Why some Indian states have grown faster than the others?
Jeffrey D Sachs, Nirupam Bajpai and Ananthi Ramiah
(Center for International Development, Harvard University)
India accounts for a meager 2.4 per cent of the world surface area yet it sustains a whooping 16.7 per cent of the world population, a little over 1 billion people residing in 29 states and 6 Union Territories.
The variation across these states and territories is enormous in regard to physical geography, culture, and economic conditions. Some states have achieved rapid economic growth in recent years, while others have languished.
To address the question of regional performance, we narrow our focus to the 14 most populous states, which excludes the North-Eastern states, and the 6 Union Territories. The included states have a combined population of 897 million, accounting for approximately 90 per cent of India's population, and 2.7 million sq kms, accounting for 83 per cent of India's total land area.
The variation in economic performance is large. The per capita state product varies from the poorest state, Bihar, at Rs 1,010 per month and population of 82 million, to the richest, Maharashtra, at Rs 4,853 per month and population of 96 million.
Growth performance has been equally varied, with the slowest growth in state per capita income in Bihar, at -0.2 per cent per year during 1992-98, compared with the fastest growth in Gujarat, at 7.8 per cent per year.
The differential performance across states has begun to raise important policy questions within India.
To what extent are the differences a manifestation of global economic forces acting upon India, especially during a period of economic liberalisation, and to what extent do they reflect differences in economic policies at the state and union level?
Will market reforms tend to make the rich states richer in relative terms, with the poor states lagging ever farther behind, or will market reforms lead to economic convergence across states?
Specifically, are the poorest states (especially the so-called BIMARU states of Bihar, Madhya Pradesh, Rajasthan and Uttar Pradesh) condemned to fall further behind the front-runners, at least in relative terms?
In the case of China, we found that the underlying drivers of economic growth, and hence the tendencies towards convergence or divergence, differed markedly across sub-periods, especially as a result of major shifts in the economic policy regime.
During the first phase of China's market reforms, for example, during 1978-84, the dismantling of the communes and the partial liberalisation of food production gave a great boost to major food producing regions.
By the late 1980s, however, international trade had become the major driver of economic growth, and as a result the coastal regions spurted ahead of the interior provinces, a pattern which obtains till now.
In India, we must similarly distinguish between policy regimes, especially the era of state planning up to 1991 and the market-reform period since 1991.
In the planning period, international trade played only a minor role and industrialisation was affected heavily by state investment plans, which attempted -- at least mildly -- to promote the laggard regions.
One great impetus to national growth came via the Green Revolution, which led to sharp increases in grain productivity in regions such as Punjab and Haryana specifically adapted for the improved crops, mainly wheat.
After 1991, market forces and international trade have played a larger role, though the insertion of India into the global economy has been much less dramatic than in the case of China. Still, we would expect that coastal regions to be advantaged relative to interior regions after 1991, as the coastal regions face much lower transactions costs in participating in global trade and investment.
Several studies of high-income market economies undertaken during the 1990s, for the US, Japan, and regions within Western Europe, found evidence for strong convergence among regions. We find little evidence of comparable convergence among Indian states, similar to the findings for China.
This raises an important question as to why some countries or regions demonstrate inter-regional convergence while others, like China and India, do not. In China and India, it appears that geographical variation across regions may block or slow the convergence of incomes.
I. Convergence among Indian States - 1980-98
We begin with the two standard ways of examining the presence or absence of unconditional convergence. The first measure is the so-called a-convergence (sigma convergence).
We measure the standard deviation across regions of the logarithm of Real Gross State Domestic Product (GSDP) per capita. There is a-convergence if the standard deviation across states tends to decline over time. The second measure is ß-convergence (beta convergence).
Here we regress the proportionate growth in per capita SDP on the logarithm of initial income. There is ß-convergence if the coefficient on initial income, denoted ß, is negative and statistically significant.
