ABSTRACT. This paper aims at offering a contribution to the understanding of the macroeconomic and distributional implications of raising interest rates after a prolonged period at near the ZLB. Our work tries to assess the interactions between monetary policy, inequality, and financial fragility, in a financialized economic system. Financialization is here portrayed as the presence in the economy of a complex financial system, in which, due to securitization, money creation has the twofold purpose of financing consumption and investment decisions on the one hand, while creating the input for the production of financial assets on the other hand. This dual nature of money creation goes hand in hand with inequality, as the indebtedness of the low-income households nourish the rent seeking activity of the rich. In order to investigate this complex relationship, we propose a hybrid Agent-Based Stock-Flow-Consistent Agent-Based (AB-SFC) model in which all sectors (non-financial firms, commercial banks, special purpose vehicles, investment funds, beyond the government and the central bank) are aggregated except the household sector, which is populated by heterogeneous agents. We build upon our previous work by refining the real side of the model and introducing a central bank steering the interest rate. The model portrays complex interactions between the sectors and among agents. Loans are sold by banks, transformed into financial assets by SPV and then, together with deposit and public bonds, enter the portfolio of financial institutions, which finance their activity issuing shares bought by richer households. Moreover, the originate-to-distribute model of lending, typical of the securitizing system, may lead banks to loosen their credit standard and diminish the rationing of credit to households and firms. The impact of the implementation of monetary policies can therefore be analysed in light of this multifaceted interrelation between behaviours and balance sheets. In particular, we simulate an increase of the policy rate motivated by the central bank’s willingness to reduce inflation. Preliminary results suggest that while the central bank appears able to meet its goal of inflation reduction - thank to higher unemployment hence lower wages - a financial dynamics unfolds. The non-financial firms make higher gross profits due to declining wages, though a larger part goes to interest payment. Therefore, the banking sector’s profit increases and more dividends are distributed to shareholders. Also considering the higher remunerations of fixed-income financial products, the central bank intervention produces a rise in financial incomes that, being more unevenly distributed, generates more income inequality. A necessary condition is that the government provides an increasing flow of transfers to unemployed people, thus partly absorbing the recessionary tendency triggered by the policy rate hike through the accumulation of public debt. Moreover, the economy becomes more financially unstable as signalled by the increasing number of households unable to pay the interest on accumulated debt. All in all, the policy rate hike results in a more unequal and unstable economic system which requires more public debt to counteract the recessionary tendency.

ABSTRACT. Input-output tables published by national statistical offices rely on data collected through consumer and business surveys, which is costly, often highly aggregated at the regional and sectoral level, and comes only unfrequently and with a considerable publication delay. The increased availability of large-scale real-time and granular economic data bears the potential to transform and/or supplement current approaches to data collection, but many biases and caveats that need to be understood before these data sets can be fully leveraged. In this talk, we present the results of an initial assessment of using large-scale payments data for the construction of National Accounts in the UK.

Recently, the UK Office for National Statistics (ONS) secured access to financial transactions data from the BACS payment system infrastructure, which is used by almost every business in the UK. We start with a prototype anonymised and aggregate dataset comprising >500m Business-to-Business payments worth >£7.5tn and covering 118k unique accounts corresponding to a considerable share of UK businesses in 2015-2022. The data comprises monthly transactions volumes and values between 85 SIC 2-digit industries at the regional level (12 NUTS-1 regions plus 2 islands). While fairly coarse, our data is available in (almost) real-time. Transactions are assigned to individual region-industry pairs, providing a much better time and cross-sectional resolution than existing Input-Output tables, which are mostly released at the national level with a time lag of 4 years.

We assess the feasibility of developing statistics consistent with national accounting standards but based on population-scale naturally occurring data in real time. In the presentation, we will show the results of a systematic comparison of payments-based IO tables with the official survey-based National Accounts, and we will provide roadmap to tackle key challenges when mapping granular account-level transaction data to the industry and macroeconomic level. We show exploratory applications to nowcasting and discuss a research agenda for the use of this dataset in downstream case studies, such as the impact of Brexit and Covid-19.

ABSTRACT. We study how income inequality affects monetary policy through the inequality-household debt channel. We design a minimal macro Agent-Based model that replicates several stylized facts, including two novel ones: falling aggregate saving rate and decreasing bankruptcies during the household's debt boom phase. When inequality meets financial liberalization, a leaning against-the-wind strategy can preserve financial stability at the cost of high unemployment, whereas an accommodative strategy, i.e. lowering the policy rate, can dampen the fall of aggregate demand at the cost of larger leverage. We conclude that inequality may constrain the central bank, even when it is not explicitly targeted.

ABSTRACT. Both macroeconomic and financial stability are linked to each other and depend on several specific factors. Workers, firms, and institutions are involved in systems where their mutual interaction leads personal behaviours to become the basic ingredient to characterize a complex aggregate dynamics. The main motivation of this paper is to analyse determinants of instability and differences in macroeconomic performance of countries.

In particular, the main focus is on the cultural attributes of populations for they influence the macroeconomic dynamics, through different attitudes in participating to the labour market and in implementing the technological advancement. Thus, possible sources of differences in macroeconomic performances of different countries may emerge, as consequences of the role played by cultural roots and customs of populations. Traditions, shared codes, beliefs, ethnicity, knowledge and the awareness level of the social ideal of 'value' influence the remuneration of labour and the social consideration of normal effort.

Culture can be considered as the predominant collective orientation of individual minds (Hofstede, 2001, Schwartz, 2017), and describes people’s attitudes, inclinations, predispositions, preferences, judgments, expectations, ways of thinking, and consequently, behaviours. Proximity will be considered as an expression of similarity among people belonging to different countries. The idea of cultural proximity can be traced in studies regarding languages (Melitz, 2008; Egger and Lassmann, 2012; Melitz and Toubal, 2014; Egger and Touba 2016; Carrère and Masood, 2018); trust (Ahern et al. 2015, Maystre et al. 2014); tastes (Felbermayr and Toubal 2010) and financial preferences (Hellmanzik and Schmitz 2017).

The empirical evidence supporting the view that culture relevantly impacts on economic outcomes is very wide (Guiso et al., 2006, 2008), and a number of contributions exits on many aspects of economic interaction: trade flows (Anderson and van Wincoop, 2003; Guiso, Sapienza, Zingales 2009; Carrère and Masood, 2018); banking activity (De Haas and van Lelyveld, 2014; Claessens and van Horen, 2015, Fisman et al., 2017; Jin et al., 2019); financial decisions (Zheng et al. 2012; Aggarwal and Godell, 2014; Boubakri and Saffar, 2016; Dowling et al. 2019); investors' reactiveness and trading behavior and proneness to follow herding (Beracha et al., 2014; Chang and Lin, 2015; Lee, Pantzalis, et. al., 2019; Tan et al. 2019); tax evasion (Bame-Aldred et al., 2013). More broadly, the concept of social capital (Putnam 1993), rooted in the results of historical experiences that each population has experienced, determines the cultural profile of each country and may have a role in explaining the attitude applied to labour market negotiations between workers and firms, unions and the productive system.

