Growth Dynamics And Regional Income Inequality In India

Published Date: 02 Nov 2017

Spatial inequality remains an important feature of Indian economy and seems to increase with economic growth and development. The regional disparities in economic growth have attracted a lot of attention since independence. The government of India has been concerned about balanced regional growth of the country since the initial period of economic planning. Obviously, the issue of regional balance has been given importance in all the plans and policies. For achieving the objective of balance regional development, Indian constitution has given the power to central government for allocating and transferring the resources among the states. Both Finance Commission and Planning Commission have the responsibility of recommending criteria for allocation and transfer of resources between the centre and the states. The disbursement of funds has been made in such a manner that the poorer states have received proportionately larger amount of funds for development purpose relative to the richer ones (Ghosh, 1998). Despite the Indian government’s concern for inequality reduction, Indian economists have highlighted regional inequalities in the level and growth of per capita income and consumption as well as in education, health and economic infrastructures (Das, Barua and Ghosh, 1993). Furthermore, these disparities have been increasing over time. Also due to severe fiscal crisis as well as external payments crisis in 1991, the Indian government implemented comprehensive economic reforms, basically liberalization and privatization programmes. These reforms have given more emphasis to market forces in the allocation of resources in the economy, which raised worries in the economy that these market forces will allocate the benefits of growth in a skewed manner to the already better states. (give studies who mentioned of increased inequality)

In this context, it seems pertinent to analyse the growth performance thoroughly and to examine whether regional economic disparity has increased or decreased in India after the economic reforms and whether a considerable degree of regional inequality is still persisting despite the Indian government’s concern for its reduction?

Ahluwalia (1996 & 2000) argued that implementation of economic reforms has led to substantial growth in India after 1992, and both the rich and poor states have experienced the benefits of economic reforms.

The empirical studies by Cashin and Sahay (1996) and by Akkina (1996) could not find any significant absolute convergence in India. However, Bajpai and Sachs (1996) were able to find evidence of convergence only during 1960s, which may be due to the agricultural sector’s growth during the "green revolution" period.

There is another important reason for studying the Regional growth dynamics and inequality in India. The major target of the Eleventh Five-Year Plan (2007-12) was ‘faster and more inclusive growth’. ‘Faster, sustainable and more inclusive growth’ has again chosen as major target of the Twelfth Five-Year Plan (2012-2017). It has also been identified that the gains of the rapid growth have not distributed equally to all parts of the country and disparities among regions have been increasing gradually. Given the present scenario of the economy, it seems relevant to investigate how far economic growth has been ‘inclusive’ and to what extent the fruits of growth have been shared by different regions of the country.

This chapter examines the growth dynamics and income inequality in India by focusing on the differencing among 15 major states of India from 1980-81 to 2009-10. These fifteen major states are Andhra Pradesh, Bihar, Gujarat, Haryana, Himachal Pradesh, Karnataka, Kerala, Maharashtra, Madhya Pradesh, Orissa, Punjab, Rajasthan, Tamil Nadu, Uttar Pradesh and west Bengal. Jharkhand, Chattisgarh and Uttaranchal have been merged with their parent states. These 15 states comprises 93 percent of total Indian population and 85 percent of total India’s geographical area and around 82 percent of total income (GDP) of India in 2009-10. So, these 15 states can be taken as representative of India.

Data and Methodology

Neoclassical growth paradigm has been extensively used in the recent years in understanding the differences in regional growth (Solow, 1965 & Cass Koopmans, 1965). In the neo classical growth theory, convergence takes place mainly due to diminishing returns to capital. Economies that having less relative capital per worker tends to have higher growth rate and over the time inter regional differences in per capita output disappear. Barro and Sala-i-Martin (1992, 1995) and Sala-i-Martin (1996) have converted the economic notion of convergence into a well-defined statistical hypothesis. Two notions of convergence 1) β-convergence and 2) σ- convergence are extensively used in empirical works. The concept of σ-convergence concerns with decline in cross-sectional dispersion of per capita income. The existence of σ-convergence implies a tendency of per capita income to be equal across regions over time. Whether the presence of σ-convergence in per capita income is due to higher growth rates of the poorer regions than the richer ones can be examined by looking into the presence of β-convergence. β-convergence has also distinguished into two types: 1) Absolute β-convergence and 2) Conditional β-convergence. β-convergence is said to exist if the poorer regions tend to grow faster than the richer ones. In this study, the regional income inequality has been measured by using both β and σ- convergence.

However, there is an alternative literature discussing the inter relationship between regional inequality and growth belongs to new economic geography (NEG). New Economic Geography has challenged the neoclassical growth theory’s way of explaining spatial variations in economic development. This new economic geography theory is more appealing because these models are based on increasing returns to scale in production, monopolistic competition, costly inter-regional factor mobility that are fundamental to a proper understanding of spatial disparities in economic development. The main proposition of NEG is the agglomeration economies, which leads to concentration of economic activities in general and industrial activity in particular to already developed geographical locations compared to other regions. This results in urban concentration in these regions whose growth rates keep diverging from those in the peripheral regions.

In the first paper of the field, Krugman (1991) showed that regional inequalities might be persistent because of the so called home market effect: it is beneficial to locate production close to a large market as this enables to increase sales and profits

. There is some empirical literature

Marshall (1920) has suggested three sources of such

agglomeration economies. These are: (i) the sharing of inputs

whose production involves internal increasing returns to scale;

(ii) labour market pooling that allows a better match between an

employer’s needs and a worker’s skills; and (iii) knowledge spillovers

between workers. Other sources of agglomeration, that

have come up in the literature, include home market effects,

economies of consumption, etc

Another group of models discussing the interaction of regional inequality and growth belongs to the field

of new economic geography (NEG; a thorough overview of the basic NEG models can be found in Baldwin

et al., 2003).

. As splitting production between several regions is not profitable due to increasing

returns to scale, each firm produces only in one region. Costly trade causes the prices of the products to

be higher in regions that are served by exporting and, thus, the firms are able to sell smaller quantities of

their products there than in the home region. Moreover, the low prices carry over to high real wages that

attract mobile workers to the region with more firms. The wages are additionally drawn up in that region due

to the competition of the firms for workers. The home market effect appears also if the workers are assumed

to be immobile, but the products of the firms are used by other firms as intermediate inputs (Krugman and

Venables, 1995). As the result, in a two region setting the firms and workers concentrate in one of the regions

(the core) if the trade costs are sufficiently low even if the regions are initially identical in their technological

level and resource endowments. In fact, the home market effect was already present in Krugman’s

(1980) trade model without labour mobility. However, that model was unable to explain the emergence of

agglomerations of economic activity in case of symmetric regions, i.e. the inequalities that can be observed

in space.

Some further developments in the field have directly tackled the interplay of growth and regional inequality.

The first paper to address this question was Baldwin (1999). Abstracting from factor mobility, he shows

that growth can affect inequality. This is achieved by the assumption that capital (utilised by the modern

sector) depreciates and has to be replaced. Also, investment into capital construction will be done only if the

present value of its expected flow of return is at least as large as the investment costs. Another assumption is

that the constructed capital can be utilised for producing the consumer goods only in the region of construction.

The spatial equilibrium is achieved if the expected return from capital covers exactly its construction

costs: in that case there will be no growth.

β-convergence

For β-convergence, analysis can be done at two levels 1) Cross sectional analysis, 2) Panel data analysis. Under cross sectional analysis, it is assumed that the production structure is common to all states. But there are certain limitations of single cross sectional analysis. First, all information is not used because in cross section analysis all the time series information is reduced to single average observation. Second, cross sectional analysis suffer from omitted variable bias. Third, one or more of the regressors may be endogenous (Hoeffler, 2002). Without accounting for this omitted variable bias and endogeneity, the speed of convergence is potentially bias and inconsistent. This omitted variable bias and endogeneity of explanatory variables can be addressed through dynamic panel growth framework of neoclassical growth paradigm (Islam 1995, Caselli et al. 1996, Aiyer 1999, Yao and Zhang 2001 and Hoeffler 2002). The regressions with both cross-section and panel data have been estimated for absolute beta convergence while conditional convergence has been estimated with dynamic panel data model. For it, the rate of growth of GDP per capita has been regressed against their initials levels respectively.

Regression Equations:

(1)

The equation (1) can be re written as

(2)

where y indicates income GDP per capita, The equation (1) will predict the relation for absolute convergence with assumption that all economies will reach to the same steady state income levels. In simple words, the poor economies will grow faster than the richer ones if all economies are structurally identical and have access to the same technology. This is the hypothesis of absolute β−convergence that is usually tested by the model given by equation (1). There will be absolute β−convergence when the estimate of β is significantly negative.

For estimating conditional convergence, assumption of identical steady state has been relaxed here i.e. the different economies reach to different steady state income levels depending on its rate of saving, depreciation, population growth and technology. There will be conditional β−convergence if the estimate of β is significantly negative after keeping the other control variables in the regression equation.

For the panel data analysis, 5 year rolling sub-periods has been used and equation takes the form

(3)

Where is a period-specific dummy and is a state-specific effect and remaining variables are same as in cross section analysis. We can re-write equation (3) as

(3a)

(3b)

This equation (3b) is known as fixed effect model. The fixed effects formulation allows controlling the unobserved differences between the steady states of regions. However, this equation has simultaneity problem and therefore the OLS estimators will be inconsistent. There is an alternative model to the fixed effects model. This is called the Generalised Methods of Moments Model (GMM). The method is to take first-difference of the basic growth equation to eliminate the fixed regional effects and then uses instrumental variables to address the correlation between the differenced lagged dependent variable and the induced first order moving average error term. The first difference model can be written as

(3c)

Here the component of is correlated withand thereby implying that OLS estimators will be inconsistent. So, GMM approach has been used to estimate it, which is based on the use of lag value of series as instruments for lagged first differences.

Sometimes the lagged levels are poor instruments for the first-differenced regressors. In this case, the augmented version- "system GMM" should use. The system GMM estimator uses the levels equation to obtain a system of two equations: one differenced and one in levels. By adding the second equation additional instruments can be obtained. Thus the variables in levels in the second equation are instrumented with their own first differences. This usually increases efficiency. So, this study uses the System GMM to estimate conditional convergence.

These all regressions have been estimated at aggregate level for GDP per capita and also for sectoral GDP per capita i.e. for Agriculture, Industry and Services separately. This will show the sectoral convergence sources i.e. which sector is converging and which are diverging.

Here all these regressions assume each state as an independent spatial unit, without considering possible spillover effects among economies of different states.

σ- convergence

From literature, it has been found that beta convergence is a necessary but not a sufficient condition for sigma convergence (Barro and Sala-i-Martin 1995). Other things being equal, beta convergence may eventually lead to sigma convergence. But if other things are not equal may be because each region is subject to random disturbances, then beta convergence need not lead to sigma convergence i.e. a reduction in dispersion of per capita income levels over time.

For measuring σ-convergence, there are several indices suggested in literature like Gini Coefficient, the coefficient of variation, the weighted coefficient of variation, the variance of logarithmic income, standard deviation of log of variables and Theil indices. However this study employs standard deviation of log of SDP per capita and log of per capita income of different sectors and their movement over time is studied.

GSDP data has been taken from CSO and converted in to same base at 1999-00 prices by splicing method for the analysis. Population data has also been obtained from CSO. Workers data has been extracted from unit level data of various quinquennial rounds of NSSO, Employment- Unemployment Survey in India, Schedule 10. The other control variables used in conditional convergence analysis are literacy rate as a proxy for human capital, agricultural productivity as a proxy for economic structure of the economy, per capita capital expenditure and per capita outstanding credits extended by All Scheduled Commercial Banks [1] (SCBs) as a proxy for investment. The literacy rate has been calculated from unit level data of NSSO, quinquennial rounds and for in between years, it has been interpolated. Agricultural productivity has been measured as agricultural SDP divided by agricultural workers. Agricultural SDP has been obtained from CSO and agricultural workers have also been calculated from unit level NSSO. Per capita outstanding credit extended by all Schedule Commercial Banks has been obtained from various issues of Banking Statistics, Basic Statistical Returns, Reserve Bank of India and Per capita state government capital expenditure has collected from various issues of State Finances, Reserve Bank of India. Both per capita government capital expenditure and per capita outstanding credit have also been measured at 1999-00 prices by using GDP implicit deflator.

The study examines the convergence in per capita GDP across major 15 states over the period from 1980-81 to 2009-10 by employing dynamic panel model. Since using annual time data on per capita real GDP are very short to study the convergence analysis and also it has the disadvantage of increasing serial correlation due to business cycle effect and shocks. Similarly using longer time intervals reduce the number of observations and increase the probability of obscuring changes in the steady state that have occurred during the period. This will accentuate the ‘Bias’ in dynamic fixed effect panel model (Nickell, 1981). Therefore keeping all these in mind, we have divided the total time period from 1980-81 to 2009-10 into five year shorter time period for applying the dynamic fixed effect panel model. We constructed six panels, with each covering the five year period: 1980-81 to 1984-85, 1984-85 to 1989-90, 1989-90 to 1994-95, 1994-95 to 1999-00, 1999-00 to 2004-05 and 2004-05 to 2009-10. However the beginning period i.e. 1980-81 to 1984-85 is kept as four year period due to lack of data. The dependent variable is the natural log of per capita income at the end point of each five year span while the independent variable is natural log of per capita income at the beginning of each five year period. Total number of observations in dynamic fixed effect panel model is 90 as pooling across 15 major states and 6 shorter time periods. Conditional convergence has been examined by using both Least Square Dummy Variable (LSDV) and Generalised Methods of Moments (GMM) estimates. In spite of Superiority of GMM estimates, the LSDV estimates are also calculated to study the dispersion of steady state of income i.e. fixed effects across the states. GMM has been estimated by "xtabond2" command written by Roodman (2006) in Stata 12.0.

Results

State wise Sectoral GSDP Growth

Growth rates of GSDP across states have large differences. Some states have witnessed extraordinary high growth, while the others lagged behind the all-India growth rate. Here we have done this analysis for 15major states only. Three newly created states namely, Chattisgarh, Jharkhand and Uttaranchal are merged with their parent states namely Bihar, Madhya Pradesh and Uttar Pradesh respectively. Therefore these three states refer to undivided states. The comparative average growth rates of SDP for 15 major states at 1999-00 prices for the decades 1980s (1980-81 to1989-90), 1990s (1990-91 to 1999-2000) and 2000s (2000-01 to 2009-10) as well for the overall period (1980-81 to 2009-10) are given in Table 1.

It can be seen that seven of the fifteen states -Andhra Pradesh, Bihar, Himachal Pradesh, Kerala, Madhya Pradesh, Uttar Pradesh and West Bengal – have less than 5 percent and other eight states have more than 5 percent growth rate during the 1980s against the all-India growth rate of 5.4 percent per annum. Haryana and Rajasthan have progressed rapidly during the 1980s with over six percent compound annual growth rate and Kerala recorded the lowest growth rate of 3.2 percent per annum. However, during the 1990s highly industrialized states like Gujarat, Karnataka and Maharashtra, grew at over 8 and nearly 7 percent per annum respectively. Among other major states, West Bengal, Tamil Nadu, Rajasthan and Himachal Pradesh have performed well with over all India 6.21 percent growth rates per annum.

It is interesting to note that West Bengal which was growing at less than all India growth during 1980s, has grown not only faster than the all-India average, but also than many other states which was growing at higher than all India average in 1980s, such as, Haryana, Punjab. These two states have in fact grown much slower than the all-India average during 1990s. The poor performance of both Punjab and Haryana may be attributed to stagnation in agriculture. Some of the high growing states, namely Gujarat, Maharashtra, Karnataka and Tamil Nadu, got large foreign investment in the 1990s. On the other hand, poor states like Bihar, Orissa and Uttar Pradesh have attracted less foreign investment. This might have held down the growth in these states.

During 2000s decade, seven states, namely, Gujarat, Maharashtra, Orissa, Haryana, Tamil Nadu, Kerala and Andhra Pradesh have grown at faster rates than all India 8 percent growth rate. Gujarat and Maharashtra has performed very well during 2000s with over 10 percent growth rate while Punjab has grown at only 6.3 percent per annum; however Punjab growth rate has increased than during 1990s. It is interesting to note here that Haryana has again attained its pre reform position of above all India average growth rate. It is also very surprising that Orissa economy has also revived during this decade and achieved 9.42 percent growth after having low growth rate for last decade.

Table 1:State Wise Sectoral Growth Rate of Gross Domestic Product

 

 

