World Bank Reprint Series: Number Seventy-three Surjit S. Bhalla The Role of Sources of Income and Investment Opportunities in Rural Savings Reprinted from Joiurnal of Development Economnics 5 (1978) Journal of Development Economics 5 (1978) 259 -281. ©3 Nortlh-Holland Publishing Company THE ROLE OF SOURCES OF INCOME AND INVESTMENT OPPORTUNITIES IN RURAL SAVINGS Surjit S. BHALLA* The World Bank, Washington1, DC 20433, U.S.A. Received ,aly 1977, revised version received December 1977 This paper investigates the effect that sources of income and investment opportllnitico have on the savings behavior of farm households in rural India. The panel nature of the data (agricultural years 1968-69, 1969-70 and 1970-71) allows for the identification of the permanent and transitory componer;'. of a household's income. It is shown that income variability (rather than investment opportunities) can account for observed differences in the propensity to save out of different sources (agricultural/noni-agricultural). A direct test of the effect of investment opportunities -- savings is offered ir. the second part of the paper. It is observed that capital market conditions have an important effect on this relationship; poor houiselholds save more, and rich households save less, in response to an increase in investment opportuntiities. 1. Introduction The traditional view that low incomes in rural areas of developing countries preclude the generation of savings has been challenlged by the contention that low rates of return to inmestment, resuilt in low savings. This 'new' view' implicitly assumes that aggregate liotiseiol(d sa\ ings r espolid positively to increases in rates of return to investments (savings). The assumption is important because of its inmplication for development stra- tegies; in particular, it suggests that the setting up of financial intermediaries, and a 'realistic' interest rate policy, will not only lead to an imprrovement in the capital market, but also to an increase in rural savings, capital formlation and growth. The absence of detailed data on financial intermiediaries impairs the testinlg of any relationship between interest rates and rural savings. However, faJrn households (a major proportion of 'saver' hotiseliolds) receive a return on *An earlier version of this paper formed part of a report written for the World Mink, Aspects of savings behavior in ruiral India. I am praieful to the inltitutioin for its financi:al s.upport. I would like to thaink Surman K. Bery, Robert Z. Lawrence atnd Plillip Musgrove for helpful discussions, and anonymous referees ft's comments on an earlier draft. The views expre,sed in this paper are mine alone. tSee Schultz 1964) for a detailed disci:'sion of this view. Also see McKinnon (1973) and Adams (1973). 26() S.S. Bhalla, Rural savings their savi-ags through on-farm investments. Evidence on household responses to differences in these rates of return (investment opportunities) can be used, therefore, to generalize to a statement on the yield sensitivity of savings, e.g. if savings are inelastic with respect to rates of return on investments, then interest rate policy will niot have a substantial effect on household savings. Differences in investment opportunities, however, are difficult to measure; consequently, a direct test of their effect on household savings has (ap- parently) never been conducted. A popular form of an indirect test is to divide income into its sources-profits and wages for aggregate data, and agricultural/non-agricultural income for rural household data.2 Observed differences in the propensities to save fromn the respective sources are then ascribed (partly) to differences in investment opportunities. Section 2 offers a critical assessment of these studies. It is argued that the emphasis on investment opportunities as an explanator of savings may be misplaced. An alternative explanation, [Friedman (1957)] - namely, that proponsities to save reflect the variability of inr,ome streams-is offered and empirically tested. The basis for these tests is the data collected by the National Council of Applied Economic Research (NCAER) on 1980 farm (cultivator) households in rural India for three years -1968- 69 to 1970 71. This data has detailed information on the production, consumption, savings, and Source. of income for these households.3 Section 3 is concerned with a direct test of the hypothesis relating imcstnient opportunities and farm household savings. The theoretical basis of the lihpothesis is examined, and it is shown that assumptions about the nature of the capital market crucially affect the conclusions regarding farm household savings: in particular, a perfect market for funds will imply a decline in household savings with an increase in investment opportunities. The attributes of a proper measure for investment opportunities are dis- cussed, and the longitudinal nature of the data exploited to constluct such an index. The section concludes with an empirical test which conforms favorably with theoretical predictions. Section 4 summarizes the conclusions of this paper. 2. Sources of income and savings The propensity to save out of different sources of income has received c01nsiderable attention in both the growth theory and economic development liteCLLratCr Identification of the source of income (prorits, wages, agricultural iiicome-ic, non-agricuiltUral income, etc.) is not relevant per st for a study of savNings belhavior. A rupee is a ruipee and presumably the lhouseldll does not 2A partial listing of these studies includes Kaldor (1955), Houthakker (1965), Kelley- Nvilliams,on (1968), Shinohara (1970), Ong Adanis Singh (1976). 'See appendix I for a brief description of the data. Also see NCAER (1974). S.S. Bhalla, Rural savings 261 determine its behavior on the basis of the source of income. Rather, its importance is derived rrom the assumption that such a classification allows one to stratify households according to differences in the economic environ- ment. Thus, sources of income become a proxy for economic unobservables, and conclusions about the effect of the latter can be drawn from observations on the former. As mentioned previously, one particular economic unobservable for which sources of income can act as a proxy is investment opportuinities. A traditional method of analysis is to divide income on the basis of occupation into two sources-profits (entrepreneurs) and wages (workers). The higher observed propensity to save out of profits is then interpreted as supporting the notion that savings are interest elastic-this under the assumnption that capitalists systematically face higher rates of return on their investments, and therefore save more.4 However, alternative explanations are possible. Profits and mnurginal saving rates may be positively correlated with levels of permanent income-5 hence, the higher observed propensity to save out of profits. Further, differences in the variability of the (profits and wages) income streams may cause differences in the observed propensities. If it is assumed that en- trepreneurial income is inherently more variable than salary incorie, then at any given level of income, transitory income will formi a larger part of total income for the self-employed. Thus, any cross-section data will show a higher saving rate for the entrepreneur.6 The above explanations for the higher propensity to save on the part of the capitalists are not mutually exclusive. Thus, there is no way to choose between the different explanations - is it the higher level of permanent irncome, oi' the greater variability, or profitable outlets for saing o r just tastes that cause entrepreneurs to save at a higher rate? The longitudinal nature of the NCAER data allows one to be more precise about the different causes of savings. In particular, a relatively rigorous test of Friedman's contention, that it is the variability in income which accounts 4The assumption that capitalists face higher rates of return on their investments is plausible. Real rates of interest on financial savings are low, if not negative, in most developing CoUntries. See McKinnon (1973) for an elaboration on this point. 