Pith. sign in

REVIEW 4 major objections 4 minor 81 references

Object-oriented Bayesian networks can turn the European Gender Equality Index into a probabilistic, scenario-testing model of how gender-equality dimensions interact — and on Italian province data the model predicts the highest equality lev

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · deepseek-v4-flash

2026-08-01 08:24 UTC pith:XYP2ZXX5

load-bearing objection A useful applied extension of Bayesian networks to the Gender Equality Index, but the policy-effect claims outrun what conditioning on observed variables can support. the 4 major comments →

arxiv 2607.21114 v1 pith:XYP2ZXX5 submitted 2026-07-23 stat.AP

Enhancing Gender Equality Assessment through Object-Oriented Bayesian networks: the European Gender Equality Index Case

Lorenzo Giammei , Fulvia Mecatti , Flaminia Musella , Paola Vicard This is my paper
classification stat.AP
keywords Gender Equality IndexObject-Oriented Bayesian networksComposite indicatorsGender Impact AssessmentScenario analysisIntersectional statisticsItalian provincesHierarchical models
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper argues that the Gender Equality Index (GEI), a composite indicator that summarizes gender equality in a single 0–100 score, hides the very interactions that produce inequality. The authors build an Object-Oriented Bayesian Network (OOBN) that mirrors the GEI's layered structure — ingredients, sub-domains, domains — and learns from 107 Italian provinces in 2019 how those layers depend on each other and on local socio-economic context. The result is a model that can answer what-if questions: instantiate a bad income-equality level and the probability of the lowest GEI class jumps; instantiate a good one and it ripples upward across Work, Knowledge, Power, and Health. The paper shows that a constrained version of the OOBN that imitates the GEI's arithmetic structure misses these cross-domain spillovers, and it reports an AUC above 0.90 for predicting the highest GEI level. The broader claim is that the GEI can be used not only for benchmarking but as a decision-support tool for ex-ante Gender Impact Assessment.

Core claim

On the paper's own terms, the central discovery is that the GEI's hidden conditional dependencies can be learned from sub-national data and used for policy simulation. Each of the five available GEI domains (Work, Money, Knowledge, Power, Health; Time is unavailable at province level) becomes a Bayesian network instance whose output node feeds a deterministic GEI node computed by a re-weighted geometric mean. The learned networks reveal substantive structure — for example, Political power and Economic power are conditionally independent given the Employment rate, and Health status and Health access are conditionally independent given the Higher education rate. Because the OOBN is a full prob

What carries the argument

The central object is the Object-Oriented Bayesian Network (OOBN), an extension of Bayesian networks that builds a large multivariate dependence model from modular sub-networks, or 'instances', connected by interface nodes. Here each GEI domain is an instance: chance nodes represent the domain's ingredient indicator variables (discretized into quartiles), orange input nodes represent exogenous 'extra' variables such as GDP, geographical area, density, and employment rate, and a green output node is a deterministic function node that computes the domain score exactly as the official GEI methodology does (unweighted arithmetic means within sub-domains, geometric means across them). A top-level

Load-bearing premise

The load-bearing premise is that the conditional dependencies learned from 107 Italian provinces in 2019 — measured with proxy variables, quartile-discretized, and with extra variables forced to be exogenous — describe actual mechanisms that would hold under a policy intervention, rather than mere correlations that might be confounded by unmeasured factors.

What would settle it

A temporal out-of-sample test would settle the core claim: train the OOBN on 2019 Italian province data (as the paper does), then apply it to 2021 or 2023 province-level data and check whether its AUC for predicting the highest GEI class remains above 0.90 and whether scenario instantiations (e.g., imposing a high income-equality level) shift the predicted GEI distribution in the direction actually observed in provinces that did improve. If the out-of-sample AUC falls toward 0.5 or the scenario predictions diverge systematically from observed changes, the paper's predictive claim would be refu

