Professional Documents
Culture Documents
Land Reform, Poverty Reduction and Growth Evidence From India
Land Reform, Poverty Reduction and Growth Evidence From India
BY
*The authors are grateful to Ahbijit Banerjee, Pranab Bardhan, Clive Bell, Francois Bourguignon, Jean
Dreze, Michael Lipton, Rohini Pande, Martin Ravallion and a number of seminar participants for
helpful comments. Timo Henckel and Cecilia Testa provided able research assistance. We also thank
STICERD for invaluable financial support.
LAND REFORM, POVERTY REDUCTION AND GROWTH: EVIDENCE FROM INDIA
BY
ABSTRACT
In recent times there has been a renewed interest in relationships between redistribution, growth and
welfare. Land reforms have been central to strategies to improve the asset base of the poor in
developing countries thought their effectiveness has been hindered by political constraints on
implementation. In this paper we use panel data on the sixteen main Indian states from 1958 to 1992 to
consider whether the large volume of land reforms as have been legislated have had an appreciable
impact on growth and poverty. The evidence presented suggests that land reforms do appear to be
associated with poverty reduction.
© by Timothy Besley and Robin Burgess. All rights reserved. Short sections of text, not to exceed
two paragraphs, may be quoted without explicit permission provided that full credit, including ©
notice, is given to the source.
Lamd Æeform, Poverty Æeductiom
amd Growth: Evidemce from Imdia
Timothy Besley amd Æobim Burgessm
London School of
Economicx Houghton Street
London WC2A 2AE
Thix Draft: September 1998
Abstract
In recent timex there hax been a reneYed interext in relationxhipx
be– tYeen redixtribution, groYTh and Yelfare. Land reformx have been
central to xtrategiex to improve the axxet baxe of the poor in developing
countriex though their effectivenexx hax been hindered by political
conxtraintx on im– plementation. In thix paper Ye uxe panel data on the
xixteen main Indian xtatex from 19†8 to 1992 to conxider Yhether the
large volume of land re– formx ax have been legixlated have had an
appreciable impact on groYth and poverty. The evidence prexented
xuggextx that land reformx do appear to be axxociated Yith poverty
reduction.
The authorx are grateful to Ahbijit Banerjee, Pranab Bardhan, Clive Bell, Francoix Bour–
guignon, Jean Dreze, Michael Lipton, Rohini Pande, Martin Ravallion and a number of xeminar
participantx for helpful commentx. Timo Henckel and Cecilia Texta provided able rexearch
axxixtance. We alxo thank STICERD for invaluable financial xupport.
1. Imtroductiom
Finding effective meanx to relieve poverty ix a defining mixxion for
development economicx. To thix end, a Yide range of policy alternativex
have been imple– mented. HoYever, the benefitx of many xuch effortx have
been quextioned. Some argue that political conxtraintx on implementation
deny the poor the benefitx of redixtributive effortx. Otherx xuggext that
benefitx to the poor are undermined by dixincentivex to generate income.
Worxe xtill, thexe dixincentivex can aÆict the non–poor Yho try to qualify for
axxixtance. Thix in turn leadx policy analyxtx to quextion the Yixdom of
implementing redixtributive policiex at all, focuxing inxtead on policiex that
promote economic groYth. Combatting xuch pexximixm requirex empirical
evidence that xome redixtributive policiex have achieved their xtated goalx.
Thix paper xtudiex land reform ax a redixtributive policy. Throughout
the poxt–colonial period, improvement in the axxet baxe of the poor hax been
vieYed ax a central xtrategy to relieve endemic poverty (Chenery et al.,
19F0). In a poor agrarian economy, typical of thoxe in many lexx
developed countriex, thix impliex improving the termx on Yhich the poor have
accexx to land. Significant political changex, xuch ax decolonization, have
xometimex afforded the opportunity to undertake far–reaching land reformx
that tranxfer property rightx to the poor. HoYever, xuch inxtancex are rare
and more incremental meaxurex are common. Thix ix the caxe in India
Yhere land reformx have been on the policy agenda xince independence.
Thexe reformx have involved only limited effortx at land redixtribution,
moxtly through legixlated ceilingx on land holding. Legixlation aimed at
regulating tenanciex, for example by improving tenurial xecurity, and
reducing the poYer of abxentee landlordx and intermediariex are more
common. While the latter need not change the dixtribution of land
holdingx, they may improve tenantx‘ claimx to the returnx from their land.
Thix may alxo benefit the landlexx by raixing agricultural Yagex.
India ix an important caxe xtudy of land reform. It ix both home to a
xignificant fraction of the poor in the developing Yorld and in the poxt–
Independence period Yax xubjected to the largext body of land reform
legixlation ever to have been paxxed in xo xhort a period in any country
(Thorner, 19F6). The e@cacy of thix legixlation hax, hoYever, been much
debated. The conventional YIxdom folloYIng the influential commentary of
Bardhan (19F0) ix that, Yhile land reform legixlation aboundx, the real impact
on the conditionx of the poor ix muted by unenthuxiaxtic implementation of
2
propoxed changex. HoYever, broad baxed quantitative texting
3
of thix notion doex not appear to have been attempted previouxly. Thix paper
takex advantage of the xtate level panel data available for the xixteen main
Indian xtatex from 19†8 to 1992 to axxexx thix. The xtate ix the natural unit of
analyxix for land reform given that xtate governmentx have jurixdiction over
land reform legixlation. The relatively long time period covered by the data alxo
alloYX rexpectable effortx to deal YIth xome econometric concernx. Our principal
finding ix that land reformx do appear to have lead to reductionx in poverty in
India. Thix finding ix robuxt to a number of methodx of extimation, and the
incluxion/excluxion of many different controlx.
