GEOGRAPHY AND AGRICULTURAL PRODUCTIVITY IN INDIA: IMPLICATIONS FOR TAMIL NADU

Submitted to: State Government of Tamil Nadu

Prepared by: Rina Agarwala Rachel Gisselquist Harvard Institute for International Development With the Direction of Dr. Nirupam Bajpai, Director, HIID’s India Project, And Dr. Jeffrey D. Sachs, Director, HIID April 6, 1999

i TABLE

OF

CONTENTS

Pages Table of Contents EXECUTIVE I.

II.

i iii

SUMMARY

Overview

1

Statement of Objectives

1

Why Study Geography and Agriculture?

2

Why Does this Analysis Focus on Foodgrains?

4

Other Analysis of Agricultural Productivity in India

6

Tamil Nadu is an Ideal Case Study

6

Overview of the paper

7

Methods

8

and

Tools

Step 1: Building the Data Set

8

Step 2: Identifying Patterns Using Geographic Information Systems

10

Step 3: Designing an Empirical Model

10

III. Regional

Foodgrain

12

Trends

National Trends (1967-1980)

12

National Trends (1981-present)

13

A Closer Look at Foodgrain Yield Trends in Tamil Nadu

14

IV. Geographic

Variation

and

Foodgrain

Yields

17

Variable 1: Koeppen Zones

17

Variable 2: Average Precipitation

19

Variable 3: Elevation

20

ii

Pages IV. Geographic

Variation

and

Foodgrain

Yields—cont.

Variable 4: Distance to the Nearest Navigable River 2 0 Variable 5: Soil Suitability Index V.

A

Model

of

Geography

and

20

Foodgrain

Yields

22

Model 1: Isolating the Effects of Koeppen Zones on Yields

22

Model 2: Isolating the Effects of Rainfall and Temperature Across Koeppen Zones

28

VI. Additional

Differences

A

Cross

Koeppen

Zones

36

Comparison of Non-Geographic Determinants Across Climate Zones

36

A Model of Fertilizer Use and Koeppen Zones

38

VII. Policy

Recommendations

41

1. Include Geographic Factors in Economic Analysis of Tamil Nadu’s Agriculture

41

2. Evaluate the Effects of Tamil Nadu’s Agricultural Input Policies on Different Agro-Climatic Production Environments Across Districts

42

3. Encourage Research on New Technologies Adapted to Tamil Nadu’s Geography

43

4. Support the Adoption of Technologies Suited to Tamil Nadu’s Geography

45

5. Address Concerns of Agricultural Risk Caused by Tamil Nadu’s Climate

46

6. Continue Investments in Tamil Nadu’s Manufacturing and Trade Sectors

48

iii

Pages BIBLIOGRAPHY

Appendices

Appendix A: Maps of Regional Foodgrain and Input Trends Appendix B: State Geography and Foodgrain Yields

49

EXECUTIVE

SUMMARY

The Problem: The growth rate of agriculture in Tamil Nadu

may be cause for concern. Recent studies by Professor Jeffrey Sachs show that tropical conditions, such as Tamil Nadu’s can slow economic growth through lower levels of agricultural productivity. In 1995-96, the negative growth of Tamil Nadu’s agricultural sector pulled the state’s growth rate down so low that Tamil Nadu was unable to meet its growth target for the Eighth Plan. Moreover, farmers in Tamil Nadu are shifting away from foodgrain production to higher value commercial crops, thereby raising substantial food security concerns. A continued decline in Tamil Nadu’s agricultural sector, particularly in foodgrains, could have grave implications for the State’s large percentage of rural poor, food supplies and long term growth. This Report: Since the Green Revolution began in the

late 1960s, India has experienced considerable regional variation in agricultural productivity. In particular, states in the Northwest have consistently performed better than central and southern states. Existing analysis on the determinants of this variation have focussed on economic policies and institutions. This report analysis the effects of geography on cross-state variations in rice, wheat, maize and total foodgrain yield within India and presents the policy implications of this analysis for the State Government of Tamil Nadu. It uses the tools of statistical analysis, Geographic Information Systems, and econometric models based on a production function approach. To measure geographic variation, this study uses rainfall and temperature data, as well as the “Koeppen Zone” measure, an indicator of agro-climatic characteristics. The study is intended to complement work done by other analysts on the effects of economic policy on agriculture. Findings: This analysis finds that the differences in foodgrain

yields among states in northern, central, and southern India are strongly linked to regional geographic variation. Geography has

2

an effect even when income, agricultural inputs, and fertilizer are held constant. Specific findings are as follows: *

An empirical model shows that tropical zones, like Tamil Nadu, have less positive effects on yields than dry zones, like Punjab and Harayana. To isolate the effects of Koeppen Zones on yields, the model holds constant labor density, fertilizer lagged NSDP, and the four primary Koeppen Zones represented in India. The model explains over 99 per cent of the variation in foodgrain yields across states in 1991.

*

A Second empirical model illustrates that precipitation and temperature patterns in tropical and dry states have the largest impact on increasing foodgrain yields above those in temperate states, like Uttar Pradesh and Madhya Pradesh. The temperate Zone’s more volatile precipitation levels help explain why its yields are lowest. To isolate the effects of rainfall and temperature across Koeppen Zones, this model uses more detailed information from the World Bank covering the period 1967-86.

*

Finally, this analysis shows that dry states have the highest agricultural input levels. This suggests that technology and inputs may be one channel through which geography affects agricultural productivity.

Recommendations: These findings raise important concerns for tropical states like Tamil Nadu. They suggest that analysis of geography and agricultural productivity within India can help Tamil Nadu improve its own agricultural investments across geographically advantaged and disadvantaged districts. It can also help Tamil Nadu better understand its position and opportunities relative to other states in India. This report concludes with the following six recommendations to assist the Tamil Nadu Government to incorporate geographic analysis into its agricultural policy: 1.

Include geographic factors in economic analysis of Tamil Nadu’s

Within Tamil Nadu, the state government will gain valuable insight into the causes of district level variations in yields by applying the methods described in this report. The Tamil Nadu Government can also build on this report to increase understanding of its geographic advantages and disadvantages relative to other states. agriculture.

3 2.

Evaluate the effects of Tamil Nadu’s agricultural input policies on different agro-climatic production environments across districts.

This report shows that some agro-climatic environments are more favourable for HYV and input use and thus have higher technology and input levels. Tamil Nadu has attempted to raise input levels throughout the state through subsidies and infrastructure projects. In evaluating the effects of these policies, the state should measure whether they are succeeding in both favourable and unfavourable production environments within the state. 3.

Encourage research on new technology adopted to Tamil Nadu’s

India has one of the world’s largest public agricultural research systems. The North, however, enjoys the largest share of research resources. The Tamil Nadu Government should promote increased research focussing on the unique needs of the South’s geography through national public councils, state agricultural Universities, the private sector, and international partnerships. geography.

4.

Support

the

adoption

of

existing

technologies

suited

to

While Tamil Nadu cannot change its geographic profile, it can increase yields through the use of existing technologies that are well-suited to its unique geography. Providing farmers with information on new technologies and releasing new technologies to the market for purchase by farmers can assist farmers in adopting technologies suited to Tamil Nadu’s geography. Tamil Nadu’s geography.

5.

Address concerns of agricultural risk caused by Tamil Nadu’s

This study illustrates that volatility in rainfall and temperature plays an important role in agricultural yields. Uncertainty has led Tamil Nadu’s farmers to favour reduced risks over high yields, thereby slowing the growth of the state’s agricultural sector. The Tamil Nadu Government can reduce the costs of uncertainty by providing farmers with information about weather and strengthening credit and insurance institutions within the state. climate.

6.

Continue investments in Tamil Nadu’s manufacturing and trade

Some effects of geography on yields can be mitigated by technology and inputs that are suited to regional geography. In the long run, however, Tamil Nadu may naturally shift away from agriculture to other, more profitable sectors of the economy. Tamil Nadu appears to be making this transition, but should not ignore agriculture in the process. sectors.

4 I.

OVERVIEW

This policy analysis falls under the Harvard Institute for International Developments’ three year contract with the State Government of Tamil Nadu, India. Under this contract, HIID provides Tamil Nadu with technical assistance and policy advice to further the state’s macroeconomic growth. Research and analysis were conducted under the guidance of Professor Jeffrey Sachs, Director of HIID, and Dr. Nirupam Bajpai, Director of HIID’s India Program. STATEMENT

OF

OBJECTIVES

The primary objective of this policy analysis is to help the Tamil Nadu Government increase its understanding of the determinants of foodgrain productivity in India. Such understanding can lead to more targeted agricultural development strategies that promote growth, alleviate poverty, and ensure food security for the state’s growing population. Traditional studies have focussed on economic policies and institutions as the primary determinants of Indian agricultural productivity. This analysis looks at geography as another potential determinant of agricultural productivity, with particular focus on foodgrains, across Indian states. The analysis provides the State Government with information to complement existing studies. It highlights the implications geography may have on Tamil Nadu’s foodgrain productivity relative to other states and concludes with six policy recommendations to assist the State Government to incorporate geographic analysis into development polices. This report uses statistical and econometric analysis and Geographic Information Systems to study the impact of geography on variations in foodgrain yield across India’s states. It represents a first step in applying the methods of recent HIID cross-country studies on geography and agricultural productivity to state-level analysis of foodgrain productivity in India. This analysis answers the following questions: *

What are the recent trends in rice, wheat, maize, and total foodgrain yields in India and in Tamil Nadu?

*

How do state variations in foodgrain yields match up with geographic variations?

5

WHY

*

Holding income, agricultural inputs, and technology constant, does geographic variation have an impact on foodgrain yields in India?

*

Through what indirect channels might geography impact foodgrain yields in India?

*

What are the policy implications of this analysis for Tamil Nadu? STUDY

GEOGRAPHY

AND

AGRICULTURE?

Recent cross-country analysis conducted by HIID have shown, that, in addition to economic policy, geography has important effects on cross-country variations in economic growth. Jeffrey Sachs points out that agricultural productivity is one of the three channels through which geography affects economic growth2. John Gallup’s paper “Agricultural Productivity and Geography” shows that agricultural output per person is much lower in tropical than in temperate regions.3 These findings have important implications for development in India. Twenty-five per cent of India’s gross domestic product (GDP) comes from agriculture, 75 per cent of the population lives in rural areas, and 40 per cent of the land is in the tropics. While India’s location in the tropics suggests that it is at a disadvantage in terms of agricultural productivity, India cannot ignore the agricultural sector. First, most of the country’s citizens still live in rural areas and depend on agriculture for their livelihood. In 1993, thirty-seven per cent of the rural population, totalling 268 million people, lived below the poverty line. Second, the country’s population of 961 million demands enormous food supplies, which must grow at least as fast as projected population growth rates. Likewise, these findings have important implications for Tamil Nadu for four reasons. First, 95 per cent of Tamil Nadu’s land is in the tropics. Second, Tamil Nadu’s foodgrain production is decreasing as farmers shift to higher value crops, thereby raising food security concerns for the state. 1

The acronym “HIID” is used to refer to the Harvard Institute for International Development throughout this report.

2

Gallup, John Luke and Jeffrey Sachs, “Geography and Economic Development”, Harvard Institute for International Development (April 1998).

3

Gallup, John Luke, “Agriculture Productivity and Geography,” International Development (1998).

Harvard Institute for

6

Third, while Tamil Nadu’s economic indicators generally compare favourably to those of other Indian states, its growth rates have been erratic and some of this variability has been driven by variable growth rates in agriculture.4 Tamil Nadu’s net state domestic product (NSDP) is highest among the southern states. However, as shown in Figure 1, negative growth in agriculture in 1995-96 pulled the state’s overall growth rate down so low that the state was unable to achieve its targeted growth rate of 5.6 per cent during the Eighth Plan. 5 The need to focus on agriculture does not mean, of course, that Tamil Nadu should ignore other, growing sectors of its economy. Nearly 80 per cent of the state’s income came from manufacturing and services in 1996, and the state’s manufacturing sector ranks second after Maharashtra in terms of value added. However, a productive agricultural sector is necessary for further growth in other sectors. It can provide stable and reasonably priced food supplies, employment opportunities, and a consumer base for urban output.

Figure

1:

Tamil

Nadu’s

NSDP

and

Agriculture

Growth

Rates.

4

Sawant, S.D., “Performance of Indian Agriculture with Special Reference to Regional Variations,” Indian Journal of Agricultural Economics 52 (July-September 1997): 354373.

5

Government of Tamil Nadu,. Evaluation and Applied Research Department, Tamil Nadu: An Economic Appraisal, 1995-96

(Mumbai: Government of Tamil Nadu, 1995).

7

Fourth, Tamil Nadu has high rural poverty rates, and in order to raise incomes in the agricultural sector, it must raise agricultural productivity. In 1994, Tamil Nadu’s NSDP per capita was the highest among India’s southern states and fifth highest in the country. However, as shown in Figure 2, Tamil Nadu’s percentage of rural poor exceeds that of the other southern states and the national average. Forty-six per cent of its rural residents live below the poverty line. Despite Tamil Nadu’s high urbanization rate of 34 per cent (compared to national average of 26 per cent), 66 per cent of the state’s population still resides in rural areas and 60 per cent of its work force is in agriculture. 6 On the positive side, Tamil Nadu has the third lowest infant mortality rate, after Kerala and Punjab, and the fourth highest literacy rate, after Kerala, Maharashtra, and Himachal pradesh. 7 Figure

2:

Poverty

in

Southern

States

(1996)

Number of poor (000,000)

Percentage of rural poor

161

45.8

Andhra Pradesh

95

20.92

Karnataka

94

32.82

Kerala

66

29.1

2294

39.06

Tamil Nadu

India WHY

DOES

THIS

ANALYSIS

FOCUS

ON

FOOD

GRAINS?

First, better understanding of the determinants of foodgrain productivity can improve food policy and food security in India. Food policy affects nutrition levels and food security for the entire population, as well as income levels for the rural population. Sound food policy thus balances welfare concerns with economic efficiency. The Indian Government at the national and state level is very involved in food policy. Since 6

1991 Indian Census.

7

Centre for Monitoring the Indian Economy, Profiles of States (Mumbai: CMIE, 1997).

8

1943, India has employed considerable food control, such as input subsidies, international and domestic trade restrictions, and subsidized distribution through the Public Distribution System8. Indian

Food

Policy

Objectives

Concerning

Foodgrains

1. Achieve self-sufficient in production. 2. Maintain price stability. 3. Assure equitable distribution of supply at reasonable prices. In recent years, food policy experts have expressed concern that farmers are shifting from foodgrain to non-foodgrain production, and that India will thus be unable to meet its growing foodgrain needs through domestic production. The 1990s marked a distinct fall in the growth of foodgrains in India to a rate barely equal to population growth. The Government of India has recognized this trend as a concern that “must be reversed.”9 The shift away from foodgrains reflects farmers’ changing production incentives and highlights the need for policy makers to better understand farmers’ production decisions. This report uses a production function approach to model the effects of geography on farmers’ production decisions. Second, the majority of agricultural land in India is devoted to foodgrains, so productivity in foodgrains is often used as a proxy for agricultural productivity. As shown in Figure 3, well over 50 per cent of the gross cropped area in all but 3 states is under foodgrain cultivation.

8

For further information on Indian food policy see Chopra, R.N., Evolution of Food Policy in India (New Delhi: Macmillan India Ltd., 1981) and Sanderson, Fred and Shyamal Roy, Food Trends and Prospects in India (Washington, DC: Brookings Institution, 1979). Policy analysts disagree as to whether India’s food policy favours poor consumers, urban consumers, large farm holders and/or powerful farm lobbies, and whether it supports or hinders agricultural productivity.

9

Government of India, Economic Survey 1998-99 (New Delhi: Government of India, Ministry of Finance Economic Division, 1999) 117.

9 Figure

3:

Percentage

of

Gross

Cropped

area

(GCA)

in

Foodgrains

OTHER IN

ANALYSIS

OF

AGRICULTURAL

PRODUCTIVITY

INDIA.

Most economic Analysis of Indian agriculture to date focus on non-geographic factors, such as economic policies on subsidies and tariffs; Green Revolution technology; or institutional constraints, such as access to credit.10 There have been some attempts to incorporate geography into economic Analysis of agriculture in India. Under the leadership of SN. Subramanium, the Madras Institute began studying geography and agriculture in India as early as 1928. Until the Green Revolution in the 1960s, however, the studies focussed on descriptive accounts of static land use and crop distribution. Since the 1960s, there has been increased interest in regional disparities in agricultural development and crop productivity. Most studies argue that policies and technology are the key factors driving regional variations in agriculture.11 In 1979, a study by the Brookings Institution argued that the largest variations in agricultural performance in the 1960s-70s were due to the cost of technology and to technology’s poor adaptation to geographic conditions.12 These studies, however, tend to control for the fixed effects of geography in order to focus on other determinants of agricultural productivity. There 10

Tiwari, P.S., Agricultural Geography (New Delhi: Heritage Publishers, 1986).

11

Chatterjee, S.P., Fifty years of Science in India: Progress of Geography, (Calcutta: Indian Science Congress Association, 1968).

12

Sanderson and Roy.

10

have been few efforts to control for other determinants to examine the effects of geographic variations on agricultural productivity across Indian states. TAMIL

NADU

IS

AN

IDEAL

CASE

STUDY

Tamil Nadu’s unique geography makes it an ideal region in which to apply cross-state analysis of the implications of geography on agricultural productivity on India. It is the southern most state in India, and ninety-five per cent of the state is in the tropics. Tamil Nadu’s northern and western boundaries are flanked by the Western Ghats, which reach a peak of 8,000 feet in the Nilgiri Hills. The southern and eastern boundaries of the state are on the Indian Ocean. Most of the south-eastern portion is comprised of plains with one major river and several small tributaries. Due to its unique location, Tamil Nadu is the only state in India that receives two monsoons. From June until September, it receives the southwest monsoon, on which most of the state’s agriculture relies. On average, the state receives 32.4 per cent of its annual rainfall during this season. From October to December, it receives the northeast monsoon, from which it receives 47.6 per cent of its annual rainfall. Tamil Nadu’s annual rainfall average is low to moderate at 943 mm per year. Its tropical climate, and temperatures ranging from 180C to 440C makes the rate of surface water run off and evaporation very high. Therefore, it is difficult to store monsoon rains in tanks. Despite the state’s considerable investment in expansion of groundwater irrigation, ground water tables are rapidly decreasing. OVERVIEW

OF

THE

PAPER

The remainder of this report is organized in six sections. Section II describes project methodology and useful tools for further analysis. Section III gives an overview of regional foodgrain trends in rice, wheat, maize, and total foodgrains since the Green Revolution. It also provides a closer look at foodgrain

11

trends in Tamil Nadu. Section IV Analysis how state variations in foodgrain yields correspond with five geographic variables in India and isolates Koeppen Climate Zones as the geographic variable of interest for this study. Section V presents two empirical models that isolate the impact of geography on state foodgrain yields in India. These models hold constant income, agricultural inputs, and technology. Section VI describes a preliminary analysis of how input levels and other factors of production differ across Koeppen Zones. Finally, this report concludes with six recommendations for the Tamil Nadu Government incorporate the policy implications of these findings.

12

II. Step

1:

Building

METHODS the

data

AND

TOOLS

set

This study combines two data sources from HIID, the Integrated India Data Set and Geographic Information Systems (GIS) India project files.13 The combined Integrated Data Set now includes over 100 economic, demographic and geographic variables for each state and most union territories from 1980-1996. In addition, this study uses the World Bank’s India Agricultural Data Set. This data set includes district level data for 271 districts in 13 states, covering 85 per cent of India for the period 1967-1986. Kerala and Assam are the two major agricultural states not covered. Because this project was intended to focus on cross-state analysis for the whole country, a significant portion of the project involved building the Integrated Data Set to include more agricultural and geographic variables. New Agricultural Variables in the Integrated Data Set: New

data was added for total foodgrain yield per hectare; rice, wheat and maize yields; area under foodgrain cultivation; tractors; electric pumps; diesel pumps; fertilizer; average rainfall per year; net and gross irrigated area under rice, wheat, maize and total foodgrain cultivation; cultivable land; and land sown.14 New Geographic Variables in the Integrated Data Set: Using GIS, data tables were built from Maps of India and added to the data set. The new data included measures for mean elevation (meters), surface temperature (average of monthly means

13

The Geographic Information Systems India Project file was built by Andrew Mellinger of HIID. He contributed greatly to this project by providing expert assistance in working with GIS.

14

The majority of the new agricultural data was drawn three publications by the Government of India, the Centre for Monitoring the Indian Economy, Agriculture, August 1997 edition (Mumbai: CMIE, 1997); Economic Intelligence Service, Agriculture (Mumbai: Economic Intelligence Service, September 1998); and Ministry of Agriculture, Area and Production of Principal Crops in India (New Delhi: Ministry of Agriculture, 1994). Recent data was also used from Ministry of Agriculture, Agricultural Statistics at a Glance (New Delhi: Image Print, March 1998) and Ministry of Agriculture, Indian Agriculture in Brief 26th edition (New Delhi: Government of India Press, May 1995).

13

in 1987), rainfall (monthly mean for 1987), Koeppen Zones15 (percentage of land area in each zone), distance to the nearest coastline (in km from the centroid), soil moisture (mean), soil temperature (mean), soil depth (mean), soil suitability (mean), irrigation suitability (mean) and Matthews Cultivated Land. Limitations

of

the

Integrated

Data

Set

The main weakness in using the Integrated Data Set to study geography and agriculture is that it lacks adequate rain, temperature and soil data for varying years. For example, the only temperature and precipitation data included are the monthly means for 1987, a drought year in India. HIID is in the process of coding almost 100 years of precipitation and temperature data from meteorological stations. This data will prove valuable in future Analysis. In addition, the Integrated Data Set’s only soil data are two soil suitability indices for all crops, making detailed foodgrain analysis difficult. In order to remedy the weaknesses of the Integrated Data Set, World Bank’s district level data set was used as well. The

World

Bank

India

Agricultural

Data

Set

The World Bank Data Set includes detailed agro-climatic data on temperature, rainfall, and soil quality. It also includes statistics on agricultural productivity, inputs, technology use, and prices for 1967-1986. Climate data is from meteorological climate and precipitation observations from 160 weather stations across India and is calculated for districts in the set using surface interpolation techniques.16

15

Koeppen Zones are a climate classification system. See Appendix B for a guide to Koeppen Zone classification.

16

Much of the work on this data set was originally organized by Robert Evenson with James McKinsey. The data set has been used extensively by the Word Bank in Studying the effects of climate change on Indian Agriculture. For more information on World Bank analysis using the World Bank Data Set, edaphic variables, and extrapolating climate data from station to district level, see Dinar, Ariel, et al, Measuring the Impact of Climate Change on Indian Agriculture (Washington, DC: World Bank, 1998), World Bank Technical Paper No. 402. The World Bank India Agricultural Data Set may be downloaded from http://www-esd.worldbank.org/indian/database.heml.

14 Step

2:

tion

Identifying

patterns

using

Geographic

Informa-

Systems

In addition to its use in providing geographic data for the Integrated Data Set, GIS is also valuable as a tool of analysis in its own right. In this study, it was useful in illustrating trends and correlations between geographic and foodgrain yield variations among states. These are complex relationship that are often difficult to study through the use of statistics alone. GIS maps helped isolate India’s Koeppen Zones as the key geographic variable of interest in this project. Appendix A includes sample GIS maps. Step

3:

Designing

an

Empirical

Model

Building on a literature review and GIS results, this study undertook a formal empirical analysis of the effect of geography on foodgrain yields. The underlying hypothesis tested was that dry Koeppen Zones have the most positive effect on foodgrain yields, even when income, Green Revolution technology, and inputs are held constant. The focus was on yields because most analysis agree that, due to limited resources, increased production in India will only come through increased yields. This hypothesis was tested in sequential steps. The following sections of this report mirror these steps. First, is there regional variation in foodgrain yields in India? Second, do foodgrain yields vary with India’s geographic variation? In this step, Koeppen Zones were identified as the geographic variable of interest. Third, holding other factors of production constant, do Koeppen Zones have different effects on foodgrain yields? Fourth, is there any evidence that non-geographic factors of production are affected by Koeppen Zones? The empirical methods used include t-tests of the equality of means to measure whether yields and variables differ in a statistically significant manner across Koeppen Zones; linear regression analysis; F-tests and t-tests of coefficient estimates; and predictions and simulations using the results of regression analysis. The regression models follow Gallup in using an agricultural production function to explain the empirical

15

relationship between inputs, geography and agricultural productivity. The dependent variables were rice, wheat, maize, and total foodgrain yields. The production function model was chosen because it works best in isolating the effects of geographic variables on the dependent variable. Many recent Analysis have followed the Ricardian model, using annual net revenue as a proxy for net rent or value of farmland. Net revenues are used because land rents are so highly controlled. However, net revenues can only serve as estimates and may thus distort results.17 Other work controls for differences in geography with fixed effects. This, however, nets out the influence of geography from the analysis.18 In contrast to other models, the use of a physical production function in Gallup’s words “avoids most of the complications of the effect of the economic policy regime on agriculture, like exchange rates, quotas, price subsidies and taxes. Nor should missing markets affect the estimation. Whatever input levels are chosen, which will be affected by price distortions and market imperfections, those inputs should have a consistent impact on output if the aggregate production function specification is tenable.”19 One drawback to the use of the production function approach is that it does not directly estimate policy effects within the model. However, direct measurement of policy effects in the empirical model are problematic in any case because comprehensive data on policy variables is often unavailable. Also, there is little variation in policy across states as many agriculture-related policies are determined on a national basis 17

For further explanation, see Dinar, Ariel et al, Measuring the Impact of Climate Change on Indian Agriculture (Washington, DC: World Bank, March 1998), World Bank Technical Paper No. 402. Another weakness of the Ricardian approach is that it will be biased if an uncontrolled factor is correlated with the variable of interest. Thus, it becomes especially important to measure and control for every variable that might affect farm economic performance and be correlated with geographic variables. This can often be difficult because of the dearth of data on developing countries.

18

Gallup, 1.

19

Gallup, 2.

16

in India. Advocates of the Ricardian method argue that using a production function may overstate the effects of geography. Nevertheless, the production function approach is ideal for this analysis because it allows for estimation of the isolated effects of geography on yields.

17 III.

REGIONAL

FOODGRAIN

TRENDS

Question: Is there regional variation in foodgrain yields in India? Answer:

Yes, foodgrain yields vary considerably across Indian states. In particular, the early years of the Green Revolution marked a period of pronounced regional variation in foodgrain yields between the Northwest and the South. Regional variation in yields has decreased since the 1980 but is still apparent.

National

Trends

(1967-1980)

The Green Revolution in the late 1960s shifted the world’s focus from increasing agricultural output through expansion of cultivated area to increasing it through higher yields. Yields were increased through the use of irrigation, fertilizer and new high yield variety seeds (HYVs). From 1966 to 1980, India increased its foodgrain yields by 63 per cent from 644 kg/ha to 1,023 kg/ha. The 1970s thus began India’s move towards self-sufficiency in foodgrains. The revolutionary improvements of the 1970s, however, also began more pronounced variation in foodgrain yields between the Northwest and the South. Output growth in 1970s was heavily concentrated in the Northwest regions of Punjab, Hariyana and Western Uttar Pradesh. As shown in Figure 4,20 Figure

4:

Regional

Variation

Agricultural

20

in

%

of

National

Output.

Government of India, Area and Production of Principle Crops in India (New Delhi: Ministry of Agriculture, 1994).

18

in 1965, before the Green Revolution began, the South and Northwest made similar contributions to national agricultural output. However, by 1980, the Northwest contributed almost twice as much as the South. Regional variation in foodgrain yield was, in part, due to research priorities focussing on regional environments. For example, until the 1980s scientists focused on developing HYVs only for certain climatic conditions. Rice technology and research in Asia has historically focused on improving yields in favourable environments because of the “higher probability of scientific success.”21 In India, early HYVs for wheat and rice required bright days and cool nights and thus fared well in the dry Northwest states. Early HYVs for rice did not fare well in the tropical South, because they were not suited for the cloudy days and warm nights of the tropical monsoon season, during which 95 per cent of rice is grown in the South.22 HYVs used during the tropical monsoon season, produced both less rice and rice of inferior quality. Thus, farmers in the South received lower returns on their investment in HYV rice than did farmers in Northwest. In addition, during the tropical dry season, successful use of HYVs required substantial investments in irrigation to provide the amount of water required for HYVs. Thus, during the 1970s, most farmers in the South opted against costly investments in modern inputs and continued to use lower productivity, traditional seed varieties.23 National Trends (1981 - Present)

Since the 1980s, use of HYVs and other new technologies have spread to the eastern, western, central and southern regions of India, resulting in more widespread agricultural growth throughout the country. This was due, in part, to the adaptation of technology to other environments and to favourable monsoons, which helped to optimize the new technology. 21

David, Cristina C and Keijiro Otsuka, “Modern Rice Technology: Emerging Views and Policy Implications,” in Modern Rice Technology and Income Distribution in Asia, eds. Cristina C. David and Keijiro Otsuka (Boulder & London: Lynne Rienner Publishers, 1994), 428.

22

Sanderson and Roy, and Food Trends and Prospects in India.

(Washington, D.C.: The

Brooking Institution, 1979) and Gillespie, Stuart and Geraldine McNeil, Food, Health, and Survival in India and Developing Countries (Delhi: Oxford University Press, 1992). 23

Gillespie, Stuart and McNeil, Geraldine.

Food, Health and Survival in India and

Developing Countries. (Oxford: University Press, 1992) p. 36.

19

As shown in Figure 5, although area under foodgrains declined during the 1980s, foodgrain production and yield grew considerably from the previous decade.24 The growth rate of the net national domestic product generated in agriculture grew from 2.09 per cent from 1968-1980 to 3.22 per cent from 19811990.25 During the 1980s, India became self-sufficient in foodgrains. Maps in Appendix B illustrate the dramatic changes Figure Food

5

:

Grains

Compound between

Annual Phase

Growth I

and

Rate Phase

of II

that took place in Indian rice, wheat, maize and total foodgrains yields between the early part of the Green Revolution (shown in maps for 1980 in Appendix B) and the later part (shown in maps for 1992 in Appendix B). Nevertheless, regional variation is still evident as states in the Northwest remain among the top food producers in India. (See, for example, crop yield maps in Appendix B). Moreover, the 1990s have been marked by a decline in annual growth rates, as there have been few new contributions to widespread growth in yields since the Green Revolution. A Closer Look at Foodgrain yield trends in Tamil Nadu.

26

Although Tamil Nadu is not among the top ten states in terms of foodgrain area or production, it is ninth in terms of foodgrain yields.27 While Tamil Nadu’s performance has been 24

Sawant, S.D. et al, “Performance of Indian Agriculture,” Indian Journal of Agricultural Economics 52.3 (July-September 1997), 354-374.

25

Sawant, S.D. 354-374.

26

Production and yield charts use data from Centre for Monitoring the Indian Economy, (Mumbai: CMIE, 1997) and from the Integrated Data Set and GIS India Project files.

27

Centre for Monitoring the Indian Economy, Agriculture (New Delhi: CMIE, September 1998).

20

generally positive, there are two causes for concern. First, Tamil Nadu has had volatile growth rates in foodgrain yields since the mid-1980s. Second, Tamil Nadu’s principal foodgrain crop, rice, has had declining growth rates in yield since the mid-1980s. These trends prompt considerable food security concerns. Rice comprises nearly 80 per cent of Taml Nadu’s total foodgrain production. Today, Tamil Nadu is fourth in rice yield after Punjab, Hariyana and Goa. During the 1980s, Tamil Nadu’s area under rice cultivation declined by almost 10 per cent to 2,228.5 ha. in 1994.28 Yield and production levels have been positive since the 1980s, but the growth rates in yields are decreasing as shown in Figure 6. During the 1980s, Tamil Nadu comprised the third largest share of national value of agricultural output at 9.5 per cent after Uttar Pradesh and Madhya Pradesh.29 However, the contribution of agriculture to NSDP declined faster in Tamil Nadu than in India, from 52 per cent in 1960 to 40 per cent in 1982, as compared to 49 to 40 per cent over the same period in India.30

28

Ministry of Agriculture, Area and Production of Principal Crops in India, Volumes 1987-94 (New Delhi: Ministry of Agriculture, 1987-94).

29

Government of Tamil Nadu, Tamil Nadu : An Economic Appraisal, 1995-96 (Chennai: Government of Tamil Nadu, 1996).

30

Perumalsamy, S. Economic Company Ltd., 1996).

Development

of

Tamil

Nadu (New

Delhi: S. Chand and

21 Figure

6:

Foodgrain

and

Rise

Trends

Yield

and

Production

22 IV.

GEOGRAPHIC

VARIATION

AND

FOODGRAIN

YIELDS

Question:

Do state foodgrain yields vary with India’s geographic variation?

Answer:

Yes, state foodgrain yields vary considerably with differences in Koeppen Climate Zone classification. Koeppen Zones serve as a summary variable for rainfall, temperature and soil quality. States in the dry Koeppen Zone have the highest average foodgrain yields. Based on statistical analysis and maps using the Integrated Data Set and GIS India Project file, this report identifies Koeppen Zones as the Key variable of interest for further analysis. Cross-country Analysis done by Sachs and Gallup have also identified agro-climatic zones as significant to variation in agricultural productivity.

This section presents a summary of India’s cross-state foodgrain yield variation across five geographic variables found in the Integrated Data Set and GIS India Project file: Koeppen Climate Zone, elevation, average precipitation, distance from the nearest navigable river and a soil suitability index. (For further information on these variables, maps and charts, see Appendix B). VARIABLE

1:

KOEPPEN

ZONES

Koeppen Zones are a climate classification system based on monthly and seasonal rainfall and temperature, and other geographic indicators. Nine Koeppen sub-zones are represented in India, but only four make up the majority of land in most of India’s states. These are the tropical monsoon type AM zone; the tropical AW zone with a distinct dry season; the dry steppe climate BS zone and the temperate CW zone with a winter dry season. (For a guide to the Koeppen classification system, classification of regions by Koeppen Zone and a map of zones and regions, see Appendix B). Thirty-eight per cent of India is in the temperate BS zone, 27 per cent in the tropical land falls within the tropical monsoon AM zone 16 per cent in the dry BS zone and 6 per cent in the temperate AW zone. More than 92 per cent of Tamil Nadu’s land falls within the tropical

23

monsoon AM zone. As in HIID’s cross-national studies, regions with over 50 per cent of their land in one Koeppen Zone were classified within that zone. Tamil Nadu was thus classified as “tropical AM”. Average

Yields

across

Koeppen

zones

Maps suggest that foodgrain yields vary by Koeppen Zone. (See, for example, maps in Appendix B.) As shown in Figure 8, the World Bank data shows that average rice, wheat and maize yields from 1967 to 1986 were higher in the dry zones than in the temperate and tropical zones. Figure

8:

A V E R A G E S,

1967-1982

(From World Bank India Agricultural Data Set) Tropical AM

31

Temperate

Dry

Highest

C W

BS

Zone

Rice Yield (kg/ha)

1380

791

1467

Dry

Wheat Yield (kg/ha)

692

1045

1186

Dry

Maize Yield (kg/ha)

947

862

1250

Dry

As shown in Figure 9, according to the Integrated Data Set, dry zones continued to outperform the other zones in rice, wheat and maize yields in the early 1990s. However, from 1991 to 1996, tropical AW states performed best in terms of total foodgrain yields. Figure

9:

AVERAGES

(From HIID Integrated Data Set) Tropical

Tropical

Dry

A M

A W

BS

C W

Foodgrain Yield, 1991-96 (kg/ha)

1328

1653

1627

1171

1293 Tropical A W

Rice Yield, 1992 (kg/ha)

2145

2020

2455

1284

1767

Dry BS

Wheat Yield, 1992 (kg/ha)

1335

.

2583

1567

1784

Dry BS

Maize Yield, 1992 (kg/ha)

1861

.

2218

1515

1767

Dry BS

31

Temperate Overall Highest

The World Bank Data Set does not include states in the tropical Aw zone.

Thus,

“tropical” is used to refer only to the AM zone when discussing the World Bank data.

24

In order to test whether mean foodgrain yields have varied across Koeppen Zones in a Statistically significant manner, null hypothesis were constructed about the equality of mean yields in different zones. Hypothesis were tested using-t-tests of means, as shown in Figure 10. In the data from 1967-86, all the differences in mean yields across Koeppan Zones are statistically significant at the 1 per cent level. In the data for foodgrain yields for the 1990s, the null hypothesis was rejected in all but one case, suggesting that mean foodgrain yields continued to differ between all zones, except between AW-BS, in a statistically significant manner. In other words, the differences in average yields between almost all Koeppen Zones since the beginning of the Green Revolution are probably not the random result of the years studied or observations included. Figure

10:T-Tests

on

difference

across

Koeppen

in

MEAN

foodgrain

yield

zones

(Using the HIID Integrated Data Set, 1990-96) Null hypothesis

Reject ? (at 95% confidence)

Ho: Mean yield in am zone = mean yield in aw zone

Reject

Ho: Mean yield in am zone = mean yield in bs zone

Reject

Ho: Mean yield in am zone = mean yield in cw zone

Reject

Ho: Mean yield in aw zone = mean yield in bs zone

Fail to reject

Ho: Mean yield in aw zone = mean yield in cw zone

Reject

Ho: Mean yield in bs zone = mean yield in cw zone

Reject

VARIABLE

2:

AVERAGE

PRECIPITATION

The amount and timing of rainfall obviously affects yields. In order to capture this relationship for use in empirical analysis, the HIID Integrated Data Set includes a measure of average monthly precipitation in 1987. There are two clear problems with this measure. First, it does not capture the effects of timing. Second, 1987 was a drought year in India and data from this period is therefore unrepresentative. Not surprisingly, average yearly precipitation for 1987 does not seem to be

25

related to foodgrain yields in the sample. Likewise, the crossstate temperature data is not informative. Given the limitations in the current Integrated Data Set, the World Bank India Agricultural Data Set was used instead to study district level rainfall and temperature effects. VARIABLE

3

:

ELEVATION

In many countries, cropping patterns are strongly linked with elevation. For example, in sub-Saharan Africa, one might observe those at the foot of mountain growing sorghum and millet while those higher up grow maize, vegetables, and beans. Aside from the Himalayas in the far north, India has relatively little variation in elevation. The HIID Integrated Data Set includes measures of mean elevation across states, based on GIS calculations. These state-level measures are not correlated with foodgrain yields. While there may be evidence within states of the effects of elevation, effects are not captured at the statelevel. VARIABLE

4:

DISTANCE

TO

THE

NEAREST

NAVIGABLE

RIVER

There is considerable variation in the distance from the centre of each state or union territory to the nearest navigable river. The average measure for Pondicherry is the closest at 1.15 km, and the farthest is for Jammu & Kashmir at 1189.72 km. One would expect that regions nearest navigable rivers might have an advantage in agricultural productivity due to superior sources of water for irrigation and, perhaps, cheaper access to agricultural inputs because of the case of transport on navigable rivers. Thus, regions nearest rivers might have higher yields on average than those farthest from rivers. This relationship is not supported in the Integrated Data Set, or in GIS maps. In fact, states farthest from navigable rivers seem to have higher yields than those that are closer. One possible explanation is that the distance to the nearest navigable variable is too imprecise for use at the cross-state level. It is likely that yields vary considerably within states in areas at different distances from rivers, but this relationship is not captured in state-level data.

26 VARIABLE

5:

SOIL

SUITABILITY

INDEX

It is clear that soils are important to agriculture. For example, one explanation of Africa’s poor agricultural growth is its poor soils. The Integrated Data Set includes two indices of Soil Suitability constructed for all crops. Soil Suitability Index # 2 has been found to be significant in some crosscountry studies. In studying India’s foodgrain yields, however, this index is found to be too aggregated to be at use. As the chart in Appendix B illustrates, states with the lowest soil suitability values have yields that are about the same as states with the highest suitability values.

27 V.

A

MODEL

OF

GEOGRAPHY

AND

FOODGRAIN

YIELDS

Question: Holding other factors of production constant, do Koeppen Zones have different effects on foodgrain yields? Answer:

Yes, Koeppen Zones have different effects on foodgrain yields. In India, dry and tropical zones have more positive effects than temperate zones. Using two empirical models that isolate the impact of geography on state foodgrain yields in India, this section shows that rain temperature and soils are important factors in agricultural productivity. It also shows that the agro-climatic conditions of dry and tropical zones have significantly more positive effects than those of temperate zones in India. Detailed rainfall and temperature analysis suggest that the temperate zone’s more volatile precipitation levels may help to explain why its yields are lowest.

This section presents two empirical models of foodgrain yield using the HIID Integrated Data Set and the World Bank India Agricultural Data Set. For each model, regression results are findings are presented, along with analysis of what drives simulated differences in mean yields between Koeppan Zones. The first model isolates the effects of Koeppen Zones on yields. The second model isolates the effects of rainfall and temperature across Koeppen Zones. MODEL #1: ISOLATING ZONES ON YIELDS

THE

EFFECTS

OF

KOEPPEN

Using the HIID Integrated Data Set, the model that provides the best fit for total foodgrain yield in 1991 includes measures for rural labor density, fertilizer use, NSDP in 1980 and the four primary Koeppen Zones represented in India. This model, shown in Figure 11,32 has an R2 value of 0.995 and an adjusted

32

In the equation, Y is the dependent variable, d is the constant term, Bx are the co efficients to be estimated, and E is the error term.

28

R value of 0.986, which suggest that it explains almost all of the variation in foodgrain yields across states in 1991. In addition, all of the seven variables included, as well as the constant term, have statistically significant effects on yields, even though the regression uses only 12 observations. The results of the regression are detailed in Figure 12. Figure

11:

The

model

in

Equation

Form

Log Y = α + βo* (log rural labor density) + β1* (Fertilizer use) + β2* (Lagged NSDP) + β 3 *(Tropical AM Zone) + β 4 *(Tropical AW Zone) + β 5* (Dry BS Zone) + β 6* (Temperate CW Zone) + ε Figure

12:

Regression

on

foodgrain

yield,

1991

(log of kg/ha)

Rural labor density, 1991 (log of persons/ha)

0.186* (0.014)

Fertilizer, 1991 (kg/ha)

0.005* (0.016)

Lagged NSDP (log of 1980-81 value in 1980 US$)

-0.270** (0.007)

Tropical AM Koeppen Zone

0.982** (0.003)

Tropical AW Koeppen Zone

-3.047* (0.014)

Dry BS Koeppen Zone

1.444** (0.007)

Temperate CW Koeppen Zone

0.998** (0.001)

Constant

7.832** (0.001)

Number of observations

12

R2 OLS regression with robust standard errors. P-values in parentheses. **=significant at 1% confidence. *=significant at 5% confidence.

0.995

29

A similar model of rice yield explains 93 per cent of the variation in rice yields in 1992. As shown in Figure 13, four of the seven variables and the constant are statistically significant. Models of wheat or maize yields for 1992, however, did not provide useful results, probably due to the lack of observations.33 Figure

13:

Regression

of

rice

yield,

1992

(log

of

kg/ha)

Rural Labor density, 1991 (log of persons/ha)

0.451** (0.003)

Irrigation, 1992 (net irrigated area as a percentage of net cropped area)

-0.013 (0.099)

Lagged NSDP (log of 1980-81 value in 1980 US$) (0.164)

0.177

Tropical AM Koeppen Zone

1.003* (0.039)

Tropical AW Koeppen Zone

-5.697* (0.049)

Dry BS Koeppen Zone

2.687* (0.023)

Temperate CW Koeppen Zone

0.565 (0.275)

Constant

8.370** (0.000)

Number of observations

12

R2

0.931

OLS regression with robust standard errors. P-values in parentheses. **=significant at 1% confidence. *=significant at 5% confidence.

33

In the Integrated Data Set, only 1980 and 1992 yields were available for rice, wheat and maize.

30 FINDINGS

:

Results suggest that states in the tropical AM zone, like Tamil Nadu, have an inherent geographic advantage in foodgrain and rice yields over those in the tropical AW zone, like Kerala. States in the dry BS zone, however, likely have an advantage over states in all other zones. The tropical AW zone is clearly the worst in all cases. The models using the Integrated Data Set show that most Koeppen Zone variables have a statistically significant impact on yields. This might reflect simply that rain, temperature and soil quality, the measures used in Koeppen Zone classification, are important in explaining yields. F-tests of joint significance were therefore used to test whether Koeppen Zones could be said to have distinct impacts on yields. The results show that most Koeppen Zones do indeed have distinct effects. A more formal presentation of F-test results is summarized in Figure 14. Figure 14: F-Tests of joint significance of coefficient estimates Null hypothesis

Reject? (i.e., is the difference of the estimates statistically significant?) Foodgrains

Rice

Reject** (0.010)

Reject* (0.045)

Could reject0 (0.070)

Reject* (0.019)

Ho:(Tropical AM zone) - Temperate CW zone) = 0

Fail to reject (0.8743)

Fail to reject (0.218)

Ho:(Tropical AW zone) - (Dry BS zone)=O

Reject** (0.010)

Reject* (0.038)

Ho:(Tropical AW zone) - Temperate CW zone) = 0

Reject**

Could reject0 (0.060)

Ho:(Tropical AM zone) - Tropical AW zone)=0 Ho:(Tropical AM zone) - (Dry BS zone)=0

(0.008) Ho:(Dry BS zone) - Temperate CW zone) = 0

Could reject (0.084)

Reject** (0.007)

P-Values in parentheses. **=Significant at 1%.. *=Significant at 5%. 0= Significant at 10%.

31

By substituting sample means values of each variable into the regressions, the model can be used to predict log of foodgrain and rice yields across Koeppen Zones.34 As shown in Figure 15, foodgrain and rice yields are highest in the dry zone in all three cases. Predictions are not included for the tropical Aw zone because data for rural labour density was unavailable.35 Figure

15:

Predicted

and

actual

mean

yields

(log of kg/ha)

Tropical AM

Dry BS

Temperate

Overall

C W

Foodgrain

7.172

7.590

7.502

7.152

Regression

(7.309)

(7.442)

(7.214)

(7.264)

Rice Yield

6.926

7.822

6.874

6.631

Regression

(7.671)

(7.860)

(7.158)

(7.477)

Actual Means in Parentheses What Drives Differences in Yields Across Koeppen Zones?

These regressions suggest that differences in foodgrains yields between zones are driven largely by agro-climatic factors. In particular, dry zones have higher yields than temperate and tropical zones mainly due to characteristics of their Koeppen Zone. Using the foodgrain yield regression, Figure 16 shows the factors that contribute to the differences in predicted foodgrain yields. For example, the -0.17 value for rural labor density in the “Dry BS-Temperate CW” column is the product of the difference between the mean values of rural labor density in each zone times the coefficient estimate generated in the foodgrain regression. One weakness of this analysis is that it yields ambiguous results about the relationship between the tropical AM and 34

Values are calculated using the mean value for each variable in each of the three Koeppen Zones for which data is available. The percentage of land in each Koeppen Zone in each zone is used, rather than a value of 1 or 0 as many states have at least some land in different zones.

35

Two of the mean values necessary for calculation are missing for the AW zone.

32

temperate CW zones. As shown above, the differences in the coefficient estimates for the two variables are not statistically significant. In addition, the actual means in the tropical AM zone were higher than those in the temperate CW zone, while the foodgrain model predicts higher yields for the temperate CW zone. The relationship among the effects of Koeppen Zones is examined further in the more complete World Bank Data Set for the period 1967-1986. Figure

16:

Components

Differences

in

of

Simulated

means

(Using Foodgrain Yield Regression) Dry BSTemperate C W

Tropical AMTemperate C W

Dry BSTropical A M

-0.17

-0.26

0.09

0.45

0.24

0.21

-0.44

-0.34

-0.10

Tropical AM Koeppen Zone

0.05

0.75

-0.70

Tropical AW Koeppen Zone

-0.02

-0.03

0.00

1.02

0.11

0.91

-0.80

-0.80

0.00

0.00

0.00

0.00

0.09

-0.3336

0.42

Rural labor density, 1991 (log of persons/ha) Fertilizer, 1991 (kg/ha) Lagged NSDP (log of 1980-81 value in 1980 US$)

Dry BS Koeppen Zone Temperate CW Koeppen Zone Constant Predicted difference Other

Factors

that

Impact

Yields

Rural labor density was estimated as the rural population

(in 100,000s) divided by land area available for cultivation (ha). Results are mixed across the two regressions. In the foodgrain regressions, increased density is related to higher yields. In the rice yield regression, the reverse holds, but is not statistically significant. This discrepancy may reflect differences in the types of foodgrains included in each regression. While increased labor 36.

Yields in the tropical AM region are actually higher than those in temperate CW region.

33

density may be related to decreased rice yields, for example, it might have the opposite effect for other foodgrain crops. Fertilizer and irrigation were used as a rough estimate

of agricultural inputs and technology.37 Fertilizer has a positive impact on foodgrain yields for states in any zone. The results for irrigation in the rice regression are not statistically significant. This may be due to the large proportion of rain-fed irrigation in India, the effects of which will be picked up by the Koeppen Zone variables. In rice regressions, fertilizer use was unpredictive, but irrigation levels were useful in the model. This difference may be due to a variety of factors including lower requirements for fertilizer in rice cultivation and higher requirements for irrigation, or lower responsiveness of modern rice varieties, as compared to other foodgrains, to fertilizer use. Lagged NSDP was included in both regressions in order

to account for the effects of income levels across states, and the 1980 NSDP value was used instead of the current figure to avoid the problem of reverse causality.38 In the foodgrain model, lagged NSDP has a significant and negative relationship with yields. The negative relationship might seem to run counter to conventional wisdom that richer states are better able to afford technology and inputs to agriculture making for higher yields. However, it might be explained in that richer states may have shifted resources from agriculture to manufacturing or services and raised the costs of labor and capital for agriculture. Lagged NSDP may, in fact, have a positive effect on yield per person, even though it has a negative value for yield per hectare, the measure used here.

37.

See, for example, Sanderson and Roy, “Fertilizers, HYVs and irrigation were found to be highly intercorrelated, so fertilizers are used as a proxy variable to represent the entire package of modern technological,” p.22.

38.

As foodgrain production clearly contributes to NSDP, the NSDP in 1991 is probably a function of foodgrain productivity in 1991. Thus, the NSDP in 1991 is endogenous, i.e., determined within the model and should not be included in the regression.

34 MODEL AND

#2: ISOLATING TEMPERATURE

THE EFFECTS OF ACROSS KOEPPEN

RAINFALL ZONES

Because adequate rainfall and temperature data are not available in the HIID Integrated data Set, a similar model was developed using the World Bank India Agricultural Data Set to isolate the effects of the different components of Koeppen Zones on rice, wheat and maize yields during the period 1967-1986.39 The components of Koeppen Zone studied include mean monthly temperature, mean monthly rainfall, aquifers and soil type. Model #2 also controlled for non-geographic variables not available in the Integrated Data Set. Figure 17 summarizes the variables used in the best-fit model and Figure 18 shows the actual and predicted mean yields to illustrate that the model is quite predictive. (For regression results, see Appendix C).

39.

The model generally followed the model for cross-state variations in yield described above. The regression results are presented in Appendix C.

35 Figure

17:

Variables

HYVs for each crop (percentage of gross cropped area)

Included

In

The

Model

For

1967-86

Literacy (log of percent- Precipitation2 age for rural males)

Tractors (log of units/ha) Temperature (log of 0C Aquifers (150m thick)40 Fertilizer (nitrogen, po- Temperature 2 tassium, and phosphorus in log of tons/ha)

Soils (laterite, red & yellow, shallow black & medium black)41

Rural labor (log of per- Precipitation (log of mm sons/ha) for monthly mean in growing season)

Figure

18:

Predicted (Log

(WORLD

and of

Actual

.

Mean

Yields

KG/Ha)

BANK INDIA AGRICULTURAL DATA SET) Tropical AM

Dry BS

Temperate

Rice

7.12 (7.23)

7.11 (7.29)

6.49 (6.67)

Wheat

6.78 (6.54)

7.06 (7.08)

6.90 (6.96)

Maize

6.99 (6.86)

7.10 (7.13)

6.56 (6.76)

CW

Actual means in parentheses.

40

Aquifer levels measures ground water.

41

The storie Index, an overall measure of soil productivity, was not used because it is also considered a proxy of temperature and rain.

36 Additional

Variables

Used

in

Model

#

2

Although the models from the two data sets are generally similar, they differed on certain variables due to the different data contained in each set. In addition to fertilizers, Model #2 also controlled for HYVs and tractors since so much attention has been paid to Green Revolution technology in India. Irrigation was not controlled for, as the aquifer and precipitation variables seemed to pick up the effects of India’s groundwater and rain-fed irrigation. The labor variable in Model# 2 is weighted by the number of days worked by rural males whose primary job classification is agricultural labor or cultivation. Model #2 also controlled for literacy as a proxy for the ability to adopt new technology. Although infrastructure could improve access to inputs and markets, and thus yields, roads and distance from sea were found to be insignificant in Model# 2, and were thus left out of the model. Temperature and precipitation measurement posed one of the biggest challenges in building Model# 2. Cropping calendars contain a complex mix of information that depends not only on absolute levels, but also on timing, seasons and crop needs. Temperature and precipitation effects can vary even between neighbouring farms. The most predictive model used the log of the monthly mean temperatures and the temperature squared for each month in the growing season, as well as the log of the monthly mean precipitation level and the precipitation squard for each month in the growing season. The growing season is defined when temperature is greater than 50C.42. A comparison of actual mean crop yields with predicted mean crop yields shown above illustrates that the model is predictive. Several different methods of including temperature and precipitation were tested, including the World Bank’s approach of four evenly spaced months, annual averages, minimums and maximums, and seasonal averages.

42

Gallup, 4.

37 Temperature Yields

and

across

Precipitation Koeppen

Drive

Differences

in

Zones.

As in the regression using the Integrated Data Set, the regressions using the World Bank Data Set show that climate and precipitation variables have a significant impact on foodgrain yields in India. In particular, temperature and precipitation differences across Koeppen Zones appear to be the largest drivers of rice, wheat, and maize yield differences between zones. They dry BS zone seems to have the most favourable temperature and precipitation patterns for all three crops. Rice.—As shown in Figure 19, tropical and dry zones have

higher rice yields than temperate zones. Temperature and precipitation patterns seem largely responsible for this difference. Dry zones have higher rice yields than tropical zones, eventhough the model predicts slightly higher tropical yields. Neither temperature nor precipitation appear to play the most important role in the difference between tropical and dry zone yields. Wheat.— As shown in Figure 20, dry zones have higher wheat yields than temperate zones. Precipitation patters seem largely responsible for this, despite the favourable temperature patterns in temperate zones. As expected, tropical zones do not produce much wheat due to their precipitation patterns. Maize.— As shown in Figure 21, dry zones have higher

maize yields than temperate zones. Again, temperature and precipitation patterns seem largely responsible for this difference. Tropical zones also have higher maize yields than temperate zones. Temperature patterns seem almost solely responsible for this difference. Dry zones have higher maize yields than tropical zones. Precipitation patterns seem largely responsible for this difference, despite the favourable temperature patterns in tropical zones.

38 Figure

19:

Rice:

Components

Differences

(WORLD

BANK

Simulated

Means

DATA

SET)

Dry BS Temperate C W

Tropical AMTemperate C W

Dry BSTropical Am 43

HYVs for each crop (per cent of gross cropped area)

0.01

0.05

(-)0.04

Rural labor (log of persons/ha)

0.03

(-)0.02

0.04

Tractors (log of units/ha)

0.01

(-)0.02

0.03

0.08

0.10

(-)0.02

Literacy (log of per cent for rural males)

0.01

0.00

0.00

Temperature (log of oC for monthly mean in growing season and value squared)

0.22

0.26

(-)0.04

Precipitation (log of mm for monthly mean in growing season and value squared)

0.24

0.21

0.03

Soils (laterite, red & yellow, shallow black, & medium black)

0.00

0.04

(-)0.03

Aquifers (150m thick)

0.01

0.00

0.01

Predicted difference

0.62

0.63

(-)0.01

Actual difference

0.62

0.56

0.06

Fertilizer (nitrogen, potassium and phosphorus in log of tons/ha)

43

In

of

The model predicts higher rice yields in the tropical zone, eventhough they are actually higher in the dry zone.

39 Figure

20:

Wheat:

Components

Differences

(WORLD

BANK

In

of

Simulated

Means

DATA

SET)

Dry BS Temperate C W

Tropical AMTemperate C W

Dry BSTropical A M

HYVs for each crop (per cent of gross cropped area)

(-)0.02

(-)0.09

0.06

Rural labor (log of persons/ha)

0.00

0.00

(-)0.01

Tractors (log of units/ha)

0.00

0.01

(-)0.01

Fertilizer (nitrogen, potassium and phosphorus in log of tons/ha)

0.06

0.07

(-)0.01

Literacy (log of per cent for rural males)

0.02

0.01

0.01

Temperature (log of 0C for monthly mean in growing season and value squared)

(-)0.15

0.03

(-)0.18

Precipitation (log of mm for monthly mean in growing season and value squared)

0.23

(-)0.08

0.32

Soils (laterite, red & yellow, shallow black, & medium black)

0.03

(-)0.07

0.10

Aquifers (>100m, 100-150m, and >150m thick)

0.00

(-)0.01

0.00

Predicted difference

0.16

(-)0.12

0.28

Actual difference

0.12

(-)0.42

0.54

40 Figure

21:

Maize:

Components

Differences

(WORLD

BANK

In

of

Simulated

Means

DATA

SET)

Dry BS- Tropical AMTemperate Temperate C W C W

Dry BSTropical A M

HYVs for each crop (per cent of gross cropped area)

0.01

0.00

0.01

Rural labor (log of persons/ha)

0.04

(-)0.03

0.07

Tractors (log of units/ha)

0.01

(-)0.02

0.02

Fertilizer (nitrogen, potassium and phosphorus in log of tons/ha)

0.08

0.00

0.09

Literacy (log of per cent for rural males)

0.02

0.01

0.01

Temperature (log of 0C for monthly mean in growing season and value squared)

0.25

0.47

(-)0.22

Precipitation (log of mm for monthly mean in growing season and value squared)

0.23

(-)0.02

0.25

Soils (laterite, red & yellow, shallow black & medium black)

(-)0.02

(-)0.02

0.00

Aquifers (>100m, 100-150m, and >150m thick)

(-)0.02

(-)0.02

0.00

Predicted difference

0.60

0.38

0.22

Actual difference

0.37

0.10

0.27

What

Makes

tation

So

the

Impacts

of

Temperature

and

Precipi-

High?

The impacts of temperature and precipitation were calculated as a product of the difference in mean monthly temperatures or precipitation between zones, times the co-efficient estimates for each monthly temperature or precipitation varilable. Thus, a high impact of temperature or precipitation can be due to either a high difference in means or to a high co-efficient.

41 The impact of temperature is especially significant in

driving temperate rice and maize yields below those in the dry and tropical zones. As shown in Figure 22, mean temperatures are not too different between the three zones. Nevertheless, temperature patterns in the temperate zones are the most distinct of the three zones. Mean tropical temperatures differ from mean dry temperatures by only 4 percent, while they differ from mean temperate temperatures by 8 percent. This is slightly couter-intuitive, as one might expect dry and tropical zones to have the largest differences in temperature. Figure

22:

Average Zones

Difference 0

C

in

in

absolute

Temperature

between

value)

Dry BSTemperate C W

Tropical AMTemperate A W

Dry BSTropical AM

1.64

2.21

1.08

The larger difference between tropical and temperate temperatures can be attributed to the higher volatility in temperate zones during summer and winter months, as shown in Figure 23. This volatility in temperate zones may have a negative impact on farmer’s ability to predict weather patterns accurately and may not be conducive to the use of some rice and maize varieties. Figure

23:

Average

Temperature,

1967-86

42 The impact of precipitation is especially significant in

driving temperate rice, wheat, and maize yields below those in the dry zones. Precipitation is also significant in driving temperate rice yields below those in the tropical zone. As with temperature, a closer look shows that precipitation patterns in the temperate zones are the most distinct of the three zones as shown in Figure 24. Mean tropical precipitation differs from mean dry precipitation by 44 percent, while it differs from mean temperate precipitation by 35 percent. Again, this is slightly counter-intuitive, as one might expect dry and tropical zones to have the largest differences in precipitation.

43 Figure

24:

Average Zones

Difference mm

in

in

Precipitation

absolute

between

value)

Dry BSTemperate C W

Tropical AMTemperate C W

Dry BSTropical AM

43.46

49.88

39.57

The larger difference between tropical and temperate precipitation can be attributed to the higher volatility in temperate zones during the monsson season, as shown in 25. This volatility in temperate zones may have a negative impact farmers’ ability to predict weather patterns accurately and may not be conducive to the use of some rice, wheat, and maize varieties. The impact of precipitation is also significant in driving tropical wheat and maize yields below those in the dry zones. This can be attributed to the higher levels of precipitation in tropical zones, which do not seem well suited for the production of wheat and maize.

Figure

25:

Average

precipitation,

1967-1986

44 Other

Factors

that

Impact

Yield

Fertilizer: As with Model # 1, Model # 2 shows that fertilizer is also a key factor in driving differences in yields across Koeppen Zones. The impact is especially apparent in driving temperate yields below dry yields for rice, wheat, and maize. Fertilizer is also significant in driving temperate rice yields below tropical rice yields. Fertilizer is also significant in driving tropical maize yields below dry maize yields. Labor: Labor differences across zones appear to cause some differences in rice and maize yields across zones, but have little impact on differences in wheat yields. The lower labor density in dry zones appears to help drive its rice and maize yields above those in temperate zones.

HYVs, tractors, soil type, fertilier, and soil type differences are about as important as temperature and precipitation in driving the differences between dry and tropical rice yields.

45 VI.

ADDITIONAL

DIFFERENCES

KOEPPEN

ACROSS

ZONES.

Question: Do non-geographic factors of foodgrain production vary across Koeppen Zones? Answer:

Yes, agricultural inputs, technology, and some economic and demographic characteristics vary across Koeppen Zones. Dry zones generally have the most favourable indicators. A preliminary empirical model of fertilizer use suggests that dry and tropical zones, as compared to temperate zones, have the most positive effects on fertilizer use in India. Such variations across Koeppen Zones may be due to a variety of factors and should be explored further in any effort to equalize yields across states.

This section first presents statistics showing how levels of agricultural inputs, technology, and some economic and demographic indicators vary across Koeppen Zones. It then presents and interprets a preliminary empirical model of the effects of geography on fertilizer use. COMPARISON MINANTS

OF ACROSS

NON-GEOGRAPHIC CLIMATE

DETER-

ZONES.

Important determinants of agricultural productivity such as irrigation, tractors, fertilizer use, and rural literacy are all highest in the dry zones according to the HIID Integrated Data Set, as shown in Figure 26. See Appendix A for maps on irrigation and crop yields.

46 Figure

26:

Average Koeppen

Input Zones

Levels

Across

#1

(Source: HIID Integrated Data Set) Tropical A M Fertilizer use, 1990-96 (Kg/ha) Tractors, 1981, ‘86 & ’91 (units)

77.786

Tropical A W

67.869

Dry BS

129.327

17876.240 1750.667 65869.130

Temperate C W

35.817

Overall

Highest

63.511

Dry

42218.820 39953.000

Dry

Irrigation, 1970, ‘80, & ‘92 (gross irrigated area as percent of gross cropped area)

24.999

14.807

49.171

22.338

28.253

Dry

Irrigation, 1970, ‘80, & ‘92 (net irrigated area as per cent of net sown area)

25.308

15.233

50.931

25.708

30.219

Dry

0.010

0.016

0.013

perate

Rural population density, 199144 (people in 100, 000/ha)

Tem0.003

Similar results are found in the World Bank India Agricultural Data Set for 1967-86 as shown in Figure 27. T-tests of the differences in means were run to compare Koeppen Zone averages. The null hypothesis that the means of nitrogen and potassium fertilizer use in the tropical versus the dry zones are equal can be rejected at the 5 per cent level. All other differences in means are significant at the 1 per cent level.

44

Variable is calculated as rural population divided by land available for cultivation.

47 Figure

27:

Average

Input

Levels

Across

Koeppen

Zones # 2

(Source: World Bank India Agricultural Data Set, 1967-86) Tropical AM

Temperate C W

Dry BS

Highest Zone

Pumps (units/ha)

9.292

9.246

17.967

Dry

Tractors (unit/ha)

0.598

1.500

2.470

Dry

2553.108

1100.771

2743.271

Dry

Literacy (per cent of rural males)

0.341

0.288

0.378

Dry

Irrigation (per cent of gross cropped area)

0.306

0.239

0.275

Tropical

154,953.5 114,278.2

Tropical

Roads (kms)

Rural labor (persons/ha)

202,342.2

Population density

3.619

2.794

94.310

397.210

HYV rice (per cent of gross cropped area)

0.124

0.032

0.048

HYV wheat (per cent of gross cropped area)

0.007

0.084

0.063 Temperate

HYV maize (per cent of gross cropped area)

0.004

0.005

0.007

Dry

Fertilizer: Nitrogen (tons/ha)

18.11

11.95

20.19

Dry

Fertilizer: Phosphorous (tons/ha)

6.42

3.37

7.84

Dry

Fertilizer: Potassium (tons/ha)

3.95

1.20

3.38

Tropical

Distance from sea (kms)

Relationship

Between

Indicators

and

2.059

Tropical

199.690 Temperate Tropical

Yields

According to both data sets, the dry BS zone has, on average, used more pumps, tractors, and nitrogen and phosphorus fertilizers than the other two zones. It also has the highest literacy rates, a rough indicator of how easily the population will be able to adapt new technologies, which may

48

require basic literacy or numeracy and familiarity or contact with non-traditional institutions. Acccording to the HIID Integrated Data Set, irrigation intensity in 1980, 1990, and 1992 was highest in dry states. However, according to the world Bank Data Set irrigation between 1967 and 1986 was highest in tropical states. The discrepancy between these figures may be explained by the World Bank Data Set’s incomplete coverage of all states or by the different time periods. HYVs: High crop yields are often assumed to be caused

primarily by HYV use. However, although wheat and rice yields were highest in dry states between 1967-86, HYV wheat use was actually highest in temperate states and HYV rice use was highest in tropical states. Population density: While the relationship between inputs and technology and yields seems quite clear, the relationship between population density and yields is less clear. Classical economic theory suggests that agricultural expansion along the internal margin results in more labor-intensive agricultural methods with higher yields. Ester Boserup has argued that agricultural yields may increase in response to population growth. However, despite the higher yields in dry states, population density was found to be highest in tropical states from 1967-86, and rural population density was found to be highest in temperate states in 1991. A

MODEL ZONES.

OF

FERTILIZER

USE

AND

KOEPPEN

Given that the levels of agricultural inputs and technology vary greatly across climate zones, it seem likely that geography might affect agricultural yields indirectly through these factors as well as those described in the previous section. Results of a simple model of fertilizer using the Integrated Data Set are shown in Figure 28. The model predicts about 86 per cent of the variations in fertilizer use. See Appendix A for maps on fertilizer consumption across states and Koeppen zones.

49 FIGURE

28:

REGRESSION

OF

FERTILIZER

USE

1992

(KG/HA)

Irrigation, 1992 (net irrigated area as a per cent of net sown area)

1.489** (0.000)

Lagged NSDP (log of 1980-81 value in 1980 US $)

1.083 (0.661)

Tropical AM zone

54.653* (0.045)

Tropical AW zone

72.650** (0.010)

Dry BS zone

68.890 (0.066)

Temperate CW zone

4.080 (0.851)

Constant

Number of observations R2

-27.987 (0.205) 25 0.864

OLS regression with robust standard errors. P-values in parentheses. *=significant at 1% **=significant at 5%.

F-tests on joint significance, shown in figure 29, illustrate that the differences in all estimates are significant, except for the difference between the AM and BS zones, both of which have very large “effects” of fertilizer use. Clearly, the temperate zone has a much lower positive effect on fertilizer use than do any of the other zones.

50 Figure 29: F-Tests of Joint Significance of coefficient

estimates

Null hypothesis

Reject? (I.e., is the difference of the estimates significant?)

Ho:(Trophical AM zone)—((Trophical AW zone)=0 Ho:(Trophical AM zone)—(Dry BS. zone)=0

Reject* (0.033) Fail to reject (0.102)

Ho:(Trophical AM zone)—(Temperate CW zone)=0

Reject** (0.006)

Ho:(Trophical AW zone)—(Dry BS zone)=0

Reject* (0.019)

Ho:(Trophical AW zone)—(Temperate CW zone)=0

Reject** (0.000)

Ho:(Dry BS zone)—(Temperate CW zone)=0

Reject* (0.020)

P-Values in parentheses.

Why

Might

Geography

Affect

Fertilizer

Use?

There are two main possibilities. First, fertilizer might work best in some climate zones. This may be due to the responsiveness of the dominant crop type to fertilizers or to whether HYVs are used which may require more fertilizers. Second, fertilizer subsidies and related policy may somehow differ across geographic zones and affect levels of fertilizer use. These findings may have important implications for increasing policy makers’ understanding of the channels through which geography impacts yields. More detailed analyses should be built on the simple model presented here.

51 VII.

POLICY

RECOMMENDATIONS

The following recommendations are options for the Tamil Nadu Government to incorporate into its existing agricultural policy. These options have been chosen with budget constrainst in mind, and are thus least-cost options. Moreover, they do not require new institutions or capacities. Rather, they involve strengthening existing institutions. 1.

INCLUDE ANALYSES

GEOGRAPHIC OF TAMIL

FACTORS IN ECONOMIC NADU’S AGRICULTURE

Cross-national studies show that geography affects growth through agricultural productivity. This study confirms that geography impacts state-level variations in agricultural productivity in India. Variations in precipitation, temperature and soil quality affect variations in foodgrain yields directly and possibly, indirectly through the effectiveness of technologies and inputs. Future analyses of economic growth and agricultural productivity should, therefore, include geography in order to fully understand agricultural variations and policy options. Analysis

at

the

District

level

Understanding the geography-related causes of variations in foodgrain yields within Tamil Nadu can assist the State Government in determining appropriate investments in geographically advantaged and disadvantaged districts. The Government can apply the methods described in this report to cross district data. Preliminary analyses were made at the district level for Tamil Nadu using the methods and data described in this report. The results show that temperature and precipitation do indeed have significant impacts on rice yields in Tamil Nadu. (See Appendix D for regression results.) There have been some attempts to do this using data from village surveys. A 1992 study of irrigation, HYV rice, and income distribution in Tamil Nadu by C. Ramasamy, P. Paramasivam and A. Kandaswamy, for example, finds that variation in rice yields across villages“strongly suggest that degree of water control is the decisive factor influencing rice

52

yields”45 The degree of water control, they find, varies across environments in different administrative districts.46 The Tamil Nadu Government might build on this village-level data or use existing cross-district data sources. Analyses

at

the

Regional

and

national

levels

Geographic analyses on agricultural productivity can also assist the State Government to better understand its geographic advantages and disadvantages in relation to other states, particularly other Southern states. Knowledge of the state’s disadvantages in agricultural productivity can help it to design appropriate investments and policies with other states and with the national government. For example, in 1995, Tamil Nadu’s agricultural sector suffered from unexpected low levels of water. Water disputes led to the “non-release of water” by Karnataka from the Cauvery, which had a large negative effect on Tamil Nadu’s agricultural productivity.47 A thorough understanding of geographic disadvantages and their impact on agriculture might have helped the State Government push for appropriate policies in advance and thus prevent such situations. 2.EVALUATE

THE

AGRICULTURAL

EFFECTS INPUT

AGRO-CLIMATIC ACROSS

OF

POLICIES

PRODUCTION

TAMIL ON

NADU’S DIFFERENT

ENVIRONMENTS

DISTRICTS

This study finds significant variations in input levels across koeppen Zones in India. Similarly, Ramasamy,. Paramasivam and Kandaswamy’s study shows differences in input levels across agro-climatic zones within Tamil Nadu.48 Input policies can greatly affect such geographic variations in input use. Thus evaluations on input policies should include separate and complete analyses for both favourable and unfavourable agro-climatic environments. 45

Ramasamy, C.P. Paramasivam, and A. Kandaswamy. “Irrigation Quality, Modern Variety Adoption, and Income Distribution: The Case of Tamil Nadu In India,” in Modern rice Technology and Income Distribution in Asia eds., David, Cristina and Keijiro Otsuka (Boulder & London: Lynne rienner Publishers, 1994), 331.

46

Ibid, 325-326

47

Government of Tamil Nadu, Tamil Nadu-An Economic Appraisal (Chennai: Government of Tamil Nadu, 1996).

48

Ramasamy,

329.

53

Both Tamil Nadu and the national government have recognized the importance of agricultural inputs to yields. In particular, government policies have attempted to increase agricultural input levels through subsidies for fertilizer, irrigation, water and power, as well as through infrastructure projects. These policies are complex and spark considerable debate. This analysis does not attempt to take sides on this debate. Rather, this analysis suggests a new way to evaluate these policies to include the impact of geography on input use. If input levels vary significantly across environments, and input levels are linked both to agricultural productivity and to income, unfavourable environments can be targeted to meet the goals of increasing growth and reducing poverty. Favourable environments may also be targeted to ensure the highest returns on input investments. Agricultural input policies should then be judged successful if they increase input levels and yields in unfavourable environments, as well as in favourable ones. The first step to this process, as described in Recommendation#1, will be to identify which areas in Tamil Nadu have geographic advantages and disadvantages for agriculture. Then the Government can determine the true effect of input policies in various regions. 3.

ENCOURAGE TECHNOLOGIES

RESEARCH ADAPTED

TO

ON TAMIL

NEW NADU’S

G E O G R A P H Y

This study finds that geographic variables are important determinants of yields. The State Government of Tamil Nadu cannot, of course, change the state’s geographic profile. It can, however, increase yields by encouraging public, private, and international research on agricultural technologies that are wellsuited to the state’s geography. India has one of the largest publicly funded agricultural research systems in the world. According to a recent study by Robert Evenson, Carl Pray and Mark Rosegrant, public agricultural research accounted for nearly 40 per cent of total

54

factor productivity growth in Indian agriculture between 1956 and 1987.49 Since the mid-1970s however, public research growth has experienced regional variation. Much of the nationallevel research funding from the Indian Council of Agricultural Research (ICAR) is channeled through state governments and state agricultural universities. Although universities cover all regions in India, agricultural research in the North enjoys the most resources in terms of levels and as a per cent of agricultural GDP. This can largely be attributed to the strong state support for research in the North.50 The Tamil Nadu Government should encourage more agricultural research for the South. Increased research in the South can help ensure research on technology that is fit for the South’s geographic profile. Currently, Tamil Nadu has the lowest levels of expenditure on agricultural research of the southern states, although its expenditure as a percentage of agricultural GDP is second only to Kerala. To help promote agricultural research for the South’s geography, the State Government can: *

Push for more ICAR resources toward the South’s state agricultural universities.

*

Expand existing public programs, such as the RiceWheat Consortium, a program of the CGIAR, to move beyond the Indo-Gangetic Plain to also include tropical regions.

*

Increase ties with existing International Institutions, such as International Crops Research Institute for the SemiArid Tropics and the International Rice Research Institute (IRRI), based in Andhra Pradesh.

*

Encourage private sector research. Private research has increased dramatically since the 1960s, it is highly efficient, and it imposes no burden on the public budget.

49

Evenson, Robert, Carl Pray and Mark Rosegrant, “Agricultural Research and Productivity Growth in India” (Washington, DC: International Food Policy Research Institute, 1999). 1-5.

50

Ibid, 5-26.

55

*

Increase ties with the national Agricultural Research System (NARS). NARS connects all ICAR institutes, state agricultural universities, and zonal research centres established under the National Agricultural Research Project (NARP).

*

Increase ties with central government departments that provide some funding for agricultural research, such as the Ministry of Commerce or the Ministry of Science and Technology.

The state might also refer to existing analyses of how to prioritize research to correct for Tamil Nadu’s geographic disadvantages. A possible starting point is the comprehensive analysis of rice research priorities in Southern India done by C.Ramasamy, T.R. Shanmugham, and D. Suresh of the Tamil Nadu Agricultural University. The study first identified “technical constraints” to yields, including insects and pests; diseases; soils/agronomy; genetic/physiological; and climatic and environmental factors. It then estimated the severity of specific constraints and potential benefits to “solving” them against the research costs of “solving” them. according to this study, Tamil Nadu’s top five yield constraints are leaf folder, ear head bug, fertilizer imbalance, rice blast, and water management51. 4. S U P P O R T

THE

TECHNOLOGIES

ADOPTION SUITED

TO

OF TAMIL

EXISTING NADU’S

G E O G R A P H Y

In addition to promoting new technology through increased research, the State Government can help increase yields by promoting the adoption of existing technology that is well suited to Tamil Nadu’s geography. To this end, the Tamil Nadu Government can review and evaluate existing technologies and their varying success in different geographic zones. Through the use of demonstration sites at the district level, the Government can increase farmers’ awareness of which technologies are most appropriate for various geographic characteristics. Because HYV 51

Ramasamy, C,. T.R. Shanmugham, and D. Suresh “Constraints to Higher rice Yields in different Rice Production Environments and prioritization of Rice Research in Southern India, “in Rice Research in Asia: Progress and Priorities,

56

success and adoption are so closely linked with water availability, the Tamil Nadu Government should continue its focus on supporting irrigation projects throughout the state. Further, the government can facilitate the adoption HYVs that are fit for the geographic conditions in Tamil Nadu simply by urging the national government to make them available on the market for purchase by farmers. Currently, India’s seed industry is highly regulated and many HYVs available in other tropical regions are not available in Tamil Nadu or in India. The Indian government expends significant resources in testing HYVs to find the best ones to release in the market, but this process can be time consuming. For example, the ICAR’s Rice Improvement Programme took from 1991-1995 to release just four high yield rice varieties for use in Karnataka, Tamil Nadu, and Andhra Pradesh52. The Tamil Nadu Government should support efforts to hasten the government’s release of HYVs that are suited to use in tropical regions. 5. A D D R E S S CAUSED

CONCERNS BY

TAMIL

OF NADU’S

AGRICULTURAL

RISK

CLIMATE

This study clearly illustrates that rainfall and temperature play important roles in agricultural yields. Moreover, volatile rainfall and temperature seem to have negative impacts on yields in India. The “Production risk” due to weather is peculiar to the agricultural sector as compared to industry53. Uncertainty from volatile weather can hurt poor farmers who lack savings and credit to protect them in bad seasons. In addition, uncertainty can lead farmers to give up higher yields in return for certainty, and can thus slow growth in the agricultural sector. To reduce risk, for example, many farmers plant two different crops with different rainfall needs, thereby reducing overall yields and profits. In addition, Many farmers opt for traditional seeds over HYVs, because HYVs have higher variability in yields. The State Government should help address farmers’ concerns eds. Evenson, R.E. and R.W. Herdt with M. Hossain (CAB International, 1996) 146-160. 52

53

Food and Agriculture Organisation, “Support to Indias, Rice improvement Programme — IND/91/008,”http://www.fao.org/ag/agp/agpc/promo/india.htm. The Programme was initiated in 1989 and approved for funding by the UNDP/FAO in1991. Mishra, Pramod K., Agricultural risk, Insurance and Income (Aldershot: Avebury, 1996), 11.

57

of agricultural risk caused by climate to maximize farmers’ efforts to attain higher yields. Two

Options

to

Reduce

Production

Risks

First, the Government can reduce the costs of uncertainty about weather by increasing information. As the Food and Agriculture organization notes, “Farmers in developing and tropical countries could benefit enormously from access to interpreted agrometeorological data and improved seasonal forecasts. An accurate idea of the rainfall already in the soil, as well as that confidently predicted, would allow them to plant the right crop at the right time”.54 In Mali, for example, under the AGRHYMET program, extension workers have used accurate weather information to advise farmers about what to plan and when. An International Workshop on Agrometeorology in the Twenty-First Century, in February 1991, identified related priority areas that are relevant to Tamil Nadu. These include the support of networks of climatological stations and the provision of “training in agrometeorology to all agriculture professionals with secondary school or higher education”55. Second, the State Government should build on existing institutions to improve access to insurance and credit to farmers. Tamil Nadu currently has the largest credit/deposit ratio in the country. The Government of Tamil Nadu can work through the Comprehensive Crop Insurance Scheme (CCIS), which the Indian national government introduced in 1985-86 and operates with the assistance of state governments through the General Insurance Corporation of India. A significant limitation of the CCIS for Tamil Nadu is its selective coverage of crops, not including Tamil Nadu’s principal crop, rice. Another drawback is that state supported insurance and credit may pose a significant drain on the public budget.56 54

Food and Agriculture Organisation, “Weather-wise farmers could improve yields,” March 30, 1999, http://www.fao.org/news/1999/990307-ehtm, 2. For further information, contact Rene.Gommes @fao.org.

55

LIbid, 2.

56

For further information on the CCIS, see Mishra, Pramod K. Agricultural Risk, Insurance and Income: A study of the Impact and Design of India’s Comprehensive Crop Insurance Scheme (Aldershot: Avebury, 1996).

58

The State Government can also increase ties with the growing body of private and non-governmental organizations (NGOs) involved in credit and insurance initiatives. The private sector has access to efficient management techniques, experience in lending, and substantial resources. The NGOs have access to large community groups, experience with capacity building, and local training initiatives that maximize the use of credit services. The partnership between the private and NGO groups is growing rapidly across the world. The State Government should make concerted efforts to remain a key actor in this partnership. 6. C O N T I N U E

INVESTMENTS

MANUFACTURING

AND

IN TRADE

TAMIL

NADU’S

SECTORS

This report argues that geography has significant effects on agricultural productivity, even holding constant income and other factors of production. Technology and higher input levels may have only limited impact on mitigating geographic effects on agricultural yields. Tamil Nadu, along with other tropical AM and AW states, is at a slight disadvantage in agricultural productivity compared to dry BS states. In the long run, Tamil Nadu may naturally shift away from agriculture to other, more profitable sectors of the economy, like manufacturing. Tamil Nadu already appears to be making this transition. The State Government should continue to support it through investments in industry and trade. At present, however, the Tamil Nadu Government is well advised to continue efforts toward increased understanding of agriculture and well-targeted investments. The agricultural sector has significant effects on rural poverty and on state and national food security. Growth in agriculture provides a firm basis for both economic growth and poverty alleviation.

59 BIBLIOGRAPHY

Athreya, Venkatesh and Sheela Rani Chunkath. “Gender and Infant Survival in Rural Tamil Nadu: Situation and Strategy.” Economic and Political Weekly (October 3, 1998): 2588-2594. Bajpai, Nirupam and Jeffrey D. Sachs. “Strengthening India’s Strategy for Economic Growth.” Harvard Institute for International Development (July 1998). Development Discussion Paper No. 641. Bhalla, G.S. and Peter Hazell, “Foodgrains Demand in India to 2020: A Preliminary Exercise.” Economic and Political Weekly (December 27, 1997): A150-A154. Bhalla, G.S. and Gurmail Singh. “Recent Developments in Indian Agriculture: A State Level Analysis”. Economic and Political Weekly (March 29, 1997): A2-A18. Bloom, David E. and Jeffrey D. Sachs. “Geography, Demography, and Economic Growth in Africa.” Harvard Institute

for

International

Development,

September

1998. First draft.

Centre for Monitoring Indian Economy, Agriculture, August 1997 Mumbai: CMIE, 1997. Centre for Monitoring Indian Economy. Agriculture, September 1998. Mumbai: CMIE, 1998. Chand, Ramesh and T. Haque. “Sustainability of RiceWheat Crop system in Indo-Gangetic Region.” Economic and Political Weekly (March 29, 1997): A26—A30. Chattopadhyay, Manabendu and Atanu Sengupta. “Farm Size and Productivity: A New Look at the Old Debate.” Economic and Political Weekly (December 27, 1997): A172-A175. Chopra R.N. Evolution of Food Policy in India, New Delhi: Macmillan India Ltd., 1981. Dagli, Vadilal, ed. A Regional Profile of Indian Agriculture, Bombay: Vora & Co., Publishers, 1974.

60

Datt, Gaurav and Martin Ravallion. “Why have some Indian states done better than others at reducing rural poverty?” Washington, DC: World Bank, Policy Research Department, January 21, 1997, Photocopied. David, Cristina C. and Keijiro Otsuka. Modern Rice Technology and Income Distribution in Asia. Boulder and London: Lynne Rienner Publishers, 1994. Desai, Bhupat M. and N.V. Namboodiri. “Determinants of Total Factor Productivity in Indian Agriculture.” Economic and Political Weekly (December 27, 1997) A165A171. Desai, Bhupat M. and N.V. Namboodiri. “Developing Agriculture in Gujarat: A Strategic Perspective for Ninth Plan.” Economic and Political Weekly (March 29, 1997): A31-A40. Datt, Gaurav and Martin Ravallion, “Farm Productivity and Rural Poverty in India.” The Journal of Development Studies 34.4. (April 1988): 62-85. Dev. S. Mahendra, “Summaries of Group Discussion: Regional Variations in Agricultural Performance in the Last Two Decades.” Indian Journal of Agricultural Economics 53.1 (January-March 1998): 15-18. Dhawan, B.D. “Production Benefits from Large-Scale Canal Irrigation.” Economic and Political Weekly (December 27, 1997): A177-A181. Dhawan, B.D. and S.S. Yadav. “Public Investment in Indian Agriculture: Trends and Determinants.” Economic and Political Weekly (April 5, 1997): 710-714. Dinar, Ariel, Robert Mendelsohn, Robert Evenson, Jyoti Parikh, Apurva Sanghi, Kavi Kumar, James McKinsey, and Stephen Lonergan. Measuring the Impact of Climate change on Indian Agriculture. Washington, DC: The World Bank, 1998. World Bank Technical paper No. 402.

61

Dreze, Jean and Amartya Sen. India Economic Development and Social Opportunity. Delhi Oxford University Press, 1995. Economist Intelligence Unit. India 1998-99,Country Profile. http://www.eiu.com. Evenson, Rob, Pray, Carl and Rosegrant, Mark, “Agricultural Research and Productivity Growth in India” (Washington, D.C.: International Food Policy Research Institute, 1999). Food and Agriculture Organisation. “Support to India’s Rice Improvement Programme—IND/91/008.” http:// www.fao.org/ag/agpc/promo/india.htm. Food and Agriculture Organisation, “Weather-wise farmers could improve their yields.” FAO News Highlights (March 30, 1999). http://www.fao.org/news/1999/ 990307-e.htm. Gallup, John Luke. “Agricultural productivity and Geography.” Harvard Institute for International Development, January 1998. Gallup, John Luke and Jeffrey Sachs, with Andrew D. Mellinger. “Geography and Economic Development.” Harvard Institute for Internal Development, 1998. Photocopied. Gillespie, Stuart and McNeil, Geraldine. Food, Health and Survival in India and Developing Countries. Oxford University Press, Delhi, 1992. Government of India, Ministry of Finance, Economic Division. Economic Survey 1998-99. New Delhi: Government of India Press, 1999. Government of Tamil Nadu, Agriculture Department. Tamil Nadu Agricultural Policy. November 1995. Government of Tamil Nadu, Evaluation and Applied Research Department. Tamil Nadu—An Economic Appraisal,

1995-96.

62

Kabra, Kamal and Ittyerah, Anil. The Public Distribution System in India. Eastern Books, New Delhi, 1992. Kalirajan, K.P. and R.T. Shand. “Sources of Output Growth in Indian Agriculture” Indian Journal of Agricultural Economics 52 (October-December 1997): 693-706. Klohn, Wulf E. and Bo G. Appelgren. “Challenges in the field of water resources management in Agriculture.” Food and Agriculture Organisation. http://www.fao.org/WAICENT/FAOINFO/AGRICULT/AGL/ AGLW/webpub/ath-kln/ATH-KLM.htm. Krugman, Paul.“The Role of Geography in Development.” 1999. Photocopied. Kumar, Praduman, Mark Rosegrant, and Peter Hazell, “Cereals Prospects in India to 2020: Implications for Policy.” IFPRI 2020 Vision Brief 23 (June 1995). http:/ /www.cgiar.org/ifprl/2020/briefs/number23.htm. Madras Institute of Development Studies, Tamil Nadu Economy, Performance and Issues. Oxford and IBH Publishing Co., New Delhi, 1988. Mishra, Pramod K. Agricultural Risk, Insurance and Income: A Study of the Impact and Design of India’s Comprehensive Crop Insurance Scheme. Aldershot: Avebury, 1996. Mishra, S.N. “Agricultural Liberalization and Development Strategy in Ninth-Plan.” Ecomomic and Political Weekly (March 29, 1997): A19-A25. Ministry of Agriculture, Area and Production of Principal Crops in India. (New Delhi: Ministry of Agriculture, Volumes 1987-1994). Ministry of Agriculture, Agricultural Statistics at a Glance. (New Delhi: Image Print, 1998). Ministry of Agriculture, Indian Agriculture in Brief, 26th Edition (New Delhi: Government of India Press, 1995).

63

Mukherjee, Amitava. Structural Adjustment Program and Food Security. Avebury, Aldeshot, 1994. Mundlak, Yair, Donald Larson, Rita Butzer, “The Determinants of Agricultural Production: A CrossCountry Analysis.” March 16, 1997. Photocopied. Parikh, Kirit S., ed. India Development Report 1997. Indira Gandhi Institute of Development Research. Delhi: Oxford University Press, 1996. Perumalsamy, S. Ecomomic Development of Tamil Nadu, S. Chand and Company Ltd. Ramnagar, New Delhi, 1996. Raj, K.N., Amartya Sen and C.H. Hanumantha Rao, eds. Studies in Indian Agriculture. Delhi: Oxford UP, 1988. Ramasamy, C., T.R. Shanmugam, D. Suresh. “Constraints to Higher Rice Yields in Different Rice Production Environments and Prioritization of Rice Research in Southern India.” In Rice Research in Aisa: Progress and Priorities, eds. R.E. Evenson and R.W. Herdt with M. Hossain, 145-160. CAB International, 1996. Rao, V.M. “Agricultural Development with a Human Face: Experiences and Prospects.” Economic and Political Weekly (June 29, 1996): A50-A62. Rawal, Vikas and Madura Swaminathan, “Changing Trajectories: Agricultural Growth in West Bengal, 1950 to 1996.” Economic and Political Weekly (October 3, 1998): 2595-2602. Sanderson, Fred and Roy, Shyamal. Food Trends and Prospects in India. The Brookings Institution, Washington, D.C.,1979. Sanghi, Apurva, K.S. Kavi Kumar and James W. McKinsey, Jr. “India Agriculture and Climate Data Set.” The world Bank. http://www-esd.worldBank/org/indian/ database.html.

64

Sawant, S.D. “Performance of India Agriculture with Special Reference to Regional Variations.” Indian Journal Of Agricultural Economics 52 (July-September 1997): 354-373. Scandizzo, Pasquale and Swamy, Gurushri, “Benefits and Costs of Food Distribution Policies; The India Case.” World Bank Staff Working Papers, No.509. Washington D.C., The World Bank, 1982. Singh, I.J., K.N. Rai and J.C. Karwasra. “Regional Variations in Agricultural Performance in India.” Indian Journal of Agricultural Economics 52 (July - September 1997): 374-386. Timmer, C. Peter and Falcon, Walter and Pearson, Scott. Food Policy Analysis. John’s Hopkins University Press, Baltimore, 1983. Varma,

Ram

K.

Economic

Geography

and

Agricultural

Growth. Jaipur; Printwell, 1996.

Winters, Paul, et al. “The Role of Agriculture in Economic Development: Visible and Invisible Surplus Transfers.” The Journal of Development Studies 34.5 (June 1998): 71-97. World Bank, Operations Evaluation Development. “The Seed Industry in South Asia.” OED Precis 112 (May 1996). http://www.worldbank.org/html/oed/pr112.htm.

65 APPENDICES

A. Maps of Regional Foodgrain Trends and Inputs B. Stete Geography and Foodgrain Yields C. Regression Results for Model#2 D. Tamil Nadu Regression

66 APPENDIX MAPS

OF

REGIONAL AND

A FOODGRAIN

TRENDS

INPUTS

This appendix corresponds with Section III: Regional Foodgrain Trends; Section V: A Model of Geography and Foodgrain Yields; and Section VI; Additional Differences Across Koeppen Zones. It includes maps for total foodgrains, rice, wheat, and maize yield and production; net sown area, irrigation and fertilizer use. Rice: * Rice Yield 1980 * Rice Yield 1992 * Rice Production 1980 * Rice Production 1996 Wheat: * Wheat Yield 1980 * Wheat Yield 1992 * Wheat Production 1980 * Wheat Production 1992 Maize: * Maize Yield 1980 * Maize Yield 1992 Total Foodgrain: * Foodgrain Yield

1980

* Foodgrain Yield 1996 * Net Sown Area 1980 * Net Sown Area 1992

67

Irrigation: * Rice Yield and Irrigation 1980 * Rice Yield and Irrigation 1992 * Wheat Yield and Irrigation 1980 * Wheat Yield and Irrigation 1992 * Maize Yield and Irrigation

1980

* Maize Yield and Irrigation 1992 Fertilizer: * Fertilizer Consumption 1992 * Fertilizer Consumption 1996 * Koeppen Zone and Fertilizer Consumption 1992

68 APPENDIX STATE

GEOGRAPHY

AND

B FOODGRAIN

YIELDS

This appendix corresponds with Section IV: State Geography and Foodgrain Yields. It includes the following tables, charts and maps for 5 geographic variables: Variable 1: Koeppen Zones * Guide to Koeppen Zone Classification * Percent of State Land in Each Koeppen Zone * States Listed by Koeppen Zone * Map of Koeppen Zones * Chart of Koeppen Zones and Yields Variable 2: Average

Precipitation

* Chart of Average Precipitation and Yields, with classification of States in Each Division Variable 3: Elevation * Chart of Elevation and Yields, with Classification of States in Each Division Variable 4: Distance to the Nearest Navigable River * Chart of distance and Yields, with Classification of States in Each Division Variable 5: Soil Suitability Index * Chart of Soil Suitability Index and Yields, with Classification of States in Each Division.

69 Variable

GUIDE

1:

TO

Guide

Dry

Temperate

Cold

Polar

Koeppen

Zone

Classification

KOEPPEN CLIMATE ZONES IN THE TEGRATED DATA SET 5 7 Zone

Tropical

to

HIID

IN-

Description

Class A

Temperature in the coldest month does not exceed 18*C.

AF

No dry season, at least 60 mm of rain in the driest month.

A M

Short dry season, but the ground remains wet throughout the year. Monsoon type.

A W

Distinct dry season with monthly rainfall 30mm.Difference between the wettest and driest months is less than for CS and CW.

CS

Summer dry season. Rainfall in the wettest month in winter is about 3 times that in the driest month in summer. Rainfall in the driest summer month10*C and in the coldest month 30 mm. Difference between the wettest and driest months is less than for CS and CW.

D W

Winter dry season. Rainfall in the wettest month of summer is 10 times than in the driest month of winter.

Class E

Average temperature in the warmest month 10*C.

ET

Average temperature in the warmest month >0*C. Tundra.

57 From Food and Agriculture Organisation, “Brief Guide to Koeppen Climate Classification System.” http://www.fao/org/WAICENT/SUSTDEV/Eidirect/CLIMATE/Eisp0066.htm. Information translated into table.

70 Variable

1:

Chart

of

Yields

by

Koeppen

Zone

Foodgrain Yield and Koeppen Zone 3000

Yield (kg/ha)

2500

2000

1500

1000

500

0

all zones

am

aw

bs

Koeppen Zone rice yield 1980 rice yield 1992

wheat yield 1980 wheat yield 1992

maize yield 1980 maize yield 1992

foodgrain yield 1980 foodgrain yield 1992

Yellow

Cyan

Magenta

Black

cw

71 Variable

2:

Chart

of

Yield

by

Average

Precipitation 5 8

“AVE_PRC” and Foodgrain Yield 3000

Yield (kg/ha)

2500

2000

1500

1000

500

0

overall

18.21-87.53

87.53-156.86

156.56-226.18

226.18-295.5

Average Monthly Precipitation (mm) rice yield 1980 rice yield 1992

wheat yield 1980 wheat yield 1992

maize yield 1980 maize yield 1992

foodgrain yield 1980 foodgrain yield 1992

58

The range of average precipitation was divided into four equal divisions in order to classify states. Only those regions with rice, wheat, and/or maize yields in the data set were included.

Yellow

Cyan

Magenta

Black

72 Variable

3:

Chart

of

Yield

by

Elevation 5 9

Elevation and Foodgrain Yield 2500

Yield (kg/ha)

2000

1500

1000

500

0

overall

5.41-485.807

5.41-966.203

966.203-1926

1926-2887.79

2887.793848.58

Mean Elevation (mm) rice yield 1980 rice yield 1992

wheat yield 1980 wheat yield 1992

maize yield 1980 maize yield 1992

foodgrain yield 1980 foodgrain yield 1992

59

. The range of elevation measurements was divided into four equal categories. The first category then divided into 1/2 of the first quarter. Only those regions with rice, wheat, and/or maize yields in the data set are included.

Yellow

Cyan

Magenta

Black

73 Variable

4: Chart

of

Yield

Navigable

River

by

Distance

to

the

Nearest

60

“DISKMRIV” and Foodgrain Yields 3000

Yield (kg/ha)

2500

2000

1500

1000

500

0

overall

1.15149.17

1.15298.35

298.35595.53

595.53892.72

892.721189.72

Distance to the nearest navigable river (km) rice yield 1980 rice yield 1992

wheat yield 1980 wheat yield 1992

maize yield 1980 maize yield 1992

foodgrain yield 1980 foodgrain yield 1992

60 The range of values was divided into four equal categories. The first category was then divided in half. Only those regions with rice, wheat, and/or maize yields in the data set are included.

Yellow

Cyan

Magenta

Black

74 Variable

5:

Chart

of

Yield

by

Soil

Suitability

61

Soil Suitability Index #2 and Yield 2500

Yield (kg/ha)

2000

1500

1000

500

0

overall

5.41-485.807 4.66-853

5.41-966.203 8.53-13.19

966.203-1926 13.19-21.73

1926-2887.79 21.73-30.26

2887.79-3848.58 30.26-38.79

“ss2_mean” rice yield 1980 rice yield 1992

wheat yield 1980 wheat yield 1992

maize yield 1980 maize yield 1992

foodgrain yield 1980 foodgrain yield 1992

61

The range of values was divided into four equal categories. and/or maize yields in the dataset are included.

Yellow

Cyan

Magenta

Only those regions with rice, wheat,

Black