IZA Newsroom

IZA – Institute of Labor Economics

Mark Fallak

Gender-based expectations influence external demands for parental involvement

by

A new IZA discussion paper explores an everyday but important question: When systems, like schools, reach out to families, who do they contact first—mothers or fathers? The study by Kristy Buzard, Laura Katherine Gee, and Olga B. Stoddard uses a large-scale field experiment to show that mothers are 1.4 times more likely than fathers to be contacted.

The findings shine a light on significant ways that gender roles continue to shape family and workplace dynamics today. The study involved sending emails to over 80,000 school principals across the United States. These messages came from fictional families—each one portrayed as a married, heterosexual couple with a school-aged child. The goal of the email was straightforward: the family was moving to the area and looking for a school for their child.

The message asked the principal to call to discuss school options and provided phone numbers for both parents. The key to the experiment was in the wording of the email. Sometimes the message was neutral—giving both phone numbers but not suggesting who should be contacted. In other cases, the email stated that one parent was more available or less available for a conversation.

The researchers then tracked who got a phone call—mom or dad. In the first version of the email, there was no suggestion of which parent should be called. It simply said, “Can you call one of us to discuss?” and listed the phone numbers for both the mother and the father.

The default to mothers

So, in a situation where there’s no stated preference, do school principals reach out to both parents equally? Not quite. When a call was made in this group, it went to the mother about 60% of the time. This is well above the 50% mark that would be expected if the choice were completely random.

In a second version of the email, the message included a clear statement: the father had a lot of availability to talk. This time, the researchers were checking whether principals would adjust their outreach accordingly and prioritize the more available parent. And they did—to some extent. In this scenario, over 70% of the calls went to the dad. So yes, principals were more likely to contact the father when told he was more available.

But here’s the twist: 26% of the time, they still called the mother, even though the father said he was very available. This suggests that even when dads state that they are clearly available, there’s a strong tendency to contact moms.

A version of the email said that the mother had a lot of availability. Now the outcome was even more skewed. A full 90% of the calls went to the mother. That’s a much stronger shift than in the previous case. When mom was described as the very available parent, principals overwhelmingly chose to contact her. So while they responded to availability in both cases, the bias in favor of contacting mothers was stronger when it aligned with traditional expectations.

Broader implications for gender equality

These differences in call patterns point to deeper societal norms and expectations. The researchers argue that the results are likely just the visible part of a much larger issue. They suspect that similar biases play out in many day-to-day situations—like when schools need to call home because a child is sick or when teachers need help organizing an event. In those cases, moms are probably contacted even more often than dads, even when both parents are equally capable or available.

This pattern of defaulting to mothers as the primary contact point—even when they’re not available—feeds into broader issues of gender inequality. For example, if mothers are always the ones being contacted by schools, they may be the ones more likely to miss work, adjust their schedules, or take time off. Over time, that can have real consequences for their careers and income.

This research helps explain why progress toward gender equality in the workplace has been so slow, even as cultural attitudes have shifted. If society continues to place more caregiving and coordination responsibilities on mothers, women will continue to face extra hurdles when it comes to balancing work and family life.

Addressing unconscious biases

The study doesn’t suggest that principals are intentionally biased. In fact, many probably aren’t even aware of these patterns in their behavior. But that’s exactly the point: many gender-based expectations are so deeply embedded that they operate beneath the surface. They’re part of our habits, assumptions, and unconscious decision-making.

By bringing this issue to light, the researchers hope to start a conversation—not just in schools, but in all workplaces and households—about how gender roles continue to influence our actions in subtle ways. They argue that recognizing and addressing these patterns is a key step toward greater fairness and equality in both the home and the workplace.

