Is Conflict Between Economic Growth and Equitable Distribution of Income Inevitable? Theory and Evidence from India

Abstract This paper examines the theoretical relationship between growth and distribution through income transfers. The ultimate effect of income transfers on growth rates is found to be a combination of two factors: accumulation and redistribution. It is not necessarily true that income transfer will invariably reduce growth rates. If the initial distribution of income is fairly unequal, growth induces greater equality. On the other hand, at high levels of per capita incomes, growth may raise inequality if the initial level of inequality is not very high. This brings a new dimension in the ‘inverted-U hypothesis’. Looking at the growth experiences of the Indian economy, which is known as one of those economies where economic inequalities are extremely high, we find that during 1981-2002 regional income inequality has increased but annual growth rates are negatively correlated with inequality measures.

JEL O15, O41, O53, C51 Key words: Economic growth, Income distribution, Income transfers, Inequality, India.

1

Is Conflict Between Economic Growth and Equitable Distribution of Income Inevitable? Theory and Evidence from India I. Introduction Income distribution can be solely determined by the state in a command economy along with a plan to achieve high rates of economic growth. But in a market economy, growth is generally believed to cause income inequalities at least at some stages of economic development. The literature dealing with growth and income distribution is fairly diversified, as it includes theoretical issues, country experiences and econometric modeling of processes that are related with growth, poverty and income distribution. It will be sufficient to cite a few key studies such as Kuznets (1955) and Williamson (1965) for the ‘inverted-U’ hypothesis, Kakwani (1986), Bhatty (1974) and Jain and Tendulkar (1990) for country studies and Das and Barua (1996) and Ravallian and Datt (1990) for econometric modeling. The theoretical relationship between economic growth and income distribution generally turns out to be a complex one. It will be prudent to look at the theoretical models of economic growth to find out if there is any discussion on growth and income distribution. The so-called Cambridge models of Kaldor (1956) and Pasinetti (1974) have discussed the relationship between growth and distribution in the framework of equilibrium growth. Two important parameters in these models are the workers’ and the capitalists’ propensities to save. In Kaldor’s model equilibrium is attained if the warranted rate of growth of income, which is the ratio between the average of the two savings propensities (s) and the incremental capital output ratio (v), is equal to the natural rate of growth which, in the absence of any technological change, is the exogenously

2

determined rate of population growth. In steady state, the per capita income stops growing and its value is determined in such a way that s/v is equal to the rate of growth of population. If per capita income happens to exceed its steady state value, then the actual growth rate of income (ΔY/Y) will be less than s/v, which also means that ΔY/Y is less than the rate of population growth resulting in a fall in the per capita income. In the process of income contraction, investment (v.ΔY) falls short of total savings (s.Y) which, under the condition of full employment, will lead to a fall in the price level and a rise in the real wage rate as well as the share of wages in national income. Thus, the decline in per capita income in the adjustment process is associated with an improvement in the income distribution. Conversely, if per capita income is less than its steady state value, investment will exceed total savings, the price level and the share of profit in national income will rise, as per capita income rises to approach its steady state value. Both Kaldor and Pasinetti, particularly the latter, insisted on the irrelevance of workers’ propensity to save. But the overall relationship between the growth of income and the extent of equality in the distribution of income between wage earners and capitalists is a negative one. Another important aspect of Cambridge growth models as well as the neoclassical growth models is that in the process of income redistribution the real rewards going to the various economic classes do not remain constant. Both in the Cambridge models and the neo-classical models one finds a relationship between growth and distribution only when the economy is off the steady state path. If one assumes that the economy is always on the steady state path, the rate of growth of income is determined by the exogenously given rate of growth of population with no change in income distribution as neither the factor shares nor the real returns to

3

factors change. An introduction of technological change will however change all this because it may cause a shift in the steady state path of income growth. Technological progress will also change income distribution even if it is of the Hicks-neutral type. The effect of technological progress on growth is unquestionably positive, but its effect on income distribution depends on the nature of the technological progress. For instance, a technological progress occurring in the capital-intensive industry may change income distribution in favour of capital. The subsequent work on income distribution and growth has become a part of development economics with the contribution of Kuznets (1955) and Williamson (1965) who have built up both the conceptual and empirical basis of the ‘inverted- U hypothesis’. According this hypothesis during the initial stages of economic development growth of income raises income inequality but reduces it in the later stages. Recently, there has been a revival of interest in the relationship between growth and distribution. Persson and Tabellini (1994) have used an overlapping generation model to show that a reduction of income inequality raises the growth rate of income, assuming that the median income class in the population makes all the political decisions relating to income transfers that are uniform. An income transfer is defined as uniform if income is transferred from the above average earners to below average earners in proportion to the difference between the income earned and average income. They implicitly assume that the median income exceeds the arithmetic mean, which implies that the income distribution is negatively skewed with the longer tail falling on the lower range of income. The median class will vote for a transfer if and only if the transfer does not curtail their incentive to accumulate more productive assets. Since by definition a transfer

