You are on page 1of 7

inter and intra-state disparities

Intra-State Disparities in Gujarat, Haryana,


Kerala, Orissa and Punjab

Amaresh Dubey

There is a large body of literature that highlights growing 1 Introduction

T
inter-regional disparities in India. However, intra-state his paper examines the intra-state disparities in five states
in India, Gujarat, Haryana, Kerala, Orissa and Punjab. I
disparities have not elicited similar attention, primarily
have chosen three indicators, consumption, inequality
due to the non-availability of comparable data at the and the incidence of poverty, to examine this issue. These indi-
sub-nss region level. This paper uses nss consumption cators taken together reflect overall well-being of the population
expenditure survey data for two recent quinquennial as they are the outcome of the interplay of a large set of
e­conomic   and policy variables. The states chosen for the analy-
rounds to calculate comparable welfare indicators and
sis of intra-state disparities had a relatively homogeneous initial
indices of inequality at the district level in five states. The level of poverty in 1973-74, the coefficient of variation (COV) of
data show that intra-state disparities are also increasing. the headcount ratio (HCR) being about 20% in 15 major states
From the policy point of view, intra-state disparities (reported in A­ppendix Table 1, p 230). In the last 30 years, the
COV of the HCR has reached close to 45%. The available data also
need the same kind of attention that rising inter-state
s­uggest that the rise in inter-state disparities has been the
inequalities have attracted in recent times. f­astest   since 1993-94, coinciding with the surge in growth of per
capita income.
The steep rise in inter-state inequality has been recognised in
policy circles and the approach paper to the Eleventh Five-Year
Plan adopted in 2006 forcefully articulates the urgency for
“bridging the gaps”.
The strategy of inclusive growth proposed in this paper can command
broad-based support only if growth is seen to demonstrably bridge
d­ivides and avoid exclusion or marginalisation of large segments of
our population. These divides manifest themselves in various forms:
between the haves and the have-nots; between rural and urban areas;
between the employed and the under-unemployed; between different
states, districts and communities; and finally between genders.1

The discussion about disparities or inequality is not new. The


well-known Kuznets curve (Kuznets 1955) relating economic
d­evelopment with regional disparities predicted that during the
initial stages of development, inter-regional disparities tend to
increase, which broadly conforms to the experience of the devel-
oping countries, including India (Figure 1, p 225). The Kuznets
curve has an inverted U-shape which implies that after an initial
rise in inequalities, they narrow down. However, several recent
studies in development economics point out that developing
countries are characterised by various kinds of heterogeneity
This is the revised version of the paper which was presented at the
(socio-­cultural and religious) that could defy the predictions im-
w­orkshop organised by the Asian Development Research Institute plied by the Kuznets curve. It is increasingly being realised that
in Patna. Helpful comments and suggestions from Atul Sarma, two the long-term economic standing of the households are also influ-
discussants and other participants in the workshop are gratefully enced by historical forces – multicultural, multiethnic and multi-
acknowledged. Usual disclaimers apply.
religion – that have shaped income levels and the possessions of
Amaresh Dubey (amareshdubey@mail.jnu.ac.in) is at the Centre for the households. Several studies underline the sharp differences in
Study of Regional Development, Jawaharlal Nehru University, standards of living across geographical domains, across castes
New Delhi.
and r­eligious communities within and across states.
224 june 27, 2009  vol xliv nos 26 & 27   EPW   Economic & Political Weekly
inter and intra-state disparities

While a large body of literature is devoted to inter-regional dis- Figure 1: Kuznets Curve
parities, intra-regional disparities have received scant attention.
One of the main reasons appears to be the unavailability of com-
parable information on variables of interest. As a result there are
very few studies which deal with intra-regional disparities.
Among the earliest studies is that by Jain et al (1988). They exam-

Inequality
ined disparities across regions identified by the National Sample
Survey Organisation (NSSO).2 They calculated six interrelated
characteristics of poverty in about 56 NSS regions and reported
the COV of poverty incidence to be around 47% in 1972-73, which
was nearly two and half times the COV of the HCR across major
states in 1973-74 (Appendix Table 1).
Among the more recent studies, Dubey and Gangopadhyay Income per capita

