Zambezia (1992), XK (ii).FERTILITY LEVELS AND TRENDS IN ZIMBABWEMARVELLOUS M. MHLOYIDepartment of Sociology, University of ZimbabweIN THIS ARTICLE I attempt to establish the fertility level in Zimbabwe for theyears 1982-4. In view of the inaccuracy of recorded data from censusesand surveys several methods of estimation are used. This period waschosen because a census was taken in 1982 (Central Statistical Office,1985) and a reproductive health survey was conducted in 1984 (ZimbabweNational Family Planning Council, 1985). The data from this period can becompared with those from 1969, when a population census was taken(Central Statistical Office, 1969), with those from 1987 when an intercensaldemographic survey was conducted (Central Statistical Office, 1991), andwith those from 1988 when a demographic health survey was undertaken(Central Statistical Office, 1989). I then consider fertility trends over a 20-year period (1969-88). Empirical analysis is based on the 1982 ten per centsample (Central Statistical Office, 1985) and data from the ZimbabweReproductive Health Survey, 1984 (Zimbabwe National Family PlanningCouncil, 1985). There is evidence that fertility declined between the 1969population census and 1984. If the data recorded in the 1987 intercensaldemographic survey (Central Statistical Office, 1991) and the 1988 demo-graphic health survey are correct, this decline seems to have continuedup to 1988. Such a decline may be the result of an increased use of modernmethods of contraception. However, little can be said about the magnitudeand the timing of the decline until further research is undertaken.BACKGROUNDConsistent with Zimbabwe's political history, a peripheralization of theAfrican population is characteristic of all censuses undertaken (at five-year intervals) between 1901 and 1962. The demography of the Africanpopulation was left to speculation until 1962 when the census provided,for the first time, data on the African population by sex and age. It is,therefore, difficult to discern historical trends in population growth rateswith any precision.The population growth rate in the 1969-82 intercensus period wasestimated at 2,93 per cent (Central Statistical Office, 1985). Such a highgrowth rate is consistent with declining mortality coupled with persistenthigh fertility. The crude death rate based on the ten per cent sample of the1982 census of population was estimated at 10,8 per thousand. The infantmortality rate was estimated at 83 per thousand, while life expectancy was7980 FERTILITY LEVELS AND TRENDS IN ZIMBABWEapproximately 57 years (both sexes), a remarkable improvement since the1950s when it was estimated to be about 42 years for both sexes.Fertility was still quite high. The crude birth rate and the total fertilityrates, as estimated from the 1982 ten per cent sample, were 39,5 and 5,6,respectively (which, as will be shown later, are underestimations). Totalfertility rate as estimated from the Zimbabwe Reproductive Health Survey,1984 (Zimbabwe National Family Planning Council, 1985) and the ZimbabweDemographic and Health Survey, 1988 (Central Statistical Office, 1989) was6,5 and 5,7, respectively. Total fertility rate as estimated from the 1969census was 7,18 (Mzite, 1981), an implied decline of 2,2 births during the13-year intercensus period Š that is, if we take the total fertility rate of 5,6from the 1982 census at its face value. Family planning services were notonly highly inadequate during the pre-Independence period but they werealso perceived as a colonial plot to limit the African population (Clarke,1972). Low usage of contraceptive methods was characteristic of thisperiod but this is inconsistent with the precipitous decline in fertilityimplied in the findings of the 1982 census. The one-child increase between1982 and 1984 was accompanied by a marked increase in the use ofcontraceptives from less than 10 per cent prior to Independence to 38 percent in 1984 (data based on the Zimbabwe Reproductive Health Survey,1984 (Zimbabwe National Family Planning Council, 1985)). However, thisimplied increase is followed by a decline of 0,8 of a child between 1984 and1988. From the 1969 census to the 1988 survey fertility seems to havedeclined by approximately 2,1 children compared to the 2,2 child declineduring the 13-year intercensus period (1969-82). What, then, is the actualtrend in fertility?OBJECTIVES OF THE STUDYThere