The sense of anticipation ahead of the Sveriges Riksbank Prize in Economic Sciences in memory of Alfred Nobel, citing the full title of the prize, is always evident among academic economists, especially those of us who are expected to be a future laureate as a doctorate. I have had the privilege of working together. student. To add to the sense of occasion, the Economics Prize, incidentally, falls on Thanksgiving Day in Canada, a major national holiday. This year’s prize was awarded jointly to David Card (half a share) and Joshua Angrist and Guido Imbens (a quarter each) for their path-breaking efforts in allowing us to make credible causal claims when analyzing data. The exhibit was up to the task.
In a sense, this year’s award is the flip side of the 2019 award, given to Michael Kramer, Indian-origin Abhijit Banerjee and Esther Duflo. For their work in popularizing the ‘randomized controlled trial’ (RCT), long a staple in the natural sciences in economics research, the trio won in good measure. RCTs allow us to make causal claims by randomly assigning subjects to a ‘control’ or ‘treatment’ group, ensuring that any differences between the two groups can be attributed to the treatment, not the treatment. that for any unseen difference; The theory is that such differences should be averaged once the subjects are randomized.
But RCTs have some major limitations, as your columnist has argued extensively in these pages (‘Experimental turning point in economics’, 30 January 2016), bit.ly/3iVXdaC) A major problem is ‘external validity’: can the finding in one context be replicated in another – a very different one? Equally important, because creating an RCT is not always possible, nor perhaps even ethical, in many situations such an approach cannot address some of the ‘big’ questions in economics, for which we need to use raw, non-random data. What needs to be analyzed, and if it exists, find some way to tease out causation.
Keep in mind that the observed statistical correlation between two variables in the data set is not, in itself, evidence of a causal relationship. Take an example near the house. During the pandemic, my classes switched online, with two hours of pre-recorded lectures followed by ‘live’ one-hour Q&A sessions. Attendance afterward was highly recommended, but not mandatory. Equally, I observed that students who participated and actively participated in discussion sessions performed better on the course. But is it because my discussion session allowed them to perform better? While it is flattering for any professor, it is equally plausible that those who were going to do well chose to participate – what we call ‘reverse causality’. Or, perhaps there were unseen differences between those who attended and those who didn’t, Internet access and reading, time available for study and discussion, poor connectivity, struggling with work and school—rather than those who didn’t—that correlated. could explain. Is it non-random selection, rather than causality, that matters more to the consequences in this case?
Economics is replete with situations where correlations in data lead us to draw a causal conclusion. This can be treacherous in the absence of randomization which, as mentioned, is impossible to achieve in most real-world situations.
David Card’s genius, working with the late Alan Krueger, was to find a clever solution, which was to investigate a real-world situation that presented a useful natural experiment—in this case, two contiguous US states that were otherwise identical. and who shared a common labor market and general macroeconomic situation, but one increased their minimum wage, while the other did not. (Readers can find a more detailed description in the fine write-up on this year’s prize by economist Alex Tabarok marginal revolution blog, bit.ly/3v69bmI, and in Tim Harford Financial Timesson.ft.com/3lCDZZk) Calculating the “differences in differences” before and after the changes in the two jurisdictions allowed them to infer that any differential effects on unemployment were driven by policy changes, not any undiscovered differences.
Similarly, research by Angrist and Imbens, again with Krueger and published in a series of papers, studied important questions such as whether schooling increases people’s earnings, a clear situation where a uni-directional causal link is needed. Any assumption can be problematic. . For example, promising students can study more and earn higher income due to better qualifications. In a seminal paper, Angrist and Kruger asked whether compulsory schooling could lead to an increase in pay, and found a brilliant technique for randomization. Given the oddities of the American school system, students born in late December would be one class behind those born in early January, and laws in some states allowed students to drop out at age 16. The result is that there will be at least some students otherwise nearly identical who received a year more schooling for a purely random reason, and, these students were found to earn higher wages, making a causal claim valid. Gaya. (Again, readers can check out the writings of Tabarok and Harford for more details).
The beauty of these contributions is that they were not based on a complex and technical mathematical or statistical result incomprehensible to the layman, but on a simple and profound intuition that even in our messy and non-random world How can randomness be found? , thus rationalizing the causal inference. Well done!
Vivek Dehjia, Associate Professor of Economics and Philosophy at Carleton University, Canada
Don’t miss a story! Stay connected and informed with Mint.
download
Our App Now!!
.