1980-81 to

1989-90

1990-91 to

2000-01

2000-01 to

2009-10

1980-81 to

2009-10

Andhra Pradesh

Agriculture

2.05

2.63

5.14

3.29

Industry

5.22

6.00

9.52

6.21

Services

5.92

6.46

9.22

6.88

GDP

4.18

5.18

8.24

5.57

Bihar

Agriculture

2.78

0.05

3.14

1.82

Industry

6.20

4.67

6.57

4.03

Services

5.50

4.59

9.03

5.65

GDP

4.60

2.98

6.75

3.89

Gujarat

Agriculture

-0.84

4.44

7.64

2.85

Industry

7.88

9.89

11.01

8.02

Services

7.55

8.82

11.01

8.02

GDP

5.05

8.30

10.50

6.77

Haryana

Agriculture

3.77

1.87

3.27

3.16

Industry

9.54

6.12

8.93

7.35

Services

7.69

7.58

12.25

8.88

GDP

6.20

4.94

9.17

6.47

Himachal Pradesh

Agriculture

1.31

0.65

3.86

2.88

Industry

7.34

11.14

8.83

8.85

Services

7.02

6.87

8.97

6.97

GDP

4.70

6.31

7.74

6.21

Karnataka

Agriculture

2.76

4.44

2.79

2.39

Industry

6.56

7.99

8.84

7.45

Services

7.48

9.12

9.50

8.42

GDP

5.36

7.32

7.89

6.35

Kerala

Agriculture

1.46

2.24

0.74

2.35

Industry

2.44

6.93

9.34

6.26

Services

4.97

7.38

9.93

7.19

GDP

3.16

5.85

8.25

5.81

Madhya Pradesh

Agriculture

1.64

2.91

4.99

2.25

Industry

5.56

7.28

8.61

7.05

Services

6.39

5.69

6.51

5.77

GDP

4.02

5.25

6.75

4.94

Maharashtra

Agriculture

3.13

4.35

4.75

3.93

Industry

6.10

6.66

10.83

6.32

Services

6.76

8.55

10.61

8.48

GDP

5.64

7.12

10.00

6.96

Orissa

Agriculture

2.78

1.76

4.05

0.89

Industry

7.74

4.89

13.04

6.77

Services

6.54

6.41

10.08

6.74

GDP

5.01

4.32

9.42

4.65

Punjab

Agriculture

4.95

2.54

2.29

3.10

Industry

7.50

6.99

10.10

6.82

Services

4.73

5.91

6.95

5.55

GDP

5.44

4.72

6.31

4.93

Rajasthan

Agriculture

3.32

3.43

3.98

3.54

Industry

7.35

9.68

7.93

8.10

Services

8.91

7.68

7.58

7.22

GDP

6.01

6.60

6.83

6.13

Tamil Nadu

Agriculture

3.40

2.95

2.80

2.75

Industry

4.73

6.01

7.55

5.69

Services

6.56

8.95

10.23

7.86

GDP

5.18

6.64

8.49

6.20

Uttar Pradesh

Agriculture

1.64

4.08

1.92

3.01

Industry

5.56

6.75

9.83

6.84

Services

6.39

4.78

7.79

5.75

GDP

4.02

5.04

6.61

5.03

West Bengal

Agriculture

6.03

5.20

2.57

4.15

Industry

3.34

5.97

6.57

5.85

Services

4.76

8.44

8.34

7.14

GDP

4.70

6.82

6.56

5.96

All India

Agriculture

2.93

3.15

2.96

2.86

Industry

6.88

6.56

8.65

6.70

Services

6.65

8.04

9.52

7.71

GDP

5.41

6.21

8.00

6.12

State wise Sectoral Composition and Sectoral Contribution to Growth

Sectoral growth rates of GSDP, composition of GDP and sectoral contribution to GDP growth are given in tables 1, 2 and 3 respectively.

Table 2: Structure of the Economy

 

 

1980-81

1990-91

2000-01

2009-10

Andhra Pradesh

Agriculture

40.21

34.88

30.16

22.48

Industry

19.34

22.41

22.61

24.70

Services

40.46

42.71

47.23

52.82

Bihar

Agriculture

43.73

36.16

31.43

20.91

Industry

20.87

24.36

21.06

21.34

Services

35.41

39.48

47.51

57.75

Gujarat

Agriculture

37.04

26.37

14.99

12.26

Industry

28.59

35.46

40.28

40.32

Services

34.37

38.17

44.73

47.42

Haryana

Agriculture

45.74

40.39

30.36

17.91

Industry

25.36

26.95

27.94

27.92

Services

28.90

32.66

41.70

54.17

Himachal Pradesh

Agriculture

43.19

36.87

24.79

16.54

Industry

26.92

27.70

36.59

40.54

Services

29.90

35.43

38.62

42.91

Karnataka

Agriculture

44.66

34.91

28.25

17.03

Industry

20.72

23.35

24.16

26.69

Services

34.62

41.73

47.59

56.28

Kerala

Agriculture

36.95

28.42

22.00

11.54

Industry

18.42

18.98

21.01

21.84

Services

44.62

52.60

57.00

66.63

Madhya Pradesh

Agriculture

44.56

35.76

22.62

20.97

Industry

19.73

23.35

28.87

31.34

Services

35.72

40.90

48.51

47.69

Maharashtra

Agriculture

23.88

19.70

15.30

10.72

Industry

30.69

32.06

27.57

27.08

Services

45.43

48.24

57.13

62.21

Orissa

Agriculture

54.92

39.22

28.80

19.52

Industry

17.00

24.65

24.50

30.47

Services

28.07

36.13

46.70

50.01

Punjab

Agriculture

43.14

43.74

36.63

25.67

Industry

16.53

20.11

23.13

30.10

Services

40.33

36.15

40.24

44.23

Rajasthan

Agriculture

45.47

42.78

28.42

21.47

Industry

18.68

20.02

27.79

30.23

Services

35.85

37.21

43.79

48.29

Tamil Nadu

Agriculture

24.55

21.54

16.71

9.74

Industry

32.35

32.45

30.45

27.09

Services

43.10

46.00

52.83

63.17

Uttar Pradesh

Agriculture

41.20

36.74

33.43

22.57

Industry

20.22

21.75

22.68

27.34

Services

38.58

41.52

43.89

50.09

West Bengal

Agriculture

34.85

33.72

29.03

20.97

Industry

19.67

18.78

18.16

17.62

Services

45.47

47.50

52.82

61.41

All India

Agriculture

38.59

32.16

23.89

15.52

Industry

21.79

25.45

25.80

26.55

Services

39.62

42.39

50.31

57.93

Table 3: State wise Sectoral Contribution to Growth

 

 

