Under what conditions are social ties salubrious? This dissertation addresses this question by examining three distinct aspects of social relationships and resources embedded in these relationships: social capital over the life course, the racial and gendered adolescent popularity effects on adult body mass index, and the relationship between network spillover and workplace food choice. While network research increasingly demonstrates that social networks manifest themselves in highly context-dependent ways, it is unclear when, where, how, and for whom people's networks become a "resource," "cost," or "nothing at all." Investigating these contexts has practical implications because sociologists and policymakers wish to understand what network interventions might more effectively improve individual well-being and reduce health inequalities. Theoretically, examining these contexts is significant because it may elucidate overlooked mechanisms that underlie the network effects for health and health inequalities. In the second chapter, I demonstrate that in contrast to prior work, which found the stability of social capital, only 16% of individuals maintained social capital throughout their lives, whereas most experienced gains or losses. Life-course social capital matters, as maintaining high or increasing social capital can improve adult general health. Further analyses demonstrate that whether people maintain, increase, decrease, or have mixed life-course social capital depends on the social positions they occupy in multiple stratification systems. In the third chapter, I show that adolescent structural prestige at the ego and structural levels are linearly associated with adult self-reported and measured BMI and waist circumference, even after adjusting for early- and late-life confounders. However, the direction of associations is dependent on the intersections between gender and race, where white women receive benefits, yet women of color receive null or penalties. In the fourth chapter, I demonstrate that when network spillover is present, even 20% of individuals receiving interventions can significantly improve population-level behavior change. However, segregated interactions undermine the effects of network spillover. Ultimately, this dissertation highlights the role of social networks in health and health inequalities and the potential of computational social science in opening up new research on social determinants of health.