Jump to ContentJump to Main Navigation
Battling CorruptionHas NREGA Reached India's Rural Poor?$

Shylashri Shankar and Raghav Gaiha

Print publication date: 2013

Print ISBN-13: 9780198085003

Published to Oxford Scholarship Online: September 2013

DOI: 10.1093/acprof:oso/9780198085003.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2017. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: null; date: 24 February 2017

(p.171) Appendices

(p.171) Appendices

Source:
Battling Corruption
Publisher:
Oxford University Press

Appendix A1

In AP and MP, the eight and ten villages, respectively, in the sub-sample were representative of the full sample in the participation variable (share of NREGA participants in population of adults), inequality of distribution of landholdings in the village (the Gini), and household size. The monthly consumption expenditure was slightly lower than in the full sample for AP and slightly higher than in the full sample for MP. TN averages in the sub-sample were similar for the participant and household size but exhibited more inequality of landholdings, more land owned, and lower PCME, as compared to the full sample. In Rajasthan, the averages for participant, Gini and household size were similar between the sub-sample and the full sample, but varied on the monthly household consumption expenditure (the sub-sample average was higher), and land owned (the sub-sample had higher mean of landholdings than the full sample). So, for instance, if the average share of NREGA participants in Rajasthan was 20 per cent, the sub-sample had a similar percentage of NREGA participants. Similarly, if the average household size in the full Rajasthan sample was 4 members, the sub-sample too had an average household size of 4 members. (p.172)

Table A1.1 Definitions of Variables Used in Probit Analysis

Dependent Variable

Definition

NREGA participation

NREGA participation (=1 if participated in NREGA; 0 otherwise)

Explanatory variables

Gender

Gender of household member or head (1 if male, 0 if female)

Age

Age of household member or head

Square of age

Square of age of household member or head

Whether married

Dummy for being married (=1 if married; 0 otherwise)

Illiterate (reference)

Dummy for no education (=1 if illiterate, 0 otherwise)

Below primary education

Dummy for primary education (=1 if literate but up to primary education, 0 otherwise)

Middle school

Dummy for middle school (=1 if passed only up to middle school, 0 otherwise)

Secondary education

Dummy for secondary education (=1 if literate but up to secondary education, 0 otherwise)

Higher secondary plus

Dummy for higher secondary and above (=1 if education up to higher secondary and above, 0 otherwise)

SC

Dummy for SC (=1 if household or member of SC, 0 otherwise)

ST

Dummy for ST (=1 if household or member of ST, 0 otherwise)

OBC

Dummy for OBC (=1 if household or member of OBC, 0 otherwise)

Others

Dummy for Others (=1 if household or member of other caste, 0 otherwise)

Amount of land owned

Amount of land owned

Square of amount of land owned

Square of amount of land owned

Household size

Size of the household

Square of household size

Square of size of the household

Number of adult male

Number of adult males in the household

Number of adult female

Number of adult females in the household

Ratio of NREG to AGR WR

Ratio of NREG wage to agricultural

Square of Ratio of NREG to

Square of ratio of NREG wage to agricultural

AGR WR

WR at village level

Land Gini index

Gini index of inequality of landholdings

Square of land Gini index

Square of Gini index of inequality of landholdings

Interaction: Ratio NREG/AGR/WR with LGI

Interaction of ratio of NREG wage to agricultural WR at village level with Gini index of inequality of landholdings

Average distance of site from the village

Average distance of site from the village

Percentage households attending meetings

Percentage households attending meetings at village level

Interaction: Percentage households meet/attend with LGI

Interaction of percentage households attending meetings with Gini index of inequality of landholdings at the village level

Percentage households with TV/cell phone

Percentage households with TV and cell phone both at village level

Interaction: Percentage households meet/attend with Percentage households with TV/cell phone

Interaction of percentage households attending meetings with percentage households with both TV and cell phone at village level

Source: Based on the household and ethnographic surveys conducted by the authors.1

(p.173)

Table A1.2 Sub-sample Means

Variables

AP

TN

MP

Rajasthan

Participant

0.41 (0.43)

0.29 (0.26)

0.19 (0.21)

0.19 (0.22)

Gini

0.56 (0.60)

0.54 (0.75)

0.58 (0.60)

0.47 (0.48)

Land owned

1.25 (1.33)

0.94 (0.64)

2.35 (2.13)

2.35 (1.72)

Household size

4.91 (4.77)

4.72 (4.23)

6.17 (6.03)

6.19 (6.22)

PCME

416 (456)

548 (614)

270 (236)

594 (487)

Note: Figures within parentheses are full sample averages.

Appendix A2

(p.174)

Table A2.1 Correlates of Poverty, Participation in NREGA, and Share of NREGA Earnings in Household Income in Rajasthan and Andhra Pradesh (percentage)

Household and village characteristics

Rajasthan

AP

Share in population

Share in NREGA participation

Share of NREGA earnings in household’s income net of NREGA

Share in population

Share in NREGA participation

Share of NREGA earnings in household’s income net of NREGA

Male-headed households

95.15

94.34 (65.25)

11.21

92.86

91.71 (77.56)

23.41

Female-headed households

4.85

5.66 (76.76)

23.69

7.14

8.29 (91.17)

26.48

SC

25.36

27.03 (70.15)

11.86

29.23

34.86 (93.69)

23.16

ST

29.55

33.57 (74.76)

13.78

9.59

11.45 (93.69)

27.25

OBC

34.19

32.96 (63.46)

9.87

48.95

49.92 (80.09)

22.86

Others

10.91

6.44 (38.85)

10.89

12.23

3.77 (24.22)

26.52

Poor

40.98

49.92 (80.17)

16.01

25.34

28.11 (87.11)

25.80

Non-poor

59.02

50.08 (55.84)

9.22

74.66

71.89 (75.62)

22.81

Landless

33.61

26.08 (51.07)

15.26

43.44

47.27 (85.46)

26.17

〉0-〈=1

26.77

30.47 (74.90)

14.67

24.86

28.67 (90.56)

22.91

〉1-〈=2

24.51

30.31 (81.38)

10.58

16.40

14.88 (71.24)

24.09

〉2-〈=5

11.16

10.40 (61.35)

6.86

11.78

47.27 (85.46)

17.74

〉5

3.95

2.75 (45.74)

15.26

3.51

28.67 (90.56)

14.17

4 and less

38.47

39.37 (67.36)

16.74

59.21

57.90 (76.81)

23.61

〉4–〈=8

55.55

53.67 (63.60)

10.03

39.92

40.99 (80.63)

23.72

〉8–〈=12

5.89

6.87 (76.85)

7.08

0.87

1.11 (100.00)

19.08

〉12

0.10

0.08 (54.02)

0.68

All

100.00 (34)

100.00 (66)

11.63

100

100

23.61

Poverty Status

Acutely Poor

30

34 (76)

18

11

12 (86)

21

Moderately Poor

11

16 (90)

13

14

16 (88)

20

Moderately non-poor

19

20 (70)

12

28

29 (82)

20

Affluent

40

30 (49)

11

47

42 (72)

18

All

100 (34)

100 (66)

13

100

100 (82)

19

Note: All calculations are at the household level. Figures in brackets represent shares within groups (row percentage).

(p.175) (p.176)

Table A2.2 Correlates of Poverty, Participation in NREGA, and Share of NREGA Earnings in Household Income in Madhya Pradesh and Tamil Nadu, 2008–9 (percentage)

Household characteristics

MP

TN

Share in population

Share in NREGA participation

Share of NREGA earnings in household’s income net of NREGA

Share in population

Share in NREGA participation

Share of NREGA earnings in household’s income net of NREGA

Gender of household head

Male

95.80

96.15 (62.81)

5.79

89.20

88.79 (76.61)

5.28

Female

4.20

3.85 (57.29)

5.80

10.80

11.21 (79.93)

11.39

Social Group

SC

14.67

14.04 (59.90)

11.21

37.02

40.4 (84.00)

6.24

ST

35.45

44.95 (79.36)

8.30

5.84

6.81 (89.76)

9.96

OBC

41.68

38.78 (58.23)

4.65

55.3

51.62 (71.84)

4.91

Others

8.21

2.23 (16.97)

1.07

1.83

1.16 (48.82)

4.99

Poverty Status

Poor

71.08

81.50 (71.76)

9.54

35.98

39.98 (85.50)

9.72

Non-poor

28.92

18.50 (40.03)

2.15

64.02

60.02 (72.17)

4.34

Land owned group ( in acres)

Landless

38.46

36.75 (59.80)

7.43

51.76

52.72 (78.39)

5.52

〉0–≤1

15.91

19.62 (77.19)

13.30

30.79

31.68 (79.19)

7.75

〉1–≤2

15.76

17.52 (69.57)

7.17

11.16

11.15 (76.89)

4.72

〉2–≤5

22.54

23.07 (64.05)

5.25

5.14

3.99 (59.79)

2.32

〉5

7.34

3.04 (25.93)

1.09

1.15

0.46 (30.82)

0.61

Household size group

4 and less

37.44

33.82 (56.52)

5.54

71.16

67.43 (72.93)

5.48

〉4–≤8

56.89

59.59 (65.54)

5.99

28.11

31.62 (86.58)

5.75

〉8–≤12

5.54

6.39 (72.21)

5.71

0.73

0.95 (100.00)

10.74

〉12

0.13

0.20 (100.00)

1.47

0.00

0.00 (0.00)

-

Poverty Status

Acutely Poor

56

64 (72)

16

20

23 (89)

17

Moderately Poor

15

17 (73)

14

16

17 (810

11

Moderately non-poor

14

11 (52)

16

25

37 (83)

12

Affluent

15

7 (29)

9

39

29 (65)

9

All

100.00

100.00 (63)

15

100.00

100.00 (77)

12

Note: S1: Share (percentage) in population, S2: Share (percentage) in NREGA Participation, S3: Share (percentage) of NREGA earnings in household’s income net of NREGA. All calculations are at the household level. Figures in brackets represent shares within groups (row percentage).

(p.177)

(p.178)

Table A2.3 Number of Days Worked in NREGA by Poverty Status

Mean

Standard Deviation

Median

Madhya Pradesh

Acute poverty

44

29

36

Moderate poverty

47

31

40

Moderate Non-poverty

41

27

30

Affluent

40

29

30

Poor

45

29

36

Non-poor

41

28

30

All

44

29

35

Tamil Nadu

Acute poverty

46

22

40

Moderate poverty

41

29

36

Moderate Non-poverty

41

23

33

Affluent

42

25

37

Poor

44

25

38

Non-poor

41

24

36

All

42

25

37

Andhra Pradesh

Acute poverty

87

53

75

Moderate poverty

96

68

78

Moderate Non-poverty

89

53

80

Affluent

82

52

74

Poor

92

52

77

Non-poor

85

62

78

All

87

55

78

Rajasthan

Acute poverty

66

33

65

Moderate poverty

58

36

60

Moderate Non-poverty

62

41

52

Affluent

61

42

45

Poor

64

34

61

Non-poor

62

42

51

All

63

38

54

(p.179)

Table A2.4 Per day NREGA Wages Earned

Mean

Standard Deviation

Median

Madhya Pradesh

Acute poverty

70

17

68

Moderate poverty

72

20

80

Moderate Non-poverty

79

13

85

Affluent

85

13

91

Poor

70

18

69

Non-poor

82

13

85

All

73

17

70

Tamil Nadu

Acute poverty

67

10

68

Moderate poverty

61

12

65

Moderate Non-poverty

65

12

67

Affluent

66

11

70

Poor

65

11

65

Non-poor

66

12

68

All

65

12

67

Andhra Pradesh

Acute poverty

78

12

78

Moderate poverty

79

9

80

Moderate Non-poverty

79

9

79

Affluent

81

10

81

Poor

79

10

78

Non-poor

80

10

80

All

79

10

80

Rajasthan

Acute poverty

55

14

57

Moderate poverty

61

13

60

Moderate Non-poverty

62

14

60

Affluent

61

11

61

Poor

57

14

60

Non-poor

62

12

60

All

59

13

60

(p.180)

Table A2.5 Estimation of NREGA Participation Equation: Probit Regression, Rajasthan and Andhra Pradesh

Models

Rajasthan

AP

Explanatory variables

Coefficients (t-value)

Elasticities (t-value)

Coefficients (t-value)

Elasticities (t-value)

Gender

−0.06 (−0.62)

−0.05 (−0.61)

0.04 (0.50)

0.022 (0.50)

Age

0.178*** (9.62)

8.53*** (8.47)

0.19*** (9.35)

6.14*** (8.16)

Square of Age

−0.002*** (−8.76)

−2.93*** (−7.38)

−0.002*** (−8.34)

−2.93*** (−7.38)

Married

−0.13 (−0.71)

−0.114 (−0.71)

−0.07 (−0.40)

−0.03 (−.40)

Below primary education

−0.40*** (−3.38)

−0.242*** (−3.36)

−0.33*** (−3.26)

−0.127*** (−3.23)

Middle school

−0.67*** (−3.75)

−0.131*** (−3.74)

−0.69*** (−4.25)

−0.09*** (−4.14)

Secondary education

−0.60** (−2.36)

−0.070** (−2.33)

−0.61*** (−4.16)

−0.06*** (−4.10)

Higher secondary plus

−0.93*** (−4.85)

−0.141*** (−4.81)

−0.45** (−2.56)

−0.04** (−2.52)

SC

0.34* (1.85)

0.16* (1.84)

0.85*** (5.62)

0.275*** (5.52)

ST

0.36* (1.87)

0.21* (1.87)

0.87*** (5.08)

0.08*** (4.97)

OBC

0.42** (2.26)

0.26** (2.26)

0.62*** (4.50)

0.33*** (4.44)

Amount of land owned

−0.02 (−0.73)

−0.041 (−.73)

−0.12*** (−5.01)

−0.165*** (−4.94)

Number of adult male

−0.13** (−2.54)

−0.47** (−2.49)

0.09** (2.07)

0.179** (2.08)

Number of adult female

−0.21*** (−3.48)

−0.704*** (−3.43)

−0.14*** (−2.94)

−0.26*** (−2.93)

Ratio of NREGA to AGR WR

−1.25* (−1.93)

−2.31* (−1.89)

4.16** (2.39)

5.16** (2.39)

Square of Ratio of NREGA to AGR WR

−1.76** (−2.42)

−2.54** (−2.41)

Land Gini index

−3.88*** (−2.85)

−4.2*** (−2.73)

0.57 (1.31)

.38 (1.31)

Interaction: Ratio NREGAGRWR with LGI

3.30** (2.54)

3.44** (2.47)

Average distance of site from the village

−0.05 (−0.59)

−0.188 (−0.59)

−0.39*** (−4.77)

−0.737*** (−4.58)

Percentage households attending meetings

0.01** (2.53)

0.466** (2.49)

0.06*** (2.89)

4.97*** (2.89)

Square of Percentage households attending meetings

−0.001*** (−3.47)

−3.68*** (−3.46)

Percentage households with both TV and cell phone

Constant

−1.58 (−2.14)

−6.72 (−5.47)

Number of observations

2,684

2,190

Pseudo R-square

0.3220

0.3512

Wald chi-square

392.63

649.76

Note: *, **, *** = significance at the 10 percentage, 5 percentage and 1 percentage level, respectively.

(p.181)

(p.182)

Table A2.6 Estimation of NREGA Participation Equation: Probit Regression, Madhya Pradesh and Tamil Nadu

Models

MP

TN

Explanatory variables

Coefficients (t-value)

Elasticities (t-value)

Coefficients (t-value)

Elasticities (t-value)

Gender

0.527*** (5.44)

0.54*** (5.14)

−0.71*** (−7.75)

−0.55*** (−7.16)

Age

0.201*** (10.55)

10.04*** (8.66)

0.146*** (8.68)

6.62*** (7.97)

Square of Age

−0.002*** (−10.12)

−4.6*** (−8.42)

−0.001*** (−7.89)

−3.27*** (−7.35)

Married

0.327* (1.91)

0.307* (1.91)

0.37** (2.57)

0.28** (2.56)

Below primary education

−0.093 (−.82)

−0.066 (−0.82)

−0.17 (−1.45)

−0.62 (−1.45)

Middle school

−0.193 (−1.21)

−0.03 (−1.21)

−0.41*** (−2.76)

−0.090*** (−2.77)

Secondary education

−0.29 (−1.42)

−0.03 (−1.42)

−0.78*** (−5.07)

−0.17*** (−5.01)

Higher secondary plus

−1.28*** (−3.29)

−0.121*** (−3.16)

−0.91*** (−4.65)

−0.161*** (−4.54)

SC

0.49** (2.00)

0.15** (1.96)

0.074 (0.74)

0.042 (0.74)

ST

0.69*** (2.79)

0.54*** (2.72)

0.64*** (2.85)

0.0533*** (2.83)

OBC

0.49** (2.01)

0.36** (1.97)

Amount of land owned

−0.05*** (−2.96)

−0.22*** (−2.92)

−0.10** (−2.59)

−0.117** (−2.57)

Number of adult male

−0.413*** (−3.68)

−1.5*** (−3.64)

Number of adult female

0.236** (2.00)

0.774** (1.99)

Ratio of NREGA to AGR WR

−2.92*** (−2.72)

−7.68*** (−2.69)

−0.079 (−0.45)

−0.09 (−0.45)

Square of Ratio of NREGA to AGR WR

1.10*** (3.08)

4.10*** (3.04)

Land Gini index

3.82* (1.85)

4.69* (1.84)

−0.79** (−2.55)

−0.78** (−2.53)

Square of Land Gini

−3.04* (−1.75)

−2.45* (−1.75)

Interaction: Ratio NREGAGRWR with LGI

Average distance of site from the village

0.166** (2.2)

0.64** (2.18)

−0.205* (−1.77)

−0.45* (−1.76)

Household size

−0.18* (−1.70)

−1.17* (−1.70)

Square of household size

0.015 (1.46)

0.48 (1.46)

Percentage households attending meetings

−0.00 (−0.17)

−0.025 (−0.17)

0.002 (0.94)

0.11 (0.94)

Square of Percentage households attending meetings

Percentage households with both TV and cell phone

0.009** (2.35)

0.71** (2.35)

Constant

−4.23 (−3.85)

−1.8 (−3.65)

Number of observations

2,163

1,855

Pseudo R-square

0.38

0.34

Wald chi-square

444.39

434.09

Note: *, **, *** = significance at the 10 percentage, 5 percentage, and 1 percentage level, respectively.

(p.183) (p.184)

Table A2.7 Distribution of the Poor (percentage)

Degrees of poverty

Rajasthan

AP

MP

TN

Total

Participants in NREGA

Total

Participants

Total

Participant

Total

Participant

Acutely Poor

33

37

12

13

57

62

23

25

Moderately Poor

11

13

16

18

15

17

16

17

Moderately Non-poor

21

21

31

32

14

12

26

28

Affluent

35

29

40

37

14

8

36

31

(p.185)

Table A2.8 Profile of Participants by Poverty Status: Rajasthan

Dependent variable:

Poverty status: 1 = Acutely poor, 2 = Moderately poor, 3 = Moderately non-poor, 4 = Affluent

Estimation methods:

Ordered Probit regression

Explanatory variables

Coefficients

Marginal effects for poverty status

All

Acutely poor

Moderately poor

Moderately non-poor

Affluent

Gender

−0.313* (−1.75)

0.110* (1.71)

0.013* (1.81)

−0.021 (−1.55)

−0.0103* (−1.78)

Age

−0.015 (−0.43)

0.005 (0.43)

0.0007 (0.42)

−0.001 (−0.43)

−0.005 (−0.43)

Square of Age

0.00004 (0.86)

−0.0001 (−0.86)

−0.00001 (−0.79)

0.00002 (0.85)

0.0001 (0.86)

Whether Married

0.04 (0.22)

−0.015 (−0.22)

0.003 (0.21)

0.003 (0.21)

0.014 (0.22)

Below primary education

−0.044 (−0.20)

0.015 (0.20)

−0.002 (−0.23)

−0.002 (−0.19)

−0.014 (−0.21)

Middle school

0.75** (2.12)

−0.21*** (−2.84)

0.002 (0.21)

−0.008 (−0.03)

0.28** (2.04)

Secondary education

0.618 (1.54)

−0.178** (−1.97)

−0.06* (−1.72)

−0.0008 (−0.03)

0.23 (1.46)

Higher secondary plus

0.835** (2.20)

−0.22*** (−3.27)

−0.077* (−1.74)

−0.019 (−0.47)

0.317** (2.15)

SC

−0.66* (−1.75)

0.24* (1.72)

−0.01 (1.35)

−0.06 (−1.42)

−0.197** (−2.00)

ST

−0.72* (−1.89)

0.26** (1.89)

0.020* (1.88)

−0.05* (−1.61)

−0.22** (−2.08)

OBC

0.384 (1.03)

−0.12 (−1.08)

−0.022 (−0.89)

0.01 (1.35)

0.132 (1.01)

Adult Male

0.223** (2.28)

−0.078** (−2.30)

−0.010* (−1.73)

0.014** (1.91)

0.07** (2.29)

Adult Female

−0.253** (−2.43)

0.08*** (2.41)

0.012** (1.93)

−0.016** (−1.94)

−0.08*** (−2.46)

Amount of land owned

0.04 (1.09)

−0.017 (−1.07)

−0.002 (−1.10)

0.003 (1.02)

0.0165 (1.09)

Land Gini index

0.06 (0.11)

−0.022 (−0.11)

−0.003 (−0.11)

0.004 (0.11)

0.02 (0.11)

Average Distance to worksite

0.076 (0.49)

−0.026 (−0.49)

−0.003 (−0.49)

0.004 (0.49)

0.02 (0.50)

District measure of participation

−0.28 (−0.09)

0.100 (0.09)

0.014 (0.09)

−0.018 (−0.09)

−0.096 (−0.09)

Log likelihood

−7633958.8

Number of observations

576

Pseudo R-square

0.1249

Wald chi-square

116.98

Probability〉chi 2

0.0000

Marginal effects after Oprobit

0.30393614

0.18305987

0.23683994

0.27616405

Note: *, **, *** = significance at the 10 percentage, 5 percentage, and 1 percentage level, respectively. Figures in parentheses are the t-values.

(p.186)

(p.187)

Table A2.9 Profile of Participants by Poverty Status: Madhya Pradesh

Dependent variable:

Poverty status: 1 = Acutely poor, 2 = Moderately poor, 3 = Moderately non-poor, 4 = Affluent

Estimation methods:

Ordered Probit regression

Explanatory variables

Coefficients

Marginal effects for poverty status

All

Acutely poor

Moderately poor

Moderately non-poor

Affluent

Gender

−0.42** (−2.24)

0.054** (1.99)

−0.034** (−1.95)

−0.01* (−1.87)

−0.006 (−1.52)

Age

0.02*** (3.17)

−0.003*** (−3.26)

0.002*** (2.98)

0.02*** (2.98)

0.0003*** (2.01)

Whether Married

−0.17 (−0.62)

0.02 (0.55)

−0.014 (−0.56)

−0.005 (−0.53)

−0.002 (−0.52)

Below primary education

0.87*** (3.74)

−0.142*** (−3.03)

0.083*** (3.07)

0.03*** (2.57)

0.023* (1.85)

Middle school

0.25 (0.54)

−0.03 (−0.47)

0.02 (0.48)

0.008 (0.45)

0.004 (0.42)

Secondary education

1.31*** (3.77)

−0.33*** (−2.69)

0.09** (2.22)

0.088* (1.61)

Higher secondary plus

0.39 (0.67)

−0.063 (−0.53)

0.03 (0.57)

0.015 (0.50)

0.009 (0.44)

SC

0.55 (1.05)

−0.086 (−0.81)

0.051 (0.86)

0.02 (0.80)

0.012 (0.64)

ST

0.02 (0.04)

−0.002 (−0.04)

0.001 (0.04)

0.0005 (0.04)

0.0002 (0.04)

OBC

0.44 (0.88)

−0.05 (−0.76)

0.03 (0.78)

0.014 (0.76)

0.007 (0.65)

Amount of land owned

0.22*** (5.33)

−0.02*** (−4.28)

0.017*** (3.73)

0.006*** (3.52)

0.003** (2.19)

Land Gini index

−6.9 (−1.58)

0.81 (1.55)

−0.53( –1.47)

−0.19 (−1.52)

−0.09 (−1.47)

Square of Land Gini index

5.5 (1.45)

−0.64 (−1.41)

0.42 (1.35)

0.15 (1.4)

0.076 (1.34)

Average Distance to worksite

−0.55*** (−3.05)

0.06*** (3.05)

−0.04*** (−3.53)

−0.015*** (−2.78)

−0.007** (−2.05)

Percentage household Attend Village Meetings

−0.01*** (−4.13)

0.001*** (3.87)

−0.001*** (−3.53)

−0.0004*** (−2.93)

−0.0002** (−2.24)

Adult Males in household

0.29 (1.37)

−0.034 (−1.37)

0.02 (1.27)

0.008 (1.46)

0.004 (1.45)

Adult Females in household

−0.43* (−1.76)

0.05* (1.73)

−0.033 (−1.47)

−0.012* (−1.83)

−0.006* (−1.80)

Ratio of NREGA wages to agricultural wages

−3.94* (−1.74)

0.46* (1.70)

−0.302* (1.68)

−0.107* (−1.63)

−0.055 (−1.37)

Square of Ratio of NREGA wage to agricultural wages

1.37* (1.83)

−0.16* (−1.80)

0.10* (1.77)

0.03 (1.70)

0.019 (1.42)

District Probability of Participation

8.4*** (2.85)

−0.99*** (−2.82)

0.64*** (2.52)

0.22*** (2.63)

0.117** (2.12)

Number of observations

446

Pseudo R-square

0.1952

Wald chi-square

69.69

Probability〉chi 2

0.0000

Log Likelihood

−1047188.8

Marginal effects after Oprobit

0.940

0.042

0.011

0.004

Note: *, **, *** = significance at the10 percentage, 5 percentage, and 1 percentage level, respectively. Figures in parentheses are the t-values.

(p.188)

(p.189)

Table A2.10 Profile of Participants by Poverty Status: Andhra Pradesh

Dependent variable:

Poverty status: 1 = Acutely poor, 2 = Moderately poor, 3 = Moderately non-poor, 4 = Affluent

Estimation methods:

Ordered Probit regression

Explanatory variables

Coefficients

Marginal effects for poverty status

All

Acutely poor

Moderately poor

Moderately non-poor

Affluent

Gender

0.033 (0.32)

−0.002 (−0.32)

−0.006 (−0.32)

−0.002 (−0.32)

0.012 (0.32)

Age

0.001 (0.18)

−0.00009 (−0.19)

−0.0002 (−0.18)

−0.00008 (−0.18)

0.0003 (0.18)

Whether Married

0.43** (2.52)

−0.29*** (−3.31)

−0.079*** (−2.72)

−0.057* (−1.78)

0.165*** (2.46)

Below primary education

−0.288** (−2.37)

0.02** (1.90)

0.05*** (2.36)

0.017*** (2.51)

−0.105*** (−2.42)

Middle school

−0.357 (−1.38)

0.04 (1.03)

0.07 (1.39)

0.007 (0.53)

−0.121 (−1.53)

Secondary education

−0.254 (−1.23)

0.02 (1.02)

0.05 (1.20)

0.01** (2.06)

−0.088 (−1.30)

Higher secondary plus

−0.65*** (−2.86)

0.08* (1.88)

0.131*** (3.11)

−0.017 (−0.54)

−0.204*** (−3.54)

SC

−0.55*** (−3.82)

0.05*** (2.90)

0.112*** (3.67)

0.025*** (2.43)

−0.195*** (−4.12)

ST

−0.85*** (−4.83)

0.12*** (2.93)

0.166*** (5.34)

−0.03 (−1.15)

−0.25*** (−6.30)

OBC

−0.65*** (−4.80)

0.058*** (3.42)

0.127*** (4.69)

0.05*** (3.96)

−0.23*** (−5.08)

HOUSEHOLD size

−1.43*** (−10.40)

0.123*** (5.83)

0.283*** (8.20)

0.118*** (3.90)

−0.525*** (−10.27)

Square of household size

0.08*** (6.65)

−0.007*** (−4.76)

−0.016*** (−6.04)

−0.006*** (−3.51)

0.030*** (6.57)

Amount of land owned

0.315*** (5.69)

−0.027*** (−4.82)

−0.06*** (−5.11)

−0.026*** (−3.33)

0.115*** (5.70)

Ratio Of NREGA wage

1.70*** (5.38)

−0.147*** (−4.19)

−0.33*** (−5.03)

−0.141*** (−3.22)

0.628*** (5.25)

Land Gini index

−0.234 (−0.50)

0.020 (0.49)

0.04 (0.50)

0.019 (0.51)

−0.086 (−0.50)

Average Distance

2.88*** (6.24)

−0.248*** (−5.43)

−0.572*** (−5.73)

−0.23*** (−3.14)

1.06*** (6.05)

Square of Average Distance

−0.713*** (−5.46)

0.06*** (4.79)

0.141*** (5.14)

0.05*** (3.02)

−0.262*** (−5.33)

District Measure of Participation

2.32*** (4.43)

−0.200*** (−3.66)

−0.46*** (−4.23)

−.193*** (−3.10)

0.855*** (4.43)

Number of observations

940

Pseudo R-square

0.2435

Wald chi-square

448.13

Probability〉chi2

0.0000

Log likelihood

−11867199

Marginal effects after Oprobit

0.039

0.165

0.451

0.342

Note: *, **, *** = significance at the 10 percentage, 5 percentage, and 1 percentage level, respectively. Figures in parentheses are the t-values.

(p.190)

(p.191)

Table A2.11 Profile of Participants by Poverty Status: Tamil Nadu

Dependent variable:

Poverty status: 1 = Acutely poor, 2 = Moderately poor, 3 = Moderately non-poor, 4 = Affluent

Estimation methods:

Ordered Probit regression

Explanatory variables

Coefficients

Marginal effects for poverty status

All

Acutely poor

Moderately poor

Moderately non-poor

Affluent

Gender

−0.101 (−0.81)

0.028 (0.81)

0.0109 (0.82)

−0.006 (−0.74)

−0.03 (−0.83)

Age

0.089** (2.48)

−0.0248** (−2.48)

−0.009** (−2.29)

0.005* (1.86)

0.029** (2.51)

Square of Age

−0.001** (−2.56)

0.0003** (2.55)

0.0001** (2.35)

−0.00006* (−1.89)

−0.00003** (−2.59)

Whether Married

0.366** (2.09)

−0.112** (−1.91)

−0.031** (−2.67)

0.034 (1.57)

0.11** (2.28)

Below primary education

0.012 (0.09)

−0.003 (−0.09)

−0.0014 (−0.08)

0.0007 (0.09)

0.004 (0.09)

Middle school

0.182 (0.85)

−0.04 (−0.91)

−0.021 (−0.79)

0.007 (1.45)

0.062 (0.82)

Secondary education

−0.051 (−0.23)

0.014 (0.22)

0.0055 (0.23)

−0.003 (−0.21)

−0.016 (−0.23)

Higher secondary plus

0.09 (0.26)

−0.025 (−0.27)

−0.011 (−0.25)

0.0044 (0.36)

0.03 (0.26)

SC

0.002 (0.02)

−0.0006 (−0.02)

−0.0002 (−0.02)

0.0001 (0.02)

0.0007 (0.02)

ST

−0.786** (−2.03)

0.267* (1.83)

0.03*** (2.73)

−0.102 (−1.53)

−0.202*** (−2.73)

HOUSEHOLD size

−0.676*** (−4.27)

0.188*** (4.38)

0.074*** (3.36)

−0.040*** (−2.64)

−0.22*** (−4.20)

Square of household size

0.046*** (2.86)

−0.012*** (−2.89)

−0.005*** (−2.53)

0.002** (2.17)

0.015*** (2.84)

Amount of land owned

0.60*** (5.01)

−0.167*** (−5.15)

−0.066*** (−3.71)

0.035*** (2.72)

0.198*** (4.96)

Land Gini index

1.92*** (4.62)

−0.53*** (−4.41)

−0.21*** (−3.93)

0.115*** (2.66)

0.634*** (4.54)

Average Distance

0.023 (.14)

−0.006 (−0.15)

−0.002 (−0.14)

0.001 (0.15)

0.007 (0.14)

Ratio of NREG wages to agricultural wages

0.39* (1.78)

−0.109* (−1.76)

−0.043* (−1.75)

0.023 (1.55)

0.129* (1.78)

Percentage HOUSEHOLD in village with TV and Cell phone

−0.006 (−0.91)

0.001 (0.91)

0.0006 (0.90)

−0.0003 (−0.91)

−0.002 (−0.91)

District measure of Participation

9.50*** (3.54)

−2.64*** (−3.53)

−1.05*** (−3.08)

.569*** (2.49)

3.13*** (3.47)

Number of observations

501

Pseudo R-square

0.1234

Wald chi-square

104.05

Probability〉chi2

0.0000

Log likelihood

−2649572.7

Marginal effects after Oprobit

0.198

0.214

0.318

0.268

Note: *, **, *** = significance at the 10 percentage, 5 percentage, and 1 percentage level, respectively. Figures in parentheses are the t-values.

(p.192)

(p.193) Appendix A3

Data and Methodology

The present analysis draws upon the household dataset, the household subset, the ethnographic interviews, and the worksite focus group interviews. As noted in the earlier chapter, the sub-samples, by and large, were similar to those in the full sample, as illustrated in Appendix 3, Table A3.1, at the end of the book. An advantage of combining these datasets with the ethnographic one is that the interviews allow us to investigate in considerable detail the effects of political reservations at the level of the village council.

We constructed three types of transfers. For all three transfers, the dependent variable is the ratio between actual earnings of a participant relative to what he/she should have earned in a specific type of transfer. So the higher the value of the ratio, the greater is the benefit to the recipient (in our case, a participant in NREGA).

Uniform Transfers

The uniform transfer was constructed by dividing the total wage earnings under the scheme in each village by the number of participants. This gives us the average wages that ought to be earned by each participant in a village per year. Actual divided by uniform transfer takes values less than or equal to or greater than 1. Whether NREGA’s direct transfers exceed uniform transfers is analysed using robust regression with village and household level variables.2 In AP, the average wages per year per participant ranged from Rs 683 to Rs 3,788. The mean uniform transfer was Rs 2,303. In TN, the figures were Rs 1,277 to Rs 2,826, and the mean was Rs 2,438. In MP, the figures were Rs 1,121 to Rs 2,134, and the mean was Rs 2,018. In Rajasthan, the range was between Rs 829–2,039, and the mean was Rs 1,827.

Proportional Transfer

The proportional transfer is the amount the SC/ST groups are supposed to receive proportional to their (p.194) population in a village. For instance, if the SC/ST population is 30 per cent of the total population of the village, their proportional earnings would be 30 per cent of the total wage earnings of that village. Our dependent variable—proportional transfer—is a ratio of the actual earnings to the proportional earnings. In other words, if the SC/ST proportion was 30 per cent, we multiplied the total wage by 30 per cent and divided it by the number of SC/ST participants to create the proportional transfer. We then constructed the proportional transfer variable by taking the actual earnings of an SC/ST participant and dividing it by the proportional earnings. One of the relationships that the robust regression allows us to examine is the performance of SC/ST living in a reserved village, as compared to SC/ST living in unreserved villages. For instance, if the average amount earned by an SC/ST living in a reserved village is 0.50 of the proportional earnings, in non-reserved villages this amount may be 0.40 (or less than 0.50) if the sign was positive and significant for SC/ST reserved villages. Note that the village level proportional transfer figures will vary from village to village depending on the proportions of SC/ST populations. At the state level, the proportional transfer figures based on the proportions of SC/ST populations in the state are: AP (Rs 2,431), TN (Rs 2,321), MP (Rs 1,437), Rajasthan (Rs 1,791).

Mandatory Transfer

The mandatory transfer was constructed in the following manner: we applied 33 per cent to NREGA earnings in a village, and divided this by the number of female participants. This is the mandatory transfer—the ratio between the actual wages earned by female participants and mandatory amount. The mandatory transfers are: AP (Rs 1,754); TN (Rs 1,093), MP (Rs 1,904), Rajasthan (Rs 1,197).

Transfer Details for Poor Participants

We have included it in this appendix for convenience. Please see Chapter 4 for the analysis.

(p.195) Proportional Transfer

The proportional transfer is the amount the poor groups are supposed to receive proportional to their population in a village. For instance, if the poor are 30 per cent of the total population of the village, their proportional earnings would be 30 per cent of the total wage earnings of that village.

Our dependent variable—proportional transfer—is a ratio of the actual earnings to the proportional earnings. In other words, if the poor’s proportion was 30 per cent, we multiplied the total wage by 30 per cent and divided it by the number of poor participants to create the proportional earnings. We then constructed the proportional transfer variable by taking the actual earnings of a poor participant and dividing it by the proportional earnings. The proportional transfer figures are: AP (Rs 2,431), TN (Rs 2,321), MP (Rs 1,437), Rajasthan (Rs 1,791).

Rawlsian Max–Min Transfer

We use the intuition developed by political philosopher, John Rawls, in the context of the Difference principle. Rawls defined this principle as the rule which states that social and economic inequalities should be arranged so that ‘they are to be of the greatest benefit to the least-advantaged members of society’. An unequal distribution can be just when it maximizes the benefit to those who have the most minuscule allocation (in our case, the acutely poor) of welfare-conferring resources (which he refers to as ‘primary goods’). Using this intuition, we distributed the total amount of NREGA wages among the acutely poor participants, and termed it Rawlsian transfer. Our dependent variable then is a ratio of the actual earnings of the acutely poor to the Rawlsian amount.

Description of the Model

Since our full sample and the subset have participants and non-participants, we ran a probit model on the full sample to derive the probability of participation in NREGA, the results of which we report below. In the second part of the analysis, we ran a robust regression on the sub-sample to analyse (p.196) NREGA earnings relative to three different norms of transfers: uniform, proportional and mandatory. The robust regression results are reported in the main body of the chapter.

OLS regression is based on the assumption that the unobservable error term, conditional on the explanatory variables, is constant and has zero mean and constant variance, that is homoscedasticity.3 Homoscedasticity fails whenever the variance of the unobservables changes across different segments of the population, where the segments are determined by the different values of the explanatory variables. Robust regression allows us to overcome this difficulty (Wooldridge 2006).

The probit participation equation on the full sample assessed the probability of participating in NREGA.4 We report some of the interesting results in the four states.

Tamil Nadu: Males were less likely to participate than women. As age increased, the probability of participation increased. Educated persons and land owners were less likely to participate than the illiterate. As the ratio of NREGA wage to agricultural wage increased, participation likelihood also increased. As inequality of landholdings in a village increased, the probability of participation decreased but this effect disappears at high inequality. The probability of participation decreased as the average distance to the worksite increased to over the maximum distance allowed by NREGA (5 kms).5 Individuals living in villages with a high incidence of attendance at public meetings had a higher probability of participating in NREGA than those living in villages with low attendance, thus demonstrating the importance of these meetings in generating awareness of the program.

Andhra Pradesh: As age increased, the probability of participation increased but the effect diminishes after a certain age. Compared to the illiterates, educated individuals were less likely to participate. SC, ST, and OBC individuals were more likely to participate as compared to others.6 Those who owned land were less likely to participate than the landless. Households with high number of adult males were more likely to participate while households with (p.197) high number of adult females were less likely to do so. As the ratio of NREGA wages to agricultural wages increased, the probability of participation increased but at a diminishing rate. As average distance to the worksite increased, the probability of participation decreased. Participation rose with higher attendance at public meetings in a village but at a diminishing rate.

Madhya Pradesh: Those more likely to participate in NREGA were women (compared to men), young and middle-aged (compared to the old), and married persons. Educated persons were less likely to participate as were those who owned land. Those belonging to households with large number of adult males were less likely to participate. Participation increased with an increase in the ratio of NREGA wage to agricultural wages, since it made NREGA work more lucrative than agricultural work. Somewhat surprisingly, participation increased with distance but at a diminishing rate, presumably because of the small range of variation of distance to worksite.

Rajasthan: Many of the results are similar to those in the other states. As age increased, the likelihood of participating in NREGA increased but diminished after a certain age. Compared to illiterates, educated persons were less likely to participate. SC and OBC castes were more likely to participate, compared to other castes.7 Those belonging to households with a large number of adult females were less likely to participate. Participation increased with an increase in the ratio of NREGA wage to agricultural wages, and in villages with high attendance in public meetings. Participation decreased with the increase in inequality of landholdings in a village, implying that villages with larger numbers of landless and/or a few with large landholdings saw lower participation.8 This result could be due to several reasons including the possibility that the powerful landlords prevent these groups from participating in NREGA. The ethnographic research in MP and Rajasthan revealed that the farmers were unhappy with the fact that they had to increase agricultural wages (the average prevailing rate was Rs 25–40) in order to compete with the mandatory NREGA wage of Rs 100. (p.198)

Table A3.1 Characteristics of Sarpanches in 25 Villages per State (percentage)

Sarpanch Characteristics

MP

Rajasthan

AP

TN

Gender

Male

64

8 (32%)

16 (64%)

14 (56%)

Female

36

17 (68%)

9 (36%)

11 (44%)

Age

Caste

SC

0

2 (8%)

4 (16%)

13 (52%)

ST

13 (52%)

13 (52%)

1 (4%)

28 (37%)

OBC

5 (20%)

7 (28%)

16 (64%)

20 (27%)

Others

7 (28%)

3 (12%)

4 (16%)

10 (13%)

Education

Illiterate

2 (8%)

2 (8%)

2 (8%)

3 (12%)

Sign only

2 (8%)

7 (29%)

4 (17%)

0

Below Secondary

5 (20%)

6 (25%)

2 (8%)

1 (4%)

Secondary

9 (36%)

6 (25%

2 (8%)

8 (32%)

Higher Secondary

6 (24%)

0

9 (38%)

8 (32%)

Graduate

1 (4%)

3 (13%)

2 (8%)

1 (4%)

Post graduate

0

0

3 (13%)

2 (8%)

Below Primary

0

0

0

2 (8%)

Not elected before

84

75

50

Heard about social audit

100

100

92

64

Social audit was conducted in GS here

100

100

68

46

If yes, how often has social audit been conducted

Once = 57

4

59

55

Twice = 30

44

24

0

Over 3 times= 13

52

16

45

Has GP formed a committee to look after NREGA work? Yes

84

100

76

If yes, specify number of members

5 (in 63% of of villages)

4 (in 72% of villages)

3 (in 86% villages)

How often does the GP NREGA committee meet per month?

Once or twice= 38

88

22

More than twice= the remaining

Four times=44

Panchayat distributes NREGA wages

80

0

96

If yes, does it distribute wages on time?

40

100

If no, why?

Funds are not disbursed from the centre on time

84

Other work to look after

16

Who assesses work in NREGA?

Pradhan

Panchayat

BDO

VLW

District officer

(p.199) (p.200)

Table A3.2 Political Reservations and Transfers

Dependent variable

Uniform Coeff (t–value)

Proportional Coeff (t–value)

Mandatory Coeff (t–value)

Explanatory variables

TN

AP

MP

Rajasthan

TN

AP9

MP

Rajasthan

TN

AP

MP

Rajasthan

Predicted Participant

0.257*

1.01***

0.793***

0.208

0.598***

1.54***

1.53***

−0.209

1.25**

2.03***

1.59***

1.61***

(1.72)

(4.56)

(4.00)

(1.01)

(2.96)

(3.63)

(4.15)

(–0.55)

(2.34)

(3.41)

(4.74)

(3.05)

Gender

−0.104*

.101

0.088

(–1.65)

(1.50)

(1.22)

Household size

−0.028*

0.032**

(–1.78)

(1.96)

Marital status: Married

−0.258*

0.272

(–1.87)

(1.12)

Illiterate

−0.118**

0.120

−0.213**

0.084

0.259**

−0.294*

−0.332*

−0.108

(–1.93)

(1.57)

(–2.04)

(0.71)

(1.97)

(–1.82)

(–1.86)

(–1.00)

Age

−0.004

0.017**

(–0.69)

(2.48)

SC

0.147***

−0.385**

−0.400**

0.240*

−0.038

−0.33**

−0.237

(2.41)

(–2.19)

(–1.94)

(1.71)

(–0.37)

(–2.39)

(–1.57)

ST

−0.441**

−0.304*

−0.269***

(–2.35)

(–1.68)

(–2.75)

OBC

−0.291*

−0.356**

(–1.64)

(–1.96)

Land owned

0.066**

0.076***

−0.002

−0.043

0.054**

0.126***

−0.041

0.079*

0.084***

−0.008

(2.60)

(4.52)

(–0.08)

(–1.26)

(1.98)

(5.63)

(–0.58)

(1.82)

(4.54)

(–0.20)

Number of adult males in household

−0.195***

−0.043

−0.112***

−.104*

(–5.08)

(–1.32)

(–2.76)

(–1.84)

Number of adult females in household

0.005

−0.019

−0.083*

−0.088

(0.12)

(–0.53)

(–1.96)

(–1.38)

All SC reserved villages

0.306***

.264**

(4.64)

(2.36)

All SC/ST reserved villages

0.519*

0.649**

(1.87)

(2.46)

Illiterate sarpanch

−0.127**

−0.162

−0.149**

−0.430***

−0.479***

(–1.98)

(–1.95)

(–2.23)

(–2.84)

(–2.76)

Below primary school edn sarpanch

−0.041

(–0.50)

Secondary school educated sarpanch

0.011

(0.09)

Education of sarpanch

0.148**

−0.063

0.069***

0.746***

(2.48)

(–1.54)

(4.45)

(4.86)

Ppartial

0.016

0.521**

0.474*

3.73***

(0.09)

(2.33)

(1.88)

(3.95)

Woman sarpanch

0.706**

1.35**

0.246**

0.388***

4.28***

(2.47)

(2.04)

(2.06)

(3.99)

(4.03)

Land Gini index

−2.57**

−0.308

0.704**

0.305*

0.715

−4.17**

−6.67*

1.78***

−1.39**

1.13**

6.65***

(–2.23)

(–0.82)

(2.31)

(1.74)

(1.22)

(–2.10)

(–1.86)

(2.97)

(–2.32)

(2.11)

(3.85)

Land Gini squared

7.80**

(2.12)

Woman sarpanch of villages with high inequality of landholdings

−2.79***

−7.47***

(–2.74)

(–4.12)

Perception of high corruption

−1.97**

0.162

−0.653*

(–2.22)

(1.06)

(–1.78)

Participant attends public meeting

2.30*

(1.70)

Large number of contestants for sarpanch post

115.06**

−158.01**

−1791.2***

(2.46)

(–2.19)

(–2.73)

Extremely large number of contestants for sarpanch post

333048.5**

(2.34)

7,253 = (3.39)

14,275 = (3.69)

10,156 = (4.97)

6,158 = (3.34)

9,236 = (7.34)

10,279 = (5.41)

7,136 = (10.32)

8,156 = (3.24)

10,250 = (4.93)

10,279 = (4.32)

8,157 = (15.72)

6,158 = (6.19)

Number of observations

261

290

167

165

246

290

144

165

261

290

166

165

Probability〉F

0.0018

0.0000

0.0000

.0040

0.0000

0.0000

0.0000

0.0019

0.0000

0.0000

0.0000

0.0000

Note: ***, **, * refer to significance at the 1 percentage, 5 percentage, and 10 percentage level, respectively; and w denotes weakly significant (〉10 % level). Figures in the parenthesis are the t-values.

(p.201) (p.202)

(p.203) Appendix A4

We asked the household head to report the per capita monthly consumption expenditure by answering a questionnaire on annual household expenditure on 13 non-food items, monthly consumption expenditure on eighteen food items and on kerosene. Based on these figures, we classified the household into acutely poor, moderately poor, moderately non-poor and affluent (Table A4.1). Please refer to Appendix A3 for details on the construction of the models. In brief, the following method was followed to construct the model. The probabilities of poor and acutely poor participating were obtained from probits run on the full household sample in each state. These probabilities along with other household and village characteristics were used in robust regressions to explain variations in NREGA earnings relative to appropriate transfer norms (uniform, proportional and Rawlsian). We only discuss the results of the robust regressions in each type of transfer. The probit results for each type of transfer are available with the authors.

Table A4.1 Definition of Different Levels of Poverty Calculated According to Per Capita Monthly Consumption Expenditure

Levels of poverty

Rajasthan

MP

AP

TN

Acute poverty

If per capita monthly consumption expenditure 〈Rs 383

If per capita monthly consumption expenditure 〈Rs 365

If per capita monthly consumption expenditure 〈Rs 299

If per capita monthly consumption expenditure 〈Rs 396

Moderate poverty

If per capita monthly consumption expenditure 〉=383 but 〈Rs 450

If per capita monthly consumption expenditure 〉=Rs 365 but〈Rs 429

If per capita monthly consumption expenditure 〉=Rs 299 but〈Rs 352

If per capita monthly consumption expenditure 〉=Rs 396 but〈Rs 466

Moderate Non-poverty

If per capita monthly consumption expenditure 〉=Rs 450 but Rs 〈585

If per capita monthly consumption expenditure 〉=Rs 429 but Rs 〈558

If per capita monthly consumption expenditure 〉=Rs 352 but Rs 〈458

If per capita monthly consumption expenditure 〉=Rs 466 but Rs 〈606

Affluent

If per capita monthly consumption expenditure 〉=Rs 585

If per capita monthly consumption expenditure 〉=Rs 558

If per capita monthly consumption expenditure 〉=Rs 458

If per capita monthly consumption expenditure 〉=Rs 606

All Poor

If per capita monthly consumption expenditure 〈Rs 450

If per capita monthly consumption expenditure 〈Rs 429

If per capita monthly consumption expenditure 〈Rs 352

If per capita monthly consumption expenditure 〈Rs 466

(p.204)

Table A4.2 Poverty Profile of the Sub-sample by State

Poverty Status

TN

AP

MP

Rajasthan

Acutely Poor

68 (113)

19 (61)

103 (286)

35 (167)

Moderately Poor

26 (76)

31 (82)

37 (74)

19 (55)

Moderately non-poor

65 (132)

50 (156)

29 (71)

29 (103)

Affluent

61 (179)

60 (201)

31 (69)

77 (175)

Total

220 (500)

160 (500)

200 (500)

160 (500)

Note: Figures in brackets represent the numbers from the full sample of 500 households per state.

(p.205)

Table A4.3 Poor and Acutely Poor Participants and Transfers

Dependent variable:

Rawlsian Marginal (t-value)

Poor Marginal (t-value)

Explanatory variables

TN

AP

MP

Rajasthan

TN

AP

MP

Rajasthan

Predicted Participation

0.328**

0.841**

0.878**

−0.223

0.813***

1.20***

1.36***

−0.216

(1.76)

(2.35)

(2.54)

(1.33)

(2.93)

(3.33)

(3.17)

(–0.49)

Gender

−0.095**

0.048

−0.0788

0.079

−0.144

−0.026

(−2.25)

(0.77)

(–0.63)

(1.20)

(–1.05)

(–0.24)

Marital status: Married

−0.287

−0.0545

−0.457**

(–1.48)

(–1.13)

(−2.25)

Illiterate

−0.091**

−0.042

−0.183**

−0.011

−0.057

0.223*

(−2.22)

(–0.38)

(−2.46)

(−0.18)

(–0.46)

(1.64)

SC

−0.140**

0.289

0.071

0.158**

−0.233

0.168

−0.114

(–2.35)

(0.64)

(1.65)

(1.90)

(–1.39)

(0.50)

(–1.10)

ST

0.0831

0.021

−0.418*

−0.140

−0.284

0.011

(–0.42)

(0.55)

(–2.28)

(–0.81)

(–1.29)

(0.09)

OBC

−0.267

(–1.55)

Land owned

1.04**

0.056

0.064**

0.129***

(2.26)

(1.60)

(2.47)

(5.04)

Number of adult males in household

−0.077**

0.401**

−0.195***

−0.246***

0.333**

(–2.25)

(2.21)

(–5.30)

(–5.88)

(2.07)

Number of adult females in household

−0.028

−0.171***

−0.321*

0.126***

−0.119***

−0.220

(–1.15)

(–3.41)

(–1.77)

(4.76)

(–2.72)

(−1.28)

Household size

−0.568

(–2.66)

Only network

1.04**

(2.26)

Land Gini index

−2.93**

−0.192**

−0.747

0.874*

−2.568*

−5.00**

(−2.62)

(−2.04)

(–1.22)

(–1.71)

(–1.75)

(–2.09)

Land Gini squared

6.53***

(2.76)

Secondary school educated sarpanch

0.084***

(3.04)

Education of sarpanch

0.080***

0.000

0.171

(3.05)

(0.03)

(1.60)

SC/ST reserved villages

0.091**

−0.487***

0.802**

0.364**

(2.22)

(–2.63)

(2.11)

(2.22)

Unreserved village

−0.425***

(–3.42)

Perception of high corruption in village officials

−0.594

−0.892***

(–1.55)

(–3.14)

Perception that sarpanch is partial to own caste

2.35**

(2.46)

Proportion of contestants in sarpanch elections

−0.287.5***

(–2.76)

Participant belongs to same party as sarpanch

−0.037

(0.53)

Constant

0.372

0.549

1.549

0.379

1.897

2.325

.114

1.88

(5.89)

(4.05)

(2.71)

(4.29)

(5.29)

(5.26)

(1.60)

(3.37)

Number of Observations

89

31

90

37

261

290

110

165

F

3.55

5.30

3.06

4.62

4.24

3.98

3.26

5.24

Prob〉F

0.0059

0.0013

0.0014

0.0009

0.0000

0.0000

0.0004

0.0000

Note: Figures in the bracket represent t-values.

*** refers to significance at the 1 percentage,

** refers to significance at the 5 percentage and

* refers to significance at the 10 percentage level.

(p.206) (p.207)

(p.208) Appendix A5

Table A5.1 The Formal Process of a Social Audit

State

Resource Team

Who Conducts SA?

Records to be Audited

Performance Evaluation

Involvement of the GS

SA Report

Process of Social Audit

AP

Use a computerised randomized process to select resource teams at the state and district levels. For each social audit, for each month, there shall be a separate team. No one shall conduct social audit in his native district. The State Resource Persons will make the payment to the VSAs.

  • – VSA who are educated youth from villages and trained by state resource team (NGOs).

  • – They are expected to stay with the ‘poorest of poor’. The VO of SHG will be requested to make their food on payment of Rs 100 per day per person. Cannot conduct audits in their own villages

Program Officer of NREGA to ensure that all required records of all implementing agencies including previous social audit reports are available to the social auditors at least fifteen days in advance of the scheduled date of meeting of the GS where the social audit findings are revealed.

A Performance evaluation of VSAs is to be conducted.. Also, Special Task Force of State Level Officers from Rural Development Dept. to cross-check the functioning of SRP’s and DRP’s and the findings of Social Audit report

District Collectors directs the Sarpanch to convene the GS on the last day of the social audit.

Reviewed by Commissioner of Rural Development and the format of the report shall be redone keeping in view the facility of follow-up action.

  • – Information collection by resource team

  • – Information dissemination

  • –Presenting the results to the GS on the last day

  • –Data entry of the SA reports within 14 days from the date of conclusion of the social audit.

  • – In August 2011, AP introduced an ordinance to criminally prosecute officers whose malpractices were exposed by social audits.

  • –Conducted once every six months

MP

Every village is supposed to create a VMC. NGOs employed by government to sensitize the VMCs.

The social auditors include members of the VMC, the Muster Roll Verification Committee, the Worksite Verification Committee and the Community Mobilisation Committee as well as block level officials. NGO representatives too are in this group.

Same as AP

Same as AP

Same as AP

All Action taken reports shall be filed within a month of convening of the Social Audit

The social audit process is similar to AP except that the VMC collects information from the Gram Panchayat on NREGA related works..

Rajasthan

Constitute VMCs in every village comprising the beneficiaries,

social workers, retired civil/ defence/ private sector officials, other retired employees like teachers, representatives from SC/ST/Women.

Representatives from VMC.

Same as AP

Same as AP

Same as AP

Same as MP

Same as MP.

(p.209) (p.210) (p.211)

Table A5.2 Social Audits in Action

Social Audits

AP

Rajasthan/MP

Jan Sunvai Process

A. SURVEILLANCE

Who audits

VSAs identified and trained by state and district resource persons (NGOs) appointed by the government. (Department of Rural Development)

VSAs, Government representatives including implementing agency

NGOs, VSAs

100 VSAs in teams of 10 cover 2–3 gram panchayats.

Autonomous agency?

Not initially.

In 2010, the government set up an autonomous SSAAT to oversee the social audits and allocated 0.5% of the total NREGA funds for the purpose.

No

NA

What is audited?

  • Financial accounts

  • Muster Rolls

  • Quantity and quality of works in relation to expenses incurred

  • Number of works/materials used

  • Selection and location of work

  • Financial Accounts

  • Muster Rolls

  • Quantity of works and expenses

  • Number of works

Same as AP

Wages: Materials ratio: 80:20

Wages:Materials ratio: 60:40

NA

Process of auditing

Survey conducted over the course of ten days by VSAs who are generally educated youth from labourers’ families.

Only books examined over the course of 1–3 days.

Same as AP

Findings from accounts cross-checked with beneficiaries

No cross checking with beneficiaries.

NGOs involved?

Yes

Initially yes, but subsequently no (except in two areas)

Yes

Nature of Govt.-NGO collaboration

Collaborative

Combative

Collaborative/Combative

B. DISCLOSURE

Disclosure Process

Jan audit manch every six months

GS and/or Jan audit manch every six months.

Jan Sunwai (but not at regular intervals)

Who reads out?

Social Auditors

Sarpanch or Social Auditors

Social Auditors

Who attends?

Majority of villagers, senior state and district officials, media, political representatives, implementing agencies

Government representatives, implementing agencies, a handful of villagers

Same as AP, but also includes eminent persons from media, law, government, academics, politicians etc.

What is in the report?

Assessment of performance and pinpointing of irregularities and assignment of responsibility. Corrective actions suggested

Assessment of books. The audit team does not note down complaints but merely scolds the culprits and exhorts them to do better the next time.

Same as AP.

C. ENFORCEMENT

Follow-up

Some but not sustained.

No follow-up since the reports do not record complaints.

Even when they record complaints, no follow-up occurs.

Same as AP

Irregularities

Decreased

No change (continue to be high)

No change (continue to be high)

Examines what?

Financial, compliance, performance

Financial, some compliance

Same as AP

Outcomes

Outputs

Mode of accountability

Substantive

Procedural (weakly so)

Substantive

Effect

Challenges unequal power relationships

Reinforces status quo (unequal power relationships)

Challenges and tries to transform unequal power relationships

Becomes a social audit

Becomes an inefficient government audit

Becomes a people’s audit

(p.212) (p.213)

(p.214) Appendix A6

Table A6.1 Estimation of Household Attendance in Village Meetings

Dependent Variable: Estimation Methods

Whether Household Attend Village Meeting: Probit Regression

Models

Rajasthan

Andhra Pradesh

Explanatory variables

Coefficients (t-value)

Marginal effects (t-value)

Coefficients (t-value)

Marginal effects (t-value)

Gender

0.52 (1.00)

0.17 (1.17)

−0.56 (−1.19)

−0.10 (−1.62)

Age

0.08 (1.52)

0.03 (1.51)

−0.01 (−0.15)

0.00 (−0.15)

Square of Age

0.00 (−1.39)

0.00 (−1.39)

0.00 (0.16)

0.00 (0.16)

Whether Married

0.03 (0.07)

0.01 (0.07)

1.09* (1.90)

0.36 (1.61)

Below primary education

0.44* (1.95)

0.17* (1.92)

−0.29 (−1.60)

−0.07 (−1.47)

Middle school

0.00 (−0.02)

0.00 (−0.02)

0.92** (2.46)

0.14*** (4.23)

Secondary education

0.43 (1.33)

0.17 (1.30)

0.20 (0.62)

0.04 (0.69)

Higher secondary plus

0.47 (1.57)

0.18 (1.54)

−0.21 (−0.39)

−0.05 (−0.36)

SC

−0.23 (−0.70)

−0.08 (−0.71)

−0.53 (−1.51)

−0.14 (−1.37)

ST

−0.10 (−0.28)

−0.04 (−0.28)

−0.08 (−0.21)

−0.02 (−0.20)

OBC

−0.13 (−0.41)

−0.05 (−0.41)

−0.18 (−0.55)

−0.04 (−0.55)

Household size

−0.07 (−0.49)

−0.03 (−0.49)

0.13 (0.56)

0.03 (0.56)

Square of household size

0.00 (0.01)

0.00 (0.01)

−0.02 (−0.71)

0.00 (−0.70)

Amount of land owned

0.03 (0.78)

0.01 (0.78)

−0.14 (−1.29)

−0.03 (−1.25)

Square of amount of land owned

0.01 (0.91)

0.00 (0.88)

Land Gini index

0.89 (0.26)

0.33 (0.26)

2.23 (0.24)

0.52 (0.24)

Square of Land Gini index

−1.00 (−0.31)

−0.37 (−0.31)

−1.64 (−0.20)

−0.38 (−0.20)

Percentage households attending meetings minus 5%

0.03*** (5.38)

0.01*** (5.50)

0.04*** (9.86)

0.01*** (8.54)

Social networking

0.56* (1.65)

0.22* (1.64)

0.37** (2.20)

0.09** (2.11)

Constant

−3.85** (−2.44)

−2.93 (−1.07)

Number of observations

499

498

Pseudo R-square

0.1058

0.3072

Wald chi-square

42.28***

162.30***

Note: *, **, *** = significance at the 10 percentage, 5 percentage, and 1 percentage level, respectively. This is an ad hoc adjustment to avoid double counting of a household.

(p.215) (p.216)

Table A6.2 Estimation of Household Participation (attendance) in Village Meetings

Dependent variable: Estimation methods

Whether Household Attends Village Meeting: Probit Regression

Models

MP

TN

Explanatory variables

Coefficients (t-value)

Marginal effects (t-value)

Coefficients (t-value)

Marginal effects (t-value)

Gender of Household head

1.74*** (3.41)

0.42 (8.89)

Age of Household head

−0.00 (−0.05)

0.00 (−.05)

Square of Age

0.00 (0.26)

0.00 (0.26)

Whether Married

−0.87*** (−2.37)

−0.33*** (−2.68)

Below primary education

0.54*** (2.97)

0.22*** (3.02)

Middle school

1.05*** (4.03)

0.39*** (4.83)

Secondary education

1.07*** (3.35)

0.39*** (4.16)

Higher secondary plus

0.13 (0.42)

0.05 (0.42)

SC

−0.38 (−1.09)

−0.14 (−1.15)

ST

−0.04 (−0.11)

−0.02 (−0.11)

OBC

−0.21 (−0.64)

−0.08 (−0.65)

Household size

0.14 (1.08)

0.05 (1.07)

Square of household size

−0.01 (−0.93)

−0.00 (−0.93)

Amount of land owned

0.02 (1.26)

0.01 (1.26)

Square of amount of land owned

Land Gini index

1.83 (0.63)

0.71 (0.63)

Square of Land Gini index

−1.48 (−0.62)

−0.58 (−0.62)

Percentage households attending meetings minus 5%1

0.02*** (7.95)

0.01*** (7.85)

Social networking

0.72*** (2.21)

0.28*** (2.39)

Constant

−3.85 (−2.44)

Number of observations

500

Pseudo R-square

0.0000

Wald chi-square

117.41

Log

−1807851

Note: *, **, *** = significance at the 10 percentage, 5 percentage, and 1 percentage level, respectively. This is an ad hoc adjustment to avoid double counting of a household.

(p.217)

(p.218) Appendix A7

Methodology

Measurement of social capital includes measuring the attitudes among community members and participation in community associations (Brehm and Rahn 1997; Jackson and Miller 1998). Associational involvement of households in Indian villages includes participation in village functions, festivals, assemblies, caste groups, self -help groups, women’s’ associations, youth groups, and occupational groups (farmers’ union, workers’ union), political parties, among others. We would like to point out here that there is a strong link between social capital (that is, belonging to a social/political network) and being cognizant of government programs. Reverse causality is ruled out as the balance of evidence suggests that such associational characteristics are accumulated over time and therefore are not endogenous to presence of government programmes. Kenneth Arrow (2000) points out that the essence of social networks is that they are built up over a period of time for reasons other than their economic value to the participant. Dasgupta (2000) says that one may think of social networks as systems of communication channels protecting and promoting personal relationship; we are born into certain networks (for example, caste networks) and we enter new ones over the course of a lifetime. Other studies suggest that participation in SHGs, caste associations, youth and womens’ groups, among others, predated their participation in NREGA. For instance, AP, TN, Kerala, and Karnataka accounted for 57 per cent of the self-help group credits linked during the financial year 2005–6 (Fouillet and Augsburg 2007).

We constructed the variables on social and political networks from Dataset 2–participant and non-participant households interviewed in the subset of eight villages in AP and eleven in TN. The reason for taking the subset was that we had detailed ethnographic information for these villages and were, therefore, able to also construct village (p.219) level indicators of political associational activity for this subset. To do so, we interviewed 20–24 respondents per village who occupied a range of positions, occupations and income levels and, among other questions, asked them whether in the previous year they had attended meetings called by political parties, and/or made financial contributions to a political party—indicators of broader and more regular political participation. We divided the numbers who said yes to each of the questions by the total number (20–25) interviewed for the ethnographic survey in each village and created two village level variables that indicated the level of political networking in each village. We used these variables as indicators of political connectedness in the mlogit equations in both states.10

To construct the social networking variable, we used responses of households in dataset–2 on their participation in caste groups, SHGs, youth and women’s’ associations, and farmers’ associations.

However, the political networking variable posed a challenge. We first assessed the general nature of political participation of TN and AP households in elections and voting by asking them if the respondent had voted in the most recent national election. Voting figures were high across social groups: about 92 per cent in AP and 98 per cent in TN had voted in the most recent state assembly election.11 Since these figures did not show much variation, we could not use them to construct the networking variable.

We, therefore, used another variable—had the individual attended a GS in the previous year—as a proxy for political participation. The quantitative indicator for political participation is attendance in the GS. As mentioned earlier, NREGA tried to ensure a role for vulnerable groups by making the GS a key body in deciding and signing off on projects. In our larger 500 household survey for each of the two states, 75 per cent in AP and 35 per cent of households in TN had attended the GS meetings. Taking the official poverty line as the cut-off point, we found that in both states, more non-poor attended the GS (72 per cent in AP and 63 per cent in TN).12 In the sub-sample, 78 per cent attended in AP and of these, only 28 per (p.220) cent were poor. In TN, 31 per cent attended, and of these 46 per cent were poor.

Our results are consistent with the findings of Rao and Sanyal (2010) who surveyed South Indian villages and found that, on average, about 83 persons attend a GS out of a population that ranges between 2000–10,000 depending on the state. One third of the participants were woman, and 37 per cent were SCs but neither group spoke up in the meetings. Upper castes, on the other hand, were less likely to attend but more likely to dominate the discussions when they did. In another study, Bardhan et al. (2010) found that villages with greater GS participation were also those that delivered more benefits to the landless and the SC/ST population; and villages with lower incidence of landlessness and ST presence exhibited greater GS participation. They point out that while this does not provide evidence of a causal impact of village meetings on targeting of government schemes, it is consistent with the hypothesis that village meetings ‘formed a channel of accountability of village councils (gram panchayat) to poor and low caste groups’. In a study of four south Indian states (including AP and TN), Besley et al. (2004) found that women and illiterates were less likely to have heard of and attended GS meetings, while in contrast, SCs/STs and landless were more likely to attend such meetings but no more likely to have heard of GS meetings; upper castes and the wealthy were more likely to have heard of GS but less likely to attend.

Since it is hard to determine whether belonging to a political network influences the probability of belonging to a social network and vice versa, our method has the merit that it does not assume priority of one over the other. We used a multinomial logit (ML) regression model because we treat our dependent variables as categorical, which allows for simultaneity of participation in various groups.13 This method allows us to compare the probability of membership in only political networks, only social networks, both political networks and social networks as compared to the probability of membership in neither political nor social networks.

(p.221) Thus, if a household head or someone in the household had attended a GS meeting the previous year, we classified the household as belonging to a political network. If someone in the household was a member of a self -help group, credit or savings group, religious groups, trade unions, caste associations, agriculture/milk associations, or a womens’/farmers’/youth association, we classified that household as belonging to a social network. If a household had attended the GS and belonged to one of the groups associated with the social network, it was classified as belonging to both, and if the household had done neither, it was classified as belonging to neither.

Table A7.1 shows that more people in AP, as compared to TN, are networked—over half the sample belonged to social and political networks in the former as compared to only 27 per cent in the latter. About half the TN households did not belong to any network as compared to only 8 per cent in AP. Those belonging only to political networks was higher in AP (27 per cent) than in TN (11 per cent).

The base category was those who did not belong to political or social networks. The associations of being in different networks with household and village characteristics are somewhat weak and uneven. We first ran the models with the same explanatory variables for all the states. However, our ethnographic work and household dataset show that the context varies for some of the variables. Hence, we ran separate regressions for each state, and report the specifications that explain the situation best in a state.

Table A7.1 Distribution of Households in Sub-set (percentage in brackets)

Social Networks

AP

TN

No networks

13 (8%)

103 (47%)

Only social networks

22 (14%)

34 (16%)

Only political network

43 (27%)

23 (11%)

Social and political networks

81 (51%)

59 (27%)

Total

159

219

(p.222) Who Belongs to Social and Political Networks?

To determine probabilities of belonging to different networks, an unordered ML model is used.14 The base category was those who did not belong to a network (political or social), and the other three categories were those who only belonged to social networks, those who only belonged to political networks, and those who belonged to social and political networks. The results given below denote the probability of being in different networks relative to the omitted category of ‘no networks’. The variables in AP were caste (OBC versus non-OBC), whether they owned land, whether the village had middle schools (a proxy for education), whether the villagers attended political meetings, and made financial contributions to political parties. In TN, the variables were illiterate (versus educated), SC/ST (versus OBC and others), whether the individual owned land, whether the distribution of landholdings was unequal in the village (Gini), percentage of households in the village with TVs and cell phones, and whether the villagers attended meetings called by political parties. Villages where attendance in political meetings is high had variations in the proportions of socially and politically networked individuals. In AP, the mlogit model reveals the following15: The probability of households belonging to a social network, relative to no networks, was lower in villages where proportions of households that contributed financially to political parties was high. By contrast, the (relative) probability of a household belonging to political and social networks was higher in villages with higher proportions of households participating in political meetings and contributing financially to political parties.

We get a richer set of results for TN. The (relative) probability of households belonging to social networks varied with the amount of land owned, negatively with inequality in land distribution, positively with higher proportions of households owning cell phones and TVs, and negatively with higher proportions of households attending political meetings in a village. The (relative) probability of being in a (p.223) political network varied with proportions of households in a village owning cell phones and TVs, and inequality in land distribution. The relative probability of being in both—political and social—networks varied with literacy, negatively with SC/ST affiliation, and landowned, positively with inequality of land distribution, negatively with the proportion of households owning cell phones and TVs, and positively with proportions of households in a village attending political meetings. We then used these probabilities in probit models to assess the relationship between being networked and being aware of the program’s components.

Probit Model on Participation in NREGA

We first generated the predicted probability of participation by running a probit on the variables influencing participation in NREGA. We ran this equation on all 500 respondents in each state who included participants and non-participants. As our sub-sample of those interviewed for assessing awareness of the decision-making process and components of NREGA is confined to participants in this scheme, the responses obtained are conditional on participation. Here, along with networking probabilities, the probabilities of participating in NREGA are used as explanatory variable in the awareness and related equations. We then used the predicted probabilities in another probit where the dependent variable was whether the individual was aware of decision- making on NREGA projects.

The probits on participation in NREGA in Table A7.2 reveal the following in AP. The second column focuses on the attributes of the participants in NREGA. As the age increased, the likelihood of participation increased, but declined for old people, which is not surprising since the work (digging and other manual labour) require physical fortitude. Educated persons (primary, middle, and secondary levels) were less likely to participate than illiterates; SCs were more likely to participate than Other (a proxy for upper) castes. (p.224)

Table A7.2 Networks and Awareness of Who Chooses NREGA Projects

AP

TN

Dependent variable

NREGA participation

Whether aware of who decides on projects

NREGA Participation

Whether Aware of who Decides on Projects

Explanatory variables

Coefficient (t-value)

Coefficient (t-value) (marginal effect)

Coefficient (t-value)

Coefficient (t-value) (marginal effect)

Gender

0.200 (1.59)

−0.777*** (−6.94)

0.382 (1.01) (0.008)

Age

0.154*** (5.28)

0.022*** (5.44)

0.000 (0.02) (8.46)

Square of age

−0.002*** (−4.49)

Marital status: Married

−0.233 (−0.96)

1.01*** (7.46)

Illiterate

−0.756 (−1.61) (–0.081)

0.039 (0.27)

0.537 (1.37) (0.015)

Primary education

−0.490*** (−3.14)

−0.276 (−0.71) (–0.033)

0.065 (0.43)

Middle school

−0.700*** (−2.77)

Secondary education

−0.683*** (−3.06)

Higher secondary and above

−0.326 (−1.24)

SC

0.703*** (2.78)

−1.51** (−2.40) (–0.230)

−0.561 (−0.91) (–0.014)

ST

0.212 (0.73)

−1.73** (−2.31) (–0.377)

OBC

0.385 (1.61)

−1.76** (−2.21) (–0.216)

−0.075 (−0.63)

Land owned

−.086** (−2.12)

0.045 (0.48) (0.004)

−0.078 (−1.38)

Land owned squared

Only socially networked

−1.97 (−0.86) (–0.213)

−1.48 (−0.92) (0.070)

Only politically networked

1.82 (1.09) (.196)

10.47* (1.80) (0.266)

Politically and socially networked

3.65** (1.89) (0.394)

2.77* (1.70) (–0.037)

Household size

−0.043 (−1.24)

−0.894** (−2.06) (–0.096)

Household size squared

0.066* (1.75) (0.007)

Ratio of NREGA to AGR WR

0.870** (1.95)

8.36** (2.15)

Ratio of NREGA to AGR WR squared

−5.46** (−2.29)

Land Gini index

−2.47*** (−4.03)

−35.95 (−1.38) (–3.88)

−11.52*** (−4.63)

Land Gini squared

26.81 (1.31) (2.89)

10.04*** (4.01)

Village level indicator of financial contribution to political parties

.038** (2.43)

Average village distance to site

−.233 (−1.47)

Lower Probability of participation (below.5)

−.364 (−0.92) (–.010)

Probability of participation

2.98*** (3.70) (.321)

Constant

−2.01 (−2.57)

13.52 (1.77)

–1.89 (−1.02)

1.02 (1.25)

Number of obs

697

250

892

242

Wald chi2 (15)

247.32

LR chi2

41.35

343.70

25.75

Pseudo R2

0.3092

0.2477

0.3172

0.2542

Log likelihood

−4206401.4

−62.773014

−369.9579

−37.780599

LR test of independent eqations. (rho = 0) that is, chi-square (1)

Prob 〉 chi2 = 0.0000

Prob 〉 chi2 = 0.0002

Prob 〉 chi2 = 0.0000

0.0012

Note: ***,**,* refer to significance at the 1 percentage, 5 percentage, and 10 percentage level, respectively; and w denotes weakly significant (〉 10 % level). Figures in the parenthesis are the t-values.

(p.225) (p.226) As landholdings increased, the likelihood of participation decreased. But as the inequality of landholdings increased in a village, participation decreased. As the ratio of NREGA wages to agricultural wages increased, participation also increased.

In Table A7.2, column 4 focuses on the attributes of the TN participants in NREGA. Unlike in AP, males were significantly less likely to participate than females in TN. Similar to AP, as the age increased, the likelihood of participation increased, but declined for old people. Married persons were more likely to participate than unmarried. This corroborates our qualitative evidence from the worksites that men tended to send their wives to NREGA worksites while they themselves looked for higher paying jobs elsewhere. Education and caste did not return significant results. As the ratio of NREGA wages to agricultural wages increased, participation also increased, but after a certain level, the effect wore off. As the distribution of landholdings in a village became more unequal, the likelihood of participation decreased, but after a certain level of inequality, the effect wore off. Villages where households made contributions to political parties were more likely to have higher participation than villages where the contributions were lower or did not occur. We assume that only those households who are interested in politics (either as party members or attendees in political rallies) would make such financial contributions. This could imply that political parties do operate as channels to disseminate information about NREGA.