are no easy answers to questions about fertility trends in Zimbabwe.Firstly one must establish what the fertility levels were in Zimbabwe atdifferent times and then interpret these estimates within their socio-economic contexts which may give more insight into the direction of thetrend. I attempt in this article to piece together both quantitative andqualitative data in order (a) to establish fertility levels and (b) to discernthe trend in fertility levels in Zimbabwe and infer their future direction.METHODOLOGYDetailed analyses of fertility levels were based on the 1982 and 1984 datasets. Such results were supported and interpreted with the help of thedata from the Demographic and Health Survey and a qualitative surveyconducted by myself in two rural settings (Mhloyi, 1985). Some indirectMARVELLOUS M. MHLOYI 81estimation techniques were used on data from the 1982 census and theReproductive Health Survey to estimate fertility levels between these twodates. Resort to as many indirect methods of estimation as possible isvital in view of the complete unreliability of fertility indices based onrecorded data. Experience from a number of African countries has shownthat recorded data tend to underestimate fertility.It must be stated at the outset that some of the methods used intro-duced a bias into the estimates, particularly those which, like the BrassP/F Ratio Method, 'assume no change in fertility in the recent past'.However, it is hoped that the bias may be systematic enough to facilitatethe assessment of the inherent bias in the original data. Four methodswere used in this analysis: (1) Brass's P/F ratio method; (2) Brass's birthorder method; (3) duration of marriage by number of children ever born;and (4) the Rele Method. A brief description of each method and itsresults follows. A synthesis of the results of the respective methods wasmade with the aim of establishing a plausible fertility level; the emphasisis not on a specific number per se but rather a range of estimates withinwhich a plausible specific number falls. A discussion of fertility trends andthe underlying factors concludes this article.FINDINGSBrass P/F Ratio MethodDescription and assumptions: The Brass Fertility Method is used to adjustthe age pattern of reported fertility derived from information on recentbirths by the level of fertility implied by the average parity of women inthe reproductive age groups (United Nations, 1983). It is commonly usedbecause it requires only data on cumulative fertility, classified by age ofmother, and births in the past year (current fertility), also classified byage of mother.The assumptions involved in the Brass method are:(i) Fertility has been constant during the recent past;(ii) The pattern of fertility is accepted; and(iii) Younger women report their fertility more completely than olderwomen do.It should be noted that this method is sensitive to changes in fertility.It was applied to the African, Coloured, Asian and European populationsusing data from the 1982 population census only since the ReproductiveHealth Survey dealt exclusively with the African population.Results: The results show that the fertility estimates are not significantlyaffected by the adjustment factor, a pattern that is consistent with the82 FERTILITY LEVELS AND TRENDS IN ZIMBABWEsimilarity of the P/F ratios (Tables I to V). However, it is important to notethat the fertility estimate from the 1982 census is only slightly higher thanthat of the Reproductive Health Survey: 7,4 and 7,2, respectively. Thesimilarity of the P,/F,- ratios does not provide evidence of a fertility declineof the magnitude portrayed by the estimates of the 1969 and 1982 censuses.Further, these results tend to suggest that overall fertility is more under-reported in the 1982 census than in the Reproductive Health Survey(whose original estimate is not affected as much as that of the 1982census). However, both estimates show higher fertility than those producedby the respective direct measures shown in the following summary.TOTAL FERTILITY RATERecorded Adjusted1982 5,6 7,41984 6,4 7,2In contrast, virtually no upward adjustment occurred for the Europeanpopulation. Corresponding moderate and small upward adjustmentsaffected the Coloured and Asian populations. These results show that,although the Brass method introduces an upward bias, the technique hasdifferentiated the magnitude of the