(1998) also look at intra-state disparities in the incidence of pov- i­ntervention, this would also facilitate the study of intra-state
erty at the NSS region level. They find that there are several states disparities. As a result, there have been several attempts to pro-
in India where the incidence of poverty (calculated from NSS con- vide estimates of well-being indicators, including poverty inci-
sumption expenditure data) across regions within a state is very dence, at the district level.5
high. For example, they reported that in 1993-94, among the For estimating indicators of well-being, some researchers use
seven NSS regions in Madhya Pradesh, poverty incidence varied the small-area estimation method. In these studies, the sub-state
from one of the lowest in the country in the western region to one or NSS region level poverty ratios are used to estimate district-
of the highest in the eastern region (now Chhattisgarh). level poverty in a given NSS region (Parikh and Radhakrishna
The analysis of disparities across districts (clearly demarcated, 2002). A basic flaw in the small-area estimation technique is that
smallest administrative units) in India is a more recent phenomenon. the estimation is carried out using the “average values” of indica-
Using data on a set of variables ranging from the incidence of pov- tors from a large area. The estimates of characteristics for the
erty to immunisation rates of children at the district level, Borooah smaller units thus have an obvious bias.
and Dubey (2007) identified the 100 most backward districts, which In the last few years there have been attempts to overcome this
fall into several states in India.3 But none of the studies have system- problem through directly estimating district-level poverty by in-
atically analysed intra-state disparities even in larger states. creasing the sample size at the district level. Murgai et al (2003)
As pointed out, in this paper I have chosen three indicators at the have tried to estimate district-level poverty incidence by combining
district level to focus on intra-state disparities in five states. The central and state sample consumption expenditure data c­ollected
rest of the paper is organised in the following fashion. In the next by the NSSO and state directorate of economics and statistics.6 Sim-
section, issues related to the data used in this paper have been dis- ilarly, Bhandari and Dubey (2003) produced district-level poverty
cussed. This is followed by a discussion of inter-state disparities estimates by combining employment and un­employment data with
(across major states only) in Section 3. In Section 4, the intra-state consumption expenditure data of the 55th   round.
disparities among the five states chosen in this study are discussed. While the problem of sample size is taken care of to some ex-
The findings of the paper have been summarised in Section 5. tent by pooling the two data sets (state and central samples) by
Murgai et al (2003), the main criticism of this exercise is that the
2 Data and Methodological Issues data collection agencies are different (state and central). Conse-
Despite well-grounded criticism of monetary indicators of well- quently, there could be problem of data quality bringing in what
being in recent times, levels of income or expenditure at the Sastry (2003) refers as “agency bias”.7 The strategy used by
household level and the proportion of population below a pre- Bhandari and Dubey (2003) does provide a reasonable sample
specified poverty norm continue to be the main indicators (of size for a fairly large number of districts, but the major issue with
well-being). In this paper I have mainly used the consumption this exercise has been non-comparability of household expendi-
expenditure survey (CES) data collected by the NSSO during two ture distribution in the two surveys. Besides these, the NSSO sam-
quinquennial rounds of surveys, the 50th (1993-94) and the 61st pling design in earlier years has been such that the urban sample
(2004-05). The NSSO CES data have been in the public domain for was decided at the NSS region level. As a result, some districts
some time now. Therefore, it is assumed that its scope, sampling went unrepresented in urban areas, thereby introducing bias in
design, and limitations are well known.4 the district-level poverty estimate.
One of the data-related issues that has relevance to this paper Suggestions from Sastry (2003) and others (Indira et al 2002)
is the calculation of some of the characteristics at the district have been to use only the central sample for estimation. Sastry
level from the NSSO data. Since, the district is emerging as the (2003) has examined the relative standard error (RSE) of average
b­asic unit for implementing and monitoring progress of these monthly per capita expenditure (MPCE) at the district level for
d­evelopment-related programmes, it is strongly felt that there 1999-2000. He reports that out of nearly 490 districts in the
should be some basic district-level indicators of well-being. 1999-2000 NSS sampling frame, the RSE is within the 0-5% range for
B­esides being able to monitor the progress of well-being at the the rural population of 451 districts. However, bias remains in the
smallest administrative unit, the district, for informed policy estimates because of the region-level stratification of u­rban   areas.
Economic & Political Weekly  EPW   June 27, 2009  vol xliv nos 26 & 27 225
inter and intra-state disparities

