Вы находитесь на странице: 1из 13

MARGINAL THE SEASONAL ESTIMATING PRODUCTS OF LABOUR IN AGRICULTURE1

By S. K. NATH Introduction whether marginal the of to THEREhave beena number studies determine in like product labourin agriculture a country India is positive. Some of of the morerecentand betterknownof thesehave been by Desai and [9], et Mazumdar Raj Krishna and Wellisz al. [19]. The lasttwofound [4], whilethe of of the estimated coefficient a man-year labourto be positive, different zero in the sub-sample from first found to be notsignificantly it thesestudiesanalysed of farms which nothireanylabour. Of course, did years. In ouropinion, different data for different ofIndia,and from parts of of product investigations the marginal theseand othersuchempirical whichis offundaof agriculture labourignore characteristic backward a the of mental importance: is thatinsuchagriculture response agricultural it partsoftheyear. This is in outputto inputs quitedissimilar different is thingat the firstly, is vital to do the right it because in cropproduction comeand go rather times variousoperations for right time-and theright of only a proportion thetotalnumber quickly; secondly, during certain and of days in a yearis the response outputto labour,and certainother of needsto be put in on suchdays. Such inputs, keen. A greatdeal ofwork constitute busyseason. the consecutive, days,which notall necessarily are the of Thenthere therestofthedaysoftheyearwhen response output are the to labouris sluggish; theseconstitute slackseason. has Indeedtheseasonal variation agricultural in activity beenfrequently noted. Thus Mazumdar[13] uses it to explainwhyinputof labourand on farms. hisarticle surplus In on output acremaybe higher thesmaller per to thataccording his modelmorelabour labour,Sen [16]-having shown in product resulting a higher hours acrewillbe usedonthepeasantfarm per seasonality farm-discusses recognizing how peracrethanonthecapitalist of authors-suchas a does not spoilthatresult. Similarly, number other and Clark Haswell[3],HansenandEl Tomy[7],Mabro[11],Nakajima[14], Robinson[15], and Wellisz[18]-have notedand analysedone or other in But implication seasonality agriculture. whenit comesto estimating of themarginal it beenthepractice ignore to of product labour, has invariably the seasonaldifferences. in the Similarly discussing uses to whichthese
1 I am indebted to the Ministry of Food and Agriculture,Government of India, for making available the data on which this study is based. Part of the researchexpenses were financed by NuffieldCollege, Oxford. I am gratefulto Ian Little, Maurice Scott, Farouk Elsheikh, and membersof a seminar at Warwick Universityforuseful comments. Earlier stages of researchwere assisted by JeremyCorbett.

376

ESTIMATING SEASONAL MARGINAL PRODUCTS

estimates put,e.g. calculating shadowwagerate,theimplications are a of the seasonalvariation nevernoticed. are It is arguedin thispaper that in production function studiesof crop production the less developedcountries particular, busy-season in in the and theslack-season labourinputsneedto be included separate, as independent inputs-each having different a marginal product.For,whilefor mostindustrial processes (and perhapsalso fortotal farm production of it thehighly mechanized, mixed farms theadvancedcountries) is reasonof in able to assumethattheresponse output variation inputs general of to in -and in thelabourinputin particular-isthe same in different partsof the year,such an assumption unwarranted is regarding agricultural an sector which characterized lowintensity cropping, is by of little no mixed or farming, labour-intensive and techniques.Even in suchagriculture there are someinputs which used onlyin thebusyor theslackweeksofthe are year,so thattheannualamount usedofsuchan inputat a farm represents theamount usedin one ortheother season. Regarding suchinputs, then, no specialdifficultyraisedby the usual specification the production is of function be fitted annualdata in a cross-section timeseries to to or study. on Labour hoursare used in the two seasonsin varying proportions different farms.The slack-season the busy-season and labourinputsare the substitutes each otherto someextent. Now if Y represents value for oftheannualtotalcropyield, thelabourhours Lb usedinthebusyseason, that and L8 thelabourusedin theslackseason,thenourarguments imply as theseasonal labourinputs needto enter production the function separate
inputs thus: Y
=

Y(Lb, Ls....).

