{"id":"05079799-f911-42a8-b87e-04c89d9a1c2f","arxiv_id":"2505.14655","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"For 39 public firms holding Bitcoin, Bitcoin returns are the dominant driver of their stock returns, with only brief episodes where the stocks influence Bitcoin, as measured by transfer entropy.","lead":"This paper studies 39 publicly traded companies that hold Bitcoin on their balance sheets and finds that their stock returns move strongly with Bitcoin, with Bitcoin usually leading the relationship. It suggests that investors and risk managers need hedging and portfolio strategies that adapt as information flows shift during events like ETF approvals and Bitcoin halvings.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Full-sample TE does not support 'BTC dominant': Table 5 significance counts are 7 vs 6 and the asymmetry rests on MSTR alone.","rationale":"I read the paper as a transparent descriptive study; Table 5 lets the reader see the balanced significance counts. The most load-bearing concern is internal: the abstract and conclusions assert consistent BTC dominance for 39 firms, but the only clear asymmetrical evidence is the MSTR rolling TE. This is not a matter of disagreeing with consensus; the paper's own table contradicts the strength of the claim. The beta estimates are contemporaneous co-movement measures and do not bear on the directional claim, so they do not rescue it. The MSTR-specific finding remains credible and could support a narrower claim about BTC-to-MSTR information flow. Multiple-testing correction and full-sample rolling TE are feasible revisions, so the reader's CONDITIONAL verdict remains appropriate; I therefore recommend UNCHANGED. I treat the time-zone alignment issue as a secondary but real validity threat, which is why my agreement with the reader's weakest assumption is only partial.","tokens_in":20653,"tokens_out":9489,"duration_ms":79880,"concrete_test":"Recompute the TE analysis with returns aligned to each exchange's UTC close (using intraday BTC prices or shifting the BTC daily return interval per ticker), apply Benjamini-Hochberg FDR at q=0.10 to the 78 p-values, and compute rolling-window significant-TE frequencies for all 39 firms rather than only MSTR. If the number of firms with significant reverse TE (X→BTC) remains comparable to the number with significant BTC→X, as in Table 5, the 'consistent BTC dominance' claim should be withdrawn or restricted to MSTR.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that 'BTC consistently acts as the dominant information driver' for the 39 sampled equities is not supported by the paper's own full-sample transfer-entropy results. In Table 5 (Appendix C), using the paper's thresholds (*: 0.05<p<0.1, **: p≤0.05), 7 firms show significant TE(BTC→X) (MSTR, CLSK, LMFA, SATO.V, RUM, 0434.HK, GNS) and 6 show significant TE(X→BTC) (HIVE, CLSK, CAN, CIFR, 0434.HK, AKER.OL). Since the empirical p-values in Eq. 5 are not corrected for multiple comparisons, and the 10% test level over 78 tests implies about 7.8 expected false positives overall (3.9 per direction), these counts are close to chance; an FDR correction would leave few or no firms significant. The global mean TEs differ by only 0.0007 bits (0.0151 vs 0.0144) with overlapping cross-sectional standard deviations (0.0066 vs 0.0060) and no confidence intervals. The clear asymmetry appears only in the rolling TE analysis of Section 5.3, which is computed for MSTR alone and is not aggregated over the 39 firms. The headline claim thus extrapolates a single-firm result to the whole sample. The time-zone synchronization issue raised by the reader is also valid and can bias one-day-lag directions, but even with perfect synchronization the reported full-sample statistics would not establish dominance.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper assembles a hand-collected dataset of 39 publicly listed firms that hold Bitcoin on their balance sheets and studies the daily return linkages between these equities and BTC over April 2023–April 2025. It reports same-day correlations (mean 0.29), single-factor BTC betas (mean 0.62, with 12 firms above 1 and MSTR at 1.37), and transfer-entropy (TE) estimates at one-day lag. The headline claim is that BTC 'consistently acts as the dominant information driver' for these equities, with reverse information flow rare and event-specific. The rolling TE analysis is performed for MSTR only, and the paper concludes that MSTR functions largely as a leveraged BTC vehicle. The authors provide code for reproduction.","tokens_in":20949,"tokens_out":2832,"duration_ms":26485,"significance":"If the central dominance claim were fully supported, the paper would make a useful empirical contribution to the literature on crypto–equity integration and on the risk profile of BTC-treasury firms. The dataset is a valuable resource, and the combination of correlation, beta, and TE measures is sensible. The paper also ships reproducible code and uses a standard, well-established TE estimator, which are clear strengths. However, the full-sample TE evidence does not establish that BTC is the dominant information driver for the 39-firm sample: the significance counts in Table 5 are nearly symmetric (7 against 6), the global mean TE difference is 0.0007 bits with overlapping standard deviations, and the asymmetry appears only in the MSTR rolling analysis. The single-firm result is interesting but does not support the cross-sectional generalization in the abstract. The time-zone misalignment of daily closing prices is an additional threat to the directional claim. These issues are fixable by reframing the claims or adding cross-firm evidence, so the paper merits a major revision.","major_comments":[{"comment":"The full-sample TE results do not support the statement that 'BTC consistently acts as the dominant information driver' for the 39 sampled equities. At the paper's own significance thresholds (*: 0.05<p<0.1, **: p≤0.05), Table 5 lists 7 firms with significant TE(BTC→X) (MSTR, CLSK, LMFA, SATO.V, RUM, 0434.HK, GNS) and 6 firms with significant TE(X→BTC) (HIVE, CLSK, CAN, CIFR, 0434.HK, AKER.OL). The global means differ by 0.0007 bits (0.0151 vs 0.0144) with cross-sectional standard deviations of 0.0066 and 0.0060, so the means are well within one standard deviation of each other. Because the empirical p-values from Eq. (5) are computed over 78 tests without any multiple-comparison correction, the expected number of false positives at the 10% level is about 7.8, which is close to the observed counts. This evidence is consistent with near-symmetric or null information flow in the full sample, not with BTC dominance. Please either restrict the dominance claim to MSTR or provide cross-firm evidence that survives multiple-testing correction.","section":"§5.1, Appendix C, Table 5"},{"comment":"The rolling TE analysis, which shows a clear asymmetry (410 significant windows for BTC→MSTR vs 179 for MSTR→BTC), is computed for MSTR alone. The abstract and introduction extrapolate this single-firm result to the whole sample ('Transfer entropy analysis consistently identifies BTC as the dominant information driver'). The paper does not aggregate rolling TE over the 39 firms, and Table 5 shows that the full-sample asymmetry is not present at the aggregate level. Please either remove the cross-firm generalization or add a cross-sectional rolling analysis (e.g., mean TE difference and significant-window counts for all 39 firms) to support it.","section":"§5.3, Figure 7"},{"comment":"The dataset uses daily closing prices from yfinance for firms trading on different exchanges and in different time zones (e.g., 0434.HK in Hong Kong, 3350.T in Tokyo, AKER.OL in Oslo, ISP.MI in Milan) along with BTC's 24/7 price. A one-day lag in return series can reflect nonsynchronous trading hours rather than genuine information flow: a stock's daily close may react to BTC movements from a different calendar day, and the direction of the resulting lagged correlation depends on the exchange's trading hours relative to UTC. The lagged correlations and TE estimates in §5.1 and Table 5 are therefore potentially biased. Please provide a robustness check that aligns observations in a common time frame (e.g., using BTC returns measured over the same local trading session) or explicitly justify that the one-day lag captures information flow for all sampled exchanges.","section":"§3, §5.1"},{"comment":"The TE significance testing is performed independently for each of the 78 directional tests, and the reported p-values are not adjusted for multiple comparisons. With a 10% test level, one expects roughly 7.8 significant results by chance across 78 tests, close to the observed 7 and 6 counts. The paper should report FDR- or family-wise-corrected p-values, or at a minimum discuss why the uncorrected counts are not treated as evidence of asymmetry. Without this, the full-sample TE significance counts in Table 5 cannot be used to support the dominance claim.","section":"§5.1, Eq. (5)"}],"minor_comments":[{"comment":"The abstract states that TE 'consistently identifies BTC as the dominant information driver' without mentioning that the asymmetry is established only for MSTR in the rolling analysis. The conclusion (§6) is appropriately MSTR-focused, so the abstract should be aligned with the evidence actually presented.","section":"Abstract and Introduction"},{"comment":"The table is labeled 'T able 3 continued' and 'T able 3' in the text, and the footnote formatting is inconsistent. Please correct the captions and ensure the table renders as a single table.","section":"Appendix A, Table 3"},{"comment":"The classification of firms into three groups is described after the figure, and the legend does not show the group boundaries explicitly. Adding clear labels or shaded regions for the three groups would improve readability.","section":"§5.2, Figure 6"},{"comment":"The rolling TE windows are described as 252 trading days with daily stride, but the text does not state the effective number of independent windows or how serial overlap affects the interpretation of the 31.6% vs 13.8% significance rates. A short discussion of this would help.","section":"§5.3"},{"comment":"The summary statistics for TE report means and standard deviations but no confidence intervals. Given the small differences, reporting bootstrap or shuffle-based confidence intervals for the mean difference would make the result more informative.","section":"§5.1, Table 2"}],"recommendation":"major_revision","confidential_remarks":"The paper is a reasonable empirical study with a useful dataset, but the main generalization in the abstract and introduction exceeds what the full-sample results support. The MSTR-specific rolling TE section is the strongest part, and the paper could be reframed around that single-firm finding. I would suggest the editor keep the manuscript under consideration for major revision, since the issues are addressable within the paper's scope. The time-zone synchronization problem is real but not fatal; a robustness check or a more cautious interpretation would suffice."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick read of Aufiero et al. on BTC-holding equities. The genuinely useful part is the dataset: 39 listed BTC-treasury firms assembled from Bitcoin Treasuries with clear inclusion thresholds, plus cross-sectional BTC betas (average 0.62, 12 firms above 1, MSTR at 1.37). That is a clean, reproducible empirical mapping. The rolling TE for MSTR also shows a real asymmetry: BTC→MSTR is significant in 31.6% of windows versus 13.8% for the reverse direction. I would trust that as a description of MSTR specifically.\n\nThe soft spot is the headline. The abstract and introduction say BTC \"consistently acts as the dominant information driver\" for the sampled equities, but the full-sample TE does not support that. In Table 5, only 7 firms show significant TE(BTC→X) and 6 show significant TE(X→BTC) at the 10% level. With 78 tests and no multiple-comparison correction, that is close to what chance would produce. The global mean difference is 0.0007 bits with overlapping standard deviations. The asymmetry appears sharply only in the MSTR rolling analysis, which is a single-firm result. So the strong claim is an overgeneralization; the defensible claim is that MSTR is driven by BTC and a handful of other BTC-treasury firms have high BTC beta.\n\nThe nonsynchronous trading concern is also valid: daily closes from yfinance for Tokyo, Hong Kong, Oslo, and Milan are aligned with a 24/7 BTC series without time-zone adjustment, which can bias one-day-lagged TE directions. That said, even with perfect alignment, the full-sample TE would not establish dominance.\n\nThe paper is a solid empirical description, not a derivation. Code is available, the regressions are simple and reproducible, and the literature review is adequate. The main fixes: tone down the abstract, apply FDR or at least discuss multiple testing, and either aggregate the rolling TE across firms or frame the result as MSTR-specific. The conclusion currently says \"our findings consistently indicate\" dominance, but the data say \"MSTR is a BTC vehicle.\"\n\nWho gets value: people working on crypto-treasury stocks, dynamic hedging, and the MSTR-BTC relationship. It deserves a serious referee — the dataset and the MSTR rolling analysis are worth publishing after major revision. I would not cite the dominance claim, but I might cite the dataset for its sample list. Recommend peer review with revision.","headline":"A useful new dataset of 39 BTC-treasury firms and a solid MSTR-specific rolling-transfer-entropy analysis, but the paper's headline claim that Bitcoin consistently dominates information flow across the whole sample is not supported by its own full-sample TE results.","tokens_in":21493,"tokens_out":2304,"would_cite":false,"duration_ms":21209,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"BTC is the dominant information driver for bitcoin-holding equities.","keywords":["bitcoin treasury","transfer entropy","BTC beta","information flow","corporate hedging","MicroStrategy","cryptocurrency equities","directional dependence"],"falsifier":"Compute the same transfer entropy and beta analysis using time-zone-aligned timestamps or intraday prices sampled at a common instant (for example, BTC at the exact equity close); if the dominant BTC-to-stock direction weakens or stock-to-BTC flow becomes comparable under alignment, the central asymmetry claim would be called into question.","tokens_in":20461,"feed_emoji":"📈","tokens_out":3965,"duration_ms":33093,"temperature":0.7,"pith_summary":"This paper asks which way information flows between Bitcoin and the shares of companies that keep Bitcoin on their balance sheets. Using daily returns for 39 listed BTC-holding firms from first acquisition through April 2025, it finds that Bitcoin is consistently the dominant driver: the average BTC beta is 0.62, twelve firms have betas above 1, and Strategy (MSTR) shows 1.37. Rolling transfer entropy confirms the asymmetry across time, with BTC-to-stock information flow significant in about 32% of windows for MSTR versus 14% in the reverse direction. The stakes are practical: if the asymmetry is real, static hedging ratios based on full-sample averages will misprice these equities during market events.","feed_headline":"Bitcoin drives the stocks that hold it, not the reverse","feed_subtitle":"Average beta 0.62 across 39 holders; MSTR moves 1.37% per 1% BTC move.","key_machinery":"The load-bearing instrument is one-day-lagged transfer entropy computed in 252-trading-day rolling windows with 1000 shuffles for significance, defined as the reduction in Shannon entropy of a firm's future returns when past BTC returns are added to the conditioning set. It is paired with a single-factor regression of each stock's daily log returns on BTC returns to estimate beta, same-day and lag-one Pearson correlations, and the Amihud illiquidity ratio, which together assign firms to exposure-liquidity groups and provide the vocabulary for interpreting who leads whom.","core_discovery":"The paper's central claim is that BTC consistently acts as the dominant information driver for the sampled BTC-holding equities, with reverse information flow from stocks to BTC rare and tied to firm-specific announcements. This is established by an average one-day-lagged transfer entropy from BTC to stocks of 0.0151 bits versus 0.0144 bits in reverse, by rolling 252-day windows in which BTC-to-MSTR transfer exceeds the reverse and is significant in 31.6% of windows against 13.8%, and by a single-factor beta of 0.62 on average with MSTR at $eta = 1.37$. The authors interpret MSTR as a leveraged financial vehicle for BTC exposure rather than a price setter.","pith_inferences":["A testable extension: intraday or time-zone-aligned data should reduce the measured BTC-to-stock information lag; if the asymmetry persists under alignment, the conclusion is robust, and if it vanishes, part of the reported dominance is a settlement-time artifact.","The results suggest the 'Bitcoin treasury' strategy is primarily a way to sell volatility and leverage, not to create alpha; firms adopting it can expect their equity to become a derivative of BTC with firm-specific noise.","By analogy, tokens other than BTC on balance sheets (e.g., Solana purchases) should exhibit weaker equity-to-token information flow because BTC is the market-wide benchmark; this is a checkable prediction."],"forward_implications":["Static hedge ratios estimated