When I was a graduate student, I was given to understand that the reason why so many Canadian economists worked with American data was that US-based journals weren’t interested in publishing work done using non-US data. This impression was of course completely wrong. If anything, US journals (and their referees!) seem to have a prejudice in favour of work that isn’t based on well-scoured US data.
The problem is getting hold of the Canadian data in the first place. Early on in my career (before the www), I learned that if you called the BEA or the BLS for data, you got it. When you called StatsCan, you got a price list.
This distinction persists to this day. If you go to the BEA website, you can get (say) the time series for US GDP at the price of a couple of mouse clicks. But if you go to StatsCan’s CANSIM page, you’ll find that the Canadian equivalent will cost you $3.00. Curiously enough, StatsCan also offers – for our convenience – US GDP data, also for the price of $3.00. (Question: why isn’t StatsCan hauling the BEA before the NAFTA tribunal? It’s as clear a case of dumping as we’ll ever see.)
Of course, for those of us who work at places such as universities or other institutions who are willing to pay for the CANSIM subscription, this doesn’t matter very much. But for those of us who are interested in raising the level at which the discussion of economic issues takes place in Canada, it matters a great deal. As it is, there aren’t enough people who would take the time to check out factoids that don’t pass the sniff test (example), so making them pay for it is just making a bad situation worse.
I spent a couple of weeks trying to get some sort of official word from StatsCan explaining why they do things this way. Here’s the official policy:
From the Policy on Dissemination, Communication and Marketing Policies: Statistics Canada provides information of broad interest to the public free of charge and recovers the costs of providing specialized information service in compliance with federal government legislation and policies on communications and external charging. The Agency strives to provide Canadians with broad, free access to basic statistical information from all of its programs. The expanding content on the web site, free electronic publications and the DSP are mechanism through which this objective is being met. However, for more detailed and specialised data needs, Statistics Canada offers specialized products and services on a cost-recovery basis to those who require them. Cost-recovery ensures that the special needs of smaller groups of users are fulfilled without subsidization by other taxpayers. Cost recovery also allows the Agency to expand its line of products and services by requiring them to be self-funding. As a result, cost-recovery imposes a rigorous discipline on the development of products and services: They must recover their costs, which in turn requires that they find a market of sufficient size to generate the revenues offsetting the costs.
(It is possible to get free GDP data – unfortunately, it only goes back to mid-2005.)
Clearly, the marginal cost of supplying GDP data is zero: the numbers are already there in the database. So why are they using average cost pricing?
Worthwhile Canadian Initiative 
I’ve often wondered the same thing. One time when I was away from Guelph while preparing some graphs for lectures I found that I’d have to pay to download the annual Canadian unemployment rates. So I downloaded the Canadian numbers free from a US government site – the Economic Report of the President, as I recall. It’s really bad in health economics – I regularly get offers of free downloads of very detailed US data from US government sites, whereas I’d have to pay a pretty hefty price for similar Canadian data, if I could access it at all.
The policy goes back to the fiscal contraction in the early 1990s, when the federal government cut budgets for all departments and agencies and asked them to find their own sources of revenue. It was an idiotic decision at the time, since if data are not public goods, then what is? But it has continued, in part, because it gives StatsCan researchers exclusive rights over the richest, most promising datasets. It’s hard to know whether this is skewing results are just slowing down research, but it is a bad thing either way.