Search

Like Feinberg and Sherman (1985) and Phillips and Mason (1992) we test experimentally whether conglomerate firms, i.e., firms competing on multiple structurally unrelated markets, can effectively limit competition. Our more general analysis assumes differentiated rather than homogeneous products and distinguishes strategic substitutes as well as complements to test this forbearance hypothesis. Rather than only a partners design we also explore a random strangers design to disentangle effects of forbearance and repeated interaction. Surprisingly, conglomerate firms do not limit competition, they rather foster it. More in line with our expectations we find more cooperation in complement markets than in substitute markets and also more cooperation in a partners than in a strangers matching.

We study a situation where two players first choose a sharing rule, then invest into a joint production process, and then split joint benefits. We investigate how social preferences determine investments. In our experiment we find that even the materially disadvantaged player cares more for social welfare than about inequality. Behavioral preferences of disadvantaged players actually increase inequality. We also investigate when players give up an advantageous sharing rule. Power-sharing can be successful in the experiment, even when it is not in a selfish world.

In this paper we are studying a multiple player two-armed bandit model with two risky arms in discrete time. Players have to find the superior arm and can learn from others' history of choices and successes. In equilibrium, there is no con?ict between individual and social rationality. If agents depart from perfect rationality and use count heuristics, they can benefit from coordination (or centralization) of search activities. We test the conjecture that agents gain from coordination with a between-subject design in two treatments. In the experiments we find no gains from coordination. Instead, we find less severe deviations from the equilibrium strategy in the non-coordinated treatment

We explore experimentally how power asymmetries between partners affect relationship-specific investments. We find that on average players' investments are larger than equilibrium investments. In contrast to social dilemma experiments, in our experiment preferences for social welfare and those for equality call for different actions. Surprisingly, even disadvantaged players care more for social welfare and less for equality. As a result social welfare increases but so does inequality. We then study conditions under which power-advantaged players give up power. Power-sharing can be successful in the experiment, even when it is not in a selfish world.

In this paper we study equilibrium- and experimental bidding behaviour in first-price and second price auctions with outside options. We find that bidders do respond to outside options and to variations of common knowledge about competitors’ outside options. However, overbidding in first-price auctions is significantly higher with outside options than without. First-price auctions yield more revenue than second-price auctions. This revenue-premium is significantly higher with outside options. In second-price auctions the introduction of outside options has only a small effect.

We introduce a new method of varying risk that bidders face in first-price and second-price private value auctions. We find that decreasing bidders’ risk in first-price auction reduces the degree of overbidding relative to the risk-neutral Bayesian Nash equilibrium prediction.This finding is consistent with the risk-aversion explanation of overbidding. Furthermore, we apply the method to second-price auctions and find that bidding behavior is robust to manipulating bidders'' risk as generally expected in auction theory.

Deviations from equilibrium bids in auctions can be related to inconsistent expectations with correct best replies (see Eyster and Rabin, 2005; Crawford and Iriberri, 2007) or correct expectations but small (perhaps quantal-response) mistakes in best replies (see Goeree et al., 2002). To distinguish between these two explanations we use a novel experimental procedure and study expectations together with best replies. We extensively test the internal validity of this setup. We find that deviations from equilibrium bids do not seem to be due to wrong expectations but due to deviations from a best reply.

We present results from a series of experiments that allow us to measure overbidding and, in particular, underbidding in first-price auctions. We investigate how the amount of underbidding depends on seemingly innocent parameters of the experimental setup. To structure our data we present and test a theory of constant markdown bids. While a fraction of bidders can be well described by Bayesian Nash equilibrium bids, a larger fraction seems to either use constant markdown bids or seems to rationally optimise against a population where some members use markdown bids and some are rational.

We use experiments to compare dynamic and static wars of attrition (i.e. second-price all-pay auctions) and first-price all-pay auctions. Many other studies find overbidding in first-price all-pay auctions. We can replicate this property. In wars of attrition, however, we find systematic underbidding. We study bids and revenue in different experimental frames and matching procedures and draw a link to the literature on stepwise linear bidding functions.

In this paper we study learning and cooperation in repeated prisoners' dilemmas experiments. We compare interaction neighbourhoods of different sizes and structure and we observe choices under different information conditions and estimate parameters of a learning model. We test robustness of the eastimator. We find that imitation, albeit a driving force in many models of spatial evolution, is often a negligible factor in the experiment. As a result, we find less cooperation.

