MARIANN OLLAR

mollar(at)ed.ac.uk

Hi! I am a Lecturer (Assistant Professor) at the School of Economics of the University of Edinburgh. Until 2018, I was an Assistant Professor at the University of Groningen and, until 2016, I was a postdoctoral researcher at the University of Pennsylvania.

My teaching fields are microeconomics, game theory, mathematics for economists and environmental economics.

I am a microeconomic theorist and my research is in Mechanism Design and Market Design. I like to think and talk about externalities, moment conditions in economic design, information aggregation, information design in exchanges, markets with centralized uniform price clearing, and game theoretic solution concepts such as rationalizability under various assumptions on beliefs and behavior.

CV: pdf

Education:

2014 - PhD in Economics (minor: Mathematics) - University of Wisconsin - Madison

2012 - Master of Science, Economics - University of Wisconsin - Madison

2009 - Master in Quantitative Economics (minor: Operations Research) - Corvinus University of Budapest

Publications:

[1] Full Implementation and Belief Restrictions, with Antonio Penta, pdf, American Economic Review, 2017.

Multiplicity of equilibria and the dependence on strong common knowledge assumptions are well-known problems in mechanism design. We address them by studying full implementation via transfer schemes, under general restrictions on agents’ beliefs. We show that incentive-compatible transfers ensure uniqueness - and hence full implementation - if they induce sufficiently weak strategic externalities. We then design transfers for full implementation by using information on beliefs in order to weaken the strategic externalities of the baseline canonical transfers. Our results rely on minimal restrictions on agents’ beliefs, specifically on moments of the distribution of types, that arise naturally in applications.

Keywords: Unique Δ-Rationalizability, Full Implementation, Belief Restrictions, Moment Conditions, Robust Mechanism Design

[2] Monotonicity and Robustness of Majority Rule, pdf, Economics Letters, 2010.

The majority voting rule is not superior on all domains when `monotonicity' replaces the requirement of `independence of irrelevant alternatives'. On some domains the Borda-rule outperforms the majority rule, nevertheless a diversity condition on the preference domain restores the superiority of the majority rule.

Keywords: Majority rule, Monotonicity, Well-working, Continuum Number of Voters, Diverse Domain

Working Papers:

[1] Implementation via Transfers under Identical but Unknown Distributions, with Antonio Penta.

In this paper we derive the implementation-related consequences of a weak informational assumption on agents' beliefs: we consider common knowledge of identicality of agents' beliefs, without assuming the knowledge of the distributions which describe these beliefs. That is, we assume that the designer knows that agents commonly know that other agents' types are identically distributed, but the designer might not know the correlations nor the distributions of these types. In single crossing environments with public concavity, we show that exploiting this common knowledge of identicality can ensure unique implementation. Full interim implementation is possible if and only if it is possible via transfers that load the canonical externalities to the agents who, in the baseline canonical direct mechanism, are least influenced by other agents' strategies. These loading transfers also minimize sensitivity to the applied moment conditions and ensure the best approximate k-level implementation. Additionally, we study the applicability of balancing transfers, which ensure equal dependence on other agents’ strategies. Under symmetric aggregators, we show that these balancing transfers are maximally robust to agents’ mistakes in play. Our results rely on a novel characterization of incentive compatible transfer schemes under various restrictions on beliefs. This characterization shows the necessity of moment conditions in transfer design for partial implementation. Further, an eigenvalue condition on the resulting strategic externalities provides sufficiency for uniqueness and hence full implementation.

Keywords: Moment Conditions, Robust Full Implementation, Rationalizability, Interdependent Values, Identical Distributions, Unknown Distributions, Uniqueness, Strategic Externalities

[2] Information Design and Stability in Exchanges.

This paper shows a limitation of information design in exchanges with uniform price market clearing. In these exchanges, improving traders' surplus through a better choice of the information structure conflicts with the stability of the exchange: traders' surplus is higher if information sources are less correlated, however, the less correlated the information sources are the closer the exchange is to the boundary of the set of stable exchanges. This paper also shows that under better information structures, traders' orders are closer to market orders.

Keywords: Information Design in Exchanges, Uniform Price Market Clearing, Price-Dependent Orders, Price Informativeness, Stability of Trade

[3] Measurement Errors and Information Aggregation in Exchanges.

