impact factor in research article

University of Illinois Chicago

University library, search uic library collections.

Find items in UIC Library collections, including books, articles, databases and more.

Advanced Search

Search UIC Library Website

Find items on the UIC Library website, including research guides, help articles, events and website pages.

• Search Collections
• Search Website

Measuring Your Impact: Impact Factor, Citation Analysis, and other Metrics: Journal Impact Factor (IF)

• Measuring Your Impact
• Citation Analysis
• Find Your H-Index
• Other Metrics/ Altmetrics
• Journal Impact Factor (IF)
• Selecting Publication Venues

About Journal Impact

Impact Factor - What is it?;  Why use it?

The  impact factor (IF)  is a measure of the frequency with which the average article in a journal has been cited in a particular year. It is used to measure the importance or rank of a journal by calculating the times its articles are cited.

How Impact Factor is Calculated?

The calculation is based on a two-year period and involves dividing the number of times articles were cited by the number of articles that are citable.

Calculation of 2010 IF of a journal:

Reliability of the Impact Factor

• Seglen, P. O. (1997). Why the impact factor of journals should not be used for evaluating research . British Medical Journal, 314(7079), 498-502.
• Johnstone, M. J. (2007). Journal impact factors: Implications for the nursing profession . International Nursing Review 54(1), 35-40.
• Ironside, P. M. (2007). Advancing the science of nursing education: Rethinking the meaning and significance of journal impact factors . Journal of Nursing Education, 46(3), 99-100.
• Satyanarayana, K. & Sharma, A. (2008). Impact factor: Time to move on . The Indian Journal of Medical Research, 127(1), 4-6.
• Greenwood, D. C. (2007). Reliability of journal impact factor rankings . BMC Medical Research Methodology, 7(48), 48.
• Howard, J. (2009). Humanities journals confront identity crisis. The Chronicle of Higher Education, 55(19), A1.

Tools to Measure Journal Impact (Impact Factor)

Journal Citation Reports ( Learn more )

SJR, CiteScore, SNIP through Scopus ( learn more )

SCImago Journal Rank (SJR)  ( Learn more )

 SNIP   (Learn more)

Journal Citation Reports

Journal Citation Reports provides ranking for journals in the areas of science, technology, and social sciences. For every journal covered, the following information is collected or calculated: Citation and article counts, Impact factor, Immediacy index, Cited half-life, citing half-life, Source data listing, Citing journal listing, Cited journal listing, Subject categories, Publisher information.

• Limited to the citation data of Journals indexed in Web of Science
• Process to determine journals included in the tool 
• Indexes over 12,000 journals in arts, humanities, sciences, and social sciences

You can enter a journal title in the Search box under "Go to Journal Profile".   Because impact factors mean little on their own, it's best to view the journal you are interested in comparison to the other journals in the same category.  To determine the impact factor for a particular journal, select a JCR edition (Science and/ or Social Science), year, and Categories, found on the left of the screen. Click Submit .  Scroll the list to find the journal you are interested in.  The list can be resorted by Journal time, Cites, Impact Factor, and Eigenfactor.

Scopus  (Elsevier)

Scopus provides three journal metrics - CiteScore, SJR (SCImago Journal Rank) and SNIP (Source Normalized Impact per Paper).  Once you are in Scopus, click on "Sources" at the top to access the journal impact data.   See below for more on SJR and SNIP

Over 22,000 active journals from over 4,000 international publishers

• Process to determine  journals included in the tools

SCImago Journal Rank (SJR)    

“The SCImago Journal & Country Rank is a portal that includes the journals and country scientific indicators developed from the information contained in the Scopus® database (Elsevier B.V.).” Scopus contains more than 15,000 journals from over 4,000 international publishers as well as over 1000 open access journals.  SCImago's "evaluation of scholarly journals is to assign weights to bibliographic citations based on the importance of the journals that issued them, so that citations issued by more important journals will be more valuable than those issued by less important ones." ( SJR indicator )

SNIP (Source Normalized Impact per Paper) 

Source Normalized Impact per Paper (SNIP) measures contextual citation impact by weighting citations based on the total number of citations in a subject field. The impact of a single citation is given higher value in subject areas where citations are less likely, and vice versa.  Unlike the well-known journal impact factor, SNIP corrects for differences in citation practices between scientific fields, thereby allowing for more accurate between-field comparisons of citation impact. CWTS Journal Indicators also provides stability intervals that indicate the reliability of the SNIP value of a journal.  SNIP was created by Professor Henk F. Moed at Centre for Science and Technology Studies (CWTS), University of L

CWTS Journal Indicators currently provides four indicators:

• P. The number of publications of a source in the past three years.
• IPP. The impact per publication, calculated as the number of citations given in the present year to publications in the past three years divided by the total number of publications in the past three years. IPP is fairly similar to the well-known journal impact factor. Like the journal impact factor, IPP does not correct for differences in citation practices between scientific fields. IPP was previously known as RIP (raw impact per publication).
• SNIP. The source normalized impact per publication, calculated as the number of citations given in the present year to publications in the past three years divided by the total number of publications in the past three years. The difference with IPP is that in the case of SNIP citations are normalized in order to correct for differences in citation practices between scientific fields. Essentially, the longer the reference list of a citing publication, the lower the value of a citation originating from that publication. A detailed explanation is offered in our scientific paper.
• % self cit. The percentage of self citations of a source, calculated as the percentage of all citations given in the present year to publications in the past three years that originate from the source itself.

See more at:  https://www.journalindicators.com/methodology

• << Previous: Other Metrics/ Altmetrics
• Next: Selecting Publication Venues >>
• Last Updated: Mar 13, 2024 12:51 PM
• URL: https://researchguides.uic.edu/if

Impact Factors

• Journal Impact Factor
• Author Impact Factor
• Article Impact
• Document Your Research Impact

Other Perspectives on Impact Factors

The San Francisco Declaration on Research Assessment (DORA) recognizes that impact factors are sometimes misused.

It begins "There is a pressing need to improve the ways in which the output of scientific research is evaluated by funding agencies, academic institutions, and other parties."

An Impact Factor is a quantitative measure of the relative importance of a journal, individual article or scientist to science and social science literature and research.

Each index or database used to create an impact factor uses a different methodology and produces slightly different results.  This is why it is important to use several sources to gauge the true impact of a journal's or scientist’s work.

This guide includes information on Journal Impact Factor , Author Impact Factor , Article Impact Factor , and Documenting Your Research Impact .

Informed and careful use of these impact data is essential, and should be based on a thorough understanding of the methodology used to generate impact factors. There are controversial aspects of using impact factors:

• It is not clear whether the number of times a paper is cited measures its actual quality.
• Some databases that calculate impact factors fail to incorporate publications including textbooks, handbooks and reference books.
• Certain disciplines have low numbers of journals and usage. Therefore, one should only compare journals or researchers within the same discipline.
• Review articles normally are cited more often and therefore can skew results.
• Self-citing may also skew results.
• Some resources used to calculate impact factors have inadequate international coverage.
• Editorial policies can artificially inflate an impact factor.

Please contact  your library liaison , or submit your question using the HSL comment box .

• Next: Journal Impact Factor >>
• Last Updated: Feb 6, 2024 10:31 AM
• URL: https://guides.lib.uw.edu/hsl/impactfactors

Be boundless

1959 NE Pacific Street | T334 Health Sciences Building | Box 357155 | Seattle, WA 98195-7155 | 206-543-3390

© 2024 University of Washington | Seattle, WA

University Library, University of Illinois at Urbana-Champaign

Introduction to Impact Factor and Other Research Metrics

• Types of Metrics

Impact Factor

• Identifying Journals
• More Resources

More Information

• About Journal Impact Factor Visit this article published by Thomson Reuters on journal impact factor to learn more about the bibliometric and how it measures importance.
• Annual Reviews Rankings in JCR Visit this page to see how Annual Review Journals currently rank in Journal Citation Reports.
• SCImago Journal and Country Rank The SCImago Journal & Country Rank is a publicly available portal that includes the journals and country scientific indicators developed from the information contained in the Scopus database.

Ask a Librarian

Impact factor,  or Journal Impact Factor,  is a measure of the frequency with which the "average article" published in a given scholarly journal has been cited in a particular year or period and is often used to measure or describe the importance of a particular journal to its field. Impact factor was originally developed by Eugene Garfield, the founder of Institute of Scientific Information, which is now a part of Clarivate Analytics. Journal Impact Factor can be found in the  Journal Citation Reports or the JCR, as it's commonly known. Over the years various organizations have been created similar journal-level metrics, such as  SCImago Journal & Country Rank .

This page describes how to find impact factor in Journal Citation Reports .

Journal Citation Reports

Clarivate Analytics (formerly Institute for Scientific Information (ISI)) ranks, evaluates, and compares journals within subject categories and publishes the results in Journal Citation Reports . Journal Citation Reports provides ranking for journals in science, technology, and the social sciences. For every journal, the database collects and/or calculates information such as:

• citation/article counts
• impact factor
• immediacy index
• cited half-life
• citing half-life
• source data listing
• citing journal listing
• cited journal listing
• publisher information
• subject categories

Find Journal Citation Reports (JCR)

Follow the instructions below to find the Journal Citation Reports using the Library's resources.

• Begin at the Library homepage .
• Click on the orange tab that says "Find Materials," then scroll down until you see a laptop icon with the words "Databases by Subject and A-Z"; click on the icon.
• Type journal citations reports in the search box on the left side of the screen and then click on the magnifying glass to search that title.
• Your result will say "Journal Citation Reports"; click on it. It might ask you to provide your Net ID and password if you are off campus.

Find the Impact Factor

• Once in the database you either search by journal title (if you know which journal you want to see) or browse by category, which will let you view journals by JIF by discipline.
• On the left side you can choose search criteria, like impact factor range, year, and if the journal is open access.
• It is important to choose the right edition based on your subject area, as you won't be able to see specific journals if you choose the wrong one. Once you have finished selecting what to search, click Submit.
• You can't access impact factors from last year because the calculations only happen every two years (i.e. if the current year is 2021 the farthest you can go back is 2020). Most people choose the most current year they can access.
• Journals limited by the subject area, publisher, or geographic region.
• View all journals in order to browse.
• Search for a specific journal if you already know its title

Once you find a journal, the JCR gives you information about the journal, including the journal's abbreviations, how often it is published each year, the publisher, and the ISSN. 

Controversy

Many people have questioned the legitimacy of impact factor. Here are a few reasons why:

• Impact factor focuses purely on the numbers. There is no consideration of qualitative elements that have become important in today's world.
• Impact factor fails to incorporate more recent ways of sharing and using research, including Twitter mentions and posts, citation management downloads, and news and community information.
• Because impact factor is based on citations in only indexed journals , it fails to incorporate statistics from journals that might not be indexed and other sources like conference papers (which are important in the social sciences).
• Basic or summary information is usually cited the most in academia. That means that journals that publish articles with basic information are more likely to have higher impact factors. Journals that publish obscure or innovative information might not have as high of an impact factor.
• Some argue that impact factor is encouraging scholars to stick with mainstream topics and research.
• Scholars don't always have to cite something for it to be influential. Sometimes researchers just read something and it influences them, regardless of if they cite it in a future paper or piece of research.
• The journals in the JCR are mostly published in English. This means that many international sources aren't included in the conversation.
• It has been argued that journals have the ability to skew impact factor for their own journal. Before publishing an author, they will ask the author to cite more articles within their journal so that their impact factor goes up. This is NOT a common occurrence but instead something we should be aware of.
• << Previous: Types of Metrics
• Next: Identifying Journals >>
• Last Updated: Feb 28, 2024 12:49 PM
• URL: https://guides.library.illinois.edu/impact

Impact factors: What they are, where to find them, how to use them

• Introduction to impact factors
• Which UC Merced databases include impact factors?
• Video tutorials on impact factors
• Need additional help?

Instruction & Outreach Librarian

More on Citation Metrics

Wikipedia has an excellent collection of articles on various types and aspects of citation metrics, including impact factors, H-indexing, citation analysis and more. Click the image below if you're interested in learning more.

This guide was created in support of the Fall 2022 UC Merced Library workshop, "Impact factors: What they are, where to find them, and how to use them." Here, we'll discuss impact factors with emphasis on their use in writing theses and dissertations for degree credit at UC Merced.

What is an impact factor?

The impact factor measures the number of times a journal article has been cited by researchers in a given year. It's used to measure the importance of a scholarly journal -- that is, its importance to the discipline or field its articles cover, and by extension, the researchers working in that discipline or field -- by measuring the number of times articles in that journal are cited.

• Generally, impact factors are the best way to determine a journal's relative importance in a particular field or discipline. Your own research will be more readily accepted if it's based on the top journals -- meaning, the journals with the highest impact factors -- in your field.
• Impact factors are not perfect, and can be gamed, so to speak. Many journals will attempt to increase their impact factors by requiring that authors whose work is accepted for publication include citations to articles published in those journals.

Creating the impact factor

• Journal impact factors are calculated on the total number of citable articles in the two most recent, previous years. So it's not possible to get a journal impact factor for the present year. In 2022, the most recent journal impact factors will have been calculated on 2020 and 2021.
• Because journal impact factors are calculated on two years of article citations, it's not possible to calculate an impact factor for new journals.
• An impact factor of 10 is an excellent impact factor and indicates that the journal is of major importance in a field or discipline.
• An impact factor of 3 is considered to be good.
• Average impact factors for most journals are less than 1. However, this doesn't indicate that a journal is of poor quality. It may be a journal that publishes research in a field that is not noted for research.
• Next: Which UC Merced databases include impact factors? >>
• Last Updated: Oct 20, 2022 10:28 AM
• URL: https://libguides.ucmerced.edu/impact-factors

Find Journal Information & Scholarly Metrics: Introduction to Impact Factors

• Introduction to Impact Factors
• ISI Web of Knowledge
• Google Scholar: Metrics and Self-Citation Counts
• Journal Directories
• Eigenfactor

Journal Impact Factors Introduction

What are impact factors.

The impact factor is a citation measure produced by Thompson Scientific's ISI Web of Knowledge database. Impact factors are published annually in ISI's Journal Citation Reports database. Impact factors are only available for journals that are indexed in ISI databases. 

One journal's impact factor on its own doesn't mean much. Instead, it's important to look at impact factors of multiple journals in the same subject area. This way, one can determine if the impact factor of the journal of interest is high or low compared to other journals in a subject area.

Impact Factor Debate

Impact factors have been much debated in the literature in terms of their value for evaluating research quality. The general consensus is that impact factors have been misunderstood and abused by many institutions that place too much value on something that is not entirely scientific or reliable. Please refer to the 'Factors that Influence Impact Factors' and 'Additional Readings' sections to find out more.

How Impact Factors are Calculated

A journal's impact factor for 2007 would be calculated by taking the number of citations in 2007 to articles that were published in 2006 and 2005 and dividing that number by the total number of articles published in that same journal in 2006 and 2005. Please see the example below.

Example: The specific calculations for  Nursing Research 's 2007 impact factor are displayed below. Articles published in 2006 that were cited in 2007: 98 Articles published in 2005 that were cited in 2007: 103 98+103=201  Total Number of articles published in 2006: 67 Total number of articles published in 2005: 48 67+48=115 201 (articles published in 2006 and 2005 that were cited in 2007) 115 (total number of articles published in 2006 and 2005) = 1.748 The 2007 Impact Factor for the journal  Nursing Research  means that, on average, articles published in this journal from one or two years ago have been cited around 1 and three-quarter times.

Factors that Influence Impact Factors 

Date of publication.

The impact factor is based solely on citation data and only looks at the citation frequency of articles from a journal in its first couple years of publication. Journals with articles that are steadily cited for a long period of time (say, 10 years) rather than only immediately lose out with this calculation.

Large vs. Small Journals

Large and small journals are compared equally. Large journals tend to have higher impact factors--nothing to do with their quality.

Average Citation

It’s important to remember that the impact factor only looks at an average citation and that a journal may have a few highly cited papers that greatly increase its impact factor, while other papers in that same journal may not be cited at all. Therefore, there is no direct correlation between an individual article’s citation frequency or quality and the journal impact factor.

Review Articles 

Impact factors are calculated using citations not only from research articles but also review articles (which tend to receive more citations), editorials, letters, meeting abstracts, and notes. The inclusion of these publications provides the opportunity for editors and publishers to manipulate the ratio used to calculate impact factor and falsely try to increase their number.

Changing / Growing Fields 

Rapidly changing and growing fields (e.g. biochemistry and molecular biology) have much higher immediate citation rates, so those journals will always have higher impact factors than nursing, for instance.

ISI's Indexing / Citation Focus

There is unequal depth of coverage in different disciplines. In the health sciences, the Institute for Scientific Information (ISI), the company which publishes impact factors, has focused much of their attention on indexing and citation data from journals in clinical medicine and biomedical research and has not focused on nursing as much. Very few nursing journals are included in their calculations (around 45). This does not mean that nursing journals they do not include are of lesser quality, and, in fact, they do not give any explanation for why some journals are included and others are not. In general, ISI focuses more heavily on journal dependent disciplines in the sciences and provides less coverage for areas of the social sciences and humanities, where books and other publishing formats are still common.

Research vs. Clinical Journals

In some disciplines such as some areas of clinical medicine where there is not a distinct separation between clinical/practitioner versus research journals, research journals tend to have higher citation rates. This may also apply to nursing.

Compiled by Heidi Schroeder, Michigan State University (used with permission)

Measure Research Impact

Researchers are often asked to demonstrate the impact of their research. Typically this means showing evidence that your work is being used / built-on / responded-to by other researchers in your field. The simplest approach to this is by listing each of your publications and then counting the number of times each work is cited. Beyond that, many departments and disciplines rely on more complicated approaches to measure impact.

A typical approach considers the quality of publication venues as well as some measure of how often your work is cited. There are also approaches that try to measure the impact of research beyond the sphere of journals; this alternative approach to metrics is sometimes called “altmetrics.”

The use of any particular measure and the importance of those measures depends on the standards within a given academic discipline and even more specifically within a particular department. You should consult your Departmental Collegial Review Document and other relevant documentation to see if and how your department measures impact.

Citation counts and metrics

The most rudimentary form of measuring impact is to count the citations for each of your publications. Since checking every subsequent article in your field to see if your publication is mentioned isn’t a practical or sustainable approach, most scholars lean on an indexing database or service to find citation counts for a given article or author. Web of Science, Scopus, and Google Scholar are the three most used general databases. Western Carolina pays for access to Web of Science and Scopus, while Google Scholar is free to the public. There are also some smaller databases that cater to specific disciplines. SciFinder is one such database to which Western Carolina has access and its focus is limited to journals in the fields of biology, chemistry, physics, and medicine.

Web of Science and Scopus have fairly straightforward search engines for tracking the citations of particular articles and authors. Google Scholar works similarly for articles, but for citation counts for authors, they use a more robust, but also more involved Scholar Profile which requires a little bit of set-up.

• Web of Science This link opens in a new window
• Google Scholar Citations

There are a few other metrics based on citation count that are popularly used and are often also calculated by these services. The most popular of these is h-index, a measure of an author’s scholarly output and impact where h represents how many publications an author has that have been cited h times. An author with an h-index of 5 has five publication that have each been cited at least five times, while an h-index of 12 would signify twelve publications each cited at least twelve times. This measure was proposed by J.E. Hirsch in 2005 and has enjoyed widespread popularity. 

There have been other proposed summary measurements of varying complexity. The g-index is a proposed improvement on the h-index, but there have been many other attempts to build on or improve what the h-index does. Anne-Wil Harzing discusses some of these variants and their various strengths and weaknesses on their website. Ultimately, the point of these measures is to demonstrate the impact of the researcher’s work and their usefulness to you will be largely dependent on your specific department’s requirements.

Publish or Perish is a software application developed by Anne-Wil Harzing specifically for researchers looking to demonstrate the research impact of their work. It is based on top of the Google Scholar database, but provides a way to look at and calculate some of these other metrics.

Measuring journal quality

There are several popular methods of measuring journal quality. Although many of these measures can be calculated independently, they are often associated with a particular commercial database or service that indexes journal articles.

CiteScore and Scimago Journal Rank

CiteScore is a number produced from the indexing done by Elsevier’s Scopus database. It is defined as the measure of the ratio of citations to documents in a journal over four years divided by the total number of citable documents that journal published over that same period. This measure (and rankings derived from this measure) are freely available on the Scopus website. 

