DISCUSSION
Understanding Economic Growth in India, Further Observations
Pulapre Balakrishnan, M Parameswaran
the identification of growth regimes endogenously. This is noteworthy and Dholakia’s effort in getting it recorded as scholarship on econo mic growth in India is to be appreciated.
Of the two other papers cited by Dholakia, the first, Dholakia and Dholakia (1993), is concerned with the sources of growth. This was not our concern, which
The authors respond to Ravindra Dholakia’s critique (August 25) of their article ‘Understanding Economic Growth in India: A Prerequisite’ (July 14) and give further evidence that the acceleration of growth in the parts quarter century has been driven by services.
We thank Ravindra Dholakia for having sent us his communication in advance, and for having responded to an earlier version of this note. Of course, we do not take his response for agreement with our views. Also, we would like to avail of this opportunity to thank, without implication, Jessica Wallack for discussions at an early stage of our research.
Economic & Political Weekly November 3, 2007
W
Relevant Literature
Dholakia opens his note by referring to some research not cited by us, work that is putatively relevant to the results reported in our article. Thus, “Had B-P gone back to 1992 for selecting their references, they would have found at least three references directly relevant to their methodology and findings” [Dholakia 2007, pp 3509]. While the thought that we may have somehow overlooked published work directly relevant to our research does concern us, having now turned to Dholakia’s references, we find that we have not missed anything of consequence. First, while it is true that Ganesh Kumar (1992) addressed himself to the same question as we do, the method adopted by us supersedes that in his paper. To be precise, Ganesh Kumar’s method, based on the work of Quandt allowed him to endogenously estimate a break at a time, while the Bai-Perron method used by us allows for the simultaneous estimation and testing of multiple structural breaks. Not only does the latter suggest itself a priori when studying the trajectory of an economy over 50 years as we do, but the method eminently justifies its use retrospectively as we detect up to four breaks in some of the series. As the method and the time period differ, the comparison of the results reported by us with those reported by Ganesh Kumar serves little purpose. Nevertheless, Ganesh Kumar’s work is of interest in that it is, as far as we know, the first to attempt was actually to identify growth regimes by sector in India, and to thereby enhance our understanding of Indian economic growth. Similar reasoning applies to the second paper cited – Dholakia (1994). We can see that the spatial dimension of growth is an interesting question but that is not an issue of interest to us just yet.
Explaining Breakdates
The main body of Dholakia’s comment appears in two sections. First, there is a commentary on our results and then, in a section entitled ‘Spatial Dimension’ there is a discussion of the results of Dholakia and Dholakia (1993) on the geography of Indian economic growth. As our paper does not address this aspect we have nothing to say on his discussion in this section. However, it must be pointed out that Dholakia’s statement, “If a panel data considering states, sectors and years are considered for identification of the national breakdates in time trends, the results could be very different from the ones reported in B-P (2007)” is speculative as no reason is provided why this must be so. It is, of course, not inconceivable that the behaviour displayed by the collection of disaggregated elements of a variable would differ from that of the aggregate. However, in macroeconomics, it is the convention to measure the growth of an economy in terms of aggregate output and, since in this paper we are interested in the growth of the Indian economy, we consider it appro priate to work with GDP.
The section ‘Breakdates by Sectors’ does address our principal concern but the exercise is mostly conducted by comparing our results with those of Ganesh Kumar. We find this artificial, as our method is less restrictive in that, as stated, it allows for multiple breaks while that used by
DISCUSSION
Kumar can accommodate only one break at a time. In any case, by way of summing up Dholakia has had the following to say regarding the results presented in our article:
The methodology used by B-P (2007) is so sensitive that adding or subtracting data may have impact on the identification of breakdates even in the past. Thus, potentially, there are three reasons for not treating as final the identification of past breakdates in India applying the methodology developed by Bai and Perron (1998, 2003). One, change of base year of the series; two, consideration of additional data points; and three, non-adjustment of the weather factor, so critical in India till recently [Dholakia 2007, pp 3511]. To this our response is as follows:
First, as breaks in the progress of a time series are defined in relation to its trajectory both prior to and subsequent to the breakdate, it is indeed plausible that altering the sample size, i e, “adding or subtracting data”, may alter an estimated breakdate. If the Bai-Perron method detects this, then, it must be recognised as a source of its strength rather than as a weakness whereby it is “sensitive” to change in the sample size.
Second, of the three reasons for caution that Dholakia advances in the above statement, we would say that while it is always healthy to exercise caution when choosing among results, it is also imporant to appreciate why one does so. Altogether, we are not for excessive caution in the present context for the following reasons. Adjustment for weather when searching for trend breaks in a time series of 50 years appears to us unnecessary, so long as the fluctuations are random. In any case, our interest was in a break in the trend of agricultural output, for which weather-induced random deviations from the trend do not matter.
Taking our first point, instances of base-year revision that involve the use of a different commodity basket would indeed alter the time series properties of the revised series, making it incompatible with the original. This may lead to a new set of estimated breakdates. This is only to be expected as we are now working with a different series1 altogether. Finally, our second point. While the possibility that an extant result pertaining to breakdates may alter as fresh observations come in has already been addressed by us, we might add that science is worth doing precisely for the feature that fresh results are pregnant with the possibility of falsification! At the same time, we hold fast to the position that till such time as divergent results are provided, or our method has been shown to be invalid, the results presented in our paper must be reckoned with.