flaw in the data for the differentpopulations in the expected direction, that is, data on Africans are expectedto be more defective than data for Europeans. One conclusion is clear: thetotal fertility rate of 5,6 based on recorded data from the 1982 census isvery much an underestimate.Brass Birth Order MethodDescription and assumptions: The Brass Birth Order Method was designedto estimate fertility by comparing reported fertility with fertility derivedfrom birth registration. This method is based on the assumption thatunder-reporting of first births is less serious than incompleteness of regis-tration (Brass, 1975). He maintains that it is generally reasonable to assumethat under-reporting is proportionally the same for first births as for allbirths or, at least, nearer to the truth than accepting registration ascomplete. Specifically, this method is based on the ratio of all birthsduring the year to first births in the same year. Assuming that reporting ofbirth order is accurate, the ratio of all births to first births represents afertility index. This index is particularly useful in societies where birthregistration is incomplete and disproportionately favours first births. Brassalso suggested the use of second births to check the complete family sizederived from first births.Table IESTIMATION OF AFRICAN TOTAL FERTIUTY BY BRASS BIRTH ORDER METHOD, ZIMBABWE, 1982Agegroup15-1920-2425-2930-3435-3940-4445-49Averageno. birthsduring past12 monthsi ft1 0,08372 0,24593 0,24714 0,22365 0,17216 0,09837 0,0398Averageno. childrenever bornPi0,24001,51053,13844,76636,03576,99027,4216Cumulativefertility atbeginningof interval0,00000,41851,64802,88354,00154,86205,3535MultiplyingfactorsWi1,97842,84403,01203,12163,24823,51804,41380,16560,69930,74430,69800,55900,34580,1757EstimatedaveragecumulativefertilityFt0,16561,11782,39233,58154,56055,20785,5292P/F,1,44931,35131,31191,33081,32351,34231,3423Adjustedage-specificfertility rates0,11310,33230,33390,30220,23260,13280,0538Cumulative Adjustedadjustedage-specificfertility ratesft0,11310,44540,77931,08151,31411,44691,5007totalfertilityrates0,56552,22703,89655,40506,57057,23457,5035Mean age of childbearing: m = 30,2403 yearsCorrection factor was calculated using P,/Fi in 20-24 age group (Source: Brass, 1968, 96).2XTable IIESTIMATION OF TOTAL FERTILITY BY BRASS TECHNIQUE, ZIMBABWE, 1984Exactage ofwomenat timeofsurvey15-2020-2525-3030-3535-4040-4545-50i1234567Average no.births inpreceedingyear perwomanft0,13080,28910,29550,26310,21960,09210,0110Average no.of childrenever bornPi0,3031,6593,2064,6316,2197,0377,464Cumulativefertilityat beginningof intervalj = o0,0000,6542,09953,57704,89255,99056,451Multiplyingfactors forestimatedaverage valuefertilityWj2,2832,8873,0333,1393,2833,6044,616EstimatedaveragecumulativefertilityPi/FtAdjustedage-specificfertility rates-nm5m5)3mIz0,29891,47962,9964,40295,61346,32246,50181,01471,12121,07011,05181,11301,11301,14800,14330,31670,32380,28830,10090,10090,01201,44570,14670,32420,29500,29500,10330,10330,01231,4590mTable IIIESTIMATION OF COLOURED TOTAL FERTILITY BY BRASS TECHNIQUE, ZIMBABWE, 1982Age group15-1920-2425-2930-3435-3940-4445-491234567Averageno. birthsduringpast 12monthsi0,05220,19720,17200,06250,07810,00000,0238AverageCumulativeno. children fertilityeverbornfi0,13080,95681,64522,40323,31254,33334,9868at beginningof intervalPi0,00000,26101,24702,10702,41952,81002,8100Multi-plyingfactors1,76302,80652,99603,10603,22833,46814,2737Wi0,09200,55340,51530,19410,25210,00000,1017Estimatedaveragecumulativefertilityw,f,0,09200,81441,76232,30112,67162,81002,9117Fi1,42171,17490,93361,04441,23991,54211,7127Adjustedage-specificfertilityratesPi/Fi0,06130,23170,20210,07340,09180,00000,0280Cumulative Adjustedadjustedage-specificfertilityrate0,06130,29300,49510,56850,66030,66030,6883totalfertilityratefi0,30651,46502,47552,84253,30153,30153,4415Mean age of childbearing: m = 27,6052 yearsCorrection factor was calculated by using P,/F, In 20-24 age group (Source: Brass, 1968,96).3rn||CO2ITable IVESTIMATION OF ASIAN TOTAL FERTILITY BY BRASS TECHNIQUE, ZIMBABWE, 1982Agegroup15-1920-2425-2930-3435-3940-4445-49i1234567Average no.birthsduringpast 12monthsfi0,02040,11110,15560,03850,12500,00000,0000Mean age of chlldbearing:(Source:Averagei Cumulativeno. children fertilityeverbornPi0,06120,36511,29551,88462,42503,35143,1923at beginningof interval0,00000,10200,65751,43551,62802,25302,2530m > 28,5158 