Search

Commuting is an unavoidable activity of daily working life, which helps us to get from home to work and vice versa. However, such activity can be stressful and reduce the time available for family and leisure. Indeed, negative effects of commuting on health and subjective well-being (SWB) have been found in the literature. However, not many studies have investigated the influence of commuting on a commuter's job satisfaction. With its link to performance at work, an examination of the effect of commuting on job satisfaction is likely to be useful in enhancing the general welfare of working-aged people. Moreover, apart from the individual connections with commuting, health, SWB and job satisfaction are potentially related to one another. Thus, exploring each of these three variables while controlling for the others would provide a clearer picture of the effects of commuting. Because of the endogeneity implied in these relationships, the literature has applied fixed effect (FE) models to control for time-invariant individual heterogeneity. However, the potential for reversed causality between the variables of interest calls for a more robust econometric technique to identify the causal relationships between commuting, health, SWB and job satisfaction. Without comprehensive control of multiple endogeneity sources, regression estimations can only provide association inference, rather than causal connections. This thesis adopts a FE panel data analysis of the effects of commuting on SWB, health and job satisfaction. The first empirical study in Chapter 4 finds a negative association between commuting time and health. Private and public transport modes are both linked with deteriorating health, whereas cycling can improve it. The findings are consistent over subsamples in which regional and gendered differences are explored. Moreover, instrumental variables (IVs), together with the Generalised Method of Moments (GMM) estimator in linear FE models are employed in Chapter 5 to control for the endogeneity issues intrinsic to the study. This approach can strengthen causal insight, which is missing in the literature, and the thesis is one of the first studies to have employed IV in the investigation of potential commuting effects on health, SWB and job satisfaction. Chapter 5 identifies that longer time spent on active commuting can causally raise commuters' health. It is also likely that active commuting can indirectly improve satisfaction at work and with overall life.

Spatial interaction and the locational structure between observations as well as availability of satellite derived data has meant a richer and more exhaustive exploration of topics relevant in development topics, particularly in areas of subnational economic activity and conflict. This research leverages thus spatial econometric techniques to dynamically decompose impacts from socio-economic determinants on conflict incidence (with setting in Sub-Saharan Africa). Later I also present a statistical framework (based on extension of Henderson’s approach (2012)) to augment official income figures at district / county level with multiple satellite derived signals, with specific context given to developing countries. In the first chapter, I look at the relationship and interplay between conflict intensity, foreign aid (in the form of geocoded World Bank Aid allocations) and economic activity (proxied by Sum of Lights, SOL, as gathered from satellite night lights sources), at the sub-national (provincial) level in Sub-Saharan Africa over 2000-13, using a Panel Vector Autoregression approach based on a multi-stage Continuous Updated Estimator GMM estimation strategy, and incorporating spatial effects amongst the concerned variables as well as in the model disturbances. I then decompose the derived impulse responses from this system into spatial direct and indirect responses. As per the findings, conflict intensity reacts (largely) positively to negative shocks in economic activity and World Bank Aid, with evidence of persistent spillover effects stemming from these aforementioned shocks. In the second chapter, following on from the first chapter, I specifically look at the impact of income inequality, derived from the spatial distribution of night lights raster and population raster data, on conflict incidence in Sub-Saharan Africa, using a Spatial Exponential Feedback Model approach (as opposed to the more standard Linear Feedback Model in the literature), based on Empirical Likelihood estimation. I also derive spatial direct and indirect impacts from changes in inequality, with direct responses fully dying away within 5 years while indirect response has an extent of in-built persistence. Thus, this chapter adds to the existing literature on conflict and income inequality by exploring the spatial dimension of the dynamics at play. Lastly, in the third chapter, a modified statistical method is presented, based on Henderson et al. (2012) where he looked at augmenting official national income growth measures by using satellite data on night lights. In the approach as presented here, a Method of Moments approach is introduced so as use multiple satellite signals, in addition to night lights, to augment income growth data at sub-national level. The two other signals are spread of non-vegetative cover and urban land cover data (derived from European Space Agency Climate Change Initiative Land Cover raster products). Three countries were studied with this approach: India, Indonesia and the U.S.

