Changes in proximate determinants of fertility in sub-saharan Africa Kazuyo Machiyama, Andy Sloggett, John Cleland London School of Hygiene & Tropical Medicine Quality and comparability of demographic data in sub-saharan Africa INED/Pôle Suds 20 March 2012 Improving health worldwide www.lshtm.ac.uk
Background Several sub-saharan African countries (SSA) seem to have slowed down the pace of fertility decline, but the findings are inconclusive. (Casterline 2001, Bongaarts 2008, Garenne 2008, Schoumaker 2009, Sharpiro 2010, Sneeringer 2009) One reason for the disagreement may be due to the lack of consideration to the DHS data quality. Age displacement of children and omission may have affected recent fertility trends. (Schoumaker 2008, Machiyama 2009)
Observed TFR from DHS reports 7,5 Total fertility rates, 17 SSA, 1986-2011 7 6,5 6 5,5 5 4,5 4 3,5 1985 1990 1995 2000 2005 2010 BF BJ CM GH KE MD ML MW NG NI NM RW SN TZ UG ZM ZW
TFR from DHS reports no decline? 8 Total fertility rates, 17 SSA, 1986-2011 7,5 7 6,5 6 5,5 5 4,5 4 3,5 1985 1990 1995 2000 2005 2010 Incl. TFR from preliminary reports BF BJ CM GH KE ML NG RW TZ UG ZM ZW
Potential non-sampling errors Digit preference in women s age Age displacement of women Age displacement of children Incomplete age reporting (women & children) Different composition of women Use of different sampling frame Omission of births Nov 1997 Birth Feb 1998 Birth Jan 1998 Cut-off Month for child health Qs Jul 2002 Birth Aug 2003 Interviewed Ghana
Data quality problems on fertility trends - unadjusted Ghana Partial TFR (15-39) 2 3 4 5 6 7 8 9 DHS1989 DHS1993 DHS2003 DHS1998 DHS2008 1980 1985 1990 1995 2000 2005 Cut-off year Year
Data quality problems on fertility trends - adjusted Ghana Partial TFR (15-39) 2 3 4 5 6 7 8 9 1980 1985 1990 1995 2000 2005 Year
TFR (15-39): Loess-based approach Ghana Partial TFR (15-39) 2 3 4 5 6 7 8 9 Ghana adjusted estimates of TFR (15-39) fitted loess curve uncertainty areas based on different values of the Loess parameter α 1980 1985 1990 1995 2000 2005 Machiyama, K., Silverwood, R. Sloggett, A., and Cleland, J. 2009. Recent Year Fertility Declines in Sub-Saharan Africa:Analysis of country trends of fertility decline. Paper prepared for the 2010 Quetelet Seminar, 24-26 November 2010, Louvain-la-Neuve, Belgium..
Changes in pace of fertility decline between 1985-95 and 1995-2005 Average annual rate of decline(1995-2 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 Burkina Faso Madagascar Niger Senegal Rwanda Cameroon Ghana Mali Uganda Nigeria Zambia Malawi Tanzania Namibia Benin Zimbabw Kenya -0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 Average annual rate of decline(1985-95)
Background- cont d Few studies examined the mechanisms of the deceleration. Fertility changes should be supported by behavioural changes of fertility, i.e. the proximate determinants of fertility. Few applied the proximate determinants model to SSA. (Stover 1998, Kiersten et al. 2011) An examination of changes in proximate determinants provide an opportunity to assess what extent observed/fitted TFR estimates are consistent with the behavioural changes of fertility Socio-economic, cultural environment variables Proximate determinants of fertility Fertility
Research objectives 1. To assess changes in the proximate determinants in 17 sub-saharan African countries over the past two decades 2. To explore what extent changes of the proximate determinants support the estimated fertility decline by using the modified Bongaarts framework
Methods Data 65 DHS surveys in 17 SSA countries with 3 or more surveys undertaken (1986-2010) Benin, Burkina Faso, Cameroon, Ghana, Kenya, Madagascar, Malawi, Mali, Namibia, Niger, Nigeria, Rwanda, Senegal, Tanzania, Uganda, Zambia and Zimbabwe
Methods: Bongaarts proximate determinants framework (1978, 1982) Fertility-inhibiting effect of: Postpartum infecundability (Ci) Total Fecundity Rate (TF) Contraception and induced abortion (Cc, Ca) Marriage (Cm) Total fertility rate (TFR) TFR = TF * Cm * Ci * Ca * Cc
Characteristics of African fertility The original Bongaarts model is not applicable to SSA (e.g. Bongaarts 1983, Caldwell 1992) Appreciable premarital sex and childbearing (Garenne et al. 2006) Lower frequency of sex within marriage (Caraël 1995, Brown 2000) Polygyny High infertility (Frank 1983, Larsen, 1999) High reliance on traditional family planning methods (Johnson-Hanks 2002, Che et al. 2004)
Methods Stover s revision (1998) Abortion (Ca) Infertility (Cf) Potential Fertility (PF) Postpartum infecundability (Ci) Contraception (Cu) Sexually active (Cx) Total Fertility Rates (TFR) TFR = PF * Cx * Ci * Ca * Cu * Cf Stover, J. (1998). Revising the Proximate Determinants of Fertility Framework: What have We Learned in the Past 20 Years? Studies in Family Planning 29(3):255-267.
