This literature review examines banking structures in agent-based economic simulations, drawing on the foundational works of Caiani et al. (2016, 2019), Cincotti et al. (2010), Poledna et al. (2023), and extensions including the CANVAS model of Hommes et al. (2022), as well as closely related contributions by Riccetti et al. (2013), Chan-Lau (2017), and Dosi et al. (2013). It focuses on the specification and evolution of capital constraints, the nature of loan restrictions, and equity dynamics. The review adopts the correct model view of banking: banks issue loans that simultaneously create balancing deposits; banks may later swap a portion of those deposits for new equity capital through seasoned equity offerings at market-determined prices; and banks issue equity proactively because equity prices reflect a trade-off between leverage and risk, with prices falling both when leverage is too low and when risk is too high (too close to the regulatory minimum or when too many risky loans have been issued).
The reviewed models correctly implement the endogenous creation of money. Banks issue loans that simultaneously generate equal deposits, thereby expanding balance sheets and the money supply in a stock-flow-consistent manner. This feature appears consistently across the decentralised credit networks of Caiani et al., the firm-bank matching processes of the EURACE family, the representative-bank structures of Poledna et al. and CANVAS, and the Schumpeterian frameworks of Dosi et al. The models therefore reproduce the accounting identity that loans create deposits without requiring prior reserves.
The models diverge from the correct view on equity management. None include a mechanism for banks to swap new deposits into permanent equity capital via seasoned equity offerings (SEO) priced by a market. No model allows banks to issue equity proactively when market valuation signals an attractive interior optimum on the leverage-risk frontier. Capital or net worth evolves solely through retained profits, loan-loss write-offs, and dividend payouts when buffers permit. Similarly, the models contain no mechanism for proactive equity retirement through buy-backs. Equity reduces only via write-offs and dividend payments. External equity issuance is either absent or limited to reactive bail-ins that restore the minimum ratio by reducing deposit liabilities.
Capital constraints take the form of hard minimum ratios of net worth to loans or risk-weighted assets. Thresholds range from 3 per cent in Poledna et al. and CANVAS – which employs the same simple 3 per cent minimum capital-to-loans leverage ratio as Poledna et al. (2023), supplemented by a firm-level loan-to-value cap, resulting in no material difference in the capital constraint or the associated quantity-rationing mechanism – to 6 to 12 per cent in Caiani et al. and EURACE. Banks evaluate each loan request against the post-grant ratio and ration credit if approval would breach the floor. This mechanism produces binding quantity restrictions: banks supply partial loans, deny requests, or curtail overall lending regardless of borrower willingness to pay higher rates. Interest rates adjust secondarily through mark-ups or risk premia, but never clear excess demand once the ratio binds. Capital evolves pro-cyclically. It grows with net interest income in expansions and contracts with defaults in downturns. No forward-looking seasoned equity offering rule modulates this evolution on the basis of equity price signals.
He et al. (2024) document precisely the behaviour implied by the correct view. US banks conduct seasoned equity offerings when already holding voluntary buffers well above minima (average equity-to-assets ratio of 8.84 per cent pre-SEO versus a 3 per cent leverage floor). Post-SEO, capital ratios rise further while total assets and deposits expand significantly. The proceeds fund additional lending, particularly for-sale and other loans, and risk measures increase. Market-to-book ratios fall, consistent with issuance occurring at relatively favourable valuations rather than distress. These patterns indicate that banks manage leverage and risk proactively to maximise equity valuation, swapping deposit liabilities for equity when the trade-off favours expansion without excessive dilution or regulatory pressure. Empirical evidence from other jurisdictions further augments these findings. Dinger and Vallascas (2016), using an international sample of bank seasoned equity offerings, show that the likelihood of issuance is higher for poorly capitalised banks and that such banks prefer seasoned equity offerings to alternative capitalisation strategies. Both market discipline and regulatory capital pressure influence issuance, with evidence that market valuation conditions materially shape the choice of seasoned equity offerings. This extends the US patterns by confirming that equity issuance responds to market pricing signals across a range of capital positions.
