DISCUSSIOQN PAPER Report No.: UDD-53 THE IMPACT OF UNBALANCED PRODUCTIVITY ADVANCE ON INDIAN URBANIZATION: SOME PRELIMINARY FINDINGS by Charles M. Becker Edwin S. Mills Jeffrey G. W-illiamson April 1984 Water Supply and Urban Development Departmnent Operations Policy Staff The World Bank the World Bank does not accept responsibility for the views expressed herein which are those of the authiors anid should not be attributed to the .World Bank or to its affiliated organizations. The findings, interpretations, and conclusions are the results of research supported by the Bank; they do not necessarily represent official policy of the Bank. The designations employed, the presentation of material, and any maps used in this document aze solely for *the convenience of the reader and,do not imply the expressidn of any opiniion whatsoever on the part of the World Bank or its affiliates co-icerning the legal status of any country, territory, city, area, or its authorities, or concerning the delimitation of its boundaries, or national affiliation. Research Prorect No.: 672-64 Analysis of Indian Urbanization ABSTRACT This paper investigates the impact of changes in sectoral productivity on output and employment patterns in a simulation model of the Indian economy. Productivity gains in major urban sectors are found to have fairly strong urban growth effects both in the short and long run. Rural productivity advances initially stem urban growth, but have little long run effect. Sv. TABLE OF CONTENTS INTRODUCTION STATIC AND DYNAMIC EFFECTS OF A ONE TIME PRODUCTIVITY SHOCK CONCLUDING REMARKS REFERENCES8 THE IMPACT OF UNBALANCED PRODUCTIVITY ADVANCE ON INDIAN URB.NIZATION: SOM PRELIMINARY FIN.DINGS Charles H. Becker* Edwin S. Mills*** Jeffrey G. Williamson** *Assistant Professor of Economics Vanderbilt University, Nasnville,TN,. 37235 **Professor of Economics, Har-vard University, Cambridge. NA. 02138 ***Professor of Economics, Princeton University, Princeton, N.J. 08544 ABSTRACT This paper investigates the impact of changes in sectoral productivity on output and employ- ment patterns in a simulation model of the Indian economy. Productivity gains in major urban sectors are found to have fairly strong urban growth effects both in the short and long run. Rural productivity advances initially stem urban growth, but have little long run effect. INTRODUCTION India's development trends differ in many critical respects from stylized growth patterns. Its urban population has grown at the moderate annual rate of 3.8%, Just less than twice the national population growth rate. Yet this persistent urbanization has not been a&companied by large structural shifts in employment and output composition. Furthermore, incremental capital/output ratios have experienced alarming growth, especially in, predominantly urban sectors. The unusual circumstances of India's growth indicate that a thorough analysis of Indian eco- comic growth and it.srelation to urbanization would be valuable. In order to provide such a Study, we have constructed a m=ti-sector model of Indian rural-urban migration, city growth and economic development (Becker, Mills and Williamson (2], hereafter "BMW"). Since many of th, critical forces of relevance to the city growth issue cannot be captured adequately by a single analytical model, we have designed a more complex model that is solved by numerical techniques. The model willx.serve eventually to forecast future urban growth under varying assumptions regarding exogenous variables' values. Counterfactual simulations can also be used to assess the contributions. to recent urban growth of productivity changes, world market conditions, capital accumulation, skilled labor formation, demand shifts and changes in fis- cal policy. This latter exercise requires us to derive a comprehensive understanding of the impacts of these forces, both in the short run and over time. This paper begins the analysis of the sources of Indian urbanization and output growth by ex- amining the impacts of sector-specific, disembodied technological change. Since sectoral factor productivity growth rates do appear to vary in India, it is important to determine the effects on urbanization of such differences. After presenting a brief description of the model, we report results from counterfactual simulations of the Indian economy for the 11 year period 1960-1970 that predict how the Indian economy would have responded at that time to chaages in productivity parametars from their estimated values. THE