Although the ElGamal scheme states that the message $M$ must be $1 \le M \le p-1$ in this paper: A Secure and Optimally Efficient Multi-Authority Election Scheme they propose a method where the message can be $-l \le M \le l$ where $|\ l\ | \lt p/2$. This is stated in page 9.
The idea behind the paper is that for 2-way voting system the available options are encoded as $m_0=1,\ m_1=-1$ and when doing additive ElGamal the result could possibly be $g^{-4}$ or any negative exponent.
I know that for computing the message in the final stage of ElGamal we execute a loop for each possible value of $M$ but in this case, although this is what the paper proposes as well, I don't see how it is done.