History, Institutions and Economic Performance: The Legacy of Colonial Land Tenure Systems in India

Size: px
Start display at page:

Download "History, Institutions and Economic Performance: The Legacy of Colonial Land Tenure Systems in India"

Transcription

1 History, Institutions and Economic Performance: The Legacy of Colonial Land Tenure Systems in India Abhijit Banerjee (MIT) Lakshmi Iyer (Harvard) The New History of Economic Inequality, UCLA April 17, 2004 (Courtesy of Lakshmi Iyer. Used with permission.)

2 Motivation Do institutions matter? --Institutions = rules of the game ; distribution of economic and political power in societies --Large and growing literature which correlates institutional measures and economic performance: North and Thomas (1973), Knack and Keefer (1995), La Porta et al (1998, 2000) --Endogenize institutions: Engerman and Sokoloff (1997, 1998), Acemoglu, Johnson and Robinson (2001) We study the impact of a very specific historical institution: British land tenure systems in India.

3 Our context Land tenure systems: arrangements for the collection of land revenue; defined who had the liability to pay the land tax to the British and by implication who had property rights. Important determinant of asset distribution and contracting environment in agriculture. Land revenue was more than 60% of total British government revenue in The choice of land tenure system was the subject of considerable policy debate. We will compare the impact of different land tenure systems on agricultural productivity and other outcomes today. These systems were formally abolished in the early 1950 s. We thus look at pure institutional overhang effects.

4 Road Map Land tenure systems in British India OLS comparisons IV results and other robustness checks How do historical factors affect current outcomes?

5 Growth of the British Empire in India 1757: Battle of Plassey and political control of Bengal : Conquests in Madras Presidency : Conquest of Bombay Presidency, Central Provinces and North-Western Provinces 1820 s: Orissa and Assam conquered 1846, 1849: Sikh Wars and conquest of Punjab 1856: Annexation of Oudh 1857: Sepoy Mutiny 1858: British Crown takes over administration; no further conquests 1947: End of British rule; Indian empire partitioned into India and Pakistan.

6 Pre-British and British land revenue systems Mughal system (16 th and 17 th century): revenue collector was a transferable state official; revenue rights nonhereditary British land tenure systems set up during Three major types of British systems: Landlord-based (zamindari, malguzari): Property rights given to big or small landlords; 1793 Permanent Settlement in some areas makes revenue payments fixed. Individual-based (raiyatwari): Property rights given to cultivators; detailed survey and record of rights maintained Village-based (mahalwari): Property rights given to village bodies with joint ownership. Depending on size of village body, could resemble landlord-based or individual-based systems.

7 Choice of land tenure system Influence of individual administrators: Munro supports raiyatwari in Madras (1812) Elphinstone implements raiyatwari in Bombay (1820s) Mackenzie recognizes village bodies in North-West Provinces (1819) Ideology prevalent in England: Anti-Jacobinism, Utilitarianism. (Stokes 1978) Beliefs about local tradition or presence of an existing landlord class (e.g. Bengal). No consensus about the original system of India. Date of land revenue control: areas conquered later were more likely to get non-landlord systems due to growth of infrastructure, changes in ideology and precedents set in other places Political events: landlords in Oudh were reinstated after the 1857 revolt, despite the initial plan of a village-based system

8 Data District-level annual data on agricultural investments and productivity Historical land systems and extent of landlord control: Land Settlement Reports (1870s, 1880s), historical accounts Current districts are matched up to old districts using old and new maps: district boundaries have changed much less than state boundaries which were reorganized in 1956 We drop districts which are currently in Pakistan or Bangladesh and those which were under administrative control of Indian rulers.

9 Empirical strategy OLS regressions: Y it = constant + α t + β NL i + X it γ + ε it Y it = outcome in district i and year t -- Agricultural investments: irrigation, fertilizer use, adoption of high-yielding varieties (HYV) -- Agricultural productivity: yield per hectare -- Investments in human capital NL i = measure of non-landlord control (proportion of district not controlled by landlords) X it = control variables (geography, date of British land revenue control)

10 Endogeneity concerns Choice of land revenue system often determined by political factors Landlord areas have better geographical characteristics e.g. greater rainfall and topsoil depth Some areas were converted from a landlord system to non-landlord; these were usually less productive areas where the landlord could not meet the revenue commitments Landlord areas were more productive in the Colonial period. Nevertheless, we compute instrumental variable estimates to address potential endogeneity concerns.

11 Results OLS results: Non-landlord areas have significantly higher agricultural investments and yields Sensitivity to alternate measures of NL Classify all districts as either landlord or non-landlord Compare landlord areas to individual-based areas We might be concerned about (i) Omitted district characteristics (ii) Measurement error in NL (iii) Endogeneity of historical institutions Use only neighboring districts which happened to have different land tenure systems Construct instrumental variable estimates

12 Instrumental variable estimates Instrument = dummy for whether the British took over revenue control between 1820 and 1856 Date of revenue control = date of conquest for most places Why is this a good instrument? Administrative considerations favor landlord systems in areas conquered earlier. Non-landlord systems established in Madras and NWP starting from 1820, forming precedents. Change in Oudh policy after revolt of 1857 We control linearly for the direct effect of a longer period of British control over land revenue.

13 Why do landlord areas fall behind? Differences widen in the mid-1960s when new agricultural technologies are introduced. Inequality Landlord areas had higher land and wealth inequality in the colonial period. Inequality has declined more in landlord areas in the post- Independence period. State policies States with more landlord areas have enacted more land reforms. States with more landlord areas have significantly lower development expenditure. Poverty reduction is lower in states with more landlord areas.

14 Direct vs indirect colonial rule Compare areas in India which were under direct colonial rule (``British Empire'') with areas which were under indirect colonial rule (``Native states''). Comparison is in the post-colonial period. British controlled foreign and defense policy but internal administration was left to local (Indian) rulers. Omitted variables bias likely to be less severe than in cross-country studies. British colonial history in India provides a source of exogenous variation in colonial status => control for selection into the British empire Uniform administrative structure after the end of colonial rule

15 Empirical Strategy and Results OLS results: --British districts have significantly higher rainfall and lesser proportion of barren or rocky areas. --British areas have significantly higher agricultural investments and productivity, and slightly fewer schools and health centers in the modern period. Use Dalhousie s Doctrine of Lapse as an exogenous determinant of British annexation: native states taken over if the ruler died without a natural heir in IV results are different from OLS: --British districts have no advantage in agriculture, but significantly fewer schools and health centers. --British districts have significantly higher infant mortality rates. --Differences smaller over time => better provision of public goods by local rulers in the past.

16 Results for public goods Coefficient on British dummy British dummy Full sample Post-1847 sample Mean OLS IV Dependent variables: Proportion of villages having public goods Middle school ** (0.034) (0.037) High school * (0.040) (0.042) Primary health center * ** (0.014) (0.013) Primary health ** subcenter (0.017) (0.021) Roads *** (0.065) (0.066) Infant mortality rate ** 1991 (6.36) (10.54) Controls Latitude, rainfall, coast, proportion sandy/barren # districts # native states 96 71

17 TABLE 2 Differences in geography Difference Latitude Altitude Mean annual rainfall (mm) Top 2 soil types: Black soil Alluvial soil Red soil Top-soil depth <25 cm 25-50cm cm cm >300 cm Population density Proportion rural Proportion of working population in farming Cash crops to cereals ratio *** (0.961) *** (25.98) *** (80.83) *** (0.072) ** (0.067) (0.090) (0.024) (0.058) (0.075) *** (0.051) *** (0.088) ** (4.02) *** (0.023) * (0.027) *** (0.048)

18 TABLE 3 DIFFERENCES IN AGRICULTURAL INVESTMENTS AND YIELDS Mean non-landlord proportion = (s.d.=0.4274) Dependent variable Mean Coefficient on NL OLS INVESTMENTS 1. Proportion of irrigated area 2. Fertilizer use (kg/ha) 3. Proportion of rice area under HYV 4. Proportion of wheat area under HYV 5. Proportion of other cereals area under HYV YIELDS 1. log(yield of 15 major crops) 2. log(rice yield) 3. log(wheat yield) Year fixed effects Geographic controls Length of British rule Clustering for errors * (0.034) *** (3.345) 0.079* (0.044) 0.092** (0.046) 0.057* (0.031) 0.157** (0.071) 0.171** (0.081) 0.229*** (0.067) 166 YES YES YES YES Standard errors in parantheses. ***, **, * denote significance at 1%, 5% and 10% level

19 DIFFERENCES IN AGRICULTURAL INVESTMENTS AND YIELDS Excluding Bengal and Bihar Dependent variable INVESTMENTS Without Bengal and Bihar OLS 1. Proportion of irrigated area 0.066* (0.035) 2. Fertilizer use (kg/ha) *** (3.406) 0.094** 3. Proportion of rice area under HYV (0.043) 0.119*** 4. Proportion of wheat area under HYV (0.045) 0.084*** 5. Proportion of other cereals area under HYV (0.024) YIELDS 1. log(yield of 15 major crops) 2. log(rice yield) 3. log(wheat yield) No. of districts Year fixed effects Geographic controls Length of British rule Clustering for errors 0.152** (0.074) 0.195** (0.081) 0.228*** (0.070) 143 YES YES YES YES Standard errors in parantheses. ***, **, * denote significance at 1%, 5% and 10% level

20 AGRICULTURAL INVESTMENTS AND YIELDS BINARY MEASURES OF NON-LANDLORD CONTROL NL=0 for landlord-based and Oudh districts NL=1 for individual-based and village-based districts other than Oudh Dependent variable INVESTMENTS Coefficient on NL dummy OLS OLS Exclude village-based 1. Proportion of irrigated area 2. Fertilizer use (kg/ha) 3. Proportion of rice area under HYV 4. Proportion of wheat area under HYV 5. Proportion of other cereals area under HYV YIELDS 1. log(yield of 15 major crops) 2. log(rice yield) 3. log(wheat yield) 0.077*** (0.027) 9.988*** (2.301) (0.032) (0.036) (0.025) 0.173*** (0.053) (0.062) 0.188*** (0.054) (0.032) *** (3.040) 0.074* (0.038) 0.107** (0.052) 0.109*** (0.041) (0.085) 0.173** (0.079) (0.098) No. of districts Year fixed effects YES YES Geographic controls YES YES Length of British rule YES YES Clustering for errors YES YES Standard errors in parantheses ***, **, * denote significance at 1%, 5% and 10% level

21 DETAILED REGRESSION SPECIFICATION Dependent variable: Log agricultural yield (based on 15 major crops) No controls Controls altitude, latitude rainfall, coast soil types British rule Proportion non-landlord * 0.157** (0.076) (0.088) (0.086) (0.069) (0.071) Altitude * 0.644*** 0.722*** (0.243) (0.245) (0.194) (0.189) lat (0.006) (0.007) (0.007) (0.007) Mean annual rainfall 0.161** (0.067) (0.060) (0.062) Black soil dummy 0.171* (0.096) (0.094) (0.093) Red soil dummy *** *** (0.085) (0.085) Alluvial soil dummy 0.137* (0.071) (0.069) Coastal dummy 0.196*** 0.190*** (0.053) (0.052) Date of British conquest *** (0.001) Constant *** *** *** *** 3.586** (0.038) (0.145) (0.181) (0.186) (1.502) No. of districts R-squared * significant at 10%; ** significant at 5%; *** significant at 1% Standard errors in parantheses, adjusted for within-district clustering. Data from Year fixed effects in all regressions.

22 TABLE 4 INSTRUMENTAL VARIABLES ESTIMATION: FIRST STAGE ESTIMATES Dependent variable: Proportion non-landlord Instrument = 1 if district was conquered between 1820 and 1856 (1) (2) (3) Instrument 0.331*** 0.430*** 0.419*** (0.086) (0.092) (0.087) R-squared No. of observations Geographic controls YES YES YES Date of British rule YES YES YES Date of British rule squared NO YES NO State fixed effects NO NO YES Standard errors in parantheses ***, **, * denote significance at 1%, 5% and 10% level

23 TABLE 4 AGRICULTURAL INVESTMENTS AND YIELDS Instrument = dummy for being conquered between 1820 and 1856 Dependent variable INVESTMENTS Coefficient on NL OLS IV Neighbors only 1. Proportion of irrigated area 2. Fertilizer use (kg/ha) 3. Proportion of rice area under HYV 4. Proportion of wheat area under HYV 5. Proportion of other cereals area under HYV YIELDS 1. log(yield of 15 major crops) 2. log(rice yield) 3. log(wheat yield) 0.101** (0.041) ** (4.979) (0.083) 0.078** (0.034) (0.024) 0.145** (0.061) (0.098) 0.253*** (0.084) (0.137) ** (13.244) 0.411** (0.163) 0.584*** (0.163) 0.526*** (0.129) (0.261) 0.554* (0.285) 0.706*** (0.214) No. of districts Year fixed effects YES YES Geographic controls YES YES Length of British rule YES YES Clustering for errors YES YES Standard errors in parantheses. ***, **, * denote significance at 1%, 5% and 10% level

24 ARE YIELDS EXPLAINED BY INVESTMENTS? Log total yield Log rice yield Log wheat yield (1) (2) (3) Proportion non-landlord Proportion of irrigated area fertilizer use (kg/ha) % area under HYV (0.053) (0.063) (0.063) 0.693** 0.439** 0.435** (0.112) (0.096) (0.117) 0.007** 0.004** (0.001) (0.001) (0.001) 4.274** 0.580** 0.618** (1.122) (0.063) (0.070) R-squared No. of districts Year fixed effects YES YES YES Geographic controls YES YES YES Length of British rule YES YES YES Clustering for errors YES YES YES

25 TABLE 6 AGRICULTURAL INVESTMENTS AND YIELDS WHEN DO THE DIFFERENCES APPEAR? Dependent variable INVESTMENTS Coefficient on NL After 1965 Difference (1) (2) (3) 1. Proportion of irrigated area 2. Fertilizer use (kg/ha) (0.033) 1.026** (0.425) 0.079** (0.036) *** (4.763) 0.033** (0.016) 14.55*** (4.44) YIELDS *** 0.135*** 1. log(yield of 15 major crops) (0.065) (0.076) (0.033) ** 0.088** 2. log(rice yield) (0.069) (0.089) (0.044) 0.146** 0.268*** 0.122* 3. log(wheat yield) (0.058) (0.079) (0.063) No. of districts Year fixed effects YES YES YES Geographic controls YES YES YES Length of British rule YES YES YES Clustering for errors YES YES YES Standard errors in parantheses ***, **, * denote significance at 1%, 5% and 10% level

26 TABLE 7 EDUCATION AND HEALTH Dependent Mean of Coefficient on NL variable dependent OLS IV variable (1) (2) EDUCATION AND HEALTH INVESTMENTS 1981 Proportion of villages having: Primary school *** ** (3.486) (14.585) Middle school *** (2.401) (9.075) High school ** (1.445) (12.953) Primary health center * (0.374) (1.481) Primary health subcenter ** (1.437) (5.385) EDUCATION AND HEALTH OUTCOMES Literacy rate (1961, 1971, 1981, 1991) ** ** (0.0190) (0.0624) Infant mortality rate (1991) *** *** (5.38) (19.00) No. of districts Year fixed effects YES YES State fixed effects NO NO Geographic controls YES YES Length of British rule YES YES Standard errors in parantheses ***, **, * denote significance at 1%, 5% and 10% level

27 APPENDIX TABLE 4 STATE DEVELOPMENT EXPENDITURE Variable Mean Landlord states Non-landlord states Difference (1) (2) (3) Real development expenditure (1960 rupees per capita) (2.45) (1.83) (2.54) Development expenditure/state domestic product (0.0071) (0.0071) (0.0074) Difference is computed after controlling for level of state domestic product per capita

28 TABLE 8 AGRICULTURAL INVESTMENTS AND YIELDS INVESTMENTS Coefficient on NL OLS OLS OLS Base State dev exp State fixed per capita effects 1. Proportion of irrigated area 0.065* (0.034) *** 2. Fertilizer use (kg/ha) (3.345) 0.079* (0.044) 3. Proportion of rice area under HYV 0.092** (0.046) 0.057* YIELDS (0.031) 1. log(yield of 15 major crops) 0.157** (0.071) 2. log(rice yield) 0.171** (0.081) 0.229*** 3. log(wheat yield) (0.067) 0.074** (0.035) (0.036) *** (3.717) (3.308) (0.040) (0.042) (0.047) (0.039) * (0.030) (0.026) 0.174** (0.076) (0.072) (0.082) (0.078) 0.243*** 0.150*** (0.072) (0.045) No. of districts Year fixed effects YES YES YES Geographic controls YES YES YES Length of British rule YES YES YES Clustering for errors YES YES YES State fixed effects NO NO YES Standard errors in parantheses ***, **, * denote significance at 1%, 5% and 10% level

29 TABLE 9 EDUCATION AND HEALTH Dependent Coefficient on NL variable OLS OLS OLS Base state dev exp state fixed per capita effects EDUCATION AND HEALTH INVESTMENTS 1981 Proportion of villages having: Primary school *** 6.778** 9.681*** (3.486) (2.99) (3.365) Middle school 8.454*** 6.116** 5.454** (2.401) (2.433) (2.438) High school 3.579** (1.445) (1.207) (1.322) Primary health center 0.717* *** (0.374) (0.337) (0.311) Primary health subcenter 3.567** 2.402* (1.437) (1.471) (1.259) EDUCATION AND HEALTH OUTCOMES Literacy rate (1961, 71, 81, 91) ** * (0.0190) (0.0171) (0.0176) Infant mortality rate (1991) *** *** *** (5.38) (5.28) (5.40) No. of districts Year fixed effects State fixed effects Geographic controls Date of British land revenue control YES YES YES NO NO YES YES YES YES YES YES YES

30 APPENDIX TABLE 8 CRIME RATES Dependent Mean of Coefficient on non-landlord proportion variable dep.var OLS OLS IV OLS Full sample Neighbors only Full sample State FE CRIME RATES: 1971, 1981, 1991 Total crime/'000 population *** ** ** (0.239) (0.423) (0.888) (0.223) Dacoities+riots/ '000 population ** (0.018) (0.022) (0.060) (0.012) Violent crime/'000 population *** * ** (0.023) (0.026) (0.083) (0.019) Stealing/'000 population *** * * (0.112) (0.316) (0.503) (0.119) Petty crimes/'000 population (0.009) (0.010) (0.039) (0.009) No. of districts a Year fixed effects YES YES YES YES Geography and length of British rule YES YES YES YES State fixed effects NO NO NO YES Clustering for errors YES YES YES YES

31 APPENDIX TABLE 6 CHANGES IN POVERTY AND INEQUALITY Dependent Mean of Coefficient on non-landlord proportion variable dep.var OLS OLS OLS IV OLS Full sample Full sample Neighbors Full sample State FE (1) (2) (3) (4) (5) CHANGE IN POVERTY AND INEQUALITY BETWEEN 1972 AND 1987 Change in Head Count Ratio from *** ** 1972 to 1987 (3.108) (2.466) (2.965) (13.235) (2.275) Change in Gini coefficient from ** 0.019* ** (0.013) (0.010) (0.010) (0.026) (0.007) No. of districts a Year fixed effects YES YES YES YES YES Geography and length of British rule YES YES YES YES YES Control for initial 1972 level b NO YES YES YES YES State fixed effects NO NO NO NO YES Clustering for errors YES YES YES YES YES

32 NEIGHBOURING DISTRICTS Group States Landlord Non-landlord Reason for difference in tenure 1 Andhra Pradesh Srikakulam East Godavari Areas put under landlords and Permanent Vishakhapatnam Settlement before the rest of Madras Presidency was converted to individual system. 2 Tamil Nadu Madurai Coimbatore Areas put under landlords and Permanent Ramanathapuram Thanjavur Settlement before the rest of Madras Tiruchirapalli Presidency was converted to individual system. Tirunelveli 3 Madhya Pradesh, Bilaspur Raigarh Sambalpur district taken over in 1849 from local Orissa Raipur Sambalpur chief and put under individual cultivator system, unlike CP which was conquered in 1818 and put under a landlord based system. 4 Maharashtra, Chandrapur Amravati Berar province taken over from Nizam in 1853 Madhya Pradesh Wardha Yavatmal and put under individual system, unlike CP' Betul which had a landlord based system. 5 Uttar Pradesh Faizabad Allahabad Oudh districts have greater % of (East) Gonda Azamgarh landlords because of change of Pratapgarh Basti policy after 1857 Mutiny. Sultanpur Gorakhpur Jaunpur 6 Uttar Pradesh Hardoi Farrukhabad Oudh districts have greater % of (West) Kheri Fatehpur landlords because of change of Rae Bareli Kanpur Nagar policy after 1857 Mutiny. Unnao Pilibhit Shahjahanpur

33 LAND DISTRIBUTION : 1885 Province Dominant % land owned by large % land owned by peasant land-ownership landlords (paying more than proprietors or village Rs.5000 in revenue) communities Oudh Landlord Central Provinces Landlord NWP Village communities Punjab Village communities Madras Peasant proprietors Bombay Peasant proprietors Berar Peasant proprietors

34 LAND DISTRIBUTION: 1948 Province Dominant % land revenue paid by large % land revenue paid by peasant land-ownership landlords (paying more than proprietors or village Rs.5000 in revenue) communities Oudh Landlord NWP Village communities

35 LAND DISTRIBUTION: 1990 Province % land holdings of size % land holdingsof size Ratio < 1 hectare 2-10 hectares (marginal) (medium) Landlord districts Village-based districts Individual system districts