'Both the permanent income hypothesis (PIH) of Friedman (1957), and the life cycle hypothesis (LCHI) of Ando Nlodigliani-Brumberg assert that saving rates are in1dept-11Cnint of the level of permanent lifetime income, This 'independcnce proposition' has proved to be con- troversial and a summary of the conflicting evidence is provided by Mayer (1972). It shouild be noted that not a single study for a developing country supports the indepenidnce prtplosiiiion. See Bhalla (1976a, 1976b) and Musgrove (1974) for a refutation of the proposition for India and Colombia. respectively. In these studies, saving rates are shown to increase with the level of permanent income. 6This requires the aissumplion that the marginal propensity to save (All'S ) out of transitory income is greater than the MPS out of permanent income -an assumption supported by most analyses of savings behavior. 26_2 S.S. Bhaha, Rural savings for higher observed propensities, is conducted in the latter half of this section. But first a 'traditional' analysis of savings and sources of income is presented. 2.1. Re.SL'arhT ivithjrrni data The easy distinction between a capitalist and a worker is lost when one attemipts to study the sources of income hypothesis for farmers, The income from a farm is a return to both labor and capital, and cannot be easily separated into its components. Farm incomes can, however, be divided into their on-farm and off-farm compone-nts, and thus a division of income into separate sources can be achieved. This differentiation is comparable to the traditional capitalist,worker dichotomy since a major component of off-farm income is labor income. Thus, it can be asserted that a household which derives all of its income from a farm systematically differs from one which derives only 50 perrent of its income in a like manner. The former is likely to have greater cointrol over physical assets, a higher ratio of proflt/income and (pe rhapzi) greater investment opportunities. This sources of income hypothesis has been tested by Kelly-Williamson (1968) for Indonesia and by Ong et al. (1976) for Taiwan.7 The empirical methodology of these studies consists essentially of first grouping households by their share of agricultural income in total income, and then estimating the following equation for each group:8 S=t±+b(Y1+ Ye), (1) where S, Ya, and YO represent savings, agricultural income and non- agricultural income, per capita, respectively. (For a complete derinition of these variables, see appendix L.) In both studies, the average and marginal saving rates are found to ilncrease with the share of agricultural income. This result is then interpreted by the authors as supporting the theory that entrepreneurs, or farms with gi-eater control over assets, save more; alternatively, the higher savings rate is presuimed to reflect the higher investment opportunities of the full time farmer. (jiven the capitalist/worker (or profits/wages) dichotomy, this result is not SUIrpiking, a priori. However, there are two problems associated with the letod(lology of these studies. Firstly, estimating an equationl like (1) forces Other studies of this kind include the following for Japan: Mizoguchi (1970), Noda (1970), andi Shinohara (1970). 8Kelly Williamson emnplh) a slighll) different estimation procedure i.e. they estimate equa- tions wlhich reflect the cumulative proportions of income that are derived from agriculture -11 percent, 21 per-cent ........ 91 percent. S.S. Bhalla, Rural saviings 263 the propensities to save to be the same within each group; thus, it is impossible to test whether propensities to save differ by source of income. Secondly, an increasing ratio of agricultural income/total income may be positively correlated with increasing total income. (The simple correlation between the two is 0.24 in the NCAER data.) If savings rates are associatc(d with the level of income, then what one observes by running regressionls of ratio groups is simply the effect of higher incomes rather than the effect of 'control over assets', 'entrepreneurial income' or 'investment opportunities'. A 'proper' test of differences in propensities to save income is to estimate an equation like S=a+baY+ b0Y, for each homogeneous group of households. Differences in b0 and b0 can then be interpreted as indicative of differing propensities to save. The NCAER data for the 1970-71 agricultural year was used to estinulte eqs. (1) and (2) for 1980 farm households. In order to duplicate earlier research efforts, the sample was divided into six groups hlouscholds group)eLdi by the share of agricultural income in total income, Sa( <50 lpercent, 50 75 percent and >75 percent), and land owning categories (<5 acres, 5 15 acres, and >15 acres).9 The results for eqs. (1) and (2) (and associated data) are presCntCeL in table 1. As also observed by Kelly-Williamson and Ong et al., the orer1l{ propensity to save income does increase with the share of agricultural income and land ownership. However, as the table makes clear, the size of income also increases and therefore attribution of differences in the mnarginal propensity to save (MPS) to sources of income is qucstionable. ;e-garding eq. (2), it is seen that except for one group (Sag, 5(Y75`"i), the MPS out of non-agricultural income, b0, is always higher than the .\IPS out of agricultural income, ba The two coefficients are significalnti) different for the following groups-all households, Sag>75 %, and land categories <5 acres and 5-15 acres. For the exception group, Sagg=50-75',0, b17 is less than ba, but the difference is not significant at the 10 percent level of confidence. These results are contrary to expectations. If it is assumed that Y1 roLuglylN corresponds to profits and Y, to wages, then one should have obscrved that ba was greater than b,. Analogously, the size of Yq, could be a proxy for investment opportunities-again, prior belief would indicate that hb,>bhe,. Though contrary to other hypotheses, the result that b0 > ba is entirely consistent with the interpretation that sources of income merely reflect the 9The classifications were chosen to create homogeneous groups it is impliciltlv assumed that the size of farm owned and/or the share of agricultural income are indicative of asset holdings, ratio of profits to wages, investment opportunities, etc. Table I Sources of income and savings -cultivators, 1970-71. Sample averages: Regression coefficients: Average Land savings Eq. (2) Eq. (1) owned