Clearly, for calculating district-level poverty, the major c­oncern Gini coefficient has been relatively stable between 1983 and
of researchers is whether the expenditure distribution of the 1993-94. The COV of the HCR for major states for 1973-74, 1977-78,
p­opulation provided by the NSS surveys will be adequate. Put 1983 and 1987-88 are 19.8%, 27.7%, 33.4% and 31.9%, respec-
d­ifferently, the central issue is whether we have enough sampled tively. Inter-state disparities were not just low but i­ncreased at a
households at the district level to obtain reliable estimates. A moderate rate between 1973-74 and 1983 and stagnated d­uring
r­elated problem is the constant reorganisation of districts over 1983 and 1993-94.
time, especially during the 1980s and 1990s.8 With the splitting In Table 2, the incidence and change of poverty (HCR) for two
up of districts, the sample size further shrinks and getting r­eliable years, 1993-94 and 2004-05, along with the annualised growth of
estimates at the district level becomes that much more difficult. gross state domestic product (GSDP) during the same period is re-
The problem of urban bias has been addressed by the NSSO in ported. There are several points to be noted. First, there is a large
its last round of CES survey (2004-05) as rural and urban areas in variation in the HCR across states in both years. Second, the COV of
the HCR has increased by about 12 percentage points b­etween
Table 1: Number of NSS Regions and Districts
States Number of NSS Regions Total Number of Districts Number of Districts
1993-94 and 2004-05. Observe that there is no change in the COV
Available for Calculation during the 1980s. Third, the inter-state variation in a­nnualised GSDP
Gujarat 5 24 13 growth is over 21% but the COV of decline in poverty incidence is
Haryana 2 19 5 close to 55%. Fourth, the large difference in the COV of reduction in
Kerala 2 14 10 poverty and annualised GSDP seems to indicate that GSDP growth is
Orissa 3 30 10
very poorly associated with poverty reduction, which is confirmed
Punjab 2 17 9
Source: Tabulated from unit record CES data for 2004-05 by the author.
by the extremely low correlation between the two (-0.06).
The substantial increase in the COV of the HCR requires a much
each of the districts in the sampling frame have been considered. closer examination which is beyond the scope of this paper. But it
In this paper, I have tried to circumvent the second problem (con- could be mentioned in passing that inequalities have started play-
stant reorganisation of the districts) by merging the contiguous ing a role now and the apparent disconnect between GSDP growth
districts within an NSS region in a state. In most cases, the newly and poverty reduction could be explained by looking at the level
created districts have been merged with the districts from which and change in the Gini coefficient.
they were carved out.
The total number of districts and the number of districts for 4 Intra-State Disparities
which calculation of relevant characteristics was carried out is re- The brief analysis of inter-state disparities in the last section sug-
ported in Table 1 (districts in the sampling frame of the NSSO in the gests that after being stagnant for over a decade during the 1980s,
2004-05 survey are listed in Appendix Table 2, p 230). It is to be inter-regional disparities in poverty incidence have risen much
mentioned here that the number of sampled households in each faster than observed during the 1970s. In the rest of this paper, I
one of the “pseudo” districts in the last column is sufficient for cal- examine intra-state disparities among the states in the sample,
culation of poverty and other characteristics as the estimated stand- Gujarat, Haryana, Kerala, Orissa and Punjab.
ard e­rrors are within acceptable limits in almost all the cases. The As relatively large geographical units, these states have differ-
table suggests that though the incidence of poverty and other re- ent levels of development that have been conditioned by their
lated characteristics is calculated for fewer districts than the number
Table 2: Poverty Incidence and Change in Major States
of districts within the states, the number of “pseudo” districts is State HCR HCR Percentage Point Reduction Annualised Trend
large enough to provide useful insights into intra-state disparities. 1993-94 2004-05 in Poverty b/w Growth in GSDP
1993-94 and 2004-51 (1993-94 prices) between
As indicated earlier, the characteristics used for investigation of 1993-94 and 2004-52
the disparities within each of the states are the incidence of poverty, Andhra Pradesh 21.8 14.8 7.0 5.9
the mean consumption and the widely used measure of inequality, Assam 41.4 20.4 21.0 3.3
the Gini coefficient. For calculating poverty incidence, the state and Bihar* 54.9 42.0 12.9 4.7
Gujarat 24.2 17.0 7.2 6.2
sector-wise poverty lines published by the Planning Commission have
Haryana 25.0 13.6 11.5 6.2
been used. In comparing consumption expenditure, the 2004-05 data
Karnataka 32.9 24.3 8.6 7.0
has been deflated using implicit price d­eflators in each state and Kerala 25.0 14.8 10.2 5.7
s­ector, derived from the state and s­ector poverty lines (PLs). Madhya Pradesh* 42.6 38.9 3.7 4.0
Maharashtra 37.0 30.6 6.4 5.3
3 Trends in Inter-State Disparities Orissa 48.7 46.6 2.1 4.5
Interstate disparities calculated from per capita state income Punjab 11.3 8.1 3.1 4.4
Rajasthan 27.5 21.4 6.0 5.7
have been investigated earlier; see for example, Mathur (1983),
Tamil Nadu 35.5 22.8 12.7 5.0
which pointed out the existence of the disparities but without any
Uttar Pradesh* 40.8 33.0 7.8 4.1
definite trend. One of the reasons for such findings could have West Bengal 37.0 24.7 12.3 7.1
been the overall inertia of the Indian economy during the 1960s and Mean 33.70 24.87 8.83 5.25
1970s. This is in part corroborated by the COV of poverty incidence SD 11.38 11.25 4.83 1.12
across 15 major states reported in Appendix Table 1. In addition, CoV 33.77 45.23 54.68 21.26
Debroy and Bhandari (2007) report that the consumption-based Correlation -0.06

226 june 27, 2009  vol xliv nos 26 & 27  EPW   Economic & Political Weekly
inter and intra-state disparities
Figure 2: Gini Coefficient and Per Capita GSDP for 2004-05 (Large States) The mean PCTE ranges from Rs 423 in Rewari to Rs 632 in
F­aridabad. But the incidence of poverty ranges from 5.1% in the
Gini (Household Consumption Expenditure), 2004-05

.4
●Ker ●Mah
●TN Ambala group to over 18% in the Rewari group. Despite having
●Kar the highest mean PCTE, the poverty ratio in Faridabad is higher
●MP
.35 ●WB ●Pu●nHar than the state average. This could be due to the high level of
●AP
●Guj i­nequality indicated by the Gini coefficient (0.431).
●UP●Ori ●HP
The COV of the HCR is more than two times the COV of the real
.3 ● Raj PCTE. The COV of the HCR increased by about 10 percentage points
over 1993-94, indicating that there is an increase in disparities in
the incidence of poverty. In 1993-94, poverty incidence was the
● Bih ●JK
.25 highest in Gurgaon and Faridabad at close to 34%, which de-
●Asm
| | | clined by 18.7 percentage points. The largest decline in poverty is
20000 10000 30000 for the group of districts clubbed along with Ambala. Though
Per capita GSDP, 2004-05
Source: Debroy and Bhandari (2007). Faridabad and Gurgaon have the highest PCTE, inequality in
these districts is also the highest. The level of vertical inequality
agro-climatic conditions and locations. While Haryana and as captured by the Gini coefficient of PCTE puts these two dis-
P­unjab in the north are leaders in improving the well-being of the tricts (Gurgaon and Faridabad) of Haryana among the districts
population with one of the highest levels of income among the that have most unequal distribution.
major (large) states in India, Kerala leads the list in achievements
other than income indicators of development, namely, health and 4.1.2 Inter-District Inequalities in Punjab
education. Between 1973-74 and 2004-05, Kerala improved its In Table 4 (p 228), the three indicators, the mean PCTE, the Gini
rank in poverty incidence from 11 to 4 (Appendix Table 1). Gujarat c­oefficient and the HCR are reported for nine pseudo districts in
is considered the most investment friendly state in recent times Punjab. Unlike Haryana where the sample sizes at the district
and has attracted large investments. Its relative ranking in the level were relatively smaller and fewer pseudo districts could be
HCR has stagnated at around 4 in the same period (Appendix formed combining contiguous districts, there are five districts in
T­able 1). Orissa is at the bottom in most of the development Figure 3: Change in Gini Coefficient and Per Capita GSDP (Large States)
i­ndicators and considered the poorest among the 17 major states, (between 1993-94 and 2004-05)
ranked 15th throughout except in 1993-94, when its rank was 14. .o8
●Ker
Improvement in the well-being of households is quite varied. For ● Pun
.o6
example, with similar levels of poverty incidence in 1993-94 and ●Guj
Kar●
identical GSDP growth, Gujarat and Haryana show quite different ●Har
● MP ● Ori
Change in Gini