Pandemic slowed German firms’ tech progress

by

A new IZA discussion paper by Melanie Arntz, Michael Johannes Böhm, Georg Graetz, Terry Gregory, Florian Lehmer, and Cäcilia Lipowski investigates how German firms adjusted their technology investments before and during the COVID-19 pandemic. The researchers aimed to answer two key questions: Did the crisis accelerate the adoption of advanced, “frontier” technologies (the latest, most integrated tech tools)? Or did it primarily shift investment focus towards solutions like remote work technologies, without boosting overall advanced technology use?

The study defines frontier technologies as state-of-the-art office tools (e.g., cloud computing, automated marketing, AI in business software) and advanced production equipment (e.g., smart robotics, integrated manufacturing systems).

Methodology: A deep dive into firm-level data

The study draws on a large, detailed survey of around 3,000 German firms, enriched by linking survey responses to administrative worker data and other official records. This allowed the researchers to capture:

  • When new technology investments were made (pre- or during-pandemic).
  • Whether these investments were explicitly motivated by the pandemic.
  • How firms’ technology use (office and production) changed over time.

By comparing current practices to data from a 2016 survey, the researchers tracked technology adoption changes over several years.

Key findings: Shifting sands of tech investment

Contrary to expectations, overall investments in frontier technologies actually dropped during the pandemic. For instance, before COVID-19, about 6% of firms annually invested in cutting-edge office technology; this rate nearly halved to roughly 3.6% during the pandemic. A similar, even steeper, decline occurred for production-related technology.

Only a small percentage of firms made their main technology investments specifically in response to the pandemic. Even when COVID-19 was cited as a reason, these pandemic-driven investments were smaller in scope. While additional secondary investments increased the percentage of pandemic-related investments somewhat, they still couldn’t match the scale of pre-crisis investments.

A clear change was firms shifting focus toward technologies supporting remote operations. Investments in communication, collaboration, and remote-work tools significantly increased, becoming crucial for maintaining connectivity and business continuity during lockdowns.

The study also examined the link between technology investments and workforce changes. Firms investing in remote work solutions saw a marked increase in employees working from home. They also relied less on government-supported short-time work schemes. This suggests that while overall tech spending declined, the investments made helped firms adjust and stabilize their workforce during the crisis.

Implications: Long-term growth and policy considerations

The pandemic appears to have slowed the overall pace of frontier technology adoption. Using models to compare outcomes with a no-crisis scenario, the researchers estimate firms lost approximately 1.4 years of technology investment activity relative to normal times. This slowdown indicates that despite the clear need for digital adjustments, large-scale technology investments decreased.

Investments in remote work tools provided an immediate remedy, helping firms adapt to COVID-19 restrictions. However, these changes were typically small-scale and more reversible, not signaling a broader, lasting shift towards a more technologically advanced operating model. The long-term plans of these firms didn’t show a strong, sustained commitment to additional frontier technology investments once the immediate crisis passed.

The findings raise important questions about how economic crises affect long-run growth. In Germany, the slowdown in adopting advanced technologies might contribute to lower productivity growth. The researchers suggest this pattern may call for counter-cyclical innovation policies—policies that boost technology investments during economic downturns to support long-term growth.

Conclusion: A complex picture of technological change

The study concludes that rather than dramatically accelerating advanced technology uptake, the COVID-19 crisis led to an overall reduction in such investments among German firms. While firms did reorient spending toward remote work tools—helping them manage operational challenges—these adjustments weren’t enough to offset the significant loss in broader technology adoption.

Consequently, while short-term adaptations stabilized employment, the long-term trajectory of technological progress may have slowed, with potential implications for future economic growth. This research underscores the complexity of technological change during crises and highlights the need for supportive policies to prevent temporary setbacks from having lasting effects on a nation’s economic dynamism.

The well-being costs of immigration in Europe

by

Despite prevailing media narratives, a new IZA discussion paper by Kelsey J. O’Connor of STATEC Research indicates that the significant increase in immigration to Europe between 1990 and 2019 has not negatively impacted the average well-being of either destination or sending countries. In fact, when considering the substantial gains experienced by immigrants themselves, the overall impact shifts from a net cost to a net benefit.

