1: \begin{abstract} % 100 or fewer: condense, add sentence on application "causal estimates"

2: % Discontinuous changes in observed outcomes based on thresholds in treatment assignments can provide causal insights. In many contexts, like geographic settings, the locations of thresholds are unknown and multivariate. We introduce a non-parametric method that can estimate multivariate discontinuities with unknown location by segmenting the estimated regression surface into smooth and discontinuous parts. The estimator relies on a convex relaxation of the Mumford-Shah functional, for which we establish identification and convergence results. We use it to analyze the effects of an internet shutdown on economic activity in India, which dropped by 50\%, much more than previous estimates. As such, the increasing global prevalence of such shutdowns presents alarming implications for economies worldwide.  

3: 

4:    Discontinuities in regression functions can reveal important insights. In many contexts, like geographic settings, such discontinuities are multivariate and unknown a priori. We propose a non-parametric regression method that estimates the location and size of discontinuities by segmenting the regression surface. This estimator is based on a convex relaxation of the Mumford-Shah functional, for which we establish identification and convergence. We use it to show that an internet shutdown in India resulted in a reduction of economic activity by 25--35\%, greatly surpassing previous estimates and shedding new light on the true cost of such shutdowns for digital economies globally. \\

5: \noindent \emph{Keywords:} boun\-ded variation, cali\-bra\-tion,  free discon\-tinuity pro\-blem, $\Gamma$-convergence, internet shutdowns, Mum\-ford-Shah functional, nonparametric regression, segmentation. 

6: \noindent \\ 

7: \emph{JEL Codes: C14, C21, L86, O33}

8: 

9: 

10: 

11: \end{abstract}

12: