The “V-factor”: Distribution, timing and correlates of the great Indian growth turnaround

https://doi.org/10.1016/j.jdeveco.2011.09.002Get rights and content

Abstract

We analyze a panel of output series for India, disaggregated by 15 states and 14 broad industry groups. Using principal components (Bai, 2004; Bai and Ng, 2004) we find that a single common “V-factor” captures well the significant shift in the cross-sectional distribution of state-sectoral output growth rates since the 2nd half of the 1980s. The timing of the turnaround implied by the V-factor is more closely related to the pattern of policy reforms than has been found in previous research. Regression-based analysis also provides some insights into the uneven distribution of the turnaround across Indian states.

Introduction

In the past two decades or so there has been a remarkable turnaround in Indian growth. From 1960 to 1987 output per capita in India (measured by real net domestic product1) grew by only 1.31% per annum, while on the same measure US output per capita grew at 2.36%, so that Indian and US output levels were steadily diverging. In marked contrast, from 1987 to 2004 Indian output per capita grew at 4.12% per annum, while US per capita growth slowed to 1.62%; thus India has been converging towards US output per capita levels at a more rapid rate than it was diverging in the earlier period. However a notable feature of the turnaround has been the distinctly uneven distribution of the growth turnaround across the major states, several of which have shown little or no increase in growth.

The turnaround in Indian economic growth has inevitably generated considerable public interest and some academic research with respect to its timing, possible causes, and unevenly distributed nature.2 In this paper we present evidence on all three issues.

Our approach exploits the fact that, amongst economies at similar income levels, India's economy is unusually well provided with data. We utilize a new panel dataset, disaggregated into 15 major states and, within each state, into 14 broad industrial sectors, over the sample 1970–2004; we can also extend the dataset back a further ten years for a subset of ten states. We first show that the shift in growth has been highly pervasive across the Indian economy, in that there has been a shift in the cross-sectional distribution of growth rates of output per capita that is highly significant in statistical terms. We then use principal components analysis (following Bai and Ng, 2002, Bai and Ng, 2004, Bai, 2004) to derive a common factor representation of the dataset. We show that a single common factor provides a powerful and parsimonious account of the distributional shift. This common factor is V-shaped, with a minimum in the second half of the 1980s.

A significant advantage of this approach is that we do not need to impose a particular date for the turnaround in growth. Nor do we need to impose that it be a deterministic shift, as in standard econometric representations of structural breaks; nor even that all series participate in the shift at identical dates.

The strong explanatory power of this common “V-factor” suggests a single common cause. Our results appear to resolve the puzzle discussed by Rodrik and Subramanian (2005), who, along with other researchers, had concluded that the turnaround in growth came in the late 1970s or early 1980s, well before any significant observable shift in policy.3 We find a later turnaround, in the second half of the 1980s, which is much more consistent with what is known about the pattern of liberalization (see Panagariya, 2004, Pursell, 1992). In particular, we show that the time profile of the V-factor is strongly correlated with the pattern of trade liberalization, as summarized by the effective tariff rate. We emphasize our results on the tariff rate because it is the closest thing we have to an indicator of a true trade policy measure, rather than of an endogenous response to policy. But we also provide evidence on other trade and non-trade indicators that are consistent with the time profile of the V-factor.4

The remainder of the paper is structured as follows. In Section 2 we provide some summary evidence of growth shifts at the sectoral and state levels. In Section 3 we carry out the statistical analysis and derive the factor representation. We examine the evidence for a shift in the second half of the 1980s, and contrast this with the results from earlier studies. In Section 4 we compare the path of the V-factor with what we know about shifts in policy. In Section 5 we use regression analysis to examine whether state characteristics can account for the very disparate performance across the states noted above. Section 6 concludes the paper. A web appendix provides details of data construction and statistical analysis.5

Section snippets

Sectoral and state-wise shifts in growth

Fig. 1, Fig. 2 give two alternative broad-brush pictures of the turnaround in growth. We compare average sub-sample growth rates before and after 1987.6 Fig. 1 shows that virtually all sectors of the private sector economy have seen substantial increases in growth, albeit from often significantly

The dataset

We analyze a panel dataset of output per capita series broken down both by state and by sector. For fifteen major states (the same group shown in Fig. 2, excluding Jammu and Kashmir) we have a sectoral breakdown into fourteen broad industrial sectors, from 1970 to 2004; for a subset of 12 states (also excluding Assam, Bihar and Orissa) we have the same sectoral breakdown from 1965, and for 10 states (also excluding Haryana and Punjab) from 1960. We eliminate three series due to clear data

The V-factor and economic policy

The contrast between our results on the timing of the turnaround and those of earlier research is of particular interest, since it suggests a resolution of a puzzle discussed by Rodrik and Subramanian (2005): while they, in line with most other research, identified a turning point in the late 1970s or early 1980s, this appeared significantly to pre-date major policy changes. Is the later turning point we identify in the V-factor more consistent with what we know about the timing of economic

Participation in the turnaround: some regression results

While the common nature of the growth turnaround, as identified by the V-factor, appears to correspond fairly well to observable shifts in India-wide economic policy, the quite disparate impact of the turnaround across the states (as illustrated in Fig. 2) is quite striking. In this section we use our panel dataset to investigate whether this disparate performance can be captured by observable state characteristics. We find that it can; however our results reveal less about the role of

Conclusions

In their international study of growth accelerations, Hausmann et al. (2005, p. 328) conclude that:

“It would appear that growth accelerations are caused predominantly by idiosyncratic, and often small-scale, changes. The search for the common elements in these idiosyncratic determinants – to the extent that there are any – is an obvious area for future research.”

This paper provides evidence of such common factors in the context of the Indian economy; we hope that the techniques we employ may

References (25)

  • A. Bhattacharjea

    Labor market regulation and industrial performance in India: a critical review of the empirical evidence

  • Deb Kusum Das

    Quantifying trade barriers: has protection declined substantially in Indian manufacturing?

  • Cited by (0)

    We are extremely grateful to Amit Sadhukhan for research assistance during the course of the project. We thank Dr. Savita Sharma and Pronab Sen of the Indian Central Statistical Office for helpful advice on the data. The co-editor, William Easterly, and two referees gave invaluable comments. We also thank Gerhard Glomm, Sanghamitra Das, Samarjit Das, Abhiroop Mukhopadhyay, George Kapetanios, Ron Smith, and seminar participants at ICRIER, DIW Berlin, the Max Planck Institute — Jena, ISI Delhi, JNU, Institute of Economic Growth, the Delhi School of Economics, the 45th Meeting of the Indian Econometric Society (TIES), Jadavpur University, Claremont Graduate University, and Indiana University (Bloomington) for comments. Stephen Wright is grateful to the Indian Statistical Institute, Delhi, and the EGP group at the Max Planck Institute — Jena for hospitality during research visits in 2007 and 2008. Both authors are very grateful to the PPRU Committee for financial assistance related to this project.

    View full text