Thesis: Ph. D., Massachusetts Institute of Technology, Sloan School of Management, Operations Research Center, 2016. / This electronic version was submitted by the student author. The certified thesis is available in the Institute Archives and Special Collections. / Cataloged from student-submitted PDF version of thesis. / Includes bibliographical references (pages 198-204). / The problem of investing over time remains an important open question, considering the recent large moves in the markets, such as the Financial Crisis of 2008, the subsequent rally in equities, and the decline in commodities over the past two years. We study this problem from three aspects. The first aspect lies in analyzing a particular dynamic strategy, called the stop-loss strategy. We derive closed-form expressions for the strategy returns while accounting for serial correlation and transactions costs. When applied to a large sample of individual U.S. stocks, we show that tight stop-loss strategies tend to underperform the buy-and- hold policy due to excessive trading costs. Outperformance is possible for stocks with sufficiently high serial correlation in returns. Certain strategies succeed at reducing downside risk, but not substantially. We also look at optimizing the stop-loss level for a class of these strategies. The second approach is more behavioral in nature and aims to elicit how various market players expect to react to large changes in asset prices. We use a global survey of individual investors, financial advisors, and institutional investors to do this. We find that most institutional investors expect to exhibit highly contrarian reactions to past returns in terms of their equity allocations. Financial advisors are also mostly contrarian; a few of them demonstrate passive behavior. In contrast, individual investors are, on average, extrapolative, and can be partitioned into four distinct types: passive investors, risk avoiders, extrapolators, and everyone else. The third part of the thesis studies how people actually trade. We propose a new model of dynamic trading in which an investor is affected by behavioral heuristics, and carry out extensive simulations to understand how the heuristics affect portfolio performance. We propose an MCMC algorithm that is reasonably successful at estimating model parameters from simulated data, and look at the predictive ability of the model. We also provide preliminary results from looking at trading data obtained from a brokerage firm. We focus on understanding how people trade their portfolios conditional on past returns at various horizons, as well as on past trading behavior. / by Alexander Remorov. / Ph. D.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Sloan School of Management, Operations Research Center, 2004. / Includes bibliographical references. / Theorists suggest the rising value of human capital will mean greater participation by employees in the ownership and governance of firms. This thesis explores aspects of these claims by analyzing the competitive effects of the allocation of ownership and authority in ad agencies and airlines. Essays 1 and 2 analyze the organizational history of the advertising industry to reconcile the stylized, theoretical views of professional service firms (PSFs) as unstable, small, private partnerships with the empirical reality of large public corporations in several professional service industries. Essay 1 uses a panel of advertising agencies and creativity awards from 1960-1980 to assess whether public ownership reduced PSF competitiveness, particularly whether it diluted employee incentives. Finding no difference in the survival, growth, and award rates of public and private ad agencies, this paper challenges the notion that allocating ownership exclusively to employees provides advantage in the PSF environment. Essay 2 draws on interviews and historical research to develop hypotheses about the structure and evolution of the industry. It proposes that agency size affects the ability to service large projects, hence the size distribution of agencies stems from heterogeneity in the units of demand. It also proposes that the industry's holding companies add value through financial intermediation. Together these essays suggest that the large public corporation is a feasible and perhaps advantaged governance form even in environments based predominantly on human capital. / (cont.) They challenge several assumptions underlying the stylized view of PSFs, and offer the speculation that the rarity of public PSFs stems from institutional barriers, not economic disadvantages. Essay 3 stems from separate research on airline labor relations and analyzes the turnaround of Continental Airlines. A case study reveals Continental's improved employee relations stem from a fundamental change in its authority system, from a traditional hierarchy to a high-involvement system. The case also discusses likely facilitators of this transformation of Continental's authority system. Taken together the essays offer a broad conjecture for future research: that allocation of authority inside the firm may be a more important factor in employee incentives than allocation of ownership to employees. / by Andrew Gustaf von Nordenflycht. / Ph.D.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Sloan School of Management, Operations Research Center, 2010. / Cataloged from PDF version of thesis. / Includes bibliographical references (p. 97-103). / This thesis presents new methods for exploring data using visualization techniques. The first part of the thesis develops a procedure for visualizing the sampling variability of a plot. The motivation behind this development is that reporting a single plot of a sample of data without a description of its sampling variability can be uninformative and misleading in the same way that reporting a sample mean without a confidence interval can be. Next, the thesis develops a method for simplifying large scatter plot matrices, using similar techniques as the above procedure. The second part of the thesis introduces a new diagnostic method for regression called backward selection search. Backward selection search identifies a relevant feature set and a set of influential observations with good accuracy, given the difficulty of the problem, and additionally provides a description, in the form of a set of plots, of how the regression inferences would be affected with other model choices, which are close to optimal. This description is useful, because an observation, that one analyst identifies as an outlier, could be identified as the most important observation in the data set by another analyst. The key idea behind backward selection search has implications for methodology improvements beyond the realm of visualization. This is described following the presentation of backward selection search. Real and simulated examples, provided throughout the thesis, demonstrate that the methods developed in the first part of the thesis will improve the effectiveness and validity of data visualization, while the methods developed in the second half of the thesis will improve analysts' abilities to select robust models. / by Rajiv Menjoge. / Ph.D.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Sloan School of Management, Operations Research Center, 2011. / This electronic version was submitted by the student author. The certified thesis is available in the Institute Archives and Special Collections. / Cataloged from student-submitted PDF version of thesis. / Includes bibliographical references (p. 153-158). / Convective weather is responsible for large delays and widespread disruptions in the U.S. National Airspace System, especially during summer. Although Air Traffic Flow Management algorithms exist to schedule and route traffic in the face of disruptions, they require reliable forecasts of airspace capacity. However, there exists a gap between the spatial and temporal accuracy of aviation weather forecasts (and existing capacity models) and what these algorithms assume. In this thesis we consider the problem of integrating currently available convective weather forecasts with air traffic management in terminal airspace (near airports). We first demonstrate how raw convective weather forecasts, which provide deterministic predictions of the Vertically Integrated Liquid (the precipitation content in a column of airspace) can be translated into reliable and accurate probabilistic fore- casts of whether or not a terminal-area route will be blocked. Given a flight route through the terminal-area, we apply techniques from machine learning to determine the probability that the route will be open in actual weather. This probabilistic route blockage predictor is then used to optimize terminal-area operations. We develop an integer programming formulation for a 2-dimensional model of terminal airspace that dynamically moves arrival and departure routes to maximize expected capacity. Experiments using real weather scenarios on stormy days show that our algorithms recommend that a terminal-area route be modified 30% of the time, opening up 13% more available routes during these scenarios. The error rate is low, with only 5% of cases corresponding to a modified route being blocked while the original route is in fact open. In addition, for routes predicted to be open with probability 0.95 or greater by our method, 96% of these routes are indeed open (on average) in the weather that materializes. In the final part of the thesis we consider more realistic models of terminal airspace routing and structure. We develop an A*-based routing algorithm that identifies 3-D routes through airspace that adhere to physical aircraft constraints during climb and descent, are conflict-free, and are likely to avoid convective weather hazards. The proposed approach is aimed at improving traffic manager decision-making in today's operational environment. / by Diana Michalek Pfeil. / Ph.D.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Sloan School of Management, Operations Research Center, 2013. / Cataloged from PDF version of thesis. / Includes bibliographical references (pages 199-204). / We consider a retailer offering an assortment of differentiated substitutable products to price-sensitive customers. Prices are chosen to maximize profit, subject to inventory/ capacity constraints, as well as more general constraints. The profit is not even a quasi-concave function of the prices under the basic multinomial logit (MNL) demand model. Linear constraints can induce a non-convex feasible region. Nevertheless, we show how to efficiently solve the pricing problem under three important, more general families of demand models. Generalized attraction (GA) models broaden the range of nonlinear responses to changes in price. We propose a reformulation of the pricing problem over demands (instead of prices) which is convex. We show that the constrained problem under MNL models can be solved in a polynomial number of Newton iterations. In experiments, our reformulation is solved in seconds rather than days by commercial software. For nested-logit (NL) demand models, we show that the profit is concave in the demands (market shares) when all the price-sensitivity parameters are sufficiently close. The closed-form expressions for the Hessian of the profit that we derive can be used with general-purpose nonlinear solvers. For the special (unconstrained) case already considered in the literature, we devise an algorithm that requires no assumptions on the problem parameters. The class of generalized extreme value (GEV) models includes the NL as well as the cross-nested logit (CNL) model. There is generally no closed form expression for the profit in terms of the demands. We nevertheless how the gradient and Hessian can be computed for use with general-purpose solvers. We show that the objective of a transformed problem is nearly concave when all the price sensitivities are close. For the unconstrained case, we develop a simple and surprisingly efficient first-order method. Our experiments suggest that it always finds a global optimum, for any model parameters. We apply the method to mixed logit (MMNL) models, by showing that they can be approximated with CNL models. With an appropriate sequence of parameter scalings, we conjecture that the solution found is also globally optimal. / by Philipp Wilhelm Keller. / Ph.D.

Thesis: Ph. D., Massachusetts Institute of Technology, Sloan School of Management, Operations Research Center, 2018. / Cataloged from PDF version of thesis. / Includes bibliographical references (pages 119-125). / This thesis addresses several prediction and estimation problems under structured decision constraints. We consider them in two parts below. Part 1 focuses on supervised learning problems with constrained output spaces. We approach it in two ways. First, we consider an algorithmic framework that is based on minimizing estimated conditional risk functions. With this approach, we first estimate the conditional expected loss (i.e., conditional risk) function by regression, and then minimize it to predict an output. We analyze statistical and computational properties of this approach, and demonstrate empirically that it can adapt better to certain loss functions compared to methods that directly minimize surrogates of empirical risks. Second, we consider a constraint-embedding approach for reducing prediction time. The idea is to express the output constraints in terms of the model parameters, so that computational burdens are shifted from prediction to training. Specifically, we demonstrate how certain logical constraints in multilabel classification, such as implication, transitivity and mutual exclusivity, can be embedded in convex cones under a class of linear structured prediction models. The approach is also applicable to general affine constraints in vector regression tasks. Part 2 concerns the estimation of a rank-based choice model under substitution constraints. Our motivating application is to estimate the primary demand for a bike-share service using censored data of commuters' trips. We model commuter arrivals with a Poisson process and characterize their trip preferences with a probability mass function (PMF) over rankings of origin-destination pairs. Estimating the arrival rate and PMF, however, is challenging due to the factorial growth of the number of rankings. To address this, we reduce the parameter dimension by (i) finding sparse representations efficiently, and (ii) constraining trip substitutions spatially according to the bike-share network. We also derive an iterative estimation procedure based on difference-of-convex programming. Our method is effective in recovering the primary demand and computationally tractable on a city scale, as we demonstrate on a bike-share service in Boston. / by Chong Yang Goh. / Ph. D.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Sloan School of Management, Operations Research Center, 2004. / Includes bibliographical references (p. 113-118). / We present a methodology for deriving robust airline schedules that are not vulnerable to disruptions caused by bad weather. In this methodology, the existing schedule is partitioned into independent sub-schedules or layers - prioritized on the basis of revenue - that provide airlines with a clear delay/cancellation policy and may enable them to market and sell tickets for flight legs based oil passenger preference for reliability. We present three different ways to incorporate degradability into the scheduling process: (1) between flight scheduling and fleet assignment (degradable schedule partitioning model), (2) with fleet assignment (degradable fleet assignment model), and (3) with aircraft routing (degradable aircraft routing model). Each problem is modeled as an integer program. Search algorithms are applied to the degradable aircraft routing model, which has a large number of decision variables. Results indicate that we can successfully assign flight legs with high revenue itineraries in the higher priority layer without adding aircraft or changing the schedule, and differentiate the service quality for passengers in different priority layers. Passengers in the high priority layers have much less delay and fewer cancellations than passengers in low priority layers even during the bad weather. In terms of recovery cost, which includes revenue lost, operational cost saving and crew delay cost, degradable airline schedules can save up to $30,000 per day. Degradable airline schedules have cost saving effect, especially when an airport with a high capacity reduction in bad weather is affected by bad weather. / by Laura Sumi Kang. / Ph.D.

Thesis: Ph. D., Massachusetts Institute of Technology, Sloan School of Management, Operations Research Center, 2018. / Cataloged from PDF version of thesis. / Includes bibliographical references (pages 127-138). / In Chapter 1, we design controlled laboratory experiments to investigate whether and how process-driven discussions may improve information sharing, information integration, and the resulting group performance in collective decision making. We first investigate three different conditions on group discussion and find that while suggesting a process (without enforcement) significantly enhances information sharing, enforcing the process is a critical enabler for the group to integrate the shared information into its final decision making and hence improves group performance. We then replicate our results with a more complex task in the context of operations management, and find evidence that the underlying mechanism for process to induce better performance is reduced difficulty and hence lowered mental effort involved in the task. Chapter 2 and 3 focus on supply-chain and governance measures to improve food safety management. In Chapter 2, we employ Heckman's sample selection framework to investigate whether and how structural properties of China's farming supply chains and the strength of governance within the regions in which the supply chains operate jointly influence the risks of economically motivated adulteration (EMA) of food products. We find that both supply chain dispersion and weak local governance are associated with higher EMA risks. In Chapter 3, we further explore a set of innovative transparency measures for quantifying the strength of governance in food safety, which are designed based on data scraped from relevant government websites. / by Shujing Wang. / Ph. D.

This electronic version was submitted by the student author. The certified thesis is available in the Institute Archives and Special Collections. / Thesis (Ph. D.)--Massachusetts Institute of Technology, Sloan School of Management, Operations Research Center, 2008. / Includes bibliographical references (p. 103-110). / (cont.) Second, for almost all 0-1 bipartite instances, we give a lower bound on the integrality gap of various linear programming relaxations of this problem. Finally, we show that for almost all 0-1 bipartite instances, all feasible schedules are arbitrarily close to optimal. Finally, we consider the problem of minimizing the sum of weighted completion times in a concurrent open shop environment. We present some interesting properties of various linear programming relaxations for this problem, and give a combinatorial primal-dual 2-approximation algorithm. / In this thesis, we study three problems related to various algorithmic and game-theoretic aspects of scheduling. First, we apply ideas from cooperative game theory to study situations in which a set of agents faces super modular costs. These situations appear in a variety of scheduling contexts, as well as in some settings related to facility location and network design. Although cooperation is unlikely when costs are super modular, in some situations, the failure to cooperate may give rise to negative externalities. We study the least core value of a cooperative game -- the minimum penalty we need to charge a coalition for acting independently that ensures the existence of an efficient and stable cost allocation -- as a means of encouraging cooperation. We show that computing the least core value of supermodular cost cooperative games is strongly NP-hard, and design an approximation framework for this problem that in the end, yields a (3 + [epsilon])-approximation algorithm. We also apply our approximation framework to obtain better results for two special cases of supermodular cost cooperative games that arise from scheduling and matroid optimization. Second, we focus on the classic precedence- constrained single-machine scheduling problem with the weighted sum of completion times objective. We focus on so-called 0-1 bipartite instances of this problem, a deceptively simple class of instances that has virtually the same approximability behavior as arbitrary instances. In the hope of improving our understanding of these instances, we use models from random graph theory to look at these instances with a probabilistic lens. First, we show that for almost all 0-1 bipartite instances, the decomposition technique of Sidney (1975) does not yield a non-trivial decomposition. / by Nelson A. Uhan. / Ph.D.

Thesis (S.M.)--Massachusetts Institute of Technology, Sloan School of Management, Operations Research Center, 2008. / Includes bibliographical references (p. 69-71). / The growth of Low Fare Carriers (LFCs) has encouraged many airlines to remove fare restrictions (such as advance purchase requirements and Saturday-night stays) on many of their fare class products, leading to the simplification of fare structures in competitive markets. In the most extreme case, these markets have fare structures that are unrestricted; the fare class products differ only by price since they AL1 lack restrictions. In these unrestricted markets, passengers buy the lowest possible fare product since there are no longer any restrictions that prevent them from doing so. A forecasting method known as "Q-forecasting" takes into account the sell- up potential of passengers in forecasting the demand in each of the fare products in such markets. Sell-up occurs when passengers upon being denied their original fare class choice, decide to pay more for the next available fare class so long as the price remains below their maximum willingness to pay. Quantifying this sell-up potential either using estimated or input values is thus crucial in helping airlines increase revenues when competing in unrestricted fare markets. A simulation model known as the Passenger Origin-Destination Simulator (PODS) contains the following 3 sell-up estimation methods: (i) Direct Observation (DO), (ii) Forecast Prediction (FP), and (iii) Inverse Cumulative (IC). The goal of this thesis is thus to investigate and compare the revenue performance of the 3 sell-up estimation methods. These methods are tested in a 2-airline (consisting of AL1 and AL2) unrestricted network under different RM fare class optimization scenarios. / (cont.) Both estimated and input sell-up values are tested on AL1 whereas only input sell-up values are tested on AL2. The findings of the simulations indicate that using FP typically results in the highest revenues for AL1 among AL1 3 sell-up estimation methods. When compared against simple RM fare class threshold methods that do not consider sell-up, using FP results in up to a 3% revenue gain for AL1. Under some fare class optimization scenarios, using FP instead of input sell-up values even results in a revenue increase of close to 1%. These findings suggest that FP is robust enough under a range of fare class optimizers to be used by airlines as a sell-up estimator in unrestricted fare environments so as to raise revenues. / by Jingqiang Charles Guo. / S.M.