Nightlight and Demonetisation - An Intertwining saga

Let’s start today by looking at a quaint little graph that Bloomberg published some time back. Look carefully at the axis. Is it the night light intensity(a complex word for total lights on at night in a country )? Yes. Demonetisation was announced on 8th Nov 2016. But why something as trivial as night light intensity is showing a dip during that period? How can something as simple as the number of night lights and something as complex as our entire economy be correlated? Or are they even correlated or is this some random coincidence? Worry not, you guys just have to continue reading and all the mysteries will be Demystified!!

Our economy has two sectors, namely the Formal sector, and the informal sector. The formal sector includes all jobs with normal hours and regular wages, and are recognized as income sources on which income tax must be paid. The informal sector, on the other hand, includes all the jobs which are not recognized as normal income sources, and on which taxes are not paid ex. vendors selling vegetables, workers, small businesses, etc.

It’s surprising that no one knows the actual size of the Indian informal workforce, least of all the government. The Economic survey of 2018-19, says almost 93% of the total workforce is informal. But a survey by Niti Aayog estimates it as 85%. What is the source of such information? The Economic Survey does not mention it. The Niti Aayog does and cites a 2014 report, 'OECD India Policy Brief: Education and Skills', which, in turn, is silent on its source of information.

Nevertheless, Gross Domestic Product (GDP) is the most important variable for economic analysis and policymaking. But the measurement of GDP in developing country like India is quite uncertain. Lack of statistical capacity, mismeasurement of economy, and the existence of informal economy, among others, can all subject GDP to substantial revision. This problem is more acute for developing country like ours where the data collection process is less sophisticated, and GDP is very often revised.  On top of this, state-level GDP i.e Gross State Domestic Product (GSDP) is often published with a substantial lag. Further district-level data on GDP are not compiled for all districts, and even if they are compiled, it is not done on a regular basis. 

Now suppose you are a policymaker and you implement a policy. In the current system, it is very hard to assess the performance of the economy at a regional level. This all has encouraged analysts and policymakers to supplement the available data on the economy with new and innovative data sources including the use of big data to produce better estimates. In recent years, economists are turning to satellite imagery as an alternative measure to assess the state of the economy even before revised estimates are available.

But why a regression model based on night light intensity predicts GDP? In other words, why night light intensity correlates high with GDP, especially in a developing country like India?

To answer this, let me take you through a case study.

The above figure compares satellite images of nighttime light intensity for mainland China and the lower 48 states of the United States. While both of them became brighter in 2013, China’s transformation is clearly visible. Variation in nighttime lights may thus contain useful information on China’s real economic growth. In contrast, the united states were already bright enough in 1992. The small change in the intensity of lights over this period may not correspond to economic growth, most of which is likely to happen on the scientific and technological frontier rather than on infrastructure development².

Let me explain this with an example.

There is a town A. People of this town have farming as their main source of income. As people earn less, their spending on consumer goods like balcony lamps is limited. Also, in an under-developed town well-functioning public infrastructure like street lights, park lamps etc. is often absent. Now a local steel company decides to set up a steel plant on the outskirts of this town, providing new job opportunities for the people. Due to jobs at the factory, the income of many will increase. As income will increase, people’s spending on necessary consumer goods like balcony lamps, and other electrical appliances will see a rise. Also, the owner of the factory has invested in street lights and other public infrastructure of the town as the factory's CSR activity. Now if we compare nighttime lights intensity per person before and after the factory has set up, clearly, the after case will dominate over before. Likewise, in developing country like ours, most of the time, increment in GDP happens due to infrastructure development and that is why nighttime light is apt proxy to capture this.

So now you might be asking, as nighttime light requires electricity so why not use electricity consumption directly as a proxy?

Technically if we can obtain electricity consumption data with the same frequency as nighttime light and with the same geographical fineness then electricity consumption will be the more accurate proxy than nighttime light. But, in reality, procuring this daily is very hard. Electricity business in India is broadly divided into 3 segments namely generation, transmission and distribution. There are about 51 electricity distribution companies and we get our electricity through one of these distribution companies. All of them have methods for continuously monitoring electricity data but the data is highly confidential. Even if someone gets access to it, aggregating this data at the national level will be time-consuming and a difficult task. On the other hand, satellite data are available at the monthly frequency with a resolution of 30 - 5cm /pixel. So people have been inclined towards using nighttime light as a proxy to economic growth.