Iftheproduction function so conceived is then though marginal the product of ofthebusy-season labour(9YI@Lb) and themarginal product theslackof are the seasonlabour(&Y/&Ls) bothdefined, marginal product theannual (total)labouris notdirectly defined becauseannuallabouris notan argumentofthe production function. the data onthehours busyof Ofcourse, suggested specification requires seasonlabourand theslack-season labourusedin a year. We are luckyin the on havingdata from which, some reasonableassumptions, seasonal labourinputscan be derived.However, regarding someother inputs(e.g. bullock labour)whichare used in bothseasons,our data do not allowus to similarly derivethe seasonalamounts.For theseinputswe are forced the annualtotalsin ourregressions. to follow usual practice usingtheir of Data, specification,and results the Ourdata comefrom first of in year,1967-8, thethree-year study the in Economics FarmManagement theFerozepur of district thePunjab. of

S. K. NATH

377

Thisstudy directed Professor S. KahlonofthePunjab Agricultural is by A. Ludhiana. The data are collected the University, through cost-accounting method 150holdings-10eachfrom sample 15villages.Theaverage for a of size of a holding 12*5hectares, is of witha standarddeviation 8-4. The holdings dividedintofivesize groups, are below6, 6-9, 9-14, 14-24,and 24 and above,hectares.The 150holdings covered about 11percentofthe total cultivated area ofthevillages. area in thedistrict, percentwas under Ofthetotalcropped 32-3 wheat, and 50*6per centundercerealsas a group(including wheat). The other majorcropofthearea is cotton.The percentage American for cotton was 10*3; for and Desi cotton, Fodders 4*5. accounted 15percent. Ferozepur for contributed percentofthetotalwheatproduction thePunjab and 21*4 of 15*6 centofthestate'scotton per production. Punjab initsturn a major is growing area in India for thesetwocrops. The percentage netirrigated of area to area sownwas 62 in the district 1967-8. Mostoftheirrigation in was by canal water, thiswas an uncertain but sourceofsupplycompared to old-style wellsand tube wells. In 1967-8,89*7 centofthe cultivated per area ofthe 150 holdings was self-cultivated, percentcashrented, percentwas fixed-kind 2*7 05 rented, and 7*1per centwas sharerented.Therewas a marked similarity the in tenurial system thedifferent in size-classes farms.Ofthetotalcultivated of area, 98-3percentwas sownarea, and only1 5 per centcurrently fallow. The average intensity cropping all holdings of for together 130percent. was There was somebutnota strong tendency thelarger for farms havelower to intensity. Our dependent value variablein each regression equationis themoney of all crops whichincludeMexicanwheat,Desi (or traditional) wheat, etc. paddy,sugar-cane, cotton, fodder and by-products, we first Turning theindependent to variables, consider area sown. This minuLs current fallowland. We have equals total area undercultivation triedto makethisvariablehomogeneous farms.Data overthe different providefigures actual rentpaid and imputed for rentforeach farm.We consider perhectare be perhaps bestsingle the measure thequality of rent to makeunitsofarea homogeneous oflandthatthedata provide.In order to and rent between we farms, worked district-level perhectare, individual out farmrentper hectare:by dividing latterfigure each farm the the for by for district-level per hectarewe obtaineda set of weights all farms. rent for These weights werethenused to adjust thefigures area sown. Irrigation in Everyholding thissamplehas someirrigation. maybe by is tubewells, canals. Separateinformationnotavailable old-style wells, or used kind about theamount each different ofirrigation by a farm.But of include we areprovided which with for figures irrigation expenses irrigation

