Getting the measure of business research downloads — and the rise of AI

For Michael Magoulias, three words describe the Social Science Research Network’s work: dissemination, visibility and accessibility. With SSRN offering free online access to research papers and “preprint” drafts, says Magoulias, its commercial director, academic authors can reach new audiences and extend the impact of their research.

In analysis produced by SSRN (see box) for the Financial Times, papers uploaded in the past 12 months are ranked here according to how frequently they have been downloaded by governments, companies and other non-academic users.

This makes it possible to see what topics and forms of research align most closely with societal concerns. “In the past it was purely citations and those are still the most important proxy,” says Magoulias. “But people also want to know they’re changing the public conversation and having an impact in the real world.”

In this year’s analysis, finance and technology dominate the papers downloaded, reflecting the interest from readers including investors, regulators and decision makers in ideas that are topical and have practical applications. Sustainability also features, but more modestly.

Unsurprisingly, the ranking indicates that the topic of the moment is artificial intelligence. Half of the papers in the top 10 list feature AI.

Research on the impact of AI on financial markets is being widely read, according to Magoulias. “People are saying that large language models and AI tools can be used in new ways to generate alpha [excess return],” he says. “That’s a very clear trend.”

Meanwhile, articles exploring the impact of AI on labour markets appear frequently in both the top 10 and the top 50. Some focus on hiring and employment trends; others look into how AI is changing the workplace itself.

For Magoulias, the question is how long the rapid evolution of artificial intelligence will consume the attention of researchers and readers. However, he sees plenty of potential to explore how these technologies are changing society and the economy. “It will be the gift that keeps on giving,” he says.

Among the most downloaded papers (see table below) were:

AI and productivity

Given the focus on artificial intelligence in this year’s ranking, it is no surprise to find this paper heading the top 10 list. “The effects of generative AI on high-skilled work: evidence from three field experiments with software developers” explores something on which economists, policymakers, employers and others are placing high hopes: that AI will spark a global productivity revival.

The academics and researchers, from Princeton, MIT, Pennsylvania: Wharton and Microsoft, assessed whether such hopes are justified when it comes to knowledge workers. They studied the use of Copilot, an AI assistant, by almost 5,000 software developers at Microsoft, Accenture and an unnamed Fortune 100 company over periods of two to four years.

They found that developers using the AI tool experienced productivity gains of more than 26 per cent — lower than the figure seen in other studies. However, the authors argue that because those studies were either observational or conducted in controlled lab-like experiments, they have limitations.

First, the previous studies do not account for the fact AI tools may not be appropriate for all coding tasks. Moreover, since coding is only part of the job, the time saved may not deliver the overall productivity gains seen in lab results. By contrast, this study was conducted in real-world settings, which, say the authors “strongly supports its external validity”.

Faces, personalities and success

While an absence of data prompted a research team to investigate the link between personality and labour market success, the problem was not a shortage of material. It was the dearth of large-scale personality surveys, leaving researchers to rely on studies with small samples or those with larger samples but limited personality proxies.

“Research that shows non-cognitive skills can be important for the labour market,” says Kelly Shue, professor of finance at Yale School of Management. “But there was nothing that was speaking to personality differences in a large sample of highly skilled college graduates.”

The team — including academics from Pennsylvania: Wharton, Reichman: Arison and Indiana: Kelley — took a different approach. Rather than survey-based personality measures the researchers harnessed a rapidly advancing field: use of artificial intelligence to extract personality characteristics from facial images.

In “AI personality extraction from faces: labor market implications”, the research team used AI and LinkedIn microdata to extract the “big five” personality traits (extroversion, agreeableness, openness, conscientiousness and neuroticism) from the facial images of 96,000 MBA graduates.

Some findings came as a surprise. For example, the team found that personality characteristics predicted the ability to secure a place at a top-ranked business school differently from labour market success or pay level.

“Being more extroverted is associated with going to a less competitive school,” says Shue. “Yet, on the whole, extroversion predicts labour market success, especially for men.”

Given the controversies associated with use of algorithms in reviewing resumes, the team trod cautiously. “We are careful not to advocate for this technology in job market screening,” says Shue. “But we thought it was important to study, as it may be on the horizon or already being used.

“I hope this inspires additional research into the role of personality in determining economic success,” she adds.

Weight-loss drugs on the menu

Known as GLP-1s, appetite-suppressing drugs have taken the world by storm. But while academics from Cornell University and Numerator, a Chicago-based market research company, saw plenty of research into the impact on weight loss and health, they saw a gap: data linking adoption of these drugs to decisions shoppers make in the supermarket.

“We wanted to know how these medications were affecting consumer goods purchases,” says Sylvia Hristakeva, assistant professor at Cornell University’s SC Johnson College of Business. “Until we started the research, there was little systematic evidence of how they changed what people buy.”

In their paper “The No-hunger games: how GLP-1 medication adoption is changing consumer food demand”, the team found that use of GLP-1 did change what some US households bought. Those with at least one GLP-1 user reduced grocery spending by 5.3 per cent within six months, with spending of higher-income households cut by 8.2 per cent. Spending at fast-food chains, coffee shops and limited-service restaurants dropped by 8 per cent.

The researchers tracked rising adoption, with more than 16 per cent of US households having at least one GLP-1 user by July 2024. “That level of penetration makes these drugs important for public health but also as a consumer force,” says Hristakeva. The team did not study the impact of the shift in purchasing patterns on food industry strategies. “It’s too early right now to understand how the industry is going to react,” she adds.

One unexpected insight emerged on learning about the different audiences for the paper. While the usual suspects — policymakers, the food industry, other researchers — engaged with the research, it received an unusual level of interest from financial forecasters and investment banks.

“They were trying to use what we had to understand how these consumer shifts would change retail revenues, supply chains and corporate earnings,” she says. “That was the one big surprise for us.”

Carbon mispricing and net zero

The only sustainability-focused article in our top 10, the findings of “Does the carbon premium reflect risk or outperformance?” have implications for global net zero goals.

The authors, from London Business School and Sabanci University in Turkey, explored why companies with more carbon emissions than others enjoy higher stock market returns, something known as the “carbon premium”.

By digging into the link between emissions and positive earnings surprises, the academics aimed to uncover whether the carbon premium resulted from outperformance or risk.

The extent to which financial markets price in carbon transition risk, argue the authors, directly affects the ability to shift to a low-carbon economy. If the risk is heavily discounted, companies have powerful incentives to reduce emissions and investors will pressure them to do so. None of this will happen, say the authors, if markets fail to price in transition risk sufficiently.

The paper suggests that the carbon premium is a result of this mispricing. Essentially, the authors write, this enables companies to get away with contributing to climate change.

“Financial markets may not be fully pricing in carbon transition risk, potentially because of doubts about the likelihood of government action,” they write.