.o4
results as far as decline in poverty is concerned. Similarly, Orissa ● TN ● AP
and Punjab had similar GSDP growth and reduction in p­overty dur- ● Asm ● UP
● Raj
.o2
ing 1993-94 and 2004-05 but the poverty levels in these states are ● Mah

the highest and lowest, respectively, among the major states. ● Bih
●HP
0
An analysis of intra-regional disparities could, therefore, pro-
vide some insight into which areas in each of these states are
–.02 ●JK
l­agging. In this section, intra-regional disparities are reported | | | | |

3 4 5 6 7
separately for each one of the five states in our sample. Change in per capita GDP
Source: Debroy and Bhandari (2007).

4.1  Haryana and Punjab Punjab where the sample size has been found to be sufficient for
These two states in northern India are considered the most deve­ calculation of these characteristics.
loped. They are typically a textbook case of economic development. From Table 4, it becomes apparent that the districts in Haryana
Initial public investment in agriculture (irrigation and i­nfrastructure) and Punjab are not too different as far as disparities in mean PCTE and
improved yields and incomes, which created a booming consumer inequality measures are concerned. But it is in the case of HCR that
market. Since 1973-74, both the states have had lowest levels of in- Punjab turns out to be the most heterogeneous. The HCR ranges from
cidence of poverty, except in 1993-94 when Haryana slipped to the 3.1% in Kapurthala to 19.7% in Firozpur. The only other district in the
fourth rank only to regain its position in 2004-05. state that has an HCR in double digits is B­hatinda (18.3%). Thus, the
south-western region has the highest incidence of poverty within the
4.1.1 Inter-District Inequalities in Haryana state. In this region (Bhatinda and Firozpur), poverty incidence actu-
Table 3 (p 228) reports three indicators, the real (at 1993-94 ally increased significantly in 2004-05 compared to 1993-94.
prices) per capita total expenditure (PCTE), the Gini coefficient The other feature of intra-regional disparity in Punjab is the ex-
calculated from real PCTE and the incidence of poverty in five istence of a very high level of inequality. Since the Gini coefficient
r­egions within Haryana. The names of the districts appearing in has been calculated from CES data, the Gini at 0.469 could be one
column 1, along with those in column 2, form the five regions for of the highest in the country. It is to be noted that inter-country
which the three indicators have been calculated. comparison of income based on Gini coefficients suggest that
Economic & Political Weekly  EPW   June 27, 2009  vol xliv nos 26 & 27 227
inter and intra-state disparities
Table 3: Mean PCTE, Gini and HCR among Districts in Haryana in 2004-05 observed. Compared to 1993-94, the COV of the HCR increased by
District Code Merged District(s) Sample Mean Gini HCR
and Name Code and Name Size PCTE Coefficient
about 25 percentage points in 2004-05. Three districts had an HCR
02 Ambala 01 Panchkula, 03 Yamunanagar, substantially higher than the all-India average while four districts
04 Kurukshetra, 05 Kaithal 600 572 0.319 5.4 had an HCR less than 10%. At 3.5%, Bhavnagar had the lowest HCR
06 Karnal 07 Panipat, 08 Sonipat,
14 Rohtak, 15 Jhajjar 680 472 0.325 15.7
among all Indian districts (major states only). The disparities ob-
19 Faridabad 18 Gurgaon 480 632 0.431 15.2 served in the mean PCTE, the Gini and the HCR are among the larg-
17 Rewari 13 Bhiwani, 16 Mahendragarh 400 423 0.315 18.1 est. Thus, among the relatively better-off states, Gujarat has a fairly
09 Jind 10 Fatehabad, 11 Sirsa, 12 Hisar 560 453 0.313 14.7 high level of r­egional disparities. This could be due to more disper-
Haryana 2,720 514 0.355 13.6
sion in the investment in the state.
Mean 510 0.341 13.8
SD 88.0 0.051 4.9
COV 17.2 14.9 35.2
4.3  Kerala
Source: As in Table 1. It has been pointed out that Kerala has the best human develop-
Table 4: Mean PCTE, Gini and HCR among Districts in Punjab ment indicators among the major states in the country. At the ag-
District Code Merged District(s) Sample Mean Gini
gregate level, there has been constant improvement in the rank-
and Name Code and Name Size PCTE Coefficient HCR
01 Gurdaspur 360 620 0.359 3.8 ing of Kerala in the case of poverty incidence. It was ranked 11
02 Amritsar 510 440 0.232 6.8 (out of 15 major states) in 1973-74. By 2004-05, Kerala’s efforts in
11 Firozpur 11 Moga, 12 Muktsar 583 407 0.298 19.7 reducing poverty had been impressive and it stood in the fourth
09 Ludhiana 559 767 0.469 6.3 position. There were a total of 14 districts in Kerala in the NSSO
04 Jalandhar 318 585 0.273 3.2
03 Kapurthala 05 Hoshiarpur, 07 Rupnagar,
sampling frame in 2004-05. With limited reorganisation of dis-
06 Nuwanshahr 720 573 0.311 3.1 tricts over the last few years, it was possible to calculate the indi-
17 Patiala 08 Fatehgar Sahib 480 658 0.365 4.8 cators used in this paper for 10 districts. There were seven dis-
16 Sangrur 320 534 0.286 5.2 tricts where merging of neighbouring districts was not required.
14 Bathinda 13 Faridkot, 15 Mansa 438 453 0.309 18.3
The incidence of poverty ranges from 4.4% in Thiruvanan-
Punjab 4,288 558 0.347 8.1
Mean 560 0.322 7.9 thapuram to about 30% in Kasaragod (Table 6, p 229). Of 10 dis-
SD 115.5 0.07 6.4 tricts, only two have an HCR higher than the national average
COV 20.6 21.3 81.4 (27.5%). The mean PCTE ranges from Rs 411 in Kasaragod to
Source: As in Table 1.
Rs 856 in Thiruvananthapuram (Table 6). While the COV of PCTE
c­onsumption inequality in Ludhiana is even higher than that ob- is modest at over 22%, disparities are high in the case of the HCR
served in several developing countries, including the US (0.408).9 As (COV over 64%), which has increased substantially from about
far as intra-state disparity in poverty incidence is concerned, it has 23% in 1993-94. Of the 10 districts, poverty has declined in seven,
increased by about 33 percentage points, as i­ndicated by the COV. with the largest reduction observed in Palakkad.
From Table 2, it can be noted that even with modest growth in
4.2  Gujarat the GSDP (annualised growth of 5.7%) Kerala appears to have
Gujarat in western India is rated as the most investor-friendly state done well in improving income and substantially reducing pov-
in India. In NSS surveys, it is divided into five regions. This is the erty. But recent growth in income has come at a cost – it has in-
only state in India where a number of districts fall into two or more creased intra-regional disparities. Another interesting feature of
NSS regions based on cropping, agro-climatic conditions and de- inequality in Kerala is that the level of consumption inequality is
mographic patterns. There are large disparities in the state’s agro- highest among the five states considered in this paper.
climatic regions. Consequently, the creation of pseudo districts has
Table 5: Mean PCTE, Gini and HCR among Districts in Gujarat
been particularly cumbersome. Following the practice in other District Code Merged District(s) Sample Mean Gini
states, I have stuck to contiguity of the districts in combining these and Name Code and Name Size PCTE Coefficient HCR