BMW MODEL OF INDIAN URBANIZATION BMW offers a dynamic general equilibrium simulation model of the Indian economy. It is oriented in particular to analyzing variables likely to influence urbanization that have been omitted in previous models. The framework is predominantly neoclassical, but migrants' mov- ing costs are recognized, as are segmented capital markets. A major difference between BMW and other computable general equilibrium ("CGE") models of LDCs involves the role of governmnt in a poor, mixed enterprise economy. In BMW, government has the option of consuming goods and services, providing transfers to households, delivering public services to households and industries, or investing in government enterprises. Most CGE models minimize government's role by assuming similar public and private investment behavior, and by ignoring differences in public and private enterprises. BMW explicitly con- fronts the effectx of non-competitive government investment allocation, and incorporates government relationships with public enterprises in its fiscal system. BMW also recognize. the critical role of public services as intermediate inputs, and thus provides a direct role f or government in th.e producrioh' --oce~s * There pU.iai services the major compon- ents of infrastructure: rcad3, -7i'o ,e tran.sct.. 'anr, *ar,.r, and healt. servic2s. We creat these as private intermedicate products uzsec in the pro duot4on of final goods, and assuime that their provis`on is biased towards rl:rn induscrios and t eal:. household:i. BMWl's characterization of India 3'could oe cons.stent with major private seccor institucions as well. These characterist4cs include high'y fragmented capital markets and a p?rtially segmented labor market. Earr.ings differ bt skill classes in BMW, and since laborers of different skills are not close substitutes, akilled and unskilled labor markets are quite distinct. Yet unskilled and sklfled labor both are present in the traditional and rural as vell as the modern and urban sector3. Models :E-zen omiz cFnsidiration of skilled labor out- side the modern urban sectors. bUt a modei designed tr^ analyre Indian rural-urban migration cannot justifiably ignore the presence of a substa;.t-al body of educated, skilled labor in rural India. Indeed, our results find that the proportion of skilled workers in migrant streams tends to be higher than their share of the national labor'force. Flnally, neither profit nor rental income accrues exclusively to one class, as many modela assume. The model distinguishes between tradeable and non-tradeable goods, and consumption of some services and housing is therefore location-specfic. This specificity, along with different consumption' patterns and the presence of location-specific taxes, generates differences in urban and rural households' living costs. Two goods are internationally tradeable:nanu- factures and agricultural products. India is a producer,consumer, importer and exporter of both goods. Following Armington li), we assum.e ext'ort den.and functions with finite price elasticlties, and also assume that imported g,'ods are imperfect substitutes for do-mestically produced goods of the same type. In view of our desire to understand Indian growth while minimizing co#plexity, BMW includes ten sectors and four productive factors. The model includes agricultural and urban manu- facturing, which together absorb most of the labor force. As with the Kelley-Williamson i4,5) CGE model of a "representative developing countrv", EMN contains rural and urban "informal" service sectors as well as a skill- and caDital-intensive service sector that produces largely public administration.1 Again foll'owinf! Kelley-'illiamson, there are three housing sectorn--rural housing, urban "informal" housing of low quality, and urban "formal" housing of higher quality. Finally7, the irndel distinguishes separately u;ban public services (power, other utilities, some transport) ann rural public servizes (largely public irrigation and rural public adm.inistration). Although these public service sectors are not large employers in India, they do utilize large portions of the nation's capital stock and provide essential intermediate goods to other sectors. The four factors in BMW are capital, skilled labor, unskilled labor, and land. Numerical solutions to the BMW program require the derivation of a set of consistent para- meter values. Insofar as is possible, these values are based on available econometric results. For values not thereby