1980-81

to

1989-90

1990-91

to

2000-01

2000-01

to

2009-10

1980-81

to

2009-10

Andhra Pradesh

Agriculture

29.92

17.28

14.14

17.98

Industry

21.90

25.59

27.26

25.82

Services

48.18

57.13

58.60

56.20

Bihar

Agriculture

14.51

-18.33

6.84

11.99

Industry

35.05

41.21

18.89

20.21

Services

50.44

77.12

74.27

67.80

Gujarat

Agriculture

12.82

3.24

10.31

7.76

Industry

39.05

48.96

41.10

42.80

Services

48.13

47.80

48.60

49.45

Haryana

Agriculture

28.10

13.68

7.48

12.45

Industry

33.12

31.36

27.81

28.77

Services

38.78

54.96

64.71

58.78

Himachal Pradesh

Agriculture

23.89

0.61

9.13

10.60

Industry

28.28

52.60

45.20

45.07

Services

47.82

46.79

45.67

44.34

Karnataka

Agriculture

24.77

25.69

3.90

11.16

Industry

26.73

23.54

30.82

28.44

Services

48.50

50.77

65.28

60.41

Kerala

Agriculture

10.65

11.24

1.80

5.85

Industry

20.35

23.11

23.91

22.76

Services

69.00

65.64

74.28

71.39

Madhya Pradesh

Agriculture

13.69

15.03

18.79

14.13

Industry

29.90

35.09

34.98

35.17

Services

56.41

49.88

46.23

50.70

Maharashtra

Agriculture

19.26

9.16

6.90

8.11

Industry

30.93

25.78

26.98

26.37

Services

49.81

65.06

66.12

65.52

Orissa

Agriculture

35.62

13.42

12.20

9.22

Industry

24.79

27.75

33.95

33.61

Services

39.59

58.83

53.86

57.18

Punjab

Agriculture

43.81

26.62

11.15

20.89

Industry

24.70

28.36

39.25

33.65

Services

31.49

45.02

49.60

45.46

Rajasthan

Agriculture

33.40

7.41

13.64

16.66

Industry

19.85

41.63

33.56

32.71

Services

46.75

50.97

52.80

50.63

Tamil Nadu

Agriculture

19.36

9.70

1.94

6.71

Industry

27.37

25.53

22.80

25.56

Services

53.27

64.77

75.27

67.73

Uttar Pradesh

Agriculture

14.73

27.99

8.08

16.10

Industry

27.70

25.93

34.18

30.38

Services

57.57

46.08

57.74

53.52

West Bengal

Agriculture

39.85

24.60

11.36

17.65

Industry

13.50

15.90

17.20

17.06

Services

46.65

59.51

71.45

65.29

All India

Agriculture

22.50

13.40

6.83

10.58

Industry

30.32

24.68

27.35

27.50

Services

47.18

61.92

65.82

61.92

It can be seen that the share of agriculture sector to total SDP has decreased drastically from about over 40 percent in the early 1980s to even less than 20 percent except for some states like Punjab (25 percent), Uttar Pradesh (22.5 percent) and Rajasthan (21.5 percent) in 2009-10. In states, such as Gujarat, Maharashtra, Kerala and Tamil Nadu, the share of agriculture sector in GSDP has come down to around 10 percent in 2009-10.

Even in the poorer states like Bihar, Madhya Pradesh, Orissa, Rajasthan and Uttar Pradesh, the share of the agriculture sector to total GDP has declined significantly over the last three decades. In Bihar the share of agriculture sector which was nearly 44 percent in GDP in the 1980-81 has now come down to about 20 percent in the late 2009-10. This change has mainly brought by poor performance of agriculture rather than a rapid growth of industry and services sector.

The contribution of agriculture sector to growth has reduced drastically in all the states except Madhya Pradesh. It has reduced to even less than half or one third during 2000s of what it was during 1980s. The drastic reduction in the contribution of agriculture sector in GDP growth during this period is partly on account of faster growth in industry and services and partly on account of decreased in share of agriculture sector in total GDP. In Madhya Pradesh, agriculture contribution to growth has increased because of increase in agriculture growth rate during this period however the share of agriculture in total GDP has also decreased over this period. The service sector contribution to GDP growth has increased tremendously over the period. This is combination of both tremendous increase in Service sector growth and its share in total GDP over time. The contribution of industrial sector in GDP growth has also increased, though small increase, except for some states likes Bihar, Haryana, Maharashtra and Tamil Nadu.

Gujarat is the only exceptional state where the industrial sector has become the largest sector with around 40 percent share in GDP in 2009-10 and its contribution to growth has also increased. This is because of fastest acceleration in the industrial sector growth in Gujarat from 7.8 to 9.8 and then further to 11 percent in 1980s, 1990s and 2000s respectively. In Himachal Pradesh also, the share of industrial sector has risen to 40 percent in 2009-10. In Maharashtra, a major industrial state, the share of the industrial sector has in fact decreased from 30 percent in 1980-81 to 27 percent in 2009-10.

In an agriculturally prosperous state of Punjab, the share of agriculture sector to GDP and also contribution to growth has also declined. This is because of slackening of agricultural growth in Punjab in the 1990s and 2000s. It is interesting to note that the industrial sector grew at higher or similar rate like services sector in all the states in 2000s except Bihar, Haryana, Karnataka, Tamil Nadu and West Bengal. As a result, the share of the industrial sector in these states has increased marginally during 2000s. For example in Uttar Pradesh, the share of the industrial sector has increased from 20 percent in 1980-81 to 27 percent in 2009-10.

During the last two decades, the services sector has recorded the fastest growth in most states. The structure of the economy in all the states has shifted away from agriculture towards Services sector. The share of the services sector now exceeds 50 percent of GDP in all states except Gujarat, Himachal Pradesh, Punjab and Rajasthan. Similarly more than 60 percent of growth has been contributed by the services sector except for Gujarat, Himachal Pradesh, Punjab and Rajasthan. So, it can be concluded that service sector has become the engine of growth in most of the states and also at all India level.

Per Capita SDP Growth Rate

Table 4: State wise Per Capita SDP Growth Rate

 

1980-81

to

1989-90

1990-91

to

2000-01

2000-01

to

2009-10

1980-81

to

2009-10

Andhra Pradesh

1.94

3.63

7.07

3.94

Bihar

2.39

0.62

4.99

1.67

Gujarat

2.97

6.28

8.78

4.78

Haryana

3.66

2.37

7.13

4.00

Himachal Pradesh

2.83

4.52

5.94

4.36

Karnataka

3.27

5.53

6.60

4.66

Kerala

1.71

4.93

7.34

4.75

Madhya Pradesh

1.60

3.04

4.79

2.74

Maharashtra

3.27

4.89

8.32

4.81

Orissa

3.14

2.73

8.25

3.08

Punjab

3.49

2.76

4.42

2.99

Rajasthan

3.28

3.98

4.85

3.66

Tamil Nadu

3.64

5.43

7.62

5.01

Uttar Pradesh

1.65

2.65

4.60

2.75

West Bengal

2.44

4.95

5.36

4.15

For a better understanding of regional disparities, we have analysed growth of per capita SDP also. The growth of per capita SDP for fifteen major states is presented in Table 4. It may be seen that in the 1980s, Haryana recorded the highest per capita SDP growth rate 3.66 percent per annum followed by Tamil Nadu and Punjab and Madhya Pradesh the lowest per capita SDP growth 1.6 percent per annum followed by Uttar Pradesh 1.65 percent per annum. In the 1990s, Gujarat has highest per capita GDP growth 6.28 percent per annum.

Karnataka and Tamil Nadu have also grown over 5 percent per annum during the post reform era. Kerala and Gujarat’s performance are particularly worth mentioning, as the growth rate of Kerala has jumped from a slow growth 1.71 in 1980s to 4.95 in the 1990s and Gujarat has jumped from a moderate 2.97 in the 1980s to highest 6.28 percent in the 1990s.

