SLIDE 1
Brazilian Cohort Fertility by Education and Color/Race: Trends and Future Scenarios for Projections
Eduardo L.G. Rios-Neto Adriana de Miranda-Ribeiro Paula de Miranda-Ribeiro I – Introduction In the first decade of this century, Brazil entered into the group of countries that had below replacement fertility. Below replacement fertility occurred about 40 years after the onset of fertility transition in the country. The transition process began in the late 1960s when the total fertility rate (TFR) was 5.8 children per woman. This transition accelerated during the 1980s, reaching an average of 2.4 children per woman by the end of the century. Data from the National Household Sample Survey (PNAD) of 2006 showed that the TFR in the country had reached two children per woman. The following PNAD confirmed this trend and, according to the 2010 Demographic Census, the TFR in Brazil was 1.9. The fertility decline in Brazil was accompanied by a fall in the mean age of childbearing (MAC), suggesting a rejuvenation of the fertility schedule. The PNAD of the second half of the 2000s and the 2010 Demographic Census both indicated that there was a reversal in this trend towards a rising MAC. The strong fertility decline first with a fall and later with a rise in the MAC makes it difficult to devise clear scenarios for future fertility in Brazil. We draw on our previous works calculating tempo and quantum fertility in Brazil to present a different approach calculating cohort fertility at ages 30, 35, and 40. We calculate cohort fertility by four education groups and color/race. Cohort fertility stratified by education presents greater variation among the different birth cohorts of women than in the case of women classified by color/race. This paper uses a dataset based on a methodology that we developed to build birth histories from Demographic Censuses. We add the 2010 census to the birth history already built to the 2000 Brazilian Demographic Census. We can move back as far as fifteen years with the calculated birth histories, a combination of this information with the total number of children ever born entail the calculation of cohort fertility at certain specific ages. After the calculation of cohort fertility, we perform some forecasting exercises with cohort fertility, TFR, adjusted TFR, tempo effect, and mean age at first birth. Some regressions of TFR with cohort fertility and tempo effect are performed to simulate predictions. A critical analysis of all results is performed to devise future fertility scenarios. II – Previous Findings In previous papers, we calculated the tempo effect and parity and tempo adjusted fertility for Brazil (Rios-Neto and Miranda-Ribeiro, 2015; and Miranda-Ribeiro, Rios-Neto, and Garcia, 2016). Following Figure 1, we can see that the Brazilian tempo effect is negative during the 1990’s, approaching zero at the end of this decade. The tempo effect becomes positive, between 5% and
SLIDE 2 10% during the first decade of this century. Over the whole period, the tempo effect goes from a 20% inflation of TFR in the early 1990s to a 10% deflation in late 2000’s. Figure 1 – Tempo effect (%) in Brazil, 1986 to 2010.
Sources: IBGE – 1991, 2000 and 2010 Brazilian Censuses in Rios-Neto and Miranda-Ribeiro (2015).
Figure 2 - Brazil, 1986 to 2010: observed TFR and adjusted PATFR.
Sources: IBGE – 2000 and 2010 Brazilian Censuses in Rios-Neto and Miranda-Ribeiro (2015).
