Quadratic Mean-Field BSDEs and Exponential Utility Maximization (2511.17214v1)

Abstract: We study real-valued mean-field backward stochastic differential equations (BSDEs) of the form [ Y_t = ξ+ \int_t^T \widetilde{\mathbb E}\, g(s, Z_s, \tilde Z_s)\, ds - \int_t^T Z_s \, dW_s, ] where $\tilde Z$ denotes an independent copy of $Z$ and $\widetilde{\mathbb E}$ the expectation with respect to $\tilde Z$. Under a \emph{separately quadratic} growth assumption (H\textsubscript{q}) on the generator $g$ in $(Z,\tilde Z)$, together with a bounded terminal condition, we prove existence and uniqueness of solutions in $\mathbb S^\infty \times \mathbb H^2_{\mathrm{BMO}}$. Our approach departs from classical fixed-point arguments and instead combines Malliavin calculus with refined BMO and stability estimates: we first obtain uniform $\mathbb S^\infty$-bounds for $Z$ in a Lipschitz setting with bounded Malliavin derivative, and then pass to the quadratic case by approximation, using a stability result in $\mathbb S^2\times\mathbb H^2$. This closes the gap between the quadratic BSDE results of Cheridito and Nam (2017) and Hao et al. (2025). In the second part of the paper, we extend the framework to generators of the form $g(t,z,\tilde z,\bar z)$ satisfying a \emph{fully coupled quadratic} condition (H\textsubscript{q}$'$). In this general regime we establish existence and uniqueness under a smallness condition on the centered terminal variable. As an application, we solve a mean-field exponential utility maximization problem with a collective liability, thereby generalizing the classical utility maximization framework of Hu et al. (2005) to a fully coupled quadratic mean-field setting.