The study, which analyzed data from 37 European countries over nearly four decades, found that Europe saw an influx of 32 million immigrants between 1990 and 2019. Much of this movement was from East to West, with the Western Bloc experiencing net immigration of approximately 30 million people, while the Eastern Bloc saw a net emigration of about 12 million. The majority of immigrants typically originated from within Europe, with exceptions noted in Northern Europe (Denmark, Finland, Norway, Sweden) and to some extent Southern Europe (Italy, Spain, Portugal, Greece). In Central and Eastern European countries, less than 20 percent of immigrants were from outside Europe on average.

Immigration significantly shaped population dynamics. Eastern Bloc countries collectively experienced a population decline of over 24 million during the period, exceeding the current population of the Netherlands. Conversely, Southern Europe’s population grew by nearly 13 million, with immigrants accounting for 11 million of that increase.

Life satisfaction as a measure of overall well-being

To assess the broader impacts of migration, O’Connor employed a comprehensive subjective measure of well-being: life satisfaction. This approach captures both the economic and non-economic effects of immigration. The rigorous regression analysis revealed no discernible impact of the immigrant share of the population on destination countries’ life satisfaction.  Furthermore, there was no reliable negative impact of the emigrant share on sending countries’ life satisfaction; if anything, a positive effect was observed, likely due to remittances sent home by those who moved away.

While no aggregate well-being costs were found for either destination or sending countries, immigrants themselves experienced considerable benefits. The study found that new immigrants likely experienced a lasting increase of 0.4 points on a 1-to-10 life satisfaction scale on average. This is a substantial gain, comparable to the negative effect of becoming unemployed, and is valued at approximately 30,000 euros for a period of five years, with benefits increasing over longer durations. Unsurprisingly, people tend to migrate from countries with lower life satisfaction to those with higher levels. Importantly, the life satisfaction of immigrants tends to converge with that of natives over time.

The research draws upon data from reputable sources including the United Nations Population Division, the World Bank, and the European Values Study. While acknowledging that effects could vary across subpopulations, the author emphasizes that the aggregate null effects on local populations suggest any potential negative impacts could be offset by redistributing benefits gained elsewhere.

Diversity can enrich non-market experiences

O’Connor highlights a key strength of this study: unlike much previous research that focused on narrower outcomes, this analysis captures all factors individuals deem relevant to their overall lives. This includes considerations such as job competition, fiscal solvency, social cohesion, safety, and the availability of less expensive and more diverse goods and services. The author also notes that diversity can enrich non-market experiences that are otherwise difficult to quantify.

In conclusion, O’Connor advocates for the liberalization of immigration policy, based on the findings that immigration yields substantial benefits to immigrants without imposing aggregate well-being costs on destination or sending countries.

Brexit’s hidden cost: Higher patient mortality in NHS hospitals

by

Brexit has claimed an unexpected victim: patient care in England’s National Health Service (NHS) hospitals. According to a new IZA discussion paper by Henrique Castro-Pires, Kai Fischer, Marco Mello, and Giuseppe Moscelli, hospitals previously reliant on European Union nurses experienced significant declines in care quality following Brexit, leading to approximately 4,454 additional deaths and 8,777 extra unplanned readmissions across England.

The study employs a difference-in-differences approach, comparing hospitals with varying levels of pre-Brexit dependence on EU nurses. This methodology isolates Brexit’s specific impact from other potential factors affecting healthcare quality.

The researchers estimate that Brexit’s immigration restrictions resulted in roughly 1,485 additional patient deaths annually. This alarming increase in mortality appears directly linked to staffing changes forced by the sudden reduction in EU nurse recruitment.

Hospitals responded to the nurse shortage by hiring less-skilled replacements, the authors note. Evidence for this skills downgrade appears in the lower pay grades assigned to new nursing hires in the post-Brexit period.

