University of Southern California Dissertations and Theses

An investigation of consumption, insurance and village institutions in India

(USC Thesis Other)

An investigation of consumption, insurance and village institutions in India

PDF

Download

Open document

Flip pages

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content INFORMATION TO USERS This manuscript has been reproduced from the microfilm master. UMI films the text directly from the original or copy submitted. Thus, some | thesis and dissertation copies are in typewriter free, while others may be i from any type o f computer printer. V I The quality o f this reproduction is dependent upon the quality of the t 1 copy submitted. Broken or indistinct print, colored or poor quality | illustrations and photographs, print bleedthrough, substandard margins, : and improper alignment can adversely affect reproduction. £ In the unlikely event that the author did not send UMI a complete manuscript and there are missing pages, these will be noted. Also, if unauthorized copyright m aterial had to be removed, a note will indicate the deletion. Oversize materials (e.g., maps, drawings, charts) are reproduced by sectioning the original, beginning at the upper left-hand comer and continuing from left to right in equal sections with small overlaps. Each original is also photographed in one exposure and is included in reduced form at the back o f the book. Photographs included in the original manuscript have been reproduced xerographically in this copy. Higher quality 6” x 9” black and white photographic prints are available for any photographs or illustrations appearing in this copy for an additional charge. Contact UMI directly to order. UMI A Bell & Howell Information Company 300 North Zeeb Road, Ann Arbor MI 48106-1346 USA 313/761-4700 800/521-0600 l ___________ ______________________________ __________ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. AN IN V EST IG A T IO N O F C O N SU M PTIO N , IN SU R A N C E AND V ILLA G E IN STITU TIO N S IN IN D IA by Pushkar M aitra A D issertation Presented to th e FACULTY OF TH E G R A D U A TE SCH O O L U N IV ER SIT Y OF SO U TH ER N C A LIFO R N IA In P artial Fulfillm ent o f th e Requirem ents for th e D egree D O C TO R O F PH IL O S O P H Y (Economics) August 1997 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. UMI Number: 9816050 UMI Microform 9816050 Copyright 1998, by UMI Company. All rights reserved. This microform edition is protected against unauthorized copying under Title 17, United States Code. UMI 300 North Zeeb Road Ann Arbor, MI 48103 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. UNIVERSITY OF SOUTHERN CALIFORNIA THE GRADUATE SCHOOL UNIVERSITY PARK LOS ANGELES. CALIFORNIA 9 0 0 0 7 This dissertation, written by PUSHKAR MAITRA under the direction of J l.A .® Dissertation Committee, and approved by all its members, has been presented to and accepted by The Graduate School in partial fulfillment of re quirements for the degree of DOCTOR OF PHILOSOPHY Deair of Graduate Studies DISSERTATION COMMITTEE Chairperson Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Committee M em ber D ep a rtm en t A ndy N eum eyer Economics (C hair-person) Jeff Nugent Economics Caroline B etts Econom ics Selo Im rohoroglu Finance an d Business Economics Ayse Im rohoroglu Finance and Business Economics Doug Joines (E xternal) Finance and Business Economics Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Acknowledgements I would like to th an k , w ithout im plicating, C aroline B etts, Ayse Im rohoroglu, Selo Im rohoroglu, D oug Joines, Jeff N ugent, J im Robinson, N irvikar Singh, A nanish Chaudhuri, L a ta G angadharan, Sunanda Roy, C .V .S.K . Sarm a, T ridib S harm a, sem inar p articip an ts a t USC, participants at th e E conom etric Society M eetings a t Cal Tech, M eetings of th e Society of C om putational Econom ics a t Stanford a n d W estern Economic A ssociation M eetings a t S eattle for th e ir com m ents and suggestions on earlier drafts. Special thanks go to A ndy N eu m ey erfo r his support, suggestions and advise thro u g h o u t m y stay at USC. I w ould also like to th an k th e In tern atio n al Crop Research In stitu te for the Sem i-Arid Tropics (ICRISA T) for allow ing m e to use their d a ta an d M ark Rosenzweig at th e U niversity of Pennsylvania fo r allowing m e to use th e A R IS-N C A E R data. This research has been p artly funded b y th e U ni versity of S o u th ern C alifornia Haynes D issertation Fellowship and th e U niversity of California, S anta-C ruz Endow m ent for R esearch on P unjab, India. ii i Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Abstract T his thesis uses d ata from ru ral India to exam ine how households sm ooth con sum p tio n against income shocks. T h e existing literatu re on risk sharing in India has focussed on three villages in th e sem i-arid tropics and finds th a t households insure against idiosyncratic incom e shocks. This result is striking in th a t no evidence of such insurance has been o b tain ed using d a ta from o th e r countries. However these villagers axe left vulnerable to aggregate village level shocks. T hese findings lead to th e th ree broad questions th a t I exam ine in m y disserta tion. F irst, How do households sm ooth consum ption to th e ex ten t th a t they do? Second, w hat type of in stitu tio n s can help households insure against aggregate vil lage level risk? Third, is it possible to generalize th e results on risk sharing in these villages to o ther regions in In d ia - particularly rural India? In answ ering the first tw o questions, I focus on th e use of labor m echanism s for insurance purposes. It is show n th a t village labor m arket plays an im portant role in insuring households (p articu larly those who are restricted in th eir access to the village credit m arkets) against idiosyncratic shocks to agricultural incom e. I show th a t som e households sm o o th consum ption by m aking com pensating changes in labor m arket participation, and reducing own farm work, in response to adverse shocks to all non-labor income. I argue th a t an optim ally designed work-fare program (like the E m ploym ent G uarantee Scheme, EGS) will enable households to insure against aggregate village level shocks. T urning to th e th ird question, I find th a t th e results on risk sharing in the IC R ISA T villages cannot be generalized to o ther regions of India. Tests on consum ption insurance from villages in P u n jab show th a t there is no evidence of com plete risk sharing by households w ithin villages, thought there is su b stan tial risk sharing across villages. iii Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Contents L ist O f T ables vi L ist O f F igu res v iii 1 In tro d u ctio n 1 1.1 D ata D e s c rip tio n ................................................................................................... 7 1.2 A ppendix: Geographical C haracteristics of th e S u r v e y s ........................... 10 2 Is C on su m p tion Sm ooth at th e C ost o f V o la tile L eisure? 13 2.1 In tro d u c tio n ............................................................................................................. 13 2.2 E xisting Village I n s titu tio n s .............................................................................. 16 2.2.1 Village Financial M a r k e t s ...................................................................... 17 2.2.2 Village Labor M a r k e t s ............................................................................. 19 2.3 O ptim al Allocations in a D ecentralized E c o n o m y ........................................ 22 2.4 Econom etric S p e c ific a tio n ................................................................................. 28 2.5 E stim ation R e s u l t s ............................................................................................... 33 2.6 C o n clu sio n ................................................................................................................. 38 2.7 A ppendix: P areto O ptim ality w ith Labor-Leisure C h o i c e .................... 39 3 A Q u an titative A nalysis o f E m p loym en t G u aran tee Program s w ith an A p p lica tio n to Rural India 42 3.1 In tro d u c tio n ..................................................................................................................42 3.2 S tru ctu re of the E c o n o m y ................................................................................. 47 3.3 C alibrating th e M o d e l ........................................................................................ 56 3.4 W elfare Analysis of the Em ploym ent G uarantee P r o g r a m ...........................60 3.4.1 An Evaluation of th e EGS in M a h a r a s h tr a ..................................... 6 6 3.5 C o n clu sio n ................................................................................................................. 71 3.6 A ppendix 1 : C om putation A lg o rith m ............................................................ 73 3.7 A ppendix 2: Sensitivity Analysis ................................................................... 74 3.7.1 Varying the P aram eter V a lu e s .............................................................. 74 3.7.2 Relaxing the R estrictions on Labor M arket P articipation . . . 77 iv i ______ ___ _____________ ______ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4 C on su m p tion S m ooth in g in R ural P u n jab in th e G reen R evolu tion Years 81 4.1 In tro d u c tio n .............................................................................................................. 81 4.2 Test for Consum ption In s u ra n c e ....................................................................... 83 4.2.1 T h e o ry ........................................................................................................... 84 4.2.2 Em pirical S p ec ific atio n ............................. 87 4.3 E stim ation R e s u l t s ................................................................................................ 93 4.4 Risk Pooling Across V illa g e s................................................................................. 100 4.5 C o n clu sio n .....................................................................................................................102 4.6 A ppendix: T a b le s ...................................................................................................... 104 5 C on clu sion 118 6 R eferen ce L ist 122 v Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. List Of Tables 1.1 G eographical Location of Sam ple IC R ISA T Villages ............................. 10 1.2 Soil, Rainfall and Crop C h a r a c te r is tic s .......................................................... 11 1.3 Households P o p u la tio n .......................................................................................... 11 1.4 N um ber of Households Surveyed in E ach Village .................................... 12 1.5 R elative Farm S iz e s ................................................................................................. 12 2.1 % of T otal C redit O btained , Share of C onsum ption an d Incom e (1 9 7 5 -1 9 8 4 )................................................................................................................ 18 2.2 % C redit by Source for Different Land C la s s e s ............................................ 20 2.3 M ean Share of Wages in Total Incom e: by Land C la s s ................................... 21 2.4 R eduction in Incom e V olatility D ue to Labor I n c o m e ................................. 21 2.5 Im pact of U nanticipated Incom e Shock on A lo g (/tf) : 7 3 35 2.6 Im pact of U nanticipated Incom e Shock on A log(ctt) : /? 3 36 2.7 S um m ary of R e s u l t s .............................................................................................. 38 3.1 B enchm ark E c o n o m ie s ...............................................................................................58 3.2 S um m ary Statistics (B enchm ark Econom y) ............................................... 61 3.3 W elfare Costs for A lternative Econom ies (B enchm ark Econom ies) . . 65 3.4 Im plications of Varying 7 , A = 0 . 1 5 ............................................................. 75 3.5 Im plications of Varying a ................................................................................... 76 3.6 Im plications of Varying h9 ................................................................................... 77 3.7 R esults for A lternative Econom ies ................................................................ 79 4.1 S um m ary Statistics on Incom e an d C onsum ption (Classified by Land H o ld in g ) ......................................................................................................................... 104 4.2 S um m ary S tatistics on Incom e an d C onsum ption (Classified by Dis trict) 105 4.3 C onsum ption Regressions: F irst Differences ................................................... 106 4.4 C onsum ption Regressions: G row th R a t e s .......................................................... 106 4.5 C onsum ption Regressions: U sing V illage D um m ies .....................................107 4.6 C onsum ption Regressions by Land Class: F irst D iffe re n c e s ...................... 10S 4.7 C onsum ption Regressions by Land Class: G row th R ates ..........................109 4.8 C onsum ption Regressions by D istrict: F irst D iffe re n c e s ............................. 110 4.9 C onsum ption Regressions by D istrict: G row th R a t e s ................................. I l l vi Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4.10 C onsum ption Regressions by Land Class: Com parison of IV and OLS E stim ation ..................................................................................................................112 4.11 C onsum ption Regressions For A ltern ativ e M easures of C onsum ption: F irst Differences ....................................................................................................... 113 4.12 C onsum ption Regressions For A ltern ativ e M easures of C onsum ption: Growth. R a t e s .............................................................................................................. 114 4.13 R isk Pooling Across Villages ...............................................................................115 4.14 V olatility of Incom e and C onsum ption ............................................................ 116 vii Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. List Of Figures Figure 1: Average Steady S tate U tility for A lternative E c o n o m ie s .............. 62 Figure 2: Potential W elfare Benefits of E G S ....................................................... 6 6 Figure 3 (Panel A): W elfare C ost of Paying w = 0 . 2 5 ....................................... 67 Figure 3 (Panel B): W elfare Cost of Paying w = 0 . 4 0 ............................................6 8 Figure 3 (Panel C): Change in W elfare C o s t ........................................................ 69 Figure 4: Risk Pooling Across V illa g e s .......................................................................117 v iii Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Chapter 1 Introduction Incom e risk is a central feature of rural areas of developing countries, w here the m ajo rity of the population is em ployed in agriculture. Household incom e in rural areas is subject to laxge variations due to different kinds of shocks. F irst, there axe shocks due to aggregate factors like adverse w eather conditions or variations in intern atio n al crop prices. Second, there axe household specific shocks due to hum an and drought anim al sickness. In the presence of com plete insurance m arkets, fluc tu atio n s in earnings need not be reflected in fluctuations in consum ption. However, in m any parts of the world, such maxkets eith er do not exist or work im perfectly. In recent years the literatu re on risk and insurance in developing countries has focussed on th e em ergence of institutions designed to rem edy these maxket failures . 1 1See surveys by Alderman & Paxson (1992), Morduch (1995) and Townsend (1995). 1 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. T h e existing lite ra tu re on risk sharing in Indian villages has focussed on three villages in the states o f A ndhra Pradesh and M a h a rash tra . 2 Tow nsend’s (1994) stu d y on risk and insurance in these villages finds th a t households insure against idiosyncratic shocks. T h is is a striking result because researchers have not ob tain ed evidence of such insurance using d ata from o th er co u n tries . 3 Lim (1991) however finds th a t households in these villages axe left v ulnerable to aggregate village level shocks. T hese findings lead to th e th ree broad questions th a t I exam ine in m y disser tatio n . First, w hat are th e institutions th a t enable villagers to insure against id iosyncratic incom e shocks? T he existing research on risk sharing in India finds th e existence of risk sharing, b u t does not specify th e m echanism by which risk is shared. Second, w hat type of in stitu tio n s can help households insure against aggregate vil lage level risk? Lim ’s (1991) stu d y finds th a t households in these villages are left vulnerable to aggregate village level risk as in stitu tio n s do not work efficiently to insure households against such risk. Third, is it possible to generalize th e results on risk sharing in these villages to other regions in In d ia - in p articu lar to villages in Punjab? T he choice of P u n jab is m otivated by th e fact th a t households in ru ral P unjab and those in th e ICRISA T villages co n stitu te two extrem es of th e rural population in India. W hile th e ICRISAT villages are characterized by prim itive 2 Data collected by the International Crop Research Institute for the Semi-Arid Tropics (ICRISAT). This data is know as the ICRISAT data. This data will be described in greater detail in section 1.1 below. 3See Mace (1991) for tests for US, Townsend (1995) for Thailand and Deaton (1990) for Cote D ’Ivoire. 9 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. m odes of production and susceptibility to th e vagaries of the monsoons, agriculture in P unjab is characterized by m odem capital intensive techniques of prod u ctio n and extensive irrigation facilities. In answ ering th e first two questions, I focus on th e use of labor mechanisms for insurance purposes . 4 I find th a t village labor m arkets play an im p o rtan t role in in suring households (particularly those who are restricted in their access to th e village credit m arkets) against idiosyncratic shocks to agricultural income. G eneralizing the argum ent o f K ochar (1995a, 19956) I show th a t som e households sm ooth consum p tion by m aking com pensating changes in labor m arket participation, and reducing own farm work, in response to adverse shocks to all non-labor incom e. U sing this fact I argue th a t an optim ally designed work-fare program (like th e E m ploym ent G uarantee Scheme, EGS) will enable households to insure against aggregate village level shocks. Turning to the th ird question, I find th a t the results on risk sharing in th e IC R ISA T villages cannot be generalized to o th er regions of India. Tests on consum ption insurance from villages in P u n jab show th a t there is no evidence of com plete risk sharing by households w ithin villages, though there is su b stan tial risk sharing across villages in Punjab. How do households sm ooth consum ption to the extent th a t they do? I argue, in ch ap ter 2 of m y dissertation, th a t depending on th eir access to credit m arkets, households use different m echanism s to insure against income shocks. T h e m edium 4Labor mechanisms refer to households making compensating changes in labor market partici pation in response to adverse shocks to crop profits. See Kochar (1995a, 19956.). 3 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. and large farm ers are fully insured. They behave as if they have access to state- contingent transfers th a t m im ic Arrow-Debreu securities, enabling them to rem ain unaffected by income shocks (neither consum ption nor leisure is affected). Small farm ers are also able to insure consum ption. However, th e results for the sm all farm ers is qualified by th e observation th a t these farm ers are constrained in the credit m arket and respond to adverse income shocks by increasing participation in the daily wage labor m ark et and reducing own farm work (thereby sm oothing consum ption by sm oothing incom e directly). T he landless are left vulnerable to such shocks. They are constrained in the credit m arket and also do not have the m argin to increase the labor m arket participation in response to adverse shocks to incom e, since even in a norm al year labor income form s th e m ain source of income for these households. T h e fact th a t some households (in particular those who are restricted in their p articip atio n in the village cred it m arkets) use particip atio n in the daily wage labor m arket to sm ooth consum ption against income shocks, has im p o rtan t im plications for designing institutions th a t will help households (specially th e sm all farmers and landless households) to sm ooth consum ption. I argue th a t an optim ally designed work-fare program can work successfully in this respect. C h ap ter 3 of m y thesis investigates one such in stitu tio n : th e Em ploym ent G uarantee Scheme (EGS) in the Indian S ta te of M aharashtra. T he EGS was originally designed to provide em ploym ent to the rural unskilled population on dem and and thereby reduce absolute 4 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. poverty in the ru ral areas and provide a safety-net for the participants. T h ere exists a laxge volum e of literatu re th at exam ines th e effect of the EGS on ru ral poverty0, but the insurance aspect of the EGS has n o t been investigated previously. I study this aspect of th e EGS by conducting a com putational experim ent in a dynam ic general equilibrium m odel. Agents in this econom y face exogenous incom e shocks and th e only way th ey can self-insure is savings in th e form of a non-interest bearing asset. T he m odel is calibrated using d a ta from th e two ICRISAT villages in th e state of M aharashtra, which had a functioning Em ploym ent G uarantee Schem e (EG S) in the period 1979-1984. I find th a t when agents participating in th e EG S are paid the optim al wage, they do not need to hold any asset as precautionary savings. All insurance in this case is provided by th e EG S. In general, because of its insurance aspect, the optim al wage rate in th e EGS is found to be higher th an th e m arginal product of labor. T h e optim al wage and th e welfare gains of the program depend on how productive the EGS is, relative to th e private sector. T h e w elfare gains of paying the optim al wage as opposed to a wage equal to the m arginal p ro d u ct of labor in th e EGS varies from 0.01% of G D P to 2.63% of GDP, depending on th e labor productivity of th e EGS. T he actual wage p aid by the EGS during th e years 1979- 1984 yields a welfare level th at is lower th a n paying a wage equal to th e m arginal product of labor. F urther I find th a t th e 1988 increase in wages in th e EG S actually 5See Bhende et al (1992), Ravallion (1991), Ravallion, Datt & Chaudhuri (1993), Gaiha (1996). • 5 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. decreased welfare, a n d an even greater increase in wages is necessary to increase the welfare of th e population. C h ap ter 4 of m y thesis extends the test of consum ption insurance to a differ ent region of India - n am ely ru ral Punjab. I exam ine th e success of households in ru ral P unjab in insuring consum ption against incom e flu ctu atio n in the green revo lution years (1968-69 - 1970-71). I find th a t for th e p o p u latio n as a whole th e null hypothesis of full consum ption insurance is rejected. T h e qualitative properties of these results are very different to those obtained from using d a ta from villages in A ndhra P radesh and M aharashtra, where there is evidence of significant risk sharing by households against idiosyncratic income shocks. T h e re is no such such evidence for households residing in P unjab. W hen th e p o p u latio n is subdivided according to Iand-ownership, th e null hypothesis is rejected for th e landless households and sm all farm ers. However, for th e m ed ium and large farm ers th e null hypothesis of full insurance cannot be rejected. Further, there exists significant regional variation in th e ability of households to insure consum ption ag ain st incom e changes. I also conduct a te st of full insurance at th e regional level an d test w hether there is risk sharing across villages. Lim (1991) conducts a test of risk sharing across villages in the IC R ISA T region and finds th a t households are left vulnerable to shocks which are insurable a t th e regional level th ro u g h in ter village level tra d e and proper financial interm ediation. However, he defines th e regional economy to consist of th e three villages A urepalle, Shirapur and K anzara pooled together. These three 6 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. villages are spatially separated an d do not form a regional economy. O n th e other hand pooling th e 20 villages spread over 9 districts in P unjab (and H aryana) does define regional economy. I te st for risk sharing across villages in ru ral P u n jab , by placing each village in a larger regional economy. I find th a t th ere is significant evidence of risk sharing across villages. So while the null hypothesis of risk sharing w ithin each village is rejected, th ere is evidence th at if the villages pool together th en th ey can insure against aggregate incom e risk. 1.1 Data Description I will use d a ta collected by th e International Crop Research In stitu te for th e Semi A rid Tropics (ICRISA T) and th e A dditional Rural Incomes Survey (A RIS) D ata, collected by the N ational Council of A pplied Economic Research (N C A E R ), New D elhi. The ICRISAT d ata consists of panel d ata collected over th e period 1975- 1984 for a set of 10 villages spread over 5 districts in the states of A n d h ra P radesh, M aharashtra, G ujarat and M adhya Pradesh. One can refer to W alker & : R yan (1990) for details on th e ICRISA T villages and Singh, Binswanger & c Jo d h a (1985) for details on th e survey. T he IC R ISA T region consists of a high risk, prim arily agrarian economy, spread over a vast geographical area in India, in th e states of A n d h ra Pradesh, M aharashtra, G u jarat and M adhya Pradesh. As a first step in the collection of d ata, several districts were selected in the particular agro-clim atic zone. 7 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The second step consisted of selecting tw o villages from each d istric t . 6 T he th ird step consisted of selecting households from each village. 40 households w ere selected from each village to ensure equal representation of all categories of households using a stratified random sam pling procedure. T hese households were from four landholding classes - landless laborers, small farm ers, m edium farmers and large farm ers. In case a household dropped out due to som e reason, it was replaced by an o th er from the sam e stra tu m from the original list o f households collected in th e village (these households are num bered differently). R esident investigators were expected to collect inform ation every three weeks and tran sfer th e d a ta into specific coding sheets. T h e m ost com plete (and continuous) d a ta exists for the villages A urepalle, Shira- pur an d K anzara. I will focus on these th re e villages. Table 1.2 presents th e rainfall, soil and crop characteristics in the th ree corresponding districts. Tables 1.3 and 1.5 present som e descriptive statistics on household population and relative farm sizes in the th re e villages, Aurepalle, S hirapur and K anzara. Note however, th a t even though th e survey was conducted over th e period 1975-1984, the labor m arket d a ta is available only for the period 1979-1984. Hence I will use d ata for th a t period in m y analysis. T h e ARIS d ata, on the other hand consists of 4118 households draw n from an annual national survey of rural landed an d landless households in th re e consecutive 6See Table 1.1 for geographical location of the villages and districts. S Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. years, 1968-69, 1969-70 and 1970-71. T h e households were selected as a representa tive cross-section of rural households. T his survey will henceforth be referred to as the A R IS-N C A E R survey. D etailed inform ation was collected on th e dem ographic com position o f the household, th e level o f incom e by source, the level of consum ption by type of expenditure, level of savings by ty p e an d th e level of asset holdings by type. O ne can refer to N C A ER (1975) for m ore details on th e survey. Since my thesis deals prim arily w ith P unjab, I consider d a ta from a set of 20 villages in the states of P u n jab and Haryana, spread over 9 d istricts, covering 334 households (See Table 1.4) . 7 T h e survey, it m ight be noted, was undertaken at a tim e w hen the G reen R evolution was ju st getting sta rte d in m any parts of th e country, including Punjab and by 1970-71, the green revolution had started m aking its im p a ct . 8 T he A RIS d a ta enables one to conduct a test of full insurance a t th e regional level and te st w hether there is risk sharing across villages. Lim (1991) conducts a test of risk sharing across villages in th e IC R ISA T region and finds th a t households are left vulnerable to shocks which are insurable a t the regional level th rough inter village level tra d e and proper financial interm ediation. However, he defines the 7Haryana became a separate state in 1966, prior to which it was a part of the province of Punjab. They still share the same capital city, Chandigarh. So for the purposes of this paper, by Punjab I mean these villages in Punjab and Haryana. 8 Following Chakravarty (1987) one can define the Green Revolution as the adoption of a new package of policies comprising of the following set o f measures (refer to Chakravarty (1987), page 25): (1) A shift from ”major” to "minor” irrigation works, which implied largely a shift from publicly financed large irrigation projects to small tube wells and energized pump sets; (2) Adequate provision of "credit” to those who were considered to be credit-worthy, which in effect meant the large farmers; (3) An alteration in the input base of agriculture, which meant an increase in the rate of fertilizer consumption along with commercial sources o f energy, such as electricity and diesel oil; (4) The development of fertilizer sensitive varieties of grains. 9 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. regional econom y to consist of th e three villages A urepalle, S hirapur and K anzara pooled together. These th re e villages are so widely sep arated geographically th a t they do not tru ly form a regional economy. O n th e o th er h an d pooling the 2 0 villages spread over 9 districts in P u n jab (and H aryana) does define regional economy. T he ARIS d a ta therefore allows us to test for risk sharing across villages, which is not possible using th e IC R ISA T d ata. I will come back to th is issue in chapter 4 of m y dissertation. 1.2 Appendix: Geographical Characteristics of the Surveys T able 1.1: G eographical Location of Sam ple IC R ISA T Villages Village D istrict S tate A urepalle M ahbubnagar A n d h ra P radesh D okur M ahbubnagar A n d h ra P radesh Shirapur Sholapur M ah arash tra K alm an Sholapur M ah arash tra K anzara Akola M ah arash tra K inkheda Akola M ah arash tra Boriya Sabarkantah G u jarat R am p u ra Sabarkantah G u jarat P ap da Raisen M adhya P rad esh R am p u ra K alan Raisen M adhya P rad esh Source: S in g h , B in sw an g er & Jo d h a (1985) 10 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 1.2: Soil, Rainfall and Crop Characteristics C haracteristic M ahbubnagar D istrict Sholapur Akola Soil Red Deep black M edium deep heavy clay black clay W ater R etention High High M edium C apacity R ainfall U nassured U nassured A ssured (31%) (35%) (2 2 %) M ajo r Crops K harif Rabi K harif and Rabi F ig u res in P aren th esis in d icate C oeffient o f V ariatio n o f R ainfall S ource: W alker Sc R yan (1990), page 4 Table 1.3: H ouseholds Population Sam pling Inform ation A urepalle Village Shirapur K anzara Laborers 146 97 54 (30.7) (32.7) (32.0) C ultivators 322 183 109 (67.7) (61.6) (64.5) O thers 8 17 6 (1.7) (5.7) (3.6) Total 476 297 169 ( 1 0 0 ) ( 1 0 0 ) ( 1 0 0 ) S ource: W alker & R yan (1990), p a g e 13 F ig u re in B rackets refer to % o f to ta l Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 1.4: Number of Households Surveyed in Each Village Village D istrict S tate N um ber of Households D alla Ludhiana P unjab 15 H eran Ludhiana P unjab 17 Jassowal Ludhiana P unjab 14 Barm i Ludhiana P unjab 17 B hadus G urgaon H aryana 17 Notki G urgaon H aryana 1 1 K alyat Jind H aryana 23 K hori Lam ba Jin d H aryana 17 N anajathaji K ap u rth ala P unjab 17 K hajurla K ap u rth ala P unjab 17 B alta Hissar H aryana 23 B hagana Hissar H aryana 17 B hojraj Hissar H aryana 18 N ayabano R ohtak H aryana 17 Sahiyatera R ohtak H aryana 17 Makowal A m ritsar P unjab 17 M ohanbhandaria A m ritsar P unjab 14 B akhushah Ferozpur P unjab 1 2 Baxmawala Ferozpur P unjab 15 K otla Ju ll under P unjab 19 Satowali Ju ll under P unjab 17 S ource: A R IS -N C A E R D a ta T able 1.5: R elative Farm Sizes Sam pling Inform ation ICRISA T A urepalle S hirapur K anzara P unjab Small 0 .2 - 1 . 2 0 .2 -2 . 0 0 .2 - 1 . 8 0.1-2.5 M edium 1.2-3.2 2.0-5.3 1.8-5.3 2.6-6.5 Large > 3.2 > 5.3 > 5.3 > 6 . 6 F arm Size is in hectares S ource fo r IC R ISA T V illages: W alker Sc R y a n (1990), page 13 S ource fo r P u n ja b : N C A E R (1975) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Chapter 2 Is Consumption Smooth at the Cost of Volatile Leisure? 2.1 Introduction In th is chapter I exam ine how villagers sm ooth consum ption to th e ex ten t th a t they do. To do this, I investigate th e functioning of the existing village m arkets and in stitu tio n s in detail. I use a decentralized general equilibrium m odel to exam ine w hat in stitu tio n s enable households to sm ooth consum ption against unanticipated incom e shocks and show th a t different households use different in stitu tio n s to sm ooth consum ption against income shocks. Because of differentiated access to village m ar kets som e villagers are able to insure against income shocks, w hereas others are left un-insured. 13 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. I generalize th e argum ent of K ochar (19956) to exam ine household responses to u n an ticip ated shocks to all non-wage incom e . 1 I show th at differentiated access to m ark ets (p articu larly financial m arkets) im plies th a t different households respond differently to sim ilar shocks. M edium an d laxge farm ers (who have access to unre stricted credit) use sta te contingent transfers. Sm all farm ers who are constrained in th eir ability to borrow in tim es of need, vary th eir labor m arket paxticipation, substi tu tin g own farm work w ith work in the daily wage labor m arket. Finally th e landless do not have credit and axe also unable to vary th eir labor maxket paxticipation in response to incom e shocks an d axe left un-insured. T h e population is subdivided by land class and for each land class I te st w hether individuals have u nrestricted access to a perfect set of A rrow -Debreu securities or not. If individuals w ithin a p articu lar lan d class have unrestricted access to a full set of Arrow D ebreu securities th ey axe unconstrained. If, on the o ther hand, individuals do not do not have unrestricted access to insurance m axkets, they axe constrained. Since these individuals cannot obtain credit to sm ooth consum ption against incom e shocks, they m ust use other m echanism s to do so. Individuals axe n o t ex-ante subdivided into th e two broad categories (constrained or unconstrained). Instead, I test w hether individuals belonging to a p articu la r land class axe constrained o r not. xKochar (19956) shows that households use the labor market to smooth unanticipated variations to crop profits. 14 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. O ptim al choices of consum ption and leisure for each subset of th e population are derived by including leisure as a com ponent of the u tility function . 2 This leads to testab le im plications of consum ption and leisure insurance, w ith and w ithout perfect m arkets. T he effect of u n an ticip ated income shocks on leisure and consum ption is estim ated separately for each land class. For th e m edium an d large farm ers, the null hypothesis of perfect m arkets cannot be rejected. They are therefore unconstrained, n either leisure nor consum ption is affected by u n anticipated incom e shocks. For th e sm all farm ers also, th e null hypothesis of perfect m arkets cannot be rejected. However, the results for th e sm all farm ers are qualified by th e observation th at these farm ers are constrained in th e credit m arket and respond to adverse shocks by increasing participation in th e daily wage labor m arket (and reducing th eir own farm work), and so sm ooth consum ption by sm oothing incom e directly. For th e landless th e null hypothesis of perfect m arkets is rejected. They are constrained in the credit m arket and in addition cannot increase their labor m arket p articipation in response to adverse shocks, and hence cannot sm ooth consum ption against such shocks. This chapter uses panel d a ta collected by th e International Crop Research Insti tu te for the Sem i-arid Tropics (IC RISA T) for a set of farm households in C entral and Southern India. T he m ost com plete d a ta exists for three villages A urepalle, Shirapur and K anzara in the states of A n d h ra Pradesh and M aharashtra. W hile th e survey 2The existing test of consumption insurance defines utility only as a function of consumption. This specification gives optimal consumption allocations which are used in the test of consumption insurance. However, if households use labor market participation to smooth consumption, then consumption might be smooth at the cost of volatile leisure. Not including leisure as a component of the utility function will lead to the model being mis-specified. 15 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. was conducted over the period 1975-1984, labor m arket d a ta is available only for the period 1979-1984. Hence, I use d a ta for th a t period in th e em pirical analysis. One can refer to W alker R yan (1990) for a detailed account of th e ICRISA T villages, and Singh, Binswanger & Jo d h a (1985) for m ore details of th e survey. T h e rest of this chapter is organized as follows. Section 2.2 describes th e exist ing village institutions. I present evidence to show th a t all households do not have equal access to financial m arkets. T h e traditional argum ent of th e financial m arkets ene.bling individuals (and households) to sm ooth consum ption does not always hold tru e for households residing in th e ICRISA T villages. I show th a t labor m arkets play an im p o rtan t role in risk m anagem ent, a t least for some sections o f th e population. Section 2.3 exam ines how consum ption insurance is a ttain ed in a decentralized m ar ket economy. This section sets up th e theoretical form ulation for th e individual’s optim ization problem and derives testab le im plications of consum ption and leisure insurance. Section 2.4 describes th e econom etric specification and the estim ation techniques. Section 2.5 presents th e estim ation results and Section 2.6 concludes. T he P areto O ptim al allocations which provide the benchm ark for our analysis axe derived in th e appendix. 2.2 Existing Village Institutions This section exam ines the m axkets and institutions in th e ICRISA T villages, in par ticular the financial and th e labor m axkets. It is generally argued th a t by providing 16 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. explicit state contingent loans credit/financial m arkets enable households to insure against incom e fluctuations (see Esw aran K otw al (1989)). I present evidence th a t relations in th e village financial m arkets in th e IC R ISA T villages axe skewed in favor of th e large an d m edium farm ers while th e landless laborers and sm all farm ers face adverse financial m axkets. As a consequence it is unlikely th a t all households sm ooth consum ption by o b tainin g state-contingent transfers from th e village financial m ar kets. W age income, on th e other hand, is an im p o rtan t com ponent of to tal household incom e for th e landless and the sm all farm ers and it also plays an im p o rtan t role in reducing the volatility of income. 2 .2 .1 V illa g e F in an cial M ark ets T he prim ary sources of credit in the ICRISA T villages are ( 1 ) Governm ent (w hich includes Food C orporation of India (FC I), Fair Price Shops (FPS) and P anchayati Sam ities); (2) Com m ercial Banks; (3) P rim ary A gricultural Cooperatives (P A C ’s); (4) M oney Lenders; (5) Landowners and (6 ) O thers (w hich includes relatives, friends, p rivate shops and m illers). T he inform al m oney lending system is quite well devel oped and extrem ely prevalent in A urepalle — or m ore generally in the entire M ahbub- nagar district of A ndhra Pradesh, as opposed to th e Sholapur and Akola d istricts in M aharashtra, where institutional credit plays a m ore im p o rtan t role. In th e absence of in stitu tio n al credit, a village typically has up to 15 professional m oney lenders, 17 1 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. and each services up to 50 custom ers. T h e villagers and the m oney-lenders are as sociated in long-term relationships. For sm all loans, individuals can approach other m oney lenders, b u t sw itching of long-term m oney lenders, th o u g h feasible, is not com m on. T hese m oney lenders provide loans for b o th consum ption and produc tion purposes. In stitu tio n al sources (th e governm ent, th e com m ercial banks and the PA C ’s) provide very little credit for consum ption purposes, (see W alker & Ryan for m ore details). Table 2.1: % of Total C redit O b tain ed , Share of C onsum ption and Incom e (1975- 1984) L and Class A urepalle S hirapur K anzara Landless % Share of C redit 4.71 13.57 0.37 % Share of C onsum ption 15.07 30.34 13.02 % Share of Incom e 9.91 18.81 9.60 Sm all Farm ers % Share of C redit 12.16 21.13 10.39 % Share of C onsum ption 22.03 22.42 19.96 % Share o f Incom e 19.16 21.79 15.58 M edium Farm ers % Share of C redit 27.64 27.74 23.05 % Share of C onsum ption 22.56 26.58 22.95 % Share of Incom e 19.58 22.62 18.86 Large Farm ers % Share of C redit 55.49 37.55 66.19 % Share o f C onsum ption 40.34 30.66 44.06 % Share o f Incom e 51.35 36.79 55.96 D a ta S ource: IC R IS A T VLS T apes D epending on their land class, borrow ers have differentiated access to both insti tu tio n al and inform al m arkets. V illagers axe differentiated in term s of land-holding . 3 Table (2.1) presents some sum m ary statistics on credit availability by land class. I 3Classification o f households in terms of land-holding is explained in Table (1.5) in chapter 1. IS Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. also present th e share of consum ption and incom e for each land class. T he m edium and large farm ers obtain m ore credit th a n th eir consum ption requirem ents w arrant and it is th e opposite for the landless households and th e sm all farm ers. For exam ple th e landless in A urepaile account for 15.07% of to tal village consum ption b u t o b tain only 4.71% of to tal credit. D etailed figures on the proportion of credit obtained by th e different landholding classes from form al and inform al sources is presented in Table (2.2). These figures clearly show th a t the landless laborers and th e sm all farm ers are restricted in their access to th e village credit m arkets. For exam ple, from Table (2.1) see th a t th e landless receive only 4.71% of value of all credit given o u t in A urepaile. T hey obtain 1.39% o f all form al sector credit given out and 6.64% of all inform al sector loans (see Table (2.2)). For m ore details on the sum m ary statistics and the functioning of the village financial m arkets, refer to W alker R yan (1990) and Bhende (1986). T he landless (and sm all farm ers to a lesser degree) are clearly excluded from th e village financial m arkets. 2 .2 .2 V illa g e L abor M a rk ets T he labor m arket is active in all three villages, w ith farm households depending heavily on hired labor. 4 T he daily wage m ark et is generally viewed to be com petitive. C ontracts are defined a t the beginning of the day regarding the num ber of work hours, and paym ents are m ade at the end of th e day. Because m ost contracts 4This section draws on the description of labor market in Walker & Ryan (1990) and summarized in Kochar (19956). Bhende et al (1992) provides a description of the impact of labor market on the rural poor. 19 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 2.2: % Credit by Source for Different Land Classes Land Class A urepaile Shirapur K anzara Form al Inform al Form al Inform al Form al Informal Landless 1.39 6.64 13.76 13.29 0.06 4.19 (783.86) (393.15) (1211.97) (530.56) (13.93) (116.17) Sm all Farm ers 2.67 17.67 10.94 37.49 10.24 12.31 (659.38) (865.81) (642.16) (1409.09) (493.29) (211.88) M edium Farm ers 24.42 29.61 35.79 15.77 23.42 18.42 (3193.67) (1800.76) (1804.16) (563.82) (788.25) (291.97) Large Farm ers 71.70 46.08 40.32 33.45 66.28 65.08 (6028.51) (2853.04) (1560.42) (1447.46) (2200.82) (1160.45) F ig u res in b rack ets a re m ean value o f lo an s from th e tw o sources for each la n d class in R s. D a ta S ource: IC R ISA T VLS T apes are extrem ely short term , th ere is no evidence of inter-linkages betw een th e labor and financial or goods m arkets. M ost cultivator households (irrespective of farm size) p artic ip a te in the daily wage m arket, both as buyers and sellers of labor. An im p o rtan t aspect of the labor m arket is th e considerable segm entation of tasks by gender, resulting in separate labor m arkets for men and women. Households in S h irap u r and K anzara, the two M ah arash tra villages have access to the M aharashtra G overnm ent’s Em ploym ent G uarantee Schem e (EGS) - a rural em ploym ent program th a t guarantees em ploym ent in public works projects at a wage slightly less th an the prevailing agricultural wage. T his adds stability to the wage incom e. Increased wage incom e, particularly from public works projects frequently constitutes th e prim ary m eans of dealing w ith aggregate village level shocks like droughts (see chapter 3 of this dissertation). 20 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 2.3: Mean Share of Wages in Total Income: by Land Class Land Class A urepaile S hirapur K anzara Pooled Landless 67.73 80.64 89.69 77.18 Sm all Farm ers 50.93 66.78 79.39 61.29 M edium Farm ers 50.25 58.56 76.68 53.66 Large Farm ers 1 2 . 2 1 36.82 30.36 18.24 D a ta S ource: IC R ISA T VLS T ap es Table (2.3) presents evidence on th e im portance of labor incom e in th e three ICRISA T villages. For the landless, wages co n stitu te 67.7% of th eir to ta l incom e in A urepaile, 80.6% in Shirapur, and 89.7% in K anzara . 5 In com parison, for th e large farm ers the corresponding figures are 12.2%, 36.8%, 30.4% respectively. W hen th e three villages are pooled, th e m ean share of wages is 77.18% for th e landless, 61.29% for th e sm all farm ers, 53.66% for the m edium farm ers and 18.24% for the large farm ers. Table 2.4: R eduction in Incom e V olatility Due to Labor Incom e Land Class A urepaile S hirapur K anzara Pooled Landless 6 6 . 8 8 141.38 115.81 163.78 Sm all Farm ers 47.91 56.86 77.60 65.56 M edium Farm ers 23.94 37.60 49.09 34.34 Large Farm ers 10.62 21.03 20.30 12.63 D a ta S ource: IC R ISA T VLS T apes 5Total household income consists of crop income, income from trade and handicrafts, income from animal husbandry and wages. 21 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table (2.4) presents evidence of th e im pact of labor incom e in reducing variability in to ta l incom e. Including wages reduces volatility (as m easured by th e coefficient of variation6) by m ore th a n 150 for th e landless in th e pooled data; th e im pact of wages in reducing v o latility of incom e is also quite high for th e sm all farm ers. R eduction in volatility is m easured as follows R eduction in V olatility = CV (Non W age Incom e) - C V (Total Income) where C V (Incom e) = S.D (Incqm e) 0 M ean (Incom e) C V (N on W age Incom e) = ^ , 100 M ean(N on W age Income) 2.3 Optimal Allocations in a Decentralized Economy T his section derives th e o p tim al allocations of consum ption and leisure, w ith and w ithout p articip atio n constraints in the village financial m arkets. The population is assum ed to consist of two broad categories. 7 T h e first set is constrained by th eir lack of access to th e existing financial maxkets. T h e second set is unconstrained, in th e sense th a t th ey have access to a com plete set of A rrow -D ebreu securities. 6 Henceforth CV. 7Once again ex-ante the individuals are not forced to fall into one category or the other. 99 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Individuals optim ally choose consum ption allocations and leisure to solve the following problem , M ax t i p s8*U(^ifa? ^itsj ® its) (2 1 ) (=0j=1 subject to th eir budget constraint. c,-tJ is th e per-adult equivalent consum ption. T he m ethodology of p er-ad u lt equivalence is discussed later. /,-* * is th e num ber of days of leisure (see section 2.4 for a form al definition). ps is th e probability of state s = 1 ,..., S , and 8 is th e discount rate, assum ed to b e th e sam e for all individuals. Oits incorporates factors th a t change tastes. If individuals are constrained in th eir ability to borrow, the relevant budget constraint is Cits = yits V i,s (2.2) where H its ~ ~ (w + ^is)(T lits ^t’ fs) “ I" ‘ ‘Tits (2-3) yit3 is per-adult equivalent household incom e, T is th e to tal tim e in a year, available to the individual (T = 366 in our em pirical analysis). is an illness shock th at affects individual labor supply by reducing the to tal tim e available for work, ir,-4 < is a m easure of non-labor incom e. N ote wages are defined to fluctuate random ly around a given m ean, th e fluctuation depending on rainfall surprise (eta)- 23 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. If on th e o th er h and individuals are unconstrained and have unrestricted access to a perfect set of A rrow D ebreu securities, th e relevant budget constraint is (2-4) t=Oa=l where t/,ta is defined in equation (2.3) above. Pts is the date 0 A rrow -D ebreu price of consum ption deliverable in year t, contingent on sta te s occurring . 8 Individuals can therefore be thought of as trading tim e-state contingent consum ption claim s at date 0 . Finally the specific form of the utility function is assum ed to be = (-ps^- + i ^ l ) e x p ( « iu) (2.5) 1 —7 1 — a Therefore there axe th ree kinds of shocks: aggregate village level shocks m anifested by variations in rainfall (ets), illness shocks specific to individuals (utt3) and prefer ence shocks (0 ft,). T he set of (form al a n d /o r informal) in stitu tio n s th a t m im ic this perfect m arket m ay be confined to th e village or m ay involve th e whole country. All one is concerned w ith, is the existence of such institutions. An equilibrium is defined by a set of consum ption and leisure choices c"t3 an d /“ t3, such th a t 8See Altug & Miller (1990). 24 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (i) T he constrained m axim ize (2.1) subject to (2.2) (ii) T h e unconstrained m axim ize (2.1) subject to (2.4) (iii) M axkets clear, V(t, s) T he first o rd er condition for th e optim ization problem for the constrained in dividuals can be w ritten as (assum ing th e specific functional form of th e u tility function in eq u atio n (2.5)): p s t f c ^ exp(9 - i t s ) = A'fs P s^lTts e x p {Oits) > A (-fa[(u; + efs)] (= if lita < T ) Here \'its is th e Lagrangian m ultiplier. Define A T a l c i n g logarithm s, first difference w ith respect to tim e and deviations from th e village level average equation (2 .6 ) gives (disregarding th e notation for the state) A logic*) = A c;a + - i A 0 it - A 0 f) - - ( A log(A*) - AA“) Vs (2.8) 7 7 where c£ * “ = } E i= i log(c*t); dt = 7 £ f= i K = / T.i=i log(^«t) 25 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (2.6 ) (2.7) Now consider equation (2.7) and assum ing equality (th e estim atio n methodology will correct for corner solution), taking logarithm s, first difference w ith respect to tim e and deviation from th e village level average gives A log(/*) = A i r + - ( AOit - AQa t ) - - ( A log(Ait) - AA“) Vs (2.9) a a w here l 'a = ~ JZLi log(/,t) E quations (2.8) and (2.9) give th e optim al allocations for constrained individuals. I now tu rn to th e case of th e unconstrained individuals. T h e first order condition for the optim ization problem for th e unconstrained individuals can be w ritten as (once again assum ing th e specific functional form of th e u tility function in equation (2.5)): cTtl exp(dits) = XisPts (2.10) exp{8its) > XisPts(w + ct5) (2.11) (= if lit3 < T) As in the constrained case equation (2.10) gives A log(cft) = A c“a + i ( A Bn - A Of) Vs (2.12) 7 26 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Assuming th a t equation (2.11) holds with equality Alog(Zft) = Al'ta + - ( A 0 ft - A et) Vs (2.13) a E quations (2.12) and (2.13) give the o p tim al allocations for unconstrained indi viduals. T hese two equations axe nested in equations (2.8) and (2.9). T h e last two have an ad d itio n al term 7 7, - t = Alog(A,t) — AA“ . T he o p tim al allocations derived in this section lead to testable im plications of consum ption insurance, with and w ithout th e assum ption of perfect m arkets. O pti m al allocations in th e constrained equilibrium can be represented by th e following set of equations Alog(cii) = f30 + (3xA c r + (32r]it + ec it (2.14) A log (/,■ * ) = 70 + 71 A i r + J2rjit + e\t where eft an d e[t incorporate both preference shocks (A0,-t — A0“) and m easurem ent error. O n th e o th er hand, individuals who have u nrestricted access to credit m arkets, choose consum ption and leisure to satisfy A log(c,t) = (30 + (3xAd;a + ec it (2.15) Alog(/,-t) = 70 + 7 iA i r + eit 27 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. N otice th a t th e allocations w hen individuals have u nrestricted access to a set of A rrow -D ebreu securities is identical to th e Pareto O ptim al A llocations in this econ om y (see A ppendix 1). A te st for Perfect maxkets is to estim ate equation (2.14) an d test Hq : (3 2 = 0 and 7 2 = 0 . 2.4 Econometric Specification To reiterate, estim ate A log(c,f) = (3 q + /?iAc*a + ^rjit + A log (/,•*) = 7 0 + 7 t A l'ta + 72 7 7 i t + At and test H0 : /?2 = 0 and 7 2 = 0. All the variables w ith th e exception 77, - £ are observable. T he first problem is to estim ate 77#. E stim ation follows th e strategy of Altonji & Siow (1987) and Jacoby & c Skoufias (19946). Here rjit is a m easure of th e revision of the logarithm o f th e m arginal utility of incom e (net of the village level average). 77,- f is related to th e unan ticip ated (and hence ex-ante un-insurable) com ponent of change in incom e. If individuals have access to a full set of state contingent transfers, consum ption and leisure should be unaffected by such u n anticipated shocks to income. 77,t is not observable. 28 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. rjit is m easured by a com ponent of change in non-wage incom e th a t is (at least p artly ) u nanticipated by th e consum er. Define 7 rt “ as this (unanticipated) com ponent o f change in to tal non labor incom e. Also, define t jh to be a scalar transform ation of 7rt “ as Tfit = &o tt,“ + R ew riting (2.14) as A log(c,-t) = Po + PiAc™ -1 - P2 (boT r?t + e^) + ec it A log(/tt) = 70 + 7i a + 72 (& ott,“ + 4 ) + e‘ it which gives A log(c,-t) = p0 + Pi Ac™ + Pztt?£ + wc it (2.16) A log(l,'t) = 7 0 -I- 7 i A/*a + 7 3 7t“ £ + w[t where wft = Pie\t + eft and w\t = 7 2 e^ + e\t. However, even 7r,“ is unobservable and I have to choose a proxy for tt“ £ which is uncorrelated w ith wft and w\t. To obtain an in stru m en t for 7rt “ first decom pose change in non-wage income (A log(7r,t)) in period t into 1 . a com ponent th at is predictable by th e consum er given his inform ation set at th e beginning of period t ; 2. a com ponent th at is unpredictable to th e consum er given his inform ation set a t th e beginning of period t; 29 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3. a com posite erro r term . It is assum ed th a t while rainfall surprise is an aggregate shock, it affects dif ferent households differently, depending on th e endow m ent characteristics o f the households. So w rite A lo g ( 7rti) = ho + h iX i't-i + ^ 2 ( ^ 1,f-i * DRt) + V {t (2.17) where is a vector of household endow m ent characteristics (lagged one period) so th a t inform ation on X ^ t- i is available to households at the beginning of period t. D Rt is th e rainfall deviation in period t an d u,t is a random com ponent consisting of unobserved factors affecting income changes an d m easurem ent error. Define th e an ticip ated com ponent of th e incom e change as the projection of the actual change in incom e on inform ation available to households available a t the beginning o f period £, so th a t 7rt n = h iX i't - 1 and the u n an ticip ated com ponent as n't = h2(X i t t - 1 * D R t ) 30 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. since (X,-,t_i * DRt) is unknow n to the household at the beginning of period t. tt“ is used as th e instrum ent for ir“. B y definition 7r“ is uncorrelated w ith both wc it and Wi f Consider therefore, th e following set of equations: A log(cit) = Pq + /?iAc*a + /?3tt£ -f wft A log(fit) = To + 7 iA /t * 3 + 737 r ,“ + w l it (2.18) Perfect Maxkets require / ? 3 = 0 and 7 3 = 0. Since not all individuals can vary their participation in th e labor m arket, correc tion for com er solutions in the first order conditions is required. log(/it) = log( 1 ^ ) + A i;a + - { A 6it - A0?} - - { A log(A,-t ) - AA“} a a = log(l7t) (say) So r lo g (r) if iog (/r*) > io g (r) log(/ft) = log(/rt) otherwise 0 if log(/“ t) > log(T) A log(/,t) = A /*“ + ^{A 0,t — A 0“} — A{Alog(Att) — AA“} otherwise 9For more on this kind o f decomposition see Jacoby & Skoufias (19946). 31 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. So sample separation is as follows. Define / “ = 6\ A/*“ + O ^ T Jit + 63 l°S(^*i— 1/-^) + eii and { 0 if /* > 0 . 1 otherw ise So f A log(cft) = 0o + ftA cj'" + 037r“ t + if / = 1; ^ (2.19) [ A log(f.t) = 7 0 + 7 iA /t ' a + 7 3 7ift + u;{t and if / = 0; A log(cft) = 0 o + 0iAc'ta + 0 3*?t + w c it (2.20) This is a sw itching regression framework, and I estim ate 01,03,71,13 using Heck m an ’s M ultistage Process, as discussed in M addala (1983). E stim ation follows in two stages. In stage 1 I define 7rt “ and o b tain predictions of 7r?t from equation (2.17). In stage 2, using 7 rft, obtained from stage 1 (above), as an explanatory variable, run the following regressions, correcting for th e possibility of a com er solution in leisure, as described in equations (2.19) and (2.20). A log (c,-t) = 0o + 0i A c p + 037 T ,“ + Wjt A log(Zit) = 7 0 + 7 iA /’“ + 7 3 ££ + w l it 32 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. and test /3 3 = 0 , 7 3 = 0. T h a t concludes our discussion on th e estim ation technique. I now tu rn to the results of estim ating th e m odel using survey d a ta from th e ICRISA T villages. 2.5 Estimation Results T he m odel is estim ated using d a ta from the IC R ISA T villages. I include only adult m ales and fem ales in th e age group 15-60. As defined above, consists of inform ation known to th e individual at th e beginning of period £; so I use Livestock, Im plem ents, A rea of Land U nder (Khaxif) cultivation (Land Area) and G rain Stock at th e end of period t — 1. 7 r,“ is defined as h, 2 {X it-i * D R t). DRf is m easured by th e deviation of rainfall from th e 16 year pooled average in Q uarters 2 and 3 of the calender y ear . 10 Leisure (/,-£ ) is defined as follows la = T — (m arket labor -f- own farm w ork), (2.21) where T = 366 is the m axim um tim e available to any individual in a given year. M ar ket labor includes days p articipating in th e daily wage labor m arket (daily wage farm labor, governm ent em ploym ent, including EGS, an d non farm private em ploym ent). 10 An adverse rainfall in Quarter 2 affects agricultural production in the pre-harvest stage and adverse rainfall in Quarter 3 affects production in the harvest stage. 33 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Individuals who axe perm anent farm servants or contractual labor are registered as non-participants in th e daily labor m arket. If own farm w ork is included as a com ponent of leisure (lit), any variation in leisure now reflects a change in p articip atio n in th e daily wage labor m arket. An increase in labor m arket p articipation w ithout a corresponding decline in leisure im plies a su b stitu tio n of own farm work by increased m arket work. Using th e change in lab o r m arket particip atio n as th e dependent variable provides an im p o rtan t insight as to how the villagers behave. T his will elab o rated later. Using th e sw itching regression fram ew ork I o b tain estim ates of th e Inverse M ill’s R atio which are significant. 11 Hence a sim ple OLS would have led to significant selection bias. L eisu re E ffects T able (2.5) presents the effects of unan ticip ated incom e shocks on change in leisure (A log(Z,t)), i.e., th e estim ate of 7 3 , (7 3 ) from equation (2.18), corresponding to / = 1 in th e previous section (equation (2.19)). Colum n (2) uses the tru e definition of leisure, i.e., own farm work is not a com p onent of leisure, and in Colum n (3) own farm work is included as a com ponent of leisure. T he coefficients in Colum n (3) th en reflect the effect of unanticipated incom e shocks on labor m arket participation. 11 These figures are not presented. 34 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 2.5: Im pact of Unanticipated Income Shock on A log(/tt) : 7 3 Land Class ( 1 ) Effect on Leisure* (2 ) Effect on Non M arket A ctivity* (3) Landless 0.09 0.09 (0.19) (0.19) Small Farmers 0.19 0.40* (0.16) (0.16) M edium Farm ers 0 . 0 2 -0.05 (0.05) (0.05) Large Farmers 0.05 0.04 (0.42) (0.50) *: Leisure defined in e q u a tio n (2.21) + : Non M arket A ctiv ity = L eisu re (from equation (2.21)) -j- O w n farm work F igures in P arenthesis in d ic a te S ta n d a rd E rrors *: t-ratio significant a t 95% level o f significance D a ta Source: IC R ISA T VLS T ap es T he results in C olum n (2) show th at individuals (irrespective of land class) do not vary their leisure in response to unanticipated incom e shocks. R esults in Column (3) show th at only th e sm all farm ers adjust their m arket participation in response to unanticipated incom e shocks. The coefficient 7 3 is positive and significant for th e sm all farmers. T his im plies th at, when they face an adverse income shock, they reduce their own farm work and increase participation in the daily wage labor m arket. The m edium and large farm ers do not need to use labor as insurance, since they have access to credit, which enables th em to insure against unanticipated shocks. The landless do not vary their m arket labor supply in response to such shocks. For these households, labor income com m ands a m uch larger share of total incom e than for any of th e o th er land classes (see (2.3)). Even in a normal year, 35 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. th e landless have to work extensively in th e labor m arket for subsistence reasons, and so when faced w ith an adverse shock, th ey do not have any m argin to increase participation in th e daily wage labor m arket to com pensate against such shocks. C o n su m p tio n E ffects We m easure consum ption in real (1983=100) per-adult equivalent term s. To obtain adult equivalent fam ily sizes, following Tow nsend (1994) th e following weights are used: 1.0 for ad u lt males; 0.9 for ad u lt females; 0.94 and 0.83 respectively for males and females aged 13-18; 0.67 for children aged 7-12, regardless of gender; 0.52 for toddlers aged 4-6; and 0.05 for infants. T able 2.6: Im pact of U nanticipated Incom e Shock on Alog(c,-f) : /? 3 Land class Labor M arket P articipation Yes No ( 1 ) (2 ) (3) Landless 0 .2 0 * 0.15* (0.09) (0.06) Sm all Farm ers 0 . 1 0 0.43“ (0.06) (0 . 1 2 ) M edium Farm ers 0.06 -0.04 (0.04) (0.06) Large Farm ers 0.03 0.06 (0 .2 1 ) (0.30) F igures in P aren th esis in d ic a te S ta n d a rd E rrors *: t-ra tio significant a t 95% level o f significance D a ta Source: IC R ISA T V LS T apes 36 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table (2.6) presents th e effects of u n an ticip ated income shock on consum ption (Alog(c,-t)). C olum n (2) presents th e estim ates of (3z {(h) from equation (2.18), corresponding to I = 1 (equation (2.19)). C olum n (3) presents th e estim ates of 03 from equation. (2.18) corresponding to I = 0 (equation (2.20)). C onsum ption of the m edium and large farm ers is unaffected by u n an ticip ated income changes, im plying th a t the Null H ypothesis of Perfect M arkets can n o t be rejected for th e m edium and laxge farm ers. Sm all farm ers who do not have access to state contingent transfers, can insure consum ption from unanticipated incom e changes by varying th e ir labor m arket paxticipation. T he Null H ypothesis of (3$ = 0 cannot be rejected in Colum n (2), but is rejected /?3 = 0 in Colum n (3), im plying th a t individuals who vary their m axket paxticipation axe able to insure consum ption against such shocks. T he rest of the sm all farm ers are left vulnerable. T his again is evidence of th e fact th at sm all farm ers use labor maxket paxticipation as an insurance m echanism . Small faxmers then sm ooth consum ption not at the cost of volatile leisure, b u t by m aking com pensating changes in maxket paxticipation. Lastly, the landless are affected by u nanticipated incom e shocks, since neither do th ey have access to sta te contingent transfers to insure against such shocks, nor axe they able to m ake com pensating changes in m axket participation in response to such shocks. Table (2.7) sum m arizes the m ain results of th e paper. Individuals belong to one of the tw o categories: Constrained or U nconstrained. T he landless and the sm all farm ers belong to the first category. T h e sm all farm ers insure consum ption 37 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 2.7: Sum m ary of Results Land Class Leisure M arket P articipation C onsum ption Constrained Landless N ot Affected Not Affected Affected (+ ) Sm all Farmers N ot Affected Affected (+ ) Not Affected* Unconstrained M edium Farm ers N ot Affected Not Affected Not Affected Large Farmers N ot Affected Not Affected Not Affected F ig u res in brackets gives d ire c tio n o f m ovem ent *: O n ly for those who m ak e c o m p e n sa tin g ch an g es to lab o r m ark et p a rtic ip a tio n by m aking com pensating changes in labor m arket participation. The direction of m ovem ent is positive, im plying th a t if they face an adverse shock they work less on th e ir own farm and work m ore in the daily wage labor m arket. Leisure is unaffected. T h e m edium and large farm ers belong to the second category. They have access to credit and can borrow, and hence insure consum ption w ithout needing to make com pensating changes in labor m arket participation. 2.6 Conclusion T his ch ap ter has developed a fram ew ork to exam ine how villagers in the ICRISAT villages sm ooth consum ption to th e ex ten t th a t they do. T he m echanism s used vary across individuals. It is incorrect to assum e th a t all households use credit to sm ooth consum ption, because, as th e d a ta shows, the credit m arkets are biased in favor 3S Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. of m edium and large farm ers. These households have access to sta te contingent transfers (credit from form al and informal sources) which enables th em to insulate them selves from th e effects of incom e shocks. Sm all farm ers a re constrained in the financial m arkets. However, some of them vary th eir m arket participation - sub stitu tin g th eir own farm work w ith increased labor m arket p artic ip atio n to smooth incom e and hence consum ption. Small farm ers then sm ooth consum ption not at the cost of volatile leisure, but by m aking com pensating changes in lab o r m arket partic ipation. T he landless do not have th e m argin to vary labor m ark et participation in response to unanticipated shocks to income, and are left vulnerable to such shocks. 2.7 Appendix: Pareto Optimality with Labor-Leisure Choice T he te st for Full C onsum ption Insurance is a test of the validity o f P areto O ptim ality for th e economy under consideration. The m ajor im plication o f risk sharing is th at individual consum ption varies positively w ith aggregate consum ption rather than w ith idiosyncratic variables such as individual income. W e w rite the social planner’s problem as / co S Max 5^ /j.is 5^ P ^ U (c,ia, lit3; 9ita) 1 = 1 «=0 3=1 39 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. subject to Y , c u , = F ( H „ ,e „ ) V(i,a) 1= 1 T h e u tility function is separable in consum ption and leisure (see equation (2.5)). Consider th e aggregate production function to be F ( H ,„ C .) = [ A { £ ( r - l » , - (2.22) i=l Aggregate production depends on aggregate labor supply, given a certain level of capital stock A. T — /,-ta — u,-ta is th e labor supply of individual i in year t when sta te is s. A ggregating over all villagers gives th e aggregate village labor supply at tim e f. 5 < 1. Form the Lagrangian L = f; H i, f; »«,) + A,»{[A{X:(r - li„ - Ui„)}< , e,.}] - £ C i , . } ] t=0 t = l 3=1 t= l t= l T he F irst order conditions: H i,U ,(c it„ h „ \ S i „ ) = \ t, (2.23) H M d u , k „ ; S o .) > A „ { [ A { £ ( r - l,„ - U i„)}"-1£,.]} (2.24) i=l = if la, < T 40 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Consider equation (2.23). Taking logarithm s, deviation from th e village level average and th e first difference over tim e give (disregarding th e notation for the state) A log(c«) = A c f + -(A0it- A 6 f ) Vs (2.25) 7 w here cf = j E L i log(c,-t ) and Of = j E i= i 0it Consider equation (2.24) w ith equality. I correct for th e possibility of corner solution in our estim ation technique (Section 4). Proceeding ex actly as above I get A log(/ft) = Alf + - ( A 0it - A Of) Vs (2.26) o c where If = \ J2L i log(/,-«) Equations (2.25) and (2.26) define the P areto O ptim al A llocations of th e economy under consideration. N otice th a t th e Paxeto O ptim al A llocations are identical to th e allocations a ttain ed when individuals face perfect m arkets (see Section 3). 41 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Chapter 3 A Quantitative Analysis of Employment Guarantee Programs with an Application to Rural India 3.1 Introduction In ch ap ter 2 of m y dissertation I have shown th a t som e households in th e ICRISA T villages use participation in th e labor m arket as a m eans of consum ption insurance. W hat is m ore im portant is th a t th e households th a t do so are restricted in th eir access to village financial m arkets. This has im p o rtan t im plications in the design of in stitu tio n s th a t will enable households to sm ooth consum ption against aggregate village level shocks. T he prevalent m echanism to deal w ith such risk is through conservative production choices and through storage. Such a technique is costly in term s of lower m ean output. I argue th a t an o ptim ally designed work-fare program , like th e Em ploym ent G uarantee Schem e (EGS) in th e Indian state of M aharashtra 42 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. can potentially insure households against aggregate village level risks. In this chapter I exam ine th e welfare effects of such a work fare program , in an econom y where agents face exogenous income shocks an d are unable to insure them selves through private credit m arkets. T he Em ploym ent G uarantee Scheme (EGS) is one of the biggest work-fare pro gram s in developing countries around the world (As Ravallion, D a tt Chaudhuri (1993) point out, in a typical year it provides about 100 m illion person-days of em ploym ent, at an average of US$1 per day - the sta te ’s aggregate rural work force including cultivators was about 20 m illion persons in the m id 1980’s). T he prim ary aim of the EGS is to provide em ploym ent to the rural unskilled population on de m and - as em bodied in its slogan Magel tyala kam (W hoever desires work will get it). T h e participants are provided em ploym ent in a variety of altern ativ e occupations - from digging, breaking rocks, shifting earth and transplanting, to paving roads and building dams for irrigation purposes. However the essential idea of th e EGS is not to build physical infrastructure, b u t to reduce poverty and provide a safety-net against incom e shocks for participating workers. There exists a su b stan tial literature docum enting the perform ance of this program in reducing absolute poverty in rural M aharashtra (Bhende et al (1992), Ravallion (1991), Ravallion, D a tt & : Chaudhuri (1993), G aiha (1996)). T he insurance aspect of the EGS, however, has not been studied. 43 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. T his chapter studies the role of th e EG S as an insurance m echanism by conduct ing a com putational experim ent in a dynam ic general equilibrium m odel. Specifi cally, I exam ine th e following questions: 1. W h at is the th e wage ra te th a t m axim izes average steady sta te utility? 2. W h at axe th e welfare effects of n o t paying a wage equal to th e optim al wage ra te (as defined in ( 1 ), above)? T he econom y is assum ed to consist of a continuum of villages. Individuals in each village axe identical and axe subject to village wide shocks, m anifested by variations in rainfall. They do not have access to private insurance m arkets, axe unable to borrow, and m ust hold their savings in th e form of a non-interest bearing asset. Such an economy is representative o f th e ICRISA T villages. W alker & R yan (1990) in their study of th e ICRISAT villages find evidence th a t the professional village m oney lenders who perform m ost of th e credit m arket operations (given th a t com m ercial banks axe of relatively little im portance), do not accept tim e deposits from households and do not trad e outside th e p articular village (or set of villages) in which each of them operates. In th e absence of financial instrum ents w ith which to self-insure, households use non-financial assets like cash and jew elery (which axe also non-productive) or drought anim als as a store of value. In addition, th ere is also evidence th a t households in these villages use labor m arket participation as a means of insuring against unanticipated shocks to incom e (see chapter 2 of th is dissertation and K ochar (19956)). In other words, households can sm ooth consum ption across 44 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. states by varying th eir participation, in th e lab o r m arket, in particular by w orking in the EGS. T h is p ap er num erically exam ines how participation in the EGS can help households increase th eir steady state u tility by insuring against aggregate village level shocks. T he first step in th e analysis is to calculate th e optim al wage rate in th e EG S. T he governm ent can eith er pay participants in th e EG S a wage equal to th eir m arginal product, or th e y can choose to pay a wage g re ater th a n th e m arginal product of labor in the EGS (should insurance requirem ents w arran t it). Com putations in this p ap er show th a t th e o p tim al wage generally exceeds th e m arginal product of labor in the EGS. T his p ay m en t in excess of the m arginal p roduct is a pure transfer (subsidy), financed by taxing th e income of the rest of th e population. W hen agents are paid the optim al wage rate in the EGS, they do n o t need to hold any non-interest bearing asset for p recau tio n ary savings. All insurance is provided by the EGS. A fter com puting th e optim al wage rate in th e EGS, I exam ine th e welfare im plications of th e program . To do this I ask th e following question: suppose th e governm ent chooses to pay the p articipants any non-optim al wage in th e EG S (in cluding a wage equal to the m arginal p ro d u ct of labor); what additional lum p-sum transfer w ould provide th e agent w ith th e sam e level of utility as in an econom y where they earn th e optim al wage in th e EG S? T his lum p-sum transfer expressed as a percentage of G D P gives a m easure of th e foregone welfare gains of this program . 