During the last 2000s decade, Gujarat, Maharashtra and Orissa progressed sharply with over 8 percent per capita SDP growth per annum. Bihar, Rajasthan, Madhya Pradesh, Uttar Pradesh and Punjab have grown even less than 5 percent growth per annum. The performance of Orissa is notable from 2.73 percent in 1990s decade to 8.25 in 2000s decade.

Convergence Analysis

β- Convergence

At the outset, we estimated the absolute or unconditional β- Convergence in both cross-section and panel dimension. It assumes that all the regions will converge to same steady state. So, statistically a common intercept is assumed for all the regions as shown by equation (1). The equation (1) broadly involves regressing "growth in income" on the "initial level of income". The equation (1) may be re-written as equation (2). The results of cross-section regressions show by equation (2) for the aggregate per capita GSDP and sectoral per capita GDP for the entire period between 1980-81 and 2009-10 and also for the sub-periods (1980-81 to 1989-90, 1990-91 to 1999-00 and 2000-01 to 2009-10) are shown by table 5. It has been found that for the total per capita GSDP for entire period, the coefficient of the initial income term, 1+β, is a statistically significant 1.076, with the implied rate of convergence, λ, -0.0026. From this the estimated β equals 0.076. The sign of beta coefficient is positive which implies unconditional divergence. Similarly, it may be seen from table that the sign of the beta coefficient for total income is negative only for the sub- period from 1990-91 to 1999-00. This implies unconditional convergence for total income during 1990-91 to 1999-00.

Table 5: Absolute Convergence or Divergence in GSDP: Cross Section Regression (OLS)

Category

Period

Coefficient of Log of Initial Income,1+β, (β)

Standard Error

P-Values

R-Square

Implied λ (rate of Convergence)

Agriculture

1980-81 to 1989-90

1.069

(0.069)

0.103

0.000

0.891

-0.0077

Constant

-0.465

0.855

0.596

1990-91 to 1999-00

0.963

(-0.037)

0.142

0.000

0.779

0.0041

Constant

0.361

1.191

0.767

2000-01 to 2009-10

0.837

(-0.163)

0.106

0.000

0.828

0.0181

Constant

1.519

0.892

0.112

1980-81 to 2009-10

0.917

(-0.083)

0.171

0.000

0.688

0.0029

Constant

0.976

1.420

0.504

Industry

1980-81 to 1989-90

1.026

(0.026)

0.087

0.000

0.915

-0.0029

Constant

0.149

0.662

0.825

1990-91 to 1999-00

1.102

(0.102)

0.094

0.000

0.914

-0.0114

Constant

-0.422

0.758

0.587

2000-01 to 2009-10

1.072

(0.072)

0.068

0.000

0.949

-0.0080

Constant

-0.023

0.578

0.969

1980-81 to 2009-10

1.267

(0.276)

0.196

0.000

0.761

-0.0092

Constant

-0.657

1.504

0.670

Services

1980-81 to 1989-90

0.931

(-0.069)

0.085

0.000

0.901

0.0076

Constant

0.943

0.695

0.198

1990-91 to 1999-00

1.242

(0.242)

0.120

0.000

0.892

-0.0268

Constant

-1.624

1.026

0.137

2000-01 to 2009-10

1.150

(0.150)

0.089

0.000

0.927

-0.0167

Constant

-0.715

0.805

0.391

1980-81 to 2009-10

1.214

(0.214)

0.279

0.001

0.593

-0.0074

Constant

-0.221

2.275

0.924

GSDP

1980-81 to 1989-90

1.062

(0.062)

0.043

0.000

0.979

-0.0069

Constant

-0.286

0.393

0.479

1990-91 to 1999-00

0.914

(-0.086)

0.075

0.000

0.920

0.0096

Constant

1.174

0.705

0.120

2000-01 to 2009-10

1.103

(0.103)

0.080

0.000

0.935

-0.0114

Constant

-0.485

0.788

0.548

1980-81 to 2009-10

1.076

(0.076)

0.130

0.000

0.839

-0.0026

Constant

0.494

1.191

0.685

For the sub-sectors, Agriculture, Industry and services, for the entire period and sub-periods, the cross section convergence estimates reveal the following. There are signs of absolute convergence in the agriculture sector for the entire period and for two sub-periods also as may be seen from table that the sign of beta coefficient is negative for agriculture sector for the entire period and also for sub periods except from 1980-81 to 1989-90. Though the evidence is in favour of convergence for agricultural sector, same does not hold for industrial and services sector for the entire period and also during any of the sub-periods except for services sector during 1980-81 to 1989-90. The sign of beta coefficient for services sector during 1980-81 to 1989-90 is negative and significant, which implies the absolute convergence in services sector for this sub-period only.

Moreover we have estimated the panel data ordinary least square regression for estimating absolute convergence for the entire period between 1980-81 and 2009-10 for total and sectoral per capita income. The table 6 shows the results for unconditional convergence on panel dimension.

Table 6: Unconditional Convergence: Panel Data Regression (OLS)

Coefficient

1+β

(β)

Standard Error

P-Values

R-Square

Implied λ (rate of Convergence)

Agriculture

Log Per Capita GDP 1980-81

0.951

(-0.049)

0.038

0.000

0.875

0.0097

Constant

0.455

0.322

0.161

Industry

Log Per Capita GDP 1980-81

1.087

(0.087)

0.023

0.000

0.961

-0.0175

Constant

-0.478

0.188

0.013

Services

Log Per Capita GDP 1980-81

1.143

(0.143)

0.021

0.000

0.971

-0.0287

Constant

-0.987

0.183

0.000

GSDP

Log Per Capita GDP 1980-81

1.141

(0.141)

0.022

0.000

0.967

-0.0283

Constant

-1.161

0.215

0.000

It may be seen from table that for the total income, the coefficient on the log of per capita income1980-81, 1+β, is 1.141, with an implied convergence rate -0.0283. However, the sign of coefficient of β (0.141) is positive, which implies absolute divergence at the rate of 2.83 percent per annum over the five year period. This reconfirms the finding of unconditional convergence from cross section regression analysis for the total income for the entire period from 1980-81 to 2009-10.

