land reforms- besley burgess
DESCRIPTION
LR in IndiaTRANSCRIPT
LAND REFORM, POVERTY REDUCTION, AND GROWTH:EVIDENCE FROM INDIA*
TIMOTHY BESLEY AND ROBIN BURGESS
In recent times there has been a renewed interest in relationships betweenredistribution, growth, and welfare. Land reforms in developing countries areoften aimed at improving the poor’s access to land, although their effectiveness hasoften been hindered by political constraints on implementation. In this paper weuse panel data on the sixteen main Indian states from 1958 to 1992 to considerwhether the large volume of legislated land reforms have had an appreciableimpact on growth and poverty. We argue that such land reforms have beenassociated with poverty reduction.
I. INTRODUCTION
Finding effective means to relieve poverty is a de�ningmission for development economics. To this end, a wide range ofpolicy alternatives has been implemented. However, the bene�tsof many such efforts have been questioned. Some argue thatpolitical constraints on implementation deny the poor the bene�tsof redistributive efforts. Others suggest that bene�ts to the poorare undermined by disincentives to generate income. Worse still,these disincentives can afflict the nonpoor who try to qualify forassistance. This in turn leads policy analysts to question thewisdom of implementing redistributive policies at all, focusinginstead on policies that promote economic growth. Combattingsuch pessimism requires empirical evidence that some redistribu-tive policies have achieved their stated goals.
This paper studies land reform as a redistributive policy.Throughout the postcolonial period, improvement in the assetbase of the poor has been viewed as a central strategy to relieveendemic poverty {Chenery et al. 1970}. In a poor agrarianeconomy, typical of those in many less developed countries, thisimplies improving the terms on which the poor have access toland. Signi�cant political changes, such as decolonization, havesometimes afforded the opportunity to undertake far-reachingland reforms that transfer property rights to the poor. However,
* The authors are grateful to Alberto Alesina, Ahbijit Banerjee, PranabBardhan, Clive Bell, Francois Bourgignon, Jean Dreze, Lawrence Katz, MichaelLipton, Rohini Pande, Martin Ravallion, a number of seminar participants, andtwo anonymous referees for helpful comments. Timo Henckel and Cecilia Testaprovided able research assistance. We also thank STICERD for invaluable�nancial support.
r 2000 by the President and Fellows of Harvard College and the Massachusetts Institute ofTechnology.The Quarterly Journal of Economics, May 2000
389
such instances are rare, and more incremental measures arecommon. This is the case in India where land reforms have beenon the policy agenda since independence. These reforms haveinvolved only limited efforts at land redistribution, mostly throughlegislated ceilings on landholding. Legislation aimed at regulat-ing tenancies, for example by improving tenurial security, andreducing the power of absentee landlords and intermediaries aremore common. While the latter need not change the distributionof landholdings, they may improve tenants’ claims to the returnsfrom their land. This may also bene�t the landless by raisingagricultural wages.
India is an important case study of land reform. It is bothhome to a signi�cant fraction of the poor in the developing worldand in the postindependence period was subjected to the largestbody of land reform legislation ever to have been passed in soshort a period in any country {Thorner 1976}. The efficacy of thislegislation has, however, been much debated. The conventionalwisdom following the in�uential commentary of Bardhan {1970} isthat, while land reform legislation abounds, the real impact on theconditions of the poor is muted by unenthusiastic implementationof proposed changes. However, broad-based quantitative testingof this notion does not appear to have been attempted previously.This paper takes advantage of the state level panel data availablefor the sixteen main Indian states from 1958 to 1992 to assessthis. The state is the natural unit of analysis for land reform giventhat state governments have jurisdiction over land reform legisla-tion. The relatively long time period covered by the data alsoallows respectable efforts to deal with some econometric concerns.Our principal �nding is that land reforms do appear to have led toreductions in poverty in India. This �nding is robust to a numberof methods of estimation, and the inclusion/exclusion of manydifferent controls.
We also use our data to investigate the relationship betweenland reform and growth. This relates to more general debatesabout how inequality and growth interact. Alesina and Rodrik{1994} and Persson and Tabellini {1995} have argued that initialinequality is bad for economic growth. The link is through thepolitical system—greater inequality encourages redistributiveactivities that blunt accumulation incentives. However, Hoff andLyon {1995}, Banerjee and Newman {1993}, and Benabou {1996},among others, have emphasized that when markets are incom-plete, then redistribution can alter the terms of agency problems
QUARTERLY JOURNAL OF ECONOMICS390
in credit markets and foster accumulation decisions, thus under-mining the standard equity efficiency trade-off. If accumulation isenhanced by redistribution along the growth path, then we wouldexpect to �nd a positive relationship between redistributiveefforts and economic growth. The existing literature has focusedpredominantly on �scal redistributions. By affecting access toland, land reform may have a more lasting effect on poverty. Thisview is consistent with the literature that points to early redistri-butions of land leading to relatively egalitarian access as being animportant precondition for high growth in East Asia (see, forexample, Rodrik {1995}).
Most existing empirical evidence on the links between redis-tribution and growth comes from cross-country data (see Perotti{1996} for a careful review). While informative, there are insur-mountable problems of comparability of data across countries anddealing with concerns about endogeneity. The fact that our datacome from one country with similar data collection strategies ineach state, and the relatively long time period, allow us to makeprogress on this.
Empirical studies of the impact of land reform are rare sincereliable estimation requires data from the pre- and postreformperiods. In India there are numerous case studies of land reform(reviewed below), but few attempts to look at the overall picture.Discussion of the theoretical impact of land reform has beendominated by the frequently found inverse farm size-productivityrelationship, whence small farmers are supposed to achievehigher yields (see Binswanger et al. {1995}). This suggests that�nding means of evening the distribution of landholding shouldlead to productivity gains in addition to redistributive bene�ts.However, land reforms in India are rarely of a form that coulddirectly exploit this possibility. Moreover, careful analyses, suchas Banerjee and Ghatak {1997} show that the theoretical effectson productivity are inherently ambiguous when assessing theimpact of tenancy reforms that allow tenants greater security.
Our main �nding is that there is a robust link between landreform and poverty reduction. Closer scrutiny reveals that, in anIndian context, this is due primarily to land reforms that changethe terms of land contracts rather than actually redistributingland. Consistent with the antipoverty impact, we �nd that landreform has raised agricultural wages. The impact of land reformon growth also depends upon the type of land reform. Overall,there is some evidence that the gain in poverty reduction did come
LAND, POVERTY, AND GROWTH: INDIA 391
at the expense of lower income per capita. We show that all ofthese results are consistent with a simple model of agriculturalcontracting.
The remainder of the paper is organized as follows. The nextsection discusses background and data issues. Section III exam-ines the impact that land reforms have had on poverty and dealswith potential problems in interpreting the basic results. SectionIV addresses the issue as to whether land reforms can havegeneral equilibrium effects by examining their impact on agricul-tural wages. Section V then turns to the issue of how land reformshave affected economic growth. In Section VI we examine theextent to which land reforms have been redistributive in terms oftheir effect on the distribution of land and income. In Section VIIwe develop a theoretical framework that allows us to interpret ourresults in the light of the literature on agricultural contracting.Section VIII concludes. A Data Appendix details the constructionand sources of the key variables used in the analysis.
II. BACKGROUND AND DATA
Under the 1949 Indian Constitution, states were granted thepowers to enact (and implement) land reforms. This autonomyensures that there has been signi�cant variation across statesand time in terms of the number and types of land reforms thathave been enacted (see Table I). We classify land reform acts intofour main categories according to their main purpose (see Mearns{1998}). The �rst category is acts related to tenancy reform. Theseinclude attempts to regulate tenancy contracts both via registra-tion and stipulation of contractual terms, such as shares in sharetenancy contracts, as well as attempts to abolish tenancy andtransfer ownership to tenants. The second category of land reformacts are attempts to abolish intermediaries. These intermediarieswho worked under feudal lords (Zamandari) to collect rent for theBritish were reputed to allow a larger share of the surplus fromthe land to be extracted from tenants. Most states had passedlegislation to abolish intermediaries prior to 1958. However, �ve(Gujarat, Kerala, Orissa, Rajasthan, and Uttar Pradesh) did soduring our data period. The third category of land reform actsconcerned efforts to implement ceilings on landholdings, with aview to redistributing surplus land to the landless. Finally, wehave acts that attempted to allow consolidation of disparatelandholdings. Although these reforms, in particular the latter,
QUARTERLY JOURNAL OF ECONOMICS392
TAB
LE
IS
UM
MA
RY
OF
MA
INVA
RIA
BL
ES
Sta
te
Rur
alpo
vert
yga
p
Rur
alh
ead
cou
nt
Agr
icu
ltu
ral
wag
es
Sta
tein
com
epe
rca
pita
Agr
icu
l-tu
ral
yiel
d
Cu
mu
lati
veto
tall
and
refo
rmle
gisl
atio
n
Cu
mu
lati
vete
nan
cyre
form
legi
slat
ion
Cu
mu
lati
veab
olit
ion
inte
rmed
iari
esle
gisl
atio
n
Cum
ula
tive
land
ceil
ings
legi
slat
ion
Cu
mu
lati
vela
nd
con
soli
dati
onle
gisl
atio
n
And
raP
rade
sh14
.87
50.5
94.
5310
0433
.40
1.52
80.
528
1.00
00
0(5
.11)
(11.
61)
(1.1
0)(2
60)
(33.
11)
(0.5
06)
(0.5
06)
(0)
(0)
(0)
Ass
am10
.69
48.9
15.
3590
350
.54
2.00
00.
611
00.
472
0.91
6(2
.67)
(9.1
6)(1
.04)
(196
)(3
7.59
)(1
.069
)(0
.494
)(0
)(0
.506
)(0
.280
)B
ihar
20.8
864
.65
4.07
633
42.6
44.
305
2.63
90
1.66
70
(4.6
7)(6
.40)
(1.0
1)(1
10)
(39.
95)
(1.9
24)
(0.9
30)
(0)
(1.0
42)
(0)
Gu
jara
t15
.81
53.4
94.
3911
7625
.21
3.05
61.
472
0.66
70.
917
0(4
.94)
(9.9
9)(0
.78)
(272
)(2
3.84
)(1
.264
)(0
.654
)(0
.478
)(0
.280
)(0
)H
arya
na7.
1130
.00
—14
4423
.22
00
00
0(2
.15)
(6.9
0)(3
57)
(20.
46)
(0)
(0)
(0)
(0)
(0)
Jam
mu
and
7.20
34.5
5—
1021
51.5
31.
333
0.47
20
00.
861
Kas
hm
ir(2
.59)
(8.1
3)(2
28)
(43.
28)
(0.7
17)
(0.5
06)
(0)
(0)
(0.3
51)
Kar
nat
aka
16.9
954
.46
3.85
1037
25.2
62.
833
1.41
70
1.41
70
(3.8
6)(8
.08)
(0.6
6)(2
16)
(24.
26)
(1.3
84)
(0.6
92)
(0)
(0.6
92)
(0)
Ker
ala
19.7
056
.92
6.24
864
65.7
55.
444
2.41
71.
972
1.05
60
(7.9
8)(1
4.53
)(1
.56)
(182
)(6
0.26
)(3
.376
)(1
.556
)(1
.000
)(0
.860
)(0
)M
adh
ya18
.03
57.2
63.
8184
317
.01
2.80
60.
944
00.
917
0.94
4P
rade
sh(4
.11)
(7.4
5)(0
.83)
(190
)(1
6.07
)(0
.710
)(0
.232
)(0
)(0
.280
)(0
.232
)M
ahar
ash
tra
19.7
163
.82
3.55
1288
20.8
41.
861
0.97
20
0.88
90
(4.3
8)(9
.64)
(0.7
1)(3
31)
(20.
57)
(0.4
24)
(1.6
67)
(0)
(0.3
19)
(0)
Ori
ssa
17.4
256
.63
4.07
873
25.0
65.
056
1.94
40.
583
1.94
40.
583
(4.6
2)(9
.53)
(0.8
5)(1
86)
(20.
23)
(3.1
16)
(1.0
93)
(0.5
00)
(1.0
93)
(0.5
00)
Pu
nja
b6.
1426
.22
8.16
1732
34.5
50.
583
0.58
30
00
(2.8
8)(8
.23)
(1.0
9)(3
84)
(29.
70)
(0.5
00)
(0.5
00)
(0)
(0)
(0)
Raj
asth
an16
.96
53.4
15.
1278
516
.27
0.94
40
0.94
40
0(3
.81)
(7.4
8)(0
.68)
(136
)(1
5.75
)(0
.232
)(0
)(0
.232
)(0
)(0
)Ta
mil
Nad
u18
.58
58.0
43.
9210
1536
.59
4.91
74.
028
00.
889
0(4
.40)
(8.5
6)(0
.52)
(272
)(3
2.66
)(2
.545
)(2
.336
)(0
)(0
.319
)(0
)U
ttar
Pra
desh
12.8
446
.70
4.71
874
4.64
3.75
01.
417
1.41
70.
917
0(3
.14)
(7.6
8)(1
.38)
(140
)(3
8.23
)(1
.251
)(0
.554
)(0
.554
)(0
.280
)(0
)W
est
Ben
gal
14.9
251
.48
6.12
1173
60.5
96.
139
3.83
30
0.61
11.
694
(5.3
2)(1
2.42
)(1
.81)
(191
)(5
7.20
)(5
.581
)(3
.476
)(0
)(0
.993
)(1
.369
)T
OTA
L15
.01
50.7
94.
799
1030
35.4
92.
910
1.45
50.
411
0.73
10.
312
(6.2
8)(1
4.08
)(1
.584
)(3
46)
(37.
36)
(2.7
49)
(1.7
07)
(0.6
92)
(0.8
25)
(0.6
35)
Sta
nda
rdd
evia
tion
sar
ein
pare
nth
eses
.—de
not
esa
mis
sin
gva
riab
le.S
eeth
eD
ata
App
end
ixfo
rd
etai
lson
con
stru
ctio
nan
dso
urc
esof
vari
able
s.T
he
data
are
for
the
sixt
een
mai
nst
ates
.H
arya
na
spli
tfr
omth
est
ate
ofP
un
jab
in19
65.F
rom
this
dat
eon
,we
incl
ud
ese
para
teob
serv
atio
ns
for
Pu
nja
ban
dH
arya
na.
Th
eex
cep
tion
isru
ral
wag
esw
her
eth
ere
isn
ose
para
tese
ries
for
Har
yan
aor
for
Jam
mu
and
Kas
hm
ir.S
tate
inco
me
per
cap
ita
isob
tain
edby
exp
ress
ing
esti
mat
esof
stat
edo
mes
tic
prod
uct
inre
alpe
rca
pit
ate
rms.
Agr
icu
ltu
raly
ield
mea
sure
sre
pre
sen
tth
era
tio
ofre
alag
ricu
ltu
ral
stat
edo
mes
tic
pro
duct
ton
etso
wn
area
mea
sure
din
thou
san
dsof
hec
tare
s.T
he
wag
ed
ata
refe
rto
the
dail
yw
age
rate
for
mal
eag
ricu
ltu
rall
abor
ers
and
isex
pres
sed
inre
alte
rms.
