Sahil Deo, Pooja Verulkar and Sanjana Krishnan, “Fast Data for Faster Decision-making: The Utility of High-frequency Economic Indicators,” ORF Issue Brief No. 372, June 2020, Observer Research Foundation.
“Essentially, all models are wrong, but some are useful.”
Box, George E. P.; Norman R. Draper
It is a little over five months since the first case of COVID-19 was reported in Wuhan, China, and the pandemic has had an enormous impact on people’s lives and livelihoods. Businesses have been forced to shut and many have lost their jobs. Millions of migrant workers in India have been forced to walk hundreds of kilometres to reach their villages after losing their source of income in the cities. Recent estimates of the Centre for Monitoring Indian Economy (CMIE) state that India’s unemployment rate rose to 24.3 percent in the week ended 24 May, slightly higher than the figures recorded in the previous eight weeks (23.3 on 5 April).
Indeed, the pandemic which started as a geographically restricted health crisis is on its way to becoming one of the world’s worst economic crises in recent history. Several months-long lockdowns imposed by countries have triggered deep economic disruptions, domestically and internationally. In a televised address in May, Shaktikanta Das, the governor of the Reserve Bank of India (RBI), said that India’s GDP growth is expected to drop to negative this financial year. While economists are comparing the pandemic-induced recession with the Great Depression, they admit that the toll of COVID-19 could in fact be higher.
Empowering Decision-makers with More Reliable Data
The importance of data in decision-making cannot be overemphasised. More so in times like the ongoing health crisis where there is massive uncertainty. World leaders are making crucial policy decisions every day to arrest the further spread of the pandemic and mitigate its economic impact. Making decisions on issues like the need for more ventilators or extending a lockdown have a direct impact on lives and livelihoods. When the stakes are this high, governments need reliable real-time data to make quick and effective decisions; do they have such data?
Realising the need for real-time[a] health and disease data, various countries have launched their contact-tracing apps to track potential cases of infection based on the location of the user. China, for instance, uses an app which not only colour-codes people based on their risk of infection but also acts as a travel pass to access public places. The real-time data from India’s contact-tracing app, Aarogya Setu, is being used by the government to track infections. Such data on health indicators is not only helping trace individuals at risk but also to plan the delivery of health resources. The UK‘s National Health Service (NHS), for its part, is testing an AI system, developed based on existing data on confirmed cases, that predicts demand for intensive-care beds and ventilators.
A similar real-time understanding of economic outcomes is unavailable due to lack of data in India. Economists and decision-makers are forced to rely on “traditional” indicators alone, which are plagued with issues of timeliness, granularity, difficulty in collection, lack of reliability.
First, due to the current disruption in economic activity, routine data collection and growth forecasting has become difficult, including for indicators like inflation and the Consumer Price Index (CPI). A recent article states that the Indian government is exploring ways to remotely collect economic data; Pravin Srivastava, the country’s chief statistician, has discussed the possibility of institutionalising alternative methods for data collection.
Second, traditional economic indicators like GDP and inflation rate are not frequently updated given the complexity in collecting and computing them. Relying on this data involves waiting for several months, as quarterly data on these indicators are released seven to eight weeks after the end of the reference quarter. While they do provide a fair picture of the state of the economy, they do not necessarily help gauge the near-term impact of an economic policy decision.
Third, the traditional macroeconomic indicators might not always be able to provide geographically granular data, limiting their ability to provide insights at the level of cities and districts. City-level GDP has been a topic on the radar of the Ministry of Housing and Urban Affairs, according to the consultation paperdrafted in February 2019.
To fill the current gaps, higher-frequency and more granular economic data could help decision-makers obtain a more real-time, localised understanding of the problems to target policy measures and resources and conduct a course correction more efficiently. This could facilitate economic decisions such as easing a lockdown, reallocating essential supplies and services, addressing distressed areas with targeted measures, creating roadmaps to stimulate the economy, and monitoring the road to recovery.
