{"paper_id":"2409.12516","title":"A Multi-agent Market Model Can Explain the Impact of AI Traders in Financial Markets -- A New Microfoundations of GARCH model","authors":["Kei Nakagawa","Masanori Hirano","Kentaro Minami","Takanobu Mizuta"],"paper_date":"2024-09-19","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2409.12516v1","page_url":"https://thequant.space/flowcharts/a-multi-agent-market-model-can-explain-the-impact-of-ai-trad/","scores":{"empirical_rigor":6,"math_complexity":8,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper proposes a multi‑agent market model that derives the microfoundations of the GARCH(1,1) model by aggregating the behavior of noise, fundamental, and AI traders.","quote":"This study proposes a multi-agent market model to derive the microfoundations of the GARCH model, incorporating three types of agents: noise traders, fundamental traders, and AI traders. By mathematically aggregating the micro-structure of these agents, we establish the microfoundations of the GARCH model.","summary":"The AI traders in financial markets have sparked significant interest in their effects on price formation mechanisms and market volatility, raising important questions for market stability and regulation. Despite this interest, a comprehensive model to quantitatively assess the specific impacts of…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"Fundamental traders are risk‑averse with a utility function satisfying U' > 0 and U'' < 0.","type":"model","stated":"explicit","quote":"These traders are risk-averse investors, meaning that their utility function satisfies U ′ > 0 and U ′′ < 0.","support":"full"},{"assumption":"AI traders are risk‑averse with a utility function M' > 0 and M'' < 0.","type":"model","stated":"explicit","quote":"These traders are assumed to be risk-averse investors, meaning that A New Microfoundations of GARCH model 7 M ′ > 0 and M ′′ < 0.","support":"full"},{"assumption":"The utility of both fundamental and AI traders depends on the mean and standard deviation of returns.","type":"model","stated":"explicit","quote":"Similar to fundamental traders, we assume that the M of rA,t is determined by its mean and standard deviation.","support":"full"},{"assumption":"The GARCH parameters satisfy ω > 0 and α + β < 1 to ensure stationarity.","type":"statistical","stated":"explicit","quote":"Furthermore, ω > 0 and α + β < 1 are necessary for satisfying the stationarity of the GARCH model. ω and α mean the constant volatility and the sensitivity to past return shocks, respectively. β is the dependence on past volatility, which is necessary to replicate volatility clu","support":"full"},{"assumption":"AI traders train a prediction model h by minimizing a squared loss over the return series.","type":"model","stated":"explicit","quote":"AI Traders AI traders are assumed to train each own prediction model h that effectively explains the past return series rt, t = 1, · · · , T. We assumed that they also have a utility M(·) based on the model. Therefore, we minimize: min h X t (rt −h(xt−1, It−1))2 (10) to train the model h.","support":"full"},{"assumption":"Fundamental variable x_t is generated as a standard normal random variable.","type":"data","stated":"explicit","quote":"The fundamental variable xt was generated by a random variable following a standard normal distribution.","support":"full"}]},"datasets":{"status":"available","items":[{"name":"Synthetic return series generated by the multi‑agent simulation","provider":null,"kind":"synthetic","access":"unknown","period":null,"universe":null,"quote":"In this section, we temporarily define those functions and other parameters and check how the nu-merical simulation works well by checking stylized facts, which are well-known phenomena in financial markets. We evaluate the validity of the multi-agent market model based on whether it can reproduce stylized facts [4].","support":"full"}],"known":[],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"synthetic"},"transaction_costs":{"status":"not_applicable","modeled":"n/a","headline_net_of_costs":null,"detail":null,"quote":null,"source":"front_matter","rule":"R3","candidates":{"front_matter":"n/a","evidence_card":"n/a","evidence_card_quoted":false},"meaning":"the paper reports no trading or execution result"},"results":{"status":"available","evaluation_type":"none","findings":[{"finding":"The simulated return series exhibit key stylized facts: negative skewness (‑1.880), high kurtosis (9.230), significant KS statistic (0.788), and positive first‑order autocorrelation of squared returns (0.367).","kind":"statistical","quote":"Skewness (< 0) -1.880*** Kurtosis (> 3) 9.230*** KS statistics 0.788*** 1st-order autocorrelation (> 0) 0.367*** Table 1 summarizes the statistics for the generated return series, with signif-icance levels indicated by asterisks. A New Microfoundations of GARCH model 13 The results from Table 1 demonstrate that the multi-agent market model successfully generates return series that exhibit key stylized facts observed","support":"full"},{"finding":"Increasing the proportion of fundamental traders (p1) raises the GARCH β parameter, implying stronger volatility clustering.","kind":"theoretical","quote":"If the ratio of fundamental traders p1 increased, β in the GARCH model would increase according to theorem 1. Therefore, a high ratio of fundamental traders p1 means that the strength of the volatility clustering becomes high.","support":"full"},{"finding":"Increasing the proportion of AI traders (p2) raises the GARCH α parameter, indicating a larger market response to shocks and greater turbulence.","kind":"theoretical","quote":"If the ratio of AI traders p2 increased, α in the GARCH model would increase according to theorem 1. Therefore, if there exist more AI traders in markets, the response to the exogenous diffusion, i.e., market shocks, becomes bigger, and the market turbulence becomes bigger than the expected market shocks.","support":"full"}],"reported_performance":null,"out_of_sample":{"method":"none","headline_out_of_sample":null}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":12,"citations":{"cited_by_in_corpus":0,"by_independent_authors":0,"how_cited":{"used_as_baseline":0,"uses_method_or_data":0,"extends":0,"discusses_findings":0,"mentions":0,"unclassified":0},"independent_papers":[],"scope":"Citations between papers in this corpus only; citations from outside it are not counted."}},"implementation":{"status":"available","public":false,"code_url":null,"note":"No repository presented by the authors as their own was found in the paper. Absence here is not proof that no code exists."},"replication":{"status":"none_found","card":null,"note":"No replication of this paper has been run by thequant.space yet."},"extraction":{"text_source":"fulltext","model":"openai/gpt-oss-120b","version":2,"extracted_at":"2026-11-14T19:55:27Z"},"provenance":{"claims":"Everything here is what the authors state in the paper. Nothing was re-run or reproduced by the platform.","quotes":"Assumptions, datasets, cost detail and findings were extracted by a language model. Every quote is verbatim paper text. Where an item has `support`, \"full\" means every figure in it appears in the quoted sentences and \"partial\" means a figure is stated elsewhere in the paper; items with a figure found nowhere in the paper were discarded.","datasets_known":"datasets.known lists well-known vendors and public datasets by a stable id, found by a dictionary over the full text where the sentence says the data is used; it is independent of the model and covers only names in the dictionary.","flags":"transaction_costs.modeled, results.evaluation_type and results.out_of_sample.method are platform classifications (structured-decision models); see https://thequant.space/score-guide/#penalty-flags","comparisons":"supports / contradicts / qualifies are a language model's reading of two papers' quoted findings, restricted to closely related papers that share no author. They are pointers to read, not verified replications, and coverage is partial.","implementation":"code_url is a repository the paper itself presents as its own (rule-scored, ambiguous cases judged by a model), checked for liveness when it was recorded."},"docs":"https://thequant.space/agents/"}