even if a district falls into two or more NSS regions. From a total of 10 Jamnagar 09 Rajkot 440 477 0.237 8.1
08 Surendranagar 01 Kachch 230 322 0.249 23.2
24 districts in the NSSO sampling frame in 2004-05, we could form
12 Junagadh 13 Amreli, 11 Porbandar 400 425 0.239 6.2
13 pseudo districts, the largest among the five states covered in 14 Bhavnagar 231 417 0.229 3.5
this paper. There are five d­istricts with a large urban population, 02 Banas Kantha 03 Patan 280 271 0.229 29.5
which have stood alone in all the calculations in this paper. 06 Gandhinagar 04 Mahasena, 05 Sabarkantha 437 431 0.367 18.7
Table 5 reports the mean PCTE, the Gini coefficient and the HCR 17 Panchmahals 18 Dohad 320 286 0.281 38.1
07 Ahmedabad 429 592 0.318 11.3
among the 13 pseudo districts in Gujarat in 2004-05. Given so much
16 Kheda 15 Anand 280 290 0.225 32.6
variation in the physical features of the state, the large variation 19 Vadodara 310 577 0.379 6.8
across districts in all three indicators is not surprising. The highest 21 Bharuch 20 Narmada 200 404 0.335 18.8
level of PCTE is in Ahmedabad and it is about 2.2 times higher than 22 Surat 438 531 0.306 13.6
the figure estimated for Banas Kantha district. Compared to two of 25 Valsad 23 The Dangs, 24 Navasari 280 497 0.292 10.1
Gujarat 4,275 434 0.328 17.0
the most developed states (Haryana and Punjab) discussed above,
Mean 425 0.284 16.9
the COV of PCTE is about 6 percentage points higher in Gujarat. The SD 109.0 0.054 11.1
COV of the Gini coefficient is larger than Haryana but lower than COV 25.7 19.1 65.2
Punjab. However, it is in the case of the HCR that large variation is Source: As in Table 1.