The reviewed agent-based models therefore align with He et al. (2024) and the supporting international evidence only in crisis or near-floor states, where quantity rationing and reactive recapitalisation occur. In normal conditions the models generate excessive credit rationing and fail to reproduce the observed asset and deposit expansion after equity issuance. They omit the market-price signal that induces proactive seasoned equity offerings and the associated deposit-to-equity swap. They also lack any facility for proactive equity retirement through buy-backs. As a result, the simulations are likely to overstate the frequency and severity of binding quantity constraints relative to empirical reality, where price-based equity issuance relaxes capital limits and supports growth.
Incorporating the correct view would require a behavioural rule in which each bank monitors its equity price as a function of current leverage and portfolio risk, then issues shares when valuation reaches a local maximum. The resulting inflow of equity would permit further loan creation while preserving stock-flow consistency. A symmetric rule for buy-backs when equity is over-valued would complete the picture. Such extensions would narrow the gap between simulated banking dynamics and the patterns documented in He et al. (2024) and Dinger and Vallascas (2016), yielding more realistic transmission of monetary and macroprudential policy.
References
Caiani, Alessandro, Ermanno Catullo, and Mauro Gallegati. ‘The Effects of Alternative Wage Regimes in a Monetary Union: A Multi-Country Agent Based-Stock Flow Consistent Model’. Journal of Economic Behavior & Organization 162 (June 2019): 389–416. https://doi.org/10.1016/j.jebo.2018.12.023 .
Caiani, Alessandro, Antoine Godin, Eugenio Caverzasi, Mauro Gallegati, Stephen Kinsella, and Joseph E. Stiglitz. ‘Agent Based-Stock Flow Consistent Macroeconomics: Towards a Benchmark Model’. Journal of Economic Dynamics and Control 69 (August 2016): 375–408. https://doi.org/10.1016/j.jedc.2016.06.001 .
Chan-Lau, Jorge and JChan-Lau@imf.org. ‘ABBA: An Agent-Based Model of the Banking System’. IMF Working Papers 17, no. 136 (2017): 1. https://doi.org/10.5089/9781484300688.001 .
Cincotti, Silvano, Marco Raberto, and Andrea Teglio. The EURACE Macroeconomic Model and Simulator . 24 January 2012. https://www.researchgate.net/publication/266291038_The_EURACE_macroeconomic_model_and_simulator .
Dinger, Valeriya, and Francesco Vallascas. ‘Do Banks Issue Equity When They Are Poorly Capitalized?’ Journal of Financial and Quantitative Analysis 51, no. 5 (2016): 1575–609. https://doi.org/10.1017/S0022109016000545 .
Dosi, Giovanni, Giorgio Fagiolo, Mauro Napoletano, and Andrea Roventini. ‘Income Distribution, Credit and Fiscal Policies in an Agent-Based Keynesian Model’. Journal of Economic Dynamics and Control 37, no. 8 (2013): 1598–625. https://doi.org/10.1016/j.jedc.2012.11.008 .
He, Liangliang, Hui Li, Hong Liu, and Tuyet Nhung Vu. ‘Why Do Banks Issue Equity?’ Research in International Business and Finance 69 (April 2024): 102256. https://doi.org/10.1016/j.ribaf.2024.102256 .
Hommes, Cars, Mario He, Sebastian Poledna, Melissa Siqueira, and Yang Zhang. ‘CANVAS: A Canadian Behavioral Agent-Based Model for Monetary Policy’. Journal of Economic Dynamics and Control 172 (March 2025): 104986. https://doi.org/10.1016/j.jedc.2024.104986 .
Poledna, Sebastian, Michael Gregor Miess, Cars Hommes, and Katrin Rabitsch. ‘Economic Forecasting with an Agent-Based Model’. European Economic Review 151 (January 2023): 104306. https://doi.org/10.1016/j.euroecorev.2022.104306 .
Riccetti, Luca, Alberto Russo, and Gallegati Mauro. Financial Regulation in an Agent Based Macroeconomic Model . October 2013. https://mpra.ub.uni-muenchen.de/51013/ .