The variable names for the percentage errors ($a, b, c$) happen to be the same as the exponents in this specific problem. Don't get confused! Usually, the formula is:
$%P = (\text{power}_x)(%x) + (\text{power}_y)(%y) + \dots$
Here, power is 'a' and error is 'a', so we get $a \cdot a = a^2$.