Sg Income, Y rate, SIY Number of (acres) (%) (Rs.) (°%) b 0 bo b observations Land categories <5 acres 2.6 63.9 2838 4.5 0.24 0.31 0.25 810 (2.73) 5-15 acres 9.3 86.4 5099 12.6 0.34 0.47 0.35 743 (3.34) >15 acres 28.8 93.0 8716 21.4 0.43 0.49 0.43 427 (0.80) Share of agricu'-.ral income (S,,g) <50° 4.3 33.2 3322 8.9 0.20 0.38 0.35 360 (0.97) 50-75% 6.8 63.5 3773 9.4 0.45 0.20 0.36 365 (1.70) >75% 13.8 96.0 5766 15.8 0.39 0.53 0.39 1255 (1.98) All observations 10.8 78.6 4959 14.1 0.38 0.43 0.38 1980 (2.30) Note: Values in pareniheses indicate the absolute value of the t-statistic for the difference in the marginal propensities to save, ba and b. ..S. BiTalla, Rtirall savings 265 c0flpositioli of inicomne i.e. their permianent tranisitory n1atuLre. Non- aoricultUral inicomiie is mainily composed of wages and salaries from outside em11ployment. Small farmers sLupplemenClt their annual income with olutside work since their own farms are not large eniough to keep them fully ciiiploved. (OLItside icomille formed otnly 7 percent of total income for households owning more than fifteeni acres.) Out'idc inicome, lo'exr, is likely to be more untcertaini than on-farm income, since it is dlepeid(iclnt on the probability of obtaining a job. Apart from its repular componlent, outside work is also resorted to ujnder 'speciat' (and transitory) conditions of the family. Tlhus, 1,, is likelx to lhave a greater transitory comiponienit than Y-, and if the MPS out of tranSitory illcoimie is higher, then the observed propensitics to sa e (b,, and h),, in eq. (2)) can he explal.in1ed by reference to trarnsitory colmlponlelnts. Though plausible, the above initerpretation remains conjectural. The per- manent and transitorv componenits of sources of inicomiie are n ot usuallN identifiable with one-period data. However, the longitudinal aspect of the NCAER data can be used to estimiate the perianent and tranisitory coMpOneits (to be e\acl. the xa riances of eacch component) of each son rce Of incomne. Thus, the '% ar iabifity of income' hivpuithesis of sah ings can be empirically tested. 'Tle 1m11tlC0LIOLrg l and empirical estillmates follNow. S.2r. .i:mg.N cidi lilt rar,'abiii 1' o/ incoine AI.nalogons to Friedmaini i (1957), each source of inicome is assutmed to be comnposedl of a permanient complronlenit Y; anid Y,,) and a transitory com- ponenit ( Y, and Y"'). Fturtlher, the tramisitor%v conuponlents are aIsslU mCLd to be u ncorrelated withi the permanaent componien t. Thluts the following relation- shiips lhold: y, Y' + y"', i a, o. (3a) co(s y.; ' y;.) i - , o. (3b) Let h'(,") represent the comllllloli propenisity to save out of pernianent (trallsilory) ilcolmc.'" Gkei' e these assumption1s, the 'true' model of sa%iings behavior is: KS' + b'( Y',+ Y,, ) 4 b (Ya, +; (4) "'It is assurmed thatl th ic propCn..i to saNe permanent income is unaffected by the source of the income. Thik &,timpm't ion implies no more thani the simple assertion thiat a (pcrmancnt) rutpee is a rupee, ieg.mmdle- oli h1e source. A test of thi, assertion is beyond the scope of thits puper. Holbrook and Suiaohrd t1971) found roughls constanit permanenlt propensities to save out of different sources of income witlh Ut.S. data. 266 S.S. Blhlla, Ruirail sarings An eqttation, estimated as in (1), S=a+h b(Y',+ Y")+b,,(Y' + Y"), (5) is nins-specified and h, and ho are biased estimnates of the propensitics to save, h' and h". llowe%er, a relationship does exist between the observed pro- pensities bh, bh and the unmieasuired propensities, h' and h" i.e. bo - h,= ( - b') (A+ B +Ct), where A= var Y' var a^ -va r Ya' var YF,, Bcov(Y", Y")(var Y -var YO), C=cov(YQ, Y,)(var Y`-var Y"). Since it can safely be assumed that h">bh' (see note 6), the observed results bha,> can only occLur if (A+B+( )>O. ff the extrme niassumlllption is made that the covariances hetwcen the perinianent anid transitory comnponenits [cov(Y(, Y,,) and cov(sY,, Y`)] are zero, then ha and h)o are equali to hh'(var Y',var Yj)+h l(var Y`/var Y)) (6a) hb h'(var Y>'var Y,)+h"(var Yr/var Y). (6b) In other words, ha and h, are wNeighted averagcs of the shares of permiianicnt and transitory in:omne in each source of incomie: and the statenment that difillrences in ohserxed propenlsities merely renlect the pernianentll/transitory nature of different income streamis is %erified if b,>1)a is observed along with var YO >var Y1, > var Yjva r Y. (6c) The exactitude of the above statement is lessened if one allows for the possibility that the two co%ariances, co)(Y', Y,) and cov(Y,',, Y') are non- /cro. In aratiCulr1i1, if cither B or (' is positirc it can no loinger be asserted that h b. ha because Ya, has a greater transitory componeint than Y,. On the other hand, if B and C are both negative, then one hazs the 'strong' resuilt that b>a h CLbecaLIusC of differences in the transitory nature of inlcomies. Ihle likClih0od Of SuIChI a result is examine( first by looking at thc expeccted sign of terms B and C, and later their mnagnitLudC. ".See Bl1aIlk ( 1976a i I;r algebraic dLt:ajk. S.S. Jlh,. Rurall Suring, 267 On an a priori basis, one would expect the term cov (Y", YO) to be negative for a given point in time. This is because a household is expected to mnaintain a given level of total incomic-unexpecled changes in one source are partially made up by opposite changes in other soLurces of income. [However, one factor that might cause the correlation to be po.wiiirc shotuld be mentionc(d namely, the weather. One componenit of non-agricUltural income is the wal'cS Cear1nCeL on algricuiltural land of non-family hiouseiholdls and it is expected that incomes from this source will move in the same direction as agriCUltural incomes.12 The net result is therefore ambiguLous, thOLuIgl it is likely that the correlationi will be negatlke. Amnongst groups classified by land holdings, one would expect the cor- relation between the permaanent componen-ts (Y', Y,) to be positive. This is because itjs expected that factors like education and access to nmarket affect both inCo0i1e simultanCoLSly. If these factors are associated positively with other deter minianilts of on-farm income (land, other assets), thell a 'high `(eLuCatioL1' household is likely to have a higher on-farm income as well as higher off-farm incomie in the form of salaries, n(on-farm busIillcSs etc. Howvever, if the cla,',ification of hoiuseholds is s,Ufficiently broad, this cor- relation can be negative. For ins,tance, a large land owning houIehelold is likely to keep its famil imembhers emnployed otn the farm, whilst a smlall frmll-I 11heLIel1(ld is likell to npplcnleent its labor in1com)ne with work off the farm. The probability of a negati ' e correlation amion tst the pernac n t coinl- p(onecnlts is reLduIcel (but not elilmlil.ItCd tby the fact that houchohllis 11e beenl clas:sified itito r latli ' el1 0mogLeCnl eoiiL grotips. 