yearsUnited Nations, 1967,36-7).Multi-plyingfactorswt1,50292,74702,97583,08773,20583,41304,0563wfi0,03070,30520,46300,11890,40070,00000,0000EstimatedaveragecumulativefertilityFi0,03070,40721,21051,55442,02872,25302,2530Pi/F,1,99350,89661,15621,21241,19531,48751,4169Adjustedage-specificfertilityrates0,02350,12800,17930,04440,14410,00000,0000Cumulative Adjustedadjustedage-specificfertilityratefi0,02350,15150,33080,37520,51930,51930,5193total: fertilityrate0,11750,75751,65401,87602,59652,59652,5965TlmE>moCOi ZIMBABWETable VESTIMATION OF EUROPEAN TOTAL FERTILITY BY BRASS TECHNIQUE, ZIMBABWE, 1982AgegroupAverage no.birthsduringpast 12monthsi fjAverageCumulativeno. children fertilityeverbornPiat beginningof intervalMulti-plyingfactorsU>i UEstimatedaveragecumulativefertilityhfi Fi PAdjustedage-specificfertilityrates\/FtCumulative Adjustedadjusted totalage-specific fertilityfertility ratesratesfi15-19 120-24 225-29 330-34 435-39 540-44 645-49 70,02260,12200,16470,11290,03500,00600,00000,04480,42941,24951,83792,20592,35322,61060,00000,11300,72301,54662,11102,28602,31601,50922,74882,97633,08813,20633,41424,06150,03410,33540,49020,34860,11220,02050,00000,03490,44841,21321,89522,22322,30652,31601,28370,95761,02990,96980,99221,02021,12720,02280,12330,16640,11410,03540,00610,00000,02280,14610,31250,42660,46200,46810,46810,02280,73051,56252,13302,31002,34052,3405ioCO2S5Mean age of childbearing: m = 27,8638 years(Source.- United Nations, 1967,73-4; Shryock and Siegel, 1976,829; Kpedekpo, 1982,188-90).88 FERTILITY LEVELS AND TRENDS IN ZIMBABWETotal fertility rate is derived by adjusting completed fertility by theproportion of women who had at least one child by the age of 49 and whowere married for a duration of at least 25 years.This method was applied to data from both the 1982 census and the1984 survey. However, since the data from the census do not includeinformation on duration of marriage, the adjustment factor was simply theproportion of women aged 49 who were mothers. This approach can biasthe proportion slightly downwards, particularly if one assumes that mostchildbirth occurs within marriage. The formulas for computing completedfertility from the respective birth orders are:First birth: Fm = 1,5 (B/B,) - 1,25Second birth: Fm = 1,5 (B - BO/Bz - 0,75whereFm = Mean completed family sizeB = All births in a yearBj = First births in a yearB2 = Second births in a yearResults: The total fertility rate estimated from the census data was 5,8 ascompared with 6,5 from the Reproductive Health Survey data using firstbirths. Using second births the total fertility rates are 5,6 and 6,1 for the1982 census and Reproductive Health Survey, respectively (Table VI).It should be noted that when the data are controlled for duration ofmarriage, the proportion of women with at least one child by the age of 45is increased. The census estimate which does not take account to durationof marriage might, therefore, be a slight underestimation of fertility. If onewere to apply the proportion that are mothers in the Reproductive HealthSurvey to the 1982 census data, fertility would be 6,0 and 5,0 for the firstand second births, respectively.An examination of births by order reflected a more complete reportingof first births compared to other subsequent births, including secondbirths. This is reflected in the total fertility rates derived using secondbirths: 6,1 for the Reproductive Health Survey and and 5,6 for the 1982census. The results from first births are more plausible and consistentwith the other estimates.This method produces fertility estimates which are almost identicalfor the Reproductive Health Survey and slightly higher for the 1982 census.It needs to be added that the results tend to support the idea that fertilitywas more greatly underestimated in the 1982 census than in the Repro-ductive Health Survey.MARVELLOUS M. MHLOYI89Table VIESTIMATION OF FERTILITY USING THE BIRTH ORDER METHODBirth order method as used in 1982 CensusFormula:First birth:Formula:Second birth:FmFmTFRFm: FmTFRBirth Order Method asFormula:First birth:Formula:FmFmTFRFmTFR1,5(B/B,) -1,251,5(B/B,) -1,251,5(28076/5824)-1,251,5(4,821) - 1,255,98 x PM = 5,98 x 0,9641 = 5,77l,5t(B/Bi)/B2]-0,75l,5[(28076 - 5824)/5070] - 0,75l,5[22252/5070] - 0,751,5(4,389) - 0,755,83 x PM = 5,83 x 0,9641 = 5,621used in the Reproductive Health Survey1,5(B/B,) -1,251,5(6/8!) -1,251,5(555/108) -1,251,5(5,13889) -1,256,46 