The model has been widely discredited by academics because it mixes apples and oranges. Treasuries are nominal assets whereas stocks are real assets (their returns are a function of inflation). The fact that these two series seem to move together doesn't prove that they have predictive ability over the mispricing of stocks compared to bonds. For a great discussion of the Fed Model, see Cliff Asness's article "Fight the Fed Model" I've also discussed this issue in my 2002 JFQA paper with Jay Ritter.
So why bring all this up today? Well, my favorite finance blogger, Felix Salmon appears to be making the Fed Model mistake all over again, long after I thought that the model was dead and buried. Felix doesn't refer to it as the Fed Model directly, but his presumption is that Treasury yields and earnings yields should track each other. After 2002 they don't.
While it looks like you'll earn a higher yield on stocks rather than bonds, this is very misleading. First earnings are not cash flows - something that we hammer home at the beginning of any stock valuation class. Second, whereas bond cash flows are pretty secure, the cash flows from stocks are much more uncertain - the risk premium is zero for bonds and who knows what stocks! Finally, the bond cash flows are fixed in nominal terms, while the stock cash flows are real - they change with inflation.
Again: apples to oranges.