Natural experiments in economics: Business Line 16th October 2021

 

The work of this year’s Economics Nobel winners is particularly relevant in public policy

David Card, Joshua Angrist and Guido Imbens, the Nobel Prize winners in Economics this year, have made a unique contribution in the field of labour economics, which is based on cause-and-effect related issues. This in turn has relevance in public policy. Their approach can be carried forward to our everyday life too.

Normally when we want to find out the cause and effect of any economic phenomenon, we carry out randomised experiments. Samples are chosen, and the question is posed to the participants without telling them the purpose. But this runs the problem of selection bias and tends to be narrow. To overcome this shortcoming, Card, Angrist and Imbens use what is called the observational data approach for answering questions of causation.

This is within the realm of natural experiments, which involves studying results of any action taken by, say, the government on the stakeholders concerned over a period. Interestingly, the 2019 prize had gone to Abhijit Banerjee, Esther Duflo and Michael Kremer for their work based on the ‘experimental approach’. Evidently, both sets of theories have their place in economics and add value. This is important in economics as distinct from medicine, where clinical trials are based on random trials that are carried out to see if medicines work. Hence a controlled experiment has yielded a preventive vaccine for Covid, but the impact of the pandemic on schooling and children is hard to analyse this way. The canvas is larger, and the experiment involves observing the patterns across countries to study the causative effects which will be within the purview of the work of this year’s winners.

Education versus income

Let us see some of the questions in labour economics which are answered by their work. The first is whether more years of education leads to higher future income? This cannot be done with a fixed set of respondents as it would be a biased one. One must observe over time in specific geographies, after understanding the socio-economic conditions. Their experiments did show that students who studied more earned higher income.

This makes sense as those with better qualifications tend to do better in life. But their focus was mostly on schooling in lesser developed countries. At times this could get distorted if people studied more just because they did not get jobs. But the take here is that this would not normally happen for long as people not interested in studying more would like to join the labour force and earn money as soon as they can.

Another issue which was studied by David Card was whether an increase in minimum wage leads to lower employment. This is a critical question for the labour market which always seeks to find a balance where labour wants a higher wage and companies which would like to be selective in paying higher wage to the more efficient. His analysis shows that employment does not suffer as the minimum wage sets the threshold for producers who must pay the higher amount as per law and there is little choice.

But Card showed that when minimum wage increases labour tends to become more efficient which helps the organisation to increase output. Therefore, labour costs do not get onerous. Also, at this level there would be unskilled labour involved which cannot be substituted by technology unlike other desk jobs or even factory skills that can be replaced. Such a conclusion is drawn by studying trends and patterns over time.

An area which Card has analysed in some depth has been the issue of immigration. Do immigrants create problems to the local labour force? This is one issue which has come up often in the US, especially under the Trump regime that alleged that they tend to take over the jobs of the local population. His analysis shows this does not necessarily happen and it depends on the level of work.

The approach of the trio also involved natural experiments which cannot be influenced by the analyst. It also answers simple questions on, say, a course which must be offered to students in an university. Or it could be a new product that is sold with special offers, including discounts and free samples. By making such announcements, can the students who take the course or consumers who buy the product be influenced? Universities as well as companies must take this chance because there would be students who would take a course even without being specifically made the offer while there could be another set who came in due to the offer while the third set would not be interested anyway. These are the kind of questions that are posed when we deal with cause and effect in such experiments.

Companies are always doing such an exercise and fine tune their marketing plans. Often, they follow the faster path of randomised controlled trials which is a market survey-based approach when new products are launched. When they do not do well it would mean going back to the approach of natural experiments to find reasons for things not working out.

The larger question is whether corporates can wait for long to get these results before launching a product? Often the product is differentiated and cannot be compared with stories of other companies. Such natural experiments tend to work better for government-driven programmes where experiences from the past provide signals for the future. Therefore, from the point of view of public policy these theories are very relevant and more likely to be pursued.