This article shows that the interstate differences in poverty rates can be largely explained by differences in the per capita gross domestic product, agricultural growth and the share of the bottom 40 per cent of the population in consumption. To eliminate poverty, economic policy therefore has to accelerate growth, focus programmes on agriculture and rural development in the poorer states and target subsidies at the bottom 40 per cent. The most critical areas distinguishing state growth performance have been modern (registered) manufacturing and commerce.
SPECIAL ARTICLEjanuary 12, 2008 Economic & Political Weekly54Growth and Poverty: Policy Implications for Lagging StatesArvind VirmaniThis article shows that the interstate differences in poverty rates can be largely explained by differences in the per capita gross domestic product, agricultural growth and the share of the bottom 40 per cent of the population in consumption. To eliminate poverty, economic policy therefore has to accelerate growth, focus programmes on agriculture and rural development in the poorer states and target subsidies at the bottom 40 per cent. The most critical areas distinguishing state growth performance have been modern (registered) manufacturing and commerce.There is a view favoured by the anti-globalisation activists, left economists and the global/international socialists that faster growth in India has not reduced poverty. A sub-set of these personalities go so far as to assert that higher growth may even have caused or contributed to the widening of interstate gaps in income and poverty. On the periphery of this group are those who assert that growth is the least important issue among dozens that they can list. On the other extreme, are a small number who assert that the faster the growth, the better, and as long as growth is fast, there is not much else that the government needs to worry about. Perhaps, a careful exami-nation and analysis of the facts can help resolve some of these issues, even though it is unlikely to convince these extremes. This paper analyses the data on interstate variations in growth and poverty, to see what we can learn about economic growth and social welfare, with a view to improving planning and policy formulation.Virmani (2005, 2006a, 2006c) has presented an extensive analysis of aggregate growth from 1951-52 to 2004-05. In this analysis, central government policies relating to the external sector and monetary, fiscal and industrial policies that are the preserve of the central government are considered. Interstate dif-ferences in growth are, however, likely to arise either from the conditions in the states that affect the impact of central policies or from the differences in policies that are the preserve of the state governments. Since, a large sample of NSS data is available for 1993-94, 1999-2000 and 2004-05 and the state gross domes-tic product (SGDP) series in 1993-94, prices are available from 1993-94 to 2004-05, we focus on this period.Section 1 analyses the first aspect of welfare, per capitaGDP and its growth. Section 2 deals with the second important deter-minant of social welfare, poverty and analyses its links to growth and other aspects. Section 3 concludes the paper.1 GrowthAs shown in Virmani (2006a, 2006c) aggregate economic growth accelerated in the 1980s from an average of about 3.5 per cent per annum between 1951-52 and 1979-80 to about 5.8 per cent per annum during 1980-81 to 2004-05. More recent data shows that aggregate growth averaged about 5.5 per cent per annum during 1980-81 to 1994-95 and has accelerated further to an average of 6.8 per cent per annum during 1995-96 to 2006-07. Ahluwalia and Bajpai and Sachs have shown that the acceleration in growth was less in the poorer states, so that interstate inequality in per capita income has increased. Any views expressed in the paper are those of the author and do not necessarily reflect those of the organisation for which he works. The author thanks K L Datta for help in generating consumption data.Arvind Virmani(virmani@nic.in) is with the department of economic affairs, ministry of finance, government of India.
1993 94 2004 05
SPECIAL ARTICLEjanuary 12, 2008 Economic & Political Weekly56Public and Quasi-Public GoodsIn our view, therefore, public and quasi-public goods that affect investment are likely to be more important determinants of interstate variation in growth than other infrastructure goods and services. This would be particularly so in the relatively poorer or backward states, which have not shared in the growth acceleration.The fifth hypothesis for interstate differ-ences in growth, pro-posed in this paper, is related to services as the drivers of growth since the 1980s. Much of the acceleration in aggregate growth has come from an acceler-ation of growth in the services sector. Cer-tain services have ac-celerated more than others in the aggre-gate. Bosworth et al (2007) have noted that though software and related services have received a lot of attention, the acceler-ation of the services sector’s growth, “has been more broadly based, including trade, transportation and community and personal services”. We hypothesise, that the eco-nomic growth in poorer states has not been accelerated because services that have propelled national growth have not accelerat-ed proportionately. This can help us identify the state policies that can help accelerate the growth of poorer states.5If we combine this with the previous hypotheses, then we would expect that services which are particularly dependent on, or, whose growth is associated with, the development of public and quasi-public goods are likely to prove important in explain-ing the interstate differences in growth. For instance, there are small but significant quasi-public goods associated with travel and tourism. These include historical monuments, cultural and religious sites, cultural (including religious) traditions, events, local art and crafts and natural attractions (such as water bodies, rivers, waterfalls; forest and animal reserves). It also includes prosaic public services items like clean drinking water, public toi-lets, lawns and flower beds at these sites. Preservation, enhance-ment and development of these quasi-public goods by the state or local governments will impact the economy through the sectors mentioned by Bosworth et al.1.2 Data and AnalysisWe use the national account data by state (and union territories (UTs)) and sector. The SGDP at constant 1993-94 prices is available for most states by sector. For a few states or UTs it is available only till 2003-04(1) or 2002-03(2).6 For each sector we calculate the compound annual rate of growth between 1993-94 and 2004-05 and use this as the basic data for analysis. In the case of some states and UTs data is missing for one or more sectors (e g, communications in Punjab, Sikkim and Jammu and Kashmir).Table 3 presents the summary statistics of economic growth across states by sector. Column 2 of the table shows that the two modern service sectors, communications and banking and insurance, are the fastest growing sectors across states. Further, communications is also considered an infrastructure, and in the days of landlines, it would have been considered on par with electricity as a candidate for characterisation as a quasi public good in rural areas. Note also that, there was little difference in the mean growth rate of the secondary and tertiary sectors. Columns 3 and 4 show the standard deviation and coefficient of variation of the growth rate across states for each sector. The electricity sector has the highest standard deviation of growth rate with a coefficient of variation of over 1. The primary sector and its four components have the highest coefficient of variation. Though one expects the variability in agriculture growth rates to be high because of rainfall variation, the high variability in mining and fishing is somewhat surprising! Growth and Total GDPFinally, column 5 of Table 3 shows the correlation of the growth of each sector with the total GDP of the state. At a broad level, the secondary sector is found to have a higher correlation (0.75) with total growth than the tertiary sector (0.7). If we look at individual sectors, “Trade, Hotels and Restaurants” is found to have the highest correlation of 0.67 with total SDP, followed by registered manufacturing with 0.5 and communication with 0.44. On the negative side, agricultural growth is almost uncorrelated with total SDP, while forestry and fishing have a negative correlation. The electricity sector has a modest correlation of 0.19.Table 3: Summary Statistics of State Growth (1993-94 to 2004-05)Sector Mean StdDv CV Correlation (%) (%)Agriculture 2.1 3.3 1.55-0.02Forestry 1.4 3.5 2.53 -0.39Fishing 3.3 6.0 1.84-0.45Mining 4.47.61.72-0.28Primary 1.37.35.82-0.24Manufacturing 5.4 4.1 0.750.41 Registered 5.8 4.3 0.74 0.50 Unregistered 4.7 3.8 0.82 0.21Construction 8.4 4.30.510.22Electricity 7.0 8.6 1.220.19Secondary 7.2 3.4 0.470.75Transport, communication, etc 10.9 3.3 0.30 0.56 Railways 5.9 3.5 0.59 -0.10 Other transport 7.3 3.2 0.44 0.05 Storage 4.5 4.1 0.90 -0.17 Communication 21.1 5.2 0.25 0.44Trade, hotels 6.3 3.3 0.53 0.67Bank, insurance 10.8 2.3 0.22 -0.06Rest, business service 5.4 2.4 0.44 0.36Govt administration 6.1 1.4 0.24 0.08Other services 6.7 2.3 0.34 0.25Tertiary 7.4 1.5 0.210.70SDP total 6.2 1.9 0.31 1.00ST stands for sub total.Table 4: Summary Table of Regression Coefficients(Dependent Variable is State GDP Growth)Variables Model 1a Model 1b Model 2a Model 2b Model 3a Model 3b Model 3b SURERate of growth of (cmpnd) Agriculture 0.42160.17860.089640.04030.0225 (3.2***) (1.8*) (1.0) (0.54) (0.42)Manufacturing 0.5473 0.1479 0.2258 0.1919 0.1911 regd (6.5***) (1.4) (3.6***) (2.6**) (3.6***)Electricity 0.19860.04160.0107-0.0205-0.0117 (3.6***) (1.1) (0.28)(-0.65) (-0.52)Communication 0.2195 0.1454 0.06450.0722 (5.2***) (5.1***) (1.6) (2.5**)Trade, hotel, 0.6031 0.3858 0.4361 0.3726 0.4329 restaurants (7.8***) (5.4***) (5.4***) (5.4***) (8.6***)Construction 0.23620.07520.20010.15330.0845 (4.2***)(1.5)(3.4***)(2.5**)(1.9*)F (Chi for SURE) 52 86 181 233 123 159 1267Prob > F (Chi) 0.0 0.0 0.0 0.0 0.0 0.0 0.0R2 0.860.940.930.970.960.980.98Adj R2 0.840.930.920.960.950.97RootMSE 0.026 0.018 0.018 0.013 0.014 0.014 0.010No of observations 29 26 29 26 29 26 261 Numbers in bracket are t (z) statistics, stars show confidence level * = 10%, ** = 5%, *** = 1%2 In the SURE regression growth rates of the variable “trade, hotels, restaurants”, communications and registered manufacturing were treated as dependent on total state GDP growth (uses z(not t) stat).
SPECIAL ARTICLEEconomic & Political Weekly january 12, 200857We do the model testing in two versions of the models (a and b), without and with communications as the latter is missing for three states/UTs. The results are summarised in Table 4 (p 56). We test the first three hypotheses by regressing the compound annual growth rate of state domestic product (GrSdp) between 1993-94 and 2004-05 on the growth rates of SDP from agriculture (GrSag), registered manufacturing (GrSmreg) and electricity (GrSelec). All three coefficients are significant at the 1 per cent level of con-fidence when communication variable is not included (Model 1a).7 The R2 is also a relatively high 0.86 with the adjusted R2 at 0.84. When the communication variable is introduced (Model 1b), it is highly significant, but both registered manufacturing and electri-city become non-significant, while agriculture is barely significant.Investment in roads, dams and canals are not captured sepa-rately in the national accounts, but are included in the construc-tion sector. However, the construction sector includes a much larger proportion of private construction. Because of their public good character, value added by roads and dams and canals is not measured separately, but is part of the value added by the users of these public goods, the road transport and agriculture sectors, respectively. In the case of roads, the externality is much broader in terms of general economic activity. For those who have tra-velled regularly down any highway over a number of years, will have noticed how economic activity, including shops and ‘dhabas’, spring up along newly built or improved/widened high-ways in three-five years. Such important externalities are proba-bly captured best by the sector, Trade Hotels and Restaurants. Besides roads, communication is also essential for growth of trade, which in turn, is necessary for the development of agri-culture, mining and manufacturing in rural areas. We, therefore, test the fourth and fifth hypotheses by regress-ingSDP growth on the growth of the sectors Trade, Hotels and Restaurants (GrTrHtlRes), construction (GrConst) without and with communications (GrCom).8 The results are shown under the columns Model 2a and 2b of the table. In the absence of commu-nication growth variable, both variables are highly significant (at 1 per cent level). The R2 and R2 (adj) are 0.94 and 0.93 respec-tively, both higher than those in Model 1a, indicating that Model 2a has higher explanatory power than Model 1a. When communi-cation variable is introduced into the Model (2b) it is highly sig-nificant but construction growth becomes non-significant. The R2 and R2 (adj) are 0.97 and 0.96 respectively, both higher than those in Model 1b, but now the gap is narrower.How do we choose between the different models? The fact that both sets of models give significant results suggests that they may be subject to missing variable bias. We, therefore, run the regres-sion by including all the variables identified. The results are given in columns marked Models 3a and 3b. Agriculture and electricity growth are not significant in either equation, indicating that they do not explain even a fraction of the difference in growth rates across states and UTs during the period following the “new eco-nomic policies” initiated in the 1990s. The only variable from the first three hypotheses that survives is the rate of growth of regis-tered manufacturing (at 5 per cent level), while both Trade, Ho-tels and Restaurants and constructions variables associated with the last two hypotheses are significant at 1 per cent level.9 We find that growth ofSGDP affects the variables, growth of communications and Trade, Hotels and Restaurants at a 1 per cent level and registered manufacturing at 10 per cent level. The coefficients of these variables could, therefore, be biased by the simultaneity problem. We, therefore, rerun the last equation (Model 3b) usingSURE (sum model of regression. This confirms that these three variables are significant, though the level of significance increases to 1 per cent for registered manufacturing and falls to 10 per cent for construction. The communication growth variable is also found to be highly significant (at 1 per cent level). The other two variables, growth of agriculture and electricity, remain non-significant. The numerical impact of the four sectors identified as drivers of state GDP(SGDP) growth is quite high. A 1 per cent dif-ference in the rate of growth of Trade, Hotels and Restaurants and 1 per cent in registered manufacturing can result in 0.43 per cent and 0.19 per cent difference, (respectively) in the rate of growth ofGDP. This impact is much larger than the average share of these sectors in all-India GDP of 0.145 per cent and 0.13 per cent respectively during 1993-94 to 2004-05. The impact of a 1 per cent difference in telecommunications and construction growth onSGDP growth is 0.07 per cent and 0.08 per cent respectively. This impact is also greater than their average share in all-India GDP of 0.025 per cent and 0.051 per cent, respectively. To illustrate the impact consider Uttar Pradesh (divided), which grew at 4.3 per cent per annum, 2.1 per cent points (or one-third) lower than the 26 state mean of 6.4 per cent per annum. If the rate of growth of trade et al, manufacturing (regd) and construction in Uttar Pradesh was raised to the mean rate of 6.9 per cent, 6.4 per cent and 8.7 per cent, respectively, Uttar Pradesh (divided)’s SGDPwould have grown at 6.4 per cent per annum. Assam and Bihar could correspondingly have grown at 5.4 per cent and 6 per cent per annum instead of 3.5 per cent and 4.5 per cent respectively.101.3 ImplicationsandRecommendationsVirmani (2005) has identified the “double dualistic” structure of the economy, predominantly in the manufacturing sector, as a drag on productivity growth.11 The current paper deepens our understanding of this phenomenon. Despite labour policy rigidi-ties and other policy distortions such as small scale industry(SSI) reservations and tax exemptions that have created this dualism, the registered manufacturing sector remains a driver of economic growth in India. Though the policy distortions have eliminated India’s competitive advantage in “labour-intensive mass manu-facturing”, vis-à-vis China and other labour-surplus countries with more flexible labour markets, this is not so for “semi-skilled labour-intensive goods”. On the contrary, (we hypothesise that) India’s entrepreneurs have responded to policy induced handicaps by;(a) Focusing on niche markets that do not require high volume production.(b) Outsourcing labour-intensive parts of the production process in-cluding service activities that were formerly part of “manufacturing”.(c) Fragmenting production into smaller units that are subject to less rigid labour rules and procedures.
SPECIAL ARTICLEjanuary 12, 2008 Economic & Political Weekly58(d) Using and/or developing technology and systems that substitute semi-skilled for unskilled labour as the labour rules are less rigid on the former.12(e) Using more capital-intensive methods and processes.Thus India has developed (in our view) a comparative advan-tage in ‘semi-skilled labour-intensive’ manufacturing that is visi-ble in automobile components and specialised chemicals, drugs and pharmaceuticals. Those states that have been able to apply labour apply rules and procedures flexibly, reduce bureaucratic red tape, curb the predation by government monopoly service providers and provide a more attractive investment environment, have more benefited. One must caution, however, that as long as labour rigidities remain, this sector cannot generate unskilled employment at a pace necessary to correct the distorted structure of employment (two-thirds of labour force in agriculture). Informal SectorThe major part of the dual economic structure is the informal, unorganised and predominantly services related sector. Most ob-servers have been too mesmerised by the information technology, IT-enabled services(ITES) and high tech services to notice that the rest of the services sector is also important. Bosworth, Collins and Virmani (2007) have showed that the fastest growth in total factor productivity has been in the services sector, and that too during the 1980s, whenIT andITES were not even a speck on the horizon. The current paper goes beyond earlier analyses to show that a humdrum sector like Trade, Hotels and Restaurants, has been a major driver of growth since 1993-94 (the market reform phase).13 If we examine these results and then a step beyond them an interesting narrative emerges of economic development through trade, commerce, tourism and urbanisation.14 With rigi-dities in labour policy and low literacy/primary education levels stymieing the growth of “labour-intensive mass manufacturing”, states have differentiated themselves in the 1990s by their rela-tive growth of trade, commerce, tourism, real estate development and construction. For rural areas (villages, habitation) to utilise their compara-tive advantage they have to be connected to markets and towns. Sometimes the comparative advantage has to be developed in the form of historical monuments, cultural/religious sites or natural resources (water bodies and courses). Roads and communica-tions are necessary conditions for this to happen. State policies, rules and regulations that facilitate this process, then lead to fast-er growth, through construction of shops, workshops, restau-rants, offices, hotels and housing both on the town roads extend-ing into the countryside as well as on roadside villages. The qual-ity range is wide and diverse from a dhaba to an air-cooled or air-conditioned restaurant to an eating place in an air- conditioned mall, as is the diversity in the education and skill levels of the employees needed to service these establishments.Roads: The importance of roads in economic development has gradually come back on the agenda over the last decade, after relative neglect for half a century. This is reflected in the central government programmes like the national highways develop-ment project (NHDP) and Prime Minister’s Grameen Rozgar Yojana (PMGRY). It is still not clear, however, whether state governments recognise their critical importance. I would recom-mend road development as the number one focus programme for all poor states and regions in the country. States must have a de-tailed operational plan for building a comprehensive road grid connecting every village and habitation within a time bound framework. At the high end, we would also find a mix of high quality offices and communication services coupled with busi-ness services (call centres).(a) All cities in the country must be connected by national high-ways that are constructed and maintained to middle income country standards.(b) All towns in every state must be connected to each other by state highways of a standard that can sustain traffic at least 30 kph during adverse conditions such as the monsoons. (c) Each village must be connected to the nearest town by an all weather metalled/surfaced district highway. Each village must be connected to all neighbouring village by a road of a standard suitable to the traffic.(d)All remaining habitations must be connected to the grid within 10 years.(e) New highways must identify and plan for the emergence of shopping and trading areas near villages and highway/road in-tersections. This will entail development of small stretches of parallel local roads, underpasses for rural traffic and designa-tion/acquisition of contiguous areas on the non-highways side of the parallel roads for commercial activity.(f) State road transport, goods and passenger, must be de-licensed to build rural and local entrepreneurship in road trans-port. At most states could have a system of registration along with registration fees. Registration should be automatic unless there is a pattern of systemic violation of road rule. Registration fees should be automatically transferred to a road maintenance organisation at the local or state level charged with maintaining district/state highways.(g) All octroi posts and other barriers to intra-state traffic must be dismantled. Communications: The communication revolution is already sweeping the country. Its impact on the rural areas of the country can be expedited if some anomalies in policy and regulation are corrected.(h) The universal services obligation (USO) fund must be used to expedite extension of telephony and internet into the rural areas. It should, for instance, hold subsidy auctions for extending the mobile footprint into areas of non-presence.(i) Optical fibre cables, landlines and all other wires going into the rural areas must be subject to open access with the Telecom Regulatory Authority of India (TARI) ensuring that reasonable rates are charged for access. The department of telecommunica-tions (DOT) and the telecom commission should stop siding with Bharat Sanchar Nigam(BSNL) if they are genuinely interested in extending communications to rural areas.(j) TRAI had previously recommended that all physical mobile infrastructure in rural areas should be sharable. We should go further and completely unbundle the local loop in rural areas so
SPECIAL ARTICLEEconomic & Political Weekly january 12, 200859that they can get the benefits of competition that urban areas are already benefiting from.Let a 100 Towns Bloom: Allow and encourage private entre-preneurs to build thousands of new towns/townships in semi-urban, semi-rural areas. Government has the responsibility of connecting these to the nearest highway and water-supply mains. Planning water reservoirs, solid waste disposal and sewage treatment is also stategovernments’ responsibility. States must scrap forth-with expropri-atory sections of the rent control act(s) and corres-ponding rules and proce-dures, so that the private sector can build and provide rental accommodation for the lower middle class and the poor.2 PovertyThe question of whether growth is a necessary or a sufficient condition for the elimination of poverty has been long debated. The focus of our investigation is precise: to what extent do in-terstate differences in average per capita income of states (as measuredby the National Accounts Statistics(NAS)) explain in-terstate differences in poverty rates as measured by the official poverty data?2.1 All India PovertyThe overall poverty picture at the national level is presented in Tables 4, 5 and 6. Over the 11-year period from 1993-94 to 2004-05, the proportion of poor below the poverty line and the absolute number of poor have declined. The poverty ratio has declined by 23 per cent from 36 per cent of population to 27.5 per cent while the number of poor has declined by 6 per cent to 301.7 million (Table 5). The other noteworthy feature is the convergence of rural and urban poverty rates from a gap of 4.9 per cent points in 1993-94 to 0.8 per cent points in 2004-05. This suggests that the rural and urban areas are getting better integrated in terms of movement of workers, goods and services and the price differentials that drive them.Some commentators have speculated that the rate of poverty reduction has slowed because poverty de-clined by about 8.5 per cent points during the 10-year period from 1983-84 and 1993-94 by about 8.4 per cent points in the 11-year period from 1993-94 to 2004-05. Despite the rise in growth rate of GDP. The decline was however 19 per cent in the former and 23 per cent in the latter, indicating that the rate of decline was virtually unchanged between these two periods. A similar picture emerges if we look at the 1990s (Table 6).2.2 DistributionFunctionGiven an income or consumption distribution f(y), and a poverty line Yp the proportion of people below the poverty line (poverty rate or headcount ratio) is given by, F(Yp), where F is the cumula-tive distribution. The headcount ratio or proportion of people below the poverty line is therefore,F(Yp) = F (Y; Ya, Ω )| Y =Yp , ...(1)where Ya = the average or mean of the income or consum-ption distribution and Ω is a distributional parameter(s) like variance. If the distribution for each state is assumed to be drawn from the same family of distributions and differ from each other in terms of the mean and distribution parameter(s) then we can differentiate this distribution around the poverty line Yp (which is the same in real value across Indian states) to obtain,d F(Yp) = F1 (Yp; Ya, Ω ) dYa + F2 (Yp; Ya, Ω ) d Ω | at Y = Yp ...(2)where F1 and F2 are respectively the differentials of the cumula-tive distribution with respect to mean income or consumption and the distributional parameter, evaluated at Y =Yp. If we divide both sides of equation (2) by F and rearrange we obtain an elas-ticity form of the equation.d F/F = (Ya F1/F) (dYa/Ya) + (Ω F2 /F) (dΩ / Ω) | at Y = Yp ...(3)2.3 EmpiricalEstimationThe estimating equation based on equation (3) can be written as,GrPovertyi = A GrPcSgdpi + B GrShrL40Ri + C GrShrL40Ui + ¤i ...(4)with i = 1…n are the states, A, B and C are parameters to be esti-mated and ¤ is the error term. GrXi denotes compound annual rate of change of the variable X in state i between 1993-94 and 2004-05, poverty is the poverty rate or headcount ratio for the state, PcSgdp is the per capita state GDP, ShrL40R is the share of the lower 40 per cent of the state’s rural population in rural con-sumption and ShrL40U is the share of the lower 40 per cent of the state’s urban population in urban consumption.For the poverty rates (dependent variable) we use the official poverty estimates for 2004-05 and 1993-99 estimated from the uniform recall period(URP) data for each state. We can also obtain the MRP estimates for 1993-94 and use these along with official MRP based estimates for 2004-05. As we estimate the equations in difference form using state panels, the new states of Jharkhand, Chhattisgarh and Uttrakhand have to be left out of the estimation.For the independent variables, we use the national account data for per capita state GDP. It is, however, also possible to use the average monthly per capita expenditure (MPCE) calculated from the survey data, by converting 1993-94 data to 2004-05 prices using the same deflators that are used to Table 5: National Poverty: Head Count Ratio 1993-94/1999-20002004-05 URPMRPURPMRPRural Poverty ratio 37.3 27.1 28.3 21.8 Number of poor (million) 244.0 193.2 220.9 170.3Urban Poverty ratio 32.4 23.6 25.7 21.7 Number of poor (million) 76.3 67.0 80.8 68.2Total Poverty ratio 36.0 26.1 27.5 21.8 Number of poor (million) 320.4 260.3 301.7 238.5URP = uniform recall period, MRP = mixed recall period.Table 6: Change in Poverty Rate between 1993-94 and 2004-05 (% Point Per Year) (Compound Annual) 1993-94 to 1990-2000 1993-94 to 1999-2000 2004-05 to2004-05 2004-05 to2004-05 URP MRP URP MRP1 Rural 0.78 1.06 2.48 4.262 Urban 0.60 0.38 2.07 1.683 Total 0.77 0.87 2.40 3.56Note: As in Table 5.
SPECIAL ARTICLEjanuary 12, 2008 Economic & Political Weekly60derive the poverty line in current prices. For the distribution vari-able, we use the share of bottom 40 per cent of the population in rural and urban areas. One of the reasons for using this variable to represent distributional factors is that in cross country regres-sions this variable has proved significant in explaining differenc-es in poverty [Virmani 2006a, 2006d]. Alternative variables such as the share of the bottom 20 per cent, etc, turn out to be non-significant in the cross-country context.A number of proxies have been used in the literature to capture distributional parameters. Bhalla (2002) has the most direct measure, which he calls “shape of distribution elasticity” (SDE). Ahluwalia (1978), Bell et al (1994), Ravallion and Datt (1996) and Datt and Ravallion (1998), have shown that agricultural output or productivity is linked to poverty.15 If agricultural and non- agricultural growth have significantly different impact on overall (combined rural and urban) poverty, it would imply that the pattern of growth can have distributional consequences. This along with the fact that a large proportion of the poor live in rural areas and are directly or indirectly linked to agricul-ture, implies rural-urban seg-mentation of markets for fac-tors (labour, capital) and goods. The labour market segmenta-tion could be due to lack of in-formation about non-agricul-tural/urban jobs, transaction costs of moving to the urban area, lack of mobility for social and other reasons or lack of re-quired education/skills. Goods market segmentation could be partly due to poor transport and communication links, while the capital/credit market segmentation is generally re-lated to information problems and rule of law (contractual arrangements, enforcement).The issue of the special link between agriculture growth and poverty suggested by the above research, can be reframed in the present context as follows: whether agriculture growth contrib-utes something additional to poverty reduction that is not cap-tured by growth in average per capita income/consumption and the distributional parameter used. We test this hypothesis by in-troducing the compound annual growth rate of state GDP from agricultural between 1993-94 to 2004-05 (GrSgAg) as an addi-tional variable in equation (3) to obtain an alternative estimation equation,GrPovertyi = A GrPcSgdpi + B GrShrL40Ri + C GrShrL40Ui + D GrPcSgAgi + ¤i ...(5) The least squares estimates of equations (4) and (5) are sum-marised in Table 7. The most important result is that the per capi-ta state domestic product is significant at the 1 per cent level of confidence in all these estimates. There is thus no factual basis to the ideological position that income growth has had a neutral or perverse effect on interstate poverty gaps. Estimating equation (4) shows that every 1 per cent increase in the per capita state GDP results in a 1 per cent reduction in the poverty rate (columns 2 and 6 of Table 7). The second result is that only the rural distri-bution is statistically significant, with a 1 per cent improvement in the distribution leading to a 0.8 per cent reduction in the pov-erty rate. This simple regression explains between 63 per cent and 67 per cent of the URP-based poverty. The results for the MRP-based poverty are similar.16 If we estimate the model with agriculture growth, this variable is highly significant and modifies the impact of the other varia-bles (column 3). Every 1 per cent increase in agriculture growth reduces the rate of URP poverty by 0.45 per cent, in addition to its effect on average per capitaGDP. This model, therefore, has about 9 per cent more explanatory power for the URP poverty. The in-troduction of agriculture growth also reduces the impact of per capita income by 14 per cent and of consumption shares of lower 40 per cent of rural population by about 25 per cent. Third, the consumption share of the lower 40 per cent of urban population (-0.83) now becomes signifi-cant and higher than that of rural shares (-0.62). Lastly, as a consequence, the model now explains a higher 74 per cent and 81 per cent of the overall difference in reduction in pov-erty rates across states. The fourth column shows the regression results of the same model if the growth of SGDP from agriculture is re-placed by per capitaSGDP from agriculture. The results are substantially the same except that the effect of per capitaGDP growth reverts to 1 and the ef-fect of rural and urban shares is now almost identical.2.4 PolicyImplicationsThus we find that though differences in agriculture growth across states provide no explanation for differences inSGDP growth, they are important in explaining differences in total poverty reduction. Agriculture – Public Goods: There are two approaches to agri-culture growth. One based on the detailed and comprehensive analysis of all the problems that have arisen in the agriculture sector/rural areas across this vast country and lists every policy, institutional and programme change needed for agriculture to thrive. I call this the symphonic approach where all the instru-ments play in harmony to produce good results. The other ap-proach, which could be called the quartet, identifies a few critical actions that government must take which are particularly impor-tant for agriculture/rural development. There are three public goods and one quasi-public good which may be particularly Table 7: Estimated Coefficients for Poverty EquationsDependent Variables = Poverty Poverty Poverty Poverty Poverty Poverty Independent Variables URP URP URP MRP MRP MRP 1 2 3 4 5 6 7Per capita state GDP -1.025 -0.863 -1.06 -0.982 -0.922 -0.902Per capita state GDP non-agri -7.0*** -6.3*** -7.8*** -6.8*** -6.6*** -7.3***Cons share of lower 40%: rural -0.821 -0.621 -0.736 -0.656 -0.583 -0.597 -2.1**-1.8*-2.0*-1.8*-1.7-1.8*Cons share of lower 40%: urban -0.655 -0.832 -0.793 0.132 -0.147 -1.6-2.3**-2.1**0.3-0.3State GDP from agriculture -0.454 -0.291 -0.275 -3.1***-1.9*-2.0*State per capita GDP from agriculture -0.447 -2.3**F 171916232121Prob > F 0.0 0.0 0.0 0.0 0.0 0.0R2 0.670.760.730.770.810.81Adj R2 0.63 0.72 0.68 0.74 0.77 0.77Root MSE 0.031 0.027 0.029 0.031 0.027 0.027No of observations 28 28 28 24 24 24
SPECIAL ARTICLEEconomic & Political Weekly january 12, 200861lacking in poor, backward regions, namely, the rule of law, per-manent all weather roads, knowledge/information and assured irrigation. Many of the poorest regions are still characterised by semi-feudal relations in land, labour and credit markets. In the absence of rule of law, there is nothing to keep traditional feudal families from transforming into tin pot oligarchs who supply spurious fer-tiliser, pesticides, etc, and use strong arm methods to collect overdue loans or extract indentured labour. Historically, the continental interior and geographically re-mote and hilly areas are least likely to have been connected to the main transport corridors, and therefore, the gaps are likely to be widest. Research and development (R&D) on crop and non-crop agri-culture and animal husbandry, including new varieties, opera-tional methods and management practices has traditionally been generated by government universities and transmitted to farm-ers by public organisations. The deterioration in volume and quality of this knowledge transfer must be reversed. The syner-gies between telecom connectivity, internet access, e-governance, e-learning and e-marketing must be exploited. Water Management: Water is essential for drinking, personal hygiene, sanitation and irrigation. From a global comparative perspective, India is a relatively water-scarce country and global environmental changes threaten to make this worse. Yet our limited water resources are either not fully utilised (flow to the sea) or are misused (depleting groundwater). Public water supply systems also need to make better use of rain water.There is an urgent need to improve the comprehensiveness and quality of water planning and management at every level (centre, state, district, town, panchayat and smallest farmer). Water harvesting, water shed development, recharge of water bodies and aquifers, must be planned and implemented in every nook and corner of the country. Education and demonstration of models with the active participation of NGOs can play an impor-tant role. Dams and canals have a place in cutting down the flow of water into the sea, recharging aquifers and supplying dry areas and parched towns. Tube wells in depleting aquifers must be discouraged through proper pricing of electricity and perhaps even water.Primary Education: Literacy can help in acquiring knowledge about job opportunities, tools and productivity. Government must ensure that every member of the labour force, every citizen, has the education that (s)he is supposed to acquire with the com-pletion of primary education. But this education must also be made more relevant by providing information on agriculture and allied subjects and training them on how to access relevant infor-mation in future. We should not declare premature victory for primary educa-tion and move on to higher levels and once again deprive the poor of their access to basics while satisfying the middle classes hun-ger for secondary education. The latter can be better achieved through a modern, transparent, regulatory system that minimises the problem of asymmetric information [Virmani 2005, 2006 b], fosters competition in supply17 and empowers them to get value for the 3 per cent of GDP that they already spent on private education.Every youth, rural or urban, after completing primary educa-tion must also have access to the 6,000 or so globally identified skills. This requires a massive joint effort by government, NGOs and private skill providers. Government must provide funding for the poor while all possible private and foreign expertise and experience is attracted to India to provide training in all these skills in the next five years.This section confirms that average per capitaGDP is an impor-tant determinant of poverty. It also shows that the higher agricul-ture growth has an impact on poverty reduction in addition to its normal contribution to overall GDP growth. A special focus on agricultural growth in poorer states and in the states with oppor-tunities for productivity improvement can therefore be justified in terms of poverty removal even though it may not have any im-pact on overall growth. The empirical results also justify an add-ed focus on rural roads and telecom connectivity (in addition to the general effects found earlier) to the extent that they promote the development of agriculture. Development of rural connectiv-ity also improves market integration and labour mobility, which in turn, will remove the differential and segmented impact of growth on rural and urban poverty.This section also shows that the consumption share of bottom 40 per cent of the population is an important determinant of pov-erty. Targeted benefit programmes should therefore focus on the bottom 40 per cent of the population. It is essential to set up a comprehensive data base with unique identifications – photo-graphs and bio-metric identifications that will eliminate fraud and help identify the poorest 30 per cent to 40 per cent of the population.18 3 ConclusionsMany sectors of the economy are directly under the purview of the states, in terms of policy or government expenditure or both. There are, however, still central government policies (e g, labour) that impede aggregate economic growth or poverty reduction. Given such growth constraining policies, each state has the option of adjusting its own rules and procedures to minimise the negative effects of these central policies as well as to improve the policies that come directly under its purview. States that have done so have been more successful in accelerating growth during the 1990s, while those that have not done so have seen little acceleration. Some states have even deteriorated because of worsening govern-ance and deteriorating investment climate in the state. This paper has concluded that the most critical areas distinguishing state growth performance have been modern (registered) manufactur-ing and commerce captured best by the GDP sector Trade, Hotels and Restaurants. To multiply the benefits of these two growth driv-ers, we must provide a positive policy environment for the growth of trade, hotels, restaurants, construction, real estate and town-ships. It must focus on urban/civic planning and connectivity. The paper, therefore, recommends that the poorer states’ expenditure allocation should put primary emphasis on roads and the country build an interconnecting road grid of a standard equal to that of the middle income countries.
SPECIAL ARTICLEjanuary 12, 2008 Economic & Political Weekly54Growth and Poverty: Policy Implications for Lagging StatesArvind VirmaniThis article shows that the interstate differences in poverty rates can be largely explained by differences in the per capita gross domestic product, agricultural growth and the share of the bottom 40 per cent of the population in consumption. To eliminate poverty, economic policy therefore has to accelerate growth, focus programmes on agriculture and rural development in the poorer states and target subsidies at the bottom 40 per cent. The most critical areas distinguishing state growth performance have been modern (registered) manufacturing and commerce.There is a view favoured by the anti-globalisation activists, left economists and the global/international socialists that faster growth in India has not reduced poverty. A sub-set of these personalities go so far as to assert that higher growth may even have caused or contributed to the widening of interstate gaps in income and poverty. On the periphery of this group are those who assert that growth is the least important issue among dozens that they can list. On the other extreme, are a small number who assert that the faster the growth, the better, and as long as growth is fast, there is not much else that the government needs to worry about. Perhaps, a careful exami-nation and analysis of the facts can help resolve some of these issues, even though it is unlikely to convince these extremes. This paper analyses the data on interstate variations in growth and poverty, to see what we can learn about economic growth and social welfare, with a view to improving planning and policy formulation.Virmani (2005, 2006a, 2006c) has presented an extensive analysis of aggregate growth from 1951-52 to 2004-05. In this analysis, central government policies relating to the external sector and monetary, fiscal and industrial policies that are the preserve of the central government are considered. Interstate dif-ferences in growth are, however, likely to arise either from the conditions in the states that affect the impact of central policies or from the differences in policies that are the preserve of the state governments. Since, a large sample of NSS data is available for 1993-94, 1999-2000 and 2004-05 and the state gross domes-tic product (SGDP) series in 1993-94, prices are available from 1993-94 to 2004-05, we focus on this period.Section 1 analyses the first aspect of welfare, per capitaGDP and its growth. Section 2 deals with the second important deter-minant of social welfare, poverty and analyses its links to growth and other aspects. Section 3 concludes the paper.1 GrowthAs shown in Virmani (2006a, 2006c) aggregate economic growth accelerated in the 1980s from an average of about 3.5 per cent per annum between 1951-52 and 1979-80 to about 5.8 per cent per annum during 1980-81 to 2004-05. More recent data shows that aggregate growth averaged about 5.5 per cent per annum during 1980-81 to 1994-95 and has accelerated further to an average of 6.8 per cent per annum during 1995-96 to 2006-07. Ahluwalia and Bajpai and Sachs have shown that the acceleration in growth was less in the poorer states, so that interstate inequality in per capita income has increased. Any views expressed in the paper are those of the author and do not necessarily reflect those of the organisation for which he works. The author thanks K L Datta for help in generating consumption data.Arvind Virmani(virmani@nic.in) is with the department of economic affairs, ministry of finance, government of India.
1993 94 2004 05 
SPECIAL ARTICLEjanuary 12, 2008 Economic & Political Weekly56Public and Quasi-Public GoodsIn our view, therefore, public and quasi-public goods that affect investment are likely to be more important determinants of interstate variation in growth than other infrastructure goods and services. This would be particularly so in the relatively poorer or backward states, which have not shared in the growth acceleration.The fifth hypothesis for interstate differ-ences in growth, pro-posed in this paper, is related to services as the drivers of growth since the 1980s. Much of the acceleration in aggregate growth has come from an acceler-ation of growth in the services sector. Cer-tain services have ac-celerated more than others in the aggre-gate. Bosworth et al (2007) have noted that though software and related services have received a lot of attention, the acceler-ation of the services sector’s growth, “has been more broadly based, including trade, transportation and community and personal services”. We hypothesise, that the eco-nomic growth in poorer states has not been accelerated because services that have propelled national growth have not accelerat-ed proportionately. This can help us identify the state policies that can help accelerate the growth of poorer states.5If we combine this with the previous hypotheses, then we would expect that services which are particularly dependent on, or, whose growth is associated with, the development of public and quasi-public goods are likely to prove important in explain-ing the interstate differences in growth. For instance, there are small but significant quasi-public goods associated with travel and tourism. These include historical monuments, cultural and religious sites, cultural (including religious) traditions, events, local art and crafts and natural attractions (such as water bodies, rivers, waterfalls; forest and animal reserves). It also includes prosaic public services items like clean drinking water, public toi-lets, lawns and flower beds at these sites. Preservation, enhance-ment and development of these quasi-public goods by the state or local governments will impact the economy through the sectors mentioned by Bosworth et al.1.2 Data and AnalysisWe use the national account data by state (and union territories (UTs)) and sector. The SGDP at constant 1993-94 prices is available for most states by sector. For a few states or UTs it is available only till 2003-04(1) or 2002-03(2).6 For each sector we calculate the compound annual rate of growth between 1993-94 and 2004-05 and use this as the basic data for analysis. In the case of some states and UTs data is missing for one or more sectors (e g, communications in Punjab, Sikkim and Jammu and Kashmir).Table 3 presents the summary statistics of economic growth across states by sector. Column 2 of the table shows that the two modern service sectors, communications and banking and insurance, are the fastest growing sectors across states. Further, communications is also considered an infrastructure, and in the days of landlines, it would have been considered on par with electricity as a candidate for characterisation as a quasi public good in rural areas. Note also that, there was little difference in the mean growth rate of the secondary and tertiary sectors. Columns 3 and 4 show the standard deviation and coefficient of variation of the growth rate across states for each sector. The electricity sector has the highest standard deviation of growth rate with a coefficient of variation of over 1. The primary sector and its four components have the highest coefficient of variation. Though one expects the variability in agriculture growth rates to be high because of rainfall variation, the high variability in mining and fishing is somewhat surprising! Growth and Total GDPFinally, column 5 of Table 3 shows the correlation of the growth of each sector with the total GDP of the state. At a broad level, the secondary sector is found to have a higher correlation (0.75) with total growth than the tertiary sector (0.7). If we look at individual sectors, “Trade, Hotels and Restaurants” is found to have the highest correlation of 0.67 with total SDP, followed by registered manufacturing with 0.5 and communication with 0.44. On the negative side, agricultural growth is almost uncorrelated with total SDP, while forestry and fishing have a negative correlation. The electricity sector has a modest correlation of 0.19.Table 3: Summary Statistics of State Growth (1993-94 to 2004-05)Sector Mean StdDv CV Correlation (%) (%)Agriculture 2.1 3.3 1.55-0.02Forestry 1.4 3.5 2.53 -0.39Fishing 3.3 6.0 1.84-0.45Mining 4.47.61.72-0.28Primary 1.37.35.82-0.24Manufacturing 5.4 4.1 0.750.41 Registered 5.8 4.3 0.74 0.50 Unregistered 4.7 3.8 0.82 0.21Construction 8.4 4.30.510.22Electricity 7.0 8.6 1.220.19Secondary 7.2 3.4 0.470.75Transport, communication, etc 10.9 3.3 0.30 0.56 Railways 5.9 3.5 0.59 -0.10 Other transport 7.3 3.2 0.44 0.05 Storage 4.5 4.1 0.90 -0.17 Communication 21.1 5.2 0.25 0.44Trade, hotels 6.3 3.3 0.53 0.67Bank, insurance 10.8 2.3 0.22 -0.06Rest, business service 5.4 2.4 0.44 0.36Govt administration 6.1 1.4 0.24 0.08Other services 6.7 2.3 0.34 0.25Tertiary 7.4 1.5 0.210.70SDP total 6.2 1.9 0.31 1.00ST stands for sub total.Table 4: Summary Table of Regression Coefficients(Dependent Variable is State GDP Growth)Variables Model 1a Model 1b Model 2a Model 2b Model 3a Model 3b Model 3b SURERate of growth of (cmpnd) Agriculture 0.42160.17860.089640.04030.0225 (3.2***) (1.8*) (1.0) (0.54) (0.42)Manufacturing 0.5473 0.1479 0.2258 0.1919 0.1911 regd (6.5***) (1.4) (3.6***) (2.6**) (3.6***)Electricity 0.19860.04160.0107-0.0205-0.0117 (3.6***) (1.1) (0.28)(-0.65) (-0.52)Communication 0.2195 0.1454 0.06450.0722 (5.2***) (5.1***) (1.6) (2.5**)Trade, hotel, 0.6031 0.3858 0.4361 0.3726 0.4329 restaurants (7.8***) (5.4***) (5.4***) (5.4***) (8.6***)Construction 0.23620.07520.20010.15330.0845 (4.2***)(1.5)(3.4***)(2.5**)(1.9*)F (Chi for SURE) 52 86 181 233 123 159 1267Prob > F (Chi) 0.0 0.0 0.0 0.0 0.0 0.0 0.0R2 0.860.940.930.970.960.980.98Adj R2 0.840.930.920.960.950.97RootMSE 0.026 0.018 0.018 0.013 0.014 0.014 0.010No of observations 29 26 29 26 29 26 261 Numbers in bracket are t (z) statistics, stars show confidence level * = 10%, ** = 5%, *** = 1%2 In the SURE regression growth rates of the variable “trade, hotels, restaurants”, communications and registered manufacturing were treated as dependent on total state GDP growth (uses z(not t) stat).