Certain countries are experimenting with new indicators based on high-frequency data, generated primarily by the private sector, to understand the state of the economy, in general, and in particular to quantify the economic impact of COVID-19. These indicators are based on data on various economic activities such as transportation, electricity use, and road traffic. None of these indicators provide a complete and reliable picture by themselves, but taken together, they present a credible measure of the state of the economy.
Table 1: Select high-frequency economic activity indicators in various countries
|Opportunity Insights Economic Tracker ||US||A real-time economic activity platform that shows microeconomic indicators such as aggregated large credit card processors, payroll firms, job posting aggregators, consumer spending, job postings, and revenue of small businesses. This is aggregated from the private sectors to get a macroeconomic understanding of the US economy and track it.||Daily|
|COVID-19 Economic Data Tracking by Federal Reserve Bank of St Louise||US||A combination of dashboards for the US that collects higher-frequency financial market variables along with monthly indicators that track expenditures, employment and unemployment, and key business and consumer surveys||Monthly|
|Capital Economics||Multiple Countries||A dashboard to show economic impact of Covid-19 in different regions and countries||Daily|
|Consumer Survey for Covid-19 by the Federal Reserve Bank of Cleveland||US||Survey data that captures how consumers’ beliefs and expectations have changed over time as the pandemic has unfolded in the US||Weekly|
|Financial Times Economic Activity Index||China||A real-time economic index for China, based on six daily, industry-based data series, that include real estate floor space sales, traffic congestion within cities, coal consumption, container freight, box office numbers and air pollution in the 10 largest cities||Daily|
|The Office for National Statistics Report||UK||Several experimental indicators for the UK created based on close-to-real-time big data, administrative data sources, rapid response surveys and experimental statistics.||Weekly/ Fortnightly|
|WeBank’s AI Moonshot China||China||A deep learning system to detect steel manufacturing activity from satellite imagery to understand China’s economic recovery from the novel coronavirus outbreak||Weekly|
Along with more conventional high-frequency indicators, several alternative and proxy indicators have also been explored in India. Most of these indicators are available at higher frequency and at more granular level (for example, indicators like a text-based index and electricity consumption are available daily). Table 2 summarises select high-frequency indicators that have been developed in the past to proxy economy activity in India.
Table 2: High-frequency economic activity indicators in India
|Indicator||Methodology||Frequency||Data Source||Developer||Unit of data collected|
|Consumer Confidence Survey||A survey of 5,100 responses on households’ perceptions and expectations on the general economic situation, the employment scenario, the overall price situation and their own income and spending.||Quarterly||Consumer Confidence Survey||Reserve Bank of India||Cities|
|Real Estate Activity||Nowcasting the sales growth of real estate companies using Big Data Analytics based on Google search data||Quarterly||Google search data||Reserve Bank of India||National|
|Business Conditions Index||
A real-time measurement of business conditions based on several economic activity indicators of different frequencies such as
1. Yield curve term premium, at daily frequency
2. Initial claims for unemployment insurance, a weekly flow variable
3. Employees on non-agricultural payrolls, a monthly stock variable
4. Real GDP, a quarterly flow variable
|Monthly||Various economic indicators||Reserve Bank of India||National|
|Dynamic Factor Model||Nowcasting the quarterly GDP using a dynamic factor model based on economic activity indicators that represent various sectors and correlate with GDP||Quarterly||Various high-frequency economic activity indicators||Reserve Bank of India||National|
|Text-based Economic Index||Various economic indices based on the proportion of economy-related articles in financial newspapers, the sentiment of newspaper articles related to economy and internet searches related to economy||Daily||Newspapers
|Reserve Bank of India||National|
|Electricity usage||Indicator for assessing economic activity based on causal relationship between energy consumption and economic growth||Daily||Central Electricity Authority||–||National and sub-national|
|Satellite-image based Economic Index||Assessing the socio-economic conditions of villages using satellite images||Yearly[b]||Google Earth Engine||NITI Aayog and IIT, Mumbai||District level|
While alternative indicators can provide a snapshot of economic activities at a higher frequency, they are not reliable by themselves. They need to be viewed as complementary to the existing/traditional indicators which are benchmarked, multidimensional, and constructed with economic, mathematical and statistical expertise. For alternative indicators, the appropriate indicators need to be selected from a pool of potential indicators for each facet of the economy and they need to be benchmarked with the traditional indicators. Comparing them with the public statistics can help understand their biases and construct aggregates that are more representative of overall economic activity.11
Given the limitations of data availability, several of these indicators are only representative of the geographies/cities they are generated from and cannot be taken as a proxy for national indicators of economic activity. Further, if data generated from the formal economy is used (such as unemployment insurance claims or aggregate credit card spending), it is important to acknowledge that they represent the economic condition of only certain population groups. Finally, real-time indicators such as manufacturing activity, interest rates and consumer inflation rates have low predictive power as they change almost at the same time as the changes they signal occur (i.e., they are “coincident” to economic activity).