45 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. T h e o p tim al wage ra te in the EGS and th e p o ten tial welfare gains of th e program depend on how productive labor is in th e EG S, relative to the private sector. Given th e fact th a t individuals employed in the EG S prim arily produce of public goods (roads, dam s etc.) w hen working for th e governm ent it is im possible to obtain th e value of th is o u tp u t. I will, therefore, consider 5 exam ple economies, varying th e p ro d u ctiv ity of th e EGS from 0.00 to 0.25, norm alizing th e productivity in th e private sector to b e one. W hile it is unreasonable to assum e th a t th e EGS is unproductive, it is also im probable th a t this productivity exceeds 25% of the pro d u ctiv ity of the private secto r . 1 For each of the 5 economies, I calculate the optim al wage rate in the EG S, and o b tain a m easure of the welfare loss (equivalently, th e foregone welfare gains) of n o t paying this optim al wage. T he o p tim al wage rate in the EGS and th e welfare gains of th e program depend on th e p aram eter values and on the productivity of the EGS. In th e event the governm ent pays participants in the EGS a wage equal to the m arginal product of labor, th e foregone welfare gains of th e program range from 0.01% of G D P to 2.63% of G D P, depending on th e productivity of th e EGS. T he actual wage paid in the EGS is o b tain ed to be less th an the optim al wage rate. I exam ine th e welfare effects of an across th e board increase in wages in th e EGS and conclude th a t such a wage 1The actual wage received by the participants in the EGS is obtained to be 25% of private sector income (This is discussed later in greater detail). This actual wage is equal to the marginal product of labor in the EGS plus a pure subsidy. A labor productivity exceeding 25% would mean that the participants in the EGS are not even paid a wage equal to their marginal product, and it would imply a negative subsidy or an negative income tax on the rest of the population. 46 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. increase, increases welfare only for th e m ost productive EGS. In the other cases, an even g reater increase in wages is required to reduce th e w elfare costs. T h e rest of this chapter is organized as follows. In Section 3.2 I describe the stru ctu re of the economy and define th e stationary com petitive equilibrium for this economy. In Section 3.3 I discuss issues on com puting th e equilibrium - the m odel is calibrated to d ata from th e ICRISA T villages. Section 3.4 discusses the results - welfare implications and policy experim ents. Section 5 presents the concluding argum ents. T he com putation algorithm and sensitivity analyses are presented in A ppendix 1 and 2 respectively. 3.2 Structure of the Economy T he m odel used in this paper is sim ilar to the one used by H ansen & c Imrohoroglu (1992). I consider an econom y populated by ex-ante identical, infinitely lived in dividuals, w ith utility defined over consum ption and leisure. Following Lim (1991) and Townsend (1994), it is assum ed th a t household specific shocks are absorbed by village level institutions, and th is enables me to consider each village to be popu lated by a single representative agent. T he economy consists of a continuum of such ex-ante identical individuals (villages). Agents face a good or a bad state according to a known stochastic process. Labor is indivisible, so th a t if th e agent chooses to work in th e EGS, he has to w ork for an exogenously determ in ed num ber of hours. M oreover, I assume th at in a good sta te the individual has to work on his own farm 47 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. for a pre-determ ined num ber of hours. T h e only source of uncertainty is th e sta te faced by the agent (w hich is also th e state experienced by the village) - th e econom y as a whole faces no aggregate shock. Let th e utility function be U(ct,lt) = _or)1_nr) / ( l — 7 ), where a is th e share of consum ption in th e u tility function and 7 is th e coefficient of risk aversion. Each individual m axim izes OO E Y , P U M ) , (3.1) t=0 subject to th e following constraints at+i = &t + V t — c t i (3-2) at+1 > 0 Vi (3.3) w here 0 < (3 < 1 is th e subjective discount factor, cf is consum ption in period t and It is leisure in period t. Individuals are unable to borrow, and have no access to private insurance m arkets. They axe, however, able to accum ulate non interest bearing assets. Let at be an agent’s asset holding a t th e beginning of period f, his asset holding a t th e beginning of period 14 - 1 is given by at+i, yf is disposable incom e in period t. Agents axe endow ed w ith one unit of tim e in each period which they can allocate to work or leisure. In a good state, the agent has to work on his own faxm. In a bad state he cannot work on his own faxm. T he EG S is an alternative em ploym ent 4S Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. o p p o rtu n ity and the agent can. always work there. In a good state, therefore, in addition to working on his own faxm th e agent m ay choose to work in th e EGS. In a good s ta te th e agent has to w ork 0 < h° < 1 hours on his own farm and, if he chooses to work on the EGS h e has to work 0 < hi < 1 hours, or not a t all. T he allocation of tim e therefore satisfies th e following condition h\ hf - + ■ It = 1 Every period, each village (an d therefore each individual) faces an village wide shock, m anifested by variations in rainfall. It therefore faces either a good rainfall year or a bad rainfall year. T h e econom y as a whole faces no aggregate shocks. T he s ta te for th e village follows a tw o -state M arkov process, w ith the tran sitio n function for th e production states given by th e following m atrix x = [ x d ; * \i « { ^ 6 } w here for exam ple Prob{st+i = < 7|s* = 6 } = Xbg is th e probability of being in the good sta te in t + 1 conditioned on being in the bad sta te in t. In a good yeax an individual works in his own faxm and produces y units of o u tp u t every tim e he works, w here y is constant over tim e. So to tal production in th e private sector (own faxm) \s Y0 = y N 0, where N 0 is th e p roportion of villages (or individuals) who face a good rainfall yeax. Therefore, y is th e m arginal product of labor in the private sector. In 49 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. ad d itio n he can work in the EGS. In a bad yeax, he cannot work on his own farm. He can eith er work in the EGS, or not work at all. T h e prod u ctio n function in th e EG S is given by the following linear form Yg = A N g, w here Yg is to tal production in th e EGS an d Ng is the to tal nu m b er of people em ployed in th e EGS. Now when individuals work in the EGS they o b tain a wage w w hich m ig h t be different from th e m arginal product of labor in th e EG S. W rite w = A + yg, w here A is the m arginal pro d u ct of labor in the EGS, an d yg is a pure subsidy th a t is paid to the workers (wage exceeding the m arginal p ro d u ct of labor). yg is assum ed to be non-negative. For th e sake o f notational sim plicity I will assum e A = By, w here 9 e [0,1]. T he subsidy (yg) is financed by taxing incom e a t th e rate r , so th a t th e governm ent budget constraint is satisfied with equality. T h e governm ent budget co n strain t will be specified later. T he sequence of events is as follows: 1 . N a tu re reveals its state (s), i.e., s = g, b. h° if s = g 0 if s = b N otice th a t the agent is not given th e choice of not working on his ow n faxm in a good state. He is, however, given th e choice of working in th e EG S in addition to w orking on his own faxm in th e good state. This is done because it has been observed in th e ICRISAT villages th a t ad u lt males in households owning land 50 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. always work on th e ir own farm in a good rainfall year, supplem enting th eir crop income by w orking as wage labor, eith er in other farm s or in th e EGS. Moreover, a g reater num ber of people p articip ate in th e EGS in a bad rainfall year. In A ppendix 2 (section 3.7.2) I check for th e robustness of the results if this assum ption is relaxed by conducting two experim ents. In the first one the agent is allowed to choose between working in the EGS and on his own farm in a good state. In th e second one th e individual is not allowed to work in the EGS in a good sta te , b u t he is allowed to choose not to work on own farm in a good state. T hese two experim ents show th a t th e results are not significantly affected by this assum ption. 2. Given s, and th e ir current asset holdings, agents choose w hether to work in the EGS or not (d), i.e., choose d = e,u. h9 f = k9 if d = e 0 if d = u 3. After the agent chooses d, he chooses asset holding (a7 ) for th e next period and current consum ption (c) subject to the asset accum ulation equation and the no borrowing constraint 51 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The household’s disposable income can be w ritten as (y + w)( 1 - r ) if i = (g,e) y( 1 - r ) if * = (g,u) (3.4) iu (l — r ) if t = (6 , e) 0 if z = (6 , u) In sum m ary, given the realization of th e sta te (s = g,b), agents choose w hether or not to work in th e EGS, (d = e, u) and also how m uch asset to hold, a'e f£+ , such th a t th e y m axim ize d = e ,u ; s = g,b t — 0 subject to a' = a + yd — c. Note a, c, yd, a' all depend on the choice of d, given s. Note, for notational simplicity, I have w ritten U(d\s) = U((c,l)\(i = s ,d )); s = g,b;d= e,u Even though agents in different villages axe ex-ante identical, ex-post they are characterized by th e state of their village s = g,b and their asset holding a. The distribution A(a, s) represents th e fraction of agents in each sta te at a given point of tim e, and also specifies the fraction of tim e a particular agent is in a given state over his entire lifetim e. 52 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. We can characterize th e solution to th e above problem using th e B ellm an ’s equa tion: Max<£ U ^ s ) = M axa' < Max^ U(e\g) + /? E a'X ( s ,s ') V ( ( a ',s ') |i = g,e): U{u\g) + ^ E a 'X O ^ - s 'M ^ O I * = 9,u) U{e\b) + /? E * x (* ,* ')V r((« ',* ,)l*'= *,«); U{u\b) + / t f E y x O b O ^ ^ s O K = hi u ) (3.5) subject to a' > 0 . Using th e budget constraint (equation (3.2)), th e u tility functions in (3.5) axe U(e\g) = U{a + {y + u>)(l — r ) — a', 1 — h° — hg) U(u\g) = U{a + y(l — r ) — a', 1 — h°) U(e\b) = U (a + w (l — t) — a ', 1 — ha) U(u\b) = U(a — a', 1) D efin ition : S tation ary C o m p etitiv e E quilib rium A stationary competitive equilibrium for this econom y consists of a set o f tim e in variant individual decision rules d(a, s), c(a, s),a'(a, s ) for em ploym ent, consum ption and asset accum ulation respectively, a statio n ary distribution A (a, s) of agents, a tax rate r , an d a param eter A such th a t 53 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1. G iven r , th e household decision rules solve equation (3.5), above. 2. T h e goods m arket clears, i.e., ]C 5)c(°> s ) = '52 M a, ff)no(a, g)y + ^ s )ng(a > S)A a 3 a a s n0(a ,g ) is th e fraction o f th e population w ith asset holding a in a good state, n ff( a ,s ) is the fraction of population in sta te s(s = < 7, 6 ), w ith asset holding a, who p articipate in the EGS. 3. T h e governm ent budget constraint is satisfied, i.e., T '51{Kai 9)(y + wdo) + A (a,6 )(u;do)} = y5 5 ^ A (a ,s )n ff( a ,s ) a s w here do = 1 if d = e and do = 0 if d = u. 4. T h e invariant distribution o f agents A(a, s) solves th e following functional equa tio n A ( a ', s ') = J 3 x ( 5 ,s')A (a ,s) 3 a c Cl(a',s') C ondition ( 1 ) is the household’s optim ization problem, i.e., given r , a, s, households choose w hether or not to work in th e EGS (d), current period consum ption (c) and asset holding for the next period (a1 ). C ondition (2) is the goods m arket clearing condition. T h e left hand side of (2) is th e aggregate consum ption in th e economy and th e right hand side the to tal production (own farm production plus production 54 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. in th e EG S). (3) is th e governm ent budget constraint which m ust be satisfied with equality. Hence to tal subsidy paym ent to participants (i.e., paym ent in the EGS in excess of the m arginal p ro d u ct) m ust equal taxes collected by th e governm ent. G iven th a t only statio n ary equilibria axe considered, condition (4) ensures th a t the distribution of agents across states is tim e invariant. T he labor supply in own farm production and th e EG S axe given by Ng = J 2 ^ 2 \(a ,s)n g(a,s) a s N0 = 5 3 A (a,g)n0(a,g) a w here Ng is the to tal num ber of agents employed in th e EGS and N0 is the total num ber of agents working on own faxm. Rem em ber th a t an agent who faces a good sta te m ust work on his own faxm. He cannot work on his own faxm if he faces a bad state. Every agent can choose to work in th e EGS, irrespective of the state he faces. Labor dem and in either sector is infinitely elastic - every agent facing a good state can work on his own farm (for a pre-determ ined num ber of hours) and everyone who w ants to, can work in the EGS (again for a pre-determ ined num ber of hours). 55 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3.3 Calibrating the Model T he m odel is calib rated using survey d a ta from th e ICRISAT villages. T he ICRISA T region consists of a high risk, prim arily ag rarian econom y spread over a vast geo graphical area in India, in the states of A n d h ra P radesh, M aharashtra, G u jarat and M adhya P radesh. T he m ost com plete d a ta exists for th e villages A urepalle (in the sta te of And h ra P radesh), and S hirapur an d K anzara (in the sta te of M aharashtra). W hile th e IC R ISA T survey was conducted over th e period 1975-1984, th e labor m arket d a ta is available only for the period 1979-1984. During the survey period, th e EGS was operational in Shirapur and K an zara and so I will calibrate th e m odel using d ata from these two villages for th e years 1979-1984. O ne can refer to W alker & Ryan (1990) for m ore details of the region and Singh, Binswanger Jo d h a (1985) for m ore details on the survey. Own farm o u tp u t of the agent in a good year is norm alized to one (y = 1 ). Since A = 9y, A then reduces to being a m easure of the productivity of th e EGS relative to th a t of the private sector (own farm ) . 2 A n exact value of A cannot be specified. T his is because of th e n ature of th e goods produced in th e EGS - it is difficult to o b tain the true value of a public good like a road, th a t is constructed using EGS labor. Therefore 5 alternative econom ies, characterized by different values of A , (A = 0.00, A = 0.10, A = 0.15, A = 0.20, A = 0.25) denoted as Econom y 1, 2, 3, 4 and 5 respectively are considered. In econom y 5, th e EGS is th e m ost productive, 2 When I am comparing productivities across the two sectors, I am actually comparing produc tivity of labor in the two sectors. 56 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. since the p ro d u ctiv ity of labor is the highest in his economy. I also assum e an annual discount rate o f 4%, so th a t /? = 0.9615 (th e risk free real interest ra te on long term governm ent bonds in India vary between 4% an d 10% per annum ). T he weight placed on consum ption (a ) is a m easure of th e consum ption-output ratio. In equilibrium it is approxim ately equal to th e average fraction of tim e spent on m arket activ ity (See K ydland & c P rescott (1996) and Cooley & P resco tt (1995)). For a rural econom y, we need to correct for tim e working on own faxm. Following Rosenzweig (1988) and Townsend (1994) th e tim e endow m ent of an ad u lt m ale is taken to be 312 days in a year, a is obtained as th e average of th e n u m b er of days in a year spent by an ad u lt m ale in work (EG S and own farm work) as a proportion of to tal tim e endow m ent. Accordingly a = 0.45. Few estim ates for coefficient of risk aversion 7 exist for the econom y u n d er con sideration. I therefore use the estim ates of M orduch (1990) and consider 7 = 1.5. However, in th a t paper u tility is a function of consum ption alone. In th e fram ew ork of this chapter u tility is a function of consum ption and leisure and, hence, 7 has to be corrected to account for this fact - the correct m easure of th e coefficient of risk aversion then is given by a ( l —7 ). 7 is set to be a value th a t solves a r ( l — 7 ) = 1 — 7 , where 7 = 1.5. Since a = 0.45. Accordingly 7 = 2.00 To obtain values of h° and h9, again assum e th a t the tim e endow m ent of an adult male is 312 days. h° is approxim ated by th e average num ber of days in a yeax spent by an ad u lt m ale on own faxm work, as a proportion of to ta l tim e 57 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. endow m ent. Sim ilarly, ha is approxim ated by th e average num ber of days in a year spent by an ad u lt m ale working in th e EGS as a proportion of to tal tim e endow m ent. Accordingly, h° = 0.40 and h3 = 0.30. T h e bench-m ark econom y is presented in Table (3.1) below. Table 3.1: B enchm ark Economies 7 a (3 h° h3 2 . 0 0 0.45 0.9615 0.40 0.30 T he tran sitio n probabilities Xss' are calculated so th at, on average villages face a bad rainfall yeax 30% of the tim e. Rosenzweig & c Wolpin (1993) find th a t over the period 1975-1984 (the period of the survey) Shirapur experienced b ad rainfall in 1977, 1978, 1983 and K anzara experienced b ad rainfall in 1976, 1977, 1979. Even though rainfall is a purely random event, th ere seems to be substantial serial corre lation in household incom e (see Lim (1991)) and a two year average d u ratio n for a bad shock is a fair approxim ation .3 T he tran sitio n m atrix is given by X g g X gb Xbg Xbb 3 For simplicity the mechanism that converts rainfall shocks into persistent income shocks is ignored. Doing this would require me to incorporate a productive asset in the model, which house holds use as a means of precautionary savings, but in doing so they affect their future production. Household ownership and sale of bullocks in the ICRISAT villages is a good example of such behavior. See Rosenzweig & Wolpin (1993). 58 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. T he elem ents of the transition m atrix axe calculated as follows: ( 1 — Xbb) = 2 , ( 2 year average d u ratio n of bad state) 0.70xff6 + 0.30x66 = 0.30, (30% probability of a bad state) Xgg Xbg 1 Xgbi 1 — Xbb, and th e transition m atrix is obtained as X = 0.7857 0.2143 0.5000 0.5000 Hence all the param eters of th e m odel (w ith the exception of A) have been calibrated using th e ICRISAT d ata. T h e com putation algorithm requires discretization of th e state-space to obtain exact num erical solution to th e discrete economy. D etails of th e algorithm axe in A ppendix 1. I next tu rn to th e results and th e welfare im plications of the EGS program . 59 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3.4 Welfare Analysis of the Employment Guarantee Program This section discusses th e welfare im plications of th e EG S. I first calculate th e op tim al w (or o p tim al yg, since every w is associated w ith a unique yg: given A). To this end, th e behavior of average steady state u tility (U ) is exam ined as the subsidy yg is varied over th e range [0,1]. U is defined as V = £ £ A ( a , . ) t f ( c ( a , i ) , J ( a , « ) ) 5 a T he optim al subsidy y“ is given by th e yg th a t m axim izes th e average steady sta te utility, U. This y“ depends on the productivity of th e EGS, relative to the private sector. N ext I calculate the welfare gains of th e EGS program s and exam ine the effects of an across the board increase in wages in the EGS (th e 1988 wage increase). Table (3.2), presents some of the sum m ary statistics describing the equilibrium for each economy I study. N otice th at when particip an ts are paid an optim al wage in the EGS, the coefficient of variation of consum ption is m uch lower and asset holdings drop to zero in th e m ore productive EGS (economies 3, 4 and 5). Agents no longer need to hold th e non interest bearing asset (a) as a m eans of precautionary savings. T he EGS provides all th e necessary insurance. 60 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 3.2: Sum m ary Statistics (Benchm ark Economy) C riterion A = 0 A = 0.10 A = 0.15 A = 0.20 A = 0.25 y ’ a 0.03 0 . 0 1 0.80 0.75 0.70 Coeff. of Variation o f Consumption ya = 0 w = 0.25 y3 = y'g 27.66 22.23 27.63 27.21 22.27 27.04 26.28 22.26 2.32 24.61 22.25 2.32 22.18 22.18 2.32 Asset Holdings y9 = 0 w = 0.25 y3 = V o 2.1027 0.9669 1.9767 1.7150 0.9913 1.6851 1.5025 1.0031 0 1.2702 1.0115 0 1.0161 1.0161 0 F igure 1 presents th e average steady s ta te u tility for Economies 1 , 3 an d 5 (i.e., econom ies corresponding to A = 0.00, A = 0.15 and A = 0.25) for different values of th e subsidy yg. T he utilities have been norm alized such th a t th e m axim um U in each econom y equals 1.00. N otice th e non-m onotonic behavior of U as yg increases over th e interval [0,1]: as yg increases, first U, increases (Region 1 ), th en decreases (Region 2) an d finally increases again (R egion 3). How does one explain th is non m onotonic behavior of U as yg increases? To do this I exam ine the behavior of U in each of th e th ree regions. C onsider Econom y 1 (i.e., A = 0.00). U tility is a function of consum ption and leisure. T he average u tility for th e economy is th e w eighted average of th e utility o f th e agents in th e good s ta te and in the bad state. 61 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 0.99 1.15 0.98 g 0.96 C O O 0.95 0.94. 0.8 0.6 0.7 0.9 0.3 0.5 Subsidy 0.1 0.4 F ig u re 1: A v e r a g e S te a d y S ta te U t i l i t y f o r A lte r n a tiv e E c o n o m ie s R e g io n 1 : S tartin g from yg = 0, as th e subsidy ra te in the EGS increases, the tax rate required to balance th e governm ent budget increases and the u tility of the agents in th e good s ta te decreases (this agent is not given the choice of not working in a good state). Now consider the agents in th e bad state. For such agents, as yg increases, consum ption increases (this is th e consumption effect) b u t leisure decreases (this is th e leisure effect). S tarting off from a low yg, th e positive consum ption effect is strong enough to dom inate both the negative leisure effect for the agents in the bad state and the decreasing utility of th e agents in th e good state, so th a t average utility for th e en tire econom y increases. R e g io n 2 : As yg increases m ore people decide to work in the EGS (for low yg agents in the bad state who s ta rt off w ith high asset holding choose not to work - instead 62 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. dis-save to satisfy their consum ption requirem ents) and consequently the tax rate on incom e increases. Beyond yg = 0.03, the decreasing u tility of th e agents in the good sta te dom inates, and average u tility for the entire econom y decreases. One m ust note th a t the utility for th e agents in the bad sta te increases - for the agents in th e bad state, th e positive consum ption effect still exceeds the negative leisure effect b u t this dom inance is n o t strong enough to overcom e th e decrease in utility of th e agents in the good state. R e g io n 3: As yg increases beyond 0.35, there is no leisure effect, and one is back at th e stan d ard insurance problem . In this range every agent in the bad state works in th e EGS and no agent in th e good state works in th e EGS. Here average utility for th e agents in th e good s ta te decreases, the average u tility of th e agents in the bad sta te increases and th e average utility for the en tire econom y increases. The EGS is equivalent to a pure transfer/insurance scheme in this range (this is the pure transfer effect). For A = 0.00 the governm ent em ploym ent program is not productive so average u tility is m axim ized a t yg = 0.03 - th e transfer effect in Region 3 is not strong enough to overcom e th e strong leisure effect in Region 2. This econom y is alm ost the same as an econom y w ith pure insurance - only worse, because agents have to work (to o btain insurance) w ithout adding to aggregate production (and consum ption) in the economy. If A > 0.10, the m axim um utility is obtained in Region 3. For exam ple for Econom y 3, optim al yg = 0.80. For Economies with A > 0.10, th e EGS is sufficiently 63 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. productive, so th a t as m ore people choose to work in the EGS, average incom e (and hence consum ption) in th e econom y increases and th e pure transfer effect in Region 3 dom inates. For these economies, th e ad d itio n al production (and consum ption) in th e EG S com pensates for th e negative leisure effect of agents having to w ork. T he yg a t w hich th e economy moves from one region to another varies across econom ies. Now consider the 5 exam ple econom ies. Row 1 of Table (3.3) presents th e o p tim al w in each of th e five economies; Row 2, th e average (steady state) utility correspond ing to optim al yg = y~ (the yg th a t m axim izes average steady state u tility ); Row 3 th e average u tility for the econom y w ith no subsidy (in this economy agents in th e EG S are paid th eir m arginal product w hich is equal to A). Row 4 presents th e tax ra te ( r “) th a t has to be paid to su stain y*. Rows 2 and 3 im ply th a t agents would prefer to live in the econom y w here they axe paid th e optim al subsidy rate ra th e r th a n in th e economy where th ey axe not paid any subsidy (i.e., an econom y w here they axe paid a wage exactly equal to th e ir m arginal product in the EG S). To ob tain a m agnitude of this welfare loss I ask th e following question: In each econom y, com pared to the case w here yg = 0 (equivalently w = A), w hat lum p-sum transfer m ust be given to the agent, so th a t he has th e sam e average u tility w hen p aid y“? T h e m easure is presented in Row 5 (th e welfare cost of paying yg = 0). T he lum p-sum transfer tu rn s o u t to be 0.01% of th e G D P for Econom y 1, 0.06% for Econom y 2, 0.27% for Econom y 3, 1.57% for Econom y 4 and 2.63% for Econom y 5. As expected the m ore pro d u ctiv e 64 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 3.3: W elfare Costs for Alternative Economies (Benchmark Economies) C riterion A = 0 A = 0.10 A = 0.15 A = 0.20 A = 0.25 y l 0.03 0 . 0 1 0.80 0.75 0.70 A verage U tility yg = y* 1 . 0 0 1 . 0 0 1 . 0 0 1 . 0 0 1 . 0 0 A verage U tility yg = 0 0.9996 0.9994 0.9978 0.9877 0.9802 T * 0 . 0 0 0 1 0.0003 0.2437 0.2284 0.2132 W elfare cost o f Paying yg = 0 0 . 0 1 0.06 0.27 1.57 2.63 A verage U tility w = 0.25 0.9713 0.9865 0.9910 0.9856 0.9802 W elfare Cost o f Paying w = 0.25 3.93 1 . 6 8 1.07 1.71 2.63 A verage U tility w = 0.40 0.9445 0.9720 0.9826 0.9828 0.9832 W elfare Cost o f Paying w = 0.40 8 . 0 1 3.77 2.48 2.34 2.31 A verage U tility w = 0.50 0.9496 0.9772 0.9876 0.9878 0.9879 W elfare Cost o f Paying w = 0.50 7.26 3.45 1 . 6 6 1.64 1.78 W elfare C o st m e a su re d a s % o f G D P the EGS the g re ater th e welfare loss by paying agents yg = 0. This lum p-sum transfer is a m easure of th e potential welfare benefits o f the em ploym ent guarantee program . Figure 2 below presents this p o ten tial welfare benefits as a % o f G D P for th e altern ativ e econom ies (benchm ark m odel). 65 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. A = 025 2.5 - a. a a 'o s R 2 - A=0-20 a > | ! . 5 h S C B l 1 a > o a. A=0.15 0.5 - A=0.10 A=0.00 2 3 4 Alternative Econom ies Figure 2: P o te n tia l W elfare B en efits o f EGS 3 .4 .1 A n E valu ation o f th e E G S in M ah arash tra Lack of knowledge of A im plies th a t th e governm ent does not pay w‘ or equivalently y*. I now look at w hat th e governm ent actually pays, and evaluate its welfare im plications. From th e ICRISAT d a ta I can arrive at a rough m easure of the w th at is paid by th e governm ent. T he average incom e from own farm activity, per unit of land owned, per adult fam ily m em b er is norm alized to be equal to one. T he average incom e from the EGS per ad u lt m ale is obtained as a fraction of th e norm alized average incom e from own farm . T his figure tu rn s out to be 0.25. So it is assum ed th a t th e governm ent pays w = 0.25, (w = A + yg). In this case the welfare cost of paying this w is m easured by asking th e following question: w hat lum p-sum transfer 66 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. in this econom y would provide the agent w ith th e sam e average utility as in the econom y w ith w ith optim al subsidy (y*)? T his welfare m easure is presented in Row 7 of Table (3.3). T h e welfare cost is different depending on how productive th e EGS is, relative to th e private sector, i.e., depending on th e value of A. T he lum p-sum transfer is 3.93% of G D P in Economy 1 , 1.68% of G D P in Economy 2, 1.07% in Economy 3, 1.71% in Economy 4 and 2.63% of G D P in Econom y 5. N ote th at U(w = 0.25) < U(w = A) w ith strict inequality for economies 1 , 2, 3, and 4. So agents living in these four economies would prefer to be paid w = A (w here A is the m arginal product of labor in the EG S), rath er th an w = 0.25. In econom y 5, U(w = 0.25) = U(w = A). 3.5 3 - Q . O 2 2 . 5 O “ 2 - 8 o ® « 1.5 a > 5 0.5 A=0.00 *=0.25 *=0.10 A=0.20 A=0.15 2 3 4 Alternative Economies F ig u r e 3 (P a n e l A ): W e lfa re C o s t o f P a y in g w = 0.25 67 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. In 1988, th e governm ent in stitu ted a su b stan tial increase of the wage ra te in the em ploym ent g u aran tee program in th e s ta te o f M aharashtra. This can b e thought of as an increase in th e subsidy ra te to w = 0.40. I m easure w hat lum p-sum transfer in such econom ies (i.e., economies w ith w = 0.40) will give the sam e u tility as in the economies w ith o p tim al subsidy. This lum p-sum transfer as a % of G D P am ounts to 8.01%, 3.77%, 2.48%, 2.34% and 2.31% in Economies 1, 2, 3, 4 an d 5 respectively. Since U(w = 0.25) > U(w = 0.40) for econom ies 1 , 2, 3 and 4, a m ovem ent from w = 0.25 to w = 0.40, actually increases th e welfare cost (decreases th e u tility of agents) in all, except Economy 5. This wage increase is a m ovem ent in th e right direction only for Econom y 5, i.e., when th e productivity of th e EGS is th e highest. 9 8 7 Q. O 6 a " o ^ 5 C O C O 0 9 (34 s JO | 3 2 1 ° 0 1 2 3 4 5 6 Alternative Econom ies F ig u r e 3 (P a n e l B ): W e lfa re C o s t o f P a y in g w = 0.40 68 A=0.00 *=0.10 A=0.15 \=0.20 A=0.25 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. W ill a fu rth er increase in w help? I experim ent w ith a fu rth e r increase in w to w = 0.50. N ote from Table (3.3) (above) U(w = 0.40) < U(w = 0.50) for all the Econom ies. Consequently, th e welfare cost of paying w = 0.50 (again m easured by a consum ption loss as a % o f G D P, of not paying the op tim al wages) is lower for all th e econom ies. So a fu rth er increase in w is a move in the rig h t direction. P an el A of Figure 3 presents th e welfare cost of paying w = 0.25 as opposed to y*. P an el B the welfare cost of paying w = 0.40 as opposed to w“ and Panel C the change in welfare cost expressed as a % of G D P as a result of increasing w from 0.25 to 0.40. Notice this change is negative only for Econom y 5, indicating th a t the welfare cost of paying yg = 0.40 com pared to th a t of paying w = 0.25 is lower only for Econom y 5. 4.5 A=0.00 3.5 a 2.5 A=0.10 A=0.15 1.5 0.5 -0.5 Alternative Economies F igure 3 (P a n e l C): C hange in W elfare C ost 69 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. To sum m arize, first note th at the optim al wage in the EGS generally exceeds the m arginal product of labor. It is therefore not optim al to pay th e workers a wage equal to th e m arginal product of labor. T h e welfare cost of paying th em a wage equal to the m arginal product of labor, as opposed to th e optim al wage ra te varies from 0.01% of G D P to 2.63% of GDP, depending on the productivity of th e EGS. T he m ore productive th e EGS, the greater is th e welfare cost of paying workers a wage equal to th e m arginal product of labor. I exam ine th e EGS in greater detail by obtaining a m easure of w hat wage is actually paid to the participants, using d ata from two IC R ISA T villages in the Indian s ta te of M aharashtra. This wage is also less th an the optim al. M oreover agents would prefer to be paid w = A ra th er than this wage rate. F inally I exam ine the welfare effects of an across the board increase in wages in the EG S (th e 1988 wage increase). This increase actually reduces welfare for all, but th e m ost productive EGS!. I also show th a t a further increase in wages in the EGS would be a m ovem ent in the right direction. This com pletes the welfare analysis of th e EGS. I test the robustness of th e results to changes in p aram eter values by conducting a series of com putation experim ents. I consider variations in 7 (the coefficient of risk aversion), a (the share of consum ption A in the utility function), and h9 (the pre-determ ined num ber of hours th e agent has to work in th e EG S, in th e event he chooses to work). We also test w hether the assum ptions th a t I have m ade regarding labor participation in the EGS and own farm work m a tter. I find th a t y'g is significantly affected by changes in a , 7 and h9, 70 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b u t changes in assum ptions on participation in EGS and own farm work do not have any significant im pact. T he results are presented, in detail, in A ppendix 2. 3.5 Conclusion T his chapter studies the role of an em ploym ent guarantee schem e in th e Indian s ta te of M aharashtra, in a m odel calibrated using d ata from two IC R ISA T villages (in M aharashtra). Agents in th is econom y are subject to exogenous incom e shocks an d are unable to insure them selves through private credit m arkets. T he only way th e y can self insure is by holding non-interest bearing assets. However if agents have access to guaranteed em ploym ent (of th e kind discussed), all insurance is provided by th e em ploym ent guarantee program . Agents do not need to hold any non-interest bearing asset for precautionary savings. T his insurance aspect of th e EGS has not been exam ined previously. Because of this insurance aspect, th e optim al wage in th e EGS generally exceeds th e m arginal product of labor in th e program . Given the fact th a t it is difficult to ob tain a tru e m easure of the productivity of labor in the EG S, th e analysis is conducted by considering five exam ple econom ies, differentiated by (labor) productivities of th e EGS. T he welfare gains of paying th e optim al wage in th e EGS (as opposed to any non-optim al wage) depend on how productive the EGS is, relative to the private sector. In particular, the welfare costs of paying a wage equal to the m arginal product of labor in the EGS (as opposed to paying th e optim al wage rate) varies from 0.01% of G D P to 2.63% of G D P, depending on 71 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. the productivity of th e EGS. I also calculate th e welfare im plications o f an across the board increase in wages in the program . Such a move is in th e rig h t direction only for th e m o st productive economy (Econom y 5). For the other econom ies, this increase is not sufficient. A further increase in wages would be m ovem ent in the right direction, since a fu rth er increase in w actu ally reduces the welfare cost. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3.6 Appendix 1: Computation Algorithm T h e co m p u tatio n algorithm requires discretization of th e state-space to obtain exact n u m erical solution to th e discrete economy. T he co m p u tatio n algorithm is as follows: 1 . G iven th e subsidy ra te w , choose a certain r an d use value function iteration to solve equation (3.5). 2. T h e invariant distribution A corresponding to th ese decision rules is found by ite ra tin g on A ( a ' ,s ' ) = ^ £ x(s,s')A(a,s) • * a t n(a',3') 3. T h e decision rules and th e corresponding invariant distrib u tio n is used to eval u a te th e governm ent budget constraint (G B C ). 4 . Stop if GBC is satisfied, else we vary the ta x ra te r and follow steps ( 1 ) — (3), u n til th e governm ent budget is balanced. T h e algorithm is discussed in detail in Im rohoroglu, Im rohoroglu & Joines (1993) an d H ansen & : Im rohoroglu (1992). T h e grid space is discretized by choosing a grid o f feasible asset holdings. T he m axim um num ber of assets th a t an agent holds is assum ed to be 8 (This figure is th e m axim um value of physical and financial assets held by an agent in a good state, norm alizing his incom e in a good sta te to be equal to one). A grid of 301 points w ith increm ents of 0.027 is used. The aggregate state for each village has two realizations, (s = g, 6 ) an d so th e to ta l num ber of points in 73 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. th e sta te space is 301 x 2 = 602. A ccordingly th e decision space also has 602 points and th e value function is also a 301 x 2 m atrix . S tart w ith an initial value function Vo( a, s) an d use th e following u p d ate to o b tain th e decision rule at the end of the first itera tio n and a value function V i(a,s). V jt+ 1 ( a ,s ) = M axa/ M ax<f Max,* U(e\g) + = ff,e); U{u\g) + P'52s>x(Sis')Vk((a/,s ,)\i = g,u); U(e\b) + /? E ,x ( a ,* ') V t ( ( a ', 3 ')[i = 6 ,e); U(u\b) + j3 E S ' x (s, s')Vk((a', s ')|i = 6, it); subject to a' > 0. Continue until convergence is attain ed . U(d\s) is defined above for each set of agents (see Section 2 ). 3.7 Appendix 2: Sensitivity Analysis 3 .7 .1 V aryin g th e P a ra m eter V alu es This section exam ines th e sensitivity of th e results to the param eter values? The results are not affected by changes in th e m axim um num ber o f assets th at the agent can hold. However, they axe significantly affected by variations in the value of 7 (the coefficient of risk aversion), a (the share of consum ption in th e utility function), and 74 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. k 9 (the pre-determ ined num ber of hours the agent has to work in the EGS, in the event he chooses to work). Im plications of increases in risk aversion is presented in Table (3.4). Consider Table 3.4: Im plications of Varying 7 , A = 0.15 7 y'a 7 = 2 0.80 7 = 5 0.75 7 = 1 0 0.70 Econom y 3, i.e., A = 0.15. N ote, y“ decreases as 7 increases, i.e., as agents become m ore risk-averse, th e optim al subsidy decreases. T his, som ew hat counter-intuitive, result can be explained as follows. Given the p articu lar form of the utility function, individuals are concerned w ith m axim izing th e com posite com m odity zt = c°l\~a. T he m ore risk averse the agents are, the m ore im portance they place on sm oothing this com posite com m odity zt. As 7 increases, given th a t they face variability in leisure, agents choose a m ore variable consum ption sequence in order to obtain a sm ooth p ath of the com posite commodity. Thus as 7 increases, agents seek a m ore volatile sequence of ct, arid hence require a sm aller yg to a tta in optim um . In th e benchm ark m odel, a. = 0.45. a is defined as th e average fraction of tim e spent in m arket work, including work on own farm . W hether tim e spent on own farm is indeed a m arket activ ity is a m atter of debate. In the event own farm work is excluded when m easuring m arket activity, a = 0.33. This implies th a t agents 75 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. p u t a g reater weight on leisure, since a decrease in a implies a decrease in th e share of consum ption in the utility function (correspondingly agents s ta rt valuing leisure m ore). T h e im plications of reducing a is presented in Table (3.5). A verage steady Table 3.5: Im plications of Varying a a y; A = 0.15 A = 0.20 A = 0.25 a = 0.45 0.80 0.75 0.70 a = 0.33 0 . 0 0 0 . 0 0 0 . 0 0 sta te u tility follows the sam e p a th for th e two choices of a. However, th e welfare im plications are quite different. For a = 0.33 none of the exam ple econom ies are productive enough so th a t the consum ption gain is not sufficient to com pensate for the loss in u tility due to decreased leisure from having to work - th e pure transfer effect in Region 3 is not strong enough to overcome the leisure effect unlike in the benchm ark economy. T he optim al t/“ is 0.00 in these economies, i.e., it is optim al to pay a wage equal to the m arginal p ro d u ct of labor. T h e im plications of an increase in h9 is sim ilar to the effect of decreasing a . An increase in h9 implies th at agents who choose to participate in th e EG S have to work for m ore (pre-determ ined) hours. I focus on an increase in h9, because this is one policy variable th a t the governm ent m ight use. If A < 0.20, th e EGS is not productive enough so th a t the leisure effect in Region 1 dom inates th e pure transfer effect of Region 3, and higher yg, an d hence higher consum ption for EG S p articip an ts 76 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. does not com pensate for th e higher num ber of hours th ey have to work. The results axe presented in Table (3.6). Table 3.6: Im plications of V arying h3 ha ym a A = 0.15 A = 0.20 A = 0.25 ha = 0.30 0.80 0.75 0.70 h3 = 0.35 0 . 0 0 0 . 0 1 0.70 3 .7 .2 R ela x in g th e R e str ic tio n s o n L abor M ark et P a rticip a tio n T h e analysis has so fax assum ed th a t in a good state, agents do not have the option o f not working on th eir own farm . In addition they can also work in th e EGS. I next ex p erim en t with relaxing th is assum ption. 1. Individuals are allowed to choose between working on his own farm and in the EGS in a good state. 2. Individuals are not allowed to work in the EGS in a good state, but he is allowed to choose not to work on his own farm in a good state. I exam ine th e results of each of these experim ents next. In th e first experim ent, I allow th e agent to choose betw een working on own farm an d in th e EGS in a good sta te . All other param eters are left unchanged (i.e., th e agent has to work a pre d eterm in ed h° fraction of tim e on own farm should he choose to, h3 fraction of tim e 77 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. in th e EG S, should he choose to). Once n a tu re reveals its state (5 ), i.e., s = g, 6, given s, an d th eir current asset holdings, agents choose w hether and w here to work (d), i.e., choose d = ( / , m , u). N ote d = / im plies th a t the agent chooses to work on own farm , d = m implies th a t the agent chooses to work in the EG S, and d = u im plies th a t th e agent chooses not to work. A fter the choice of d, all th a t is left for the individual is to choose asset holding (a') for th e next period and current consum ption (C ) su b ject to th e budget constraint, an d th e asset accum ulation equation. This experim ent is further interesting because it brings the work-fare program closest to a stan d a rd transfer based welfare (unem ploym ent insurance) program . In a such a program (see Hansen & Im rohoroglu (1992)) agents are offered an em ploym ent o p p o rtu n ity according to a known stochastic process (or equivalently face a good sta te o f n a tu re according to a known stochastic process). If they are not offered an em ploym ent opportunity they can collect unem ploym ent insurance (if they face a bad s ta te of n atu re they can obtain a transfer from the governm ent). However, in th e case stan d ard unem ploym ent insurance program the agent m ight stan d to gain by rejecting the em ploym ent o p p o rtu n ity (choosing not to work on own farm ) and pretending to be unem ployed (pretending to face a bad state of n a tu re), thereby collecting unem ploym ent insurance and enjoying leisure. In the work-fare program of th e k in d I discuss, to collect insurance th e agent has to work. T herefore only agents who face the bad state will choose to work in the EGS. Those facing a good state will choose to work on own farm . In principle therefore this kind of a work-fare 78 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. program can provide a solution to the m oral hazard problem th a t plagues a standard transfer based insurance problem . Table (3.7), presents th e results for th e altern ativ e economies. The param eter values used axe th e sam e as in Table 1. N otice th e results are unaffected for the low productivity econom ies (i.e., Economies 1 and 2). For Economies 3, 4 and 5, optim al y “ is lower in this economy. I obtain th e welfare costs of paying a wage equal to th e m arginal product of labor in the EGS, paying a wage w = 0.25 in the EGS, and increasing wages in th e EGS to w — 0.40. F irst, th e m ore productive the EGS, the greater is the welfare cost of paying a wage equal to th e m arginal product (i.e., paying yg = 0 ) . T he rest o f th e implications are the sam e as in th e benchm ark model (discussed previously), including the argum ent th a t an across th e board increase in wages is a move in th e right direction only Econom y 5. Table 3.7: Results for A lternative Economies Criterion o I I < A = 0.10 A = 0.15 A = 0.20 A = 0.25 y'g 0.03 0 . 0 1 0.65 0.60 0.55 W elfare Cost ya = 0 0 . 0 1 0.06 0.13 1.25 2.25 W elfare Cost w = 0.25 3.93 1 . 6 8 0.82 1.54 2.25 W elfare Cost w = 0.40 8 . 0 1 3.77 2.08 2.07 2.03 In th e second experim ent th e agent in a good s ta te is not allowed to work in the EGS, but he is given the option of choosing not to work on own farm in a good year. Once n atu re reveals its s ta te (s), i.e., s = g,b, given th eir current asset holdings, agents choose w hether to work or not (d), i.e., choose d = f ,m , u. N ote d = f implies 79 ( t i ___ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. th a t th e agent chooses to work on ow n farm in a good year, d = m im plies th a t the agent chooses to work in th e EGS in a bad year, and d = u im plies th a t th e agent chooses not to work, irrespective o f s. Finally after he chooses d, th e individual has to choose asset holding (a') an d consum ption (C) subject to th e budget constraint, an d th e asset accum ulation equation. T h e results are the sam e as in th e benchm ark econom y (Table (3.3)). Hence preventing agents from p a rtic ip a tin g in th e EGS in a good sta te (but providing th em th e option of not working on th eir own farm in a good state) does not add an y th in g to our analysis of the E m ploym ent G uarantee program . SO Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Chapter 4 Consumption Smoothing in Rural Punjab in the Green Revolution Years 4.1 Introduction Tests for consum ption insurance in In d ia have focussed prim arily on low incom e regions, characterized by traditional m odes of production - the ICRISA T villages (Townsend (1994)). T h e purpose of this ch ap ter is to test for consum ption insur ance in an o th er region in India, nam ely ru ral P u n jab . T he choice of P u n jab is m otivated by th e fact th a t households in ru ral P u n jab and those in th e ICRISA T villages co n stitu te tw o extrem es of th e ru ral population in India. W hile agricul tu ra l production in th e ICRISAT villages is characterized by trad itio n al m odes of production, susceptible to th e vagaries o f th e monsoons, agriculture in P u n jab is characterized by m odem , capital intensive farm ing techniques. C om parison of th e SI Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. results of tests of consum ption insurance in th e tw o regions will provide useful infor m ation on th e role of in stitu tio n s in th e two regions and how successful they are in enabling households to sm ooth consum ption. In p articular, this chapter exam ines th e success of households in rural Punjab in insuring consum ption against incom e fluctuation in the green revolution years. This chapter uses th e A dditional R ural Incom es Survey (ARIS) D ata, collected by th e N ational C ouncil of Applied Economic R esearch (N C A E R )1, New Delhi, to test for consum ption insurance in rural P u n jab . 2 To sum m arize th e m ain conclusions of this chapter, I find th a t for th e population as a whole th e null hypothesis of full consum ption insurance is rejected. These results are very different from those obtained using d a ta from th e ICRISAT villages. T he null hypothesis of com plete risk sharing by households w ithin each village is rejected for households in ru ral P u n jab risk sharing by households w ithin each village. W hen the population is sub-divided according to landow nership, the null hypothesis of com plete risk sharing is rejected for the landless households and the sm all farm ers. However, for the m edium and large farm ers the null hypothesis of full insurance cannot be rejected. F u rth er, there exists significant regional variation in the ability of households to insure consum ption against incom e changes. Finally I test for risk sharing across villages and find th at there is significant evidence in its favor. xThe ARIS-NCAER Survey. 2By rural Punjab, we mean villages in Punjab and Haryana. See Table (1.4) in Chapter 1. S2 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. T h e rest of this chapter is organized as follows. In section 4.2 I describe the em pirical specification and estim atio n procedure used to test for consum ption insur ance T h e estim ation results axe presented in Section 4.3. In section 4.4 I present the test for risk shaxing across villages. Section 4.5 presents the concluding rem arks. 4.2 Test for Consumption Insurance Tables (4.1) and (4.2) present th e m ean and volatility of incom e and consum ption over th e survey period, classified by land holding class and d istrict. V olatility is m easured by the coefficient o f variation (C.V )3. T otal household incom e is defined to include crop income, incom e from livestock, oth er agricultural incom e, agricultural wages, non-agricultural wages, incom e from salaries and incom e from self employ m ent. N otice first th a t the volatility of income is higher than th e volatility o f con sum ption for the population as a whole (com paring colum ns (3) and (5) in Table (4.1). Second, for the population as a whole, notice th a t th e volatility of income and consum ption has declined over th e survey period. C.V. of incom e decreased from 90.014% in 1968-69 to 73.79% in 1970-71 and the C.V. of consum ption de creased from 66.004% in 1968-69 to 56.992% in 1970-71. See Table (1.5) in C hapter 1 for a classification of households on the basis of land holding class. Table (4.2) presents d a ta on regional variation of income and consum ption, w ithin P u n jab . The 3CV- = S t - fcrgr ' * 100 S3 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. d a ta shows th a t th e region under consideration is fairly heterogeneous. However one th in g is clear, consum ption is less volatile th a n incom e across different regions of the state. 