David Rosenberg, chief economists at Gluskin Sheff, has the following interesting graph:

The graph (upper left corner), shows a value R = 82%. The title of the graph refers to an 82% correlation. Both can not be right.

The correlation is R-squared. If the correlation is 82%, then the graph is mislabeled. If the graph label is correct, then R-squared = 67% = correlation and the title is wrong.

Note: I classify R-squared 80% to 90% as good correlation and 60% to 80% as fair correlation. You are welcome to contribute your own characterizations.

Other characterizations:

1. 90% to 100% is excellent. Of course, 100% is perfect correlation.

2. 40% to 50% is weak.

3. 20% to 40% is poor.

4. Below 20% is no significant correlation. Of course, 0% is random scatter, exactly zero correlation.**Disclosure: **No stocks mentioned.