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • The GEI can be used as a decision-support tool: policymakers can simulate an intervention on any ingredient or extra variable and see the full posterior shift of the GEI and its domains before committing resources.
  • Because the model flags cross-domain spillovers that the index formula ignores, policy evaluations based solely on the GEI score may be underestimating systemic effects; the OOBN provides a way to correct that.
  • Domain-level instances make multi-level governance possible: a regional administration can test a policy within its own domain instance while the top level aggregates the combined national effect.
  • The modular architecture transfers to other countries, sub-groups, or other composite indicators, as long as one has enough observations to learn the dependency structure.
  • The reported AUC above 0.90 is a concrete, checkable claim: the OOBN can sort provinces into the highest GEI class better than random chance, which is what you need if you want to prioritize interventions.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Going beyond the paper, if the learned dependencies are stable over time the same OOBN could forecast GEI levels under demographic or economic projections, turning a scenario tool into a forecasting tool.
  • Going beyond the paper, the modularity invites adding age, disability, or ethnicity as extra variable nodes or instances, making the intersectional claim operational beyond geography.
  • Going beyond the paper, one could rank individual ingredients by their system-wide marginal effect on the GEI posterior, producing a data-driven priority list to compare against expert-weight rankings.
  • Going beyond the paper, a natural validation is to compare the OOBN's scenario probabilities against a real policy change; the gap between constrained and unconstrained versions would quantify the value of modelling interactions.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

4 major / 4 minor

Summary. The paper proposes Object-Oriented Bayesian Networks (OOBNs) as a multivariate probabilistic complement to the European Gender Equality Index (GEI). The authors build an OOBN on 107 Italian provinces for 2019, using five of the six GEI domains because Time-domain data are unavailable at the provincial level, and include extra province-level socio-economic variables. The network is learned with the NPC algorithm on quartile-discretized variables, with the GEI aggregation encoded as deterministic function nodes. The paper reports a high AUC for predicting the highest GEI level, and presents scenario analyses in which nodes are instantiated to simulate income-equality and employment/segregation policies. A constrained version of the OOBN, in which only direct ingredient-to-domain paths are active, is used to contrast the method with a traditional composite-indicator calculation. The central claims are that OOBNs capture conditional dependencies and intersectional structure, add predictive and scenario-based capabilities, and thereby support Gender Impact Assessment.

Significance. If the claims were fully supported, the paper would provide a useful template for embedding a composite indicator into a hierarchical probabilistic model: the OOBN formulation is a natural modular extension of earlier Bayesian-network work on the GEI, and the explicit encoding of the index formula as function nodes is a practical design choice. The comparison between the unconstrained and constrained networks helpfully demonstrates how indirect associative paths change the implied GEI distribution. However, the paper's two headline contributions — predictive strength and policy-relevant scenario evaluation — are not yet established. The AUC is computed on data generated from the learned network itself, and the scenario analyses are based on probabilistic conditioning rather than causal intervention. The manuscript is therefore best read as a descriptive dependency-modeling exercise; the predictive and policy claims require major reframing and additional validation.