The Scimago Journal Rank (SJR) indicator uses the same set of Scopus-indexed articles, but uses a process based on Google’s PageRank algorithm. Essentially, the entire interconnected network of citations is analyzed so that journals that are cited often impart more prestige when they cite an article than the same citation in a less-cited journal. Citations in Nature would count more than citations in Bob’s Funtime Quarterly, an imaginary and presumably much less prestigious publication. SJR indicators for journals are freely available on the Scimago Journal Rank website.

Impact Factor and its variants

In its most simple formulation, impact factor is calculated as the ratio of number of citations received by that journal in a given year to the total number of citable items in the journal over the most recent two years. While this formula is platform agnostic, the term “impact factor” and “journal impact factor” are most closely associated with the calculations done using the articles indexed on the Clarivate Web of Science platform and then reported in the InCites and Journal Citation Reports products. Both of these are proprietary products that Western Carolina does not currently subscribe to.

Eigenfactor is a proprietary measure that attempts to account for the differences between a citation in a highly-cited journal and one in a less-cited journal. If this sounds a lot like what the SJR indicator does, it’s because the two measures share the same approach. For a further discussion of the differences between SJR indicator and Eigenfactor there is a solid explanation and breakdown on the Society for Scholarly Publishing’s blog. The source data used to calculate Eigenfactors is again from the Clarviate platform, but for a time, Eigenfactors were made publicly available on Eigenfactor.org. However, it looks like any values from 2016 or later are not available.

Google Scholar takes a different approach to measuring journal quality, adapting the author metric h-index and applying it to journals. Using h-index (and a number of its variants), Google Scholar ranks journals and also offers rankings for particular fields and areas of expertise, using the same suite of h-index measurements. 

• What is the Eigenfactor of a Specific Journal
• Google Scholar Metrics

“Altmetrics” is an expansive term that encompasses approaches to measuring impact that don’t map onto the more traditionally used approach of citation counts in peer-reviewed journals. The approach is an acknowledgment of the basic truth that actual scholarly impact isn’t limited to this fairly restricted dataset. Often hinging on uses of publicly available data on the internet, many altmetrics track mentions and discussion in other publications beyond peer-reviewed academic journals such as in newspapers, blogs, and even on social media. Some approaches try to quantify other measures of engagement beyond mentions (e.g. times an article is viewed, downloaded, and shared).

Our Research has several websites that can give you a taste of two approaches to altmetrics. Impactstory is the better known and is notable for its integration with ORCID, while Paperbuzz is a newer project powered by using the Crossref API. The websites detail the methodology used by each of these projects, but the purpose is the same as more well-known and traditional methods: to demonstrate the impact of your research.

Altmetrics, by nature, is an ever-evolving and malleable approach. But like any of the more traditional measures, its purpose is to help demonstrate and quantify scholarly impact. If it’s useful for showing the reach of your work, it’s worth further investigation.

• << Previous: Home
• Next: ISI Web of Knowledge >>
• Last Updated: Aug 7, 2023 11:17 AM
• URL: https://researchguides.wcu.edu/scholarlymetrics

HUNTER LIBRARY

176 Central Drive Cullowhee, NC 28723 Administration: 828-227-7485 Reference: 828-227-7465 Circulation: 828-227-7485

QUICK LINKS

Ask-A-Librarian Reserve a Study Room My Account Library Catalog Article Databases Interlibrary Loan

Evaluating Information Sources

• Evaluate Your Sources
• Publication Types and Bias
• Reading Scholarly Articles

Journal Impact Factors

Links summary.

• Author Impact / Citations
• Author H-index
• Author h-index Options
• Author Citation Reports in Web of Science
• What are metrics?
• Cited Articles = Confusing Statistics
• Predatory Publishing

According to Journal Citation Reports (JCR) , an  impact factor is a ratio focusing on original research. 

Impact factor = # of citations to all items published in that journal in the past two years (divided by) # of articles and reviews published over those past two years referencing those citations

For example, if a journal has an impact factor of 2.5, this means in the indexed year each article published was cited on average 2.5 times in the previous two years in that journal.

Impact factor is used for journals only.

JCR only includes  12,000 journals and conference proceedings from over 3,300 publishers.

• InCitesTM: Journal Citation Reports® This Web of Science hosted database is a citation-based research evaluation tool for journal performance metrics with the goal of offering a systematic and objective means to evaluate the journals based on citation data.
• Tips for Using JCR Tips for using the Web of Science InCitesTM Journal Citation Reports

Metrics on the Web

• Eigenfactor® Project
• Google Scholar - Metrics
• InCitesTM: Journal Citation Reports®
• PlumX Metrics

Research Guides 

• Impact Metrics and Scholarly Attribution (UCLA)
• Research Impact Metrics (UM)
• Spreading the Word: Publishing Your Research & Extending Your Impact (USC)
• << Previous: Reading Scholarly Articles
• Next: Predatory Publishing >>
• Last Updated: Mar 8, 2024 1:17 PM
• URL: https://libguides.usc.edu/evaluate

• Harvard Library
• Research Guides
• HKS Library & Research Services

Measuring Your Scholarly Impact

Author & article impact, journal rankings & evaluation, visualization software, statements on responsible research metrics.

Article & Author Impact

• Web of Science (Harvard Login) A multidisciplinary database, with searchable author abstracts, covering the journal literature of the sciences, social sciences and arts. Create personal account to generate citation reports by author and create citation maps for articles.
• Google Scholar Citations Track citations to your publications. Determine which authors are citing to your publications. Graph your citations over time. Note: You must register for a Google account using an academic email account. Authors of scholarly articles should claim their Google Scholar page to verify that your publications listings are accurate and complete. You can also create an author profile by following the instructions on the Google Scholar Citation page
• Publish or Perish A free author and journal impact metrics software program developed by Anne Wil Harzing that retrieves and analyzes citations to articles and books. The software uses Google Scholar to obtain the raw citations. The tool can be used to locate most cited articles and books by searching in general citations field.
• ORCID Register for an ORCID number. An ORCID is a unique id number that distinguishes you from every other researcher. This id is essential if you have a common last and first name as it distinguishes you from other scholars. Read more about ORCID at the Library's ORCID page. You can also use Harvard ORCID Connect to allow Harvard to access your ORCID more easily. This means internal scholarly and administrative systems can pull in your public ORCID data, saving you time, maintaining the accuracy and consistency of your data, and creating meaningful connections between systems. The id can be integrated into the research workflow such as manuscript and grant submissions and it supports automated linkages between you and your professional activities.

Journal Metrics

Journal metrics are used to identify key journals in a research field.  This identification may be most useful to authors who are considering which journals to submit manuscripts to for future publication.

The Impact Factor may be the most familiar metric in academics. Eugene Garfield of Thomson Scientific first introduced this idea in the 1950s. Impact Factor calculations are now available through Thomson’s Journal Citation Reports (JCR) and the Elsevier product, Scopus.

Despite their merits, journal metrics can be misused for evaluating individual authors. Altmetrics is an alternative for measuring scholarly impact. Altmetrics measures the use of social media tools such as bookmarks, links, blog postings, and tweets to gauge the importance of scholarly output by authors.  Using altmetrics as a measure of scholarly impact is controversial as social media tweets and mentions can be gamed by authors.

Journal Evaluation Resources

• Journal Citation Reports (Harvard Login) Database for journal evaluation, using citation data drawn from over 8,400 scholarly and technical journals worldwide in the sciences and social sciences. Coverage is both multidisciplinary and international, and incorporates journals from over 3,000 publishers in 60 nations.
• Google Scholar Metrics Defaults to top 100 publications in English, ordered by their five-year h-index and h-median metrics. Use the search box to search for individual journal titles. Compare the publications that are of interest to you. Explore publications by subject area by going to the left column, selecting your language, and picking a general search category. You can refine your results further by clicking on the subcategory link under each general subject category. more... less... For additional information about google inclusion and exclusion policies for metrics, go to https://scholar.google.com/intl/en/scholar/metrics.html#overview
• SJR (Scimago Journal & Country Rank) A free ranking tool for journals. Data from Elsevier product, Scopus. You can limit results by country and geographic region.
• SciRev Provides information on journal response times and review duration based on feedback from individuals. Heavy focus on science journals but includes some social science and business/economics journals.

Conducting Literature Reviews

Scholars will often publish journal articles that evaluate the top ranked journals in their discipline. Conduct a search in a large interdisciplinary database such as Proquest Social Sciences Premium Collection or Ebsco's Academic Search Premier using keywords such as "top journals" or "highly ranked journals" and the field. For example, you could search in Academic Search Premier using the terms "top ranked journals" and "economics". You can also select a narrower subject specific database such as EconLit .  Alternatively search across a number of different databases for full-text articles using the Google Scholar search option.

Defining Altmetrics

The term "altmetrics" (alternative metrics) is used to describe approaches to measure the impact of scholarship by using new social media tools such as bookmarks, links, blog postings, inclusion in citation management tools, mentions and tweets to measure the importance of scholarly output.

Proponents of altmetrics believe that using altmetrics will help measure the impact of an article in a more comperhensive and objective way than was done with more traditional scholarly impact measures such as journal impact factor.  However, there are limits to this approach and caution should be used to not rely on any one particular measure in evaluting the importance of scholarship.

• Altmetrics, A Manifesto Web site devoted to altmetrics. Started by group of librarians and researchers who are active in promoting altmetrics as an alternative to more traditional forms of tracking article impact.
• Impact Story ImpactStory aggregates altmetrics measures from articles, datasets, blog posts, and more.
• Altmetric (Free Tools) Private company that sells access to altmetrics products. Individual researchers can download a free bookmarklet to check altmetrics on individual articles.
• Open Syllabus Project Creators of site scraped college Web sites and have put together the metadata for over 1 million syllabi. New metric based on this project, the "Teaching Score" (TS), is a numerical indicator of the frequency with which a particular work is taught.
• Leiden Madtrics Leiden Madtrics is the official blog of the Centre for Science and Technology Studies (CWTS) at Leiden University. The blog looks specifically at the processes of evaluating research, developing research policy, and the myriad ways academic research makes an impact in society

Free Software for Visualizing Citations

• CitNetExplorer A product of CWTS of Leiden University, this tool allows citation networks to be imported directly from the Web of Science database and used to visualize and analyze citation networks of scientific publications.
• Gephi An open-source software for visualizing and analysing large networks graphs. Gephi uses a 3D render engine to display graphs in real-time and speed up the exploration. You can use it to explore, analyse, spatialise, filter, cluterize, manipulate and export all types of graphs.
• Sci2 Network visualization tool that allows for easy data import from standard comma-separated lists and generates network analytics as well as visualizations.
• VOSviewer Created by Leiden University's CWTS, a software tool for constructing and visualizing bibliometric networks. Offers text mining functionality that can be used to construct and visualize networks of important terms extracted from a body of scientific literature.
• Local Citation Network Construct and visualizes citation networks to identify the most influential papers in a given topic or field.
• Citation Gecko Start from a small set of 'seed papers' that define an area you are interested. Gecko will search the citation network for connected papers allowing you to quickly identify important papers that you may have missed.You can connect to your Zotero libraries.
• Leiden Manifesto for Research Metrics
• San Francisco Declaration on Research Assessment
• Last Updated: Aug 1, 2023 2:18 PM
• URL: https://guides.library.harvard.edu/hks/scholarlyimpact

Harvard University Digital Accessibility Policy

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings

Preview improvements coming to the PMC website in October 2024. Learn More or Try it out now .

• Advanced Search
• Journal List
• World J Nucl Med
• v.13(2); May-Aug 2014

Journal Impact Factor: Its Use, Significance and Limitations

Mohit sharma.

Department of Conservative Dentistry, Shree Bankey Bihari Dental College and Research Centre, Ghaziabad, India

Anurag Sarin

Priyanka gupta.

1 Department of Conservative Dentistry, Maharana Pratap College of Dentistry and Research Centre, Gwalior, India

Shobhit Sachdeva

2 Department of Pedodontics, ITS Dental College, Ghaziabad, India

Ankur V. Desai

3 Department of Conservative Dentistry, Vaidik Dental College and Research Centre, Daman, India

Dear Editor,

The impact factor (IF) is frequently used as an indicator of the importance of a journal to its field. It was first introduced by Eugene Garfield, the founder of the Institute for Scientific Information.[ 1 ] Although IF is widely used by institutions and clinicians, people have widespread misconception regarding the method for calculating the journal IF, its significance and how it can be utilized. The IF of a journal is not associated to the factors like quality of peer review process and quality of content of the journal, but is a measure that reflects the average number of citations to articles published in journals, books, thesis, project reports, newspapers, conference/seminar proceedings, documents published in internet, notes, and any other approved documents (by Indian Council of Medical Research or similar body).[ 2 ]

Impact factor is commonly used to evaluate the relative importance of a journal within its field and to measure the frequency with which the “average article” in a journal has been cited in a particular time period. Journal which publishes more review articles will get highest IFs. Journals with higher IFs believed to be more important than those with lower ones.[ 3 ] According to Eugene Garfield “impact simply reflects the ability of the journals and editors to attract the best paper available.”[ 4 ] Journal which publishes more review articles will get maximum IFs.

Impact factor can be calculated after completing the minimum of 3 years of publication; for that reason journal IF cannot be calculated for new journals. The journal with the highest IF is the one that published the most commonly cited articles over a 2-year period. The IF applies only to journals, not to individual articles or individual scientists unlike the “H-index.” The relative number of citations an individual article receives is better evaluated as “citation impact.” In a given year, the IF of a journal is the average number of citations received per article published in that journal during the 2 preceding years. IFs are calculated each year by Thomson scientific for those journals that it indexes, and are published in Journal Citation Reports ( http://www.thomsonreuters.com/products_services/science/science_products/a-z/journal_citation_reports/ ). For example, if a journal has an IF of 3 in 2008, then its papers published in 2006 and 2007 received three citations each on average in 2008. The 2008 IFs are actually published in 2009; they cannot be calculated until all of the 2008 publications have been processed by the indexing agency (Thomson Reuters). The IF for the biomedical journals may range up to 5-8%.[ 5 ] The IF of any journal may be calculated by the formula;

2012 impactfactor =A/B

Where A is the number of times articles published in 2010 and 2011 were cited by indexed journals during 2012. B is the total number of citable items like articles and reviews published by that journal in 2010 and 2011.

The calculation of IF for the journal where in a person has published articles is a contentious issue. Nevertheless, this have been already warned; “misuse in evaluating individuals” because there is “a wide variation from article to article within a single journal” therefore, “In an ideal world, evaluators would read each article and make personal judgments,” said by Eugene Garfield.[ 1 ]

• Journal Article Publishing Support Center

To post social content, you must have a display name. The page will refresh upon submission. Any pending input will be lost.

• Research & Preparation

Where do I find the Impact Factor of a journal?

The journal Impact Factor is an index that measures how often a journal's articles are cited in other research. This is calculated by the number of citations received by articles published in that journal during the two preceding years, divided by the total number of articles published in that journal during the two preceding years. You can find the journal Impact Factor on the journal homepage.

Follow these steps to find the Impact Factor of a journal:

• Go to the journal's homepage .
• When you're on the journal homepage, scroll down the page until you see the 'Journal Metrics' section on the left hand side. The journal Impact Factor is located under the 'Journal Metrics' section.

For more information on journal metrics, please visit the Measuring a journal's impact page from our website .

If there is no Impact Factor available for your journal, it's likely that it's new within the last 2 years and so the data aren't available yet

To determine if this is the case for your journal, you can find out how long a journal has been publishing for on ScienceDirect:

• Go to ScienceDirect.com
• Start typing the name of the journal into the search field ‘Journal/book title’, as you start typing titles will be suggested. When the title of your journal appears, click on it then click the magnifying glass to search.
• On the search results page, click the title of the journal.
• You are now on the journal’s publication page, where you can see its previous Volumes listed on the left-hand page. Check the dates against these volumes to determine the journal’s years of publication.

Was this answer helpful?

Thank you for your feedback, it will help us serve you better. If you require assistance, please scroll down and use one of the contact options to get in touch.

Help us to help you:

Thank you for your feedback!

• Why was this answer not helpful?
• It was hard to understand / follow.
• It did not answer my question.
• The solution did not work.
• There was a mistake in the answer.
• Feel free to leave any comments below: Please enter your feedback to submit this form

Related Articles:

• How can I find the right journal for my paper in Journal Finder?
• Is there a template available for my manuscript file?
• How do I find out if my journal complies with the policies of my funding body?
• Where can I find the submission site for a journal?
• Can you check if my manuscript is suitable for submission?

For further assistance:

ORIGINAL RESEARCH article

This article is part of the research topic.

Green Finance & Carbon Neutrality: Strategies and Policies for a Sustainable Future

Evaluation and influencing factors of green transformation in three major urban agglomerations in China Provisionally Accepted

• 1 Zhou Enlai School of Government, Nankai University, China

The final, formatted version of the article will be published soon.

Since the 21st century, the world has increasingly focused on the issue of sustainable development, and the green transformation issues have become a new hot topic worldwide. Urban agglomerations are important connections between urban development and regional coordination, as well as important spatial carriers for economic activities. This article focuses on 48 cities in the three most mature and influential urban agglomerations in China from 2011 to 2019, namely the Beijing-Tianjin-Hebei urban agglomeration, the Yangtze River Delta urban agglomeration, and the Pearl River Delta urban agglomeration. The three-stage DEA model and Malmquist index model are used to measure the green transformation efficiency of the three urban agglomerations from both dynamic and static perspectives, and a Tobit regression model is constructed to explore the influencing factors of green transformation efficiency in urban agglomerations. Research has found that: (1) From a static perspective, the overall efficiency of green transformation in the three major urban agglomerations is at a high level, but from a temporal perspective, it shows a downward trend. The Pearl River Delta urban agglomeration is known for its green development, with the highest average efficiency of green transformation, followed by the Yangtze River Delta urban agglomeration. The Beijing-Tianjin-Hebei urban agglomeration has the lowest level of green transformation; (2) From a dynamic perspective, technological progress is the main driving factor for improving the efficiency of green transformation in the three major urban agglomerations. Therefore, the government should pay special attention to the progressiveness of technology when formulating relevant policies to promote urban green transformation; (3) From the perspective of spatiotemporal differences, there are significant differences in the spatiotemporal characteristics of green transformation among the three major urban agglomerations, and there are significant differences in green transformation strategies among different urban agglomerations. Eliminating environmental factors and random interference is necessary for accurately measuring the efficiency of green transformation in urban agglomerations; (4) From the perspective of influencing factors, factors such as industrial structure upgrading, green innovation level, and environmental regulation intensity jointly affect the efficiency of green transformation in urban agglomerations.

Keywords: urban agglomerations, green transformation, efficiency measurement, Influencing factors, Low carbon development

Received: 27 Feb 2024; Accepted: 18 Apr 2024.

Copyright: © 2024 Ma and Lin. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY) . The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

* Correspondence: Dr. Zhihan Lin, Nankai University, Zhou Enlai School of Government, Tianjin, 300071, China

People also looked at

• Open access
• Published: 14 April 2024

The potential impact fraction of population weight reduction scenarios on non-communicable diseases in Belgium: application of the g-computation approach

• Ingrid Pelgrims 1 , 2 , 3 ,
• Brecht Devleesschauwer 3 , 4 ,
• Stefanie Vandevijvere 3 ,
• Eva M. De Clercq 1 ,
• Johan Van der Heyden 3 &
• Stijn Vansteelandt 2  

BMC Medical Research Methodology volume  24 , Article number:  87 ( 2024 ) Cite this article

191 Accesses

1 Altmetric

Metrics details

Overweight is a major risk factor for non-communicable diseases (NCDs) in Europe, affecting almost 60% of all adults. Tackling obesity is therefore a key long-term health challenge and is vital to reduce premature mortality from NCDs. Methodological challenges remain however, to provide actionable evidence on the potential health benefits of population weight reduction interventions. This study aims to use a g-computation approach to assess the impact of hypothetical weight reduction scenarios on NCDs in Belgium in a multi-exposure context.

Belgian health interview survey data (2008/2013/2018, n  = 27 536) were linked to environmental data at the residential address. A g-computation approach was used to evaluate the potential impact fraction (PIF) of population weight reduction scenarios on four NCDs: diabetes, hypertension, cardiovascular disease (CVD), and musculoskeletal (MSK) disease. Four scenarios were considered: 1) a distribution shift where, for each individual with overweight, a counterfactual weight was drawn from the distribution of individuals with a “normal” BMI 2) a one-unit reduction of the BMI of individuals with overweight, 3) a modification of the BMI of individuals with overweight based on a weight loss of 10%, 4) a reduction of the waist circumference (WC) to half of the height among all people with a WC:height ratio greater than 0.5. Regression models were adjusted for socio-demographic, lifestyle, and environmental factors.