4

will benefit at least fifty percent of the population who would accumulate more productive assets, one tautologically gets a positive relationship between growth and equity. The model simply provides an alternative explanation of the forces behind the falling segment of the inverted-U curve, where there is no conflict between growth and reduction of income inequality. However, the most restrictive assumption of this model is the constancy of the rate of return accruing to the asset holders in the process of income transfers. In this paper we have combined the idea of uniform income transfer in Persson and Tabellini (1994) with a growth process driven by accumulation and shown that growth necessarily leads to income convergence in the class of below-average earners and income divergence in the class of above- average earners. In a highly unequal income distribution the number of below-average earners will exceed that of above-average earners by a big margin. Thus higher the inequality in income distribution the greater is the possibility of economic growth causing a reduction in inequality. This adds a new dimension to the Kuznets-Williamson thesis. Whether income inequality rises at low levels of per capita income or not depends on the initial level of inequality. If the level of inequality is high at low per capita incomes, growth may reduce inequality if uniform income transfers take place. On the other hand growth may increase inequality even at high per capita incomes if the initial distribution is not very unequal. We have also found strong empirical support from Indian experience during 1980-81 to 2001- 02. Economic reforms introduced in 1991 have not reversed the rising trends in inter-regional inequality in any sector of the Indian economy and the levels of inequality, measured by Theil’s (1967) entropy index, are fairly high. But we find

5

significant negative correlation between annual growth rates and levels of inequality in practically all sectors of the Indian economy which can be characterized as a highinequality economy. The paper is organized in the following way. The following section deals with Theil’s decomposable measure of inequality and our interpretation of the measure. In section III we discuss the arithmetic of uniform income transfer and show that it leads to an improvement of income distribution at every level of disaggregation. The model of growth and income distribution is developed in section IV to discuss the apparent conflicting relationship between the two. In section V we have examined the pattern of regional inequality in India in the distribution of total income as well as of its sectoral components among 29 states of the Indian union during 1980-2002. The Central Statistical Organization (CSO) provides the database of this study. In an earlier study using CSO data for the period: 1970-92, Das and Barua (1996) found significant negative correlation between growth and regional inequality only in the case of agriculture and infrastructure sectors. The present study shows that in all sectors higher annual growth rates are associated with lower inequality. In contrast to the earlier study, it can now be clearly established that there is no conflict between growth and equity in the Indian economy. The main conclusions are summarized in the last section.

II. A Decomposable Inequality Measure There are many ways in which economic inequality can be measured and the index of inequality that one uses depends largely on the context. If we are interested inequality among regions/states of a federal economy, then it is necessary to compare one

6

region’s shares in country’s income and population with those of the other regions. If, for instance, it turns out that a single region having only five percent of the country’s population accounts for fifty percent of its income, then regional inequality must be considered rather acute. If, on the other hand, for most regions the share in population and the share in the country’s income are close, then inequality is not very large. The entropy measure we have used in this study is based on this principle. Let yk be the share of the k-th region in the country’s GNP and pk its share in the total population. Then an absolute equality in the regional distribution of income is represented by a situation in which the ratio, yk/pk is unity for all regions. Any deviation of this ratio from unity indicates inequality. It is however not possible to examine all these ratios and come to any conclusions regarding the degree of inequality prevailing in a country. Additionally, inter-temporal or inter-country comparison of inequality becomes impossible if one has to deal with all these ratios between income and population shares. What is required is a summary of the inequality scenario represented by these ratios. Measuring the ratios between income and population shares on the logarithmic scale, the following measure summarizes economic inequality among K regions of a country. K

TY = ∑ y k log( y k p k )

(1)

k =1

where, K is the total number of states/regions and ∑ yk = ∑ pk = 1. The ratio between the k-th state’s income share and population share, i.e., yk / pk , is nothing but the ratio between its per capita income and the country’s per capita income. We may therefore designate yk /pk

as the k-th state’s relative income. The welfare of

the k-th state is assumed to be a function of its relative income and let log (yk / pk )

7

be the k-th state’s welfare function. If income share yk is taken as the probability of achieving the relative income yk / pk , then TY is the expected welfare of the k-th state. Any increase in TY would indicate an increase in inter-state inequality. Theil has given a proof that TY ≥ 0 and we will not repeat it here. TY has a minimum equal to zero which is attained (subject to the condition that the shares add up to unity) when, for all states, the income share and the population share are identical. The situation of maximum inter-state inequality arises when, for a given assignment of non-zero population shares to all states, the income share of one state, say the h-th state, tends to unity, while the income shares of all other states tend to zero. In this case TY tends to − log(ph ) > 0 which is the maximum value of the measure or maximum inequality. Our choice of the Theil index, TY , as a measure of interstate income inequality is based on the consideration that the measure in (1) is decomposable and that, as we shall now demonstrate, with uniform income transfers, the overall income inequality in a country as well as the inter-state and intra-state inequality move in the same direction. The decomposition of the Theil index is discussed in the context of international income inequality only for the purpose of an illustration. The model is described as follows. Let n countries in the world economy be grouped into G groups, with the G

number of countries in the g-th group being ng. Therefore, n = ∑ n g . Let y and p denote g =1

income and population shares. Then the decomposition works out as follows:

8

∑ y log( y p ) = ∑ y log(y n

i =1

G

i

i

i

g =1

g

g

pg

)+ ∑ y G

g =1

g

Eg

(2)

ng

where,

E g = ∑ y kg log( y kg p kg ) k =1

The term on the left hand side is the Theil entropy index of inequality among all countries. This is decomposed as a summation of several entropy measures appearing on the right hand side. The first term on the right hand side is the measure of inequality between groups with yg and pg denoting the income and population shares of the g-th group in world totals. The first term, therefore, measures inter-group inequality. The second term is the weighted average of inequality levels within groups or intra-group inequality . The inequality within the g-th group is measured by the entropy Eg. In the expression for Eg , the terms yk g and pk g are the income and the population shares of the k-th country in the totals of the g-th group. The first term is referred to as a measure of ‘between inequality’, whereas the second term stands for ‘within inequality’. We shall now discuss an application of this decomposition. The model is applied to inter-country data on 94 countries for the period, 1979-93. We applied the above decomposition on the World Development Report grouping of 94 countries according to per capita income. There are four groups: (i) Low-income countries, (ii) Middle-income countries, (iii) Upper-middle income countries and (iv) High-income countries. The list of countries is given in WDR tables. The results of the decomposition are given in Table1.

9

TABLE 1: FOUR GROUP DECOMPOSITION Year

94

Low

Countries Income

Middle

Upper

High

4-Group

Income

Middle

Income

Entropy

Countries Countries Income

Countries (Between (Within

Countries (1)

(2)

Inequality) Inequality)

(3)

(4)

(5)

(6)

(7)

(8)

1979

42.53

2.54

5.49

1.91

1.41

40.85

1.68

1980

44.73

2.74

4.55

17.95

1.59

41.71

3.03

1982

57.04

1.08

3.14

2.14

11.83

46.63

10.41

1983

44.52

0.85

3.26

2.21

3.74

41.10

3.41

1984

45.67

0.92

4.19

2.30

2.39

43.30

2.37

1985

46.76

0.57

2.83

2.34

2.63

44.25

2.51

1986

57.74

0.61

4.52

2.59

9.02

49.59

8.16

1988

49.74

0.93

3.30

2.41

1.42

48.21

1.53

1990

50.00

0.81

4.17

3.35

0.86

48.83

1.17

1991

53.63

8.34

3.42

19.40

1.01

50.91

2.72

1992

51.00

2.33

4.66

6.25

2.84

47.81

3.19

1993

51.22

1.48

6.63

6.24

1.92

48.74

2.48

Source: World Development Report.

The second column of the table gives the entropy index of inequality among 94 countries. Columns (3) to (6) provide the measures of inequality within each of the four groups of countries. An weighted average of these ‘within inequalities’ is reported in column (8). The weights used are the shares of the respective groups in world GDP as in the decomposition model presented above. Column 7 gives the index of inequality between the four groups. The decomposition ensures that the sum of columns (7) and (8) is column (2). The values of all entropies are multiplied by 100 for visual convenience.

10

The table shows that ‘within inequality’ is a small fraction of total inequality. In other words, the levels of inequality within each group are very low and it simply implies the WDR grouping by per capita income creates homogeneous groups of countries. Most of the inequality is between the groups. As the table shows, ‘within inequality’ has no clear time trend, whereas ‘between inequality’ has a rising trend along with the 94-country inequality index.

III. Uniform Income Transfers As mentioned earlier, income transfer is uniform if the transfer is from the aboveaverage earners to below-average earners, the size of the transfer being proportional to the actual income earned and average income. We propose to demonstrate that if a policy of uniform income transfers is followed, then inter-personal, inter-regional and intraregional income distributions change in the same direction. The policies that result in uniform redistribution of income without discriminating between regions or between groups within a region are the direct taxes such as income, wealth and property taxes. Redistribution through indirect taxes is not uniform, because these taxes are based on expenditure and therefore discriminate against an individual or a region having a lower-than-average income but a higher-than-average expenditure. The bias in income transfer brought about by the presence of indirect taxes is however partly offset by the fact that the inequality in consumption expenditure can be expected to be uniformly lower than the inequality in income, as the proportion of income saved rises with the size of income across income-earners. While this may be a partial justification for restricting our analysis only to uniform transfers, there are other considerations that

11

suggest that the income transfers actually taking place in a society are not uniform. A deviation from uniformity occurs when public spending does not discriminate between the rich and the poor. Because of the character of non-exclusivity of public goods, such non-uniformity cannot be avoided. However, progressive direct tax rates that prevail in almost all countries can be expected to correct some of this non-uniformity. On the whole, it may not be too unrealistic to base a conceptual framework to study income distribution on income transfers that are uniform. If income transfers are uniform, such transfers unfailingly improve income distribution irrespective of the size of the transfer. Uniform income transfers, taken as a benchmark, can be used to find out how far the actual transfers have deviated from uniformity, thus indicating the failure of the policy in reducing inequality and poverty in the country. We have already discussed the meaning of uniform income transfer and the forces acting for and against uniformity in the redistribution of income. What remains to be established is the arithmetic of income transfers that cause unidirectional movements in the inter-personal, inter-state and intra-state income inequality. Such transfers can take place at various levels: interpersonal, interregional and interclass. It is a simple arithmetic exercise to show that if the transfers that take place at the interpersonal level are uniform, then the transfers are also uniform at higher degree of aggregation, i.e., at the interregional and interclass levels. Let Yijk be the income earned by the i-th person belonging to the j-th income class and residing in the kth state and ⎯Y is the per capita income. Superscripts are used to date the variables. A uniform redistribution of incomes between time t-1 and time t is defined as

12

(

Yijkt = Yijkt −1 − λ Yijkt −1 − Y t −1

)

