Businesses, governments, and their respective leaders utilize academic writing to shape policy and strategy. This is especially the case in the field of economics, where research informs central bank policy, antitrust policy, and the design of taxes and regulation. Economists rely heavily on formal modelling, quantitative analysis, and peer review to make implied policy prescriptions. However economists, like all people, are subject to partisan or political preferences of their own.
While economists are expected to be non-partisan in their recommendations, new research from Bruce Kogut, Sanford C. Bernstein & Co. Professor of Leadership and Ethics at Columbia Business School, has found that economists’ political affiliations actually bleed through into these implied policy prescriptions.
He and his co-researchers, Columbia University Professor Suresh Naidu and University of Maryland Professor Zubin Jelveh, used machine learning to develop a "political slant" index that measures the left-right ideological orientation of economists based on their published work, campaign contributions, and signed letters.
“Economics has frequent debates between left and right, which is healthy," Kogut says.
He noted that the diversity of thought in economics was a strength in shaping policy and strategy, since it allows for real debate and discussion on important issues, rather than a field dominated by a single ideological perspective. Other social sciences do not have the ideological diversity of economists. Still, the discovery has significant implications for economists' policy recommendations, such as on tax rates or the minimum wage.
Partisan Prescriptions
To create their political slant index, the researchers gathered a large dataset of articles, papers, and signed letters. Next, they utilized machine learning, training a computer model to look for patterns in the words economists used in their writing that predicted real world political behavior, like campaign contributions or petition signing.
Republicans were more likely to use terms like "gold standard" or "right-to-work,” whereas Democrats used terms like "gender equality" or "minimum wage." The researchers found that an economist's political slant can move their policy estimates by 12-15 percent – a “big number” that shouldn’t be overlooked, according to Kogut.
"People are human, and the doing of science often entails inspiration,” Kogut says. “That inspiration could be your political beliefs."
For example, Kogut and his team determined that going from the most left-wing economist’s estimate of the taxable top income elasticity to the most right-wing economist’s estimate decreases the optimal tax rate from 77 percent to 60 percent. Similarly, the researchers were able to find a significant correlation between an economist's political bias and the estimates they produce on minimum wage policies — more left-leaning economists tended to have higher estimates for the effects of minimum wage increases, while more right-leaning economists had lower estimates.
Preventing Bias
Awareness of political bias in economic research is key for policymakers, industry leaders, and the public to critically evaluate the conclusions being drawn, Kogut noted. Business leaders especially should know that even experts may bring subjective perspectives that can lead to more informed and balanced decisions, but also more bias.
As Kogut and his team found, machine learning can be an excellent tool for uncovering such bias in humans. However, AI can also be a double-edged sword that makes it easier for researchers to manipulate data to get desired results.
“AI is blowing up the potential for manipulation,” Kogut says. He stressed the importance for continuing to strengthen institutional measures to address AI-related bias, such as requiring transparency in methodologies and pre-registering research plans. Stronger governance in academic journals and institutions can also help reduce bias, by way of requiring open access to data sharing.
The Need for Ideological Diversity
Though economists have their biases, economics as a field has more political diversity than many other social sciences, according to Kogut. Lack of political diversity in academia can heavily skew the research landscape, leading to some political perspectives being wholly underrepresented in research and policy prescriptions.
Having a diversity of political views represented in economic research can help provide a more balanced, less biased perspective on policy issues, Kogut noted. Plus, maintaining these diverse political perspectives can ultimately strengthen the credibility of the field.
“Science happens within a context of humans, and they behave as humans might be expected to do,” Kogut says. “The good news is despite political slant in language, there is a lot of discipline imposed by transparency and openly acknowledging, rather than hiding, these biases that can help leaders better question and interpret the findings and policy implications.”
