No description available.

This thesis comprises four essays in the economics of networks and market design. The common thread in all these essays is the presence of complementarities or externalities. Chapter 2 presents a unified model of networks and matching markets. We build on a contribution by Pycia (2012). We show that strong pairwise alignment of agents’ preferences is a necessary and sufficient condition for the existence of strongly stable networks and strongly stable allocations in multilateral matching markets with finite contracts. Strongly stable networks are not necessarily efficient. Although we use a demanding stability concept, strong pairwise alignment allows for complementarities and externalities. In Chapter 3, we generalise the gross substitutes and complements condition introduced by Sun and Yang (2006). Our new condition guarantees the existence of competitive equilibrium in economies with indivisible goods. Competitive equilibrium can be found using an extension of the double-track adjustment process (Sun and Yang, 2009). In this chapter, we also study contract networks (Ostrovsky, 2008). We show that chain-stable contract allocations can exist even in cyclical contractual networks, such as electricity markets, as long as they are appropriately segmented. In Chapter 4, we run a series of experiments to compare the performance of four auctions – first-price, Vickrey, Vickrey-Nearest Rule (Day and Cramton, 2008), and Reference Rule (Erdil and Klemperer, 2010). In our setting, there are two items and three bidders. Two local bidders want an item each, but the global bidder wants both items. We introduce various exposure and package-bidding treatments. We find that the first-price auction always revenue-dominates all the other auctions without any loss in efficiency, strengthening the results of Marszalec (2011). Exposure affects global bidders only in the first-price auction. In other auctions, global bidders often do not take into account the effect of their own bids on their payments. We find no evidence of threshold effects. Finally, in Chapter 5, we develop a new model of online social network formation. In this model, agents belong to many overlapping social groups. We derive analytical solutions for the macroscopic properties of the network, such as the degree distribution. We study the dynamics of homophily – the tendency of individuals to associate with those similar to themselves. We calibrate our model to Facebook data from ten American colleges.

This thesis consists of two parts: Part I (Chapters 2 and 3) critically assesses a set of methodological tools that are widely used in the literature and that are applied to the empirical analysis in Part II (Chapters 4 and 5). Using a randomised experiment, the first chapter compares pen-and-paper interviewing (PAPI) with computer-assisted personal interviewing (CAPI). We observe a large error count in PAPI, which is likely to introduce sample bias. We examine the effect of PAPI consumption measurement error on poverty analysis and compare both applications in terms of interview length, costs and respondents’ perceptions. Next, we formalise an unproven source of ordinary least squares estimation bias in standard linear-in-means peer effects models. Deriving a formula for the magnitude of the bias, we discuss its underlying parameters. We show when the bias is aggravated in models adding cluster fixed effects and how it affects inference and interpretation of estimation results. We reveal that two-stage least squares (2SLS) estimation strategies eliminate the bias and provide illustrative simulations. The results may explain some counter-intuitive findings in the social interaction literature. We then use the linear-in-means model to estimate endogenous peer effects on the awareness of a community-based development programme of vulnerable groups in rural Tanzania. We denote the geographically nearest neighbours set as the relevant peer group in this context and employ a popular 2SLS estimation strategy on a unique spatial household dataset, collected using CAPI, to identify significant average and heterogeneous endogenous peer effects. The final chapter investigates social network effects in decentralised food aid (free food and food for work) allocation processes in Ethiopia, in the aftermath of a serious drought. We find that food aid is responsive to need, as well as being targeted at households with less access to informal support. However, we also find strong correlations with political connections, especially in the immediate aftermath of the drought.