Although option market is becoming increasingly important, its complex structure and abundant underlying assets make it very difficult to obtain relatively accurate pricing formulas or algorithms. The first part of this thesis focuses on a theoretical stochastic volatility model for option pricing. We propose a continuous-time stochastic volatility model based on an arithmetic Brownian motion: a one-parameter extension of the normal stochastic alpha-beta-rho (SABR) model. Using two generalized Bougerol's identities in the literature, the study shows that proposed model has a closed-form Monte-Carlo simulation scheme and that the transition probability for one special case follows Johnson's SU distribution---a popular heavy-tailed distribution originally proposed without stochastic process. It is argued that the SU distribution serves as an analytically superior alternative to the normal SABR model because the two distributions are empirically similar. The second part of this thesis gives a bias corrected algorithm of the least square Monte Carlo (LSM) algorithm for pricing American options. The traditional LSM estimator contains undesirable look-ahead bias, and the conventional technique of removing it necessitates doubling simulations. We present the leave-one-out LSM (LOOLSM) algorithm for efficiently eliminating look-ahead bias. We also show that look-ahead bias is asymptotically proportional to the regressors-to-simulation paths ratio. Our findings are demonstrated with several option examples, including the multi-asset cases that the LSM algorithm significantly overvalues. The LOOLSM method can be extended to other regression-based algorithms that improve the LSM method. In addition to option pricing, the third part of this thesis studies item response theory (IRT) and its application to foraging data analysis. A fundamental problem in biology is to understand how the differences among components, or variation in function, contribute to collective behavior. To tackle this problem, we first give a new approach to latent trait modeling in IRT, then design corresponding experiments and finally apply IRT to data analysis. We also give detailed explanation and further discussion for experiment results

The innovation performance of entrepreneurial ventures critically depends upon their management of internal and external relationships. Within organizational boundaries, how cofounders collectively contribute resources and efforts can significantly impact the process of technological ideation and commercialization. Beyond organizational boundaries, how entrepreneurial ventures manage their external relationships with partners, such as product users, is also vital for innovation and overall performance. Across three inter-linked papers, this dissertation investigates collaborative innovation both within and beyond the organizational boundary of entrepreneurial ventures, focusing on the antecedents, consequences, as well as valuation of technological resource contributions. Using a quasi-natural experiment, the first paper examines how innovation-driven collaborations between ventures and their product users can be undermined by legitimacy concerns stemming from another domain within the multiplex relationship between the two parties. The second paper analyzes the organizational consequences of venture-user collaborations and suggests heterogeneous effects on different types of successes (initial public offering vs. acquisitions). Finally, the third paper explores how the type of resources contributed (technological vs. non-technological) shapes initial equity share division among cofounders as well as equity dilution over time. Together, this dissertation contributes to the literature on interorganizational relationships, innovation, and entrepreneurship and provides policy implications

As the threats of climate change and environmental degradation worsen, policy makers and non-profits worldwide have developed a plethora of incentive-based solutions that induce environmental efforts from individuals. This dissertation studies these incentive-based solutions using the theoretical modeling tools of contract theory, mechanism design and queueing theory. In the first two sections of this dissertation, I ask how to better design contracts and payment structures to increase the efficiency and effectiveness of environmental payment programs. First, I develop a stylized principal-agent model to study Payment for Ecosystem Services programs, which are used by governments worldwide to pay forest owners to conserve forest. I focus on simple and robust contracts that have performance guarantees even when the government has limited information. Next, I investigate the optimal payment structures in long-term and large-scale afforestation programs with smallholder farmers, finding that paying farmers early can increase the environmental benefits generated per dollar spent. Lastly, motivated by operational challenges in the palm oil value chain, I study the impact of travel delay and real-time announcements on queues at a congested server, finding that travel delay can cause reduction in customer welfare in a queueing system