(3)

(0 < λ < 1) Equation (3) shows that income is transferred from persons with more-thanaverage incomes to persons with less-than -average incomes. It shows that if a person’s income is higher than average in time t−1, then the person’s income decreases between time t−1 and t. It may be checked by aggregating (3) with respect to all i, j and k that Yt = Yt-1 with Y representing GNP, so that there is no income growth between the two periods. Also, an increase in λ makes income distribution more equal by raising the size of the transfer. The maximum value of λ is unity in which case everyone has the same income. Aggregating (3) with respect to i and k and writing Yj to represent the total income of the j-th income class and nj to represent the number of people in the j-th income class , we get

(

Y tj = Y tj −1 − λ Y tj −1 − n j Y t −1

)

(4)

Equation (4) , written in terms of per capita incomes, is Y tj nj

=

Y tj −1 nj

⎛ Y tj −1 ⎞ − λ⎜ − Y t −1 ⎟ ⎜ nj ⎟ ⎝ ⎠

(4a)

which shows that an increase in λ improves income distribution by income class. Now aggregating (3) with respect to i and j and writing Yk for the total income of the k-th region and nk for the number of people located in the k-th region, we get the same result for inter-state income distribution:

⎞ ⎛ Ykt −1 Ykt Ykt −1 = − λ ⎜⎜ − Y t −1 ⎟⎟ nk nk ⎠ ⎝ nk

(5)

13

Since the basis of inter-state transfer according to the uniformity rule is the difference between a state’s per capita income (Yk / nk ) and the country’s per capita income (⎯Y), the Theil measure of interstate inequality defined in (1) will capture the extent to which inter-state transfers have deviated from the uniformity rule. Finally, aggregating (3) with respect to i and writing Yjk to denote the aggregate income earned by all earners belonging to the j-th income class and residing in the kth state and njk to denote the number of persons in the j-th income class and residing in the k-th state, we get the redistribution rules within each state : Y tjk n jk

=

Y tjk−1 n jk

⎛ Y tjk−1 ⎞ − λ⎜ − Y⎟ ⎜ n jk ⎟ ⎝ ⎠

(6)

An increase in λ, therefore, is seen to improve personal, class-wise, inter-state and intrastate income distributions, provided that the income transfers are uniform. The decomposition of the Theil measure to capture the effects of uniform income transfers on income distribution at different levels in a federal economy is what remains to be seen. We have shown that the Theil measure is decomposable and total inequality can be broken into components. We have also made one application of the measure to show how international income inequality can be broken up into components so that one can see the trends in ‘within group’ and ‘between group’ inequalities. If income transfers are uniform, then the trends in total inequality and the trends in the inequality in the components will be similar. However, one cannot expect the income transfers to be always uniform at every level. Theil index of inequality therefore would measure the extent to which income transfers have been uniform.

14

IV. Growth and Distribution The framework of analysis we propose to apply in this study is an extension of the model of uniform income transfers where GNP is assumed to remain fixed. We now assume that the people earning below average do not save and those earning incomes above average save a fixed portion of their net incomes and earn interest or profit on their savings. Steady state equilibrium may be disturbed by the government’s policy of income transfers from the above-average to the below-average earners. Given the fixed rate of savings, this redistribution of income will reduce total savings and investment. But redistribution may raise the rate of return on investment, which may make up for reduction in the level of investment. Technological progress is another reason why incomes may rise at both above-average and below average levels. But the important question is whether this growth process combined with uniform income transfer will improve income distribution or not. Equation (3) can be changed in the following manner:

) [{

(

(

Yijkt = Yijkt −1 − λ Yijkt −1 − Y t −1 + μ s Yijkt −1 − λ Yijkt −1 − Y t −1

)}]

(7)

where s is the rate of savings and,

μ > 0 if Yijkt −1 ≥ Y t −1 = 0 otherwise μ is the rate of interest in equilibrium which may be positively related to the distribution parameter λ, as a policy induced reduction of savings, in the absence of technical progress, may increase its rate of return.

15

As may be recalled, there was no provision for income growth in equation (3). The simple transfer mechanism illustrated in equation (3) has been modified in equation (7) to include an income growth component It is simply assumed that a part of income minus transfer is saved by earners who earn more than the average income and earn a return on this saving in the next time period. It is assumed for the sake of simplicity that the below-average earners do not save. An understanding of the growth process represented by (7) will require a distinction to be made between the below-average earners and the above-average earners. Let yt and xt be the incomes of the below-average and above-average earners respectively in any income class and residing in any state at time t. These incomes can be defined as follows: y t = y t −1 + λ (Yt −1 − y t −1 )

(8)

x t = x t −1 − λ (x t −1 − Yt −1 ) + μ s[x t −1 − λ (x t −1 − Yt −1 )]

(9)

The difference equations in (8) and (9) can be solved subject to an assumption regarding the extent of deprivation (⎯Yt−1 − yt−1 ) of below-average earners and the extent of affluence ( xt−1 − ⎯Yt−1 ) of the above-average earners. Obviously, the former declines and the latter rises with any increase in personal income. We hypothesize a linear relationship between deprivation or affluence and personal income: Yt −1 − y t −1 = a − by t −1

(10)

(a > 0, 0 < b < 1) x t −1 − Yt −1 = c + dx t −1

(11)