major comments (4)
  1. [Section 4.2, goodness-of-fit and AUC] The predictive claim is load-bearing but circular. The text states that 500 data points were "generated from the learned network" and then evaluated against that same network. The reported AUC>0.90 therefore measures internal self-consistency, not the ability to predict real province-level GEI levels on new or held-out data. Since the abstract and Section 5 claim a "predictive dimension" and the AUC is the only quantitative support for it, this needs to be replaced with out-of-sample evaluation (e.g., cross-validation on the 107 provinces), a baseline comparator (e.g., a naive or logistic model on the same variables), and ideally calibration metrics.
  2. [Section 4.3, scenario simulation and causal language] The scenarios are implemented by instantiating nodes, i.e., by conditioning: P(GEI | Employment rate = high, Work segregation = high), not by intervention: P(GEI | do(Employment rate = high), do(Work segregation = high)). The paper repeatedly says the model "simulate[s] an action" and that the results support "ex-ante policy evaluation" and "Gender Impact Assessment." With data from 107 provinces and no explicit causal identification, these posterior shifts in Tables 3 and 4 are associations that may reflect unmeasured confounding. The constraint that extra variables are not influenced by GEI ingredients is an edge-direction assumption, not a justification for a causal reading. The paper should either use do-calculus or clearly relabel all scenario results as conditional predictions, removing policy-effect language.
  3. [Section 4.1 and Section 4.2, Time domain and renormalized weights] The model uses five domains instead of the six-domain GEI, with weights renormalized to 0.22, 0.18, 0.26, 0.12, and 0.22. The resulting node is not the GEI as defined by EIGE, but a five-domain modified index. Calling it the "GEI" and claiming the framework "preserves the hierarchical structure of the index" is an overstatement. The limitation is acknowledged, but the manuscript should either include a Time-domain proxy at the provincial level, or explicitly and consistently describe the target as a modified five-domain GEI and temper conclusions that refer to the actual European index.
  4. [Tables 3 and 4, uncertainty quantification] The scenario results are reported to two decimal places without any uncertainty. With n=107, empirical quartile discretization, and many learned CPTs, the posterior probability differences highlighted in the text (e.g., GEI lowest-level probability moving from 0.91 to 0.68) may be well within sampling variability. Bootstrap confidence intervals, credible intervals, or a sensitivity analysis over the structure-learning constraints are needed before substantive claims about the size of scenario effects can be evaluated.
minor comments (4)
  1. [Throughout] Typos and style: "North-Est" should be "North-East" (Section 4.3); "istantiated" should be "instantiated" (Section 4.3); "synthetizes" in the abstract should be "synthesizes"; the phrase "preliminary position" in Section 4.2 is unclear and should be rephrased.
  2. [Section 2.3, Eq. (1)] The adjustment coefficient is typeset as ∝; this should be a clear symbol such as α, and the formula should be explicitly defined in the text to avoid confusion with the proportionality sign.
  3. [Section 3.1, Figure 4] The description of the OOBN example is dense. The distinction between input and output interface nodes is important for the later case study, so a short explicit definition in the caption or text would improve readability.
  4. [Section 4.2, software] The analysis uses Hugin Expert; the version and the specific NPC settings (significance levels, constraints, prohibited edges) should be reported for reproducibility. The reference to Steck and Tresp (1999) should be checked, since the title listed does not directly identify the NPC algorithm.

Circularity Check

1 steps flagged

The reported AUC>0.90 'predictive accuracy' is computed on 500 points simulated from the learned network itself, making it a self-consistency check rather than an out-of-sample prediction; the rest of the modeling framework is not circular.

specific steps
  1. fitted input called prediction [Section 4.2, 'An OOBN Model for GEI at the Italian Province level', goodness-of-fit/ROC paragraph]
    "In line with this, the overall goodness-of-fit of the model was assessed by generating a sample of 500 data points from the learned network and subsequently evaluating how well these data conform to the learned network structure. ... To predict the highest level of GEI, the model achieved an AUC score greater than 0.90, which, according to the classification by (Swets, 1988), indicates optimal predictive accuracy."

    The test data are generated from the fitted network and then classified with that same fitted network. With GEI and domain nodes being deterministic function nodes computed from their parents, the model will classify its own samples very accurately by construction. Therefore AUC>0.90 measures internal self-consistency, not predictive validity on held-out provinces or years. The paper's predictive claim is supported only by this circular metric, which is equivalent to evaluating the model on its own inputs.

full rationale

The central methodological contribution—using OOBNs to mirror the GEI hierarchy and learn conditional dependencies from provincial data—is a legitimate modeling exercise and is not circular: the DAG and CPTs are estimated from data, and the GEI aggregation architecture is taken from EIGE's published methodology. No load-bearing self-citation or imported uniqueness theorem is present; the prior BN work by the same authors is used only as motivation. The scenario-based 'policy' illustrations are conditional-probability queries on the fitted network, not independent predictions; concerns about conditioning versus intervention (do-calculus) are validity/identifiability limitations rather than definitional circularity. However, the paper's headline predictive validation is circular: the 500-point sample used for the ROC/AUC is drawn from the learned network itself, so the reported 'optimal predictive accuracy' is forced by construction. The limitation section acknowledges data sparsity and the missing Time domain but does not disclose this self-generated evaluation. Because one central predictive claim reduces to its own inputs while the modeling framework still has independent content, the circularity score is 6.

