The census long form - A critical tool for better understanding and policy

The following was written by Russell Davidson, Charles Beach and Thomas Lemieux on behalf of the Canadian Economics Association. A condensed version appeared in Saturday's Globe and Mail.

With a bill to reinstate the mandatory Census long form currently before Parliament, we wish to express our views, on behalf of the executive council of the Canadian Economics Association, on the importance of reinstating a mandatory long form for the 2016 Census.

What the mandatory long form offered was a large sample and good coverage of detailed local neighbourhood areas in combination with household income and workers' labour market involvement and earnings data. By good coverage is meant high response rates (93.5 percent in the 2006 long form). The impacts of the government dropping the long form from the 2011 Census can be seen in four broad areas.

First, historical comparability means that researchers can use the Census to analyze long-run economic, geographical and social trends in Canada, and to make comparisons to U.S. Census-based patterns and results for other countries. For example, studies of income distribution (ie., who gets what) and inequality and how these have changed over time are inhibited because results from the replacement vehicle, the new National Household Survey (NHS) are not comparable to previous Census results up to 2006.

Not only is the response rate on the voluntary NHS down to 68.6 percent, but this response rate is not uniform across the income distribution. It is well known that lower response rates are more common among low-income and high-income households (as well as many rural areas). If someone doesn't respond, they are not included in the Census data. If non-responses shoot up in the 2011 Census at the lower and upper ends of the income distribution, and you want to look at, say, what has happened to the middle class or to middle-class workers in the labour market between 2006 and 2011, you can't tell whether the changes are due to a continuation of fundamental trends in the Canadian economy or simply due to the change in Statcan methodology and the dramatic rise of non-response rates at the two ends of the income distribution. We don't know on a comparable basis what has happened here since the 2008-09 Great Recession, and have to look south to U.S data to get a better understanding. Similarly, we can't tell the extent to which middle-class jobs are losing out to low-paid and high-paid employment or formally test what underlying factors may be driving these major on-going changes to the Canadian economy (eg., IT-based technological change, off-shoring and globalization, or various policy/structural changes in the Canadian labour market). If we don't know what is happening, who the winners and losers are, and what is causing such changes, governments cannot develop appropriate policies to deal with such issues (see Barrie McKenna's column, January 30: "Statcan forces decision-makers to drive blindfolded").

The second area of impact of eliminating the Census long form in 2010 has been on the coverage of low-income recipients and hence on the analysis of poverty rates - the breadth, severity and location of low-income households and the unemployed - again because of unreliable coverage and low response rates. These include the most vulnerable portions of society who are most dependent on key government services within their communities. This has been an especial concern since the Great Recession of 2008-09, and yet research and better understanding of the situation of the have-nots in the economy have been limited. As has often been said, if you can't measure it, it is hard to formulate good policy to address such economic and social concerns.

The third area is the lack of coverage of household income and labour market involvement at the local neighbourhood level for policy and decision-making purposes. Because of the large size of the Census, the long form provides reliable and consistent fine geographical detail for local policy analysis and for the best provision of local government services. For example, regional health authorities need to know which neighbourhoods are likely to have the greatest needs for certain health resources. Schools, local transportation, and hospital planning all depend upon such information for their planning purposes. Where does one locate offices for government services such as EI, income assistance, or immigrant settlement services across a large region or urban area? The recent article by Tavia Grant (January 29: "Damage from cancelled census as bad as feared, researchers say") documents the planning changes and additional costs faced by municipal authorities and the private sector as a result of their no longer being able to rely on long-form local community data.

Fourth, the loss of the long form can have a long-term effect on data quality from major Statistics Canada surveys. The Labour Force Survey, for example --- the source of critical statistics on unemployment rates and changes in employment levels --- is a survey that periodically adjusts its methodology to reflect benchmark coverage (and hence response rates) in the Census in order to make the LFS results representative and accurate of the Canadian population as a whole. But in the absence of a mandatory long form and with the low and uneven response rates of the replacement NHS, this adjustment becomes less reliable. So, if there are concerns about the reliability of recent LFS results (see again McKenna's column), these concerns may well be set to rise.

We urge members of Parliament of all parties to consider these serious concerns about the impact of not having a Canadian Census long form on the bill to reinstate the long form currently before Parliament.

 

Russell Davidson, Professor of Economics, McGill University, and President of the Canadian Economics Association
Charles Beach, Professor of Economics Emeritus, Queen's University, and President-Elect of the Canadian Economics Association
Thomas Lemieux, Professor of Economics, University of British Columbia, and Past-President of the Canadian Economics Association.

Comments

[Avon Barksdale]

John Cochrane has an interesting take on this issue:

"For the moment, the Federal Government does have a unique role in creating and supplying economic data. We can't study what we can't measure. This really is a public good, reasonably well created managed, and starved for resources. But most of our data sources are decades old, and have not been adequately re-thought or expanded in that time. Especially with the internet, there is more and more private collection and supply of data, but for the moment it cannot supplant the Federal government...Economics is not, yet, "big science" requiring massive infrastructure to produce research. Economic data collection is "big," and best directed by researchers not government officials. Data can be sold, so it's not a pure public good. But I'm willing to go with the idea that not enough good data is produced."

He goes on to point out that a lot of economic research is just bad and that subsidies should go to data collection and not analysis. More than enough research is being done. Is it true that not enough important research in economics would get done without direct government subsidies? I doubt it. But we all suffer without data. Cochrane continues,

"But much of the federal research subsidy to economists does not go to creating new, publicly useful data sets. So, I think we can agree on research support for researchers to produce new data, but we don't need support to analyze that data."

The full post is at, http://johnhcochrane.blogspot.com.au/2012/08/subsidies-for-economists.html

[rjs]

Wondering whether the Canadian Government will make the results of the long form data (including PUMS) freely available. If not, screw em -- Canadians should demand to be paid to answer those questions if the government is going to charge them to view the results.