(c > 0, d > 0)

16

Applying (10) and (11) in (8) and (9) and solving we get,

y t = y e + ( y0 − y e )(1 − λb )

t

(12)

y e ≡ a/b

xt =

⎡ f f ⎤ t + ⎢ x0 − g (g − 1) ⎣ (g − 1)⎥⎦

g ≡ (1 + μ s )(1 − λ d )

(13)

f ≡ λ c(1 + μ s )

We assume that for below-average earners y0 < ye. For the above-average earners, unless a drastic redistribution is taking place with λ taking high values, g is likely to be greater than one. 1 The annual growth rates of personal incomes 2 , Gbt and Gat, in the below-average and above average categories can be calculated from (12) and (13) which also brings us to the propositions relating to income convergence. G tb =

λ b( y e − y 0 ) 1−t y 0 + y e (1 − λb ) − 1

[

]

⎛ f ⎞ ⎜⎜ x 0 − ⎟(g − 1) g − 1 ⎟⎠ ⎝ a Gt = f x0 + (g − 1) g1−t − 1

(

(14)

(15)

)

Proposition I Incomes of below-average earners converge and those of above-average earners diverge over time. Thus, if the initial income distribution is unequal with the number of below-

The precise condition for g to be greater than one is: μs > λd / (1−λd) which is satisfied if either μ is high or λ is low. 2 Growth rates are defined as the difference of incomes at t and t−1 divided by income at t−1. 1

17

average earners far exceeding the number of above-average earners, growth will cause greater equality in income distribution. It is obvious from (14) that ∂ Gbt / ∂ y0 < 0. It is also easy to show from (15) that ∂Gat / ∂ x0 > 0. In other words, lower the initial level of personal income of a belowaverage earner, higher is the rate at which income grows over time causing convergence of income. The opposite is true for above-average earners. There is nothing in the growth process that prevents income distribution to turn more equal while the economy is on a high growth path. An increased intensity of transfer (higher value of λ) does not necessarily raise the growth rates of incomes of below-average earners just as it does not necessarily reduce the growth rates of incomes of above-average earners. It can be easily ascertained that the derivatives, ∂ Gbt / ∂λ and ∂Gat / ∂λ have indeterminate signs. The reason is quite simple. If the intensity of transfer is raised, a below-average earner will get higher income at t that may take the earner closer to the average earner. Thus in t+1, less income will be transferred causing a reduction in the income growth rate of the below-average earner. The process operates in the opposite direction for above-average earners whose incomes may grow at higher rates at t+1 when they pay lower tax, having made a transfer at t. The following proposition summarizes the arguments:

Proposition II The effect of the intensity of transfers on growth rates of incomes in both the aboveaverage and below-average categories is indeterminate. In other words, there is no systematic effect of the uniform income transfers on the growth process.

18

Now we look at the process of accumulation represented by two parameters, the savings rate(s) and the rate of return (μ). The expressions for g and f, figuring in Gat and defined in (13), depend on both accumulation and redistribution. There is no direct relationship between Gbt and the accumulation parameters because the income growth among below-average earners depends only on redistribution. But growth is sure to raise average income which means accumulation does affect the income growth rates of below-average earners indirectly and causes more transfers from above-average to below average earners. Whether this has an adverse effect on the process of accumulation or not is difficult to say. The effect of a change in accumulation parameters, s or μ, on Gat is indeterminate. An increase in the savings rate or rate of return will surely raise the incomes of the above-average earners but they also pay more in taxes. They do experience a net increase in incomes but it is difficult to say if their income growth rates will increase as a result of higher savings rate or rate of return on savings. Thus, theoretically it is possible to have higher growth rates of income along with greater equality in the distribution of income.

Proposition III The effect of higher savings rate or higher return on investment on the rates of growth of incomes is indeterminate among the above-average earners. But in so far as accumulation raises average income, greater transfer raises the growth rates of incomes among the below-average earners, as ∂ Gbt / ∂ye > 0. In view of the propositions stated above what clearly appears is that high growth rate may result from the process of accumulation and that the growth process will

19

certainly lead to higher equality in the distribution of incomes provided that the initial level of income distribution is characterized by high degrees of inequality. Whether this theory is valid or not is an empirical question. We now turn to the growth experiences of the Indian economy, which is known as one of those economies where economic inequalities have traditionally been extremely high.

V. Inequality and Growth: The Indian Experience There is a very large literature on various aspects of economic inequality in the Indian economy. 3 As mentioned by Kuznets (1955), structural changes in the process of growth lie at the heart of the relationship between growth and distribution. It is, therefore, only logical that we follow the growth process of the Indian economy during 1981- 2002. During the sixties and seventies the growth rates in the Indian economy have been abysmally low, often characterized by the term ‘Hindu rate of growth’. Some minor changes in policy were introduced during eighties but major economic reforms started in 1991 which have altered the structure of the Indian economy in a drastic way. The following table gives the growth rates of the major sectors of the Indian economy for three different periods. It shows higher growth rates for per capita income, NDP and services during the second period, i.e., the post-liberalization period.

3

See Das and Barua (1996) for studies on growth and inequality in the Indian economy.