228 june 27, 2009  vol xliv nos 26 & 27  EPW   Economic & Political Weekly
inter and intra-state disparities
Table 6: Mean PCTE, Gini and HCR among Districts in Kerala i­ntra-state and inter-state comparisons is that all three have    been
District Code Merged District(s) Sample Mean Gini
and Name Code and Name Size PCTE Coefficient HCR
calculated from the same data source, thus mini­mising biases.
01 Kasaragod 02 Kannur, 03 Wayanad 790 411 0.334 29.8 A comparison of the three characteristics at the state level
04 Kozhikode 460 425 0.335 29.1 throws up a big surprise. Among the five states considered in this
05 Malappuram 550 508 0.402 20.5 paper, Kerala has the highest mean PCTE, even higher than the
06 Palakkad 400 561 0.390 12.8
most prosperous states in India. In the rural sector, the mean
07 Thrissur 480 587 0.375 13.6
08 Ernakulam 480 639 0.388 14.2
PCTE in Kerala was highest in 2004-05 among the major states.
09 Idukki 10 Kottayam 550 697 0.358 5.7 However, this comes at a price – the highest vertical inequality is
11 Alappuzha 370 684 0.439 7.1 also observed in Kerala as indicated by the consumption
12 Pathanamthitta 13 Kollam 630 610 0.335 7.0 e­xpenditure-based Gini coefficient.
14 Thiruvananthapuram 540 856 0.363 4.4
On intra-state inequalities, the picture that emerges is that the
Kerala 5,250 593 0.389 14.8
Mean 598 0.372 14.4
highest level of disparities in poverty incidence is in Punjab, fol-
SD 133 0.034 9.3 lowed by Gujarat and Kerala. Haryana has the least disparities, only
COV 22.3 9.14 64.4 marginally lower than that in Orissa. Compared to 1993-94, dispari-
Source: As in Table 1. ties in poverty incidence have increased in all the states but the
highest increase is in Kerala where the COV of the HCR has increased
4.4  Orissa by close to 40 percentage points. The second highest increase is
Orissa is one of the poorest states in India and on a few indicators seen in Punjab, by 33 percentage points. The COV in Gujarat in-
it is comparable to some of the most underdeveloped areas in the creased by about 25 percentage points. The lowest i­ncrease was in
world. With a persistently high level of poverty for the last few Orissa (6%), which was lower than the increase in Haryana (10%).
decades, Orissa is considered one of the toughest challenges for The increase in the COV observed in 2004-05 is accompanied by
development economists. As apparent from Appendix Table 1, it a substantial increase in the COV of the mean PCTE compared to
has remained the poorest state in India for most of the 30 years 1993-94. In 1993-94, the COV of the mean PCTE across d­istricts in
since 1973-74. During an era of unprecedented economic growth Gujarat, Haryana, Kerala, Orissa and Punjab was 13.4%, 11.8%,
and modest annualised growth of GSDP, the slowest reduction in 14.8%, 13.8% and 13.6%, respectively. As apparent from these
poverty in more than 10 years (between 1993-94 and 2004-05) n­umbers, the level of intra-regional disparities was similar in all
was seen in Orissa (only 2.2 percentage points). During the same the states in the sample. However, over the last 10 years, along
period, the annualised GSDP growth in Orissa at 4.5% per annum with an increase in the GSDP, the mean PCTE in most cases   has
was marginally higher than in Punjab. Table 7: Mean PCTE, Gini and HCR among Districts in Orissa
With this background information, one would expect a large in- District Code Merged District(s) Sample Mean Gini
ter-state variation in the PCTE and the Gini coefficient. But contrary and Name Code and Name Size PCTE Coefficient HCR
03 Sambalpur 01 Baragarh, 02 Jharsuguda, 04 Debgarh,
to expectations, Table 7 suggests that the mean PCTE ranges from 23 Sonapur, 24 Balangir 757 227 0.314 63.5
Rs 195 in Phulabani to Rs 319 in Baleshwar, that is, the highest PCTE 05 Sundargarh 06 Kendujhar 440 274 0.339 51.5
in the state is only 1.6 times compared to the lowest. In the other four 07 Mayurbhanj 240 274 0.351 50.9
states included in this study, the ratio of the mean PCTE between the 08 Baleshwar 09 Bhadrak 440 319 0.310 28.7
12 Cuttack 10 Kendrapara, 11 Jagatsinghapur,
poorest and richest districts is more than 2. It is, therefore, not sur-
13 Jajapur 830 314 0.271 20.8
prising that the COV of PCTE in Orissa is among the lowest at about 14 Dhenkanal 15 Anugul 318 233 0.253 54.5
17%. The COV of the Gini coefficient is fairly low, only marginally 21 Phulabani 22 Baudh, 25 Nuapada, 26 Kalahandi 460 195 0.288 70.8
higher than the COV of the Gini c­oefficient in Kerala. 29 Koraput 27 Raygada, 28 Nabarangapur,
30 Malkangiri 540 196 0.356 71.7
As said earlier, Orissa has had the highest level of poverty
19 Ganjam 20 Gajapati 418 266 0.276 39.5
i­ncidence since 1973-74. It ranged from close to 21% in Cuttack to
18 Puri 16 Nayagarh, 17 Kordha 580 291 0.292 36.3
about 72% in Koraput in 2004-05. The COV of the HCR is among the Orissa 5,023 263 0.320 46.6
lowest in Orissa at 35.5%, which has increased marginally by   about Mean 259 0.305 48.8
6 percentage points from its level in 1993-94. Thus, Orissa is char- SD 44.6 0.035 17.3
acterised by low levels of regional disparities in all the measures COV 17.2 11.53 35.5
Source: As in Table 1.
used here but continues to have the highest levels of deprivation
and poverty among the five states considered in this paper. i­ncreased but the rates have been different among the districts in
each state. As a result, the COV of the mean PCTE increased in all