restricted, we compel the model to exactly replicate obsery- edd1 national accounts and other data for a "benchmark equilibrium" year, chosen to be 1960. That is, we choose unknown parameter aad exogenous variable values to force the model's solu- tion to provide estimates of endogenous variable values identical to t;ose actually observed for 1960. Once base year exogenous variablea are known, they can be systematically 'revised for succeeding time peric-.s, 'Fo' example, 1961 camital stocks equal 1960 capital stocks plus endogenously determined 1960 net in-:estment levels. A small set of technological change parameters are not restricted by the static data con- sistency requirement. These parameters are derived bt assuming that India experienced an "equilibrium 'growth era" during the 1960s. Given static parameter values and initial exogenous variable values along with hi-torL:zl ser!qs for exogenous variables such as world prices, BMW is solved for ics endogenous variable vilues in the absence of any technological progress. We then compare actual and simulared vjaIues -f key endogenous variables and cal- culate residual productivity -ai.nr i;~ot Pxp'a nec by capital accuu=Lation and labor force skill acquisition. Once we hnve dcriv:: dynamic parameter values by an iterative process from our equilibrium growth era restrictions, th.. model can be run; forward, with its per- formance judged by its ability to rerli.-ate output an,' emplovr.:ent patterns from the 1970s and carlv 1980s. The dy-namic parameters are now in the process of being derit e.. Those values used in the sim- ulations reported here were derive- unt, r the x' mmlifying assump: ion that all productivity changes occurred as Hicks-neutral factor-disoarb,di2ad multiplicative shifts in the sectoral production functions. These parameter. ar;- jaliminary, as the Hicks-neutrality restriction enables the model to successfully track rectcral cutput - ths, but generates greater urban However, BMW's RS sector includes a large rural marnufacturing component in addition to rural private services. -9 - employ-ent creation than that actuallv recorded. This r ndm- 'ndicates that a labor-saving bias has been present in urban manufacturitfg (anid pop6ibly i- orher jrzbJt sectors' techno- logical change). Given the rapid growth of registered manufacturing relative to unregistered manuracturing, the presence of s.cah a bias at an aggregate level is hardly surprising. The dynamic para=eter values employed here provide good measures of overall sectoral pro- ductivity advance (though urban technological progress may be slightly understated and rural gains overstated: see [ 3]) liowever, presence o' labor-saving bias in some urban sectors, along with needed demographic data refinements, force these figures to be preliminary. The productivity parameters chosen also differ eor the periods 1960-1964 and 1965-1970, reflect- *ing the latter periods' dramatic decline in the urban manufacturing and in public service sectorts growth. Simulations beyond 1970 will assume that long term producitivity growth is an average of the extremely successful early 1960s and unsuccessful late 19603.2 STATIC AND DYNAMIC EFFECTS OF A ONE TIME PRODU'CTIVITY SHOCK. Table 1 reports elasticity values for k,ely urbanizatior and macroeconomic variables with re- spect to productivity parameter changes. Each counterfactual simulation involves a sector- specific,.Ricks-neutral technological improvement. Specifically, the simulations increase the ith sector production function'smultiplicative constant term Ai for the entire period 1960-1970. The impact of this technology shock involves immediate reactions in labor, pro- duct, and foreign trade markets. Investment behavior responds as well, but stock adjustments occur with a one year gestation lag. Cost of living differential changes are a8so assumed to have a lagged effect on migration decisions, The figures presented in Table 1 reflect full general equilibrium influences, and will differ from the partial equilibrium multipliers commonly used, but the impact of the disembodied technological progress on output and relative price of the affected sector's good depends critically on this general price elasticity of demand. If output demand is relatively price elastic, then productivity growth will be reflected in large output supply responses in the affected sector. If output demand elasticities are low, then cost-reducing innovations will be passed on to users via falling prices. Resources will flow to