Thus, while working with alternative indicators, it is important to have clear definitions, to check them for their robustness, indicate the representativeness of the indicator, and clearly state their limitations.
Meeting Economic Data Needs: What could it look like?
This brief suggests developing a real-time economic dashboard that is based on alternative indicators; the dashboard will be accessible to the public. To achieve this, a dedicated team of economists and data experts at the Ministry of Statistics and Programme Implementation could be deployed to explore new high-frequency indicators, undertake timely data collection and analyses, and develop and maintain the economic tracking portal. To widen the data collection net, data from the private sector could also be used. Collecting real-time data without significant delays is undoubtedly a challenging task with the traditional methods of data collection. Thus, alternative ways for data collection such as mobile application and reliable telephone surveys could also be explored to maintain a regular flow of data.
The following paragraphs, including the tables and graphs, outline the potential high-frequency indicators representing each sector. Select indicators are also visualised.
Table 3: Potential High-Frequency Economic Indicators
|Industry and Construction||Index of Industrial Production (IIP) – Core||Monthly||IIP can be used to track manufacturing activities in different sectors of economy. Higher IIP indicates better economic growth|
|Electricity Consumption||Daily||Higher electricity consumption suggests increased economic activities|
|Steel Production||Monthly||Tracking production of steel as a measure of industrial activities|
|Cement Production||Monthly||Tracking production of cement as a measure of industrial activities|
|Capacity Utilisation||Monthly||Higher capacity utilisation at industrial firms suggests higher demand|
|Consumption||IIP – Consumer Goods||Monthly||Higher production of consumer goods suggests higher demand and consumption of consumer goods|
|Auto Sales (subset of cars motorcycles, tractors etc.)||Monthly||Higher auto sales show higher consumer demand of vehicles, higher agricultural demand for tractors|
|Employment||Unemployment Rate (CMIE)||Monthly||Higher unemployment shows poor economic growth|
|MGNREGA Employment||Monthly||Higher employment under MGNREGA suggests lower economic activity in rural labor market|
|Inflation||Consumer Purchase Index (CPI)||Monthly||Higher prices of goods and services suggests higher inflation|
|Credit and Finance||Bank Credit Growth||Monthly||Higher credit growth suggests pick-up in economic growth|
|Stock Market Index||Daily||Higher stock market indexes indicate better economic growth|
|Bond Yields||Daily/ weekly||Falling bond yields can mean the market expects turbulence|
|Interest Rates||Daily||Lower interest rates signal muted economic growth|
|Other||Pollution Levels||Daily||Lower levels of Nitrogen Oxides (NOx) suggest decline in economic/ industrial/ vehicular activity at source|
|Road Traffic||Daily||Lower road traffic suggests decline in economic activity|
|Community Mobility||Daily||Lower community mobility suggests decline in economic activity|
|Nighttime light satellite imagery||Daily||Lower night-time lights indicate low economic activity|
Figure 1: 10 Year Government Bond Yield
The graph shows the % year-over-year change in the 10-year government bond yield for the past 5 months compared to the previous year.