'rlTe ternms l'ar Y, -var Y,) and (var Y,-var Y,) need to be e\alined before the sin. of B and C' are dcectrinied. If the (reasonable) asssulmlptionl is made that each souLrCe of income. has appiloxinia;telv thle sallme coefliciei of x ariation,l i.e. var }Y (, )2 a N zIr y, (}, } , i;, are mean i alties. and ir agricultllllral incomne is a mal jor part of incomnc, (i.e. , > 1,), then it fllows that (var Y,, - ar Y,,) is positix c when lnon-agriCulturllal income has a ,reater tra nsitorn coniponeii *4 Additional restrictions are necessar) to 1assign a sign to (var Y";,' var ): the emipirical cidence, ho( ever, ,tr rogl\ siuggets that tisio is llegaziie whien noni-agricul itral inicomiie is miorc t ransitor\ (See table 2). T'hls, (\,r Y,, --var Y' ) is positike and (%a ar var -- is Siici k .pe..king the sign of the correlatttin is nlrik.ear and dependent on the relaliwe shift itn tile labor diendiiii(i curse oniwtvardl) iar A :h1 alaor supply ctrve (ouitwa.rd) thatil is caused by the had 0calieLr 'I \cept for otne laisificationl (Sn,,, Sw%,) 1, tis is the case for all grouips. See table 1. var 1<;, var 1,, - var Y" uar t,, tihen w.' ),, ar I, -var Y' var Y,, or -ar Y' var Y;, sar 1, var 1I, or air Y' var Y,,} i0 ) 1. 268 S'S. Blialla. Rurat'l savings, evpectcL d to be Ineati%e. TIhis Couplled with the assupinlrli160l that cov(Y", Y,,) is negative and c () Yr,) is po>iti%e, iniiplies that B and C are bothl Ileg'ative. Thus. unider 'expected' confilioin, one obtains the 'strong' resull that b, >J7 11Cfl nonii-agricult ural ineomiie is inmore t ranitoly than agricultural income. cii.J if the correla tionis are of the 'wronig' sign, it is likely that the Magiiiiii(les of 1B and C are n11ulcl smiialler than 4, atid tl heefore negligible. T\pandiiio ternms A1, B, and C otne obtains 'tar Y,,(var Yl+cov(1 . Y);;)+var Y,,(var ,','-+cov( Y,, Y, )) i cmW(1a. Y,,)(var Y: --- var Y},) >, for he,> b,. (7) If the varianices are m11ucih larger thiath the cs ariance. as expected, then the positive niatulre of cov( Y,', Y" ) IaLdd oiilv a iiCgligible a1m1101.unt to thle first two terms. Further. tile (small) no.ti\c covai.iiance betweeiI the perIm1aInInt comiipit lenits is mLultiplied by thle dliffeicine betwveen the tranlsitory varianices. TIhus it appears tlwi a strlonIg test of the ;%ariahilitN' hypothesis of savings is the associatioll of the obscred lpropenisities h,, >b,,. with the relationiship var IY ^ var Y, var Var1T' v , Ta%) avumptionhos are incccssNai\ In order to isolate the %iirianIce,s ol the perm111a;neit' and tran1itt)rv coiimponent . Firstly, it is assued thlat tri";11tory ICoILes. dLI of 6, . L tCCes, are Li11correlated withl each other for the fir-st anltd thlird 'e.ar olf the survey. TIhlle secolnd aSiiMpt ionl ConlCeL11rS thle l;atioushlip betmwell thle permanent coinpoiicniis or thie first and third year. ( r)Wtll I iniLCnes is especLtd to causeIL a , steC1a11tiC rLeC% iaIOtI between the two ai iales. (CoNkth in actual iiicoiniies for illdikidL111 u hoLusCehl0ds ca oimot be ulsedl to st mllate c\pecttiih ins about pimainen ticomlle since these incomles are 'coitainiated by tlle tii1I1icLIreld) trailsitt)ry comllpolllt. 1urthlllel, thle ',romh Ii x periemiccdl duritlg tIle tll. i eers by a1n d i sL lidi lIKal hiouiselhold mnay- not be a .alid iudicalto of its t -kleted prlofile of permancnl t ilncomne. A pliatsible assIu i.nptionll is that far1tm 1hous101 ehOlds adjust thieir expectatioils about giot 11i in perlI;niilent incoie bN the arerage growth experience of similar lotuselldiLs". Thlis is tlhe asstiiiiptimn uised in this paper-, though qeLstiSoable, it hlould be cilnphaWi;i/ that the cla;ssiricationi of farm 11ouNseh0olds into ri ait 9i c lv h 8 iii'genon grotups ict-rase, (tlhe % alid ityv of the assuii inptioni. If the fi, st Na i si almes are repriesented by lower case letters, the assumnp- tions Zare as follows: cw° (1.l'iY" ) t - ) i a,1l O. (Xa) I.& zero) correlation between transitorv terms Lwo vears aplait imiplies a ii. orinii of iii, years. f \c,' 'rkhno to lFri-edmnan 1. ' p. ti. oe definition of the hohrion is 'ihe lenIgthI of titme a tactor mlutist aitc Income before ii 1s cotisidlere(I pernii nint'.1 rmpirical ,upport for this assuniptioni is proxided in Blhalla (9t. -h) where it iS shlown that three years is an upper hound estiniate of the horizon for farm households in India. S.S. Bhialla, Rtural savitngs 269 'a-=gaY" g.= E(y.)IE(Y.), (8b) Y = g0 YO g o = E(y0 )/E(t )'). (8c) Eqs. (3) and (8), and the covariances of income in the first and third year of the survey are enough to isolate the variances of permanaent and transitory components of income: Co V (*\a Y.) -ct)v(.la, + Vil, Ya +ya ) ( =''hcovb(i., YUY), =ga var Y'. (9a) Analogously', eov(y,s Y,)=g0var Y'. (9b) Eqs. (9a) and (9b) allow one to cstinuite the variances of the permiianeilt com- ponents. Estimlates of the transitory variances are obtained fromll the relationship, var(Yi)=var Y' -Fvar Y', i - a,. (10) Table 2 Ipreseniis the results for the estimated prioportions of inlcome Nariance, by soLurce and type of inlcomne. Estimnates of h)0 and b0, rieported in table 1, are also prcsented. It is observed that the result b0 > ha is alwva3ys associated with the result var Y"' Ic Y<. var YI/var Y. In the one case that bh> b,0 the opposite (and consistenlt) result that var Y",,,\'ar Y, 15 All acres acres acres <50 ',, 50-75% >75% observations Mvean, I, 319 659 1060 159 330 816 607 Std. dev., YI, 315 585 906 150 260 722 652 Mean, Ye, 153 89 63 318 188 26 109 Std, dev., Y, 176 189 205 267 157 94 191 Mean, ya 316 572 859 228 343 670 529 Std. dev., ya 363 582 797 294 355 672 601 Mean, v, 186 123 117 281 177 100 147 Std. dev., vy 383 221 246 486 222 237 304 corr (Ya, y) 0.564 0.537 0.453 0.593 0.548 0.494 0.559 corr ()K,,) 0.241 0.499 0.411 0.290 0.593 0.148 0.357 var Y":'var Y, 0.34 0.38 0.51 0.19 0,28 0.44 0.41 var Y',var YO, 0.57 0.58 0.73 040 ().11 0.90 0,58 NIPS Y., hla 0.24 o.34 0.43 0.29 0.45 0.39 0.38 MPS Y,, ha 0.31 0.47 0.49 0.38 0.20 0.53 0.43 Notes. i1) Upper case lower case represent third year,first year values. (2) Y',, Y' represent the transitory components of agricuiltural and non-agricultural income, respectively. the respoonse of investment opportunities to saving. The importance of this subject was emphasized by Schultz (1964), who, in his classic study on traditional agriculture, stated that 'alhllough there has been a long standing concern about the effects of the level of per fanmily inconme upon percentage of incoime that is saved, there has been no comiiparable concern about the effect of difference in relative prices of new income streams upon savings and investment' (1964, p. 74). It is the purpose of this section to test directly the effect of investment opportullities on savings. The term 'investment opportLuliLy' was first formalized by Irving Fisher. He defined it as follows: 'The conicept of investment opportunities rests on that of ani "option". An option is any possible income stream open to an inldividual by .itiliz.inig his resources, capital, labor, land, money, to produce or secure said income stream. An investnment opportuniity is the opportunity to shift irom one such option, or optional income streanm, to aniotlher' (1934, p. 151). For farm households sLuch as thlose analyzed in this paper, an investinent opportLunity can be defined to be a perceived shift in their produictiori possibilities. The greeni revoltitioni years of the late sixties provide a rare and ideal representationi of shifts in procluction possibilities. The introduction of high yielding varieties (HYV) afforded farmers the chance to increase significantly their rates of return from investment; this investmiienit opportunity meant S.S. Bhalla, Rural savinigs 271 increased profits from both old capital (irrigation) and from new investments. But did this improvement in the return to existing and additional capital result in more short-run savings on the part of farm households? The effect of investment opportunities on saving: is analogous to the effect of interest rate on savings, e.g. Wright (1967), Weber (1970). These studies, however, were done in the context of time series data wvith all households assumed to face the same rate of interest. In this paper, cross section variation in the investment opportunities faced by otherwise identical households plays the role of interest rate variation in time-series studies. The variation in investment opportunities is best illustrated by fig. t, which shows t2 Fn, t.1+ T0 F0 , . . TT0 z Yr Y1 tl Fig. 1. Investment opportunity loci faced by F. (old technology farmer) and Fn (new technology farmer). Al PN A oi Z T0 z IT T, YI B B' C (a) Perfect Capital Market (b) No Capital Market Fig. 2. Investment opportunities anld savings. 