Health care costs in the United States exceed $3.5 trillion annually, with between $760 billion and $935 billion considered waste. Data-driven analytics could reduce costs and provide higher quality care to patients by more efficiently allocating limited resources, just as analytics has done in other industries such as logistics, manufacturing and aviation. In this dissertation, I demonstrate three levels at which analytics provide value in health: clinical decision making, healthcare operations management and public health policy. Clinical decision making refers to decisions at the individual patient level: for example, determining which treatment to provide a patient or predicting an individual's risk of disease. Healthcare operations management refers to decisions about the system that delivers care to patients: for example, determining how to organize patient flow through a hospital or schedule procedures. Finally, public health policy refers to decisions about the overall health of a population: for example, determining how to control an infectious disease or distribute limited resources across different diseases

This dissertation develops, estimates, and analyzes a deep learning model for home prices using an unprecedented dataset containing over 150 million sale transactions between 1970 and 2019 in the United States. We contrast our model's superior performance with methods commonly used in the literature, and uncover previously unstudied pricing factors. The last sale amount figures as the most influential variable, and we identify localized statistics, house characteristics, mortgage rates, and macroeconomic conditions to be among the most salient features. We also contribute to the literature in demonstrating our model's capacity to capture non-linear effects, exposing, for example, the model's increased sensitivity to local statistics in times of high and low economic prosperity. In the second part, we use the trained network to scrutinize the efficiency of housing markets —which assumes all information is contained in home prices— in a set of novel approaches. We challenge the efficiency notion by finding that model-induced price trends can be predictive of future returns, and devise a trading strategy that exploits such effects to produce unparalleled profits. In an out-of-sample simulation of house trading from 1995 to 2019, our deep learning-induced strategy produces the highest average returns (at an annualized rate of over 25%) when compared to benchmark methods, and we find that short term resales yield the highest performance. Finally, we find evidence that suggests the market is not efficient (there are large gains to be made using only past information), and expose the areas with the highest price exploitability, such as California, Washington State, Wyoming, Colorado, New Mexico, Massachusetts, New York, Connecticut, and New Jersey

In the past few years there has been an enormous amount of progress in machine learning, and one of the biggest contributing factors, especially for deep learning, is the vast amount of data that we have been able to collect, due to digitization and the internet. Harder and more ambitious problems in general artificial intelligence that that will enable agents to learn on their own and to act autonomously in the environment remain largely open. Initial breakthroughs include training an agent to play a complicated board game, or training agent to drive a car demonstrate that these problems require a lot of data even more data, even more compute than ever before, and possibly more than what we currently have available. This motivates several algorithmic challenges, namely how do we design algorithms that make the best use of the data that is available, and how do we design algorithms that are empirically and theoretically effective on the kinds of data that we often see in practice, for example, data with temporal dependencies and data that follow distributions that are hard to describe. This thesis proposes and analyzes a few algorithmic solutions along this theme, which is an important step to more reliably deploying general artificial intelligence into society

This dissertation consists of two parts, with the central topic on efficient simulation. The first part focuses on the problem of optimal stratified sampling with finitely many and infinitely many strata. We present a more general framework for analysis and design of adaptive sampling algorithms achieving optimal convergence rate. We show that asymptotic optimal variance can be achieved. The second part focuses on hospital-level COVID demand forecast, where we provide a method to generate consistent forecast intervals. We show that no stationarity assumption of the underlying point process is required for the proposed method. Furthermore, the model is computationally lightweight to estimate, as compared to epidemiology and machine learning based models

Entrepreneurial finance is a cornerstone of economic growth. Prior work has done much to explore the financial outcomes of entrepreneurial financing. However, many gaps remain in our understanding of the organizational effects of these events. This dissertation addresses these gaps by applying organizational theory lenses to entrepreneurial finance in three tightly-linked papers. The first is a literature review on entrepreneurial finance, with a focus on IPOs and emerging fundraising mechanisms. The second paper uses over 1 million reviews from Glassdoor.com to assess how the addition of public shareholders affects employees when an organization goes public. Finally, the third paper examines how the composition of accelerator cohorts affects entrepreneurial learning and startup performance. Together, these papers contribute several findings to the literature on organizations and entrepreneurship, as well as insights for managers and entrepreneurs

