In Introduction to Fixed Income Analytics by Frank Fabozzi, p. 41, there is an example how to calculate the theoretical spot rate of a 1.5 year treasury bond with a 3.5% annual interest and semiannual coupons.
\begin{array} {|c|c|c|c|c|} \hline \text{Period} & \text{Years} & \substack{\text{Annual Yield to}\\\text{Maturity (BEY) (%)}} & \text{Price} & \substack{\text{Spot Rate}\\\text{(BEY) (%)}}\\ \hline 1 & 0.5 & 3.00 &-& 3.0000 \\ 2 & 1.0 & 3.30 &-& 3.3000 \\ 3 & 1.5 & 3.50 &100.00& \text{?}\\ \hline \end{array}
The first two rows are the zero-coupon treasury bills.
The cash flow of the 1.5 year treasury bond is obviously:
0.5 year: 0.035 × \$100 × 0.5 = \$1.75
1.0 year: 0.035 × \$100 × 0.5 = \$1.75
1.5 year: 0.035 × \$100 × 0.5 + 100 = \$101.75
He now claims that the present value of the cash flows is:
$$\mathrm{PV}(z_1, z_2, z_3) = \frac{1.75}{(1+z_1)^1} + \frac{1.75}{(1+z_2)^2} + \frac{101.75}{(1+z_3)^3}$$ where
$z_1 =$ one-half the annualized 6-month theoretical spot rate
$z_2 =$ one-half the annualized 1-year theoretical spot rate
$z_3 =$ one-half the annualized 1.5-year theoretical spot rate.
If we solve $\mathrm{PV}(3.00, 3.30, z_3) = 100$ for $z_3$, we are supposed to get the theoretical spot rate for the 1.5 year treasury bond as described above.
But why can he do the whole example on semiannual intervals? The 1-year treasury bill does not pay any semiannual coupons, right? Where does that compounding come from?
Why isn't it: $$\mathrm{PV}(z_1, z_2, z_3) = \frac{1.75}{1+z_1} + \frac{1.75}{2\cdot z_2} + \frac{101.75}{(1+z_3)^3}\; ?$$
OK, BEY is given as effective semiannual rate. The convention is simply defined that way. So the question has been resolved.