The first scenario resulted in preventing a proportion of cases ranging from 32.3% for diabetes to 6% for MSK diseases. The second scenario prevented a proportion of cases ranging from 4.5% for diabetes to 0.8% for MSK diseases. The third scenario prevented a proportion of cases, ranging from 13.6% for diabetes to 2.4% for MSK diseases and the fourth scenario prevented a proportion of cases ranging from 36.4% for diabetes to 7.1% for MSK diseases.

Implementing weight reduction scenarios among individuals with excess weight could lead to a substantial and statistically significant decrease in the prevalence of diabetes, hypertension, cardiovascular disease (CVD), and musculoskeletal (MSK) diseases in Belgium. The g-computation approach to assess PIF of interventions represents a straightforward approach for drawing causal inferences from observational data while providing useful information for policy makers.

Peer Review reports

By affecting almost 60% of adults and nearly one in three children in the European Region, excess body weight is the fourth most common risk factor for NCDs, after high blood pressure, dietary risks, and tobacco use [ 1 ]. In Belgium, as in many high-income countries, average body mass index (BMI) has continuously increased over the past decades among both children and adults [ 2 ]. According to the most recent Belgian Health Interview Survey (BHIS) conducted in 2018, 48% of the adult population suffered from overweight (BMI > 25) and 14% from obesity (BMI > 30), compared to respectively 41% and 11% in 1997 [ 3 ]. Tackling obesity is therefore one of the greatest long-term health challenges in Belgium, such as in other countries, and is vital to successfully achieve the Sustainable Development Goals with regards to the reduction of premature mortality from non-communicable diseases [ 4 ].

Assessing the contribution of excess weight status as risk factor for NCDs and evaluating the potential health impact of policies for the prevention of overweight presents certain challenges and methodological issues, especially when using observational and cross-sectional data. To capture the association between a risk factor exposure and a health outcome, typical approaches in epidemiological studies use linear or logistic regression models, which estimate the differences between outcomes associated with a change in the risk factor exposure. This approach, relying on stratum-specific estimates, is however limited because it is not informative on how the burden of disease might change by modifying the risk factor exposure in the population. Furthermore, in the case of logistic regression models, interpretation of the obtained odds ratio is subtle because of non-collapsibility: it tends to move further away from 1 when adjusting for more and more variables, even in the absence of confounding [ 5 ]. The population attributable fraction (PAF), a concept introduced by Levin [ 6 ], is a measure widely used by epidemiologists to estimate the proportion of a disease attributable to a risk factor in a given population [ 7 ]. The PAF is typically calculated using the relative risk and the prevalence of the risk factor in the population and is often interpreted as the proportion of cases in the population avoidable if a particular risk factor was eliminated. The PAF based on Levin’s formula [ 6 ] was originally unadjusted for co-existing (risk) factors but methods such as adjusted and average attributable fraction (AAF) or attribution methods have been developed since to account for multi-causal situations (i.e. when a given disease is caused by more than one causal mechanism) [ 8 , 9 , 10 , 11 ]. However, for the PAF/AAF to have a valid causal interpretation, strong assumptions are required. This is because the excess cases seen in people with overweight need not all be “attributable” to overweight: they may not all be overweight-induced but rather the effect of other risk factors prevalent in those people [ 12 , 13 ]. Unfortunately, those assumptions are often disregarded or misreported in articles [ 12 ]. In addition, the PAF assumes that there is an optimal intervention which completely eradicates the risk factor in the population which is often unrealistic because a part of the population will often continue to be exposed to the risk factor, even with the most effective intervention. The potential impact fraction (PIF), also called the generalized impact fraction, is another measure that allows to estimate the fractional reduction of cases that would occur from changing the current level of exposure in the population to some modified level [ 14 ]. The PAF and the PIF, both affected by the strength of the association between the disease and the risk factor as well as the prevalence of the risk factor, estimate the disease risk in the population in case of “complete withdrawal” and “partial reduction” of the exposure [ 15 , 16 ]. The application of the traditional PAF or PIF for policymaking in this context is strongly limited by the rigors of complete elimination of the risk factor as well as the disadvantages of traditional methods based on standard regression models [ 7 , 17 ].

To overcome those limitations, the use of causal inference methods has been suggested by several authors [ 18 , 19 , 20 , 21 , 22 , 23 ]. In particular, the g–computation approach (a model-based direct standardization) has the advantage that it can handle continuous risk factors and predict the causal impact of public health interventions on the population burden of disease, using cross-sectional data [ 18 , 24 , 25 ]. Unlike traditional regression models, the method allows the estimation of population parameters, where the population average causal effect is estimated as the difference in the health outcome that would have been observed in the population if there had been a specific intervention as opposed to no intervention (everything else remained equal). Those population intervention parameters allow determining which hypothetical intervention may have the greatest impact on the disease. The method requires to clearly specify the causal effect of interest and to explain all assumptions needed to identify this effect from the available data. This can be achieved using a directed acyclic graph (DAG) which is a graphical representation used to illustrate the hypothesized causal structure of the processes under study [ 23 , 26 ]. Compared to standard analytic techniques, the method also enables modelling the impact of dynamic interventions, where different subjects can receive different levels of the exposure under study [ 27 ]. Although causal inference methods, and in particular the g-computation approach, have already been well described in the literature as a useful tool for assessing intervention effects and producing policy-relevant findings [ 28 , 29 , 30 ], their application in public health remains however limited [ 31 ]. In particular, g-computation has not yet been extensively used for studying the health impact of excess weight [ 32 , 33 , 34 ].

Other common methodological issues in observational studies aiming to evaluate the potential health impact of exposure-reducing interventions are related to the validity of self-reported data. Although a large body of literature already exists on methods to obtain more accurate surveillance data by correcting for measurement error related to self-reported data in health interview surveys, few epidemiologic studies use them in practice [ 35 , 36 ]. The measurement biases are however not without consequence because when exposures are not valid, the PIF estimates may be severely biased.

This study aims to use a g-computation approach to quantify the effects of different population-based weight reduction interventions on important NCDs in Belgium in a multi-exposure context (taking into account lifestyle, metabolic, and environmental exposures). The research relies on cross-sectional data from the Belgian Health Interview Survey and Health Examination Surveys, addressing measurement bias due to self-reported health and anthropometric data through a random-forest multiple imputation method [ 37 ]. Additionally, this paper aims to provide a didactic application of the g-computation approach to assess PIF from cross-sectional data.

Study area, study population and data

The study area is the entire Belgian territory with a population of 11.6 million inhabitants in 2023. The study sample consists of 27 536 participants of different waves of the Belgian Health Interview Survey (BHIS 2008, 2013, and 2018) all aged 18 years and above. Additionally, it includes a subset of 1,184 participants who also took part in the Belgian Health Examination Survey in 2018 (BELHES 2018). The information from BELHES 2018 was primarily used to address measurement errors in self-reported health and anthropometric data.

The BHIS is a national cross-sectional population survey carried out every five years by Sciensano, the Belgian institute for health, in partnership with Statbel, the Belgian statistical office [ 38 ]. Data are collected through a stratified multistage, clustered sampling design and weighting procedures are applied to obtain results which are as representative as possible of the Belgian population [ 39 ]. In the BELHES, objective health information was collected among a random subsample of the BHIS participants. The BELHES included a short additional questionnaire, a physical examination, and the collection and analysis of blood- and urine samples. Details on the data collection are available in the BELHES publication [ 40 ].

Based on the geographical coordinates of the residential address of participants and using Geographical Information Systems (GIS), the dataset was further enriched with objective measures of the residential environment related to long-term exposure air pollution (Black carbon), green space (vegetation coverage in a 1 km buffer), and noise from road traffic (Lden, day–evening–night noise level).

Abdominal obesity and non-communicable disease indicators

BMI and waist circumference were used as continuous variables, the latter to assess abdominal obesity. Four NCDs were considered: diabetes (type 1 & 2), hypertension, cardiovascular disease (CVD), and musculoskeletal (MSK) disease. The variables used to construct these indicators are displayed in Table  1 .

Socio-demographic and lifestyle indicators

The following variables were used to describe each participant’s socio-demographic status:  age (years) ,  sex (male vs female) ,  household composition (single, one parent with child(ren), couple without child(ren), couple with child(ren), other or unknown) ,  highest educational level in the household (No diploma/primary school, low secondary, high secondary, higher) , reported household income (quintiles), birth country (Belgian, Non Belgian EU, non-Belgian non EU) , and civil status (single, married, widow, divorced). To describe the participant’s lifestyle, we used the variables:, smoking status (daily smoker, occasional smoker, former smoker or never smoked) , indoor smoking (yes vs no), alcohol consumption, and level of physical activity (≥  4 h sport or intensive training per week, < 4 h sport or light activities per week or sedentary behavior) . This last variable is based on the WHO indicator describing leisure time activity in the last 12 months [ 41 ], where sedentary behavior is defined as the complete absence of physical leisure activities. To assess alcohol consumption, we transformed the ordinal variable representing the average number of alcoholic beverages per week into a numeric variable. The numeric values assigned were as follows: 1 = Abstainers and occasional drinkers, 2 = 1 to 7 glasses, 3 = 8 to 14 glasses, 4 = 15 to 21 glasses, 5 = 22 + glasses. One glass stands for a “standard unit” which varies according to the type of alcohol (for example, 0,33 l beer, 0,125 l wine, 4 cl spirits, etc.).The reported household income, defined by the quintile distribution was also converted into a numeric variable. The binary variable Indoor smoking describes household where at least one person smokes inside the dwelling on most days.

Environmental indicators

The selection of environmental factors in our study, including air pollution, green spaces, and noise, was guided by their well-established associations with NCDs. They represent a good proxy of the individual exposure since they were derived from the geographical coordinates of the survey participants' residential addresses.

Air pollution was assessed through the annual average of exposure to black carbon (BC). BC represents one of the most health-relevant components of particulate matter (PM) and is a valuable indicator to assess the health effects of air quality dominated by primary combustion particles [ 42 ]. BC exposure was obtained as a continuous grid through the Belgian Interregional Environment Agency (IRCEL – CELINE) which supervises the national monitoring system assessing air pollutant concentrations through a dense network of stations, and estimates local exposure through interpolation, taking into account land cover data in combination with a dispersion model [ 43 , 44 ]. BHIS data of 2008, 2013, and 2018 were respectively linked to BC exposure data of 2010, 2013, and 2018. Exposure to green spaces was assessed based on CORINE Land Cover (CLC) data [ 45 ]. The vegetation coverage was obtained at the neighborhood level in a 1 km buffer around the respondent’s dwelling. This 1 km buffer of vegetation coverage is justified by the need to capture the immediate neighborhood environment that individuals are likely to interact with regularly and aligns with common practices in environmental epidemiology, where this scale is frequently used to assess the impact of neighborhood characteristics on health. Lifestyle factors, including physical activity and stress reduction, are influenced by the accessibility of these spaces in one's daily life. BHIS data of 2008, 2013, and 2018 were respectively linked to green space data of 2006, 2012, and 2018. Noise pollution, approached through the road traffic noise (Lden, day–evening–night noise level), was obtained from published noise maps, as required by the European Noise Directive (2002/49/EC) [ 46 , 47 , 48 ]. Noise data are created at the regional level and downloaded from the regional portals for environmental data [ 49 , 50 , 51 ]. BHIS data of 2008, 2013, and 2018 were respectively linked to noise data of 2016. Noise from the road traffic, is recognized as a significant environmental stressor associated with various health issues, including cardio-vascular diseases and a lower quality of life. The Lden metric provides a comprehensive measure of overall noise exposure and the 55 dB used cut-off aligns with the recommended WHO threshold, acknowledging the detrimental health impact above this threshold.

Statistical analyses

All variables were described with their 95% confidence interval and the missing data pattern was displayed for the merged BHIS/BELHES dataset (additional file 1 , 2 , 3 , 4 , 5 ).

Database compiling

In a first step, the measurement error related to self-reported height, weight, diabetes, and hypertension in the BHIS database was corrected based on the objective information included in the BELHES and using a random forest multiple imputation method. A MICE algorithm [ 52 ] was used to multiply impute the missing values of the merged dataset. The imputation model included all the variables of the dataset, including variables used in the weighting procedure associated with the survey sample design (province, number of persons by household, age, and sex). All missing values of the covariates included in the imputation models were imputed in the same process. Details on the application of this correction method in the BHIS is found in a previous publication [ 37 ]. The number of iterations of the random-forest multiple was set to 500 and the defined number of trees was set to 100. The convergence of the algorithm was monitored by plotting the mean and standard deviation of the synthetic values against the iteration number for the imputed B HIS data (Additional file 2 ). The number of imputations was limited to 10, which was found satisfactory: using infinitely many imputations instead of 10 was estimated to reduce the variance of the estimators by at most 1%.

Population impact fractions

In a second step, a g-computation approach was used in each of the 10 completed datasets to assess the PIF of four weight reduction scenarios:

a distribution shift where, for each person with overweight, a counterfactual weight was randomly drawn from the distribution of persons with a “normal” BMI (> = 18.5 and < 25)

a one-unit reduction of the BMI of people with overweight

a modification of the BMI of people with overweight based on a weight loss of 10%.

a reduction of the waist circumference (WC; cm) to half of the height (cm), among all people with a WC:height ratio greater than 0.5 [ 53 ].

The selection of these four scenarios aims to provide a comprehensive exploration of potential BMI reduction strategies and was guided by a combination of practical relevance and existing literature supporting their potential impact on health outcomes.

The impact of the first three scenarios on each NCD was evaluated for two target populations: people with overweight (BMI > 25) and people with obesity (BMI > 30). The fourth scenario was applied to the specified population.

The mean reduction in BMI was calculated for the first three scenarios, and the mean reduction in waist circumference was calculated for the fourth scenario. This was determined by subtracting the counterfactual BMI (or WC) under the intervention from the actual BMI (or WC) of each individual and then averaging these differences.

In each of the ten imputed datasets, standard errors of the PIF were obtained using 1000 nonparametric bootstrap samples. The imputation steps of the g-computation approach [ 28 ] are described in Fig.  1 :

Steps of the g-computation approach

The association between excess weight and each NCD was modelled based on the “backdoor criterion” which is specific to the causal inference theory [ 54 ]. The DAG displayed in Fig.  2 illustrates the postulated causal structure of the association between excess weight and NCDs. Confounding factors such as socio-economic, environmental, and lifestyle factors influence this association. Excess weight affects NCDs through metabolic risk factors. Models were not adjusted for the metabolic risk factors (hypercholesterolemia and hypertension) since they were colliders or lying on the causal pathway between excess weight and the disease. Two logistic regression models were performed for the four NCDs considered. The first model included BMI and the second model included WC to assess the excess weight status. Models were adjusted for socio-economic, lifestyle, environmental factors, region, and year. Interactions were tested between BMI (and WC) and each of the covariates. The performance of the models was assessed by randomly splitting each of the ten imputed dataset into a training dataset (70%) and a test dataset (30%) and by evaluating the Area under the curve (AUC). The ten obtained AUC values were then averaged. In order to account for the potential indirect effect of BMI on chronic diseases through physical activity, sensitivity analyses were performed by fitting models without adjustment for physical activity.

Directed acyclic graph of the causal association between excess weight and each of the four non-communicable diseases (diabetes, hypertension, cardiovascular disease, and musculoskeletal disease)

The PIF of each scenario was calculated in each of the ten imputed dataset and results of the multiple analysis were pooled using the standard Rubin rules [ 55 ]. Standard errors of the prevalence estimates were obtained as the square root of the total variance (taking into account the within and between imputation variance and a correction factor for using 10 imputations). PIF were reported as percentage indicating the proportion of disease cases that would be avoided under the hypothetical weight reduction scenarios. The degree to which all the underlying assumptions required to draw a causal inference [ 56 ] (temporal ordering, exchangeability, no-interference, experimental treatment assignment, consistency, no model misspecification, no measurement error) is addressed in the Discussion section.

Statistical analyses were performed taking into account the survey sample design. The multistage sampling method was accommodated by incorporating weights, calculated to reflect the likelihood of being selected in the sample, based on the geographical stratification, the selection of clusters within each stratum, the choice of households within each cluster, and the selection of individuals within each household.

All analyses were fit and evaluated using the statistical software R, version 4.2.1 (R Development Core Team, 2006) and the “mice” package [ 57 ]. The R code used for the implementation of the G-computation to assess the PIF (for the diabetes example) is available in Additional file 6 .

Data description

A total of 27,536 participants from the 2008, 2013, and 2018 Belgian Health Interview Surveys (BHIS), aged 18 years and above, were included in the analysis, with 1,184 of them participating in the 2018 Belgian Health Examination Survey (BELHES).The missing data pattern and summary statistics of all considered variables in the merged BHIS/BELHES dataset are displayed in Additional files 2 – 6 . The impact of the four weight reduction scenarios on the BMI and WC distribution are visualized in Fig.  3 .

BMI and waist circumference distribution under the four weight-reduction scenarios

Association between excess weight and diabetes, hypertension, CVD, and MSK disease

Results of the multivariable logistic regression models showed a significant association between both BMI and WC and each of the four NCDs that were considered (Table  2 ). A stronger association was found for diabetes and hypertension compared to CVD and MSK disease. The four models for diabetes, hypertension, CVD, and MSK demonstrated a good predictive performance with AUC of 77%, 80%, 80%, and 72%, respectively. Forest plots of the logistic regression models for each NCD are displayed in Additional files 7 , 8 , 9 , 10 . The results of the sensitivity analysis without adjustment for physical activity showed similar estimates (additional file 11 ).

Potential impact fractions of the four weight reduction scenarios

The PIFs of the four weight reduction scenarios on diabetes, hypertension, CVD, and MSK disease in Belgium are visualized in Fig.  4 . The average BMI reduction under the first three scenarios are respectively 4.2, 1 and 1.6 units and the average WC reduction under the last scenario is 9.9 cm. These amount to less than 1 SD (the conditional SD of BMI, given all the covariates equals 5.3 units, and of WC equals 13.2 cm).

Bar plots illustrating the potential impact fraction (PIF) of the four weight reduction scenarios (1. distribution shift where, for each person with overweight, a counterfactual weight was drawn from the distribution of persons with a “normal” BMI, 2. One-unit reduction of the BMI of individuals with overweight, 3. modification of the BMI of individuals with overweight based on a weight loss of 10%, 4. reduction of the WC to the half of the height among all people with a WC/height ratio greater than 0.5) on A . diabetes, B . hypertension, C . cardiovascular disease, and D . musculoskeletal disease in Belgium. Error bars represent the 95% confidence intervals

The fourth scenario, where the waist circumference was reduced to half of the height had the highest impact on the four diseases considered, with nearly one third of the diabetes cases and one fourth of the hypertension cases that could have been avoided in the Belgian population. By contrast, the second scenario, where the BMI of people with excess weight was reduced by one unit had only a marginal impact on the four diseases considered. PIF were higher when the scenarios applied to people with overweight compared to people with obesity only (Table  2 ). The PIFs were all significantly different from 0, except for scenarios 3 related to CVD.

Main findings

In this study, we presented a g-computation approach to evaluate the potential impact of hypothetical weight reduction scenarios on the burden of four NCDs in Belgium. We examined what would be the risk of suffering from diabetes, hypertension, CVD, and MSK disease if we could manipulate the BMI or the WC of Belgian adults and set them to values determined by hypothetical scenarios. The predicted risk was then compared to the risk under the “status quo” scenario, where no intervention would be implemented to the population. This is in contrast with the estimates we would have obtained using traditional regression models which produced stratum-specific odds ratios.

Our findings suggest that implementing weight reduction scenarios among individuals with excess weight could lead to a substantial and statistically significant decrease in the prevalence of diabetes, hypertension, CVD, and MSK diseases in Belgium. A major benefit was found for the fourth scenario, where the WC was lowered to half of the height for all Belgians with a ratio WC:height ratio above 0.5. Under this scenario, the prevalence of diabetes and hypertension would be drastically reduced, with respectively 36% and 25% of avoidable cases. The reduction was less pronounced for CVD and MSK diseases with a PIF of respectively 11% and 7%. A recent guideline report from the National Institute for Health and Care Excellence (NICE) mentioned that a waist measurement of more than half of a person’s height was a better indicator of increased fat in the abdomen compared to BMI and could better predict the risk of developing NCDs such as type 2 diabetes or CVD [ 53 ]. BMI remains however a useful practical measure to define overweight and obesity but should be interpreted with caution especially among older people and adults with high muscle mass, since it is less accurate to determine body fatness in these groups [ 58 ].