Prior literature emphasizes that aligning with regulatory, normative, and cognitive institutional arrangements aids organizational formation, resource gathering, and performance. However, during institutional changes, the old and the new rules often coexist and interact, resulting in conflicting institutional arrangements. These conflicts create strategic dilemmas for entrepreneurs, but we don't have a systematic understanding of entrepreneurial strategies to tackle the conflicts. To address this gap, my dissertation focuses on three aspects: (1) adapting to the transition from old rules to new rules, (2) leveraging the new rules to replace the old rules, and (3) combining the new rules and the old rules. I examine these aspects in three empirical settings respectively: marketization, digitization, and tokenization. My first paper analyzes how entrepreneurs change growth strategies during China's institutional change from a government-dominated to a more market-based economy. My second paper draws on institutional intermediary and network tie formation literature to examine entrepreneurial fundraising strategies on online platforms rather than offline. My third paper draws on optimal distinctiveness theory to explore how blockchain entrepreneurs combine new and old elements in framing to balance differentiation and legitimation. Empirically, I use machine learning models to create measures from big data and econometric models to identify causal relationships. Overall, my dissertation contributes to institutional theory by examining entrepreneurial agency in tackling institutional pressure and contributes to strategy literature by analyzing the institutional effects on entrepreneurial strategies

In this thesis we study two applications of machine learning to estimate models that explains asset prices by harnessing the vast quantity of asset and economic information while also capturing complex structure among sources of risk. First we show how to build a cross-section of asset returns, that is, a small set of basis or test assets that capture complex information contained in a given set of characteristics and span the Stochastic Discount Factor (SDF). We use decision trees to generalize the concept of conventional sorting and introduce a new approach to robustly recover the SDF, which endogenously yields optimal portfolio splits. These low-dimensional investment strategies are well diversified, easily interpretable, and reflect many characteristics at the same time. Empirically, we show that traditional cross-sections of portfolios and their combinations, especially deciles and long-short anomaly factors, present too low a hurdle for model evaluation and serve as the wrong building blocks for the SDF. Constructed from the same pricing signals, our cross-sections have significantly higher (up to a factor of three) out-of-sample Sharpe ratios and pricing errors relative to the leading reduced-form asset pricing models. In the second part of the thesis, I present deep neural networks to estimate an asset pricing model for individual stock returns that takes advantage of the vast amount of conditioning information, while keeping a fully flexible form and accounting for time-variation. The key innovations are to use the fundamental no-arbitrage condition as criterion function to construct the most informative test assets with an adversarial approach and to extract the states of the economy from many macroeconomic time series. Our asset pricing model outperforms out-of-sample all benchmark approaches in terms of Sharpe ratio, explained variation and pricing errors and identifies the key factors that drive asset prices

As entrepreneurial funding supports the growth of entrepreneurial firms, it can be viewed as the fuel that enhances the creation, development, and growth of new technologies, industries, and markets. However, investing in entrepreneurial firms is highly risky. For example, Venture Capitalists (VCs) may receive a large number of business plans/proposals every year but only a few of them can be successful. Since VCs typically employ a small number of people, they do require a more effective and efficient screening and evaluation process. Prior research has mainly focused on identifying evaluation criteria to help VCs predict the business success of these firms. Nevertheless, relying on VCs' self-reporting and small regional datasets, these earlier studies have little agreement on the evaluation criteria. Therefore, it is difficult to employ those criteria to predict business success in practice. In this work, we propose data-driven approaches using machine learning methods as a complementary methodology to help VCs predict companies' success when they screen and evaluate investment deals. We start by verifying new evaluation criteria with large datasets and then focus on applying machine learning methods to predict business success. We compare different machine learning methods and discuss how VCs can benefit from the prediction. We also apply deep neural networks and few-shot learning methods to two challenging scenarios faced by the VCs: (1) when the companies are just founded and don't have any funding history; (2) when the companies are from an emerging industry that doesn't have a lot of historical data to learn from

