This is a fairly obvious point, but I don't recall seeing it discussed anywhere. When we examine the costs of a public policy, we not only need to consider the financial costs, but other opportunity costs as well. Take, for instance, the costs of an election. We not only need to consider the financial costs of the election, but we need to consider the fact that it takes time for people to get to the polling station and vote - time they could have spent doing something else.
Now what does this have to do with the long form census?
We can calculate the average time cost per person (or more accurately, per household) of the long form census (LFC) by considering the following three variables:
x = The amount of time it takes, on average, to fill out the LFC (in minutes)
y = The percentage of people who receive the LFC
z = The response rate, as a percentage of the LFC for people who receive it.
The average time cost is simply: avg[TC] = x * y * z
Under the old system we know that: y = .2 (20% of households receive it) and the response rate is around 95% so z = .95. I don't have solid data on x, so we will leave it as a variable. The average time cost, as function of x is avg[TC] = .19x. This means that if it takes, say, 30 minutes on average to fill out the LFC, the average time cost per person is .19*30 = 5.7 minutes. Not a lot of time, sure, but when we're talking about millions of households, it adds up.
Under the new system, the replacement for the LFC will be mailed to 33% of households, so y = .33. We do not know for sure what the response rate will be, but Statscan thought they might be able to get it up to 70%, so we will use z = .7. Then the average time cost, as a function of x is avg[TC] = .231x. Note this is over 23% higher than the time cost under the old system (.19x). In other words, the census changes have not just increased the financial costs to Canadians, they have also increased their time burden!
Of course, the 70% response rate is just a guess/optimistic estimate. The response rate may end up being lower. For the two systems to have an equivalent time burden, the response rate needs to be 57.6% (.7 * .576 * x= .19x). But would anyone anywhere consider a 57% response rate to the voluntary census to provide useful data?
We haven't considered the x variable, however. It could be that the long-form census will only be filled out by people who are particularly efficient at filling out the form. This would mean that the value of x under the old system is higher than the value of x under the new system. But of course, this would also mean that we're admitting that the new system provides a biased sample, and is therefore unreliable!
In other words, the changes to the census have either increased the administrative burden on Canadians or provide near worthless data or both. Increasing the time cost of the census to Canadians by over 20% should not be ignored.