High PIFs were also observed under the first scenario, where the distribution of the BMI of all people with overweight would be shifted to the distribution of the BMI of people fallen in the “normal” BMI category. While this scenario may not be highly realistic, it is nonetheless valuable in defining the boundaries within which realist policy interventions could have an impact. This very theoretical scenario has the advantage to estimate the global burden of excess weight on NCDs and is closest to traditional PAF which estimate the risk of disease with a complete removal of the risk factor in the population. Under this first scenario, PIF for diabetes, hypertension, CVD, and MSK disease were 32%, 23%, 9%, and 6%, respectively. Those estimates were however lower in comparison to the PAF estimates obtained from the last Global Burden of Disease (GBD) study where the PAFs attributable to high BMI in Belgium were respectively of 50% for diabetes, 20% for ischemic heart diseases, 25% for stroke, 7% for back pain, and 13% for osteoarthritis [ 59 ].

It must be noted that those estimates cannot directly be compared to the estimates presented in this article. The g-computation approach is tailored to our data by estimating, for each individual, the conditional probability of developing a chronic disease given the variables included in the model, and subsequently averaging it at the population level.

In contrast, the GBD study's PAF estimates consider the overall contribution of high BMI to diseases across the entire population. They are not calculated directly from the specific population but often rely on relative risk estimates from external studies. These differences in data sources, methodologies, and the underlying framework for estimating population-level burden versus individual causal effects make direct comparisons between the two sets of estimates complex.

In addition, the variables for CVD and MSK used in this study were constructed based on a group of diseases (Table  1 ), which is difficult to compare with the GBD estimates, where PAFs are calculated for each disease separately.

The second and third scenario, where the BMI was respectively reduced by one unit and modified based on a ten percent reduction of the person’s weight, represent more realistic scenarios but had a smaller impact on the prevalence of the four diseases. A weight loss of 5–10% is considered by guidelines from the UK and the USA a the minimum weight loss to be achieved to have a clinical impact on health outcomes [ 58 , 60 ]. To achieve this goal, evidence-based interventions include dietary modifications, physical activity, psychological interventions, pharmacotherapy, and bariatric surgery, for individuals with severe obesity [ 61 , 62 ]. There is substantial evidence demonstrating that these interventions not only contribute to weight loss but also have a statistically significant impact on reducing the risk of obesity-related outcomes [ 63 ]. A one-unit reduction in BMI within the Belgian population would result in a reduction of 4.5% of the cases of diabetes, 3% of the cases of hypertension, 1.5% of the cases of CVD, and 1% of the cases of MSK disorders.

Strengths and limitations

An important strength of this study lies in the didactic application of the g-computation approach and the description of the steps required to estimate the population effect of a potential intervention in cross-sectional data. The methodological tool used in this present study, based on a g-computation approach and a random-forest multiple imputation method, allows the assessment of the potential effects of any well-defined intervention and targeting of any subgroup of interest, while also addressing the bias related to self-reported data and the missing data issue in health interview surveys. This paper contributes to familiarizing a public health audience with the g-computation approach enabling them to estimate policy-relevant effects of hypothetical health interventions. Compared to standard analytic techniques, the g-computation approach has the advantage to provide flexibility in simulating real world interventions. It enables modeling the impact of dynamic interventions, where different subjects can receive varying levels of the exposure under study, as well as joint interventions, where the values of multiple exposures can be modified simultaneously. Another additional benefit of the g-computation approach, lies in its ability to handle time-varying confounders (i.e., confounders whose value changes over time), especially in situations where there's treatment confounder feedback (i.e., when the confounder is affected by the exposure) [ 64 ]. However, the cross-sectional nature of the data in this study did not allow us to take full advantage of this benefit.

This study also represents the first application of the random-forest multiple imputation method to address the bias related to self-reported health and anthropometric data in the BHIS. This method has been recently identified as a more adequate approach for valid measurement error correction in comparison to regression calibration [ 37 ]. Whenever feasible, self-reported information from health interview surveys should be combined with objective information from health examination surveys to address the bias related to self-reported anthropometric data and therefore provide more accurate PIF. A second important strength of the present study is the consideration of the potential confounding role of the environmental factors in the association between excess weight and chronic diseases. In particular, the linkage of the BHIS data with objective environmental factors at the residential address of the participants provides a significant improvement on the state of the art, as most studies do not consider environmental factors in the link between BMI and chronic diseases. Also, environmental factors are often assessed on a broad scale, using exposure e.g. in administrative units. Our study used the residential address, thus considerably refining the spatial scale. The limits of this approach are discussed further in the section measurement error .

Findings of this study must nevertheless be seen in the light of some limitations. If the g-computation approach allowed to evaluate the PIF of several weight reduction scenarios, the obtained estimates should however be treated with caution and several assumptions need to be met to interpret them causally. The first assumption is the “temporal ordering assumption” where we assume that the exposure precedes the outcome and the confounding factors precede the exposure. Unfortunately, this required assumption is not met by the cross-sectional structure of the data and is undoubtedly the most questionable assumption in this present study. While we can reasonably assume that fixed variables such as age, sex or education are causes rather than effects of the excess weight risk factor, it is not that obvious that the excess weight risk factor precedes chronic disease or that lifestyle factors precede the weight status. Making the distinction between unintentional weight loss, which may result from chronic disease, and intentional weight loss can be challenging [ 65 ]. People suffering from chronic disease could also be physically less active and therefore be at greater risk of gaining weight. For instance, individuals with CVD, MSK disorders or diabetes may exhibit weight gain due to factors like reduced mobility (leading to a decrease in calorie expenditure), medications, or fluctuations in blood sugar levels. Another challenge with cross-sectional data is the inability to differentiate whether covariates function as mediators or confounders. In this study, physical activity was considered as a confounding factor but it cannot be ruled out that excess weight may impact physical activity and indirectly the risk of chronic disease. One possible consequence could be underestimation of the true causal effect because the PAF would not incorporate all burden for the disease that is attributable to the excess weight risk factor. Physical activity could also function as a collider variable (a variable that is a common effect of both the exposure and the outcome) and adjusting for it may have introduced collider bias, potentially generating a spurious association between excess weight and chronic diseases.

The second assumption is the “exchangeability” assumption which assumes that there are no unmeasured confounding factors in the exposure-outcome association. Indeed, the exposure may only be considered as randomized within each stratum of the confounders if all confounders are considered in the model. This assumption is also very difficult to meet in the available cross-sectional study. Although we included in our analyses all the confounders identified in the literature that were available in our data, there remain several potential unmeasured confounding factors, such as genetic factors or nutritional habits which can both play an important role in the association between excess weight and chronic disease. Even though the variables related to nutritional habits were available in the BHIS, it was decided to not include them in the model because they were highly prone to a reverse causation effect.

The third assumption, known as the “no-interference assumption”, asserts that the outcome of each individual is not affected by the exposures and outcomes of the other individuals. We can reasonably expect that this assumption is fully met in our study for the reason that chronic diseases are not contagious. This, however, may vary depending on the intervention and study group. For instance, the implementation of a dietary intervention to reduce BMI of participants, such as changing the cooking style in the family, could potentially influence members of the same family similarly.

The fourth assumption, the “experimental treatment assignment” assumption, also called the positivity assumption [ 66 ], assumes that the exposure to the risk factor is possible for all individuals in each stratum of the covariates. In the context of this study, it means that the BMI values generated under the considered scenarios must be attainable for all individuals in which the scenario took place. This assumption is closely related the realism of the scenario and is therefore more likely violated for the first and fourth scenarios, which requires changes in the BMI or in the WC that are rarely observed in the population (e.g. a drop in the BMI from 35 to 25). In concrete terms, this means that each stratum of the covariates that contains overweight individuals should also contain individuals with a normal BMI. To evaluate the positivity assumption, we compared the probability of individuals being overweight among the two populations groups under study (individuals with overweight and individuals with a “normal” BMI). We built a model for BMI based on all confounders, and predicted, for each individual with overweight, the probability of being overweight. This process was repeated for individuals with a normal BMI. The observed overlap between the two probability distributions suggests that this assumption is plausible (Additional file 12 ).

The fifth assumption is the “consistency” assumption, which assumes that “an individual’s potential outcome under his observed exposure history is precisely his observed outcome” [ 19 ]. While consistency is plausible for medical treatments, because it is easy to manipulate hypothetically an individual's treatment status, consistency may however be problematic when the exposure is a biologic feature and the manipulation difficult to conceive [ 67 ]. Violations of consistency assumption often occur when there is ambiguity in the definition of interventions to change exposure. In the context of this study, BMI interventions remain vague because they specify attributes rather than specific behaviors. The main limitation of our approach lies in the highly theoretical nature of the hypothetical scenarios considered, which do not accurately mirror real-world interventions. Ambiguity arises from the fact that there are many competing approaches to decrease an individual’s BMI and each of these approaches may have a different causal effect on the outcome [ 68 ]. By presenting an estimate for the effect of a “BMI reduction”, we implicitly assume that all interventions on BMI have the same effect on the risk of suffering from a chronic disease, which is unlikely to hold. Another difficulty arising from ill-defined interventions is the challenge of selecting the confounding factors required to achieve conditional exchangeability. Firstly, the set of confounding factors to be considered may vary for different versions of the intervention. Secondly, because BMI is not an intervention in itself but rather a physiological risk factor, identifying all the confounders becomes a practically impossible task due to the necessity of also considering genetic factors. Even if we manage to account for all potential confounding factors including genetic factors, there is a high likelihood that the positivity assumption will be violated. Certain genetic traits could exert such a strong influence on body weight that all subjects possessing them automatically become obese [ 68 ]. Another issue with interventions on BMI is that the better we adjust for confounders that determine both excess weight and chronic diseases, the more we narrow our focus to the remaining factors that have a direct effect on BMI (such as genetic predispositions). Consequently we isolate a potential intervention that changes the remaining determinants of BMI. In this study, we compared the risk of suffering from a chronic disease of overweight vs non overweight individuals conditional on their physical activity level, smoking status, environmental and alcohol consumption. This means that our estimates correspond to the effect of other versions of the intervention “BMI reduction”, such as healthy diet or genes. However, other versions of the intervention may not be manipulable and not be of primary interest for policymakers. Successful interventions with evidence for effective weight reduction are multifactorial and it is unrealistic to assume that BMI in the population could be modified without considerable changes to all other aspects of lifestyle. Our findings may therefore be underestimated, since our analyses adjusted for possible confounding by physical activity or alcohol consumption and thereby do not entirely take into account the co-benefits of weight reduction intervention via changes in physical activity or alcohol consumption.

The sixth underlying assumption of g-computation approach is the “no model misspecification” assumption. A necessary condition (but not sufficient) for the absence of model misspecification is that the model should be able to accurately predict the outcome under no intervention. Variables from the model were selected based on their theoretical relevance and guided by a DAG that reflects the hypothesized causal structure. Non-linear relationships were assessed by testing the quadratic terms, while interactions were examined using the StepAIC algorithm (a variable selection method that iteratively adds or removes variables from a model based on their impact on the Akaike Information Criterion, aiming to find the most parsimonious model with a good fit). The AUC demonstrated a good predictive performance for the four NCDs models.

Lastly, like other studies based on observational data, the validity of our results relies on the key assumption of no measurement error. It can however be challenging to accurately assess the exposure to risk factors of NCDs through observational studies, such as abdominal obesity or environment. Although we applied a correction method to address the bias of self-reported anthropometric data and used both BMI and waist circumference separately to approximate abdominal obesity, another measure that could have been used is the Body Shape Index (ABSI), a comprehensive indicator of body shape integrating both waist circumference and BMI [ 69 ]. For the environment also, it is important to keep in mind that air pollution exposure is extrapolated from the mean annual concentration of a given area to individual exposure, and does not take into account the time spent in this area. Personal mobility could be integrated in dynamic exposure assessments, but determining individual buffer values to delimit a person’s neighborhood is still an active field of research. Other methods to determine environmental exposure are human biomonitoring or deploying wearable sensors, but this is unfortunately impossible to apply for large samples, over long time periods or for past studies. There was also a time lag between health data collection and environmental data. However, as environmental change is slow, we do not expect a strong impact on our results. A certain degree of measurement error also applies for the diseases. While the bias related to self-reported diabetes and hypertension could be addressed based on clinical information from the BELHES, the same correction could not be applied for self-reported CVD and MSK diseases, as the relevant clinical information was not available in the BELHES.

Furthermore, our estimates apply to the Belgian population and may not be generalizable to other populations characterized by different NCDs risk factor distributions. For example, we estimated the risk of diabetes in the Belgian population for a distribution shift of the BMI of individuals with overweight to the distribution of individuals with a normal BMI, but the BMI distribution may be very different in other populations. Our PIF estimates may also vary a lot for different diseases within the same CVD or MSK group, limiting the possibility of comparing our results with the GBD estimates.

A final limitation of our study lies the lack of detailed analysis regarding the differential effects of BMI on different types of diabetes. While our findings demonstrate a significant association between BMI and diabetes, it must be recognized that the impact of BMI may vary between type 1 and type 2 diabetes. While the link between obesity and type 2 diabetes is well-established, emerging evidence suggests a link between obesity and type 1 diabetes as well [ 70 ]. Future research could explore this aspect further to elucidate whether BMI affects both types of diabetes similarly.

Whilst obesity is widely considered as a major modifiable risk factor for many chronic diseases, nevertheless, a rigorous examination of the mentioned assumptions underscores the challenge in determining its causes and consequences. Addressing this is however important, as the prevention of any disease requires that interventions focus on causal risk factors. Although all the required assumptions of the g-computation approach may not be fully met, based on the literature knowledge regarding the relationship between excess weight and NCDs, the evidence from literature supports the direction of causality investigated in this study.

Conclusions

This study gives a demonstration of the use of a g-computation approach to assess the benefits of hypothetical weight reduction scenarios on NCDs in Belgium in a multi-exposure context. Results suggest that implementing weight reduction scenarios among individuals with excess weight could lead to a substantial and statistically significant decrease in the prevalence of diabetes, hypertension, cardiovascular disease (CVD), and musculoskeletal (MSK) diseases in Belgium. The g-computation based approach to assess PIF of interventions represents a straightforward approach in epidemiology for making causal inference from observational data while providing also useful information for policy makers. Future epidemiological and health impact assessment studies should be conducted in ways that are more informative for policymakers and should consider all the underlying assumptions explicitly in order to better evaluate the possibility of a causal effect. In particular, we acknowledge the importance of the consistency assumption in ensuring the validity of the study’s findings, especially within the field of obesity epidemiology. Ideally, longitudinal studies including time-varying data should be used in the future to address the “temporal ordering assumption” in the association between excess weight and chronic diseases.

Availability of data and materials

The data that support the findings of this study are not publicly available. Data are however available from the authors upon reasonable request and with specific permission ( https://www.sciensano.be/en/node/55737/health-interview-survey-microdata-request-procedure ). Legal restrictions make that BHIS and BELHES data can only be communicated to other parties if an authorization is obtained from the sectoral committee social security and health of the Belgian data protection authority.

WHO Europe. WHO European Regional Obesity Report 2022. 2022. Available from: https://apps.who.int/iris/bitstream/handle/10665/353747/9789289057738-eng.pdf

Abarca-Gómez L, Abdeen ZA, Hamid ZA, Abu-Rmeileh NM, Acosta-Cazares B, Acuin C, et al. Worldwide trends in body-mass index, underweight, overweight, and obesity from 1975 to 2016: a pooled analysis of 2416 population-based measurement studies in 128·9 million children, adolescents, and adults. Lancet. 2017;390(10113):2627–42.

Article   Google Scholar  

Drieskens S, Charafeddine R, Gisle L. Enquête de santé 2018 : Etat nutritionnel. Bruxelles, Belgique : Sciensano [Internet]. Report No.: D/2019/14.440/62. Available from: www.enquetesante.be . Cited 2024 Apr 8.

Ralston J, Cooper K, Powis J. Obesity, SDGs and ROOTS: a framework for impact. Curr Obes Rep. 2021;10(1):54–60.

Article   PubMed   PubMed Central   Google Scholar  

Pang M, Kaufman JS, Platt RW. Studying noncollapsibility of the odds ratio with marginal structural and logistic regression models. Stat Methods Med Res. 2016;25(5):1925–37.

Article   PubMed   Google Scholar  

Levin ML. The occurrence of lung cancer in man. Acta Unio Int Contra Cancrum. 1953;9(3):531–41.

CAS   PubMed   Google Scholar  

Mansournia MA, Altman DG. Population attributable fraction. BMJ. 2018;22(360):k757.

Eide GE. Attributable fractions for partitioning risk and evaluating disease prevention: a practical guide. Clin Respir J. 2008;2(s1):92–103.

Nusselder WJ, Looman CWN. Decomposition of differences in health expectancy by cause. Demography. 2004;41(2):315–34.

Rückinger S, von Kries R, Toschke AM. An illustration of and programs estimating attributable fractions in large scale surveys considering multiple risk factors. BMC Med Res Methodol. 2009;23(9):7.

Rothman KJ, Greenland S. Causation and Causal Inference in Epidemiology. Am J Public Health. 2005;95(S1):S144–50.

Greenland S. Concepts and pitfalls in measuring and interpreting attributable fractions, prevented fractions, and causation probabilities. Ann Epidemiol. 2015;25(3):155–61.

Greenland S, Robins JM. Conceptual problems in the definition and interpretation of attributable fractions. Am J Epidemiol. 1988;128(6):1185–97.

Article   CAS   PubMed   Google Scholar  

Morgenstern H, Bursic ES. A method for using epidemiologic data to estimate the potential impact of an intervention on the health status of a target population. J Community Health. 1982;7(4):292–309.

Saatchi M, Mansournia MA, Khalili D, Daroudi R, Yazdani K. Estimation of Generalized Impact Fraction and Population Attributable Fraction of Hypertension Based on JNC-IV and 2017 ACC/AHA Guidelines for Cardiovascular Diseases Using Parametric G-Formula: Tehran Lipid and Glucose Study (TLGS). Risk Manag Healthc Policy. 2020;5(13):1015–28.

Drescher K, Becher H. Estimating the generalized impact fraction from case-control data. Biometrics. 1997;53(3):1170–6.

Khosravi A, Mansournia MA. Recommendation on unbiased estimation of population attributable fraction calculated in “prevalence and risk factors of active pulmonary tuberculosis among elderly people in China: a population based cross-sectional study.” Infect Dis Poverty. 2019;8(1):75.

Hernán MA, Robins JM. Estimating causal effects from epidemiological data. J Epidemiol Community Health. 2006;60(7):578–86.

Robins JM, Hernán MÁ, Brumback B. Marginal Structural Models and Causal Inference in Epidemiology. Epidemiology. 2000;11(5):550.

Dahlqwist E. Method developments for the attributable fraction in causal inference. Inst för medicinsk epidemiologi och biostatistik / Dept of Medical Epidemiology and Biostatistics; 2019. Available from: http://openarchive.ki.se/xmlui/handle/10616/46672 . Cited 2021 Feb 23

Breskin A, Edmonds A, Cole SR, Westreich D, Cocohoba J, Cohen MH, et al. G-computation for policy-relevant effects of interventions on time-to-event outcomes. Int J Epidemiol. 2020;49(6):2021–9.

Article   PubMed Central   Google Scholar  

Hernán MA. A definition of causal effect for epidemiological research. J Epidemiol Community Health. 2004;58(4):265–71.

Igelström E, Craig P, Lewsey J, Lynch J, Pearce A, Katikireddi SV. Causal inference and effect estimation using observational data. J Epidemiol Community Health. 2022;76(11):960–6.

Palazzo C, Yokota RTC, Ferguson J, Tafforeau J, Ravaud JF, Van Oyen H, et al. Methods to assess the contribution of diseases to disability using cross-sectional studies: comparison of different versions of the attributable fraction and the attribution method. Int J Epidemiol. 2019;48(2):559–70.

Robins J. A new approach to causal inference in mortality studies with a sustained exposure period—application to control of the healthy worker survivor effect. Math Model. 1986;7(9):1393–512.

Greenland S, Pearl J, Robins JM. Causal diagrams for epidemiologic research. Epidemiology. 1999;10(1):37–48.

Robins J, Hernan M. Estimation of the causal effects of time-varying exposure. In: Longitudinal data analysis. 2008. p. 553–99.

Chapter   Google Scholar  

Ahern J, Hubbard A, Galea S. Estimating the effects of potential public health interventions on population disease burden: a step-by-step illustration of causal inference methods. Am J Epidemiol. 2009;169(9):1140–7.