In addition to looking at a-convergence and ß-convergence across the 14 states, we also divide the states into two groups based on GSDP per capita, and examine convergence within these two subgroups.
Group I includes the high-income states of Maharashtra, Punjab, Haryana, Gujarat, Tamil Nadu, West Bengal, Karnataka and Kerala. Group II includes the poor states of Rajasthan, Andhra Pradesh, Madhya Pradesh, Uttar Pradesh, Orissa, and Bihar.
When all the 14 states are pooled together they show an increasing standard deviation between 1980-1990 and an increased rise in that standard deviation in the post-reform period. There was an overall rise in inequality (between 1980 and 1998) of 2.40 per cent per annum.
Of the two periods within that time frame, it is the 1992-98 period that caused the most divergence 2.14 per cent per annum compared to 1.24 per cent per annum in 1980-1990. Thus there was no a-convergence during either of the periods under question.
It is interesting to see that the Group II states exhibit much greater volatility in dispersion than the Group I states, in addition to exhibiting the greater absolute dispersion (even though Group II states started off with considerably lower standard deviation than their counterparts). This is the case for the both the periods under question.
From 1980-90, Group I states see an increase in dispersion of a mere 1.24 per cent per annum compared to 2.51 per cent per annum within the Group II states. During 1992-98, Group I states see a decrease in dispersion of 0.35 per cent per annum while Group II states witness an increase of 4.61 per cent per annum.
Thus, it appears that the richer states experienced a degree of convergence during the post-reform period, while the poorer states did not.
The lack of a-convergence is mirrored by a lack of ß-convergence. During 1980-90 growth patterns were divergent. As just one example, the state with the highest GSDP per capita level in 1980 was Punjab at Rs 3,020 per month and the state with the lowest GSDP per capita level was Bihar at Rs 1,062 per month.
In terms of growth rates from 1980-81 to 1990-91, Punjab grew at 3.78 per cent per annum and Bihar at 2.94 per cent per annum. More generally, the richest states also had the highest growth rates.
The only notable exception is Rajasthan, which had the second lowest initial GSDP per capita level, but was the fastest growing state during the 1980s. Apart from Rajasthan, all states have generally grown in a manner that perpetuates divergent trends during the pre-reform period.
Do states exhibit convergent trends during the post-reform period? The fastest growing state is Maharashtra, but it also has one of the highest GSDP levels of the 14 states. Overall there is a significant positive relationship between initial income in 1990 and growth during the 1990s, indicating divergence.
The regression evidence suggests that b-divergence was especially marked for Group I states in the post-reform period.
By both standards of convergence, India demonstrated overall divergence during the period 1980-98, as well as during both the pre-reform and post-reform sub-periods. Divergence was most notable within the poorer group of states.
This finding is consistent with the experience of China in the post-reform period, but differs from the findings for the US, Japan, and European regions.
II. Economic performance of the states
We hypothesise that regional differences in growth reflect regional differences in the marginal productivity of investments by sub-sector.
To some extent, the relative returns to investment in each sub-sector depend on the general business environment, but to an important extent, they also depend on specific geographical factors.
Agriculture can occasionally be a leading sector in economic growth, either on the basis of a spurt in agricultural productivity or on the basis of cash-crop exports ("vent for surplus"). In the case of India, agricultural-productivity-led growth occurred in one major historical period, the Green Revolution, dating from 1965-66 to the early 1980s.
The Green Revolution was centered on short-stemmed, high-yield wheat, and to a lesser extent paddy rice, with both crops depending on irrigation and intensive application of fertiliser. The epicenter of the Green Revolution was Punjab and Haryana, and to a lesser extent other states of the North Indian Plains (as far east as Bihar) and southward to Rajasthan, Gujarat, and Maharashtra.
High-yielding rice varieties made their impact most powerfully in West Bengal and Tamil Nadu. Note that China, like India, experienced one short-lived burst of agriculture-led growth, with the dismantling of the communes and consequent jump in food output during 1978-84.