1 Results: Index of sexually active 0,9 0,8 0,7 0,6 0,5 0,4 BJ BF CM GH ML NI NG SN KE MD MW RW TZ UG ZM ZW NM 1st DHS 2nd DHS 3rd DHS 4th DHS 5th DHS
Recent sex by co-residence Proportions of married women (15-39) who had sex in the last 28 days by co-residential status, 17 SSA 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% married women livign with husbands/partners 0% 1985 1990 1995 2000 2005 2010 married women staying apart from husbands/partners 1985 1990 1995 2000 2005 2010 BF BJ CM GH KE MD ML MW NG NI NM RW SN TZ UG ZM ZW
Recent sex by marital status Proportions of women (15-39) who had sex in the last 28 days by marital status non-virgin never married formerly married women women 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 1985 1990 1995 2000 2005 2010 1985 1990 1995 2000 2005 2010 BF BJ CM GH KE MD ML MW NG NI NM RW SN TZ UG ZM ZW
1,00 Index of postpartum infecundability 0,90 0,80 0,70 0,60 0,50 0,40 BJ BF CM GH ML NI NG SN KE MD MW RW TZ UG ZM ZW NM 1st DHS 2nd DHS 3rd DHS 4th DHS 5th DHS
1,00 Index of infertility 0,90 0,80 0,70 0,60 0,50 0,40 BJ BF CM GH ML NI NG SN KE MD MW RW TZ UG ZM ZW NM 1st DHS 2nd DHS 3rd DHS 4th DHS 5th DHS
Contraceptive prevalence 0,70 Contraceptive prevalence 0,60 0,50 0,40 0,30 0,20 Traditional Modern 0,10 0,00 BJ BF CM GH ML NI NG SN KE MD MWRW TZ UG ZM ZW NM
Average effect of contraceptive 1,00 0,90 DHS1 DHS2 DHS3 DHS4 DHS5 0,80 0,70 0,60 0,50
Index of contraception 1 0,9 DHS1 DHS2 DHS3 DHS4 DHS5 0,8 0,7 0,6 0,5 0,4 0,3 BJ BF CM GH ML NI NG SN KE MD MW RW TZ UG ZM ZW NM
Projected vs Loess TFRs 8 Cameroon 8 Senegal 6 6 4 4 2 1991 1998 2004 2 1986 1993 1997 2005 Projected TFR loess TFR
Projected vs Loess TFRs 8 Ghana 8 Uganda 6 6 4 4 2 2 1988 1993 1998 2003 2008 8 Madagascar 1995 2000 2006 6 Projected TFR loess TFR 4 2 1992 1997 2003 2008
Projected vs Loess TFRs 8 Benin 8 Nigeria 6 6 4 4 2 1996 2001 2006 2 1990 1999 2003 2008 8 Tanzania Projected TFR loess TFR 8 Zambia 6 6 4 4 2 1992 1996 1999 2004 2010 2 1992 1996 2001 2007
Projected vs Loess TFRs 8 Kenya 8 Malawi 6 6 4 4 2 1989 1993 1998 2003 2008/9 8 6 Projected TFR 4 loess TFR 2 Rwanda 1992 2000 2004 2010 2 1992 2000 2005 2010
Projected vs Loess TFRs 8 Namibia 8 Zimbabwe Projected TFR 6 6 loess TFR 4 4 2 1992 2000 2006 2 1988 1994 1999 2005
Summary: comparison (between 1985-95 and 1995-2005) Trends of projected TFR based on the proximate Increase determinants framework Deceleration/ No change Decline Trends of loess estimates Deceleration Acceleration/ Constant decline Pre-transition Benin Nigeria Kenya Ghana Malawi Tanzania Zambia Mali Niger Zimbabwe Burkina Faso Cameroon Madagascar Namibia Rwanda Senegal Uganda
Limitations The proximate determinants model is not intended to provide accurate estimates. There may be (more) biases and errors in proximate determinants data Because only the data at the time of the surveys were available, there are 3-5 point estimates on proximate determinants in each country to assess the trends. Description of the TFR trends has to rely on the slope between the two point estimates. The slope may be affected by different levels of data quality across the surveys.
Discussion Overall, the TFR estimates based on the proximate determinants framework were consistent with the loess estimations. Specifically, the deceleration was supported by the changes in proximate determinants. The changes in each indicators significantly varied across the countries. The discrepancies in TFRs estimates between the projected and loess TFRs might be resulted from: (a) high contraceptive discontinuation rates (b) higher incidence of abortion (c) low or high proportion of unmarried women among all sexually active women (e.g. Namibia)
Discussion Deceleration is likely to be resulted from: (a) stagnation of increase in contraceptive prevalence (b) decrease in duration of postpartum abstinence (and amenorrhea) (c) reduced sterility The proximate determinants model is useful to validate the observed/fitted TFRs. The revised model seem to reasonably capture the changes.
Thank you! Contact: Kazuyo Machiyama Kazuyo.machiyama@lshtm.ac.uk