5 10 15 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 Tempo Effect (%) 25,5 26 26,5 27 27,5 28 28,5 29
TEMPO (2000) TEMPO (2010) Mean age at childbearing 1,00 1,50 2,00 2,50 3,00 3,50 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 Total Fertility Rate
PATFR_ADJUSTED TFR_OBSERVED
SLIDE 3 We evaluate the distortions caused by tempo and parity composition effects on TFR by the comparing observed TFR with PATFRADJUSTED. The PATFRADJUSTED is the value of TFR without changes in the mean age at childbearing and in the parity composition. See the TFR and PATFRADJUSTED series for Brazil in Figure 2. The difference between TFR and PATFRADJUSTED diminishes over time, while both TFR and the pure index (quantum) reach replacement level. III – A Cohort Fertility Perspective Our technique to construct birth histories out of the period demographic censuses allows us to go back in time 15 years for period fertility. Thus, using the age-period-cohort identity, it is possible to calculate the number of children ever born at age 30 (CEBc30) for all women in the 30-44 age
- group. It is also possible to calculate CEBc35 for all women in the 35-49 age group and CEBc40
for all women in 40-49 age group. These cohort measures will refer to different single year cohorts at different periods or time references In Table 1, Figures 3 and 4, we observe a decline in cohort fertility. While the decline in cohort fertility at age thirty (CEBc30) may result from tempo and quantum fertility, cohort fertility at age forty (CEBc40) is clearly an indicator of quantum fertility. The youngest cohort fertility at age forty (for the 1970 birth cohort, age forty in 2010) is 2.34 (above replacement). The youngest cohort fertility at age thirty (for the 1980 birth cohort, age thirty in 2010) is 1.5. If we apply the ratio between cohort fertility at age forty and thirty for the 1970 birth cohort, we find 1.27. Applying this ratio to the 1980 birth cohort, we find a cohort fertility at age forty of 1.91 in 2020 (below replacement). Figure 3:
0.5 1 1.5 2 2.5 3 3.5 4 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980
Cohort Fertility Birth Cohorts
COHORT FERTILITY AT AGES 30, 35, AND 40 BY BIRTH COHORTS - BRAZIL
CEBc30 CEBc35 CEBc40
SLIDE 4 Figure 4:
0.5 1 1.5 2 2.5 3 3.5 4
Cohort Fertility BIRTH COHORTS
COHORT FERTILITY AT AGES 30, 35, AND 40 BY BIRTH COHORT - BRAZIL
CEBc30 CEBc35 CEBc40
SLIDE 5
Table 1: Cohort Fertility at Ages 30, 35, and 40 by Birth Cohorts - Brazil
Sources: IBGE – 2000 and 2010 Brazilian Censuses – Own calculation Birth Cohort CEBc30 CEBc35 CEBc40 1951 3.219812183 3.626485498 1952 3.174274744 3.544310425 1953 3.089868857 3.423922755 1954 3.030046682 3.341859807 1955 2.988515947 3.277960611 1956 2.333149415 2.938857104 3.213732747 1957 2.303984496 2.85908914 3.122417035 1958 2.304564469 2.840243602 3.0930195 1959 2.216850973 2.711209692 2.945388884 1960 2.221954664 2.700628487 2.931480828 1961 2.158376353 2.633498119 2.748370792 1962 2.121837713 2.579278491 2.69652723 1963 2.098956799 2.539665747 2.641316747 1964 2.062763372 2.490369949 2.600298029 1965 2.002442669 2.329977059 2.509266603 1966 1.939051242 2.199765617 2.482466694 1967 1.930217616 2.185388296 2.469764345 1968 1.904483281 2.162484874 2.426488422 1969 1.868873178 2.117422178 2.375546979 1970 1.849622981 2.089574554 2.342615689 1971 1.718580985 2.0320373 1972 1.713219667 2.026090404 1973 1.711634458 2.019270805 1974 1.685976031 1.990509 1975 1.665241369 1.964023779 1976 1.651681508 1977 1.632020745 1978 1.595841927 1979 1.562460699 1980 1.502399085
SLIDE 6 IV – Cohort Fertility by Education and Color/Race Due to the large sample sizes in the demographic censuses, it is also possible to perform these calculations for specific attributes such as mothers´ education groups and race/color. With these calculations, not only we can highlight some socioeconomic gradients of low fertility, but we can also devise auxiliary tools for scenarios associated with future cohort fertility in Brazil. We can see in Figure 5 that cohort fertility at age 30 among all women in the 1956 birth cohort was 2.25 (a little above 2.1). Total cohort fertility in this cohort was almost the same level as the cohort fertility in the group of women with primary education (4-8 years of study). As we approach the cohort of women born in 1980, then total cohort fertility approaches the level of cohort fertility in the group of women with secondary education (9 to 11 years of schooling). As the cohort fertility for each education group declines less than cohort fertility for all women, we may infer that education composition played an important role in cohort´s fertility decline. Indeed, Figure 6 shows that roughly 30% of the 1956 birth cohort had secondary or tertiary education, while the number for the 1980 cohort corresponds to 66%. This major composition shift is likely to explain a great portion of the cohort fertility decline. Cohort fertility is more associated with “quantum” than with “tempo” effects, but it is also true that the cut line at age 30 may favor the operation of some tempo effect, particularly in the case of women with tertiary education. Figure 5: Cohort Fertility at Age 30 by Women´s Education Groups – Brazil
Source: IBGE, 2000 and 2010 Demographic Censuses, Microdata, own calculations.