the affected sector if the full general equilibrium rise in output is proportionately greater than the productivity gain itself. If demand elasticities for urban goods are high and urban sectors experience relatively large total factor productivity ("tfp") growth, then final demand will shift toward urban goods. GNP will become dominated increasingly by urban sectors, new urban em- ployment opportunities will appear, and city growth will take place. It will also take place if productivity advances in rural sectors occur, but are met instead by low price and income demand elasticities; In Table l,disembodied 1mprovements in all "modern" urbaii industries are met with sufficient- ly buoyant short run demand elasticities that the own sector equilibrium supply response elasticity exceeds unity. Since the output growth rate for urban manufacturing (N), public service (PSU) and modern services (KS) exceeds tfp growth, input use must grow. With capital fixed initially,labor use therefore increases. As in the Kelley-Williamson (KW) small, open economy,l% A. growth generates a strong supply response: 74.000 workers migrate to cities, of whom 29',000 find M sector employment. But the price of M sector value added also declines.0.89% in the partially closed Indian economy, resulting in a smaller urbanization impact than if the economy were fully open.3 Short run output, own sector employment and total urbanization effects are even greater in the case of KS tfp growth, although KS gross output is 45% smaller than M's. Moreover, the literature known as economic base theory tends to describe service sector output growth as responding to growth in manufacturing (see Mills and Becker 16 3, Ch, 6). A major reason for the greater KS urbanization elasticity is that M is capital intensive relative to KS, so that its short run supply curve will be steeper. KS also faces a highly elastic demand curve: government tax reveiiue rises by 0.16%, most of which is automatically spent on pur- chasing KS goods, In addition, an HKS housing investment spiral takes place. As KS value added price falls proportionately less than the technology gain, factor marginal value pro- duct schedules increase; labor then migrates to urban areas. In particular, a high skilled labor growth elasticity (0.49) is recorded in response to skilled labor intensive KS's tfp gain, driving up the demand for high quality urban housing (HKS), encouraging investment in 2Total non-houising productivity advance declined from 1.15% in 1960 to -0.3% in 1968. The productivity parameters calculated for the 1960s can also be compared with other residuals estimated in growth accounting exercises. It is interesting to note that BMW does not find a strong "green revolution" spurt in agricultural productivity during the late 1960s once out- put levels have been corrected for variations in rainfall, labor quality and intermediate input use. 3Relative price elasticities reported are deflated by a domestic production based Laspeyres price inidex. -3- T.ASLE 1. STATIC AND !' ! 2F!ACTS I L: '*-T' AD,'.'CE: C:. UNE3AT;A X .:M .- -LA>T:0 *t; IN t:TA )t-f.t,45-L ; (The counterfactual sinbu.-aticncc::c-~.: ,produc;'-it.,Lv advance in sect.r i involves a 1% increase in Lt. :± i-loied -utr.I.::tl1':a ciou;ical progress term A. in the. fth secror' s p:--c'iuction runcr.1'n,s ElasticitieJ and ______' :.: ..S*e reo.r:.:: . to benc-mark- (no technoi2c~zv _ c Endogenous Counterfactual 1', 4increase . Variable Simulation: A A -. A A. A -"RS KE ~' 1. IncrP-qe in Urban Labor Force (thoW:sand wourkers) 1960 . o -232 *-l.3 7, 4 2' 66 -62 112 1961 155 81 157 12.7 537 2., 34 1962 - 49 - 21 - 73 156 1.71 -123 -26 . 1963 28? 341 188 3'3 ' 2t5 251 1964 -122 - 28 - 43 i.4 1.3 -, 3 -143 1965 -108 - 48 -32 12 I ' a -:. 2 S3 1966 - 41 - 47 64 149 54 57 7& 1967 - 7 -44 -228 126 321 -- 8 40 1968 - 7 - 70 11 0 225 337 l ' 50 1969 35 - 56 - 54 202 329 -37 48 1970 -158 -141 1 328 45 105 4a 2. Own Sector Constant (1960 pre-shock) Price Outpuc Elasticity with. Respect to P--ductfvity Advance lY9b 0.91 1.84 i.44 1.17 1.88 0.S5 1,80 1961 0.77 1.53 0.35 1.23 1.44 2.08 1 33 1962 0.39 10.52 1.57 2.07 :).56 -o.06 0.03 1463 0.94 0.37 1.11 1.89 12 1.85 1.02 1964 1.09 0.61 1.59 1.61 2 .t O.:7 ..1' 1965 0.96 0.58 1.79 1.25 1.1.2 1.09 1.GY 1966 0.97 0.63 1.07 1.30 ..64 0.77, 1.05 1967 0.93 0.67 1.11 1.20 0.95 O..3b 1.0,5 1968 0.94 0.75 1.07 1.48 1.02 1.21 1. 