People interact face-to-face on a frequent basis if (i) they live nearby and (ii) make the choice to meet. The first constitutes an availability of social ties; the second a propensity to interact with those ties. Despite being distinct social processes, most large-scale human interaction studies overlook these separate influences. Here, we study trends of interaction, availability, and propensity across US cities for a critical, abundant, and understudied type of social tie: extended family that live locally in separate households. We observe a systematic decline in interactions as a function of city population, which we attribute to decreased non-coresident local family availability. In contrast, interaction propensity and duration are either independent of or increase with city population. The large-scale patterns of availability and interaction propensity we discover, derived from analyzing the American Time Use Survey and Pew Social Trends Survey data, unveil previously-unknown effects on several social processes such as the effectiveness of pandemic-related social interventions, drivers affecting residential choice, and the ability of kin to provide care to family.