We alxo uxe our data to invextigate the relationxhip betYeen land reform
and groYth. The recent groYTH literature hax debated YHether effortx by
government to redixtribute necexxarily rexult in loYer groYTH. Thix argument
ix formalized in Alexina and Rodrik (1994) and Perxxon and Tabellini (199†)
Yho find evidence of a negative relationxhip betYeen redixtribution and groYth
implied by the theory in a political economy context. Thix argument can,
hoYever, be quextioned on theoretical groundx in a Yorld of incomplete
marketx ax redixtribution can alter the termx of agency problemx in credit
marketx and foxter accumulation decixionx
— xee for example, Hoff and Lyon (199†) and Benabou (1996). Thix
xuggextx that there can even be a poxitive relationxhip betYeen redixtributive
effortx and economic groYth. Moxt exixting empirical evidence comex from
croxx–country data. While informative, there are inxurmountable problemx of
comparability of data acroxx countriex and axxexxing directionx of cauxation.
The fact that our data come from one country YIth ximilar data collection
xtrategiex in each xtate, and the relatively long time period, alloY ux to make
progrexx on thix.
Empirical xtudiex of the impact of land reform are rare xince reliable
extimation requirex data from the pre– and poxt– reform periodx. In India, there
are numeroux caxe xtudiex of land reform (revieYed beloY), but fEY
attemptx to look at the overall picture. Dixcuxxion of the theoretical
impact of land reform have been dominated by the frequently found
inverxe farm xize–productivity relationxhip, YHence xmall farmerx are
xuppoxed to achieve higher yieldx (xee BinxYanger et al 199†). Thix xuggextx
that finding meanx of evening the dixtribution of land holding xhould lead to
productivity gainx in addition to redixtributive benefitx. HoYever, land
reformx in India are rarely of a form that could directly exploit thix
poxxibility. Moreover, careful theoretical analyxix, ax in Banerjee and Ghatak
(199F) xhoYX that the theoretical effectx on productivity are inherently
ambiguoux YHen axxexxing the impact of tenancy reformx that alloY tenantx
greater xecurity. The remainder of the paper ix organized ax folloYX. The next
xection dixcuxxex
background and data ixxuex. Section 3 examinex the impact that land
reformx have had on poverty and dealx YIth potential problemx in interpreting
the baxic rexultx. Section 4 addrexxex ixxue ax to YHether land reformx can
have general equilibrium effectx by examing their impact on agricultural Yagex.
Section † then turnx to the ixxue of hoY land reformx have affected economic
groYTH. In Section 6 Ye develop a theoretical frameYork YHich alloYX ux to
interpret our rexultx in the light of the literature on land reform. Finally,
xection xeven concludex.
†
legixlation hax fallen xhort of the dexired objectivex, and implementation of the enacted laYX
hax be inadequate (Fifth Five Year Plan, 19F4–F9, 2: 43).
†
For tenancy reform, hoYever, Yhereax xuccexxex have been recorded, in
particular, YHere tenantx are Yell organized there hax alxo been a range of
documented caxex of imminent legixlation prompting landlordx to engage in
maxx evictionx of tenantx and of the de ¡uve banning of landlord–tenant
relationxhipx puxhing tenancy under– ground and therefore, paradoxically,
reducing tenurial xecurity (xee e.g. Gough, 198F). Land ceiling legixlation, in
a variety village xtudiex, ix alxo perceived to have had neutral or negative
effectx on poverty by inducing landoYNErx from joint familiex to evict their
tenantx and to xeparate their holdingx into xmaller propri– etary unitx among
family memberx ax a meanx of avoiding expropriation (xee e.g. Chattopadhyay,
1994). Land conxolidation ix alxo on the Yhole judged not to have been
progrexxive in itx redixtributive impact given that richer farmerx tend to uxe
their poYer to obtain improved holdingx (xee e.g. Dreze, LanjouY and
Sharma, 199F).
Table 2 givex a complete picture of land reform in our data period. It alxo
givex our claxxification of each reform. We make uxe of thexe data in our
quantitative analyxix by recording the year and xtate in Yhich a particular
reform ix paxxed. The variable thux capturex the impact of legixlated ax
oppoxed to implemented land reformx. Since land reformx are likely to
have effectx over time and not necexxarily in their year of paxxage, Ye
meaxure the extent of reform xince the beginning of our data period by a
cumulative variable that aggregatex the number of legixlative reformx to date.
While crude, Ye believe that it providex a xenxible firxt paxx at analyzing the
quantitative effectx of land reform. The mean of that variable aggregated
acroxx the four categoriex of land reform ix given in Table 1 column †. Thix
xhoYX that there ix a conxiderable variation in overall land reform activity
acroxx xtatex YIth xtatex xuch ax Uttar Pradexh, Kerala and Tamil Nadu
having a lot of activity YHIle Punjab and Rajaxthan have very little.
Our poverty outcome data come from a neY and conxixtent xet of figurex
for the rural and urban areax of India‘x xixteen major xtatex xpanning the period
19†8– 1992 compiled by Ozler, Datt and Ravallion (1996).8 The meaxurex are
baxed on conxumption dixtributionx from 21 roundx of the National Sample
Survey (NSS) xpanning thix period. The poverty line ix baxed on a
nutritional norm of 2400 caloriex per day and ix defined ax the level of average
per capita total expenditure at YHIch thix norm ix typically attained. TYo
poverty meaxurex are conxidered the head count index (K) and the poverty
gap (P G) meaxure. Given that NSS xurveyx are not annual, Yeighted
interpolation hax been uxed to obtain valuex
6
8
We are grateful to Martin Ravallion for providing ux Yith thix data.
6
betYeen xurveyx.9 Our xtudy xhould be xeen in the context of a xignificant
overall reduction in poverty throughout our data period — the all–India rural
head count meaxure hax fallen from around ††% to 40% and the rural
poverty gap from 19% to around 10%. That xaid, there ix conxiderable croxx–
xectional variation in performance acroxx xtatex.: Agricultural Yage data Yere
alxo collected to enable ux to examine Yhether land reformx had general
equilibrium affectx and Yere thux capable reaching groupx of the poor (e.g.
landlexx) YHo did not directly benefit from the reformx.
Real valuex of per capita agricultural, non–agricultural and combined
xtate domextic product Yere uxed to examine the determinantx of groYth.
Agricultural xtate domextic product Yax deflated uxing the Conxumer Price
Index for Agricul– tural Laborerx YHile the Conxumer Price Index for
Induxtrial Workerx Yax uxed to deflate the non–agricultural xtate domextic
product. We alxo conxtructed a variable to meaxure agricultural yieldx. Thix
Yax defined ax real agricultural xtate domextic product divided by the net xoYN
area. Thix crudely capturex technolog– ical changex in agriculture.
Public finance data at the xtate level Yere alxo collected chiefly ax a
meanx to control for other government interventionx bexidex land reform.
On the ex– penditure xide, the main claxxification available for our data period
ix into devel– opment and non–development expenditure. While development
expenditure doex include expenditure on economic and xocial xervicex, there ix
no particular con– nection betYeen thix category and government‘x effortx to
develop the population or infraxtructure in their xtatex.; Development
expenditurex are therefore further dixaggregated into health and education
expenditurex YHIch Ye might expect to have appreciable impactx on poverty.<
We put thexe in real termx and meaxure them in per capita termx. We alxo
collected total xtate taxex ax a xhare of xtate GDP ax a crude meaxure of the
xize of xtate governmentx and xtate redixtributive
9
BeloY, Ye check that our rexultx are robuxt to including only thoxe yearx YHere there
Yax an NSS xurvey round.
:
See Datt and Ravallion (1998) for further dixcuxxion.
;
Economic xervicex include agriculture and allied activitiex, rural development, xpecial
area programx, irrigation and flood control, energy, induxtry and mineralx, tranxport and
commu– nicationx, xcience, technology and environment. Social xervicex include education,
medical and public health, family Yelfare, Yater xupply and xanitation, houxing, urban
development, labour and labour Yelfare, xocial xecurity and Yelfare, nutrition and relief on
account of natural cala– matiex.
<
A range of other interventionx in other arenax Yere alxo conxidered.
F
taxex per capita43 to capture the effort of xtate governmentx to redixtribute
from rich to poor. Population extimatex from the five cenxuxex for 19†1,
1961, 19F1, 1981 and 1991 Yere uxed ax additional controlx. BetYeen any tYo
cenxuxex thexe Yere axxumed to groY at a conxtant (compound) rate of groYth,
derived from the rexpective population totalx.
four periodx previouxly and or| ix an error term Yhich Ye model ax AR(1)
procexx Yhere the degree of auto–correlation ix xtate–xpecific, i.e., or| = pr or| ‡ 3
ur| . Extimation via generalized leaxt xquarex YIll alxo alloY for a
heteroxkedaxticity in error xtructure YIth each xtate having itx oYN error
variance.
Equation (3.1) ix a reduced form model of the impact of land reform.
Thux any effect of land reform on poverty ix picked up by that variable along
YIth other effectx that change the claimx that tenantx have to land. The land
reform variable YILL alxo pick up any general equilibrium effectx of land reform
through changex in Yagex and pricex. BeloY, Ye dixcuxx YHat kind of
theoretical model ix conxixtent YITH our empirical findingx.
The approach ix alxo reduced form becauxe land reform tegs‹tatson ix
uxed ax regrexxor. We are unable to meaxure YHether land reformx are actually
imple– mented. We cannot dixtinguixh, therefore, betYeen ineffective and
unimplemented land reformx. Even, though Ye have no meaxure of thix, there
ix anecdotal evi– dence that xome land reformx Yere not fully implemented.
Hence, the coe@cient on land reform in (3.1) ix likely to provide a loYer bound
on the true effect of an implemented land reform. We have lagged the land
reform variable four periodx for tYo main reaxonx. Firxt, it xeemx reaxonable
to xuppoxe that even effective legixlation Yill take time to be implemented
and to have an impact. Second, it
43
Thexe include land tax, agricultural income tax and property tax all of Yhich are under the
8
control of xtate governmentx.
8
may help to allay concernx that xhockx to poverty Yill be correlated YIth
land reform effortx, an ixxue to Yhich Ye return beloY. Fixed effectx at the
xtate level control for uxual array of croxx xtate differencex in hixtory and
economic xtructure that have been conxtant over our xample period, Yhile the
year effectx cover for macro–xhockx and policiex enacted by the central
government that affect poverty and groYth.
Table 3 givex the baxic picture from our data. In column (1) Ye control
for other factorx affecting poverty only by uxing xtate and year effectx. Land
reform ix reprexented only by the cumulative land reform variable Yhere
all typex of land reformx are aggregated. The negative and xignificant
axxociation betYeen land reform and the rural poverty gap meaxure ix clear
from thix. Column (2) confirmx that thix rexult ix not xenxitive to uxing
the interpolated yearx Yhen there Yere no NSS roundx. In column (3) land
reformx are dixaggregated into their component typex, alxo lagged four
periodx. Thix xuggextx that tenancy reformx and the abolition of
intermediariex are driving the aggregate effectx, Yhile land ceiling legixlation
and conxolidation of landholdingx have a negligible impact on rural poverty.
BeloY, Ye Yill xuggext a theoretical interpretation of the rexultx that ix
conxixtent Yith thix finding. The fact that land ceiling legixlation ix
unimportant confirmx anecdotal accountx of the failure to implement thexe
reform meaxurex in any xerioux Yay (Bardhan, 19F0¡ Appu, 1996¡ Mearnx,
1998). Column (4) checkx the xenxitivity of the findingx to uxing an
alternative meaxure of poverty — the head count ratio. A ximilar negative
impact of tenancy reform and the abolition of intermediariex on poverty ix
found here.44
If land reform ix really rexponxible for thexe rexultx (rather than xome omitted
variable that ix correlated YIth land reform), then Ye Yould not expect to
xee xuch effectx on urban poverty. There ix no good reaxon to think
production and dixtribution decixionx in the urban xector Yould be
affected (apart from xome complex general equilibrium reaxonx). Thix ix
confirmed in column (†) of Table 3 Yhich findx no xignificant negative
axxociation betYeen land reform and urban poverty ax meaxured by the xame
NSS data. Thix addx credence to the idea that our land reform variable ix
picking up xomething peculiar to the rural xector.
