Peter Bossaerts - US grants

General equilibrium theory underlies our understanding of how complex economies manage risk. It is the basis of the tools used to make investment decisions and to guide policy in applications as diverse as regulated industries (utilities, telecommunications, etc.) and pension funds. In spite of this widespread use, scientific support for such applications has been mixed. The research proposed here seeks to build and test a more robust implementation of general equilibrium theory. Our work integrates theory, experiment, and econometrics in an essential way: experiments suggest the shape the theory should have, the theory suggests which further experiments should be carried out, experiments provide tests of the final theory; econometrics links theory and experiment in a formal way.

We have developed and succesfully conducted laboratory experiments that test the principles of general equilibrium theory in the context of markets with risk. The model that guided our inference is the Capital Asset Pricing Model (CAPM). The experiments confirm the complex pricing relationships predicted by the theory. Since we used the same measurement tools as in the analysis of historical data in the field, our findings suggest that the controversy surrounding historical evidence need not be attributed to a failure of the basic principles of the theory, but to the auxiliary assumptions added to the models to make them testable on field data. Still, we reject the portfolio (allocation) implications of the same theory. The latter finding is particularly perplexing because the traditional theory that explains and supports the prices rests explicitly on the allocational predictions.

Our theory explains the puzzle. It perturbs traditional general equilibrium models in ways that better accomodate the types of behavior of individuals when observed in isolation, away from markets. We provide a unified approach to study the effects of such perturbations. We show, for instance, that random perturbations need not always wash out in large economies; in the case of information aggregation, they cause clearcut biases. Preliminary experiments confirm our predictions. We plan to build on these initial successes, to further our understanding of pricing and allocation of risk in competitive markets. In addition, the theory suggests experiments that will uncover the origin of the perturbations that are needed to make general equilibrium models explain the data. One of these will determine whether risk is like any other commodity and that no special principles must be invoked in order to understand market behavior, a fundamental premise of extant theory. Actual attitudes towards and beliefs about risk may be at odds with this view.

In addition to its scientific and policy implications, this proposal also has an important educational component. Our experiments are web-based and have involved many subjects from more than a dozen undergraduate, graduate, and professional schools across the country. These experiments provide the subjects with a unique educational experience. Our experiments have shown that access to such a large pool of subjects with diverse backgrounds is necessary to achieve results, and the theory explains why. At the same time, it provides exposure to complex financial markets to students who would otherwise not be given the opportunity, for various geographical and socio-economic reasons.

This is a collaborative proposal involving Peter Bossaerts and Charles Plott (both of Caltech) and William Zame (of UCLA).

Prop ID: 0317715
P I: Plott, Charles R.
Organization: California Institute of Technology
Title: Collaborative Research: The Evolution of Prices and Allocations in Markets: Theory and Experiment
CO-PI: Peter Bossaerts/Cal Tech; William Zame/UCLA (Collaborative)

The goal of the proposed research is to understand the dynamics of prices and choices in multi-security/multi-good markets. Do markets get to equilibrium, and how? This research concentrates on a simple environment: trade takes place during some period of time, but information is revealed and consumption takes place only at the end of the period. Within this context, the investigators study price and allocation dynamics experimentally and discover a number of regularities in the data: (i) prices eventually settle close to the theoretical equilibrium; (ii) price changes for a particular good are correlated with excess demand of all goods; (iii) trades are small. Motivated by these empirical findings, the investigators propose a local adjustment model, in the spirit of local models suggested by Smale, to explain these phenomena. The purpose of this project is to study the theoretical properties of the model and to link it to data through a series of experiments.

Because agents in the model make adjustments that are only locally optimal, the model is
in the spirit of much work on limited rationality. However, the end process of the adjustment process is quite rational, and the adjustment process is more robust than most models in the literature. Thus the model illustrates how markets can be smart even though they are populated by human beings of limited capacity. The proposed research is meant to provide a more solid empirical foundation for our theoretical understanding of the evolution of prices and choices in competitive markets for several goods/securities. At the core of the research is a dialogue between theory and data, with econometrics providing the link.

Broader Impacts: The issues studied here are critical for the many important existing markets in which several goods (including services and risky assets) are traded at the same time and in which the supply, demand and price of each of these goods depends in a complicated way on the supply, demand and price of other goods. The proposed work should lead to a better understanding of such interdependencies and therefore to a better understanding of the management of risks in such markets. The experiments proposed here depend on a web-based experimental technology (using unique facilities to which the research team has access) that permits the interaction of many participants in real time. Although technical challenges will continue to arise, there is no doubt that this experimental technology can be adapted to meet them. In addition to its usefulness as a research tool, this experimental technology will be useful as a teaching tool, and for providing broad based demonstrations of the science.

