What AI Can Do in Economics?
Paper Session
Friday, Jan. 6, 2023 2:30 PM - 4:30 PM (CST)
- Chair: Christopher Carroll, Johns Hopkins University
Deep Learning: Solving HANC and HANK Models in the Absence of Krusell-Smith Aggregation
Abstract
An incomplete-market heterogeneous-agent neoclassical model (HANC) studied by Krusell andSmith (1998) has savings through capital. This model has a remarkable feature of approximate
aggregation: the mean of wealth distribution can be accurately predicted with the mean of past
wealth distribution. However, if savings are done through bonds, the HANC model does not have
this feature (because the mean of bond holding is zero). We solve such model using deep
learning solution method in which the decision function and price functions are approximated in
terms of true state space of individual and aggregate state variables. The problem has high
dimensionality (hundreds of state variables) and ill-conditioning but neural network reduces
dimensionality and restores numerical stability. Our deep learning method delivers accurate and
reliable solutions. We also show how to solve a heterogeneous-agent new Keynesian (HANK)
model with savings through bonds and a zero lower bound on nominal interest rates in the
absence of Krusell and Smith's (1998) aggregation.
Programming FPGAs for Economics: An Introduction to Electrical Engineering Economics
Abstract
We show how to use field-programmable gate arrays (FPGAs) and their associated highlevelsynthesis (HLS) compilers to solve heterogeneous agent models with incomplete markets
and aggregate uncertainty (Krusell and Smith, 1998). We document that the acceleration
delivered by one single FPGA is comparable to that provided by using 74 CPU cores
in a conventional cluster. We describe how to achieve multiple acceleration opportunities
-pipeline, data-level parallelism, and data precision- with minimal modification of the C
code written for a traditional sequential processor, which we then deploy on FPGAs easily
available at Amazon Web Services. We quantify the speedup and cost of these accelerations.
Our paper is the first step toward a new field, electrical engineering economics,
focused on designing computational accelerators for economics to tackle challenging quantitative
models.
Testing for Asymmetric Information with Neural Networks
Abstract
The positive correlation test for asymmetric information developed by Chiappori and Salanié (2000) has been applied in many contexts. Most of the literature only allows for a constant correlation; and they use parametric specifications for the choice of coverage and the occurrence of claims. Estimating conditional covariances non-parametrically is hard; and testing restrictions on variable correlation coefficients is a highly multiple testing problem. We relax these limitations using deep learning methods. We use a neural network to classify insurees into joint choice of coverage and claim occurrence categories. Then we use the double-debiasing methods of Chernozhukov et al. (2018) to obtain a consistent and asymptotically normal estimator of the correlation function. Finally, we apply the intersection test of Chernozhukov et al. (2013) to the hypothesis that this correlation function only takes non-negative values. We conclude with an examination of the power of our procedure against small negative correlations when the correlation coefficient is assumed to be constant.Discussant(s)
Alexander Chinco
,
CUNY-Baruch College
Yucheng Yang
,
Princeton University
Toni Whited
,
University of Michigan
Vira Semenova
,
University of California-Berkeley
JEL Classifications
- C6 - Mathematical Methods; Programming Models; Mathematical and Simulation Modeling
- E0 - General