Little RJ, Rubin DB. Causal effects in clinical and epidemiological studies via potential outcomes: concepts and analytical approaches. Annu Rev Public Health. 2000;21(1):121–45.

Hubbard AE, Laan MJVD. Population intervention models in causal inference. Biometrika. 2008;95(1):35–47.

Snowden JM, Rose S, Mortimer KM. Implementation of G-Computation on a Simulated Data Set: Demonstration of a Causal Inference Technique. Am J Epidemiol. 2011;173(7):731–8.

Taubman SL, Robins JM, Mittleman MA, Hernán MA. Intervening on risk factors for coronary heart disease: an application of the parametric g-formula. Int J Epidemiol. 2009;38(6):1599–611.

Danaei G, Pan A, Hu FB, Hernán MA. Hypothetical midlife interventions in women and risk of type 2 diabetes. Epidemiology. 2013;24(1):122.

Garcia-Aymerich J, Varraso R, Danaei G, Camargo Carlos A, Hernán MA Jr. Incidence of Adult-onset Asthma After Hypothetical Interventions on Body Mass Index and Physical Activity: An Application of the Parametric G-Formula. Am J Epidemiol. 2014;179(1):20–6.

Jurek AM, Maldonado G, Greenland S, Church TR. Exposure-measurement error is frequently ignored when interpreting epidemiologic study results. Eur J Epidemiol. 2006;21(12):871–6.

Shaw PA, Deffner V, Keogh RH, Tooze JA, Dodd KW, Küchenhoff H, et al. Epidemiologic analyses with error-prone exposures: review of current practice and recommendations. Ann Epidemiol. 2018;28(11):821–8.

Pelgrims I, Devleesschauwer B, Vandevijvere S, De Clercq EM, Vansteelandt S, Gorasso V, et al. Using random-forest multiple imputation to address bias of self-reported anthropometric measures, hypertension and hypercholesterolemia in the Belgian health interview survey. BMC Med Res Methodol. 2023;23(1):69.

Demarest S, Van der Heyden J, Charafeddine R, Drieskens S, Gisle L, Tafforeau J. Methodological basics and evolution of the Belgian health interview survey 1997–2008. Arch Public Health. 2013;71(1):24.

Health Interview Survey protocol. Available from: https://his.wiv-isp.be/SitePages/Protocol.aspx . Cited 2021 May 6.

Nguyen D, Hautekiet P, Berete F, Braekman E, Charafeddine R, Demarest S, et al. The Belgian health examination survey: objectives, design and methods. Arch Public Health. 2020;78(1):50.

de Bruin A, Picavet HSJ, Nossikov A. Health interview surveys: towards international harmonization of methods and instruments. World Health Organization. Regional Office for Europe; 1996;Xiii:161. Available from: https://apps.who.int/iris/handle/10665/107328 . Cited 2023 Apr 10

Janssen Nicole AH, Hoek G, Simic-Lawson M, Fischer P, van Bree L, Ten Brink H, et al. Black carbon as an additional indicator of the adverse health effects of airborne particles compared with PM10 and PM2.5. Environ Health Perspect. 2011;119(12):1691–9.

Article   CAS   PubMed   PubMed Central   Google Scholar  

Janssen S, Dumont G, Fierens F, Mensink C. Spatial interpolation of air pollution measurements using CORINE land cover data. Atmos Environ. 2008;42(20):4884–903.

Article   CAS   Google Scholar  

Lefebvre W, Vranckx S. Validation of the IFDM-model for use in urban applications. 2013. p. 208.

Google Scholar  

EEA. CLC CORINE Land Cover 2012, Version 18.5.1 2012. Available from: https://land.copernicus.eu/user-corner/technical-library/clc-country-coverage-v18.5

Leefmilieu Brussel-BIM. 49. Doelstellingen EN Methodologie Van de Geluidskadasters in het brussels hoofdstedelijk gewest. Collectie Factsheets, Thema Geluid; 2018. Available from: https://document.environnement.brussels/opac_css/elecfile/Geluid_49 . Cited 2020 Dec 23

Basner M, McGuire S. WHO Environmental Noise Guidelines for the European Region: a systematic review on environmental noise and effects on sleep. Int J Environ Res Public Health. 2018;15(3):519.

Directive 2002/49/CE du Parlement européen et du Conseil du 25 juin 2002 relative à l’évaluation et à la gestion du bruit dans l’environnement; 2002 p. 12–25. Report No.: Journal Officiel n° L 189. Available from: http://publications.europa.eu/resource/cellar/0354e2a3-4ee8-45a2-aa4a-090036045111.0010.04/DOC_1 . Cited 2021 Jan 6

Acouphen Environnement.  Carte de multi-exposition Bruxelles Environnement. Cadastre du bruit des transports routier, ferroviaire, aérien, trams et métro aérien de la Région de Bruxelles-Capitale. 2008. Available from: https://document.environnement.brussels/opac_css/elecfile/IBGE_Multi_2006_1.pdf . Cited 2020 Dec 23

Bruxelles Environnement. Rapport 2011–2014 van de staat van het leefmilieu: Exposition de la population au bruit des transports. 2011. Available from: https://environnement.brussels/lenvironnement-etat-des-lieux/rapports-sur-letat-de-lenvironnement/rapport-2011-2014/bruit-0 . Cited 2021 Jan 6

Digitaal Vlaanderen. Vlaams geoportaal. Available from: https://geopunt.be . Cited 2024 Jan 22.

van Buuren S. Flexible imputation for missing data, Second Edition (2nd ed.). Chapman & Hall/CRC; 2018. Available from: https://doi.org/10.1201/9780429492259 . Cited 2024 Apr 8.

Wise J. Advise adults to keep waist size to less than half their height, says NICE. BMJ. 2022;8(377):o933.

Arif S, MacNeil MA. Predictive models aren’t for causal inference. Ecology Letters. 2022;25(8):1741–5.

Campion WM, Rubin D. Multiple Imputation for Nonresponse in Surveys. 1989.

Hernan MA, Robins JM. Causal Inference: What If. Boca Raton: Chapman&Hall/CRC; 2020.

van Buuren S. Package ‘mice’; 2021. Available from: https://cran.r-project.org/web/packages/mice/mice.pdf

Overview | Obesity: identification, assessment and management | Guidance | NICE. NICE; 2014. Available from: https://www.nice.org.uk/guidance/CG189 . Cited 2023 Apr 13

Global burden of 369 diseases and injuries in 204 countries and territories, 1990–2019: a systematic analysis for the Global Burden of Disease Study 2019. Lancet. Available from: https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(20)30925-9/fulltext . Cited 2023 Apr 7

Jensen MD, Ryan DH, Apovian CM, Ard JD, Comuzzie AG, Donato KA, et al. 2013 AHA/ACC/TOS guideline for the management of overweight and obesity in adults: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines and The Obesity Society. Circulation. 2014;129(25 Suppl 2):S102-138.

PubMed   Google Scholar  

Garvey WT. New tools for weight-loss therapy enable a more robust medical model for obesity treatment: rationale for a complications-centric approach. Endocr Pract. 2013;19(5):864–74.

Wharton S, Lau DCW, Vallis M, Sharma AM, Biertho L, Campbell-Scherer D, et al. Obesity in adults: a clinical practice guideline. CMAJ. 2020;192(31):E875–91.

Schwingshackl L, Dias S, Hoffmann G. Impact of long-term lifestyle programmes on weight loss and cardiovascular risk factors in overweight/obese participants: a systematic review and network meta-analysis. Syst Rev. 2014;3(1):130.

Von Cube M, Schumacher M, Timsit JF, Decruyenaere J, Steen J. The population-attributable fraction for time-to-event data. Int J Epidemiol. 2023;52(3):837–45.

Haase CL, Lopes S, Olsen AH, Satylganova A, Schnecke V, McEwan P. Weight loss and risk reduction of obesity-related outcomes in 0.5 million people: evidence from a UK primary care database. Int J Obes. 2021;45(6):1249–58.

Petersen M, Porter K, Gruber S, Wang Y, Laan M van der. Diagnosing and Responding to Violations in the Positivity Assumption. UC Berkeley Division of Biostatistics Working Paper Series. 2010. Available from: https://biostats.bepress.com/ucbbiostat/paper269

Cole SR, Frangakis CE. The Consistency Statement in Causal Inference: A Definition or an Assumption? Epidemiology. 2009;20(1):3.

Hernán MA, Taubman SL. Does obesity shorten life? The importance of well-defined interventions to answer causal questions. Int J Obes (Lond). 2008;32(Suppl 3):S8-14.

Bertoli S, Leone A, Krakauer NY, Bedogni G, Vanzulli A, Redaelli VI, et al. Association of Body Shape Index (ABSI) with cardio-metabolic risk factors: a cross-sectional study of 6081 Caucasian adults. PLoS One. 2017;12(9):e0185013.

Kueh MTW, Chew NWS, Al-Ozairi E, le Roux CW. The emergence of obesity in type 1 diabetes. Int J Obes (Lond). 2024;48(3):289–301.

Download references

Acknowledgements

Not applicable

This work was conducted as part of the WaIST project (Contribution of excess weight status to the societal impact of non-communicable diseases, multimorbidity and disability in Belgium: past, present, and future) supported by Sciensano, the Belgian institute for health.

Author information

Authors and affiliations.

Department of Chemical and Physical Health Risks, Sciensano, Rue Juliette Wytsman 14, 1050, Brussels, Belgium

Ingrid Pelgrims & Eva M. De Clercq

Department of Applied Mathematics, Computer Science and Statistics, Ghent University, Krijgslaan 281, S9, BE-9000, Ghent, Belgium

Ingrid Pelgrims & Stijn Vansteelandt

Department of Epidemiology and Public Health, Sciensano, Rue Juliette Wytsman 14, 1050, Brussels, Belgium

Ingrid Pelgrims, Brecht Devleesschauwer, Stefanie Vandevijvere & Johan Van der Heyden

Department of Translational Physiology, Infectiology and Public Health, Ghent University, Salisburylaan 133, Hoogbouw, B-9820, Merelbeke, Belgium

Brecht Devleesschauwer

You can also search for this author in PubMed   Google Scholar

Contributions

IP performed the analysis and wrote the manuscript. BD, JVH, EDC, StefV were involved in the conception of the study. StijnV, JVH advised and helped the interpretation of data. BD, EDC, StijnV, StefV, JVDH provided critical revision of the manuscript. All authors read and approved the final manuscript.

Corresponding author

Correspondence to Ingrid Pelgrims .

Ethics declarations

Ethics approval and consent to participate.

The study was approved by Ethics committee of Ghent University Hospital and a positive advice was obtained (advice with registration number B670201734213 and advice with registration number B670201834895). An informed consent was obtained for every BHIS and BELHES participant. All methods were carried out in accordance with relevant guidelines and regulations.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Supplementary material 1., supplementary material 2., supplementary material 3., supplementary material 4., supplementary material 5., supplementary material 6., supplementary material 7., supplementary material 8., supplementary material 9., supplementary material 10., supplementary material 11., supplementary materail 12., rights and permissions.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ . The Creative Commons Public Domain Dedication waiver ( http://creativecommons.org/publicdomain/zero/1.0/ ) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Cite this article.

Pelgrims, I., Devleesschauwer, B., Vandevijvere, S. et al. The potential impact fraction of population weight reduction scenarios on non-communicable diseases in Belgium: application of the g-computation approach. BMC Med Res Methodol 24 , 87 (2024). https://doi.org/10.1186/s12874-024-02212-7

Download citation

Received : 20 July 2023

Accepted : 04 April 2024

Published : 14 April 2024

DOI : https://doi.org/10.1186/s12874-024-02212-7

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Non-communicable diseases
• g-computation
• Potential impact fractions
• Health policy
• Health impact assessment

BMC Medical Research Methodology

ISSN: 1471-2288

• General enquiries: [email protected]

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

• View all journals
• My Account Login
• Explore content
• About the journal
• Publish with us
• Sign up for alerts
• Open access
• Published: 26 November 2023

Impact of industrial robots on environmental pollution: evidence from China

• Yanfang Liu 1  

Scientific Reports volume  13 , Article number:  20769 ( 2023 ) Cite this article

3589 Accesses

1 Citations

2 Altmetric

Metrics details

• Environmental sciences
• Environmental social sciences

The application of industrial robots is considered a significant factor affecting environmental pollution. Selecting industrial wastewater discharge, industrial SO 2 emissions and industrial soot emissions as the evaluation indicators of environmental pollution, this paper uses the panel data model and mediation effect model to empirically examine the impact of industrial robots on environmental pollution and its mechanisms. The conclusions are as follows: (1) Industrial robots can significantly reduce environmental pollution. (2) Industrial robots can reduce environmental pollution by improving the level of green technology innovation and optimizing the structure of employment skills. (3) With the increase in emissions of industrial wastewater, industrial SO 2 , and industrial dust, the impacts generated by industrial robots are exhibiting trends of a “W” shape, gradual intensification, and progressive weakening. (4) Regarding regional heterogeneity, industrial robots in the eastern region have the greatest negative impact on environmental pollution, followed by the central region, and the western region has the least negative impact on environmental pollution. Regarding time heterogeneity, the emission reduction effect of industrial robots after 2013 is greater than that before 2013. Based on the above conclusions, this paper suggests that the Chinese government and enterprises should increase investment in the robot industry. Using industrial robots to drive innovation in green technology and optimize employment skill structures, reducing environmental pollution.

Similar content being viewed by others

The impact of industrial robot adoption on corporate green innovation in China

Lin Liang, Liujie Lu & Ling Su

Air pollution reduction and climate co-benefits in China’s industries

Haoqi Qian, Shaodan Xu, … Libo Wu

The impact of smart city construction (SCC) on pollution emissions (PE): evidence from China

GuoWei Zhang, XianMin Sun & Shen Zhong

Introduction

Since the reform and opening up, China’s rapid economic growth has created a world-renowned “economic growth miracle” 1 . With the rapid economic growth, China’s environmental pollution problem is becoming more and more serious 2 . According to the “ Global Environmental Performance Index Report ” released by Yale University in the United States in 2022, China’s environmental performance index scores 28.4 points, ranking 160th out of 180 participating countries. The aggravation of environmental pollution not only affects residents’ health 3 , but also affects the efficiency of economic operation 4 . According to calculation of the General Administration of Environmental Protection, the World Bank and the Chinese Academy of Sciences, China’s annual losses caused by environmental pollution account for about 10% of GDP. Exploring the factors that affect environmental pollution and seeking ways to reduce environmental pollution are conducive to the development of economy within the scope of environment.

Industrial robots are machines that can be automatically controlled, repeatedly programmed, and multi-purpose 5 . They replace the low-skilled labor force engaged in procedural work 6 , reducing the raw materials required for manual operation. Industrial robots improve the clean technology level and energy efficiency of coal combustion, reducing pollutant emissions in front-end production. Industrial robots also monitor the energy consumption and sewage discharge in the production process in real time. The excessive discharge behavior of enterprises in the production process is regulated, reducing the emission of pollutants in the end treatment. Based on the selection and coding of literature (Appendix A ), this paper uses the meta-analysis method to compare the impacts of multiple factors such as economics, population, technology, and policy on environmental pollution. As shown in Table 1 , compared to other factors, industrial robots demonstrate greater advantages in reducing environmental pollution. There is a lack of research on the relationship between industrial robots and environmental pollution in China. With the advent of artificial intelligence era, China’s industrial robot industry has developed rapidly. According to data released by the International Federation of Robotics (IFR), from 1999 to 2019, China’s industrial robot ownership and installation shows an increasing trend year by year (Fig.  1 ). In 2013 and 2016, China’s industrial robot installation (36,560) and ownership (349,470) exceeds Japan for the first time, becoming the world’s largest country in terms of installation and ownership of industrial robots. Whether the application of industrial robots in China contributes to the reduction of environmental pollution? What is the mechanism of the impact of China’s industrial robots on environmental pollution? Researching this issue is crucial for filling the gaps in existing research and providing a reference for other countries to achieve emission reduction driven by robots.

Industrial robot installations in the world’s top five industrial robot markets from 1999 to 2019.

Based on the above analysis, this paper innovatively incorporates industrial robots and environmental pollution into a unified framework. Based on the panel data of 30 provinces in China from 2006 to 2019, this paper uses the ordinary panel model and mediating effect model to empirically test the impact of industrial robots on China’s environmental pollution and its transmission channels. The panel quantile model is used to empirically analyze the heterogeneous impact of industrial robots on environmental pollution under different environmental pollution levels.

Literature review

A large number of scholars have begun to study the problem of environmental pollution. Its research content mainly includes two aspects: The measurement of environmental pollution and its influencing factors. Regarding the measurement, some scholars have used SO 2 emissions 7 , industrial soot emissions 8 and PM2.5 concentration 9 and other single indicators to measure the degree of environmental pollution. The single indicator cannot fully and scientifically reflect the degree of environmental pollution. To make up for this defect, some scholars have included industrial SO 2 emissions, industrial wastewater discharge and industrial soot emissions into the environmental pollution evaluation system, and used the entropy method to measure environmental pollution level 10 . This method ignores the different characteristics and temporal and spatial trends of different pollutants, which makes the analysis one-sided. Regarding the influencing factors, economic factors such as economic development level 11 , foreign direct investment 12 and income 13 , population factors such as population size 14 and urbanization level 15 , energy consumption 16 all have an impact on environmental pollution. Specifically, economic development and technological innovation can effectively reduce environmental pollution 17 . The expansion of population size can aggravate environmental pollution. Income inequality can reduce environmental pollution, but higher income inequality may aggravate environmental pollution 18 . There are “pollution heaven hypothesis” and “pollution halo hypothesis” between foreign direct investment and environmental pollution 19 . Technological factors also have a non-negligible impact on environmental pollution 20 .

With continuous deepening of research, scholars have begun to focus on the impact of automation technology, especially industrial robot technology, on the environment. Ghobakhloo et al. 21 theoretically analyzed the impact of industrial robots on energy sustainability, contending that the application of industrial robots could foster sustainable development of energy. Using data from multiple countries, a few scholars have empirically analyzed the effect of industrial robots on environmental pollution (Table 2 ). Luan et al. 22 used panel data from 73 countries between 1993 and 2019 to empirically analyze the impact of industrial robots on air pollution, finding that the use of industrial robots intensifies environmental pollution. Using panel data from 66 countries from 1993 to 2018, Wang et al. 23 analyzed the impact of industrial robots on carbon intensity and found that industrial robots can reduce carbon intensity. On the basis of analyzing the overall impact of industrial robots on environmental pollution, some scholars conducted in-depth exploration of its mechanism. Based on data from 72 countries between 1993 and 2019, Chen et al. 5 explored the impact of industrial robots on the ecological footprint, discovering that industrial robots can reduce the ecological footprint through time saving effect, green employment effect and energy upgrading effect. Using panel data from 35 countries between 1993 and 2017, Li et al. 24 empirically examined the carbon emission reduction effect of industrial robots, finding that industrial robots can effectively reduce carbon emissions by increasing green total factor productivity and reducing energy intensity. Although the above studies have successfully estimated the overall impact of industrial robots on environmental pollution and its mechanisms, they have not fully considered the role of technological progress, labor structure and other factors in the relationship between the two. These studies all chose data from multiple countries as research samples and lack research on the relationship between industrial robots and environmental pollution in China, an emerging country.

The above literature provides inspiration for this study, but there are still shortcomings in the following aspects: Firstly, there is a lack of research on the relationship between industrial robots and environmental pollution in emerging countries. There are significant differences between emerging and developed countries in terms of institutional background and the degree of environmental pollution. As a representative emerging country, research on the relationship between industrial robots and environmental pollution in China can provide reliable references for other emerging countries. Secondly, theoretically, the study of the impact of industrial robots on environmental pollution is still in its initial stage. There are few studies that deeply explore its impact mechanism, and there is a lack of analysis of the role of technological progress and labor structure in the relationship between the two.

The innovations of this paper are as follows: (1) In terms of sample selection, this paper selects panel data from 30 provinces in China from 2006 to 2019 as research samples to explore the relationship between industrial robots and environmental pollution in China, providing references for other emerging countries to improve environmental quality using industrial robots. (2) In terms of theory, this paper is not limited to revealing the superficial relationship between industrial robots and environmental pollution. it starts from a new perspective and provides an in-depth analysis of how industrial robots affect environmental pollution through employment skill structure and green technology innovation. This not only enriches research in the fields of industrial robots and the environment, but is also of great significance in guiding the direction of industrial policy and technology research and development.

Theoretical analysis and hypothesis

Industrial robots and environmental pollution.