x 1 = 6,46 = 6,5l,5[(B-B,)/B2]-0,751,5[(555 - 108)/98] - 0,751,5(4,5612) - 0,756,09 x 1 = 6,09 = 6,1PM = proportions married.Duration of marriage by number of children ever bornDescription and assumption: In natural fertility regimes where fertility isnot parity dependent and where marriage occurs early, fertility patternsare fairly similar by duration of marriage across populations (Shryock andSiegel, 1976). Hence, the sequence of average parities by duration ofmarriage should also be similar. Thus the reported sequence of averageparities (for durations under 5 years, 5-9 years and 10-14 years) can becompared with a standard embodying the natural fertility for respectivepopulations. Marital fertility is therefore estimated as equal to standard90 FERTILITY LEVELS AND TRENDS IN ZIMBABWEnatural fertility multiplied by the estimated level in the population. Theoverall fertility schedule can be estimated as the product of the proportionmarried and estimated marital fertility for each age group. The selection ofthe standard fertility schedule is based on the age of entry into marriage,assumed to range between 12 and 15 years. The assumptions that must bemet in using this method require that there is minimal contraception in agiven population and that most of all fertility behaviour occurs withinmarriage.This method is ideally used for continously and currently marriedwomen. However, since the data do not reflect the number of marriages,natural fertility can still be calculated for currently married women.However, 'dead time' between unions will affect the ratios of observedparities by duration of marriage groups from the parities derived from thestandard PXi. The 'dead time' duration increases with age since theprobability of marriage since first marriage, Pi/jq will also tend to decreasewith age. However, where marriage dissolution is minimal this effect isalso negligible.This method was used for those women in the Reproductive HealthSurvey only where information on duration of marriage was available.However, most respondents did not have dates of marriage, thus durationof marriage could not be calculated for such cases. An examination offertility, controlling for availability of data on duration of marriage, did notshow a clear difference between those with duration of marriage reportedand those without. Cumulative fertility is slightly higher for those withoutdurations of marriage compared to those with duration of marriagereported. However, current fertility is higher for those with duration ofmarriage up to age 35 and the reverse is true beyond that age. Taking intoconsideration the effect of widowhood, the pattern is consistent withminimal remarriage after widowhood, and thus poses minimal bias, if any,to the calculations. Two definitions of marriage were used: married andcohabitating (as is used throughout the study) and married only.Results: The first definition of fertility yielded a total fertility rate of 6,75while the conservative definition produced an estimate of 6,6 (Table VII)-The definition of marriage is a problem in African countries (includingZimbabwe) where marriage is a process rather than an event. Since mar-riage is a proxy for the period of exposure to the risk of getting pregnant,the liberal definition which yielded an estimate of 6,75 is relevant to ourstudy.Rele MethodDescription and assumptions: Rele (1976) proved that in most populationsat a given level of mortality there is a nearly linear relationship betweenMARVELLOUS M. MHLOYI91Table VIIESTIMATION OF FERTILITY FROM DURATION OF MARRIAGE AS SHOWN INTHE ZIMBABWE REPRODUCTIVE HEALTH SURVEY, 1984Agegroup15-1920-2425-2930-3435-3940-4445-49TFRIndexJ1234567Adjustedmarital fertilityrate g(j)0,33300,38000,35800,32300,26100,13500,02001,81009,0500Estimatedage-specific fertilityrate fQ,)0,10050,25800,30800,29400,23600,11900,01801,35056,7500Adjusted age-specific fertilityrate fQz)0,08900,27500,30100,28200,23470,11880,01801,31906,6000the child-woman ratio and the gross production rate. The relationshipwas expressed as:GRR = a + b CWRwhereGRR is gross reproduction rate,CWR is child-woman ratio (expressed per woman) andb is the level of mortality.The coefficients a and b determine the relationship for the populationsat mortality level n. The coefficients were computed by selecting a sampleof 36 stable populations with different combinations of fertility. It wasproved that