4 .2 .1 T h e o ry T he test for full consum ption insurance is th e te st o f th e validity of P areto O ptim ality for th e econom y under consideration. T he problem for th e social planner is I oo S M ax S H E H i s P s ^ U ( C i t s ', O its ) t=l t= 0 3— 1 subject to I I E = E yit” v (*’s ) f = i t = i All th e term s are th e sam e as before. To solve th is problem , I form the Lagrangian as L = ^ 2 ^ [ ^ 2 H i s E P * U ( c i t 3 ', Q its ) + A t s - Q ^ ( y tt s - Cit a ) } ] t=0 «=I s= 1 i= 1 I will consider two alternative preference specification - an exponential utility func tion, where th e specific form of U is U(cit3; 0its) = - - e x p { - a ( c t - ts - 0tts)} (4.1) a S4 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. and a pow er u tility function, where th e specific form of U is c - - 7 U(c*a; 6it3) = exp(6its) (4.2) 1 — 7 Using th e preference specification in equation (4.1), th e first order condition (after m anipulations, and disregarding th e notation for th e sta te - see chapter 1 ) can be w ritten as Ac* = Ac? + (Adit - A0?) where 1 V ' 1 i=l = 7 X > . 1 i=l U nder th e assum ption of Full C onsum ption Insurance, individual consum ption c* depends only on aggregate consum ption c?. The em pirical Specification then is Ac* = (3 q + ,#iAc? + /? 2 Ay* + Vitt+i (4.3) N ote u,-tt+i incorporates b o th preference shocks (5*) and m easurem ent error. The com plete risk sharing m odel predicts th a t (3 1 = 1 and /? 2 = 0 . 85 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Using the preference specification as in equation (4.2), th e first order condition (after m anipulations, and disregarding the notation for th e state) can be w ritten as Alog(cft) = Ac-* + - A0?) 7 where cT = t E 1 ^ ) 1 i=i < n = t E * . 1 t=i I can now w rite th e em pirical Specification as A log(c,'t) = (3 q + /?iAc”° + (32 A log(y,-t) + u f , £ + 1 (4.4) T he com plete risk sharing m odel predicts th a t 0i = 1 and / ? 2 = 0. Note the only difference betw een th e two specifications is th a t in the case of power preferences (w ith m ultiplicative preference shocks) th e growth rates of con sum ption, net of preference shocks are equalized across individuals, whereas in the case of exponential preferences, changes in consum ption net of preference shocks are equalized across individuals, yu (or correspondingly log(yti)) includes any id iosyncratic variable, household income being one of th e m any possible. T he joint hypothesis j3\ = l , / ? 2 = 0 tests a stronger proposition of optim al allocation w ithin S6 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. the village. A weaker hypothesis /? 2 = 0 only tests w hether households are able to insure against idiosyncratic shocks - here consum ption allocation w ithin the village need not be o p tim al (i.e., there is less th an full insurance). How is th e P areto O ptim al A llocation attain ed in a decentralized econom y? Let m e assum e th e existence of a com plete set of Arrow-Debreu securities. T he exis tence of such securities allows one to decentralize th e economy and exam ine w hether full consum ption insurance can be attain ed through m arket m echanism s in such an economy. T his enables m e to test for consum ption insurance for subsets of the pop ulation - in p articu lar I test for consum ption insurance for households classified by land holding and location. One can show th a t if th ere exists a com plete set of Arrow- D ebreu securities, th e equilibrium consum ption allocation will be identical to those obtained under th e social planner’s problem (see chapter 2 of this dissertation). This com pletes th e discussion on th e theoretical foundations of th e test for con sum ption insurance. I now turn to the em pirical specification. 4 .2 .2 E m p irica l S p ecification I now discuss th e em pirical specification an d th e estim ation procedure. Let c,„t denote the consum ption of household i in village v and in year t(i = 1 , . . . , / ; u = 1 = 0, ...oo). U nder the assum ption of com plete risk pooling, I can w rite S7 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. change in individual consum ption to be a function of its village level average alone, so th a t the o p tim al consum ption allocations can b e represented in th e following form Ac,vt = Po + /?i A -+ - Vijt+i In each case I also consider a growth rate specification, and results are presented for both the first difference and the grow th ra te specification. The sta n d a rd te st for full consum ption insurance th en involves th e following regression: A c,vj = /?o /? iA c ^ + ^ 2 A i/iv t + 1 (4 -5 ) where * /,-„ * is th e gross incom e of household i in village v and in year t . T he altern ativ e is the regression Ac,-„t = /?0 + /3i A c“f + /? 2 A Eivt + Ut\*+i (4.6) where A E{vt is th e change in num ber of earners in household i in village v, betw een years t and t + 1. Full consum ption insurance requires = l , / ? 2 = 0, i.e., household consum ption tracks average village level consum ption and nothing else, household specific vari ables does not affect household consum ption.4 4As I have noted, in each case I also consider a growth rate specification. Hence I also consider the following regressions A log(c,„t) = 0o + 0 i A + 02 A log(y,-ut) -1 - (4.7) A log(c,„£) = 0o + /?iAc*f + 02 A E ivt 4- (4.8) ss Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. T h e problem w ith th is specification is th a t th e village m eans are not observable from th e survey d a ta . N ote in th e A R IS-N C A ER d a ta th e num ber of households surveyed in each village does not exceed 23, and so one cannot assum e c“t (or equiv alently c £ ) as know n. N otice if one were to use th e ICRISA T d a ta as in Townsend (1994) this problem does not arise because th e sam ple size in each village is big ger and so one can use the sam ple m ean to approxim ate th e population m ean - replace by th e sam ple m ean for the village excluding th e consum ption of th e ith household. Hence estim atio n of equations (4.5) and (4.6) will lead to inconsistent estim ates since c“t is m easured w ith error. O ne way to get rid of this problem is to consider th e following regression (see D eaton (1990)): v = 5 3 ljdvj,t + 02 A y i v t + v i,t + i (4-9) i=i or alternatively having the change in th e num ber of earners w ithin the household as an explanatory variable v A Civt = 5 3 'Ifjd-uj't + 02&Eivt + Vij+l (4.10) j = i S9 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Note dvj't is a dum m y such th at dvj,t — 1 if v = j 0 otherw ise The coefficient 0 2 = 0 if there is com plete risk sharing w ithin th e village. So I estim ate equations (4.9) and (4.10) an d test 0 2 = 0. The altern ativ e hypothesis implies th a t consum ption changes are n o t independent of income changes (or changes in th e num ber of earners in the fam ily). In stru m en ta l V ariable E stim a tio n Rejection of th e jo in t hypothesis (3X ~ l , / ? 2 = 0 (or the weaker hypothesis fi2 = 0) in th e regression5 Ac,ut = 0o + /?iA c“t + 02Ayivt + e,„it+i implies th a t th e household cannot perfectly sm ooth consum ption against income fluctuations (th ere is less than full consum ption insurance). However there m ight be occasions where households are sm oothing income directly (possibly through conservative consum ption and production choices or through variations in m arket participation). In such a situation th e null hypothesis will be rejected because con sum ption will track income, b u t then incom e itself is sm ooth. R ejecting the null hypothesis in this case is equivalent to com m itting a Type I error. T h e argum ent ’For this subsection I examine only the first difference formulation. 90 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. is th a t agents look at the fu tu re when deciding how m uch to consum e, they will save m ore when they have reason to expect th a t they are likely to face a decline in incom e in th e future. Further, if agents have private inform ation about their future incom e, th en one could possibly use th eir current saving to pred ict th e ir future in com e changes m ore accurately th a n it would be possible using public inform ation. So if households are able to sm ooth consum ption (through incom e sm oothing), one should be able to detect this by exam ining th eir savings behavior. Following Deaton (1992), I estim ate the sam e equation A C i-u t = (3 q -+ - /?i A c“t + ^ A t H vt + b u t using instrum ents for A I use lagged savings and incom e, and respectively as th e instrum ents. It is argued th a t lagged savings and lagged income will reveal private inform ation o f th e household which is otherw ise not publicly observed. So th e appropriate tw o stage process is to consider th e following set of regressions: A yivt = ao + Qi'Siv.t-i + 022/tv,t-i + utv,t+i (4-11) Ac,uf = (3 0 -f- (3i A c“ t + faAyivt + etv,t+i (4.12) 91 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. This is regressed, as a stan d a rd two stage least squares (2SLS). I com pare the results of this regression to th e stan d a rd test of considering an OLS A Civt = 0 o -h PiACrf -f- 0 2A y ivt + e,Vtt+i O nce again th e null hypothesis of full insurance is Ho : 0\ = 1,02 = 0. Now define 0 i v = \0 o 0 \ 02\rv as th e vector of coefficients obtained from the IV estim atio n and 0 O L S = [A) 01 0 2 ]o L S as th e vector of coefficients obtained from the OLS estim ation. I need to test if indeed one gains any additional inform ation from using th e instrum ents, i.e. are th e coefficients obtained using th e instrum ents different from those obtained from stan dard OLS estim ation? To test for equality of estim ates Ho : 0 i v = 0 o l s construct th e following W ald statistic (see Greene (1990)) w = (Po l s - M ' K X ' X ) - 1 - (X 'X )-L ] - ‘(/?ots - M (4 13) s2 and W ~ X2(2) 92 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. H ere s 2 is th e estim ate of a 2 o b tain ed using th e IV estim ator, and X is th e set of fitted values w hen the corresponding colu m n o f X is regressed on ail th e colum ns of Z , w here Z is th e set of in strum ents. So X is given by from eq u atio n (4.11), and Xt/ivt is th e predicted value o f A7/tut from equation (4.11), which is used as an in stru m en t in equation (4.12). R eject Ho : firv = Pols if IV > x 2(2)- T h is concludes the discussion on th e econom etric specification of th e m odel. I now e stim a te th e m odel to d a ta from rural P unjab using the N C A ER -A R IS d ata. I present th e results of each of th e regressions in Tables (4.3) - (4.12). 4.3 Estimation Results B a s ic R e su lts T h e basic results are presented in Tables (4.3) and (4.4). All variables axe m easured in real units. Since I do not have d a ta for district level price indices, I use th e All India price indices for agricultural laborers, base 1968-69 = 100, (C M IE (1993)). T able (4.3) presents th e results for th e first difference form ulation (from equations (4.5) an d (4.6)) and Table (4.4) th e results for th e grow th rate specification (from equations (4.7) and (4.8)). In each case I consider two idiosyncratic variables, change in household incom e (change in log of household income) and the change in the n u m b er o f earners in th e household. In colum n (2) I report the intercepts, in colum n (3) th e coefficient on th e village level average consum ption and in colum ns (4) and (5) 93 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. the coefficient on household incom e and earners respectively. T h e standard errors of th e estim ated coefficients are in parenthesis. In colum n (6) I present the F-R atio for th e jo in t test (/3X = l , / ? 2 = 0). N ote the jo in t test (change in household consum ption is correlated w ith village level average consum ption and is unaffected by household level idiosyncratic variables) is always rejected.6 T he real interest in th e result however revolves around the significance or in significance of the coefficient of the household level idiosyncratic variable (change in household incom e or change in the num ber of earners w ithin the fam ily). This is tested by the null hypothesis 02 = 0. T he estim ated coefficients are in columns (4) and (5) in Tables (4.3) and (4.4). However even this null hypothesis is always rejected. So for th e population as a whole the null hypothesis of full insurance is always rejected. T he literatu re on risk sharing in Indian villages has thus far focussed only on the ICRISA T villages. Tow nsend (1994) in his study of risk sharing in these villages finds significant evidence of risk sharing and concludes th a t households are able to insure against idiosyncratic incom e shocks fairly well. This result has been regarded as an exception in the literatu re on risk sharing in developing economies. Is it possible to generalize Tow nsend’s results to the rest of the country? T he results presented above answers this question in th e negative. For th e villages in Punjab, the null hypothesis of risk sharing is rejected and one can conclude th a t households in rural 6The average village level consumption c“t is the sample average from the observed households in each village. Since the sample consists of at most 23 households from each village, there is a measurement error problem here. 94 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. P u n jab are not able to fully insure th eir consum ption against idiosyncratic income shocks. U sin g V illa g e D u m m ies T he results for th e test using village dum m ies from equations (4.9) an d (4.10) (Ho : ( 3 - 2 = 0), are presented in Table (4.5). Ho is rejected in th e regression w ith income (b o th for th e first difference and th e grow th rate specification). However Ho is not rejected in th e regression w ith num ber of earners in the first difference specification b u t rejected for th e grow th rate specification. R eg r essio n b y L and H o ld in g C lass To exam ine if the results vary across land holding class, I classify th e population according to land holding. The purpose of this regression is to test w hether richer households are able to insure consum ption m ore th an the sm all farm ers and landless households. If this is true, then it im plies th a t the richer households have access to specific institutions, which enable them to insure successfully. In Tables (4.6) and (4.7) I present th e results of these regressions, classifying the d a ta on th e basis of land holding.7 In colum n (3) of Table (4.6) I present the F -ratio for th e test of th e jo in t hypothesis, (3i = l,/?2 = 0, (A y{vt is th e independent variable). Notice I cannot reject th e null hypothesis for th e large land owners (w ith land holding of 6.6 7There is insufficient data for households with land holding exceeding 14.5 hectares. Hence I could not run the regressions for this particular land class. 95 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. hectares and above). In colum n (5) of the sam e table I present th e F -ratio when the independent variable is th e change in th e num ber of earners in th e household. T he jo in t hypothesis is rejected for th e landless an d th e m arginal farm ers, w ith land holding not exceeding 1.0 hectares, (which is expected) but surprisingly also for the m edium farm ers w ith land holding 4.6 - 8.5 hectares. However the sim ple hypothesis is not rejected for all landed households. T urning to consum ption regressions on grow th rates (Table (4.7)), th e jo in t hy pothesis is not rejected for 3 o u t of th e 8 land classes when I use log(y,-vf) as the explanatory variable (the jo in t hypothesis is rejected for the landless, for those house holds in th e 1.0 — 6.5 hectares an d in the 4.6 — 6.5 hectares - see colum n (3)). On the o th er hand when I use A E{vt as th e independent variable, th e jo in t hypothesis is not rejected for all but th e landless and those households w ith 6.6-8.5 hectares (see c o lum n (5)). The im m ediate im plication is th at the larger farm ers are more successful in full consum ption insurance. Sm all farm ers are able to insure against changes in th e num ber of earners in the household (possibly insure against illness shocks and m igration shocks) b u t are unable to insure against fluctuations in total household income. The landless are left vulnerable in both respects. It would be interesting to exam ine w hat in stitu tio n s enable larger households to insure against such shocks and leave the landless vulnerable. B u t a m ore detailed exam ination of this n a tu re is not feasible using th e A R IS-N CA ER data. 96 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. R e g r e ssio n by D istric t To exam ine if th ere is regional variation in the ability o f households to insure con sum ption, I carry o u t the sam e test for consum ption insurance by classifying the d a ta on th e basis of th e d istricts of th e region. T he results are presented in Table (4.8) for th e first difference specification and in Table (4.9) for th e grow th rate spec ification. O nce again I ru n two sets of regressions for each, one using changes in gross household incom e and th e other as th e change in th e num ber of earners in the household as explanatory variables. T he F -ratio for the jo in t test /?i = l , / ? 2 = 0 are presented in colum ns (3) and (5) of th e two tables. Using Ay,,,* as the explanatory variable th e null hypothesis is n o t rejected for only households in th e districts Am ritsar and Jullunder (see colum n (3), T able (4.8)). O n th e o th er hand using A E{vt as th e explanatory variable, th e null hypothesis is accepted for all th e districts (see colum n (5), Table (4.8)). T urning to th e grow th ra te specification, I find th a t the the null hypothesis cannot be rejected for all districts b u t Ludhiana, for the second set of regressions (num ber of earners - see colum n (5), Table (4.9)). Using changes in household incom e as th e explanatory variable th e null hypothesis is not rejected for th e d istricts G urgaon, K apurthala, H issar and R ohtak (see colum n (3), Table (4.9)). 97 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. In stru m en ted V ariable E stim a tio n T able (4.10) compaxes the results of th e OLS and the IV (2SLS) estim atio n . Notice first, th e jo in t hypothesis (/?i = l,/? 2 = 0) is not rejected for th e larger landowners in b o th sets of estim ates. T he W ald test for th e equality of estim ates (3iv = Pols is also presented in Table (4.10).8 T he W ald statistic, com puted as in equation (4.13) is W = 0.9653. So W < x 2(2) and th e null hypothesis of equality o f estim ates from th e OLS and IV estim ation cannot be rejected. U sin g A ltern a tiv e M easu res o f C on su m p tion R esults are also reported for alternative m easures of consum ption. I use T otal House hold E xpenditure, Food Expenditure, N on-Food Expenditure E x p en d itu re on Con sum er D urables, Expenditure on Fuel and Lighting, E xpenditure on C lothing and Toiletry, E xpenditure on Education, M edical Expenditure, E x penditure on Services an d E xpenditure on E ntertainm ent as th e alternative measures of consum ption. In Table (4.11) results are reported for th e first difference specification. In c o lu m n s (2) and (3) I have the results from using the first difference in household in com e as an independent variable, and colum ns (4) and (5) presents th e results from using change in th e num ber of earners in th e household as an independent variable. N ote th e jo in t test of full insurance (change in household consum ption is unaffected by change in idiosyncratic household specific variables and is perfectly correlated 8The Wald Statistic is computed only for the entire sample. 98 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. w ith village level average consum ption) is not rejected for 4 out of th e 10 m easures of consum ption (consum er durables, fuel-light, education and services - see colum n (3) in T able (4.11)) and for 8 out of 10 m easures in column (5) (th e null hypothesis is rejected only for total expenditure an d clothing - see colum n (5) in T able (4.11)). Likewise th e sim ple test /32 = 0 is not rejected for 4 out of th e 10 m easures in the first regression (consum er durables, fuel-light, education and services - see colum n (2) in T able (4.11)) and for 7 out o f th e 10 m easures in the second regression - food expenditure, consum er durables, fuel-light, education, m edicine, services and en tertain m en t (see colum n (4) in T able (4.11)). R esults axe reported in Table (4.12) for th e corresponding grow th rate speci fication. For th e first specification (log(yiut) as th e explanatory variable) th e joint hypothesis, j3\ = l,/? 2 = 0, cannot be rejected i.e., the results are consistent w ith full insurance, for 4 out of the 10 m easures of consum ption - consum er durables, educa tion, services and entertainm ent (see colum n (3) in Table (4.12)). A sim ilar result is obtained for th e second regression w ith change in the num ber of earners, where the jo in t hypothesis is rejected for 5 out o f th e 10 alternative m easures of consum ption (consum er durables, education, m edicine, services and entertainm ent - see colum n (5) in Table (4.12)). Likewise the sim ple test /?2 = 0 is not rejected for 5 out of the 10 m easures in th e first regression (consum er durables, education, m edicine, services 99 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. and en tertain m en t - see colum n (2) in Table (4.12)) and for 6 out of the 10 m ea sures in th e second regression - consum er durables, fuel-light, education, m edicine, services and e n tertain m e n t (see column (4) in T able (4.12)). One im p o rtan t im plication of th e results presented above is th a t households are able to insure relatively well against changes in th e to ta l num ber of earners in th e household, b u t less against changes in to tal household income. Are the results sensitive to th e definition of incom e used? I also conduct a test using change in to tal non-labor incom e (A7Ttul or A log(7r,vt) as th e case m ay be) as th e idiosyncratic variable of interest in equations (4.5) and (4.7).9 T hese results are not presented. For the first difference specification th e results are exactly th e sam e (the null hypothesis /5i = 1 and /?2 = 0 is not rejected for 4 out of th e 10 m easures of consum ption). However for th e grow th ra te specification, the null hypothesis is not rejected for all but 2 of th e 10 m easures. T he growth rate specification therefore does much b e tte r in predicted consum ption insurance, particularly w hen we use non labor incom e as an explanatory variable. 4.4 Risk Pooling Across Villages A final issue th a t I w ant to discuss is risk pooling across villages in rural P unjab. Lim (1991) in his stu d y of risk sharing across villages using th e ICRISAT d ata finds th a t households are left vulnerable to aggregate village level risk, even though such 9The use of non-labor income is also motivated by the fact that in chapter 2 ,1 examine household response to shocks to non-labor income. 100 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. risk is insurable at th e regional level. However, he defines the region to com prise of th ree geographically separated villages, w hich implies th a t the pow er of his tests are extrem ely low. As I have argued in ch ap ter 1, 20 villages spread across 9 districts can be thought of as constituting a regional economy. I can now use th e same fram ew ork to throw some light on risk sharing across villages w ith in P unjab. There exists evidence in favor of such risk sharing. In Figure 4 I present th e deviation (from th e pooled average) of th e village level average income and consum ption. Average village level consum ption is alm ost always less volatile than th e average village level. T his result is reinforced in T able (4.14). Define c„t as the average consum ption in village u, in year t. To test risk pooling across villages I consider th e regression A C ut = 0o + /?iA c“ 4- yvt + £v,t+i (4-14) w here c“ = y j J2V H« °ivt is th e pooled average of consum ption and c^t = \ H i C ivt is th e village level average of household consum ption. If th ere is com plete risk sharing across villages, then /3i = 1 and (3i = 0. A test of incom plete risk sharing is given by th e weaker test = 0. T able (4.13) presents results from th e test of risk pooling across villages. If th ere is risk sharing across villages, average village consum ption should be correlated w ith th e pooled average consum ption and not affected by average incom e.1 0 T h e F -ratio for the joint hypothesis is presented 10 Alternatively I have the growth rate specification A log(cv() = /?0 + /f?iAcja + /3 2A log(yut) +£„, t+i (4-15) 101 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. in colum n (7). Note th a t the joint test of com plete risk shaxing across villages cannot be rejected, b o th in th e first difference specification (see equation (4.14)) and in th e grow th rate specification (see equation (4.15)). This has im p o rtan t policy im plications, particularly since th e null hypothesis of risk shaxing w ithin villages is rejected (see the results in Tables (4.3) and (4.4)). So even though th ere is no risk shaxing w ithin villages, if th e villages do pool, then it is possible to share risks in an effective m an n er. T he m agnitude of welfare gains obtained by such risk shaxing is left for future research. 4.5 Conclusion From th e consum ption regressions I find th a t for th e population as a whole th e null hypothesis of full consum ption insurance is rejected. This is very different from the results obtained using d a ta from the ICRISA T villages, where th e null hypothesis of full insurance cannot be rejected. W hen I subdivide the population according to landow nership, the null hypothesis is rejected for the landless households and the sm all farm ers. However, for the m edium and large farm ers th e null hypothesis of full insurance cannot be rejected. The landless axe left vulnerable and th e m edium and large farm ers axe insured in both cases. R ural institutions therefore seem to operate sim ilarly across regions. Further, there exists significant regional variation where c” t a = p-y Y2v Y li l°g(c«t/t) is the pooled average of log consumption and log(cut) = j Iog(c,-„t) is the village level average of the log o f household consumption. 