Large Bets and Stock Market Crashes
Contrary to conventional thinking, stock market crashes are neither random nor unpredictable, according to research from Albert “Pete” Kyle, the Smith School's Charles E. Smith Chair Professor of Finance, and Smith Assistant Professor of Finance Anna Obizhaeva.
The researchers developed a formula that can help predict the magnitude of price declines based on the size of bets relative to other market conditions. The framework, “market microstructure invariance,” sheds light on past crashes, and could help prevent future crashes if traders and policymakers take heed.
The volume and velocity of securities now traded in one day dwarf what was once traded in a full year. This makes it crucial for market players to follow reasonable trading algorithms and better understand how market depth is related to the volume and volatility of the markets in which they participate, Kyle says.
The working paper is “Large Bets and Stock Market Crashes.”
Social Influence Creeps into Movie Reviews
Film critics sometimes react not just to the film itself, but also to one another — which can alter the critics’ rating of the film and content of the review, according to research from Smith Associate Professor of Management David Waguespack and Smith PhD Daniel Olson.
The researchers analyzed 20,000 film reviews from Metacritic.com and looked at reviews of the same movie in pairs. They found when critics’ reviews came out on the same day, they were less likely to have read and been influenced by other’s reviews. But reviews filed on different days showed critics were influenced by other reviews they read, with ratings in 1 in 10 of the sets of reviews diverging by at least half a star on a four-star scale.
The researchers attribute the divergence to “a tension between wanting to be accurate — to provide the best, objective information — and wanting to differentiate oneself from peer reviewers.”
The working paper is “Competition and Social Influence among Critics.”
Flu or Something More Sinister? Using Computer Models to Find Out
A computer model developed by Sean Barnes, Smith assistant professor of operations management, aims to differentiate bioterrorism attacks from flu outbreaks — which have alarmingly similar symptoms — by their very different transmission dynamics.
Barnes built his original simulation model for his dissertation as a mathematics PhD student at the University of Maryland (2012) to help public health officials seeing the two scenarios play out and determine which they are dealing with. Now he is working with Bruce Golden, the France-Merrick Chair in Management Science, and another Maryland doctoral student to improve the model to incorporate additional aspects such as attacks in multiple locations, bioterrorism agents that spread between humans and information uncertainty.
The Centers for Disease Control track a number of real-time health data points to look for red flags. Barnes says, ideally, this model could be included as a module in that tracking system and potentially save lives in the event of an attack.
“Early Detection of Bioterrorism: Monitoring Disease Diffusion Through A Multilayered Network,” was published in Proceedings of the 2013 Industrial and Systems Engineering Research Conference.
Judging Borrowers by the Company they Keep
Your friends say a lot about you — and can even determine whether you can get a loan, according to new research from Siva Viswanathan, Smith associate professor of decision, operations and information technologies.
Viswanathan has been studying crowdfunding markets, websites that allow borrowers to seek loans from many individual lenders, and how well untrained investors that make up the “crowd” make funding decisions. His study looks at one of the first crowdfunding markets, Prosper.com, in which people seek loans typically less than $25,000.
Like traditional banks, crowdfunding lenders use “hard” data, such as credit history and FICA scores, to make the right choices. In Viswanathan’s study (conducted with Smith PhD Mingfeng Lin, now at the University of Arizona, and Smith associate finance professor Nagpurnanand Prabhala), borrowers with better scores got better outcomes. The researchers found that the crowd is also very good at using “soft” information about borrowers to make investment decisions.
This information includes the borrowers’ self-created profiles, details information about why they need the loan, and lists connections to “friends” in their social network. Having a lot of friends doesn’t necessarily matter for borrowers.
What does matter is whether a borrower’s friends are actually lending to them, or are at least willing to lend to them. These people are more likely to get funded and get better interest rates, and are less likely to default on loans, says Viswanathan. The researchers also found that default seems contagious: If a borrower’s friends have defaulted, that increases the likelihood he or she will default.
“Judging Borrowers by the Company they Keep: Friendship Network and Information Asymmetry in Online Peer-to-Peer Lending,” was published in Management Science in January 2013.