5  Conclusions and Policy Implications the states but the quantum of increase is modest.
In this paper I have considered intra-state disparities in Gujarat, In sum, the analysis on three indicators carried out in this
Haryana, Kerala, Orissa and Punjab. For investigating them, p­aper suggests that intra-state disparities, which were modest in
three indicators, the mean PCTE, the Gini coefficient and the 1993-94, have increased. The highest increase in disparity is
i­ncidence of poverty, have been considered. Taken together, these found in case of the HCR. Inter-district disparity in real mean
three variables indicate the level of well-being that is a result of PCTE in 2004-05 has also increased in each of the states consid-
the inter-play of a large number of economic and policy variables. ered in this paper but the quantum of increase is modest, in the
One of the advantages of the indicators used in this paper for range of 6 to 12 percentage points.
Economic & Political Weekly  EPW   June 27, 2009  vol xliv nos 26 & 27 229
inter and intra-state disparities
Notes Economic Development and Cultural Change, Sastry, N S (2003): “District Level Poverty Estimates:
1 Towards Faster and More Inclusive Growth: An Vol 35(3), 475-505. Feasibility of Using NSS Household Consumer
Approach to the Eleventh Five-Year Plan, Plan- Murgai, R, M H Suryanarayana and S Zaidi (2003): E­xpenditure Survey Data”, Economic & Political
ning Commission, Government of India, Decem- “Measuring Poverty in Karnataka: The Regional Weekly, 25 January, 409-12.
ber 2006, Chapter 5. Dimension”, Economic & Political Weekly, 25 Janu- Sundaram, K and S D Tendulkar (2003): “NAS-NSS
2 The NSSO identifies the regions combining contig- ary, 404-408. Estimates of Private Consumption for Poverty
uous districts within a state which have similar Parikh, K and R Radhakrishna (2002): India Develop­ Esti­mation: A Further Comparative Examina-
agro-climatic and demographic characteristics. In ment Report 2002 (New Delhi: Oxford University tion”, Economic & Political Weekly, 25 January,
its 2004-05 survey, the NSSO identified 78 regions. Press). 376-84.
3 Among other studies that look at district level in- Rajaraman, I, O P Bohra and V Renganathan (1996): UNDP (2006): “Human Development Report 2006:
dicators are state human development reports “Augmentation of Panchayat Resources”, E­co­ Beyond Scarcity: Power, Poverty and the Global
and state development reports. However, the nomic & Political Weekly, 4 May, 1071-83. Water Crisis”, UNDP and Macmillan.
scope and coverage of these studies is often limit-
ed due to unavailability of standardised and com- Appendix Table 1: Incidence of Poverty (HCR) among Major States
parable outcome variables at the district level.
States (major) 1973-74 1977-78 1983 1987-88 1993-94 2004-05
4 Details about the NSSO CES data are available in HCR Rank HCR Rank HCR Rank HCR Rank HCR Rank HCR Rank
the reports brought out by the Ministry of Statis-
tics and Programme Implementation, Govern- Punjab 28.2 1 19.3 1 16.2 1 13.2 1 11.8 1 8.1 1
ment of India. Haryana 35.4 2 29.6 2 21.4 2 16.6 2 25.1 4 13.6 2
5 See, far example, Debroy and Bhandari (2004) for
district level indicators on hunger, poverty, school
Andhra Pradesh 48.9 5 39.3 4 28.9 3 25.9 3 22.2 2 14.8 3
enrolment, child mortality and immunisation. Kerala 59.8 11 52.2 8 40.4 7 31.8 5 25.4 5 14.8 4
Another source for district level information is Gujarat 48.2 4 41.2 5 32.8 4 31.5 4 24.2 3 17.0 5
the census report, which has basic demographic
information at the district level. Bhalla and Singh Assam 51.2 6 57.2 11 40.5 8 36.2 7 40.9 12 20.4 6
(2001) report some selected district level agricul- Rajasthan 46.1 3 37.4 3 34.5 5 35.2 6 27.4 6 21.4 7
tural outputs.
Tamil Nadu 54.9 9 54.8 9 51.7 12 43.4 12 35.0 8 22.8 8
6 The consumption expenditure data collected by
the NSSO is known as the central sample. There is Karnataka 54.5 8 48.8 6 38.2 6 37.5 8 33.2 7 24.3 9
at least a matching sample size (number of house- West Bengal 63.4 14 60.5 12 54.9 13 44.7 13 35.7 9 24.7 10
holds) that each state or union territory collects
independently in each round of NSS survey. It is, Maharashtra 53.2 7 55.9 10 43.4 9 40.4 9 36.9 10 30.6 11
therefore, possible to pool the data from two Uttar Pradesh* 57.1 10 49.1 7 47.1 10 41.5 10 40.9 11 33.0 12
sources. But to the best of our knowledge, com- Madhya Pradesh* 61.8 12 61.8 14 49.8 11 43.1 11 42.6 13 38.9 13
bining state and central sample data to calculate
district level HCR has been attempted in only two Bihar* 61.9 13 61.6 13 62.2 14 52.1 14 55.0 15 42.0 14
states, Karnataka and Uttar Pradesh. Most of the Orissa 66.2 15 70.1 15 65.3 15 55.6 15 48.6 14 46.6 15
states do not even validate the state sample data.
Mean 52.7 49.2 41.8 36.6 33.6 24.9
7 See Indira et al (2002) for more details on this issue.
8 Among the states considered in this paper, be- SD 10.4 13.6 13.9 11.7 11.2 11.2
tween the 1981 Census and the 1991 Census, four COV 19.8 27.7 33.3 31.9 33.3 45.2
new districts were created in Haryana and two in *These are undivided states, they include Uttarakhand, Chhattisgarh and Jharkhand, respectively.
Kerala. Between the 1991 Census and the 2001 Source: Planning Commission, Government of India (1997, 2007).
Census, 17 new districts were carved out from the
existing districts in Orissa. Appendix Table 2: Name and Code of Districts in the States
9 UNDP (2006).
District Name Code District Name Code District Name Code District Name Code