102 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. in the ability of households to insure consum ption against incom e changes. Finally, th ere exists significant evidence in favor of risk sharing across villages in the region. In th e crop year 1981-82 N C A ER re-surveyed th e 1970-71 households, th e Rural Economic D evelopm ent Survey (RED S). T h e A R IS-N CA ER an d th e REDS d ata com bine to form tw o alternative panels - one for th e green revolution years and one for the post green revolution years. It will be interesting to com pare th e ability of households to pool risk in th e two separate periods. This exercise could provide im portant inform ation on the developm ent of rural institu tio n s in light of overall economic growth. B u t this is left for future research. 103 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4.6 Appendix: Tables T able 4.1: S um m ary S tatistics on Incom e and C onsum ption (Classified by Land H olding) Land Class Income C onsum ption M ean C.V. M ean C.V. (1) (2) ( 3 ) ( 4 ) ( 5 ) All Households 5903.99 (82.0745) 4489.73 (59.6842) No Land 2621.02 (69.4623) 2597.47 (60.3984) < 1.0 3588.74 (74.5058) 3155.53 (46.1270) 1.0-2.5 4763.27 (57.1891) 4169.42 (47.1459) 2.6-4.5 6009.66 (57.8592) 4950.09 (49.2807) 4.6-6.5 8574.83 (62.4588) 5482.87 (45.8450) 6.6-8.5 9174.18 (57.0113) 6097.27 (46.9809) 8.6-10.5 9719.50 (67.6237) 6123.85 (53.2722) 10.6-14.5 8859.35 (64.4592) 5880.18 (44.3401) > 14.5 13233.58 (53.6255) 8444.22 (36.9011) M ean in FIs. 104 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 4.2: Sum m ary Statistics on Income and Consumption (Classified by District) Land Class Incom e C onsum ption M ean C.V. M ean C.V. (1) (2) (3) (4) (5) All Households 5903.99 (82.0745) 4489.73 (59.6842) Ludhiana 5268.08 (74.0672) 4291.28 (63.6633) G urgaon 3484.58 (82.7348) 3411.68 (53.8740) Jin d 3485.01 (71.4727) 3154.28 (62.5224) K apurthala 5811.88 (89.1773) 4378.03 (73.8805) Hissar 6622.68 (78.4857) 4981.91 (48.4722) R ohtak 5604.74 (71.6039) 4687.79 (55.0829) A m ritsar 6533.57 (72.5008) 4722.19 (61.7965) Ferozpur 8975.73 (75.1912) 5265.20 (52.4641) Ju ll under 6809.87 (75.7389) 4872.38 (54.7626) M ean in R s. 10-5 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 4.3: Consumption Regressions: First Differences C onsum ption Intercept Cv t A yivt A Eivt F -R atio (1) (2) (3) (4) (5) (6) -1.37 0.98 0.18 na 43.56 (0.84) (0.07) (0.02) A art 0.06 1.01 n a 3.63 3.38 (0.84) (0.08) (1.40) S ta n d a rd E rro rs in P arenthesis *: D o n o t R eject H q : 01 = 1,02 = 0 a t 95% D o n o t R eject H q i : / ? 2 = 0 a t 95% Table 4.4: Consum ption Regressions: Growth Rates C onsum ption Intercept A log(ytV f) A B{vt F -R atio (1) (2) (3) (4) (5) (6) A log(c,„t) -0.03 0.91 0.15 na 29.28 (0.02) (0.06) (0.02) A log(c,„t ) 0.00 0.96 na 0.09 5.81 (0.02) (0.07) (0.03) S ta n d a rd E rro rs in P arenthesis D o n o t R eject Ho : 0 i = 1 ,0 ? = 0 a t 95% D o n o t R eject Ho : (3 -2 — 0 a t 95% 106 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 4.5: Consum ption Regressions: Using Village Dummies Cons. M easure (1) A j/t'v t (2) A E i v t (3) Alog(y,vt) (4) A i w (5) Ac,'^ 0.21 3.00* na na (0.02) (1.57) (na) (na) A log(civt) na na 0.21 0.08 (na) (na) (0.02) (0.03) S ta n d a rd E rro rs in P aren th esis *: Do n o t R eject H q : (3^ = 0 a t 95% Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 4.6: Consumption. Regressions by Land Class: First Differences Land Class A y ivt F-R atio A E i v t F-R atio (1) (2) ( 3 ) ( 4 ) ( 5 ) A ll Households 0.18 43.56 3.63 3.38 (0.02) (1.40) No Land 0.42 32.00 3.51 8.85 (0.06) (1.28) < 1.0 0.21** 4.52 4.61** 3.46 (0.12) (3.87) 1.0-2.5 0.28 12.65 2.00** 0.79* (0.06) (4.24) 2.6-4.5 0.22 9.47 8.60** 2.28* (0.05) (5.33) 4.6-6.5 0.05** 3.77 3.54** 3.41 (0.05) (7.91) 6.6-8.5 0.11** 1.67* 90.23 10.99 (0.07) (19.52) 8.6-10.5 0.12** 2.78* 0.84** 1.60* (0.08) (6.34) 10.6-14.5 0.12** 1.03* -2.02** 0.02* (0.08) (13.59) S ta n d a rd E rrors in P aren th esis *: D o n o t R eject H o • P i = 1 ,/3j = 0 a t 95% Do n o t R eject Ho : f e = 0 a t 95% 10S Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 4.7: Consumption Regressions by Land Class: Growth Rates Land Class A log(yivt) F-R atio A Eivt F-R atio (1) (2) ( 3 ) ( 4 ) ( 5 ) All Households 0.15 29.28 0.09 5.81 (0.02) (0.03) No Land 0.21 12.51 0.11 3.90 (0.04) (0.04) < 1.0 -0.04** 0.19* 0.11** 0.48* (0.12) (0.13) 1.0-2.5 0.23 9.43 0.08** 1.45* (0.06) (0.10) 2.6-4.5 0.11 3.22 0.12** 0.77* (0.05) (0.10) 4.6-6.5 0.12 4.37 -0.05** 0.63* (0.04) (0.12) 6.6-8.5 0.19 4.28 0.84 4.29 (0.07) (0.32) 8.6-10.5 0.10** 1.24* 0.07** 0.75* (0.08) (0.09) 10.6-14.5 0.05** 1.14* -0.00** 0.81* (0.06) (0.23) S ta n d a rd E rrors in P a re n th e sis D o n o t R eject H q : 0 1 — l< 0 7 — 0 a t 95% D o not R eject H q -.0 2 = 0 a t 95% Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 4.8: Consumption Regressions by District: First Differences D istrict A yivt F -R atio A Eivt F-Ratio (1) (2) ( 3 ) ( 4 ) ( 5 ) All Households 0.18 43.56 3.63 3.38 (0.02) (1.40) L udhiana 0.24 8.74 9.42 2.45* (0.29) (4.26) G urgaon 0.33 10.31 2.43** 0.68* (0.07) (2.09) Jin d 0.50 24.60 2.95** 0.94* (0.07) (2.15) K ap u rth ala 0.29 8.61 7.03** 1.23* (0.07) (4.48) Hissax 0.19 9.90 -14.71** 0.77* (0.04) (11.88) R ohtak 0.23 4.30 7.79** 1.12* (0.08) (5.21) A m ritsar 0.02** 0.03* -1.28** 0.01* (0.09) (8.15) Ferozpur 0.21 19.04 -8.14** 0.48* (0.33) (8.31) Ju ll under 0.049** 0.14* 9.43** 0.69* (0.09) (8.04) S ta n d a rd E rro rs in P arenthesis D o n o t R eject H q : 0 \ = — 0 a t 95% D o n o t R eject H q : = 0 a t 95% 110 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 4.9: Consumption. Regressions by District: Growth Rates D i s t r i c t A l o g (yivt) F - R a t i o A E ivt F - R a t i o (1) (2) ( 3 ) ( 4 ) ( 5 ) A l l H o u s e h o l d s 0.15 29.28 0.09 5.81 (0.02) (0.03) L u d h i a n a 0.28 10.78 0.24 3.83 (0.06) (0.09) G u r g a o n 0.15 2.19* 0.09** 1.15* (0.07) (0.06) J i n d 0.40 9.09 0.07** 0.71* (0.09) (0.06) K a p u r t h a l a 0.12** 1.79* 0.15** 1.55* (0.06) (0.08) H i s s a r 0.07 2.17* -0.28** 0.94* (0.03) (0.21) R o h t a k 0.18** 1.64* 0.17** 0.97* (0.10) (0.12) A m r i t s a r 0.26 5.65 -0.09** 0.26* (0.08) (0.14) F e r o z p u r 0.42 23.77 -0.19** 0.08* (0.06) (0.15) J u l l u n d e r 0.32 4.98 0.30 2.27* (0.10) (0.14) S ta n d a rd E rro rs in P aren th esis *: D o not R eject Ho : P l = 1,02 = 0 a t 95% Do n o t R eject Ho : 02 = 0 a t 95% Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. T able 4.10: C onsum ption Regressions by Land Class: Com parison of IV and OLS E stim ation IV E stim ation OLS E stim ation Land Class A j/iu £ F -R atio Ay,-„t F-R atio (1) (2) ( 3 ) ( 4 ) ( 5 ) All Households'1 " 0.15 5.29 0.18 43.56 (0.05) (0.02) No Land 0.12** 5.84 0.42 32.00 (0.13) (0.06) < 1.0 0.19** 3.49 0.21** 4.52 (0.17) (0.12) 1.0-2.5 0.31 6.39 0.28 12.65 (0.09) (0.06) 2.6-4.5 0.24 Ol to 00 0.22 9.47 (0.08) (0.05) 4.6-6.5 -0.01** 3.30 0.05** 3.77 (0.09) (0.05) 6.6-8.5 0.09** 0.86* 0.11** 1.67* (0.08) (0.07) 8.6-10.5 0.21** 1.22* 0.12** 2.78* (0.12) (0.08) 10.6-14.5 0.11** 0.25* 0.12** 1.03* (0.15) (0.08) + : W = 0.9653 S ta n d a rd E rro rs in P aren th esis *: D o n o t R eject Ho : i3j = 1 ,0 2 = 0 a t 95% Do n o t R eject Ho : .0 2 — 0 a t 95% 112 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 4.11: Consum ption Regressions For A lternative M easures of Consum ption F irst Differences C onsum ption M easure A yttrt F-R atio AE{vt F -R atio (1) (2) (3) (4) (5) Total E xpenditure 0.18 43.56 3.63 3.38 (0.02) (1.39) Food E xpenditure 0.13 75.21 1.52** 1.88* (0.01) (0.79) Non Food E xpenditure 0.05 5.69 2.10 1.95* (0.02) (1.07) Consum er Durables -0.00** 0.004* 0.09** 0.18* (0.002) (0.16) Fuel-Light 0.001** 1.78* 0.06** 1.67* (0.001) (0.03) C lothing 0.02 12.49 0.86 9.10 (0.003) (0.20) Education -0.002** 0.38* 0.08** 0.10* (0.003) (0.18) M edicine 0.01 8.00 0.12** 0.34* (0.002) (0.14) Services -0.001** 0.01* 0.04** 0.02* (0.003) (0.17) E ntertainm ent 0.002 11.42 0.04** 1.39* (0.00) (0.03) S ta n d a rd E rro rs in P aren th esis *: Do n o t R eject Ho : = l , / ? 2 = 0 a t 95% Do n o t R eject Ho : /Tj = 0 a t 95% 113 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 4.12: C onsum ption Regressions For A lternative M easures of C onsum ption Growth Rates C onsum ption M easure (1) A log(j/tVi) (2) F -R atio AEivt (3) (4) F-Ratio (5) Total E xpenditure 0.15 29.28 0.09 5.81 (0.02) (0.03) Food E x penditure 0.15 33.15 0.08 5.13 (0.02) (0.03) Non Food E xpenditure 0.13 8.47 0.12 3.69 (0.03) (0.05) C onsum er D urables 0.05** 1.69* 0.03** 0.47* (0.03) (0.04) Fuel-Light 0.11 7.28 0.11** 3.70 (0.03) (0.04) C lothing 0.13 6.41 0.27 15.97 (0.04) (0.05) E ducation 0.02** 0.27* -0.004** 0.16* (0.04) (0.05) M edicine 0.11** 3.97 0.11** 2.17* (0.04) (0.06) Services 0.08** 2.19* -0.04** 0.70* (0.04) (0.06) E n tert ainm ent -0.004** 0.22* -0.04** 0.63* (0.03) (0.05) S ta n d a rd E rro rs in P arenthesis *: Do n o t R eject Ho : = l . f t = 0 a t 95% Do n o t R eject Ho : 07 = 0 a t 95% 114 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 4.13: Risk Pooling Across Villages Consum ption Intercept (1) (2) A c“ ( 3 ) A c '“ ( 4 ) < 1 i2. A log(j/vt) (6) F-R atio ( 7 ) A c* -0.71 0.90 na 0.11** n a 0.78* (2.12) (0.68) (0.10) A log(cvt) -0.01 na 0.81 na 0.08** 0.50* (0.05) (0.17) (0.07) S ta n d a rd E rro rs in P aren th esis Do N ot R eject H q : 0 i = 1,&2 = 0 a t 95% Do N ot R eject H o : 02 = 0 a t 95% Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 4.14: Volatility of Income and Consumption Village Incom e C onsum ption (1) (2) (3) All Households 82.074 59.684 D alla 54.774 42.841 H eran 79.562 62.517 Jassow al 64.557 78.473 B arm i 57.406 64.453 B hadus 84.001 59.172 N otki 70.861 43.780 K alyat 71.473 62.522 N an ajath an 116.225 103.349 K h aju rla 66.271 48.112 B alta 63.978 39.632 B hagana 65.951 50.648 B hojraj 105.545 51.238 N ayabano 57.989 44.788 S ahiyatera 76.282 62.668 M akowal 67.324 63.811 M ohanbhandaria 72.177 57.430 B akhushah 57.004 40.991 B annaw ala 85.256 57.053 K otla 67.969 49.105 Satowali 82.584 59.754 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. m fO F igure 4: R isk P o o lin g A cross V illa g es Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Chapter 5 Conclusion M y thesis adds to th e literature on th e role of institutions in enabling households in rural In d ia sm ooth consum ption when th eir incom e is subject to different kinds of shocks. Using d a ta from two different regions in India (the ICRISAT villages and villages in P unjab) I exam ine the following issues. F irst, I exam ine how households in th e ICRISA T villages sm ooth consum ption to th e extent th a t they do. Second, I ask w hat type of institutions enable households to sm ooth consum ption against aggregate village level shocks? Finally I exam ine w hether the results on risk and insurance for th e ICRISAT villages can b e generalized, by extending the test of consum ption insurance to a region of India nam ely rural Punjab. I find th a t depending on their access to credit maxkets, households differ in their response to sim ilar shocks to income. M edium and large farm ers who have un restricted access to credit (both form al and inform al) use institutions th a t m im ic state-contingent transfers to insure against incom e shocks. Small farm ers and land less households are excluded from th e credit m arkets. However small farm ers are 118 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. a b l e t o s m o o t h , c o n s u m p t i o n , b y m a k i n g c o m p e n s a t i n g c h a n g e s i n m a r k e t l a b o r s u p p l y - t h e y r e s p o n d t o n e g a t i v e i n c o m e s h o c k s b y w o r k i n g l e s s o n t h e i r o w n f a x m a n d w o r k i n g m o r e i n t h e d a i l y w a g e l a b o r m a x k e t . T h e l a n d l e s s a r e u n a b l e t o m a k e t h i s c o m p e n s a t i n g c h a n g e a n d a r e l e f t v u l n e r a b l e t o i n c o m e s h o c k s . The fact th a t som e households (in particular those who axe restricted in their paxticipation in th e village financial maxkets) use paxticipation in th e daily wage labor maxket in response to incom e shocks has im portant im plications for th e design of institutions th a t will enable households to sm ooth consum ption against aggregate village level shocks. T h e prevalent m echanism to deal w ith such risk is through con servative production choices and through storage. Such a technique is costly in term s of lower m ean o u tp u t. I argue th a t an optim ally designed work-fare program , like the EGS can successfully insure households against aggregate village level risk. So I exam ine th e insurance aspects of such a program using a com putational experim ent in an economy w here agents face exogenous income shocks and axe unable to insure themselves through private credit maxkets. T he only way they can self-insure is sav ings in th e form of a non-interest bearing asset. The model is calibrated using d ata from the two ICRISA T villages in the sta te of M aharashtra, which had a functioning Em ploym ent G uarantee Schem e (EGS) in the period 1979 - 1984. W hen agents are paid the optim al wage, th ey do not need to hold any asset as precautionary savings. All insurance in this case is provided by the EGS. In general, because of its insurance aspect, the optim al wage ra te in the EGS is found to be higher th an th e m arginal 119 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. product of labor. T h e o p tim al wage and the welfare gains o f th e program depend on how productive th e EG S is, relative to the private sector. Because of th e n atu re of production in th e EG S, it is not possible to obtain a tru e m easure of th e m arginal product of labor em ployed in th e EGS. T he welfare gains o f paying the optim al wage as opposed to a wage equal to th e m arginal product of labor in th e EGS varies from 0.01% of G D P to 2.63% of G D P, depending on th e labor productivity of the EGS. T he actual wage paid by th e EGS during the years 1979-1984 yields a welfare level th a t is lower th a n paying a wage equal to the m arginal pro d u ct of labor. F urther I find th a t th e 1988 increase in wages in the EGS actually decreased welfare, and an even greater increase in wages is necessary to increase th e welfare of th e population. My thesis also extends th e test o f consum ption insurance to a different region of India - nam ely ru ral P u n jab . I exam ine the success of households in rural Punjab in insuring consum ption against incom e fluctuation in th e green revolution years (1968-69 - 1970-71). I find th a t for the population as a whole the null hypothesis of full consum ption insurance is rejected. T he qualitative properties of these results are very different from those obtained from using d a ta from villages in A ndhra Pradesh and M ah arash tra, where there is evidence of significant risk sharing by households against idiosyncratic incom e shocks. T here is no such such evidence for households residing in P unjab. W hen the population is subdivided according to landow nership, th e null hypothesis is rejected for th e landless households and sm all farm ers. However, for th e m edium and large farm ers th e null hypothesis of full 120 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. insurance cannot be rejected. Further, there exists significant regional variation in th e ability of households to insure consum ption against income changes. I also conduct a test of full insurance at th e regional level and test w hether there is risk sharing across villages. Lim (1991) conducts a test of risk sharing across villages in th e IC R ISA T region and finds th a t households are left vulnerable to shocks w hich are insurable a t the regional level through inter village level tra d e and proper financial interm ediation. He defines th e regional economy to consist of th e three villages A urepalle, Shirapur and K anzara pooled together. These th ree villages are sp atially separated and do not form a regional economy. On th e o th er hand pooling th e 20 villages spread over 9 d istricts in P unjab (and H aryana) does define regional economy. I test for risk sharing across villages in ru ral P unjab, by placing each village in a larger regional econom y and find th at there is significant evidence of risk sharing across villages. So while the null hypothesis of com plete risk sharing w ithin each village is rejected, th ere is evidence th at if th e villages pool together th en th ey can insure against aggregate incom e risk. 121 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Chapter 6 Reference List A lderm an, EL & Paxson, C. (1992), ‘Do th e Poor Insure? A Synthesis of the Liter atu re on R isk and Consum ption in Developing Countries’, Mimeo, Princeton Uni versity. A ltonji, J. G. Sz Siovv, A. (1987), ‘Testing th e Response of C onsum ption to Income Changes w ith (Noisy) Panel D a ta ’, Quarterly Journal o f Economics pp. 293-328. A ltug, S. Sz Labadie, P. (1994), Dynamic Choice and Asset M arkets, A cadem ic Press. A ltug, S. Sz M iller, R. (1990), ‘Household Choice in E quilibrium ’, Econometrica 58, 543-570. Besley, T . (1995), ‘N onm arket Institutions for Credit and Risk Sharing in Low- Incom e C ountries’, Journal of Economic Perspectives 9(3), 115— 128. Bhende, M. J. (1986), ‘Credit M axkets in R ural South India’, Economic and Political Weekly X X I(38-39), A119-A124. Bhende, M . J., W alker, T. S., Lieberm an, S. S. & V enkataram , J . V. (1992), ‘EGS and th e Poor: Evidence from Longitudinal Village Studies’, Economic and Political Weekly pp. A19— A22. Carroll, C. D. (1990), ‘Buffer Stock Saving and the Life C ycle/P erm anent Income H ypothesis’, Mimeo, Johns Hopkins University. Carroll, C. D. (1991), ‘Buffer Stock Saving and the P erm anent Incom e H ypothesis’, Mimeo, Johns Hopkins University. C arroll, C. D. (1992), ‘The Buffer Stock Theory of Saving: Som e M acroeconomic Evidence’, Brookings Papers on Economic Activity (2), 61-156. 122 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. C arroll, C. D. (1996), ‘Buffer Stock Saving: Some T heory’, Mimeo, Johns Hopkins University. C hakravarty, S. (1987), Development Planning: The Indian Experience, Oxford Uni versity Press. C M IE (1993), Basic S tatistics R elating to th e Indian Economy, Technical rep o rt, C enter for M onitoring Indian Economy, M um bai, India. C ochrane, J. H. (1991), ‘A Sim ple Test of C onsum ption Insurance’, Journal o f Po litical Economy 99(5), 957-976. Cooley, T . F. & P rescott, E. C. (1995), Econom ic G row th and Business Cycles, in T . F. Cooley, ed., ‘Frontiers of Business Cycle Research’, Princeton U niversity Press, pp. 1-38. C rucini, M. J. (1996), ‘On International and N ational Dimensions of Risk S haring’, Mimeo, Ohio State University. D eaton, A. (1990), ‘O n Risk, Insurance and Intra-village C onsum ption Sm oothing’, Mimeo, Princeton University. D eaton, A. (1992), ‘Saving and Incom e Sm oothing in Cote D’Ivoire’, Mimeo, Prince ton University. Dreze, J. P. & : M ukherjee, A. (1989), Labour C ontracts in Rural India: Theories and Evidence, in S. Chakravarty, ed., ‘T he B alance betw een Industry and A griculture in Econom ic D evelopm ent’, Vol. 3, M acm illan, pp. 233-265. Esw aran, M. & : Kotwal, A. (1989), ‘C redit As Insurance in Developing Econom ies’, Journal o f Development Economics 31, 37-53. Foster, A. D. & Rosenzweig, M. R. (1996a), Household Division, Inequality and R ural Econom ic G row th’, Mimeo, University of Pennsylvania. Foster, A. D. & Rosenzweig, M. R. (19966), ‘Technical Change and H um an-C apital R eturns and Investm ent: Evidence from th e G reen Revolution’, American Economic Review 86(4), 931-953. G aiha, R. (1996), ‘How D ependent Are th e R ural Poor on the Em ploym ent G uar antee Schem e in India’, Journal o f Development Studies 32(5), 669-694. G reene, W . H. (1990), Econometric Analysis, international edn, M acm illan P ublish ing Com pany. 123 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. H ansen, G. D. (1985), ‘Indivisible L abor an d th e Business Cycle’, Journal o f Mon etary Economics 16, 309-327. H ansen, G. D. & Imrohoroglu, A. (1992), ‘T h e Role of U nem ploym ent Insurance in an Econom y w ith Liquidity C onstraints and Moral H azard’, Journal o f Political Econom y 100(1), 118-142. Im rohoroglu, A., Im rohoroglu, S. & Joines, D. (1993), ‘A N um erical A lgorithm For Solving M odels W ith Incom plete M ark ets’, International Journal O f Supercomputer Applications 7(3), 212-230. Jacoby, H. Skoufias, E. (1994a), ‘T esting Theories of C onsum ption B ehavior Using Inform ation on Aggregate Shocks: Incom e Seasonality and R ainfall in R ural India’, Mimeo, University o f Rochester. Jacoby, H. & Skoufias, E. (19946), ‘R isk, F inancial M arkets and H um an C apital in a D eveloping C ountry’, Mimeo, University o f Rochester. K ochar, A. (1995a), ‘Explaining H ousehold Vulnerability to Idiosyncratic Incom e Shocks’, American Economic Review 8 5 (2 ), 159— 164. K ochar, A. (19956), ‘Hours-of-Work Responses to Idiosyncratic A gricultural Shocks: E m pirical Evidence From R ural In d ia’, Mimeo, Stanford University. K ydland, F. E. & : Prescott, E. C. (1996), ‘T he C om putational E xperim ent: An Econom etric Tool’, Journal o f Economic Perspectives 10(1), 69— 85. Lam b, R . L. (1996), Off-Farm Labor Supply and Fertilizer Use, Technical R eport 49, Federal Reserve Board, W ashington D .C. Ligon, E. (1996), ‘Risk-Sharing and Inform ation: Theory and M easurem ent in Vil lage E conom ies’, University o f California, Berkeley Lim, Y . (1991), ‘Disentangling P erm an en t Income and Risk Sharing: A G eneral E quilibrium Perspective on C redit M arkets in R ural South In d ia’, Mimeo. Lim, Y . & Townsend, R. M. (1996), ‘C urrency: Theory and M easurem ent in Village E conom ies’, Mimeo, University o f Chicago. M ace, B. J . (1991), ‘Full Insurance in th e Presence of Aggregate U n certain ty ’, Jour nal o f Political Economy 99(5), 928— 956. M addala, G. S. (1983), Limited Dependent and Qualitative Variables in Economet rics, E conom etric Society M onographs, C am bridge U niversity Press. 124 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. M orduch, J. (1990), ‘Risk P roduction and Saving: Theory and Evidence From Indian H ouseholds’, Mimeo, Harvard. University. M orduch, J. (1995), ‘Incom e Sm oothing and C onsum ption Sm oothing’, Journal o f Economic Perspectives 9(3), 103— 114. N C A E R (1975), Changes in R ural Incom e in India: 1968-69 1969-70 1970-71, Tech nical rep o rt, N ational Council of A pplied Econom ic Research, New Delhi. N elson, J . A. (1994), ‘On Testing for Full Insurance Using Consum er E xpenditure Survey D a ta ’, Journal o f Political Economy 102(2), 384-394. Paxson, C. & C haudhuri, C. (1994), ‘C onsum ption Sm oothing and Incom e Season ality in In d ia’, Mimeo, Princeton University. Rao, S. V. (1984), ‘R ural Labour: Case S tudy of a K arnataka Village’, Economic and Political Weekly 19(18), 766— 776. Ravallion, M. (1991), ‘Em ploym ent G uarantee Schemes: Are They a Good Idea?’, The Indian Economic Journal 39(2), 50— 65. R avallion, M ., D a tt, G. & C haudhuri, S. (1993), ‘Does M aharashtra’s EGS G uaran tee Em ploym ent? Effects of the 1988 W age Increase’, Economic Development and Cultural Change 41(2), 251-276. Rogerson, R. (1988), ‘Indivisible Labor, Lotteries and E quilibrium ’, Journal o f Mon etary Economics 21, 3-16. Rose, E. (1995), ‘Ex A nte and Ex Post Labor Supply Responses to Risk in a Low Incom e A rea’, Mimeo, University of Washington, Seattle. Rosenzweig, M. R. (1988), ‘Risk, Im plicit C ontracts and the Fam ily in R ural Areas of Low-Income C ountries’, Economic Journal 9 8, 1148-1170. Rosenzweig, M. R. & c Binswanger, H. (1993), ‘W ealth, W eather Risk and th e Com po sition an d P rofitability of A gricultural Investm ents’, Economic Journal 103, 56-78. Rosenzweig, M. & W olpin, K. (1993), ‘C redit M arket Constraints, C onsum ption Sm oothing, and th e A ccum ulation of D urable Assets in Low-Income C ountries: In vestm ents in Bullocks in India’, Journal o f Political Economy 101(2), 223-244. Singh, R. P., Binswanger, H. P. & : Jo d h a, N. S. (1985), M anual of Instructions For Econom ic Investigators in ICRISA T’s Village Level Studies, Technical report, IC R ISA T, P atancheru,A ndhra Pradesh 502324, India. 125 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Townsend, R. M. (1994), ‘Risk and Insurance in Village India’, Econometrica 6 2 (3 ), 539- 591. Townsend, R. M. (1995), ‘Consum ption Insurance: A n Evaluation of Risk B earing Systems in Low-Income Econom ies’, Journal o f Economic Perspectives 9(3), 83— 102. Udry, C. (1990), ‘C red it M arkets in N orthern Nigeria: C redit as Insurance in a R ural Economy’, The World Bank Economic Review 4(3), 251-269. Udry, C. (1994), ‘R isk and Insurance in a R ural C redit Maxket: An Em pirical In vestigation in N orthern N igeria’, Review o f Economic Studies 61, 495-526. W alker, T. S. & R yan, J. G. (1990), Village and Household Economies in In d ia ’ s Semi-arid Tropics, T h e Johns Hopkins U niversity Press, Baltim ore and London. Zim m erm an, F. & : C axter, M. R. (1996), ‘R ethinking th e Dem and for In stitu tio n al Innovation: Land R ights and Land Maxkets in th e W est African Sahel’, Mimeo, Stanford University. 126 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Inflation stabilization, monetary policy instruments and borrowing constraints