The dissertation studies the property of transitivity in the social choice theory. I explain why we should care about transitivity in decision theory. I propose two social decision theories: redistribution regret and ranking regret, study their properties of transitivity, and discuss the possibility to find a best choice for the social planner. Additionally, in the joint work, we propose a general method to construct a consistent estimator given two parametric models, one of which could be incorrectly specified. In “Why Transitivity”, to explain behaviors violating transitivity, e.g., preference reversals, some models, like regret theory, salience theory were developed. However, these models naturally violate transitivity, which may not lead to a best choice for the decision maker. This paper discusses the consequences and the possible extensions to deal with it. In “Redistribution Regret and Transitivity”, a social planner wants to allocate resources, e.g., the government allocates fiscal revenue or parents distribute toys to children. The social planner cares about individuals' feelings, which depend both on their assigned resources, and on the alternatives they might have been assigned. As a result, there could be intransitive cycles. This paper shows that the preference orders are generally non-transitive but there are two exceptions: fixed total resource and one extremely sensitive individual, or only two individuals with the same non-linear individual regret function. In “Ranking Regret”, a social planner wants to rank people, e.g., assign airline passengers a boarding order. A natural ranking is to order people from most to least sensitive to their rank. But people's feelings can depend both on their assigned rank, and on the alternatives they might have been assigned. As a result, there may be no best ranking, due to intransitive cycles. This paper shows how to tell when a best ranking exists, and that when it exists, it is indeed the natural ranking. When this best does not exist, an alternative second-best group ranking strategy is proposed, which resembles actual airline boarding policies. In “Over-Identified Doubly Robust Identification and Estimation”, joint with Arthur Lewbel and Jinyoung Choi, we consider two parametric models. At least one is correctly specified, but we don't know which. Both models include a common vector of parameters. An estimator for this common parameter vector is called Doubly Robust (DR) if it's consistent no matter which model is correct. We provide a general technique for constructing DR estimators (assuming the models are over identified). Our Over-identified Doubly Robust (ODR) technique is a simple extension of the Generalized Method of Moments. We illustrate our ODR with a variety of models. Our empirical application is instrumental variables estimation, where either one of two instrument vectors might be invalid.

This thesis focuses on developing and implementing new statistical methods to address some of the current difficulties encountered in the analysis of high-dimensional correlated biomedical data. Following the divide-and-conquer paradigm, I develop a theoretically sound and computationally tractable class of distributed statistical methods that are made accessible to practitioners through R statistical software. This thesis aims to establish a class of distributed statistical methods for regression analyses with very large outcome variables arising in many biomedical fields, such as in metabolomic or imaging research. The general distributed procedure divides data into blocks that are analyzed on a parallelized computational platform and combines these separate results via Hansen’s (1982) generalized method of moments. These new methods provide distributed and efficient statistical inference in many different regression settings. Computational efficiency is achieved by leveraging recent developments in large scale computing, such as the MapReduce paradigm on the Hadoop platform. In the first project presented in Chapter III, I develop a divide-and-conquer procedure implemented in a parallelized computational scheme for statistical estimation and inference of regression parameters with high-dimensional correlated responses. This project is motivated by an electroencephalography study whose goal is to determine the effect of iron deficiency on infant auditory recognition memory. The proposed method (published as Hector and Song (2020a)), the Distributed and Integrated Method of Moments (DIMM), divides responses into subvectors to be analyzed in parallel using pairwise composite likelihood, and combines results using an optimal one-step meta-estimator. In the second project presented in Chapter IV, I develop an extended theoretical framework of distributed estimation and inference to incorporate a broad range of classical statistical models and biomedical data types. To reduce computational speed and meet data privacy demands, I propose to divide data by outcomes and subjects, leading to a doubly divide-and-conquer paradigm. I also address parameter heterogeneity explicitly for added flexibility. I establish a new theoretical framework for the analysis of a broad class of big data problems to facilitate valid statistical inference for biomedical researchers. Possible applications include genomic data, metabolomic data, longitudinal and spatial data, and many more. In the third project presented in Chapter V, I propose a distributed quadratic inference function framework to jointly estimate regression parameters from multiple potentially heterogeneous data sources with correlated vector outcomes. This project is motivated by the analysis of the association between smoking and metabolites in a large cohort study. The primary goal of this joint integrative analysis is to estimate covariate effects on all outcomes through a marginal regression model in a statistically and computationally efficient way. To overcome computational and modeling challenges arising from the high-dimensional likelihood of the correlated vector outcomes, I propose to analyze each data source using Qu et al.’s quadratic inference funtions, and then to jointly reestimate parameters from each data source by accounting for correlation between data sources.

