Spatial Heterogeneity and Population Mobility in India. Jajati Keshari Parida S Madheswaran

Size: px
Start display at page:

Download "Spatial Heterogeneity and Population Mobility in India. Jajati Keshari Parida S Madheswaran"

Transcription

1 Spatal Heterogenety and Populaton Moblty n Inda Jajat Keshar Parda S Madheswaran

2 ISBN , Copyrght Reserved The Insttute for Socal and Economc Change, Bangalore Insttute for Socal and Economc Change (ISEC) s engaged n nterdscplnary research n analytcal and appled areas of the socal scences, encompassng dverse aspects of development. ISEC works wth central, state and local governments as well as nternatonal agences by undertakng systematc studes of resource potental, dentfyng factors nfluencng growth and examnng measures for reducng poverty. The thrust areas of research nclude state and local economc polces, ssues relatng to socologcal and demographc transton, envronmental ssues and fscal, admnstratve and poltcal decentralzaton and governance. It pursues frutful contacts wth other nsttutons and scholars devoted to socal scence research through collaboratve research programmes, semnars, etc. The Workng Paper Seres provdes an opportunty for ISEC faculty, vstng fellows and PhD scholars to dscuss ther deas and research work before publcaton and to get feedback from ther peer group. Papers selected for publcaton n the seres present emprcal analyses and generally deal wth wder ssues of publc polcy at a sectoral, regonal or natonal level. These workng papers undergo revew but typcally do not present fnal research results, and consttute works n progress.

3 SPATIAL HETEROGENEITY AND POPULATION MOBILITY IN INDIA Jajat Keshar Parda and S Madheswaran Abstract Moblty s one of the mportant aspects of human nature, whch s often guded by socoeconomc, poltcal as well as envronmental factors. The nature, pattern and drecton of populaton moblty may vary across the space. The dynamcs of nternal mgraton n Inda plays an mportant role n the process of economc development and socal transformaton and shows an ncreasng trend of rural to urban flow over the years. At the same tme, t shows fallng trends n all other streams of mgraton and are regsterng negatve growth rates as well. The determnants of rural to urban mgraton nclude a set of soco-economc, demographc, geographcal and envronmental varables. The emprcal results establsh the Gravty Model of mgraton n Inda; where as the Harrs-Todaro Model of rural urban mgraton has lmted applcablty n both nter-state and ntra-state mgraton n Inda. 1. Introducton The concept of nternal mgraton now connotes much wder dmenson that vares from daly commutng to nearby places on one hand, to permanent shft of resdence to dstant places on the other. The nature and pattern of mgraton vares from one socal group of mgrants to another because of the fact that the rural mgrants are not a homogenous group (Mtra and Murayama, 2008). Rural to urban mgraton has attracted the attenton of academcans as well as the polcymakers throughout the world n recent years, because of ts wde range of soco-economc, poltcal, demographc, ecologcal and envronmental mplcatons. It plays an mportant role n the process of economc development and socal transformaton because the shftng of the work force from prmary to secondary and tertary sectors that result n structural change. The mprovement n nfrastructure and communcaton accompanyng economc development could result n ncreased populaton moblty. But from the earler studes (Kundu and Gupta, 1996; Sngh, 1998; and Srvastava, 1998), t s evdent that mgraton as a percentage of total populaton has been declnng up to 1991 census, where as the rural to urban mgraton s showng an ncreasng trend over the perod 1971 to 2001(Lusome & Bhagat, 2006) The recent report of Unted Natons Development Programme (UNDP, 2009) enttled Moblty and Human Development stated that populaton shft from vllages to ctes s natural and should be encouraged and that t results n concentraton of producton, moblty of people and economc ntegraton to lft rural people out of poverty n Inda. Hence, any long-term plan amng at an ncrease n labour productvty and rural employment opportuntes should consder the possblty of mgraton of a relatvely large proporton of populaton away from areas wth relatvely low agronomc potental (Dhar, 1980; and Kundu & Sarang, 2007). The avalablty of good agrcultural land contnues to be one of the most powerful economc factors determnng the magntude and drecton of populaton mgraton. The Research Scholar, Centre for Economc Studes and Polcy, ISEC, Bangalore. Professor, Centre for Economc Studes and Polcy, ISEC, Bangalore. We are thankful to Prof Abdul Ajz, Prof K S James, Prof R S Deshpande, Prof M R Narayana, Smt B P Van, Prof Meenaksh Rajeev and Dr K Gaythr for ther valuable suggestons. Thanks are also due to the unknown referees for ther thought provokng comments and suggestons. 1

4 agrcultural unemployment and underemployment, lower agrcultural wage and poverty are the major factors pushng labourers towards job opportuntes n urban areas, whle the avalablty of employme nt opportuntes, relatvely hgher wages, better educatonal opportuntes, better avalablty of transportaton and nfrastructure etc. are the major factors n the rural-urban mgraton (Harrs and Todaro, 1970; Barro & Sala--Martn, 1991; Gupta, 1993; Bhattacharya, 1998; Pekkala and Tervo, 2002; Bhattacharya, 2002; and Josh & Lobo, 2003; Andrenko and Serge, 2003). Therefore, n ths context, an attempt has been made to examne the broad trends and patterns of nternal mgraton and to study the determnants of rural to urban mgraton n Inda. The paper s organsed as follows. Secton 2 provdes a bref revew of lterature. Secton 3 gves the data sources used n the present study both for the estmaton of trends and patterns, and determnants of mgraton. Secton 4 analyses the broad trends and patterns of nternal mgraton and the ntercensal growth of mgrants for three ntercensal perods as well as provdes the reasons for mgraton. The emprcal model for the estmaton of mgraton functon s gven n secton 5. Secton 6 provdes the emprcal results and secton 7 presents the concluson. 2. A Bref Revew of Lterature Internal mgraton s defned as populaton movement across a poltcal or admnstratve boundary, whch entals a change of usual resdence (Blsborrow, 1998). The mgraton lterature provdes a dverse range of related defntons based on three dmensons, whch are explaned by Kothar (2002). The frst dmenson entals the motves for mgratons. The reasons to mgrate are n many cases not just economc, socal, poltcal and cultural aspects also play ts part. A mgrant from an mpovershed backward regon may be attracted to more prosperous regons because of better publc servces, hgher wages, more busness opportuntes, more employment opportuntes etc. The second dmenson s the geographcal aspect of populaton movement. Mgraton flows form a certan spatal pattern, whch depends on the dstance the mgrant's travel and the drecton of ther movements. The last dmenson s tme, whch s for how long does the mgrant stay n the destnaton area and how often does he or she travel between dfferent destnatons. The Indan Populaton Census classfes a person as a mgrant f ether (a) hs/her place of brth was dfferent from the place of enumeraton or (b) hs/her place of last contnuous resdence was dfferent from the place of enumeraton. Hence, the census provded data on both lfetme (brthplace) mgraton and last-move mgraton. The former defnton s less practcal because of the fact that t defnes a person as a mgrant even f he or she stays n the new destnaton area permanently. The Populaton Census of Inda collected nformaton regardng nternal mgraton flows at t hree levels of spatal aggregaton,.e., mgraton flows between Indan states (nterstate mgraton), mgraton flows between dstrcts (nter-dstrct mgraton) and mgraton flows wthn a dstrct (Intra-dstrct). The concept of nternal mgraton (from rural to urban) attracted the attenton of both academcans and polcymakers only after the semnal work of Lews (1954). The nter-sectoral allocaton of labour s the centerpece n the dual economy analyss of Lews (1954) and subsequent works by Rans and Fe (1961). The man message of these studes s that n the process of development, labour moves to the modern sector whch facltates development. However, n developng 2

5 ths dea, t s assumed that the modern sector faces perfectly elastc labour supply orgnatng n the tradtonal or rural sector (.e., there s unlmted supply of labour from rural to urban sectors). Ths vew s nconsstent wth the dea that mgraton s determned by varyng ncome dfferentals and that labour s productve n all sectors of the economy. But labour mgraton from rural to urban sector ncreases labour productvty and hence t s always desrable and should be encouraged. Rans and Fe (1961) are of the opnon that technologcal progress n agrculture sector wll make mgraton process slow by ncreasng agrcultural labour productvty. The Todaro (1969) and Harrs and Todaro (1970) models explan the mgraton flows n the presence of rsng unemployment rates n the destnaton (urban) regon whch n turn results n the development of nformal sector. The nternal mgraton s drven by the hgher expected wage n urban sector. The expected wage s nothng but the product of hgher urban (nsttutonally fxed) wage and probablty of fndng a job (calculated by dvdng total avalablty of urban jobs and total urban labour force) n the urban sector. Further developments on mcro modelng approach are the new economcs of mgraton whch focus on famly as the agent that maxmze the famly utlty functon by mnmsng the rsks assocated wth the agrcultural sector (Stark & Bloom 1985; Stark & Katz 1986; Banerjee, 1998; Taylor & Martn 2001; Stark 1991; Mncer, 1978). Smlarly, the networks created by mgrants n the destnaton regon reduce the mgraton costs and hence encourages new mgraton flows to the destnaton places (Bhattacharya, 1998; Mtra, 2003). Internal mgraton at macro level studes the relatonshp between aggregate mgraton flows and the presence of spatal heterogenety among the dfferent locatons. The geographcal dfferentals n terms of economc growth and development, s the man determnant of nternal mgraton. The Gravty Model s the most common theoretcal framework used n emprcal analyss to study the spatal determnants of mgratons. It argues that mgraton s drectly correlated wth populaton sze and nversely correlated wth the dstance between the orgn and the destnaton regons. Dstance s a key varable, the proxy for all the mgraton costs, both psychologcal and monetary, that s spatally related to the sendng and destnaton regon. The populaton sze s the sgn of the rate of urbansaton or the growth of urban sector (Greenwood, 1985; Greenwood and Hunt, 2003; Larson and Mundlack, 1995). However, not all people react n the same way to dfferences between places. The selectvty nfluence of mgraton characterses the dfferent propensty to mgrate for dfferent categores of people. Young people n the workng age have a hgher propensty to mgrate than people n other age groups. Moreover, educaton s also an mportant selectve factor. Hgh sklled people tend to mgrate more than low sklled people (Greenwood, 1997). The majorty of emprcal works tres to nvestgate the mpact of some economc varables on nternal mgraton. A hgh economc prosperty also means more actvtes, servces and opportuntes for people lvng n that area. Moreover, dynamc centres attract mostly young people, who are wdely recognsed to be hghly moble. The most representatve economc varable s the per capta gross domestc product (GDP) as explaned by Greenwood (1997). Internal mgraton has greater potental for poverty reducton, meetng the Mllennum Development Goals (MDGs) and contrbutng to economc growth n developng countres than nternatonal mgraton (Deshngkar, 2004, 2006; Deshngkar and Start, 2003; Srvastava and Saskumar, 2004; Deshngkar et al, 2006). The relatvely poor and backward states show large populaton moblty, whch s prmarly n search of lvelhood. The moblty of male 3

6 populaton s also found to be promnent n the relatvely advanced states lke Maharashtra and Gujarat as explaned by Mtra and Murayama (2008). The labour market varables, whch are often ncluded as explanatory varables n mgraton analyss, are the unemployment rate and wage rate. The ncreasng level of educaton for rural resdents unaccompaned by ncreased opportuntes for hgher earnngs mples an ncrease n the gap between rural earnngs and hence a cause of mgraton (Dehl, 1966; Barkley, 1990; Perloff et al, 1998 and Agesa, 2001; Waddngton and Sabates-Wheeler, 2004). The flow of mgraton to the major ctes s the result of rural urban dchotomes n ncome, employment opportunty and absorptve capacty as stated by Ullah (2004). Hossan (2001) found that whle poverty, job searchng and famly nfluence were the man push factors for out -mgraton, better opportunty, pror mgrants and avalablty of jobs were the man factors behnd mgraton. The reason why people decde to move from one regon to another s not just related to economc factors. A host of other factors play a role as well. The group of varables that can affect nternal mgraton flows s qute broad and s related wth the qualty of lfe. All these factors concern publc safety, socal servces, envronmental qualty, as well as poltcal factors (Adrenko and Gurev, 2003). Accordng to Adrenko and Gurev (2003), people move from poorer and job scarce regons wth worse publc good provson to areas that are rcher and more prosperng both n terms of employment prospects and publc goods. In ths context, an attempt has been made to study (at the macro level) the trends, patterns and determnants of nternal mgraton n Inda usng the grouped data from Indan Census. To study the determnants of nternal mgraton, a Gravty Model s used as the theoretcal background for the emprcal estmaton. In the Gravty Model, a set of other explanatory varables (besdes populaton sze and dstance) are ncluded to test the applcablty of Harrs-Todaro Model n both nter-state and ntra-state mgraton n Inda. 3. Data Sources The paper seeks to address two man questons, the frst beng the current patterns and trends of nternal (both nter-state and ntra-state) mgraton n Inda and the second one: What are the determnants of rural to urban mgraton n Inda? The frst queston s analysed usng the Census data. The mgraton tables (Census of Inda) are used to analyse the trends and broad patterns of nternal mgraton at nter-state, nter-dstrct and ntra-dstrct levels and for dfferent sex categores. The ntercensal growth of mgrants s also estmated usng the same mgraton tables. The data analyss for the study of determnants of mgraton s lmted to males (Census of Inda, mgraton tables), snce a large proporton of female mgraton n Inda s for non-economc reasons such as marrage. The data for the explanatory varables are taken from other sources lke Mnstry of Agrculture, Government of Inda (GoI); Lok Sabha Unstarred Queston No. 193, 15 th July, 2002; Department of Secondary and Hgher Educaton, Mnstry of Human Resource Development, GOI; Wage Rate n Rural Inda, Labour Bureau and Annual Survey of Industres, , Statstcs on Employment and Labour, Mnstry of labour, GOI; Central Statstcal Organsaton (CSO). Most of the data are collected from the webste ( The dstance data between the captal ctes s collected from ( 4

7 4. Trends, Patterns and Intercensal Growth of Mgraton n Inda 4.1. Trends of Internal Mgraton: The total (n mllons) and percentage of lfetme mgrants based on the place of last resdence crteron, by sex and type of resdence n Inda from 1971 to 2001 s gven n Table 1 and 2, respectvely. In 1971 census, mllon people, comprsng of 49.6 mllon males and 110 mllon females, were termed as mgrants on the bass of place of last resdence crteron. Ths consttutes 30.6 per cent of the total populaton of the country. In terms of total volume of mgraton, the fgure ncreased to mllon n 1981, mllon n 1991 and mllon n Table 1: Internal Mgrants by Sex, Inda ( ) Year Lfetme Mgrants (n mllon) Total Males Females Source: Calculated from Census of Inda, Mgraton Tables. The percentages of mgrants to total populaton however declned to 30.3 n 1981 and further to 27.4 n It could be suspected that Census of Inda underestmates the total number of mgrants. It s partally recurrng seasonal mgraton and commutng (n the less than one year duraton of last resdence), whch s the most sgnfcant emergng pattern of nternal mgraton n Inda. The percentage of mgrants s however ncreased to per cent n 2001 and ths ncrease may be due to the rapd expanson of the urban areas n Inda. The percentages of male mgrants declned from 18.9 n 1971 to 17.8 n 1981 and further to 14.6 n It has however ncreased to per cent n The phenomenal ncrease n the urban nformal sector, partcularly after the economc reforms (1991), could be responsble for ths. The percentages of female mgraton ncreased to 48.3 n 1981 as compared to 42.3 n 1971 and n 1991, t was 41.2 per cent whch agan ncreased to per cent n It s obvously due to the socal-cultural setup n Inda where females normally mgrate after marrage. In comparson to the rural and urban categores, the percentage of mgrants n rural areas ncreased from 28.2 per cent n 1971 to per cent n 2001, whle the percentages of urban mgrants declned from 28.2 per cent n 1971 to 17.3 per cent n In both rural and urban areas, the share of female mgrants s agan found to be domnant. 5

8 Table 2: Percentage of Lfetme Mgrants n the Total Populaton by Sex and Type of Resdence n Inda, ( ). Type of Resdence Year Total Male Female Total Rural Urban Source: Calculated from Census of Inda, Mgraton Tables Reasons for Mgraton: Table 3: Reasons for Internal Mgraton accordng to Sex and dfferent Streams of Mgraton n 2001 Census (fgures n %) Male Female Reasons for Mgraton Rural- Rural Rural- Urban Urban- Rural Urban- Urban Rural- Rural Rural- Urban Urban- Rural Urban- Urban Work or Employment Busness Educaton Marrage Moved after brth Moved wth Household Other reasons Total Source: Calculated from Census of Inda, Mgraton Tables. The reasons for mgraton have been classfed nto seven broad groups (vz., work/employment, busness, educaton, marrage, moved at brth, moved wth famly and others) n 2001 Census. The data on reasons for mgraton are useful to understand the motvatonal factors behnd the movement of people. It s observed from Table 3 that employment among males and marrage among females are the man reasons for mgraton. About 29 per cent of the male mgrants moved form rural to rural areas due to work or employment reasons. Smlarly, employment or work caused around 26.5 per cent from rural to urban, per cent from urban to rural and per cent from urban to urban mgraton n Inda. Whle n the case of females, per cent mgrated from 6