LAND, POVERTY, AND GROWTH: INDIA 393
were justi�ed partly in terms of achieving efficiency gains inagriculture, it is clear from the acts themselves and from thepolitical manifestos supporting the acts that the main impetusdriving the �rst three reforms was poverty reduction. It istherefore interesting to assess whether these reforms were effec-tive in achieving their stated aims.
Existing assessments of the effectiveness of these differentreforms are highly mixed. Although promoted by the center invarious Five Year Plans, the fact that land reforms were a statesubject under the 1949 Constitution meant that enactment andimplementation was dependent on the political will of stategovernments {Bandyopadhyay 1986; Radhakrishnan 1990; Appu1996; Behuria 1997; Mearns 1998}. The perceived oppressivecharacter of the Zamandari (and their intermediaries) and theirclose alliance with the British galvanized broad political supportfor the abolition of intermediaries and led to widespread implemen-tation of these reforms most of which were complete by the early1960s {Appu 1996; Mearns 1998}.1 Centre-state alignment on theissue of tenancy reforms was much less pronounced.2 With manystate legislatures controlled by the landlord class, reforms thatharmed this class tended to be blocked, although where tenantshad substantial political representation, notable successes inimplementation were recorded. Despite the considerable publicityattached to their enactment, political failure to implement wasmost complete in the case of land ceiling legislation. Hereambivalence in the formulation of policy and numerous loopholesallowed the bulk of landowners to avoid expropriation by distrib-uting surplus land to relations, friends and dependents {Appu1996; Mearns 1998}. As a result of these problems, implementa-tion of both tenancy reform and land ceiling legislation tended tolag well behind the targets set in the Five Year Plans {Bandy-opadhyay 1986; Radhakrishnan 1990}.3 Land consolidation legis-lation was enacted less than the other reforms and, owing partly
1. There were nonetheless some major design �aws, most notably the failureto limit the size of home farms of Zamindars or to protect short-term tenants.
2. Warriner {1969} commented that the Congress party (the main politicalforce for most of our period) ‘‘provided both the motivation for land reform and theopposition to it, as a socialist head with a conservative body.’’
3. The Fifth Plan gives a frank assessment of the situation which is directly inline with that of Bardhan {1970}: ‘‘A broad assessment of the programme of landreform adopted since Independence is that the laws for the abolition of intermedi-ary tenures have been implemented fairly efficiently whilst in the �elds of tenancyreforms and ceilings on holdings, legislation has fallen short of the desiredobjectives, and implementation of the enacted laws has been inadequate’’ {FifthFive Year Plan, 1974–79, 2: 43}.
QUARTERLY JOURNAL OF ECONOMICS394
to the sparseness of land records, implementation has beenconsidered to be both sporadic and patchy only affecting a fewstates in any signi�cant way {Radhakrishnan 1990; Appu 1996;Behuria 1997; Mearns 1998}.
Village level studies also offer a very mixed assessment of thepoverty impact of different land reforms (see Jayaraman andLanjouw {1997}). Similar reforms seemed to have produced differ-ent effects in different areas leaving overall impact indeterminate.There is some consensus that the abolition of intermediariesachieved a limited and variable success both in redistributingland toward the poor and increasing the security of smallholders(see, e.g., Wadley and Derr {1990}). For tenancy reform, however,whereas successes have been recorded, in particular, wheretenants are well organized, there has also been a range ofdocumented cases of imminent legislation prompting landlords toengage in mass evictions of tenants and of the de jure banning oflandlord-tenant relationships pushing tenancy underground andtherefore, paradoxically, reducing tenurial security (see, e.g.,Gough {1989}). Land ceiling legislation, in a variety of villagestudies, is also perceived to have had neutral or negative effects onpoverty by inducing landowners from joint families to evict theirtenants and to separate their holdings into smaller proprietaryunits among family members as a means of avoiding expropria-tion (see, e.g., Chattopadhyay {1994}). Land consolidation is alsoon the whole judged not to have been progressive in its redistribu-tive impact given that richer farmers tend to use their power toobtain improved holdings (see, e.g., Dreze, Lanjouw, and Sharma{1998}).
Table II gives a complete picture of land reform legislation,and its classi�cation, during our data period. Our empiricalanalysis aggregates reforms within each category. If land reformshave any effect, then we doubt that this would be instantaneous.Thus, we cumulate land reforms over time, generating a variablethat aggregates the number of legislative reforms to date in anyparticular state. While crude, we believe that it provides asensible �rst pass at analyzing the quantitative effects of landreform. The mean of that variable aggregated across the fourcategories of land reform is given in column 6 of Table I. Similarmeans for the different categories of reform are given in columns7–10. The table demonstrates considerable variation in overallland reform activity across states with states such as Uttar
LAND, POVERTY, AND GROWTH: INDIA 395
TA
BL
EII
IMP
OR
TA
NT
EV
EN
TS
INL
AN
DR
EF
OR
MS
ININ
DIA
NS
TA
TE
SS
INC
E19
50
Sta
teY
ear
Tit
leD
escr
ipti
onC
lass
And
hra
Pra
desh
1950 (a
men
ded
1954
)
(Tel
enga
na
Are
a)Te
nanc
yan
dA
gri-
cult
ura
lLan
dsA
ctTe
nan
tsre
ceiv
edpr
otec
ted
ten
ancy
stat
us;t
enan
tsto
hav
em
inim
um
term
ofle
ase;
righ
tof
purc
hase
ofn
onre
sum
able
lan
ds;t
rans
fer
ofow
ner
ship
topr
otec
ted
ten
ants
inre
spec
tof
non
resu
mab
lela
nds
;as
are
sult
13,6
11pr
otec
ted
ten
ants
decl
ared
own
ers.
1
1952
Hyd
erab
adA
boli
tion
ofC
ash
Gra
nts
Act
Abo
liti
onof
allt
he
975
jagi
rsin
Tele
nga
na.
2
1954
Inam
Abo
liti
onA
ct(a
bsor
bed)
encl
aves
Abo
liti
onof
inam
s(w
ith
few
exce
ptio
ns).
2
1955
(Hyd
erab
adJa
gird
ars)
Act
Abo
liti
onof
allt
he
975
jagi
rsin
Tal
enga
na.
219
56In
am(A
boli
tion
and
Con
vers
ion
into
Ryo
twar
i)A
ctA
cqu
isit
ion
of11
,137
esta
tes;
abol
itio
nof
1.06
mil
lion
min
orin
ams.
2
1956 (a
men
ded
1974
)
Ten
ancy
Act
Ten
ancy
con
tin
ues
up
to2/
3of
ceil
ing
area
;law
does
not
prov
ide
for
con
fer-
men
tof
own
ersh
ipri
ght
onte
nant
sex
cept
thro
ugh
righ
tto
purc
has
e;co
nfe
rsco
ntin
uou
sri
ght
ofre
sum
ptio
non
lan
dow
ner
s.
1
1957
Inam
Abo
liti
onA
ctA
boli
tion
ofin
ams
(wit
hfe
wex
cept
ions
),st
ruck
dow
nby
the
Hig
hC
ourt
in19
70.
2
Ass
am19
51S
tate
Acq
uis
itio
nof
Zam
inda
riA
ctA
boli
tion
ofin
term
edia
ryri
ghts
invo
lvin
g0.
67m
illi
onh
ecta
res.
219
54L
ush
aiH
ills
Dis
tric
t(A
cqu
isit
ion
ofC
hie
fsR
igh
ts)A
ctS
ame
asab
ove.
2
1956 (a
men
ded
1976
)
Fix
atio
nof
Cei
lin
gon
Lan
dH
oldi
ngs
Act
Sel
f-ex
plan
ator
y.3
1960
Con
soli
dati
onof
Hol
din
gsA
ctIn
trod
uct
ion
ofco
mpu
lsor
yco
nso
lida
tion
.4
1971
Ten
ancy
Act
Cla
ssi�
este
nan
tsin
tooc
cupa
ncy
and
non
occu
pan
cyte
nan
ts;f
orm
erh
asse
curi
tyof
ten
ure
,may
acqu
ire
lan
dlor
d’s
righ
tof
hol
din
gby
payi
ng
50ti
mes
the
lan
dre
ven
ue;
subl
etti
ngis
disa
llow
ed.
1
QUARTERLY JOURNAL OF ECONOMICS396
Bih
ar19
50L
and
Ref
orm
sA
ctA
boli
tion
ofza
min
dari
;im
plem
enta
tion
ofth
isac
tve
rysl
ow.
219
57H
omes
tead
Ten
ancy
Act
Con
fers
righ
tsof
perm
anen
tte
nan
cyin
hom
este
adla
nds
onpe
rson
sh
oldi
ng
less
than
one
acre
ofla
nd.
1
1961 (a
men
ded
1973
)
Lan
dR
efor
ms
Act
Pro
hibi
tssu
blet
tin
g,pr
even
tin
gsu
bles
see
from
acqu
irin
gri
ght
ofoc
cu-
pan
cy.
1
1961
Lan
dC
eili
ng
Act
Impo
siti
onof
ceil
ing
onla
ndh
oldi
ngs
of9.
71–2
9.14
hec
tare
s(1
960–
1972
)an
dof
6.07
–18.
21h
ecta
re(a
fter
1972
).3
1973 (a
men
ded
1982
)
Act
12(a
men
dmen
tto
Lan
dR
efor
ms
Act
)In
trod
uce
dpr
ovis
ion
sre
lati
ng
toth
evo
lun
tary
surr
ende
rof
surp
lus
lan
d.3
1976
Act
55P
rovi
ded
for
the
subs
titu
tion
ofle
galh
eir;
ceil
ing
area
shal
lbe
rede
ter-
min
edw
hen
clas
si�c
atio
nof
land
chan
ges;
orde
red
that
the
lan
dhol
der
nec
essa
rily
reta
inla
ndtr
ansf
erre
din
con
trav
enti
onof
the
Act
.
3
1986
Ten
ancy
(Am
endm
ent)
Act
Pro
vide
sde
�nit
ion
ofpe
rson
alcu
ltiv
atio
n;p
rovi
des
for
acqu
isit
ion
ofoc
cu-
pan
cyri
ghts
byu
nde
rrai
yats
.1
Gu
jara
t19
48 (am
ende
d19
55an
d19
60)
Bom
bay
Ten
ancy
and
Agr
icu
ltur
alL
ands
Act
Ten
ants
enti
tled
toac
quir
eri
ght
ofow
ners
hip
afte
rex
piry
ofon
eye
aru
pto
ceil
ing
area
;con
fers
own
ersh
ipri
ght
onte
nan
tsin
poss
essi
onof
dwel
lin
gsi
teon
paym
ent
of20
tim
esan
nu
alre
nt;
law
does
not
con
fer
any
righ
tson
subt
enan
ts.
1
1960
Agr
icu
ltur
alL
ands
Cei
lin
gA
ctIm
pose
dce
ilin
gon
lan
dhol
din
gsof
4.05
–53.
14he
ctar
es(1
960–
1972
)an
dof
4.05
–21.
85h
ecta
res
(aft
er19
72).
3
1969
Dev
asth
anIn
ams
Abo
liti
onA
ctA
boli
shes
allg
rade
sof
inte
rmed
iary
ten
ure
s,bu
tla
ww
aspa
rtia
lly
inju
nct
edfr
omim
plem
enta
tion
byor
der
ofS
upr
eme
Cou
rt.
2
1973
Am
endi
ng
Act
Pro
vide
sop
port
un
ity
toac
quir
eow
ner
ship
ofho
ldin
gsbu
tla
rgel
yov
er-
ridd
enby
nu
mer
ous
prov
isio
ns.
1
Har
yan
a19
53P
un
jab
Sec
uri
tyof
Lan
dTe
nu
res
Act
Pro
vide
sco
mpl
ete
secu
rity
ofte
nure
for
ten
ants
inco
nti
nu
ous
poss
essi
onof
lan
d( ,
15ac
res)
for
12ye
ars;
gran
tste
nan
tsop
tion
alri
ght
ofpu
r-ch
ase
ofow
ner
ship
ofn
onre
sum
able
lan
d;n
oba
ron
futu
rele
asin
g.
1
1955
Pep
suTe
nan
cyan
dA
gric
ult
ura
lL
and
Act
Sam
eas
abov
e.1
(con
tin
ued
onne
xtpa
ge)
LAND, POVERTY, AND GROWTH: INDIA 397
TA
BL
EII
(CO
NT
INU
ED
)
Sta
teY
ear
Tit
leD
escr
ipti
onC
lass
Jam
mu
1962
Con
soli
dati
onof
Hol
din
gsA
ctIn
trod
uct
ion
ofco
mpu
lsor
yco
nso
lida
tion
.4
and
Kas
hm
ir19
76A
grar
ian
Ref
orm
sA
ctA
llri
ghts
,tit
les,
and
inte
rest
sin
land
ofan
ype
rson
not
cult
ivat
ing
itpe
r-so
nal
lyin
1971
are
exti
ngu
ish
edan
dtr
ansf
erre
dto
the
stat
e;pr
ovid
esfo
rco
nfe
rmen
tof
own
ersh
ipri
ghts
onte
nan
tsaf
ter
allo
win
gre
side
ntla
ndl
ord
tore
sum
ela
nd
for
pers
onal
cult
ivat
ion
.
1
Kar
nat
aka
1954
Mys
ore
(Per
son
alan
dM
isce
llan
eous
)In
ams
Abo
liti
onA
ctA
boli
shed
allt
hela
rge
inam
dari
inte
rmed
iari
es;p
roce
ssof
impl
emen
ta-
tion
very
slow
.2
1955
Mys
ore
(Rel
igio
usan
dC
har
itab
le)
Inam
sA
boli
tion
Act
Sam
eas
abov
e.2
1961
Lan
dR
efor
ms
Act
Pro
vide
sfo
r�
xity
ofte
nu
resu
bjec
tto
lan
dlor
d’s
righ
tto
resu
me
1/2
leas
edar
ea;g
ran
tste
nant
sop
tion
alri
ght
topu
rcha
seow
ner
ship
onpa
ymen
tof
15–2
0ti
mes
the
net
rent
;im
posi
tion
ofce
ilin
gon
lan
dhol
din
gs.
1,3
1974
Lan
dR
efor
ms
(Am
endm
ent)
Act
Impo
siti
onof
ceil
ing
onla
ndh
oldi
ngs
of4.
05–2
1.85
hec
tare
s(a
fter
1972
);re
mov
alof
allb
uton
eof
the
exem
ptio
ns
from
tena
ncy
legi
slat
ion
.1,
3
Ker
ala
1960
Agr
aria
nR
elat
ion
sA
ctA
boli
shes
inte
rmed
iari
es,b
ut
law
stru
ckdo
wn
byS
upr
eme
Cou
rt.