the estimate of fertility is relatively insensitive to the level ofmortality hence a and b were computed for six levels of mortality. Grossreproduction rate for a mortality which falls between two of the six levelsis estimated by computing the gross reproduction rates at adjacent levelsand then interpolating the results for the exact levels. Total fertility rate isthus approximately:2,05 GRR = 2,05 (a + b CWR)Although the mathematical relationship was derived from the stablepopulations theory, the relationship between the child-woman ratio andthe fertility rate is insensitive to normal variations in a population's agestructure (Hanenberg, 1983).92 FERTILITY LEVELS AND TRENDS IN ZIMBABWERele's method is attractive because of its ability to produce twofertility estimates referring to two periods from the same data, that is, atotal fertility rate computed using a child-woman ratio with children aged0-4 in the numerator refers to fertility during the 0-4 years preceding thesurvey, and that with children aged 5-9 in the numerator refers to fertility5-9 years prior to the survey. Thus such estimates can provide an indicationof the trend of fertility over the decade preceding the census. The regressionmethods are simply variations on the standard techniques of reversesurvival.Results: Rele's method could be used on the census data only since theReproductive Health Survey does not have data for the calculation of thechild-woman ratios. Total fertility for the period 0-4 years prior to 1982 ata life expectancy of 57 years was estimated at approximately 6,0, whilethat referring to the period 5-9 years prior to 1982 was 7,0 (Table Vlir).These results tend to suggest that there has been a decline in the totalfertility rate of about one child between 1972 and 1982. It is interesting tonote that the total fertility rate of 7,0 is close to the estimated 7,8 of the1969 census (Mzite, 1988). Also, the total fertility rate for the 0-4-yearperiod preceding the census is higher than the estimated 5,6.Table VIIIESTIMATION OF FERTILITY USING THE RELE METHODCWRTFR50TFR60TFR57CWRTFR50TFR60TFR57Children 0-4/women 15-440,84392,05[- 0,1529 + 3,7375(0,8439)]2,05[- 0,1645 + 3,5556(0,8439)]6,20 + 0,7(5,87-6,20) = 5,976,0Children 5-9/women 20-49 = 1210,98612,05[- 0,1529 + 3,7375(0,9861)]2,05[- 0,1645 + 3, 5556(0,9861)]7,24 + 0,7(6,85-7,24) = 6,907,0= 6,20= 5,87926/123 647= 7,24= 6,85MARVELLOUS M. MHLOYI 93SYNTHESISNotwithstanding the possible bias in the estimates emanating from aviolation of the assumptions particularly that of 'no change in fertility' afew generalizations can be made from an assessment of the results fromthe different estimation techniques (see Table IX). Firstly, direct estimatesderived from both the 1982 census and Reproductive Health Survey givean underestimation of the fertility level in Zimbabwe at least for the period1982-4. However, this underestimation is larger for the 1982 censuscompared to the Reproductive Health Survey regardless of the estimationtechnique used. Secondly, the fertility estimates range between 5,6 and 7,5which suggest that the Reproductive Health Survey direct estimate is acloser approximation of the total fertility rate. Thus, if one is willing todiscard the 1982 census estimate as a gross underestimation of fertilityand accept the 1984 direct estimate of 6,5 as a closer approximation ofreality, then one can conclude that there has been a decline in fertilityfrom a total fertility rate of 7,8 in 1969 to 6,5 in 1984. It is generally true thatsurveys yield better fertility estimates than censuses. On that basis, the1969 census estimate is also highly questionable, particularly if oneconsiders the incompleteness of that particular census.Table IXESTIMATED TOTAL FERTILITY RATESEstimation techniqueDirect MethodBrass P/F Ratio MethodBrass Birth Order MethodRele Child-Woman Ratio MethodDuration of MarriageCumulative Fertility for 45-49-year olds1982 census5,67,45,86,07,41984 survey6,57,26,5_6,87,2Based on the above the following conclusions may be made:1) Using the recorded data alone it is highly improbable that fertilityhas risen between 1982 and 1984 (from 5,6 to 6,5 live births).There is little doubt that the above rates support the view thatcensus data suffer from severe undercount in comparision withsurvey data. In other words the figure of 5,6 from the 1982census is no doubt an underestimate.2) Average completed family