Almost all the regions among the Group II states (except perhaps much of western Rajasthan and parts of Western Madhya Pradesh and southern Uttar Pradesh) have the agro-climatic potential to yield high returns in agriculture because of reasonable to high rainfall and availability of perennial river waters.
Much of the reason for poverty in these states is thus a human failure rather than it being a result of natural factors. While it may be useful to identify states with high/low incidences of poverty, there are states, which have high variations within them both owing to historical/economic antecedents and agro-climatic factors.
This is typically true of the larger states though such variation exists in smaller states as well.
A more disaggregated, NSS region-wise picture of poverty (head count ratio) shows that, to a significant extent, there are heterogeneities in each state except perhaps Bihar, which is uniformly poor. Sharp contrasts are witnessed in Andhra Pradesh, Karnataka and Maharashtra, though variations can be seen in smaller states like Haryana, and Punjab as well.
The regions have been segregated by low (up to 20 per cent), medium (21-40 per cent), high (41- 60 per cent) and very high (more than 60 per cent) levels of poverty. Southern Bihar, southern Orissa, southwestern Madhya Pradesh and southern Uttar Pradesh fall in the very high poverty bracket. These regions are composed of the districts in Chhota Nagpur and Santhal Parganas in Bihar, Koraput and Phulbhani districts in Orissa, the Jhansi region in Uttar Pradesh and its adjacent regions in Madhya Pradesh, including Betul, Khandwa and Hoshangabad.
Two peculiar features of these regions are, that either they are mainly tribal (except Jhansi) or rocky and dry, yet densely populated because of their agro-climatic features. The one major inference drawn here is that tribal areas are predominantly and distinctly poor.
The high priority (41-60 per cent) areas are in Bihar, portions of Madhya Pradesh, inland Maharashtra, northern Tamil Nadu, eastern and central Uttar Pradesh, and parts of West Bengal. The reasons here are similar; that tribal, thickly populated semi-arid areas, and those, which have been neglected historically, are poor.
Parts of West Bengal have made strides in poverty alleviation. Medium level poverty persists in regions of western states; a few regions have made more progress than others, compared to the eastern ones where there is uniform poverty.
Typical examples are, Madhya Pradesh, Maharashtra, Tamil Nadu and Uttar Pradesh. Lastly, the western coastal regions, entire Andhra Pradesh, Punjab, parts of Madhya Pradesh and Rajasthan, which are a continuum of a north-south belt having experienced green revolution, are pockets of low poverty.
The manufacturing sector is a much more consistent engine of growth, and it is likely to play a growing role after 1991 with the opening of the economy.
As China's experience demonstrates, trade liberalisation in a low-wage, surplus-labor environment permits a rapid expansion of export-oriented industry, which can absorb large numbers of workers to provide goods for the world market.
India's insertion into the world economy has been much less dramatic, and successful, than China's, but it has been real nonetheless. The share of exports of goods and services in GNP was stable at 7 per cent in 1980 and 1990, and after reforms rose to 11 per cent in 1999.
In China the comparable share rose from 6 per cent in 1980 to 18 per cent in 1990 and 22 per cent in 1999. There are many differences in this experience between China and India. China's reforms were bolder in promoting both FDI and manufacturing exports, and China benefited from the vast inflows of FDI from overseas Chinese investors, especially in Hong Kong, Taiwan, China, and Southeast Asia.
The most likely site for sustained manufacturing growth in India, like China, is along the coast, especially at the four large port cities of Mumbai (Maharashtra), Kolkata (West Bengal), Chennai (Tamil Nadu), and Kandla (Gujarat).
Coastal, urban-based industry can serve both the internal market and the international market, and can more readily make logistical links with foreign suppliers and customers than can interior-based enterprises.
New export-oriented units are therefore heavily concentrated on the coast. Manufacturers in interior regions can of course service the domestic market, particularly in consumer goods such as processed foods, but the potential for rapid growth based on the internal market tends to be more limited than the growth based on exports to the world market.
