This thesis empirically studies two themes in the field of decision making. First, it investigates firm and worker behavior in an ``on-demand'' manufacturing environment (first chapter); second, it explores the effect of family spillovers on health care decision making (second chapter). The first chapter develops and estimates a dynamic equilibrium model of firm and worker behavior in a production environment, by which manufacturing and supplying of products happens immediately after receiving orders. The firm solves a dynamic discrete choice cost minimization model in which it faces uncertainty about future product demand and workers' productive capacity. The firm chooses to employ two types of workers -- gig and permanent -- and it sets parameters of a compensation scheme that is a mix of salary and performance-based incentives to elicit worker effort. Heterogeneous workers solve a daily effort choice problem given the compensation scheme offered by the firm. I estimate the model using panel data from an online, global manufacturer that produces customized items. I find that gig workers are much more responsive to incentive pay. I show that varying compensation scheme over time and using a mix of gig and permanent workers provide the flexibility that the firm needs to effectively operate an on-demand production process. The second chapter studies how family spillovers shape health care consumption through two sources: a learning channel whereby family members share information regarding their health insurance plan, and a behavioral channel whereby risk perception and habits are shared and transmitted. I exploit two types of sudden health shocks to identify a causal effect operating through each channel, and incorporate these shocks into event-study frameworks to quantify the effect of spillovers on non-injured adult family members. I find a significant behavioral spillover effect of a more than 70% increase in medical spending of preventive care over a two-year horizon. Moreover, I find a persistent spillover effect associated with the learning channel that amounts to an average increase in medical spending of more than 100% relative to prior to the health shock. While the first result is in line with previous findings in the literature, the second is novel.