1 1969 0.94 0.74 1.15 1.48 1.02 0.04 .l( 1970 1.05 0.73 1.02 1.68 1,21 0.28 1.09 3. Own Sector Unskilled Labor Employment ':lasticir, 1960 -0.13 1.01 0.75 0.37) 1.48 -0.47 1.30 1961 -0.28 0.30 -0.74 0.35 1.00 I..50 0.44 1962 -0.03 -0.68 0.89 1.49 0.11 -1.38 -1.49 1963 . -0.06 -0.5j 0.14 1.21. 0.69 1.36 0.05 1964 0.10 -0.34 0.72 0.76 0.66 -0.52 0.19 1965 -0.09 -0.46 0.84 0.35 0.66 0.26 0.16 1966 -0.09 -0.38 0.14 0.42 1.33 -0.10 0.08 1967 -o0.11 -0.33 0.14 0.30 0.52 -0.63 0.12 1968 -0.10 -0.25 0.27 0.64 0.55 0.49 0.13 1969 -0.09 -0.25 0.27 0.61 0. 52 -0.31 0.13 1970 0.04 -0.26 0.00 0.67 0.71 0.24 0.13 4. Own Sector Skilled LabQt Elmploymert Elastici Cy 1960 -0.32 1.00 0.46 0.27 0;97 -0.43 1.00 1961 -0.56 0.48 -0.76 0.18 0.53 0.80 0.48 1962 . -0.42 -0.46 0.66 1.08 -0.32 -1.16 -0.13 1963 -0.18 -0.50 0.14 0.84 0.30 0.74 0.00 1964 0.06 -0.36 0.69 0.57 0.29 -0.35 0.38 1965 -0.27 -0.29 0.92 0.20 0.21 0.00 0.00 1966 -0.30 -0.25 0.06 0.19 0.77 -0.31 0.00 1967 -40.30 -0.22 0.12 0.12 0.05 -0.60 O.00 1968 -0.29 -0.18 0.06 0.38 0.14 0.00 0.00 1969 -0.36 -0.17 0.22 0.04 0.14 -0.53 0.27 1970 -0.09 -0.16 0.05 0.67 0.30 0.00 0.00 Glossary or sectors: A - Agriculture and Raw Materials RS - Rural Services and Manufacturing PSR - Rural Public Services and Government Administration M - Urban Manufacturing KS - Urban Capital or Skill Intensive Services and Government Administration US - Urban Informal Sector Services PSU - Urban Public Services HRS - Rural Housing Services HUS - Urban High Quality Housing Service7 HKS - Urban Low Quality Housing Services -4- that sector. As RKS scrucrures are prcducec i. the KS sezteor, cost declines further induce investment, so that these secr.ndary cffec:s are -rz o irorcing. The presence of a srcck :t f rural sklQled liabor .5's cri-icil in determining urbanizaticn - pacts .of KS tfp gains. In the absence o: a rurall ski'ted 'abor fo-ce-, shQrt run Y.S supply responses would be muted by akilLed-labor bottZneeck.-, Shi.rt run deanr.d for KS sinilarly would be restrained' by the absence c, an :iS hau-'..g bou.. Modern sdr-:J.ces consequently have far more "mportzrx urban gro"'th e' enta in - ta.an in earlier genera; equilib;ri`um- models that ignore the possib^lity of an infLu:i of euca:-ec: !orkers fre, rural areas, Longer run effects differ considerably, from short rurn izpacts. Strong cyclIcal behavior emerges as the KS housing bocor. dissipates rapidly. Over the ten years follcwing the in±tial reaction (1961-1970), the average KS output increase over its boase value In the k, growth simulation is only 1.01%, while the M sector response in the A>> growth sim;ulacion averages a 1.52% increase. These differences reflect a continued bcuyant demand for phvsical invest- ment goods (produced in the WI sector), in contrast with a dissipation of initial housing booms generated by the initial flow of urqan immigrants. The initial capital constraint, which affects M more than KS, also is reduced over time, while the presence of rural skilled workers prevent a skil l constraint from strongly Limiting Initial KS growth. This skill con- straint becomes a more important curb to KS growth than does the capital constraint to N1 growth in later years though, as capital accumulates relatively rapidly, especially given the tfp gain in the capital goods (x) S.ctor. ' Although the KS expansion recedes int the A.,c advance case, its total urbanizalton effect continues to be the largest, though urbanizacion effects of A, and AKS growth are much closer in the long run than they are initially. Since own. sector employment elasticities after 1960 are typically higher in the Al, growth than in the A., growth case., and since M employs more workers, it is apparent that KS tfp growtn genera;'es much employment growth in other urban sectors. This success in part reflects a change in consumption patterns in favor of urbarn goods due to a r'se of real incomes cf the very top groups ir. the . shift relative to the A shift and to the benchmark. Somewhat surprisingly, though, real incomes of skilled workers exhibit little long run changes in response to the A, gain. A second reason for the large long run urbanizing influence of KS tfp gains rel-ative to the A,, gain involves the substitucicn of capital for labor. While the M tfp gain is Hicks- neutral, it has th,e economy-wide bias of making capital goods inexpensive relative co skill, ed and unskilled labor. The shift in relative prices enables the urban savings pool to support a surge in the urban capital stock that to some extent in aggregate substitutes for labor inputs, thus limiting urban i=migration. Since urban sectors are capital