|
on Computational Economics |
|
Issue of 2013‒09‒28
six papers chosen by |
| By: | Bobby Gramacy; Mike Ludkovski |
| Abstract: | We propose a new approach to solve optimal stopping problems via simulation. Working within the backward dynamic programming/Snell envelope framework, we augment the methodology of Longstaff-Schwartz that focuses on approximating the stopping strategy. We reinterpret the corresponding partitions of the state space into the continuation and stopping regions as statistical classification problems with noisy observations. Accordingly, a key new objective that we pursue is efficient design of the stochastic grids formed by the simulated sample paths of the underlying state process. To this end, we introduce active learning schemes that adaptively place new grid points close to the stopping boundaries. We then discuss dynamic regression algorithms that can implement such recursive estimation and local refinement of the classifiers. The new algorithm is illustrated with a variety of numerical experiments, showing that an order of magnitude savings in terms of total grid size can be achieved. We also compare with existing benchmarks in the context of pricing multi-dimensional Bermudan options. |
| Date: | 2013–09 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:1309.3832&r=cmp |
| By: | F. J. Fernández-Díaz; C. Patxot; G. Souto |
| Abstract: | This paper presents the results of DyPeS, the first dynamic microsimulation model of the retirement pensions system applied to the Spanish case. The simulation of the reform approved in 2011 shows that only the delay in retirement age (from 65 to 67) would have a significant effect on pension expenditure, while other measures changing the computation of the initial pension for new retirees have a limited impact. Paradoxically, it is found that the consideration of more contribution years in the computation of the initial pension amount, despite fostering the Bismarckian nature of the system, has a positive impact on redistribution. |
| Date: | 2013–09 |
| URL: | http://d.repec.org/n?u=RePEc:fda:fdaddt:2013-06&r=cmp |
| By: | Philippe Mathieu (LIFL - Laboratoire d'Informatique Fondamentale de Lille - CNRS : UMR8022 - Université Lille I - Sciences et technologies - Université Lille III - Sciences humaines et sociales - INRIA, LIFL - SMAC - Université Lille I - Sciences et technologies - CNRS : UMR8022); Sébastien Picault (LIFL - Laboratoire d'Informatique Fondamentale de Lille - CNRS : UMR8022 - Université Lille I - Sciences et technologies - Université Lille III - Sciences humaines et sociales - INRIA, LIFL - SMAC - Université Lille I - Sciences et technologies - CNRS : UMR8022, LIP6 - Laboratoire d'Informatique de Paris 6 - CNRS : UMR7606 - Université Pierre et Marie Curie (UPMC) - Paris VI) |
| Abstract: | L'usage croissant de la simulation multi-agents pour modéliser des systèmes pourvoyeurs de grandes quantités de données, suppose l'identification automatique des paramètres pertinents ou l'extraction de connaissances à partir des données réelles, faute de quoi la fiabilité des prédictions et des explications fournies par la simulation est sujette à caution. Dans cet article, nous proposons une méthode pour extraire automatiquement des profils comportementaux à partir de mesures statistiques, dans le cadre de comportements de consommateurs dans un magasin. Dotés des mêmes capacités globales d'interaction, les agents sont munis de profils différents issus de l'exploration des données. Placés dans un magasin virtuel réaliste, dans lequel tous leurs objectifs peuvent ne pas être atteignables, ils effectuent néanmoins des achats qui reflètent la diversité des clients réels ainsi que les profils initiaux. Nous défendons l'idée que de telles techniques sont nécessaires pour faire des simulations multi-agents un puissant outil d'aide à la décision. |
| Keywords: | simulation multi-agents ; exploration de données ; marketing ; interactions |
| Date: | 2013 |
| URL: | http://d.repec.org/n?u=RePEc:hal:journl:hal-00826405&r=cmp |
| By: | Peter C.B. Phillips (Cowles Foundation, Yale University); Shu-Ping Shi (Australian National University); Jun Yu (Singapore Management University) |
| Abstract: | This paper provides the limit theory of real time dating algorithms for bubble detection that were suggested in Phillips, Wu and Yu (2011, PWY) and Phillips, Shi and Yu (2013b, PSY). Bubbles are modeled using mildly explosive bubble episodes that are embedded within longer periods where the data evolves as a stochastic trend, thereby capturing normal market behavior as well as exuberance and collapse. Both the PWY and PSY estimates rely on recursive right tailed unit root tests (each with a different recursive algorithm) that may be used in real time to locate the origination and collapse dates of bubbles. Under certain explicit conditions, the moving window detector of PSY is shown to be a consistent dating algorithm even in the presence of multiple bubbles. The other algorithms are consistent detectors for bubbles early in the sample and, under stronger conditions, for subsequent bubbles in some cases. These asymptotic results and accompanying simulations guide the practical implementation of the procedures. They indicate that the PSY moving window detector is more reliable than the PWY strategy, sequential application of the PWY procedure and the CUSUM procedure. |
| Keywords: | Bubble duration, Consistency, Dating algorithm, Limit theory, Multiple bubbles, Real time detector |
| JEL: | C15 C22 |
| Date: | 2013–09 |
| URL: | http://d.repec.org/n?u=RePEc:cwl:cwldpp:1915&r=cmp |
| By: | Kang, Jee Eun |