For this reason we have seen much faster growth in coastal China than in the interior of China.
The tourist sector can also be a source of export-led growth, but in a country the size of India, it is likely to play a secondary role except in some local niches.
Tourist potential is, of course, very much geographically determined, as it depends on the physical environment (e.g. beachfront), the presence of historical sites, and easy access to transport nodes, especially international airports.
Rajasthan has been the major state with the most significant growth and scale of the tourist industry, based on the popularity of tourist visits to Jaipur and Udaipur and the proximity to Delhi.
High-tech services, such as information-and-communications-based industry (e.g. software production), or financial services, are almost always reliant on a network of universities and an urban labor market.
These sectors are much less dependent on coastal access, however, since much of their business can be transacted over telephone and Internet connections. A high quality of life of the location, as an attraction for highly mobile skilled workers, probably looms larger in these sectors than in other sectors of the economy.
The most important state for service-sector activities is surely Maharashtra, as it combines the country's financial center with an important IT-based industry. Other key states include Tamil Nadu, Karnataka, Delhi, and to a lesser extent Andhra Pradesh.
Foreign investors have multiple motivations: to service the domestic market; to exploit site-specific natural resources (e.g. mining); and in low-wage countries to establish export platforms in labor-intensive goods, or labor-intensive stages of the production process, or in standardised technologies that are easily transferable to lower-wage settings.
In general, coastal access is a huge benefit for all export-platform manufacturing, as we've seen clearly in the case of China. More generally, FDI is attracted to urban areas and to natural resource deposits. Interior cities (such as Bangalore and Hyderabad) may be attractive for IT-based activities, which do not depend on coastal access.
A simple regression confirms that FDI flowed mainly to the urbanised states and to the states with large mining sectors as a per cent of GNP (especially Orissa and to a lesser extent Madhya Pradesh).
Taken in total, these considerations suggest that urbanisation is likely to be a key determinant of economic growth in the 1980s and 1990s, as we would expect that already existent urban areas would be the preferred location for new investments in manufactures and services.
The extent of urbanisation varies widely, between a low of 13 per cent in Bihar and Orissa and 38 per cent in Maharashtra as of 1991, with the relative proportions of urbanisation by state relatively constant over the past 30 years.
The degree of urbanisation itself depends on underlying geographical factors, especially the location of the main national ports (with their origins in history and the geography of natural harbors), as well as the productivity of agriculture in the region. Regions of high agricultural productivity tend to support a larger proportion of the local population in an urban setting, while regions of low agricultural productivity tend to have a high proportion of the population in peasant, subsistence agriculture.
Empirically, we find, for example, that two factors - having a major port (Maharashtra, Tamil Nadu, West Bengal, and Gujarat) and having a dry steppe climate (Bs) suitable for wheat production - accounts for two-thirds of the variation in urbanisation rates across the fourteen states:
As we would expect, the rate of growth of GSDP per capita is highly correlated with the extent of urbanisation at the beginning of the period. In a regression of growth during 1980 to 1998 on initial income in 1980 and urbanisation as of 1981, the urbanisation coefficient is highly significant with a coefficient 0.13 and t-statistic 5.3.
A remarkable 82 per cent of the cross-state variation in growth is explained by just this variable, and with no hint of any conditional convergence after controlling for the degree of urbanisation.
The regression estimate shows that a 10-percentage-point higher rate of urbanisation is associated with 1.3 percentage points per year higher annual growth. A simple bivariate regression without initial income shows the same results.
During 1980 - 90 the Green Revolution continued to play a role in growth differentials across states. To capture the effect of the Green Revolution, we construct a dummy variable equal to 1.0 in Punjab and Haryana (the epicenter of high-yield wheat), 0.25 in Gujarat, Maharashtra, and Rajasthan, and 0 elsewhere.