219
63L
and
Ref
orm
sA
ctC
once
des
ten
ant’s
righ
tto
purc
hase
the
lan
dfr
omla
ndow
ner
s.1
1969 (a
men
ded
1979
)
Lan
dR
efor
ms
(Am
endm
ent)
Act
Con
ferm
ent
offu
llow
ner
ship
righ
tson
ten
ants
;2.5
mil
lion
ten
ants
cou
ldbe
com
ela
ndo
wn
ers;
righ
tof
resu
mpt
ion
expi
res;
alth
ough
far-
reac
hin
gon
pape
r,la
w‘‘n
otco
ndu
cive
toso
cial
just
ice’
’bec
ause
ofco
nce
aled
ten
-an
cy;i
mpo
siti
onof
ceil
ing
onla
ndh
oldi
ngs
of6.
07–1
5.18
hec
tare
s(1
960–
1972
)an
dof
4.86
–6.0
7h
ecta
res
(aft
er19
72);
abol
itio
nof
inte
rme-
diar
yri
ghts
.
1,2,
3
1974
Agr
icu
ltur
alW
ork
ers
Act
Cal
led
for
empl
oym
ent
secu
rity
,�xe
dh
ours
,min
imu
mw
ages
,etc
.1
Mad
hya
Pra
desh
1950
Abo
liti
onof
Pro
prie
tary
Rig
hts
(Est
ates
,Mah
als,
Ali
enat
edL
ands
)Act
Abo
liti
onof
inte
rmed
iary
righ
ts.
2
1951
Un
ited
Sta
tes
ofG
wal
ior,
Indo
re,
and
Mal
wa
Zam
inda
riA
boli
tion
Act
Sam
eas
abov
e.2
QUARTERLY JOURNAL OF ECONOMICS398
1951
Abo
liti
onof
Jagi
rA
ctS
ame
asab
ove.
219
52V
indh
yaP
rade
shA
boli
tion
ofJa
girs
and
Lan
dR
efor
ms
Act
Sam
eas
abov
e.2
1959
Lan
dR
even
ueC
ode
Lea
sin
gpr
ohib
ited
;en
titl
esoc
cupa
ncy
tena
nts
toow
ner
ship
righ
tsof
non
-re
sum
able
area
onpa
ymen
tof
15ti
mes
the
lan
dre
ven
ue;
impl
emen
ta-
tion
ofre
form
inef
fici
ent,
one
reas
onbe
ing
that
shar
ecro
pper
san
dte
n-an
tsar
en
otre
cord
ed.
1
1959
Con
soli
dati
onof
Hol
din
gsA
ctIn
trod
uct
ion
ofco
mpu
lsor
yco
nso
lida
tion
.4
1960
Cei
lin
gon
Agr
icu
ltu
ralH
oldi
ngs
Act
Impo
sed
ceil
ing
onla
ndh
oldi
ngs
of10
.12
hec
tare
s(1
960–
1972
)an
dof
4.05
–21.
85h
ecta
res
(aft
er19
72).
3
Mah
aras
htr
a19
50H
yder
abad
Tena
ncy
and
Agr
icu
ltu
ral
Lan
dsA
ctP
rovi
des
for
suo
mot
totr
ansf
erof
owne
rsh
ipto
ten
ants
ofno
nre
sum
able
lan
ds(a
ppli
esto
Mar
atha
wad
are
gion
).1
1958
Bom
bay
Ten
ancy
and
Agr
icu
ltur
alL
and
Act
Pro
vide
sfo
rtr
ansf
erof
owne
rsh
ipto
ten
ants
wit
hn
onre
sum
able
land
s(w
ith
effe
ctfr
om1-
4-96
).1
1961
Agr
icu
ltur
alL
and
(Cei
lin
gon
Hol
d-in
gs)A
ctIm
posi
tion
ofce
ilin
gon
land
hol
din
gs.
3
Ori
ssa
1951
Est
ate
Abo
liti
onA
ctA
imed
atab
olis
hin
gal
lin
term
edia
ryin
tere
sts.
219
72L
and
Ref
orm
sA
ctE
nti
tled
ten
ants
toac
quir
ery
otir
igh
tsov
eren
tire
lan
dh
eld
byth
em.
219
60 (am
ende
d19
73an
d19
76)
Lan
dR
efor
ms
Act
Pro
vide
sfo
r�
xity
ofte
nu
reof
non
resu
mab
lear
ea;p
roh
ibit
ssu
blet
ting
;im
plem
enta
tion
poor
;�n
anci
alhe
lpfo
rpu
rch
ase
ofow
ners
hip
righ
tla
ckin
g;m
ost
leas
esin
form
ofsh
arec
ropp
ing
but
shar
ecro
pper
sno
tre
cord
edas
ten
ants
;im
posi
tion
ofce
ilin
gon
lan
dhol
ding
sof
8.09
–32.
37h
ecta
res
(196
0–19
72)a
nd
of4.
05–1
8.21
hect
ares
(aft
er19
72).
1,3
1972
Con
soli
dati
onof
Hol
din
gsan
dP
re-
ven
tion
ofF
ragm
enta
tion
ofL
and
Act
Intr
odu
ctio
nof
com
puls
ory
con
soli
dati
on.
4
Pu
njab
1953
Pu
nja
bS
ecu
rity
ofL
and
Ten
ure
sA
ctP
rovi
des
com
plet
ese
curi
tyof
tenu
refo
rte
nan
tsin
con
tin
uou
spo
sses
sion
ofla
nd
( ,15
acre
s)fo
r12
year
s;gr
ants
ten
ants
opti
onal
righ
tof
pur-
chas
eof
own
ersh
ipof
non
resu
mab
lela
nd;
no
bar
onfu
ture
leas
ing.
1
(con
tin
ued
onne
xtpa
ge)
LAND, POVERTY, AND GROWTH: INDIA 399
TA
BL
EII
(CO
NT
INU
ED
)
Sta
teY
ear
Tit
leD
escr
ipti
onC
lass
1955
Pep
suTe
nan
cyan
dA
gric
ult
ura
lL
and
Act
Sam
eas
abov
e.1
1972
Lan
dR
efor
ms
Act
Per
mis
sibl
eli
mit
(cei
ling
)is
7h
ecta
res;
5ac
res
ofla
nd
are
secu
red,
the
rest
may
bere
sum
ed;o
ptio
nal
righ
tof
purc
has
eof
own
ersh
ip;s
har
e-cr
oppi
ng
not
con
side
red
ten
ancy
;ten
ants
ofte
nco
erce
dto
‘‘vol
unta
rily
surr
ende
r’’l
and;
lan
dle
ases
not
regi
ster
edu
nde
rpr
ovis
ion
ofte
nanc
yla
ws.
1
Raj
asth
an19
52L
and
Ref
orm
san
dR
esu
mpt
ion
ofJa
gir
Act
Abo
lish
esal
lin
term
edia
ryri
ghts
.2
1953
Bom
bay
Mer
ged
Terr
itor
ies
and
Are
a(J
agir
Abo
liti
on)A
ctS
ame
asab
ove.
2
1954
Hol
ding
s(C
onso
lida
tion
and
Pre
ven
-ti
onof
Fra
gmen
tati
on)A
ctIn
trod
uct
ion
ofco
mpu
lsor
yco
nso
lida
tion
.4
1955
Ajm
erA
boli
tion
ofIn
term
edia
ries
and
Lan
dR
efor
ms
Act
Abo
lish
esin
term
edia
ryin
tere
sts
inot
her
area
s.2
1955
Ten
ancy
Act
Con
fers
secu
rity
ofte
nu
reto
tena
nts
and
subt
enan
ts;o
wn
ersh
ipri
ghts
can
betr
ansf
erre
d;pr
ovis
ions
ofvo
lun
tary
surr
ende
rm
ade
legi
slat
ion
‘‘mer
efa
rce.
’’
1
1959
Zam
inda
rian
dB
isw
edar
iAbo
liti
onA
ctA
boli
shes
inte
rmed
iary
inte
rest
sin
oth
erar
eas.
2
Tam
ilN
adu
1948
Est
ates
(Abo
liti
onan
dC
onve
rsio
nin
toR
yotw
ari)
Act
XX
VI
Ase
ries
ofla
ws
enac
ted
(th
roug
hlo
ng
inte
rval
s)fo
rth
eab
olit
ion
ofva
riou
sty
pes
ofin
term
edia
ries
.2
1952
Th
anja
vur
Tena
nts
and
Pan
naiy
alP
rote
ctio
nA
ctP
rovi
des
grea
ter
secu
rity
ofte
nu
re.
1
1955 (a
men
ded.
1965
)
Mad
ras
Cu
ltiv
atin
gTe
nant
sP
rote
c-ti
onA
ctP
rohi
bits
any
cult
ivat
ing
ten
ant
from
bein
gev
icte
dbu
tal
low
sfo
rre
sum
p-ti
onu
pto
1/2
ofla
nds
leas
edou
tto
tena
nt.
1
1956
Cu
ltiv
atin
gTe
nan
ts(P
aym
ent
ofF
air
Ren
t)A
ctA
boli
shes
usu
ryan
dra
ck-r
enti
ng.
1
QUARTERLY JOURNAL OF ECONOMICS400
1961
(am
.71)
Pu
blic
Ten
ants
Act
Pro
vide
sth
atn
opu
blic
tru
stca
nev
ict
its
cult
ivat
ing
tena
nts.
119
61L
and
Ref
orm
s(F
ixat
ion
ofC
eili
ng
onL
and)
Act
Impo
siti
onof
ceil
ing
onla
ndh
oldi
ngs
of12
.14–
48.5
6h
ecta
res
(196
0–19
72)
and
of4.
86–2
4.28
hec
tare
s(a
fter
1972
).3
1969
Agr
icu
ltur
alL
and-
Rec
ords
ofTe
n-
ancy
Rig
htA
ctP
rovi
des
for
prep
arat
ion
and
mai
nte
nan
ceof
com
plet
ere
cord
ofte
nan
cyri
ghts
.1
1971
Occ
upa
nts
ofK
udi
yiru
ppu
Act
Pro
vide
sfo
rac
quis
itio
nan
dco
nfe
rmen
tof
owne
rsh
ipri
ghts
onag
ricu
ltu
r-is
ts,a
gric
ultu
rall
abor
ers,
and
rura
lart
isan
s.1
1976
Ru
ralA
rtis
ans
(Con
ferm
ent
ofO
wn
-er
ship
ofK
udi
yiru
ppu)
Act
Sam
eas
abov
e.1
Utt
arP
rade
sh19
50 (am
ende
d19
52,1
954,
1956
,195
8,19
77)
Zam
inda
riA
boli
tion
and
Lan
dR
efor
ms
Act
All
ten
ants
are
give
nco
mpl
ete
secu
rity
ofte
nu
rew
itho
ut
any
righ
tof
resu
mpt
ion
for
the
lan
dow
ner
;lea
ses,
inge
ner
al,a
reba
nned
;law
pro-
vide
dfo
rtr
ansf
erri
ng
and
vest
ing
ofal
lzam
inda
ries
tate
s;za
min
dari
was
abol
ish
edov
er60
.2m
illi
onac
res
(ou
tof
tota
lsta
tear
eaof
72.6
mil
-li
onac
res)
.
1,2
1953
Con
soli
dati
onof
Hol
din
gsA
ctIn
trod
uct
ion
ofco
mpu
lsor
yco
nso
lida
tion
.4
1960
Impo
siti
onof
Cei
lin
gson
Lan
dhol
d-in
gsA
ctIm
posi
tion
ofce
ilin
gon
land
hol
din
gsof
16.1
9–32
.37
hec
tare
s(1
960–
1972
)an
dof
7.30
–18.
25h
ecta
res
(aft
er19
72).
3
Wes
tB
enga
l19
50B
arga
dars
Act
Sti
pula
ted
that
the
barg
adar
and
the
lan
dow
ner
cou
ldch
oose
any
prop
or-
tion
acce
ptab
leto
them
.1
1953
Est
ates
Acq
uis
itio
nA
ctL
andh
olde
rsli
mit
edto
ace
ilin
g;pr
ovid
edfo
rab
olit
ion
ofal
lin
term
edia
ryte
nur
es.
1,2,
3
1955 (a
men
ded
1970
,197
1,19
77)
Lan
dR
efor
ms
Act
Pro
vide
sth
atla
ndow
ner
can
resu
me
land
for
pers
onal
cult
ivat
ion
such
that
tena
ntis
left
wit
hat
leas
t1
hec
tare
;sh
arec
ropp
ing
not
con
side
red
ten
ancy
(in
Wes
tB
enga
lmos
tte
nan
tsar
esh
arec
ropp
ers)
;pro
vide
sfo
rla
nd
cons
olid
atio
nif
two
orm
ore
lan
dow
ner
sag
ree.
1,4
1972
Acq
uis
itio
nan
dS
ettl
emen
tof
Hom
e-st
ead
Lan
d(A
men
dmen
t)A
ctTe
nan
tsof
hom
este
adla
nds
are
give
nfu
llri
ghts
.1
1975
Acq
uis
itio
nof
Hom
este
adL
and
for
Agr
icul
tura
lLab
orer
s,A
rtis
ans
and
Fis
her
men
Act
Ove
r25
0,00
0pe
ople
wer
egi
ven
hom
este
adla
nd
(abo
utei
ght
cen
tsea
ch)
up
toJa
nu
ary
1991
.1
(con
tin
ued
onne
xtpa
ge)
LAND, POVERTY, AND GROWTH: INDIA 401
TA
BL
EII
(CO
NT
INU
ED
)
Sta
teY
ear
Tit
leD
escr
ipti
onC
lass
Wes
tB
enga
l(c
ont.
)
1977
Lan
dR
efor
ms
(Am
endm
ent)
Act
‘‘Rai
ses
pres
um
ptio
nin
favo
rof
shar
ecro
pper
s’’{
Yuga
ndh
aran
dIy
er19
93,
p.48
}.1
1981
Lan
dR
efor
ms
(Am
endm
ent)
Act
Des
ign
edto
plu
gth
elo
oph
oles
inth
eea
rlie
rA
cts
rela
tin
gto
the
ceil
ing
ofla
ndh
oldi
ngs
.3
1986
Lan
dR
efor
ms
(Am
endm
ent)
Act
Sou
ght
tobr
ing
allc
lass
esof
land
un
der
the
ceil
ing
prov
isio
ns
byw
ith
-dr
awin
gpr
evio
us
exem
ptio
ns;p
rovi
ded
for
regu
lato
rym
easu
res
toch
eck
indi
scri
min
ate
conv
ersi
onof
land
from
one
use
toan
oth
er;l
awn
otye
tfu
lly
impl
emen
ted.
3
1990
Lan
dR
efor
ms
(Am
endm
ent)
Act
Sam
eas
abov
e.3
Th
eco
nte
nt
ofla
nd
refo
rmac
tsar
ecl
assi
�ed
into
fou
rca
tego
ries
(15
ten
ancy
refo
rm,
25
abol
itio
nof
inte
rmed
iari
es,
35
ceil
ings
onla
ndh
oldi
ngs
,4
5co
nso
lida
tion
ofla
nd
hol
din
gs),
wh
ere
itis
poss
ible
for
agi
ven
act
tobe
lon
gto
mor
eth
anon
eca
tego
ry.I
nth
eza
man
dar
ilan
dte
nu
resy
stem
,wh
ich
cove
red
56p
erce
nt
ofpr
ivat
ely
own
edla
nd
inB
riti
shIn
dia,
the
lan
dw
asve
sted
inth
ela
nd
lord
know
nas
Zam
inda
r.B
etw
een
him
and
the
real
cult
ivat
orth
ere
wer
ese
vera
llay
ers
ofre
ntr
ecei
vin
gin
term
edia
ries
.Jag
irs
and
inam
sw
ere
free
gran
tsof
subg
ran
tsfr
omth
est
ate
wit
hth
eri
ght
toco
llec
tan
dap
prop
riat
ela
nd
reve
nu
e,th
ough
wit
hth
epa
ssag
eof
tim
e,ja
gird
ars
and
inam
dars
beca
me
the
virt
ual
own
ers.
Inth
eir
con
cept
ion
the
ryot
war
ian
dm
ahal
war
ilan
dte
nu
resy
stem
sdi
dn
otre
cogn
ize
any
inte
rmed
iary
betw
een
the
stat
ean
dth
ecu
ltiv
ator
(th
ough
ryot
san
dm
ahal
sdi
dh
ave
full
righ
tsto
sale
,le
asin
gan
dtr
ansf
erof
lan
d).I
n�
ltra
tion
ofm
oney
len
ders
and
trad
ers
into
agri
cult
ure
and
the
leas
eof
them
tote
nan
tsle
dto
crea
tion
ofan
inte
rmed
iary
clas
sev
enin
area
sty
pi�
edby
thes
ela
nd
ten
ure
syst
ems.
QUARTERLY JOURNAL OF ECONOMICS402
Pradesh, Kerala, and Tamil Nadu having a lot of activity whilePunjab and Rajasthan have very little.
Our poverty data come from a consistent set of �gures for therural and urban areas of India’s sixteen major states spanning theperiod 1958–1992 compiled by Ozler, Datt, and Ravallion {1996}.The measures are based on consumption distributions from 22rounds of the National Sample Survey (NSS) spanning thisperiod. The poverty line is based on a nutritional norm of 2400calories per day and is de�ned as the level of average per capitatotal expenditure at which this norm is typically attained. Twopoverty measures are considered: the headcount index and thepoverty gap.4 Given that NSS surveys are not annual, weightedinterpolation has been used to obtain values between surveys.5
Our study should be seen in the context of a signi�cant overallreduction in poverty throughout our data period—the all-Indiarural headcount measure has fallen from around 55 percent to 40percent, and the rural poverty gap from 19 percent to around 10percent. That said, there is considerable cross-sectional variationin performance across states.6 Agricultural wage data were alsocollected to enable us to examine whether land reforms hadgeneral equilibrium effects and were thus capable of reachinggroups of the poor (e.g., landless laborers) who did not directlybene�t from the reforms.
Real values of per capita agricultural, nonagricultural, andcombined state domestic product are also available to examine thedeterminants of growth. Agricultural state domestic product wasde�ated using the Consumer Price Index for Agricultural Labor-ers while the Consumer Price Index for Industrial Workers wasused to de�ate the nonagricultural state domestic product. Wealso constructed a variable to measure agricultural yields. Thiswas de�ned as real agricultural state domestic product divided bythe net sown area. This crudely captures technological changes inagriculture.
Public �nance data at the state level were also collectedchie�y as a means to control for other government interventionsbesides land reform. On the expenditure side, the main classi�ca-
4. The headcount index is the proportion of the population living below thepoverty line. The poverty gap is the average distance below the line expressed as aproportion of the poverty line, where the average is formed over the wholepopulation (counting the nonpoor as having zero distance below the line).
5. Below, we check that our results are robust to including only those yearswhere there was an NSS survey round.
6. See Datt and Ravallion {1998} for further discussion.
LAND, POVERTY, AND GROWTH: INDIA 403
tion available for our data period is into development andnondevelopment expenditure. While development expendituredoes include expenditure on economic and social services, there isno particular connection between this category and government’sefforts to develop the population or infrastructure in their states.7
Development expenditures are therefore further disaggregatedinto health and education expenditures that we might expect tohave appreciable impacts on poverty. We put these in real percapita terms. We also collected total state taxes as a share of statedomestic product as a crude measure of the size of state govern-ments and state redistributive taxes per capita8 to capture theeffort of state governments to redistribute from rich to poor.Population estimates from the �ve censuses for 1951, 1961, 1971,1981, and 1991 were used as additional controls. Between any twocensuses these were assumed to grow at a constant (compound)rate of growth, derived from the respective population totals.
III. LAND REFORM AND POVERTY REDUCTION
A. Basic Results
The empirical approach is to run panel data regressions of theform,
(1) xst 5 a s 1 b t 1 g yst 1 c lst 2 4 1 e st,
where xst is some measure of poverty in state s at time t, a s is astate �xed effect, b t is a year dummy variable, yst is a vector ofvariables that we treat as exogenous (detailed below), lst 2 4 is thestock of past land reforms four periods previously, and e st is anerror term which we model as AR(1) process where the degree ofautocorrelation is state-speci�c; i.e., e st 5 r s e st 2 1 1 ust. Estimationvia generalized least squares will also allow for heteroskedasticityin the error structure with each state having its own errorvariance.
Equation (1) is a reduced-form model of the impact of landreform. Thus, any effect of land reform on poverty is picked up by
7. Economic services include agriculture and allied activities, rural develop-ment, special area programs, irrigation and �ood control, energy, industry andminerals, transport and communications, science, technology, and environment.Social services include education, medical and public health, family welfare, watersupply and sanitation, housing, urban development, labor, and labor welfare,social security and welfare, nutrition, and relief on account of natural calamities.
8. These include land tax, agricultural income tax, and property tax all ofwhich are under the control of state governments.
QUARTERLY JOURNAL OF ECONOMICS404
that variable along with other effects that change the claims thattenants have to land. The land reform variable will also pick upany general equilibrium effects of land reform through changes inwages and prices. Below, we discuss what kind of theoreticalmodel is consistent with our empirical �ndings.
The approach is also reduced form because land reformlegislation is used as regressor—we are unable to measurewhether land reforms are actually implemented. We cannotdistinguish, therefore, between ineffective and unimplementedland reforms. Even though we have no measure of this, there isanecdotal evidence that some land reforms were not fully imple-mented. Hence, the coefficient on land reform in (1) is likely toprovide a lower bound on the true effect of an implemented landreform. We have lagged the land reform variable four periods fortwo main reasons.9 First, because even effective legislation willtake time to be implemented and to have an impact. Second, itmay help to allay concerns that shocks to poverty will be corre-lated with land reform efforts, an issue to which we return below.Fixed effects at the state level control for the usual array ofcross-state differences in history and economic structure thathave been constant over our sample period, while the year effectscover for macro-shocks and policies enacted by the central govern-ment that affect poverty and growth.
Table III gives the basic picture from our data. In column (1)we control for other factors affecting poverty only by using stateand year effects. Land reform is represented only by the cumula-tive land reform variable where all types of land reforms areaggregated. The negative and signi�cant association betweenland reform and the rural poverty gap measure is clear from this.Column (2) con�rms that this result is not sensitive to using theinterpolated years when there were no NSS rounds. In column (3)land reforms are disaggregated into their component types, alsolagged four periods. This suggests that tenancy reforms and theabolition of intermediaries are driving the aggregate effects, whileland ceiling legislation and consolidation of landholdings have anegligible impact on rural poverty. Below, we will suggest atheoretical interpretation of the results that is consistent withthis �nding. The fact that land ceiling legislation is unimportantcon�rms anecdotal accounts of the failure to implement thesereform measures in any serious way {Bardhan 1970; Appu 1996;
9. The results are not sensitive to the exact lag speci�cation chosen here.
LAND, POVERTY, AND GROWTH: INDIA 405
Behuria 1997; Mearns 1998}. Column (4) checks the sensitivity ofthe �ndings to using an alternative measure of poverty—thehead-count index. A similar negative impact of tenancy reformand the abolition of intermediaries on poverty is found here.
If land reform is really responsible for these results (ratherthan some omitted variable that is correlated with land reform),
TABLE IIILAND REFORM AND POVERTY IN INDIA: BASIC RESULTS
Model
Ruralpoverty
gap
Ruralpoverty
gap
Ruralpoverty
gap
Ruralheadcount
Urbanpoverty
gap
Povertygap
difference
Povertygap
difference
Head-count
difference
(1) (2) (3) (4) (5) (6) (7) (8)
GLSAR(1)
GLSAR(1)
GLSAR(1)
GLSAR(1)
GLSAR(1)
GLSAR(1)
GLSAR(1)
GLSAR(1)
Four-year laggedcumulative landreformlegislation
2 0.281(2.18)
2 0.443(3.21)
0.085(1.05)
2 0.534(5.24)
Four-year laggedcumulativetenancy reformlegislation
2 0.604(2.52)
2 1.378(3.13)
2 0.736(3.27)
2 1.916(4.37)
Four-year laggedcumulativeabolition ofintermediarieslegislation
2 2.165(4.08)
2 4.354(4.11)
2 1.327(2.59)
2 3.364(3.73)
Four-year laggedcumulative landceilinglegislation
0.089(0.11)
0.734(0.86)
0.230(0.61)
0.888(1.14)
Four-year laggedcumulative landconsolidationlegislation
0.456(0.82)
2 0.208(0.19)
2 0.210(0.42)
2 1.737(1.62)
State effects YES YES YES YES YES YES YES YESYear effects YES YES YES YES YES YES YES YESNumber obser-
vations507 300 507 507 507 507 507 507
z-statistics are in parentheses. See the Data Appendix for details on construction and sources of thevariables. The data are for the sixteen main states. We use data 1961–1992 for fourteen states. For Haryanawhich split from the Punjab in 1965, we use data 1965–1995 and for Jammu and Kashmir we use data1961–1991 as there was no NSS survey in 1992. This gives us a sample size of 507. The sample size in column(2) is smaller as it is only run for years when NSS surveys were carried out. Poverty measures in otherregressions have been interpolated between survey years. The GLS AR(1) model allows a state-speci�c AR(1)process—see equation (1) in the text for details. In columns (6) and (7) the poverty gap difference is thedifference between the rural and urban poverty gap. In column (5) the headcount difference is the differencebetween the rural and urban head-count index.
QUARTERLY JOURNAL OF ECONOMICS406
then we would not expect to see such effects on urban poverty.There is no good reason to think production and distributiondecisions in the urban sector would be affected (apart from somecomplex general equilibrium reasons). This is con�rmed in col-umn (5) of Table III which �nds no signi�cant negative associationbetween land reform and urban poverty as measured by the sameNSS data. This adds credence to the idea that our land reformvariable is picking up something peculiar to the rural sector.
Columns (6)–(8) use the difference between rural and urbanpoverty as the left-hand-side variable. As we observed fromcolumn (5), urban poverty does not respond to land reform. Thishelps to control for any omitted variables that have commoneffects on poverty in both places.10 Column (6) con�rms our �ndingthat aggregate cumulative land reforms lagged four periods arenegatively associated with a reduction in rural-urban povertydifference. Results broken out by type of land reform are consis-tent with those for rural poverty: tenancy reforms and theabolition of intermediaries have had a signi�cant impact inclosing rural-urban poverty gap while the impact of the other twotypes of land reform are insigni�cant (column (7)). Using the gapbetween rural and urban head-count index yields similar �ndings(column (8)).
Taken together, these results demonstrate a consistent pic-ture.11 Land reform in general appears to be associated withreductions in rural poverty, with these effects most strongly
10. Unlike poverty levels, it is also a variable that does not trend downwardovertime. In the levels regression the cumulative nature of our land reformvariable makes it difficult to identify its effect separately from a state-speci�c timetrend. Indeed, including state-speci�c time trends as regressors in a poverty levelsregression leads to the land reform variable becoming less signi�cant. However,when the poverty difference is included as the left-hand-side variable, the effectof land reforms remains signi�cant even when state-speci�c time trends areincluded.
11. These results assume that the effects of each land reform work indepen-dently from one another. To re�ect the possibility that packaging of certain reformsis important, we ran our basic speci�cations including interactions between thedifferent types of land reforms. No general pattern emerges from this exercise,although there is some suggestive evidence that undertaking both tenancy reformand abolition of intermediaries together enhances the impact of land reformfurther. However, this �nding is somewhat sensitive to the exact measure ofpoverty used and the inclusion of particular control variables. We also consideredwhether there was a difference between land reforms enacted recently comparedwith those more than ten years ago. To this end, we reran the main resultsseparating out a variable cumulating recent land reforms and those more than tenyears old. We found that both enter negatively and signi�cantly in the povertyregressions, with the older land reforms frequently taking an (absolutely) largecoefficient. Following Moene {1992}, we also investigated whether land reforms inmore densely populated states appeared to have a larger impact on poverty. Formost of the speci�cations that we looked at, this was indeed the case.
LAND, POVERTY, AND GROWTH: INDIA 407
associated with land reforms that seek to abolish intermediariesand reform the conditions of tenancies.
B. Robustness
While these results are clean, they leave two signi�cantconcerns unmet. First, they make no effort to allow for otherpolicies that affect poverty—land reform may be proxying forother policies that are correlated with poverty reduction. Second,land reform could be endogenous and responding to the sameforces that drive poverty. We now address both of these concerns.
Table IV reports results that include an array of additionalcontrols. All regressions now include the population growth rateand agricultural yield lagged four periods. The latter may proxyfor other policies that could have enhanced agricultural productiv-ity and are correlated with land reform. It may also pick upexogenous technological change. Our policy measures are in twocategories: re�ecting the expenditure and tax policies of stategovernments. Our expenditure variables are health expendituresper capita, education expenditures per capita, and other expendi-tures per capita.12 The former two might be thought to beimportant determinants of poverty reduction efforts. On the taxside, we have two rather crude measures that give a picture of thegeneral policy stance of the government in office. State taxesexpressed as a share of state domestic product crudely serve tomeasure the size of the state government. We can also measurehow much the government is intent on designing a tax systemthat is geared toward taxing the better off. We create a measure ofthe progressiveness of the tax system under state control. This isthe sum of land taxes, agricultural income taxes, and propertytaxes expressed in real per capita terms. All policy variables arelagged four periods to give the same timing structure as the landreform variables and to minimize concerns about the possibleendogeneity of these policy variables.
In columns (1)–(6) of Table IV we replicate the regressions ofland reform on poverty including these other policies.13 Irrespec-tive of the speci�cation, state redistributive taxes and state taxshare exert signi�cant negative impacts on rural poverty whereas
12. That is total expenditure excluding health and education.13. We experimented with an array of speci�cations that included a larger
array of controls for government expenditure including those on food security,famine relief, rural infrastructure, and other social services and �ner disaggrega-tions of taxes. Including these variables did not affect our key results in anysigni�cant way so we have decided to use a more parsimonious speci�cation.