378

ESTIMATING

SEASONAL MARGINAL PRODUCTS

charges (actually paid or imputed) and depreciation on irrigationequipfor expenses,the figures otherinputs,except labour, ment. Like irrigation are in termsof moneyvalue. We now considerthe data about labour. We are provided with hours hiredlabour, and and moneyvalue per year forfamilylabour,permanently casually hired labour. Money value of family labour was apparently imputedon the basis of the marketwage rate forlike labour. Permanently the hiredlabour and familylabour are available forworkthroughout year. Casual labour is hiredon a day-to-daybasis as thereis need to supplement the permanently available workers. Of the 150 farms,49 did not use any hiredlabour, buteach ofthe150 farmsused somecasual labour. permanently The mean value, forthe sample, of the ratio of the casual labour to total labour hours was 0*26,with a standard deviation of 0 13. The quality of labour in each of these categoriesmay vary fromfarmto farm. In orderto make the hoursoflabour morehomogeneouswithinthese categories,we weightedthe hours of each kind of labour on a farmwith the ratio ofthe estimatedwage rate on that farmofa certainkind oflabour wage rate. We assume that these adjusted to the estimated district-level hours of each of the three categories of labour are homogeneousover the worker of i.e. threecategories, that the personalefficiency the typicalfamily of the personal efficiency the typical permanentlyhired is the same as workeror that ofthe typical casually hiredworker. This assumptionseems wage rates forthe threekinds justifiedbecause the estimateddistrict-level of labour are almost identical. We assume that casual labour is used only duringthe busy season. This seems a reasonable assumption for two reasons. First, this is what the farmers themselvessay. Secondly it seems reasonable to expect that on a typical slack day the marginalreal cost of applying labour of permanent workerscomesintoequilibriumwiththeirmarginalproductat a level lower than the slack-seasonreserveprice of the labourerswho mightbe casually hired. A casual workeris unlikelyto accept a wage lowerthan a livingwage forworkeven on a slack day-not least because he usually has some work, albeit low-productivity work,ofhis own to do duringthe slack weeks. That work might be making music at weddings and other such celebrations, tendinghis own small plot, and so on. We have added togetherthe hours of familyworkersand permanently hiredworkers hiredworkersbecause both familyworkersand permanently to the are available forworkthroughout year. We refer this amalgamated input as permanentlabour. Of the annual total ofpermanentlabour hours that the data thus provide us with,some were utilized in the busy season and some in the slack season. The data do not give us any guidance on seasons. how to divide this total into the parts utilized in the two different

S. K. NATH

379

However, if we are willingto assume that the proportion(say A) of the total permanentlabour utilizedin the busy season is the same on all farms, guess as to what this proportionis, then and further, make an informed we can for each farmapportion a part of the annual total of permanent labour hoursto the busy season, and anotherpart to the slack season. We have tried threevalues of P in our calculations: 0-25,0 5, and 0 75. As we for shall see, the estimated coefficients the seasonal inputs of labour are not very sensitiveto a fairlybig change in the value of Pi. Once we have estimatedthe numberofhoursofpermanentlabour which are utilized in the busy season, we add them to the casual labour hours for each farmto arriveat the labour input forthe busy season. This assumes substitutesin the busy that casual and permanentlabour hoursare perfect season. Since only permanentlabour is utilized in the slack season, the labour input forthe slack season is given by whateveris leftof the annual total hoursof permanentlabour once those hourswhich are assumed to be utilized in the busy season are subtracted. Since we shall be using a single equation estimationprocedure,we are assuming that there is no joint determinationof inputs and outputs; i.e. the factor amounts used are assumed to be exogenous. We also assume membersof that the factorsused and the outputs produced by different the sample are homogeneous and that they all have access to the same technical knowledge.' We triedlinear, Cobb-Douglas, and transcendental specificationsof the productionfunction. It will be noticed (see Tables I labour hoursused in the slack regressions, and II) that in the odd-numbered season and labour hours used in the busy season are included as separate independentvariables; but forthe sake ofcomparisonin the even-numbered all we regressions have followedthe usual practiceofadding together labour hoursused duringthe year,and includingthemas one independentvariable total labour hours forthe year. which is described as undifferentiated triedsome simplelinearregressions We first (whichare notreportedhere). We foundthat area sown (A) is fairlyhighlycorrelatedwith a number of other variables; the residuals of these regressionsalso indicated that the was roughly variance of the disturbance proportionalto area sown. Therefore,we normalizedall the variables forarea sown in all regressions. Once for we had obtained the regressioncoefficients the linear equations, we multipliedeach side of the regressionequations by A, so that the regresof sion coefficient 1/A became the estimated constant, and the constant of became theestimatedcoefficient A. The resultsincludedin Table I forthe were obtained in this way. In the double-log equations, linear regressions
1 These are the standard assumptions of the estimationprocedurewe shall use. It is not our purpose to review these assumptions here. For excellent discussions see Griliches [5] and Heady and Dillon [8].