Axiom & Free-Parameter Ledger

3 free parameters · 5 axioms · 0 invented entities

The central claim rests on the GEI's official hierarchy and weights as a valid measure, on province-level proxies standing for the official ingredients, on an exogeneity ordering imposed on extra variables, and on finite-sample structural learning assumptions; the paper acknowledges data limitations but not the causal identifiability issue.

free parameters (3)
  • Renormalized domain weights (Work/Money/Knowledge/Power/Health) = 0.22/0.18/0.26/0.12/0.22
    Chosen to compensate for the missing Time domain; used in the GEI function node; no sensitivity analysis.
  • Empirical quartile cut points for all variables = per-variable 25/50/75 percentiles
    All variables categorized into four quartile levels before structural learning; this discretization affects edges and conditional probabilities.
  • NPC algorithm thresholds/constraints = not reported
    The NPC structure-learning parameters and user-imposed edge constraints are not fully specified, yet they determine the reported graph.
axioms (5)
  • domain assumption GEI architecture and official weights validly measure gender equality
    The OOBN takes the six-domain/fourteen-subdomain/31-ingredient structure and the official expert weights as given; if this measurement is invalid, the target node is not gender equality.
  • domain assumption Province-level proxies measure official GEI ingredients
    Tables 1-2 replace income, poverty, power participation, and health access with local proxies; no validation against official GEI scores at national level.
  • ad hoc to paper Extra variables are exogenous to gender-equality ingredients
    Section 4.2 constrains extra variables so they are not influenced by GEI ingredients; this causal ordering is imposed, not tested.
  • domain assumption NPC structural learning is consistent with n=107 and quartile-discretized data
    The analysis relies on finite-sample conditional-independence tests; no bootstrap or stability check is provided.
  • ad hoc to paper Time domain omission does not change the meaning of the GEI
    Section 4.1 says modularity lets the model use available data; the renormalized five-domain GEI is treated as comparable to the official index without empirical calibration.

pith-pipeline@v1.3.0-alltime-deepseek · 18746 in / 11633 out tokens · 105445 ms · 2026-08-01T08:24:04.158628+00:00 · methodology

0 comments
read the original abstract

A novel data-driven framework is introduced to assess gender equality by complementing and empowering a widely used European gender composite indicator, the Gender Equality Index (GEI). The GEI synthetizes the latent construct of gender equality into a single score and is extensively employed for cross-country comparison and monitoring. While effective for communication and benchmarking, this practice is affected by conceptual and methodological limitations, including marginal analysis that leaves interactions and conditional (in)dependencies unmeasured, and a lack of predictive capability. To address these limitations, this paper proposes the use of Object-Oriented Bayesian Networks (OOBNs) to model the GEI. By preserving the hierarchical structure of the index, OOBNs extend Bayesian Networks and enable a multivariate and probabilistic representation of interdependencies among the components of gender equality. This approach advances intersectional gender statistics by shifting the focus from computing a single composite score to modelling the underlying mechanisms that shape gender inequalities. The proposed methodology enhances the assessment and monitoring of gender equality and adds a predictive dimension through scenario-based evaluation, thereby supporting Gender Impact Assessment and policy decision-making. An application to Italian official statistics illustrates the practical relevance of the framework and its applicability to other national contexts and policy needs.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

81 extracted references · 5 canonical work pages

  1. [1]

    International Journal of Preventive Medicine , volume =

    Mohsen Bayati and Vahid Yazdi Feyzabadi and Arash Rashidian , title =. International Journal of Preventive Medicine , volume =

  2. [2]

    arXiv preprint arXiv:1409.0473 , year =

    Dzmitry Bahdanau and Kyunghyun Cho and Yoshua Bengio , title =. arXiv preprint arXiv:1409.0473 , year =

    Pith/arXiv arXiv

  3. [3]

    Lestón and Bertrand Mollard and Justina Reingardė , title =

    Daniele Barbieri and Marta Karu and Ignacio R. Lestón and Bertrand Mollard and Justina Reingardė , title =. 2017 , doi =

    2017

  4. [4]

    Territorial Impact Assessment of National and Regional Territorial Cohesion in Italy

    Simone Bozzato and Franco Salvatori and Alessandro Ricci and others , title =. Territorial Impact Assessment of National and Regional Territorial Cohesion in Italy. Place Evidence and Policy Orientations Towards European Green Deal , pages =

  5. [5]