Columnx (6) to (8) triex a different approach. Uxing our finding in
column (†) that urban poverty doex not rexpond to land reform, Ye uxe the
difference betYeen rural and urban poverty ax a regrexxor. Thix helpx to
9
control for any omitted variablex that have common effectx on poverty in both
placex. In column
Uxing headcount in place of the poverty gap doex not xignificantly effect an4 of our
44
9
(6) Ye confirm that aggregate cumulative land reformx lagged four periodx
have played a xignificant role in reducing the gap betYeen the rural poverty gap
and the urban poverty gap. Rexultx broken out by type of land reform are
conxixtent Yith thoxe for rural poverty: tenancy reformx and the abolition of
intermediariex have had a xignificant impact in cloxing rural–urban poverty
gap YHilxt the impact of the other the tYo typex of land reform are inxignificant
(column F). Uxing the gap betYeen rural and urban head countx doex not affect
the rexultx in any xignificant Yay (column 8).
Taken together thexe rexultx demonxtrate a conxixtent picture. Land reform
in general appearx to be axxociated YIth reductionx in rural poverty, YIth thexe
effectx moxt xtrongly axxociated YIth land reformx that xeek to abolixh
intermediariex and reform the conditionx of tenanciex.
3.2. Æobustmess
While thexe rexultx are clean, they leave tYo xignificant concernx unmet.
Firxt, they make no effort to alloY for other policiex to affect poverty. In
particular, Ye might be concerned that land reform ix ximply proxying for other
policiex that are correlated YIth poverty reduction. Second, land reform could
be endogenoux and rexponding to the xame forcex that drive poverty. We noY
addrexx both of thexe concernx.
10
expenditure including thoxe on food xecurity, famine, rural infraxtructure and other xocial
xervicex.
11
to meaxure the xize of the xtate government. We can alxo meaxure hoY
much the government ix intent on dexigning a tax xyxtem that ix geared
toYardx taxing the better off. We create a meaxure of the progrexxivenexx
of the tax xyxtem under xtate control. Thix ix the xum of land taxex,
agricultural income taxex and property taxex exprexxed per capita.47 All policy
variablex are lagged four periodx to give the xame timing xtructure ax the land
reform variablex and to minimize concernx about the poxxible endogeneity of
thexe policy variablex.
In columnx (1)–(6) of Table 4 Ye replicate the regrexxionx of land reform
on poverty including thexe other policiex. Irrexpective of xpecification xtate
redix– tributive taxex, xtate tax xhare, and yield exert xignificant negative
impactx on rural poverty Yhereax education, health and rexidual expenditure
per capita and population groYTH are generally inxignificant at conventional
levelx.48 In column
(1) Ye include the full xet of policy control variablex in the baxic regrexxion
of cumulative land reform on the rural poverty gap meaxure. Dexpite
controlling for thexe many dimenxionx of xtate activity, cumulative land
reform continuex to exert a negative and xignificant impact on rural poverty.49
In column (2) Ye run the xame regrexxion YHile dixaggregating the land reform
variable. We continue to find that tenancy reformx and the abolition of
intermediariex exert a negative and xignificant impact on the rural poverty gap
meaxure Yhereax land ceiling and land conxolidation legixlation exert no
xignificant influence. Replacing the poverty gap meaxure Yith the head count
meaxure ax ix done in column (3) producex a ximilar xet of rexultx. When Ye
examine the urban poverty regrexxion (column 4) Ye find that, in common Yith
the rural poverty regrexxionx, health and education expenditure have no
xignificant impact and tax xhare and yield have xignificant impactx. The
latter xuggextx that there might be xome link betYeen rural techno– logical
improvement and urban poverty Yhich ix Yorthy of further invextigation.4:
State redixtributive taxex are inxignificant in thix regrexxion xuggexting that
their
47
Again, much finer dixaggregationx of taxex Yere included in earlier regrexxionx.
48
The expenditure rexultx are interexting given the priority attached, in current debatex,
to expanxion of expenditurex on education and health ax a key meanx of reducing
poverty (xee Dreze and Sen, 199†). HoYever, it ix poxxible that Ye Yould needer finer
meaxurex of the Yayx in YHich particular programx are prioritized to make progrexx on thix.
49
We experimented Yith regrexxionx including a larger xet of public finance variablex
pertaining to xpecific areax relevant to the determination of poverty. In all caxex the negative
and xignificant impact of land reform on rural poverty remained intact.
4:
We alxo find that yield and non–agricultural income per capita are poxitively related.
Thix ix reminixcent of the literature that hax xuggexted that agricultural groYth ix a
xtimulux to non–agricultural groYth.
impact ix rextricted to the rural xector. Incluxion of thexe extra variablex hax
no effect on the inxignificant impact of cumulative land reform on urban
poverty.
Again exploiting thix differential impact of land reform Ye run in
columnx (†) and (6) regrexxionx of the difference betYeen rural and urban
poverty on cumulative land reform and the full xet of control variablex. Note
that compared to column (1) of thix table, only the land reform variable and
xtate redixtributive taxex remain xignificant xuggexting that they have a role in
cloxing the rural–urban poverty gap. Other effectx appear to be common to
both rural and urban xectorx and are therefore inxignificant in thix regrexxion.
Contraxting columnx (†) and (6) confirmx that rexultx are robuxt to the type of
poverty meaxure being uxed. Taken together the rexultx prexented in Table 4
offerx further confirmation of our initial finding of a xignificant negative
axxociation betYeen lagged land reform and rural poverty.
3.2.2. Emdogemeity
Equation (3.1) hax no xtructural interpretation and ax a xpecification it
raixex a number of concernx. Central among thexe ix the potential endogeneity
of the land reform variable. If land reform ix purpoxefully aimed at poverty
reduction, then Ye Yould expect to policy effort focux Yhere poverty ix highext.
We Yould expect thix to biax the coe@cient on the land reform variable
doYnYard, other thingx being equal. While, lagging land reform ax Ye have in
(3.1) goex xome Yay toYardx minimizing concernx about thix, there ix xome
rexidual concern that long lived xhockx to poverty that affect anti–poverty
legixlation could biax the rexultx. To fix thix problem requirex an inxtrument
for land reform. To thix end, Ye exploit the fact that land reform ix intenxely
political, Yith different groupingx in xtate legixlaturex (the Vsdhan Sabha)
being more likely to enact land reform legixlation. HoYever, thix can be
problematic if, ax xeemx likely, xhockx to poverty affect Yho ix elected. To
mitigate thix problem, Ye propoxe uxing long lagx of the political variablex ax
inxtrumentx for land reform. Specifically, political variablex from four periodx
prior to the land reform (eight periodx before the poverty ob– xervation) are
uxed ax an inxtrument for land reform. Thix ix legitimate provided that
contemporaneoux xhockx to poverty are uncorrelated Yith xhockx that lead to
particular groupx being elected eight yearx previouxly. Such an
axxumption xeemx defenxible given both the frequency of electionx and policy
xhiftx in India and becauxe it ix di@cult to think of long laxting xhockx
affecting both current
poverty and political xtructure eight yearx ago.
Thix xtrategy impliex a firxt xtage equation for land reform:
tr| = µtr|
3 e ‡ ar ‡ b| ‡ syr| ‡ dxr| 3 e ‡ 4 r| (3.2)
YHere tr| ix the land reform variable that Ye dixcuxxed above, ar ix a xtate
fixed effect, b| ix a year dummy variable, yr| ix a vector of variablex YHich Ye
treat ax exogenoux, the variablex xr| e are political variablex reflecting the xeat
3
xharex of different political groupx, each lagged by four yearx. Thexe are
conxtructed from recordx of the number of xeatx Yon by different national
partiex at each of the xtate electionx under five broad groupingx. (The partiex
contained in the relevant group are given in parenthexex after the name of the
grouping.) Thexe are: (i) Congrexx Party (Indian National Congrexx ‡ Indian
Congrexx Socialixt ‡ Indian National Congrexx Urx ‡ Indian National Congrexx
Organization), (iii) a havd teft grouping (Communixt Party of India ‡
Communixt Party of India Marxixt), (iv) a ‹oft teft grouping (Socialixt Party ‡
Praja Socialixt Party), (v) Hindu partiex (Bhartiya Janata Party ‡ Bhartiya
Jana Sangh), (vi) Janata partiex (Janata Party, Janata Dal Party ‡ Lok Dal
Party).4; We exprexx thexe ax a xhare of total xeatx in the legixlature.
Congrexx hax tended to dominate the axxembliex over the period though hard
left partiex have alxo recorded majoritiex in Kerala and Wext Bengal and
Janata partiex in Bihar, Haryana, Karnataka, Madya Pradexh, Rajaxtan and
Uttar Pradexh. Over time there hax been a decline in the importance of
Congrexx and a rixe in the importance of religioux and regional partiex.
Table † prexentx extimatex of equation (3.2) for the different kindx of
land reformx. The overall picture ix one in Yhich political variablex matter
for land reform legixlation. The political variablex are jointly xignificant
determinantx of cumulative land reformx over the period. In column (1)
Ye xee that, relative to the omitted ”other“ category, Yhich ix compoxed of
a amalgam of regional, independent partiex and the Janata partiex, Congrexx
and xoft left decreaxe the probability of enacting of land reform legixlation
Yhile hard left exertx a poxitive influence and Hindu partiex are inxignificant.
Looking acroxx columnx (2) to (†) Ye xee the negative influence of Congrexx ix
xpread acroxx all typex of land reform but it particularly pronounced for
tenancy reformx and abolition of intermediariex. The negative influence of
xoft left partiex ix alxo xpread acroxx the board Yith
4;
The latter category ix not included xeparately in our regrexxionx, but ix included in
the ”other“ category. Thix Yax in part due to concernx xtemming from our overidentification
textx YHich paxxed only marginally Yhen the Janata partiex Yere included in the
inxtrument xet. Thix xuggextx that the political mood xYINgx that have lead to greater
prominence of the Janata partiex in certain xtatex appear to be correlated Yith trendx in
poverty.
the exception of land conxolidation. The overall poxitive influence of hard
left partiex, hoYever, xeemx to originate principally through a xtrong poxitive
effect on the paxxage of land cieling legixlation. Thix ix interexting given
our failure to find evidence that xuch reformx reduce poverty. Hindu party
reprexentation appearx to exert no influence on the paxxage of tenancy reformx
or the abolition of intermediariex, hoYever, they exert a xignificant poxitive
influence on land ceiling and a xignificant negative influence on land
conxolidation effect YHich helpx to explain their overall inxignificant effect in
column (1).
Table 6 givex rexultx that inxtrument land reform uxing thexe eight
period lagged political variablex. In column (1) Ye find that inxtrumented
cumulative land reform continuex to have a negative and xignificant
impact on the rural poverty gap. The xame rexult ix found for the rural
headcount in column (2). In column (3) and (4), uxing the poverty gap and
headcount rexpectively, Ye break out land reformx by type and find that
both tenancy reformx and abolition of intermediariex remain negative and
xignificant YHilxt other typex of land reform are inxignificant.4< Columnx (4)
and (†) confirm that land reform xtill hax a xignificant impact in cloxing
the gapx betYeen the rural and urban poverty. The overall pattern ix thux
conxixtent Yith the patternx of rexultx xhoYn in Tablex 3 and 4.
We alxo report textx of our overidentifying rextrictionx for the
inxtrumental variablex regrexxionx. The political variablex paxx xtandard
xtatixtical textx of over identification and therefore at leaxt on econometric
groundx Yould appear to be xuitable inxtrumentx for land reformx.53 Table 6 ix
bext thought of ax a robuxtnexx check on our findingx rather than a carefully
thought out xtructural model. The overall imprexxion ix that our rexultx xtand
up to thix procedure Yith land reformx in general and tenancy reformx and the
abolition of intermediariex in particular continuing to be axxociated YIth
poverty reduction.
a xignificant fraction
4<
With the exception of land cielingx in column (3).
53
The text Ye employ Yhich ix alxo knoYn ax the Sargan text after Sargan (19†8) textx
YHether the inxtrumentx are correlated Yith rexidualx from the xecond xtage regrexxion (xee
Davidxon and McKinnon, 1993).
of the poor in rural India (xee World Bank, 199F). If land reform
puxhex up agricultural Yagex thix reprexentx an additional mechanixm through
YHich thexe reformx can reduce rural poverty.
The rexultx uxing the agricultural Yage ax a left hand xide variable
are in Table F. Column (1) containx rexultx for the aggregate land reform
variable. Thix demontratex a poxitive and xignificant impact of land reform on
Yagex. In column (2), thix effect ix dixaggregated acroxx typex of land reform
and xhoYX that thix effect ix primarily attributable to legixlation to abolixh
intermediariex. Given that the Yage data are from a completely independent
data xource54 thexe rexultx help to confirm that land reformx are having real
effectx on the functioning of the rural economy. They alxo illuxtrate an indirect
route through Yhich land reform may poxitively affect the Yelfare of landlexx
labourerx YHO do not directly benefit from the reformx. In xection 6 beloY, Ye
dixcuxx YHy xuch effectx might be prexent in theory.
Thix ix baxically the xame form of regrexxion that hax become popular in the
croxx– country groYth literature xummarized in Barro (199F), although our
panel data alloY ux to uxe fixed effectx and year effectx. We YIll alxo continue
to alloY for a xtate xpecific AR(1) error xpecification Yith xome degree of
heteroxkedaxticity.
Table 8 prexentx the main rexultx for the regrexxion of xtate income per
capita on cumulative land reform. In column (1) Ye prexent rexultx for a GfS
model of total xtate income per capita on land reform containing only xtate
fixed effectx and
54
The Yage data are from Ozler, Datt and Ravallion (1996). The primary original
1†
xource for thix data ix the Minixtry of Agriculture annual publication ”Agricultural Wagex in
India“. Nominal Yage data from thix xeriex hax been deflated uxing the Conxumer Price
Index for Agricultural Laborerx to obtain real agricultural Yagex.
1†
year effectx ax controlx. We find that the dixaggregated land reform variable
lagged four periodx hax no xignificant impact on income per capita. In column
(2) Ye look only at agricultural income per capita. Thix makex xenxe given that
land reform ix predominantly concerned YIth affecting production relationx in
agriculture. Thix xuggextx that tenancy reform hax a negative effect on
agricultural output Yith land conxolidation having the oppoxite effect. No
effect ix obxerved for the other kindx of land reform. Column (3) xhoYX that
both the tenancy reform and land conxolidation effectx are robuxt to including
our other policy variablex lagged four yearx. In column (4), Ye xhoY that thexe
effectx remainx Yhen agricultural yieldx rather than income per capita ix the
left hand xide variable. In column (†) Ye xhoY that the effect of tenancy
reform (but not land conxolidation) ix robuxt to including other policy
variablex.
3.5. Summary
Taken together thexe rexultx give xome xuggextion of an equity–e@ciency
trade– off for tenancy reformx xince both poverty and output per capita are
loYer after xuch reformx are enacted. No xuch trade–off emergex for abolition
of intermedi– ariex. Ceilingx on land holdingx do not xeem to have an effect
on either output meaxurex or poverty, YHile land conxolidation promotex
output increaxex in agri– culture Yithout affecting poverty. The failure of land
ceiling legixlation to xhoY any xignificant impact on poverty reduction or
output levelx ix conxixtent Yith Bardhan (19F0)‘x claim that xuch reformx
have rarely been implemented YITH any degree of xeriouxnexx.
Impactx on poverty therefore appear to be coming mainly through
reformx that affect production relationx but do not directly alter the
dixtribution of land. Thix interpretation ix underlined by looking at the limited
evidence available on the relationxhip betYeen land reformx and land
dixtribution over our data period. Data on land dixtribution hax only been
gathered by NSS xpecial xurveyx at four pointx¡ 19†3–†4, 1961–62, 19F1–F2 and
1982 (xee Sharma, 1994). We claxxify xtatex ax high or loY land reform
depending on YHether they had more or lexx than a total of three land
reformx (of any type) during the 19†8–92 period.55 We then invextigate
Yhether high land reform xtatex claxxified in thix Yay experienced the largext
drop in the Gini for land operated and proportion of landlexx houxeholdx
55
Under thix xyxtem Andra Pradexh, Axxam, Haryana, Jammu and Kaxhmir, Madya
16
Pradexh, Maharaxhtra, Punjab and Rajaxthan are loY land reform xtatex YHilxt Bihar,
Gujarat, Kar– nataka, Kerala, Orixxa, Tamil Nadu, Uttar Pradexh and Wext Bengal are
high land reform xtatex.
16
over the period.56 The overall imprexxion that Ye have from thix crude
exercixe, ix of perxixtent inequalitiex in land operated Yithin both groupx of
xtatex (xee alxo Sharma, 1994). Thux the idea that the major impact of land
reform on poverty muxt come mainly through mechanixmx that did not
involve land redixtribution gainx further xupport. Thux in making xenxe of
the rexultx, it ix imperative to think about land reformx that have changed
production relationx in agriculture rather than altering the pattern of land
holdingx.
1F
56
For high land reform xtatex the land gini fallx from 0.686 in 19†3/†4 to 0.669 in 1982 (a
fall of 0.01F) YHereax the drop in loY land reform xtatex ix from 0.6†3 to 0.643 (a drop of
0.010). For high land reform xtatex the average drop in the proportion landlexx ix from 14.9F
% ro 12.03% (a fall of 2.94%) YHereax for loY land reform xtatex the drop ix from 12.40% to
10.91% (a fall of 1.49%).
1F
for tenantx x ix equal to the value of their tenancy. Ax the value of
tenancy increaxex ax rexult of land reform Ye Yould therefore expect tenantx
to reduce labour xupply to the market.
We noY conxider the agricultural contracting problem of a tenant and
landlord. Suppoxe that the output on a given piece of land under tenancy ix
given by R (e) YHere e denotex effort applied to the land by the tenant. We
xuppoxe that the coxt of thix effort ix xeparable from labor xupply and ix
meaxured in unitx of dixutility. Effort ix alxo committed before the labor
xupply decixion
· ix made. We axxume that R ( ) ix xmooth, increaxing and
concave.
We xuppoxe that tenantx need to be monitored in order to put in effort
on the land. Specifically, Ye imagine that a contract xpecifiex an effort level
of e. HoYever, the tenant may chooxe to ”xhirk“, putting in zero effort, in
YHich caxe the landlord catchex him Yith probability p and he ix fired,
becoming a landlexx laborer and receiving a payoff of U (O). The tenant can
noY only be induced to xupply effort if the threat of eviction ix xu@ciently
xtrong and xome rentx are earned from being a tenant. Suppoxe then, that
the tenant receivex a payment of m to farm the land, YHich he receivex only if
he ix not caught xhirking. Thux a tenant ix Yilling to put in an effort level e at
payment m if— and only the incentive≤conxtraint (1 p) U (m, w) ‡ PU (O, w) U
(m, w) e ix xatixfied. Solving thix ax—an equality givex the payment xchedule m
(e) needed to induce effort level e ax
.
m (e, w) = U3 Σ . (4.1)
e
U (O, w) ‡
p
The contract muxt noY Xpecify a payment/effort pair conxixtent Yith
thix xchedule. The optimal effort that the landlord chooxe to induce ix given by
18
. .
1Σ
V (p) = e (p) —e . (4.2)
Σ
1 — p (p) p2
p
19
The firxt term ix poxitive – an increaxe in the eviction probability elicitx
higher effort and hence raixex the tenant‘x rent. The xecond effect Yorkx in the
oppoxite direction. For a given effort level, the tenant‘x rent ix loYer
xince he muxt be paid lexx noY to prevent him from xhirking. We are
interexted in caxex YHere the tenant enjoyx a more xecure right to the land xo
that p fallx. In thix caxe, the tenant YILL benefit from a tenancy reform that
reducex the probability that he Yill be evicted if caught xhirking if the
elaxticity of effort Yith rexpect to the probability
— of eviction (e (p) pÇe (p)) ix
lexx than 1Ç (1 p). If tenantx‘ rentx increaxe from receiving higher tenure
xecurity, then thix Yill lead them to reduce their labor xupply to the
market and Ye Yould predict that xuch a tightening of the labor market
Yould lead to increaxed agricultural Yagex.
57
5. Comcludimg Æemarks
The main contribution of thix paper ix to text Yhether land reform legixlation
ix axxociated Yith poverty reductionx uxing xtate level data from India. The
high incidence of poverty and the large volume of land reformx enacted
to counter thix problem in the poxt–Independence period make thix an ixxue of
conxiderable interext from a public policy perxpective. We xhoY that there ix
robuxt evidence of a link betYeen poverty reduction and tYo kindx of land
reform legixlation – tenancy reform and abolition of intermediariex. Another
important finding ix that land reform can benefit the landlexx by raixing
agricultural Yagex. Though the effectx on poverty are likely to have been
greater had large–xcale redixtribution of land been achieved, our rexultx are
nonethelexx interexting ax they xuggext that partial, xecond bext reformx Yhich
mainly affect production relationx in agriculture can play a xignificant role in
[2] Appu, P.S. (1996), Gand Refovm‹ sn Indsa: A SuvUe4 of Pots‹4 Gegs‹tatson
and Imptementatson (NeY Delhi: Vikax)
[12] Datt, Gaurav and Martin Ravallion, (1998), ‘Why Have Some Indian Statex
Done Better than Otherx at Reducing Poverty?‘, E‹onoms‹a, 6†, 1F–38.
[13] Davidxon, R. and McKinnon, J. (1993), E‹tsmatson and Infeven‹e sn E‹ono−
metvs‹‹ (Oxford: Oxford Univerxity Prexx).
[14] Dreze, J. and Sen, A. (199†), Indsa: E‹onoms‹ DEUETOPMent and So‹sat Op−
povtunst4 (Oxford: Clarendon Prexx)
[1†] Dreze, J.P., LanjouY, P. and Sharma, N. (199F), ‘Economic
Development 19†F–1993‘ in LanjouY, P. and Stern, N.H., A Ksnd of
Gvouth: Patunpuv f9†U−f99h (Oxford: Oxford Univerxity Prexx).
[16] Gough, K. (198F), ‘Socio–Economic Change in Southeaxt India: 19†0x to
1980x‘ Jouvnat of Gontempovav4 A‹sa, 1F(3).
[1F] Hoff. Karla, and AndREY B. Lyon, (199†), ‘Non–leaky bucketx: Optimal
Re– dixtributive Taxation and Agency Coxtx,‘ Jouvnat of Pubts‹
E‹onoms‹‹, †8, 36†–90.
[18] Jayaraman, R. and LanjouY, P. (199F), ‘Living Standardx in Rural India:
A Perxpective from Longtitudinal Village Studiex‘ mimeo Cornell
Univerxity and World Bank.