The research team developed a theoretical framework predicting that Brexit would force hospitals previously dependent on EU nurses to lower their hiring standards, subsequently reducing care quality. This effect would occur regardless of whether overall nurse numbers were maintained.

Importantly, the negative impact varied based on pre-Brexit staffing patterns. Hospitals with higher reliance on EU nurses before the referendum experienced more severe quality declines than those less dependent on European staff.

The researchers took extensive measures to ensure their findings weren’t influenced by confounding variables, controlling for pre-Brexit changes in hospital finances, patient volumes, bed occupancy, and workforce composition.

While nursing was particularly affected, the study suggests higher-paying medical positions faced less severe consequences, as the relative wage advantage of UK employment continued to attract qualified candidates even after Brexit.

This research carries profound policy implications, highlighting how immigration restrictions can severely impact essential services like healthcare. As countries worldwide debate immigration policies, this cautionary tale demonstrates how limiting skilled worker mobility can produce unintended—and potentially deadly—consequences.

How community networks shape elections after a crisis

by

When a natural disaster strikes before an election, it can alter political preferences in profound and often unexpected ways. A new IZA discussion paper by Giovanni Gualtieri, Marcella Nicolini, Fabio Sabatini, and Marco Ventura examines the electoral consequences of the 2009 L’Aquila earthquake, which devastated central Italy just two months before the European Parliament elections. While previous research has shown that incumbent governments often benefit electorally from disaster response, this study highlights a crucial yet overlooked factor: the role of social capital in shaping voter behavior.

Using a unique dataset that combines high-resolution seismic intensity data, electoral results, and indicators of local civic engagement, the authors find that the earthquake significantly influenced voting patterns—but in a highly heterogeneous way. The national government, led by the center-right coalition of Silvio Berlusconi, gained electoral support in the most affected areas, while local administrators saw no such benefit. However, the extent of this electoral boost depended heavily on pre-existing levels of social capital within communities.

How social capital shields communities from political manipulation

The study reveals that where social capital was low—meaning fewer civic associations—and citizens relied entirely on government institutions for aid, the Berlusconi government successfully capitalized on the disaster response, securing a clear electoral advantage.

Conversely, in municipalities with stronger social capital, where local organizations were active in coordinating relief efforts, the electoral effect of the earthquake was negligible. This suggests that when citizens can rely on their own communities for support, they are less likely to reward the government for disaster relief efforts, as they do not perceive state assistance as their only safety net.

This is a crucial insight, as it challenges the common assumption that voters react uniformly to post-disaster aid. Instead, the effect of government response is mediated by social structures: in fragmented communities, political leaders can leverage crises to consolidate electoral support, while in well-connected societies, this mechanism fails to operate in the same way.

A temporary boost, followed by disillusionment

Interestingly, this effect was short-lived. In subsequent elections, support for Berlusconi’s coalition dropped sharply in the very municipalities where it had initially surged—particularly those with low levels of social capital. This suggests a “post-disaster disillusionment effect”: in the short term, voters may be swayed by high-visibility government interventions, but if the long-term recovery process fails to meet expectations, electoral gains can quickly turn into losses. In essence, promises of swift reconstruction
may generate immediate political rewards, but failing to deliver on those promises erodes trust in the long run.

Policy implications: Rethinking disaster response and governance

These findings have important policy implications for governments and international institutions dealing with disaster response and recovery. First, they highlight the importance of investing in social capital before disasters strike. Stronger community networks reduce dependence on government aid, making disaster response more resilient and less prone to political opportunism.

Second, policymakers should recognize that while high-profile emergency interventions may yield short-term political gains, long-term recovery efforts are crucial in maintaining public trust. Governments that fail to sustain reconstruction efforts risk facing a severe electoral backlash once the initial wave of support fades.

