Contradictions from Stats Can

Statistics Canada has been in the news lately because of an error it made in measuring the consumer price index. The agency understated inflation because it had hotel room rates decreasing when they actually were increasing. Rather than looking for an error in the data, they rationalized the strange result on the basis of how the market might work with declining visits from the U.S.

Statistics Canada has an excellent reputation amongst international organizations that monitor the reliability of national statistical agencies. Nevertheless, one of the lessons that should be learned is to double and triple check when the numbers look unusual. Another lesson is that Statistics Canada shouldn't be beyond criticism; it needs to take responsibility for the mess that has been made out of this year's census.

For an example of how the agency publishes apparently contradictory data, consider average construction wages in British Columbia. Statistics Canada publishes two surveys that measure average wages. Its Labour Force Survey (LFS) is quoted monthly as the measure for unemployment rates, but that survey also samples wage rates. The agency's Survey of Payroll Employment, Earnings and Hours (SEPH) also reports on average wages, and it has been doing so for longer. As its names suggests SEPH looks at employees (payrolls) while LFS includes the self-employed. Statistics Canada publishes dozens of tables from its LFS in a three disk CD set titled "Labour Force Historical Review".

annual data average weekly BC construction wages Annual data from the Historical Review on weekly construction wages in BC, shown in the graph, indicate that construction wages declined between 2001 and 2005. On an hourly basis, the average rate was $20.32 in 2001 and $19.97 in 2005. On a weekly basis, the rate was $827.09 in 2001 and $820.65 in 2005. It is hard to understand how construction wages would be declining, or stagnant, at a time when everyone acknowledges that BC is experiencing a construction boom which accounts for the largest share of BC's job growth.

monthly data average weekly BC construction wages The second graph compares monthly data for average weekly wages in BC construction as measured by both SEPH and LFS. The black line is the LFS numbers; the red line is the SEPH numbers. Both lines have had trend lines drawn through them using Microsoft's Excel option for fitting a polynomial. The green line is the trend through the SEPH data; the blue line is the trend through the LFS data.

According to the LFS data, average construction wages in January 2006 were 1.4% lower than in January 2001; in February 2006 they were only 0.3% higher than in February 2001, and in March 2006 they were only 0.1% higher than in March 2006. According to the SEPH data, average construction wages in January 2006 were 10.2% higher than they were in January 2001. There's a big difference between minus 1.4% and plus 10.2%! The SEPH data show wages 13.5% higher in February 2006 than in February 2001, and 11.1% higher in March 2006 than in March 2001.

The trend line for the LFS data between May 2001 and October 2006 is virtually flat, while the trend line for the SEPH data shows a steady increase. The LFS data show a sudden takeoff in construction wages between October 2005 and May 2006 (although they've dropped from $847.23 in May 2006 to $790.55 in July 2006); however, the SEPH data show a drop from $834.77 in October 2005 to $788.08 in May 2006.

There may be understandable reasons why Statistics Canada's two surveys that measure wages show apparently contradictory results. Other examples of contradictions between the two surveys are easy to find. What is hard to understand is why Statistics Canada offers no commentary on the contradictions and behaves as if its separate divisions which are responsible for the two surveys are distinct empires that don't communicate with each other. Statistics Canada needs to be held accountable. Most economists who regularly work with detailed data provided by Statistics Canada are aware of difficulties with various data series. In light of the CPI fiasco, it is time for Statistics Canada to provide a clear explanation as to why its two wage surveys indicate opposite trends.