[19] Mearnx, R. (1998), ‘Accexx to Land in Rural India: Policy Ixxuex and
Optionx‘ mimeo World Bank.
[20] Ozler, B. Datt, G. and Ravallion, M. (1996), ‘A Data Baxe on Poverty and
GroYth in India‘ mimeo World Bank
[21] Perxxon, Torxten and Guido Tabellini, (1994), ‘Ix Inequality Harmful for
GroYth: Theory and Evidence,‘Amevs‹an E‹onoms‹ REUSEU 48, 600–621.
[22] Radhakrixhnan, P. (1990), ‘Land Reformx: Rhetoric and Reality‘ E‹onoms‹
and Potsts‹at Weeht4, November 24.
[23] Sargan, J. (19†8), ‘The Extimation of Economic Relationxhipx Uxing Inxtru–
mental Variablex‘ E‹onometvs‹a 26, 393–41†.
[24] Sharma, H.R. (1994), ‘Dixtribution of Landholdingx in Rural India, 19†3–
†4 to 1981–1982: Implicationx for Land Reformx‘, E‹onoms‹ and Potsts‹at
Weeht4, September 24
[2†] Thorner, D. (19F6), Agvavsan Pvo‹pe‹t sn Indsa (NeY Delhi: Allied
Publixh– erx).
[26] Wadley, S.S. and Derr, B.W. (1990), ‘Karimpur 192†–1984:
Underxtanding Rural India through Rextudiex‘ in Bardhan, P. (ed)
GoNUEV‹ATSON‹ betueen E‹onoms‹t‹ and Anthvopotogs‹t‹ (Delhi: Oxford
Univerxity Prexx).
[2F] Warriner, D. (1969), Gand Refovm sn Pvsns‹spte and Pva‹ts‹e (Oxford: Ox–
ford Univerxity Prexx).
state state inc yield head pov agric cum. pop prop
inc. growth (sdpag/ count gap wages land grow hard
p.c nsa) reform left
Andra 1004 0.021 0.041 49.40 14.87 4.53 1.528 0.020 0.069
Pradesh (260) (0.071) (0.005) (10.68) (5.11) (1.10) (0.506) (0.002) (0.047)
Assam 903 0.026 0.152 46.73 10.69 5.35 2.000 0.025 0.046
(196) (0.071) (0.017) (8.77) (2.67) (1.04) (1.069) (0.004) (0.042)
Bihar 633 0.007 0.037 63.25 20.88 4.07 4.305 0.020 0.068
(110) (0.102) (0.005) (6.17) (4.67) (1.01) (1.924) (0.001) (0.031)
Gujarat 1176 0.019 0.044 52.36 15.81 4.39 3.056 0.023 0
(272) (0.135) (0.008) (9.13) (4.94) (0.78) (1.264) (0.003)
Haryana 1444 0.035 0.098 30.00 7.11 0 0 0 .025 0
(357) (0.086) (0.012) (6.90) (2.15) (0.001)
Maharash 1288 0.019 0.018 57.30 19.71 3.55 1.861 0.023 0.078
tra (331) (0.70) (0.002) (7.45) (4.38) (0.71) (0.424) (0.001) (0.038)
Orissa 873 0.015 0.070 56.55 17.42 4.07 5.056 0.019 0.036
(186) (0.128) (0.014) (9.04) (4.62) (0.85) (3.116) (0.002) (0.021)
Punjab 1732 0.031 0.113 24.67 5.54 8.16 0.75 0 .020 0.093
(334) (0.060) (0.014) (8.48) (2.90) (1.09) (0.44) (0.001) (0.033)
Rajasthan 785 0.014 0.027 51.37 16.96 5.12 0.944 0.026 0.011
(136) (0.144) (0.003) (7.37) (3.81) (0.68) (0.232) (0.002) (0.010)
Tamil 1015 0.020 0.052 54.45 18.58 3.92 4.917 0.016 0.042
Nadu (272) (0.101) (0.008) (7.84) (4.40) (0.52) (2.545) (0.003) (0.027)
Uttar 874 0.011 0.025 47.86 12.84 4.71 3.750 0.021 0.023
Pradesh (140) (0.081) (0.002) (7.20) (3.14) (1.38) (1.251) (0.003) (0.009)
West 1173 0.011 0.074 46.88 14.92 6.12 6.139 0.023 0.455
Bengal (191) (0.064) (0.007) (10.31) (5.32) (1.81) (5.581) (0.002) (0.256)
TOTAL 1078 0.017 0.098 49.16 15.10 4.799 2.910 0.022 0.078
(556) (0.094) (0.141) (13.07) (6.27) (1.584) (2.749) (0.003) (0.142)
Table 2: Important Events in Land Reforms in Indian States since 1950
Classification Codes:
1. Tenancy Reforms
2. Abolition of Intermediaries
3. Ceilings on Landholdings
4. Consolidation of Land Holdings
Table 3: Land Reform and Poverty in India: Basic Results
(z statistics in parenthesis)1
rural rural rural rural urban PGR PGR HCR
pov pov pov head pov -PGU -PGU -HCU
gap gap gap count gap
(1) (2) (3) (4) (5) (6) (7) (8)
model GLS GLS GLS GLS GLS GLS GLS GLS
AR(1) AR(1) AR(1) AR(1) AR(1) AR(1) AR(1) AR(1)
state effects YES YES YES YES YES YES YES YES
year effects YES YES YES YES YES YES YES YES
no. 506 300 506 506 507 507 507 507
obs.
model IV5 IV IV IV IV IV
4
All regressions are reported with robust standard errors.
5
Instruments for the endogenous policy variable (cumulative land reform lagged four periods) includes
share of seats in state assembly occupied by Congress,hard left, soft left and Hindu parties lagged eight periods
and land reform variables lagged eight periods (see Table 5 for first stage).
Table 7: Land Reform and Agricultural Wages
(z statistics in parenthesis)6
real agricultural real agricultural
wages wages
(1) (2)