Financial markets have long been known to play a crucial role in societal re-allocation and diffusion of risk. Recently, financial markets have been observed contributing to social cognition as well. Information is transmitted, problem solving is influenced, and individual inference is affected. The mechanics by which financial markets contribute to social cognition are not well understood. Neoclassical economic theory assumes that market participants can rationally infer information from others through transaction prices. But the very rationality on which such inference is based should make market participants wary of trading. Unfortunately, if there is no trade, there are no transaction prices, and hence, nothing is revealed. Social cognition is impossible.

People trade -- in fact they trade a lot -- but we do not know why. Correlation analysis of order and trade flows and subsequent actions has not provided much insight. Nor have surveys helped much, suggesting that actions may be largely sub-conscious. If so, direct measurement of sub-conscious changes in perceived risk and reward may be a necessary first step towards resolving the trading puzzle and eventually understanding the role of asset markets in social cognition. Recently, scientists have discovered how changes in expected reward and risk induces specific responses in certain sub-cortical parts of the brain. The PIs plan to reverse this approach, exposing subjects to market activity while monitoring brain activity. The goal is to detect features in order and trade flows that trigger changes in perceived risk and reward as reflected in brain activity. The approach borrows from the neuroscience of vision, where scientists have successfully been able to identify the sources of changes in visual perception even in environments as complex as full-feature movies.

Competitive financial markets play an absolutely central role in the economy. They provide the means by which society shares risk and allocates capital. Existing theoretical models of such markets are elegant and beautiful, but fare badly when confronted with actual data. The work proposed uses laboratory experiments to explore several possible explanations for the disconnection between theory and fact: (a) that some agents are better informed than others, (b) that some agents do not use available information as well as others, and (c) that some agents suffer from cognitive biases.

This work shows that experiments can play a role in finance entirely analogous to the role they play in the physical sciences, making it possible to create and study complicated systems in a controlled setting, eliminating "frictions" and noise, studying the effect of changing a few variables while holding others constant, and exploring counterfactuals. The insights to be gained from this research have implications for investment and for government policy and regulation, including social security.

In this research project, the PIs investigate the possibility that social rationality explains the emergence of one type of bubble in competitive asset markets: a bubble referred to as a "credit market bubble." The bubble is defined as a situation where (i) the debt is priced above its commonly known intrinsic value and (ii) the debt is rolled over even though each creditor should cash in because everyone knows that the debtor will never be able to repay. Building on evidence from behavioral game theory, the PIs conjecture that such credit market bubbles emerge whenever the debtor's payment ability, although never sufficient, grows over time. Pilot experimental data confirm the emergence of bubbles in this setting. The researchers will conduct experiments to further examine the robustness of bubbles in this environment and test the hypothesis that norms are driving the observed behavior.

In terms of broader impacts, this research will provide a better understanding of price bubbles. Credit bubbles and accompanying asset price run-ups re-occur with alarming frequency in the real world. This research suggests that tension between individual and social rationality is the root cause for their existence. It will lead to a better understanding of this ubiquitous phenomenon in modern capitalist society, and inspire novel and effective government policy and regulation to minimize their negative effects.

US-German Collaboration: Computational and Neural Mechanisms of Inference over Decision-Structure

PI: John O?Doherty, Co PI: Peter Bossaerts

ABSTRACT

In this project the Principal Investigators will determine how the brain is able to identify the relevant rules that apply to a given decision-making problem in order to effectively make decisions. In most cases, features of the decision structure are hidden variables, i.e. they can be inferred only through discrete observations of outcome variables (such as reward feedback). Understanding how inferences over decision-structure are performed in a noisy and partially observable environment is therefore a fundamental yet almost unaddressed issue in the computational neurobiology of human decision-making. Here, we conceive of inferences over decision problems as a form of hierarchical inference in which the higher level of the hierarchy represents probabilistic beliefs over which decision structure is currently in place, while the lower level of the hierarchy encodes beliefs over which actions are currently rewarded within a specific decision structure. We will compare and contrast a variety of computational models deploying different strategies to solve this problem. We will combine these models with behavioral and functional magnetic resonance imaging (fMRI) data from human participants in order to address whether dynamic signals are present in the brain pertaining to the implementation of such hierarchical models, and whether different brain regions are involved in performing inference at different levels of the hierarchy. This project could potentially lead to a new understanding of the contribution of the prefrontal cortex and other brain regions in decision-making. This project will also provide insight into the neural implementation of a fundamental missing part of the picture concerning the neurobiology of human decision-making: decision-structure inference.