To summarise, If the nighttime light intensity per person increases, consumption of goods and services will increase, thus economic growth, GDP will increase.

Now back to the case study. The World Bank has developed a regression model in which they have calibrated the sum of nightlight intensity obtained from satellites to the total economic activity (formal plus informal) for countries. But again, how did they estimate total informal activity? As it was mentioned earlier that measuring it is very difficult. Again, there are many proxies to measure this. Some of them are

• Currency Demand Analysis³: Most of the transactions in the informal economy are made with cash. So an increase in the activity of the informal economy will lead to an increase in cash transactions. Thus increasing the demand for cash. Monitoring this demand will give a good estimate on the size of the informal economy.

• Electricity Consumption³: This method defines the growth rate of the informal economy as “the difference between the growth of official or measured GDP and the growth rate of electricity consumption.” Ex. If the growth in GDP was 5% but the growth in electricity was 10%. This method assumes the extra 5% in electrical consumption was due to the presence of the informal economy.

Both the proxies mentioned above are not 100% accurate and have received a lot of criticism from the economic community. Nevertheless, they are time-consuming. So creating the training dataset for the model will be a time-consuming task. But once the model is trained, the prediction will be fast as it only requires nighttime light as an input.

Coming back to our original demonetisation graph, at the aggregate level, the comparison suggests only a small dip in the economic activity occurred after demonetization. However, areas that were more informal experienced a drop in GDP in the range of 4.7 to 7.3 percentage points(That’s a lot!).

A major portion of the Indian workforce is part of the informal economy. They use cash to meet all their expenses and demonetization has resulted in a lot of them losing their jobs. According to CMIE’s Consumer Pyramids Household Surveys (CPHS), approximately 1.5 million jobs were lost during the final quarter of the financial year 2016-17.

Take the case of this news-report in The Times of India on the glass and bangle industry in Firozabad in Uttar Pradesh. The report quotes a union leader as saying: "Around 65% of bangle factories are shut and the majority of those which are operating are exploiting workers, paying them wages of Rs 30 to Rs 50 per day from Rs 400 per day that an average worker used to get".

Or take the case of this survey, carried out by All India Manufacturers Organisation(AIMO), which has projected a loss in revenue of 55% and a drop in employment of 60% before March 2017, for small businesses. And According to Indian Express reports "The AIMO represents over 3 lakh micro, small and medium scale industries(MSME) engaged in manufacturing and export activities." So the loss in revenue at this scale was huge!

That’s too much negative talk. But there were some positive outcomes too!

"What demonetization has done for digital payments is more than what any other initiative could achieve before." The absence of liquid cash has led to people making transactions using mobile wallets or account transfers. They have also switched to virtual payments wallets like Paytm even in tier-II and tier-III cities.

Demonetisation has pushed our economy towards formalization. The demonetization year 2016-2017 recorded a 29% increase in a number of income tax filers as compared to the financial year 2015-2016, followed by a 25% increase in the financial year 2017-2018. This all led to an increase in indirect tax collection from 7.4 Lakh Crore in the financial year 2015-2016 to 10 Lakh Crore in the financial year 2017-2018.

So, did demonetization work for us? Based on the current knowledge and reports I have, I believe the negative side dominates over the positive. If you feel like my views are biased, please reach me out or do comment. I would love to listen to your arguments and maybe even correct myself.

Enough of politics, coming back to science, nighttime light can be used as a proxy to measure the economic growth of developing countries. But it's clearly not a universal proxy as it lacks the capability of capturing economic growth on technological and scientific frontiers.

I myself came across this in a news about two months back which discussed the use of nighttime light to predict the performance of the economy. I was blown by the fact that how can something as trivial as the nighttime light intensity can reveal so much about the economy. But this example is just a use case of the much broader family “Alternative Datasets”. Investors, hedge funds make millions by using these kinds of data-driven insights along with their conventional ones. Who could have imagined, something as trivial as nighttime light can tell so much about the economy? But again, out there people have taken satellite imagery to the next level. People track manufacturing units using drones and satellites to get the signal of changing stock prices. More on this in the next blog.

Eagerly waiting to hear your views! and Do not forget to subscribe to get regular updates.

Rushikesh Rathod.

rushikeshbrathod@gmail.com