{"paper_id":"2509.05760","title":"Rethinking Beta: A Causal Take on CAPM","authors":["Naftali Cohen"],"paper_date":"2025-09-06","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2509.05760v3","page_url":"https://thequant.space/flowcharts/rethinking-beta-a-causal-take-on-capm/","scores":{"empirical_rigor":7,"math_complexity":8,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper casts the CAPM regression as a structural causal model using a minimal three‑node graph (Z, market return Rm, asset return Ri) to make explicit the assumptions and identify when beta reflects a causal effect versus an associational proxy.","quote":"First, we cast CAPM explicitly as a structural causal model using a minimal three-node core (𝑍, 𝑅𝑚, 𝑅𝑖), making visible the assumptions that are only implicit in standard treatments.","summary":"The CAPM regression is typically interpreted as if the market return contemporaneously \\emph{“causes”} individual returns, motivating beta-neutral portfolios and factor attribution. For realized equity returns, however, this interpretation is inconsistent: a same-period arrow R_“m,t” → R_“i,t”…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"The market return Rm is a value‑weighted aggregate of its constituent returns.","type":"market","stated":"explicit","quote":"Themarketreturnisavalue–weighted aggregate of its constituents; for an index such as the S&P 500, 𝑅𝑚,𝑡= ∑︁ 𝑗 𝑤𝑗,𝑡−1 𝑅𝑗,𝑡.","support":"full"},{"assumption":"The external driver Z is exogenous and independent of the market return X (Rm) and the asset‑specific shock.","type":"model","stated":"explicit","quote":"Over a single sampling interval Δ (time indices suppressed), 4 introduce mutually independent exogenous shocks (𝑈𝑋,𝑈𝑌,𝑈𝑍) with E(𝑈·) = 0, and posit the SCM 𝑍:= 𝑓𝑍(𝑈𝑍), 𝑋:= 𝑓𝑋(𝑈𝑋), 𝑌:= 𝑓𝑌 \u0000𝑋, 𝑍,𝑈𝑌 \u0001, (1) so that 𝑋⊥𝑍and 𝑋⊥𝑈𝑌(Fig. 1).","support":"full"},{"assumption":"The causal graph must be acyclic (no feedback loops).","type":"model","stated":"explicit","quote":"Second, we formalize what we call the Aggregator Contradiction: because 𝑅𝑚,𝑡is defined as a value-weighted aggregate of its constituents, a same-period arrow 𝑅𝑚,𝑡→𝑅𝑖,𝑡cannot co-exist with acyclicity unless the constituent is excluded (leave-one-out) or the causal effect is zero.","support":"full"},{"assumption":"Temporal priority holds: any genuine market‑to‑stock effect must respect time ordering (lagged or leave‑one‑out).","type":"model","stated":"explicit","quote":"Second, when a market→stock mechanism exists at all, it must respect time ordering (e.g., 𝑅𝑚,𝑡−1 →𝑅𝑖,𝑡) or a leave–one–out construction (𝑅−𝑖 𝑚,𝑡), and any such claim must be supported by lead–lag evidence rather than assumed.","support":"full"},{"assumption":"OLS beta is interpreted as a population projection slope under the specified DAG.","type":"statistical","stated":"explicit","quote":"OLS will always deliver a projection coefficient; whether that slope can be read as a causal path coefficient or merely as an associational proxy depends entirely on the underlying DAG and the time indices we feed it.","support":"full"},{"assumption":"The sample consists of large‑cap U.S. equities and publicly available macro series.","type":"data","stated":"explicit","quote":"We use U.S. equities and macro data from public sources (ALFRED; Yahoo Finance), 2015–2025.","support":"full"}]},"datasets":{"status":"available","items":[{"name":"U.S. large‑cap equity returns","provider":"public sources (Yahoo Finance, ALFRED)","kind":"market-prices","access":"public","period":"2015–2025","universe":"≈100 large‑cap U.S. tickers, balanced across SPDR sectors","quote":"We use U.S. equities and macro data from public sources (ALFRED; Yahoo Finance), 2015–2025. Monte Carlo estimates (points) track the analytical values (dashed lines). comprises ∼100 large-cap U.S. tickers, balanced across the 11 SPDR sectors.","support":"full","known_id":"yahoo-finance"},{"name":"SPY log return (market proxy)","provider":"public (Yahoo Finance)","kind":"market-prices","access":"public","period":"2015–2025","universe":"S&P 500 index","quote":"The market proxy 𝑅𝑚is the SPY log return.","support":"partial","known_id":"yahoo-finance"},{"name":"VIX daily change (ΔVIX)","provider":"public (CBOE)","kind":"macro","access":"public","period":"2015–2025","universe":null,"quote":"As observable proxies for the driver 𝑍, we include the daily change in VIX (ΔVIX), the change in the 10y Treasury yield (ΔDGS10), and the DXY log return; we also add