QUARTERLY JOURNAL OF ECONOMICS408
TAB
LE
IVL
AN
DR
EF
OR
MA
ND
PO
VE
RT
YIN
IND
IA:
CO
NT
RO
LL
ING
FO
RO
MIT
TE
DP
OL
ICY
EF
FE
CT
S
Mod
el
Ru
ral
pove
rty
gap
Ru
ral
pove
rty
gap
Ru
ral
hea
dco
un
tU
rban
pove
rty
gap
Pov
erty
gap
diff
eren
ceH
ead
cou
nt
diff
eren
ce
(1)
(2)
(3)
(4)
(5)
(6)
GL
SA
R(1
)G
LS
AR
(1)
GL
SA
R(1
)G
LS
AR
(1)
GL
SA
R(1
)G
LS
AR
(1)
Fou
r-ye
arla
gged
cum
ulat
ive
lan
dre
form
legi
slat
ion
20.
378
(3.7
8)0.
037
(0.0
42)
20.
539
(4.6
3)2
1.29
8(5
.04)
Fou
r-ye
arla
gged
cum
ulat
ive
tena
ncy
refo
rmle
gisl
atio
n2
0.56
5(2
.32)
20.
897
(1.9
8)F
our-
year
lagg
edcu
mul
ativ
eab
olit
ion
inte
rmed
iari
esle
gisl
atio
n2
1.79
0(2
.81)
23.
14(2
.48)
Fou
r-ye
arla
gged
cum
ulat
ive
lan
dce
ilin
gle
gisl
atio
n2
0.35
2(0
.82)
20.
121
(0.1
4)F
our-
year
lagg
edcu
mul
ativ
ela
nd
con
soli
dati
onle
gisl
atio
n0.
164
(0.3
2)2
1.00
0(1
.02)
Pop
ula
tion
grow
thra
te2
90.6
1(1
.14)
297
.99
(1.2
1)2
87.5
9(0
.50)
274
.32
(1.2
2)74
.81
(0.9
1)2
145.
05(0
.90)
Fou
r-ye
arla
gged
per
capi
taed
uca
tion
expe
ndi
ture
0.06
3(2
.04)
0.07
0(2
.24)
0.07
6(1
.10)
0.04
1(1
.73)
0.07
7(2
.18)
0.03
4(0
.42)
Fou
r-ye
arla
gged
per
capi
tah
ealt
hex
pend
itur
e0.
038
(0.8
8)0.
041
(0.9
1)0.
072
(0.7
6)2
0.00
3(0
.09)
0.04
2(0
.83)
0.21
8(1
.76)
Fou
r-ye
arla
gged
per
capi
taot
her
expe
ndi
ture
0.02
0(2
.69)
0.01
7(2
.31)
0.02
6(1
.56)
0.01
2(2
.40)
0.00
09(0
.12)
20.
008
(0.4
0)F
our-
year
lagg
edpe
rca
pita
redi
stri
buti
vest
ate
taxe
s2
0.13
0(2
.70)
20.
142
(2.9
2)2
0.36
4(3
.25)
20.
045
(1.2
5)2
0.18
2(3
.53)
20.
422
(3.2
1)F
our-
year
lagg
edst
ate
taxe
sas
ape
rcen
tage
ofst
ate
dom
esti
cpr
oduc
t2
49.5
9(2
.99)
249
.11
(2.9
4)2
87.3
3(2
.46)
227
.70
(2.2
3)16
.43
(0.9
7)4.
790
(0.1
3)F
our-
year
lagg
edag
ricu
ltu
raly
ield
0.00
1(0
.05)
20.
003
(0.0
2)2
0.50
7(1
.19)
20.
006
(0.4
2)0.
031
(1.4
5)2
0.01
3(0
.30)
Sta
teef
fect
sY
ES
YE
SY
ES
YE
SY
ES
YE
SYe
aref
fect
sY
ES
YE
SY
ES
YE
SY
ES
YE
SN
um
ber
ofob
serv
atio
ns
436
436
436
436
436
436
z-st
atis
tics
are
inpa
ren
thes
es.S
eeth
eD
ata
App
end
ixfo
rde
tail
son
con
stru
ctio
nan
dso
urc
esof
the
vari
able
s.T
he
data
are
for
the
sixt
een
mai
nst
ates
.We
hav
eda
ta19
64–1
992
for
nin
est
ates
.F
orP
un
jab
we
hav
eda
ta19
69–1
992,
for
Har
yan
aw
hic
hsp
lit
from
the
Pu
nja
bin
1965
we
hav
eda
ta19
70–1
975
and
1978
–199
2,fo
rJa
mm
uan
dK
ash
mir
1968
–199
1,fo
rB
ihar
and
Gu
jara
t19
64–1
975
and
1977
–199
2,fo
rT
amil
Nad
u19
64–1
975
and
1978
–199
2an
dfo
rB
ihar
1964
,196
9an
d19
72–1
977.
Th
isgi
ves
us
ato
tals
amp
lesi
zeof
436.
Inco
lum
n(5
)th
epo
vert
yga
pd
iffe
ren
ceis
the
diff
eren
cebe
twee
nth
eru
ral
and
urb
anpo
vert
yga
p.In
colu
mn
(6)
the
hea
d-co
un
tdi
ffer
ence
isth
ed
iffe
ren
cebe
twee
nth
eru
ral
and
urb
anh
ead-
cou
nt
inde
x.T
heG
LS
AR
(1)m
odel
allo
ws
ast
ate-
spec
i�cA
R(1
)pro
cess
—se
eeq
uati
on(1
)in
the
text
for
deta
ils.R
edis
trib
uti
veta
xes
are
agri
cult
ura
linc
ome
taxe
s,la
nd
taxe
s,an
dpr
oper
tyta
xes.
LAND, POVERTY, AND GROWTH: INDIA 409
education and health expenditure, per capita yield, and popula-tion growth are generally insigni�cant at conventional levels.14 Incolumn (1) we include the full set of policy control variables in thebasic regression of cumulative land reform on the rural povertygap measure. Despite controlling for these many dimensions ofstate activity, cumulative land reform continues to exert a nega-tive and signi�cant impact on rural poverty. In column (2) we runthe same regression while disaggregating the land reform vari-able. We continue to �nd that tenancy reforms and the abolition ofintermediaries exert a negative and signi�cant impact on therural poverty gap measure whereas land ceiling and land consoli-dation legislation exert no signi�cant in�uence. Replacing thepoverty gap measure with the head-count index as is done incolumn (3) produces a similar set of results. When we examine theurban poverty regression (column (4)), we �nd that, in commonwith the rural poverty regressions, health and education expendi-ture and yield have no signi�cant impact and tax share has asigni�cant impact. State redistributive taxes are insigni�cant inthis regression suggesting that their impact is restricted to therural sector. Inclusion of these extra variables has no effect on theinsigni�cant impact of cumulative land reform on urban poverty.
Columns (5) and (6) regress the difference between rural andurban poverty on cumulative land reform and the full set ofcontrol variables. Note that compared with column (1) of thistable, only the land reform variable and state redistributive taxesremain signi�cant in this speci�cation.15 Other policy effectsappear to be common to both rural and urban sectors, becominginsigni�cant in this regression. Contrasting columns (5) and (6)con�rms that results are robust to the type of poverty measurebeing used. Taken together, the results presented in Table IV offerfurther con�rmation of our initial �nding of a signi�cant negativeassociation between lagged land reform and rural poverty.
A further concern with the speci�cation in equation (1) is thepotential endogeneity of the land reform variable. It is notpossible to ascertain the direction of bias due to this a priori. If
14. The expenditure results are interesting given the priority attached, incurrent debates, to expansion of expenditures on education and health as a keymeans of reducing poverty (see Dreze and Sen {1995}). If anything, educationexpenditures per capita seem to exert a positive in�uence on the rural poverty gap(columns (1) and (2)) but not on the rural head-count ratio (column (3)). However, itis possible that we would need �ner measures of the ways in which particularprograms are prioritized to make progress on this.
15. An exception is education expenditure per capita in the poverty gapspeci�cation (see column (5)).
QUARTERLY JOURNAL OF ECONOMICS410
land reform is purposefully aimed at poverty reduction, then wewould expect policy effort to focus on where poverty is highest,leading to downward bias. However, if responsiveness to landreform is greater where poverty is highest, then the effect may goin the other direction. While lagging land reform as we have in (1)goes some way toward minimizing concerns about this, there issome residual worry that long-lived shocks to poverty that affectantipoverty legislation could bias the results.
To �x this problem requires an instrument for land reform.16
To this end, we exploit the fact that land reform is intenselypolitical, with different groupings in state legislatures (the Vid-han Sabha) being more likely to enact land reform legislation.However, this can be problematic if, as seems likely, shocks topoverty affect who is elected. To mitigate this problem, we proposeusing long lags of the political variables as instruments for landreform. Speci�cally, political variables from four periods prior tothe land reform (eight periods before the poverty observation) areused as an instrument for land reform. This is legitimate providedthat contemporaneous shocks to poverty are uncorrelated withshocks that lead to particular groups being elected eight yearspreviously. Such an assumption seems defensible given both thefrequency of elections and policy shifts in India and because it isdifficult to think of long-lasting shocks affecting both currentpoverty and political structure eight years ago.
This strategy implies a �rst-stage equation for land reform:
(2) lst 5 µlst2 4 1 as 1 bt 1 cyit 1 dzst2 4 1 h st,
where lst is the cumulative land reform variable, as is a state �xedeffect, bt is a year dummy variable, yst is a vector of variables thatwe treat as exogenous, and the variables zst2 4 are politicalvariables re�ecting the seat shares of different political groups,each lagged by four years. These are constructed from records ofthe number of seats won by different national parties at each ofthe state elections under four broad groupings. (The partiescontained in the relevant group are given in parentheses after thename of the grouping.) These are (i) Congress Party (IndianNational Congress 1 Indian Congress Socialist 1 Indian Na-tional Congress Urs 1 Indian National Congress Organization),(ii) a hard left grouping (Communist Party of India 1 Communist
16. This will also help to deal with measurement error which is a concerngiven that we measure only legislated land reforms.
LAND, POVERTY, AND GROWTH: INDIA 411
Party of India Marxist), (iii) a soft left grouping (SocialistParty 1 Praja Socialist Party), and (iv) Hindu parties (BhartiyaJanata Party 1 Bhartiya Jana Sangh). We express these as ashare of total seats in the legislature. Congress has tended todominate the assemblies over the period, although hard leftparties have also recorded majorities in Kerala and West Bengal.Over time there has been a decline in the importance of Congressand a rise in the importance of religious and regional parties.
Table V presents estimates of equation (2) for the differentkinds of land reforms. As we would expect, all cases �nd laggedland reforms to be strongly signi�cant. The political variables alsomatter for land reform legislation, and are jointly signi�cant in allcolumns. In column (1) we see that, relative to the omitted ‘‘other’’category, which is composed of a amalgam of regional, indepen-dent, and Janata parties, Congress and soft left decrease theprobability of enacting land reform legislation, while hard leftexerts a positive in�uence and Hindu parties are insigni�cant.Looking across columns (2) to (5), we see that the negativein�uence of Congress is spread across all types of land reform, butit is particularly pronounced for tenancy reforms and abolition ofintermediaries. The negative in�uence of soft left parties is alsospread across the board with the exception of land consolidation.The overall positive in�uence of hard left parties, however, seemsto originate principally through a strong positive effect on thepassage of land ceiling legislation. This is interesting given ourfailure to �nd evidence that such reforms reduce poverty. Hinduparty representation appears to exert no in�uence on the passageof tenancy reforms or the abolition of intermediaries. However,they exert a signi�cant positive in�uence on land ceiling and asigni�cant negative in�uence on land consolidation.
Table VI presents our results from instrumental variablesestimation. Column (1), which uses political variables and laggedland reforms as instruments, continues to �nd a negative andsigni�cant impact of land reform on the rural poverty gap. We �nda similar result when we use the head-count poverty measure inplace of the poverty gap measure in column (2). Columns (3) and(4) follow a similar instrumentation procedure but break out totalland reforms into constituent types. This con�rms our earlierresults; both tenancy reforms and abolition of intermediariesremain negative and signi�cant while other types of land reform
QUARTERLY JOURNAL OF ECONOMICS412
TABLE VLAND POLICY DETERMINATION
Model
Cumulativetotal land
reformlegislation
Cumulativetenancyreform
legislation
Cumulativeabolition of
intermediarieslegislation
Cumulativeland ceilinglegislation
Cumulativeland
consolidationlegislation
(1) (2) (3) (4) (5)
OLS OLS OLS OLS OLS
Four-year laggedcumulativeland reform leg-islation
0.406(12.23)
Four-year laggedcumulative ten-ancy reformlegislation
0.693(16.26)
2 0.002(0.16)
2 0.009(0.38)
0.021(1.13)
Four-year laggedcumulative abo-lition of inter-mediaries legis-lation
0.041(0.53)
0.664(14.21)
0.109(1.51)
2 0.029(1.06)
Four-year laggedcumulativeland ceiling leg-islation
2 0.131(2.11)
2 0.172(0.65)
0.631(15.60)
2 0.045(1.44)
Four-year laggedcumulativeland consolida-tion legisla-tion
0.694(5.06)
2 0.038(1.14)
0.174(2.93)
0.772(7.85)
Four-year laggedcongress partyshare of seats
2 0.460(2.81)
2 0.472(4.78)
2 0.098(2.37)
2 0.066(1.85)
2 0.075(1.85)
Four-year laggedhard left shareof seats
2.837(2.95)
0.476(0.72)
0.149(0.97)
1.437(5.46)
2 0.302(0.73)
Four-year laggedsoft left share ofseats
2 3.921(3.09)
2 2.363(3.25)
2 1.101(2.60)
2 1.990(3.63)
2 0.426(1.06)
Four-year laggedhindu partiesshare of seats
0.270(0.33)
2 0.089(0.19)
2 0.045(0.15)
0.556(2.01)
2 0.410(2.08)
State effects YES YES YES YES YESYear effects YES YES YES YES YESNumber of obser-
vations474 474 474 474 474
t-statistics are in parentheses. All regressions are reported with robust standard errors. All monetaryvariables are in real terms. See the DataAppendix for details on construction and sources of the variables. Thedata are for the sixteen main states. We have data 1962–1992 for eight states. Punjab and Haryana split intoseparate states in 1965. For Punjab we have data 1962–1989 while for Haryana we have data 1969–1991. ForJammu and Kashmir we have data 1965–1991, for Kerala and West Bengal 1962–1991, for Gujarat andMaharashtra 1963–1992 and for Bihar 1962–1989. This gives us a total sample size of 474. The partiescontained in the relevant group are given in parentheses after the name of the grouping. These are (i)Congress Party (Indian National Congress 1 Indian Congress Socialist 1 Indian National CongressUrs 1 Indian National Congress Organization), (ii) a hard left grouping (Communist Party ofIndia 1 Communist Party of India Marxist), (iii) a soft left grouping (Socialist Party 1 Praja Socialist Party),and (v) Hindu parties (Bhartiya Janata Party 1 Bhartiya Jana Sangh).