20

TABLE 2: EXPONENTIAL GROWTH RATES (%) AT 1980-81 PRICES

4

1981 – 2002

1981 – 91

1992 – 2002

POPULATION

1.75

2.15

0.99

PER CAPITA

3.8

2.69

3.77

NDP

5.52

5.04

5.36

AGRICULTURE &

3.09

2.99

2.45

MANUFACTURING

6.44

6.70

5.58

INFRASTRUCTURE

3.45

5.48

1.43

SERVICES

7.73

6.42

8.37

INCOME

PRIMARY

SOURCE : Central Statistical Organization

We have applied the entropy measure of inequality defined in (1) to compute inter-state inequality levels relating to the sectors of the economy. The sectors are however the components of gross domestic products in national income accounting. This has been done for the period 1980-81 to 2001-2002 for 23 states and union territories at constant prices with base year 1980-81. For visual convenience, the Theil measure in equation (1) has been multiplied by 100 and the estimates are shown in Table 3.

4

All growth rates are statistically significant. INFRASTRUCTURE includes electricity, gas and water supply plus transport, storage, communication and construction. PRIMARY includes forestry & logging, fishing and mining and quarrying. SERVICES include trade, hotels and restaurants plus banking and

21

TABLE 3: ENTROPY ESTIMATES AT CONSTANT PRICES FOR 23 STATES & UNION TERRITORIES (1980-81 = 100) MANUFACT

1980-81 1981-82 1982-83 1983-84 1984-85 1985-86 1986-87 1987-88 1988-89 1989-90 1990-91 1991-92 1992-93 1993-94 1994-95 1995-96 1996-97 1997-98 1998-99 1999-00 2000-01 2001-02

INFRASTRUC

NSDP

URING

SERVICES

AGRI& PRIMARY

TURE

2.1019 2.0929 2.2076 2.0532 2.0994 2.3389 2.3149 2.4578 2.3574 2.7842 2.6477 2.2908 3.5198 3.6917 4.3826 4.98 5.1465 5.194 5.3777 4.4999 6.6792 2.1258

9.4901 8.2898 7.8504 7.9194 7.5418 8.4214 9.0413 7.6155 7.4516 7.7201 8.232 8.1045 9.7471 10.2239 9.9742 10.4022 11.0653 9.9883 10.3747 10.6221 9.1035 9.2101

4.23837 4.17424 4.36075 4.20923 4.00868 4.46859 4.33804 4.18438 4.12219 4.2792 4.21944 4.99189 5.26846 5.75874 4.7726 5.24353 5.12283 5.3758 5.70349 5.92515 3.9623 3.71584

1.521631 1.6805 1.9645 1.644653 1.4974 1.7863 1.7705 2.1403 1.9361 2.0575 1.944227 2.1869 2.3689 2.3837 2.0846 2.3415 2.274 2.3758 2.1164 2.4732 0.2139

9.2554 9.4826 10.0106 9.9273 10.2422 9.8685 9.5227 9.6512 9.7061 10.2589 9.7396 10.0861 10.5024 10.9026 30.3603 35.8657 39.0874 39.1119 37.4632 55.4753 90.952 88.5431

SOURCE: CSO

Regional inequality measures show increasing trend in all cases except in agriculture and primary. Estimates of exponential growth rates of inequality during 19812002 are as follows. Annual average growth rate of regional inequality is 4.67%, for NSDP, 1.27% for manufacturing, 0.93% for services and 10.87% for infrastructure. All insurance plus real estate, ownership of dwellings and business services plus public administration and

22

these growth rates of inequality are statistically significant. Growth of infrastructure in India has been lopsided and very high levels of inequality show this in the inter-state distribution of infrastructure facilities as well as a very high growth rate of inequality in this sector. Inequality growth rate in agriculture and primary is negative but not significant. Manufacturing inequality shows a negative but insignificant trend growth rate during 1981-1991. We have also looked at the post-liberalization period with a different base at 1993-94 prices. The entropy measures are shown in table 4. Increasing trend in regional inequality is apparent in all categories except in the last few years. In the new series, manufacturing has overtaken both infrastructure and primary sectors in terms of levels of inequality. Primary and agricultural sectors clubbed together show lowest levels of regional inequality. TABLE 4: ENTROPY ESTIMATES AT CONSTANT PRICES (1993-94 = 100) 5

YEAR

NSDP MFG

SER

AGRI

PRI

INFRA

AGRI& PRI

1993-94 6.8771 27.0286 13.56664

6.1679 25.07785

14.8964 5.27998

1994-95 7.0523 26.2913 13.90672

6.1532 25.59443

15.7323 5.43146

1995-96 7.9923 27.5681 15.21837

6.9323 24.82762

15.8942 6.14492

1996-97 7.8178 27.9117 14.97477

6.9568 23.51134

16.1627 6.02813

1997-98 8.328

7.258

24.80574

17.4228 6.49976

6.6929 21.06397

16.8351 5.84232

1999-00 8.7233 27.4686 16.85401

7.2635 21.8554

15.3921 6.1779

2000-01 9.1994 22.075

2.9785 17.81359

12.1728 1.65694

26.4641 15.70642

1998-99 8.7168 29.153

16.32404

12.94616

2001-02 8.8996 19.6086 8.92856

7.12173

8.7881

2.55842

other services. All income categories are the same as defined by CSO. AGRI = AGRICULTURE, INFRA = INFRASTRUCTURE MFG = MANUFACTURING, PRI = PRIMARY, SER = SERVICES, NSDP = NET STATE DOMESTIC PRODUCT. 5

