By Philip Cross, August 7, 2019
Do we really need a census every five years?
A recurring news story this year is President Donald Trump’s attempt to add a question about citizenship to the 2020 US census. Most news reports have noted the US has a census every 10 years, as its constitution mandates. But Canadian commentators have not raised the obvious question of why we go to the expense and trouble of conducting a full census every five years. Our more frequent census is emblematic of our penchant for costly government interventions and also shows how easily a transitory government program becomes a permanent fixture without any close scrutiny of its benefits.
Before 1986, Statistics Canada conducted a full census every decade, as in the US. Long-form questions supplementing a population count every five years were mandated by the Constitution for some western provinces since 1906 and were extended to the whole country in 1956 (the first example of mission creep in the census). However, after the severe 1981-1982 recession, the Liberal government was persuaded that a special long census was needed to identify and measure supposed structural changes resulting from the recession. It was scheduled for 1986. Though the US economic downturn was just as severe as Canada’s, nobody there thought a census was needed to understand its consequences.
After the landslide victory of Brian Mulroney’s Progressive Conservatives in 1984, the new government shocked Statcan by accepting a proposal to cut the 1986 census. After frantic lobbying by the same academic and government groups that surfaced to defend the 2011 census (with the notable exception of the business community, which increasingly uses its own data in place of the census), the 1986 census was reinstated with the important proviso that Statcan pay for it out of its existing budget. This was accomplished with a creative mixture of program cuts, user fees and funding from other departments, a useful lesson on how government departments benefit from being prodded to regularly rethink their fundamental goals and methods. This prod, however, is usually external — the bureaucratic reflex is to remain with the routine and the habitual.
The 1986 census ultimately proceeded as planned. But its proponents have never demonstrated how it provided insights into the recession’s impact that were not available from existing data sources. The 1981-1982 recession was certainly a watershed, especially for economic policy, which shifted to high interest rates to subdue persistently high inflation. The desire for a free trade deal with the US was stoked by the vulnerability of our manufacturing exports during the recession, while, more broadly, the global recession spurred the retrenchment of government intervention in the economy and the privatization of government assets spearheaded by Margaret Thatcher and Ronald Reagan. But none of these seismic economic changes originated in findings from the census.
About the only lasting legacy of the 1986 census has been more frequent censuses. Since that year full censuses have been conducted every five years, not every 10 years as in many other countries. This represents a major commitment of resources: about $700 million per census. As Statcan likes to boast, the census is the largest non-military operation a Canadian government conducts. Planning it requires a minimum of seven years, so work on the next one begins before the current one is complete. This means senior management at Statcan is constantly preoccupied with census matters. Careers literally are built on the never-ending work the census generates, which of course is one reason Statcan is its strongest advocate.
No one ever asked whether the 1986 census was worth the expense, just as no one questions how Canada benefits from permanent quinquennial rather than decennial censuses. The true total cost is not just the government resources it absorbs that could be spent elsewhere or returned to taxpayers. It also includes the never-measured response burden on Canadians, especially on the one in four unlucky enough to receive the long form, which now requires nearly 100 responses. The shift to a full census every five years is a textbook example of how government programs perpetuate themselves without anyone considering whether the original rationale still applies or, for that matter, ever was valid.
In retrospect, the whole controversy around the replacement of the long-form portion of the 2011 census with the voluntary National Household Survey was a missed opportunity to ask whether a decennial census isn’t enough to meet the information needs of government planners and academics. Even if the 2011 survey results weren’t useful the information gap between the 2006 and 2016 censuses isn’t large. The burden should be on census advocates to show how policymaking was hampered by not having a census in 2011 and why Canada shouldn’t return to a census every 10 years, thus freeing up respondents’ time and saving taxpayers millions of dollars.
Philip Cross is the former chief economic analyst at Statistics Canada and Munk senior fellow at the Macdonald-Laurier Institute.