Significant attention in the financial economics literature is given to the usage of aggregated factors in their ability to explain variability in asset returns. Whereas the Capital Asset Pricing Model (CAPM) considers the excess return on the market portfolio as the dominant source of systematic variability in asset returns, the framework of Arbitrage Pricing Theory (APT) suggests that systematic risk can be further decomposed into numerous common risk factors that underlie co-movement in asset returns. Chen, Roll, and Ross (1986) popularized empirical evaluation of macroeconomic indicators in their relation to asset returns, finding that macro-economic indicators can be useful to price assets and carry statistically significant risk premiums in sample. Following the intuition of the Roll (1977) critique, I consider the pricing of risk derived from unexpected shocks, or innovations, to a wider set of macroeconomic and capital market variables. I find that information contained in shocks to common risk factors is significantly priced in the cross-section of asset returns and differs from information contained in the Fama-French-Carhart factors.

This thesis is focused on measuring and assessing country-level trends in average farm size and farm size distributions over space and time and across income levels worldwide. We develop and implement a new variant of the Generalized Method of Moments approach to estimate farm size distributions derived from world census data. Notwithstanding a major data collection effort, global generalizations are difficult due to incomplete and inconsistent farm size samples over time. However, we did find that average farm sizes have been increasing overall for both high and low to middle income countries. We also detected complex structural differences between countries stratified by income class. These differences were revealed by estimating farm size distributions (and a range of associated summary statistics) that would otherwise be masked when considering the global dynamics of farm size based only on changes in the average size of farm per country.

