ABSTRACT. The neoclassical growth model (Ramsey [1928], Koopmans [1960], Cass [1966]) serves as the natural benchmark for evaluating alternative hypotheses on expectation formation and economic interaction. In our current setup, the household can only predict the effect of a saving rate choice on capital for a short time horizon. However, the household ignores the true capital law of motion for a longer horizon, resulting in a label of “myopic”. This myopia reflects the household’s inability to see the future despite its rational behavior and intention to optimize intertemporal utility. As a result, rational households face the challenge of predicting the consequences of their choices on future income or capital.

Among the many hypotheses on the way myopic households form expectations on the future state of the economy, it is reasonable to assume that households attempt to leverage the sole information at their disposal, namely, their short-term expectations of capital growth. For instance, households may extend their perceived short-run growth rate of capital to the long run. This assumption aligns with the primary objective of the paper, which is to explore a scenario in which households lack the ability to make reliable predictions about the future trajectory of the economy, rendering them uninformed and forcing them to behave according to intuitive or unsophisticated strategies. An alternative to the notion that households assume a constant capital growth rate is that they assume a constant level of future capital. Both assumptions are simplistic and indicative of a paucity of information.

The household will have the opportunity to verify the accuracy of its prediction and adjust it based on the latest information, such as the actual variation of the capital, at each subsequent date. In this framework, the agent is required to solve a static optimization problem at every time step, disrupting the standard intertemporal relationship between the control variable (consumption) and the state variable (capital). Because the agent assumes that its current decision will last indefinitely, there is no need to estimate the feedback from the expected interest rate (which is dependent on future capital intensity) to consumption. Therefore, the static optimization problem-solving approach in this framework disregards the standard nexus that operates in the pure neoclassical growth model, leading to a breakdown of the inter-temporal relationship between consumption and capital intensity.

Extensive research in the literature has focused on studying myopic behavior in neoclassical models, particularly in relation to the time inconsistency of the optimal plan. However, it is important to note that this type of myopia stems from changes in individual preferences over time rather than from limited capabilities in the expectation formation process.

Our work is linked to this body of literature because the sequential optimizing process indirectly creates time inconsistency due to the interaction between rule-based expectations and impatience. Biased forecasts on capital growth lead to time-varying utility plans, even if we assume homogeneity of discount factors, because they alter the perceived discounted future income.

When agents apply time-varying discount factors, they can delay saving and increase current consumption, which may lead to sub-optimal outcomes without institutional settings that force agents to correct such behavior (via precommitment, for example).

Barro [1999] extended the Ramsey model along these lines and found that the interplay between interest rates and variable rates of time preference generates dynamic effects similar to those resulting from differences in the rate of time preference in the standard model.

However, our results deviates strongly from the no-commitment RKC model as consumption and welfare levels depends in steady state from the pure patience rate in a non-monotonic way. The Barro’s no-commitment setup leads to sub-optimal postponed saving, higher interest rate, lower capital accumulation and lower consumption and welfare levels, for every impatience level compared to the pure Ramsey case. In our myopic setup, constant growth expectations lead to a slight over-saving compared to Ramsey, due to higher “effective” impatience, leading to higher consumption-welfare levels. Such over-saving deepens for extreme low impatience leading to consumption and welfare shrinkage. We identify an analytical impatience threshold that leads higher capital accumulation compared to the Solow golden rule. This threshold is critical because of is located in an implied interest rate levels that is empirically relevant. This means that in a world where agent does not perform accurate (rational) forecasting, sub-optimal-accumulation is a probable outcome and even a small shock could lead to a good or bad welfare region. Our conclusions has relevant policy implications as, for standard model parametrization and impatience level matching observed interest rates, it would suggest a policy intervention to reduce excessive saving or, equivalently, warns policy-makers to take into account the intrinsic human behavioral nature of expectations formation mechanism against the dangerous belief that agents are entirely rational in the RKC sense due to such adverse long-run consequences. Apart from such relevant policy implications, the model displays a profound shift in the dynamical properties. On the opposite of the Ramsey model, the myopic assumption leads the stable arm to gain global stability, that is, without imposing any perfect foresight ability the consumer can use any initial condition point of the consumption-capital phase-space leading the economy to the a unique steady state. Such property is gained without increasing the model dimension (Guerrini [2010]). Such behavioral modification to the RKC model conciliates better the empirically observed speed of convergence and helps solve the Euler equation puzzle. Another important new property of the model is a biased input factor remuneration that leads to a different optimal allocation of factors leading to a biased consumption-investment mix compared to the RKC model. In particular the model suggest that wage income share finance to a larger extent saving and investment compared to the RKC model.

Our approaches differentiate deeply to other attempts to introduce distorted rational expectations as for example learning. In some way, we share some results in attempts done modifying expectation formation processes performed in rational expectation models inheriting the basic neoclassical underlying structure. In later model vintages as in New-Keynesian DSGE model, heuristics expectations have been widely studied (Jang and Sacht [2022]), however, without evaluating the global properties of the economy in terms of welfare implications, saving behavior, growth performances, capital deepness and interest rate levels.

We introduce a fixed network to compare the role of a sparse set of agents transacting on the goods, labor and capital market. We operate over a three-dimensional set of hypotheses: i) the degree of rationality ii) the initial condition, iii) the degree of scale and completeness of the network. We use the heterogeneous RKC agent model as a benchmark for the final evolution: the case of heterogeneous interacting agents. It is worth noting that the representative agent Ramsey and Myopic model are special cases of this general model and the heterogeneous versions when the network is complete and dense.

When the number of agents is two with symmetric preferences (same time preferences), the rationality is fully Ramsey. The agents act in perfect foresight (they set initial conditions such as they start on the stable arm) we demonstrate the existence of an aggregate equilibrium where agents follow its stable arm converging to the individual/aggregate steady state that is equivalent to the representative agent Ramsey case. Otherwise, if time preference is asymmetric, the equilibrium displays specialization with the more patient agent absorbing capital and the impatient providing labor services.

In the case of full Ramsey model, non-perfect foresight of two agent symmetric time preferences, i.e. starting freely both in consumption and capital initial levels and the same degree of patience, we obtain that agents with a divergent path as in the aggregate Ramsey model cannot coexist, leading to a stabilization of the economy, with a lower aggregate welfare steady state w.r.t the representative Ramsey model. Indeed, the agents constraining each other on the goods market lead consumption trajectory to stabilize.

The same combinations are used in the entirely Myopic model. The intrinsic stability of the model transfers to the individual agents with higher levels of consumption and capital. Increasing the number of agents, we explore the degree of network completeness. In the case of a complete network and symmetric time preference, the equilibrium is equivalent to the Ramsey or Myopic cases. When the network is incomplete we can measure the loss in welfare and the asymmetry of final capital/consumption distribution.

References

Robert J. Barro. Ramsey meets laibson in the neoclassical growth model. The Quarterly Journal of Economics, 114(4):1125–1152, 1999. ISSN 00335533, 15314650.