0,0000 0,5000 1,0000 1,5000 2,0000 2,5000 3,0000 3,5000
CEB30Cohort Birth Cohort
Birth Cohort Fertility at Age 30 (CEB30Cohort) by Groups of Education - Brazil
Total 0 to 3 Years of Study 4 to 8 Years of Study 9 to 11 Years of Study 12+ Years of Study
SLIDE 7 Figure 6: Education Structure of Women by Birth Cohorts – Brazil Source: IBGE, 2000 and 2010 Demographic Censuses, Microdata, own calculations.
Figure 7: Cohort Fertility at Age 35 by Women´s Education Groups – Brazil
Source: IBGE, 2000 and 2010 Demographic Censuses, Microdata, own calculations.
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
% BIRTH COLHORT
EDUCATION STRUCTURE OF WOMEN BY BIRTH COHORT - BRAZIL
12+ Years of Study 9 to 11 Years of Study 4 to 8 Years of Study 0 to 3 Years of Study 0.0000 0.5000 1.0000 1.5000 2.0000 2.5000 3.0000 3.5000 4.0000 4.5000 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975
Cohort Fertility at age 35 Birth Cohort
COHORT FERTILITY AT AGE 35 BY BIRTH COHORT AND EDUCATION ATTAINMENT - BRAZIL
0 to 3 years of study 4 to 8 years of study 9 a 11 years of study 12+ years of study Total
SLIDE 8 Figures 7 and 8 present the same education stratification as before for the analysis of cohort fertility at ages 35 and 40. There is a striking decline of cohort fertility for the most uneducated group of
- women. This decline is even stronger in the CEBc40 measure. As Table 2 indicates, CEBc40 of
uneducated women declined from 4.9 for the 1951 cohort to 3.02 among the women in the 1970
- cohort. According to this sample, the 1970 cohort reaches CEBc40 of 2.06 (replacement level).
Figure 8: Cohort Fertility at Age 40 by Women´s Education Groups – Brazil
Source: IBGE, 2000 and 2010 Demographic Censuses, Microdata, own calculations.
0.00 1.00 2.00 3.00 4.00 5.00 6.00
Cohort Fertility at age 40 Birth Cohort
COHORT FERTILITY AT AGE 40 BY BIRTH COHORT AND EDUCATION ATTAINMENT - BRAZIL
0 to 3 years of study 4 to 8 years of study 9 a 11 years of study 12+ years of study Total
SLIDE 9
Table 2: Cohort Fertility at Age 40 – CEBc40 by Education - Brazil
Source: IBGE, 2000 and 2010 Demographic Censuses, Microdata, own calculations.
Table 3: Cohort Fertility at Age 40 – CEBc40 by Color/Race - Brazil
Source: IBGE, 2000 and 2010 Demographic Censuses, Microdata, own calculations.
Birth Cohort 0 to 3 years of study 4 to 8 years of study 9 a 11 years of study 12+ years of study Total 1951 4.89 3.19 2.24 1.84 3.63 1952 4.79 3.17 2.22 1.79 3.54 1953 4.70 3.12 2.23 1.77 3.42 1954 4.61 3.11 2.18 1.76 3.34 1955 4.57 3.04 2.17 1.73 3.28 1956 4.52 3.03 2.14 1.70 3.21 1957 4.43 2.98 2.11 1.66 3.12 1958 4.39 2.96 2.09 1.65 3.09 1959 4.27 2.88 2.04 1.61 2.95 1960 4.18 2.86 2.01 1.59 2.93 1961 4.08 2.78 2.03 1.63 2.75 1962 3.97 2.72 1.97 1.58 2.66 1963 3.87 2.66 1.94 1.56 2.59 1964 3.76 2.59 1.89 1.49 2.52 1965 3.57 2.51 1.82 1.46 2.40 1966 3.48 2.47 1.78 1.42 2.35 1967 3.42 2.43 1.74 1.37 2.30 1968 3.27 2.38 1.69 1.33 2.22 1969 3.14 2.32 1.65 1.27 2.13 1970 3.02 2.26 1.56 1.21 2.06 Birth Cohort White Black Brown Other Total 1951 3.09 4.04 4.49 3.46 3.63 1952 3.02 3.99 4.35 3.37 3.54 1953 2.93 3.79 4.23 3.06 3.42 1954 2.87 3.73 4.08 3.29 3.34 1955 2.81 3.67 4.00 3.28 3.28 1956 2.75 3.66 3.92 3.00 3.21 1957 2.68 3.54 3.79 3.08 3.12 1958 2.65 3.52 3.73 3.04 3.09 1959 2.55 3.41 3.53 2.88 2.95 1960 2.52 3.37 3.51 3.02 2.93 1961 2.35 3.04 3.22 3.07 2.75 1962 2.31 2.94 3.15 2.97 2.70 1963 2.28 2.86 3.07 2.86 2.64 1964 2.23 2.85 3.02 2.97 2.60 1965 2.18 2.74 2.89 2.79 2.51 1966 2.15 2.69 2.86 2.72 2.48 1967 2.14 2.69 2.83 2.78 2.47 1968 2.13 2.64 2.74 2.73 2.43 1969 2.09 2.57 2.68 2.74 2.38 1970 2.06 2.56 2.62 2.73 2.34
SLIDE 10 Figure 9: Cohort Fertility at Age 40 by Women´s Color/Race – Brazil
Source: IBGE, 2000 and 2010 Demographic Censuses, Microdata, own calculations.
