Economics for a Better World

Economists like David Van Dijcke, a Rackham Ph.D. candidate, spend a lot of time on a deceptively simple question, applied to a number of scenarios: What is the expected outcome of a proposed policy? To know this, economists need methods of understanding impact in the most holistic ways possible.

Traditional data collection methods like surveys have limitations in offering policymakers the insights they need to be responsive to emerging crises—especially during disasters and war, when sending survey takers into the field is too dangerous. Working with traditional data distributions have pitfalls too: When analysis focuses on averages, a narrative can emerge that is influenced by the poles of the data set—and if that influence goes unacknowledged, policy makers have an incomplete and often incorrect picture of the issues at hand. 

“Bringing in alternative data, which economists often call ‘unstructured data,’ can give us the best sense of how policies might impact various areas,” Van Dijcke says. “This kind of data can be many things: satellite images, mobile phone data, networks, and other ways to look at full distributions of data to show a larger picture of what’s happening.”

Van Dijcke’s research brings forward frameworks and methods for working with unstructured data in hopes that it can inform policy and have a positive impact on the lives of everyday people. “Economics is usually really about making better policies for society, to make better use of resources,” he says.

According to Van Dijcke, policymakers are called to make some of their most far-reaching, high-stakes decisions under intense uncertainty and time pressure, a reality he first experienced during his internship with the Bank of England’s advanced analytics team during the height of the COVID-19 pandemic.

During COVID, he says, his team used unstructured data in the form of online job postings to produce “a daily update” on the job market—information that traditional sources might only provide monthly. 

“In these situations, it’s important to gather detailed data sources with the most up-to-date information,” he says. “You can’t wait for traditional reports.”

After his internship, during his Ph.D. studies, Van Dijcke consulted with Ipsos, an international market research firm with a public affairs arm that supports governments and institutions during wars, crises, and natural disasters. There, he applied his framework and methods to mobile phone data to gain a better understanding of the economic impacts of internet shutdowns in India.

Each year in Rajasthan, officials shut down mobile internet across much of the state to prevent cheating on a large, state-wide teacher exam. Van Dijcke and colleagues used mobile data and a boundary-based design—comparing areas just inside and outside the shutdown zone—to estimate the economic impact of the shutdowns. The result was stark: a “sharp drop in economic activity,” with losses he describes as massive in a region where transactions and communication were conducted primarily using mobile data. Van Dijcke’s use of unstructured data measured harms precisely enough to challenge blunt policy tools—and to encourage cheaper, more targeted alternatives: “If it comes down to 25 percent of GDP that day, you can afford to implement very high security measures and still come out with a net profit, right?”

While working with Ipsos, Van Dijcke also applied his framework and methods to gain a better understanding of the Ukrainian government’s mobile alert system designed to warn civilians of incoming Russian attacks. 

At the beginning of the war, Van Dijcke observed a behavioral response to the alert system in real time via mobile data: “The alarm goes off and suddenly we see a big jump in people moving around, running for shelter.” 

But the pattern changed as the war continued. “The response to an alert went down,” he says, and the data helped rule out comforting explanations, like people simply getting more efficient. Instead, they documented a fatigue effect in places with extremely frequent alerts. “If you get 20 alerts a day, at some point you stop responding,” he says.

Worse, when Van Dijcke looked at the targeted geographic areas, he saw a strong correlation between mortality in an area and the decreasing response to the alerts. According to Van Dijcke, the alert system prevented about 35 to 45 percent of potential civilian casualties, but if engagement with the alert system could have been maintained, “an additional 8 to 15 percent of casualties could have been avoided.” Their findings were communicated to officials, and Ukraine later adjusted messaging to make alerts more engaging.

Van Dijcke’s new research tackles a politically charged question in the United States: How does having a Democratic versus Republican governor affect a state’s income distribution?

Acknowledging that states differ in many ways besides party control, Van Dijcke’s research specifically leverages data from close elections, where outcomes hinge on a tiny margin. When a candidate wins with, say, 50.5 percent of the vote, “it’s almost random,” he says—close enough that small, chance events could have changed the result. That setup supports a cleaner look at one of Van Dijcke’s driving questions: What would the income distribution have looked like if the other party had narrowly won?

In his application, Van Dijcke finds evidence consistent with what is sometimes called an “equality-efficiency trade-off:” narrowly elected Democratic governors have a negative impact on pre-tax income for the top 10 percent of earners, with suggestive gains at the very bottom of the distribution. An inverse economic impact is true for narrowly elected Republican governors. Regardless of who is in office, middle income households don’t experience much of a shift to their financial picture.

Van Dijcke is careful about what his research means—and what it does not mean. “I don’t have a stance on what the right policy is,” he says. But he argues that credible measurement can “help dispel some notions of what is going on” and give policymakers better evidence within their own priorities.

“The goal with these methods is to get at the truth to some extent,” Van Dijcke says. “Anyone can make stories about what the effects of policies are going to be, but I try to estimate what the impact actually will be—and hopefully that will help policymakers adjust and do things that are ultimately better for the public.”