380

ESTIMATING SEASONAL MARGINAL PRODUCTS

.o

I
w-

I o

co *

*
o

^+*
o N O

o?i**

R
+sqe-Xoe nbmb* Vt m

ko C be-s

Io

-~
00
oo

0D o
66

~
oooooNq

* c

_N
C coo

q ~~~~

MN

-f~m

* q~~~~~~~o. o ^ 00~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~0
00

-E-*

oo D* Ds- c:* c sc

--

oo0

* CqM ,d4

Cq

I .poioXXXeb

tb c

cce C: qe nCo Xo Oo

h_

O0

00~~~~~~~~~~~~~
*

r-

_q

OOOO

OO

.s e.?Oi 11 z*

-*-* X~~~~a c. tb

t t0 ? * as* 0

4s 6

Ot-

w r- M *=

6 6 16 O

aq

m aq

,..

st

c: 1C:4 C; cn 4 6 O OXC O

O CO

w N

X_

4 o ^ o

?n

en~~00
0 c0

t-

t-N

w0

wO

w ct

.?~~~~a 1r-Z O o

cO0IOmbOXX

o~

O-:O O; O~O aq

bXsmXO

o-

ooooooooooo-o

tb C;

o ejo

~~~~~~~~~

o
aq

O
00 tE

aq

- -*
0

S. K. NATH

381

the specificationchosen implies that we have restrictedthe sum of the elasticities with respect to all inputs (including area sown) to be equal to one. This procedure seemed justifiedbecause we firstestimated two regressionequations in double logs without normalizingany variable for area sown. In the first themP was assumed to be equal to 0-25; and, in of the second, 05. The sums of input elasticitiesworked out as 1H1in each different from1. case; in neithercase was the sum significantly Equations (5a), (5b), (5c), and (6) are the transcendental production function-a generalization of the Cobb-Douglas function suggested by Halter et al. [6]. Though it too, like the Cobb-Douglas, restrictsthe elasticityofsubstitution betweenany two inputsto be equal to one, it does, unlike the former, permitthe elasticityof output with respect to an input to vary. Thus it permitsthe marginal product of an input to increase, stages of production; whereas decrease, and be constant over different according to the Cobb-Douglas, marginal product of an input increases, decreases, or stays constant in all stages of production. This propertyof the transcendentalproductionfunctionis obviously greatlydesirable for our purposes. In regressions (5a), (5b), (5c), and (6) all inputs and outputs have again been normalizedforarea sown. In these equations we grouped inputs other than labour and area sown, into two categories: those used mainly or only in the busy season, and those used in both the busy and slack seasons. The two groups were: 'Seeds, etc.', which includes seeds, manure, fertilizer,and insecticides and pesticides; and 'Total Capital
Notes to TABLE I 1. i = proportionof total annual hours of permanentlyavailable labour which is assumed to be used in the busy season. 2. The numbers in parentheses are the calculated standard errorsof the respective coefficients. An asterisk indicates that the estimated coefficient significantat least at the is five per cent level. are 3. Regressions(la), (lb), (ic), and (2) are linear; the otherfourregressions in double logs. 4. All the variables have been normalizedfor the size of the area sown. 5. The dependent variable, total crops, is measured in money value. Notation to TABLE I C = constant L' = hours of labour used in the slack season Lb = hours of labour used in the busy season Lt = annual total of hours of labour used A = area sown in hectares B = value of bullock labour used in a year S = value of seeds used in a year M = value of manure used in a year F = value of fertilizer used in a year I = value of insecticidesand pesticides used in a year Kw = intereston workingcapital fora year Kf = intereston fixed capital fora year D = depreciationon buildings fora year R = irrigationexpenses fora year E = miscellaneous expenses fora year

382

ESTIMATING SEASONAL MARGINAL PRODUCTS TABLE II

Punjab, 1967-8 results totalcrops: Ferozepurdistrict, for Regression (transcendental specifications)


(5a) ==0.25 (5b) 3=0.5 6-20* (0.20) -0-016 (0.012) -0-0008 (0 00065) 0.086 (0.047) 0-00083 (0.00047) (5c) 3=0.75 6.18* (0.021) -0-017 (0-013) -0-0026 (0-0016) 0-084 (0.047) 0.00087 (0 00046) (6) 6.49* (0.13)

a
Log L, Ls Log Lb
Lb

Log Lt Lt Log S* S* Log K* K*

6.20* (0.196) -0-014 (0.012) -0-0002 (0.00034) 0.089 (0.047) 0.00077 (0.00048)

F test R2

00022 (0-0021) 0.0018* (0.00036) 0.027 (0 0082) 0-000056 (0.00016) 30-36 0.61

0-0022 (0.0021) 0.0018* (0*00036) 0-027 (0 0082) 0 000053 (0 00016) 30-28 0.61

0.0022 (0.0021) 0.0018* (0 00036) 0-027 (0 0082) 0.000051 (000016) 30-22 0-61

0 0035 (0.0088) 0-000013 (0 0000070) 0-0026 (0.0022) 0.0018* (0 00038) 0.030 (0 0084) 0-00022 (0-00016) 33-61 0-57

Notes to TABLE II available labour which is 1. /3= proportion total annual hours of permanently of assumed to be used in the busy season. 2. The numbersin parenthesesare the calculated standard errorsof the respective is coefficients. asterisk indicates that the estimated coefficient significant An at least at the fiveper cent level. 3. All the variables have been normalizedforthe size of the area sown. is 4. The dependentvariable in these regressions the logarithmof the moneyvalue of the total crops. Notation to TABLE II C = constant Ls = hours of labour used in the slack season Lb = hours of labour used in the busy season Lt = annual total of hours of labour used and S* = the sum of the values of seeds, manure,fertilizer, insecticidesand pestiof cides used in a year. Log S* equals the sum ofthe logarithms the foregoing inputs on and fixed K* = the sum of the annual values of bullocklabour, interest working capital, irrigationexpenses, depreciation on buildings, and miscellaneous of inputs expenses. Log K* equals the sum of the logarithms the foregoing

S. K. NATH

383

Services', which includes value of bullock labour, interest on working capital, interest on fixed capital, irrigation expenses, depreciation on buildings, and miscellaneous expenses. Where these inputs enter the equations in their logarithms,their logarithmswere added together to formthe groups. Table I gives the regressionresults for the linear and the double-log specifications;and Table II forthose of the transcendentalspecifications. include the labour inputs forthe busy and The odd-numberedregressions the slack seasons as separate inputs. The proportionof the annual total of permanentlabour hourswhichis assumed to be used in the busy season, i.e. A, is equal to 0*25,0.5, and 0 75 respectivelyin equations marked a, b, and c. of It will be noticedthat the estimatedcoefficient the busy-seasonlabour input is positive and highly significanteach time. Moreover, it is not values given to /P.We expect positive marginal affectedby the different product in the busy season because at that time of the year all the farms in our sample hiresome casual workers;therewould be littleeconomicsense in hiringthemifthe marginalproductof labour in the busy season was not of positive. On the otherhand,the estimatedcoefficient slack-seasonlabour fromzero. But different is always negative and mostlynot significantly in regression(1c) it is both negative and significant.This casts doubt on either the realism of assuming Pi= 0.75 or the suitability of the linear or specification the productionfunction, both. But we are not surprised of of when the estimated coefficient slack-season labour turns out to be not been assumed from zero. Though it has commonly significantly different

a typical day in the slack season, ifat the pointwherethe marginalproduct of labour becomes zero, the marginaldisutilityof effort the permanent by workersof the farmis also zero (or perhaps negative), then application of labour hours duringthe slack season may well be carried up to the point where its marginalproduct is zero.' in The othercoefficients whichare significant all or nearlyall regressions insecticides and pesticides, are those for bullock labour, seeds, fertilizer, to intereston workingcapital, and irrigation charges. It is interesting note the that in the double regressions inputswiththe highestestimatedelasticities are seeds, irrigation, busy-season labour hours, and workingcapital. In regressions(2), (4), and (6), we ignored the seasonal variation in