Finally, the study’s results suggest that political leaders may have incentives to focus on immediate relief efforts at the expense of long-term resilience strategies. Understanding how voters respond to disasters can help design policies that prioritize sustainable recovery over short-term political gains, ensuring that crisis management serves public welfare rather than electoral interests.

How do future elites view inequality?

by

The United States stands as the most unequal country in the OECD, despite most Americans expressing a preference for more equal wealth distribution. This presents a puzzle: if most citizens want greater equality, why does inequality persist and even grow?

Theories from political science offer one compelling explanation. While the average voter might prefer more redistribution of wealth, those who exert disproportionate influence on policy—business and economic elites—may hold dramatically different views.

For this explanation to hold water, economic elites would need to be more tolerant of inequalities than typical citizens. However, researching the preferences of economic elites has proven challenging, as they are often difficult to reach using standard surveys. In a new IZA discussion paper, Marcel Preuss, Germán Reyes, Jason Somerville, and Joy Wu address this gap by studying the redistributive preferences of tomorrow’s economic leaders—MBA students from an elite Ivy League program.

Finding 1: Future business leaders accept greater inequality

Using an experimental design where impartial spectators (MBA students) choose how to allocate earnings between pairs of workers, the researchers found that MBA students implement substantially more unequal distributions than the average American.

When worker earnings were randomly assigned, MBA students implemented distributions with a Gini coefficient (a common measure of inequality where 0 represents perfect equality and 1 represents maximum inequality) of 0.43, which is notably higher than the 0.36 Gini observed in previous work using representative US samples.

To put this magnitude in perspective, the gap between MBA students and average Americans represents about 35% of the inequality gap between Americans and Scandinavians in analogous experiments.

Finding 2: Future elites are highly responsive to efficiency costs

The most dramatic difference between future business leaders and the general population emerged in their sensitivity to efficiency costs. When redistribution comes with efficiency costs—meaning the total available income decreases when earnings are redistributed—MBA students sharply reduce their redistribution efforts.

Introducing even modest efficiency costs increased the Gini coefficient they implemented by 0.20 points. This responsiveness is an order of magnitude larger than what researchers observe in studies of average Americans, who typically show an elasticity close to zero—meaning they continue to redistribute even when it reduces the total economic pie.

This suggests a fundamental difference in values: while the average American prioritizes fairness concerns over efficiency when considering redistribution, future business leaders place much higher weight on maximizing the total economic output, even if it means accepting greater inequality.

Why this matters

These findings illuminate why the United States maintains high levels of inequality despite most citizens expressing preferences for more equal distributions. The individuals who disproportionately influence policy—current and future economic elites—implement more unequal earnings distributions than the average American and place much greater weight on efficiency considerations.

This research has important implications for understanding policy debates around taxation, social programs, and economic inequality. If those with the most influence over economic policy consistently prioritize efficiency over equality, it becomes clearer why policies aimed at reducing inequality face significant hurdles, even when they have popular support.

As wealth concentration continues to increase in many developed economies, understanding these divergent preferences becomes crucial for addressing one of today’s most pressing economic challenges: how to create broadly shared prosperity in a system where decision-makers may fundamentally value different outcomes than the majority of citizens.

Would you read a story written by a machine?

by

By Martin Abel and Reed Johnson

While much recent research has focused on how people say they prefer creative works made by humans over AI, we don’t yet know whether these stated preferences translate into consumer behavior. And, as the saying goes, talk is cheap. But for the millions of people worldwide employed in creative industries, the concern of whether people will be willing to pay more for human-made art is more than an idle academic question. Amidst the coming avalanche of AI-generated work, it is a question of real livelihoods—and more broadly, a question of what will remain ours in this most human of endeavors.

An experiment with AI-generated creative writing

To investigate these questions, we turned to AI-generated creative writing. We asked OpenAI’s GPT-4  to generate a short story in the style of the critically acclaimed author Jason Brown. We then recruited a nationally representative sample of over 650 people and asked participants to read and assess this story. Crucially, only half the participants were told that the story was written by AI, while the other half were misled into believing it was the work of Jason Brown. This design allows us to isolate the effect of (perceived) authorship and to test if consumers value human over AI writing.

