Freedom to information is what all econometricians hope for - that is, readily accessible and reliable data that you don't have to pay for.
Canada's statistical agency, Statistics Canada, has finally come into line with most of its international counterparts by announcing that starting next year much of its data will be available for free. Thank goodness for that! Finally!
Tavia Grant's post in the Globe and Mail a couple of days ago sets out some of the details very nicely. If you're a user of StatsCan data, ranging from their CANSIM collection, to Census data, this is good news. It's especially good news if you're a non-academic used.
We're fond of saying that the econometric methods we use are only as good as the data we can access. True enough, but let's not forget that free data can still be used inappropriately, and are only as good as the econometric analysis they're subjected to.
© 2011, David E. Giles
Oh dear. I was so excited about the fact that the data were going to be free that I completely forgot about the risk of it being badly used.ReplyDelete
Stephen: No free lunch, after all? :-)ReplyDelete
This has always bugged me. What doubly bugged me was how data the Bank of Canada collects had to be bought back by us. 3 dollars a series is a lot of money when you start getting into CMA/NACIS or Occupation level data.ReplyDelete
Andrew: Darn right! Where's the value-added to justfiy the cost?ReplyDelete
My own pet peeve with CANSIM is that it's very difficult in many cases to get long-run continuity of data. Series are "discontinued", or simply disappear. As an academic I get free access to this particular database, but I often feel that I'm getting what I pay for! I know I'm not alone in feeling this way.
I have had big problems with that, trying to make a wage series for construction sector. There was a change from NACIS versus SIC. Contacting them, they told me that it would take too much work to do properly and I should just splice the series my self. Kinda disconcerting.ReplyDelete
Andrew: VERY disconcerting, and a great example of some of the issues associated with CANSIM.ReplyDelete