We generated the predicted probability of participation from the above equation and inserted it into a second probit where the dependent variable was whether the beneficiary knew who made decisions on the choice of NREGA projects. This is discussed in the main body of this chapter. (p.227)

Table A7.3 Networking and Awareness of Facilities

Dependent variable

AP

TN

Whether they were aware of facilities promised by NREGA++

Aware of safe drinking water

Aware of provision of first aid

Explanatory variables

Coefficient (t-value) (marginal effect)

Coefficient (t-value) (marginal effect)

Coefficient (t-value) (marginal effect)

Gender

−0.255 (1.11)

−0.612** (−2.00)

(0.069)

(–0.171)

Age

−0.007 (−0.81)

0.027** (2.28)

(–0.002)

(0.007)

Square of age

Marital status: Married

Illiterate

−0.186 (−0.73)

0.074 (0.31)

0.267 (0.85)

(–0.058)

(0.021)

(0.070)

Primary education

Middle school

Secondary education

Higher secondary and above

SC

0.293 (1.11)

1.17*** (3.12)

(0.084)

(0.317)

ST

0.579 (1.64)

(0.158)

OBC

−0.319 (−1.58)

(–0.102)

Land owned

0.334*** (3.39)

(0.106)

Land owned squared

Percentage of adults in household

Politically and socially networked

4.15*** (5.09)

1.03* (1.74)

3.41*** (4.42)

(1.31)

(0.294)

(0.878)

Only socially networked

2.99*** (2.99)

1.73* (1.72)

(0.083)

(1.286)

Only politically networked

0.041 (0.05)

0.294 (0.14)

4.99* (1.77)

(.013)

(.847)

(.445)

Household size

Ratio of NREGA to AGR WR

Land Gini index

Probability of participation in NREGA

−0.538 (−0.90)

0.417 (0.77)

−4.02*** (−5.03)

(–.171)

(0.118)

(–1.03)

Social group of sarpanch same as individual

1.24*** (4.72)

(0.439)

Percentage of TVs/cell phones in village

0.040***(4.46)

(0.012)

Average village distance to site squared

Constant

−4.305 (−4.59)

−0.335 (−0.59)

−0.368(–0.51)

Number of obs

291

242

241

Pseudo R2

0.2670

0.0758

0.5274

Log Likelihood

−129.21744

−118.72715

−67.314808

LR chi2

109.18

19.48

150.26

Prob〉chi2

0.0000

0.0125

0.0000

Note: ***,**,* refer to significance at the 1 percentage, 5 percentage, and 10 percentage level, respectively; and w denotes weakly significant (〉10 % level). Figures in the parenthesis are the t-values.

++ In TN, we ran this equation on the probability of knowing about 2 facilities: safe drinking water and first aid.

(p.228) (p.229) (p.230)

Table A7.4 Networking and Complaints

AP

TN

Dependent variable

Whether they complain Coefficient (T-value) (Marginal effects)

Whether they complain

Explanatory variables

Gender

.239 (0.90)

(.060)

Age

−.005 (0.010)

(–0.001)

Illiterate

−0.168 (−0.64)

−.202 (−0.78)

(–0.061)

(–.050)

SC/ST

0.245 (1.04)

−.081 (−0.33)

(0.088)

(0.019)

Land owned

−0.155** (−1.97)

(–0.056)

Politically and socially networked

3.10** (1.90)

1.66*** (2.52)

(1.13)

(0.404)

Only socially networked

9.31*** (4.52)

2.07** (2.33)

(3.40)

(0.505)

Only politically networked

4.79*** (3.56)

3.55** (1.73)

(1.75)

(0.868)

Household size

0.214*** (3.24)

Land Gini index

−0.468 (−0.20)

(.171)

Same caste Sarpanch as the household

0.567 (1.53)

(0.193)

Predicted probability of participation

1.58** (2.32)

1.28** (1.76)

(0.578)

(0.311)

Constant

−2.488 (−0.46)

−2.64 (−3.67)

Number of obs

225

239

Prob〉chi2

0.0000

0.0078

Pseudo R2

0.2180

0.0887

Log Likelihood

−116.75203

−106.68023

LR Chi2

65.26

20.76

Note: ***,**,* refer to significance at the 1 percentage, 5 percentage, and 10 percentage level, respectively; and w denotes weakly significant (〉10 % level). Figures in the parenthesis are the t-values.

(p.231) Appendix A8

Methodology

Political Competition and NREGA Participation

To determine the answer to the first question about targeting of NREGA, we used a probit model. First, we generated the political competition variable, which we describe below. For each village in our sub-sample, we had information on the votes the winner and the runner up got and the total voting population.16 The political competition index was designed as follows:

Political competition index 1= 1-((S12+S22+S32)/N)

S1= votes received by the winner/total voting population, S2=votes received by runner up/total voting population, and S3= 1-(S1+S2) that is, proportion of population that are voting for the residual parties and those who have not voted at all, and N is the number of contestants.

This measure of political competition takes into account the number of contestants but the disadvantage is that it also includes the non-voters. The higher the political competition index, the higher the competition.

Political Competition Index 2 = S1/S2.

This measure takes into account only the voting population voting for the top two candidates. The advantage of this measure is that non-voters are excluded, but the disadvantage is that number of contestants is not included.

For the poverty variable, we computed the ratio of the monthly per capita expenditure of each individual to the poverty line.17 While expenditure and poverty may be the result of individual choices, this ratio is less so. Hence we take it as given (that is, not endogeneous) in the participating equation for NREGA. To assess the influence of political competition on participation in NREGA and on poor (p.232) people’s participation, a probit model was employed (Tables A8.2 and A8.1). The dependent variable was the probability of participation (that is, the result). The independent or explanatory variables (those that explain the result) were the political competition index, ratio of PCME to the poverty line, the combination of these two variables, the individual’s demographic and socio-economic characteristics (age, gender, household size, caste, education, social network, and landholdings), instruments for NREGA participation (ratio of NREGA wages to agricultural wages and/or average distance to the worksite), and the inequality of distribution of landholdings in the village (the Gini). For the other two poverty indices described below, the intensity and severity of poverty, tobit models were used.18

Poverty Indices and Political Competition

We computed three types of poverty indices.19 The first index, the HI measures whether one is below the poverty cutoff point or not. The household survey collected information on the monthly consumption expenditure of each household. We grouped households into poor and non-poor based on whether they were below or above the government’s poverty line figure (see the annex for details on the classification). The second poverty index, the poverty income gap index (PIGI) measures the intensity of poverty. How poor is the poor person? Assume that there are two villages, village A where there are a large number of poor persons and one rich person but the average income is Rs 50 for the village, while in village B there are few poor persons and many non-poor persons and the average income is also Rs 50 for the village. Village A is poorer than village B as the gap between incomes of the poor and the poverty cut-off point is larger. The same principle applies in the case of persons: A is Rs 100 below the poverty line and B is Rs 10 below the poverty line. A is ten times poorer than B and experiences a higher intensity of poverty. The third poverty index, the distributionally sensitive poverty gap (DSPG) measures the severity of poverty. The index counts the poorest differently from those who are moderately poor. Here, greater weight is given to the income (p.233) gap of the poorest, as compared to that of the less poor relative to the poverty line. To construct this index, we take the square of the difference in the second poverty index. Suppose for two persons the gaps are 0.20 and 0.10. For measuring the severity of poverty, the gaps will be (0.20)2 and (0.10)2. So those who are farther away from the poverty line, that is, (0.20) squared and suffering more severe poverty are given higher weights in calculating this index.

The difference between the three indices is that in the headcount ratio, we count one for one (each poor person has the same weight). In PIGI, proportional income gaps or intensities of poverty relative to the poverty line are added up. In SPIG, the income gaps of the poorest would be given highest weight.20 (p.234)

Table A8.1 Political Competition and the Poor’s Participation in NREGA Probit Results (marginal effects)

Dependent variable: Poor Person’s Participation

Marginal Effect (z-value)

Explanatory variables

TN

AP

MP

Rajasthan

Political Competition

327.06** (2.21)

5.24** (2.20)

−322.55*** (−4.81)

507.95* (1.75)

Political competition squared

−180.27** (−2.21)

−6.39*** (−2.75)

182.91*** (4.78)

−273.70* (−1.75)

Gender

−0.196*** (−2.91)

0.066 (0.76)

0.185*** (3.71)

−0.130 (−1.54)

Age

0.024* (1.66)

0.109*** (4.86)

0.043*** (3.67)

0.001 (0.54)

Square of age

−0.0004 (−1.40)

−0.001*** (−4.57)

−0.0005*** (−3.23)

Marital status: Married

−0.010 (−0.09))

Illiterate

0.004 (0.55)

0.223** (2.08)

0.069 (1.15)

0.169 (1.42)

Primary education

0.103 (0.79)

SC

−0.023 (0.27)

−0.057 (0.56)

0.138 (1.29)

0.029 (0.12)

ST

−0.193* (−1.74)

0.525* (3.44)

0.096 (1.28)

OBC

0.106 (0.37)

Land owned

−0.055 (−0.54)

−0.013 (−0.46)

−0.065*** (−4.44)

−0.023 (−0.73)

Number of adult males in household

−0.007 (−0.17)

Number of adult females in household

−0.019 (−0.46)

Household size

−0.016 (0.53)

−0.014 (−1.15)

−0.035** (−1.95)

Social Network

0.072 (0.83)

0.015 (0.11)

0.241** (2.18)

−0.040 (−0.28)

Ratio of NREG to AGR WR

0.371 (0.84)

0.422*** (4.74)

1.36 (1.39)

Average Distance from site

−0.256*** (−2.94)

0.190 (1.70)

0.316 (1.10)

Land Gini index

−0.753***(–2.97)

−2.13 (1.59)

−0.832** (−2.43)

10.86* (1.83)

Land Gini squared

−12.02* (−1.78)

Proportion of villagers who say sarpanch is partial to own caste

0.003 (0.09)

Sarpanch is same caste

0.148(0.98)

Number of Observations

259

184

427

186

LR (Chi2

40.83

64.96

107.07

43.54

Prob〉chi2

0.0000

0.0000

0.0000

0.0000

Pseudo R2

0.1137

0.2549

0.1938

0.1896

Log likelihood

−159.064

−94.961

−222.644

−93.655

Note: Figures in the bracket represent t-values.

*** refers to significance at the 1 percentage,

** refers to significance at the 5 percentage and

* refers to significance at the 10 percentage level.

(p.235) (p.236)

Table A8.2 Political Competition and General Participation in NREGA Probit Results (marginal effects)

Dependent variable: General Participation

Marginal Effect (z-value)

Explanatory variables

TN

AP

MP

Rajasthan

Political Competition

0.163** (1.89)

5.73** (2.49)

−302.53*** (−4.93)

−223.71** (−2.49)

Political competition squared

−90.62** (−1.91)

−4.97*** (−3.72)

171.52*** (4.89)

117.40** (2.48)

Ratio of PCME to Poverty Line

−0.061* (−1.77)

−0.109 (−1.47)

−0.267*** (−2.88)

−0.097** (−2.11)

Gender

−0.296*** (−7.07)

0.113** (2.32)

0.183*** (4.07)

−0.118** (−2.55)

Age

0.041*** (4.20)

0.049*** (4.36)

0.046*** (4.36)

0.046*** (4.60)

Square of age

−0.0004*** (−3.59)

−0.000*** (−4.09)

−0.0005*** (−3.90)

−0.0005*** (−4.63)

Marital status: Married

−0.114 (−1.62)

Illiterate

0.089 (1.61)

0.264*** (4.88)

0.102** (1.88)

0.051 (1.00)

Primary education

SC

0.048 (0.85)

0.136** (2.49)

0.160 (1.60)

0.164* (1.68)

ST

−0.078 (−0.96)

0.240* (1.94)

0.123* (1.87)

OBC

0.151* (1.66)

Land owned

−0.004 (−0.18)

−0.005 (0.34)

−0.029*** (−2.92)

0.0006 (0.06)

Number of adult males in household

−0.017 (−.71)

Number of adult females in household

−0.032 (−1.11)

Household size

−0.015 (0.76)

−0.032*** (−2.74)

−0.033*** (−3.05)

Social Network

0.078 (1.47)

−0.010 (−0.14)

0.252** (2.54)

0.221*** (2.65)

Ratio of NREG to AGR WR

0.558** (2.40)

0.402*** (5.25)

0.008 (0.03)

Average Distance from site

−0.186*** (−3.08)

0.153 (1.43)

Land Gini index

−0.874*** (−4.65)

−24.63 (−1.08)

−0.651** (−2.20)

0.535* (−1.85)

Land Gini squared

22.67 (1.10)

SC/ST reserved village

0.228*** (2.74)

Number of Observations

624

516

502

567

LR Chi2

153.35

135.74

135.25

112.52

Prob〉chi2

0.0000

0.0000

0.0000

0.0000

Pseudo R2

0.1808

0.1919

0.2123

0.1645

Log likelihood

−347.473

−285.813

−250.973

−285.669

Note: Figures in the bracket represent t-values.

*** refers to significance at the 1 percentage,

** refers to significance at the 5 percentage and

* refers to significance at the 10 percentage level.

(p.237)

(p.238) Poverty Indices and Political Competition (8.3–8.6)

Table A8.3 Rajasthan

Dependent Variable Marginal Effects (t-value)

Headcount Probit Gap Probit

Poverty Income

Distributionally Sensitive Poverty Income Gap Probit

Political Competition Household Head’s

−13.38** (−2.58)

−1.01*** (−3.78)

−0.457*** (−3.92)

Education—illiterate

Illiterate

0.299*** (3.39)

0.036*** (3.78)

0.015*** (3.82)

Upto primary education

0.162 (3.34)

0.038*** (2.94)

0.013*** (2.49)

Total number of earners in the household

−0.712*** (−3.57)

−0.009*** (−2.99)

Household size

1.47*** (4.99)

0.014*** (6.31)

0.005*** (5.76)

SC/ST

0.658*** (6.99)

0.083*** (8.64)

0.034*** (8.26)

Distance to nearest town

0.633** (2.78)

0.001*** (3.25)

0.0008** (3.41)

Distance to Block office

Percentage of households that have electricity

Distance to market

0.129 (1.15)

−0.0006 (−1.09)

−0.004* (−1.73)

Land owned

−1.87*** (−6.50)

−0.033*** (−8.38)

−0.014*** (−8.21)

Social network

0.017 (0.66)

−0.011 (−0.82)

−0.005 (−0.93)

SC/ST Reserved village

0.502** (2.31)

0.040*** (3.63)

.0017*** (3.52)

Number of observations

567

567

567

LR chi2

188.01

220.95

215.15

Prob〉 chi2

0.0000

0.0000

0.0000

Pseudo R2

0.2620

0.3982

0.9234

Log likelihood

−264.981

−166.950

−8.927

Note: Figures in the bracket represent t-values.

*** refers to significance at the 1 percentage,

** refers to significance at the 5 percentage and

* refers to significance at the 10 percentage level.

(p.239)

Table A8.4 Andhra Pradesh

Dependent Variable Marginal Effects (t-value)

Headcount Probit Marginal Probability

Poverty Income Gap Tobit

Distributionally Sensitive Poverty Income Gap Tobit

Political Competition

−8.63*** (−4.05)

−0.427*** (−6.28)

−0.131*** (−5.64)

Square of Political Competition

8.27*** (4.46)

0.493 (6.83)

0.153*** (6.19)

Illiterate

−0.001 (−0.02)

0.006 (1.24)

−0.002 (−1.32)

Upto primary education

0.148** (−2.36)

0.009* (1.79)

0.003* (1.89)

Total number of earners in household

−1.22*** (−4.18)

−0.009*** (−4.78)

−0.003*** (−4.94)

Household size

5.13*** (8.04)

0.027*** (16.15)

0.009*** (16.39)

SC/ST

0.062 (0.63)

0.005 (1.11)

0.001 (0.86)

Pucca road

−0.335*** (−4.18)

−0.038*** (−5.22)

−0.011*** (−4.21)

Land owned

−0.322*** (−4.17)

−0.007*** (−5.33)

−0.002*** (−5.71)

Social network

−0.804** (−4.40)

−0.032*** (−4.61)

−0.011*** (−4.83)

SC/ST Reserved village

0.281* (1.87)

0.012 (1.30)

0.005 (1.62)

Percentage of villagers attending public meetings

−3.19*** (−4.51)

−0.001*** (−7.32)

−0.0004*** (−6.95)

Number of observations

516

516

516

LR chi2

255.60

351.72

366.49

Prob〉chi2

0.0000

0.0000

0.0000

Pseudo R2

0.3802

0.9872

−12.9899

Log likelihood

−208.33909

−2.279878

197.35035

Note: Figures in the bracket represent t-values.

*** refers to significance at the 1 percentage,

** refers to significance at the 5 percentage and

* refers to significance at the 10 percentage level.

Table A8.5 Madhya Pradesh

Dependent Variable marginal Effects (t-Value)

Headcount Probit

Poverty Income Gap Tobit

Distributionally Sensitive Poverty Income Gap Tobit

Political Competition

−1.419*** (−3.86)

−0.574*** (−3.00)

−0.308*** (−3.30)

Total number of earners in household

−0.079*** (−2.58)

−0.014*** (−3.68)

−0.008*** (−4.20)

Household size

0.055*** (5.68)

0.033*** (11.69)

0.015*** (10.77)

Illiterate household head

0.024 (1.69)

0.017 (1.55)

0.009* (1.83)

Distance to nearest town

0.122*** (3.47)

0.002** (2.09)

0.001 (2.75)

Land owned

−0.070*** (−4.87)

−0.017*** (−8.19)

−0.007*** (−7.45)

Social network

−0.006 (−2.37)

−0.061*** (−3.45)

−0.034*** (−4.14)

Piped water

−0.055*** (−5.42)

−0.122*** (−8.43)

−0.05*** (−7.93)

SCST reserved village

0.086*** (4.11)

0.062*** (5.18)

0.022*** (3.71)

Lowgini21

0.004 (0.56)

0.007 (0.52)

0.001 (0.28)

Number of observations

623

623

623

LR chi2

219.24

325.98

267.74

Prob〉chi 2

0.0000

0.0000

0.0000

Pseudo R2

0.4043

0.8450

−1.4777

Log likelihood

−161.49005

−29.900961

224.46161

Note: Figures in the bracket represent t-values.

*** refers to significance at the 1 percentage,

** refers to significance at the 5 percentage and

* refers to significance at the 10 percentage level.

(p.240)

Table A8.6 Tamil Nadu

Dependent Variable marginal Effects (T-Value)

Headcount Probit

Poverty Income Gap Tobit

Distributionally Sensitive Poverty Income Gap Tobit

Political Competition

−6.04*** (−2.15)

−0.589*** (−2.56)

−0.238*** (6.11)

Total number of earners in household

−0.505*** (−3.26)

−0.0134*** (−3.00)

−0.006*** (−3.35)

Illiterate household head

0.235*** (3.81)

0.027*** (3.17)

0.010*** (2.92)

Household size

2.66*** (4.32)

0.038*** (3.96)

0.014*** (3.42)

Householdsize.2

−0.967*** (−3.37)

−0.002*** (−2.70)

−0.008*** (−2.21)

Distance to market

0.277*** (2.33)

0.002 (3.38)

0.001*** (3.93)

Land owned

−0.494*** (−5.62)

−0.039*** (−6.89)

−0.014*** (−6.14)

Social network

−0.073 (−1.20)

−0.023*** (−2.60)

−0.012*** (−3.22)

SC/ST

−0.025 (.38)

−0.005 (−.54)

−0.003 (−0.78)

Gini – Land

−0.54 (−0.37)

0.443*** (2.31)

0.307*** (3.75)

Gini – Land- squared

−0.042 (−0.05)

−0.502 (−2.78)

−0.322*** (−4.18)

Pucca road

−0.304*** (−3.03)

−0.073*** (−5.67)

−0.037*** (−6.60)

Number of Observations

624

624

624

LR Chi2

126.79

159.89

166.86

Prob〉chi2

0.0000

0.0000

0.0000

Pseudo R2

0.1497

0.2685

1.0284

Log likelihood

−360.08023

−217.86209

2.3060406

Note: Figures in the bracket represent t-values.

*** refers to significance at the 1 percentage,

** refers to significance at the 5 percentage and

* refers to significance at the 10 percentage level.

(p.241)

Notes:

(1) This is the source for all tables in this chapter.

(2) Ideally, the direct transfer benefit from NREGA should be net of opportunity cost of time (net of earnings foregone in alternative employment). Here we have opted for the assumption that the opportunity cost of time is 0. As the share of NREGA participants with alternative employment opportunities is non-negligible, these estimates may treated as first order approximations (Jha et al. 2011).

(3) Robustness, as Bruce Western (1995) points out, refers to the stability of conclusions in the face of departures from the model’s assumptions. In practice, this means that small changes in the distribution of the data do not result in large changes in the variance of the estimates,. In a regular (OLS) regression, all observations for a particular variable are treated as being of equal value. The problem is that an outlier would have the same value as a non-outlier and would skew the results. A robust regression weighs the observations differently based on how well behaved these observations are. robust regression allows us to take care of heteroscedasticity or, of unequal variances of the unobservable error term in a regression, conditional on the explanatory variables.

(4) For each state, and each type of transfer, we first ran a probit participation model on the full sample to derive the predicted (p.242) probability of participating in NREGA. We report the results of the probit generated for the uniform transfer in each state. The results of the other probits that derived the predicted probability of participation for SCs/STs and women are available on request with the authors.

(5) We use two variables, distance to worksite and the ratio of NREGA wages to agricultural wages, to capture whether transaction costs (for example, travel time) and (relative) attractiveness of NREGA influence participation in it. Both are assumed to be exogenous (Wooldridge 2006).

(6) In our full sample, 31 per cent were SC, 11 per cent were ST, 46 per cent were OBCs and 13 per cent were others.

(7) In our full Rajasthan sample, 27 per cent were SCs, 32 per cent were ST, 35 per cent were OBCs and 6 per cent were others.

(8) Inequality in the distribution of land could take different forms of which these are two (not necessarily) overlapping forms.

(9) There are no villages with ST sarpanches in our sub-sample in AP.

(10) Only 7 per cent of households in MP and Rajasthan were socially networked. Therefore we could not construct this variable for those two states and had to drop them from the analysis. We merged the household datasets of the two states but even this did not work since only 3 per cent were socially and politically networked, another 3 per cent were socially networked, and 68 per cent did not belong to any networks.

(11) Other studies (Krishna 2006) found similar figures for Rajasthan, but rightly point out that such high self-reported voting figures are not reliable indicators of high levels of political participation or active involvement in processes of local self government.

(12) Also see Chattopadhyay et al. (2010).

(13) A ML regression model is a regression that allows more than two discrete outcomes. The model can predict the probabilities of the different possible outcomes of a categorically distributed dependent variable—in this case no networks, only social, only political, both social and political networks—when a set of independent variables is given. An unordered categorically distributed dependent variable is one where the categories cannot be ranked in a meaningful way. In our case, we cannot rank no networks as being worse than belonging to only social networks, and so on (Greene 2007).

(14) For a detailed exposition, see Greene (2007).

(15) As the coefficient estimates could be misleading, it is appropriate to confine our comments to marginal effects. The results of the Mlogits are available with the authors.

(16) This is similar to the Herfindahl Index used to measure the degree of competition in an industry.

(p.243) (17) The factors underlying poverty or expenditure are easy to understand. What causes the relative affluence or relative poverty of a person is less obvious, and hence less endogeneous. Therefore we take the ratio as given.

(18) For details, see Wooldridge (2006).

(19) Sen (1976) refers to the counting of the poor (as in the headcount ratio) as the identification exercise, and the poverty income gap and DSPG as aggregation exercises.

(20) The algebraic version of our model is as follows. We have used three poverty indices: the HI, the poverty gap and the squared poverty gap. The HI simply measures the proportion of the population that is counted as poor, denoted as P 0. Algebraically, P 0 = N p N , where Np is the number of poor and N is the total population in the sample. A person is counted as poor if his/her per capita income/expenditure, yi is less than the poverty line z. A merit of this measure is that it is simple to construct and easy to understand. However, a limitation is that it does not take into account the intensity of poverty or how poor are the poor. A supplementary measure is the poverty gap index. This index adds up the extent to which individuals on average fall below the poverty line, and expresses it as a percentage of the poverty gap. More specifically, define the poverty gap Gi as the difference between the poverty line z and actual income/expenditure yi for poor individuals; and the line is zero for all others.

Then the poverty P 1 = 1 N i = 1 N G i z gap. Some people find it helpful to think of this measure as the minimum cost of eliminating poverty (relative to the poverty line) as it shows how much would have to be transferred to the poor to raise income/expenditure to the poverty line (as a proportion of the poverty line). A limitation of this measure, however, is that it fails to reflect greater poverty in a village with the same poverty gap as another but with more extremely poor persons (or severity of poverty).

To overcome this difficulty, a squared poverty gap is used. This is simply a weighted sum of poverty gaps, where the weights are the proportionate poverty gaps themselves. In contrast with the poverty gap index, where the gaps are weighted equally, the squared poverty gap puts more weight on gaps that are well below the poverty line. Algebraically P 2 = 1 N i = 1 N ( G i z ) 2 .

(21) The mean of the Gini is 0.57, which is very high. So, we created a new variable, low Gini village, which comprised villages with Ginis that were less than 0.543. The Ginis of villages in our sample ranged from 0.37 to 0.73. (p.244)