## _wp0854

## Source details

**Canonical URL:** [_wp0854](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2008/_wp0854.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2008/_wp0854.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2008/_wp0854.pdf.json)

---

### Introduction and scope
- India experienced "For the past four fiscal years, real GDP has risen at over 8.5 percent per year."
- Central questions:
  - Did high aggregate growth benefit the bottom of the income distribution over the last two decades?
  - How did growth and changes in inequality contribute to poverty reduction?
  - Did the inclusiveness of growth vary across Indian states and relate to state characteristics (financial development, education, labor policies)?
- Data and measurement context:
  - NSS quinquennial rounds used: 1983 (38th), 1987/88 (43th), 1993/94 (50th), 1999/00 (not relied upon for key comparisons), and 2004/05 (61st).
  - Official deflators: CPIAL for rural sector and CPIIW for urban sector.
  - Consumption captured in NSS: "62 percent and 41 percent of the private consumption reflected in the NAS in 1993/94 and 2004/05 respectively."

### Aggregate growth versus household survey growth
- Growth rates:
  - "The annual growth rate of real GDP per capita accelerated from 3 percent in 1983–1993/94 to an average of 4½ percent in the post reform period 1993/94–2004/05."
- Divergence between NAS and NSS:
  - Over 1983 to 1993/94, average annual growth of per capita expenditure in NSS was "only half as large as the rate measured in the NAS."
  - In the 1990s NSS per-capita growth "shrunk to 40 percent of NAS growth rate."
- Explanations reviewed:
  - Survey underestimation of consumption (Bhalla, 2003).
  - Concentration of gains among the very rich underrepresented in surveys:
    - "Banerjee and Piketty (2005) find that growth of the top 1 percent of incomes can explain 20–40 percent of the gap between average consumption growth in survey-based and NAS data."
    - Ravallion (2001) notes "income effect on underreporting and noncompliance"; NSS–NAS gap largest for commodities consumed more by higher income groups.

### Poverty trends and incidence (1983–2004/05)
- Poverty line definition:
  - Minimum subsistence income supporting consumption of 2400 calories in rural areas and 2100 calories in urban areas; poverty line first established in 1979 and periodically adjusted for inflation.
- Key poverty statistics (Table 2):
  - Poverty Rate (share below poverty line)
    - 1983/84: All India 45.2%, Rural 46.2%, Urban 42.1%
    - 1987/88: All India 39.3%, Rural 39.3%, Urban 39.2%
    - 1993/94: All India 35.8%, Rural 36.8%, Urban 32.8%
    - 2004/05: All India 27.5%, Rural 28.0%, Urban 25.8%
  - Poverty Depth (poverty rate × average shortfall)
    - 1983/84: All India 12.6%, Rural 13.0%, Urban 11.5%
    - 1987/88: All India 9.6%, Rural 9.4%, Urban 10.4%
    - 1993/94: All India 8.4%, Rural 8.4%, Urban 8.3%
    - 2004/05: All India 5.7%, Rural 5.5%, Urban 6.2%
- Aggregate changes highlighted:
  - "Over the last two decades, the incidence of poverty fell by nearly 20 percentage points."
  - As of 2004/05: "25¾ percent of people in urban areas and 28 percent of people in rural areas lived below the poverty line."
  - State extremes: Orissa >45 percent below poverty line in 2004/05; Punjab "just 8 percent in 2004/05."
- Alternative consumption measure:
  - Places India’s poverty rate at 21 percent in 2004/05 down from 26.1 percent in 1999/00 (comparable earlier-year estimates not available).

### Inequality and wealth concentration
- Inequality indicators (selected entries from Table 3):
  - Variance of Log Consumption / Gini / Theil Index / Log (PCE 95 / PCE 5) reported for 1982/83, 1987/88, 1993/94, 2004/05 (values preserved in sequence as presented).
    - 1982/83: 0.3190 / 0.3120 / 0.3400 / 0.1980 / 0.1910 / 0.215 / 1.774 / 1.740 / 1.866 / 0.303 / 0.290 / 0.342
    - 1987/88: 0.3130 / 0.3010 / 0.3490 / 0.2000 / 0.1860 / 0.241 / 1.688 / 1.621 / 1.887 / 0.279 / 0.258 / 0.348
    - 1993/94: 0.3030 / 0.2850 / 0.3430 / 0.1910 / 0.1710 / 0.235 / 1.638 / 1.537 / 1.878 / 0.263 / 0.234 / 0.345
    - 2004/05: 0.3250 / 0.2980 / 0.3780 / 0.2280 / 0.1960 / 0.288 / 1.692 / 1.541 / 2.029 / 0.285 / 0.240 / 0.402
  - Trend: "significant increase in overall inequality in the 1990s, particularly in urban areas."
- Wealth concentration and ownership:
  - Ahya and Sheth (2007) estimate "increase in wealth of over 100 percent of GDP in the past four years" from equity market, residential property, and gold.
  - Participation/ownership concentrations:
    - "4–7 percent of the population participating in the stock market."
    - "47 percent of the population owning a ‘pucca’ house."
    - "The top 34 percent of households holding 71 percent of the value of consumer durables (including gold and jewelry)."
  - Jayadev et al. (2007) find a "small but perceptible rise in interpersonal wealth inequality in the 1990s" with representativeness caveats.

### Decomposing poverty reduction: growth versus distribution
- Decomposition method:
  - Poverty rate Pt = P(mt, lt) where mt is mean consumption and lt is Lorenz curve; decomposition follows Dhongde (2007) averaging two growth components and two distribution components to obtain a path-independent, complete decomposition.
- Counterfactual simulation results:
  - 1983–1993/94:
    - Distributional changes enhanced the effect of growth on poverty reduction overall.
    - Rural: poverty reduction from "growth alone" would have been 27 percent lower had distribution not changed in favor of the poor.
    - Urban: "growth alone" accounts for entire poverty decline.
  - 1993/94–2004/05:
    - Changes in the distribution of consumption moderated the extent to which growth reduced poverty.
    - Distribution-neutral growth would have generated poverty decline that was 22 percent higher in rural India.
    - Distribution-neutral growth would have generated poverty decline that was 76 percent higher in urban areas.
    - Indicates a marked shift in how gains from growth were distributed under the newer market-oriented framework.
- Selected numerical entries from Table 4 (preserved exactly):
  - 1983–1993/94:
    - Rural: Initial Level of Poverty 0.4617; Change in Poverty -0.0933; Contribution of Growth -0.0683; Change in Distribution -0.0249
    - Urban: Initial Level of Poverty 0.4208; Change in Poverty -0.0925; Contribution of Growth -0.0973; Change in Distribution 0.0047
    - All India: Initial Level of Poverty 0.4524; Change in Poverty -0.0940; Contribution of Growth -0.0808; Change in Distribution -0.0132
  - 1993/94–2004/05:
    - Rural: Initial Level of Poverty 0.3684; Change in Poverty -0.0880; Contribution of Growth -0.1071; Change in Distribution 0.0191
    - Urban: Initial Level of Poverty 0.3283; Change in Poverty -0.0702; Contribution of Growth -0.1237; Change in Distribution 0.0536
    - All India: Initial Level of Poverty 0.3585; Change in Poverty -0.0837; Contribution of Growth -0.1151; Change in Distribution 0.0314

### Patterns of consumption growth and inclusiveness
- Growth incidence curves:
  - 1983–1993/94: Growth at the bottom outpaced growth at the top, especially in rural India; urban growth distribution-neutral.
  - 1993/94–2004/05: Reversal — far faster growth at the top than the bottom, stronger pro-rich bias in urban areas.
  - "In almost all states growth became less equalizing in the 1990s."
- Quantitative highlights:
  - 1993/94–2004/05:
    - Consumption of the richest grew by an average of 3 percent every year.
    - Consumption growth for the poorest was slightly above 1 percent every year.
  - Over an 11 year period in NSS: consumption of the richest grew by 40 percent.
  - Banerjee and Piketty (1987/88–1999/00 income-tax data): incomes of the top 1 percent increased 70 percent; incomes of the top 0.01 percent tripled.
  - Absolute change for the poorest: consumption of the poorest increased by 8 percent.

### Real wages and labor patterns
- Wage construction:
  - Daily real wages from NSS Schedule 10 for 15–65 year olds engaged in regular salaried or casual wage labor; pooled rural and urban; only non-zero wages used; deflated as monthly per capita consumption.
- Wage patterns:
  - 1983–1993/94: real wage growth at the bottom generally outpaced growth at the top.
  - 1993/94–2004/05: real wage growth more pronounced at the top.
- Caveat:
  - Wage analysis overlooks changes in labor participation and unemployment; focus remains on per capita consumption growth.

### Heterogeneity across household characteristics
- Education:
  - 1980s: consumption growth relatively balanced; with exception of college graduates, growth rates declined with education level.
  - 1990s: households with completed middle school and above rose substantially more; fastest growth for college education and above.
- Sector of employment:
  - 1990s: manufacturing and services households experienced much larger consumption increases relative to agriculture.
- Caste:
  - 1980s: consumption growth similar across SC/ST and other castes.
  - 1990s: divergence with non-SC/ST experiencing faster consumption growth.

### Defining inclusiveness and cross-state empirical approach
- Inclusiveness metric:
  - Defined as the difference between the consumption growth rate of the poorest and richest 30 percent of the population; robustness checks use bottom/top 10 percent, 20 percent, and below/above median.
- Cross-state analysis:
  - 15 large Indian states (95 percent of India’s population in 2004) across three periods: 1983–1987/88, 1987/88–1993/94, 1993/94–2004/05.
- Panel regression setup:
  - Dependent variable y_{t,t-1,s}: inclusiveness (bottom 30 minus top 30 consumption growth).
  - Policy variables X_{t-1,s}: log total real credit per capita; percent of agricultural laborers with bank loan; labor regulation index; percent with secondary education and above; infrastructure index; log socio-economic expenditure per capita.
  - Controls Z_{t-1,s}: log real NSDP per capita; percent of workforce in agriculture.
  - Period and state fixed effects; regressions weighted by square root of number of observations within a state; N around 43 in specifications.

### Sectoral composition and inclusiveness (correlations)
- Overall NSDP growth:
  - Coefficients on Real Per capita NSDP growth close to zero (examples: -0.016, 0.005, 0.001, 0.001 with reported standard errors).
- Sectoral correlations (selected coefficients preserved):
  - Real Per capita Agriculture growth: 0.029, 0.032, 0.021, 0.013 ([0.072][0.051][0.040][0.026]) — no strong positive correlation with inclusiveness.
  - Real Per capita Industry growth: 0.139, 0.145, 0.142*, 0.105** ([0.109][0.087][0.071][0.049]) — higher industry/secondary growth correlated with larger pro-poor bias for broader inclusiveness definitions.
  - Real Per capita Services growth: -0.591***, -0.496***, -0.421***, -0.292*** ([0.182][0.143][0.121][0.083]) — faster service growth associated with larger gap favoring the rich.
- Contextual fact: service sector share increased from 40 percent in 1983/84 to 54 percent in 2004/05, with expansion concentrated in banking, insurance and communications.

### Policy correlations with inclusiveness (key regression findings)
- Financial development:
  - Log Real Credit per Capita: 3.780**, 5.429***, 7.073**, 6.152**, 3.865*** ([1.801][1.837][2.845][2.290][1.264]) — higher real credit per capita associated with more pro-poor growth.
  - Percent of Laborers with Bank Loan: 0.190** ([0.090]) — larger initial share of agricultural laborers with formal loans associated with more pro-poor subsequent growth.
- Labor regulation:
  - Labor Laws coefficients: -0.682, -1.138**, -1.347, -1.287*, -0.807** ([0.674][0.500][0.803][0.637][0.342]) — amendments toward employer flexibility associated with more pro-poor consumption growth conditional on other controls.
- Education:
  - Percent Primary Graduates and Above: 0.058** ([0.024]).
  - Percent Secondary Graduates and Above: 0.253*, 0.235*, 0.445**, 0.321**, 0.139* ([0.137][0.113][0.177][0.140][0.074]) — higher secondary/higher education associated with more inclusive growth.
- Infrastructure:
  - Infrastructure Index: 2.990*, 2.502**, 4.069**, 2.840*, 1.654** ([1.540][0.915][1.482][1.418][0.624]) — better infrastructure associated with more inclusive growth.
- State socio-economic expenditure:
  - Log Socioec. Exp. per Capita: coefficients include 2.1, -1.506, -1.92, -1.78, -1.02 ([1.718][1.428][2.029][1.707][1.003]) — no consistent statistically significant correlation; possible reverse causality, targeting/leakage, or administrative issues.
- Controls and sample notes:
  - Log Real NSDP per Capita and Percent Farmers coefficients vary and are generally imprecise across specifications.
  - Number of observations: N 43 (varies across regressions); regressions include state and period fixed effects and are weighted by square root of number of observations within a state.
- Interpretation caution:
  - Results are conditional correlations without exogenous variation; causal inference limited and potential unobserved time-varying state factors may confound associations.

### Summary conclusions and policy implications
- Structural change and distribution:
  - 1990s structural shift toward services coincided with less inclusive consumption growth; top incomes captured larger gains relative to the 1980s.
- Indicative policy associations:
  - Promoting financial development (greater real credit per capita; higher share of agricultural laborers with bank loans) is associated with more pro-poor growth.
  - Expanding secondary and higher education correlates with more inclusive growth.
  - Greater labor market flexibility (amendments reducing rigid pro-worker regulations) correlates with a more pro-poor pattern, potentially via facilitating industry/secondary sector expansion.
  - Improved infrastructure correlates with more inclusive growth.
  - State socio-economic spending does not show a robust positive correlation with inclusiveness in these specifications (potential reverse causality, targeting/leakage, or implementation gaps).
- Policy focus recommendation (based on empirical associations):
  - Increase access to formal credit, expand secondary and higher education, improve infrastructure, and consider labor market reforms that facilitate industrial expansion — these are associated with stronger ability of the poor to gain from growth.
- Broader reflection:
  - Distinguish "bad inequalities" that block market participation and investment in human/physical capital from "good inequalities" that reflect incentives for education and investment; policy should aim to reduce bad inequalities by improving access to schooling, social services, credit, and removing exclusionary barriers.

### Data sources (Appendix I)
- State income: Economic and Political Weekly States database and Central Statistics Office (National state domestic product by agriculture, industry, services).
- Household surveys: NSS Schedule 1 and Schedule 10 for 1983, 1987/88, 1993/94, 2004/05; CPIAL and CPIIW used for deflation; Planning Commission poverty lines for poverty measures.
- Population: World Bank’s Fiscal Database; Central Statistical Organization of India primary source.
- Credit: RBI Basic Statistical Tables, “State-Wise Classification of Outstanding Credit of Scheduled Commercial Banks according to Place of Sanction and Utilization.”
- Labor regulation: Besley and Burgess (2004) updated with Malik (2002).
- Education: 1981, 1991, 2001 Census of India; linear interpolation for annual series.
- Infrastructure: principal component of installed electricity per capita, surfaced roads per state area, share of households with access to drinking water; sources include Purfield (2007), Basic Road Statistics of India, Economic Survey of India 2005/06 Table 95.
- State social spending: World Bank’s States Fiscal Database; Reserve Bank of India’s Annual Report on State Finances.

*Source: _wp0854 - References..............................................................................................................*

### References..............................................................................................................

### _wp0854 - References..............................................................................................................

### Introduction and scope
- India experienced a period of unprecedented growth: "For the past four fiscal years, real GDP has risen at over 8.5 percent per year."
- Central questions addressed:
  - Did high aggregate growth benefit the bottom of the income distribution over the last two decades?
  - How did growth and changes in inequality contribute to poverty reduction?
  - Did the inclusiveness of growth vary across Indian states and relate to state characteristics (financial development, education, labor policies)?
- Data context:
  - National Sample Survey (NSS) quinquennial rounds used: 1983, 1987/88, 1993/94, 1999/00, and 2004/05.
  - 1999/00 NSS round has measurement issues; analysis relies on rounds with consistent methodology (1983, 1987/88, 1993/94, 2004/05).
  - Official state-level deflators used: Consumer Price Index for Agricultural Labor (CPIAL) for rural sector and Consumer Price Index for Industrial Workers (CPIIW) for urban sector.
  - Consumption reported in the NSS captured "62 percent and 41 percent of the private consumption reflected in the NAS in 1993/94 and 2004/05 respectively."

### Key findings on growth, poverty, and consumption measurement
- Aggregate growth vs household survey growth:
  - "The annual growth rate of real GDP per capita accelerated from 3 percent in 1983–1993/94 to an average of 4½ percent in the post reform period 1993/94–2004/05."
  - Household survey (NSS) evidence showed substantially slower improvement in monthly expenditures per capita compared with National Account Statistics (NAS).
  - Over 1983 to 1993/94, the average annual growth rate of per capita expenditure in NSS was "only half as large as the rate measured in the NAS."
  - In the 1990s NSS per-capita growth "shrunk to 40 percent of NAS growth rate."
- Explanations for divergence between NAS and NSS:
  - One view: survey underestimates consumption of all respondents (Bhalla, 2003).
  - Alternative: a very large share of growth went to the very rich, who are underrepresented in expenditure surveys. Evidence:
    - Using income tax data, "Banerjee and Piketty (2005) find that growth of the top 1 percent of incomes can explain 20–40 percent of the gap between average consumption growth in survey-based and NAS data."
    - Ravallion (2001) notes an "income effect on underreporting and noncompliance" and that the NSS-NAS gap is most pronounced for commodities consumed more heavily by higher income groups (Sundaram and Tendulkar, 2001).

### Poverty and inequality trends (1983–2004/05)
- Poverty incidence and depth:
  - "Over the last two decades, the incidence of poverty fell by nearly 20 percentage points."
  - As of 2004/05: "25¾ percent of people in urban areas and 28 percent of people in rural areas lived below the poverty line."
  - "Poverty depth decreased by more than 50 percent during this time period."
  - State heterogeneity: Orissa remained the poorest state from 1983 to 2004/05 (more than 45 percent below the poverty line in 2004/05); Punjab remained the least poor with a poverty rate of "just 8 percent in 2004/05."
- Inequality:
  - "All measures point to a significant increase in overall inequality in the 1990s, particularly in urban areas, and within all but one state (in Bihar inequality remained flat)."
  - Trend reversal: inequality was stable in urban India and declining in rural India in the 1980s; the 1990s saw rising inequality, which helps explain "the lower than expected poverty reduction."
  - Despite faster consumption growth in the 1990s versus the 1980s, the decline in poverty incidence was similar:
    - "The poverty rate fell by 9.4 percentage points (or 20.8 percent) in the 1983–1993/94 period..."

### Methodological notes and limitations highlighted
- NSS deflators and consumption measurement:
  - Use of CPIAL and CPIIW to convert nominal expenditures to base year (1993/94) values; concerns exist regarding appropriateness of these deflators (Deaton, 2003).
- Survey comparability:
  - 1999/00 NSS round has "substantial differences in the measurement of per capita expenditure" and is not relied upon for comparisons spanning the 1990s.
- Role of top incomes:
  - Survey-based measures may understate gains accruing to the top percentiles; income-tax based analyses suggest the top 1 percent played a large role in explaining NAS–NSS divergence.

### Analytical agenda and empirical approach
- The paper documents:
  - Evolution of growth, poverty and inequality in India (Section II).
  - Growth pattern of household consumption across the income distribution over two decades (Section III).
  - Cross-state analysis exploiting heterogeneity to identify state characteristics that influence growth inclusiveness (Section IV), focusing on factors such as financial development, education, labor policies, and sectoral composition.
- Literature linkage:
  - Builds on Ravallion and Datt (1996), Datt and Ravallion (2002), Besley, Burgess and Esteve-Volart (2007), Deaton and Dreze (2002), Banerjee and Piketty (2005), Ahluwalia (2002), Aiyar (2001), Bajpai and Sachs (1996), Bhattacharya and Sakthivel (2004), Purfield (2006), and others referenced for sectoral composition, state heterogeneity, and measurement issues.

*Source: _wp0854 - References..............................................................................................................*

### 8.4 percentage points (or 23.4 percent) during the slightly longer 1993/94–2004/05 period. With

### _wp0854 - 8.4 percentage points (or 23.4 percent) during the slightly longer 1993/94–2004/05 period. With

### Poverty trends and incidence
- Poverty line definition:
  - Minimum subsistence income supporting consumption of 2400 calories in rural areas and 2100 calories per person in urban areas. The poverty line was first established in 1979 and since then it has been periodically adjusted for inflation.
- Table 2: India: Evolution of Poverty — Poverty Rate 1/ (share of population below the poverty line) and Poverty Depth 2/ (poverty rate multiplied by average shortfall from the poverty line)
  - 1983/84: All India 45.2%, Rural 46.2%, Urban 42.1%; Poverty Depth: All India 12.6%, Rural 13.0%, Urban 11.5%
  - 1987/88: All India 39.3%, Rural 39.3%, Urban 39.2%; Poverty Depth: All India 9.6%, Rural 9.4%, Urban 10.4%
  - 1993/94: All India 35.8%, Rural 36.8%, Urban 32.8%; Poverty Depth: All India 8.4%, Rural 8.4%, Urban 8.3%
  - 2004/05: All India 27.5%, Rural 28.0%, Urban 25.8%; Poverty Depth: All India 5.7%, Rural 5.5%, Urban 6.2%
- Alternative consumption measure:
  - Places India’s poverty rate at 21 percent in 2004/05 down from 26.1 percent in 1999/00. Comparable earlier-year estimates are not available under this alternative definition.
- Caveats:
  - Survey data under-represent consumption of the rich; divergence between NAS and NSS data may reflect upper-end income growth. These inequality measures may provide a lower bound of true income-distribution changes.

### Inequality and wealth concentration
- Table 3: India: Evolution of Inequality (selected indicators)
  - Variance of Log Consumption / Gini / Theil Index / Log (PCE 95 / PCE 5) 1/ (as presented in sequence for 1982/83, 1987/88, 1993/94, 2004/05)
    - 1982/83: 0.3190 / 0.3120 / 0.3400 / 0.1980 / 0.1910 / 0.215 / 1.774 / 1.740 / 1.866 / 0.303 / 0.290 / 0.342
    - 1987/88: 0.3130 / 0.3010 / 0.3490 / 0.2000 / 0.1860 / 0.241 / 1.688 / 1.621 / 1.887 / 0.279 / 0.258 / 0.348
    - 1993/94: 0.3030 / 0.2850 / 0.3430 / 0.1910 / 0.1710 / 0.235 / 1.638 / 1.537 / 1.878 / 0.263 / 0.234 / 0.345
    - 2004/05: 0.3250 / 0.2980 / 0.3780 / 0.2280 / 0.1960 / 0.288 / 1.692 / 1.541 / 2.029 / 0.285 / 0.240 / 0.402
  - 1/ Log of the ratio of the per capita expenditure of the 95th percentile relative to the 5th percentile.
- Wealth evidence and concentration:
  - Ahya and Sheth (2007) estimate India witnessed an increase in wealth of over 100 percent of GDP in the past four years from three key sources: the equity market, the residential property market, and gold.
  - Participation and ownership concentrations cited:
    - 4–7 percent of the population participating in the stock market.
    - 47 percent of the population owning a ‘pucca’ house.
    - The top 34 percent of households holding 71 percent of the value of consumer durables (including gold and jewelry).
  - Jayadev et al. (2007) find a small but perceptible rise in interpersonal wealth inequality in the 1990s based on household NSS surveys on debt and investment (with representativeness caveats).

### Decomposing poverty reduction: growth vs distribution
- Methodology:
  - Poverty rate Pt expressed as P(mt, lt) where mt is mean consumption and lt is Lorenz curve; decomposition follows Dhongde (2007) averaging two growth components and two distribution components to obtain a path-independent, complete decomposition.
- Counterfactual simulation findings:
  - 1983–1993/94:
    - Distributional changes enhanced the effect of growth on poverty reduction overall.
    - In rural India, poverty reduction from “growth alone” would have been 27 percent lower had the distribution of income not changed in favor of the poor.
    - In urban India, “growth alone” accounts for the entire poverty decline.
  - 1993/94–2004/05:
    - Changes in the distribution of consumption moderated the extent to which growth reduced poverty.
    - Distribution-neutral growth would have generated a poverty decline that was 22 percent higher in rural India.
    - Distribution-neutral growth would have generated a poverty decline that was 76 percent higher in urban areas.
    - This indicates a marked change in how gains from growth were distributed under the newer market-oriented framework.
- Table 4: Contribution decomposition (selected numerical entries)
  - 1983–1993/94:
    - Rural: Initial Level of Poverty 0.4617; Change in Poverty -0.0933; Contribution of Growth -0.0683; Change in Distribution -0.0249
    - Urban: Initial Level of Poverty 0.4208; Change in Poverty -0.0925; Contribution of Growth -0.0973; Change in Distribution 0.0047
    - All India: Initial Level of Poverty 0.4524; Change in Poverty -0.0940; Contribution of Growth -0.0808; Change in Distribution -0.0132
  - 1993/94–2004/05:
    - Rural: Initial Level of Poverty 0.3684; Change in Poverty -0.0880; Contribution of Growth -0.1071; Change in Distribution 0.0191
    - Urban: Initial Level of Poverty 0.3283; Change in Poverty -0.0702; Contribution of Growth -0.1237; Change in Distribution 0.0536
    - All India: Initial Level of Poverty 0.3585; Change in Poverty -0.0837; Contribution of Growth -0.1151; Change in Distribution 0.0314

### Patterns of consumption growth and inclusiveness
- Growth incidence curves (real monthly consumption per capita):
  - 1983–1993/94:
    - Growth at the bottom of the income distribution outpaced growth at the top, especially in rural India.
    - Urban growth was distribution-neutral.
  - 1993/94–2004/05:
    - Shape reversed: far faster growth at the top than the bottom, with stronger pro-rich bias in urban areas.
    - In almost all states growth became less equalizing in the 1990s.
- Key quantitative points on consumption growth:
  - Between 1993/94–2004/05:
    - Consumption of the richest grew by an average of 3 percent every year.
    - Consumption growth for the poorest was slightly above 1 percent every year.
  - NSS over an 11 year period: consumption of the richest grew by 40 percent.
  - Using income tax data (1987/88–1999/00), Banerjee and Piketty (2005) find:
    - Incomes of the top 1 percent increased 70 percent.
    - Incomes of the top 0.01 percent tripled.
  - Aggregate implication:
    - Though aggregate growth was significantly higher in the 1990s, the bottom 50 percent experienced faster consumption growth in the previous decade.
  - Absolute change for the poorest:
    - Consumption of the poorest increased by mere 8 percent.

### Real wage trends
- Construction:
  - Daily real wages constructed from NSS Employment and Unemployment Schedules (Schedule 10) for 15–65 year old adults engaged in regular salaried or casual wage labor, pooling rural and urban samples. Only observations with non-zero wage values used.
- Patterns:
  - 1983–1993/94: Real wage growth at the bottom generally outpaced growth at the top.
  - 1993/94–2004/05: Real wage growth was a lot more pronounced at the top than at the bottom.
- Caveat:
  - This wage analysis overlooks changes in labor participation and unemployment which influence the observed wage sample; hence the study focuses on per capita consumption growth rather than wages for remaining analysis.

*Source: NSSO various rounds; and Fund staff estimates.*

### 0.5 rupee. Such low values of real wages were not observed in any of the two earlier rounds (in 1983 and 1987/88),

### _wp0854 - 0.5 rupee. Such low values of real wages were not observed in any of the two earlier rounds (in 1983 and 1987/88),

### Patterns of consumption growth across household characteristics
- Data: NSSO various rounds and Fund staff estimates; comparisons across periods 1983–1987/88, 1987/88–1993/94, and 1993/94–2004/05.
- Education (Figure 5, Panel A):
  - 1980s: consumption growth was relatively balanced; with the exception of college graduates, growth rates of consumption were declining with the education of the household head (tightening distribution).
  - 1990s: households with completed middle school and above rose substantially more than those with lower education; fastest growth for college education and above.
- Industrial affiliation (Figure 5, Panel B):
  - 1990s: households in manufacturing and services experienced much larger consumption increases relative to agriculture; growth in these sectors was also substantially faster relative to the 1980s.
- Caste (Figure 5, Panel C):
  - 1980s: consumption growth remarkably similar across SC/ST and other castes.
  - 1990s: divergence emerges with non-SC/ST experiencing faster consumption growth.

### Defining and measuring inclusiveness of growth
- Inclusive growth (pro-poor bias) defined as: the difference between the consumption growth rate of the poorest and richest 30 percent of the population.
- Robustness checks: also analyze bottom/top 10 percent, 20 percent, and below/above median.
- Analysis scope: 15 large Indian states (comprising 95 percent of India’s population in 2004); three periods 1983–1987/88, 1987/88–1993/94, 1993/94–2004/05.

### Sectoral composition and inclusiveness (Table 5 key correlations)
- No evidence of correlation between the speed of overall per capita NSDP growth and inclusiveness: coefficient on Real Per capita NSDP growth close to zero (examples shown: -0.016, 0.005, 0.001, 0.001 with standard errors in brackets [0.202][0.159][0.134][0.095]).
- Agriculture: Real Per capita Agriculture growth coefficients 0.029, 0.032, 0.021, 0.013 ([0.072][0.051][0.040][0.026]) — no strong positive correlation with inclusiveness in presented specifications.
- Industry: Real Per capita Industry growth coefficients 0.139, 0.145, 0.142*, 0.105** ([0.109][0.087][0.071][0.049]) — higher industry/secondary growth correlated with larger pro-poor bias for broader inclusiveness definitions.
- Services: Real Per capita Services growth strongly negatively associated with inclusiveness: -0.591***, -0.496***, -0.421***, -0.292*** ([0.182][0.143][0.121][0.083]) — faster service growth associated with larger gap favoring the rich.
- Context: service sector share increased from 40 percent in 1983/84 to 54 percent in 2004/05; expansion within services concentrated in banking, insurance and communications, benefiting higher-educated/top-income groups.

### Role of policy for inclusiveness of growth — empirical strategy
- Panel regression framework:
  - Dependent variable y_{t,t-1,s}: inclusiveness (bottom 30 minus top 30 consumption growth) over periods 1983–1987/88, 1987/88–1993/94, 1993/94–2004/05.
  - Policy variables X_{t-1,s}: financial development (log total real credit per capita; percent of agricultural laborers with bank loan), labor regulation, human capital (share with secondary education and above), infrastructure index (principal component of installed electricity per capita, surfaced roads per state area, share of households with access to drinking water), log socio-economic expenditure per capita.
  - Controls Z_{t-1,s}: log real NSDP per capita, percent of workforce in agriculture. Period and state fixed effects included. Data for 15 major states; regressions weighted by square root of number of observations within a state.

### Policy correlations with inclusiveness (Table 8 key findings)
- Financial development:
  - Log Real Credit per Capita coefficients (columns shown): 3.780**, 5.429***, 7.073**, 6.152**, 3.865*** ([1.801][1.837][2.845][2.290][1.264]) — higher real credit per capita associated with more pro-poor growth.
  - Percent of Laborers with Bank Loan: 0.190** ([0.090]) — larger initial share of agricultural laborers with formal loans associated with more pro-poor subsequent growth.
- Labor laws:
  - Labor Laws coefficients: -0.682, -1.138**, -1.347, -1.287*, -0.807** ([0.674][0.500][0.803][0.637][0.342]) — relaxed labor protection (amendments towards employer flexibility) associated with more pro-poor consumption growth when controlling for other policies.
- Education:
  - Percent of Primary School Graduates and Above: 0.058** ([0.024]) — primary completion linked to more pro-poor growth.
  - Percent of Secondary School Graduates and Above: 0.253*, 0.235*, 0.445**, 0.321**, 0.139* ([0.137][0.113][0.177][0.140][0.074]) — higher secondary/higher education associated with a more pro-poor pattern.
- Infrastructure:
  - Infrastructure Index coefficients: 2.990*, 2.502**, 4.069**, 2.840*, 1.654** ([1.540][0.915][1.482][1.418][0.624]) — better infrastructure associated with more inclusive growth.
- Socioeconomic expenditure:
  - Log Socioec. Exp. per Capita coefficients shown: 2.1, -1.506, -1.92, -1.78, -1.02 ([1.718][1.428][2.029][1.707][1.003]) — no consistent statistically significant correlation; possible reverse causality, targeting/leakage issues, or administrative gaps.
- Controls and sample:
  - Log Real NSDP per Capita coefficients vary across specs (examples: -1.622, 0.069, -0.641, -0.935, -1.648, -3.03, -1.351, -3.799, -4.667, -4.547, -2.589 with various standard errors as reported).
  - Percent Farmers coefficients generally small and imprecise (examples: 0.232, 0.236, 0.224, 0.19, 0.215, 0.161, 0.238, 0.281, 0.723**, 0.406, 0.152 with reported standard errors).
  - Number of observations: N 43 (varies across regressions; note table footnote: data from Schedule 1 of NSS 38th, 43rd, 50th, and 61st rounds; regressions include state and period fixed effects and are weighted by square root of number of observations within a state).
- Interpretation cautions:
  - Results are conditional correlations; lack exogenous variation so causal inference limited.
  - Potential unobserved time-varying state factors could affect both policies and inclusiveness.

### Summary conclusions and policy implications
- Structural change in the 1990s toward services expansion coincided with less inclusive consumption growth: top incomes captured larger gains relative to the 1980s.
- Indicative evidence that policy can influence distributional outcomes:
  - Promoting financial development (greater real credit per capita; higher share of agricultural laborers with bank loans) is associated with more pro-poor growth.
  - Expanding secondary and higher education correlates with more inclusive growth.
  - Greater labor market flexibility (amendments reducing rigid pro-worker regulations) correlates with a more pro-poor pattern, potentially by facilitating industry/secondary sector expansion.
  - Improved infrastructure correlates with more inclusive growth.
  - State socio-economic spending does not show a robust positive correlation with inclusiveness in these specifications, possibly due to reverse causality or implementation/targeting issues.
- Policy focus recommendation (based on empirical associations):
  - Increase access to formal credit, expand secondary and higher education, improve infrastructure, and consider labor market reforms that facilitate industrial expansion — these are associated with stronger ability of the poor to gain from growth.
- Broader reflection on inequality:
  - Distinguish “bad inequalities” that block market participation and investment in human/physical capital from “good inequalities” that reflect incentives for education and investment; policy should aim to reduce bad inequalities (improve access to schooling, social services, credit, and remove exclusionary barriers).

*Source: NSSO various rounds and Fund staff estimates; data and regression results as presented in the provided content.*

### APPENDIX I: DATA SOURCES

### APPENDIX I: DATA SOURCES

### State-level income data
- National state domestic product (in real and nominal terms) is derived from the Economic and Political Weekly States database and Central Statistics Office.
- The data provides a breakdown of the gross state domestic product into broad industrial categories: agriculture, industry (manufacturing and mining) and services.

### Household level data
- Surveys used: 1983 (38th), 1987/88 (43th), 1993/94 (50th) and 2004/05 (61st) rounds of the National Sample Survey, Schedule 1 and Schedule 10.
- Monthly per capita consumption (Schedule 1) were deflated using the officially provided deflators by the Planning Commission:
  - Consumer Price Index for Agricultural Labor (CPIAL) for the rural sector.
  - Consumer Price Index for Industrial Workers (CPIIW) for the urban/industrial sector.
- Deflation purpose: to be comparable across states, over time and across urban and rural areas.
- Poverty measures: poverty rates and poverty depth constructed using the official Planning Commission poverty lines.
- Wages: Daily wages were constructed from the weekly time disposition questionnaire in Schedule 10 for 15–65 year adults engaged in work activities as regular salaried/wage employee and casual wage labor and were deflated in the same manner as monthly per capita consumption.

### Population counts
- State population information is from the World Bank’s Fiscal Database; the primary source is the Central Statistical Organization of India.

### Credit
- Total credit of Scheduled Commercial Banks utilized by state is from the RBI Basic Statistical Tables, “State-Wise Classification of Outstanding Credit of Scheduled Commercial Banks according to Place of Sanction and Utilization.”

### Labor regulation
- State labor regulation indicators are from Besley and Burgess (2004), updated using the Handbook of Industrial Law (Malik, 2002).
- Indicators classify whether a state has a pro-employer, pro-worker, or neutral labor market regulation based on amendments to the 1947 Industrial Disputes Act.

### Education
- Educational attainment: percentage of the total state population with completed primary education and above, and completed secondary education and above, is from the 1981, 1991, and 2001 Census of India.
- To construct annual series, the number of people in each educational group is linearly interpolated.

### Infrastructure
- The infrastructure indicator is constructed as the main factor from a principal component analysis of:
  - installed electricity capacity per capita,
  - kilometers of surfaced roads per state area,
  - share of households with access to drinking water.
- Annual data sources:
  - installed electricity capacity and kilometers of surfaced roads: Purfield (2007).
  - primary source for statistics on roads: various issues of Basic Road Statistics of India.
  - access to drinking water: Economic Survey of India 2005/06, Table 95.
- Annual series for access to drinking water were completed by linearly interpolating the existing data.

### State social spending
- State spending on social services is derived from the World Bank’s States Fiscal Database; the primary source is the Reserve Bank of India’s Annual Report on State Finances.

*APPENDIX I: DATA SOURCES*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2008/_wp0854.pdf_