Tabellini (2007) presents a model of cultural transmission of values, by extending the previous work of Bisin and Verdier (2000, 2001) and Bisin, Topa, and Verdier (2004), in which parents select values to be tranferred to their children and, in so doing, influence their children’s perspective welfare. This mechanism explains, at a greater extent, that the operative attitude in working activities may be perceived as a sort of \textit{imprinting}, received from the societal environment in which life is conducted. Guiso, Sapienza and Zingales (2015) show that within firms, when employees observe ethical and trustworthy behaviours in managers' conducts, the performance of the firms is increased, thus giving support to the idea that societal engagement in common constructive values may reveal determinant, also in productive environments.

Thus, populations with similar traditions in culture and history could be expected to exhibit similar labour market characterizations and, in turn, similar attitude in generating value, once considered possible technological gaps. Provided that such a similarity is detected, the assessment of policy measures has to observe difference in values to weight macroeconomic performance and establish priorities of actions devoted to foster the human capital development, by exploiting optimally the role of cultural identity of populations. This configuration can be helpful in comparing macroeconomic dynamics of countries, while describing the determinants of growth and makes it possible to obtain intuitions on possible future scenarios or predictions, with the final aim of tailoring specific social policies fostering economic development.

The adoption of agent-based models in macroeconomic research has become increasingly popular (as shown in detailed surveys by Dawid and Delli Gatti, 2018 and in Nikiforos and Zezza, 2017) since they offer the chance to replicate true mechanisms at work in economic situations. Agent-based models also provide the possibility to study the emergence of complex properties deriving from the interaction of individual constituents at different levels of interconnected hierarchical systems (Gallegati and Richiardi, 2009).

From this perspective, the present paper discusses the impact of different cultural values in terms of GDP, unemployment, inflation and the composition of the economy (capital goods vs consumption goods) in the context of the \textsc{kcx} macro model (Biondo, 2020), thus providing theoretical background to the empirical evidence and suggesting a novel approach to policy assessment.

ABSTRACT. For centuries, national economies have been engaging in international trade and production. The resulting international supply networks not only increase wealth for countries, but also allow for economic shocks to propagate across borders. Using global, firm-level supply network data, we estimate a country's exposure to direct and indirect economic losses caused by the failure of a company in another country. We show the network of international systemic risk-flows. We find that rich countries expose poor countries stronger to systemic risk than vice-versa. The risk is highly concentrated, however higher risk levels are not compensated with a risk premium in GDP levels, nor higher GDP growth. Our findings put the often praised benefits for developing countries from globalized production in a new light, by relating them to risks involved in the production processes. Exposure risks present a new dimension of global inequality that most affects the poor in supply shock crises.

ABSTRACT. Social networks play a crucial role in diffusing information, influencing individuals’ behaviour, and affecting the stability of economic expectations. However, the effects of social networks on economic expectations remain understudied in the literature. This study examines the effects of social networks on the formation of economic expectations of bounded rational agents by incorporating a network component into a heuristic switching framework.

The findings indicate that the behaviour of a highly central agent can affect the simulation outcomes in both complex and simple network structures, albeit in varying strengths. Additionally, the complexity of network structure has a substantial impact on the speed and strength of the propagation of specific behaviours across the network, amplifying the influence of highly connected agents.

The study concludes that incorporating network effects provides valuable insights into the formation of economic expectations. The model used in this study can serve as a starting point for future research aimed at better capturing realistic expectation formation processes and their implications for economic behaviour. This could assist policymakers in creating more efficient monetary policies that account for the impact of social networks.

ABSTRACT. Phosphorus is one of the key elements in the production of fertilizers and thus the production of food. Since phosphorus is constantly removed from the soil in the process of agricultural production, its reliable availability in the form of fertilizers is essential for food security and economic development. In this paper we present a new method to trace the flows of phosphorus from the countries where it is mined to the counties where it is used in agricultural production. We achieve this by combining data on phosphorus production with data on fertilizer use and data on international trade of phosphorus-related products. We show that by making certain adjustments to value-based data on net exports we can re-construct the matrix of material phosphorus flows to a large degree, a results that is important for devising measures on sustainable development and environmental accounting, since it allows to connects research on material flows to the analysis of international trade networks and supply chains.

ABSTRACT. The European Union Emissions Trading System (EU ETS) is a crucial component of the European climate policy aiming at reducing emissions. The policy-maker sets a gradually decreasing cap on emissions and distributes emission permits to firms, which can then trade them in the market. The objective of this market mechanism is to incentivise firms to reduce their emissions efficiently and for this reason how the carbon price is formed is key. Additionally, the ETS carbon price affects not only emissions, but also economic aggregates, and inequality (Kanzig, 2021). Therefore, a comprehensive understanding of carbon price formation and the role of speculation in the ETS is essential to improve the efficient operation of the system.

Empirical evidence suggests that speculative bubbles in permit prices have occurred (Creti and Joets, 2017; Friedrich et al., 2020). This raises questions about the impact speculation has on the efficient allocation of permits across firms and over time of the ETS. We believe it is important to address such questions from a theoretical perspective, taking into account the heterogeneity of behavioural rules and motifs of the agents operating in the market with different behavioural rules. In order to address these questions, we develop a dynamic model with heterogeneous agents where the policy-maker controls the supply of permits and auction them to firms. The permit demand of firms is determined by the permit price, the cost of abatement, and their need to hedge against potential price fluctuations. Within the secondary market, futures contracts are traded by speculators and market makers, each with their own unique trading rules. Speculators participate in the market with the goal of trading futures to speculate on future price movements, while market makers, such as banks, act as counterparts for both firms and speculators. Our model considers the impact of these different agents and their various trading strategies on the efficient allocation of permits across firms and over time within the system.

Our model shows that speculative bubbles in permit prices can lead to inefficiencies in the allocation of permits, resulting in lower welfare level for all agents. We find that speculators who follow trend-following rules tend to exacerbate permit price volatility, leading to larger bubbles and bursts. In contrast, fundamentalist speculators tend to stabilise permit prices by responding to changes in market fundamentals, such as shifts in the supply and demand for permits. Furthermore, our analysis shows that the behaviour of speculators has a significant impact on the hedging demand of firms, and their overall emissions reduction strategy. We find that firms that face greater uncertainty about future permit prices tend to increase their abatement efforts, reducing their emissions in the short and long run. Overall, our results suggest that the efficient operation of the EU ETS critically depends on the behaviour of speculators in the market. Policymakers should consider the impact of speculation on permit price formation when designing the cap and trade system, in order to ensure its efficient operation and achieve the desired emissions reductions targets.

References Creti, A. and M. Joets (2017): “Multiple bubbles in the European Union Emission Trading Scheme,” Energy Policy, 107, 119–130.