In financial and commodity exchanges, traders gain information from shared information sources, such as commonly accessed forecasts and standardized reports, thus traders' information is likely to exhibit common bias in measurement errors. This paper shows in a linear normal model (following Kyle (89)) that when forecast errors are correlated, then this presence of non-iid. errors may prevent full aggregation of information, moreover, segmented trading can improve informativeness of the price. Because of learning about errors, the price is informative even on markets with independent values. The price is uninformative if the error correlations are the same as value correlations. This paper illuminates previous monotonicity results for iid. errors and common values, and more generally shows that the relationship between error and value correlations is key to monotonicity of information aggregation in market size.

Keywords: Information Aggregation, Correlated Forecast Errors, Price-Dependent Orders, Price Informativeness, Market Size

[4] Privacy in Markets, with Marzena Rostek and Ji Hee Yoon.

[5] Consumption Externalities and Competing Sellers: Efficiency Ranking of Market Structures.

The efficiency ranking of market structures is qualitatively affected by the nature of buyer side peer interactions. To show this, I formulate a model for a two-sided market with competing sellers and a network of buyers whose willingness to purchase varies according to their peers' purchases. Examples of such markets include competing small businesses, on-line music and video streaming services and publishers in advertisement exchanges. I allow for the joint presence of positive and negative peer influences and define overall positive and overall negative peer influence based on a vector aggregate of the exerted marginal influence on others' choices. I show that typical market structures (here competition in prices, competition in fees and price setting by a centralized platform) do not internalize the efficiency distortions caused by the externalities. In particular, with overall positive peer influence competition in membership fees dominates competition in prices while with overall negative peer influence this conclusion reverses.

Keywords: Efficiency Ranking, Network Externalities, Price Competition, Social Interactions

[6] Polynomial Equation Systems from Divisible Good Auctions, with David Dynerman.

Heterogenous information structures are prevalent in financial exchanges and divisible good auctions, where information may be propagated through core-periphery networks or originate from central forecasting agencies. We consider a model with uniform price market clearing that gives rise to a polynomial equation system. In the special case when each trader’s information relates to the market in the same way (`equicommonal’ information structure) this system simplifies to quadratic polynomials and has a unique closed form solution. However, in the general case this system exhibits multiple solutions. Contrary to the literature with fundamental values, price can no longer be expected to aggregate all dispersed information. The algebraic properties of the polynomial equation system and software tools (Bertini) help us identify the extent of multiplicity and its implications for information aggregation and trade efficiency.

[7] Rationalizable Implementation and Delta-Restrictions, with Antonio Penta.

We introduce a framework to study full implementation under general restrictions on agents' conjectures, which we call Delta-Implementation. These general restrictions are set-valued Delta-restrictions which, for instance, include Belief Restrictions as well as Behavioral Restrictions. We provide a characterization of the properties of general mechanisms that achieve Full Delta-Implementation and relate this to known notions of implementation.

Old Working Papers:

Shared Information Sources in Exchanges, pdf.

In financial and commodity exchanges, traders gain information from shared information sources, such as commonly accessed forecasts and standardized reports, which induce interdependence in forecast errors. In a linear normal model with noisy signals about values, I show that the presence of non-iid. errors improves trade stability when it amplifies order shading, and an increase in error interdependence improves price informativeness when it exceeds value interdependence. From a practical information-based market design perspective, source restrictions can prevent market collapse, and segmentation of trading venues can improve price informativeness. The gain in informativeness depends on the relative trader-to-trader interdependence of values and errors.

Keywords: Gaussian Normal Information Structure, Uniform Price Market Clearing, Common Bias, Correlated Errors

Privacy-Preserving Market Design, with Marzena Rostek and Ji Hee Yoon, pdf.

Preserving privacy of traders' information is a common concern in auctions and exchanges. Maintaining uniform-price market clearing, we study the joint design of transparency of auction outcomes (observables), bid schedules (contingent variables), and pre-trade communication. We show that additional learning from observable price and traded volumes is in conflict with stability of the exchange, hence there is a necessity for strong informational privacy. However, unlike observables, diverse contingent variables allow for both full informational efficiency and privacy and can be designed in a rich bidding language, for example, one with intercept-dependent bid functions.

Keywords: Intercept-Dependent Bid Functions, Divisible Good Auctions, Uniform Price Double Auction