{"id":"8ec1a876-dbd6-4785-80f5-36a472dbd50d","arxiv_id":"2601.08340","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":3.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Most German research organizations record publication costs only partially, with non-binding procedures and limited strategic use.","lead":"A survey of 258 German research organizations finds most track publication costs only partially and without binding rules, and few use the data for strategic decisions. The findings provide a baseline for the open access transformation and the \"information budget\" debate in Germany.","discovery_kind":"incremental","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Unweighted analysis with strongly variable response rates threatens generalization of headline percentages to German RPOs.","rationale":"The paper is transparent about its methods, provides the full questionnaire, and makes anonymized data available via Zenodo. The individual percentages reported are likely correct as summaries of the 258 respondents. However, the central claim that this is a comprehensive picture of how German RPOs record publication costs requires the respondent sample to stand in for the 583-institution population. It does not, as analyzed, because response rates differ sharply by stratum and no weighting or non-response adjustment is reported. This is not a question of internal inconsistency; it is a question of external validity. The check proposed would settle whether the concern changes the point estimates materially. Even if it does, the reader's CONDITIONAL verdict remains appropriate: the findings are useful as a descriptive snapshot of engaged institutions, but population-level generalizations need qualification.","tokens_in":21302,"tokens_out":5695,"duration_ms":57096,"concrete_test":"Recompute Q3 ('systematic overview achieved'), Q14 ('binding workflows'), and Q22 ('used for strategic decisions') using inverse-probability weights by institution type (weight = number of institutions in the stratum on the mailing list / number of complete responses in that stratum, from Table 1), using the Zenodo data. If any weighted percentage moves by more than 5 percentage points from the reported unweighted value, the generalization to German RPOs should be explicitly qualified; report both weighted and unweighted estimates.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The main descriptive results (Q3, §3.2; Q14, §3.6; Q22, §3.8) are reported as simple respondent percentages, but the survey's sampling frame is a stratified population of 583 RPOs in which response rates vary from 78.4% (universities) to 6.6% (Fraunhofer) (Table 1, §2.5). The raw respondent sample overrepresents universities (69/258 = 26.7%) relative to their population share (88/583 = 15.1%) and underrepresents Fraunhofer institutes (5/258 = 1.9% vs. 76/583 = 13.0%). If cost-recording practices are correlated with institution type, the unweighted proportions in the abstract and §3 are not valid estimates for 'RPOs in Germany.' The limitations section (§4.6) acknowledges variable response rates but does not provide a non-response analysis or design-weighted estimates. The single-respondent issue (84.1% library staff, §3.1) compounds this: one library contact may not know finance-department or third-party-fund workflows, as suggested by 16.9–18.5% non-response on the non-central funding items in Q20. Since the central claim is that this is the first comprehensive picture of cost-recording across German RPOs, the unweighted analysis is the load-bearing weak point.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports a 2024 survey of German research performing organizations (RPOs) on how they record publication costs. Among 583 invited institutions, 258 completed the questionnaire (44.3%). The authors describe their sampling frame, questionnaire development (including an 11-person pretest), field phase, and open-ended coding. Main findings: 26.7% of respondents report a systematic overview of publication costs already achieved (Q3), 42.7% of those with an overview have binding workflows (Q14), 59% use the collected data for strategic decisions (Q22), and 80% view cost recording as important for the open access transformation (Q23). The paper interprets these results as evidence that German RPOs lack uniform, binding cost-transparency workflows, while central publication funds are recorded more reliably than departmental or third-party funds (Q20). The anonymized data are deposited on Zenodo.","tokens_in":21599,"tokens_out":3657,"duration_ms":35824,"significance":"This is a well-documented empirical contribution to the open access cost-transparency debate. The survey is, to my knowledge, the first comprehensive attempt to map cost-recording workflows across all major German RPO types. The methodology is transparent: the mailing-list construction is described in detail, the questionnaire was pretested and is provided in an online appendix, the field phase is documented, and the data are openly deposited. The results, if generalizable, would be directly useful for science policy and for institutions considering information budgets. The main caveat is that the population-level generalizability of the unweighted percentages is not established, given strongly heterogeneous response rates and the reliance on a single respondent per institution.","major_comments":[{"comment":"The central descriptive claims (Q3: 26.7% systematic cost overview; Q14: 42.7% binding workflows; Q22: 59% strategic use) are reported as simple unweighted percentages. Table 1 shows response rates from 6.6% (Fraunhofer) to 78.4% (universities). The raw sample overrepresents universities (69/258 = 26.7%) relative to their population share (88/583 = 15.1%) and underrepresents Fraunhofer institutes (5/258 = 1.9% vs 76/583 = 13.0%). If cost-recording practices are correlated with institution type, these unweighted estimates may not generalize to 'RPOs in Germany.' §4.6 acknowledges variable response rates but does not provide design-weighted estimates