The analysis of Table 3 and Figure 9 shows that cohort fertility at age forty (CEBc40) declines in all color/race segments, although white women present the lower cohort fertility level and brown women are the ones presenting the fastest cohort fertility decline. Regarding color composition by birth cohort, there is more stability than the one found in women’s education, but there is a decline in the proportion of white women and an increase in the proportion of brown women. When we compare the cohort fertility by women’s education attainment with color, we notice that the differentials are more pronounced in education than in the color segments. The color stratification is more stable and can provide a more conservative description of the cohort fertility dynamics, while the education stratification presents a strong shift among the different birth cohorts of women. Another difference is that total cohort fertility in the education stratification converges to the higher education segments, while in the color there is a convergence process among the segments. An interaction between these two dimensions may provide further information for the limits for cohort fertility decline among Brazilian women. Notice a difference between total cohort fertility in education and race (Tables 2 and 3). This difference takes place
- nly in the series built from the 2010 Demographic Census, it is caused by missing cases in the
algorithm for calculating completed years of study which bias total cohort fertility downwards.
0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 5.00 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970
Cohort Fertility at age 40 Birth Cohort
COHORT FERTILITY AT AGE 40 BY BIRTH COHORT AND RACE/COLOR - BRAZIL
White Black Brown Other Total
SLIDE 11 IV – Forecasting Cohort Fertility and its Impact on TFR We take cohort fertility at age 40 as a reasonable measure for quantum fertility, the impact of tempo effect would take place in cohort fertility at earlier ages. In general, the best adjustments were obtained with the ARIMA (1,1,0) model, including one autoregressive and first difference
- components. The random walk model with a drift (0,1,0) is also presented in Table 4 to give a
- comparison. In this case the forecasted values are similar. Figure 10 displays the forecast of the
ARIMA (1,1,0) model with 95% confidence interval. Table 5 displays a simple linear regression between period TFR and cohort fertility at age 40. The implying period of TFR corresponds to the birth cohort date plus forty. Of course there is a strong correlation between these two variables, since there is a translation between cohort and period
- fertility. In Table 6 we show that a simulation of TFR based on our forecast of cohort fertility at
age 40 seems to exaggerate period fertility decline beyond a reasonable value. Notice that our TFR estimated series presented in Table 6 displays values that are downward biased, because it was estimated by birth history smoothed values without indirect estimation. If we compare our value for 2010 (1.56) with the one estimated by the Brazilian Census Bureau (1.9), we have a 1.218 inflation factor. Even correcting for this bias, we would predict a TFR of 1.07 for the year 2020.