size for women aged 45-49 was foundto be 7,4 and 7,2 from the 1982 census and the Reproductive94 FERTILITY LEVELS AND TRENDS IN ZIMBABWEHealth Survey, respectively. These rates may be underestimatesof the average completed family size since it is generally acceptedthat women in the older age groups tend to undercount theirlive births. Often older women do not count dead children aspart of the family, especially if the deaths occur at very youngages. Furthermore older women tend to forget those childrenwho died very young.3) The recorded total fertility rate from the Reproductive HealthSurvey was found to be 6,5 live births. Brass's birth order methodgave a total fertility rate of 5,8 for the 1982 census and 6,5 for theReproductive Health Survey. Rele's method (based on 0-4 year)provided a total fertility rate of 6,0 from the 1982 census. Allthese methods point to the total fertility rate being about 6,5 livebirths for the period 1982-4.4) The Brass P/F ratio method gives total fertility rates of 7,4 and7,2 for 1982 and 1984, respectively. The method corrects formisreporting. The total fertility rates given by the Brass P/Fratio method are the same as those given by the CumulativeFertility for 45-49 year olds. If the total fertility rates given bythe Cumulative Fertility for 45-49 year olds, which is usuallycontaminated by misreporting errors, then it can be concludedthat the total fertility rates for 1982 and 1984 were 7,4 and 7,2respectively. The other methods have, however, shown thatthere has been an underestimation of the reported fertility.So far I have attempted to construct a picture for fertility trends up to1984. What is the direction of fertility between 1984 and 1988?IS FERTILITY IN TRANSITION IN ZIMBABWE?Birth histories from the 1988 demographic and health survey, providetotal fertility rates by calendar periods and thus throw light on fertilitytrends (see Central Statistical Office, 1989, Tables 3.1 to 3.3). For theperiod 1982-4 the total fertility rate was estimated at 6,7 live births and forthe period 1985-8 it was estimated at 5,5 live births. Furthermore, theZimbabwe Demographic and Health Survey, 1988 (Central Statistical Office,1989) provides a recorded total fertility rate of 5,1 from the IntercensalDemographic Survey 1987 (Central Statistical Office, 1991). While it is truethat birth histories data cannot be accepted without further adjustments(as a result of misallocation of births to different calendar periods), andMARVELLOUS M. MHLOYI 95while the figure quoted from the Intercensal Demographic Survey is basedon recorded current fertility (probably underestimated), these may beconsidered as indications that fertility in Zimbabwe has by-and-large beendeclining between 1984 and 1988. However, the magnitude and the timingof the decline need further investigation.This possible decline is supported by the fact that the proportion ofwomen desiring more children declined between 1984 and 1988. The 1984/85 qualitative survey referred to earlier also showed that couples desirefewer children than before. Parents stated that children were becomingexpensive to bring up. While it is also true that children are perceived aseconomic assets, particularly because they give their parents security intheir old age, the respondents maintained that, in view of limited access toland, children had to be educated if they were to be useful.It may be pointed out that in rural areas it might be education of thechildren which affects a change in fertility rather than education of theparents. Most parents remarked that they did not want their children tosuffer without education as they had. Yet educating children was reportedas becoming more and more expensive Š hence the need to limit thechildren to an educatable number. It is interesting to note that the costs ofraising children were given, in order of priority, as education, clothing andfood for males. This order was reversed for females, reflecting the divisionof responsibilities within the African household. Generally couplesarticulated the increased monetization of the economy. Couples remarkedstrongly about the fact that they had to buy food (this was at the end ofthe 1981/2-1983/4 drought period); they argued that this was a newphenomenon. Couples also showed a tendency towards the need forquality rather than quantity in children. They emphasized the need toclothe and feed their