272 S.S. Bhtltia, Rural sarings the technology frontier (T0TJ) faced by an old technology farmer, F0 (e.g. farmer in a Bihar district) and the frontier (TnTn) faced by a new technology farmer, F,1 (e.g. farmer in a Punjab district). Prior to the introduction of the HYV's, in time period t- I , Fn earns an income Y- which is less than the inconme earned by F,. (Implicit assumption is that F,, ownis a smaller acreage of quality adjusted land than Fo.) With partial and inmmcdiate Ldoption of HYV's F,1 now faces a new technology friotltier, T0Tn, in time period t-l1 and beyond. Since the analysis pertains to 'otherwise identical' houscholds, both farnmer s are shown to have the same first period incomne. Y;. With investments, each far;ner can enjoy a higher income in pcriod 2, Y, by moving along his teclinolotty frontier. (Note that in%estmnents only affect income in futuire period: Z is second period income with zero first period investment.) TnTn is shown'o to have a higher slope at each investment point in order to incorporate the assumption that the new technology is more profitLible. The tlcoretical relaLtioinsliip between investmcnt opportu nities and sa-.inlgs can be shown by uise of diapgams similar to fig. 1. These two-period diagrams illustrate the optimal consum11zrption amid in'estnicni decisiotn oin the part of a farm lioiishold.`' These decision- are crtucizally affected by the nature of the capital imarket in whliclh the lhousehold operates: tlius, two polar cases are nalvly/ed (1) a perfect capital niarket (richi farmiers), fig. 2a, mllid (2) no capital market or a Robinson Crusoc cconoiny (poor farlersl,g ri. 2b. In both figures, the conmparison is between similar farmers (eulL;2l first period inicomiie) facing differenit productive (investment) opporttllities. C( and C2 Irepresenit present and future consumption, P' and P", the ecluLilihrilli points, and AB the (exogenous) rate of exclh:min. hetween the twto period-,. In the case of a perfect capital marel,t. tne Shift in thlc opportJ0 nlitV locus (T0T0 to TnTn) implies an increase in wealtlh (AB to A'B') with the relative price of consumption in eachl period (i.e. the intcre.ret rate) held colnstant. TuLIs, there is no substitution erfect (change in relative prices) inwollved with a chalnge in the locus. If it is now ahsumed that Consupnltion in each period is a superior good, then the point of tlangency for farnmer F15,P", has to be in the triangle, P'QR, represenitinig increases in co1nsUmption in botlh time periods (wealth effect). Tlhuls, rirst period comstimiption is higher, and saIvings lOWer. for lhouseholdlo with greater investment opportunities: i.e. rich new tech- lnolony farmiters sas se less than rich old technology farmers if credit is freely a'. alilable. In the self-finance worldl of Robillson01 Cruoe,C thle efectC of a slift in the tchnolot10gy locus0 on first perid)(l in\cstnment (savings) is a priori ambiguous. As shown in fig. 2b, consulmption in the First period can cithecr inlcreae or "TlheT heor'. of inter-temporal choice topmlimil investnient decision) was dexeloped by Fishier and elaborated by Iiirshilcifer Ioi5x, 1970). The reader is refer-ed to lliiihleifer for a detailed anialysis of investment choice in the context of a ti%0-PLerl model. S.S. Bliallk, Ruralshatvings 273 decrease. The reason for the ambiguity is the simultaneous presence of both wealth and substitution effect associated with the change in technology. The point of tangency for farmer F, will be the left of P' (more first period savings) if the substitution effect iominiates the wealth effect.' The arnbiguity in the savings behavior can be removed if one makes the (realistic) assumption that 'lumpy' investments are niecded for a successful adoption of the new technology. In this instance, T.Tn might not be the appropriate representation of the technology faced by the Punjabi farmer. Even though Fn has increased the flow of income by adoption of the new technology on available irrigated land, further 'lumpy' investments in irri- gation and land improvement may be necessary to successfully adopt the new technology. In a Robinisoni CrlLsoc economy, the resources for this additional investrnent can only come from increased savings. Thus, in rig. 2b, if savings greater than Y1 -P' are needed for transfer to TnT, then those households planning to use the new technology have to save greater than Y1 -P' in the 'transformation' years. Hence, poor new technology farmers are likely to save more than poor old teclhnology farmers. Thus, if discrete investments in the new technology are possible, the effect of investmllent opportunities on savings is arnbigLuOLs: if sllulfiCiClt 'ILll1JliniesN' is allow\ed for. the effect is positive.'8 In summary, the theoretical results indicate that the effects of investment opportunities on savings depend on as;Uniptions about the capital market. In a perfect capital market case, the effect is an uiianabigLIuos decline in savings."9 In a self-finance economy, the effect is ambiguous, unless lumpy investments are assumed. In the next section, an index reflecting the different investment opportunities is constructed, and the propositions outlined above are empirically tested. 3.1. Investment opportunity: Cmn.strict ion of itidex antil empirical rIes'ults The hiouseholds covered by the NCAER survey presumably face different marginal rates of return on any new investment. This likelihood is enhanced by the fact that the period covered by the survey- 1968-69 to 1970-71 - encompases th- years when the new technolog,y was being adopted by farms in India. However, the construction of an index whlich will reflect "7The wealth effect of an increase in investment opportunities is depenident on the size of savings. For poor people, this, and consequently the wealth effect, is Ei(cl) to be small. This is another argument for expecting the n?et effect of a change in productive opportunities to he positive for poorer farmers. "MNlcKinnon (1973, p. 13) also uses the lumpiness in investment LIrniption to argue that a lack of access to external funds implies that 'the conistraint of Nelf-fin:nce sharply biases investment strategy toward mnarginal variations within the traditional technology'. "In the case of an imperfect capital market, the effect is anibigmmow,h a priori, and dependent on asstimptions about the relative costs of borrowing and returns from investment. 