After reading the first half of the AI-generated story, participants were asked to rate the quality of the work along various dimensions, such as whether they found it predictable, emotionally engaging, atmospheric, etc. We also measured participants’ willingness to pay in order to read to the end of the story. To capture a fuller picture of participants’ investment in the story, we included both monetary measures (giving up a portion of their pay for participating) and time measures (their willingness to work doing a routine transcription task).

Subjective assessments vs. actual behavior

So, were there differences between the readers who knew the work was AI-generated versus those who didn’t? The short answer: yes. But a closer analysis reveals some startling results.

To begin with, the group that knew the story was AI-generated had a much more negative assessment of the work, rating it more harshly on dimensions like predictability, authenticity, and atmosphere. These results are largely in keeping with a nascent but growing body of research that shows bias against AI in areas like visual art, music, or poetry. It seems that, at least at present, consumers reflexively ascribe lower judgments of quality to AI-labeled work.

However, notwithstanding these perceptions of lower quality, participants were ready to invest the same amount of money and working time to finish reading the story, whether or not it was labeled as AI. Participants also did not spend less time actually reading the AI-labeled story. When asked afterwards, almost 40% of participants said they would have paid less if the same story had been written by AI versus a human, highlighting that many are not aware of the discrepancies between their subjective assessments and actual choices.

Implications for the future of the creative industry

These findings provide timely evidence that widely documented professed AI biases may not be reliable indicators of willingness to pay for human creative labor. The potential implications for the future of human-created work are profound, especially in market conditions where AI-generated work can be orders of magnitude cheaper to produce. With the technology still in its infancy, AI-made books are already flooding the market, recently prompting the Authors Guild to instate its own labeling guidelines. However, our research raises questions about whether these labels are effective in stemming the tide.

True, attitudes toward AI are still forming, and we may well see a backlash against AI-generated creative works. After all, similar shifts occurred in the wake of mass industrialization, such as the Arts and Crafts Movement. This response may take the form of market segmentation, where some consumers will be willing to pay more based on the process of creation, while others may be interested only in the product.

Regardless of how these scenarios play out, our findings indicate that the road ahead for human creative labor might be more uphill than previous research suggests. At the very least, we see that, while consumers may hold beliefs about the intrinsic value of human labor, many seem unwilling to put their money where their beliefs are.

Strategic Shift at DPS: Implications for IZA

by

On February 24, 2025, the Deutsche Post Foundation released the following statement:

+++

Strategic Realignment of the Deutsche Post Stiftung: Prioritizing Climate and Sustainability

Following the end of funding from DHL Group (formerly Deutsche Post AG), the Deutsche Post Stiftung (DPS) is undergoing a strategic realignment. Under its new name, Stiftung Globale Nachhaltigkeit (SGN), it will focus on climate, nature, and sustainability issues.

As part of this transition, DPS will become a grant-making foundation, supporting external projects and organizations through the Stiftungsfonds Umweltökonomie und Nachhaltigkeit (SUN). This means it will no longer operate its own research institutes. In this context, DPS has made the difficult decision to discontinue operations at the Forschungsinstitut zur Zukunft der Arbeit (IZA) as of December 31, 2025.

At the same time, DPS recognizes the importance of the international IZA research network, which has established itself as a leading and highly respected global hub for labor market research. To preserve this legacy, efforts are underway to explore alternative funding sources, partnerships, and organizational models for the network. The IZA Network Advisory Panel, who were not part of the decision, will be asked to assist in this. The goal is to present the results in June 2025.

DPS fully recognizes the uncertainty this transition creates for all employees at the Bonn-based institute. Individual agreements will be developed to avoid compulsory redundancies and to help minimize the impact on employees’ careers and livelihoods.