23

Relative contribution of sectors in the growth of regional inequality in NSDP can be discussed by applying Theil-type decomposition presented in equation (2). Instead of applying an arithmetic formula we have opted for regression analysis in which entropy measures of inequality in NSDP has been regressed on inequality measures for sectors. Table 5 shows that manufacturing and infrastructure are the sectors that significantly raise inequality in inter-state income distribution. Inequality in the inter-state distribution of services and agriculture plus primary reduces inter-state income distribution but the coefficients are not significant. As noted earlier, inequality levels in these sectors are low relative to manufacturing and infrastructure during 1980-81 to 2001-02. TABLE 5: REGRESSION RESULT: ENTROPY OF NSDP AT 1980-81 PRICES ON ENTROPIES OF SERVICES, MANUFACTURING, AGRICULTURE & PRIMARY ACTIVITIES, INFRASTRUCTURE AND TIME, 1980 – 81 TO 2001 - 02

Net state domestic product Manufacturing

Coefficient

t

P>|t|

.4143899

Standard Error .1443941

2.87

0.012

95% confidence interval .106621 .722159

Services

-.3648315

.364086

-1.00

0.332

-1.140862

.411199

Agriculture &

-.1137079

.4039614

-0.28

0.782

-.9747311

.747315

Infrastructure

.02187

.011721

1.87

0.082

-.0031128

.046853

Time

.0012249

.0003615

3.39

0.004

.0004543

.001995

Constant

-.0023999

.008694

-0.28

0.786

-.0209306

.016131

Primary

No. of observations=21. F(5,15)= 52.03 R-squared= 0.9455 Adjusted R-Squared= 0.9273 Root MSE= 0.00384

24

The relationship between economic growth and income inequality has traditionally been explored in terms of the presence or the absence of a U-type relationship between the level of inequality and per capita income. The existence of a U-type relationship in the Indian context has been explored by Mathur 6 and his study has by and large supports the Kuznets-Williamson thesis. Though such a relationship has a relevance in development policy in so far as it prescribes the social safety nets to be used at the levels of high per capita income to check the growth of income inequality in an economy that is becoming increasingly prosperous, it does not really deal with the question of growth versus distribution. We have calculated the correlation coefficients between the level of inequality, measured by an entropy, and the annual (year-to-year) growth rates. These are shown in Table 6. The correlation between growth and inequality has been found for the entire period (1981-2002) as well as for the two sub-periods, 1981-91 and 1992-2002. The entropy measures the level of regional inequality and therefore, as the table shows, there does not seem to be any conflict between growth and inequality. This is an alternative approach to the U-hypothesis and the results in the two sub periods do not indicate any turning point in the relationship between growth and inequality at the aggregate as well as at the sectoral levels. In agriculture growth and inequality are significantly and negatively correlated. In manufacturing, Das and Barua (1996) found negative but insignificant coefficients for an earlier period, but the present study shows A. Mathur, “Regional Development and Income Disparities in India : A Sectoral Analysis”, Economic Development and Cultural change, Vol. 31, No.3, 1983. Also see A. Mathur, “Why Growth Rates differ Within India : An Alternative Approach “ Journal of Development Studies, Vol. 23, No.2, 1987. Das and Barua(1996), looking at a different time period, have identified a different pattern of relationship between growth and regional disparity in the Indian economy. According to them the inverted-U may be followed by rising trend of inequality with the growth of per capita income.

6

25

negative and significant correlation. In fact, the negative correlation between growth and inequality in manufacturing is stronger in the post-liberalization period. The negative sign of the correlation coefficients is seen in all other cases and these are also statistically significant, while they were negative and insignificant in the earlier study. TABLE 6: ECONOMIC GROWTH AND REGIONAL INEQUALITY: CORRELATION COEFFICIENT BETWEEN ENTROPY MEASURES OF INEQUALITY AND THE CORRESPONDING ANNUAL GROWTH RATES

CATEGORY

NSDP

MANUFACTURING

SERVICES

AGRI & PRIMARY

INFRASTRUCTURE

1981-2002

1981-91

1992-2002

-0.36699

-0.4712

-0.67825

(-4.24103)

(-6.05683)

(-12.5609)

-0.34162

-0.74854

-0.75405

(-3.8675)

(-17.0245)

(-17.4788)

-0.35886

-0.54927

-0.33322

(-4.119)

(-7.8658)

(-3.74835)

-0.30321

-0.55158

-0.20561

(-3.17209)

(-7.53133)

(-2.03955)

0.029659

-0.70014

-0.29142

(0.296855)

(-13.7338)

(-3.18464)

Note: Figures in the parenthesis are t-values.

Source: CSO

Tables 3 and 4 show generally high levels of inequality in the Indian economy, particularly in the manufacturing and infrastructure sectors. Table 6 shows that the negative correlation between growth and inequality in these two sectors is the strongest.

26

It appears that there is less of a conflict between growth and inequality in the Indian economy if the level of inequality is sufficiently high. Table 5 shows weak correlation (though still negative and significant) between growth and inequality in agriculture where the level of regional inequality is relatively low. The pattern in the service sector is similar to the manufacturing sector. In a study for an earlier period Das and Barua (1996) saw no conflict between growth and distribution in the Indian economy but also detected some indication that higher growth rates might raise regional inequality. The growth performance of the Indian economy has been much better during the nineties. But as things stand today there is no evidence of a conflict between growth and social justice and high inequality seems to make achievement of social justice easier.