Collaboration across boundaries is critical for most creative work. Most research to date has documented the challenges of cross-boundary collaboration and knowledge transfer and translation across boundaries. Limited research has investigated the ways in which ideas are integrated into a cohesive whole. This dissertation addresses this gap by examining how ideas get integrated in creative cross-boundary projects. Through an in-depth field study of a Thai real estate development company, I empirically examine the process of integration and the practices that support integration. I introduce a theory of integration as a complex, dynamic and recursive process that occurs over time and requires idea elaboration. Ideas are first introduced, then elaborated and transformed by the cross-boundary project team and either integrated into the cohesive whole or rejected. As an idea is introduced or transformed it stimulates the transformation of other ideas in a continuous process until the project is complete. I identify four activities that support idea elaboration and integration across boundaries: instantiating ideas, representing others, balancing interests, and anticipatory activities. I also identify the conditions that enable project members to perform these integrative activities: multiple identities, big picture thinking, and a formal coordination role. I propose therefore that integration in creative cross-boundary projects is a complex process requiring idea elaboration and attention to project members' opinions, interests, and perspectives, and that a core activity of creative cross-boundary projects is how team members bring different perspectives to bear. This dissertation advances theory on cross-boundary teams, integration, and creativity by showing how integration occurs and the activities that enable integration

With the growing amount of medical data available and the increasing reliance of the medical community on artificial intelligence (AI) tools, there is an emerging demand for techniques that support medical decisions based on patient outcomes. To be adopted, machine learning (ML) tools must be accurate, trustworthy, and interpretable. Clinicians and patients should be able to understand the reasoning due to the high-stakes and sensitive nature of most medical decisions. Ultimately, medical ML tools will inform the decision-making process, empower clinicians and patients, and help improve clinical outcomes. We focus on using electronic medical records (EMRs) to improve retrospective comparative effectiveness research for reliable and interpretable decision-making in oncology. Decision making is fundamentally causal, intervening to improve outcomes. EMRs are a rich source of information that can be used to inform those decisions. Natural language processing (NLP) can analyze the unstructured clinical notes. Our research adapts causal inference, ML, and NLP techniques to discover insights from high-dimensional and high-noise EMR data. We study how we can 1) use NLP to identify cancer treatments, 2) reduce selection bias in observational studies, and 3) build ML tools to supplement radiologist decision making. In the first study, we use EMR notes to identify cancer treatments. We apply the method to prostate, esophagus, and oropharynx cancer datasets. It achieves over 90% accuracy for treatment identification. The method can be used to supplement cancer treatment records and help with future research on treatment planning and comparison. In the second study, we show how clinical notes can be used to uncover potential confounders and adjust for bias in retrospective comparative effectiveness studies. We apply our approach to prostate and lung cancer cohorts and found that we reduce the amount of bias when compared against established randomized control trials. In the third study, we develop a tool that could aid radiologist decision-making for mammogram diagnosis. It quantifies the decision threshold of each radiologist and can improve radiologist consistency and practice

Fueled by internet-based technologies that enable seamless transactions between globally dispersed individuals, new online platforms are emerging that compose "flash" teams of remote freelancers to collaborate on complex projects in diverse industries. In these provisional teams, relationships are transient, virtual, and often algorithmically mediated. This presents organizational scholars with new intriguing puzzles. First, how do flash teams achieve effective coordination despite their lack of shared context and minimal opportunities for rich interactions? Second, how do workers navigate the uncertainties associated with this way of working, including those arising from being evaluated by algorithmic systems? To answer these questions, I draw on extensive online communication data (2 million Slack messages), the performance records of 117 teams, and over 18 months of field observation from a single "gig" labor platform company that composes virtual project teams of freelance software development professionals. Study 1 develops a natural language processing-based measure of teams' cognitive diversity as expressed in their online interactions and demonstrates its task-contingent relationship with performance. In a complementary inductive approach, study 2 theorizes how leaders' situational behaviors aimed at creating role clarity explain variation in team performance outcomes. Study 3 inductively theorizes the conditions under which freelancers developed heightened levels of trust and engagement in response to an algorithmic ranking system. My findings contribute to research on team effectiveness, team temporal dynamics, and leadership, as well as to the nascent literature on algorithms at work. They also point to several empirical and methodological opportunities for organizational scholars seeking to study novel technology-enabled organizational forms