Table: Sectoral Contribution to Change in Rate of Growth of GDP
Contribution in % of Change From 1979-80 From 1991-92
I Primary sector 20.5 -30.3
Our Findings
Dholakia’s comment overlooks what we consider the main conclusion of our study. Note that having established the breakdate for aggregate GDP as 1978-79 we had gone on to decompose by sector the contribution to the higher rate of growth from 1979-80. We had found that it was the tertiary sector that had contributed overwhelmingly to this turnaround, and had flagged recognition of this feature as a prerequisite to understanding recent economic growth in India. This last finding is independent of the methodo logy of breakdate estimation. In any case, our estimated break-date for GDP growth is very close to that of other researchers who have used a similar method [Wallack 2003; Rodrik and Subramanian 2005]. So, once again, we do not see grounds for excessive caution.
Conclusion
As part of a review of our results, we have now undertaken a re-estimation of turning points for growth using the latest available series on GDP from the Central Statistical Organisation’s national accounts statistics. Based on 1999-2000 prices, this series extends up to 2005-06 while the earlier series, in 1993-94 prices, extended only up to 2003-04. Note then that we are now working with a virtually new data set. We report that when we confine the series to the period 1950-51 to 2003-04 – the period in Balakrishnan and Parameswaran (2007) – we find one
Agriculture -23.8 -25.8
break, as before, in 1978-79. When the
Forestry and logging -9.9 2.6
series is extended to 2005-06, we detect
Fishing 0.6 -0.1 Mining and quarrying 12.5 -6.9 two breaks dated 1978-79 and 1990-91,
II Secondary sector 17.9 26.1
implying three growth regimes 1950-51 to
Manufacturing 14.1 16.4
1978-79, 1979-80 to 1990-91, and 1991-92
Registered 24.5 7.8 Unregistered -10.4 8.6 to 2005-06, respectively.
Electricity, gas and water supply 7.8 -0.5
Having established the growth transi-
Construction -4.0 10.1
tions, we undertake2 a decomposition by
III Tertiary sector 102.6 104.2 Trade, hotels and restaurants 12.6 35.5 sector of the contribution to the change
Trade 11.4 31.7
in the growth rate for each of the two
Hotels and restaurants 1.2 3.8
transitions found. The results are presen-
Transport, storage and communication 13.2 36.1 Railways -0.1 -0.6 ted in the table. We confine our comment
Transport by other means 11.0 8.6
to the one main finding that strikes us as
Storage -0.1 -0.1
most interesting. Notice that, as with the
Communication 2.4 28.2 Financing, insurance, real estate and business services 45.2 25.0 transition of 1979-80, the growth transi-
Banking and insurance 18.3 19.0
tion of 1991-92 is also led by the rise in the
Real estate, ownership of dwellings and business services 26.9 6.0
rate of growth of services.3 With this find-
Community and personal services 31.6 7.7 Public administration and defence 16.5 -3.4 ing, the main conclusion of Balakrishnan
Other services 15.1 11.1
and Parameswaran (2007) needs no
October 27, 2007 Economic & Political Weekly
DISCUSSION
revision. The accelerations in the growth of the economy over the last quarter century have consistently been led by services. On the other hand, the contribution of manufacturing to these transitions has been relatively small, despite this having been the sector most targeted by changing policy. By comparison, note the extraordinary contribution of communications to the rise in the growth rate since 1991-92. Clearly, the response of the sectors to the policy changes has not been uniform.
Email:pbkrishnan@yahoo.com
Notes
1 For a discussion of the procedure adopted by the CSO in linking national income-series with differing bases see EPW Research Foundation (2002), pp 21-22.
2 See Balakrishnan and Parameswaran (2007) for the methodology used.
3 See the figure for tertiary sector in the table.
References
Bai, J and P Perron (1998): ‘Estimating and Testing Linear Models with Multiple Structural Changes’, Econometrica, Vol 66, pp 47-78.
Balakrishnan, P and M Parameswaran (2007): ‘Understanding Economic Growth in India: A Prerequisite’, Economic and Political Weekly, Vol 42, No 27, pp 2915-22.
EPW Research Foundation (2002): National Accounts Statistics of India 1950-51 to 2000-01, EPW Research Foundation, Mumbai.
Dholakia, R H (2007): ‘Understanding Indian Economic Growth: Some Observations’, Economic and Political Weekly, August, 25, pp 3509-11.
– (1994): ‘Spatial Dimension of Acceleration of Economic Growth in India’, Economic and Political Weekly, Vol 29, No 35, pp 2303-09.
Dholakia, R H and B H Dholakia (1993): ‘Growth of Total Factor Productivity in Indian Agriculture’, Indian Economic Review, Vol 27, pp 25-40.
Ganesh, Kumar N (1992): ‘Some Comments on the Debate on India’s Economic Growth in the 1980s’, Indian Economic Journal, Vol 39, No 4, pp 102-11.
Rodrik, D and A Subramanian (2005): ‘From ‘Hindu Growth’ to Productivity Surge: The Mystery of Indian Growth Transition’, IMF Staff Papers, 52, pp 193-228.
Sivasubramonian, S (2000): The National Income of India in the Twentieth Century, Oxford University Press, New Delhi.
Wallack J S (2003): ‘Structural Breaks in Indian Macroeconomic Data’, Economic and Political Weekly, Vol 38, No 41, pp 4312-15.
Economic & Political Weekly October 27, 2007