intensive relacive to rural sectors, and since the urban formal sector savings pool is the largest relative to value added, the fact that savings pools will purchase more capital goods in the A gain case does have the effect of greatly expanding pfnduction in urban sectors relative to rural sectors. The impact of this labor displacement can be seen by comparing combined output and capital use in the four non-housing urban sectors at a point well after the establishment of the productivity shift. Table 2 1968 Urban Output and Capital Stocks in the M and KS Productivity Gain Simlulations (Values are for all urban non-housing sectors combined in 1960 benchmark Rs.billion) Capital (K) Output (Q) K/O Benchmark 183.416 77.194 2.376 M tfp increase 184.571 77.864 2.370 ..KS tfp increase 183.717 77.708 2.364 Thus, while there are 112,000 more urbar. workers in 1968 in the Ai gain case compared to the A, gain case, total 1968 urban output is greater in tha latter simulation due to the greater accumulation of capital goods, The uxban public service (PSU) 1 productivity advance has surprisingly large urbanization and own sector output initial impact multipliers. This response stems partly from the 4 Thus, in the M tfp advance case, capital's (K) share initiallv rises relative to benchmark values (as factor substitution elasticities are below unity in the modern urban sectors, and skilled labor's (S) falls. By the mid-1960's, however, this situation has reversed itself (though S and K shares both rise in on KS sector). -5- -ulastic demand curves it ig assumed to Face . -i a hr c;i, r ate c.' *-ere '!s also a strong goverrment cax rev:enue et* ()c 'i.' resu ,-r grvenenc x-: pendirures have a strong urbar. bias. The -a ' in ..is ca Ital ar.d s.l irenjive 'eector alsc redistributes incomes toward s-illad .azcr "', . classes, awa fromn uinsk-2'ed urban labor. ' The ratio of ur-an skilled to 'ur*" '1.'u ,arninJ.s rises -!cre than ..5L. As wealthv classes have relatVel'.7- Lrban-ir.C=-ns:-e : ac..ens, r oE. creasing i.equalicy is to further incrzase ur.-a.. * -,.O,S. i: t 1. sec; trs. These strong PSU t-p gain impact erfects are no, sLtaaireu. Aitar f'Pi.t-l cycLical ei`t-cts abate, PSC own output elasticities ran,e ar,otnd 1.1, implyang lif-tle fdr.'nal factor usage beyond the benchmark valaues. The initial urban innfu-. of 1'1 OC2, wcrkers fluctuates-consider- ably and then falls to an average gain of !,0-5Q,00c0 over th., enchmarr'k, nearliy an order of magnit'ide lower than in the KS or Y. tfp gain, cases. Tie re6ucedl Icn,;; run eirploTment effect reflects adjustments to changes in cost-of-liiung diff!.rentials as well acs negligible lon- run capital accumulation relative to the benchmark (the 1968 urban non-hcusing capital stock elasticity is 0.04 in the PSU tfp gain simula2Jon, as coxxared with 0.165 and 0.63 in the KS and X productiv±tv gains, respectively). Th4i. lew capital accumulation elasticity occurs despite large public revenue gains, as government investment pptterns influence movements away from investment in the capital goods sector towards accumulation in KS and PSU. The absence of additional capital to furthei raise labor productivity beyond the initial urban productivity gain therebv restrains addiLional immigration. Alone among the urban sectors, productivity advance in in4formal sector services (UiS) does not increase its own workforce or the total urban work-orcs in the short run. lts m-rost strikinz effects are the creation of rural services Investment ane oucntpt booms due to the urban emigration and to create urban housing booms due tto inp.t, co05t declines. Reverse migration effects are severe enough to generate a slight decline in the demand schedule for US gcods, and hence to a price fall proportionately greater (1.04%) than the output increase (t.64%). This perverse irstability does not last, tho.ugh. As .housing and capital investments are completed, they raise labor productivity and lower costs of arban goods, while declines in relative urban living costs due co the US tf? gain further encourage urban irmigration. Fin- ally, a slight decline in the 1961 rainfall index itself provokes rural-urban migration: in the US productivitv gain case these rainfall effects are magnified because the 1960 rural population had grown in comparison with the benchmarki. Abstracting from the cycles provoked, the long run effect of a l" CUS tf-p gain is a 40-50,000 increase in the urban labor force. Most of this gain is in sectors other than US, which tends to