1 For further development of this point, see Uppal [17], who, however, does not distinguish between the two seasons. 4520.3 Cc

product thaton a farmwhere somecasual labouris hired,themarginal thebusyin be of labourcannot zero,thatreasoning fact applies onlyto season marginal product: there no reasonwhyevenon suchfarmsthe is of marginal product labourin theslack season maynot be zero. For, on

384

ESTIMATING SEASONAL MARGINAL PRODUCTS

activity,and includedthe annual sum of labour hours as one of the inputs. Its estimated coefficient positive and significant, is but rather small, in it is positive though not significantly regression(2); while in the others different fromzero. If we had tried regressionsof only this kind-with annual labour hours added together as one input-we might well have inferred from resultsthat in our sample the marginalproductoflabour the was eithervery small or not significantly different fromzero. But, as we have argued above, the resultsofregressions whereslack-season and busyseason labour inputsenteras separate variables are moremeaningful.They suggestthat the marginalproduct of labour in the slack season was zero in our sample, but that of labour in the busy season was positive. Results of the transcendental specificationare reported in Table II. They are broadlysimilarto the previousresults. Each coefficient slackof season labour is negative; and none is significantly different fromzero. All the coefficients the busy-seasonlabour are positive,thoughit is only the of coefficient the logarithmof an input which needs to be positive in the of transcendental productionfunctionforthe marginalproduct of that input to be positive. However, only one of the four coefficients estimated in connectionwith the busy-season labour in regression(5a), (5b), and (5c) is significantly different fromzero at the 5 per cent level; the others just failto be so. Few othercoefficients significant Table II. This suggests are in that the transcendentalspecificationdoes not fitthe data so well. Of the threespecifications, the Cobb-Douglas productionfunctionthen seems to fitthe data best. We have used the results of this specification to estimatethe seasonal marginalproductsoflabour. However, since none oftheestimatedcoefficients the slack-seasonlabour in the Cobb-Douglas of different fromzero, we have not used them to specification significantly is derive any estimates of the marginal product of the slack-season labour. The estimatesofan hourofthebusy-seasonlabour obtainedfrom regressions and Rs 0O8. Each (3a), (3b), and (3c) are, respectively: Rs 1-5, Rs 1X3, value of3, the proportion annual oftheseestimatesis based on a different of the busy season. In the permanentlabour that is assumed to be used in absence of any firmbasis on which to choose among the different values to of3, it is difficult decide whichofthe threeestimatesis the most accurate. Some policy implications An estimate of the marginalproduct of labour, forthe year as a whole, is meant to tell us the impact on the annual output if a unit of labour (say an hour or a man-year) is withdrawnforthe year as a whole, everything the elseremaining same. Our arguments implythat fora traditionalagriculturemarkedby a strongseasonal variationin activity,the annual marginal product of a unit of labour is best estimated indirectly first by estimating

S. K. NATH

385

