I am looking for a reference for the identity $$\chi_\lambda(C)=\frac{\dim(V_\lambda)}{|C|}\sum_{p\in P_\lambda,\,q\in Q_\lambda,\,pq\in C}\operatorname{sgn}(q)$$ for the irreducible characters of the symmetric group. It seems quite different from the two results normally seen in textbooks: the Frobenius character formula and the Murnaghan-Nakayama rule. It's a lot easier to prove than either of these, so I am sure it must be known.

Let me first explain the standard notation used in the identity and then indicate how I derived it.

The irreducible character $\chi_\lambda$ is labelled by a partition $\lambda$ of $n$ and $C$ is a conjugacy class in $\mathrm S_n$. The corresponding representation space $V_\lambda$ can be realized as the left ideal $\mathbb C[\mathrm S_n]c_\lambda$ in the group algebra. Here, $$c_\lambda=\sum_{p\in P_\lambda,\,q\in Q_\lambda}\operatorname{sgn}(q)pq $$ is the Young symmetrizer. The subgroups $P_\lambda$ and $Q_\lambda$ in $\mathrm S_n$ fix, respectively, the rows and columns in the canonical tableau of shape $\lambda$. The dimension $\dim(V_\lambda)$ is known from the hook length formula.

The idea is that, up to normalization, $\chi_\lambda$ is the intertwining projection of $c_\lambda$ onto the space of class functions. For general finite groups, we have two expressions for this projection: \begin{equation}\label{p}(P\phi)(g)=\frac{1}{|G|}\sum_{h\in G}\phi(hgh^{-1})=\sum_{\chi\in\operatorname{Irr}(G)}\langle \phi,\chi\rangle \chi(g), \end{equation} where $$ \langle \phi,\psi\rangle =\frac{1}{|G|}\sum_{h\in G}\phi(h)\overline{\psi(h)}.$$ I take $\phi=c_\lambda$ considered as a function on the group, that is, $c_\lambda=\sum_g c_\lambda (g)g$. It follows from well-known facts that $\langle c_\lambda,\chi_\mu\rangle=\delta_{\lambda\mu}/\dim(V_\lambda)$, so \begin{align*}\chi_\lambda(g)&=\frac{\dim(V_\lambda)}{n!} \sum_{h\in \mathrm S_n}c_\lambda(h gh^{-1})\\ &=\frac{\dim(V_\lambda)}{n!} \sum_{h\in \mathrm S_n,\,p\in P_\lambda,\,q\in Q_\lambda,\,hpqh^{-1}=g}\operatorname{sgn}(q). \end{align*} Using the stabilizer-orbit theorem applied to the adjoint action, this can be written as above.