As shown in Fig.  2 , the impact of industrial robots on environmental pollution is mainly reflected in two aspects: Front-end production and end treatment. In front-end production, industrial robots enable artificial substitution effects 25 . Manual operation is replaced by machine operation, reducing the raw materials needed for manual operation. Through the specific program setting of industrial robots, clean energy is applied to industrial production 26 . The use of traditional fuels such as coal and oil is reduced. In terms of end treatment, the traditional pollutant concentration tester only measures a single type of pollutant. Its data cannot be obtained in time. It is easy to cause pollution incidents. Industrial robots can measure a variety of pollutants, and have the function of remote unmanned operation and warning. It reflects the pollution situation in time, reducing the probability of pollution incidents. The use of robots can upgrade sewage treatment equipment and improve the accuracy of pollution treatment, reducing pollutant emissions. Based on the above analysis, this paper proposes hypothesis 1.

The impact of industrial robots on environmental pollution.

Hypothesis 1

The use of industrial robots can reduce environmental pollution.

Mediating effect of green technology innovation

Industrial robots can affect environmental pollution by promoting green technology innovation. The transmission path of “industrial robots-green technology innovation-environmental pollution” is formed. Industrial robots are the materialization of technological progress in the field of enterprise R&D. Its impact on green technology innovation is mainly manifested in the following two aspects: Firstly, industrial robots classify known knowledge, which helps enterprises to integrate internal and external knowledge 27 . The development of green technology innovation activities of enterprises is promoted. Secondly, enterprises can simulate existing green technologies through industrial robots. The shortcomings of green technology in each link are found. Based on this, enterprises can improve and perfect green technology in a targeted manner. Industrial robots can collect and organize data, which enables enterprises to predict production costs and raw material consumption. Excessive procurement by enterprises can occupy working capital. Inventory backlog leads to warehousing, logistics and other expenses, increasing storage costs 28 . Forecasting the consumption of raw materials allows enterprises to purchase precisely, preventing over-procurement and inventory backlog, thereby reducing the use of working capital and storage costs 29 . The production cost of enterprises is reduced. Enterprises have more funds for green technology research and development.

The continuous innovation of green technology is helpful to solve the problem of environmental pollution. Firstly, green technology innovation helps use resources better 30 , lowers dependence on old energy, and reduces environmental damage. Secondly, green technology innovation promotes the greening of enterprises in manufacturing, sales and after-sales 31 . The emission of pollutants in production process is reduced. Finally, green technology innovation improves the advantages of enterprises in market competition 32 . The production possibility curve expands outward, which encourages enterprises to carry out intensive production. Based on the above analysis, this paper proposes hypothesis 2.

Hypothesis 2

Industrial robots can reduce environmental pollution through green technology innovation.

Mediating effect of employment skill structure

Industrial robots can affect environmental pollution through employment skill structure. The transmission path of “industrial robots-employment skill structure-environmental pollution” is formed. Industrial robots have substitution effect and creation effect on the labor force, improving the employment skill structure. Regarding the substitution effect, enterprises use industrial robots to complete simple and repetitive tasks to improve production efficiency, which crowds out low-skilled labor 6 . Regarding the creation effect, industrial robots create a demand for new job roles that matches automation, such as robot engineers, data analysts, machine repairers, which increases the number of highly skilled labor 33 . The reduction of low-skilled labor and increase of high-skilled labor improve employment skill structure.

High-skilled labor is reflected in the level of education 34 . Its essence is to have a higher level of skills and environmental awareness, which is the key to reducing environmental pollution. Compared with low-skilled labor, high-skilled labor has stronger ability to acquire knowledge and understand skills, which improves the efficiency of cleaning equipment and promotes emission reduction. The interaction and communication between highly skilled labor is also crucial for emission reduction. The excessive wage gap between employees brings high communication costs, which hinders the exchange of knowledge and technology between different employees. The increase in the proportion of high-skilled labor can solve this problem and improve the production efficiency of enterprises 35 . The improvement of production efficiency enables more investment in emission reduction research, decreasing pollutant emissions. Based on the above analysis, this paper proposes hypothesis 3.

Hypothesis 3

Industrial robots can reduce environmental pollution by optimizing employment skills structure.

Model construction and variable selection

Model construction, panel data model.

The panel data model is a significant statistical method, first introduced by Mundlak 36 . Subsequently, numerous scholars have used this model to examine the baseline relationships between core explanatory variables and explained variables 37 . To test the impact of industrial robots on environmental pollution, this paper sets the following panel data model:

In formula ( 1 ), Y it is the explained variable, indicating the degree of environmental pollution in region i in year t . IR it is the core explanatory variable, indicating the installation density of industrial robots in region i in year t . X it is a series of control variables, including economic development level (GDP), urbanization level (URB), industrial structure (EC), government intervention (GOV) and environmental regulation (ER). \(\lambda i\) is the regional factor. \(\varphi t\) is the time factor. \(\varepsilon it\) is the disturbance term.

Mediating effect model

To test the transmission mechanism of industrial robots affecting environmental pollution, this paper sets the following mediating effect model:

In formula ( 2 ), M is the mediating variable, which mainly includes green technology innovation and employment skill structure. Formula ( 2 ) measures the impact of industrial robots on mediating variables. Formula ( 3 ) measures the impact of intermediary variables on environmental pollution. According to the principle of mediating effect 38 , the direct effect \(\theta 1\) , mediating effect \(\beta 1 \times \theta 2\) and total effect \(\alpha 1\) satisfy \(\alpha 1 = \theta 1 + \beta 1 \times \theta 2\) .

Panel quantile model

The panel quantile model was first proposed by Koenke and Bassett 39 . It is mainly used to analyze the impact of core explanatory variables on the explained variables under different quantiles 40 . To empirically test the heterogeneous impact of industrial robots on environmental pollution under different levels of environmental pollution, this paper sets the following panel quantile model:

In formula ( 4 ), \(\tau\) represents the quantile value. \(\gamma 1\) reflects the difference in the impact of industrial robots on environmental pollution at different quantiles. \(\gamma 2\) indicates the different effects of control variables at different quantiles.

Variable selection

Explained variable.

The explained variable is environmental pollution. Considering the timeliness and availability of data, this paper selects industrial wastewater discharge, industrial SO 2 emissions and industrial soot emissions as indicators of environmental pollution.

Explanatory variable

According to production theory, industrial robots can enhance production efficiency 41 . Efficient production implies reduced energy wastage, which in turn decreases the emission of pollutants. Industrial robots can upgrade pollution control equipment, heightening the precision in pollution treatment and reducing pollutant discharge. Referring to Acemoglu and Restrepo 25 , this paper selects the installation density of industrial robots as a measure. The specific formula is as follows:

In formula ( 5 ), Labor ji is the number of labor force in industry j in region i . IR jt is the stock of industrial robot use in industry j in the year t .

Mediating variable

Green technology innovation. Industrial robots can increase the demand for highly-skilled labor 42 , subsequently influencing green technology innovation. Compared to ordinary labor, highly-skilled labor possesses a richer knowledge base and technological learning capability, improving the level of green technology innovation. Green technology innovation can improve energy efficiency 43 , reducing pollution generated by energy consumption. The measurement methods of green technology innovation mainly include three kinds: The first method is to use simple technology invention patents as measurement indicators. Some of technical invention patents are not applied to the production process of enterprise, they cannot fully reflect the level of technological innovation. The second method is to use green product innovation and green process innovation as measurement indicators. The third method is to use the number of green patent applications or authorizations as a measure 44 . This paper selects the number of green patent applications as a measure of green technology innovation.

Employment skill structure. The use of industrial robots reduces the demand for labor performing simple repetitive tasks and increases the need for engineers, technicians, and other specialized skilled personnel, improving the employment skill structure 45 . Compared to ordinary workers, highly-skilled laborers typically have a stronger environmental awareness 46 . Such environmental consciousness may influence corporate decisions, prompting companies to adopt eco-friendly production methods, thus reducing environmental pollution. There are two main methods to measure the structure of employment skills: One is to use the proportion of employees with college degree or above in the total number of employees as a measure. The other is to use the proportion of researchers as a measure. The educational level can better reflect the skill differences of workers. This paper uses the first method to measure the employment skill structure.

Control variable

Economic development level. According to the EKC hypothesis 47 , in the initial stage of economic development, economic development mainly depends on input of production factors, which aggravates environmental pollution. With the continuous development of economy, people begin to put forward higher requirements for environmental quality. The government also begins to adopt more stringent policies to control environmental pollution, which can reduce the level of environmental pollution. According to Liu and Lin 48 , This paper uses per capita GDP to measure economic development level.

Urbanization level. The improvement of urbanization level has both positive and negative effects on pollution. Urbanization can improve the agglomeration effect of cities. The improvement of agglomeration effect can not only promote the sharing of public resources such as infrastructure, health care, but also facilitate the centralized treatment of pollution. The efficiency of environmental governance is improved 49 . The acceleration of urbanization can increase the demand for housing, home appliances and private cars, which increases pollutant emissions 50 . This paper uses the proportion of urban population to total population to measure the level of urbanization.

Industrial structure. Industrial structure is one of the key factors that determine the quality of a country’s environmental conditions 51 . The increase in the proportion of capital and technology-intensive industries can effectively improve resource utilization efficiency and improve resource waste 52 . This paper selects the ratio of the added value of the tertiary industry to the secondary industry to measure industrial structure.

Government intervention. Government intervention mainly affects environmental pollution from the following two aspects: Firstly, the government can give high-tech, energy-saving and consumption-reducing enterprises relevant preferential policies, which promotes the development of emission reduction technologies for these enterprises 53 . Secondly, the government strengthens environmental regulation by increasing investment in environmental law enforcement funds, thus forcing enterprises to save energy and reduce emissions 54 . This paper selects the proportion of government expenditure in GDP to measure government intervention.

Environmental regulation. The investment in environmental pollution control is conducive to the development of clean and environmental protection technology, optimizing the process flow and improving the green production efficiency of enterprises 55 . Pollutant emissions are reduced. This paper selects the proportion of investment in pollution control to GDP to measure environmental regulation.

Data sources and descriptive statistics

This paper selects the panel data of 30 provinces in China from 2006 to 2019 as the research sample. Among them, the installation data of industrial robots are derived from International Federation of Robotics (IFR). The data of labor force and employees with college degree or above are from China Labor Statistics Yearbook . Other data are from the China Statistical Yearbook . The descriptive statistics of variables are shown in Table 3 . Considering the breadth of application and the reliability of analysis capabilities, this paper uses Stata 16 for regression analysis.

Results analysis

Spatial and temporal characteristics of environmental pollution and industrial robots in china, environmental pollution.

Figure  3 a shows the overall trend of average industrial wastewater discharge in China from 2006 to 2019. From 2006 to 2019, the discharge of industrial wastewater shows a fluctuating downward trend, mainly due to the improvement of wastewater treatment facilities and the improvement of treatment capacity. Figure  3 b shows the changing trend of average industrial wastewater discharge in 30 provinces of China from 2006 to 2019. Industrial wastewater discharge in most provinces has declined. There are also some provinces such as Fujian, Guizhou and Qinghai, which have increased industrial wastewater discharge. Their emission reduction task is very arduous.

Industrial wastewater discharge from 2006 to 2019.

Figure  4 a shows the overall trend of average industrial SO 2 emissions in China from 2006 to 2019. From 2006 to 2019, industrial SO 2 emissions shows a fluctuating downward trend, indicating that air pollution control and supervision are effective. Figure  4 b shows the trend of average industrial SO 2 emissions in 30 provinces of China from 2006 to 2019. Similar to industrial wastewater, industrial SO 2 emissions decrease in most provinces.

Industrial SO 2 emissions from 2006 to 2019.

Figure  5 a shows the overall trend of average industrial soot emissions in China from 2006 to 2019. Different from industrial wastewater and industrial SO 2 , the emission of industrial soot is increasing year by year. From the perspective of governance investment structure, compared with industrial wastewater and industrial SO 2 , the investment proportion of industrial soot is low. From the perspective of source, industrial soot mainly comes from urban operation, industrial manufacturing and so on. The acceleration of urbanization and the expansion of manufacturing scale have led to an increase in industrial soot emissions. Figure  5 b shows the trend of industrial soot emissions in 30 provinces in China from 2006 to 2019. The industrial soot emissions in most provinces have increased.

Industrial soot emissions from 2006 to 2019.

Figure  6 shows the spatial distribution characteristics of industrial wastewater, industrial SO 2 and industrial soot emissions. The three types of pollutant emissions in the central region are the largest, followed by the eastern region, and the three types of pollutant emissions in the western region are the smallest. Due to resource conditions and geographical location, the central region is mainly dominated by heavy industry. The extensive development model of high input and consumption makes its pollutant emissions higher than the eastern and western regions. The eastern region is mainly capital-intensive and technology-intensive industries, which makes its pollutant emissions lower than the central region. Although the leading industry in the western region is heavy industry, its factory production and transportation scale are not large, which produces less pollutants.

Spatial distribution characteristics of industrial wastewater, industrial SO 2 and industrial soot.

Industrial robots

Figure  7 a shows the overall trend of installation density of industrial robots in China from 2006 to 2019. From 2006 to 2019, the installation density of industrial robots in China shows an increasing trend year by year. The increase of labor cost and the decrease of industrial robot cost make enterprises use more industrial robots, which has a substitution effect on labor force. The installation density of industrial robots is increased. Figure  7 b shows the trend of installation density of industrial robots in 30 provinces of China from 2006 to 2019. The installation density of industrial robots in most provinces has increased. Among them, the installation density of industrial robots in Guangdong Province has the largest growth rate. The installation density of industrial robots in Heilongjiang Province has the smallest growth rate.

Installation density of industrial robots from 2006 to 2019.

Figure  8 shows the spatial distribution characteristics of installation density of industrial robots. The installation density of industrial robots in the eastern region is the largest, followed by the central region, and the installation density of industrial robots in the western region is the smallest. The eastern region is economically developed and attracts lots of talents to gather here, which provides talent support for the development of industrial robots. Advanced technology also leads to the rapid development of industrial robots in the eastern region. The economy of western region is backward, which inhibits the development of industrial robots.

Spatial distribution characteristics of industrial robots.

Benchmark regression results

Table 4 reports the estimation results of the ordinary panel model. Among them, the F test and LM test show that the mixed OLS model should not be used. The Hausman test shows that the fixed effect model should be selected in the fixed effect model and random effect model. This paper selects the estimation results of the fixed effect model to explain.

Regarding the core explanatory variable, industrial robots have a significant negative impact on the emissions of industrial wastewater, industrial SO 2 and industrial soot. Specifically, industrial robots have the greatest negative impact on industrial soot emissions, with a coefficient of -0.277 and passing the 1% significance level. The negative impact of industrial robots on industrial wastewater discharge is second, with an estimated coefficient of -0.242, which also passes the 1% significance level. The negative impact of industrial robots on industrial SO 2 emissions is the smallest, with an estimated coefficient of -0.0875 and passing the 10% significant level. Compared with industrial wastewater and SO 2 , industrial robots have some unique advantages in reducing industrial soot emissions. Firstly, in terms of emission sources, industrial soot emissions mainly come from physical processes such as cutting. These processes can be significantly improved through precise control of industrial robots. Industrial SO 2 comes from the combustion process. Industrial wastewater originates from various industrial processes. It is difficult for industrial robots to directly control these processes. Secondly, in terms of source control and terminal treatment, industrial robots can reduce excessive processing and waste of raw materials, thereby controlling industrial soot emissions at the source. For industrial SO 2 and industrial wastewater, industrial robots mainly play a role in terminal treatment. Since the terminal treatment of industrial SO 2 and industrial wastewater often involves complex chemical treatment processes, it is difficult for industrial robot technology to fully participate in these processes. This makes the impact of industrial robots in the field of industrial SO 2 and industrial wastewater more limited than that in the field of industrial soot.

Regarding the control variables, the level of economic development has a significant inhibitory effect on industrial SO 2 emissions. The higher the level of economic development, the stronger the residents’ awareness of environmental protection, which constrains the pollution behavior of enterprises. The government also adopts strict policies to control pollutant emissions. The impact of urbanization level on the discharge of industrial wastewater, industrial SO 2 and industrial soot is significantly negative. The improvement of urbanization level can improve the efficiency of resource sharing and the centralized treatment of pollutants, reducing environmental pollution. The industrial structure significantly reduces industrial SO 2 and industrial soot emissions. The upgrading of industrial structure not only reduces the demand for energy, but also improves the efficiency of resource utilization. The degree of government intervention only significantly reduces the discharge of industrial wastewater. The possible reason is that to promote economic development, the government invests more money in high-yield areas, which crowds out investment in the environmental field. Similar to the degree of government intervention, environmental regulation has a negative impact on industrial wastewater discharge. The government’s environmental governance investment has not given some support to the enterprise’s clean technology research, which makes the pollution control investment not produce good emission reduction effect.

Mediation effect regression results

Green technology innovation.

Table 5 reports the results of intermediary effect model when green technology innovation is used as an intermediary variable. Industrial robots can have a positive impact on green technology innovation. For every 1% increase in the installation density of industrial robots, the level of green technology innovation increases by 0.722%. After adding the green technology innovation, the estimated coefficient of industrial robots has decreased, which shows that the intermediary variable is effective.

In the impact of industrial robots on industrial wastewater discharge, the mediating effect of green technology innovation accounts for 8.17% of the total effect. In the impact of industrial robots on industrial SO 2 emissions, the mediating effect of green technology innovation accounts for 11.8% of the total effect. In the impact of industrial robots on industrial soot emissions, the mediating effect of green technology innovation accounts for 3.72% of the total effect.

Employment skill structure

Table 6 reports the results of intermediary effect model when the employment skill structure is used as an intermediary variable. Industrial robots have a positive impact on the employment skill structure. For every 1% increase in the installation density of industrial robots, the employment skill structure is improved by 0.0837%. Similar to green technology innovation, the intermediary variable of employment skill structure is also effective.

In the impact of industrial robots on industrial wastewater discharge, the mediating effect of employment skill structure accounts for 6.67% of the total effect. In the impact of industrial robots on industrial SO 2 emissions, the mediating effect of employment skill structure accounts for 20.66% of the total effect. In the impact of industrial robots on industrial soot emissions, the mediating effect of employment skill structure accounts for 15.53% of the total effect.

Robustness test and endogeneity problem

Robustness test.

To ensure the robustness of the regression results, this paper tests the robustness by replacing core explanatory variables, shrinking tail and replacing sample. Regarding the replacement of core explanatory variables, in the benchmark regression, the installation density of industrial robots is measured by the stock of industrial robots. Replacing the industrial robot stock with the industrial robot installation quantity, this paper re-measures the industrial robot installation density. Regarding the tail reduction processing, this paper reduces the extreme outliers of all variables in the upper and lower 1% to eliminate the influence of extreme outliers. Regarding the replacement of samples, this paper removes the four municipalities from the sample. The estimation results are shown in Table 7 . Industrial robots still have a significant negative impact on environmental pollution, which confirms the robustness of benchmark regression results.

Endogeneity problem

Logically speaking, although the use of industrial robots can reduce environmental pollution, there may be reverse causality. Enterprises may increase the use of industrial robots to meet emission reduction standards, which increases the use of industrial robots in a region. Due to the existence of reverse causality, there is an endogenous problem that cannot be ignored between industrial robots and environmental pollution.

To solve the impact of endogenous problems on the estimation results, this paper uses the instrumental variable method to estimate. According to the selection criteria of instrumental variables, this paper selects the installation density of industrial robots in the United States as the instrumental variable. The trend of the installation density of industrial robots in the United States during the sample period is similar to that of China, which is consistent with the correlation characteristics of instrumental variables. The application of industrial robots in the United States is rarely affected by China’s economic and social factors, and cannot affect China’s environmental pollution, which is in line with the exogenous characteristics of instrumental variables.

Table 8 reports the estimation results of instrumental variable method. Among them, the column (1) is listed as the first stage regression result. The estimated coefficient of instrumental variable is significantly positive, which is consistent with the correlation. Column (2), column (3) and column (4) of Table 8 are the second stage regression results of industrial wastewater, industrial SO 2 and industrial soot emissions as explanatory variables. The estimated coefficients of industrial robots are significantly negative, which again verifies the hypothesis that industrial robots can reduce environmental pollution. Compared with Table 4 , the absolute value of estimated coefficient of industrial robots is reduced, which indicates that the endogenous problems caused by industrial robots overestimate the emission reduction effect of industrial robots. The test results prove the validity of the instrumental variables.

Panel quantile regression results

Traditional panel data models might obscure the differential impacts of industrial robots at specific pollution levels. To address this issue, this paper uses a panel quantile regression model to empirically analyze the effects of industrial robots across different environmental pollution levels.