In the widely used over-identified econometric model, the two-step Generalized Methods of Moments (GMM) estimator and inference, first suggested by Hansen (1982), require the estimation of optimal weighting matrix at the initial stages. For time series data and clustered dependent data, which is our focus here, the optimal weighting matrix is usually referred to as the long run variance (LRV) of the (scaled) sample moment conditions. To maintain generality and avoid misspecification, nowadays we do not model serial dependence and within-cluster dependence parametrically but use the heteroscedasticity and autocorrelation robust (HAR) variance estimator in standard practice. These estimators are nonparametric in nature with high variation in finite samples, but the conventional increasing smoothing asymptotics, so called small-bandwidth asymptotics, completely ignores the finite sample variation of the estimated GMM weighting matrix. As a consequence, empirical researchers are often in danger of making unreliable inferences and false assessments of the (efficient) two-step GMM methods. Motivated by this issue, my dissertation consists of three papers which explore the efficiency and approximation issues in the two-step GMM methods by developing new, more accurate, and easy-to-use approximations to the GMM weighting matrix.The first chapter, "Simple and Trustworthy Cluster-Robust GMM Inference" explores new asymptotic theory for two-step GMM estimation and inference in the presence of clustered dependence. Clustering is a common phenomenon for many cross-sectional and panel data sets in applied economics, where individuals in the same clusterwill be interdependent while those from different clusters are more likely to be independent. The core of new approximation scheme here is that we treat the number of clusters G fixed as the sample size increases. Under the new fixed-G asymptotics, the centered two-step GMM estimator and two continuously-updating estimators have the same asymptotic mixed normal distribution. Also, the t statistic, J statistic, as well as the trinity of two-step GMM statistics (QLR, LM and Wald) are all asymptotically pivotal, and each can be modified to have an asymptotic standard F distribution or t distribution. We also suggest a finite sample variance correction further to improve the accuracy of the F or t approximation. Our proposed asymptotic F and t tests are very appealing to practitioners, as test statistics are simple modifications of the usual test statistics, and the F or t critical values are readily available from standard statistical tables. We also apply our methods to an empirical study on the causal effect of access to domestic and international markets on household consumption in rural China. The second paper "Should we go one step further? An Accurate Comparison of One-step and Two-step procedures in a Generalized Method of Moments Framework$''$ (coauthored with Yixiao Sun) focuses on GMM procedure in time-series setting and provides an accurate comparison of one-step and two-step GMM procedures in a fixed-smoothing asymptotics framework. The theory developed in this paper shows that the two-step procedure outperforms the one-step method only when the benefit of using the optimal weighting matrix outweighs the cost of estimating it. We also provide clear guidance on how to choose a more efficient (or powerful) GMM estimator (or test) in practice.While our fixed smoothing asymptotic theory accurately describes sampling distribution of two-step GMM test statistic, the limiting distribution of conventional GMM statistics is non-standard, and its critical values need to be simulated or approximated by standard distributions in practice. In the last chapter, "Asymptotic F and t Tests in anEfficient GMM Setting" (coauthored with Yixiao Sun), we propose a simple and easy-to-implement modification to the trinity (QLM, LM, and Wald) of two-step GMM statistics and show that the modified test statistics are all asymptotically F distributed under the fixed-smoothing asymptotics. The modification is multiplicative and only involves the J statistic for testing over-identifying restrictions. In fact, what we propose can be regarded as the multiplicative variance correction for two-step GMM statistics that takes into account the additional asymptotic variance term under the fixed-smoothing asymptotics. The results in this paper can be immediately generalized to the GMM setting in the presence of clustered dependence.

The first essay investigates the relationship between natural disasters and poverty at the district level in Indonesia. System generalized method of moments (GMM) and regional fixed effects models are employed, and the results suggest that damage to manufacturing facilities, hospitals, education centers, and religious buildings are important sources of increased poverty. The results also suggest that disasters associated with real losses can reduce inequality among the poor by primarily harming the relatively less poor. Disasters are also associated with a lower poverty line in the case of real losses, suggesting the estimates obtained using that measure are biased downward. In the second essay, data on enrollment rates in primary and lower secondary school are used to explore the ways natural disasters influence enrollment in education in Indonesia. The estimated coefficients are obtained using regional fixed effects regressions and suggest that disasters are generally (but not always) associated with lower enrollment. Damage to the employment sector is more important for primary school enrollment, while damage to agriculture and educational institutions is more important for lower secondary school enrollment. Damage to crops is associated with higher enrollment in lower secondary school. Additional regressions indicate that higher poverty exacerbates the negative impact of disasters on enrollment. In the third essay, household survey data is used to examine the impact of earthquakes on investment in education in El Salvador. Investment in education is measured using enrollment in and expenditures on education. Applying a difference-indifferences approach, it is estimated that being directly affected by the earthquakes leads to larger expenditures on education but that being located in a treated region is not associated with expenditures. A direct impact of the earthquakes is not associated with enrollment, but being located in a treated region leads to lower enrollment in the year of the earthquake. The negative impact of the earthquakes on enrollment dissipates quickly as there is no association between treatment and enrollment in the year following the earthquake.