As compared to May 2019, in 2020, the bond yields have fallen by 26.4 percent.
Bond yields is a leading indicator and is released weekly. Falling bond yieldscan mean the market expects turbulence.
Figure 2: Non-food Bank Credit Growth
The graph shows the % of non-food bank credit growth from Nov to Apr this year and the previous year. The red bars show data for Nov-18 to Apr-20 and the blue bars show data for Nov-19 to Apr-20.
The graph shows that as compared to Apr 2019, in 2020, the % of non-food bank credit growth had fallen from 11.9 percent to 7.3 percent, a fall of 63 percent.
Credit Growth is a leading indicator and is released monthly. Falling credit rates indicates lower economic activity.
Figure 3: NOx levels in Mumbai
The graph shows the change in the NOx levels in Mumbai compared to the baseline (baseline is set as values from 2019). The dotted line shows the actual value while the solid line shows the moving average for periods of seven days each.
The graph shows that as compared to the baseline, NOx had fallen by 60 percent one month after the lockdown.
Lower levels of Nitrogen Oxides (NOx) suggest decline in economic/ industrial/ vehicular activity at source
Figure 4: Community Mobility Change
Figure 4 shows the community mobility change in retail spaces compared to the baseline from 1 March 2020 to 20 April 2020. The vertical dotted line represents the day when the national lockdown as imposed. The dotted line shows the actual values while the solid line is the moving average for periods of seven days each.
The graph shows that as compared to the baseline, community mobility had fallen by 80 percent one month after the lockdown.
These high-frequency indicators such as the Google Mobility and pollution data can be used in the forecasting models for real economic indicators. For example, the State Bank of India has constructed an analytical framework to estimate the GDP loss in each state by building on the GDP estimate for FY20. This model uses the high-frequency indicators such as Google Mobility and vehicle registrations data to measure the impact of COVID-19 on the state’s economic activity. In an article published in June, Rathin Roy and Amey Sapre, economists at the National Institute of Public Finance and Policy argue for an approach that is rooted in intuition and inductive as well as iterative in nature to estimate the simultaneous supply and demand shocks caused by Covid-19. Meanwhile, a June article by Pranob Sen, Programme Director of the International Growth Centre (IGC) India, develops a dynamic factor model to project the economic growth trajectory for the next four years.
Countering what is being called India’s “greatest economic emergency since independence” requires data for making informed and urgent decisions and monitoring the road to recovery. Given the complex data-gathering and computation processes, there are several issues with traditional indicators such as GDP. To account for the lack, lag and a growing distrust in government statistics, these traditional indicators must be complemented with alternatives to get real-time data on the economic health of the nation. To improve and make these alternative indicators more reliable, the concerned government agencies, along with relevant private sector entities, must work on their limitations and develop more sturdy, representative and benchmarked proxy indicators, while simultaneously pushing for more robust, more frequent, and more localised traditional indicators.
This brief recommends the development of a dashboard that will capture various real-time economic indicators. This will help the country’s policymakers make better, more targeted and more informed decisions, both at the macro and micro level.
About the Authors
Sanjana Krishnan is Partner and Sahil Deo is Co-Founder at CPC Analytics, a data-driven policy consulting firm with offices in Pune and Berlin. Pooja Verulkar is a graduate student of Policy Analytics at the University of Exeter, UK and an intern with CPC Analytics.
The views expressed above belong to the author(s). ORF research and analyses now available on Telegram! Click here to access our curated content — blogs, longforms and interviews.
Non-resident fellow at ORF. Sahil Deo is also the co-founderRead More +
Pooja Verulkar is a graduate student of Policy AnalyticsRead More +
Sanjana is a partner at CPC Analytics. As partRead More +