore concentrations in mines. Conversely, construction saw a decline in variability from the previous decade, reflecting the absence of a boom-bust cycle since 2002 and a trend towards fewer strikes. House sales rose every year from 2002 to 2007, while time lost to labour disputes fell in half between the 1990s and the 2000s.

Some of the unrelieved variability of monthly GDP reflects seasonal fluctuations in the Canadian economy that have become more pronounced in recent years. In 2002, the seasonal factors for monthly GDP ranged from a low of 94.5 (in January) to a high of 105.1 (in September), with a standard deviation of 0.030. By 2007, seasonal factors ranged from 94.3 to 106.5 with a standard deviation of 0.0322. This increase may reflect the growing weight of seasonal industries like resources and construction as well as more pronounced seasonality in other sectors (increased spending at Christmas, for example, until the proliferation of gift cards after 2006 shifted sales into the new year).

One way for analysts to minimize the impact of volatile monthly data movements is to put them in a longer-term context. Most large monthly changes in output are caused by short-term supply disruptions, which are usually soon reversed. The best example was Ontario’s electrical blackout in August 2003, which reduced GDP 0.9% only to recover 1.3% in September. Other notable examples include the ice storm in January 1998 and the retooling of auto output in December 2007.

Simply averaging growth over two months, for example, lowers the standard deviation of GDP growth by about a third, from 0.34 to 0.23 after 2002. ⁵  This is because most large monthly movements are often reversed the next month: for example, the 0.7% hike in July 2008 was followed by a 0.5% drop in August.

The standard deviation of the 12-month growth rate of real GDP is even smoother, having fallen (even as monthly volatility has not) from 1.76 before 2002 to 0.84 thereafter. ⁶  And the deviation of year-over-year growth is small compared with 0.34 for monthly changes, despite the much higher values for annual growth (where increases of 3% or more are common, while monthly changes of 0.5% or more are rare).

Employment

While monthly changes in output remain as volatile as in the previous decade, monthly job growth has become slightly more steady. The standard deviation of monthly growth in LFS employment fell to 0.17 after 2002 from 0.20 between 1992 and 2002. ⁷ 

The drop in the volatility of monthly employment changes was most pronounced in the primary sector, where the standard deviation fell from 0.76 to 0.65 after 2002. This decline largely reflects the shift from farming to other primary industries over this period, as the volatility within each of these sub-sectors was essentially unchanged over time. Mining drove the shift in primary jobs from farming to other primary industries, especially in the oil and gas industry.

The drop in the volatility of jobs even as it was unchanged for output coincided with a pronounced shortage of labour after 2002, especially in the mining industry. These shortages may have led employers to keep workers even as output fluctuated in the short-term, rather than risk losing them to other firms after a layoff. This raises the possibility that the increased stability of employment may be ephemeral as the economy turns down, as demonstrated by the sharp drop in jobs in November and December 2008.

Housing starts

Housing has always been one of the more volatile sectors of the economy. This reflects both its sensitivity to the overall business cycle and its susceptibility to weather fluctuations.

Since 2003, housing has shifted significantly from single-family homes to multiple units (especially condos and retirement homes). A decade ago, singles accounted for 60% of housing starts in Canada, and remained high at 57% in 2002. But their share has fallen steadily to 40% so far in 2008. This decline reflects a small drop in ground-breaking on new homes, and a large increase in starts of multiple units.

The shift to multiple units by itself increases the volatility of housing construction. After 2002, the standard deviation of monthly growth in multiple units was 27.8, more than three times the 9.0 for single-family dwellings. This shift to a more volatile sector, however, has been outweighed by a drop in the variability of singles by a half. This reflects the sustained boom in Canada’s housing market after 2002, as well as less severe winter weather in some parts of the country (at least until 2007). One reason for the greater stability of starts of single homes than multiples before and after 2002 is that they usually require pre-purchase, while there is a larger speculative component for condos. As well, taller buildings are subject to stricter safety rules during high winds, which force more shutdowns.

Manufacturing

The monthly volatility of nominal manufacturing sales has been little changed since 1992, despite the structural shifts induced by rising input costs and the soaring dollar after 2002. The standard deviation of monthly shipments growth has been steady at about 2.0 for the periods before and after 2002.

The overall stability in manufacturing reflects offsetting structural shifts. Since 2002, manufacturing sales growth has been driven by aerospace, primary metals and petroleum refining. These industries are more volatile than the manufacturing average, with a standard deviation of 25.1 for aerospace and 3.0 for metals and petroleum. Moreover, their volatility increased after 2002, partly a refection of the growing role price fluctuations play in monthly movements.

But even as manufacturing activity has shifted to some industries with above-average volatility, this has been offset by the declining importance of other industries also showing above-average instability. This is particularly true of autos and clothing and related industries. Meanwhile, manufacturers in the consumer and capital goods sectors, which have the lowest volatility, have also seen their importance grow in overall manufacturing.

Conclusion

This paper found that monthly changes in output have not become more stable since 2003, despite the steadiest annual growth rates on record back to 1961. While cyclical changes in the economy have become smoother from 2002 to 2007, a combination of structural changes have maintained the volatility of monthly changes and increased the seasonality of our economy.

The variability of monthly output has implications for how economists conduct their analysis and how the media report on these commentaries. It suggests that monthly changes need to be put in the context of recent trends. Simply averaging growth over two months reduces the volatility of growth by a third.

Next | Previous