Table 9 shows that industrial robots have a negative impact on industrial wastewater discharge. With the increase of the quantile of industrial wastewater discharge, the regression coefficient of industrial robots shows a W-shaped change. Specifically, when the industrial wastewater discharge is in the 0.1 quantile, the regression coefficient of industrial robot is − 0.229, and it passes the 1% significant level. When the industrial wastewater discharge is in the 0.25 quantile, the impact of industrial robots on industrial wastewater discharge is gradually enhanced. Its regression coefficient decreases from − 0.229 to − 0.256. When the industrial wastewater discharge is in the 0.5 quantile, the regression coefficient of industrial robot increases from − 0.256 to − 0.152. When the industrial wastewater discharge is at the 0.75 quantile, the regression coefficient of industrial robot decreases from − 0.152 to − 0.211. When the industrial wastewater discharge is in the 0.9 quantile, the regression coefficient of industrial robot increases from − 0.211 to − 0.188. For every 1% increase in the installation density of industrial robots, the discharge of industrial wastewater is reduced by 0.188%.

When industrial wastewater discharge is at a low percentile, the use of industrial robots can replace traditional production methods, reducing energy waste and wastewater discharge. As industrial wastewater discharge increases, the production process becomes more complex. Industrial robots may be involved in high-pollution, high-emission productions, diminishing the robots’ emission-reducing effects. When industrial wastewater discharge reaches high levels, pressured enterprises seek environmentally friendly production methods and use eco-friendly industrial robots to reduce wastewater discharge. As wastewater discharge continues to rise, enterprises tend to prioritize production efficiency over emission control, weakening the negative impact of industrial robots on wastewater discharge. When wastewater discharge is at a high percentile, enterprises should balance production efficiency and environmental protection needs, by introducing eco-friendly industrial robots to reduce wastewater discharge.

Table 10 shows that with the increase of industrial SO 2 emission quantile level, the negative impact of industrial robots on industrial SO 2 emissions gradually increases. Specifically, when industrial SO 2 emissions are below 0.5 quantile, the impact of industrial robots on industrial SO 2 emissions is not significant. When the industrial SO 2 emissions are above 0.5 quantile, the negative impact of industrial robots on industrial SO 2 emissions gradually appears.

When industrial SO 2 emissions are at a low percentile, the application of industrial robots primarily aims to enhance production efficiency, not to reduce SO 2 emissions. Enterprises should invest in the development of eco-friendly industrial robots, ensuring they are readily available for deployment when a reduction in industrial SO 2 emissions is necessary. As industrial SO 2 emissions continue to rise, both the government and the public pay increasing attention to the issue of SO 2 emissions. To meet stringent environmental standards, enterprises begin to use industrial robots to optimize the production process, reduce reliance on sulfur fuels, and consequently decrease SO 2 emissions. Enterprises should regularly evaluate the emission reduction effectiveness of industrial robots, using the assessment data to upgrade and modify the robots’ emission reduction technologies.

Table 11 shows that with the increase of industrial soot emissions quantile level, the negative impact of industrial robots on industrial soot emissions gradually weakens. Specifically, when industrial soot emissions are below 0.75 quantile, industrial robots have a significant negative impact on industrial soot emissions. This negative effect decreases with the increase of industrial soot emissions. When the industrial soot emissions are above 0.75 quantile, the negative impact of industrial robots on industrial soot emissions gradually disappears.

When industrial soot emissions are at a low percentile, they come from a few sources easily managed by industrial robots. As industrial soot emissions increase, the sources become more diverse and complex, making it harder for industrial robots to control. Even with growing environmental awareness, it may take time to effectively use robots in high-emission production processes and control industrial soot emissions. Enterprises should focus on researching how to better integrate industrial robot technology with production processes that have high soot emission levels. The government should provide financial and technical support to enterprises, assisting them in using industrial robots more effectively for emission reduction.

Figure  9 intuitively reflects the trend of the regression coefficient of industrial robots with the changes of industrial wastewater, industrial SO 2 and industrial soot emissions. Figure  9 a shows that with the increase of industrial wastewater discharge, the regression coefficient of industrial robots shows a W-shaped trend. Figure  9 b shows that with the increase of industrial SO 2 emissions, the regression coefficient of industrial robots gradually decreases. The negative impact of industrial robots on industrial SO 2 emissions is gradually increasing. Figure  9 c shows that with the increase of industrial soot emissions, the regression coefficient of industrial robots shows a gradual increasing trend. The negative impact of industrial robots on industrial soot emissions has gradually weakened. Figure  9 a, b and c confirm the estimation results of Tables 9 , 10 and 11 .

Change of quantile regression coefficient.

Heterogeneity analysis

Regional heterogeneity.

This paper divides China into three regions: Eastern, central and western regions according to geographical location. The estimated results are shown in Table 12 . The industrial robots in eastern region have the greatest negative impact on three pollutants, followed by central region, and the industrial robots in western region have the least negative impact on three pollutants. The use of industrial robots in eastern region far exceeds that in central and western regions. The eastern region is far more than central and western regions in terms of human capital, technological innovation and financial support. Compared with central and western regions, the artificial substitution effect, upgrading of sewage treatment equipment and improvement of energy utilization efficiency brought by industrial robots in eastern region are more obvious.

Time heterogeneity

The development of industrial robots is closely related to policy support 56 . In 2013, the Ministry of Industry and Information Technology issued the “ Guiding Opinions on Promoting the Development of Industrial Robot Industry ”. This document proposes: By 2020, 3 to 5 internationally competitive leading enterprises and 8 to 10 supporting industrial clusters are cultivated. In terms of high-end robots, domestic robots account for about 45% of the market share, which provides policy support for the development of industrial robots. Based on this, this paper divides the total sample into two periods: 2006–2012 and 2013–2019, and analyzes the heterogeneous impact of industrial robots on environmental pollution in different periods. The estimation results are shown in Table 13 . Compared with 2006–2012, the emission reduction effect of industrial robots during 2013–2019 is greater.

The use of industrial robots can effectively reduce environmental pollution, which is consistent with hypothesis 1. This is contrary to the findings of Luan et al. 22 , who believed that the use of industrial robots would exacerbate air pollution. The inconsistency in research conclusions may be due to differences in research focus, sample size, and maturity of industrial robot technology. In terms of research focus, this paper mainly focuses on the role of industrial robots in reducing pollutant emissions during industrial production processes. Their research focuses more on the energy consumption caused by the production and use of industrial robots, which could aggravate environmental pollution. In terms of sample size, the sample size of this paper is 30 provinces in China from 2006 to 2019. These regions share consistency in economic development, industrial policies and environmental regulations. Their sample size is 74 countries from 1993 to 2019. These countries cover different geographical, economic and industrial development stages, affecting the combined effect of robots on environmental pollution. In terms of the maturity of industrial robots, the maturity of industrial robot technology has undergone tremendous changes from 1993 to 2019. In the early stages, industrial robot technology was immature, which might cause environmental pollution. In recent years, industrial robot technology has gradually matured, and its operating characteristics have become environmentally friendly. Their impact on environmental pollution has gradually improved. This paper mainly conducts research on the mature stage of industrial robot technology. Their research covers the transition period from immature to mature industrial robot technology. The primary reason that the use of industrial robots can reduce environmental pollution is: The use of industrial robots has a substitution effect on labor force, which reduces the raw materials needed for manual operation. For example, in the industrial spraying of manufacturing industry, the spraying robot can improve the spraying quality and material utilization rate, thereby reducing the waste of raw materials by manual operation. Zhang et al. 57 argued that energy consumption has been the primary source of environmental pollution. Coal is the main energy in China, and the proportion of clean energy is low 58 . In 2022, clean energy such as natural gas, hydropower, wind power and solar power in China accounts for only 25.9% of the total energy consumption, which can cause serious environmental pollution problems. Industrial robots can promote the use of clean energy in industrial production and the upgrading of energy structure 24 . The reduction of raw materials and the upgrading of energy structure can control pollutant emissions in front-end production. On September 1, 2021, the World Economic Forum (WEF) released the report “ Using Artificial Intelligence to Accelerate Energy Transformation ”. The report points out that industrial robots can upgrade pollution monitoring equipment and sewage equipment, which reduces pollutant emissions in end-of-pipe treatment. Ye et al. 59 also share the same viewpoint.

The use of industrial robots can reduce environmental pollution through green technology innovation, which is consistent with hypothesis 2. Industrial robots promote the integration of knowledge, which helps enterprises to carry out green technology innovation activities. Meanwhile, Jung et al. 60 suggested that industrial robots can lower production costs for companies, allowing them to invest in green technology research. The level of green technology innovation is improved. Green technology innovation reduces environmental pollution through the following three aspects: Firstly, the improvement of energy utilization efficiency. China’s utilization efficiency of traditional energy sources such as coal is not high. The report of “ 2013-Global Energy Industry Efficiency Research ” points out that China’s energy utilization rate is only ranked 74th in the world in 2013. Low energy efficiency brings serious environmental pollution problems 61 . Du et al. 62 found that the innovation of green technologies, such as clean coal, can enhance energy efficiency and decrease environmental pollution. Secondly, the production of green products. Green technology innovation accelerates the green and recyclable process of production, thereby reducing the pollutants generated in production process. Thirdly, the improvement of enterprise competitive advantage. Green technology innovation can enable enterprises to gain greater competitive advantage in green development 63 . The supply of environmentally friendly products increases, which not only meets the green consumption needs of consumers, but also reduces the emission of pollutants.

Industrial robots can reduce environmental pollution by optimizing the structure of employment skills, which is consistent with hypothesis 3. Autor et al. 64 contended that industrial robots would replace conventional manual labor positions, reducing the demand for low-skilled labor. Industrial robots represent the development of numerical intelligence. With the continuous development of digital intelligence, the demand for high-skilled labor in enterprises has increased. Koch et al. 65 demonstrated that the use of industrial robots in Spanish manufacturing firms leads to an increase in the number of skilled workers. In February 2020, the Ministry of Human Resources and Social Security, the State Administration of Market Supervision and the National Bureau of Statistics jointly issues 16 new professions such as intelligent manufacturing engineering and technical personnel, industrial Internet engineering and technical personnel, and virtual reality engineering and technical personnel to the society. These new occupations increase the demand for highly skilled labor. The reduction of low-skilled labor and increase of high-skilled labor optimize the structure of employment skills. The optimization of employment skill structure narrows the wage gap between employees, reducing the communication cost of employees. Employees learn and exchange technology with each other, which not only improves the absorption capacity of clean technology. It also improves the production efficiency of enterprises and increases corporate profits, so that enterprises can use more funds for clean technology research and development, thereby reducing environmental pollution.

Conclusions and policy recommendations

Based on the panel data of 30 provinces in China from 2006 to 2019, this paper uses the panel data model and mediating effect model to empirically test the impact of industrial robots on environmental pollution and its transmission mechanism. This paper uses panel quantile model, regional samples and time samples to further analyze the heterogeneous impact of industrial robots on environmental pollution. The conclusions are as follows: (1) Industrial robots can significantly reduce environmental pollution. For every 1% increase in industrial robots, the emissions of industrial wastewater, industrial SO 2 , and industrial dust and smoke decrease by − 0.242%, − 0.0875%, and − 0.277%. This finding is contrary to that of Luan et al. 22 , who argued that the use of industrial robots exacerbates air pollution. The results of this paper provide a contrasting perspective, highlighting the potential value of industrial robots in mitigating environmental pollution. (2) Industrial robots can reduce environmental pollution by improving green technology innovation level and optimizing employment skills structure. In the impact of industrial robots on industrial wastewater discharge, the mediating effect of green technology innovation accounts for 8.17% of total effect. The mediating effect of employment skill structure accounts for 6.67% of total effect. In the impact of industrial robots on industrial SO 2 emissions, the mediating effect of green technology innovation accounts for 11.8% of total effect. The mediating effect of employment skill structure accounts for 20.66% of total effect. In the impact of industrial robots on industrial soot emissions, the mediating effect of green technology innovation accounts for 3.72% of total effect. The mediating effect of employment skill structure accounts for 15.53% of total effect. While Obobisa et al. 66 and Zhang et al. 67 highlighted the role of green technological innovation in addressing environmental pollution. Chiacchio et al. 68 and Dekle 69 focused on the effects of industrial robots on employment. The mediating impact of technology and employment in the context of robots affecting pollution hasn’t been addressed. Our research provides the first in-depth exploration of this crucial intersection. (3) Under different environmental pollution levels, the impact of industrial robots on environmental pollution is different. Among them, with the increase of industrial wastewater discharge, the impact of industrial robots on industrial wastewater discharge shows a “W-shaped” change. With the increase of industrial SO 2 emissions, the negative impact of industrial robots on industrial SO 2 emissions is gradually increasing. On the contrary, with the increase of industrial soot emissions, the negative impact of industrial robots on industrial soot emissions gradually weakens. (4) Industrial robots in different regions and different periods have heterogeneous effects on environmental pollution. Regarding regional heterogeneity, industrial robots in eastern region have the greatest negative impact on environmental pollution, followed by central region, and western region has the least negative impact on environmental pollution. Regarding time heterogeneity, the negative impact of industrial robots on environmental pollution in 2013–2019 is greater than that in 2006–2012. Chen et al. 5 and Li et al. 24 both examined the overarching impact of industrial robots on environmental pollution. They did not consider the varying effects of robots on pollution across different regions and time periods. Breaking away from the limitations of previous holistic approaches, our study offers scholars a deeper understanding of the diverse environmental effects of industrial robots.

According to the above research conclusions, this paper believes that the government and enterprises can promote emission reduction through industrial robots from the following aspects.

Increase the scale of investment in robot industry and promote the development of robot industry. China’s industrial robot ownership ranks first in the world. Its industrial robot installation density is lower than that of developed countries such as the United States, Japan and South Korea. The Chinese government should give some financial support to robot industry and promote the development of robot industry, so as to effectively reduce environmental pollution. The R&D investment of industrial robots should be increased so that they can play a full role in reducing raw material consumption, improving energy efficiency and sewage treatment capacity.

Give full play to the role of industrial robots in promoting green technology innovation. Industrial robots can reduce environmental pollution through green technology innovation. The role of industrial robots in innovation should be highly valued. The advantages of knowledge integration and data processing of industrial robots should be fully utilized. Meanwhile, the government should support high-polluting enterprises that do not have industrial robots from the aspects of capital, talents and technology, so as to open up the channels for these enterprises to develop and improve clean technology by using industrial robots.

Give full play to the role of industrial robots in optimizing employment skills structure. The use of industrial robots can create jobs with higher skill requirements and increase the demand for highly skilled talents. China is relatively short of talents in the field of emerging technologies. The education department should actively build disciplines related to industrial robots to provide talent support for high-skilled positions. Enterprises can also improve the skill level of the existing labor force through on-the-job training and job competition.

Data availability

The datasets used or analyzed during the current study are available from Yanfang Liu on reasonable request.

Liu, Y. & Dong, F. How technological innovation impacts urban green economy efficiency in emerging economies: A case study of 278 Chinese cities. Resour. Conserv. Recycl. 169 , 105534 (2021).

Article   Google Scholar  

Wang, Y. & Chen, X. Natural resource endowment and ecological efficiency in China: Revisiting resource curse in the context of ecological efficiency. Resour. Policy 66 , 101610 (2020).

Kampa, M. & Castanas, E. Human health effects of air pollution. Environ. Pollut. 151 , 362–367 (2008).

Article   CAS   PubMed   Google Scholar  

Feng, Y., Chen, H., Chen, Z., Wang, Y. & Wei, W. Has environmental information disclosure eased the economic inhibition of air pollution?. J. Clean. Prod. 284 , 125412 (2021).

Article   CAS   Google Scholar  

Chen, Y., Cheng, L. & Lee, C.-C. How does the use of industrial robots affect the ecological footprint? International evidence. Ecol. Econ. 198 , 107483 (2022).

Krenz, A., Prettner, K. & Strulik, H. Robots, reshoring, and the lot of low-skilled workers. Eur. Econ. Rev. 136 , 103744 (2021).

Xu, C., Zhao, W., Zhang, M. & Cheng, B. Pollution haven or halo? The role of the energy transition in the impact of FDI on SO 2 emissions. Sci. Total Environ. 763 , 143002 (2021).

Article   ADS   CAS   PubMed   Google Scholar  

Yuan, H. et al. Influences and transmission mechanisms of financial agglomeration on environmental pollution. J. Environ. Manag. 303 , 114136 (2022).

Liu, G., Dong, X., Kong, Z. & Dong, K. Does national air quality monitoring reduce local air pollution? The case of PM 2.5 for China. J. Environ. Manag. 296 , 113232 (2021).

Ren, S., Hao, Y. & Wu, H. Digitalization and environment governance: Does internet development reduce environmental pollution?. J. Environ. Plan. Manag. 66 , 1533–1562 (2023).

Zhao, J., Zhao, Z. & Zhang, H. The impact of growth, energy and financial development on environmental pollution in China: New evidence from a spatial econometric analysis. Energy Econ. 93 , 104506 (2021).

Wang, H. & Liu, H. Foreign direct investment, environmental regulation, and environmental pollution: An empirical study based on threshold effects for different Chinese regions. Environ. Sci. Pollut. Res. 26 , 5394–5409 (2019).

Albulescu, C. T., Tiwari, A. K., Yoon, S.-M. & Kang, S. H. FDI, income, and environmental pollution in Latin America: Replication and extension using panel quantiles regression analysis. Energy Economics 84 , 104504 (2019).

Li, K., Fang, L. & He, L. How population and energy price affect China’s environmental pollution?. Energy Policy 129 , 386–396 (2019).

Liang, L., Wang, Z. & Li, J. The effect of urbanization on environmental pollution in rapidly developing urban agglomerations. J. Clean. Prod. 237 , 117649 (2019).

Sharma, R., Shahbaz, M., Kautish, P. & Vo, X. V. Does energy consumption reinforce environmental pollution? Evidence from emerging Asian economies. J. Environ. Manag. 297 , 113272 (2021).

Chen, F., Wang, M. & Pu, Z. The impact of technological innovation on air pollution: Firm-level evidence from China. Technol. Forecast. Soc. Change 177 , 121521 (2022).

Hao, Y., Chen, H. & Zhang, Q. Will income inequality affect environmental quality? Analysis based on China’s provincial panel data. Ecol. Ind. 67 , 533–542 (2016).

Liu, Q., Wang, S., Zhang, W., Zhan, D. & Li, J. Does foreign direct investment affect environmental pollution in China’s cities? A spatial econometric perspective. Sci. Total Environ. 613 , 521–529 (2018).

Article   ADS   PubMed   Google Scholar  

Mughal, N. et al. The role of technological innovation in environmental pollution, energy consumption and sustainable economic growth: Evidence from South Asian economies. Energy Strat. Rev. 39 , 100745 (2022).

Ghobakhloo, M. & Fathi, M. Industry 4.0 and opportunities for energy sustainability. J. Clean. Prod. 295 , 126427 (2021).

Luan, F., Yang, X., Chen, Y. & Regis, P. J. Industrial robots and air environment: A moderated mediation model of population density and energy consumption. Sustain. Prod. Consump. 30 , 870–888 (2022).

Wang, Q., Li, Y. & Li, R. Do industrial robots reduce carbon intensity? The role of natural resource rents and corruption control. Environ. Sci. Pollut. Res. https://doi.org/10.1007/s11356-023-29760-7 (2023).

Li, Y., Zhang, Y., Pan, A., Han, M. & Veglianti, E. Carbon emission reduction effects of industrial robot applications: Heterogeneity characteristics and influencing mechanisms. Technol. Soc. 70 , 102034 (2022).

Acemoglu, D. & Restrepo, P. Robots and jobs: Evidence from US labor markets. J. Polit. Econ. 128 , 2188–2244 (2020).

Liu, J., Liu, L., Qian, Y. & Song, S. The effect of artificial intelligence on carbon intensity: Evidence from China’s industrial sector. Socio Econ. Plan. Sci. 83 , 101002 (2022).

Lee, C.-C., Qin, S. & Li, Y. Does industrial robot application promote green technology innovation in the manufacturing industry?. Technol. Forecast. Soc. Change 183 , 121893 (2022).

Riza, M., Purba, H. H. & Mukhlisin,. The implementation of economic order quantity for reducing inventory cost. Res. Logist. Prod. 8 , 207–216 (2018).

Google Scholar  

Tang, Z. & Ge, Y. CNN model optimization and intelligent balance model for material demand forecast. Int. J. Syst. Assur. Eng. Manag. 13 , 978–986 (2022).

