'Nowcasting' the Indian economy: A new approach to know the now-GDP

12 Jan 2018

This column describes NCAER’s new ‘nowcasting’ model which seeks to predict India’s GDP numbers at frequent intervals − typically on a monthly or quarterly basis − by exploiting the incremental information in published data on economic indicators.

In the words of Dr. Lawrence Klein¹ “Models go out of date fast. Data base gets revised new kinds of economic problems come to the fore; and sometimes behavior changes. India appears to be in need of a serious model building effort designed to function as an ongoing effort over many years turning out rolling forecasts every month or two on the most refined time period possible…“

Over the years NCAER (National Council of Applied Economic Research) has developed several models for forecasting and policy analysis. At present NCAER relies on an econometric model for forecasting quarterly GDP (gross domestic product) and overall prices. In addition NCAER has also developed what is essentially a structural model of the Indian economy with a medium-term perspective. A salient feature of the model is that it incorporates sources of productivity growth which enables assessing the impact of foreign capital boost in agricultural growth and infrastructure development on the economy. This model is essentially used to forecast GDP for the medium term and for policy analysis. Both these models have primarily used databases from the last decade. The Central Statistics Office (CSO) has replaced the earlier 2004-05 base year with 2011-12 and updated the National Account Statistics (NAS) methodology to align with more recent international guidelines. Subsequently the databases on the Indian economy have undergone substantial revisions in the wake of recent changes in the GDP methodology and measurement issues with regard to price indices. Currently both these models are fully operational. 

However the first release of quarterly GDP/GVA² (gross value added) growth is published approximately 7-8 weeks after the end of the reference quarter. Importantly CSO’s growth estimates avoid assessing the current state of the economy. In order to reduce the time lag between the actual growth during a reference quarter and the official release of the quarterly growth estimates we are developing a new ‘nowcasting’ model at NCAER. Nowcasting models try to capture the flow of data releases in real time throughout a week and month within a quarter. Accordingly we update the GDP/GVA growth estimates by using relevant data releases that are published within the same quarter. Our model exploits a large number of correlated data series and tries to extract potentially relevant signals about the state of the economy. The Dynamic Factor Model (DFM) can capture such relevant signals from large number of data releases and at the same time help avoid the so-called ‘curse of dimensionality’³  problem. It has received increasing attention in the recent macro-econometrics literature and can help identify a few dynamic factors that can represent co-movements in a large number of time series data⁴. The time series data can typically represent different measures of economic activities within an economy and include data on domestic production prices manufacturing turnover construction business surveys labour market statistics interest rate monetary aggregates stock market indices and miscellaneous indicators that are available at more frequent intervals (weekly/monthly). Econometric estimation is undertaken to identify dynamic factors for each of the sectors based on the realised high-frequency data that are available for the relevant sectors. 

Our new nowcasting model-building effort is aimed at successfully estimating such dynamic factors using data-shrinkage techniques and nowcasting Indian GVA numbers at frequent intervals − typically on a monthly/quarterly basis − by exploiting the incremental information in published data on economic indicators. In other words nowcasting involves an exercise of predicting the present the very near future and the very recent past − and that means it is more effective in shorter horizon forecasting. The model is expected to accurately estimate and forecast the rate of growth of the Indian economy.  

Sectors	Indicators
Mining and quarrying	Mining and quarrying Index monthly production of crude oil and coal
Manufacturing	Manufacturing index monthly production of steel and fertilisers purchasing managers’ Index (PMI) – manufacturing commercial vehicles production two-wheeler production passenger car production Organization for Economic Co-operation and Development (OECD) – business confidence index OECD – composite leading indicator merchandise non-oil imports foreign direct investment
Electricity gas and water supply  and construction	Electricity index monthly production of cement monthly production of crude oil oil imports
Trade hotels transport communication and services related to broadcasting	Railway freight traffic of major commodities cargo traffic − ports cargo traffic − air  foreign tourist arrivals in India telecommunication subscriber base PMI-services merchandise trade (exports and imports)
Financial real estate and professional services	Bank credit to commercial sectors national stock exchange (NSE) trading volume personal loan for housing (including priority sector housing) PMI – services foreign institutional investment FOREX (foreign exchange)
Public administration defence and other services	Expenditure of the central government net of interest payments Reserve Bank of India (RBI) net credit to government receipts of central government