The lottery ticket hypothesis (LTH) claims that a deep neural network (i.e.,\nground network) contains a number of subnetworks (i.e., winning tickets), each\nof which exhibiting identically accurate inference capability as that of the\nground network. Federated learning (FL) has recently been applied in LotteryFL\nto discover such winning tickets in a distributed way, showing higher accuracy\nmulti-task learning than Vanilla FL. Nonetheless, LotteryFL relies on unicast\ntransmission on the downlink, and ignores mitigating stragglers, questioning\nscalability. Motivated by this, in this article we propose a personalized and\ncommunication-efficient federated lottery ticket learning algorithm, coined\nCELL, which exploits downlink broadcast for communication efficiency.\nFurthermore, it utilizes a novel user grouping method, thereby alternating\nbetween FL and lottery learning to mitigate stragglers. Numerical simulations\nvalidate that CELL achieves up to 3.6% higher personalized task classification\naccuracy with 4.3x smaller total communication cost until convergence under the\nCIFAR-10 dataset.\n