LAND, POVERTY, AND GROWTH: INDIA 413
TAB
LE
VI
LA
ND
RE
FO
RM
AN
DP
OV
ER
TY
ININ
DIA
:IN
ST
RU
ME
NT
AT
ION
Mod
el
Rur
alpo
vert
yga
p
Rur
alh
ead
cou
nt
Rur
alpo
vert
yga
p
Rur
alhe
adco
unt
Pov
erty
gap
diff
eren
ce
Rur
alpo
vert
yga
p
Ru
ral
head
coun
t
Rur
alhe
adco
unt
Ru
ral
pove
rty
gap
Ru
ral
head
cou
nt
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
IV1
IV1
IV1
IV1
IV1
IV2
IV2
IV2
IV3
IV3
Fou
r-ye
arla
gged
cum
ula
tive
land
refo
rmle
g-is
lati
on2
0.73
2(6
.02)
21.
360
(5.6
8)2
0.43
8(3
.60)
20.
659
(4.0
9)2
1.19
2(3
.67)
20.
599
(3.1
8)2
1.26
3(3
.24)
Fou
r-ye
arla
gged
cum
ula
tive
ten
ancy
refo
rmle
gisl
atio
n2
0.99
8(3
.16)
22.
404
(3.6
7)2
4.59
5(4
.69)
Fou
r-ye
arla
gged
cum
ula
tive
abol
itio
nof
inte
rmed
iari
esle
gisl
atio
n2
2.27
1(2
.58)
25.
701
(3.6
4)2
7.40
8(4
.10)
Fou
r-ye
arla
gged
cum
ula
tive
land
ceil
ing
leg-
isla
tion
21.
372
(2.3
4)0.
432
(0.3
8)2
1.99
8(1
.89)
Fou
r-ye
arla
gged
cum
ula
tive
land
cons
olid
a-ti
onle
gisl
atio
n1.
624
(1.7
2)1.
969
(1.0
0)2
4.02
7(1
.45)
Ove
ride
nti�
cati
onte
stp-
valu
e0.
930.
980.
990.
980.
990.
930.
980.
980.
920.
96St
ate
effe
cts
YE
SY
ES
YE
SY
ES
YE
SY
ES
YE
SY
ES
YE
SY
ES
Year
effe
cts
YE
SY
ES
YE
SY
ES
YE
SY
ES
YE
SY
ES
YE
SY
ES
Num
ber
ofob
serv
atio
ns41
041
041
041
041
041
041
041
041
041
0
t-st
atis
tics
are
inpa
ren
thes
es.A
llre
gres
sion
sar
ere
port
edw
ith
robu
stst
anda
rder
rors
.See
the
Dat
aA
ppe
ndi
xfo
rde
tail
son
con
stru
ctio
nan
dso
urc
esof
the
vari
able
s.T
he
data
are
for
the
sixt
een
mai
nst
ates
.We
hav
eda
ta19
66–1
992
for
twel
vest
ates
.For
Har
yan
a,w
hic
hsp
lit
from
Pu
nja
bin
1965
,we
hav
ed
ata
1973
–199
2,fo
rJa
mm
uan
dK
ash
mir
1969
–199
2,an
dfo
rG
uja
rat
and
Mah
aras
htr
a19
67–1
992.
Th
isgi
ves
us
ato
tals
ampl
esi
zeof
410.
Inco
lum
n(5
)pov
erty
gap
dif
fere
nce
isth
edi
ffer
ence
betw
een
rura
lan
du
rban
pove
rty.
IV1:
Inst
rum
ents
for
the
endo
gen
ous
poli
cyva
riab
le(c
um
ula
tive
lan
dre
form
sla
gged
fou
rpe
riod
s)ar
esh
are
ofse
ats
inst
ate
asse
mbl
yoc
cupi
edby
Con
gres
s,h
ard
left
,sof
tle
ft,a
nd
Hin
dup
arti
esla
gged
eigh
tp
erio
dspl
us
lan
dre
form
vari
able
sla
gged
eigh
tpe
riod
s.IV
2:In
stru
men
tsfo
rth
een
dog
enou
spo
licy
vari
able
(cu
mu
lati
vela
nd
refo
rms
lagg
edfo
ur
peri
ods)
are
shar
eof
seat
sin
stat
eas
sem
bly
occu
pie
dby
Con
gres
s,h
ard
left
,sof
tle
ft,a
nd
Hin
dup
arti
esla
gged
eigh
tp
erio
dsp
lus
cum
ula
tive
lan
dre
form
spa
ssed
inge
ogra
phic
ally
con
tigu
ous
stat
esla
gged
eigh
tpe
riod
s.IV
3:In
stru
men
tsfo
rth
een
doge
nou
spo
licy
vari
able
(cu
mu
lati
vela
nd
refo
rms
lagg
edfo
ur
per
iods
)ar
esh
ares
ofse
ats
inst
ate
asse
mbl
yoc
cup
ied
byC
ongr
ess,
har
dle
ftan
dso
ftle
ftp
arti
esla
gged
eigh
tpe
riod
s(l
agge
dla
nd
refo
rmva
riab
les
are
excl
ud
edh
ere)
.Th
eov
erid
enti
�ca
tion
test
we
empl
oyis
due
toS
arga
n{1
958}
.Th
en
um
ber
ofob
serv
atio
ns
tim
esth
eR
2
from
the
regr
essi
onof
the
stat
e2
resi
dual
son
the
inst
rum
ents
isd
istr
ibu
ted
x2
(T1
1),w
her
eT
isth
en
um
ber
ofin
stru
men
ts.
QUARTERLY JOURNAL OF ECONOMICS414
are insigni�cant.17 Column (5) con�rms that land reform still hasa signi�cant impact in closing the gaps between rural and urbanpoverty. We also report tests of our overidentifying restrictions forthe instrumental variables regressions. The political and laggedland reform variables pass standard statistical tests of overidenti-�cation and therefore, at least on econometric grounds, wouldappear to be suitable instruments for land reforms.18 Thus, whenthe instrument set includes political variables and lagged landreforms, the picture is consistent with the patterns of resultsshown in Tables III and IV.
The remainder of Table VI experiments with alternatives tousing lagged land reforms as instruments. In columns (6)–(8) weuse cumulative land reforms passed in geographically neighbor-ing states (lagged eight periods) in our instrument set in place ofcumulative land reforms lagged eight periods. These neighboringland reforms could proxy for regional waves of support for landreform. Using these, together with the political variables asinstruments, yields robust results. Aggregate land reforms con-tinue to exert a negative and signi�cant impact on the ruralpoverty gap (column (6)) or rural head count (column (7)). Whenwe break out land reforms by type, we again �nd tenancy reformsand abolition of intermediaries exerting the strongest negativein�uences on the head-count index (column (8)). The overidenti�-cation tests are also passed. In columns (9) and (10) of Table VI, wedrop lagged land reforms completely from the instrument set. Wecontinue to obtain a negative impact of land reforms on the ruralpoverty gap (column (9)) or the rural headcount (column (10)) foraggregate land reforms.19
Taken together, Table VI �nds a pattern of results that isconsistent with those presented in Tables III and IV.20 It is alsoreassuring that the magnitude of coefficients remains roughly
17. With the exception of land ceilings in column (3).18. The test we employ is due to Sargan {1958} and tests whether the
instruments are correlated with residuals from the second-stage (poverty) regres-sion. See notes to Table VI.
19. We did not obtain signi�cant effects for disaggregated land reforms in thisspeci�cation.
20. We also experimented with a fourth instrumentation procedure where theendogenous policy variable (cumulative land reforms lagged four periods) isinstrumented using a ‘‘simulated’’ cumulative land index created by cumulatingvalues from a linear probability model which predicts whether a land reform takesplace in a given year based on political composition of the state parliament (laggedfour periods) and year and state effects. As with the three other procedures, wefound that instrumented aggregate land reform had a signi�cant negative impacton the rural poverty gap.
LAND, POVERTY, AND GROWTH: INDIA 415
constant across the different instrumentation procedures. On thewhole, the instrumented coefficients on land reform are largerthan the baseline results of Tables III and IV. Overall, theseresults are best thought of as a robustness check on our earlier�ndings rather than trying to present a carefully thought outstructural model.
IV. LAND REFORM AND AGRICULTURAL WAGES
It would be surprising if land reforms that affected povertydid not impact on other aspects of the rural economy. We nowconsider whether such reforms have an effect on agriculturalwages. The wage data are for the daily agricultural wages of malelaborers expressed in real terms.21 Agricultural wages are arobust indicator of the welfare of landless laborers who comprise asigni�cant fraction of the poor in rural India (see World Bank{1997}). If land reform pushes up agricultural wages, this repre-sents an additional mechanism through which these reforms canreduce rural poverty.
The results using the agricultural wage as a left-hand-sidevariable are in Table VII. Column (1) contains results for theaggregate land reform variable. This demonstrates a positive andsigni�cant impact of land reform on wages. In column (2) thiseffect is disaggregated across types of land reform and shows thatthis effect is primarily attributable to legislation to abolishintermediaries. These results con�rm the impact of land reformson the rural economy. They also illustrate an indirect routethrough which land reform may positively affect the welfare oflandless laborers even if they do not bene�t directly from thereforms. In Section VII we discuss why such effects might bepresent in theory.
V. LAND REFORM AND GROWTH
Even if land reform does help the poor, it could do so at a costto economic performance. We turn now, therefore, to exploringwhether land reform has a positive or negative effect on agricul-tural output per capita. In this case, we use the log of real stateincome per capita as the left-hand-side variable with the right-hand side augmented by lagged log real state income per capita to
21. See the Data Appendix for details.
QUARTERLY JOURNAL OF ECONOMICS416
model dynamics in a very simple way and to allow for convergenceover time.22 We therefore have a regression of the form,
(3) xst 5 l xst 2 1 1 a s 1 b t 1 g yst 1 c lst 2 4 1 e st.
This is basically the same form of regression that has becomepopular in the cross-country growth literature summarized inBarro {1997}, although our panel data allow us to use �xed effectsand year effects. We will also continue to allow for a state-speci�cAR(1) error speci�cation with heteroskedasticity.
Table VIII presents the main results for the regression ofstate income per capita on cumulative land reform. In column (1)we present results for a GLS model of total state income per capitaon land reform containing only state �xed effects and year effectsas controls. We �nd that the disaggregated land reform variables
22. Statewise estimates of total and agricultural state domestic product areavailable for the 1960–1992 period. See the Data Appendix. These state domesticproduct estimates are used as our proxy of state income.
TABLE VIILAND REFORM AND AGRICULTURAL WAGES
Model
Real agriculturalwages
Real agriculturalwages
(1) (2)
GLSAR(1)
GLSAR(1)
Four-year lagged cumulative land reformlegislation
0.081(2.71)
Four-year lagged cumulative tenancyreform legislation
0.049(0.88)
Four year lagged cumulative abolition ofintermediaries legislation
0.339(2.61)
Four-year lagged cumulative land ceilinglegislation
0.069(0.09)
Four-year lagged cumulative consolida-tion of land holdings legislation
0.018(0.13)
State effects YES YESYear effects YES YESNumber of observations 441 441
z-statistics are in parentheses. The wage data refer to the daily wage rate for male agricultural laborersexpressed in real terms. See the Data Appendix for details on construction and sources of the variables. Thedata are for fourteen states—data for Haryana and Jammu and Kashmir are not available. For thirteen ofthese states we have data 1961–1992 and for Rajasthan we have 1967–1991. This gives a sample size of 441.GLS AR(1) model allows a state-speci�c AR(1) process—see equation (1) in the text for details.
LAND, POVERTY, AND GROWTH: INDIA 417
TAB
LE
VII
IL
AN
DR
EF
OR
MA
ND
GR
OW
TH
ININ
DIA
Mod
el
Log
ofst
ate
inco
me
per
capi
ta
Log
ofag
ricu
ltu
rals
tate
inco
me
per
capi
ta
Log
ofag
ricu
ltu
rals
tate
inco
me
per
capi
taL
ogof
agri
cult
ura
lyie
ldL
ogof
agri
cult
ura
lyie
ld
(1)
(2)
(3)
(4)
(5)
GL
SA
R(1
)G
LS
AR
(1)
GL
SA
R(1
)G
LS
AR
(1)
GL
SA
R(1
)
On
e-ye
arla
gged
log
ofst
ate
inco
me
per
capi
ta0.
497
(12.
53)
On
e-ye
arla
gged
log
ofag
ricu
ltu
rals
tate
inco
me
per
capi
ta0.
195
(4.1
7)0.
167
(3.2
9)F
our-
year
lagg
edcu
mul
ativ
ete
nan
cyre
form
s2
0.00
2(0
.43)
20.
037
(4.5
4)2
0.03
3(2
.94)
20.
050
(6.5
5)2
0.03
8(3
.92)
Fou
r-ye
arla
gged
cum
ulat
ive
abol
itio
nof
inte
rmed
iari
es2
0.00
5(0
.54)
0.00
5(0
.27)
20.
016
(0.7
6)2
0.00
2(0
.12)
20.
013
(0.4
9)F
our-
year
lagg
edcu
mul
ativ
ela
nd
ceil
ing
legi
slat
ion
20.
002
(0.2
2)0.
019
(1.2
6)0.
012
(0.6
4)0.
015
(0.9
5)0.
015
(0.8
8)F
our-
year
lagg
edla
nd
cons
olid
atio
nle
gisl
atio
n2
0.01
3(1
.29)
0.06
5(3
.31)
0.05
7(2
.12)
0.07
4(3
.87)
0.05
4(2
.15)
Pop
ula
tion
grow
thra
te2
2.56
7(0
.75)
4.16
6(1
.11)
Fou
r-ye
arla
gged
per
capi
taed
uca
tion
expe
ndi
ture
s0.
003
(1.4
8)0.
003
(1.6
7)F
our-
year
lagg
edpe
rca
pita
hea
lth
expe
ndit
ures
20.
005
(1.9
7)2
0.00
2(0
.77)
Fou
r-ye
arla
gged
per
capi
taot
her
expe
ndi
ture
s2
0.00
04(0
.99)
20.
0002
(0.4
0)F
our-
year
lagg
edpe
rca
pita
tax
reve
nu
efr
omre
dist
ribu
tive
taxe
s2
0.00
4(1
.51)
20.
003
(1.0
5)F
our-
year
lagg
edst
ate
taxe
sas
ape
rcen
tage
ofst
ate
dom
esti
cpr
odu
ct0.
474
(0.5
4)0.
278
(0.3
1)F
our-
year
lagg
edlo
gof
agri
cult
ura
lyie
ld0.
010
(0.1
7)2
0.01
8(0
.32)
Sta
teef
fect
sY
ES
YE
SY
ES
YE
SY
ES
Year
effe
cts
YE
SY
ES
YE
SY
ES
YE
SN
um
ber
ofob
serv
atio
ns
484
484
433
488
433
z-st
atis
tics
are
inpa
ren
thes
es.S
tate
inco
me
per
cap
ita
isob
tain
edby
expr
essi
ng
esti
mat
esof
stat
edo
mes
tic
pro
du
ctin
real
per
capi
tate
rms.
Agr
icu
ltu
raly
ield
mea
sure
sre
pres
ent
the
rati
oof
real
agri
cult
ura
lsta
ted
omes
tic
pro
duct
ton
etso
wn
area
.Red
istr
ibu
tive
taxe
sar
eag
ricu
ltu
rali
nco
me
taxe
sla
nd
taxe
s,an
dp
rop
erty
taxe
s.S
eeth
eD
ata
Ap
pen
dix
for
deta
ils
onco
nst
ruct
ion
and
sou
rces
ofth
eva
riab
les.
