[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-119266-en":3,"doc-seo-119266-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":4,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":27,"seo_description":14,"update_tm":28,"read_time":29},119266,7971461740886,"Theodore","https://ap-avatar.wpscdn.com/davatar_3d24733baf745e90a7e4bdd5f77d97b2",8,"Research & Report","Refining public policies with machine learning - The case of tax auditing","We study how machine learning techniques can improve tax auditing efficiency using administrative data without relying on randomized audits. Using Italy’s population data on non-incorporated small businesses’ tax returns and audit outcomes from 2007 to 2012, the approach addresses the problem that prediction models must be trained on data selected by human audit decisions. The results show substantial revenue potential: replacing the 10% least promising audits with algorithm-selected cases increases detected tax evasion by up to 39%, and evasion that is actually recovered by 29%.","Journal of Econometrics xxx (xxxx) xxx  \n| Refining public policies with machine learning: The case of tax auditing✩\u003Cbr>Marco Battaglini a, Luigi Guisob,∗, Chiara Lacava c, Douglas L. Miller a, Eleonora Patacchini a,b\u003Cbr>a Cornell University, United States of America b EIEF, Italy\u003Cbr>c Goethe University Frankfurt, Germany |  |\n| --- | --- |\n| A R T I C L E I N F O\u003Cbr>JEL classification: C55\u003Cbr>H26\u003Cbr>Keywords:\u003Cbr>Tax enforcement\u003Cbr>Tax evasion\u003Cbr>Policy prediction problems | A B S T R A C T |\n|  | We study how machine learning techniques can be used to improve tax auditing efficiency using administrative data without the need of randomized audits. Using Italy’s population data on sole proprietorship tax returns and audits, our new approach addresses the challenge that predictions must be trained on human-selected data. There are substantial margins for raising revenue from audits by improving the selection of taxpayers to audit with machine learning. Replacing the 10% least promising audits with an equal number selected by our algorithm raises detected tax evasion by as much as 39%, and evasion that is actually paid back by 29%. |\n\n1. Introduction  \nTax authorities routinely collect deep datasets from tax returns that can be used to identify audit targets. Consequently, the choice of auditing strategy is a prime candidate for applying machine learning techniques (henceforth, ML). The promise of these techniques is that they can be deployed to exploit available information efficiently, consistently and transparently. While both tax authorities and researchers are aware of these opportunities, the opacity of the audit selection processes followed by most tax authorities makes it unclear the extent to which they operate at the ‘‘production possibility frontier’’ or whether there are margins for improvements by a more efficient use of data.  \nIn this paper, we exploit a novel dataset from the Italian Revenue Agency (henceforth, IRA) to explore whether ML techniques can be used to improve audit selection policies. The dataset includes tax returns from the universe of non-incorporated small businesses in Italy from 2007 to 2012. For these tax returns, we know whether or not they were audited, and the results of the audit, including information on whether the taxpayer appealed against the audit as well as all the statistical information available to the IRA concerning the tax return and filer.  \nThe general idea behind ML techniques is to exploit data on realized outcomes to train a predictive algorithm. In our setting, the data are on audits that have occurred, and the outcome is, for example, detected tax evasion. Ideally, after validation procedures, the algorithm can be used to guide future policy (in our case, the choice of which returns to audit).  \n✩ We are very grateful to the Italian Revenue Agency for granting us access to the data. We are solely responsible for the ideas expressed in the paper. We thank Franco Peracchi, Edoardo Di Porto, Matteo Paradisi, and seminar and conference participants at the Italian Presidency of the Council of Ministers, Cornell University, ETH Zurich, University of Cambridge, Goethe University Frankfurt, the EUTO/IEB Workshop on the Economics on Taxation and the BSE Summer Forum in Public Economics for valuable discussions. This research was supported by a Cornell Center for Social Sciences Grant, United States of America.  \n∗ Correspondence to: Via Sallustiana, 62, 00187, Rome, Italy.  \nE-mail addresses: [mb2457@cornell.edu](mb2457@cornell.edu) (M. Battaglini), [luigi.guiso@eief.it](luigi.guiso@eief.it) (L. Guiso), [lacava@econ.uni-frankfurt.de](lacava@econ.uni-frankfurt.de) (C. Lacava), [dlm336@cornell.edu](dlm336@cornell.edu)[ ](dlm336@cornell.edu)(D.L. Miller), [ep454@cornell.edu](ep454@cornell.edu) (E. Patacchini).  \n[https://doi.org/10.1016/j.jeconom.2024.105847](https://doi.org/10.1016/j.jeconom.2024.105847)  \nReceived 22 April 2023; Received in revised form 25 May 2024; Accepted 2 August","cbCais4JRjRT8BfC","https://ap.wps.com/l/cbCais4JRjRT8BfC","pdf",1076065,1,16,"English","en",105,"# Introduction\n## Selective labels and omitted payoff bias\n## Data and audit selection objective\n## Identifying poorly performing audits","[{\"question\":\"How does the paper use machine learning to improve tax auditing efficiency?\",\"answer\":\"It trains predictive algorithms on realized audit outcomes from administrative data and uses them to guide future audit selection decisions.\"},{\"question\":\"What data source and period does the analysis rely on?\",\"answer\":\"The study uses Italy’s administrative population data on tax returns and audits for non-incorporated small businesses covering 2007 to 2012.\"},{\"question\":\"What is the main impact of replacing the 10% least promising audits?\",\"answer\":\"Replacing them with audits selected by the algorithm raises detected tax evasion by up to 39% and recovered evasion by up to 29%.\"}]","Refining public policies with machine learning - The case of tax auditing | PDF",1785723404,40,{"code":4,"msg":31,"data":32},"ok",{"site_id":24,"language":23,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"refining-public-policies-with-machine-learning-the-case-of-tax-auditing","",{"@graph":36,"@context":85},[37,54,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/refining-public-policies-with-machine-learning-the-case-of-tax-auditing/119266/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":23,"description":14,"dateModified":62,"datePublished":62,"encodingFormat":61,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-08-03",true,{"@type":65,"interactionType":66,"userInteractionCount":4},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"How does the paper use machine learning to improve tax auditing efficiency?","Question",{"text":75,"@type":76},"It trains predictive algorithms on realized audit outcomes from administrative data and uses them to guide future audit selection decisions.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What data source and period does the analysis rely on?",{"text":80,"@type":76},"The study uses Italy’s administrative population data on tax returns and audits for non-incorporated small businesses covering 2007 to 2012.",{"name":82,"@type":73,"acceptedAnswer":83},"What is the main impact of replacing the 10% least promising audits?",{"text":84,"@type":76},"Replacing them with audits selected by the algorithm raises detected tax evasion by up to 39% and recovered evasion by up to 29%.","https://schema.org",{"og:url":52,"og:type":87,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":89,"canonical":52},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,119,122,127,130,134],{"id":20,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":29,"slug":118},7,"Healthcare","healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":120,"slug":121},30,"research-report",{"id":123,"doc_module":4,"doc_module_name":46,"category_name":124,"show_sort_weight":125,"slug":126},9,"Religion & Spirituality",20,"religion-spirituality",{"id":125,"doc_module":4,"doc_module_name":46,"category_name":128,"show_sort_weight":125,"slug":129},"World Cup","world-cup",{"id":131,"doc_module":4,"doc_module_name":46,"category_name":132,"show_sort_weight":131,"slug":133},10,"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":46,"category_name":136,"show_sort_weight":106,"slug":137},19,"General","general"]