Across the sub-sectors, Agriculture, Industry and services, for the entire period, the pooled regression convergence estimates reveal that the β coefficient is negative only for agriculture sector, -0.049, with an implied convergence rate 0.0097. This implies the absolute convergence in agriculture sector at the rate of 0.97 percent per annum over a five year period. For the other two sectors; industry and services, the absolute divergence is taking place at the rate of 1.75 and 2.87 percent per annum respectively. The rate of divergence in service sector is comparatively higher than industrial sector which indicates the distribution of growth in service sector is more unequal comparatively.

In sum, both panel data and cross section regression estimates provide the same pattern of absolute β-convergence or divergence. For the total income over the entire period both do not provide any evidence of absolute β-convergence, which indicates that states are not converging to same steady states.

Conditional β-Convergence

Conditional convergence assumes that different states do not converge to a single steady state income rather they converge to their own steady state income. For measuring it we include control variables in regression equation as shown by equation (3). For estimating conditional convergence equation (3b) has been measured with dynamic panel fixed effect model both LSDV (Least square Dummy Variable) and System GMM for total income as well as sectoral incomes. First we will discuss fixed effect model i.e. LSDV for total and sectoral incomes. The result for total income has been shown by table 7. It may be seen from table that the coefficient of lagged per capita GSDP is 0.798 and given the way the equation is specified, this gives estimate of β equals -0.202. Hence, the coefficient of lagged income, in case of growth of GSDP as dependent variable as shown in equation (3) is negative and statistically significant. This indicates the presence of conditional β-convergence i.e. poor states grow faster than the richer ones when the factors that affect steady state are controlled. The implied rate of conditional convergence is 4.04 percent per annum. The coefficient of per capita investment is positive and statistically significant at 5 percent level of significance. This indicates that the per capita investment has positive effect on steady state level of income and hence on growth of the economy. The result is similar to as suggested by economic theory. The coefficient of literacy rate and agricultural productivity are also positive and significant at 10 percent level of significance. The signs of both coefficients also in line with economic intuition which suggest a positive relationship between income levels of the economy and both of these variables separately. The overall explanatory power of model measured by R-square is 98 percent.

Table 7: Fixed Effect Regression Model: Least Square Dummy Variable

Dependent Variable→

Log Per capita GSDP t

Independent Variables

↓

Coefficient

1+β

(β)

Standard

Error

P-Values

Log Per capita GSDP t-4

0.798

(-0.202)

0.060

0.000

Literacy Rate

0.037

0.082

0.653

Agricultural Productivity

0.299

0.052

0.000

Per Capita Investment

0.111

0.026

0.000

Bihar

0.058

0.032

0.078

Gujarat

0.032

0.040

0.418

Haryana

-0.175

0.042

0.000

Himachal Pradesh

0.021

0.058

0.718

Karnataka

-0.031

0.026

0.246

Kerala

-0.176

0.049

0.001

Madhya Pradesh

0.101

0.034

0.004

Maharashtra

0.002

0.035

0.945

Orissa

0.029

0.045

0.527

Punjab

-0.318

0.057

0.000

Rajasthan

-0.007

0.032

0.815

Tamil Nadu

0.051

0.039

0.200

Uttar Pradesh

-0.023

0.033

0.489

West Bengal

-0.145

0.037

0.000

Constant

-1.880

0.376

0.000

R-Square

0.9859

Implied λ (rate of Convergence)

0.0404

Note: Standard Errors are heteroskedasticity Robust. Investment = Per capita outstanding credit by SCBs + Per Capita State Govt. Capital Expenditure. Andhra Pradesh is the bench mark state.

Table 8: Fixed Effect Regression Model: Least Square Dummy Variable

Dependent Variable →

Log Per capita Agricultural GSDP t

Independent Variables

↓

Coefficient

1+β

(β)

Standard

Error

P-Values

Log Per capita Agricultural GSDP t-4

0.320

(-0.680)

0.096

0.001

Literacy Rate

0.101

0.116

0.387

Per Capita Investment

0.063

0.031

0.045

Bihar

-0.408

0.094

0.000

Gujarat

-0.132

0.078

0.093

Haryana

0.256

0.067

0.000

Himachal Pradesh

0.057

0.078

0.468

Karnataka

-0.058

0.065

0.375

Kerala

-0.168

0.078

0.036

Madhya Pradesh

-0.209

0.062

0.001

Maharashtra

-0.269

0.067

0.000

Orissa

-0.130

0.073

0.082

Punjab

0.390

0.080

0.000

Rajasthan

-0.053

0.049

0.284

Tamil Nadu

-0.337

0.080

0.000

Uttar Pradesh

-0.241

0.060

0.000

West Bengal

-0.056

0.060

0.359

Constant

4.919

0.722

0.000

R-Square

0.9289

Implied λ (rate of Convergence)

0.1361

The result of LSDV for sectoral incomes, agricultural, industrial and services have been shown by tables 8, 9 and 10 respectively. It may be seen from table 8 that the coefficient of lagged per capita agricultural GSDP is 0.320 and this gives estimate of β equals -0.680. Hence, the coefficient of lagged income, in case of growth of GSDP as dependent variable as shown in equation (3) is negative and statistically significant. This indicates the presence of conditional β-convergence in agricultural sector. The implied rate of conditional convergence is 13.6 percent per annum. The coefficient of per capita investment is positive and statistically significant at 10 percent level of significance. This indicates that the per capita investment has positive effect on steady state level of income and hence on growth of the economy. The result is similar to as has been found for total income. The coefficient of literacy rate is positive but not significant even at 10 percent level of significance. However, the sign of the coefficient is same as suggested by economic theory. The overall explanatory power of model measured by R-square is 92 percent.