9 rural to rural areas due to marrage. Agan per cent from rural to urban, per cent from urban to rural and per cent from urban to urban females mgrated due to marrage. Assocatonal reasons lke movement on account of accompanyng parents or any other members of the famly s regarded as the second most mportant reason among both male and female ntercensal mgrants Patterns of Internal Mgraton: The total mgraton flow can broadly be dvded nto four dfferent streams vz., rural to rural, rural to urban, urban to urban and urban to rural. Smlarly from the pont of vew of dstance, t can be classfed as nter-state and ntra-state (nter-dstrct & ntra-dstrct) mgraton. Table 4 depcts the percentage dstrbuton of lfetme mgrants n dfferent streams for males and females from 1971 to 2001 at ntra-dstrct, nter-dstrct, nter-state dstance categores, respectvely. The results show a consderable declne n the proporton of ntra-dstrct mgrants for both the sex categores, where as both nter-dstrct and nter-state mgratons are ncreasng over the same perod of tme. As explaned by Bhagat (2005), droughts and floods whch occurs n some parts of the country every year renders many people homeless. Ths has been recognsed as one of the reasons for mgraton n the censuses of 1981 and And at the same tme, the phenomenal expanson of urban nformal sector durng the 1990s could be the reason for mgraton n 2001 Census. At the natonal level, rural to rural mgraton consttutes the most domnant category. But ts share has been declnng over tme for both males and females n all the three streams. The declne was greater n case of males as compared to females. There s a large share of female ntra-dstrct rural-to-rural mgraton, whch s normally explaned n terms of marrage mgraton (Census of Inda). Rural-to-rural ntra-dstrct mgraton of males seems to be due to ther mgraton from areas of low agrcultural productvty to sparsely populated areas wth new developmental actvtes. In addton, Lusome & Bhagat, (2006) stated that apart from employment among males and marrage among females, moved wth household emerged as another mportant factor for mgraton among males as well as females. The rural to urban streams show ncreased percentage share of mgrants over the perod, 1971 to 2001, whch s manly on account of employment or work. Combnng all the above three dstance categores of mgraton (.e., ntra-dstrct, nter-dstrct and nter-state mgraton), t s suggested that rural-to-urban male mgraton has ncreased substantally over the perod, 1971 to The creaton of modern sector n major metropolses and bg ctes and the subsequent development of the urban nformal sector as s the major reason for rural to urban mgraton. A smlar trend s found for urban-to-urban mgraton. However, the proporton of rural-tourban lfetme mgraton of females has decreased wth ncreasng dstance between orgn and destnaton places. The share of urban-to-urban mgraton of both males and females was comparatvely low n the ntra-dstrct stream, but t ncreased substantally n the nter-dstrct and nter-state streams. The reason could be that the nsttutons of hgher learnng, partcularly professonal and techncal nsttutons, are not avalable n all dstrcts and the motves for hgher educaton drves both the urban dwellers as well as the rural folks to mgrate over long dstances. 7

10 Table 4: Internal Mgraton flows by Sex and dfferent dstance categores n Inda, (fgures n %). Mgraton Categores Males Females Intra-dstrct Level Rural to Rural Rural to Urban Urban to Urban Urban to Rural Sub Total Inter-dstrct Level Rural to Rural Rural to Urban Urban to Urban Urban to Rural Sub Total Inter-state Level Rural to Rural Rural to Urban Urban to Urban Urban to Rural Sub Total Grand Total All dstance categores Rural to Rural Rural to Urban Urban to Urban Urban to Rural Source: Calculated from Census of Inda, Mgraton Tables Intercensal Growth of Mgrants n Inda: The percentage growth of ntercensal ( , and perods) mgrants for all categores are presented n ntra-dstrct, nter-dstrct, nter-state streams and the combnaton of all streams n Table 5. Durng , there was a postve growth of total and female mgrants for all mgraton categores whle a negatve growth of male mgraton from rural-rural areas n both ntradstrct and nter-state cases. Smlarly, the growth rate of urban-urban male mgraton n nter-state cases s negatve n the same perod. Agan durng the decade , both male and female mgraton decreased as regstered by the negatve growth rates. The growth rate of male mgrants n all streams s negatve except rural-urban nter-dstrct and nter-state cases, whle the growth rate of female mgraton from urban to rural areas s negatve except n nter-state case. The resultng growth of total mgraton s postve n rural-urban areas and s negatve n all other cases. The growth rate of ntercensal mgrants durng , for both male and female mgrants s postve except urban-urban category n nter-dstrct stream and rural-rural and rural to urban categores n ntra-dstrct male cases. 8

11 Male nter-state rural to urban mgraton stream shows an ncrease of nearly 90 per cent whle females ncreased by about 61 per cent. On the other hand, nter-state urban to urban mgraton stream for males shows an ncrease of 28 per cent whle females have ncreased by 22 per cent. These ncreases n rural - urban and urban urban cases are greater than the ncreases n other streams of mgraton for both male and female categores. The man reasons for the above mgraton streams could be explaned n terms of employment and marrage. Assocatonal reasons ncludng movement on account of accompanyng parents or any other members of the famly and for learnng and hgher educaton are consdered t to be the second most mportant reason for both male and female mgraton. The reasons for the negatve growth rates are mostly because of the nablty of Indan Census to collect nformaton regardng both seasonal mgrants and commuters, whch are the two emergng mgraton patterns n Inda durng the last two decades. The Census of Inda collects the nformaton at a partcular pont of tme and there are no such questons n the Census questonnare whch can collect nformaton regardng the seasonal mgraton and commutng. The reasons for the postve growth rate could be due to the creaton of new dstrcts or states durng the ntercensal perod. The number of nternal mgrants could be ether overstated or under estmated dependng upon the creaton of new dstrcts or states durng the ntercensal perod. Recently, three new states vz., Uttaranchal, Jharkhand and Chhattsgarh were created n a few months before the latest census held n February Table 5: Intercensal growth rates of mgrants by sex and dstance categores, Inda Intercensal Growth Rates Mgraton Categores Total Males Females Total Males Females Total Males Females Intra-dstrct Level Rural to Rural Rural to Urban Urban to Urban Urban to Rural Inter-dstrct Level Rural to Rural Rural to Urban Urban to Urban Urban to Rural Inter-state Level Rural to Rural Rural to Urban Urban to Urban Urban to Rural Total Internal Mgraton Rural to Rural Rural to Urban Urban to Urban Urban to Rural Source: Calculated from Census of Inda, Mgraton Tables. 9

12 5. Econometrc Framework In the present context, to study the factors affectng the nternal mgraton process n Inda, the sngle equaton regresson model s used. The theoretcal background for the emprcal model has been borrowed from the Gravty Model of mgraton (whch ncludes the demographc varables vz., populaton sze n both destnatons and places of orgn, and the dstance between the places). At the same tme, a vector of soco-economc varables s ncluded n the same model to test the applcablty of another mportant theory of nternal mgraton (.e., the Harrs-Todaro model). The orgn of most emprcal research of ths type of Gravty Model of mgraton, n whch the gross flow of mgrants are assumed to be drectly proportonal to orgn and destnaton populaton and nversely proportonal to the dstance between orgn and destnaton, n addton to beng dependent on a number of other orgn and destnaton varables s: M j α = P P β j D γ j f ( X, X j ) (1) {, j = 1,2, n} Where M j s the gross flow of mgrants from regon to regon j, P s the populaton of regon, and P j s the populaton of regon j. D j s the physcal dstance between regon and regon j. And X, X j are the vectors of orgn and destnaton varables, respectvely. Estmaton of equaton (1) s generally done n the double logarthmc form to get the coeffcents as elastctes. In ths model, the explanatory varables n the mgraton functon typcally nclude the urban expected wage rates (outcome of the product of urban fxed wage rate and probablty of gettng an urban job), the rural (agrcultural) wage rate, ncome (Per Capta Net State Domestc Product) levels of both orgn and destnaton places, unemployment or employment rates, educaton or lteracy rates, and urbansaton and populaton levels, as well as the physcal dstance between regons of orgn and destnaton. Although there s a consensus that the types of economc varables affect mgraton, there s no unformty concernng the manner n whch demographc varables are entered nto the mgraton functon. A varety of dependent varables are used even n the ordnary least square (OLS) estmaton; for example, M j /P, M j /(P.P j ), M j. The normalsaton most frequently adopted n the actual estmaton s to dvde mgraton flows by the orgn populaton. Strct adherence to the gravty model would requre the ncluson of both orgn and destnaton populaton on the rght -hand sde of the equaton. The model suggested by Levy and Wadycky (1972) s: M P j = P α 1 P β j D γ j f ( X, X j ) (2) However, populaton varables ncluded on the rght-hand sde often tend to be hghly correlated wth other explanatory varables, and the methods of ther ncluson vary wth the problems encountered whle usng dfferent data sets. Other authors have omtted both populaton varables from the rght-hand sde or mposed a constant return to scale property wth respect to orgn and destnaton populatons (Mundlak, 1979)- the ratonale beng that doublng the orgn and destnaton 10

13 populaton should double the flow of mgraton, wth factors remanng constant. Imposng homogenety (a = 1-ß) n equaton (1) produces M P j Pj = P β D γ j f ( X, X j) (3) Smlarly, the normalsaton (M j /P.P j ) was used by Greenwood (1971) whch mposes the restrcton of ß=1 (n addton to a=1) f P, P j are left out of the rght -hand sde. M j P Pj = D γ j f ( X, X j ) (4) In the present study, the dependent varable for all the mgraton equatons (nter-state and ntra-state) s the gross male mgraton flow dvded by the relevant (rural) orgn populaton. It s because of the fact that most of the females mgrate because of non-economc reasons (.e. marrage). The explanatory varables ncluded n the models are not the same, because of the unavalablty of data at the ntra-state level. The equaton 5 and 6 represent the ordnary least square (OLS) estmaton model for nter-state and ntra-state mgratons, respectvely. A detaled descrpton of the all the explanatory varables are gven n appendx-a. M j ln P β ln Y 7 j p j = β0 1 ln 2 ln Dj 3 ln U 4 ln We p ln L ln Lt ln GFCF ln Wat j 5 12 ln W LD + ε 1 6 ln Y + (5) M j ln P β ln Y 7 j p j = β0 1 ln 2 ln Dj 3 ln U 4 ln We p ln L ln Lt ln GFCF ln Wat j 5 12 ln W LD + ε 2 6 ln Y + (6) Estmaton of determnants of mgraton usng sngle equaton methods has been the practce n almost all prevous works and hence s gven prme mportance n the present study. In the present study, an attempt has also been made to study the smultaneous relatonshp between mgraton rate and rural wage, usng a two-stage least square (2SLS) method. The exstence of smultaneous relatonshp between the rural wage and nternal mgraton was establshed by Harrs-Todaro (1970), where they have assumed that n the urban sector, there exsts an nsttutonal fxed wage, whch s above the mnmum subsstence level. The nternal mgraton s drven by the hgher expected wages n urban sector, and the equlbrum rural wage and mgraton flows are determned smultaneously n the economy. The Harrs-Todaro condton for long run equlbrum s: W W = 0 (7) e 11

14 The equaton 8 and 9 are the structural equatons for both mgraton and rural wage functons estmated by the Two-Stage Least Square (TSLS) method at nter-state and ntra-state levels. A detaled descrpton of the all the explanatory varables are gven n appendx-a. M j ln P β ln Y 7 j p j = β0 1 ln 2 ln Dj 3 ln U 4 ln We p ln L ln Lt ln GFCF ln Wat j 5 12 ln W LD + ε 1 6 ln Y + (8A) ln W β 6 = β 0 ln GFCF ln U ln Emp ln Y + ε 2 3 ln L 4 M ln P j 5 ln Lt + (8B) M j p j ln = β0 1 ln P p β ln L ln Lt + ε D 3 ln U 4 ln W e 5 ln W 6 ln Emp + (9A) ln W β 6 = β 0 ln Emp ln U 1 + ε 2 2 ln L 3 M ln P j 4 ln Lt 5 ln GFCF + (9A) 6. Rural to Urban Mgraton: Emprcal Results: The mgraton functon estmated on the bass of aggregated data may not be preferable to base our expectatons about the results on a sngle hypothessed response. Because of the fact that people who belongs to dfferent economc sectons wll respond dfferently to changes n economc varables. Agan ndvduals n the same mgraton stream may have wdely dfferent reasons for movng. Accordngly, an attempt s made here to offer, wherever necessary, competng explanatons of a partcular result. Snce all the varables are n the form of logarthms, the parameter estmates represent elastctes. The T- statstcs are presented n parentheses along wth the coeffcent values. A sngle, double and trple star mples the statstcal level of sgnfcance at 10%, 5% and 1%, respectvely Inter-state Rural to Urban Mgraton Both Ordnary Least Square (OLS) and Two-Stage Least Square (TSLS) results of the less than one year and one to four years duratons of nter-state rural-urban mgraton are presented n Table 6 & 7 to facltate comparsons. The OLS results are presented n Column 2 n both the tables, where as the TSLS results are gven n Column 3 & 4 respectvely. The OLS results presented here are the robust result whch was corrected for both heteroscedastcty and possble seral correlaton problems. The mean of varance nflaton factors fgures are gven n the tables and they ndcate that the degree of multcollnearty s very low, though 12

15 t does not affect the estmated coeffcents. The R 2 (the measure of goodness of ft) for both the equatons are qute reasonable along wth hgh sgnfcance levels of F-statstcs. The estmated results suggest that n both duratons of mgraton, the gravty varables (populaton sze and dstance) are statstcally (at 1% level) as well as economcally sgnfcant, wth postve and negatve sgns (as expected), and absolute values of the coeffcents are 0.58 & (populaton elastctes of mgraton) and -1.3 & (dstance elastctes of mgraton) respectvely. The orgn populaton sze acts as a pushng factor where as the sze of urban populaton s an ndcaton for the rate of urbansaton (as a pullng factor). Both orgn and destnaton populaton have postve roles to play n the mgraton process, whch s evdent from the emprcal results. The dstance s the proxy for all mgraton costs (ncludng the psychc cost) and has played a deteroratng role n the nternal mgraton process n Inda. Table 6: Interstate Rural-Urban Mgraton (Duraton of less than 1 year) Varables OLS 2SLS M j/p M j/p W Constant (-5.01)*** (-5.0)*** 0.33(0.72) P j/p 0.58(16.29)*** 0.65(15.14)*** --- D j -1.3(-14.44)*** -1.41(-12.07)*** --- U 0.49(4.83)*** 0.38(3.08)*** 0.032(1.80)* W e 0.009(0.15) (-0.37) --- W 0.13(0.63) 4.14(3.46)*** --- Y -0.55(-2.44)*** -1.69(-3.87)*** 0.27(6.91)*** Y j 1.35(10.05)*** 1.42(7.92)*** --- L -0.07(-0.78) 0.55(2.51)** -0.16(-9.90)*** Lt -0.7(-1.13) -0.63(-0.72) 0.18(1.36) GFCF 0.24(5.47)*** 0.36(5.99)*** (-5.35)*** Wat j 0.32(1.73)* 0.21(0.87) --- LD 0.71(4.34)*** 0.84(3.94)*** --- M j/p (-1.59) Emp (-3.4)*** P (5.38)*** R F-Statstcs 76.37*** 48.97*** 34,49*** Root MSE Mean VIF N Note: The T-statstcs are presented n parentheses and *, **, and *** mples the statstcal level of sgnfcance at 10%, 5% and 1%, respectvely. The orgn (rural) unemployment has the postve coeffcents (0.49 and 0.43) as expected and are hghly sgnfcant (at 1% level), suggestng that the orgn unemployment s one of the major pushng factors responsble for rural-urban mgraton. But the most crucal fndng of the study s the nsgnfcant expected wage, the most mportant economc varable as explaned n the famous Harrs- 13

16 Todaro model. The rapd expanson of the urban nformal sector n Inda could be responsble for ths, whch n tself attracts the workers. There hardly exsts any labour moblty from urban nformal to the urban formal sectors whch s shown by the fact that about 93% of the total work force s engaged n nformal sector n Inda (Sengupta, 2007). The coeffcent of rural agrcultural real wage has a lmted role to play n the nter-state mgraton, snce t s statstcally nsgnfcant n all the equatons. It may be due to the fact that the mgraton s not just by the very poor durng tmes of crss for survval and copng but has ncreasngly become an accumulatve opton for both poor and non-poor (Deshngkar and Start, 2003).The coeffcents of both orgn and destnaton states per capta net state domestc product are hghly statstcally (at 1% level) as well as economcally sgnfcant (havng expected negatve and postve sgns, respectvely). Ths fndng s smlar to Greenwood study (1997) whch stated that the per capt a gross domestc product (GDP) s the most representatve macro-economc varable responsble for mgraton of people as hgh economc prosperty means more actvtes, servces and opportuntes for people lvng n that area. The results suggest that orgn ncome elastctes of mgraton are less than one (-0.55 and -0.56) where as the destnaton ncome elastctes of mgraton are greater than one (1.35 n both the equatons). Ths suggests that the rural to urban mgraton n Inda s bascally to t he relatvely developed states. Table 7: Interstate Rural-Urban Mgraton (Duraton of 1 to 4 years) Varables OLS 2SLS M j/p M j/p W Constant -8.89(-3.43)*** (-4.43)*** 0.94(2.26)** P j/p 0.608(18.29)*** 0.7(14.45)*** --- D j -1.48(-16.54)*** -1.64(-11.46)*** --- U 0.43(4.02)*** 0.079(0.48) 0.06(3.81)*** W e (-0.57) (-0.76) --- W 0.34(1.52) 6.82(4.39)*** --- Y -0.56(-2.44)*** -2.56(4.40)*** 0.29(8.03)*** Y j 1.35(9.53)*** 1.45(6.53)*** --- L -0.18(-1.74)* 0.82(2.96)*** -0.16(-10.65)*** Lt -1.06(-1.62) 0.24(0.23) 0.002(0.02) GFCF 0.23(5.41)*** 0.36(5.57)*** (-4.99)*** Wat j 0.49(2.42)*** 0.45(1.57) --- LD 0.64(3.54)*** 0.74(2.80)*** --- M j/p (-1.49) Emp (-3.62)*** P (5.49)*** R F-Statstcs 91.3*** 41.18*** 40.17*** Root MSE Mean VIF N Note: The T-statstcs are presented n parentheses and *, **, and *** mples the statstcal level of sgnfcance at 10%, 5% and 1% respectvely. 14