PDF

Essays on the transition from centrally-planned to market economy in Central and Eastern Europe

PDF

Essays on social security

PDF

Context is everything: An empirical study of institutions, instability and economic growth

PDF

Inter-temporal allocation of human capital and economic performance

PDF

Essays on agrarian institutions

PDF

Computer-based scenarios in strategy formulation: The case of a large insurance company

PDF

Essays on consumption behavior, economic growth and public policy

PDF

A life cycle model of fertility choice

PDF

A contingent claims analysis of the equity interest in property-casualty insurance companies

PDF

An evolution in listening: An analytical and critical study of structural, acoustic, and phenomenal aspects of selected works by Pauline Oliveros

PDF

"Vesperae solennes de Confessore", K.339, W. A. Mozart: A study in performance and analysis

PDF

Accommodation in small groups: Patterns and consequences of adjustments in group member communication style over time

PDF

How financial statements in Mexican firms reflect changes in the financial/economic environment from 1978 to 1996

PDF

Collective action and institutions in self-governing residential communities in Seoul, Korea

PDF

An investigation of information requirements determination and analogical problem solving

PDF

Categories of duty and universalization in Kant's ethics.

PDF

The Pricing Of Initial Public Offering And Privatizations: An Influence-Seeking Theory And An Empirical Investigation With New Issues In The Nigerian Equity Market

PDF

An investigation of the distribution and mechanisms of deposition of exogenous cholesterol in the rat

PDF

A study of the labor philosophy developed within the Church of Jesus Christ of Latter-day Saints

Asset Metadata

Creator Maitra, Pushkar (author)

Core Title An investigation of consumption, insurance and village institutions in India

Degree Doctor of Philosophy

Degree Program Economics

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag economics, labor,Economics, Theory,OAI-PMH Harvest

Language English

Contributor Digitized by ProQuest (provenance)

Advisor [illegible] (committee chair), Betts, Caroline (committee member), Imrohoroglu, Ayse (committee member), Imrohoroglu, Selo (committee member), Joines, Doug (committee member), Nugent, Jeff (committee member)

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c17-292786

Unique identifier UC11350101

Identifier 9816050.pdf (filename),usctheses-c17-292786 (legacy record id)

Legacy Identifier 9816050.pdf

Dmrecord 292786

Document Type Dissertation

Rights Maitra, Pushkar

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA