We develop a theoretical model of network formation in the overnight interbank market, where banks manage liquidity under reserve uncertainty by strategically forming bilateral lending relationships. The model incorporates counterparty risk and the central bank’s corridor system, yielding endogenously determined equilibrium networks. A key result, relevant for systemic stability policy, is that the equilibrium network is bipartite: active banks act either as lenders or borrowers, and no strategic (interbank) intermediation arises. We also show that, via temporal aggregation of equilibrium networks, apparent intermediation and a core-periphery structure emerge. We validate these predictions using e-MID market data, showing that the model reconciles the frequency-dependent network features documented in the empirical literature for this market.

In this paper, we explore the efficiency of the Leaning Against the Wind (LAW) policy within the New Keynesian framework with heterogeneous expectations. To do this, we add a financial sector to the model, linking it with the real sector via the financial accelerator channel. We find that the range of parameters in the Taylor rule that enable the stability of the targeted equilibrium is reduced with the financial accelerator. However, expanding the Taylor rule via the LAW policy fails to counteract this effect and may even exacerbate it if the policy reacts to any mispricing. If applied conditionally on high mispricing, the LAW policy leads to co-existing stable targeted and non-targeted equilibria. Our simulations suggest that while the LAW policy can reduce the amplitude of endogenous fluctuations, it is inefficient in dealing with exogenous shocks and results in larger average deviations from the target.

The continuous double auction (CDA) is a well-studied and widely used trading institution. However, there is no universally accepted theory regarding the dynamics of price formation, especially within the first period, that has endured experimental testing. In this paper, we introduce a behavioral model called IEL-CDA, which builds upon the Individual Evolutionary Learning (IEL) model of Arifovic et al. (2022). We enhance IEL by (a) incorporating Marshallian Selection, favoring traders with a higher expected surplus in making offers, and (b) allowing a trader's hypothetical reasoning to depend on the history of transactions. Using new experimental data, we test the hypothesis that efficiencies and average prices observed in the experiments follow the same distribution as those produced by simulations with various models. The hypothesis is rejected for both the Zero Intelligent theory of Gode and Sunder (1993) and IEL; however, it is not rejected for IEL-CDA. Therefore, IEL-CDA emerges as a more accurate theory of behavior in the continuous double auction.

The Individual Evolutionary Learning (IEL) algorithm was proposed as a portable learning model for games with large strategy spaces. In principle, IEL benchmark simulations could substitute or supplement expensive experiments with human subjects. We evaluate the ability of the IEL model to replicate experimental findings observed in repeated Keynesian Beauty Contest (KBC) games, which have a large strategy space. The IEL specification with standard parameter values is able to capture major dynamical features and differences between treatments in both one-dimensional (Nagel, 1995; Duffy and Nagel, 1997) and two-dimensional (Anufriev et al., 2022b) versions of KBC games. We compare IEL with some other simple learning models and find that it performs relatively better across multiple treatments. We also use IEL to predict behavior in repeated KBC games that have not yet been conducted experimentally.

We are examining social learning in networks, where agents aim to minimize cognitive dissonance resulting from disagreement by adjusting their statements in conversations to align with those of their associates, rather than truthfully sharing their beliefs. Our analysis investigates the impact of this adjustment, known as audience tuning, on belief revision, limiting beliefs, consensus conditions, and convergence speed. Our findings demonstrate that audience tuning facilitates extensive belief propagation beyond immediate associates, resulting in faster convergence in most of the societies considered. It also leads to a redistribution of influences on long-run beliefs, favoring agents with lower dissonance sensitivity. We also show that endogenous changes in the network, driven by dissonance minimization, can impede society from reaching a consensus.

Starting from the work of Hommes et al. (2005a), we propose an alternative version of their asset pricing model with heterogeneous agents and asynchronous updating of beliefs. In particular, we assume that the predictors are selected based on their accuracy in predicting the market price of the risky asset, measured as an absolute prediction error, and not on the net profits made by fundamentalists and trend followers. From a mathematical point of view, the deterministic skeleton of the present model is a two-dimensional piecewise-smooth map. We present an analytical study of the fundamental equilibrium and the coexisting non-fundamental equilibria and propose a comparison with the results in Hommes et al. (2005a). A robustness check is also conducted by considering an accuracy measured by squared prediction errors, a fitness measure often adopted in theoretical and experimental studies because of its smoothness and for being equivalent to risk-adjusted profits. The comparisons reveal that these different fitness measures do not modify the stability of the fundamental equilibrium. However, non-fundamental equilibria, their stability and the out-of-equilibrium dynamics are affected.

We analyze trading in a modified continuous double auction market. We study how more or less information about trading in a prior round affects allocative and informational efficiency. We find that more information reduces allocative efficiency in early rounds relative to less information but that the difference disappears in later rounds. Informational efficiency is not affected by the information differences. We complement the experiment with simulations of the Individual Evolutionary Learning model which, after modifications to account for the CDA, seems to fit the data reasonably well.

We extend the beauty contest game to two dimensions: each player chooses two numbers to be as close as possible to certain target values, which are linear functions of the averages of the two number choices. One of the targets depends on the averages of both numbers, making the choices interrelated. We report on an experiment where we vary the eigenvalues of the associated two-dimensional linear system and find that subjects can learn the Pareto-optimal Nash Equilibrium of the system if both eigenvalues are stable and cannot learn it if both eigenvalues are unstable. Interestingly, subjects can also learn it if the system has the saddlepath property – with one stable and one unstable eigenvalue — but only if the one unstable eigenvalue is negative. We show theoretically that our results cannot be explained by homogeneous level-k models where all agents apply the same level k depth of reasoning to their choices, including the naïve learning model. However, our results can be explained by a mixed cognitive-levels model, including the adaptive learning model. We also run a horserace between many models used in the literature with the winner being a simple mixed model with levels 0, 1, and equilibrium reasoning.

We study the effects of the investment horizon on asset price volatility using a Learning to Forecast laboratory experiment. We find that, for short investment horizons, participants coordinate on self-fulfilling trend-extrapolating predictions. Price deviations are then reinforced and amplified, possibly leading to large bubbles and crashes in asset prices. For longer investment horizons such bubbles do not emerge and price volatility tends to be lower. This is due to the fact that, for longer horizons, there is more dispersion in participants’ forecasts, and participants extrapolate trends in past prices to a lesser extent. We also show that, independent of the investment horizon, if the initial history of asset prices is already relatively stable before participants start their prediction task, price volatility remains small, with prices close to their fundamental values for the duration of the experiment.

The current intense food production-consumption is one of the main sources of environmental pollution and contributes to anthropogenic greenhouse gas emissions. Organic farming is a potential way to reduce environmental impacts by excluding synthetic pesticides and fertilizers from the process. Despite ecological benefits, it is unlikely that conversion to organic can be financially viable for farmers, without additional support and incentives from consumers. This study models the interplay between consumer preferences and socio-environmental issues related to agriculture and food production. We operationalize the novel concept of extended agro-food supply chain and simulate adaptive behavior of farmers, food processors, retailers, and customers. Not only the operational factors (e.g., price, quantity, and lead time), but also the behavioral factors (e.g., attitude, perceived control, social norms, habits, and personal goals) of the food suppliers and consumers are considered in order to foster organic farming. We propose an integrated approach combining agent-based, discrete-event, and system dynamics modeling for a case of wine supply chain. Findings demonstrate the feasibility and superiority of the proposed model over the traditional sustainable supply chain models in incorporating the feedback between consumers and producers and analyzing management scenarios that can urge farmers to expand organic agriculture. Results further indicate that demand-side participation in transition pathways towards sustainable agriculture can become a time-consuming effort if not accompanied by the middle actors between consumers and farmers. In practice, our proposed model may serve as a decision-support tool to guide evidence-based policymaking in the food and agriculture sector.

An asset pricing model with chartists, fundamentalists and trend followers is considered. A market maker adjusts the asset price in the direction of the excess demand at the end of each trading session. An exogenously given fundamental price discriminates between a bull market and a bear market. The buying and selling orders of traders change moving from a bull market to a bear market. Their asymmetric propensity to trade leads to a discontinuity in the model, with its deterministic skeleton given by a two-dimensional piecewise linear dynamical system in discrete time. Multiple attractors, such as a stable fixed point and one or more attracting cycles or cycles and chaotic attractors, appear through border collision bifurcations. The multi-stability regions are underlined by means of two-dimensional bifurcation diagrams, where the border collision bifurcation curves are detected in analytic form at least for basic cycles with symbolic sequences LRn$${\hbox {LR}}^{n}$$ and RLn$${\hbox {RL}}^{n}$$. A statistical analysis of the simulated time series of the asset returns, generated by perturbing the deterministic dynamics with a random walk process, indicates that this is one of the simplest asset pricing models which are able to replicate stylized empirical facts, such as excess volatility, fat tails and volatility clustering.

Abstract In this paper we address the question of how individuals form expectations and invent, reinforce, and update their forecasting rules in a complex world. We do so by fitting a novel, parsimonious, and empirically validated genetic algorithm learning model with explicit heterogeneity in expectations to a set of laboratory experiments. Agents use simple linear first order price forecasting rules, adapting them to the complex evolving market environment with a Genetic Algorithm optimization procedure. The novelties are: (1) a parsimonious experimental foundation of individual forecasting behavior; (2) explanation of individual and aggregate behavior in three different experimental settings, (3) improved one- and 50-period ahead forecasting of experiments, and (4) characterization of the mean, median, and empirical distribution of forecasting heuristics. The median of the distribution of GA forecasting heuristics can be used in designing or validating simple Heuristic Switching Models.

We present a laboratory experiment designed to investigate the effect of the fee structure on mutual fund choice. We find that subjects tend to ignore periodic and small operating expenses fees and base their decisions on gross, instead of net, returns. A fee in the form of a, much larger, front-end load leads to lock-in into one of the funds. It is used by some subjects as a commitment device, but exacerbates the decision errors of other subjects. Although past returns do not convey information about future returns, return chasing helps explain subjects’ behavior.

This paper offers an overview of the literature on the economic and financial applications of theory of nonlinear dynamics, especially bifurcation theory. After a short introductory discussion of the first nonlinear dynamic models in social sciences and the economic relevance of the zoo of bifurcations and complicated dynamics that such models can generate, we present an overview of the literature on nonlinear dynamic models in the areas of underdevelopment, environmental poverty traps, the management of common goods, industrial organization and financial markets. The review of the literature is enriched by reflections and ideas for future research.

We present results from the first laboratory experiment on the seminal heuristic switching model introduced by Brock and Hommes (1997, 1998). Subjects choose between two alternatives, a sophisticated and stabilizing, but costly, heuristic, and a destabilizing, but cheap, heuristic, and are paid according to the performance of the chosen heuristic. Aggregate choices determine the evolution of a state variable and, consequently, the performance of both heuristics. Theoretically, an increase in the costs for the stabilizing heuristic generates instability and leads to endogenous fluctuations in both the state variable and the fraction of agents using that heuristic. We vary the costs of the stabilizing heuristic in the experiment and find that the predictions of the heuristic switching model are partially confirmed. For low costs the dynamics are stable. For high costs, the dynamics initially are unstable and exhibit the type of bubbles and crashes emblematic for the heuristic switching model. However, over time the pattern of bubbles and crashes disappears and the dynamics become more stable. We estimate a standard discrete choice model on aggregate choice data and observe that subjects have a tendency to become less sensitive to payoff differences when the environment is less stable, which has important implications for the application of heuristic switching models.

The paper studies an oligopoly game, where firms can choose between price-taking and price-making strategies. On a mixed market price takers are always better off than price makers, though the profits of both types decline in the number of price takers. We investigate and confront two possibilities of firms’ decisions about their types: forward-looking equilibrium reasoning and backward-looking individual learning. We find that the Cournot outcome is the only equilibrium prediction and it is learnable if firms are sufficiently sensitive to profit differences. However, with a larger number of firms, a unilateral deviation from Cournot behavior becomes profitable. Under learning this incentive creates a space for permanent oscillations over different markets with a positive but low number of price takers.

The moment of order submission plays an important role for the trading outcome in a Continuous Double Auction; submitting an offer at the beginning of the trading period may yield a lower profit, as the trade is likely to be settled at the own offered price, whereas late offers result in a lower probability of trading. This timing problem makes the order submission strategy more difficult. We extend the behavioral model of Individual Evolutionary Learning to incorporate the timing problem and study the limiting distribution of submission moments and the resulting offer function that maps submission moments to offers. We find that traders submit different offers at different submission moments the distribution of which uni-modal with a peak moving from late to early as the market size increases. This behavior exacerbates efficiency loss from learning. If traders evaluate profitability of their strategies over longer history, orders are submitted later with the same effect of market size.

We run a laboratory experiment to study how human subjects switch between several profitable alternatives, framed as mutual funds, in order to provide a microfoundation for so-called heterogeneous agent models. The participants in our experiment have to choose repeatedly between two, three or four experimental funds. The time series of fund returns are exogenously generated prior to the experiment and participants are paid for each period according to the return of the fund they choose. For most cases participants’ decisions can be successfully described by a discrete choice switching model, often applied in heterogeneous agent models, provided that a predisposition toward one of the funds is included. The estimated intensity of choice parameter of the discrete choice model depends on the structure of the fund returns. In particular, it increases with correlation between past and future returns. This suggests human subjects do not myopically chase past returns, but are more likely to do so when past returns are more predictive of future returns, a feature that is absent in the standard heterogeneous agent models.

This paper connects variance–covariance estimation methods, Gaussian graphical models, and the growing literature on economic and financial networks. We construct the network using the concept of partial correlations which captures direct linear dependence between any two entities, conditional on dependence between all other entities. We relate the centrality measures of this network to shock propagation. The methodology is applied to construct the perceived network of publicly traded Australian banks and their connections to domestic economic sectors and international markets. We find strong links between the big four Australian banks, real estate and other sectors of the economy, and determine which entities play a central role in transmitting and absorbing the shocks.