This is a very high quality thread, unusual on this forum.
My take on this is to look at the relevant metric. And the relevant metric for each company varies.
David Einhorn gave a good paper on dangers of ROE. It can be found on Google. In essence, he is saying to make sure that E is relevant to R. He cited service based companies as examples vis a vis manufacturing companies.
CSL is worth examining in the context of flawed reliance on ROE. Its earnings are substantially derived by R & D spend. However, since CSL expensed all R & D, these spending never shows up on the balance sheet, and hence looking at ROE gives a totally wrong picture.
Of course, as cited by a few others, make sure the Return is a real return, rather than an accounting construct. Case on point- MOC, where half the "earnings" comes from book revaluation. Also potential danger here with property holding companies doing constant revaluations.
Some posters have rightfully stated that cash is a truth serum to the earnings statement.
PP has raised the example of prepaid funeral services. A further classic example is Amex when Buffett bought in. In such cases, the liability on the balance sheet is actually not a liability. Also if the company only recognises revenue when service is delivered (eg prepaid storage), and not when it receives the cash, the real earnings will be understated. Usually this is evident from cashflow which will consistently exceed revenue.
I know of at least one prominent newsletter which postulates ROE as the be-all and end-all. Amongst my circle, we call the main character Mr ROE.