the seasonal marginal products of labour. Once we have these seasonal estimates,we can sum themto obtain the composite marginalproduct' for the year as a whole; but we cannot obtain an estimate ofthis annual (composite) marginal product directlyfromthe production function,because whenthe productionfunction specified take note ofthe seasonal nature is to of agriculture,annual labour input does not figureas an argumentof the productionfunction. The resultsof our analysis are also relevant to the controversies about the existenceofsurpluslabour and to the problemofestimating shadow the wage rate oflabour. Regardingthe concept of surpluslabour, Sen [16] has convincingly argued that the criterion its existence should be whether for total agricultural output declinesor not when one workeris withdrawnand compensatingadjustments in work supply by those who are left behind are permitted. In other words, examining just the marginal product of labour by itselfis not enough. With the help of a model in which family welfaredepends on income as well as leisure and in which equal sharing of workand incomeis assumed, he shows that even with positive marginal product of labour, if the withdrawal of some labour takes place froma familyfarm,total output will be maintained at the same level if forthose who remainon the farmmarginalrate of substitutionbetweenincome and work (or what he calls the 'real cost oflabour') is constantover the relevant range. Our analysis suggests that at the time of the year (i.e. the slack season) when such adjustments by the remainingmembersof familyare possible, there may be little or no need to increase the amount of effort supplied when some labour is withdrawn (for that season) because the marginal product of labour forthat season is low or nearly zero anyway: and that at the time of the year (i.e. the busy season) when the marginal product of labour is positive and high,people on the farmare likelyto be workingrathernear theirphysical maximum on a typical day (that being the reason why in some regionseven the smallestfarmers hire some casual labour on busy days)-so that the marginal'real cost oflabour' is not likely to be constant. The theoreticalissues involved in making estimates of a shadow wage rate have been recently surveyedby Lal [10]. He mentionsmanyproblems which must be taken into account in estimating a shadow wage rate. Seasonality ofagriculturalactivityis anothersuch problem. Indeed, ifthe in marginal products of labour are as strikinglydifferent the different seasons as our findings suggest,then forsuch regionsthere may well be a
1 As far as I can ascertain, Mazumdar [13] is the firstauthor to have used the term 'composite marginal product'. However, he did not develop the argumentto the stage of pointing out that it cannot be directlyestimated by includingtotal annual input of labour in the productionfunctionto be estimated. Indeed in his later joint article [4], there is no mention of the concept.

386

ESTIMATING SEASONAL MARGINAL PRODUCTS

needto estimate different shadow wageratesfor different oftheyear. parts that Ouranalysis in suggests thedisguised unemployment agriculture may be mainly seasonal. Thus theremay be significant social advantagein certain and encouraging kinds mixed of farming, cropping schemes, cottage industries which would helpprovide more employment thehitherto in slack weeks. It is worth that according MahatmaGandhithiswas recalling to one of the strongest and arguments protecting encouraging for cottage industries India. The findings in also supportarguments providing for seasonalemployment nearby in towns agricultural the for migrants during slackseason.' In an agricultural sectorwhichis similar our sample,if to somelabourdoes moveoutforthewholeyear(including busyseason) the thenouranalysis showsthatagricultural outputis boundto suffer unless laboursubstituting methods machinery simultaneously and are introduced forthe busyseason. It is possiblethat someofthe current technological in developments Indian agriculture-e.g.improved seeds, moreregular supply water, of and