International Monetary Fund, 2004, World Economic Outlook, April 2004: Advancing Structural Reforms, World Economic and Financial Surveys (Washington).
Rodrick, Dani and Arvind Subramanian, 2004 “From Hindu Growth to Productivity Surge: The Mystery of the Indian Growth Transition,” IMF Working Paper No. 04/77 (Washington: International Monetary Fund).
Salgado, Ranil, 2002, “Recent Trends in Growth and Investment,” India: Selected Issues (Washington: International Monetary Fund).
Virmani, Arvind, 2004, “India’s Economic Growth: From Socialist Rate of Growth to Bharatiya Rate of Growth,” Working Paper No. 122 (New Delhi: India Council for Research on International Economic Relations).
Prepared by Catriona Purfield.
At factor costs. Expenditure-based GDP data are not yet available.
The procedure begins with a general unrestricted model containing three lagged values of rainfall and GDP. It uses both a top-down and bottom-up approach to recursively eliminate insignificant variables. At each stage mis-specification tests are re-computed, and if any test fails that particular reduction it is disregarded as invalid until a parsimonious model is identified where all the remaining variables are significant. The dependent variable, GDP, is non-stationary and so is specified in growth rates. The final model specification, with all variables significant at the 1 percent level, is:
LNGDPt = –1.7 + 0.008t + 0.17 LNRAIN + ut,
where t is time, LNRGDPt is real GDP growth (all variables in logarithms), and LNRAINt is rainfall. The R-squared is 0.46.
The normalized rainfall series is calculated by the Indian Metrological Department as a long-run moving average of actual rainfall.
It may also be optimal to include a measure of the dispersion of rainfall across Indian states in the model to capture the fact the distribution of rainfall can have differential impact on growth depending on the importance of agriculture and irrigation in each state. However, preliminary investigations found the measure of dispersion across India’s metrological districts (calculated using the standard deviation or the coefficient of variation) turned out to be insignificant and was rejected in the general-to-specific modeling process. Future work will examine the potential of this, and the possibility of an interaction rainfall-time variable that captures the structural decline of the importance of agriculture in the Indian economy in recent years.
A difficulty with this approach is that the trends tend to become poorly defined at the sample end-points. To cross-check the robustness of the results, trend real GDP growth, in levels and adjusted for rainfall, is also estimated by extending the sample period using staff forecasts of GDP for 2004/05–2009/10 period. The results are broadly similar.
The deceleration in trend growth from the late 1990s onwards was also found when the sample for the HP filter was extended outwards by including staff forecasts of GDP growth for the 2004/05–2009/10 period.
Applying Chow tests to a Hodrick-Prescott real (non-rain adjusted) GDP series yielded similar results, with the exception that it identified an additional structural break in 1995/96 when underlying trend growth accelerated to its highest level (just over 6 percent).
F-tests on the rain-adjusted HP filtered real growth series and on the HP filtered real GDP series are significant at the 5 and 10 percent confidence levels. The corresponding log-likelihoods tests are each significant at the 5 percent level of significance.
For China integration is defined as starting in 1979 when major economic reforms began. For India and all other regions, integration is defined as starting when the three-year moving average of constant price export growth first exceeded 10 percent: 1955 for Japan, 1967 for the NIEs, 1973 for the ASEAN-4 and 1995 for India. The NIEs consist of Hong Kong SAR, Korea, Singapore and Taiwan Province of China. ASEAN-4 consists of Indonesia, Malaysia, the Philippines, and Thailand.