This thesis is about estimation of classic and modified versions of the rational expectations competitive storage model in the tradition of Gustafson (1958) (the storage model for short), an important economic theory of price determination of storable primary commodities. The first chapter proposes and evaluates a procedure for approximating the optimal instruments under the context of the classic storage model. This procedure involves calibrating the unknown, true conditional variance function of price disturbance in the optimal instrument using the counterpart of an auxiliary storage model. Monte Carlo simulation suggests that this procedure brings small-sample efficiency gain relative to the benchmark Generalized Method of Moments (GMM) estimator of Deaton and Laroque (1992) and a few other alternatives at the sample size of 100. Its performance is also robust to parameterizations of the auxiliary model within moderate range from the true. This chapter also studies the estimators that require no preliminary estimation, which provide preliminary estimates for the proposed and other infeasible estimators. I find that a well-performing preliminary estimator does not contain instruments of lagged-more-than-one price or increasing transformations of lag prices, and does not use estimated optimal weighting matrix or adopt the continuous-updating approach of Hansen, Heaton and Yaron (1996). Therefore, an instrument of a constant plus reciprocal of lag one price and an identity weighting matrix in general form a good preliminary estimator.Chapter 2 addresses two concerns about the usefulness of the theory of storage. While commodity speculators can induce serial dependence in price, Deaton and Laroque (1992, 1995 and 1996) argue that speculation explains only a small fraction of the observed autocorrelation in the actual data. Furthermore, the expected rate of return on storage implied by previous econometric estimates is implausibly small. This chapter addresses these two concerns about the validity of the theory of speculative storage by recognizing the downward trend in real price. The existence of a unique non-stationary equilibrium is proved for a rational-expectations competitive-storage model with a trend, and testable implications of the model are also derived. I show that, when a downward price trend in part or all of the sample is ignored, the autocorrelation coefficient in price tends to be overestimated while the expected rate of return tends to be underestimated. Finally, I offer an empirical illustration of the trending storage model using annual corn price over the period from 1961 to 2005.Chapter 3 discusses the empirical implications of the distributional misspecification of two nonlinear least squares estimators of a modified storage model with unbounded prices. The existence of infrequent, extremely low harvest generates extremely high cutoff price which is difficult to pass in finite periods. Meanwhile, due to the tiny chance of such events, it is easy for the practitioners to ignore them during the estimation and apply a false storage model with relatively low cutoff price. This chapter studies how such misspecification can affect the empirical implications of estimating the storage model. Surprisingly, I find that misspecified econometric models yield better estimates for the real interest rate; and the estimated cutoff price, actually captures the sharp turning point of the equilibrium price function. Therefore, though misspecified, the estimates are practically useful. Nevertheless, it is also emphasized that such interesting property of the two estimators should by no means be understood as a defense of ignoring the infrequent influential event in the asset pricing problems. Mathematical proofs for general results and further discussions of a few econometric issues can be found in the Appendices. While a few theoretical results have been derived, this thesis relies heavily on Monte Carlo simulation and numerical functional approximation. Numerical methods turn out to be a convenient and many times necessary tools to study small-sample econometric problems when asymptotic results cannot provide an accurate approximation to the exact sampling.

This dissertation consists of three essays concerning the effects of human capital and health capital on the labor market. Chapter 1 presents a structural model that incorporates a health capital stock to the traditional learning-by-doing model. The model allows health to affect future wages by interrupting current labor supply and on-the-job human capital accumulation. Using data on sick time from the Panel Study Income of Dynamics the model is estimated using a nonlinear Generalized Method of Moments estimator. The results show human capital production exhibits diminishing returns. Health capital production increases with the current stock of health capital, or better current health improves future health. Among prime age working men, the effect of health on human capital accumulation is relatively small. Chapter 2 explores the role of another form of human capital, noncognitive skills, in explaining racial gaps in wages. Chapter 2 adds two noncognitive skills, locus of control and self-esteem, to a simple wage specification to determine the effect of these skills on the racial wage gap (white, black, and Hispanic) and the return to these skills across the wage distribution. The wage specifications are estimated using pooled, between, and quantile estimators. Results using the National Longitudinal Survey of Youth 1979 show these skills account for differing portions of the racial wage gap depending on race and gender. Chapter 3 synthesizes the idea of health and on-the-job human capital accumulation from Chapter 1 with the idea of noncognitive skills in Chapter 2 to examine the influence of these skills on human capital and health capital accumulation in adult life. Chapter 3 introduces noncognitive skills to a life cycle labor supply model with endogenous health and human capital accumulation. Noncognitive skills, measured by degree of future orientation, self-efficacy, trust-hostility, and aspirations, exogenously affect human capital and health production. The model uses noncognitive skills assessed in the early years of the Panel Study of Income Dynamics and relates these skills to health and human capital accumulation during adult life. The main findings suggest individuals with high self-efficacy receive higher future wages.

