{"id":"fe06fa12-c094-4a10-9ee9-b871d71fe894","arxiv_id":"2509.00101","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"Consumers in Great Britain are willing to pay for anonymised smart meter data sharing, and information about privacy risks increases their reluctance to share non-anonymised data.","lead":"A survey of 965 British bill-payers finds that people value anonymising smart meter data, and telling them about privacy risks changes their willingness to share. The results suggest current consent procedures may not support truly informed choices.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified: the positive WTP/A for anonymisation is conservatively estimated and robust to acknowledged sample issues.","rationale":"I reviewed the DCE design, the WTP/A estimation framework, the treatment effect analysis, and the acknowledged limitations. The reader's weakest assumption—that stated preferences are valid and that the sample skew could bias estimates—is reasonable, but the paper provides strong internal evidence that the bias would be conservative. Over-represented groups (smart meter owners, TP attitude) exhibit lower WTP/A, so correcting toward the national population would raise the headline estimates. The manipulation check on anonymisation comprehension shows a 43% failure rate, but this would add noise and attenuate the estimate; the positive signal survives. The treatment effects are heterogeneous and weaker in the full sample, but the central claim about positive demand is based on the control group and is robust. I found no internal inconsistency or methodological error that would warrant changing the ACCEPT verdict. The proposed verification would further strengthen confidence by testing sensitivity to the main sample and comprehension concerns.","tokens_in":58779,"tokens_out":12110,"duration_ms":140980,"concrete_test":"Re-estimate the control-group mixed logit model (Table E.1) after (a) excluding respondents who failed manipulation check 3 (anonymisation comprehension) and (b) re-weighting to national smart-meter ownership (44%) and privacy-attitude benchmarks. If the mean WTP/A for anonymised half-hourly data remains significantly positive, the central claim is robust to these concerns.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central claim is a positive, measurable demand for anonymisation of smart meter data. The main estimates come from the control group's discrete choice experiment, which follows standard practice and is supported by extensive robustness checks (Tables E.1–E.7). The acknowledged sample limitations—over-representation of smart meter owners and respondents comfortable with third-party data sharing—are associated in the data with lower WTP/A (e.g., Table E.5: SM No > Yes; DSA TP < BA), so the headline figures are likely lower bounds for the general population rather than upward-biased. The high failure rate on the anonymisation comprehension check (43% incorrect, Table B.7) is a potential concern, but if anything it would attenuate the estimated value of anonymisation; the finding of positive WTP/A persists despite this noise. No internal inconsistency or modelling flaw was identified that would overturn the qualitative conclusion.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper estimates British consumers' valuation of anonymisation for smart meter data using a nationally targeted online sample of 965 bill payers, a discrete choice experiment (DCE) with an embedded randomised control trial, and mixed logit modelling. It reports a mean willingness-to-pay (WTP) for anonymised half-hourly data of about 5.4% of the monthly bill and a mean willingness-to-accept (WTA) of about 12.4%, with WTA significantly exceeding WTP. It further finds that providing information about privacy risks reduces willingness to share non-anonymised data, particularly among privacy-cautious subgroups, and that demand for smart meters remains below 70% even when anonymisation is offered. The paper claims to be the first to provide explicit monetary valuations for anonymisation of smart meter data.","tokens_in":58932,"tokens_out":11802,"duration_ms":128720,"significance":"If the central results hold, the paper fills a real gap in the energy-privacy literature: prior work has measured willingness to share data or willingness to avoid sharing, but not willingness to pay for anonymisation per se. The study is policy-relevant for the GB Market-Wide Half-Hourly Settlement and the debate on privacy-by-design versus opt-out models. Methodologically, the DCE is carefully constructed: priors from a pilot, D-efficiency selection, dominance restrictions, a manipulation check for price misinterpretation, and an extensive battery of robustness checks. Hypotheses are stated ex ante, and the paper includes both quantitative and qualitative evidence. The authors also commit to releasing code for reproducibility. These strengths make the paper a potentially useful reference for regulators and for future valuation studies of privacy-preserving techniques.","major_comments":[{"comment":"The monetary attribute distribution is described inconsistently, and the reported WTP/A estimates are not reproducible from the information given. Section 3.3.1 states that monetary parameters follow a symmetric zero-bounded triangular distribution, but Table E.1 notes that Fee(%) and Discount(%) are fixed. More importantly, for the control full sample, Eq. (2) and Table E.1 give an anonymised half-hourly WTP of (β_HH + β_Anon + β_Anon×HH)/(−α_Fee) = (0.196 + 0.510 − 0.016)/0.178 = 3.88% of bill, whereas Table E.2 reports 5.42%. The analogous WTA calculation gives 8.85%, not 12.35%. The discrepancy presumably arises from the price-misinterpretation correction or from the actual distributional assumption, but the manuscript does not show the exact computation. Please clarify the model actually estimated and verify that the headline WTP/A values are consistent with Eq. (2) and the reported","section":"§3.3.1, Eq. (2), Table E.1/E.2"},{"comment":"43% of respondents in both the control and treatment groups answered the anonymisation comprehension check incorrectly, indicating they believed anonymised data could still be linked back to them. The central DCE attribute is precisely 'anonymised', so a large share of the sample may not have understood the key concept. The paper attributes this to ambiguous wording, but provides no sensitivity analysis excluding those who failed the comprehension check or interacting comprehension with the anonymisation attribute. Given the central claim is a positive demand for anonymisation, this is a construct-validity concern. Please add a robustness check that excludes respondents who failed the anonymisation check or controls for comprehension.","section":"§3.4, Table B.7"},{"comment":"The sample over-represents smart meter owners (52.0% and 55.7% versus 44% national) and respondents with permissive data-sharing attitudes, and the paper acknowledges this. The authors argue that the estimates are lower bounds because these over-represented groups show lower WTP/A. However, no nationally reweighted WTP/A estimates are presented, and the 'lower bound' claim is not directly tested. Since the quantitative magnitudes (e.g., WTA of 12% of bill) are a central policy output, a reweighted or otherwise calibrated sensitivity analysis is needed to substantiate the claim that the headline results are conservative.","section":"§3.4, Tables B.1–B.4, §5"}],"minor_comments":[{"comment":"The notation in Eq. (3) is inconsistent: β4,j is used for both the Anon×TR interaction and the Anon×IWTS interaction. The second should be a different index (e.g., β5,j).","section":"§3.3.2, Eq. (3)"},{"comment":"Figure 7 appears as four separate figure panels with repeated captions. It should be combined into a single multi-panel figure or each panel should have a distinct caption to avoid confusion.","section":"Figure 7"},{"comment":"The text reports a WTA/WTP ratio of 3.21, but Table E.2 shows lower ratios for individual options (e.g., 2.28 for anonymised half-hourly). This is presumably because the 3.21 is the mean of the ratio distribution, whereas the table gives ratios of means. Please clarify which definition is used and ensure the text is not read as inconsistent.","section":"§4.1 and Table E.2"},{"comment":"Typo: 'technology savy' should be 'technology savvy'.","section":"§2.4"}],"recommendation":"major_revision","confidential_remarks":"The paper is a solid empirical contribution with a carefully designed DCE and rich robustness analysis. My main concerns are that (i) the WTP/A values cannot be reproduced from the reported equation and tables, and (ii) the high failure rate on the comprehension check for the key attribute is not addressed with a sensitivity analysis. Both are fixable within the manuscript's scope, but they are load-bearing for the quantitative claims. The sample-bias concern is acknowledged and argued to be conservative, but a reweighting exercise would strengthen the paper. I recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThis paper is worth your time if you work on energy data governance or privacy valuation. The genuinely new thing is that it puts a number on anonymisation as a distinct attribute: mean WTA to anonymise half-hourly data is about 12% of the monthly bill, WTP about 5.4%, with a clear WTA/WTP ratio that they read as an endowment effect. No prior GB or EU study, as far as their Table 1 shows, has priced anonymisation itself rather than general data sharing or smart meter adoption. The embedded RCT on privacy-risk information is a real improvement over the observational studies in this space.\n\nMethodologically it is careful work. The DCE follows standard practice: pilot priors, D-efficient design, dominance restrictions, mixed logit with a price mis-interpretation correction, and a long list of robustness checks (segmented models, alternative error distributions, combinatorial tests). The qualitative responses are used sensibly to interpret the numbers, not as decoration. The limitations are disclosed rather than buried: the sample over-represents smart meter owners and people comfortable sharing data, and 43% failed the anonymisation comprehension check. The authors are right that both of those issues would attenuate, not inflate, the headline WTP/A figures, so the central qualitative conclusion — positive demand for anonymisation — is probably a lower bound.\n\nSoft spots, in proportion. The treatment effect on WTP/A is weak at the full-sample level; the strong effects only appear among privacy-cautious subgroups, and the negative treatment effect among the third-party-sharing group is odd but honestly reported. The smart meter demand results show no significant treatment effect, which contradicts the hypothesis and is reported as such. Stated preferences from an online panel are always one step from real behaviour, but that is a common limitation, not a fatal one.\n\nWho is this for? Regulators working on MHHS and the GB DAPF review, and empirical privacy economists who want a comparable estimate. It deserves a serious referee: the evidence is transparent, the methods are fit for purpose, and the findings are actionable. I would send it to a good energy economics or policy journal, and expect it to survive review with revisions. I would also cite it in my own work. I'd bring it to a reading group, though probably not as the year's most exciting paper.\n\nRecommendation: engage with it.","headline":"A solid, policy-relevant DCE that delivers the first explicit monetary values for anonymisation of smart meter data; the central estimates are credible and likely conservative.","tokens_in":59429,"tokens_out":1485,"would_cite":true,"duration_ms":20740,"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":"Anonymised smart meter data is worth 12% of a UK electricity bill","keywords":["smart meters","data privacy","willingness to pay","informed consent","randomised control trial","discrete choice experiment","anonymisation","endowment effect"],"falsifier":"Offer a randomly selected set of GB households a real choice between sharing non-anonymised or anonymised half-hourly data with an actual bill discount or fee, and compare opt-in rates with the paper's predicted market shares (41% high-resolution sharing without anonymisation rising to 64% with it). A cheaper check: re-run the survey on a sample whose privacy attitudes match national benchmarks, where most adults report concern about third-party data sales, and see whether the 12% WTA and 5.4% WTP figures rise as the paper's own sample-limitation logic predicts.","tokens_in":58649,"feed_emoji":"🔒","tokens_out":10772,"duration_ms":106885,"temperature":0.7,"pith_summary":"This paper sets out to establish that British electricity bill payers have a real, measurable demand for anonymisation of their smart meter data, not just for privacy in general. Based on choices made by 965 nationally representative bill payers in an online discrete choice experiment, it estimates that households would on average require compensation of about 12% of their monthly electricity bill (roughly £7) to consent to sharing anonymised half-hourly data, and would pay about 5.4% (roughly £3) to have it anonymised, with the large gap between the two revealing a strong endowment effect. The paper also argues that consumers currently consent in ignorance: in a randomised experiment, informing people about what smart meter data can reveal roughly doubled willingness-to-pay for anonymisation among the privacy-cautious and made a substantial share less willing to share non-anonymised data. The authors conclude that current opt-out arrangements and information asymmetries suppress the true demand for anonymisation, and that privacy-by-design defaults would better match consumer preferences.","feed_headline":"Anonymised smart meter data is worth 12% of a UK electricity bill","feed_subtitle":"Compensation demands triple what consumers would pay, and privacy education doubles valuations","key_machinery":"The carrying mechanism is a discrete choice experiment answered by 965 GB bill payers and modelled with a mixed logit. Respondents repeatedly chose between two electricity contracts differing only in three attributes: the change to their monthly bill, the resolution of data shared (real-time, half-hourly, or daily), and whether the data was anonymised. Estimating the fee (bill increase) and discount (bill decrease) parameters separately lets the model separate willingness-to-pay to avoid non-anonymised real-time sharing from willingness-to-accept compensation to endure it, which is what exposes the endowment effect. The experiment embeds a randomised information treatment: half the sample wa","core_discovery":"The central discovery is that anonymisation of smart meter data has a positive, quantifiable value to consumers, and that this value is strongly asymmetric. In the control group, mean willingness-to-accept to share anonymised half-hourly data rather than non-anonymised real-time data was 12.40% of the monthly bill (£7.04), against a willingness-to-pay of 5.42% (£3.09), a WTA/WTP ratio of 3.21 read as the endowment effect applied to privacy: people treat anonymisation as a right, not a commodity to buy. Simply presenting anonymisation as an option shifted behaviour: 41.7% of respondents became more willing to share anonymised data and 26.8% less willing to share non-anonymised data. The embed","pith_inferences":["The headline valuations are plausibly lower bounds for the GB population: the online panel was more comfortable sharing data than national benchmarks (roughly half routinely pass data to third parties) and smart meter ownership was over-represented, and the paper's own logic implies a sample matching national privacy attitudes would yield higher WTP/A and stronger treatment effects.","A direct opt-in versus opt-out framing comparison within the same choice design, rather than the cross-scenario simulation used here, would quantify the default effect the paper invokes through the endowment-effect literature; this is the natural next experiment.","Because the information effect concentrates in the privacy-cautious subgroup, privacy education as currently designed may widen rather than narrow the gap between the concerned and the indifferent; targeting the least-cautious groups is the testable corollary.","Since a stubborn minority rejects smart meters regardless of anonymisation and cites supplier distrust, the next lever to test is institutional trust, for example an independent data steward, an