over full samples will lag real exposures, because information flow between BTC and holdings stocks concentrates in bursts around market events.","For MSTR, a 1% BTC move is associated with a 1.37% equity move, so hedges need a beta of about 1.37 and must be re-estimated as BTC's share of valuation changes.","Equity proxies for BTC, like MSTR, transmit BTC risk but only rarely transmit firm-specific news back to BTC, so diversification strategies treating them as BTC substitutes inherit market beta plus idiosyncratic firm risk.","Firms with large BTC holdings relative to market cap but low liquidity, such as Fold, show dampened correlations with BTC, implying price-impact effects mask underlying exposure in daily data."],"supporting_citations":[{"why":"Supplies the universe of publicly listed BTC-holding firms that defines the 39-company sample.","marker":"[6]"},{"why":"R package used to estimate transfer entropy and its significance in the rolling analysis.","marker":"[5]"},{"why":"Defines Wiener-Granger causality, the conceptual basis for interpreting lagged transfer entropy as directional influence.","marker":"[24]"},{"why":"Pioneering application of transfer entropy to financial time series that motivates the one-day-lagged TE design.","marker":"[36]"},{"why":"Explains MSTR's decoupling and premium monetization, used to interpret the episodic stock-to-BTC peaks.","marker":"[48]"},{"why":"Defines the Amihud illiquidity ratio used to group firms by liquidity in the three-cluster classification.","marker":"[1]"}],"fun_headline_variants":["Bitcoin leads the stocks that hold it, not the reverse","For corporate Bitcoin holders, BTC sets the pace","Bitcoin information flow outpaces its corporate stocks","MSTR and peers follow Bitcoin's information lead","Direction of info: Bitcoin to its holders, not back"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The analysis assumes that daily closing prices retrieved for firms on different exchanges are synchronized with Bitcoin's around-the-clock trading, so that a one-day lag in transfer entropy measures genuine information flow rather than artifacts of nonsynchronous trading hours.","fun_headline_variants_meta":{"raw":{"variants":["Bitcoin leads the stocks that hold it, not the reverse","For corporate Bitcoin holders, BTC sets the pace","Bitcoin information flow outpaces its corporate stocks","MSTR and peers follow Bitcoin's information lead","Direction of info: Bitcoin to its holders, not back"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000203,"raw_usage":{"total_tokens":1359,"prompt_tokens":889,"completion_tokens":470,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":505,"completion_tokens_details":{"reasoning_tokens":394}},"tokens_in":505,"tokens_out":470,"duration_ms":13397,"temperature":1.0,"reasoning_tokens":394,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T15:29:22.552115+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the same transfer entropy and beta analysis using time-zone-aligned timestamps or intraday prices sampled at a common instant (for example, BTC at the exact equity close); if the dominant BTC-to-stock direction weakens or stock-to-BTC flow becomes comparable under alignment, the central asymmetry claim would be called into question.","supporting_citations":[{"cited_title":"Bitcoin treasuries – current crypto assets held by institutions.https: //bitcointreasuries.net","cited_arxiv_id":null,"evidence_quote":"Supplies the universe of publicly listed BTC-holding firms that defines the 39-company sample."},{"cited_title":"Peter, and David J","cited_arxiv_id":null,"evidence_quote":"R package used to estimate transfer entropy and its significance in the rolling analysis."},{"cited_title":"Investigating causal relations by econometric models and cross-spectral methods","cited_arxiv_id":null,"evidence_quote":"Defines Wiener-Granger causality, the conceptual basis for interpreting lagged transfer entropy as directional influence."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Pioneering application of transfer entropy to financial time series that motivates the one-day-lagged TE design."},{"cited_title":"Microstrategy, bitcoin yield, complete markets.Available at SSRN 5038109, 2024","cited_arxiv_id":null,"evidence_quote":"Explains MSTR's decoupling and premium monetization, used to interpret the episodic stock-to-BTC peaks."}],"review_version":1}