本セミナーは、東京理科大学経営学部ビジネスエコノミクス学科において、先端的な経済学の研究者をお招きし、研究に関する講演をしていただくものです。参加希望の方は担当者(菅原, shinya_sugawara at rs.tus.ac.jp)までご連絡ください
富士見校舎へのアクセスはこちらから
担当者: 東京理科大学ビジネスエコノミクス学科准教授 菅原慎矢 (shinya_sugawara AT rs.tus.ac.jp)
We theoretically discuss why deep neural networks (DNNs) performs better than other models in some cases by investigating statistical properties of DNNs for non-smooth functions. While DNNs have empirically shown higher performance than other standard methods, understanding its mechanism is still a challenging problem. From an aspect of the statistical theory, it is known many standard methods attain optimal convergence rates, and thus it has been difficult to find theoretical advantages of DNNs. This paper fills this gap by considering learning of a certain class of non-smooth functions, which was not covered by the previous theory. We derive convergence rates of estimators by DNNs with a ReLU activation, and show that the estimators by DNNs are almost optimal to estimate the non-smooth functions, while some of the popular models do not attain the optimal rate. In addition, our theoretical result provides guidelines for selecting an appropriate number of layers and edges of DNNs. We provide numerical experiments to support the theoretical results.
Equilibria and optima generally differ in imperfectly competitive markets. While this is well understood theoretically, it is unclear how large the welfare distortions are in the aggregate economy. Do they matter quantitatively? To answer this question, we develop a multi-sector monopolistic competition model with endogenous firm entry and selection, productivity, and markups. Using French and British data, we quantify the gap between the equilibrium and optimal allocations. In our preferred specification, inefficiencies in the labor allocation and entry between sectors, as well as inefficient selection and output per firm within sectors, generate welfare losses of about 6–10% of GDP.
This study investigates the causal effect of exporting on working conditions in Myanmar by drawing on a new firm survey. For identification, I use rapid opening of Myanmar to foreign trade alongside the firm’s products and proximity to airports that have generated variations in the access to foreign markets. The results show that exporting has large positive impacts on fire safety, healthcare management, and freedom of negotiation, weakly positive effects on wages, and insignificant effects on excessive working hours. I also find that exporting raises firm size, improves management practices, and increases the likelihood of receiving a labor audit.
発表タイトル: Large Matching in Large Market with Flexible Supply
要約
I study truthful mechanisms that achieve a large size of matching (expected number of agents matched to some objects) in an environment where the planner can decide which and how many objects to provide subject to linear upper bound constraints. Naive extensions of the classical serial mechanisms may generate an arbitrarily small matching in such environments. Assuming the market to be large (in that the variety of objects is fixed but the capacity to be large), I establish several mechanisms that have a constant guarantee for the size achieved. As a main result, I propose the adaptive parallel polymatroid serial mechanism, which (i) is not too computationally difficult to implement, (ii) achieves 1-1/e ≈ 63.2% of the maximum feasible size even in the worst case, and (iii) keeps agents' choice sets as large as possible.
(DPへのリンク)
アブスト：The theories of retail cluster formation suggest that stores perform better when surrounded by other stores of diverse complementary products because diverse stores attract consumers with love of variety preference. We analyze the impact of the diversity of neighboring stores among intermediate wholesalers located in the Tokyo Tsukiji Fish Market by exploiting a unique feature of their shop locations within the market in which their locations are determined every 4-10 years by relocation lotteries, generating exogenous variation in the diversity of neighboring stores. First, we confirm that these intermediate wholesalers’ shop locations are indeed randomly distributed. Then, we find that the diversity of the types of neighboring firms positively affect the performance of small-sized and specialized firms. We find no effect of the characteristics of close neighbors not facing the same corridor and thus not sharing the flow of shoppers. This provides evidence that our results are not due to factors other than shopping behavior, such as technology spillovers. Finally, to illustrate the general applicability of the mechanism we find, we use the Census of Commerce covering all the retailers in Tokyo to show that smaller and more specialized retailers are more likely to be located together while larger and standardized ones are isolated. Overall, our analysis shows that the complementarity of products between specialized diverse stores is an important factor behind urban agglomeration.
We consider an incomplete contracting model with a principal and an agent, in which the decision process consists of the project choice and the execution effort, and investigate the optimal allocation of authority of the project choice. Each party has an imperfectly informative private signal on the promising project, and successful execution requires the agent's effort. When the principal has authority, her project choice reflects her private signal, which in turn uncertainly influences the agent's confidence of the promising project choice. As the agent's confidence affects his incentive to exert execution effort, it may be optimal to allocate authority to the party who has less precise information on the promising project. Furthermore, we demonstrate that there may be a substantial gap between contractible delegation and informal empowerment even though the parties have no conflict on the promising project.