2174 S.S. 1 halla, Rural savings differences in the perceived rate of return on in'mestnments is not an casy task. Any attempt to constrUct such a meaClSlur'e iS fraught with d1ifficultieCs: ne%cirtheless, the index should incorporate the following factors: (ai) .4doption staltts (fiY ci i iilii iitionj i,i io -H Y V cutira tion) (?I lhotuselolds: Thie reasoni for inicludinlig this in formiation is ohviotus. The niew teclhnology, if adopted piroper-l\, cani signuificantly increase thle reCtLrnis from cultivation. (Ib) Extent and type of irr'igaitionz: The new techlnology is hcavily de- peiidleiit for its slccess on the availability and controlled supply of water. A tube well investment may tlherefu oie be expected to yield diffcrcnt returns thlan a canal irrig:ition s% stein. (c) Regionatil loction.: Lanid dLiffereilces are likely to aiffect the pro- fitability of certain in estmients: fuirtlher, the percept ions of a farmer in a suLccessrl] area of the green revolutioni (e.g. Punjab) probably dliffers sys- tematically fiotm those of ian identical farmer in an u1nsUccessful region (e.g. Bihar). (d) Cropping pattern: The cUlLtiVation of wlheat has pro%ed to be imiore profit.hle than the cltivat6ion of rice. Idenitical on-farm iVestillmen t. wlhell used in the pr-od tintitn of different cr-ops, are likely to yield differenit returns. (eL) Fertili;eer and cr1edtlit: SiLCCesfL1l aidoption of the niew techlinolo\g is dependent on the zi%.tilahilitv anid proper application of fertili/Clrs, PCStijL;OS etc. Workin- capital ]reltiircnlents incrcazse significantly witlh the new tect- nology: thils, acclv, to working capital, and its cost, affects the perceived profitability ofr the niew tec nolio,y. The abo\ e list of relevant factors affecting an i\ estmient oppoirtulity index is nceessalily inconinplete : it, iie crhlielel,s, pohints, olut the diffictulties in ol 'CLl in thle c'onIStructiOn or suchi an index. A proper definition of an index ihtl\\ithstanldiii, the pLurpose of this section remainis the testing of an in' esin1lent opportu nity effect oni sa\inus. The previOlIs discussion has made clear that the adoption statLus of a household should be included in any a.s sessiment of the rates of retturn perceived by a h1ouselhold. Tlhus, it might seen-i appropriiale to classirf households according to whether they have, or havxec not, adopted the technlology. This index', howe r Clr has sever al draw- backs. It a1ttributCs eqU'al op1portunitiCs to an adopter regardless of the crop) oro\n. or its regional location. Most iinportan-11t ly, it a IttribU tes no iii est lmlet opporulnity to a farmer who might ery well be on the verge of adoption. Anl alternative in(Lex of dIifferingg pelrceptions of in\et111mLet oppol'1rt nit ies is thle adOlotion l'ate of the niew techinolog) in the district in wx'liicli a holelh:old residel&. Soil quality, p:nttern of crops priod ced (whieat, rice, etc.) priesence of cxtension programis, credit a\ ailahilitv (povernmient. co-oper;atiies etc.) are all variables whlich vary more ainongu'd districts than amongst hotuseholds within a district. Dilferences in profitlabilit) of investmient hetween. regions should be S.S. Blialla, Rturtal savings 275 reflected in differences in district adoption rates. Moreover, use of this index assigns the same investment opportunity (IO) to all hoLuselholds within a district, regardless of whether any individLual 11osehold1 had actually adopted the technology. Comiiparing non-adopters. it is likely that a farm hoLIseChold in a Punjab district perceives a greater investment yield on its opportunlilties than a household in a Bihar distr-ict, and differences in district adoption rates should reflect this perception. Though not perfect, the district adoption rate comes closest to a desired index of investment opportunities. The NCAER survey contains data for 2952 cultivator households in 1970 71. These households were agnlrCu,atcd into 100 districts, and a weighted percenltLige of adopter households was estimated for each district.2" This percentage was then assigned as all I0 index to each household within the district. Since the analysis of savings is for the third year of the survey, perceived opportuniities of a houisehold are likely to be based on past rates of adoption. Consequently, district adoptioni rates for the second year of the survey were chosen for analysis. The hypothesis to be tested iswhether, ceteris paribus, investment opportu- nities have a positive effect on savings. An imiportant determinant of savings which one would like to control is the priinclmewt inicoimle of a housceh(ld, Ye,. (This roughly corresponds to the flow of inicomiie proportionlal to Y; in fig. 2.) This, however, causes additional problems. Whtil mieasure sholdl(i onle Lise for permanent income'? In Bhalla (1976a), two conceptually different ineaisures of permanent income (Yp) were constructed, Considerations of discount rates, expected income and expected growth in income dictated one measure, Y,. Thus, a discount rate of 35 percenlt, (corresponding to a three year horizon see note (15), and an expected growth rate in incomes of 3,5 peicelnt per year, yields y>=0.437Ys,+ 0.374Y Y +±0.275K ,, where Y0 is measured income, 197(>71. An earnings function approach dictated the other measure of permanielnt income, Ypx. This method rielates the deLerminmants of inlcome (X) to meIsuLreed income, Y; i.e. the regrecssion Y=boi +h1X1 +b2-4± . .....h is estimal.ted and pIredicted values of Y, a are taken to be estimate,s of permnlanenlt inicome, TihouigIi aiccepltable, there is a ser-iouIs problemll with this approiach. The residLuals (Y- Y eY ) rather than idlenitifyinig the assuilled 20The NCAER survey did not sample its respondents on a random basis: conic(iuenitly, population veihlits are used to obtain eiLin:tcs on a district basis. 276 S.S. Bhlotllto. Rural svtings transitory components of income, might reflect just the opposite. i.e. differ- ences in permanent levels. The panel nature of the NCAER data amoids this bias and makes possible an imnproved 'earnings function' estimate of Y,, in particular, :t allows for the deconmposition of the error term ei into an Linobserved perm;anent comi7ponienit ,8i and a 'true' error term, ei. This fixcd effects' error components model yields ya- + i0.1821I + 0.33K ± 188.7I + 5.67', where the determninLants of income are individual conimtants ( 1a. lnd value (H), capital assets (K), famiiil) labor (L), and the level of techlnologv (T). LFor details on the construction of the measuLres Yi and Y, see Blhalla (1976ab.] In order to provide a partial clheck on the robutisness of the rsullts, both measures of pernminent income are used in tlhu cnmpirical analysis. For ak of a 'complete' anaksk. tests using current income (airguied( by somiie to be the relevant 'ariable for savings behavior) for 197kY 71 are also prrc.ented. One problem1 renmains how shiould hiouisehlolds be cklasiried into 'R obilnso Crlusoe' and 'perfect capital market' econiomiiies. It is difficult to assert that any rural hlouLeholl in India fazces a perficct capital mairket. lo't . 