SPECIAL ARTICLEEconomic & Political Weekly january 12, 200857We do the model testing in two versions of the models (a and b), without and with communications as the latter is missing for three states/UTs. The results are summarised in Table 4 (p 56). We test the first three hypotheses by regressing the compound annual growth rate of state domestic product (GrSdp) between 1993-94 and 2004-05 on the growth rates of SDP from agriculture (GrSag), registered manufacturing (GrSmreg) and electricity (GrSelec). All three coefficients are significant at the 1 per cent level of con-fidence when communication variable is not included (Model 1a).7 The R2 is also a relatively high 0.86 with the adjusted R2 at 0.84. When the communication variable is introduced (Model 1b), it is highly significant, but both registered manufacturing and electri-city become non-significant, while agriculture is barely significant.Investment in roads, dams and canals are not captured sepa-rately in the national accounts, but are included in the construc-tion sector. However, the construction sector includes a much larger proportion of private construction. Because of their public good character, value added by roads and dams and canals is not measured separately, but is part of the value added by the users of these public goods, the road transport and agriculture sectors, respectively. In the case of roads, the externality is much broader in terms of general economic activity. For those who have tra-velled regularly down any highway over a number of years, will have noticed how economic activity, including shops and ‘dhabas’, spring up along newly built or improved/widened high-ways in three-five years. Such important externalities are proba-bly captured best by the sector, Trade Hotels and Restaurants. Besides roads, communication is also essential for growth of trade, which in turn, is necessary for the development of agri-culture, mining and manufacturing in rural areas. We, therefore, test the fourth and fifth hypotheses by regress-ingSDP growth on the growth of the sectors Trade, Hotels and Restaurants (GrTrHtlRes), construction (GrConst) without and with communications (GrCom).8 The results are shown under the columns Model 2a and 2b of the table. In the absence of commu-nication growth variable, both variables are highly significant (at 1 per cent level). The R2 and R2 (adj) are 0.94 and 0.93 respec-tively, both higher than those in Model 1a, indicating that Model 2a has higher explanatory power than Model 1a. When communi-cation variable is introduced into the Model (2b) it is highly sig-nificant but construction growth becomes non-significant. The R2 and R2 (adj) are 0.97 and 0.96 respectively, both higher than those in Model 1b, but now the gap is narrower.How do we choose between the different models? The fact that both sets of models give significant results suggests that they may be subject to missing variable bias. We, therefore, run the regres-sion by including all the variables identified. The results are given in columns marked Models 3a and 3b. Agriculture and electricity growth are not significant in either equation, indicating that they do not explain even a fraction of the difference in growth rates across states and UTs during the period following the “new eco-nomic policies” initiated in the 1990s. The only variable from the first three hypotheses that survives is the rate of growth of regis-tered manufacturing (at 5 per cent level), while both Trade, Ho-tels and Restaurants and constructions variables associated with the last two hypotheses are significant at 1 per cent level.9 We find that growth ofSGDP affects the variables, growth of communications and Trade, Hotels and Restaurants at a 1 per cent level and registered manufacturing at 10 per cent level. The coefficients of these variables could, therefore, be biased by the simultaneity problem. We, therefore, rerun the last equation (Model 3b) usingSURE (sum model of regression. This confirms that these three variables are significant, though the level of significance increases to 1 per cent for registered manufacturing and falls to 10 per cent for construction. The communication growth variable is also found to be highly significant (at 1 per cent level). The other two variables, growth of agriculture and electricity, remain non-significant. The numerical impact of the four sectors identified as drivers of state GDP(SGDP) growth is quite high. A 1 per cent dif-ference in the rate of growth of Trade, Hotels and Restaurants and 1 per cent in registered manufacturing can result in 0.43 per cent and 0.19 per cent difference, (respectively) in the rate of growth ofGDP. This impact is much larger than the average share of these sectors in all-India GDP of 0.145 per cent and 0.13 per cent respectively during 1993-94 to 2004-05. The impact of a 1 per cent difference in telecommunications and construction growth onSGDP growth is 0.07 per cent and 0.08 per cent respectively. This impact is also greater than their average share in all-India GDP of 0.025 per cent and 0.051 per cent, respectively. To illustrate the impact consider Uttar Pradesh (divided), which grew at 4.3 per cent per annum, 2.1 per cent points (or one-third) lower than the 26 state mean of 6.4 per cent per annum. If the rate of growth of trade et al, manufacturing (regd) and construction in Uttar Pradesh was raised to the mean rate of 6.9 per cent, 6.4 per cent and 8.7 per cent, respectively, Uttar Pradesh (divided)’s SGDPwould have grown at 6.4 per cent per annum. Assam and Bihar could correspondingly have grown at 5.4 per cent and 6 per cent per annum instead of 3.5 per cent and 4.5 per cent respectively.101.3 ImplicationsandRecommendationsVirmani (2005) has identified the “double dualistic” structure of the economy, predominantly in the manufacturing sector, as a drag on productivity growth.11 The current paper deepens our understanding of this phenomenon. Despite labour policy rigidi-ties and other policy distortions such as small scale industry(SSI) reservations and tax exemptions that have created this dualism, the registered manufacturing sector remains a driver of economic growth in India. Though the policy distortions have eliminated India’s competitive advantage in “labour-intensive mass manu-facturing”, vis-à-vis China and other labour-surplus countries with more flexible labour markets, this is not so for “semi-skilled labour-intensive goods”. On the contrary, (we hypothesise that) India’s entrepreneurs have responded to policy induced handicaps by;(a) Focusing on niche markets that do not require high volume production.(b) Outsourcing labour-intensive parts of the production process in-cluding service activities that were formerly part of “manufacturing”.(c) Fragmenting production into smaller units that are subject to less rigid labour rules and procedures.
SPECIAL ARTICLEjanuary 12, 2008 Economic & Political Weekly58(d) Using and/or developing technology and systems that substitute semi-skilled for unskilled labour as the labour rules are less rigid on the former.12(e) Using more capital-intensive methods and processes.Thus India has developed (in our view) a comparative advan-tage in ‘semi-skilled labour-intensive’ manufacturing that is visi-ble in automobile components and specialised chemicals, drugs and pharmaceuticals. Those states that have been able to apply labour apply rules and procedures flexibly, reduce bureaucratic red tape, curb the predation by government monopoly service providers and provide a more attractive investment environment, have more benefited. One must caution, however, that as long as labour rigidities remain, this sector cannot generate unskilled employment at a pace necessary to correct the distorted structure of employment (two-thirds of labour force in agriculture). Informal SectorThe major part of the dual economic structure is the informal, unorganised and predominantly services related sector. Most ob-servers have been too mesmerised by the information technology, IT-enabled services(ITES) and high tech services to notice that the rest of the services sector is also important. Bosworth, Collins and Virmani (2007) have showed that the fastest growth in total factor productivity has been in the services sector, and that too during the 1980s, whenIT andITES were not even a speck on the horizon. The current paper goes beyond earlier analyses to show that a humdrum sector like Trade, Hotels and Restaurants, has been a major driver of growth since 1993-94 (the market reform phase).13 If we examine these results and then a step beyond them an interesting narrative emerges of economic development through trade, commerce, tourism and urbanisation.14 With rigi-dities in labour policy and low literacy/primary education levels stymieing the growth of “labour-intensive mass manufacturing”, states have differentiated themselves in the 1990s by their rela-tive growth of trade, commerce, tourism, real estate development and construction. For rural areas (villages, habitation) to utilise their compara-tive advantage they have to be connected to markets and towns. Sometimes the comparative advantage has to be developed in the form of historical monuments, cultural/religious sites or natural resources (water bodies and courses). Roads and communica-tions are necessary conditions for this to happen. State policies, rules and regulations that facilitate this process, then lead to fast-er growth, through construction of shops, workshops, restau-rants, offices, hotels and housing both on the town roads extend-ing into the countryside as well as on roadside villages. The qual-ity range is wide and diverse from a dhaba to an air-cooled or air-conditioned restaurant to an eating place in an air- conditioned mall, as is the diversity in the education and skill levels of the employees needed to service these establishments.Roads: The importance of roads in economic development has gradually come back on the agenda over the last decade, after relative neglect for half a century. This is reflected in the central government programmes like the national highways develop-ment project (NHDP) and Prime Minister’s Grameen Rozgar Yojana (PMGRY). It is still not clear, however, whether state governments recognise their critical importance. I would recom-mend road development as the number one focus programme for all poor states and regions in the country. States must have a de-tailed operational plan for building a comprehensive road grid connecting every village and habitation within a time bound framework. At the high end, we would also find a mix of high quality offices and communication services coupled with busi-ness services (call centres).