(DPへのリンク)
We propose a new specification test for assessing the validity of fuzzy regression discontinuity designs (FRD-validity). We derive a new set of testable implications, characterized by a set of inequality restrictions on the joint distribution of observed outcomes and treatment status at the cut-off. We show that this new characterization exploits all the information in the data useful for detecting violations of FRD-validity. Our approach differs from, and complements existing approaches that test continuity of the distributions of running variables and baseline covariates at the cut-off since ours focuses on the distribution of the observed outcome and treatment status. We show that the proposed test has appealing statistical properties. It controls size in large sample uniformly over a large class of distributions, is consistent against all fixed alternatives, and has non-trivial power against some local alternatives. We apply our test to evaluate the validity of two FRD designs. The test does not reject the FRD-validity in the class size design studied by Angrist and Lavy (1999) and rejects in the insurance subsidy design for poor households in Colombia studied by Miller, Pinto, and Vera-Hernandez (2013) for some outcome variables, while existing density tests suggest the opposite in each of the cases.
Using unique individual-level panel data, we investigate whether preventive medical care triggered by health checkups is worth the cost. We exploit the fact that biomarkers just below and above a threshold may be viewed as random. We find that people respond to health signals and increase physician visits. However, we find no evidence that additional care is cost effective. For the “borderline type” (“pre-diabetes”) threshold for diabetes, medical care utilization increases but neither physical measures nor predicted risks of mortality or serious complications improve. For efficient use of medical resources, cost effectiveness of preventive care must be carefully examined.
This paper assesses the importance of market thickness of input trade as a mechanism of agglomeration. Using a panel of firm-to-firm input trade data of over a million Japanese firms, I empirically show that, upon exogenous bankruptcies of suppliers (identified by the reported reasons of bankruptcy), buyer firms are less likely to exit and more likely to match with alternative suppliers in thicker supplier market. To quantify the implication of the reduced-form effect on the degree of agglomeration, I build a general equilibrium model of dynamic supplier-customer matching with endogenous entry of firms, where key structural parameters are estimated to match the reduced-form causal parameters. Counterfactual equilibrium simulation of shutting down the thick-market externality shows substantial part of industry agglomeration observed in Japan is explained by thick-market externality. I also discuss policy implications.
概要：
Lung transplantation is the only treatment for patients in the last stage of chronic lung diseases. Before April 2015, there were only two types of transplantation available: cadaveric-donor transplants and live-donor transplants. Ergin, Sonmez, and Unver (2017) have proposed the idea of exchanging donors only for live-donor lung transplantation. The new technology, called hybrid transplantation, is now available as Dr. Oto and his team at Okayama University Hospital succeeded transplanting a cadaveric lung and a lobe of live lung to one patient at the same time. We point out that the new technology plays a key role in operating the cadaveric- and live-donor markets at the same time. In particular, the hybrid transplantation opens up a new type of exchanging donors, and investigate a mechanism of organizing transplants in terms of efficiency, fairness and incentive-compatibility.
A growing body of literature emphasizes the role of childhood health in shaping labor market outcomes. Little is yet known whether poor health during childhood influences later outcomes by restricting skill formation, or by mainly affecting future health. To address this issue, this paper formulates and estimates a model of the joint dynamics of skills and health over the lifecycle. The estimated model is used to quantify the relative importance of the channels through which childhood health conditions affect labor market outcomes. In the model, individuals are endowed with a multi-dimensional human capital bundle that consists of skills and health. The human capital bundle evolves over time according to a production technology, with influences from endogenous decisions regarding schooling, labor supply, and occupations. The results indicate that the most important channel accounting for mental health-related earnings gaps is the skill channel. About 60-65% of the earnings gaps can be explained by the effects of childhood mental health conditions on skill formation. The effects of childhood health status on health formation are also found to play important roles.
This paper quantifies the welfare consequences of the medical arms race in the context of MRI adoption. We build and estimate a model of the vertical structure of the industry where MRI manufacturers sell high- and low-quality MRIs in the up- stream market, whereas medical institutions provide medical services to patients in the downstream market. Simulation results suggest that the current free-entry policy in Japan leads to excess MRI adoption. Furthermore, regulating medical institutions' MRI adoption, taxing MRI purchases, or softening competition among MRI manufacturers would increase social welfare substantially by mitigating the business-stealing effect in the downstream market.
We develop a novel approach to detect Beckerian positive assortative matching (PAM) of exporters and importers by capability. Conventional approaches examining firm characteristics across matches in cross-sectional data suffer from an endogeneity problem when firm characteristics reflect unobserved partner characteristics. Instead, using the entry of new exporters induced by trade liberalization as an exogenous shock to the capability rank of incumbent exporters, we investigate resulting re-matching patterns among incumbent exporters and importers. Examining Mexico–US textile/apparel trade that experienced a surge in Chinese exporters after the Multi-Fibre Arrangement’s end, we provide the first evidence for Beckerian PAM in exporter-importer relationships.