Wang, Q. & Ren, S. Evaluation of green technology innovation efficiency in a regional context: A dynamic network slacks-based measuring approach. Technol. Forecast. Soc. Change 182 , 121836 (2022).

Chang, K., Liu, L., Luo, D. & Xing, K. The impact of green technology innovation on carbon dioxide emissions: The role of local environmental regulations. J. Environ. Manag. 340 , 117990 (2023).

Tu, Y. & Wu, W. How does green innovation improve enterprises’ competitive advantage? The role of organizational learning. Sustain. Prod. Consum. 26 , 504–516 (2021).

Dauth, W., Findeisen, S., Südekum, J. & Woessner, N. German robots-the impact of industrial robots on workers (2017).

Berger, N. & Fisher, P. A well-educated workforce is key to state prosperity. Economic Policy Institute 22 , 1–14 (2013).

Bourke, J. & Roper, S. AMT adoption and innovation: An investigation of dynamic and complementary effects. Technovation 55 , 42–55 (2016).

Mundlak, Y. On the pooling of time series and cross section data. Econometrica J. Econom. Soc. 46 , 69–85 (1978).

Article   MathSciNet   MATH   Google Scholar  

Sun, B., Li, J., Zhong, S. & Liang, T. Impact of digital finance on energy-based carbon intensity: Evidence from mediating effects perspective. J. Environ. Manag. 327 , 116832 (2023).

MacKinnon, D. P., Warsi, G. & Dwyer, J. H. A simulation study of mediated effect measures. Multivar. Behav. Res. 30 , 41–62 (1995).

Koenker, R. & Bassett, G. Jr. Regression quantiles. Econometrica J. Econom. Soc. 23 , 33–50 (1978).

Akram, R., Chen, F., Khalid, F., Ye, Z. & Majeed, M. T. Heterogeneous effects of energy efficiency and renewable energy on carbon emissions: Evidence from developing countries. J. Clean. Prod. 247 , 119122 (2020).

Pham, A.-D. & Ahn, H.-J. Rigid precision reducers for machining industrial robots. Int. J. Precis. Eng. Manuf. 22 , 1469–1486 (2021).

Du, L. & Lin, W. Does the application of industrial robots overcome the Solow paradox? Evidence from China. Technol. Soc. 68 , 101932 (2022).

Sun, H., Edziah, B. K., Sun, C. & Kporsu, A. K. Institutional quality, green innovation and energy efficiency. Energy Policy 135 , 111002 (2019).

Wang, X., Su, Z. & Mao, J. How does haze pollution affect green technology innovation? A tale of the government economic and environmental target constraints. J. Environ. Manag. 334 , 117473 (2023).

Tang, C., Huang, K. & Liu, Q. Robots and skill-biased development in employment structure: Evidence from China. Econ. Lett. 205 , 109960 (2021).

Cicatiello, L., Ercolano, S., Gaeta, G. L. & Pinto, M. Willingness to pay for environmental protection and the importance of pollutant industries in the regional economy. Evidence from Italy. Ecol. Econ. 177 , 106774 (2020).

Xie, Q., Xu, X. & Liu, X. Is there an EKC between economic growth and smog pollution in China? New evidence from semiparametric spatial autoregressive models. J. Clean. Prod. 220 , 873–883 (2019).

Liu, K. & Lin, B. Research on influencing factors of environmental pollution in China: A spatial econometric analysis. J. Clean. Prod. 206 , 356–364 (2019).

Wang, Y. & Wang, J. Does industrial agglomeration facilitate environmental performance: New evidence from urban China?. J. Environ. Manag. 248 , 109244 (2019).

Cheng, Z. & Hu, X. The effects of urbanization and urban sprawl on CO 2 emissions in China. Environ. Dev. Sustain. 25 , 1792–1808 (2023).

Hu, W., Tian, J. & Chen, L. An industrial structure adjustment model to facilitate high-quality development of an eco-industrial park. Sci. Total Environ. 766 , 142502 (2021).

Hao, Y. et al. Reexamining the relationships among urbanization, industrial structure, and environmental pollution in China—New evidence using the dynamic threshold panel model. Energy Rep. 6 , 28–39 (2020).

Guo, Y., Xia, X., Zhang, S. & Zhang, D. Environmental regulation, government R&D funding and green technology innovation: Evidence from China provincial data. Sustainability 10 , 940 (2018).

Ouyang, X., Li, Q. & Du, K. How does environmental regulation promote technological innovations in the industrial sector? Evidence from Chinese provincial panel data. Energy Policy 139 , 111310 (2020).

Zhang, W. & Li, G. Environmental decentralization, environmental protection investment, and green technology innovation. Environ. Sci. Pollut. Res. https://doi.org/10.1007/s11356-020-09849-z (2020).

Cheng, H., Jia, R., Li, D. & Li, H. The rise of robots in China. J. Econ. Perspect. 33 , 71–88 (2019).

Zhang, X. et al. Evaluating the relationships among economic growth, energy consumption, air emissions and air environmental protection investment in China. Renew. Sustain. Energy Rev. 18 , 259–270 (2013).

Jia, Z. & Lin, B. How to achieve the first step of the carbon-neutrality 2060 target in China: The coal substitution perspective. Energy 233 , 121179 (2021).

Ye, Z. et al. Tackling environmental challenges in pollution controls using artificial intelligence: A review. Sci. Total Environ. 699 , 134279 (2020).

Jung, J. H. & Lim, D.-G. Industrial robots, employment growth, and labor cost: A simultaneous equation analysis. Technol. Forecas. Soc. Change 159 , 120202 (2020).

Liu, H., Zhang, Z., Zhang, T. & Wang, L. Revisiting China’s provincial energy efficiency and its influencing factors. Energy 208 , 118361 (2020).

Article   PubMed   Google Scholar  

Du, K. & Li, J. Towards a green world: How do green technology innovations affect total-factor carbon productivity. Energy Policy 131 , 240–250 (2019).

Li, G., Wang, X., Su, S. & Su, Y. How green technological innovation ability influences enterprise competitiveness. Technol. Soc. 59 , 101136 (2019).

Autor, D. H., Levy, F. & Murnane, R. J. The skill content of recent technological change: An empirical exploration. Q. J. Econ. 118 , 1279–1333 (2003).

Article   MATH   Google Scholar  

Koch, M., Manuylov, I. & Smolka, M. Robots and firms. Econ. J. 131 , 2553–2584 (2021).

Obobisa, E. S., Chen, H. & Mensah, I. A. The impact of green technological innovation and institutional quality on CO 2 emissions in African countries. Technol. Forecast. Soc. Change 180 , 121670 (2022).

Zhang, M. & Liu, Y. Influence of digital finance and green technology innovation on China’s carbon emission efficiency: Empirical analysis based on spatial metrology. Sci. Total Environ. 838 , 156463 (2022).

Chiacchio, F., Petropoulos, G. & Pichler, D. The impact of industrial robots on EU employment and wages: A local labour market approach (Bruegel working paper, 2018).

Dekle, R. Robots and industrial labor: Evidence from Japan. J. Jpn. Int. Econ. 58 , 101108 (2020).

Download references

Author information

Authors and affiliations.

Harbin Vocational College of Science and Technology, Harbin, 150300, Heilongjiang, People’s Republic of China

Yanfang Liu

You can also search for this author in PubMed   Google Scholar

Contributions

Y.L.: Conceptualization, Resources, Supervision, Methodology, Software. I have read and agreed to the published version of the manuscript.

Corresponding author

Correspondence to Yanfang Liu .

Ethics declarations

Competing interests.

The author declares no competing interests.

Additional information

Publisher's note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Supplementary information., rights and permissions.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Cite this article.

Liu, Y. Impact of industrial robots on environmental pollution: evidence from China. Sci Rep 13 , 20769 (2023). https://doi.org/10.1038/s41598-023-47380-6

Download citation

Received : 24 July 2023

Accepted : 13 November 2023

Published : 26 November 2023

DOI : https://doi.org/10.1038/s41598-023-47380-6

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

By submitting a comment you agree to abide by our Terms and Community Guidelines . If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.

Quick links

• Explore articles by subject
• Guide to authors
• Editorial policies

Sign up for the Nature Briefing: Anthropocene newsletter — what matters in anthropocene research, free to your inbox weekly.

• SUGGESTED TOPICS
• The Magazine
• Newsletters
• Managing Yourself
• Managing Teams
• Work-life Balance
• The Big Idea
• Data & Visuals
• Reading Lists
• Case Selections
• HBR Learning
• Topic Feeds
• Account Settings
• Email Preferences

6 Common Leadership Styles — and How to Decide Which to Use When

• Rebecca Knight

Being a great leader means recognizing that different circumstances call for different approaches.

Research suggests that the most effective leaders adapt their style to different circumstances — be it a change in setting, a shift in organizational dynamics, or a turn in the business cycle. But what if you feel like you’re not equipped to take on a new and different leadership style — let alone more than one? In this article, the author outlines the six leadership styles Daniel Goleman first introduced in his 2000 HBR article, “Leadership That Gets Results,” and explains when to use each one. The good news is that personality is not destiny. Even if you’re naturally introverted or you tend to be driven by data and analysis rather than emotion, you can still learn how to adapt different leadership styles to organize, motivate, and direct your team.

Much has been written about common leadership styles and how to identify the right style for you, whether it’s transactional or transformational, bureaucratic or laissez-faire. But according to Daniel Goleman, a psychologist best known for his work on emotional intelligence, “Being a great leader means recognizing that different circumstances may call for different approaches.”

• RK Rebecca Knight is a journalist who writes about all things related to the changing nature of careers and the workplace. Her essays and reported stories have been featured in The Boston Globe, Business Insider, The New York Times, BBC, and The Christian Science Monitor. She was shortlisted as a Reuters Institute Fellow at Oxford University in 2023. Earlier in her career, she spent a decade as an editor and reporter at the Financial Times in New York, London, and Boston.

Partner Center

Numbers, Facts and Trends Shaping Your World

Read our research on:

Full Topic List

Regions & Countries

• Publications
• Our Methods
• Short Reads
• Tools & Resources

Read Our Research On:

Key facts about Americans and guns

Guns are deeply ingrained in American society and the nation’s political debates.

The Second Amendment to the United States Constitution guarantees the right to bear arms, and about a third of U.S. adults say they personally own a gun. At the same time, in response to concerns such as rising gun death rates and  mass shootings , President Joe Biden has proposed gun policy legislation that would expand on the bipartisan gun safety bill Congress passed last year.

Here are some key findings about Americans’ views of gun ownership, gun policy and other subjects, drawn primarily from a Pew Research Center survey conducted in June 2023 .

Pew Research Center conducted this analysis to summarize key facts about Americans and guns. We used data from recent Center surveys to provide insights into Americans’ views on gun policy and how those views have changed over time, as well as to examine the proportion of adults who own guns and their reasons for doing so.

The analysis draws primarily from a survey of 5,115 U.S. adults conducted from June 5 to June 11, 2023. Everyone who took part in the surveys cited is a member of the Center’s American Trends Panel (ATP), an online survey panel that is recruited through national, random sampling of residential addresses. This way nearly all U.S. adults have a chance of selection. The survey is weighted to be representative of the U.S. adult population by gender, race, ethnicity, partisan affiliation, education and other categories. Read more about the  ATP’s methodology .

Here are the  questions used for the analysis on gun ownership , the questions used for the analysis on gun policy , and  the survey’s methodology .

Additional information about the fall 2022 survey of parents and its methodology can be found at the link in the text of this post.

Measuring gun ownership in the United States comes with unique challenges. Unlike many demographic measures, there is not a definitive data source from the government or elsewhere on how many American adults own guns.

The Pew Research Center survey conducted June 5-11, 2023, on the Center’s American Trends Panel, asks about gun ownership using two separate questions to measure personal and household ownership. About a third of adults (32%) say they own a gun, while another 10% say they do not personally own a gun but someone else in their household does. These shares have changed little from surveys conducted in 2021  and  2017 . In each of those surveys, 30% reported they owned a gun.

These numbers are largely consistent with rates of gun ownership reported by Gallup , but somewhat higher than those reported by NORC’s General Social Survey . Those surveys also find only modest changes in recent years.

The FBI maintains data on background checks on individuals attempting to purchase firearms in the United States. The FBI reported a surge in background checks in 2020 and 2021, during the coronavirus pandemic. The number of federal background checks declined in 2022 and through the first half of this year, according to FBI statistics .

About four-in-ten U.S. adults say they live in a household with a gun, including 32% who say they personally own one,  according to an August report based on our June survey. These numbers are virtually unchanged since the last time we asked this question in 2021.

There are differences in gun ownership rates by political affiliation, gender, community type and other factors.

• Republicans and Republican-leaning independents are more than twice as likely as Democrats and Democratic leaners to say they personally own a gun (45% vs. 20%).
• 40% of men say they own a gun, compared with 25% of women.
• 47% of adults living in rural areas report personally owning a firearm, as do smaller shares of those who live in suburbs (30%) or urban areas (20%).
• 38% of White Americans own a gun, compared with smaller shares of Black (24%), Hispanic (20%) and Asian (10%) Americans.

Personal protection tops the list of reasons gun owners give for owning a firearm.  About three-quarters (72%) of gun owners say that protection is a major reason they own a gun. Considerably smaller shares say that a major reason they own a gun is for hunting (32%), for sport shooting (30%), as part of a gun collection (15%) or for their job (7%). 

The reasons behind gun ownership have changed only modestly since our 2017 survey of attitudes toward gun ownership and gun policies. At that time, 67% of gun owners cited protection as a major reason they owned a firearm.

Gun owners tend to have much more positive feelings about having a gun in the house than non-owners who live with them. For instance, 71% of gun owners say they enjoy owning a gun – but far fewer non-gun owners in gun-owning households (31%) say they enjoy having one in the home. And while 81% of gun owners say owning a gun makes them feel safer, a narrower majority (57%) of non-owners in gun households say the same about having a firearm at home. Non-owners are also more likely than owners to worry about having a gun in the home (27% vs. 12%, respectively).

Feelings about gun ownership also differ by political affiliation, even among those who personally own firearms. Republican gun owners are more likely than Democratic owners to say owning a gun gives them feelings of safety and enjoyment, while Democratic owners are more likely to say they worry about having a gun in the home.

Non-gun owners are split on whether they see themselves owning a firearm in the future. About half (52%) of Americans who don’t own a gun say they could never see themselves owning one, while nearly as many (47%) could imagine themselves as gun owners in the future.

Among those who currently do not own a gun:

• 61% of Republicans and 40% of Democrats who don’t own a gun say they would consider owning one in the future.
• 56% of Black non-owners say they could see themselves owning a gun one day, compared with smaller shares of White (48%), Hispanic (40%) and Asian (38%) non-owners.

Americans are evenly split over whether gun ownership does more to increase or decrease safety. About half (49%) say it does more to increase safety by allowing law-abiding citizens to protect themselves, but an equal share say gun ownership does more to reduce safety by giving too many people access to firearms and increasing misuse.

Republicans and Democrats differ on this question: 79% of Republicans say that gun ownership does more to increase safety, while a nearly identical share of Democrats (78%) say that it does more to reduce safety.

Urban and rural Americans also have starkly different views. Among adults who live in urban areas, 64% say gun ownership reduces safety, while 34% say it does more to increase safety. Among those who live in rural areas, 65% say gun ownership increases safety, compared with 33% who say it does more to reduce safety. Those living in the suburbs are about evenly split.

Americans increasingly say that gun violence is a major problem. Six-in-ten U.S. adults say gun violence is a very big problem in the country today, up 9 percentage points from spring 2022. In the survey conducted this June, 23% say gun violence is a moderately big problem, and about two-in-ten say it is either a small problem (13%) or not a problem at all (4%).

Looking ahead, 62% of Americans say they expect the level of gun violence to increase over the next five years. This is double the share who expect it to stay the same (31%). Just 7% expect the level of gun violence to decrease.

A majority of Americans (61%) say it is too easy to legally obtain a gun in this country. Another 30% say the ease of legally obtaining a gun is about right, and 9% say it is too hard to get a gun. Non-gun owners are nearly twice as likely as gun owners to say it is too easy to legally obtain a gun (73% vs. 38%). Meanwhile, gun owners are more than twice as likely as non-owners to say the ease of obtaining a gun is about right (48% vs. 20%).

Partisan and demographic differences also exist on this question. While 86% of Democrats say it is too easy to obtain a gun legally, 34% of Republicans say the same. Most urban (72%) and suburban (63%) dwellers say it’s too easy to legally obtain a gun. Rural residents are more divided: 47% say it is too easy, 41% say it is about right and 11% say it is too hard.

About six-in-ten U.S. adults (58%) favor stricter gun laws. Another 26% say that U.S. gun laws are about right, and 15% favor less strict gun laws. The percentage who say these laws should be stricter has fluctuated a bit in recent years. In 2021, 53% favored stricter gun laws, and in 2019, 60% said laws should be stricter.

About a third (32%) of parents with K-12 students say they are very or extremely worried about a shooting ever happening at their children’s school, according to a fall 2022 Center survey of parents with at least one child younger than 18. A similar share of K-12 parents (31%) say they are not too or not at all worried about a shooting ever happening at their children’s school, while 37% of parents say they are somewhat worried.

Among all parents with children under 18, including those who are not in school, 63% see improving mental health screening and treatment as a very or extremely effective way to prevent school shootings. This is larger than the shares who say the same about having police officers or armed security in schools (49%), banning assault-style weapons (45%), or having metal detectors in schools (41%). Just 24% of parents say allowing teachers and school administrators to carry guns in school would be a very or extremely effective approach, while half say this would be not too or not at all effective.

There is broad partisan agreement on some gun policy proposals, but most are politically divisive,   the June 2023 survey found . Majorities of U.S. adults in both partisan coalitions somewhat or strongly favor two policies that would restrict gun access: preventing those with mental illnesses from purchasing guns (88% of Republicans and 89% of Democrats support this) and increasing the minimum age for buying guns to 21 years old (69% of Republicans, 90% of Democrats). Majorities in both parties also  oppose  allowing people to carry concealed firearms without a permit (60% of Republicans and 91% of Democrats oppose this).

Republicans and Democrats differ on several other proposals. While 85% of Democrats favor banning both assault-style weapons and high-capacity ammunition magazines that hold more than 10 rounds, majorities of Republicans oppose these proposals (57% and 54%, respectively).

Most Republicans, on the other hand, support allowing teachers and school officials to carry guns in K-12 schools (74%) and allowing people to carry concealed guns in more places (71%). These proposals are supported by just 27% and 19% of Democrats, respectively.

Gun ownership is linked with views on gun policies. Americans who own guns are less likely than non-owners to favor restrictions on gun ownership, with a notable exception. Nearly identical majorities of gun owners (87%) and non-owners (89%) favor preventing mentally ill people from buying guns.

Within both parties, differences between gun owners and non-owners are evident – but they are especially stark among Republicans. For example, majorities of Republicans who do not own guns support banning high-capacity ammunition magazines and assault-style weapons, compared with about three-in-ten Republican gun owners.

Among Democrats, majorities of both gun owners and non-owners favor these two proposals, though support is greater among non-owners. 

Note: This is an update of a post originally published on Jan. 5, 2016 .

• Partisanship & Issues
• Political Issues

About 1 in 4 U.S. teachers say their school went into a gun-related lockdown in the last school year

Striking findings from 2023, for most u.s. gun owners, protection is the main reason they own a gun, gun violence widely viewed as a major – and growing – national problem, what the data says about gun deaths in the u.s., most popular.

1615 L St. NW, Suite 800 Washington, DC 20036 USA (+1) 202-419-4300 | Main (+1) 202-857-8562 | Fax (+1) 202-419-4372 |  Media Inquiries

Research Topics

• Age & Generations
• Coronavirus (COVID-19)
• Economy & Work
• Family & Relationships
• Gender & LGBTQ
• Immigration & Migration
• International Affairs
• Internet & Technology
• Methodological Research
• News Habits & Media
• Non-U.S. Governments
• Other Topics
• Politics & Policy
• Race & Ethnicity
• Email Newsletters

ABOUT PEW RESEARCH CENTER  Pew Research Center is a nonpartisan fact tank that informs the public about the issues, attitudes and trends shaping the world. It conducts public opinion polling, demographic research, media content analysis and other empirical social science research. Pew Research Center does not take policy positions. It is a subsidiary of  The Pew Charitable Trusts .

Copyright 2024 Pew Research Center

Terms & Conditions

Privacy Policy

Cookie Settings

Reprints, Permissions & Use Policy