David Cass. Optimum growth in an aggregative model of capital accumulation: A turnpike theorem. Econometrica, 34(4):833–850, 1966. ISSN 00129682, 14680262.

Luca Guerrini. The ramsey model with a bounded population growth rate. Journal of Macroeconomics, 32(3):872–878, 2010. ISSN 0164-0704.

Tae-Seok Jang and Stephen Sacht. Macroeconomic dynamics under bounded rationality: on the impact of consumers’ forecast heuristics. Journal of Economic Interaction and Coordination, 17(3):849–873, July 2022.

ABSTRACT. This study examines the impact of contractionary monetary policy on income and wealth inequality, by analyzing how a policy rate increase transmits to the economy through various channels, using an agent-based model (ABM). The study aims to fill gaps in the literature by analyzing the impact of monetary policy measures on the economy through almost all transmission channels identified in the literature. The model identifies five transmission channels of monetary policy: the income composition channel, the earning heterogeneity channel, the saving redistribution channel, the portfolio composition channel, and the household debt channel. This study uses two Monte Carlo simulations, in which the first assumes no change in the policy rate, and the second assumes a policy rate increase. The results suggest that a contractionary monetary policy leads to an increase in income and wealth inequality. Regarding the income composition channel, a policy rate hike decreases disposable income for all households due to a decrease in net wages, dividends, and interest paid on loan stocks widening the gap between the lowest and highest incomes. In real terms, the decrease in the net wage produces statistical differences in the mean and median values of the two simulations only for net debtors. Furthermore, net savers' increased financial income contributes to worsening income inequality. The earning heterogeneity channel has also led to more income inequality, as a rise in interest rates increases the unemployment rate, leading to a larger gap between households with a labour income and those with unemployment benefits. The saving redistribution channel has worsened income inequality, as the rise in interest rates has led to an increase in interest paid by net debtors, which translates to more financial income for net savers through the securitization process. The portfolio composition channel shows that a rise in the policy rate leads to lower disposable income for all households in nominal terms, resulting in reduced savings and a lower accumulation of wealth stocks in deposits and shares. Net savers can benefit from lower price levels, increasing the real value of their deposits and IF's shares stocks relative to net debtors, further exacerbating wealth inequality. Finally, the study finds that the household debt channel improves wealth inequality in nominal values but worsens it in real values. Indeed, net savers reduce their real loan stocks more than net debtors, improving their balance sheets and leading to greater wealth inequality. Overall, the study suggests that a restrictive monetary policy can have significant negative effects on both income and wealth inequality.

ABSTRACT. This paper investigates the asymmetry in the power of monetary policy within an agent-based stock-flow consistent macro model with financially fragile heterogeneous households. Motivated by the current inflationary environment and the series of sharp central bank policy rate increases in response, we propose a novel theoretical framework to study how policy rate increases may differ from policy rate decreases in terms of their distributive and real effects when both the macroeconomic impact of household heterogeneity and credit are taken into account. Our preliminary results indicate that contractionary monetary policy is not simply expansionary monetary policy in reverse, but its effects crucially hinges upon the distribution of households along a number of dimensions.

ABSTRACT. Most agent-based economic models usually consist of thousands of agents. The number of agents to use is generally determined by computational capacity, the developers’ goal is to be able to run their model many times quickly enough. However, in case of economic agent-based models the role of model size has not yet been investigated nor from a computational efficiency point of view, nor from the impact on model results and consequences. Our paper examines the optimal choice of scale taking into consideration the trade-offs between runtime and accuracy of the model. For our purposes we used a complex, empirical model of the Hungarian housing market presented in Mérő et al (2022), which applied 1:1 scaling of 4 million households. Therefore, we could investigate opti-mal model size ranging from 10 thousand to 4 million agents.

According to our results, the greater the model size is, the less the role of uncertainty in the outputs is, though this relation is not linear. Two theoretical reasons could explain this phenomenon: (1) Ri-gidities are more influential with smaller size, because agents are informed via market transactions, and fewer agents take fewer transactions, therefore market clearing is slower or not reached at all. (2) Idiosynchratic shocks affect a greater part of agents both directly and indirectly.

In our model, two main mechanisms react to market disequilibria, pricing of flats and construction sector activity, while non-perfect information of agents and the time for constructing new buildings are the most important rigidities. Less agents taking less transaction strengthens the role of both ri-gidities. The goodness-of-fit of the model changed marginally in case of transaction numbers and credit amounts depending on model size, however, it improved with 85 percentage points on aver-age in case of house prices and number of newly built flats if we increased the number of agents from 10 thousand to 1 million.

Because of idiosynchratic shocks, Monte Carlo simulations are needed to eliminate the effect of randomness. In our case, averaging 15 simulations with 4 million households resulted to decrease uncertainty to less than 2 percent – which was almost the same with 1 million households. However, with 10 thousand households we needed 500 runs to reach the same level of accuracy. According to these results, although our model’s runtime increases linearly with the number of agents, the runtime to evaluate a scenario takes just 3-4 times longer with 100 times more agents. Therefore, while the impact of idiosyncratic shocks depends non-linearly on model size, increasing the number of households from 1 million to 4 million only marginally diminishes the uncertainty of outputs.

To conclude, the optimal choice of model size could be greater than what is commonly used in the literature. While all agent-based models are affected by idiosyncratic shocks and rigidities, the good-ness of fit could be improved significantly by increasing model size with just a moderate increase in model runtime.

ABSTRACT. The decline of business dynamism in the US has been abundantly documented in the empirical literature, yet there is little agreement on its underlying causes. Empirical studies indicate that this decline is correlated to a number of long-term trends, including rising market concentration, increasing mark-ups, decreasing labor share, and declining productivity growth. Akcigit and Ates (2021) suggest that the decline in technological knowledge diffusion across firms can account for most of these trends.

Following this intuition, in this paper we explore the structural causes of declining business dynamism and its macroeconomic and industrial consequences by means of a macroeconomic agent-based model centered upon knowledge accumulation. Technical change originates among capital good producers (or innovators) who produce and sell machine tools to producers of consumption goods (or entrepreneurs). Capital goods are heterogeneous with respect to built-in productivity, whose improvements depend upon a stochastic innovation/imitation process. Technical progress therefore is embodied in capital goods but entrepreneurs must accumulate sufficient technical knowledge to incorporate the newest and most innovative machines into their production process. Hence entrepreneurs perform R&D in order to accumulate technological knowledge, which enhances their ability to identify and employ the best machines produced by innovators.

Entrepreneurs are heterogeneous with respect to their technical knowledge and therefore have different capabilities to access innovations depending on their knowledge gap, i.e., the difference between capital goods' technical advancement and firms' accumulated knowledge. We first calibrate the baseline high knowledge diffusion scenario to reproduce the macroeconomic behavior of the US economy until the Eighties. Then, by modifying the parameters related to: (i) intensity of knowledge constraints in technology adoption; (ii) ability to imitate; (iii) role of technological distance in imitation process, we construct an alternative low knowledge diffusion scenario.