Th
ed
ata
are
for
the
sixt
een
mai
nst
ates
.For
colu
mn
s(1
)an
d(2
)we
hav
eda
ta19
61–1
992
for
twel
vest
ates
,for
Pu
nja
ban
dH
arya
na,
wh
ich
spli
tin
1965
,we
hav
eda
ta19
66–1
992;
for
Jam
mu
and
Kas
hm
ir19
65–1
992;
and
for
Bih
ar19
61,1
966,
and
1969
–199
2.T
his
give
sa
sam
ple
size
of48
4.T
he
sam
ple
size
inco
lum
n(4
)is
slig
htl
yla
rger
asit
does
not
con
tain
ala
gged
depe
nde
nt
vari
able
asre
gres
sor.
For
colu
mn
s(3
)an
d(5
)w
eh
ave
dat
a19
64–1
992
for
nin
est
ates
,fo
rP
un
jab
1969
–199
2,fo
rH
arya
na
1970
–197
5an
d19
78–1
992,
for
Jam
mu
and
Kas
hm
ir19
68–1
992,
for
Bih
aran
dG
uja
rat
1964
–197
5an
d19
77–1
992,
for
Ass
am19
69an
d19
72–1
992
and
for
Tam
ilN
adu
1964
–197
5an
d19
78–1
992.
Th
isgi
ves
asa
mpl
esi
zeof
433.
GL
SA
R(1
)m
odel
allo
ws
ast
ate-
spec
i�c
AR
(1)p
roce
ss—
see
equ
atio
n(1
)in
the
text
for
deta
ils.
QUARTERLY JOURNAL OF ECONOMICS418
lagged four periods have no signi�cant impact on total stateincome per capita. In column (2) we look only at agricultural stateincome per capita. This makes sense given that land reform ispredominantly concerned with affecting production relations inagriculture. This suggests that tenancy reform has a negativeeffect on agricultural output with land consolidation having theopposite effect. No effect is observed for the other kinds of landreform. Column (3) shows that both the tenancy reform and landconsolidation effects are robust to including our other policyvariables lagged four years. In column (4) we show that theseeffects remain when agricultural yields rather than income percapita is the left-hand-side variable.23 In column (5) we show thatthis effect of tenancy reform is robust to including other policyvariables.
VI. LAND REFORM AND LAND INEQUALITY
Taken together, our results hint at an equity-efficiency trade-off for tenancy reforms—both poverty and output per capita arelower after such reforms are enacted. No such trade-off emergesfor abolition of intermediaries. Ceilings on landholdings do notseem to have an effect on either output measures or poverty, whileland consolidation promotes output increases in agriculture with-out affecting poverty. The failure of land ceiling legislation to showany signi�cant impact on poverty reduction or output levels isconsistent with Bardhan’s {1970} claim that such reforms haverarely been implemented with any degree of seriousness.
Overall, these results suggest that the impact on povertycomes mainly through reforms that affect production relations,rather than by altering the distribution of land. This interpreta-tion is underlined by looking at the limited evidence available onthe relationship between land reforms and land distribution overour data period. Data on land distribution has only been gatheredby NSS special surveys at four points; 1953–1954, 1961–1962,1971–1972, and 1982 (see Sharma {1994}). We classify states ashigh or low land reform depending on whether they had more orless than a total of three land reforms (of any type) during the1958–1992 period.24 We then investigate whether high land
23. Our measure of yield is real agricultural state domestic product dividedby net sown area. See the Data Appendix for details.
24. Under this system Andra Pradesh, Assam, Haryana, Jammu and Kash-mir, Madya Pradesh, Maharashtra, Punjab, and Rajasthan are low land reform
LAND, POVERTY, AND GROWTH: INDIA 419
reform states classi�ed in this way experienced the largest drop inthe gini for land operated and proportion of landless householdsover the period.25 The overall impression that we have from thiscrude exercise, is of persistent inequalities in land operatedwithin both groups of states (see also Sharma {1994}). Thus, theidea that the major impact of land reform on poverty must comemainly through mechanisms that did not involve land redistribu-tion gains further support. In further con�rmation of this, wefailed to �nd a signi�cant effect of aggregate land reform on thegini coefficient for rural per capita consumption.26
Thus, in making sense of the results, it is imperative to thinkabout land reforms that have changed production relations inagriculture rather than altering the pattern of landholdings. All ofthis notwithstanding, there is evidence that the impact of landreforms on poverty is greatest in those states where land inequal-ity was greatest in 1953–1954. To test this, we interacted thepercentage of landless individuals and the land gini coefficientwith our land reform variable. This interaction term was negativeand signi�cant in every case when we looked at aggregate landreform activity.27
VII. MAKING SENSE OF THE RESULTS
Our empirical analysis suggests that poverty reduction isassociated with land reform and this is primarily attributable tolegislation that has abolished intermediaries and reformed theterms of tenancies. The role of land redistribution per se seems tohave been of limited importance in the Indian context. Theempirical analysis also uncovers some evidence of general equilib-rium effects on wages. Our theoretical model focuses on twothings: a model of agricultural contracting and a model of labor
states, while Bihar, Gujarat, Karnataka, Kerala, Orissa, Tamil Nadu, UttarPradesh, and West Bengal are high land reform states.
25. For high land reform states the land gini falls from 0.686 in 1953/54 to0.669 in 1982 (a fall of 0.017), whereas the drop in low land reform states is from0.653 to 0.643 (a drop of 0.010). For high land reform states the average drop in theproportion landless is from 14.97 percent to 12.03 percent (a fall of 2.94 percent),whereas for low land reform states the drop is from 12.40 percent to 10.91 percent(a fall of 1.49 percent).
26. To look at this issue, we ran the basic regression shown in column (1) ofTable III but replaced the rural poverty gap with the gini coefficient for rural percapita consumption. Our inability to �nd a signi�cant impact on rural inequalitycould be explained by the fact that land reform may be shifting income from themiddle income groups to the poor rather than from the rich to the poor.
27. The results are available from the authors on request.
QUARTERLY JOURNAL OF ECONOMICS420
supply by tenants. The former focuses on how rents to tenantsshift in response to land reforms, and the latter gives rise to effectson agricultural wages. This focus on contractual problems cap-tures the spirit of recent models of the inequality-growth relation-ship that emphasize agency problems, particularly in creditmarkets. (See Benabou {1996} for a survey.) Here, we emphasizeagency problems in tenancy contracts and how they can be alteredby land reform, even if the ownership pattern is unchanged.
There are three groups: landlords who rent out land as well asfarming some of the land themselves, tenants who rent land, andlandless laborers. The poor are made up predominantly of thelatter two groups. Tenants and landless laborers supply labor tothe labor market where it is demanded by landlords who choose tobe owner-cultivators. Tenants and landless laborers care aboutconsumption c and labor supply l. Their preferences are u(c) 2f (l), where u(·) is increasing and concave and f (·) is increasingand convex. Suppose that the agricultural wage is v . Then, anindividual with nonlabor income x, has optimal labor supply of
l*(x,v ) 5 larg max u(x 1 v l) 2 f (l) .
It is straightforward to check that labor supply is decreasing in x.Now de�ne v(x,v ) 5 u(x 1 v l*(x,v )) 2 f (l*(x, v )) as the indirectutility of a tenant with nonlabor income x. Hence, we expectlandless laborers to supply l*(0, v ), while for tenants x is equal tothe value of their tenancy. As the value of tenancy increases as aresult of land reform, we would expect tenants to reduce laborsupply to the market.
We now consider the agricultural contracting problem of atenant and landlord. Suppose that the output on a given piece ofland under tenancy is given by R(e) where e denotes effort appliedto the land by the tenant. We suppose that the cost of this effort isseparable from labor supply and is measured in units of disutility.Effort is also committed before the labor supply decision is made.We assume that R(·) is smooth, increasing, and concave.
We suppose that tenants need to be monitored in order to putin effort on the land. Speci�cally, we imagine that a contractspeci�es an effort level of e. However, the tenant may choose to‘‘shirk,’’ putting in zero effort, in which case the landlord catcheshim with probability p and he is �red, becoming a landless laborerand receiving a payoff of v(0, v ). The tenant can now only beinduced to supply effort if the threat of eviction is sufficientlystrong and some rents are earned from being a tenant. Suppose,
LAND, POVERTY, AND GROWTH: INDIA 421
then, that the tenant receives a payment of w to farm the land,which he receives only if he is not caught shirking. Thus, a tenantis willing to put in an effort level e at payment w if and only theincentive constraint (1 2 p)v(w,v ) 1 pv(0, v ) # v(w, v ) 2 e issatis�ed. Solving this as an equality gives the payment schedulew(e, v ) needed to induce effort level e as
w(e, v ) 5 v 2 1(v(0, v ) 1 e/ p).
The contract must now specify a payment/effort pair consis-tent with this schedule. The optimal effort that the landlordchooses to induce is given by
e( p,v ) 5 arg maxe
5 R(e) 2 w(e,v ) .
It is easy to verify that e( p,v ) is increasing in p. The tenant’sequilibrium payoff is V( p, v ) 5 v(0, v ) 1 e( p, v ) (1 2 p)/ p which islarger than the payoff from being a landless laborer.
It is straightforward to calculate the impact of changes in p onoutput and the tenant’s payoff. An increase in p will increasenet-output since e( p, v ) is increasing. The effect on the tenant’spayoff (and hence poverty) is given by
V( p,v )
p5
e( p,v )
p
1 2 p
p2 e( p, v )
1
p2 .
The �rst term is positive—an increase in the eviction probabilityelicits higher effort and hence raises the tenant’s rent. The secondeffect works in the opposite direction. For a given effort level, thetenant’s rent is lower since he must be paid less now to preventhim from shirking. We are interested in cases where the tenantenjoys a more secure right to the land so that p falls. In this case,the tenant will bene�t from a tenancy reform that reduces theprobability that he will be evicted if caught shirking if theelasticity of effort with respect to the probability of eviction( e( p, v )/ p · p/e( p,v )) is less than 1/{(1 2 p) p2}. If tenants’ rentsincrease from receiving higher tenure security, then this will leadthem to reduce their labor supply to the market, and we wouldpredict that such a tightening of the labor market would lead toincreased agricultural wages.28
This framework can be applied to the cases of abolition ofintermediaries and tenancy reform. To include an intermediary in
28. These changes in wages would also be expected to reinforce reductions inoutput on farms that hire in labor.
QUARTERLY JOURNAL OF ECONOMICS422
the analysis, we suppose that there are three parties to theagricultural contract: a tenant, landlord, and an intermediary. Webegin by making the strong assumption that intermediaries havea very strong bargaining position can make take-it-or-leave-itoffers to the landlord and tenant. This is very much in line withthe view that intermediaries captured the surplus from the land.In this world the tenant will receive a payoff of V( p, v ), and thelandlord will receive his reservation payoff which we denote by vL.The intermediary receives the surplus {R(e( p,v )) 2 e( p, v )} 2V( p,v ) 2 vL.
After the intermediary is abolished, this surplus is nowavailable for distribution provided that p remains the same. Onlyif the tenant obtains no bargaining power at all with his landlord,would we expect to observe no effect on the tenant’s payoff.Otherwise, we would expect to see the tenant’s payoff rise.Assuming that tenants are a signi�cant group of poor in India,this is consistent with our �nding that poverty is reduced by theabolition of intermediaries. We would not expect to see any changein effort and hence output unless p were different when landlordsand intermediaries negotiated contracts. Rent increases for ten-ants also would be associated with higher agricultural wages, bythe general equilibrium mechanism we have identi�ed.
We now turn to the impact of tenancy reforms. Such reformsare multifarious which make it difficult to offer a de�nitivetheoretical account. This would require much more institutionalcontent as in the analysis of West Bengal’s land reforms byBanerjee and Ghatak {1997}. Nonetheless, it is still useful to thinkthrough a simple model in order to check that our empirical�ndings conform to the predictions of the theory laid out above.Suppose therefore that the landlord has all the bargaining powerand can make take-it-or-leave-it offers to tenants before and afterthe tenancy reform. We shall model the effect of a tenancy reformas making it more difficult to evict tenants if they shirk. In termsof our model this is equivalent to a fall in p. As we have alreadyargued, this has two effects. First, we expect effort, and therefore,output to fall. Second, we expect a change in the payoff to thetenant as his rent could go up or down. We showed that this ispositive under reasonable conditions, and thus we would expectpoverty to be reduced which is what we found in the data. This isalso consistent with higher agricultural wages if increased rentsto tenants lead them to reduce their labor supply.
To summarize, the empirical �ndings are consistent with a
LAND, POVERTY, AND GROWTH: INDIA 423
stylized model of agricultural contracting and labor supply bytenants. While many complicating features could be added to thetheory, the general thrust of the trade-off captured here isrelevant.29 It is well-known that in a variety of contexts, rents areused to motivate tenants. Thus, land reforms that affect howagency problems are solved will typically generate both outputand distributional effects. We would expect these rents to affectlabor supply and result in changes to agricultural wages.
VIII. CONCLUDING REMARKS
The main contribution of this paper is to test whether landreform legislation is associated with poverty reductions usingstate-level data from India. The high incidence of poverty and thelarge volume of land reforms enacted to counter this problem inthe post-Independence period make this an issue of considerableinterest. We show that there is robust evidence of a link betweenpoverty reduction and two kinds of land reform legislation—tenancy reform and abolition of intermediaries. Another impor-tant �nding is that land reform can bene�t the landless by raisingagricultural wages. Although the effects on poverty are likely tohave been greater if large-scale redistribution of land had beenachieved, our results are nonetheless interesting as they suggestthat partial, second-best reforms which mainly affect productionrelations in agriculture can play a signi�cant role in reducingrural poverty.
As well as being important to policy debates in India, such�ndings may help to diffuse the more general pessimism that canundermine redistributive effort in developing countries. In arecent study {World Bank 1997}, much emphasis was placed onthe role of economic growth in explaining the decline of poverty inIndia. While our results are consistent with this �nding, theyemphasize that redistributive effort has also played its part.Using the average number of land reforms implemented, our �rstcoefficient in Table III implies that a reduction of the all-Indiapoverty gap of 1 percent can be explained by land reform. This isone-tenth of the actual reduction in poverty over the period of ourdata. This remains true even after factoring in the possibility thatoutput per capita is reduced by some kinds of land reform (Table
29. Following Banerjee and Ghatak {1997}, it would be possible to introduceinvestment into the model. In general, we would expect increased tenurial securityto increase investment.