In terms of broader impacts, this research could provide fundamental new insights into understanding situations where human learning or decision-making fails or breaks down. Sometimes poor learning or decision-making may be due to a failure to infer the correct rules governing a decision-problem rather than a difficulty in learning or deciding per se. Such insights will not only impact on academic fields studying decision making but could also be used to develop novel methods to help individuals and organizations make better decisions (by focusing on improving inference over structure). The findings could provide relevant data for the development of artificial agents capable of autonomous, flexible and adaptive decisions. The proposal also has high potential clinical relevance: disorders with delusional beliefs such as schizophrenia and borderline personality disorder might involve in part a difficulty in performing inference over decision structure so as to rule out inappropriate (decision) structures in lieu of more appropriate ones. The present research might yield novel tools to study this question in clinical populations. Furthermore, there are substantial impacts on teaching and training. The PI and co-PI teach courses at undergraduate and graduate level and involve undergraduate researchers directly in their research programs. The work proposed here could lead to the development of new software to enable the analysis of brain imaging data using computational models. A companion project is being funded by the German Ministry of Education and Research (BMBF).

Starting with Chamberlin in the 40s, experiments with markets have made economists pessimistic about the merits of decentralized markets, in sharp contrast with centralized markets. The centralized continuous double auction has emerged as the mechanism most advocated by economists to generate the beneficial outcomes associated with competitive equilibrium. Buttressed by recent advances in theoretical modeling of decentralized markets, our experiments have started to paint a different picture, with decentralized markets generating outcomes that are not much different from those of centralized markets. Here, we propose to investigate the very dimension in which decentralized markets have been proposed to improve upon centralized markets, namely, in providing sustained incentives to pay for (inside) information (within the right economic setting, of course) the theory does not claim that decentralized markets will always be better). This contrasts with centralized markets, which theory and experiments have argued lead to the Grossman-Stiglitz paradox (if information is costly, prices cannot be informative). The negativity with which economists generally depict the workings of decentralized markets has affected policy making, certainly since the Great Financial Crisis, to the extent that such markets are now generally shunned, or even, as in the recent European directive MiFID 2, disallowed. Our experiments are meant to bring hard evidence to the table. They should illustrate the possibility of evidence-based policy making in finance.


Decentralized financial markets have been deemed detrimental to efficient and fair pricing because of their lack of transparency (whence the synonym "dark markets"). Recent legislation in both the U.S. and in the E.U. is gradually forcing all trading onto centralized markets, or multilateral trading platforms, where everyone can see, in a timely fashion, pretty much everything that is going on (order submission, executed trades, etc.). However, controlled experiments with financial markets have confirmed a theoretical prediction from the 70s, which is that centralized markets fail to provide incentives to collect information (when this is costly) and hence, centralized markets can at best generate only noisy prices. In contrast, more recent theoretical modeling argues that in certain settings, decentralized markets would actually generate the right incentives, and as a result, prices would be more accurate than in centralized markets. If this is true, the gradual elimination of "dark markets" might not have been a good policy. We propose to study price discovery with costly information acquisition in decentralized markets through controlled experimentation. Our experiments would advise whether recent financial markets regulation may have to be re-examined. While inspired by theoretical reasoning, our recommendation will be evidence-based, and as such, would break with the tradition in rule making in finance, which has been almost entirely model-based.

Asset pricing theory has recently been criticized for its inability to explain historical data from field markets. Yet its poor record does not mean that it is scientifically invalid. Our experience (as experimentalists) is that theorists are not aware of the scientific record of asset pricing theory (how well it works in a controlled setting). The aim of the workshop is to sensitize theorists to this record. It should make them aware not only of the successes (of the theory), but also of where the theory fails (and how the failures can be addressed). It is hoped that the workshop heralds a new era where theorists work in closer collaboration with experimentalists. The PIs have a long track record not only in financial markets experimentation, but also in the experimental method in general (having contributed successfully to neuroscience), and they have published in asset pricing theory. As such, they are uniquely positioned to moderate the proposed dialogue between theorists and experimentalists.

Financial economics is rather abstract and mathematical, and its value is difficult to ascertain from merely observing real-world financial markets, which operate in a complex environment where many key variables either remain unobserved or cannot be measured reliably. In the last decade, however, tools have been developed to study financial markets in the laboratory, where real people trade for real money. The controlled setting is designed to emulate the theoretical context and as such has proven to be an ideal testing ground. The workshop aims at bringing together theorists and experimentalists in order to start a dialogue. Experimentalists are to be taught what aspects of the theory are defining, and hence, need to be tested, and theorists are to be encouraged to explain their models in terms of experiments with which to gauge scientific validity, which requires theorists to understand the experimental approach. We thereby import into finance a longstanding tradition from the physical sciences. The goal is to move finance to an evidence-based discipline. This will ultimately benefit financial markets policy formulation and rule making, which until now have primarily been model-based, or in the rare occasions where the data were available, informed only by empirical analysis of historical data.