Friedrich, M., E.-M. Mauer, M. Pahle, and O. Tietjen (2020): “From fundamentals to financial assets: the evolution of understanding price formation in the EU ETS,” Publisher: Kiel, Hamburg: ZBW-Leibniz Information Centre for Economics.

Kanzig, D. R. (2021): “The Unequal Economic Consequences of Carbon Pricing,” SSRN Scholarly Paper 3786030, Social Science Research Network, Rochester, NY.

ABSTRACT. What are the most effective innovation policies to engage manufacturing industries in the transition away from fossil fuels? How can these policies impact retail prices and market concentration? Are policy instruments complements or substitutes?

As climate change continues to unfold, finding tools to answer these questions becomes increasingly necessary. The sustainability of the economy needs a fast and bold greening process in manufacturing industries for which innovation is required, including both the invention and diffusion of low-carbon technologies. Innovation policies are gaining momentum in the literature and among policymakers as key stimulating factors to change the direction of technological development towards environmentally friendly techniques.

However, potential resulting increases in retail prices or in market concentration could generate political resistance to the adoption of such policies. Therefore, policymakers should be able to design policy mixes that mitigate such effects.

The purpose of the present study is to provide a generalized agent-based tool to explore policy scenarios for green innovation in industrial manufacturing oligopolies. We test five policy instruments that influence the green transition through different channels. Namely, we simulate R&D subsidies, green product subsidies, carbon taxes, command and control, and research parks. Our policy menu is broad and combines rich contributions from market and system oriented approaches to innovation policy.

The results of our simulations identify specific contributions of each policy instrument to the overall outcome in productivity, market concentration, retail prices, and greenhouse emissions. Firstly, with a mild policy cost, R&D subsidies are particularly helpful in reducing market concentration by speeding the diffusion of successful technologies. Secondly, despite being costly, green product subsidies are a powerful tool to incentivize firms to adopt low-carbon technologies while keeping retail prices close to pre-transition levels. Thirdly, carbon taxes are less powerful than green subsidies in motivating the use of green machines, but they have a strong impact on reducing total carbon emissions due to its price effect while reducing the policy cost. Command and control is the most effective intervention to reduce carbon intensity and total carbon emissions with a negligible cost. Finally, research parks strongly reduce market concentration and speed up the transition.

Our results also suggest that policy combinations can be complementary or undermining depending on the target variable of interest. For example, green subsidies usually reduce the extent of carbon emission reduction as the decrease in retail prices generates rebound effects. On the other hand, R&D subsidies and research parks are reinforcing complements for all target variables. Furthermore, the marginal contribution to key outcomes from adding an extra policy instrument compared to the contribution to the policy cost is decreasing. This result suggests that policymakers should be warned to find effective complements in their policy-mix when facing financial constraints.

Methodologically, this study differs from other agent-based models that dealt with innovation policies for green transition in three aspects. Firstly, our model is partial and supply-sided and thus abstracts from macroeconomic interaction with endogenous demand or credit cycles. Whilst these channels are by all means relevant, we believe our choice limits the noise in the interpretation of the results. Notwithstanding, our model provides a richer view on the innovation process by "opening the black-box" and making the search process an explicit combinatory problem of several technological components. We use the well-established NK models developed in the field of evolutionary biology and likewise established in management. This added feature highlights the relevance of guided search in the context of technological complexity. Finally, the undiscovered technologies of our model are limited and attached to predefined features, allowing the possibility of not having "win-win" potential technologies. As a result, innovation policies could be forced to impose a trade-off on the economy.

For the rest, our model follows the canonical features of evolutionary economics such as bounded rationality, uncertainty, and creative destruction. Likewise, the validation of our model is in line with the canonical replication of several mesoeconomic stylized facts.

Overall, this research provides a simulation tool for innovation policies in the context of the green transition of manufacturing industries. We combine canonical features from evolutionary economics with NK landscapes to represent a bounded and complex process of innovation search. After validating the model against an ensemble of stylized facts, we test an exhaustive combination of policy experiments and evaluate their impact on carbon efficiency, market concentration, retail prices and total carbon emissions. Individual and complementary policy outcomes are identified, emphasising the decreasing relation of "value for money" against additional policy instruments. However, succesful welfare enhancing transitions are possible when incentives, research and regulation complement one another.

ABSTRACT. The development, employment and growth of economies is crucially determined by firm-level production networks (FPNs), as is their resilience to disturbances propagating along corporate supply chains. The networks’ resilience fundamentally depends on the details of firms' input-output relations. However, widely used input-output models (IO), are almost exclusively calibrated with highly aggregated industry-level production networks (IPN), in the form of input-output tables. This raises the question of what the limits of predicting economic outcomes with industry-level models are. Here we leverage a nearly complete nationwide FPN containing 243,399 Hungarian firms with 1,104,141 supplier-buyer-relations, and self-consistently compare the production losses from COVID-19 related shocks propagating on the aggregated IPN and the granular FPN. Industry- and firm-level shocks are of the same size, but the latter affect firms within industries differently and realistically represent the initial effects of the COVID-19 pandemic. The shock size is inferred from Hungarian firms' actual employment reductions in the course of the early phase of the pandemic. To arrive at the shocked production levels at the industry-level, for every NACE2 industry, k, we aggregate the shock, ψ, and obtain the remaining industry-level shock, ϕ_k. To quantify the size of mis-estimations if the initial shocks on the firm-level were slightly different, we sample 1,000 different, synthetic realizations of the COVID-19 shock, Ψ, that are of the same size when aggregated to the industry-level, but affect different firms within industries Following the initial COVID-19 shock, we simulate how the adaptation of firms' supply and demand propagate downstream and upstream along the production network, once on the firm-level and once on the industry-level. We employ the simulation model of Diem et. al 2022 [1], where each firm (industry) is equipped with a generalized Leontief production function. The simulation continues up to time, T, when the production levels of firms have reached a new stable state. The final production level represents the fraction of the original production a firm (sector) maintains after the shock has propagated. We define the FPN-based economy-wide production-loss, L_firm(ψ) as the fraction of the overall revenue in the network (measured in out-strength, s_i^out) that is lost due to the initial shock and the consequences of its propagation. The IPN-based economy-wide production-loss, L_ind(ϕ), is defined accordingly Our findings reveal that using aggregated IPNs leads to large estimation errors of economy-wide production losses of up to 37%. While the industry model yields a 9.6% loss, the FPN-based losses range from 10.5% to 15.3% (see Fig. 1), suggesting that sector-level IO-models have a natural limitation in forecasting economic outcomes of firm-level production networks. We ascribe the discrepancy to the large heterogeneity of firms within industries, as firms within the same sector on average only sell 23.5% to and buy 19.3% from the same industries (see Fig. 2). This underlines the inadequacy of industries for representing the firms they include. Similar errors are likely when estimating economic growth, CO2 emissions, and policy interventions with industry-level IO-like models. Our study emphasizes that using granular data is key for understanding and predicting the behavior of economic systems.

ABSTRACT. I consider heterogeneous agents who seek to uncover predictable patterns in their environment while not overfitting the data available to them. This data is itself driven collectively by their predictions. The traders are modeled as chartist who make their forecasts of future returns strictly based on patterns in past returns. These agents are uncertain about the environment as well as about the beliefs of other agents and so are willing to entertain a wide range of patterns in the data; i.e., they are willing to entertain a wide range of autoregressive predictors and prediction architectures. They employ machine learning methods to navigate the bias-variance tradeoff that they face in making such predictions from limited historical data.

I structure the environment so that the fundamental rational expectations equilibrium corresponds to a strong simplicity heuristic for these adaptive boundedly rational traders. Previous work (Georges and Pereira, 2021) has shown in a similar context that agents who use Lasso to estimate a polynomial autoregressive forecasting rule will often select a simple heuristic such as this. However, some agents will occasionally identify an apparent pocket of predictability, deviating from that heuristic and setting off a burst of volatility, with other agents jumping on the bandwagon. There we concluded that even a high degree of attention to overfitting on the part of traders who are engaged in data mining is unlikely to entirely eliminate destabilizing speculation and that their efforts to navigate their overfitting dilemma may help us explain empirical regularities in financial market data.

Overfitting is particularly relevant in financial markets today. Financial market participants increasingly find themselves in data rich environments and have access to increasingly sophisticated prediction methods and computational resources, and the problem of overfitting is widely recognized among both data scientists and market practitioners. Both also are concerned about the opacity and lack of interpretability of complex machine learning methods. Hansen (2020) found in his interviews that quants routinely employ a simplicity heuristic, preferring simpler ML models over more complex ones, both to avoid overfitting and to increase control and interpretability. Such observations led us in the 2021 paper to favor Lasso as a good way of modeling the tradeoff facing our agents. Lasso mitigates overfitting with a penalty that can be tuned dynamically via cross validation. This level of attention to overfitting on the part of the agents reduced, but did not eliminate, destabilizing speculation in our model and indeed proved to be a powerful driver of fat tailed returns and volatility clustering.

Here I consider yet more flexible prediction methods, in a similar context, focussing on neural networks and gradient boosting machines. I investigate the scope for these more flexible technologies to yield greater prediction accuracy and/or reduce excess volatility in the model simulations. I also investigate the nature of the deviations from the strong simplicity heuristic that are required to match empirical regularities in a calibrated version of the model, by tracking the population of forecasts and fitted forecast models. Under each of the three ML implementations, the tension between the agents' goals of identifying and exploiting pockets of predictability and their interest in avoiding overfitting so as not to chase spurious signals in the data is modeled by their dynamic adaptation of regularization methods via cross validation. I continue to find, in this context, that the agents' inability to eliminate overfitting generates substantial excess volatility. Further, under each of the three implementations, this dynamic adaptation of regularization methods by the individual agents acts as an important generator of empirical regularities.

ABSTRACT. Game theory is a natural way to analyze the dynamics of coalitions between fiscal and monetary policy-makers, as it allows to consider the strategic interactions of heterogeneous players. In this paper we analyze dynamic interactions in a monetary union with three fiscal players (governments of countries) and a common central bank in presence of exogenous shocks (Blueschke et al. (2023)). As we consider an extended version of the model in Blueschke et al. (2023) consisting of several additional restrictions, it is not possible to obtain equilibrium solution of the game using classical solution techniques of linear-quadratic games. For this reason, we employ a meta-heuristic solution technique, based on Differential Evolution method. This idea was originally proposed by Blueschke et al. (2013) and later adopted to dynamic games in Blueschke et al. (2020). The main advantage of heuristic methods is their flexibility that allows to incorporate various constraints on the objective functions of the players. Proposed by Storn and Price (1997), Differential Evolution (DE) is an optimization technique designed to provide ways of tackling complex optimization problems and detecting global optima for various objective functions (eligible for any constraints and particularly suited to continuous search spaces). Like many other heuristics, DE is a population-based method that relies on recombination and mutation to produce new solutions in its population. In contrast to other heuristics (such as the Genetic Algorithm or Simulated Annealing), DE applies mutation to its candidate solutions based on the differences between individuals selected from the population. Hence, the direction and magnitude of the search is tuned by the distribution of solutions and not by a pre-specified probability. Thanks to its high efficiency and only a low number of tuning parameters, DE is attracting increasing attention in evolutionary computation. The proposed meta-heuristic approach for obtaining Nash equilibrium of a nonlinear dynamic game is designed with two linked procedures: a heuristic approach (the Differential Evolution algorithm) for finding individual optimal strategies and an approach for finding an approximate game equilibrium. The main advantage of using a heuristic (evolutionary-based) framework consists in its flexibility. Using the new methodology we are able to obtain Nash equilibrium for a dynamic coalition game with various restrictions like nonsymmetric objective function and endogenous participation of players in different coalitions.

ABSTRACT. The maxim first stated by roman philosopher Cicero in his work "De Legibus" (106 BC), "Noxiae poena par esto", aka, "Let the punishment fit the crime" has become one of the fundamental principles of criminal law that draws attention from various fields including law and economics, especially after the seminal work of Becker (1968). While it is not always feasible to adhere to this principle in all circumstances due to human error, other mechanisms in criminal law aim to approximate this standard, such as pardons, amnesties, and paroles. However, the question of how early-release policies affect the general population remains a disputed topic. Does it lead to a change in criminal behavior, and if so, in what direction? Does it also influence non-criminal individuals, or does it have an impact on recidivism rates?

This research presents an analytical model that uses evolutionary game theory to examine how pre-release policies like pardons and amnesties impact crime rates and cooperative behavior in a population. Following Colman and Wilson (1997), the model defines a crime as violating social norms, which is punishable when detected. Cooperative behavior is considered the societal norm, and acting selfishly is considered a crime that incurs sanctions upon detection. The model employs the hawk-dove game to illustrate the choice between prosocial and antisocial behavior, where acting as a Hawk is considered a crime, and acting as a Dove is a norm. If a criminal is detected and convicted, they are prevented from participating in further games until their release, which may occur if there is amnesty. During imprisonment, an agent receives a zero payoff and must wait for a pardon or amnesty to re-enter the population. After each round, non-convicted and newly-released agents may change their strategy by imitating others based on their latest payoff compared to a randomly selected agent's payoff. If the agent's latest payoff is higher than or equal to the randomly selected agent's payoff, they keep their strategy; otherwise, they adopt the randomly selected agent's strategy. Additionally, agents following societal norms must pay a fee from their payoffs, similar to an income tax, which helps to fund the legal and incarceration systems and directly affects detection rates. We explore the impact of amnesties on criminal activity in three sections. The first section presents a hawk and dove model that iterates without amnesty but with a conviction mechanism. The second section introduces amnesty to the model with an exogenous probability. Lastly, an agent-based model is used to study the effects of pardon, as it necessitates information on agents' crime history.

This paper is the first to study the effects of sanctions in a game theoretical context. The primary outcomes of our study are twofold. Firstly, implementing incarceration policies acts as a deterrent and provides additional protection to those who comply with social norms. Additionally, efficient detection and conviction measures incentivize individuals to follow societal norms instead of engaging in criminal activities. Secondly, introducing even a minor amnesty measure can undermine the deterrence effect of incarceration policies. Amnesty reduces the effectiveness of sanctions and encourages more individuals to engage in criminal activities, as the cost of committing a crime becomes lower, even if the reward is not necessarily higher than before.

ABSTRACT. Interpersonal conflicts are common in organizations due to limited resources such as personnel, funding, information, or prestige. The negative impacts of these struggles can be seen in team performance, organizational culture, and structure (Kang, 2022). Understanding the dynamics of power struggles can provide valuable insights into how conflicts arise and how they can be resolved. There is often a difference in power between conflicting parties, either formal or informal. In such cases, agents might believe that a group of people are weaker than one another due to initial biases, which eventually lead to stereotypes and discrimination (Taylor et al., 1978).

Following Amadae and Watts (2022), we use agent-based models to study culture as an evolutionary game. In randomly matched asymmetric hawk and dove games, the stronger side wins a hawk-hawk escalation with a probability based on the power difference between players. Agents may categorize their opponents based on their observable traits to predict their opponents' possible struggling powers. Each agent has two categorizable traits: prestigious education and sex. The prior gives agents a higher struggling power, while the latter is void of effect. An agent can be: (1) a fine-categorizer, categorizing her opponents based on both their gender and education; (2) a regular-categorizer, only categorizing her opponents based on their educations; and (3) a coarse-categorizer, not caring about either education or sex.

Our results indicate that fine-categorizers gain an advantage when the cost of fighting is low. In contrast, coarse-categorizers enjoy an advantage when the cost of fighting is high. In both cases, miscategorization proves helpful compared to correct categorization. Moreover, we find that gender discrimination emerges in the prior. The significance of this result comes from the fact that we do not implement any differences between sexes, compared to previous studies where an initial bias is often implemented (Martell et al., 2012). Indeed, an agent discriminates and traits a group differently because he uses cognitive categorization. Moreover, we observe that the discriminated gender becomes less self-confident in their conflicts, playing more doves compared to the other sex. This result can explain why virtually created categories change culture, as observed by Brooks et al. (2018).

ABSTRACT. In this paper, we describe an agent-based model of (myopic) investors who aim at keeping a given share of equities in their allocation. Traders are willing to revise this share and they are going to do so mainly imitating more successful traders and through random mutation.

The unitary price of the bond is perfectly inelastic whereas the stock is available in finite quantity and its price is determined by matching supply and demand at each time.

We obtain two main results from the simulations of the NetLogo model: first, large swaths of the population invest a limited amount in equities, curbing the risk of "financial extinction" of their strategy and often holding shares in in the range [40%,60%] of the wealth, an amount that is frequently observed in practice.

Second, we obtain significant average equity premiums (of about 1% to 2%) for some realistic values of the parameters. While this result stresses that common allocation rules may have a role in explaining the equity premium puzzle, our results fall short of reaching the level that is historically estimated (around 4%) and, therefore, cannot fully resolve the puzzle.

ABSTRACT. In this paper we present the results of exploratory experiments and computational analyses of trading in decentralized markets with asymmetric information. We consider three trading networks, namely the ring, the small-world, and the Erd\"os-R\'enyi random network, which allow us to introduce heterogeneity in nodes degree, centrality and clustering, while keeping the number of possible trading relationships fixed. By analyzing how trading activity, market efficiency and the distribution of traders' profits are affected by the structure of the underlying trading network, we are able to infer how traders learn and spread information about the value of a risky asset through the market. Experimental results point to a learning dynamic driven by interdependent, successive trading events, enhanced locally by clustering rather than degree. By calibrating a behavioural agent-based model to the experimental data we confirm, via a series of simulations, that learning is consistent with a complex contagion process, enabled by synergistic interactions, rather than with a simple contagion process, driven by independent exposures to informed traders.

ABSTRACT. In recent years, many researches have focused on computer simulations of group economic dynamics, but the results of the simulations depend on assumptions about the behavioral patterns of individuals. Usually these are taken from theoretical studies, but as part of this project we identify actual behavioral strategies in one of the economic games and share the results with scientific community, which could improve the external validity of theoretical analyzes and simulations in the future. We propose that one way to make the outcomes of agent-based models more realistic is to base them on empirically detected behaviors, such as behavioral strategies used by human subjects in behavioral experiments.

In this project, we investigate reciprocal helping behavior in a laboratory experiment with subjects playing a long-term economic game of indirect reciprocity, called helping game. Using mixture model-based classification method, we manage to classify the behavior of almost 90% of subjects. We compare this classification to the classifications of three other existing methods in the literature and demonstrate that it is the closest to a consensus measure.

We then propose that previous classifications neglect an important type of subjects, motivated by personal experience rather than reciprocity based on reputation. We indeed find evidence for this, as more than half of the subjects use one such “experiential” strategy in our less cognitively demanding environment, and some even when reputation based reciprocity is possible. Experiential behavior is not driven by an immediate reaction to the most recent experience, but by an accumulated experience over many rounds. We also show that concern for own reputation diminishes in the final rounds of the experiment, which can explain the end-game decline in helping behavior. Finally, we compare our statistical classification to classification based on subjects' post-experimental self-reports and show that these are not a very reliable source of information.

ABSTRACT. Over the last three decades, a fast-growing literature within Behavioral Economics and Finance has departed from the representative rational agent assumption and focused on the implications of allowing for heterogeneous interacting agents. These works show how the interactions between heterogeneous agents may explain the evolution of economic variables, such as the trends of stock prices as well as exchange rates. Agents' sources of heterogeneity are more often than not, related to the information that they have regarding key variables, usually in the long run. Fundamentalists are agents who often pay a fee to have access to information regarding the fundamental value either of the asset in which they want to invest or of some other key variable of the economic system. Specifically, in financial markets this value is assumed to be the present discounted value of future dividends.

Having information about the fundamental value of an asset is implicitly assumed to be profitable, which is why in models like the seminal model of Brock and Hommes (1998), agents pay a premium to be able to be fundamentalist. However, it is far from obvious whether having fundamentalist expectations is actually more profitable than not. Using the model of Brock and Hommes (1998), we analyse the long run profitability of fundamentalists and trend followers and find that fundamentalists do not make on average higher profits than trend followers under any assumptions regarding the model parameters. More specifically, using the key parameter values of Brock and Hommes (1998), we find that trend followers are on average more profitable than fundamentalists. This seems as a paradoxical result as the choice between a fundamentalist and a trade following strategy is made on the basis of which is more profitable.

ABSTRACT. Expectations play a crucial role in the evolution of economic systems. Economic agents’ decisions influence the realization of economic variables, like prices. The state of the economy influences agents’ expectations about the future and, therefore, their decisions. Understanding how agents form their expectations in a feedback system has become a relevant issue in economics. Due to the lack of detailed empirical information about agents’ expectations, data has been traditionally collected via surveys and, more recently, conducting laboratory experiments in which the experimenters can monitor the information available to the subjects. Learning-to-forecast experiment (LtFE) design (Marimon and Sunder, 1993; Hommes et al., 2008) has been widely used to explore individual predictions about the future evolution of a specific variable. This design has been implemented to study expectations in different contexts like financial markets (Hommes et al. [2005], Bao et al. [2017]), commodity markets (Hommes et al. [2007], Bao et al. [2013]) and macroeconomic framework (Adam [2007], Cornand and M’baye [2018], Assenza et al. [2021])

This paper explores agents’ expectation formation in a financial market framework. The primary objective of this paper is to study the effect of predicting two variables, prices and dividends, on individual expectations and price dynamics. We implement a LtFE based on a standard asset pricing model. The standard model assumes that agents have heterogeneous expectations about future prices but common and correct expectations about dividend values. Our experimental design breaks this duality in modelling agents’ expectations by combining predictions about an exogenous i.i.d. variable (dividend) with predictions about an endogenous variable (price). The secondary objective of our experiment is to study which dynamic prevails when individuals predict series of different natures, a stationarity dividend series, and are almost self-fulfilling foresight price expectations.

We implement two treatments with variations on subjects’ prediction task. In the baseline treatment, subjects only predict the price for the next period. Given the price predictions in the period the realized market price is computed. In the second treatment, subjects predict the price and the dividend for the next period. Given the price and the dividend predictions in the period the realized market price is computed.

Our preliminary results are quite surprising. In the baseline treatment, we found price dynamics are in line with the literature in LtFE, large bubbles occur in many markets. However, in the second treatment, the dynamic completely changes. Prices remain stable under heterogeneous expectations on dividends and prices. We conjecture that (i) predictions about dividend values significantly affect the price predictions and (ii) to predict dividend subjects employ an adaptive rule. This suggests that subjects’ expectations about future asset prices are influenced by the stability of the dividend series. If this is true, the stable effect of the exogenous series prevails over the instability of the endogenous series.

ABSTRACT. Market-based instruments such as a carbon tax have been shown to effectively reduce greenhouse gas emissions (e.g., Sterner 2012) and are integral in the structural transition to a climate neutral economy. While existing mitigation policies are insufficient to meet the European Union’s (EU) emission reduction targets (Climate Action Tracker 2022), ramping them up to further increase firms’ relative profitability expectations of green investments requires an appropriate institutional configuration to achieve long-term credibility (comp. Campiglio et al. 2022). Committed to respond to constituencies and corporate leaders with vested interests, politicians face difficult trade-offs (Peng et al. 2021) which may impede timely implementation of substantial policies – justified by perceived high transition risks including surging energy bills, unemployment in carbon-intensive industries and financial instability (Campiglio et al. 2022) – with the adverse side effect of exacerbated path-dependence inherent in technical change (Lamperti et al. 2020). However, integrated assessment models (IAMs) typically applied to evaluate the cost-effectiveness of different policies (Weyant 2017) do not integrate potential socio-political barriers to policy implementation, questioning their suitability to assess politically durable strategies (Beckage et al. 2020, Peng et al. 2021, Brutschin et al. 2021). Against this background, we aim to provide a comprehensive framework to address transition risks and political feasibility concerns related to the green transition by integrating a joint economic evaluation of climate impacts and mitigation policy and a socio-political feasibility assessment. In line with a consequentialist analysis of strategy towards immense risk (Stern & Stiglitz 2021), we consider different mitigation policy pathways consistent with the EU’s remaining carbon budget to improve the understanding of relative trade-offs and the role of fiscal policy in developing politically durable policy strategies. To that end, we develop an agent-based opinion dynamics model of policy support over the transition period which is based on social influence (i.e., social conformity) and multiple sources of external information (i.e., climate change evidence, business cycle and lobbying influence), acknowledging the embeddedness of individuals in natural and socio-economic systems. First, the social conformity feedback represents the influence of social networks in which individuals are embedded (e.g., at home, work, leisure), where current social norms, such as the dominant opinion within a social group, are costly to violate but can shape public opinion over the long run (Moore et al. 2022). Second, the climate change perception feedback accounts for the influence of increasingly emerging signals of climate change in individuals’ everyday experience (e.g., increased temperature and/or extreme weather events) on climate policy support via changes in risk perception (Moore et al. 2022). Third, the business cycle feedback captures individuals’ tendency to adjust their climate change concern based on short-term economic conditions such as unemployment and real income (growth), indirectly affecting policy support (Benegal 2018, Kachi et al. 2015). Fourth, the political interest feedback incorporates balancing and reinforcing processes according to which the introduction of climate policies may activate powerful incumbents in the energy sector to lobby against more stringent policies, e.g., by spreading misinformation about climate change, but may also establish powerful interests of the renewable energy sector to lobby for further change (Stokes 2020, Frumhoff et al. 2015, Moore et al. 2022). Linking the opinion dynamics model to the ‘Dystopian Schumpeter meeting Keynes’ model (Lamperti et al. 2018), an agent-based IAM, allows us to study the interrelationship between endogenous growth, business cycles and climate dynamics in the presence of multiple market failures on the one hand, and public acceptance on the other hand. Comparing the co-evolution of the climate, the economy and public support of climate policy associated with different mitigation policy pathways (e.g., carbon tax schemes and revenue recycling options), we draw conclusions about the favorable design of a carbon tax and bundling with fiscal policies to be environmentally and politically stable.

References

Beckage, B., Lacasse, K., Winter, J. M., Gross, L. J., Fefferman, N., Hoffman, F. M., ... & Kinzig, A. (2020). The Earth has humans, so why don’t our climate models?. Climatic Change, 163(1), 181-188. Benegal, S. D. (2018). The impact of unemployment and economic risk perceptions on attitudes towards anthropogenic climate change. Journal of Environmental Studies and Sciences, 8(3), 300-311. Brutschin, E., Pianta, S., Tavoni, M., Riahi, K., Bosetti, V., Marangoni, G., & Van Ruijven, B. J. (2021). A multidimensional feasibility evaluation of low-carbon scenarios. Environmental Research Letters, 16(6), 064069. Campiglio, E., & van der Ploeg, F. (2022). Macrofinancial Risks of the Transition to a Low-Carbon Economy. Review of Environmental Economics and Policy, 16(2), 173-195. Climate Action Tracker. (2022, November 5). EU Country summary. Retrieved 15 February 2023, from https://climateactiontracker.org/countries/eu/ Frumhoff, P. C., Heede, R., & Oreskes, N. (2015). The climate responsibilities of industrial carbon producers. Climatic Change, 132(2), 157-171. Kachi, A., Bernauer, T., & Gampfer, R. (2015). Climate policy in hard times: are the pessimists right?. Ecological Economics, 114, 227-241. Lamperti, F., Dosi, G., Napoletano, M., Roventini, A., & Sapio, A. (2018). Faraway, so close: coupled climate and economic dynamics in an agent-based integrated assessment model. Ecological Economics, 150, 315-339. Lamperti, F., Dosi, G., Napoletano, M., Roventini, A., & Sapio, A. (2020). Climate change and green transitions in an agent-based integrated assessment model. Technological Forecasting and Social Change, 153, 119806. Moore, F. C., Lacasse, K., Mach, K. J., Shin, Y. A., Gross, L. J., & Beckage, B. (2022). Determinants of emissions pathways in the coupled climate–social system. Nature, 603(7899), 103-111. ISO 690 Peng, W., Iyer, G., Bosetti, V., Chaturvedi, V., Edmonds, J., Fawcett, A. A., ... & Weyant, J. (2021). Climate policy models need to get real about people—here’s how. Stern, N., & Stiglitz, J. E. (2021). The social cost of carbon, risk, distribution, market failures: An alternative approach (Vol. 15). Cambridge, MA, USA: National Bureau of Economic Research. Sterner, T. (2012). Distributional effects of taxing transport fuel. Energy Policy, 41, 75-83. Stokes, L. C. (2020). Short circuiting policy: Interest groups and the battle over clean energy and climate policy in the American States. Oxford University Press, USA. Weyant, J. (2017). Some contributions of integrated assessment models of global climate change. Review of Environmental Economics and Policy.

ABSTRACT. The transition towards low-carbon technologies requires ambitious climate objectives, but these have to be credible enough for firms to incorporate them in their investment choices. However, the short-term nature of electoral cycles and the pressures of lobbying groups can incentivise policy-makers to deviate from long-term commitments, especially under perceivedly high transition risks. This creates policy uncertainty that affects firms’ beliefs and investment decisions. The interactions between short-term electoral cycles, policy-makers’ credibility, and long-term investment decisions remain largely under investigated.

Building on Campiglio et al. (2023), we develop a model where firms are subject to behavioural frictions and take investment decisions depending on their heterogeneous beliefs on the credibility of climate policy announcements. Our model features two types of policy-makers competing in elections every five years. The first type is an environment-friendly policymaker who is strongly committed to meeting climate policy targets, even if it entails high transition risks. The second type, on the other hand, is less interested in mitigating climate change and is more focused on reducing potential transition risks. We propose that the probability of the environmental policy-maker winning the election is contingent upon the salience of climate change, which is determined by the magnitude of climate damages and their interaction with economic outcomes. This election probability incorporates the well-established empirical finding that climate damages amplify the salience of climate change, while economic downturns attenuate this effect.

We employ our model to investigate the evolution of firms’ beliefs in the presence of long-term policy commitments and short-term implementation and how this affects firms’ investment decisions and the low-carbon transition. We show that our model can exhibit multiple equilibria and that the long-run behavior of the economy is shaped by the strictness of climate policy, the relative importance of climate damages against economic outcomes in the election probability, and firms’ behavioral frictions. Moreover, our findings demonstrate that the commitment level of the green party has a larger influence on the likelihood of a successful transition than the non-green party’s commitment level. Additionally, we identify a critical window of time where election outcomes play a crucial role in determining the final outcome of the transition. Overall, our paper contributes to the literature on the low-carbon transition by incorporating heterogeneous forward-looking expectations, climate policy uncertainty, and electoral cycles.

ABSTRACT. The announcement of the first tariff increases on imports from China to the USA destabilized the global economy. Two of the world’s largest economies began imposing punitive tariffs on each other in early 2018. This paper aims to identify the effects, i.e., the changes in M&A transactions, due to the trade dispute between China and the USA. This leads to the main research question: “How have the trade tensions and their punitive tariffs affected the cross-border M&A deals in China, the US, and beyond?” Since the M&A behavior can be described by the value and number of transactions accurately, a quantitative approach is used. The data used was obtained from the secondary data source “Mergermarket.” The results were conclusive. Both countries were affected by the trade dispute. The M&A behavior of China has changed permanently. Both the value and number of transactions from China into the USA have changed significantly since the first punitive tariffs were imposed. In contrast, the M&A behavior from the USA to China experienced only a short-term negative shift. The M&A behavior from the USA remained the same after the Phase One agreement as before, with an increase in certain M&A activities. Furthermore, investment behavior has changed in terms of deal size distribution. Since the trade dispute, the USA has increasingly executed “Large” deals, while China has completed significantly fewer “Large” sized deals. As a result, China is investing in smaller M&A deals, and the USA is investing more in more significant M&A deals.

ABSTRACT. To combat climate change, policy makers and scientists all over the world have been advocating for the transition from a fossil-fuel economy to one based on renewable technologies. Consumers have a key role to play in this process. Whether it be solar panels, heat pumps or electric vehicles, the consumer is the one who ultimately decides whether they will adopt a product or not. As a full energy transition is needed by 2030 to meet climate objectives, governments have been encouraging adoption of environmentally-friendly products. However, uptake of such products has been shown to be unequal and dependent on income (Carley and Konisky, 2020). Lower income groups find themselves “locked out” of the benefits they offer, including a lower carbon footprint, increased convenience and reduced costs. To make matters worse, commonly used policies to encourage adoption of environmentally-friendly products, such as feed-in tariffs and marketing campaigns, have been shown to replicate or even exacerbate existing inequalities (Stewart, 2022). To further a just energy transition called for by the energy justice movement, it is crucial to create a better understanding of how different mechanisms drive (un)equal uptake and how governments can not only boost adoption of environmentally-friendly products, but to also do so fairly without leaving behind lower income groups.

I therefore conduct a study with an agent-based model that simulates green product diffusion. It incorporates three types of elements: psychological mechanisms, product characteristics and community characteristics. Agents are heterogeneous and interact with each other. To reflect the current state of knowledge in the literature around green consumption, I introduce a sophisticated utility function that considers the effects of environmental concern, product visibility, status competition, peer effects and financial constraints, as well as their interactions. I vary the green product’s cost and visibility, as well as the income inequality and environmental concern in the community. I study the adoption ratio after 10 time steps (years) with a one- factor-at-a-time analysis and a global sensitivity analysis that estimates main and interaction effects for each model parameter. Due to the bimodal distribution of the outcome variable, I use moment-independent approaches in combination with machine learning. To differentiate between income quartiles, I also conduct a subgroup analysis and consider each income quartile separately.