17 The coeffcent of the average sze of operatonal holdngs s sgnfcant only n 1 to 4-year duraton case wth a negatve sgn (-0.18), mplyng the fact the people wth small operatonal holdngs prefer to mgrate to the urban sector and ether settles down there permanently or send remttance regularly back home. The coeffcents of Gross Fxed Captal Formaton (GFCF) n agrculture are hghly sgnfcant, wth postve sgns (0.24 and 0.23), mplyng that the people from agrculturally advanced regons mgrate to urban sectors. The reasons could be ether that they can bear the mgraton costs and have a greater propensty to mgrate for better opportuntes or to carry out ther busnesses. The coeffcents of rural lteracy rate are not sgnfcant n all the regresson equatons. Ths prompts us to thnk agan about the urban nformal sector n whch a large proporton of workers are unsklled and llterate. The coeffcent of the water facltes at the destnaton s postve and sgnfcant n both the cases (0.32 and 0.49), ndcatng urban santaton faclty s one the factors responsble for rural-urban mgraton n Inda. The coeffcents language dummes are postve and hghly sgnfcant (0.71 and 0.64) n both equatons suggestng that language plays an mportant role n nter-state mgraton n Inda. Out of the total mgrants, 71% of mgrants n the less than 1 duraton and 64% of mgrants n the 1 to 4 years of duraton scenaros prefer to mgrate to states wth smlar language. Two-Stage Least Square estmates are presented above to facltate the comparson wth the OLS results. After satsfyng both the necessary and suffcent dentfcaton condtons (.e., the rank and order condtons for dentfcaton) for all equatons, t s clear form both Table 6 and 7 that the smultaneous relatonshp between mgraton rate and the average rural wage s not establshed even though the Hausman Test suggests the possblty of the exstence of smultanety between mgraton rate and rural wage. The coeffcents of mgraton rates are nsgnfcant n both the rural wage equatons even f rural wage rates are hghly sgnfcant n both the cases. Ths 2SLS results agan put a queston mark on the applcablt y of Harrs-Todaro model n the case of a developng country lke Inda 6.2. Intrastate Rural to Urban Mgraton: The estmated results n ntra-state rural-urban mgraton of both ordnary least square (OLS) and twostage least square (2SLS) for less than one year and one to four years duratons are presented n Table 8 & 9 respectvely. Agan, the OLS results presented here are robust result wth very low degree of multcollnearty, consderable R 2 and hghly sgnfcant F-statstcs. In the ntra-state case too, the Gravty Model s well establshed n Inda. The coeffcents of the destnaton and orgn populaton rato are postve and hghly sgnfcant, whch supports the theoretcal argument that the nternal mgraton s drectly proportonal to the populaton sze of both orgn and destnaton places. And the coeffcents of dstance dummes are negatve (-0.35 & ) n both the less than 1 year and 1 to 4 years of duraton cases, as expected and hghly sgnfcant. Ths ndcates the fact that dstance factor has a deteroratng mpact on populaton moblty. The coeffcent of Harrs-Todaran expected wage rate s hghly sgnfcant n less than 1 year duraton of mgraton where as nsgnfcant n 1 to 4 year duraton of mgraton cases. Ths result, n fact, s very nterestng snce most of the short dstance and duraton mgrants are seasonal mgrants and commuters, often guded by the hgher expected wages n the destnaton places. But as the duraton of mgraton ncreases, there are a host of factors responsble for rural-urban mgraton where the role of 15

18 expected wage s almost restrcted. The coeffcent of orgn unemployment s statstcally sgnfcant havng the expected negatve sgns (-0.22 and -0.31) n both less than one year and one to four years of duraton cases, whch suggests the fact that the orgn unemployment s one of the major factors for ntra-state (short dstance) rural-urban mgraton. But the most nterestng result that appears n wthn state mgraton s that the coeffcent of the destnaton employment s statstcally sgnfcant wth the negatve sgns (-5.08 and -1.34). Ths mples the fact that the short dstance mgraton n Inda s dscouraged by the destnaton employment because the urban nformal sector attracts both unsklled and semsklled workers from the rural areas. Both the self employed and wage workers face competton n the nformal job market and t s very dffcult for them to get a lvelhood f the destnaton places are already crowded. Table 8: Rural-Urban Mgraton (Duraton of less than 1 Year) Varables OLS 2SLS M j/p M j/p W Constant 10.11(1.61) 9.23 (1.28) 4.48(6.10)*** P j/p 0.16(1.93)* 0.14(1.71)* --- D -0.35(-4.17)*** -0.35(-4.17)*** --- U -0.22(-1.80)* -0.21(-1.72)* 0.044( 0.74) W e 0.98(4.25)*** 1.02( 3.82)*** --- W -0.01(-0.10) (-0.39) --- Emp j -5.08(-3.25)*** -4.82(-2.66)*** --- L 0.08(1.23) 0.081( 0.95) 0.081( 0.95) Lt 0.84(7.58)*** 0.86( 6.14)*** 0.18(2.01 )** M j/p (-0.53) R F-Statstcs 46.75*** 21.58*** 4.87*** Root MSE Mean VIF N Note: The T-statstcs are presented n parentheses and *, **, and *** mples the statstcal level of sgnfcance at 10%, 5% and 1% respectvely. The coeffcents of the average sze of operatonal holdngs are nsgnfcant n both equatons suggestng that land holdng does not have any mpact on the ntra-state mgraton flows. The coeffcents of rural lteracy are hghly sgnfcant, wth postve sgns (0.84 and 0.86) n all the regresson equatons, ndcatng that lteracy plays a crucal role n ntra-state mgraton. Ths could be due to the reason that among the class of mgrants, the educated people have a greater chance of mgratng, for purpose of ether hgher educaton or n search of better job opportuntes. Smlarly, the Two-Stage Least Square estmates presented above n Table 8 and 9 fal to establsh smultaneous relatonshp between mgraton rate and the average rural wage as the coeffcent of rural wage s nsgnfcant n both the mgraton equatons. The coeffcent of mgraton rate s not sgnfcant n the rural wage equaton n both duratons even though Hausman Test 16

19 suggested the possblty of the exstence of smultanety between mgraton and rural wage. The OLS results tself provde consstent estmator for the parameters estmated n the ntra-state mgraton. Table 9: Rural-Urban Mgraton (Duraton of 1 to 4 Years) Varables OLS 2SLS M j/p M j/p W Constant -1.9(-0.32) -3.22(-0.45) 4.48(6.2)*** P j/p 0.11(1.27) 0.08(1.0) --- D (-4.86)*** (-4.85)*** --- U -0.31(-2.56)*** -.299(-2.42)*** 0.04(0.48) W e 0.28(1.28) 0.34(1.3) --- W 0.05(0.53) -0.06(-0.28) --- Emp j -1.34(-0.9) -0.95(-0.54) --- L 0.064(0.95) 0.061(0.73) -0.03(-0.80) Lt 0.86(7.77)*** 0.899(6.48)*** 0.19(1.85)* M j/p (-0.55) P (-0.8) R F-Statstcs 46.16*** 19.15*** 4.84*** Root MSE Mean VIF N Note: The T-statstcs are presented n parentheses and *, **, and *** mples the statstcal level of sgnfcance at 10%, 5% and 1% respectvely. 7. Summary and Concluson The rural to rural mgraton has been the most mportant mgraton flow that ndcates a steadly declnng proporton, whle the proportons of other categores have ncreased over the perod. In the same way, the proporton of short dstance mgrants has decreased whle the proportons of medum and long dstance mgrants have ncreased, suggestng that the long dstance movements are more urban-orented than short dstance movements. The short dstance mgraton s man mgraton pattern among Indan females whle long dstance mgraton s often undertaken by the males. Marrage s the reason for a large proporton female mgraton where as n the case of males, the most mportant reasons are employment, busness etc. Agan, t s evdent from the Indan Census that there has been a sgnfcant ncrease n mgraton to urban areas both among males and females durng The rural to urban mgraton may be due to the rapd growth of urban nformal sectors n the recent years, whch comprses about 93 per cent of the total employment n the economy. The movements from urban to urban areas are also ncreasng consderably. Ths may be due to globalsaton and the quck expanson of the servce sector. From the current trends and patterns of nternal mgraton n Inda, t can be antcpated that long dstance rural to urban and urban to urban streams wll be the domnant mgraton streams n the future. 17

20 In ths context, the study analysed the determnants of mgraton to provde a sound understandng of the factors responsble for rural to urban mgraton. The estmated results from OLS method suggests that n both nter-state and ntra-state levels and duratons (less than 1 year and one to four years) of mgraton, the gravty varables (populaton sze and dstance between the places) are hghly sgnfcant, wth the expected sgns whle the Harrs-Todaran varable dealng wth the urban expected wages, has a lmted role to play n the nternal mgraton process n Inda. At the same tme, the possblty of the exstence of smultaneous relatonshp between the mgraton rates and rural (agrcultural) wage rate s rejected n both the categores of mgraton n Inda. But there exsts other push and pull factors (lke unemployment, level of ncome, average sze of operatonal holdngs, urban water and santaton facltes, smlar language etc.) whch operatng mutually both at orgn and destnaton places, are the man determnants for nternal mgraton. The role of the thrd sector (urban nformal sector) n the nternal mgraton process s not explctly analysed here, whch s one of the lmtatons of the present macro-level study. The man reason for ths s the use of the group-level data of Indan Census, whch fals to provde nformaton about a partcular mgrant, whether he/she s mgratng to urban nformal sector or formal sector. However, these lmtatons of Census data can be tackled f they nclude a few more questons n addton to the queston of the place of last resdence. It would be better f they ask whether they are stayng at the same place or have ever changed ther place of resdence; f yes snce how long? Wth regard to the employment queston, t s worthwhle to nclude an addtonal queston whch can explctly defne the nformal sector employment snce t consttutes the major share n employment n Indan labour market. In order to get nformaton regardng seasonal mgraton, a queston can be asked whether the partcular respondent moves elsewhere n a partcular year for the purpose of employment durng the agrcultural off seasons. Informaton about commutng can easly be collected by askng whether hs place of work s same as the place of resdence. Nevertheless, a detaled and n-depth study of mgraton behavour (at the mcro-level) and the labour market dynamcs n Inda s urgently requred. The forthcomng Natonal Sample Survey (NSS 64 th Round) data pertanng to mgraton detals and n-depth mcro survey related to new drectons and ssues may be useful to examne the complexty arsng out of the context of mgraton patterns n Inda. Notes The urban expected wage W e= W j (Emp j/total urban labour force) s equated to the rural wage W (here n the orgn wage rates) at the long run equlbrum where the mgraton flow from rural to urban sector comes to a halt. Therefore, the rural wage rates and mgraton rates n short-run are determned smultaneously (Harrs J. & M. Todaro, 1970). Hausman test s performed to fnd out whether a partcular varable s exogenous or endogenous n the model. A varable x s exogenous f the condtonal dstrbuton of y gven x does not change wth modfcatons of the process generatng x. And a strctly exogenous varable s one that s ndependent of all contemporaneous, future and past errors (Wooldrdge, 2006; and Brooks, 2008) 18