This variable has a coefficient of 3.54 (t = 4.01) in the growth regression for the period 1980-90, holding constant the initial income, the degree of urbanisation in 1981. Rajasthan is an outlier in this regression, with growth more than 1 percentage point per year faster than is otherwise explained.
This may be due to the boom in tourism, or to the rapid electrification of the state in the 1980s, or to a more significant effect of the Green Revolution than is captured by the value 0.25, or to some other unmeasured effect.
Interestingly, holding constant urbanisation, the Green Revolution variable, and the Rajasthan variable, there is evidence of conditional convergence, with the slower states achieving faster growth than the richer states. This is the only regression result in which we find this conditional convergence.
By the 1990s, however, the Green Revolution effect has disappeared entirely, as has the conditional convergence and the fast growth of Rajasthan. The only variable that accounts for cross-state growth in the 1990s is urbanisation as of 1991, with a point estimate of the urbanisation coefficient that is somewhat higher than in the 1980s (0.30 compared with 0.13).
A simple bivariate regression of growth during 1991-98 on urbanisation in 1991, with this single variable accounting for 71 per cent of the variation. It is plausible that economic liberalisation has given an added boost to the growth-promoting benefits of urbanisation, especially for the coastal cities and the main cities engaged in IT exports (Bangalore, Chennai, Hyderabad, and Delhi).
There are major differences across states in the area of policy reform. A few of the Indian States have been more reform-oriented, such as Maharashtra, Tamil Nadu, Gujarat, Karnataka and Andhra Pradesh, but states, such as Haryana, Kerala, Orissa, Madhya Pradesh, Punjab, Rajasthan and West Bengal have wide ranging unfinished reform agendas to deal with.
Of course, Bihar and Uttar Pradesh are even further behind. Data on real annual average growth rates of per capita gross state domestic product bear testimony to the fact that four out of the five states that are more reform-oriented (with the exception of Andhra Pradesh) are also the fastest growing states in India in the post-reform period.
Interestingly enough, amongst the Southern states, both in Karnataka and Tamil Nadu, per capita incomes began to surge and exceed the national average since 1991-92. On the other hand, amongst the relatively poor reformers, Bihar, Madhya Pradesh, and Uttar Pradesh, and to a certain extent Orissa, have lagged far behind the all-India average, as also in the growth of SDP per capita of other states.
With the initiation of economic reforms in 1991, the role of private investment has acquired a great deal of significance. States are now in competition with one another to attract private investment, both domestic and foreign.
Within states, the flow of investment has tended to be skewed in favour of the urban areas. State-level data on FDI approvals suggest once again that the relatively fast moving reformers have tended to attract higher levels of foreign direct investment.
Gujarat, with a population of 50 million, received over a fifth of private investment proposals, whereas Bihar with a population of 83 million barely managed a share of 5 per cent of such proposals. Maharashtra and Gujarat account for 37 per cent of total investment proposals, while Bihar, Madhya Pradesh, Orissa, Rajasthan and Uttar Pradesh, taken together, were able to attract only 26 per cent of investment proposals.
Accounting for the lack of convergence
It is surprising, but robustly the case, that after controlling for urbanisation alone there is no evidence whatsoever of conditional convergence. We did not find any candidate explanatory variables that once controlled for left signs of conditional convergence. This poses a major issue of interpretation.
Why is it that the US states displayed unconditional convergence in most decades of US history, and similarly for Japanese prefectures and European regions, but India and China do not show signs of conditional convergence much less unconditional convergence?
There are several possible hypotheses for the lack of unconditional convergence:
- The geographical differences are larger in India and China than in the United States, Europe, and Japan;
- Population movements in the United States, Europe, or Japan more readily arbitrage differences across regions;
- Policies of the national or regional governments prevented convergence;
- Economic convergence is easier at higher levels of economic development than in China and India.
We find some merit in each of these possibilities. Certainly the intrinsic economic advantages or disadvantages of Japanese prefectures and West European regions are much smaller than in the case of either India or China.
Consider coastal access, for example. In Japan, 97 per cent of the population lives within 100km of the coast, and in Europe, more than half the population of every country in the European Union lives within 100km of the coast or an ocean-navigable waterway (like the Rhine or the Danube).
On average, 51 per cent of the population lives within 100km of the coast, and 89 per cent lives within 100km of the coast or a navigable waterway. The US, it would seem is much more like China and India, with a large proportion of the land area far from the sea or ocean navigability.
Yet a surprisingly high proportion of the US population lives within 100km of the coast or ocean-navigable waterway: an estimated 65 per cent.
Of course one of the reasons that the US has such a high proportion of the population at the coasts and along navigable waterways is that it has highly efficient agriculture, which can feed the entire population (and much more) with just 2 per cent of the labor force.
For this reason, few people in the United States are "bound to the land" in the economic sense of needing to be in the place where food is grown. With much lower food productivity in China and India (itself a reflection in part of the long history of much higher man-land ratios in Asia), a much larger part of the population is needed to produce food.
This in turn means that populations are "stuck" in the interior of the country, much less able to participate in international trade and globalised production systems (e.g. outsourcing for multinational firms). Large numbers of poor near- subsistence farmers therefore live in the hinterland of China and India, regions that are not part of convergent growth except to the extent that households migrate in large numbers.
Note that climatic variability is also much lower in Europe, the US, and Japan than it is in India and China. India has substantial proportions of population in tropical, arid, sub-tropical, and highland ecozones, whereas the overwhelming proportion of US, European, and Japanese populations are in temperate ecosones.
China has large variations as well, though China has only a small proportion of the population in tropical ecosones, which have proved most difficult for development in other parts of the world (including India).
Just as with the variation in access to the sea, the climatic variation mostly likely puts a brake on cross-regional convergence.
This brings us to the question of migration. In China, migration is limited by the household registration system, which has blocked the legal migration of families from the hinterland to urban areas.
In India migration is not restricted, yet poor families without social safety nets apparently face such high costs and risks of migration that internal labor flows are not powerful enough to create forces of convergence.
Exactly why this is the case is beyond the scope of this paper, but is certainly worthy of much closer investigation. We know too that in the case of China, the policy regime in the 1980s and 1990s favored the already-favored coastal provinces, and this accounted for part of the continuing divergence between coast and interior.
In India, no such preferential policies are readily discernible.
We also examine whether social and demographic factors could account for cross-state growth patterns. The results were surprisingly negative.
The states vary considerably in social indicators, such as infant mortality rates and adult literacy. Generally, the Southern States have outperformed the Northern states by a wide margin, and Kerala has outperformed all of the rest of the country, including its Southern neighbors.
Yet, this superior performance, while undoubtedly raising the standard of well being, has not translated during this period into discernibly higher rates of economic growth at the state level.
The simple correlation of growth with literacy, for example, is positive, but disappears once we control for urbanisation, which is correlated both with growth and with the degree of literacy.
Kerala has reaped its returns to improved human capital (literacy and health) through increased migration, and a large flow of remittance income back to the state.
This remittance income is counted as part of the state's income but not its Gross State Domestic Product, which is our measure of economic development used in our analysis. Thus, Keralites have a higher income standard that is measured by production within the state. We speculate next on why this is so.
Part II: Unravelling the mysteries of state-level performance
Jeffrey D Sachs is Director of the Center for International Development at Harvard University and the Galen L Stone Professor of International Trade at the Department of Economics.
Nirupam Bajpai is Development Advisor at CID and the Director of the Harvard India Program.
Ananthi Ramiah was a Summer Intern at CID when this study was undertaken in 2001.
YOU MAY ALSO WANT TO READ:
The Rediff Budget Special
The Rail Budget 2002-03
The Economic Survey 2001-02
Run-Up To The Budget