(a) All cities in the country must be connected by national high-ways that are constructed and maintained to middle income country standards.(b) All towns in every state must be connected to each other by state highways of a standard that can sustain traffic at least 30 kph during adverse conditions such as the monsoons. (c) Each village must be connected to the nearest town by an all weather metalled/surfaced district highway. Each village must be connected to all neighbouring village by a road of a standard suitable to the traffic.(d)All remaining habitations must be connected to the grid within 10 years.(e) New highways must identify and plan for the emergence of shopping and trading areas near villages and highway/road in-tersections. This will entail development of small stretches of parallel local roads, underpasses for rural traffic and designa-tion/acquisition of contiguous areas on the non-highways side of the parallel roads for commercial activity.(f) State road transport, goods and passenger, must be de-licensed to build rural and local entrepreneurship in road trans-port. At most states could have a system of registration along with registration fees. Registration should be automatic unless there is a pattern of systemic violation of road rule. Registration fees should be automatically transferred to a road maintenance organisation at the local or state level charged with maintaining district/state highways.(g) All octroi posts and other barriers to intra-state traffic must be dismantled. Communications: The communication revolution is already sweeping the country. Its impact on the rural areas of the country can be expedited if some anomalies in policy and regulation are corrected.(h) The universal services obligation (USO) fund must be used to expedite extension of telephony and internet into the rural areas. It should, for instance, hold subsidy auctions for extending the mobile footprint into areas of non-presence.(i) Optical fibre cables, landlines and all other wires going into the rural areas must be subject to open access with the Telecom Regulatory Authority of India (TARI) ensuring that reasonable rates are charged for access. The department of telecommunica-tions (DOT) and the telecom commission should stop siding with Bharat Sanchar Nigam(BSNL) if they are genuinely interested in extending communications to rural areas.(j) TRAI had previously recommended that all physical mobile infrastructure in rural areas should be sharable. We should go further and completely unbundle the local loop in rural areas so 
SPECIAL ARTICLEEconomic & Political Weekly january 12, 200859that they can get the benefits of competition that urban areas are already benefiting from.Let a 100 Towns Bloom: Allow and encourage private entre-preneurs to build thousands of new towns/townships in semi-urban, semi-rural areas. Government has the responsibility of connecting these to the nearest highway and water-supply mains. Planning water reservoirs, solid waste disposal and sewage treatment is also stategovernments’ responsibility. States must scrap forth-with expropri-atory sections of the rent control act(s) and corres-ponding rules and proce-dures, so that the private sector can build and provide rental accommodation for the lower middle class and the poor.2 PovertyThe question of whether growth is a necessary or a sufficient condition for the elimination of poverty has been long debated. The focus of our investigation is precise: to what extent do in-terstate differences in average per capita income of states (as measuredby the National Accounts Statistics(NAS)) explain in-terstate differences in poverty rates as measured by the official poverty data?2.1 All India PovertyThe overall poverty picture at the national level is presented in Tables 4, 5 and 6. Over the 11-year period from 1993-94 to 2004-05, the proportion of poor below the poverty line and the absolute number of poor have declined. The poverty ratio has declined by 23 per cent from 36 per cent of population to 27.5 per cent while the number of poor has declined by 6 per cent to 301.7 million (Table 5). The other noteworthy feature is the convergence of rural and urban poverty rates from a gap of 4.9 per cent points in 1993-94 to 0.8 per cent points in 2004-05. This suggests that the rural and urban areas are getting better integrated in terms of movement of workers, goods and services and the price differentials that drive them.Some commentators have speculated that the rate of poverty reduction has slowed because poverty de-clined by about 8.5 per cent points during the 10-year period from 1983-84 and 1993-94 by about 8.4 per cent points in the 11-year period from 1993-94 to 2004-05. Despite the rise in growth rate of GDP. The decline was however 19 per cent in the former and 23 per cent in the latter, indicating that the rate of decline was virtually unchanged between these two periods. A similar picture emerges if we look at the 1990s (Table 6).2.2 DistributionFunctionGiven an income or consumption distribution f(y), and a poverty line Yp the proportion of people below the poverty line (poverty rate or headcount ratio) is given by, F(Yp), where F is the cumula-tive distribution. The headcount ratio or proportion of people below the poverty line is therefore,F(Yp) = F (Y; Ya, Ω )| Y =Yp , ...(1)where Ya = the average or mean of the income or consum-ption distribution and Ω is a distributional parameter(s) like variance. If the distribution for each state is assumed to be drawn from the same family of distributions and differ from each other in terms of the mean and distribution parameter(s) then we can differentiate this distribution around the poverty line Yp (which is the same in real value across Indian states) to obtain,d F(Yp) = F1 (Yp; Ya, Ω ) dYa + F2 (Yp; Ya, Ω ) d Ω | at Y = Yp ...(2)where F1 and F2 are respectively the differentials of the cumula-tive distribution with respect to mean income or consumption and the distributional parameter, evaluated at Y =Yp. If we divide both sides of equation (2) by F and rearrange we obtain an elas-ticity form of the equation.d F/F = (Ya F1/F) (dYa/Ya) + (Ω F2 /F) (dΩ / Ω) | at Y = Yp ...(3)2.3 EmpiricalEstimationThe estimating equation based on equation (3) can be written as,GrPovertyi = A GrPcSgdpi + B GrShrL40Ri + C GrShrL40Ui + ¤i ...(4)with i = 1…n are the states, A, B and C are parameters to be esti-mated and ¤ is the error term. GrXi denotes compound annual rate of change of the variable X in state i between 1993-94 and 2004-05, poverty is the poverty rate or headcount ratio for the state, PcSgdp is the per capita state GDP, ShrL40R is the share of the lower 40 per cent of the state’s rural population in rural con-sumption and ShrL40U is the share of the lower 40 per cent of the state’s urban population in urban consumption.For the poverty rates (dependent variable) we use the official poverty estimates for 2004-05 and 1993-99 estimated from the uniform recall period(URP) data for each state. We can also obtain the MRP estimates for 1993-94 and use these along with official MRP based estimates for 2004-05. As we estimate the equations in difference form using state panels, the new states of Jharkhand, Chhattisgarh and Uttrakhand have to be left out of the estimation.For the independent variables, we use the national account data for per capita state GDP. It is, however, also possible to use the average monthly per capita expenditure (MPCE) calculated from the survey data, by converting 1993-94 data to 2004-05 prices using the same deflators that are used to Table 5: National Poverty: Head Count Ratio 1993-94/1999-20002004-05 URPMRPURPMRPRural Poverty ratio 37.3 27.1 28.3 21.8 Number of poor (million) 244.0 193.2 220.9 170.3Urban Poverty ratio 32.4 23.6 25.7 21.7 Number of poor (million) 76.3 67.0 80.8 68.2Total Poverty ratio 36.0 26.1 27.5 21.8 Number of poor (million) 320.4 260.3 301.7 238.5URP = uniform recall period, MRP = mixed recall period.Table 6: Change in Poverty Rate between 1993-94 and 2004-05 (% Point Per Year) (Compound Annual) 1993-94 to 1990-2000 1993-94 to 1999-2000 2004-05 to2004-05 2004-05 to2004-05 URP MRP URP MRP1 Rural 0.78 1.06 2.48 4.262 Urban 0.60 0.38 2.07 1.683 Total 0.77 0.87 2.40 3.56Note: As in Table 5.