fertilizers-will to increase number crops tend the of thatcan be grown a year,and wouldthusincrease number busy in the of daysintheyear.2Ontheother hand,ifsuchdevelopments notincrease did thenumber daysonwhich of sowing, watering, weeding, spreadploughing, ing fertilizer, harvesting threshing, need to be done because of and etc., raised more crops hectare year,and other per per reasons, but,instead only the marginal product labouron a typicalbusyday without of increasing thenumber suchdays,theywouldmakethe contrast of between busy the seasonandtheslackseasonevenmore the without striking reducing relative of duration theslackseason. the and publish also highlights needto gather Ouranalysis agricultural datadifferentiated to seasons.In a farm according theagricultural manageare in mentsurvey particular, dailyrecords keptofthe inputsused,etc. if It wouldbe extremely useful a weekly, at least monthly, or tabulation the which also indicated kindofworkdone,was made available. ofthem, University Warwick of
REFERENCES B. 1. AGARWAL, L., and VARSHNEY, R. G., 'Seasonal variationsin employment in India', Economicand Political Weekly, Bombay, 29 November 1969,pp. 1845-9. 2. BIMLINGS,MARTIN H., and SINGH, AJAN, 'Labour and the greenrevolutionthe experience Punjab', Economicand Political Weekly, in Review of Agriculture, Bombay, December 1969, pp. A221-4.
1 It would seem that this already happens to some extent. See Agarwal and Varshney [1]. Some of the examples of seasonal industriesthey give are: production of sugar, coffee,and tobacco products; processingof cotton, jute, and wool; making bricks,tiles, and salt; and manufacturing ice, aerated and mineral water, and umbrellas. 2 There is evidence that this is already happening to some extent. See Billings and Singh [2].

S. K. NATH

387

3. CLARK, C., and HASWELL, M., The Economicsof Subsistence Agriculture, 6, ch. 4th edn., London, 1970. 4. DESAI, M., and MAZUMDAR,D., 'A test ofthe hypothesis disguisedunemployof ment',Economica, 1970. Journal 5. GRILICHES, Z., 'Specification bias in estimatesof productionfunctions', of Farm Economics,1957. 6. HALTER,A. N., CARTER, H. O., and HOCKING, J. G., 'A note on the transcendental productionfunction', JournalofFarm Economics, 1957. 'The seasonal employment in 7. HANSEN, B., and EL TOMy, MONA, profile Egyptian JournalofDevelopment agriculture', Studies,July 1965. Iowa, 1961. 8. HEADY, E. O., and DILLON,J. L., Agricultural Production Functions, 9. KRISHNA,R., 'Subsistence and familyfarms: some theoreticalmodels of suband jective equilibrium',in C. R. Wharton, Jr. (ed.), SubsistenceAgriculture EconomicDevelopment, London, 1970. 10. LAL, D., 'Disutility of effort, migration,and the shadow wage-rate', Oxford EconomicPapers, 1973. 11. MABRO, R., 'Employment and wages in dual agriculture',OxfordEconomic Papers, 1971. 12. MATHUR, A., 'The anatomydisguisedunemployment', Oxford EconomicPapers, 1964. 13. MAZUMDAR,D., 'Size of farmand productivity:a problem of Indian peasant agriculture', Economica,1965. 14. NAKAJIMA,C., 'Subsistenceand familyfarms: some theoreticalmodels of suband jective equilibrium',in C. R. Wharton, Jr. (ed.), SubsistenceAgriculture EconomicDevelopment, London, 1970. and some policy impli15. ROBINSON, W., 'Types of disguisedruralunemployment EconomicPapers, 1969. cations', Oxford 16. SEN, A. K., 'Peasants and dualism with or withoutsurpluslabour', Journalof Political Economy,1966. 17. UPPAL, J., 'Work habits and disguisedunemployment underdevelopedcounin EconomicPapers, 1969. tries: a theoreticalanalysis', Oxford 18. WELLISZ, S., 'Dual economies, disguised unemploymentand the unlimited supply of labour', Economica, 1968, esp. pp. 46-7. 19. WELLISZ, S., MUNK, B., and HUMMER, C., 'Resource allocation in traditional agriculture:a study of Andhra Pradesh', JournalofPolitical Economy,1970.

Вам также может понравиться