QUARTERLY JOURNAL OF ECONOMICS424
VIII). To put this in perspective, we compared the effect of landreforms on poverty with the effect of changes in per capita income.This comparison suggests that implementing a land reform has asimilar effect on poverty reduction to a 10 percent increase in percapita income, or around four to �ve years growth at the all-Indiaaverage growth rate over this period.30
Since the effects of redistributive intervention on poverty andgrowth are not known a priori, a signi�cant literature has testedthese links using cross-country data. Benabou {1996} reviews thisliterature and emphasizes the diverse �ndings. While adding toour general understanding, the difficulties of �nding reliablecross-country measures of redistribution is a signi�cant drawbackin this research agenda. There seems little doubt, therefore, thatexploiting policy variation due to the federal structure of somedeveloping countries may be an important additional source ofevidence on policy incidence. It will also help to get behind broadbrush policy categories such as education or health expendituresthat mask important policy variations. Our study underlines that,even within a particular area of government intervention (i.e.,land reform), the empirical effects may vary depending on theexact form that the intervention takes. This is true, moreover,even though our policy measures are themselves fairly broad.Future efforts to quantify the empirical relationship betweengrowth, poverty, and redistribution will doubtless bene�t evenmore from a detailed speci�cation of how particular policy inter-ventions are structured and implemented across space and time.
DATA APPENDIX
The data used in the paper come from a wide variety ofsources.31 They come from the sixteen main states listed in TableI. Haryana split from the state of Punjab in 1965. From this dateon, we include separate observations for Punjab and Haryana.
30. Thus, we regressed poverty on per capita income (along with state effectsand year effects) and compared the coefficient on per capita income with thatobtained on land reform. (It made essentially no difference whether we did this byincluding both land reform and per capita income in one regression, or ranseparate regressions in one case including only land reform and in the other onlyincome per capita.) Results are available from the authors on request.
31. Our analysis has been aided by Ozler, Datt, and Ravallion {1996} whichcollects published data on poverty, output, wages, prices indices, and populationto construct a consistent panel data set on Indian states for the period 1958 to1992. We are grateful to Martin Ravallion for providing us with these data, towhich we have added information on land reform, public �nance, and politicalrepresentation.
LAND, POVERTY, AND GROWTH: INDIA 425
Land Reform
To construct the land reform variable used in the regressions,we begin by recording all land reform acts that were passed in agiven state and year. By examining the content of each landreform, we then classify each land reform act into four categories(1 5 tenancy reform, 2 5 abolition of intermediaries, 3 5 ceilingson landholdings, 4 5 consolidation of landholdings) where a singleland reform can belong to several types (see Table II). For eachland reform type this gives us a variable that is 0 or 1 in givenstate s and year t. We cumulate these variables over time to giveus four cumulative land reform variables that capture the stock ofland reforms passed to date in each of the four categories. We alsoaggregate across all four land reform categories to give us anaggregate cumulative land reform variable that gives us a mea-sure of the total stock of land reforms passed in state s by year t.Amendments to acts are treated as new pieces of legislation. TheIndex to Central and State Enactments (Ministry of Law andJustice, Government of India) was used to identify Acts pertainingto land reform in different states. To examine the exact content ofthese acts, we mainly used Haque and Sirohi {1986} and Zaidi{1985}, although a range of secondary sources were used todouble-check the correctness of the information provided by thesetwo books and to �ll in and update the detail regarding speci�clegislations. The secondary sources included Appu {1996}, Behu-ria {1997}, Bonner {1987}, Borgohain {1992}, Kurien {1981},Mearns {1998}, Pani {1983}, Singh and Misra {1964}, and Yugand-har and Iyer {1993}.
Poverty Data
We use the poverty measures for the rural and urban areas ofIndia’s sixteen major states, spanning 1957–1958 to 1991–1992put together by Ozler, Datt, and Ravallion {1996}. These measuresare based on 22 rounds of the National Sample Survey (NSS)which span this period. Not all 22 rounds of the survey can becovered for each of the 22 rounds for each of the 16 states.32 TheNSS rounds are also not evenly spaced: the average interval
32. For 11 states (Andra Pradesh, Assam, Bihar, Karnataka, Kerala, MadhyaPradesh, Orissa, Rajasthan, Tamil Nadu, Uttar Pradesh, and West Bengal), all 22rounds have been covered. Because Haryana only appears as a separate state fromPunjab in 1965, we have adopted including separate series for these two statesfrom this date onward. For Gujarat and Maharashtra, twenty rounds are included,beginning with the sixteenth round in 1958–1959 (before 1958–1959, separatedistributions are not available for these two states as they were merged under the
QUARTERLY JOURNAL OF ECONOMICS426
between the midpoints of the surveys ranges from 0.9 to 5.5 years.Surveys were carried out in the following years: 1958, 1959, 1960,1961, 1962, 1963, 1965, 1966, 1967, 1968, 1969, 1970, 1971, 1973,1974, 1978, 1983, 1987, 1988, 1990, 1991, and 1992. Because otherdata are typically available on a yearly basis, weighted interpola-tion has been used to generate poverty measures for years wherethere was no NSS survey. The poverty lines used are thoserecommended by the Planning Commission {1993} and are asfollows. The rural poverty line is given by a per capita monthlyexpenditure of Rs. 49 at October 1973–June 1974 all-India ruralprices. The urban poverty line is given by a per capita monthlyexpenditure of Rs. 57 at October 1973–June 1974 all-India urbanprices. See Datt {1995} for more details on the rural and urbancost of living indices and on the estimation of the povertymeasures. The headcount index and poverty gap measures areestimated from the grouped distributions of per capita expendi-ture published by the NSS,33 using parameterized Lorenz curvesusing a methodology detailed in Datt and Ravallion {1992}.
Agricultural Wages
The primary source for the data is Agricultural Wages inIndia (Ministry of Agriculture, Government of India). Nominalwage data from this series have been de�ated using the ConsumerPrice Index for Agricultural Laborers to obtain real agriculturalwages. No agricultural wage data are available for the state ofJammu and Kashmir, and no separate wage data are available forthe state of Haryana.
Income Data
The primary source for data on state income is an annualgovernment publication Estimates of State Domestic Product(Department of Statistics, Department of Statistics, Ministry ofPlanning). The primary source for the Consumer Price Index forAgricultural Laborers (CPIAL) and Consumer Price Index forIndustrial Workers (CPIIW) which are used to de�ate agriculturaland nonagricultural state domestic product respectively is anumber of Government of India publications which include In-
state of Bombay). For Jammu and Kashmir, only eighteen rounds can be included,beginning with the sixteenth round for 1960–1961, due to a lack of data.
33. Reports from the National Sample Survey Organisation, Department ofStatistics, Ministry of Planning, Government of India and Sarvekshena, Journal ofthe National Sample Survey Organisation, Department of Statistics, Ministry ofPlanning, Government of India.
LAND, POVERTY, AND GROWTH: INDIA 427
dian Labour Handbook, the Indian Labour Journal, the IndianLabour Gazette, and the Reserve Bank of India Report on Currencyand Finance. Ozler, Datt and Ravallion {1996} have furthercorrected CPIAL and CPIIW to take account of interstate cost-of-living differentials and have also adjusted CPIAL to take accountof rising �rewood prices. Using their data allows us to puttogether a consistent and complete series on real total, agricul-tural, and nonagricultural state income for the period 1960 to1992. Our measure of agricultural yield is obtained by dividingreal agricultural state domestic product by net sown area for allcrops which is obtained from a government publication Area andProduction of Principal Crops in India (Directorate of Economicsand Statistics, Ministry of Agriculture).
Public Finance Data
The primary source for state level information on taxes andexpenditures is an annual publication, Public Finance Statistics(Ministry of Finance, Government of India). This information isalso collated in the Reserve Bank of India’s annual publicationReport on Currency and Finance.
Population Data
The population estimates are constructed using Census datafrom the �ve censuses for 1951, 1961, 1971, 1981, and 1991(Census of India, Registrar General and Census Commissioner,Government of India). Between any two successive censuses, thestate-sectoral populations are assumed to grow at a constant(compound) rate of growth, derived from the respective populationtotals.
Political Variables
Political variables are the main instruments used in thepaper. Data on the number of seats won by different nationalparties at each of the state elections are from Butler, Lahiri, andRoy {1991}. These primary data are aggregated into four politicalgroupings which are de�ned in the text and expressed as shares ofthe total number of seats in state legislatures. State politicalcon�gurations are held constant between elections.
DEPARTMENT OF ECONOMICS,LONDON SCHOOL OF ECONOMICS
QUARTERLY JOURNAL OF ECONOMICS428
REFERENCES
Alesina, Alberto, and Dani Rodrik, ‘‘Distributive Politics and Economic Growth,’’Quarterly Journal of Economics, CIX (1994), 465–490.
Appu, P. S., Land Reforms in India: A Survey of Policy Legislation and Implemen-tation (New Delhi: Vikas, 1996).
Bandyopadhyay, D., ‘‘Land Reforms in India: An Analysis,’’ Economic and PoliticalWeekly, June 21–28, 1986.
Banerjee, Abhijit, and Maitreesh Ghatak, ‘‘Empowerment and Efficiency: TheEconomics of Tenancy Reform,’’ mimeo, Massachusetts Institute of Technol-ogy, and University of Chicago, 1996.
Banerjee, Abhijit, and Andrew Newman, ‘‘Occupational Choice and the Process ofDevelopment,’’ Journal of Political Economy, CI (1993), 659–684.
Bardhan, Pranab, ‘‘India,’’ in Chenery et al., eds., Redistribution with Growth(Oxford: Oxford University Press, 1970).
Barro, Robert, Determinants of Economic Growth: A Cross-Country EmpiricalStudy (Cambridge: MIT Press, 1997).
Behuria, N. C., Land Reforms Legislation in India: A Comparative Study (NewDelhi: Vikas, 1997).
Benabou, Roland, ‘‘Inequality and Growth,’’ NBER Macroeconomics Annual(Cambridge: MIT Press, 1996).
Binswanger, Hans, Klaus Deininger, and Gershon Feder, ‘‘Power, Distortions,Revolt and Reform inAgricultural Land Relations, in Jere Behrman, and T. N.Srinivasan, eds., Handbook of Development Economics (Amsterdam: Elsevier,1995).
Bonner, Jeffrey P., Land Consolidation and Economic Development in India—AStudy of Two Hariana Villages (Riverdale: The Riverdale Company, 1987).
Borgohain, Rooplekha, Politics of Land Reforms in Assam (New Delhi: ReliancePublishing House, 1992).
Butler, David, Ashok Lahiri, and Prannoy Roy, India Decides: Elections 1952–1991(New Delhi: Aroom Purie for Living Media India, 1991).
Chattopadhyay, S. N., ‘‘Historical Context of Political Change in West Bengal: AStudy of Seven Villages in Bardhaman, Economic and Political Weekly, March28, 1992.
Chenery, Hollis, Montek Ahluwalia, Clive Bell, John Duloy, and Richard Jolly,Redistribution with Growth (Oxford: Oxford University Press, 1970).
Datt, Gaurav, ‘‘Poverty in India 1951–1992: Trends and Decompositions,’’ PolicyResearch Department, World Bank, 1995.
Datt, Gaurav, and Martin Ravallion, ‘‘Growth and Redistribution Components ofChanges in Poverty Measures: A Decomposition with Applications to Braziland India in the 1980s,’’ Journal of Development Economics, XXXVIII (1992),275–295.
Datt, Gaurav, and Martin Ravallion, ‘‘Why Have Some Indian States Done Betterthan Others at Reducing Poverty?’ ’ Economica, LXV (1998), 17–38.
Dreze, Jean, andAmartya Sen, India: Economic Development and Social Opportu-nity (Oxford, Clarendon Press, 1995).
Dreze, Jean, Peter Lanjouw, and Naresh Sharma, ‘‘Economic Development 1957–1993,’’ in Peter Lanjouw and Nicholas Stern, Economic Development inPalanpur over Five Decades (Oxford: Oxford University Press, 1998).
Gough, Kathleen, Rural Change in Southeast India: 1950s to 1980s (Delhi: OxfordUniversity Press, 1989).
Haque, Tajamul, and Amar Singh Sirohi, Agrarian Reforms and InstitutionalChanges in India (New Delhi: Concept Publishing Company, 1986).
Hoff, Karla, and Andrew B. Lyon, ‘‘Non-Leaky Buckets: Optimal RedistributiveTaxation and Agency Costs,’’ Journal of Public Economics, LIII (1995),365–390.
Jayaraman, Raji, and Peter Lanjouw, ‘‘Living Standards in Rural India: APerspective from Longitudinal Village Studies,’’ mimeo, Cornell Universityand World Bank, 1997.
Kurien, Christopher Thomas, Dynamics of Rural Transformation—A Study ofTamil Nadu: 1950–1975 (Madras: Orient Longman, 1981).
Mearns, Robin, ‘‘Access to Land in Rural India: Policy Issues and Options,’’ mimeo,World Bank, 1998.
LAND, POVERTY, AND GROWTH: INDIA 429
Moene, Karl Ove, ‘‘Poverty and Land Ownership,’’ American Economic Review,LXXXII (1992), 52–64.
Ozler, Berk, Gaurav Datt, and Martin Ravallion, ‘‘A Data Base on Poverty andGrowth in India,’’ mimeo, World Bank, 1996.
Pani, Narendar, Reforms to Pre-empt Change—Land Legislation in Karnataka(New Delhi: Concept Publishing Company, 1983).
Perotti, Roberto, ‘‘Income Distribution, Democracy, and Growth: What Does theData Say,’’ Journal of Economic Growth, I (1996), 149–187.
Persson, Torsten, and Guido Tabellini, ‘Is Inequality Harmful for Growth: Theoryand Evidence,’’ American Economic Review, LXXXIV (1994), 600–621.
Planning Commission, Report on the Expert Group on the Estimation of theProportion and Number of Poor (New Delhi: Government of India, 1993).
Radhakrishnan, P., ‘‘Land Reforms: Rhetoric and Reality,’’ Economic and PoliticalWeekly, November 24, 1990.
Rodrik, Dani, ‘‘Getting Interventions Right: How South Korea and Taiwan GrewRich,’’ Economic Policy, XX (1995), 53–97.
Sargan, John Denis, ‘‘The Estimation of Economic Relationships Using Instrumen-tal Variables,’’ Econometrica, XXVI (1958), 393–415.
Sharma, H. R. ‘‘Distribution of Landholdings in Rural India, 1953–54 to 1981–1982: Implications for Land Reforms,’’ Economic and Political Weekly, Septem-ber 24, 1994.
Singh, Baljit, and Shridhar Misra, A Study of Land Reforms in Uttar Pradesh(Honolulu: East-West Center Press, 1964).
Thorner, Daniel, Agrarian Prospect in India (New Delhi: Allied Publishers, 1976).Wadley, S. S., and B. W. Derr, ‘‘Karimpur 1925–1984: Understanding Rural India
through Restudies,’’ in Pranab Bardhan, ed., Conversations between Econo-mists and Anthropologists (Delhi: Oxford University Press, 1990).
Warriner, Doreen, Land Reform in Principle and Practice (Oxford: Oxford Univer-sity Press, 1969).
World Bank, ‘‘Poverty in India: 50 Years after Independence,’’ World Bank, mimeo,1997.
Yugandhar, B. N., and K. Gopal Iyer, Land Reforms in India: Bihar—InstitutionalConstraints (London: Sage Publications, 1993).
Zaidi, A. Moin, Not by a Class War—A Study of Congress Policy on Land Reformsduring the Last 100 Years (New Delhi: Document Press, 1985).
QUARTERLY JOURNAL OF ECONOMICS430