experience employment decline: its average output elasticity over the period 1961- 1970 is only .89, despite a slight increase in long run capital stock. US capital formation is severely limited by the isolation of the informal sector capital market, though, so that capital accumulation is constrained by the size of che market's savings pool; US employment is indirectly constrained. In summary, then, it appears that all urban sector productivity shifts have positive long run urbanization impacts, but that these effects are small for the US and PSU shifts. Rural productivity gains have weak long run migration effects but large impact multipliers. Productivity gain in India's largest sector, agriculture (initially employing 71% of the labor force and producing 48% of non-housing GDP) leads co an initial declIne in the agri- cultural labor force, since constant price value added rises proportionately, less than the productivity gain. Agricultural products are used primarily in final consumption, or as - intermediates into the M sector agricultural processing industry. Low demand elasticities, especially from India's rural poor, thus prevent demand for A from rising rapidly, The A, productivity shift nonetheless does have the impact effect of increasing rural incomes con- siderably (real per capita consumption of the rural poor rises 0.,41%), and much of this income growth is spent on other rural consumer goods (RS-and HRS). Real income growth is sufficiently great and demand leakages for urban goods sufficiently limited, then, that the aggregate effect of an agricultural productivitv shift is to raise demand for factors of production in rural areas. This demand increase induces a moderate urban emigration despite the 173,000 worker decline in agricultural employment. The urban emigration generated at the A tfp shift does nct persist into the long run. The initial decline in urban living costs relative to rural living costs operates with a lag to reverse the urban emigration. These effects are reinforced by the end of the initial rural construction boom, though both sets of effects continue to generace cyclical behavior into the mid-1960s. Excluding 1964 and 1970, urban emigration in response to the A shift is very small (especially given the size of the agricultural sector) after the cyclical effects have moderated, averaging only 26,000 in the period 1965-1969. l904 and 1970 were exceptionally good crop years, so that the productivity shift and natural circumstances interact to magnify migration trends. Rural services and manufacturing (RS) productivity advance has short run impaccs symmetric to the KS sector in the A.S growth case. Demand elastic±ties dre sufficiencly. high that RS's :,arsinal -Jalue prcduz: LS5s t.he . ;a ? factar , a:; AS r aces '-icth .,aer rural in; escnE.n anid consurnpcion goors RS c ptnsumpcr: oCo-56 alsc- bt_.efit 'from assu-.ec cnge±s effects, as these goo s ar" thle major r;ral al cerna;tl-e to Cood cCsu.mStip:un. As Capital is .. cile "'ut'Cty-- v", U -o 'ncr3oe 262,C,30 wo -ker.s in RS and 7"C.000 wocrkers n other rra' 5ectnrs3 A hou.: i: i-..b tso. -.-os. ze rural Iigr-nio, -rt:er r--iui.ng demand for lS-produae rJ.c trucc' n. Thre eRS sector's .high lanor sharZe 7.1) a'; enLSUreS that it will enju a:: elastio s;-ort run supply curve, so thlat demand growth uste nIr-E from increaaed Income and new rurai workers (espec4aily skilled) will be miet by risin; output instead of higher pri ces. hiis high elas- ticity also distin.guishes ?.S ronm A, =ich 'nas l.,rZe inic4a'1y irobile capital (16V) and land (30%) shares. As the investment boon in rural housinS ;-klses, and as rhw urban/rural living cost differen- tial declines, the initial urban emigratinr. e£fect iz coitrac-tco-:d stronzly. The average urban emigration level in response tc the ;uRS tap Shl.f is orliy 32,Q0QO, a far weaker resvonse than the long run KS urbanizing influence.. -The asymr;cn:try st,!s.s in part from the governrment expenditure effect: productIvity gains raIse Zaxt revenue, which itt turn are spent on urban products. It also stems from rural demand leakages; as risin3 incczmes and population in- crease the rural savIngs pcol relative to ch- benchmar1-h, de-Mand for urban invest-ment and con- sumption goods grows. In the A,,5 growth case, investmrenc growth effects reinforce initial urbanizing tendencies. However, rate ot return insensitivity of rural' and u-zban^ (both formal and informal) savings pools moderate such investr.ent effects. The mnessage from these