    R. G. Cowell and A. P. Dawid and S. L. Lauritzen and D. J. Spiegelhalter , title =

  6. [6]

    A. P. Dawid and J. Mortera and P. Vicard , title =. Forensic Science International , volume =

  7. [7]

    Regional Studies , volume =

    Enrico di Bella and Lucia Leporatti and Luca Gandullia and Filomena Maggino , title =. Regional Studies , volume =

  8. [8]

    Measuring Gender Equality: A Multidisciplinary Analysis of Some EU Countries , pages =

    Sandra Fachelli and Pedro López-Roldán , title =. Measuring Gender Equality: A Multidisciplinary Analysis of Some EU Countries , pages =

  9. [9]

    Expert Systems with Applications , volume =

    Maxime Gasse and Alex Aussem and Haytham Elghazel , title =. Expert Systems with Applications , volume =

  10. [10]

    Erkenntnis , volume =

    Clark Glymour and Peter Spirtes and Richard Scheines , title =. Erkenntnis , volume =

  11. [11]

    Nature Reviews Genetics , volume =

    Steve Horvath and Kenneth Raj , title =. Nature Reviews Genetics , volume =

  12. [12]

    Imboden and Matthew P

    Mary T. Imboden and Matthew P. Harber and Mitchell H. Whaley and W. Holmes Finch and Derron L. Bishop and Leonard A. Kaminsky , title =. Journal of the American College of Cardiology , volume =

  13. [13]

    IEEE Transactions on Pattern Analysis and Machine Intelligence , volume =

    Shuiwang Ji and Wei Xu and Ming Yang and Kai Yu , title =. IEEE Transactions on Pattern Analysis and Machine Intelligence , volume =

  14. [14]

    N. K. Kitson and A. C. Cobstantinuou and Z. Guo and Y. Liu and K. Chobtham , title =. Artificial Intelligence Review , volume =

  15. [15]

    Proceedings of the Thirteenth Annual Conference on Uncertainty in, Artificial Intelligence , pages =

    Daphne Koller and Avi Pfeffer , title =. Proceedings of the Thirteenth Annual Conference on Uncertainty in, Artificial Intelligence , pages =

  16. [16]

    Daphne Koller , title =

  17. [17]

    Hinton , title =

    Alex Krizhevsky and Ilya Sutskever and Geoffrey E. Hinton , title =. Advances in Neural Information Processing Systems , pages =

  18. [18]

    S. L. Lauritzen , title =

  19. [19]

    Nature , volume =

    Yann LeCun and Yoshua Bengio and Geoffrey Hinton , title =. Nature , volume =

  20. [20]

    U. B. Kiaerulff and A. L. Madsen , title =

  21. [21]

    Significance , volume =

    Fulvia Mecatti and Paola Vicard and Flaminia Musella and Lorenzo Giammei , title =. Significance , volume =

  22. [22]

    C. E. Metz , title =. Seminars in Nuclear Medicine , volume =. 1978 , publisher =

    1978

  23. [23]

    Mortera and P

    J. Mortera and P. Vicard and C. Vergari , title =. The Annals of Applied Statistics , volume =

  24. [24]

    Adjeroh and Gianfranco Doretto , title =

    Saeid Motiian and Marco Piccirilli and Donald A. Adjeroh and Gianfranco Doretto , title =. Proceedings of the IEEE International Conference on Computer Vision , pages =

  25. [25]

    Murphy , title =

    Kevin P. Murphy , title =

  26. [26]

    Quality & Quantity , volume =

    Flaminia Musella and Paola Vicard , title =. Quality & Quantity , volume =

  27. [27]

    Judea Pearl , title =

  28. [28]

    Pyrkov and Evgeny Getmantsev and Boris Zhurov and Konstantin Avchaciov and Mikhail Pyatnitskiy and Leonid Menshikov and Kristina Khodova and Andrei V

    Timothy V. Pyrkov and Evgeny Getmantsev and Boris Zhurov and Konstantin Avchaciov and Mikhail Pyatnitskiy and Leonid Menshikov and Kristina Khodova and Andrei V. Gudkov and Peter O. Fedichev , title =. Aging (Albany NY) , volume =

  29. [29]

