High frequency has its merits but such data is prone to being misinterpreted in ways that may mislead policy
The new wave of covid infections led by the Omicron variant of the Sars-CoV-2 virus has once again restricted mobility in most parts of the country. People are spending more time at home, rather than at work places or retail stores or parks, going by the anonymized data released for India by Google based on mobile phone locations.
Such new forms of data have been immensely useful to track economic activity on a regular basis during the pandemic. They give us a quick sense of what is happening in the economy, rather than having to wait for government statisticians to collate the usual quarterly estimates of activity across the economy with structured surveys as well as administrative data. These established procedures can sometimes come in too slowly to understand a rapidly evolving situation. The past two years have seen various types of big data step into the gap.
However, it is also important to be careful while using big data to make anything more than tentative guesses about economic growth. In this context, let us stick with mobility data for some time.
There have been several studies that show how Google mobility data moves in sync with underlying economic activity in normal times. But what about uncertain times such as these?
Economists at the Organisation For Economic Co-operation and Development (OECD) took a close look at the link between mobility trends and economic growth. They examined the data for 51 countries in the second and third quarters of 2020, and 43 countries in the fourth quarter of that year. The researchers found out that the impact of mobility indicators on economic growth weakened with each successive quarter. There are two possible reasons for this. First, policymakers have learnt to target specific types of economic activity, rather than impose blanket bans on movement. Second, both citizens as well as enterprises have learnt how to adapt to newer forms of work and leisure.
What is common to both factors is that we have adapted to the pandemic as a society. The relationship between mobility indicators and economic growth has not been the same across time. In more technical language, the regression coefficients have changed as economic agents have largely learnt to live with the virus. The OECD data indicates that a 10 percentage point change in mobility was associated with a 2.2 percentage-point change in economic growth in the second and third quarters of 2020, but only a 0.9 percentage-point change in economic growth in the fourth quarter of that year. That is a sharp drop.
It also means that an analyst trying to forecast the impact of changes in mobility data on quarterly economic activity based on the coefficients for the first and second quarters of 2020 will get very different results from another analyst working with the coefficient from the fourth quarter. This fact matters when trying to estimate the impact of the third wave on the Indian economy, especially when mobility data is an important element being considered.
There are challenges with some of the other types of big data as well.
Consider night lights, which are now being used by some economists as a proxy for economic activity. The input data on lights kept on past sundown comes from a number of satellites that can help pick up the intensity of lights generated by human beings in an area. Ayush Patnaik, Ajay Shah, Anshul Tayal and Susan Thomas of research firm xKDR have highlighted, in a recent working paper, that clouds interfere with the way data on night lights can be captured by satellites that hover kilometres above the land. The four researchers show that there is a downward bias in readings during cloudy months, and have built an algorithm to at least partially correct this downward bias.
Another problem with interpreting big data is the context in which it is read. The pandemic months, for example, have seen consumer demand shift from services to goods in many categories, either because of lockdowns or fear of stepping out. The e-way bills generated when goods move around the country are a very useful advance indication of economic activity. But such bills do not need to be generated for services. So a broad shift in demand from services to goods will likely lead to a sharper rise in e-way bills than can be explained by total economic activity. Similarly, a shift of demand back to services may show that e-way bills growth has come down. This does not necessarily mean that the economy has slowed down.
The easy availability of new forms of big data is definitely an opportunity for economic analysts. The examples cited above— on mobility data, night lights and e-way bills —have been used to show that such data still needs to be handled with care when it is used to make broader guesses about the economy as a whole. Some of these data points are also aggregated into easy to follow indexes of current economic activity. They have been immensely useful during the pandemic, but aggregating data with different reporting frequencies as well as potentially different seasonality patterns is a tricky problem.
The use of big data in economic analysis is welcome. There is no turning back. However, there are still a bunch of analytical puzzles that need to be sorted out.
Niranjan Rajadhyaksha is CEO and Senior Fellow at Artha India Research Advisors, and a member of the academic advisory board of the Meghnad Desai Academy of Economics.
Subscribe to Mint Newsletters
* Enter a valid email
* Thank you for subscribing to our newsletter.
Never miss a story! Stay connected and informed with Mint.
our App Now!!