The results show that inequality has the by far greatest impact on adoption rates, followed by product cost and environmental concern. This is due to the fact that inequality induces a phase transition. Phase 1 and phase 2 have distinct steady states: phase 1 at no adoption (0%), and phase 2 at full adoption (100%) after 10 years. When and if the transition happens is also partially influenced by other model parameters and, above all, the agent’s own income. During the transition there seems to be no steady state. For lower income groups, the phase transition happens at lower inequality levels and the window in which the transition phase happens is more narrow than for higher income groups. As expected, inequality itself has a mostly negative impact on lower income groups, but can benefit higher income groups as it concentrates wealth at the top. Similar to inequality, the importance of product cost increases as income decreases. It has a negative and mostly linear effect on adoption rates.

Higher income groups are much more responsive to shifts in psychological factors, such as peer pressure or a shift in environmental concern. Because higher income groups have a prolonged phase transition without fixed point, this widens the window in which they are highly susceptible to any changes in model parameters. This includes product visibility and environmental concern. Both consistently boost adoption rates, but the effect is stronger for higher income groups. For one, product visibility almost exclusively impacts adoption rates in the transition phase. As this is longer for higher income groups, they will be most affected by this. Environmental concern is even the most important factor determining the adoption rates of the top quartile, but is only on third place for all income groups combined. Similar to product visibility, its importance increases with income because financial constraints play a smaller role and the decision of whether to adopt or not mainly hinges on purchase motives. Its effect is non-linear, as it remains constant once medium environmental concern levels are surpassed. The adoption behaviour of lower income groups is governed by financial aspects, most of all inequality. Due to the sudden onset of the phase transition, triggered by low to medium inequality, and the therefore large parameter space characterised by a fixed point at 0% adoption, the degree to which other factors can affect their purchase behaviour is limited. Their financial constraints are the single most important factor keeping them from adopting environmentally-friendly products, and without addressing these, it is unlikely that any policy interventions will be able to boost their adoption rates. Product cost further compounds negative effects from inequality, and psychological factors, such as environmental concern or status competition only have a limited impact on their adoption rates.

The findings have significant implications for policies aimed at increasing uptake of environmentally-friendly products. The higher the inequality in a community, the greater the need for a tailor-made policy package that is designed to ensure lower income groups aren’t left behind. While policy interventions based on nudges or raising awareness of the environmental benefits of green products can effectively boost adoption for those who can already comfortably afford them, they will do fairly little for those who struggle to do so. Whereas focusing on psychological factors may be a cost-effective and useful way of increasing adoption among higher income groups, lower income groups will need targeted financial support in addition to that to achieve similar effects. Therefore, to design an inclusive policy, interventions that address the financial barriers of lower income groups in addition to any community-wide interventions are needed. Otherwise, it will be difficult to meaningfully boost adoption rates of lower income groups.

ABSTRACT. In this paper, we consider the interaction between a perfectly informed provider and a representative consumer who has taste uncertainty regarding a set of goods and learns through consumption. We characterize the conditions under which the consumer learns or does not learn about their own preferences or qualities of goods, and the consequent optimal strategies of a provider who wants to maximize the perceived preferences of the consumer regarding the goods. We show that the provider can succeed in manipulating the consumer’s perceived preferences by exploiting strategic bundling and initial biases. In particular, we assume that, over time, the provider offers different consumption bundles that form a dynamic network of co-occurrences among the goods. After consuming each bundle, the consumer updates their estimations of their own preferences through maximum likelihood. We demonstrate that the provider may want to hide the quality of overestimated goods through bundling, while displaying the quality of underestimated goods through single-item consumption. The manipulation is even more effective when the marginal profits for goods are heterogeneous. In such cases, the provider uses underestimated low-margin goods to boost the selling of high-margin goods. In the long run, the consumer learns, unless there is persistent multicollinearity (induced by the provider) in the network of cooccurrences or if consumption stops. In the case of multicollinearity, the consumer learns their preference for the whole bundle but not the marginal utility of each good, which allows the provider to exploit different profitability. Strategic bundling can also be used to slow down the learning process and earn higher profits by exploiting temporary positive biases.

ABSTRACT. In this paper, we develop a novel theoretical framework combining staggered update of information together with asymmetries in the agents' costs of over- and underpredicting macroeconomic variables. The combination of these two features leads an individual with asymmetric loss aversion, conscious of having outdated information, to rationally under(over)-estimate a variable in such a way to hedge against a costly over(under)-estimation. At the aggregate level, the framework implies aggregate expectations that are rationally biased and sticky, together with cross-sectional dispersion of individual expectations. The implications of the theoretical model for individual expectations are then tested using micro-level panel data from the newly introduced ECB Consumer Expectations Survey; the findings are consistent with the model predictions. Moreover, through the use of simulations we check whether selected stylized facts observed in survey data emerge from the model; we find that a model that combines sticky information and asymmetric loss aversion is sufficient to jointly explain considered stylized facts. More parsimonious models are not.

ABSTRACT. The literature on behavioral expectation formation demonstrates that learning often leads to violation of the Efficient Market Hypothesis and non-equilibrium asset prices (Brock and Hommes, 1998). These theoretical results have been corroborated by a number of experimental (Hommes, 2011) and empirical studies (Cornand and Hubert, 2020). Anufriev et al. (2019) present a model, in which agents use Genetic Algorithms to calibrate a generalized linear forecasting heuristic. The authors confirm an important insight from the original Brock-Hommes model, namely that in asset markets agents focus on trend rules in response to the positive feedback between expectations and realized prices, resulting in self-fulfilling cycles of non-fundamental optimism and pessimism.

These studies typically focus on expectation formation and thus assume exogenous market fundamentals. In particular, the non-fundamental price dynamics bear no impact on productivity of the underlying firm. An alternative approach was proposed by Dawid et al. (2019). The authors study an agent-based model with an explicit investment optimization problem of a manager, who needs to balance the trade-off between stock buyback, dividend stream and firm’s productivity.

The aim of this study is to combine these two threads of literature. We use the investment optimization framework of Dawid et al. (2019) to model behavior of a large number of initially homogenous firms. Boundedly rational traders learn how to forecast the prices of the firms’ stocks through the social learning variant of the Genetic Algorithm, independently for each firm. Depending on the version of the model, they can learn how to respond to the announced buybacks, the past observed price trend, or both of these variables. Preliminary results point to persistent heterogeneity between firms, as well as heterogenous and typically non-fundamental forecasting. Finally, we study how the monetary authority can influence the market dynamics by manipulating the risk-free interest rate.

ABSTRACT. In recent years, decentralized organizational structures have become increasingly popular, challenging traditional top-down design methods. As a result, certain design elements, such as task allocation, now emerge as bottom-up properties that cannot be fully controlled from the top. This poses the question of how to guide bottom-up task allocation towards effective organizational outcomes. This paper proposes an agent-based model that addresses this question, using an incentive mechanism to motivate agents' task allocation decisions. The mechanism offers a range of incentives that span from highly individualistic to highly altruistic. Our findings show that short-term oriented agents, when incentivized correctly, outperform their long-term oriented counterparts in terms of organizational performance. Additionally, we find that the significance of mirroring in task allocation is reduced when altruistic incentives are present within the organization.