References Gujarat Punjab Jind 9 Jharsuguda 2


Kachch 1 Gurdaspur 1 Fatehabad 10 Sambalpur 3
Bhalla, G S and Gurmail Singh (2001): Indian Agricul­
ture: Four Decades of Development (Sage Publica- Bans Kantha 2 Amritsar 2 Sirsa 11 Debagarh 4
tion: New Delhi). Kapurthala 3 Sundargarh 5
Patan 3 Hisar 12
Bhandari, Laveesh and Amaresh Dubey (2003): “Inci- Kendujhar 6
Jalandhar 4 Bhiwani 13
dence of Poverty and Hunger in the Districts of Mahesana 4
India”, GCIS Working Paper, Rajiv Gandhi Insti- Hoshiarpur 5 Mayurbhanj 7
Sabar Kantha 5 Rohtak 14
tute for Contemporary Studies, New Delhi. Nawanshahr 6 Baleshwar 8
Gandhinagar 6 Jhajjar 15
Borooah, V and Amaresh Dubey (2007): “Measuring Bhadrak 9
Rupnagar 7 Mahendragarh 16
Regional Backwardness: Poverty, Gender and Ahmedabad 7 Kendrapara 10
Children in the Districts of India”, Margin – The Fatehgarh Sahib 8 Rewari 17
Journal of Applied Economic Research, Vol 1 (4), Surendranagar 8 Jagatsinghapur 11
Ludhiana 9 Gurgaon 18
403-440. Rajkot 9 Cuttack 12
Debroy, Bibek and Laveesh Bhandari (2004): “District Moga 10 Faridabad 19
Jamnagar 10 Jajapur 13
Level Deprivation in the New Millennium” (New Firozpur 11 Kerala
Delhi: Rajiv Gandhi Institute for Contemporary Dhenkanal 14
Porbandar 11 Muktsar 12 Kasaragod 1
Studies). Anugul 15
– (2007): “Exclusive Growth–Inclusive Inequality?”,
Junagadh 12 Faridkot 13 Kannur 2
Nayagarh 16
FISME Occasional Paper 1, September. Amreli 13 Bathinda 14 Wyanad 3
Khordha 17
Dubey, A and S Gangopadhyay (1998): “Counting the Bhavnagar 14 Kozhikode 4
Mansa 15 Puri 18
Poor: Where are the Poor in India?”, Sarvekshana:
Analytical Report, No 1, February. Anand 15 Sangrur 16 Malappuram 5 Ganjam 19
– (2006): “Towards Faster and More Inclusive Kheda 16 Patiala 17 Palakkad 6 Gajapati 20
Growth, An Approach to the 11th Five Year Plan”, Haryana Thrissur 7
Planning Commission, December. Panch Mahals 17 Kandhamal 21
Panchkula 1 Ernakulam 8 Baudh 22
– (2007): “Poverty Estimates for 2004-05”, Planning Dohad 18
Commission, Press Information Bureau, March. Ambala 2 Idukki 9 Sonapur 23
Indira, A, M Rajeev and V Vyasulu (2002): “Esti­mation Vadodara 19
Yamunanagar 3 Kottayam 10 Balangir 24
of District Income and Poverty in Indian States”, Narmada 20
Economic & Political Weekly, June 1, 2171-77. Kurukshetra 4 Alappuzha 11 Nuapada 25
Bharuch 21
Jain, L R, K Sundaram and S D Tendulkar (1988): Kaithal 5 Pathanamthitta 12 Kalahandi 26
“D­imensions of Rural Poverty: An Inter-regional Surat 22 Karnal 6 Rayagada 27
Kollam 13
Profile”, Economic & Political Weekly, 15 Novem- The Dangs 23
ber, 2395-2408. Panipat 7 Thiruvananthapuram 14 Nabarangapur 28
Kuznets, S (1955): “Economic Growth and Income Navsari 24 Sonipat 8 Orissa Koraput 29
I­nequality”, American Economic Review, Vol 45. Bargarh 1 Malkangiri 30
Mathur, Ashok (1983): “Regional Development and
Income Disparities in India: A Sectoral Analysis”, Source: GoI (2004).

230 june 27, 2009  vol xliv nos 26 & 27  EPW   Economic & Political Weekly