simulations is disheartening to indian policymakers interesced in stemming urban growth by increasing rural prat!uctivi-ry. Wqhile short -ur. impacts of rural tfp gains are considerable, their abilicy to steem urh)an growth in the:, long run is quite small. Barring dramatic shifts in government expenditurn patterns or the development c a rural cap- ital goods industry, demand patterns witn large long run leakages lImit ..he extent to which rural technological i'mprovements forestall urbarnzation.5 Productivity advance in rural public -servizes also fails to generate large long run ruraL labor force increases. InicIal increases irn the rural labor force are registered, but this gain largely reflects a rural- construction boom and rapid RS output growch as new nigranta consume rural services in place of urban services. Price movements that generate rising relative costs of rural living and an end to the rursl housing construction boom again make these large multipliers short-lived. While PSR's output multiplier consistency exceeds unity and its employrnent multipliers are positive, the total rural employment creacion effect of APSR advance is small, averaging only 13,000 workers in the period 1951-1970. CONCLUDING REMARKS The asymmetric effect of t-p advance In urban and rural sectors on urbanization found in other CGE models also appear in BVW. However, urban manufacturing is joined by moderns ser- vices in naving large urbanization effects in response to productivity advance in that sector. The partially closed nature of BMW prevents dramatic migration effects from tfp gains in the tradeable goods' sectcrs. Instead, household demard patterns, government expendity_re par- terms and the urban location of investment gocds production combine to make locational leak- ages much more severe for rural sector productivity gains than for urban tfp ir.creases. One might also inquire as to whether sustained increases in the tfp growtth rate generate different results. Suppose, for example, that a more universal green revoluticn. than was actually experienced led to a rise in the annual increase in agricultural disembodied tech- nological progress from 0.6% to 1.6%--an increase certainly close to the plausible maximum gain, If this gain is sustained throughout the period 1960-1970, the counterfactual simu- lation yields large A sector output gains relative to the benchmark (11% by 1970). Yet even this sustained growth increase generates only a modest average (1964-1969) decrease in the utban labor force of 4S,000 workers relative to benchmark values. This shift is minor compared with the average urbanization impact of a one-time improvement to productivity in the M or KS sectors. In summary, unless agricultural technology gains contain strong factor- saving biases not considered h.ere, even sustained agricultural productivity growtih increases will have relatively minor (less than 0.2% of the urban Labor force) urban emigration effects. BMW assumes that all rural non-construction investment goods are produced in the .I sector. To the extent that this is an unrealistic assumption, then, the urbanization asymmetries between rural and urban tfp gains will be overstat.-d. We are grareful co Perrv eidcer and Pr:sad Ch:>*ctamaneni :For e:xcellent ss. .istarice, and to Lee j;rrard for diligenz resear.ih siStaticL. 7!is aeh;e.1Sh /S partihll*- 'fu;.cdad by Th; World Bank as part o. re,s*-iarch vroj-:cr :,-,'72-i!', .Aa-sfis . P. . RENC S 1. Arm,rngton, Paul S., "A vneorv of Uemanc for Pro.iucts Distinguish .2d by ?alce of Pto- duction," IMF Staff Pauers, Vol. IL, No. 1, January 1969, pp. 159-1;'8. 2. Becker, Charles M., Mills, Edwin S. and Williamson, Jeffrey G., ", -Publ'lc Policy, Crb2az ization and Development: A Comrputable General Equilibrium Model o- the n.dian EconomyNV Vanderbilt University Wcrking Faper N'o. 33-W.14. Nasloxille, Tj.,, Jul'y 19d3. 3. _"Dynamic Parameteri.ation of Computable Gener.il Equilibrium Models," unpublished manuscript. .Nashville., TN., April 1984. 4. Kelley, A.llen C. and Williamson, Jeffrey G., .hat Dri.-< Third Wor-d Ci tv Grcwth7^ A DVnani( General Eauilibr±um Aoproac'n. Princecon Univecrr-t.y Fress. PriTnceton. N!J, 1984, forthcoming. 5. , "The Limits to Urban Growth: Stggesticn.s ',or '!acromodeling Third World Economies," Economic De%relopment ar.r3 Cul.tural .ange, to. 30). No. 3, April 1982, pp.. 595-624. 6. Mills, Edwin S. and Becker, Charles M., 2t;:dias in n-dian Urb,,in T elonnent, unpublished manuscript. Princeton, NJ, March l9V4, C