When the economy is characterized by low knowledge diffusion, the model generates a tendency to increasing concentration driven by widening productivity gap across firms, resulting in an increasing aggregate markup and declining labor share and GDP growth. In fact, if imitation is weak and knowledge constraints are binding, the persistent technological differences among capital goods will translate into a wider productivity gap between knowledge intensive entrepreneurs (leaders) and entrepreneurs with low accumulated knowledge (followers) because only the former are able to adopt the best techniques, thus gaining a competitive advantage over the latter. Consequently, as their market shares increase, the leaders use their enhanced market power to charge higher markups, generating a shift of income distribution from labor to profit, with negative effects on demand and growth.

To provide a more concrete interpretation of the decline in knowledge diffusion, we introduce a patent system in our model and analyze the effects of different types of patent regimes on competition and growth. A patent regime is characterized by a certain length, i.e., number of periods the patent remains valid, and breadth, i.e., the size of the protected area. We model patents so as to capture the different effects on firms' behavior documented in the literature: (i) protection effect, that inhibits imitation and blocks innovations falling within the protected area; (ii) disclosure effect that makes it easy to imitate innovations whose patents have expired; (iii) incentive effect that stimulates the innovation activity of patenting firms. We find that patents have different effects on growth depending on the stringency of the patent regime. In particular, for low values of length and breadth, patents have positive effects on GDP, especially if the incentive effect channel is active. Otherwise, if the patent regime is too stringent, output growth declines.

ABSTRACT. This paper is part of the efforts being made to analyze the impacts of shocks on the price system of the economy. In this work, we seek to know how variations in retail prices, in the exchange rate and policy interventions affect the diffusion of inflationary pressures in Mexico, based on open data from the Consumer Price Index (CPI). ). This project proposes to analyze the processes of price formation and dissemination through network techniques, applied to the case of Mexico. Empirical findings show a dynamic network structure, who evolves according to core-peripheric structure, with important implications for the policies that are usually applied against inflation.

ABSTRACT. A large body of research suggests that natural resource wealth triggers social conflicts in countries with weak institutions (for a literature review see Bayramov, 2018). When property rights cannot be adequately established, resource wealth can lengthen or increase the intensity of a conflict if it enables the groups to fund themselves and hence continue fighting instead of being forced to negotiations (Ross, 2004). Important questions are how to properly measure the intensity of a conflict and how it relates to natural resources wealth. The intensity of the conflict is (in most of the cases) measured as the total amount of fighting efforts to contest the resource and often relates to the total number of battles, deaths or attacks (Nillesen & Bulte, 2014). However, the literature has hardly analyzed it in other contexts. This paper aims to determine the intrinsic economic losses associated to network in conflict when contesting a common natural resource. To this end, we develop a metric that depends not only on the intensity of the conflict but also on the extraction capacity of agents/groups and the state of the resource.

Standard models in conflict economics are based on natural resource appropriation through contest success functions (Tullock, 1980 & Hirshleifer, 1995) and main results set manifest that a positive shock to the resource (i.e., more abundant natural resources or higher market prices) causes rent-maximizing groups to rationally allocate more fighting efforts to conflict. However, these models often ignore the economic rationality underlying resource extraction, that is, the way the resource is extracted and the main drivers of resource extraction (e.g., resource scarcity, market prices, access to technology, complete/incomplete information, etc.) and the relationship among agents/groups involved into the conflict (alliances or enmities).

Our model approaches a two-stage game. The first stage relates to resource appropriation, where agents/groups in conflict sharing a natural resource try to hoard the major quantity of it using their own strength (fighting effort) and their relationships with the other agents in the dispute. The distribution of alliances and enmities in a contest is formalized by a signed network, that is, a directed and connected type of networks that allows characterizing positive and negative interactions among agents/groups. The second stage relates to the optimization process and equilibrium. Agents' extraction depends on (i) a payoff function, (ii) their distribution of alliances and enmities, (iii) their own fighting effort, and (iv) spillover effects.

The presented model allows to fine-tune many potential scenarios. For illustrative purposes, we pay attention to the two most extreme cases. The first one is called the "Utopian" world where there exist well-defined property rights, strong institutions and therefore the resource is allocated optimally among all agents/groups. In absence of conflict, the resource can be allocated efficiently (according to Pareto) which allows maximizing social welfare. The second scenario, however, represents the opposite case where there do not exist well-defined property rights and agents/groups are myopic (according to Nash). Resource extraction in this case eventually depends on a constellation of factors such as the payoff function, spillover effects due to the network and the state of the resource.

Once these two extreme scenarios are characterized, we introduce the metric called Rent Dissipation (RD) which consists of the difference between the maximal social welfare achievable (case one) and the private (Nash) equilibrium under conflict (case two). We show that RD is always maximal in a conflict due to the Nash equilibrium profile under conflict is always lower than the Nash equilibrium without a conflict. This result implies that the sustainability of the natural resource is less compromised under conflict scenarios up to certain points.

In order to find the main driver of RD, we conduct several numerical tests of our model. Concretely, we compare RD results with different network configurations (e.g., balanced versus unbalanced networks) as well as the variation on fighting effort profiles. The scarcity of the stock and its relationship with climate change and access to international markets are also analyzed. The paper discusses some policy implications of the model.

ABSTRACT. Modern economies are characterized by the presence of highly integrated and tightly synchronized networks, that involve the global interaction of a complex nexus of heterogeneous agents. Among them, the network of trade flows is of primary relevance. Trade is, indeed, the channel through which growth and prosperity are transmitted between countries; nevertheless it could become a potential threat for economic stability, since it may convey international contagion. Thus, we aim at disentangling the mechanisms that drive trade flows and at analyzing how shocks spread throughout linkages among countries. Albeit empirical studies have documented the relevance of the trade channel in ensuing contagion among countries, this topic has received yet not enough attention in the theoretical literature. Therefore, we deem it could be useful to propose a model featuring the international propagation of shocks and policies that springs from global trade relationships. To this end, we develop a multi country general equilibrium model of trade under asymmetric information, which underpins the endogenous formation of the international network. The model encompasses two layers of firms’ heterogeneity and countries differences in fundamentals, in an environment characterized by financial frictions. Thereby, we propose a micro founded framework in which the international network endogenously emerges as a consequence of countries characteristics and firms level decisions, stemming from the interplay of differences in their productivity and financial soundness. By simultaneously embedding between and within countries heterogeneity, the theoretical setting is consistent with a number of stylized facts on globalization, capturing the characteristics of the trade networks, which would be otherwise neglected. With the resulting setup, we assess network-based spillovers, which are further amplified through the financial accelerator. In particular, the above-mentioned model is employed to explore the consequences of changes in interest rate on world trade and their indirect effects, triggered by international linkages.

