![Show Menu](styles/mobile-menu.png)
![Page Background](./../common/page-substrates/page0470.png)
INFORMS Nashville – 2016
468
WD41
207C-MCC
Risk in Financial Markets
Sponsored: Financial Services
Sponsored Session
Chair: Daniel Mitchell, University of Minnesota, University Avenue,
Minneapolis, MN, 55455, United States,
damitche@umn.edu1 - Systemic Risk Of High-frequency Trading
Agostino Capponi, Columbia University,
ac3827@columbia.eduWe introduce a dynamic high-frequency trading model which accounts for the
costs of overnight inventory. The HFT optimally and continuously chooses bid and
ask prices in order to maximize end-of-day expected profits, net of inventory
costs. The model pits the HFT’s profit maximizing motives against its desire to
avoid carrying inventory overnight, which effectively generates a tradeoff. We
show that the tradeoff, which is unique to the business model of HFTs, leads to
destabilizing price dynamics.
2 - Determining Estimation Risk Using Distributional Properties Of
Portfolio Weights
Luis Chavez-Bedoya, Esan Graduate School of Business,
lchavezbedoya@esan.edu.peUsing the expected loss function of Kan and Zhou (2007), we find closed-form
expressions to determine the impact of parameter uncertainty on the
performance of the minimum-variance and the optimal mean-variance portfolio
but when these portfolios are fully invested in risky assets. The mathematical
proofs of the closed-form expressions are based on distributional properties of the
portfolio weights instead of distributional properties of the sample mean and
covariance matrix. In the numerical experiments, we assess the impact on
estimation risk when the risk-free asset is not included in the portfolio
construction.
3 - Modeling Limit Order Books With Neural Networks
Justin Sirignano, Stanford,
jasirign@gmail.comThis paper develops a new neural network architecture for modeling spatial
distributions (i.e., distributions on R^d) which is computationally efficient and
takes advantage of local spatial structure. We find statistical evidence for local
spatial structure in limit order books, motivating the new neural network’s
application to limit order books. The neural network is trained and tested on
nearly 500 stocks. The neural network uses information from deep into the limit
order book (i.e., many levels beyond the best bid and best ask). Techniques from
deep learning such as dropout are employed to improve performance. Due to the
computational challenges associated with the large amount of data, GPU clusters
are used for training. The “spatial neural network” is shown to outperform other
models such as the naive empirical model, logistic regression (with nonlinear
features), and a standard neural network architecture.
4 - Liquidation Risk
Daniel Mitchell, University of Minnesota,
damitche@umn.edu,
Jingnan Chen
We examine risk management in a portfolio liquidation setting. We consider a
model of market and limit order execution and investigate trading profiles of risk
averse traders. Our primarily interest is to determine when market orders are
preferred to limit orders in execution. Market orders can reduce variation in price
but also come at a higher expected cost.
WD42
207D-MCC
Sharing Economy, Mechanism Design and Networks
II
Sponsored: Revenue Management & Pricing
Sponsored Session
Chair: Ozan Candogan, University of Chicago, Chicago, IL,
United States,
ozan.candogan@chicagobooth.eduCo-Chair: Santiago Balseiro, Duke University, Durham, NC,
United States,
sbalseiro@gmail.com1 - The Impact Of Platform Control Capabilities On The Performance
Of Rideshare Networks
Zhe Liu, Columbia Business School, 511 W 112th Street, Apt 24C,
New York, NY, 10025, United States,
liuzhe821@gmail.com,Costis Maglaras, Philipp Afeche
We are motivated by the rise of rideshare platforms such as Uber and Lyft, that
match service providers (drivers) with demand (riders) over a network. A key
challenge is that such platforms face supply/demand imbalances. To manage
performance, the platforms have several control capabilities, specifically, they can
decide a) which demand requests to accept at each location, and b) which
capacity to reposition from one location to another. This paper studies within a
stylized network model the impact of these control levers on key performance
measures, including the revenue rate, congestion, lost demand (riders), and
idleness time (drivers), taking into account the network’s flow dynamics.
2 - Competitive Equilibrium And Trading Networks:
A Network Flow Approach
Ozan Candogan, University of Chicago,
ozan.candogan@chicagobooth.edu, Markos Epitropou,
Rakesh Vinay Vohra
In trading networks where agents exchange indivisible goods (or indivisible
contracts), recent literature has established that under a full substitutability
condition on agents preferences, a competitive equilibrium exists. Moreover,
competitive equilibria of trading networks are also stable outcomes, which is
equivalent to the seemingly weaker chain stability condition. This paper’s
contribution is to show that under the full substitutability assumption, all these
results can be obtained simply and directly from the optimality conditions of a
generalized submodular flow problem in an appropriately defined network.
3 - Mean Field Equilibria For Competitive Exploration In Resource
Sharing Settings
Krishnamurthy Iyer, Cornell University,
kriyer@cornell.edu,
Pu Yang, Peter Frazier
Inspired by crowdsourced transportation services and other location-based
activities, we consider a model comprising of a group of nomadic agents and a set
of locations each endowed with a dynamic stochastic resource process. Each agent
derives a periodic reward based on the overall resource level at her location, and
the number of other agents there. Each agent is free to move between locations,
and at each time decides whether to stay at the same location or switch to
another one. We study the equilibrium behavior of the agents as a function of
dynamics of the stochastic resource process and the nature of resource sharing in
the limit where the number of agents and locations increase proportionally.
4 - On The Efficacy Of Static Prices For Revenue Management In The
Face Of Strategic Customers
Yiwei Chen, Singapore University of Technology and Design,
Singapore, Singapore,
ywchen@mit.edu, Vivek Farias
We consider a revenue management problem wherein a monopolist seller seeks
to maximize revenues from selling a fixed inventory of a product to customers
who arrive over time. Customers are forward looking and strategize their times of
purchase. We consider a general class of customer utility models that allow for
multi-dimensional customer types. We also allow for a customer’s disutility from
waiting to be positively correlated with his valuation. We show that static prices
are asymptotically optimal. We further show that irrespective of regime, an
optimally set static price captures at least 63.2% of revenue under an optimal
dynamic mechanism.
WD44
208B-MCC
Advances In Risk Modeling Theory:
Nonlinear Systems
Sponsored: Decision Analysis
Sponsored Session
Chair: Ghorbanmohammad Komaki, Case Western Reserve University,
Cleveland, OH, United States,
gxk152@case.eduCo-Chair: Behnam B Malakooti, Case Western Reserve University,
Cleveland, OH, United States,
bxm4@po.cwru.edu1 - Storage Impact On Micro-grids With Renewable Energy Sources
Shaya Sheikh, New York Institute of Technology, 1855 Broadway,
New York, NY, United States,
ssheik11@nyit.eduIntegrating renewable energy sources and energy storages in micro-grid has
captured the attention of researchers in recent years. We investigate the impact of
energy storages on energy costs and thermal comfort in a micro-grid with
heterogeneous buildings. Our proposed model features two electricity generators
(e.g., wind and solar). Due to the stochastic nature of both renewable energy
sources and energy demand, a simulation approach is proposed to analyze this
model. The proposed model reduces total energy cost while it achieves the
thermal comfort requirements of residents.
2 - A Brief Survey Of Recent Decision-making Models
And Experiments
Mohammad Komaki, Case Western Reserve University,
komakighorban@gmail.com, Behnam Malakooti
Decision-making under risk has a long history and is one of the challenging areas
in many fields including economics, finance and engineering. Technically,
decision-making is the selection of an alternative among group of alternatives.
Several models have been developed to assist decision-makers (DMs) in the
presence of risk, for instance, Expected Utility Theory, Cumulative prospect
theory and so on. Recently, several models have been proposed. In this study, we
investigate these models and their properties. Also, we investigate their
performances in term of resolving the well-known paradoxes.
WD41