July 14-18: HERBERT HOOVER MEMORIAL BUILDING
Monday, July 14: Room 130 (Enter front door, turn right)
************************************************************************************************
10:00am-11:30: Simon Scheidegger “(Peta-) Scalable High-Dimensional Dynamic Stochastic Economic Modeling”
Abstract:
We present a highly parallelizable and flexible computational method to solve high-dimensional stochastic dynamic economic models. By exploiting the generic structure of such problems, we propose a parallelization scheme that favors hybrid massively-parallel computer architectures. Within a novel parallel nonlinear time-iteration framework, we interpolate policy functions partially on GPUs using an adaptive sparse grid algorithm with piecewise linear hierarchical basis functions. Due to the high arithmetic intensity of the interpolation, GPUs accelerate this part of the computation one order of magnitude, thus reducing overall computation time by 50%. The developments in this paper include the use of a fully adaptive sparse grid algorithm and the use of a mixed MPI- Intel TBB – CUDA/Thrust implementation to improve the interprocess communication strategy on massively parallel architectures. Numerical experiments on “Piz Daint” (Cray XC30) at CSCS show that our code is very scalable and flexible. In the case of our intermediate-sized, 8-dimensional IRBC benchmark, we found very good strong scaling properties up to the order of 10,000 threads. Weak scaling tests indicate a nearly perfect behavior up to 41,024 threads. This all suggests that our framework is very well suited for large-scale economic simulations on massively parallel high-performance computing architectures.
************************************************************************************************
************************************************************************************************
1:30-3:00pm: Philipp Renner: “Quantity precommitment & price competition: A dynamic approach”
Abstract:
In this paper we look at a new way to combine both quantity precommitment and price competition in a dynamic games framework. In each period players choose to augment or dismantle production capacities and also engage in a price competition. Production is free up to capacity and has steeply increasing costs if exceeded. The demand is allocated according to a quadratic utility function, which allows for product differentiation. By using this approach we avoid rationing rules, which are frequently used in these kind of models. We also show that if costs for exceeding production go to infinity, then the equilibrium converges to a degenerate case, where production is free up to capacity and excess has infinite marginal cost. This limiting approach allows us to compare our model to the ones used in the literature, with homogeneous goods and rationing.
************************************************************************************************
Wednesday, July 16: Room 130 (Enter front door, turn right)
************************************************************************************************
10:45am-noon: TJ Canann: “Parallel Implementation of the Modular Approach to Grobner Bases”
Abstract:
************************************************************************************************
************************************************************************************************
2:00-3:00pm: Phillipp Ott “A parallelization wrapper for AMPL”
Abstract:
************************************************************************************************
************************************************************************************************
3:15-4:00pm: Yongyang Cai: “Nonlinear Certainty Equivalent DP Solutions: Death to LogLinear Approximations!”
Abstract:
************************************************************************************************
Friday, July 18: Room 330 (Enter front door, go up two floors, turn left)
************************************************************************************************
9-10:15am:
************************************************************************************************
************************************************************************************************
10:45am-noon: Brett Clarke: “Acceleration Methods for Dynamic Programming”
Abstract:
************************************************************************************************
************************************************************************************************
2:00-3:15pm: Karl Schmedders: “Replicating Portfolios”
Abstract:
************************************************************************************************
************************************************************************************************
3:45-5:00pm:
************************************************************************************************
Every day: Judd: coming up with pesky questions
====================================================
====================================================
July 21-25: HERBERT HOOVER MEMORIAL BUILDING
Monday, July 21: 350 HHMB
Due to various conflicts, no suitable room was available. We have our presentations begin on Tuesday.
Otherwise, you may want to come to Room 350 HHMB.
Tuesday, July 22: 238 LHH
************************************************************************************************
9-10:15am: Bulat Gafarov and Konstantin Kucheryavyy: “Software for GSSA”
Abstract:
************************************************************************************************
************************************************************************************************
10:45am-noon: Ben Tengelsen: “When is Risk Sharing Beneficial for Everyone?”
Abstract:
************************************************************************************************
************************************************************************************************
2:00-3:15pm: Sevin Yeltekin: “Computing Equilibria of Dynamic Games”
Abstract:
************************************************************************************************
************************************************************************************************
3:45-5:00pm: Max Werner
************************************************************************************************
Wednesday, July 23: Room 110, Hoover Tower
************************************************************************************************
9:15-10:30am: Chris Sleet
************************************************************************************************
************************************************************************************************
10:45am-noon: Gregor Reich: “Adaptive Grids for Estimation of Dynamic Models using Constraint Optimization Approaches”
Abstract:
This paper develops a method to flexibly adapt interpolation grids of value function approximation in dynamic models, such as dynamic discrete choice models, when the estimation is done using constraint optimization, namely the MPEC approach. While often being more efficient compared to the classical nested fixed point (NFXP), the MPEC approach needs the structure of the value function approximation to be hardcoded into the constraints of the likelihood optimization problem. As a consequence, one cannot use iteratively adaptive procedures for grid refinement in every iteration of the optimization, as it is possible (and has been reported to be numerically favorable) in NFXP estimation. In this paper, we show how to adapt the interpolation grid by moving the nodes, a technique called r-adaptive refinement. We demonstrate how to obtain optimal grids (given a fixed number of nodes), and show how to integrate this approach into the likelihood maximization problem using the equi-oscillation principle. The method is applied to the bus engine replacement model of Rust (1987), modified to feature a continuous mileage state.
************************************************************************************************
************************************************************************************************
2:00-3:15pm: Alena Miftakhova, “Statistical climate emulation via MAGICC complexity reduction”
Abstract:
************************************************************************************************
************************************************************************************************
3:45-5:00pm
Abstract:
************************************************************************************************
Friday, July 25: HHMB 330
************************************************************************************************
9-10:15am: Karl Schmedders: “Polynomial Optimization Approach to Principal-Agent Problems”
************************************************************************************************
************************************************************************************************
10:45am-noon: Richard Cottle: “William Karush and the KKT Theorem: A Chicago Story”
Abstract:
************************************************************************************************
Noon-1:30pm: Group Lunch: Annenberg Room in Lou Henry Hoover Building
************************************************************************************************
2:00-3:15pm: Larry Kotlikoff
Abstract:
************************************************************************************************
************************************************************************************************
3:45-5:00pm: Felix Kubler: “Finding the right state”
Abstract: When applying recursive methods to non-optimal dynamic economies with heterogeneous agents one needs to specify an endogenous state space.
I discuss four possible choices: Beginning-of-period asset holdings across agents, beginning-of-period financial wealth across agents, recursive instantaneous Negishi-weights and current spot prices. I try to compare these approaches in terms of existence of Markov equilibrium, computational advantages/disadvantages, and economic interpretation.
************************************************************************************************
====================================================
====================================================
July 28-Aug. 1: SITE WEEK, ECON DEPARTMENT
See SITE website
====================================================
====================================================