The second equation is the first equation, post #1, just the += expanded.
So yes, not feasible, except by, as MATLAB complained, adding in all paths a unit delay.
Which I did and got it to solve. But then, and here is where I am over my head, I think it
just causes group delay, but magnitude response still accurate......not sure.
The above is why I recommended to OP to do a recursive solution in Excel to look at samples
generated, and hence plots.
Makes me wonder if there is a typo here from OP or his source.