children adequately and to educate them, a precursorto a fertility decline.One would posit that the socio-economic context in Zimbabwe duringthe 1980s has been conducive to a fertility decline. It can be noted thatcontraceptive knowledge increased by nine per cent between 1984 and1988 (Central Statistical Office, 1989, Table 4.1). Knowledge of the use ofcondoms and inter-uterine devices increased most; a possible explanationfor the former is their widespread promotion in the prevention of AIDS.The decline in knowledge of contraceptive injections is consistent withthe banning of this method.Current use of contraception has also increased slightly by about 5per cent (see Zimbabwe Demographic and Health Survey, 1988 (CentralStatistical Office, 1989), Table 4.14). What is interesting is the shift fromtraditional methods towards modern methods of contraception. Use ofmodern methods increased from 27 per cent in 1984 to 36 per cent in 1988,a nine per cent increase (Central Statistical Office, 1989).96 FERTILITY LEVELS AND TRENDS IN ZIMBABWECONCLUSIONThe scantiness of data in Zimbabwe makes the establishment of fertilitylevels and trends difficult and risky. This article has attempted to coverfertility levels for the 20-year period between the 1969 census and the 1988demographic survey, placing special emphasis on estimates from the 1982census and the 1984 survey. Fertility estimates for 1982 and 1984 from thedifferent estimation techniques ranged from 5,8 to 7,4. One may thusconclude that the total fertility rate of 6,5 as estimated from the Repro-ductive Health Survey is a close approximation of fertility for the period1982-4. This would mean that there has been a steady decline of fertilityfrom 1969 to 1988. The magnitude of the decline however, requires furtheranalysis.The decline in fertility is consistent with couples' declining desire tohave children. The decline in the demand for children is affected by theincreased monetization of the economy and the consequent increase inthe costs of feeding, clothing and educating children. The desire to limitand space children is facilitated by contraception, the use of which hasincreased from less than 10 per cent prior to 1980 to 38 per cent in 1984and 43 per cent in 1988 (for married women only) (Central StatisticalOffice, 1989).ReferencesBRASS, W. 1968 The Demography of Tropical Africa (Princeton, PrincetonUniv. Press).BRASS, W. 1975 Methods for Estimating Fertility and Mortality from Limitedand Defective Data (Chapel Hill, Univ. of North Carolina, Laboratoriesfor Population Statistics).CENTRAL STATISTICAL OFFICE 1969 Population Census 1969 (Salisbury,CSO).CENTRAL STATISTICAL OFFICE 1985 Main Demographic Features of thePopulation of Zimbabwe: An Advance Report Based on a Ten Per CentSample (Harare, Govt. Printer).CENTRAL STATISTICAL OFFICE 1989 Zimbabwe Demographic and HealthSurvey, 1988 (Harare, CSO and Institute for Development).CENTRAL STATISTICAL OFFICE 1991 Intercensal Demographic Survey 1987(Harare, CSO).CLARKE, D. G. 1972 'Problems of family planning amongst Africans inRhodesia', Rhodesian Journal of Economics, VI, ii, 36-48.HANENBERG, B. 1983 'Estimates of the total fertility rate based on thechild-woman ratio', Asian and Pacific Census Forum, X, ii, 5-11.KPEDEKPO, G. M. K. 1982 Essentials of Demographic Analysis of Africa(Heinemann, London).MARVELLOUS M. MHLOYI 97MHLOYI, M. M. 1985 'Couples' Perception on Family Planning Formation:Changes and Prospects' (Harare, Univ. of Zimbabwe, Dept. of Sociology,Working Paper).MHLOYI, M. M. 1991 'Fertility Levels and Determinants in Zimbabwe'(New York, Report to the Rockefeller Foundation).MZITE, D. J. K. 1981 'A Demographic Profile of Zimbabwe' (Canberra,Australian National Univ., MA thesis).RELE, J. R. 1967 Fertility Analysis through Extension of Stable PopulationConcepts (Berkeley, Univ. of California, Institute of InternationalStudies).SHRYOCK, H. S. and SIEGEL, J. S. 1976 The Methods and Materials ofDemography (New York, Academic Press).UNITED NATIONS 1967 Methods of Estimating Basic Demographic Measuresfrom Incomplete Data (New York, Dept. of Economic and Social Affairs,Manual IV Population Studies).UNITED NATIONS 1983 Indirect Techniques for Demographic Estimation(New York, Dept. of Economic and Social Affairs, Manual X PopulationStudies).ZIMBABWE NATIONAL FAMILY PLANNING COUNCIL 1985 Zimbabwe Repro-ductive Health Survey, 1984 (Harare, ZNFPC).