or a non-response analysis; the 'maximum sampling error' of 4.56% (§2.5) assumes simple random sampling and is not directly applicable. Please add weighted estimates or a bounding exercise (e.g., worst-case assumptions for non-responders) and","section":"§2.5, Table 1; §3.2; §3.6; §3.8"},{"comment":"Single-respondent validity is a second load-bearing issue. §3.1 reports that 84.1% of respondents are library staff, and §2.2 prioritized open access officers, publication department staff, and library directors. Q20 asks about reliability of data collection for departmental and third-party funds; the 16.9–18.5% non-response on those sub-items suggests limited knowledge outside the library domain. Additionally, 91.9% of respondents are in DEAL contracts (§3.4), and the survey was promoted via DEAL mailing lists (§2.4), likely oversampling institutions already engaged with centralized cost workflows. The manuscript should discuss how single-respondent error and this selection channel could bias the reliability ratings and the Q3/Q14 distributions, and ideally validate a subset of responses against independent institution-level data.","section":"§2.2; §2.4; §3.1; Q20 (Figure 7)"}],"minor_comments":[{"comment":"Typo: 'RPO’s' should be 'RPOs' in 'researchers at German RPO’s'.","section":"§1.4"},{"comment":"The 'Institutions on mailing list' column sums to 600 while 583 were invited; the text describes removals, but a brief note in the table caption would help readers understand the discrepancy.","section":"Table 1"},{"comment":"Q14 percentages (42.7%, 38.8%, 16.9%) are conditional on respondents who had at least a planned or achieved overview (n=178). A sentence clarifying the conditional denominator when these figures are cited in the abstract or discussion would prevent misinterpretation.","section":"§3.6"},{"comment":"In the reference list, 'Haustein et al. (2024=' contains a typo ('2024=' should be '2024').","section":"References"},{"comment":"The estimated number of publications with costs has a median of 35, a mean of 226.4, and a standard deviation of 407.6; consider reporting a trimmed mean or a log-scale visualization to reduce the impact of outliers.","section":"§3.7, Figure 6"}],"recommendation":"major_revision","confidential_remarks":"The authors have produced a carefully documented survey with publicly deposited data, which is a real strength. The main barrier to acceptance is the gap between the unweighted sample percentages and the population-level conclusions ('RPOs in Germany,' 'first comprehensive picture'). If the authors can supply design-weighted estimates or a credible non-response bounding analysis in revision, I would be glad to see this published. The single-respondent issue is also worth discussing in more depth, but it is a limitation rather than a fatal flaw."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Brief take: this is a clean, honest write-up of a useful survey, but it is an English re-presentation of results already published in German (Strecker et al., 2025b), and the abstract does not flag that. If a journal takes it, a referee should weigh the duplicate-publication issue. The methodological backbone is solid: pretested questionnaire, documented mailing list, variable response rates reported, data on Zenodo. What is actually new is the international framing, which is fine.\n\nNow the soft spots. The stress-test concern about unweighted percentages is real. Response rates range from 78.4% for universities to 6.6% for Fraunhofer. The raw sample overrepresents universities and underrepresents Fraunhofer, and if practices correlate with institution type, the headline 'majority of surveyed RPOs' is misleading as an estimate for all German RPOs. The limitations section acknowledges variable response rates but offers no weighting or non-response analysis. That is a minor-to-moderate flaw, not fatal. Also, single respondents, mostly library staff, may not know all funding streams, but that's common in this type of survey and not a deal-breaker.\n\nThe bigger issue is the prior publication. The authors disclose it in section 1.4, but not in the abstract. If the paper is meant as an English-language contribution, that's acceptable with clear cross-referencing, but the abstract should say so. This is not a case of hiding the prior work; it's just under-disclosed in the summary.\n\nWho is this for? Open access / scholarly communication researchers, university administrators, library staff, and anyone tracking cost transparency. It provides a baseline for Germany and can inform international comparisons. The descriptive findings are plausible and supported by the data. I'd cite it if I worked in this area. It deserves a serious referee, not a desk reject, because the survey itself is valuable and the methodology is transparent. The referee should ask for a design-weighted sensitivity analysis or at least a clearer statement that results apply to respondents, not all RPOs, and for the abstract to mention the German-language version.","headline":"Straightforward, transparent survey write-up whose main caveat is that the results already appeared in German; the variable response rates are a real but secondary limitation.","tokens_in":22041,"tokens_out":2734,"would_cite":true,"duration_ms":27073,"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":"This survey finds that German research institutions record publication costs only partially and mostly through non-binding workflows, despite viewing such data as important for the open access transition.","keywords":["open access transformation","publication costs","cost transparency","information budget","research performing organizations","Germany","survey","monitoring"],"falsifier":"Compare survey answers against actual financial records for a random sample of both responding and non-responding institutions: if a substantial share of non-respondents or departmental finance offices turn out to have binding cost-recording workflows, the paper's conclusion that procedures are 'often non-binding' would not