ABSTRACT. Introduction Human aggregation and segregation patterns are the natural fabric of society (Clark and Fossett, 2008), determining how social phenomena evolve influencing the whole surrounding landscape. All the human activity is embedded within social networks, where agents perpetually form and change their opinions and their subsequent actions. Consequently, the formation of social networks might impact the development of virtuous collective action, or they might negatively promote vicious human values, disharmony, and undesired social outcome (Sugiarto et al., 2017). The way people interact is therefore crucial in defining the evolution and the achievement of several societal goals, such as the sustainable use of natural resources, preserving their long-term usage for present and future generations. A large branch of ecological modelling literature highlighted the need to integrate environmental models with social aspects (Grant and Thompson, 1997), understanding how people virtuously coordinate for cooperation – reducing the intensive exploitation of resources – or viciously and selfishly defect – overexploiting the actual limited resources. This social dilemma has been conceptualized by Hardin (1968) as the Tragedy of Commons, and it describes the case where shared common resources are destroyed and limited due to rational agents following their own short-term interest, and the social optimum of slow resource utilization -guarantying their regeneration and their maintenance over time (i.e., the sustainable consumption)- is not achieved. The solution of this social dilemma requires “a fundamental extension in morality”, as the author suggested, and then the spread of virtuous pro-environmental behavior. Since the Hardin’s seminal paper, different efforts have been made in opening the black box of humans’ interactions and actions related with the resource exploitation. Dietz and Henry (2008) analyzed the relationship between “the context and the commons”, emphasizing the importance of the co-evolution of social networks and the surrounding environment and remarking the role of the social influence in promoting virtuous collective actions. The theories of contagious behavior have gained momentum with the rocketing increase of connections experienced in the last decade. For instance, Eriksson et al. (2015) defined the contagious resource exploitation as the rapid spread of exploitative practices across ecosystems facilitated by intense interconnections and increasing globalization (Brockman and Helbing 2013). Building on existing literature, this work investigates how social influence of behaviors can undermine or underpin the intensity of resource exploitation. To do this, we will provide a simulated agent-based model based on the spread of behavior, where the cost of influencing agents in cooperating (or defecting) is given by the state of the environmental resource. In what follow, we summarize the different aspects composing the methodological framework.

Methodological Framework Centola and Macy (2007) coupled the behavior spread with the network features, considering two forms of contagion: simple, requiring exposure to a single agent to adopt his/her behavior; and complex, requiring social reinforcement from multiple agents. Putting it differently, the behavior propagation can be fast-paced in the simple contagion case, while it is slow-paced in the opposite case. As the authors showed, clustered reiterated interactions can favor the social reinforcement required by the diffusion of complex and costly ideas, while open random interactions reduce their propagation. The inter-relationship between opinions propagation and social networks has been widely analyzed in network theory, modelling how different opinions spreaders (the so called “zealots”) compete in diffusing their model of behavior (see, for instance, Verma et al., 2014). Combining this literature, in this work we develop a simulation model considering the co-evolutionary game theoretical approach of Weitz et al. (2016), and integrating it with the features of social contagion (Centola and Macy, 2007) governing the formation of opinions in society. As a result, we replicate the agent-based model of Centola and Macy (2007) integrating it with two competing ideas supported by two groups of initial spreaders: cooperators, opting for the slow-paced exploitation of resources, and defectors leading to a fast-paced resource over-exploitation. As in Weitz et al. (2016), the cost of adopting a strategy depends on a feedbacking system where agents are constantly updated about the state of depletion of the environment. Therefore, costly complex contagion will occur in convincing agents in cooperating where resources are abundant and unused, while convincing agents in reducing the excessive exploitation will be less costly -following the route of simple contagion- during the depletion phase, where resources are expiring. The opposite applies for the spread of defecting behavior. Network characteristics can be helpful in defining the ideal social structure guaranteeing the diffusion of virtuous behavior even when it is costly, reducing the overall over-exploitation of resources and, therefore, the related environmental damages. For instance, Clark and Fossett (2008) discussed how social segregation and the formation of communities can isolate groups of cooperators and defectors in the exploitation of a common resource pool, reducing the spread of defecting behavior in the virtuous communities. However, as discussed above, actual society are characterized by higher social interconnections, requiring a deeper study about the network feature that can facilitate the spread of cooperative behavior, containing and contrasting the diffusion of defective anti-environmental attitudes.

Expected results With the current study, we aim at shedding light on the agents and networks characteristics that can limit the impact and the number of defectors. This will contribute to the studies on the evolution of resilience thinking in socio-ecological systems (Folke, 2016). According to the author, complex networks are prominent in describing the human social dynamics and the geological process (World-Earth models) via the interactions between social networks of resource harvesting agents. Therefore, the study of the network characteristics explaining the regime shifts in socio-ecological can be a promising tool in crafting policies favoring the spread of virtuous resource-saving behaviors, achieving sustainable goals.

References Brockmann, D., & Helbing, D. (2013). The hidden geometry of complex, network-driven contagion phenomena. science, 342(6164), 1337-1342. Centola, D., & Macy, M. (2007). Complex contagions and the weakness of long ties. American journal of Sociology, 113(3), 702-734. Clark, W. A., & Fossett, M. (2008). Understanding the social context of the Schelling segregation model. Proceedings of the National Academy of Sciences, 105(11), 4109-4114. Folke, C. (2016). Resilience (republished). Ecology and society, 21(4). Grant, W. E., & Thompson, P. B. (1997). Integrated ecological models: simulation of socio-cultural constraints on ecological dynamics. Ecological Modelling, 100(1-3), 43-59. Hardin, G. (1968). The tragedy of the commons: the population problem has no technical solution; it requires a fundamental extension in morality. science, 162(3859), 1243-1248. Sugiarto, H. S., Lansing, J. S., Chung, N. N., Lai, C. H., Cheong, S. A., & Chew, L. Y. (2017). Social cooperation and disharmony in communities mediated through common pool resource exploitation. Physical review letters, 118(20), 208301. Weitz, J. S., Eksin, C., Paarporn, K., Brown, S. P., & Ratcliff, W. C. (2016). An oscillating tragedy of the commons in replicator dynamics with game-environment feedback. Proceedings of the National Academy of Sciences, 113(47), E7518-E7525.

ABSTRACT. Intra-urban residential migration is influenced by numerous factors related to both individual and societal characteristics. Various forms of location attractiveness have been introduced in the literature to quantify a specific location's role in relocation decisions. These approaches, however, are informed by aggregate data and do not consider individual-specific characteristics and constraints, such as income or transportation availability, and therefore do not capture heterogeneity in individual preferences and decisions. To address these considerations, we use a spatial agent-based housing market model calibrated to pricing data to show how heterogeneous characteristics affect residential migration. Financial constraints of the agents drive their decision to relocate based on market conditions (e.g., affordability) and spatial considerations. Using real-world data from the Greater Sydney region, we demonstrate how the interaction of financially and spatially constrained agents with heterogeneous characteristics in a simulated urban housing market can accurately depict key actual out-of-sample migration patterns.

