|
In this thesis, I study an increasingly-common feature of modern labor markets: the decentralization of workers away from firms and towards independent labor provision. I provide new insights into the type of labor-market organization behind the gig-economy, freelancing, and independent contracting, looking at how these flexible workers search for and complete jobs, and how they interact with each other in team settings.The first chapter, Freelancing and the Value of Flexible Work provides new information on how these flexible workers search for, select, and complete jobs, and quantifies the value that freelancers gain from being able to choose their hours flexibly. The chapter is motivated by a new empirical observation: freelancers have large week-to-week variation in the hours that they work, even when working on the same project for many months. To explain this high-frequency variation in hours-worked, I develop a structural model of freelancing, built around a particular type of work arrangement unique to freelancers – a job is a fixed quantity of work, and the freelancer chooses an optimal time-horizon and hours-schedule over which to complete the work. The choice of hours can vary from week to week, in response to unpredictable shocks to the disutility of work. The model is estimated using earning data for a panel of New Zealand freelancers. I then use an estimated version of the model to quantify the trade-offs that freelancers face, including which jobs to accept and which to reject, and how to optimally complete tasks. I show that freelancers value the flexibility of shorter-term jobs versus the stability of longer-term jobs. I also conduct welfare analysis, comparing the ‘flexible work schedule’ freelancing model to a ‘fixed-hours’ baseline. I show that freelance workers would require an 8.5 percent increase in compensation to complete their jobs under a fixed-hours schedule, on average, and that if they know their shocks ex-ante and could adjust their hours, they would require 62.2 percent higher compensation to complete their jobs under a fixed-hours schedule. I show the value of flexibility varies across age cohorts in a manner consistent with fertility decisions, and across industries in predictable and sensible ways. I document a subset of freelancers termed ‘Flex-or-Quit’ workers, whose value placed on flexibility is so high that they would be unwilling to take on a traditional job with a fixed-hours work schedule.In the second chapter, Five Facts about Task-Based Work I present a series of new facts about this modern type of decentralized work, using a detailed and comprehensive administrative dataset of earnings for ‘task-based workers’ in New Zealand. I establish the following facts: (1) task-based workers often hold multiple simultaneous jobs, and engage in significant on-the-job search; (2) task-based jobs can be characterized as ‘primary’ and ‘secondary’ jobs for workers, along the dimensions of wages and job length; (3) task-based workers optimally allocate their time across these types of jobs, often in counter-intuitive ways; (4) there is an upward-sloping relationship between acceptable wages and acceptable job length; and (5) there is considerable within-worker variability in wages from period to period, which reflects both supply-side and demand-side factors. I also document a range of important descriptive statistics on task-based work, including calculating transition probabilities, steady-state unemployment rates, and marginal distributions of acceptable wages. These empirical findings provide useful guidance for models of labor-market search and job holding, particularly in the gig-work and freelancing sectors.The third chapter, Peer Effects in Expert Decision-Making studies a particular outcome of decentralized labor: the formation of teams of independent experts. In the paper, I study how these experts can have spill-overs effects on their peers’ decision-making. The chapter uses data on Major League Baseball (MLB) umpire crews to estimate the effect of an expert’s peer network on decision-making. I quantify decision-making by using the quality of strike / ball calls made by home-plate umpires. I then empirically analyze the extent to which an umpire paired to work with high-quality peers experiences an improvement in decision-quality. MLB umpire data has two advantages that make this analysis possible: 1) umpires are assigned to a crew each season / game in a way that is independent of decision-making quality, and have significant churn within these crews; and 2) decisions can be uniquely attributed to a single umpire, and are able to be quantified in a detailed way. Using data on these umpires, I show that a one-standard deviation improvement in the decision-making quality of an umpire’s peer network raises their own decision-making quality by around 0.03-0.1 standard deviations. The peer-effects results are particularly strong when looking only at decisions which require large amounts of expert judgment (close calls). Experience is a key driver of the peer effect, with more-experienced umpires having smaller peer effects than less- experienced umpires. I also show that there is some persistence to these peer effects, and that peer effects have an intensive margin, with experts gaining more from very-high-quality networks. Then I consider expert decisions in a Bayesian framework, and show that decision-making spill-over can be characterized in terms of two types of accuracy: bias and precision. I then demonstrate that both types of accuracy transmit through peer networks.
|
