The SCC’s “Most Cited”

Recently CanLII announced that they have implemented a new sorting order for the results of searches. Now along with the ability to sort by relevance and date, you can sort by “most cited.” I’ve been interested for quite a while in looking at Supreme Court of Canada (“SCC”) judgments ranked by their impact, and this seemed like a handy tool to give that examination a start.

Accordingly, I ran a search for all SCC judgments in the CanLII database, producing 4160 results. The CanLII SCC database contains all SCC reported cases from January 1, 1985 to the present and as well incomplete sets of decisions for the years 1876 to 1984.

Essentially, then, this is an examination of cases from the last two decades. I took the first 1000 in rank order according to the number of judgments that have cited each and created a simple list. That list, as a flat html file, is available on my website; the case names are linked to the CanLII report of the judgment, and the number of citations for a case is linked to the CanLII search for references to the case in question. As well, I have put the data from the 1000 cases into a database that is available online, making for a somewhat more finished presentation.

It is important to note that the algorithm CanLII uses to determine when one case cites another checks for citations (and not the name) for the referenced case in the text of the relevant database. With respect to cases citing SCC judgments, CanLII searches for at least CanLII citations (e.g. R v Nova Scotia Pharmaceutical Society, 1992 CanLII 72 (S.C.C.)), citations to the official Supreme Court Reports and the neutral citations that are widely used by courts now. However, it is clear that some citations to SCC judgments are not caught by the algorithm, particularly those that are botched in some way, such as by the use of square brackets instead of round. (See the discussion of this matter on Slaw.) Accordingly, the list of most-cited cases will be generally accurate, but not precisely so.

The most-cited SCC judgment is R v W(D), [1991] 1 SCR 742, a case on “charge to the jury” decided by a five person bench, Cory J. writing the short judgment for the majority (Cory, Gonthier, Iacobucci JJ.), with Sopinka and McLachlin JJ. dissenting. R v W(D) has been cited by 2336 cases.

The next most-cited case is Pushpanathan v Canada (Minister of Citizenship and Immigration), [1998] 1 SCR 982, referred to in 1791 cases. What is interesting to note here is that the steep drop between citations for the first fifty or so case has already begun. This descent continues, gradually leveling out in a shape that is colloquially known as the long tail and more technically as the power law graph. The graph of the citations to the 1000 cases is available via a recent post of mine to Slaw.

Let me take you quickly in leaps of a hundred through the data, giving you first the ranking of the judgment and then in parentheses the number of cases citing that judgment: 1 (2336); 100 (334); 200 (231); 300 (174); 400 (142); 500 (116); 600 (95); 700 (81); 800 (68); 900 (58); 1000 (49). A peek at the other end of the full list of 4160 shows that after the judgment ranked number 3045, R v McKay, [2007] 1 SCR 793, which has till now been cited once, no citations can be found to the remaining 1115 judgments.

This is unsurprising given that most if not all of these last cases are 19th century judgments only recently added to the database and to which, therefore, there would be few if any citations in the form congenial to the search algorithm. Given the limitations of the CanLII algorithm for finding citations, this small experiment can only be a start in what could be an interesting analysis of the influence of SCC decisions as measured by their use in other cases. It should not be too difficult to modify searches on the SCC databse in order to sweep in more of the older or “irregular” citations, to sophisticate the notion of citation and differentiate among the reasons a case gets cited, and to carve out subsets of the data according to case type.