ABSTRACT. We estimate the Alfarano-Lux-Wagner model, for the first time, on high frequency (HF) data. We opt for a brute force approach to (approximate) likelihood maximization, leveraging on abundant computational resources. The estimation employs sequential Monte Carlo sampling in order to recover the hidden variable, aka sentiment, from data. Our extensive simulation scheme is applied firstly over an artificial time series and subsequently over a real time series which covers one year (2016) of EUR/USD market data sampled at the 5 minute frequency. The first exercise shows that our extensive computation allows us to recover unbiased estimates of the true model parameters. The second exercise shows that the returns on the EUR/USD market are driven by the shifts between strong positive or negative attitudes towards the two currencies involved, since sentiment follows a bimodal distribution. Moreover, sentiment and the EUR/USD exchange rate are highly correlated. Finally, long term shifts of sentiment coincide with major events in 2016, like the Brexit referendum or the election of Donald Trump.

ABSTRACT. Short-selling restrictions (SSR) have been widely implemented during the latest phases of financial stress to maintain market stability. Policymakers are still reviewing the impact of recent policy means. Financial market plummets are understood to be behavior-driven, hence the impact on different trading behavior is at the center of the analysis. For this purpose, an agent-based financial market model, that incorporates extrapolate momentum and valuation based trading, herd behavior as well as leverage dynamics, is set-up. Furthermore, this is used to examine numerically the impact of temporarily introduced SSR, after significant price plummets. It can be seen that positive price distortion is on mean higher and negative price distortion is on mean lower. Furthermore, the cause for this upward bias can be identified. In the model, selling efforts are impeded and therefore negative returns are on mean reduced as SSR bind. Noteworthy, this suggests both: that market efficiency is impaired and that prices are prevented to fall excessively low. In addition, when opening the black box behind the market dynamics, it becomes apparent that SSR can curb the detrimental behavioral effects of herding toward speculative trading and deleveraging. Suggesting that while market efficiency is harmed, when prices are in danger of falling far under their intrinsic values, SSR are effective, temporary options to stabilize markets.

ABSTRACT. This paper proposes an heterogenous asset pricing model in which different classes of investors coexist and evolve, switching among strategies over time according to a fitness measure. In the presence of boundedly rational agents, with biased forecasts and trend following rules, rational or fundamentalist expectations do not coincide with perfect foresight ones which are not analytically obtainable. The first contribution of this paper is to propose the use of a Long-Short Term Memory Model (LSTM) to approximate the non linear and unknown functional form imposed by the presence of heterogenous investors. It is shown that when speculators use LSTM in their forecast, instead of being fundamentalists, they add instability to the system. The second contribution of the paper is empirical. Although the presence of so called noise traders in financial markets has been intensely studied, few attempts have been made in measuring their bias. Focusing on the Bitcoin market, I propose to capture the bounded rationality of noise traders by constructing an index of their bias based on textual data from Twitter. Using a dataset of more than ten million tweets containing the word “Bitcoin” I construct the Bitcoin Twitter Sentiment Index (BiTSI) through sentiment analysis in the form of the Valence Aware Dictionary and sEntiment Reasoner (VADER). The BiTSI is shown to be uncorrelated with the main factors capturing expected cryptocurrency returns identified in the literature. This suggests that the index is capturing a unique dimension of the Bitcoin market that is not accounted for in traditional financial models. Finally the heterogenous asset pricing model is estimated on daily prices by non-linear least squares, and the results confirm the switching among forecasting rules and the presence of boundedly rational investors. The model captures a significant proportion of the variation in daily returns of the cryptocurrency.

ABSTRACT. This paper presents an agent-based model with a community of investors trying to optimize their portfolios. The study compares two different methods: maximizing the Sharpe ratio and minimizing the Expected Shortfall. Traders who use the first approach are distinguished into twelve heterogeneous types, based on the methodology they use to form expectations. The portfolios performances of all thirteen techniques are evaluated and compared.

ABSTRACT. In this paper, we present a multi-country, multi-sector agent-based model that incorporates the behavioral nature of the exchange-rate dynamics and its interaction with the real side of the economy. The analysis is undertaken in an open economy that extends the model in Dosi et al. (2019). In this reformulated version, the exchange rate dynamics is influenced not only by trade flows but also by the demand of foreign currencies of traders on the exchange rate market. The complex interplay between trade flows, behavioral trade operators, and exchange rate dynamics has important implications for aggregate outcomes. Indeed, different beliefs bring feedback mechanisms and non-linear patterns. Simulations show that the introduction of speculative sentiment behavior reflects important stylized facts of the univariate bilateral exchange rate series. Furthermore, the findings show that speculative behavior substantially amplifies aggregate fluctuations and contributes to real fluctuations.

ABSTRACT. Natural hazard risks exacerbated by climate change and rapid urbanization call for accelerated climate change adaptation (CCA) at all levels of society, from governments to individuals. Empirical evidence on individual adaptation to climate-induced hazards reveals diverse behavioral and social factors affecting economic considerations. In addition, diverse adaptive capacities arise from diversity in education, incomes, experiences and institutions. Despite growing calls to model human behavior realistically, effects of adaptation strategies are still regularly quantified assuming representative rational household(s) with perfect information, producing an unaccounted gap between optimal behavior estimated by a perfectly-rational decision-maker and realistic CCA behavior. Using an evolutionary economic agent-based model parameterized by detailed survey data on households flood adaptation from Miami, USA, we quantitatively evaluate this 'adaptation deficit' under various assumptions regarding household adaptation behavior to flooding. We find that adaptation diffusion among households is significantly overestimated when assuming a representative rational agent over boundedly-rational household decision-making. The adaptation deficit is explained by heterogeneity in incomes and, more significantly, by alternative decision heuristics that account for behavioral diversity and social influences, highlighting the importance of accounting for bounded-rationality in quantifying household CCA behavior

ABSTRACT. This proposed paper brings a simulation-based analysis of the role of countercyclical capital buffer in a standard agent-based model of bank credit. The approach taken in this paper extends the standard model with a preference-based credit matching process between banks and firms. We see the credit matching as a two way classification and not as an optimization. Thus, the credit matching depends on the scoring mechanism, which is in our case based on particular feature vectors and implemented by the radial scoring function. In addition to the standard model, banks are now able to adopt a dynamic countercyclical credit policy to reflect the state in the business cycle.

The main preliminary findings confirm the stabilizing role of the countercyclical capital buffers for the economy. In general, we show how the countercyclical credit policy by banks might counteract economic cycles by impacting the distribution of clients that get their credits approved during different stages of the business cycle. In addition to this main finding, we also show, how the inclusion of some minor noise in the decisionmaking on part of banks alters the risk distribution of firms that get their credits approved. This noise might reflect the impact of uncertainty or incopmplete information on the distribution of credits that are ultimately approved for financing.

ABSTRACT. We set up an endowment based asset pricing model in which agents have heterogeneous expectations about future price levels. Expectations are a function of fundamentals, trends, and sentiment. Agents are allowed to switch between expectation formation functions based on past performance as well as sentiment. Estimation results on the S&P500 reveal that there is heterogeneity between agents, with substantial switching between groups. We find that sentiment has a direct as well as indirect effect on expectations. Specifically, heterogeneity between groups is increasing in sentiment, and sentiment reduces the frequency of switching between functions. Our results imply that the true expectation formation process is a dynamic process based on multiple information sources.

