Earlier this summer, Canada’s minister of families, children and social development, Jean-Yves Duclos, predicted the new Canada Child Benefit (CCB) would reduce child poverty in Canada by 40%, bumping 284,000 children above the poverty line between 2013 and 2017 (using the LICO-AT version of poverty). This was similar to the projected impact of the CCB in the 2015 Liberal election platform of 315,000 children.
Now I like the CCB. It is certainly a far more progressive program than what it replaced on July 1 this year. There is no doubt the new program will have an impact on child poverty as a result of its increased payouts, and 40% is an impressive number. However, once I dove into the details, the CCB looked more good than great.
In some ways this is a battle of tax models. Statistics Canada maintains a complex simulator of the entire Canadian tax/transfer system called the Social Policy Simulation Database and Model (SPSD/M). It is public, free and available to all. Employment and Social Development Canada maintains its own parallel model that does the same thing starting from the same 2010 data, but it is private, internal and not available to all, making its conclusions difficult to verify.
From working quite a bit with SPSD/M, I can say its projections on child poverty rates were not stellar up to 2013. For example, SPSD/M is wearing rose-coloured glasses when it estimates a rate of only 8.6% in 2013 instead of the actual value of 11.2%. So whenever I use SPSD/M to estimate the impact of a program like the CCB, I make sure I am not comparing the actuals (11.2% child poverty) to a projection (8.6%), as the government could have done nothing and it would still look like success (due simply to a modelling error).
Child poverty rates (LICO-AT)
|2013||2014||2017 pre-CCB||2017 post-CCB|
|Modelled by SPSD/M||8.6%||8.7%||7.6%||6.0%|
|Modelled by ESDC||unknown||unknown||Unknown||6.7%|
As you can see in the table, the Statscan and ESDC models produce similar post-CCB poverty rates in 2017 of 6.0% and 6.7% respectively, which is either 338,000 or 284,000 children lifted out of poverty (on a LICO-AT basis). Not a perfect match, but pretty similar given totally different models.
The kicker, though, is that we have no idea what the ESDC model’s starting point was. Like SPSD/M, does it underestimate child poverty from the get-go, making any reduction mostly due to modelling error? There’s really no way to know, since it’s a completely internal model, not accessible to outsiders like myself.
If you use a similar approach but based on the SPSD/M, you see a 46% decline in child poverty between the 2013 actuals (11.2%) and the 2017 post-CCB rate (6.0%). Of those 46 points, 32 are from underestimating the extent of child poverty to begin with and 14 due to the actual impact of the CCB in 2017.
That shows the program having an impact for sure, but it is much more limited in scope when the CCB is specifically isolated like that. There is a difference, in other words, between saying “child poverty will decline by 40% between 2013 and 2017,” and “because of the CCB, child poverty will decline by 40% between 2013 and 2017.”
After everyone did their analysis using the 2013 actuals, it turns out that child poverty dropped fairly dramatically in 2014—from 11.2% to 8.5% (the data quality for child poverty isn’t stellar, so you’d expect these figures to bounce around). Against the 2014 data, the SPSD/M projections are bang on, if slightly pessimistic this time.
So maybe we will see something close to a 40% decline in child poverty by 2017 (I certainly hope so). But it’s important to note that all of that drop isn’t from the CCB, which obviously did not even come into effect until 2016.
In any case, I may be totally wrong on my SPSD/M projections and ESDC totally right. We’ll all find out in the summer of 2019 when the poverty data from 2017 becomes available.
David Macdonald is a senior economist at the Canadian Centre for Policy Alternatives. Follow him on Twitter @DavidMacCdn.