Why natural experiments are a natural choice for a Nobel prize

While the Nobel Committee’s abiding focus has been on economic theory, the 2021 prize throws the limelight on how empirical analysis is done. As argued by our previous article for Mint Views, ‘Exploit the potential of natural experiments for policymaking’, 16 September 2021 (bit.ly/3ANVFoR), natural experiments seem to be the ‘natural successors’ of randomized control trials (RCTs). This view seems to have resonance in Swedish corridors: In 2019, the Nobel Prize for Economics was given to Abhijit Banerjee, Esther Duflo and Michael Kremer for their work on RCTs.

Two papers—one by Card on the effect of minimum-wage laws on employment and another by Angrist on the effect of schooling on earnings—have become the talk of the town. The duo of Alan B. Krueger and David Card conducted a famous natural experiment by observing employment trends at fast-food restaurants in New Jersey and Pennsylvania before and after the former raised the applicable minimum wage while the latter didn’t. They discovered, foxing conventional text-book economists, that employment in New Jersey’s fast-food outlets rose slightly relative to employment in Pennsylvania’s. This triggered a churn in thinking about the way economists would imagine their theoretical models working in the real world; after intense contestations over the hypotheses involved, it was finally concluded that the raising of New Jersey’s minimum wage actually had no impact on total employment in the state’s fast-food industry. The key takeaway from the episode was that the reigning thoughts of by-the-book economists about the way the world functions in the context of economic matters merited a relook. Tersely put, there was a need to verify if standard economic models of demand-and-supply really work as they have been conceptualized.

Krueger and Angrist confirmed an age-old notion that the lifetime earnings of people who attend school for longer are higher. But their use of a natural experiment set-up—compulsory-schooling laws that allowed drop-outs only after age 16 resulted in a situation where students born in the first quarter of the calendar year would be able to leave school earlier than students born in the fourth quarter—made it possible to isolate a group of individuals who stayed in school longer than others merely because of a school-leaving-rule. So, the economic outcomes of those who attended school for longer would be attributed not to any inherent difference in motivation or capability, but to an accident of birth. Thus, the usual factors that bedevil a clean flow of causal logic from longer schooling to higher earnings (as observed differences could also be on account of superior capabilities or academic inclinations) are purged this way and we have credible evidence that links time spent in school and the economic outcomes thereafter.

The two studies above broadly explain why natural experiments occupy the place of pre- eminence that they have come to acquire in policy spheres. They provide evidence rooted in real-world conditions, thereby helping formulate policies that are bottom-up (as opposed to top-down approaches that flow from an economist’s make-do world of theory), and they provide robust evidence that is less contestable.

It may often seem that situations amicable for utilization as natural experiments are unusual. However, research done over the last 30 years shows that a keen eye on geography, existing or altered policy frameworks and/or landmark events can provide us opportunities to ‘naturally’ divide our sample into treatment and control groups to establish important causal relationships. For example, Fayer and Sacerdote (bit.ly/2YSnkbe) argued that wind speed and direction distinguished the length of colonial experience in islands throughout Atlantic, Pacific and Indian Oceans, throwing light on how the number of years spent as a European colony influences contemporary gross domestic product per capita. In the sphere of political science, Pettersson-Lidbom (bit.ly/3j5UTxL) used a natural-experiment setting in Finland and Sweden to show that a large-sized legislature lowers the size of government, contrary to conventional wisdom. Chen, Harford and Lin (bit.ly/3BLWyj5) studied the impact of analysts on corporate governance. Our previous op-ed spoke of other examples. Today, natural experiments are no longer the preserve of labour economics, but are used widely in various other research fields.

Work by Angrist and Imbens in 1994 also holds significance in radically altering researchers’ approach to empirical questions using data from natural experiments. Though closer to reality, these need thorough tests of internal validity (for an untangled link of logic from cause to effect), which have increased with the technique’s expanding horizon. Imbens’ elaborate work on carving out and simplifying the empirical methodology for natural experiments has revolutionized empirical research in several domains across myriad disciplines. Institutionalizing the use of natural experiments for policy purposes seems a fitting tribute to those who have perfected this wonder tool and thereby broadened our understanding of the world.

Anshuman Kamila & Yashaswini Saraswat are, respectively, assistant director at the economic division, ministry of finance, and assistant director at the development monitoring and evaluation office, Niti Aayog.  These are the authors’ personal views.