ABSTRACT. Agent-based modelling (ABMing) is a powerful and intuitive approach to modelling complex systems; however, the intractability of ABMs' likelihood functions and the non-differentiability of the mathematical operations comprising these models present a challenge to their use in the real world. These difficulties have in turn generated research on approximate Bayesian inference methods for ABMs and on constructing differentiable approximations to arbitrary ABMs, but little work has been directed towards designing approximate Bayesian inference techniques for the specific case of differentiable ABMs. In this work, we aim to address this gap and discuss how generalised variational inference procedures may be employed to provide misspecification-robust Bayesian parameter inferences for differentiable ABMs. We will demonstrate with experiments on a differentiable ABM of the COVID-19 pandemic that our approach can result in accurate inferences, and discuss avenues for future work.

ABSTRACT. We examine the relationship between salmon-related news and trading behaviour in the salmon market. For this, we create a share price index (SPI) based on five salmon aquaculture companies trading on the Oslo Stock Exchange (OSE) with the highest market cap. We use the Latent Dirichlet Allocation (LDA) machine learning algorithm to obtain topics that format the news data set. Moreover, we utilise a lexicon-based sentiment analysis where each article has been given a sentiment score depending on the number of negative and positive words appearing in the article. These methods impose structure on the otherwise unstructured text data, enabling the application of standard econometric analyses to identify effects of news on stock returns. We find that topics relating to salmon prices across different weight classes, costs, COVID-19, and sustainability have a significant negative impact on the salmon market, while topics on land-based aquaculture, and feeding have a significant positive impact. The sentiment series based on the Loughran-McDonald lexicon is found to have a negative and insignificant effect on stock returns. Hence, utilizing a lexicon solely based on financial-related words is not sufficient for an industry-specific study, which can however be overcome by adding relevant words to the dictionary. A negative shock to sentiment within the news related to competitors foreshadows a significant increase in returns. We argue that this reflects the competitive nature of salmon markets.

ABSTRACT. Calibrating agent-based models (ABMs) in economics and finance typically involves a derivative-free search in a very large parameter space. In this work, we benchmark a number of search methods in the calibration of a well-known macroeconomic ABM on real data, and further assess the performance of ``mixed strategies'' made by combining different methods. We find that methods based on random-forest surrogates are particularly efficient, and that combining search methods generally increases performance since the biases of any single method are mitigated. Moving from these observations, we propose a reinforcement learning (RL) scheme to automatically select and combine search methods on-the-fly during a calibration run. The RL agent keeps exploiting a specific method only as long as this keeps performing well, but explores new strategies when the specific method reaches a performance plateau. The resulting RL search scheme outperforms any other method or method combination tested, and does not rely on any prior information or trial and error procedure.

ABSTRACT. Recent advances in computing power have led to vigorous research in agent-based models (ABMs) in an attempt to provide a more realistic approach to modeling the economy. Despite the growing use of ABMs in various domains, the methodology for validation and calibration of these models was initially received less attention. The need for principled approaches to calibration and validation led to vigorous research activity resulting in a number of sophisticated methods for quantitative validation and calibration in recent years. The seminal work of Grazzini et. al integrated the likelihood estimates into a Bayesian framework using a kernel based density estimator to form an approximation to the likelihood function. Notably, in a recent article that compares calibration methods, Grazzini's Bayesian estimation approach outperformed minimum distance methods in providing reasonable parameter estimates in various experiments with different modes. However, the performance of the Bayesian algorithm deteriorated in tests with large-scale models with high dimensional parameter vectors. In this paper, we present an alternative likelihood-free approach to Bayesian parameter fitting to ABMs that is scalable to models with high dimensional parameter vectors. Our algorithm employs a variational autoencoder (VAE) to directly approximate the posterior over the parameters of the ABM, without explicitly constructing a likelihood function. In this context, we train VAE to learn a mapping between standard normal distributed variables (latent factors) and model parameters conditioned on the simulation output data from ABM. Specifically, using simulations we train a conditional encoder that maps model parameters to the latent factors with input from the simulation output of the ABM and a conditional decoder that maps latent factors back to model parameters with input from the output of the ABM. Once trained we can use only the conditional decoder to directly generate posterior samples from the model parameters given an output vector. We validate the proposed method empirically on data from an ABM first studied by Brock and Hommes that models interaction of heterogeneous traders that follow various trading strategies. We experiment with different number of trading strategies to assess the effect model size on estimation performance. We compare our results to a computationally expensive approximate Bayesian computation (ABC) method baseline. Our results show that the proposed conditional VAE method is effective in estimating model parameters even for large number of model parameters.

ABSTRACT. We built an agent based model (Catullo et al., 2015; Caiani et al., 2018) with heterogeneous firms which can effectuate carry trade transactions. Firms make carry trade because they borrow from the foreign bank sector in order to buy domestic government bonds, exploiting the interest spread between foreign credit and domestic one. Carry trade can be convenient for firms but at the same time it increases the instability of the economy because exchange rate fluctuations may lead to huge losses in case of devaluation of the local currency. Indeed, we tested the impact of different macroprudential policies as increasing domestic bank capital requirements or reducing the probability of providing loans to more indebted firms. As expected these measures reduce aggregate income volatility but at the cost of a credit contraction that impacts negatively on production levels. An effective measure that is able to reduce the trade-off between instability and output can be made exploiting the heterogeneity of firms in terms of their size: if just the largest are able to borrow less and have to maintain an higher liquidity reserves the economy gains stability without a reduction of production.

ABSTRACT. The goal of the paper is to analyze the macroeconomic risks associated with green transition policies. In order to do so, we design a coupling between two well-established models in the ecological economics literature: the DSK Agent-Based Macro Model (ABM) and the WiTCH Integrated Assessment Model (IAM). By bridging the two modeling approaches, we aim at exploiting their main strengths. WiTCH can be seen as an experimental laboratory, where a carbon tax is imposed on the energy sector, so to achieve a green transition through structural change within the energy sector. Moreover, WiTCH accurately describes many key variables (energy price, energy mix, emission efficiencies, etc) along the energy transition, subject to various carbon budgets and related carbon price scenarios. The aforementioned variables feedback in the macroeconomic dynamics affecting unemployment, growth, and the system stability at large. In order to analyze such feedback, we need a fully-fledged macroeconomic model, allowing for out-of-equilibrium dynamics and path dependence. This is where the DSK model comes to our aid. Our coupling is designed such that a variety of WiTCH inputs - chiefly, the energy price, the carbon tax, and the energy mix - are integrated into the DSK, in order to study the macroeconomic adjustments occurring along the transition. We, therefore, analyze the macroeconomic risks associated with three different transition patterns defined by different carbon budgets and related energy and carbon prices: an aggressive carbon tax scenario (low carbon budget), a mild carbon tax scenario (large carbon budget), and a zero carbon tax scenario. Our exercise shows that the energy transition might generate adverse macroeconomic outcomes, measured by increasing unemployment and GDP loss. The key mechanism behind this result is a redistributive one: rises in energy prices compress the wage share and since the propensity to consume out of labor income is larger compared to other sources of income and wealth, such redistribution effectively depresses aggregate demand, therefore slowing down economic activity. We also use our framework to study how fiscal and monetary policy interact with the transition. We model fiscal policy as a lump sum transfer to households of the carbon tax revenue, showing that redistributing the carbon tax collected by the government to households can alleviate the unemployment costs along the transition. In the baseline scenario, we implement a single mandate Taylor rule aimed at inflation stabilization, which is then enriched by an employment stabilization component. We show that the latter mechanism can contain unemployment along the transition, without causing large inflation relative to the baseline scenario.