VI. Summary and Conclusions We have examined the theoretical relationship between growth and distribution and for this purpose we have first looked at the literature on economic growth in the Cambridge as well as the neo-classical tradition. Growth models emphasise on equilibrium growth or steady state and therefore do not directly deal with the effect of growth on income distribution. However, once the economy is off the steady state path and adjustments take place to get the economy back on the path, it clearly appears that growth and equitable distribution have a conflicting relationship. This perhaps is the basis of the hypothesis that in a market economy economic growth would change the income distribution adversely. In running the welfare state the policy makers do not entirely depend on the markets. The extent of state intervention in the market forces varies from country to

27

country and depends crucially on the political systems prevalent in the countries. Many observers would subscribe to the view that welfare state in Europe is stronger than in North America. In the developing world, state intervention has traditionally been very strong until recently with economic reforms and liberalisation of economic policies demanding less of government and more of free enterprise in running the economy. But the government’s role in the developing world has been quite different from that in the industrial countries. The governments in the developing countries do much more than just redistribute income, as they have been engaged directly in the production processes and performing functions that are normally in the domain of the private sector. At one time it was expected that direct involvement of the government in production would bring about an equitable distribution of income. There is ample evidence that this has not happened. In the post-liberalisation phase, the governments of the developing countries have still to perform their redistributive functions. The crucial question that arises is whether the redistributive functions of the government will come in conflict with the growth process that will now be almost entirely conducted by the market forces. We have taken this question as the main concern of the paper. Therefore, we have retained the growth mechanism of the traditional growth models in which the market absorbs the savings of the accumulators to raise the capital stock leading to growth of aggregate output in the economy. But we have introduced a concept of ‘uniform income transfers’ by which income is transferred from an above-average earner to a below-average earner with the size of transfer being proportional to the difference between the actual and average incomes. We have shown that a particular measure of inequality, namely Theil’s entropy measure, is able to capture deviations from uniform income transfers. We have discussed

28

the tax structure that can achieve such uniform income transfers. Theil’s measure happens to be additively decomposable. This means that the measure can break up total inequality into inequality levels among groups of people. In a federal economy like India or the United States, such a measure can present a disaggregate picture in which one can see which sectors or regions are reporting higher inequality even though the total inequality in the country may be under control. In one application of this measure we have decomposed international income inequality into inequality among groups of countries. The result of this exercise shows that there is much more inter-group income inequality than intra-group inequality due to the obvious fact that within each country group the governments follow income redistribution policies whereas in the absence of a world government no such policy can be implemented between country groups. We have then demonstrated that if income transfers are kept at the uniform level, then inequality will fall at every level of disaggregation, i.e., there will be an improvement in personal, inter-group and intra-group income distribution. In the last part of our theoretical analysis we have tried to place uniform income transfers in the context of growth. The ultimate effect of income transfers on growth rates is found to be a combination of two factors : accumulation and redistribution. It is not necessarily true that income transfer will invariably reduce growth rates. In fact, if the initial levels of inequality are high, high growth rates may be associated with reduction in inequality. This hypothesis is different from the well-known ‘inverted- U’ hypothesis and it deserves to be tested with the use of inter-country data on growth and inequality. Finally, we have looked at the experience of the Indian economy during 19802002 in respect of inter-regional inequality in GDP as well as in the various sectors or

29

components of GDP. We have not found any evidence to show India was confronted with a trade-off between growth and inter-state inequality during this period. There are, however, indications that high levels of economic inequality may have resulted in favourable movements in income distribution along with the growth of the Indian economy.

30

References I. Z. Bhatty, “ Inequality and Poverty in Rural India” in T. N. Srinivasan and P. K. Bardhan (eds.), Poverty and Income Distribution in India, Statistical Publishing Society, Calcutta, 1974. S. K. Das and A. Barua, “Regional Inequalities, Economic Growth and Liberalization : A Study of the Indian Economy”, Journal of Development Studies, 1996,Vol.32, No.1. L. R. Jain and S.D. Tendulkar, “Role of Growth and Distribution in the Observed Change of Headcount Ratio-Measure of Poverty”, Technical Report No.9004, 1990, Indian Statistical Institute. N. Kakwani, Analyzing Redistribution Policies : A Study Using Australian Data, Cambridge : Cambridge University Press, 1986. N. Kaldor, “Alternative Theories of Distribution”, Review of Economic Studies, 1956, Vol.23. S. Kuznets, “Economic Growth and Income Inequality”, American Economic Review, 1955,vol. XLV. L.L. Pasinetti, Growth and Income Distribution : Essays in Economic Theory, Cambridge University Press, 1974. T. Persson and G. Tabellini, “ Is Inequality Harmful for Growth?”, American Economic Review, 1994 Vol.84, No.3. M. Ravallion and G. Datt, “Growth and Distribution Components of Changes in Poverty Measures”, LSMS Working Paper No.83, 1990 World Bank. H. Theil, Economics and Information Theory, Amsterdam, 1967. J.G. Williamson, “Regional Inequality and the Process of National Development : A Description of the Patterns”, Economic Development and Cultural Change, 1965,Vol.13, No.4, Part II.

31