    2019 IEEE EMBS International Conference on Biomedical & Health Informatics (BHI) , pages =

    Syed Ashiqur Rahman and Donald Adjeroh , title =. 2019 IEEE EMBS International Conference on Biomedical & Health Informatics (BHI) , pages =. 2019 , publisher =

    2019

  30. [30]

    IEEE Journal of Biomedical and Health Informatics , volume =

    Daniele Ravì and Charence Wong and Fani Deligianni and Melissa Berthelot and Javier Andreu-Perez and Benny Lo and Guang-Zhong Yang , title =. IEEE Journal of Biomedical and Health Informatics , volume =

  31. [31]

    Riphahn and Caroline Schwientek , title =

    Regina T. Riphahn and Caroline Schwientek , title =. Applied Economics , volume =

  32. [32]

    Rosenberg and Kathleen Wilson , title =

    Mark W. Rosenberg and Kathleen Wilson , title =. Social Science , year =

  33. [33]

    Stuart Russell and Peter Norvig , title =

  34. [34]

    L. C. Sayer , title =. Social Forces , volume =

  35. [35]

    D. Sch. The uses and misuses of the gender-related development index and gender empowerment measure: A review of the literature , journal =

  36. [36]

    Peter Spirtes and Clark Glymour and Richard Scheines , title =

  37. [37]

    Harald Steck , title =

  38. [38]

    Harald Steck and Volker Tresp , title =

  39. [39]

    Stiglitz and Amartya Sen and Jean-Paul Fitoussi , title =

    Joseph E. Stiglitz and Amartya Sen and Jean-Paul Fitoussi , title =. France: OFCE , volume =

  40. [40]

    J. A. Swets , title =. Science , volume =

  41. [41]

    Brown and Constantin F

    Ioannis Tsamardinos and Laura E. Brown and Constantin F. Aliferis , title =. Machine Learning , volume =

  42. [42]

    Tonutti and A

    G. Tonutti and A. Garnero and G. Bertarelli and M. Pratesi , title =. Statistical Methods & Applications , volume =

  43. [43]

    Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , pages =

    Zongwei Wang and Xu Tang and Weixin Luo and Shenghua Gao , title =. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , pages =

  44. [44]

    arXiv preprint arXiv:1805.10445 , year =

    Ke Zhang and Na Liu and Xingfang Yuan and Xinyao Guo and Ce Gao and Zhenbing Zhao , title =. arXiv preprint arXiv:1805.10445 , year =

    Pith/arXiv arXiv

  45. [45]

    and Giammei, L

    Musella, F. and Giammei, L. and Romio, S. and Vicard, P. and Mecatti, F. , title =. Intermediate Scientific Meeting of the Italian Statistical Society (Ancona, 21--23 June, 2023), Book of Short Papers - SIS 2023 , year =

    2023

  46. [46]

    and Norlen, H

    Papadimitriou, E. and Norlen, H. and Del Sorbo, M. , title =. 2020 , publisher =. doi:10.2760/39645 , note =

    work page doi:10.2760/39645 2020

  47. [47]

    Journal of European Social Policy , year =

    Plantenga, Janneke and Remery, Chantal and Figueiredo, Hugo and Smith, Mark , title =. Journal of European Social Policy , year =

  48. [48]

    2020 , publisher =

    Saltelli, Andrea and Tarantola, Stefano and Campolongo, Francesca , title =. 2020 , publisher =

    2020

  49. [49]

    Handbook on Constructing Composite Indicators: Methodology and User Guide , year =

  50. [50]

    2022 , publisher =

    Hannan, Carolyn , title =. 2022 , publisher =

    2022

  51. [51]

    Gender Mainstreaming: Conceptual Framework, Methodology and Presentation of Good Practices , year =

  52. [52]

    and Ammons, S

    Dahlin, E. and Ammons, S. K. and Rugh, J. S. and Sumsion, R. and Hebertson, J. , title =. International Journal of Sociology and Social Policy , year =

  53. [53]

    Gender Mainstreaming Toolkit , year =

    Gender Impact Assessment. Gender Mainstreaming Toolkit , year =

  54. [54]

    2023 , url =

    Nedera, Steliana , title =. 2023 , url =

    2023

  55. [55]