ABSTRACT. The smooth functioning of overnight interbank markets is crucial for the effectiveness of central banks' transmission mechanisms and monetary policy implementation. After the European crisis, overnight unsecured interbank exchanges almost disappeared because of banks' non-accommodative behaviours. Empirical findings suggest that agents' asset choices strongly depend on individual cultural traits. Despite this, existing studies of maturity-varying interbank portfolio decisions still neglect bankers' heterogeneous personal values' role in influencing the interbank market structure. This paper develops a hybrid Agent-Based Stock-Flow model that investigates how households' and bank managers' value-based liquidity preferences affect the dynamics of the interbank market. According to our analysis, the effects of monetary policy are influenced by prevailing cultural values in different countries.

ABSTRACT. Anthropogenic climate change will have economic consequences. These include both the cost of climate damage and the cost of mitigating emissions through climate policy. Although this issue has been extensively analyzed in macroeconomic modeling, there is still significant uncertainty in several components, both in terms of functional shape and parametrization. This work investigates these costs using a macroeconomic agent-based framework with significant flexibility in the specification of the climate change and climate policy components. This is obtained by extending the Multi-Agent model for Transition Risks (MATRIX), a stock-flow consistent model endowed with an endogenous energy sector. The extensions include a climate module, a climate policy, an abatement sector, and green preferences by consumers. Supply-side climate damages will cause a significant contraction in aggregate production and real wages (between 2 and 7%), as well as an increase in nominal prices (between 3 and 11%) by 2100.In the long run, demand-side climate damages appear to reduce only nominal prices.We conduct a set of climate policy experiments to assess the economic impact of a low-carbon transition relying on a carbon tax or on consumers' green preferences (and allowing for abatement investment). Both measures contribute to reducing emissions. Nonetheless, the carbon tax is more effective at reaching more ambitious targets due to a more straightforward price signal. Overall, we observe different economic effects in the short, medium, and long term.

ABSTRACT. The spread of SARS–CoV–2 virus also known as COVID–19 is an example of an emerging pathogen resulting in a very contagious disease. Its expansion among the population has not been predictable, although the World Health Organisation has been warning for 30 years of the inevitable emergence of new pathogens that could pose a major threat to public health. The world has already witnessed emerging pathogens such as: HIV/AIDS, Ebola, West Nile virus, Hantavirus, Zika and other diseases caused by coronaviruses including SARS and MERS (Mayer and Lewis 2020). Due to the rapid contagiousness of COVID–19, most of the world economy and public life stopped in the first half of 2020, resulting in a major global economic slowdown. Since the end of 2019, the economy has been plagued by uncertainties, which was particularly evident in the hospitality and tourism sectors. As an industry based on human mobility and close contacts, tourism and hospitality have felt the consequences of the global COVID-19 pandemic (Hao et al. 2020).

Slovenia was not an exception since its economy has been similarly affected as in the rest of the world. For the most part of 2020, all catering facilities, hotel and other accommodation were closed, which was a major blow to the country’s hospitality sector, which until 2019 was largely dependent on foreign tourist arrivals and overnight stays (STO 2015, STO 2016, STO 2017, STO 2018, STO 2019, STO 2020). In 2019, 72% of all overnight stays in Slovenia were made by foreign tourists and only 28% by domestic population (STO 2020). The picture changed drastically in 2020, as 63.5% of all overnight stays were made by the Slovenian guests and only 36.4% by foreigners, which has never happened in the last twenty years (STO 2021). In addition, it should be noted that Slovenia has traditionally depended mainly on tourist arrivals and overnight stays from Italy, Germany and Austria. These countries, particularly neighbouring Italy, recorded a significant increase in infections in the first half of the year 2020. The epidemiological situation also affected the efficiency of the Slovenian hospitality sector as its capacity were largely closed for most of the year 2020. To mitigate the effects of the epidemic, the Slovenian government introduced a measure of tourist vouchers with the third anti-COVID package, which helped to save the operation of the hospitality sector only in certain parts of the country (FURS 2020; GOV 2020).

Our article aims to investigate the impact of tourist vouchers on the efficiency of the Slovenian hospitality sector in the year 2020 and the relations between redeemed vouchers and the location of the accommodation company. The survey included 609 companies that operated in the hospitality sectors and differed according to size measured by the number of beds, type of accommodation and its location by municipality. The study is divided into two parts: in the first part with the help of Data Envelopment Analysis we determine technical, pure technical and scalar efficiency, and in the second part we examine the relationship between the location factor and pure technical efficiency through cluster and correlation analyses. The location is introduced into the research using the Slovenian statistical classification by the types of the tourist municipalities. According to this classification, we classified hotel companies into six groups of municipalities: seaside, mountain, the capital of Ljubljana, city, spa and other municipalities. The individual group reflects not only the location, but also the natural and cultural attractions of a certain municipality and the related type of tourism. In 2020, the analysed hospitality enterprises recorded a higher estimate of pure technical efficiency in parts of Slovenia where most tourist vouchers were cashed. This was largely due to the decision to introduce tourist vouchers before the start of the summer tourist season and the fact that Slovenia was in complete lockdown before and after the summer of 2020.

References:

FURS (finančna uprava Republike Slovenije). Turistični BON in BON21. https://www.fu.gov.si/drugo/posebna_podrocja/turisticni_boni/ GOV (državna uprava Republike Slovenije). 2020k. Vlada potrdila tretji protikoronski paket. - https://www.gov.si/novice/2020-05-20-vlada-potrdila-tretji-protikoronski-paket/ Hao, F., Xiao, Q., & Chon, K. (2020). COVID-19 and China’s Hotel Industry: Impacts, a Disaster Management Framework, and Post-Pandemic Agenda. International Journal of Hospitality Management, 90. https://doi.org/10.1016/j.ijhm.2020.102636 Mayer, J. D., & Lewis, N. D. (2020a). An inevitable pandemic: Geographic insights into the COVID-19 global health emergency. Eurasian Geography and Economics, 61(4–5), 404–422. https://doi.org/10.1080/15387216.2020.1786425 STO (Slovenska turistična organizacija). 2015. Turizem v številkah 2014. Ljubljana: Javna agencija Republike Slovenije za trženje in promocijo turizma – Slovenska turistična organizacija. STO (Slovenska turistična organizacija). 2016. Turizem v številkah 2015. Https://www.slovenia.info/uploads/news/99/sto_tvs_2015_slo_21365_21383.pdf (16. 10. 2020). STO (Slovenska turistična organizacija). 2017. Turizem v številkah 2016. Https://www.slovenia.info/uploads/dokumenti/raziskave/2017_06_sto_tvs_2016_a4_slo_web_1.pdf (16. 10. 2020). STO (Slovenska turistična organizacija). 2018. Turizem v številkah 2017. https://www.slovenia.info/uploads/dokumenti/tvs/tvs_turizem_v_stevilkah_2017.pdf (16. 10. 2020). STO (Slovenska turistična organizacija). 2019. Letna publikacija: Turizem v številkah 2018. Https://www.slovenia.info/uploads/dokumenti/raziskave/tvs_2018/tvs_interactive.pdf (16. 10. 2020). STO (Slovenska turistična organizacija). 2020. Letna publikacija: Turizem v številkah 2019. Https://www.slovenia.info/uploads/dokumenti/raziskave/tvs_2018/tvs_interactive.pdf (16. 10. 2020). STO (Slovenska turistična organizacija). 2021. Letna publikacija: Turizem v številkah 2019. Https://www.slovenia.info/uploads/dokumenti/raziskave/tvs_2018/tvs_interactive.pdf (16. 10. 2020).

ABSTRACT. We study the impact of “vaccine nationalism”, i.e. unequal access to vaccines on the global North and the glibal South in a theoretical setting using an agent-based approach. We develop a multi-regional agent-based model that allows for a co-evolutionary perspective on a) the spread and evolution of a virus, b) vaccine development and adaption, c) immune response, c) human behavior, and d) economic outcomes. Our model builds on the viral evolution model by Mellacher (2022) and extends it in various ways: First, we add a second country that is interconnected with the first through travel. Second, human behavior is now adaptive and depends on the current danger posed by the virus as perceived by the agents, i.e., social distancing and containment policies are endogenous. Third, we introduce vaccinations that may be adapted to the currently dominant viral strain. Finally, we add a parsimonious economic dimension where Gross Domestic Product depends on social contacts within each country, as well as on international travel. While containment measures reduce GDP, so do symptomatic infections (as agents are unable to work) and voluntary social distancing (as agents stay at home fearing an infection), hence departing from an overly simplified trade-off between economy and health. Our model is able to capture salient facts about the evolution of the SARS-CoV2 virus, including its non-linear antigenic evolution as exemplified by the Omicron variant. In contrast to related literature, our model does not hinge on a fixed number of vaccine doses or new viral variants and thus offers a dynamic view on viral evolution, which allows for a longer-term perspective on the crisis. Our results indicate highlights a potential conflict of interest between the North and the South, as well as its resolution. Vaccine nationalism always While it is beneficial for the North to adopt a “vaccine nationalist” strategy if the number of vaccine doses is relatively scarce (i.e. relative to the speed of viral evolution), a “win-win” situation is possible, if the speed of the vaccination campaign can be increased and the vaccine doses can be shared equitably. This is possible due to the fact that increased vaccine coverage in the South decelerates the speed of viral evolution, reduces the spread of the virus from the South to the North and improves the viability of trade between the two regions. This result stands in contrast to a “zero-sum-game” created by a situation in which relatively few vaccine doses are available (see Figure 1, which shows global health outcomes after 1000 simulated days starting from March 2020).

ABSTRACT. Will the world’s countries continue to take action to mitigate against climate change? Since the late 1980s, countries across the world have taken steady but insufficient action to prevent climate change. Such action has taken many forms, from participating in international climate agreements (IEAs) to implementing national climate laws (de Silva and Tenreyro, 2021). However, current mitigation pathways put the world on a turbulent path to exceed 1.5°C by the 2030s (IPCC, 2022), despite increasing numbers of countries signing up to the three major IEAs to reduce greenhouse gas emissions – the 1997 Kyoto Protocol, the 2009 Copenhagen Accord and the 2015 Paris Agreement. In such a world, the effects of climate change will become increasingly strong, impacting even country's willingness to take action. Given that the effects of climate change are heterogeneous across space (Peri and Robert-Nicoud, 2021; Yohe and Schlesinger, 2002) , the geography of climate change is central for understanding how climate action will evolve over the coming decades.

Our paper tackles this question by building a simple dynamic discrete choice model to study the relationship between spatial heterogeneity and climate action. In our model, heterogeneous countries make a decision of whether to take climate action or not. Building on relevant recent empirical literature, the agents' (countries) preferences for taking action are influenced by three types of factors. The first of these is the growth rate of the stock of greenhouse gas (GHG) emissions, which acts as a proxy for future expected damages due to climate change. The second factor is climate actions of other countries. High participation increases the probability for taking action as it implies not only lower potential costs for action for each country but also due to peer pressure effects (Fankhauser et al., 2016). Finally we take into account spatially heterogeneous factors that influence costs and preferences for participation (Peri and Robert-Nicoud, 2021; Yohe and Schlesinger, 2002). The economic geography of climate change has focused on the spatially heterogenous effects of climate change, analysing its differential impact on population, fertility, migration, urbanisation, conflict, and key macroeconomic variables (Conte et al. 2021; Castells-Quitana et al, 2021; Bosetti et al., 2021, Grimm, 2021). We discuss how several of these factors influence the costs and preferences for taking climate action, including: (i) climate damages; (ii) economic resources; (iii) fossil fuel endowments; (iv) within-country inequality; and (iv) quality of institutions. For example, countries which are particularly vulnerable to climate damages are more willing to take action to mitigate future costs, while those with particularly high fossil fuel endowmenets may be disincentivised due to the importance of fossil fuel rents in their economy. Given that each country has a particular combination of these factors, influencing preferences in potentially competing ways, it is not possible to impose a simple binary between countries with either a high and low preference for participation. For example, several countries are both relatively vulnerable to climate damages but also endowed with fossil fuels. This means an alternative model is required to analyse the distribution of preferences to take action. We assume that the aggregate preferences for participation across all countries can be modelled using a logistic distribution, where the scale parameter of the logistic function (i.e. a measure of its variance) captures the degree of spatial heterogeneity across countries.

Our model gives rise to various qualitative behaviours characterising the dynamic cooperation for climate action across countries. The different cases depend on the degree of heterogeneity across countries and the initial conditions. More specifically, assuming initial conditions which resemble the current level of participation in agreements for climate action we observe three main types of possibilities. A relatively high degree of heterogeneity implies a low maximum level of cooperation for climate action in the short run, followed by an even lower level in the medium/long run. A relatively medium degree of heterogeneity leads to cyclical variations between very high and very low levels of participation for climate action. This highlights the possibility of increasing climate action in the short run followed by a decline later on. The same pattern is observed even when the degree of heterogeneity is high but the effect of peer pressure is relatively low. Finally, when the degree of heterogeneity is low and also the relative effect of peer pressure is high, then there is the possibility of sustained high levels of participation.