audit regime, or a user-facing data dashboard, rather than further privacy-preserving techniques."],"forward_implications":["Offering anonymisation roughly doubles consumers' willingness to share high-resolution data (from 41% to 64% in the control group's simulated market shares), and the financial incentive needed to push uptake to 90% falls dramatically once anonymisation is available, showing bill discounts and privacy protection act as substitutes.","Anonymisation is worth more than data resolution to consumers: valuations for half-hourly and real-time sharing are statistically indistinguishable once anonymisation is offered, so regulators can include higher-resolution options without much added privacy cost.","Consent based on consumers understanding what smart meter data reveals is unlikely to be truly informed: the information treatment moved mainly those already privacy-cautious, so the least cautious remain the least informed under current disclosure practices.","Framing matters as much as price: with compensation demands about three times willingness-to-pay, an opt-out regime exploits inertia and loss aversion, while an opt-in default with anonymisation aligns with the preferences the study measures.","A sizeable minority (about a fifth of the control group, a quarter of the treatment group) would refuse a smart meter even with anonymisation available; the qualitative responses attribute this to distrust of suppliers, indicating trust-building is a separate lever from privacy engineering."],"supporting_citations":[{"why":"Closest prior stated-choice study of smart meter data privacy valuations; supplies the WTP/A estimation approach and the complete-combinatorial test used to compare distributions.","marker":"von Loessl (2023)"},{"why":"OFGEM survey that measured willingness-to-share and the change in willingness-to-share when anonymisation is offered; its 41% figure is the direct benchmark for the paper's 41.7%.","marker":"Knight (2018a)"},{"why":"GB study of willingness-to-pay for smart service contracts; provides the comparison WTP estimates and the quota approach for the nationally representative sample.","marker":"Richter and Pollitt (2018)"},{"why":"Briefing paper mapping the personal information inferable from smart meter data; provides the content of the information treatment and the privacy-utility trade-off framing.","marker":"Teng et al. (2022)"},{"why":"Source of the endowment effect and status-quo bias concepts that underpin the WTA-greater-than-WTP hypothesis and the opt-in versus opt-out discussion.","marker":"Kahneman et al. (1991)"},{"why":"Empirical evidence of WTP/WTA disparities and moral outrage in data privacy; motivates the endowment-effect hypothesis and the interpretation of qualitative responses.","marker":"Winegar and Sunstein (2019)"},{"why":"UK experimental evidence that privacy fears drive smart meter rejection and that privacy-sensitive consumers demand compensation to accept installation; baseline for the smart meter demand analysis.","marker":"Gosnell and McCoy (2023)"},{"why":"UK study showing trust determines willingness to share energy data; supports the paper's interpretation that distrust, not privacy alone, explains the stubborn minority rejecting smart meters.","marker":"Grunewald and Reisch (2020)"}],"fun_headline_variants":["Smart meter users value anonymisation at 12% of their energy bill","Privacy premium: consumers demand £7 a month to share non-anonymised data","Offering anonymisation boosts smart meter data sharing by 42%","Endowment effect on privacy: consumers want 3x compensation to give up anonymisation","Smart meter data: consumers pay £3 for privacy, demand £7 to share raw data"],"cache_read_input_tokens":2688,"weakest_assumption_plain":"The load-bearing premise is that stated choices in an online survey predict real contracting behaviour for a service that does not yet exist; the sample's below-average privacy concern and above-average smart meter ownership mean the estimates could understate true demand.","fun_headline_variants_meta":{"raw":{"variants":["Smart meter users value anonymisation at 12% of their energy bill","Privacy premium: consumers demand £7 a month to share non-anonymised data","Offering anonymisation boosts smart meter data sharing by 42%","Endowment effect on privacy: consumers want 3x compensation to give up anonymisation","Smart meter data: consumers pay £3 for privacy, demand £7 to share raw data"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001292,"raw_usage":{"total_tokens":5135,"prompt_tokens":791,"completion_tokens":4344,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":535,"completion_tokens_details":{"reasoning_tokens":4240}},"tokens_in":535,"tokens_out":4344,"duration_ms":34872,"temperature":1.0,"reasoning_tokens":4240,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-05T15:11:47.838334+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Offer a randomly selected set of GB households a real choice between sharing non-anonymised or anonymised half-hourly data with an actual bill discount or fee, and compare opt-in rates with the paper's predicted market shares (41% high-resolution sharing without anonymisation rising to 64% with it). A cheaper check: re-run the survey on a sample whose privacy attitudes match national benchmarks, where most adults report concern about third-party data sales, and see whether the 12% WTA and 5.4% WTP figures rise as the paper's own sample-limitation logic predicts.","supporting_citations":[],"review_version":1}