SPECIAL ARTICLEjanuary 12, 2008 Economic & Political Weekly60derive the poverty line in current prices. For the distribution vari-able, we use the share of bottom 40 per cent of the population in rural and urban areas. One of the reasons for using this variable to represent distributional factors is that in cross country regres-sions this variable has proved significant in explaining differenc-es in poverty [Virmani 2006a, 2006d]. Alternative variables such as the share of the bottom 20 per cent, etc, turn out to be non-significant in the cross-country context.A number of proxies have been used in the literature to capture distributional parameters. Bhalla (2002) has the most direct measure, which he calls “shape of distribution elasticity” (SDE). Ahluwalia (1978), Bell et al (1994), Ravallion and Datt (1996) and Datt and Ravallion (1998), have shown that agricultural output or productivity is linked to poverty.15 If agricultural and non- agricultural growth have significantly different impact on overall (combined rural and urban) poverty, it would imply that the pattern of growth can have distributional consequences. This along with the fact that a large proportion of the poor live in rural areas and are directly or indirectly linked to agricul-ture, implies rural-urban seg-mentation of markets for fac-tors (labour, capital) and goods. The labour market segmenta-tion could be due to lack of in-formation about non-agricul-tural/urban jobs, transaction costs of moving to the urban area, lack of mobility for social and other reasons or lack of re-quired education/skills. Goods market segmentation could be partly due to poor transport and communication links, while the capital/credit market segmentation is generally re-lated to information problems and rule of law (contractual arrangements, enforcement).The issue of the special link between agriculture growth and poverty suggested by the above research, can be reframed in the present context as follows: whether agriculture growth contrib-utes something additional to poverty reduction that is not cap-tured by growth in average per capita income/consumption and the distributional parameter used. We test this hypothesis by in-troducing the compound annual growth rate of state GDP from agricultural between 1993-94 to 2004-05 (GrSgAg) as an addi-tional variable in equation (3) to obtain an alternative estimation equation,GrPovertyi = A GrPcSgdpi + B GrShrL40Ri + C GrShrL40Ui + D GrPcSgAgi + ¤i ...(5) The least squares estimates of equations (4) and (5) are sum-marised in Table 7. The most important result is that the per capi-ta state domestic product is significant at the 1 per cent level of confidence in all these estimates. There is thus no factual basis to the ideological position that income growth has had a neutral or perverse effect on interstate poverty gaps. Estimating equation (4) shows that every 1 per cent increase in the per capita state GDP results in a 1 per cent reduction in the poverty rate (columns 2 and 6 of Table 7). The second result is that only the rural distri-bution is statistically significant, with a 1 per cent improvement in the distribution leading to a 0.8 per cent reduction in the pov-erty rate. This simple regression explains between 63 per cent and 67 per cent of the URP-based poverty. The results for the MRP-based poverty are similar.16 If we estimate the model with agriculture growth, this variable is highly significant and modifies the impact of the other varia-bles (column 3). Every 1 per cent increase in agriculture growth reduces the rate of URP poverty by 0.45 per cent, in addition to its effect on average per capitaGDP. This model, therefore, has about 9 per cent more explanatory power for the URP poverty. The in-troduction of agriculture growth also reduces the impact of per capita income by 14 per cent and of consumption shares of lower 40 per cent of rural population by about 25 per cent. Third, the consumption share of the lower 40 per cent of urban population (-0.83) now becomes signifi-cant and higher than that of rural shares (-0.62). Lastly, as a consequence, the model now explains a higher 74 per cent and 81 per cent of the overall difference in reduction in pov-erty rates across states. The fourth column shows the regression results of the same model if the growth of SGDP from agriculture is re-placed by per capitaSGDP from agriculture. The results are substantially the same except that the effect of per capitaGDP growth reverts to 1 and the ef-fect of rural and urban shares is now almost identical.2.4 PolicyImplicationsThus we find that though differences in agriculture growth across states provide no explanation for differences inSGDP growth, they are important in explaining differences in total poverty reduction. Agriculture – Public Goods: There are two approaches to agri-culture growth. One based on the detailed and comprehensive analysis of all the problems that have arisen in the agriculture sector/rural areas across this vast country and lists every policy, institutional and programme change needed for agriculture to thrive. I call this the symphonic approach where all the instru-ments play in harmony to produce good results. The other ap-proach, which could be called the quartet, identifies a few critical actions that government must take which are particularly impor-tant for agriculture/rural development. There are three public goods and one quasi-public good which may be particularly Table 7: Estimated Coefficients for Poverty EquationsDependent Variables = Poverty Poverty Poverty Poverty Poverty Poverty Independent Variables URP URP URP MRP MRP MRP 1 2 3 4 5 6 7Per capita state GDP -1.025 -0.863 -1.06 -0.982 -0.922 -0.902Per capita state GDP non-agri -7.0*** -6.3*** -7.8*** -6.8*** -6.6*** -7.3***Cons share of lower 40%: rural -0.821 -0.621 -0.736 -0.656 -0.583 -0.597 -2.1**-1.8*-2.0*-1.8*-1.7-1.8*Cons share of lower 40%: urban -0.655 -0.832 -0.793 0.132 -0.147 -1.6-2.3**-2.1**0.3-0.3State GDP from agriculture -0.454 -0.291 -0.275 -3.1***-1.9*-2.0*State per capita GDP from agriculture -0.447 -2.3**F 171916232121Prob > F 0.0 0.0 0.0 0.0 0.0 0.0R2 0.670.760.730.770.810.81Adj R2 0.63 0.72 0.68 0.74 0.77 0.77Root MSE 0.031 0.027 0.029 0.031 0.027 0.027No of observations 28 28 28 24 24 24
SPECIAL ARTICLEEconomic & Political Weekly january 12, 200861lacking in poor, backward regions, namely, the rule of law, per-manent all weather roads, knowledge/information and assured irrigation. Many of the poorest regions are still characterised by semi-feudal relations in land, labour and credit markets. In the absence of rule of law, there is nothing to keep traditional feudal families from transforming into tin pot oligarchs who supply spurious fer-tiliser, pesticides, etc, and use strong arm methods to collect overdue loans or extract indentured labour. Historically, the continental interior and geographically re-mote and hilly areas are least likely to have been connected to the main transport corridors, and therefore, the gaps are likely to be widest. Research and development (R&D) on crop and non-crop agri-culture and animal husbandry, including new varieties, opera-tional methods and management practices has traditionally been generated by government universities and transmitted to farm-ers by public organisations. The deterioration in volume and quality of this knowledge transfer must be reversed. The syner-gies between telecom connectivity, internet access, e-governance, e-learning and e-marketing must be exploited. Water Management: Water is essential for drinking, personal hygiene, sanitation and irrigation. From a global comparative perspective, India is a relatively water-scarce country and global environmental changes threaten to make this worse. Yet our limited water resources are either not fully utilised (flow to the sea) or are misused (depleting groundwater). Public water supply systems also need to make better use of rain water.There is an urgent need to improve the comprehensiveness and quality of water planning and management at every level (centre, state, district, town, panchayat and smallest farmer). Water harvesting, water shed development, recharge of water bodies and aquifers, must be planned and implemented in every nook and corner of the country. Education and demonstration of models with the active participation of NGOs can play an impor-tant role. Dams and canals have a place in cutting down the flow of water into the sea, recharging aquifers and supplying dry areas and parched towns. Tube wells in depleting aquifers must be discouraged through proper pricing of electricity and perhaps even water.Primary Education: Literacy can help in acquiring knowledge about job opportunities, tools and productivity. Government must ensure that every member of the labour force, every citizen, has the education that (s)he is supposed to acquire with the com-pletion of primary education. But this education must also be made more relevant by providing information on agriculture and allied subjects and training them on how to access relevant infor-mation in future. We should not declare premature victory for primary educa-tion and move on to higher levels and once again deprive the poor of their access to basics while satisfying the middle classes hun-ger for secondary education. The latter can be better achieved through a modern, transparent, regulatory system that minimises the problem of asymmetric information [Virmani 2005, 2006 b], fosters competition in supply17 and empowers them to get value for the 3 per cent of GDP that they already spent on private education.Every youth, rural or urban, after completing primary educa-tion must also have access to the 6,000 or so globally identified skills. This requires a massive joint effort by government, NGOs and private skill providers. Government must provide funding for the poor while all possible private and foreign expertise and experience is attracted to India to provide training in all these skills in the next five years.This section confirms that average per capitaGDP is an impor-tant determinant of poverty. It also shows that the higher agricul-ture growth has an impact on poverty reduction in addition to its normal contribution to overall GDP growth. A special focus on agricultural growth in poorer states and in the states with oppor-tunities for productivity improvement can therefore be justified in terms of poverty removal even though it may not have any im-pact on overall growth. The empirical results also justify an add-ed focus on rural roads and telecom connectivity (in addition to the general effects found earlier) to the extent that they promote the development of agriculture. Development of rural connectiv-ity also improves market integration and labour mobility, which in turn, will remove the differential and segmented impact of growth on rural and urban poverty.This section also shows that the consumption share of bottom 40 per cent of the population is an important determinant of pov-erty. Targeted benefit programmes should therefore focus on the bottom 40 per cent of the population. It is essential to set up a comprehensive data base with unique identifications – photo-graphs and bio-metric identifications that will eliminate fraud and help identify the poorest 30 per cent to 40 per cent of the population.18 3 ConclusionsMany sectors of the economy are directly under the purview of the states, in terms of policy or government expenditure or both. There are, however, still central government policies (e g, labour) that impede aggregate economic growth or poverty reduction. Given such growth constraining policies, each state has the option of adjusting its own rules and procedures to minimise the negative effects of these central policies as well as to improve the policies that come directly under its purview. States that have done so have been more successful in accelerating growth during the 1990s, while those that have not done so have seen little acceleration. Some states have even deteriorated because of worsening govern-ance and deteriorating investment climate in the state. This paper has concluded that the most critical areas distinguishing state growth performance have been modern (registered) manufactur-ing and commerce captured best by the GDP sector Trade, Hotels and Restaurants. To multiply the benefits of these two growth driv-ers, we must provide a positive policy environment for the growth of trade, hotels, restaurants, construction, real estate and town-ships. It must focus on urban/civic planning and connectivity. The paper, therefore, recommends that the poorer states’ expenditure allocation should put primary emphasis on roads and the country build an interconnecting road grid of a standard equal to that of the middle income countries.
