Forecasting GDP – Econ Lab by Dr. Tushar Poddar


Forecasting GDP is one of the more important tools in macroeconomics since GDP is a very good indicator to gauge the health of an economy. Moreover, GDP forecasts set the starting point for many other forecasts that are central for macroeconomic management. Every good prediction requires an understanding of the past trends and behaviour of the variable in question. In the case of India, the filtered data for GDP shows an increasing trend between 2004 and 2014. However, when looking at the filtered data from 1980 to 2013, we see an increasing trend till 2008, and a slight decreasing trend afterwards, indicating the slowdown post 2008 global financial crisis.

There are various methods of forecasting GDP, and we will discuss them in turn. The first approach is the Production Function Approach, which uses the concept of potential output or potential GDP and essentially leads us to a supply-side forecast. Potential GDP specifies the aggregate supply of goods and services, given the country’s endowments of productive factors which is consistent with a stable rate of inflation. Potential output is usually estimated using the production function from the Solow framework. Under this framework, the key endowments are capital, labour, and total factor productivity (TFP) (which captures factors such as technology, human capital and institutions and enable more efficient use of labour and capital). Therefore, to forecast potential GDP, one needs to make assumptions about the future trend growths of these three endowments.

The second approach is the Production Approach. Here, one forecasts the value addition of each sector, namely Agriculture, Industry and Services, and combines them to get the total value addition. Each sector has different indicators that influences its forecast. For example, in the case of India, forecasts for agricultural production are influenced by rainfall whereas forecasts for industrial production are affected by monthly industrial production data. For the services sector, the indicators range from tax data to sales data of companies listed in NSE/BSE.

The third approach is the Expenditure Approach which gives us a demand-side forecast. Again, here, one forecasts the individual components, namely private consumption, investment, government expenditure, exports and imports, (which have different indicators affecting the predictions) and add them to get the overall GDP. For instance, In India, consumption and investment are affected by real interest rates, income etc. whereas exports and imports are influenced by exchange rate variations and global growth. Forecasting government expenditure can be tricky because many decisions here are political in nature. However, it can be predicted on the basis of fiscal policy decisions made by policy-makers.

Another way of forecasting GDP is by analysing the output gap, that is, the difference between the level of actual output and potential output.  To estimate the output gap, one uses different statistical filters such as Hodrick-Prescott Filter, Credit-neutral filter and Christiano-Fitzgerald filter. All these filters remove the cyclical component from the raw data to produce a non-linear, smooth representation of the time series. One can also obtain estimates of output gap using structural methods which rely on a specific economic theory. There methods allow for more persistent estimates of the output gap since most of the underlying theories treat trend and cycle independently. Examples of structural methods are multivariate state space models, structural vector autoregression models, and the production function approach, discussed earlier.

Forecasting can also be done by using various econometric techniques. One such technique is using the Random Walk model, which assumes that in each period the variable takes a random step away from its previous value, and the best guess for GDP growth in the next period is GDP growth in the current period. Another is the Mixed-frequency Bayesian Vector Autoregression model, which uses information from several time series and exploits high-frequency, more timely macroeconomic data for estimation. Finally, another approach of forecasting GDP is based on analysing data from various economic indicators that could predict economic activity.

In sum, there are various methods of forecasting GDP, and the choice of a specific method can be made with the help of different criteria, depending on the concrete goal of the research.

Map Unavailable

Date(s) - 24/10/2015
10:00 am - 12:00 pm

Request a Callback


Contact a course advisor at

Which course you want apply for

Economics Data Science