My dissertation research emphasizes estimation methods in evaluating the extent of social, strategic and spatial interactions among economic agents. My first essay, based on my joint research with Lung-fei Lee and John Kagel, generalizes Heckman’s (1981) dynamic discrete-choice panel data models by introducing time-lagged social interactions and proposes simulation based methods to implement the maximum likelihood estimation. We use this generalized model to investigate learning from peers in experimental signaling games. We find that subjects’ decisions are significantly influenced by the past decisions of their peers in the experiment. Hence the imitation of peers’ strategies is an important component of the learning process of strategic play. My second essay explores the robustness of Guerre, Perrigne and Vuong’s (2000) two-step nonparametric estimation procedure in first-price sealed-bid auctions with a large number of risk-averse bidders. With an asymptotic approximation of the intractable equilibrium bidding function of risk-averse bidders, I demonstrate that Guerre et al.’s two-step nonparametric estimator based on the equilibrium bidding behavior of risk-neutral bidders is still uniformly consistent even if bidders are risk-averse as long as the number of players in an auction is sufficiently large and derive the uniform convergence rate of the estimator. Furthermore, I show in Monte Carlo experiments that the two-step nonparametric estimator performs reasonably well with a moderate number of risk-averse bidders like six. In my third essay, which is based on my joint research with Lung-fei Lee and Christopher Bollinger, we consider the GMM estimation of the regression model with spatial autoregressive disturbances and the mixed-regressive spatial autoregressive model. We derive the best GMM estimator within the class of GMM estimators that are based on linear and quadratic moment conditions. Our best GMM estimator has the merit of computational simplicity and asymptotic efficiency. We show that it is asymptotically as efficient as the conventional maximum likelihood estimator under normality and asymptotically more efficient than the quasi-maximum likelihood estimator when the normality assumption does not hold. We show in Monte Carlo studies that, with moderate sample sizes, the proposed best GMM estimator has its biggest advantage when the disturbances are asymmetrically distributed.

This dissertation studies the adjustment of production worker employment and reports estimates of the costs and speed of employment adjustment in a broad range of disaggregated manufacturing industries using monthly data. Chapter two develops a model that identifies adjustment costs using a condition of cost minimization. The estimated median half-life of employment adjustment in the industries studied is two months, near the low end of estimates from previous studies using monthly data. The dissertation finds significant interindustry heterogeneity in adjustment costs and the speed of employment adjustment both across and within two-digit SIC manufacturing industries. This heterogeneity has been obscured in the earlier studies that estimated employment dynamics using data for the whole economy, the manufacturing sector, or broad industry aggregates. Chapter three extends the model to allow firms to make tradeoffs between efficiency and flexibility in the choice of production process. In the model, firms that face greater volatility in expected employment should have a high marginal benefit of additional flexibility. These firms choose production processes that are less efficient at the statically optimal level of output but entail relatively low costs of adjusting employment when moving between different levels of production. The chapter shows that much of the interindustry heterogeneity in adjustment costs can be explained by differences in the degree of fluctuations in labor demand at the industry level. Among four-digit industries within each two-digit industry, those with more volatile employment tend to have lower costs of adjusting employment. Chapter four examines the sensitivity of generalized method of moments (GMM) estimates of adjustment costs to the choice of parameter normalization. In one normalization, adjustment costs are economically trivial; in the alternative normalization, estimated adjustment costs tend to be more than an order of magnitude larger and similar to those previously found in the literature. The key difference between these normalizations is that the adjustment cost parameter multiplies the rational expectations error in the first but not in the second. The estimates from the second normalization should be preferred because they are similar, both theoretically and empirically, to those obtained from quasi-maximum likelihood estimation, which is invariant to the choice of normalization in finite samples and, like GMM, has desirable large sample properties.