the corresponding SPDR sector ETF return as a coarse sector control.2 Consumer Price Index (CPI) release dates provide exogenous macro “shock” days.","support":"partial","known_id":"cboe"},{"name":"10‑year Treasury yield change (ΔDGS10)","provider":"public (FRED)","kind":"macro","access":"public","period":"2015–2025","universe":null,"quote":"As observable proxies for the driver 𝑍, we include the daily change in VIX (ΔVIX), the change in the 10y Treasury yield (ΔDGS10), and the DXY log return; we also add the corresponding SPDR sector ETF return as a coarse sector control.2 Consumer Price Index (CPI) release dates provide exogenous macro “shock” days.","support":"partial","known_id":"fred"},{"name":"DXY log return","provider":"public (FRED)","kind":"macro","access":"public","period":"2015–2025","universe":null,"quote":"As observable proxies for the driver 𝑍, we include the daily change in VIX (ΔVIX), the change in the 10y Treasury yield (ΔDGS10), and the DXY log return; we also add the corresponding SPDR sector ETF return as a coarse sector control.2 Consumer Price Index (CPI) release dates provide exogenous macro “shock” days.","support":"partial","known_id":"fred"},{"name":"CPI release dates (macro shock days)","provider":"public (ALFRED)","kind":"macro","access":"public","period":"2015–2025","universe":null,"quote":"As observable proxies for the driver 𝑍, we include the daily change in VIX (ΔVIX), the change in the 10y Treasury yield (ΔDGS10), and the DXY log return; we also add the corresponding SPDR sector ETF return as a coarse sector control.2 Consumer Price Index (CPI) release dates provide exogenous macro “shock” days.","support":"partial"}],"known":[{"id":"yahoo-finance","name":"Yahoo Finance","provider":"Yahoo","kind":"market-prices","access":"public","quote":"We use U.S. equities and macro data from public sources (ALFRED; Yahoo Finance), 2015–2025.","mentions":1,"used_in":1}],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"daily"},"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":"unclear","findings":[{"finding":"OLS beta in a same‑period CAPM regression does not measure a causal market‑to‑stock effect but serves as an attenuated proxy for the shared driver Z.","kind":"theoretical","quote":"In this setting, OLS beta reflects not a causal transmission, but an attenuated proxy for how well $R_m$ captures the underlying driver $Z$. Using stylized models and large-cap U.S.\\ equity data, we show that contemporaneous betas act like proxies rather than mechanisms; any genuine market-to-stock channel, if at all, appears only at a lag and with modest economic significance.","support":"full"},{"finding":"Beta‑neutral portfolios remain exposed to macro or sectoral shocks, and hedging on the market index can introduce index‑specific noise.","kind":"theoretical","quote":"Consequently, ``beta-neutral'' portfolios can remain exposed to macro or sectoral shocks, and hedging on $R_m$ can import index-specific noise.","support":"full"},{"finding":"Simulation shows that in a fork graph the estimated beta attenuates smoothly as the market’s noise level (σX) increases, while the residual’s loading on Z grows; in a chain graph beta recovers the true causal coefficient and hedging removes the driver.","kind":"statistical","quote":"In the fork, 𝛽attenuates smoothly with 𝜎𝑋and the residual’s 𝑍-loading grows toward 𝑏as 𝜎𝑋rises. In the chain, 𝛽recovers 𝑐for all 𝜎𝑋and the hedge removes the driver completely.","support":"full"},{"finding":"Empirical checks on U.S. equity data reveal that contemporaneous CAPM betas behave like proxies: adding observable shocks (VIX, rates, USD, sector) collapses the market slope on CPI days, and beta‑hedged residuals still load on macro/sector shocks.","kind":"other","quote":"Figure 4: Pillar 1—contemporaneous CAPM behaves like a proxy. This is inconsistent with a stable same-day 𝑅𝑚→𝑅𝑖mechanism and consistent with a fork.","support":"full"},{"finding":"Any genuine market‑to‑stock transmission, if present, appears only with a lag and at a much smaller economic magnitude.","kind":"other","quote":"Any genuine market→stock actuation, if present, appears only at a lag and at much smaller scale. This is exactly the empirical signature implied by the minimal DAG where 𝑍 drives both 𝑅𝑚and 𝑅𝑖.","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":18,"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":"2026-11-14T19:48:12Z"},"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/"}