A growing body of empirical work shows that social recognition can meaningfully influence individuals’ choices. This paper investigates whether social recognition is a socially efficient lever for influencing individuals’ decisions. Theoretically, we show that whether social recognition is more efficient than financial incentives depends on the shape of the social recognition utility function, which determines whether social signaling is a positive-sum, negative-sum, or zero-sum game. To quantify the theoretical findings, we report results from a novel field experiment on promoting attendance to the YMCA. We quantify the impact of social recognition on attendance and elicit people’s willingness to pay for social recognition in different future states of the world. We find that social recognition increases YMCA attendance by 23% over a one-month period. We then estimate people’s willingness to pay to be recognized at varying levels of attendance, and find that social signaling is a moderately positive-sum game. Unlike previous attempts at evaluating “nudges”, we are able to generate our welfare estimates using non-parametric methods that do not rely on the assumption that individuals hold rational expectations about their future behavior. Our experimental framework provides a general template for robustly evaluating the welfare effects of non-financial policy levers.