The foundation expresses its deep appreciation for the exceptional dedication of all researchers and staff who have shaped IZA over the years. Their work has not only advanced research and international collaboration in labor economics but has also provided valuable contributions to labor market and social policy debates worldwide. Looking back on more than 25 years of achievements supported by the private funding and general leadership of DPS, we are incredibly proud of what has been accomplished. As we move forward, we are committed to supporting efforts to ensure that the spirit and impact of this research community continues into the future.

The President and the Board of Trustees

+++

Award-winning research on the labor market impact of artificial intelligence

by

We are delighted to announce that the paper AI, Task Changes in Jobs, and Worker Reallocation (IZA DP No. 17554) by Christina Gathmann (LISER and IZA), Felix Grimm (LISER), and Erwin Winkler (University of Erlangen-Nürnberg and IZA) been selected for the 2025 IZA Award for Innovative Research on a Pressing Public Issue (IRPPI). As AI continues to transform industries worldwide, the IRPPI Award recognizes research that deepens our understanding of these rapid technological shifts.

The study provides a crucial analysis of how artificial intelligence is reshaping labor markets, focusing on the reallocation of workers due to AI-driven changes in job tasks. It addresses one of today’s most critical economic concerns: the impact of AI on employment structures and worker transitions.

Using novel patent-based measures of AI and robot exposure, the authors analyze individual survey data on job tasks and administrative data on worker careers to understand these shifts. The study finds that while robots primarily reduce routine tasks, AI has a distinct impact by reducing non-routine abstract tasks, such as information gathering, while increasing the demand for ‘high-level’ routine tasks like monitoring processes.

These task shifts mainly occur within detailed occupations and become more pronounced over time. Although displacement effects are relatively small, workers respond by switching jobs, often moving to less AI-exposed industries. The study also highlights differences in wage effects: while low-skilled workers tend to experience wage losses due to AI-induced changes, high-skilled incumbent workers generally see wage gains.

The award-winning research “represents the best of modern labor economics,” according to the selection committee consisting of Kristina McElheran (University of Toronto) and Andrew Oswald (IZA and University of Warwick). The award comes with a prize of 5,000 euros.

+++
Download the paper here: AI, Task Changes in Jobs, and Worker Reallocation

For more information on the labor market implications of AI, see also:

IZA Spotlight newsletter special issue on Automation and AI: The Future of Work?

IZA World of Labor article by Nick Drydakis on AI and Labor Market Outcomes

Can AI solve the peer review crisis in economics?

by

The peer review process in economics suffers from persistent inefficiencies, including a shortage of qualified reviewers and long decision times, often exceeding six months. A new IZA Discussion Paper by Pat Pataranutaporn, Nattavudh Powdthavee, and Pattie Maes examines whether large language models (LLMs) can alleviate these challenges.

To rigorously assess AI’s role, the study conducted a large-scale experiment using 9,030 submissions derived from 30 recently published economics papers. These included research from top-five economics journals, mid-tier and lower-ranked publications, as well as AI-generated papers designed to mimic top-tier research. The researchers systematically varied author attributes—such as institutional affiliation, reputation, and gender—to analyze AI’s decision-making patterns.

Findings show that AI can effectively distinguish between different quality levels, suggesting its potential to reduce editorial workload. However, AI evaluations exhibit systematic biases, favoring submissions from prestigious institutions, well-known economists, and male authors, even when research content is identical. Additionally, AI struggles to differentiate genuine top-tier research from high-quality AI-generated papers, raising concerns about its ability to assess novelty and originality.

The authors advocate for a hybrid peer review model, where AI assists in initial screening but final decisions remain with human reviewers. To ensure fairness, they recommend bias mitigation strategies such as training AI on anonymized data and refining evaluation criteria. The study highlights both the promise and risks of AI in academic publishing, emphasizing the need for careful integration to enhance efficiency without compromising research integrity.