1IL'Cr. anll approxirmation to (ilffereniccs in capital inniakcis caini be achieved by a di isiori of households into self-ftinance anid 'openi alccess' cilatories. If thle wealth (perim.ilne i1come) of a h)ousleold is inidicatki e of its atbility to boriow, tlheni the arbitrair- (but plausible) diN ision of hotuselolds intito 'ihsistence, nion- subsistence categories is nitinginful. lidenitificaittiol of a slubsistenlce leVel, however, is a difficult matter. The subject has beenl discussed at lenig,th in the Indian literature and the general consensus seems to be that an annual income of Rs 450 per capita. 1970 71 prices (correspondinLg to Rs 15 Rs 21) per month, 196(> 61 prices) a1deqL.ately des1C>ribdes the sUbistlence (or pov,erty) level. Tlhus, ho1uselholds have been clas.sified according to % lietller their aVerage per capita incomiie, Y' , wt as abowe or belor- Rs 500 a conscrxvti%e estimate of the subsistence le'el 21 Tlhouiglh the 1ion1-sLubsistencie groulp is unlikely to face a perfect caLital market, it is likely that its access is considerably greaiter than the suLb.sistC1lCe gronlp. If the index of imc stment opportunities, 10, is presumlaed to be roughly ilndiCa1tiVe of (difcrenices, in rates of return earned by the lhouseltold, then the effect oni oavinSf 10 can be tetled by estimiaiting, eq. (11): 22 21The clhoice of llr, (ratio of the sum nof iii ec year iniconmes aindt family sizes) i as a Clla'ssiiciiioii variable was dictated by the need to halve a com1rmon sample of lhtoselhtolds for comnlpirkon of the estimates yielded by the twot 'preferred' measures of permanent intcomle Ypi and Yp, "A level iegre.siitm in sa imm: estimated as in etq. tlll resuilts in errors whiotse *amtnce increase witlh permanent inicomiie. If thle .1iiL.imioii1 is made that the Nariance of thle residluals increases with the sqtuare olf permanlenit income. thent the liciciwced1a,iiiciiy present in eq. l11) can be corrected by detlni!ine all variables by permanetit incorne. All etl.iions have been estimated with this cotrectio,n for heteroscedasticitv. S.S. BliaIlla, Ruirail %arinzg% 277 S=a1 +a2IO+b1Yp+b2yl+e, (11) where all variables are in per capita terms, and Yp(Y1)=permanent (transitory income) of household, 1970- 71, IO=investment opportunities index (weigh- ted district average of the adopters of new technology, 1969--70). This equation (with the rate of interest on savings, r, replacing IO) is similar to those estimated for the U.S. [see C. Wright (1967), Weber (1970) and L. Taylor (1971)]. These authors make the assumption that r reflects both the cost of borrowing and the return on lending. Due to the presence of opposing income and substituitioni effects, this causes the coefficient of r to be ambiguous on an a priori basis. The results pertaining to eq. (11) are not strictly comparable. The ass,umptionis regarlding the capital market-in parti- cular, that subsistence h1ouLseholds are likely to obtain little credit for investment, and that non-subsistence households obtain such credit at relatively constant rates of interest - remove some ambiguity about the sign of a2. The coefficient is predicted to be positive for poor households and negative for non-subsistence. The empirical results strongly support these aLssertionis (see table 3).3 For the subsistence oroup, coefficienlt a2 is positivc: lholdlinig constanlt the level of permanent income, poor households with greater investment opportunities save more. (Coefficieiit ai is signiricant at the 5 percenit level; for Yp, significance is at the 10 percent level, two-tailed test.) Members of the non- subsistence group ihetter access to a capital market) decrease their savinigs in response to an increase in investment opportunities.24 (CocefFicienit a is negative and significant at the 1 percent level). It should be emphasized that this is a partial effect: it is not contended that richer households save less, or at a lower rate (note that the nion-subsistenice groutp has a MPS out of Y of p 0.36: the correspondinig figure for the uuhskitece group is 0.11). The results are invariant with respect to the par,ticular meCLsuLe used to represent permanent income. This robuStneCS.N in the results, as u,ell as-the fact that the signi of the IO coefficient, a,, changes in a predictable fashion for different income groups, siipports the conitention that the effect of investment opportunlities on savings cannot be viewed in isolation. The firm h soluclo.d nature of farm families makes it imperative that capital mliarket colditiOns be explicitly incorporalted into the analysis. 2 'Similar restilts ar'e obtained if the marginal pi opeii,it% to save out of permanent income is allowed to vary witlh investment opportunities [addition of' term b2 JOYl ? in eq. I1)] and ai, assumed to be zero. AIltmilng for clhiieg.s in both c0, and b12Lt2)0. 1X2 Oi) introduuces niliticollineariiv anid causes some (J0) co ici lclcini to be sitlisticallk insgnficaiint. "4If non-suLlsistencC households are suib-divided into gr,Luip earning above and below Rs. 15t0 per capita, then lie c1c lhciciii (I, is negative, but not siunificant (at the 5 percent levetl) for tihe richest group. Thlis 'incon,ilicni resuilt for the ricli households (top I percent of the rural population) may be due to the oiijlil >aple si/e t125 o(bserBtionsi and the relatively lower variation in the Investment opportunities iindex. 00 Table 3 Effect of investment opportunities on savings. Coefficients (eq. 11) Investment Permanent Transitory opportunity income income Standard Constant a2 b, bJ2 error Subsistetnce YaRs.500 (N= 1065) Yp,= Yp, -181.3 -62.0 0.38 0.33 0.31 0d1984 (13.4) (3.7) (22.0) (16.9) Yp= Ypi -163.6 -57.4 0.36 0.36 0.31 0.1927 (12.6) (3.5) (22.0) (16.6) Yp= Y -130.6 -42.3 0.31 0.19 0.2014 (14.2) (2.9) (23.1) Notes: (1) Figures in parentheses represent the absolute value of the t-statistic. (2) Results are based on eq. (11) of text; Yp,, Yej, Y represent permanent income (earnings function), permanent income (three year weighted average) and measured income 1970-71, respectively. 4. Summary and conclusions 'This paper has examined the effects of sources of income, and invest ment opportunities, on the savings behavior of farm households in India. Tle sources (composition) of income effect on savings was critically examinte(d by relating savings to the agricultUrfl and 11011-agriCUItUrill COmIlpOInen.'L of income. It was observed that the propensity to save out of non-a,gricultural income was liighler than the propensity to save out of agriculturl; inicomiie. This result is 'unexpected' if one presumes that agricU1ltural inconmes reflect 'entrepreneurial income', investment opportunities and 'control over s.sets'. However, the permanent income hypothesis offers an alternative explaniation for the observed result--the MPS out of non-agricultural income is higher because this income has a larger transitory componenit. The longitudinal nature of the data was used to isolate the permanent and transitor X components of income v ariance for each source of income. The results consistently supported the 'variability' hypothesis i.e. MPS out of non- agricultural income was higher when its transitory component was larger. In the second part of th. paper the theoretical relationship between savings and investment opportunities was discussed and it was established that savings rates could either increase or decrease with an increase in these opportunities. The result depended