Cost-effectiveness analyses can quantify and compare the benefits, harms, and costs of potential health interventions. Often, researchers will model a single disease for a cost-effectiveness analysis. However, some interventions can prevent multiple infectious diseases. For example, the Aedes aegypti and Aedes albopictus mosquitos transmit chikungunya, Zika, dengue, and yellow fever, and thus controlling these mosquitos can prevent cases of all four diseases. This dissertation focuses on applications of and methods for modeling multiple infectious diseases in cost-effectiveness analyses. First, I investigate if the results of a cost-effectiveness analysis can depend on the set of diseases that are modeled if some interventions prevent more than one disease. Next, I model both chikungunya and dengue to conduct a cost-effectiveness analysis of prevention measures for both diseases in Colombia. Finally, I develop conditions under which it is necessary to model multiple diseases when conducting a cost-effectiveness analysis and propose methods for using parallel modeling to simplify multi-disease modeling

Donated blood is a critical component of health systems around the world, but its collection and transfusion involve risk for both donors and recipients. Transfusion-transmitted diseases and non-infectious adverse events pose a risk to transfusion recipients, and repeat blood donation can cause or exacerbate iron deficiency among donors. This dissertation describes four decision-analytic modeling projects that inform blood safety policy. In Chapter 2, I integrate epidemiological, health-economic, and biovigilance data to estimate the efficacy and cost-effectiveness of a 2016 policy mandating that all blood donations are screened for Zika virus in the U.S. The analysis uses a novel microsimulation of individual transfusion recipients that captures the relationship between disease exposure risk and the number and type of blood components transfused. In Chapter 3, I develop the first health-economic assessment of whole blood pathogen inactivation. The analysis is for Ghana and improves on prior blood safety assessments for sub-Saharan Africa by considering the likelihood and timing of clinical detection for chronic viral infections. In Chapter 4, I develop an optimization-based framework for identifying the optimal portfolio of blood safety interventions that overcomes some limitations of traditional cost-effectiveness analyses for blood safety. By applying this framework retrospectively to evaluate U.S. policies for Zika and West Nile virus, I show that the optimal policy can vary by geography, season, and year. Chapter 5 focuses on how frequently donors are allowed to give blood. I develop a machine learning-based decision model that tailors the inter-donation interval to each donor's risk of iron-related adverse outcomes to balance risks to donors against risks to the sufficiency of the blood supply. Together, these model-based analyses introduce novel methods and provide guidance for efficient and effective use of resources for blood safety

Understanding why networks form and evolve the way they do is a core goal of many scientific disciplines ranging from the social to the physical sciences. Across these disciplines, many kinds of formation models have been employed, several of which can be subsumed under a choice framework, using conditional logit models from discrete choice and random utility theory. Each new edge is viewed as a ``choice'' made by a node to connect to another node, based on (generic) features of the other nodes available to make a connection. This perspective on network formation unifies existing models such as preferential attachment, triadic closure, and node fitness, which are all special cases, and thereby provides a flexible means for conceptualizing, estimating, and comparing models. The lens of discrete choice theory also provides several new tools for analyzing sUnderstanding why networks form and evolve the way they do is a core goal of many scientific disciplines ranging from the social to the physical sciences. Across these disciplines, many kinds of formation models have been employed, several of which can be subsumed under a choice framework, using conditional logit models from discrete choice and random utility theory. Each new edge is viewed as a ``choice'' made by a node to connect to another node, based on (generic) features of the other nodes available to make a connection. This perspective on network formation unifies existing models such as preferential attachment, triadic closure, and node fitness, which are all special cases, and thereby provides a flexible means for conceptualizing, estimating, and comparing models. The lens of discrete choice theory also provides several new tools for analyzing social network formation. In large network data logit models run into practical and conceptual issues, since large numbers of alternatives make direct inference intractable and the assumptions underlying the logit model cease to be realistic in large graphs. Importance sampling of non-chosen alternatives reduces the data size significantly, while, under the right conditions, preserving consistency of the estimates. A model simplification technique called ``de-mixing'', whereby mixture models are reformulated to operate over disjoint choice sets, reduces mixed logit models to conditional logit models. This opens the door to the other approaches to scalability and provides a new analytical toolkit to understand the underlying processes. The flexibility of the logit framework is illustrated with examples that analyze several synthetic and real-world datasets, including data from Flickr, Venmo and a large citation graph. The logit model provides a rigorous method for estimating preferential attachment models and can separate the effects of preferential attachment and triadic closure. A more substantial application is the identification of the persistent and changing parts of the networking strategies of U.S. college students as they go through their college years. This analysis is done using a rich and large data set of digital social network data from the Facebook platform.ocial network formation. In large network data logit models run into practical and conceptual issues, since large numbers of alternatives makes direct inference intractable and the assumptions underlying the logit model cease to be realistic in large graphs. Importance sampling of non-chosen alternatives reduces the data size significantly, while, under the right conditions, preserving consistency of the estimates. A model simplification technique called ``de-mixing'', whereby mixture models are reformulated to operate over disjoint choice sets, reduces mixed logit models to conditional logit models. This opens the door to the other approaches to scalability and provides a new analytical toolkit to understand the underlying processes. The flexibility of the logit framework is illustrated with examples that analyze several synthetic and real-world datasets, including data from Flickr, Venmo and a large citation graph. The logit model provides a rigorous method for estimating preferential attachment models and can separate the effects of preferential attachment and triadic closure. A more substantial application is the identification of the persistent and changing parts of the networking strategies of U.S. college students as they go through their college years. This analysis is done using a rich and large data set of digital social network data from the Facebook platform

Stochastic coupling is a powerful technique in probability theory that facilitates comparison between two probability distributions. The coupling technique is useful to bound discrepancy between two distributions, design simulation algorithms, and study asymptotic behavior of stochastic processes. In this thesis, we study the applications of stochastic coupling in the context of exact simulation and distributionally robust optimization (DRO). In the first part, we provide the first generic exact simulation algorithm for multivariate diffusions. Current exact sampling algorithms for diffusions require the existence of a transformation which can be used to reduce the sampling problem to the case of a constant diffusion matrix and a drift which is the gradient of some function. Such transformation, called Lamperti transformation, can be applied in general only in one dimension. Therefore, completely different ideas are required for exact sampling of generic multivariate diffusions. Our strategy combines techniques borrowed from the theory of rough paths, on one hand, and multilevel Monte Carlo on the other, in which the random level is coupled with the diffusion process to simulate. The second part of the thesis is dedicated to optimal transport based DRO, where the optimal transport distance is defined as the expected transportation cost under the maximal coupling between two probability measures. The DRO is a systematic modelling approach to mitigate the effect of modelling error and distributional ambiguity in stochastic optimization problems. For optimal transport based DRO, we adopt a minimax framework that minimizes the expectation of random function under the worst distribution in a distributional uncertainty set characterized by optimal transport distance. We propose a methodology which learns the optimal transport cost function in a natural data-driven way. Then, under affine decision rules and conventional convexity assumptions on the underlying loss function, we obtain structural results about the objective value function, the optimal policy, and the worst-case optimal transport adversarial model. These results expose a rich structure embedded in the DRO problem (e.g. strong convexity even if the non-DRO problem was not strongly convex, a suitable scaling of the Lagrangian for the DRO constraint, etc. which are crucial for the design of efficient algorithms). As a consequence of these results, one can develop efficient optimization procedures which have the same sample and iteration complexity as a natural non-DRO benchmark algorithm such as stochastic gradient descent