Table 9: Fixed Effect Regression Model: Least Square Dummy Variable

Dependent Variable →

Log Per capita Industrial GSDP t

Independent Variables

↓

Coefficient

1+β

(β)

Standard

Error

P-Values

Log Per capita Industrial GSDP t-4

0.864

(-0.136)

0.101

0.000

Literacy Rate

0.367

0.217

0.095

Per Capita Investment

0.083

0.039

0.035

Bihar

-0.001

0.083

0.994

Gujarat

0.056

0.053

0.297

Haryana

0.014

0.049

0.772

Himachal Pradesh

0.027

0.100

0.791

Karnataka

-0.041

0.040

0.308

Kerala

-0.212

0.126

0.096

Madhya Pradesh

0.026

0.065

0.684

Maharashtra

-0.130

0.055

0.021

Orissa

0.021

0.058

0.719

Punjab

-0.049

0.053

0.360

Rajasthan

0.080

0.055

0.146

Tamil Nadu

-0.111

0.047

0.020

Uttar Pradesh

0.001

0.068

0.986

West Bengal

-0.158

0.082

0.056

Constant

-0.794

0.336

0.021

R-Square

0.9704

Implied λ (rate of Convergence)

0.0271

Table 9 shows that the coefficient of lagged per capita industrial GSDP is 0.864 and this gives estimate of β coefficient equals -0.136, which is negative and statistically significant. This indicates the presence of conditional β-convergence in industrial sector. The implied rate of conditional convergence is 2.71 percent per annum. The coefficient of per capita investment is positive and statistically significant at 5 percent level of significance. This indicates that the per capita investment has positive effect on growth of the economy. The result is similar to as has been found for both total income and agricultural income. Similar to agricultural sector, the coefficient of literacy rate is positive but not significant even at 10 percent level of significance. However, the sign of the coefficient is same as suggested by economic theory. The overall explanatory power of model measured by R-square is 97 percent.

Table 10: Fixed Effect Regression Model: Least Square Dummy Variable

Dependent Variable →

Log Per capita Services GSDP t

Independent Variables

↓

Coefficient

1+β

(β)

Standard

Error

P-Values

Log Per capita Services GSDP t-4

0.979

(-0.021)

0.066

0.000

Literacy Rate

0.184

0.111

0.103

Per Capita Investment

0.107

0.037

0.006

Bihar

0.120

0.050

0.020

Gujarat

-0.007

0.039

0.866

Haryana

0.021

0.048

0.670

Himachal Pradesh

-0.067

0.063

0.291

Karnataka

0.001

0.041

0.976

Kerala

-0.087

0.066

0.191

Madhya Pradesh

0.010

0.054

0.853

Maharashtra

-0.092

0.040

0.026

Orissa

0.056

0.045

0.215

Punjab

-0.164

0.036

0.000

Rajasthan

0.048

0.061

0.439

Tamil Nadu

-0.039

0.049

0.435

Uttar Pradesh

0.021

0.039

0.593

West Bengal

-0.021

0.049

0.664

Constant

-1.179

0.256

0.000

R-Square

0.9811

Implied λ (rate of Convergence)

0.0041

It can be seen from Table 10 that the coefficient of lagged per capita services GSDP is 0.979 and this gives estimate of β coefficient equals -0.021, which is negative and statistically significant. This indicates the presence of conditional β-convergence in services sector. The implied rate of conditional convergence is 0.41 percent per annum. The results of coefficients of per capita investment and literacy rate are similar to total, agricultural and industrial income. The overall explanatory power of model measured by R-square is 98 percent.

Because the lagged dependent variable is used as an explanatory variable in the estimation, the GMM estimation of convergence coefficients has also been computed to obtain more efficient results. The results of GMM for total as well as sectoral incomes have been shown by table 11.

Table11: Generalised Methods of Moments Estimation

Dependent Variable →

Log Per capita GSDP t

Independent Variables

↓

Coefficient

1+β

(β)

Standard Error

P-Values

Implied λ (rate of Convergence)

Log Per capita GSDP t-4

0.577

(-0.423)

0.084

0.000

0.085

Literacy Rate

0.343

0.140

0.014

Agricultural Productivity

0.405

0.071

0.000

Per Capita Investment

0.101

0.025

0.000

Constant

-2.000

0.477

0.000

Dependent Variable →

Log Per capita Agricultural GSDP t

Independent Variables

↓

Coefficient

1+β

(β)

Standard Error

P-Values

Implied λ (rate of Convergence)

Log Per capita GSDP t-4

0.478

(-0.522)

0.116

0.000

0.104

Literacy Rate

0.216

0.164

0.187

Per Capita Investment

0.017

0.045

0.710

Constant

3.432

0.798

0.000

Dependent Variable →

Log Per capita Industrial GSDP t

Independent Variables

↓

Coefficient

1+β

(β)

Standard Error

P-Values

Implied λ (rate of Convergence)

Log Per capita GSDP t-4

0.828

(-0.172)

0.092

0.000

0.034

Literacy Rate

0.614

0.204

0.003

Per Capita Investment

0.046

0.048

0.336

Constant

-1.200

0.303

0.000

Dependent Variable →

Log Per capita Services GSDP t

Independent Variables

↓

Coefficient

1+β

(β)

Standard Error

P-Values

Implied λ (rate of Convergence)

Log Per capita GSDP t-4

0.651

(-0.349)

0.147

0.000

0.070

Literacy Rate

0.827

0.286

0.004

Per Capita Investment

0.153

0.044

0.000

Constant

-1.305

0.253

0.000

The table shows that the coefficient of lagged per capita total income is 0.577 and therefore the calculate β coefficient is negative (-0.423), which implies condition beta convergence at the rate of 8.5 percent per annum. The implied rate of convergence predicted by GMM is much higher than predicted by fixed effects method-LSDV. The Coefficient of all three explanatory variables, literacy rate, investment and agricultural productivity are positive and significant. The sign of coefficients are similar to as found in LSDV and also as suggested by economic theory. GMM estimators indicate conditional β-convergence for all the sectors however it provides different picture of the speed of convergence. The implied rate of convergence in agricultural sector is 10.4 percent per annum, which is lower than as suggested by LSDV estimator. The implied rate of convergence in industrial sector is 3.4 percent per annum and for services sector is 7 percent per annum. GMM estimates indicate higher rate of convergence for both industrial and services sector compare to LSDV estimates.

σ-Convergence

On the basis of two alternative panel data estimation methods, we can conclude that there is evidence of conditional convergence. However, this finding of conditional convergence does not imply that poor states are actually catching up with the richer ones i.e. the dispersion in per capita real income levels decreases over time. The cross sectional dispersion in real per capita income has been measured by standard deviation of log of real per capita income. There will be σ-convergence if dispersion of per capita real income decreases over the time. Dispersion of per capita real income has been measured by both weighted and un-weighted standard deviation of log per capita income. Both figures 1 and 2 (Table A.1 and A.2 in appendix) clearly show that there is an increase in dispersion of per capita real income between states for the entire period. So, there is evidence of σ- divergence. In simple words, inequality in income levels between states has been increasing over time.