Experiments with first price auctions find typically a substantial amount of overbidding which is often related to risk aversion. If this could be generalised then first-price auctions should generate a higher revenue then second-price auctions. We show that at least part of this overbidding is an artefact of common experimental setups which prevents bidders from underbidding. We compare various setups and identify conditions for underbidding and overbidding and study the effect on revenue.

We introduce and experimentally test an auction model that allows for outside options of bidders as substitutes for the auctioned object under the private values assumption. Theoretically and in the experiments, bidders respond to their individual outside options and to variations of common knowledge about competitors outside options. Interestingly, private outside options induce concave equilibrium bidding functions with uniformly distributed valuations. The bidding data does support this property. As theoretically predicted, lower-valued outside options lead individuals to bid more aggressively in the experiments.

We run an experiment that implements a finitely repeated version of the trust game in which players can choose in each period with whom to interact. Change in trust and trustworthiness in terms of previous experience is statistically investigated where confounding factors are controlled for. Motives such as reinforcement learning, reciprocity and rationality are useful to explain findings. Overall we find a high persistence of choice and uncover more trust and trustworthiness than in the one shot experiments. Towards the end of the game the degree of trust and trustworthiness decline.

We study with the help of experiments a two-player second-price all-pay auction. Such an auction describes e.g. the situation of a country were stabilisations are achieved through tax increases that eliminate a budget deficit. If these tax increases have distributional implications then stabilisation may be delayed (Alesina and Drazen, 1991). We find (1) under-dissipation and not over-dissipation of rents which is in contrast to other all-pay auction experiments. (2) Underdissipation decreases with increasing cost of distortionary taxation and increases with bidding cost. (3) Bidding is closer to the equilibrium on the individual than on the aggregate level. (4) The speed of stabilisations is smaller than the risk neutral Bayesian equilibrium and reacts less sensitively to changes in the cost of distortionary taxation.

We generate observable expectations about fiscal variables through laboratory experiments using real world data from several European countries as stimuli. We compare a VAR model of expectations for data which is presented in a fiscal frame with one for neutrally presented data. We measure the degree to which participants use fiscal data for their forecasts. Agents' expectations are found neither to be consistent with rational nor with purely adaptive expectations but, instead, follow an augmented-adaptive scheme. Methodological implications of the present approach for experiments in macroeconomics are also discussed.

This study disentangles experimentally imitation, reinforcement, and reciprocity in repeated prisoners' dilemmas. We compare a simple situation in which players interact only with their neighbours (local interaction) with one where players interact with all members of the population (group interaction). We observe choices under different information conditions and estimate parameters of a learning model. We find that imitation, while assumed to be a driving force in many models of spatial evolution, is often a negligible factor in the experiment. Behaviour is predominantly driven by reinforcement learning.

We generate observable expectations about fiscal variables through laboratory experiments using real world data from several European countries as stimuli.We estimate an econometric model of individual expectations for fiscal policy, which nests various theories of expectations-forming and encompasses both micro- and macro- economic lines of research on fiscal policy. Agents' expectations are found neither to be consistent with rational nor with purely adaptive expectations. Expectations follow an augmented-adaptive scheme, which embodies the "spend and tax hypothesis" on the relationship between taxes and expenditure to a greater extent than in real world data. We relate this findings to current research on the effects of fiscal policy. Methodological implications of the present approach for experiments in macroeconomics are also discussed.

This study disentangles experimentally imitation, reinforcement, and reciprocity in repeated prisoners' dilemmas. We compare a simple situation in which players interact only with their neighbours (local interaction) with one where players interact with all members of the population (group interaction). We observe choices under different information conditions and estimate parameters of a learning model. We find that imitation, while assumed to be a driving force in many models of spatial evolution, is often a negligible factor in the experiment. Behaviour is predominantly driven by reinforcement learning.

We investigate and compare different approaches to derive strategies from laboratory data in prisoners' dilemmas experiments. While theory suggests more cooperation in spatial structures than in spaceless ones, we find in our experiments either the opposite or no difference. In this paper we investigate to which degree learning and reinforcement explains this dependence on structure and information. Starting from a very simple model we gradually develop a setup where players use repeated game strategies and choose among these strategies using a simple reinforcement rule. We then measure to which degree this model explain players' behaviour.

To understand the effect of fiscal policy on the private sector we have to comprehend how expectations about fiscal variables are formed. However, little is known about the way people form expectations about fiscal variables: no undercutting theory exists, not to say empirical evidence. The problem is that ``expectations are unobservable'' (Bertola and Drazen 1993). We generate observable expectations using a combination of laboratory experiments and real world data from several European countries. Based upon these observable expectations we develop a model of formation of expectations. We fit real world data and study the effect of fiscal policies. We show that we can study the relationship between private consumption and government consumption.

We investigate and compare different approaches to derive strategies from observed data in spatial and spaceless prisoners' dilemmas experiments. We start with a model where players choose a fixed action that remains constant for all repetitions of a stage game. As an extension we then allow players to choose simple repeated game strategies that, however, remain fixed over the course of the game. We then discuss a method how to identify changing repeated game strategies. This method is used to study a simple reinforcement model. We find that in a spatial structure reinforcement plays a more important role than in a spaceless structure.

We investigate and compare different approaches to derive strategies from observed data in spatial and spaceless prisoners' dilemmas experiments. We start with a model where players choose a fixed action that remains constant for all repetitions of a stage game. As an extension we then allow players to choose simple repeated game strategies that, however, remain fixed over the course of the game. We then discuss a method how to identify changing repeated game strategies. This method is used to study a simple reinforcement model. We find that in a spatial structure reinforcement plays a more important role than in a spaceless structure.

We experimentally study the effects of locality in a repeated prisoners' dilemma. We compare players located on a circle with those in neighbourless groups. On circles players interact only with neighbours, but not with distant players. In groups all players interact with each other in the same way. We observe less cooperation on circles than in (neighbourless) groups. This is explained through different learning behaviour and a different way to use cooperation as a signal.

We study experiments in an auction setting with interdependent valuation. Groups of three players receive private signals and then bid for a single, indivisible item. Valuations for the item differ within groups and depend asymmetrically on a bidder's own and other bidders' signals. Theoretically, the English auction yields efficient allocations, while other standard auction formats fail to do so. Consistent with equilibrium predictions, we find that an English auction yields significantly more efficiency than a second price sealed bid auction. We also study the seller's expected revenue and the bidders expected payoff, and find that the experimental results are close to the theoretical predictions.

If migrants return to their origin countries, two questions arise which are of immediate economic interest for both immigration and emigration country: What determines their optimal migration duration, and what are the activities migrants choose after a return. Little research has been devoted to these two issues. This paper utilises a unique survey data set which records activities of returned migrants. We first illustrate the activities of immigrants after returning. We show that more than half of the returning migrants are economically active after return, and most of them engage in entrepreneurial activities. We then develop a model, where migrants decide simultaneously about the optimal migration duration, and their after-return activities. Guided by this model, we specify and estimate an empirical model, where the after-return activity, and the optimal migration duration are simultaneously chosen.

This study investigates experimentally the effects of locality on learning and strategic behaviour in a repeated prisoners’ dilemma. We compare a simple situation that models space (players interact only with their neighbours) with one that does not (players interact with all members of the population). Within this context different information conditions are studied. We find that imitation, while assumed to be a driving force in many models of spatial evolution, is a negligible factor in the experiment. Behaviour is driven by reinforcement learning and, under specific conditions, strategic behaviour.

In this document we sketch some techniques that are useful when developing www moderated experiments. We describe a modular and, hence, very flexible approach. On all levels (clients, server, programming language, database, etc.) we compare alternatives that can be integrated. We argue that a very wide range of experiments can be covered. There is no restriction to particular hardware, clients, operating systems, programming languages, etc. Laboratories from all over the world can be linked together with the help of the internet.

We study a model of local evolution. Agents are located on a network and interact strategically with their neighbours. Strategies are chosen with the help of learning rules that are based on the success of strategies observed in the neighbourhood. The standard literature on local evolution assumes learning rules to be exogenous and fixed. In this paper we consider a specific evolutionary dynamics that determines learning rules endogenously. We find with the help of simulations that in the long run learning rules behave rather deterministically but are asymmetric in the sense that while learning they put more weight on the learning players' experience than on the observed players' one. Nevertheless stage game behaviour under these learning rules is similar to behaviour with symmetric learning rules.

We study experiments of an overlapping generations model where inflation is determined by the monetary policy and by the amount of average saving within each period. We use a new experimental setup that allows us to observe more details of the process of expectation forming and separate this process from the actual saving process. In contrast to experimental findings by Lim, Prescott, Sunder; Marimon, Spear, Sunder; and Marimon, Sunder we find that (1) agents do not form first-order adaptive expectations; (2) subjects ‘over-save’ for precautionary reasons; as a result (3) the so-called Friedman conjecture holds, i.e. monetary policies which are equivalent in static equilibrium exhibit different levels and different volatility of inflation in the experiment. This may generate important policy trade-offs between monetary regimes. We discuss our findings and relate them to current research on adaptive learning and the role it may have in ranking alternative monetary policies.

Für Kooperation via direkte Reziprozität müssen Menschen das Handeln ihrer Partner nachverfolgen, um Ausbeutung zu verhindern. Störungen (d.h. Gedächtnis- oder Wahrnehmungs-/Entscheidungsfehler) beeinträchtigen aber das Nachverfolgen. In meiner Dissertation erkunde ich, ob Strategien zur Modellierung menschlichen Verhaltens kognitiv umsetzbar sind und wie sie mit Störungen umgehen und erforsche umsetzbare störrobuste Alternativen. Tit-For-Tat, berühmtestes Beispiel traditioneller 1-Schritt-Gedächtnis-Strategien, ist nicht störrobust, weil kleinste Störungen seinen Erfolg verringern. Da Störungen alltäglich sind, ist Tit-For-Tat kein ideales Modell menschlichen Verhaltens. Kapitel 1 zeigte hohe Gedächtnisfehlerraten, wenn Versuchspersonen die letzte Handlung ihrer Partner wiedergeben sollten (1-Schritt-Gedächtnis). In einer evolutionären Simulation ging Kooperation bei diesen Raten unter. Sich an die letzte Handlung zu erinnern, ist weder störrobust noch umsetzbar. In Kapitel 2 untersuchte ich, ob Menschen die kognitiv eher umsetzbare Strategie benutzen, ihre Partner in Typen einzuteilen und Kooperierer/Betrüger zu unterscheiden. Verglichen damit, sich die letzte Handlung jedes Partners zu merken, sollte das die Gedächtnisanforderungen senken. Die Ergebnisse deuten an, daß Menschen Partnertypen unterscheiden und ihre Strategie an die Typenverteilung in der Umgebung anpassen. Kapitel 3 erforschte Strategien, die den Vorgang der Einteilung von Partnertypen modellieren, indem sie einen Eindruck bilden. In einer Simulation waren eindrucksbasierte Strategien beim Aufrechterhalten von Kooperation störrobuster als traditionelle Strategien. Auch sagten eindrucksbasierte Strategien Versuchspersonenverhalten besser vorher als traditionelle Strategien. Gewinner von Simulation und Kreuzvalidierung waren jedoch partnerunabhängige Strategien. Menschen scheinen kognitiv noch einfacher umsetzbare störrobuste Strategien zu benutzen. For cooperation to evolve via direct reciprocity, individuals must track their partners’ behaviour to avoid exploitation. Noise (i.e., memory errors or perception/decision errors) compromises tracking, however. In my thesis, I investigate whether strategies proposed to model human behaviour are cognitively feasible and how they cope with noise, and explore feasible noise-robust alternatives. Tit-For-Tat, the most prominent example of 1-step memory strategies, is not robust to noise, because even little noise decreases its success. Since noise is quite common in everyday life, Tit-For-Tat is not an ideal candidate to model human behaviour. Chapter 1 showed that participants, when asked to remember their partners’ previous behaviour (1-step memory), had high memory error rates. In an evolutionary simulation, these rates let cooperation vanish. Remembering a partners’ previous behaviour is neither noise-robust nor cognitively feasible. In Chapter 2, I investigated whether people use the cognitively more feasible strategy of categorizing partners into types, distinguishing cooperators and cheaters. Compared to remembering each partners’ previous behaviour, this would reduce memory effort. The results indicate that people differentiate partner types and adjust their strategy to the proportion of types in their environment. Chapter 3 explored strategies that model the process of categorizing partners into types by building an impression. In a simulation, impression-based strategies were more robust to noise in maintaining cooperation than 1-step memory strategies. A cross-validation of strategies on data from Chapter 2 confirmed that impression-based strategies better predict participants’ behaviour than 1-step memory strategies. The winner of the simulation and the cross-validation were non-contingent strategies, though, indicating that people use cognitively even simpler noise-robust strategies.