The Internet is strengthening democracy by empowering individuals. Social media enables real-time dissemination of information. Online social networks enable people to come together in support of common causes. But despite the prevalence of Internet technologies, there has yet to be an online platform where large-scale democratic decision making occurs regularly as a society. Multiple obstacles remain in the way of building such platforms. For example, places where users can exercise free-form speech often degenerate into vitriol. The unstructured, large-scale nature of the Internet means that it is not obvious what the optimal way to aggregate opinions is and how to act on them. In this dissertation, we discuss our work on building platforms that tackle these issues from theoretical and practical perspectives. Specifically, - Participatory Budgeting (PB): PB is a democratic process where citizens vote on public budgets. We propose Knapsack Voting, a vote aggregation method in PB, and show that it is welfare-maximizing and strategy-proof under certain utility models. We also discuss the Participatory Budgeting Platform (pbstanford.org), our open-source platform for running online PB elections. It has been used throughout the United States, including Chicago, Boston, Seattle, and New York City, and has distributed over 60 million of public budgets. - Sequential Deliberation: Sequential deliberation is a vote aggregation method that uses rounds of small group interaction and can be used in decision spaces that are too complex for ordinal voting. We show that this method converges to the optimal point quickly under certain utility models. We also show how sequential deliberation can be used in PB to handle project interdependencies. - Online Deliberation Platform: We discuss our platform for conducting large-scale Deliberative Polls with an automated moderator that can support over 1,000 simultaneous participants (stanforddeliberate.org). The platform has been successfully deployed in the US, Canada, Chile, Hong Kong, and Japan, where people deliberated on complex issues such as health care, renewable energy, and foreign policy. We discuss the technical challenges and our study on the efficiency of the platform

Decarbonizing the electricity sector will require a massive increase in variable solar and wind generation. Maintaining the stability of that future grid will require new sources of flexibility to keep supply and demand in balance without relying on traditional fossil-fired generation. Battery-based storage and flexible demand are two promising sources of flexibility that could each play a large role in future electricity grids. New policies are necessary to support the growth and value of these resources, and to integrate them into existing markets, because they work very differently from the traditional electricity generators that markets were originally designed to coordinate. This dissertation contributes to ongoing policy discussions by identifying how policy and market design can efficiently support the growth of each of these up-and-coming resources. In doing so, it advances our knowledge of exactly how these resources can most effectively provide value to the electricity grid. This dissertation begins with a novel taxonomy of uncertainty analysis approaches available to macro-energy systems modeling and includes a summation of best-practice advice that is applied in the following chapters. The third chapter examines the design of Demand Response (DR) programs, which could provide flexibility to the electricity grid through programs that incentivize customers to reduce or shift their electricity usage. It examines the implications of several common DR program design parameters for the system value and customer experience, revealing which parameters may severely limit the value of DR, and which offer smart tradeoffs between utility and customer outcomes. Results show that advance notice requirements as well as duration, total-time and total energy limits could be valuable tools for providing good system value while minimizing customer impacts. DR program designers should avoid creating programs with poor reliability and time-of- day limits, as these can significantly reduce system value. The fourth chapter examines the economic outcomes, as well as greenhouse gas and criteria air pollutant emissions resulting from grid-scale lithium-ion battery storage across three different Independent System Operators in the US. It identifies situations where smart policy can support the growth of this industry and its value to society and shows that the financial attractiveness of batteries varies greatly by region. The frequency regulation (FR) market currently drives the profitability of batteries. Results suggest that policy supporting learning-by-doing and attractive financing terms can help batteries remain financially attractive despite expected declines in FR prices. Finally, it shows that batteries can increase the net emissions of the electricity system, but studies that do not incorporate ancillary services into battery dispatch decisions may not correctly estimate the magnitude of this effect