critically on the possibilities available to the household for transactions in the capital market. Two polar cases were studied: (1) perfect capital market and (2) no capital market. Opposite effects of investment opportunities on savings are predicted for the two groups savings should decline for the former group and increase for the latter. The conditions necessary for a proper investment opportunities index was exten- sively discussed, and an index proposed and tested. The empirical results supported the theoretical predictions: investment opporttunities increased savings, ceteris paribus, for the subsistence group of households (no capital market) and had a negative effect for the non-subsistence gi oup (relatiely perfect capital market). Appendix 1. Data The National Council for Applied Economic Research (NCAER) under- took a survey (known as the Additional Rural Income Survey (ARIS)) of 5,115 households in 1968-69 to gather data oni the distribujti(n of incomle. and the pattern of consumption, savings and investment of these householdX,. The sample was selected according to a multi-stage stratified probabilitI design; higher income households were oversampled. The survey was re- peated in 1969-70 and 1970+7l on the same households, and the final version of the data refers to core sample of 4,118 households. c 280 S.S. Blaflla, Ruiral savings For purposes of analysis only households that were cultirators for all three years of the survey were selected. A household was derined as a cultivator if it engaged in any kind of self-cultivation on owned or leased land that was greater than 0.05 acres for all the three years of the survey. (There were 2,952 cultivators in 1970-71: the requirement that households have been culti- vators for all three years reduced the sample size to 2,532.) Further, households were selected on the basis of occupational structure (nlo transac- tions in the land market for any year of the survey) logical consistentcy (savings numerically less than income) and a (possible) lack of transcription measurement error (gross income greater than Rs 500, and a saviing rate of - 150 percent to 75 percent). These 'restrictions' reduce the sample size from 2,532 to 1,980 farm households. 2. Definitions (a) Incomne (Y): The income of a household is defined as the total of the earnings of all the members of a h1oLusehold during a reference period. This income can be business income (farm or otherwise), wages, rents (land and house property), interest and dividends on financial investments and penisionls and regular contributions. (b) Agricuiltturtalf 2non-aigricltu-arl1 incoin ( I ): Any income obtained from work on owned farm (crop incoie, rental of land and agricultural implements) is considered to be agricultural income, Y. All other illcolle (self-employment noni-Farmiiing income, salaries, wages etc.) is considered to be non-agricultural income, Y. (c) Savings S: The savings of a household is defined as the change in net worth and computed as the difference between the change in the value of assets and the chanige in liabilities. This figure is adjusted for capital transfers. In other woirds, houmsclhold savings, S is defined to be: S=dA-dL-dK, wvhere: dA =-Gross change in the value of physical and finaniciaLl as¢iets, dL =Net change in liabilities, dK = Net in/flo w of capital transfers. The savings estimate includes via dA any purchlase of consumner dLirables, and non-miionietized investment that is undertaken by the lhtoisehioldl. Savings in the form of cLu-rcrncy or gold and silvcr are not included due to lack of reliaible data: nor has any adj,:istment been made for capital gains or losses incturred by the 1houLseholdk. Depreciationi on assets is also ignored. S.S. Bhalla, Rural savings 281 References Adams, Dale W., 1973, The case for voluntary savings mobilization: Why rural capital markets flounder, Spring Review 19, June. Bhalla, Surjit S., 1976a, Aspects of savings behavior in rural India, Studies in Domestic Firi.1i1cC 31, World Bank. Bhalla, Surjit S., 1976b, Measurement errors and the permanent income hypotheses Evidence from rural India, American Economic Review, forthcoming. Fisher, Irving, 1930, The theory of interest (Macmillan Publishing Co.). Friedman, Milton, 1957, A theory of the consumption function, National Bureau of F-conoinic Research (Princeton). I-lirshleifer, Jack, 1958, On the theory of optimal investment decisions, Journal of Political Economy, Aug. Hirshleifer, Jack, 1970, Investment, interest and capital (Prentice Hall). Holbrook, Robert, 1967, The three year horizon: An analysis of the evidence, Journal of Political Economy, Oct. Holbrook, R. and F. Stafford, 1971, The propensity to consume separate types of income: A generalized permanent income hypothesis, Econometrica, Jan. Houthakker, H.S., 1965, On some determinants of savings in developed and underdceveloped countries, in: E.A.G. Robinson, ed., Problems in economic development (London). Kaldor, Nicholas, 1955--56, Alternative theories of distribution, Review of Economic Studies 23, 83-100. Kelly, Allen E. and Jeffrey WVliliamson. 1968, Household saving behavior in the developing economies, Economic Development and Cultural Change, April. Lewis, W.A., 1954, Economic development with unlimited supplies of labor, Manchester School of Economic and Social Studies, May. Mayer, Thomas, 1972, Permanent income, wealth and consumprion (University of (Californila Press). McKinnon, Ronald I., 1973, Money .:d capital in economic developmert (Brookings Institution). Mizoguchi, T., 1970, Personal savings & consumption in post-war Japan (Tokyo). Modigliani, Franco and R. Brunberg, 1954, Utility analysis and the consumption function: An interpretation of cross-section data, in: R. Kurihara, ed., Post Keynesian economics. Modigliani, Franco, R. Brumberg and A. Ando, 1963, Life cycle hypothesis of savings: Aggregate implications and tests, American Economic Review, March. Musgrove, P., 1974a, Determination and distribution of permanent household income in urban South America, unpublished Ph.D. diss,zrtation, M.I.T., June. Musgrove, P., 1974b, Determinants of urban household consumption and savings in Latin America: A summary of evidence from the ECTEL survey, Brookings Insiiutiion, Sept,, mimeo. National Council for Applied Economic Research, 1974, Credit requirements for agriculture INew Delhi). National Council for Applied Economic Research, 1974. Additional rural incomes survey, Final Report (New Delhi). Noda, T., 1970, Savings of farm households, in: K. Ohkawa et al., ed., Agricultural and economic growth: Japan's experience (Princeton Uni%ersiiy Press). Ong, Marcia L., D. Adams and I.J. Singh, 1976, Voluntarv rural savings cap.acities in Taiwan 1960-1970, American Journal of Agricultural Econonmics, Aug. Schultz, T., 1964, Transforming tradifional agriculture (Yale tTniversitv Prcss). Shinohara, Miyohei, 1970, Structural changes in Japanese economic developnment, [.conomic Research Series 11, flitoisuhashi University, Tokyo. Taylor, Lester D., 1971, Saving out of different types of income, Brookir.gs Papers on Economic Activities. Weber, W.E., 1970, Effect of interest rates on aggregate conqunirpion American Economic Review, Sept. Wright, C., 1967, Some evidence on the interest elasiiciiy of consumption, American Economic Review, Sept.