    , title =

    Berik, G. , title =. 2022 , institution =

    2022

  56. [56]

    and Crippa, F

    Mecatti, F. and Crippa, F. and Farina, P. , title =. International Statistical Review , year =

  57. [57]

    European Institute for Gender Equality , year = 2025, url =

    2025

  58. [58]

    , year = 1987, journal =

    Almquist, Elizabeth M. , year = 1987, journal =. Labor. 189634 , eprinttype =

    1987

  59. [59]

    Azmat, Ghazala and G. Gender. 2006 , month = jan, journal =. doi:10.1086/497817 , urldate =

    doi:10.1086/497817 2006

  60. [60]

    Geographical Disparities in the Health of Iranian Women:

    Bayati, Mohsen and Feyzabadi, VahidYazdi and Rashidian, Arash , year = 2017, journal =. Geographical Disparities in the Health of Iranian Women:. doi:10.4103/ijpvm.IJPVM_67_16 , urldate =

    work page doi:10.4103/ijpvm.ijpvm_67_16 2017

  61. [61]

    Gender Segregation in the Labour Market:

    Bettio, Francesca and Verashchagina, Alina and Mairhuber, Ingrid and. Gender Segregation in the Labour Market:. 2009 , publisher =

    2009

  62. [62]

    and Kahn, Lawrence M

    Blau, Francine D. and Kahn, Lawrence M. , year = 2017, month = sep, journal =. The. doi:10.1257/jel.20160995 , urldate =

    doi:10.1257/jel.20160995 2017

  63. [63]

    Education,

    Chevalier, Arnaud , year =. Education,. Oxford Bulletin of Economics and Statistics , volume =. doi:10.1111/j.1468-0084.2007.00483.x , urldate =

    arXiv 2007

  64. [64]

    Scandinavian Journal of Public Health , volume =

    Gender Inequality at Home Is Associated with Poorer Health for Women , author =. Scandinavian Journal of Public Health , volume =. 2015 , month =. doi:10.1177/1403494814562598 , urldate =

    work page doi:10.1177/1403494814562598 2015

  65. [65]

    and Lawless, Jennifer L

    Fox, Richard L. and Lawless, Jennifer L. , year =. Uncovering the. American Political Science Review , volume =. doi:10.1017/S0003055414000227 , urldate =

    work page doi:10.1017/s0003055414000227

  66. [66]

    Occupational

    Gauchat, Gordon and Kelly, Maura and Wallace, Michael , year =. Occupational. Gender & Society , volume =. doi:10.1177/0891243212453647 , urldate =

    work page doi:10.1177/0891243212453647

  67. [67]

    Gender Inequality in Politics:

    Lombardo, Emanuela , year =. Gender Inequality in Politics:. International Feminist Journal of Politics , volume =

  68. [68]

    Social Science , langid =

    Gender, Poverty and Location: How Much Di€erence Do They Make in the Geography of Health Inequalities? , author =. Social Science , langid =

  69. [69]

    Sayer, L. C. , year = 2005, month = sep, journal =. Gender,. doi:10.1353/sof.2005.0126 , urldate =

    arXiv 2005

  70. [70]

    2024 , issn =

    The gender gap in life expectancy and lifespan disparity as social risk indicators for international countries: A fuzzy clustering approach , journal =. 2024 , issn =. doi:https://doi.org/10.1016/j.seps.2023.101712 , url =

    arXiv 2024

  71. [71]

    The Measurement of Multidimensional Gender Inequality: Continuing the Debate , urldate =

    Iñaki Permanyer , journal =. The Measurement of Multidimensional Gender Inequality: Continuing the Debate , urldate =

  72. [72]

    Transforming our World: the 2030 Agenda for Sustainable Development , year =

    2030

  73. [73]

    Treaty on the European Union (Consolidated Version) , year =

  74. [74]

    Gender Equality , url =

  75. [75]

    Economic Benefits of Gender Equality , url =

  76. [76]

    How UN Women Turns Gender Data into Impact , year =

  77. [77]

    Gender and Health , year =

  78. [78]

    The Evolution of Gender Statistics , year =

  79. [79]

    Gender Data Portal , year =

  80. [80]

    Gender Data Hub , year =

Showing first 80 references.