hold.","tokens_in":21203,"feed_emoji":"🧾","tokens_out":3403,"duration_ms":35576,"temperature":0.7,"pith_summary":"The paper reports the first comprehensive survey of how German research-performing organizations record their publication costs. Based on 258 responses from 583 invited institutions, it claims that a majority record publication costs at least in part, but that procedures are often non-binding and data reliability varies sharply by funding source. Only about one in four institutions has a systematic overview of publication costs, and fewer than half of those have binding workflows. Although 80% of respondents call cost recording important for the open access transformation, only 59% of institutions with such data use it for strategic decisions. The paper concludes that the information budget concept, recommended for 2025, remains largely unrealized.","feed_headline":"Cost tracking widespread at German research bodies, binding rules rare","feed_subtitle":"Only 27% of 258 surveyed institutions have a full cost overview, and most workflows stay voluntary.","key_machinery":"The survey questionnaire itself is the key instrument, built around three loaded concepts: 'systematic overview of publication costs' (Question 3), 'binding workflows, procedures or instructions' (Question 14), and a funding-source reliability matrix (Question 20). These operationalizations let the authors measure transparency along two axes—whether data are collected at all, and whether collection is mandatory and reliable across different funding channels. The institution-type sampling frame and response-rate table carry the argument that the results approximate the German research landscape, despite uneven participation.","core_discovery":"The central empirical claim is that German research-performing organizations lack uniform, binding cost-transparency practices. As of 2024, 26.7% of surveyed institutions report having achieved a systematic overview of publication costs, and 43.8% say one is planned but not yet achieved. Among institutions with at least some overview, only 42.7% have binding workflows, while 38.8% have workflows without binding specifications. Reliability ratings are high for funds managed centrally—74.2% call them very reliable—but much lower for third-party funds acquired by research departments and for departmental budget funds. The paper also shows that 80% of respondents view collecting publication cost","pith_inferences":["Interpretation: the large gap between perceived importance (80%) and actual strategic use (59%) suggests that reported attitudes may overstate institutional commitment; testing actual behavior against survey answers would sharpen the picture.","Extrapolation: because response rates vary widely by institution type, non-response bias may be substantial; a quick audit of non-responding institutions' finance offices could shift the headline percentages.","Implication: pooled, centrally managed funding channels are where transparency is easiest; expanding such pooling, even before full information budgets exist, would directly improve data reliability.","Consequence: national or international cost-transparency aggregators will likely have incomplete data as long as departmental and third-party funds are not consistently captured at the source."],"forward_implications":["If the survey's estimates are correct, Germany's open access transformation is proceeding without a reliable national baseline for total publication expenditures.","Cost figures based only on centrally managed publication funds understate true spending because third-party and departmental funds are recorded far less reliably.","Most collected cost data are not yet feeding into strategic decisions, so the potential of transparency for reshaping publisher negotiations remains unrealized.","The 2025 information budget recommendation will not be met on a broad scale without stronger mandates or institutional support.","Integrating publication-output monitoring with cost monitoring could improve both, since output tracking is more advanced and can serve as a scaffold for cost data."],"fun_headline_variants":["Most German research bodies track costs, but rules stay voluntary","Only 27% of German research bodies have full cost overview","German research bodies: cost tracking common, binding rules scarce","Cost transparency seen as key, but German research bodies rarely use it","German research bodies: few have full cost picture, most rules optional"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The survey assumes that the 258 responding institutions represent all 583 invited German research-performing organizations, and that a single respondent—most often a library staff member—can accurately report how costs are recorded across finance, research departments, and central administration.","fun_headline_variants_meta":{"raw":{"variants":["Most German research bodies track costs, but rules stay voluntary","Only 27% of German research bodies have full cost overview","German research bodies: cost tracking common, binding rules scarce","Cost transparency seen as key, but German research bodies rarely use it","German research bodies: few have full cost picture, most rules optional"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00066,"raw_usage":{"total_tokens":2836,"prompt_tokens":704,"completion_tokens":2132,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":448,"completion_tokens_details":{"reasoning_tokens":2046}},"tokens_in":448,"tokens_out":2132,"duration_ms":15104,"temperature":1.0,"reasoning_tokens":2046,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-03T10:50:28.204959+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compare survey answers against actual financial records for a random sample of both responding and non-responding institutions: if a substantial share of non-respondents or departmental finance offices turn out to have binding cost-recording workflows, the paper's conclusion that procedures are 'often non-binding' would not hold.","supporting_citations":[],"review_version":1}