| Abstract: | This dissertation focuses on the integration of the Household Activity Pattern Problem (HAPP) with various locational decisions considering both supply and demand sides. We present several methods to merge these two distinct areas—transportation infrastructure and travel demand procedures—into an integrated framework that has been previously exogenously linked by feedback or equilibrium processes. From the demand side, travel demand for non-primary activities is derived from the destination choices that a traveler makes that minimizes travel disutility within the context of considerations of daily scheduling and routing. From the supply side, the network decisions are determined as an integral function of travel demand rather than a given fixed OD matrix. First, the Location Selection Problem for the Household Activity Pattern Problem (LSP-HAPP) is developed. LSP-HAPP extends the HAPP by adding the capability to make destination choices simultaneously with other travel decisions of household activity allocation, activity sequence, and departure time. Instead of giving a set of pre-fixed activity locations to visit, LSPxviii HAPP chooses the location for certain activity types given a set of candidate locations. A dynamic programming algorithm is adopted and further developed for LSP-HAPP in order to deal with the choices among a sizable number of candidate locations within the HAPP modeling structure. Potential applications of synthetic pattern generation based on LSP-HAPP formulation are also presented. Second, the Location – Household Activity Pattern Problem (Location-HAPP), a facility location problem with full-day scheduling and routing considerations is developed. This is in the category of Location-Routing Problems (LRPs), where the decisions of facility location models are influenced by possible vehicle routings. Location-HAPP takes the set covering model as a location strategy, and HAPP as the scheduling and routing tool. The proposed formulation isolates each vehicle’s routing problem from those of other vehicles and from the master set covering problem. A modified column generation that uses a search method to find a column with a negative reduced price is proposed. Third, the Network Design Problem is integrated with the Household Activity Pattern Problem (NDP-HAPP) as a bilevel optimization problem. The bilevel structure includes an upper level network design while the lower level includes a set of disaggregate household itinerary optimization problems, posed as HAPP or LSP-HAPP. The output of upper level NDP (level-ofservice of the transportation network) becomes input data for the lower level HAPP that generates travel demand which becomes the input for the NDP. This is advantageous over the conventional NDP that outputs the best set of links to invest in, given an assumed OD matrix. Because the proposed NDP-HAPP can output the same best set of links, a new OD matrix and a detailed temporal distribution of activity participation and travel are created. A decomposed xix heuristic solution algorithm that represents each decision makers’ rationale shows optimality gaps of as much as 5% compared to exact solutions when tested with small examples. Utilizing the aforementioned models, two transportation sustainability studies are then conducted for the adoption of Alternative Fuel Vehicles (AFVs). The challenges in adopting AFVs are directly related to the transportation infrastructure problems since the initial AFV refueling locations will need to provide comparable convenient travel experience for the early adopters when compared to the already matured gasoline fuel based transportation infrastructure. This work demonstrates the significance of the integration between travel demand model and infrastructure problems, but also draws insightful policy measurements regarding AFV adoption. The first application study attempts to measure the household inconvenience level of operating AFVs. Two different scenarios are examined from two behavioral assumptions – keeping currently reported pattern and minimizing the inconvenience cost through HAPPR or HAPPC. From these patterns, the personal or household inconvenience level is derived as compared to the original pattern, providing quantified data on how the public sector would compensate for the increases in travel disutility to ultimately encourage the attractiveness of AFVs. From the supply side of the AFV infrastructure, Location-HAPP is applied to the incubation of the minimum refueling infrastructure required to support early adoption of Hydrogen Fuel Cell Vehicles (HFCVs). One of the early adoption communities targeted by auto manufacturers is chosen as the study area, and then three different values of accessibility are tested and measured in terms of tolerances to added travel time. Under optimal conditions, refueling trips are found to be toured with other activities. More importantly, there is evidence xx that excluding such vehicle-infrastructure interactions as well as routing and scheduling interactions can result in over-estimation of minimum facility requirement. |
| Keywords: | Engineering, household activity pattern problem |
| Date: | 2013–09–01 |
| URL: | http://d.repec.org/n?u=RePEc:cdl:uctcwp:qt3sb124zz&r=cmp |
| By: | Nicolas R. Blancher; Srobona Mitra; Hanan Morsy; Akira Otani; Tiago Severo; Laura Valderrama |
| Abstract: | There has recently been a proliferation of new quantitative tools as part of various initiatives to improve the monitoring of systemic risk. The "SysMo" project takes stock of the current toolkit used at the IMF for this purpose. It offers detailed and practical guidance on the use of current systemic risk monitoring tools on the basis of six key questions policymakers are likely to ask. It provides "how-to" guidance to select and interpret monitoring tools; a continuously updated inventory of key categories of tools ("Tools Binder"); and suggestions on how to operationalize systemic risk monitoring, including through a systemic risk "Dashboard." In doing so, the project cuts across various country-specific circumstances and makes a preliminary assessment of the adequacy and limitations of the current toolkit. |
| Keywords: | Financial sector;Financial risk;Financial institutions;Spillovers;Financial crisis;Cross country analysis;Sytemic Risk; Risk Indicators; Risk Monitoring; Macroprudential Policy |
| Date: | 2013–07–17 |
| URL: | http://d.repec.org/n?u=RePEc:imf:imfwpa:13/168&r=cmp |