IMO the population polled skews the results. The analyst community is only of value if their personal models of reality are different from accrual accounting. When you are trained to model cash flows, when you are paid to produce cash flows, you will believe that cash flows measure reality.

But I have never heard a rational argument why some analyst's (or mgmt) $$ for maintenance capX is a better measure of long-run requirements than the accountant's depreciation. Certainly lumpy capX from one particular year is not valid for any valuation that is supposed to see into the future.

I disagree completely with the author's claim that it is accrual accounting that "relies on abstraction and human judgment". Accrual accounting uses actual transaction prices and uses judgement only to allocate the transactions over time. But over the long run those reported rev/exp will always add up exactly to the sum of the actual transactions.

Contrast that with cash models used by analysts where the $$ input are invented with no necessity that they get validated/corrected by actual transactions - ever. If the analysts' estimates of maintenance capX were booked in the accounts instead of depreciation, the size of assets on the balance sheet would balloon into improbably high values, that would only correct when the asset was sold/chucked out.

I agree with the first comment above about today's prevalence of mark-to-market accounting (which I hate). It has been my impression that the movement toward this non-transaction accounting has been BECAUSE of lobbying by the financial analysts community. I believe they have done us a great dis-service.