SLIDE 12
Table 4: Forecast Cohort Fertility at Age 40 Figure 10: Forecast – Cohort Fertility at age 40 by Birth Cohort – ARIMA (1,1,0)
COHORT (0,1,0) (1,1,0) Period Age40
1971 2.28 2.26
2011
1972 2.21 2.20
2012
1973 2.14 2.13
2013
1974 2.07 2.06
2014
1975 2.00 2.00
2015
1976 1.94 1.93
2016
1977 1.87 1.86
2017
1978 1.80 1.79
2018
1979 1.73 1.72
2019
1980 1.67 1.66
2020
1981 1.60 1.59
2021
1982 1.53 1.52
2022
1983 1.46 1.45
2023
1984 1.40 1.38
2024
1985 1.33 1.32
2025
1986 1.26 1.25
2026
1987 1.19 1.18
2027
1988 1.13 1.11
2028
1989 1.06 1.05
2029
1990 0.99 0.98
2030
SLIDE 13 Table 5: Regression – Period TFR (dependent) – Total (CohF40, Ind.) Table 6: Predicted TFR based on Forecasted Cohort Fertility
B
(Constant)
0.078 TOTAL 0.897 0.027 Model R R Square Adjusted R Square
Estimate Durbin- Watson 1 .992a 0.984 0.983 0.04836729 0.802
- a. Predictors: (Constant), TOTAL
- b. Dependent Variable: TFR
Model 1
PERIOD COHORT TFR CohortFert PERIOD COHORT Pred-TFR Forec-Cohort
1991
1951 2.58 3.6 2011
1971
1.42 2.26
1992
1952 2.52 3.5 2012
1972
1.36 2.20
1993
1953 2.48 3.4 2013
1973
1.30 2.13
1994
1954 2.45 3.3 2014
1974
1.24 2.06
1995
1955 2.41 3.3 2015
1975
1.18 2.00
1996
1956 2.33 3.2 2016
1976
1.12 1.93
1997
1957 2.24 3.1 2017
1977
1.06 1.86
1998
1958 2.14 3.1 2018
1978
1.00 1.79
1999
1959 2.06 2.9 2019
1979
0.94 1.72
2000
1960 1.97 2.9 2020
1980
0.88 1.66
2001
1961 1.87 2.7 2021
1981
0.82 1.59
2002
1962 1.77 2.7 2022
1982
0.75 1.52
2003
1963 1.70 2.6 2023
1983
0.69 1.45
2004
1964 1.66 2.6 2024
1984
0.63 1.38
2005
1965 1.64 2.5 2025
1985
0.57 1.32
2006
1966 1.61 2.5 2026
1986
0.51 1.25
2007
1967 1.59 2.5 2027
1987
0.45 1.18
2008
1968 1.57 2.4 2028
1988
0.39 1.11
2009
1969 1.56 2.4 2029
1989
0.33 1.05
2010
1970 1.56 2.3 2030
1990
0.27 0.98
SLIDE 14
V – Forecasting TFR and the Role of Tempo Effect on TFR In this item we apply the methodology developed in the previous item, we start presenting the forecast of TFR and the tempo effect. Then, we move on to design a simple regression model including TFR as the dependent variable and tempo effect as the independent variable. The forecasted tempo effect applied in the regression model will predict TFR. Table 7: Forecasted TFR based on ARIMA (1,1,0) Model Table 7 and Figure 11 present the TFR forecast using the ARIMA (1,1,0) model. An alternative estimation of the ARIMA (2,1,0) model provides a similar adjustment, but it estimates a stronger and less reliable declining trend in TFR. In 2020 the model predicts a TFR of 1.41, inflated to 1.72 to be comparable with 1.90 em 2010. Figures 12 and 13 present the forecasted tempo effect. In Figure 12 the forecasted tempo effect is not linearly rising in the first six projected years, it operates more strongly after 2018. Figure 13 presents the linearly increasing trend of the tempo effect under ARIMA (1,1,0) model.
PERIOD TFR Adj-TFR PERIOD Pred-TFR Adj-Pred-TFR
1991
2.58 3.14 2011 1.56 1.90
1992
2.52 3.07 2012 1.55 1.89
1993
2.48 3.02 2013 1.54 1.88
1994
2.45 2.98 2014 1.53 1.86
1995
2.41 2.94 2015 1.51 1.84
1996
2.33 2.84 2016 1.50 1.82
1997
2.24 2.73 2017 1.48 1.80
1998
2.14 2.61 2018 1.46 1.78
1999
2.06 2.50 2019 1.43 1.75
2000
1.97 2.40 2020 1.41 1.72
2001
1.87 2.27 2021 1.38 1.68
2002
1.77 2.16 2022 1.35 1.65
2003
1.70 2.07 2023 1.32 1.61
2004
1.66 2.02 2024 1.29 1.57
2005
1.64 2.00 2025 1.25 1.53
2006
1.61 1.96 2026 1.22 1.48
2007
1.59 1.93 2027 1.18 1.44
2008
1.57 1.91 2028 1.14 1.39
2009
1.56 1.90 2029 1.10 1.34
2010
1.56 1.90 2030 1.05 1.28
SLIDE 15
Figure 11: Forecast – TFR – ARIMA (1,1,0) Figure 12: Forecast – Tempo Effect – ARIMA (2,1,0)
SLIDE 16 Figure 13: Forecast – Tempo Effect – ARIMA (1,1,0) Table 8 presents the regression model that will base the simulation of TFR based on tempo effect´s
- forecast. The two simulations are in Table 9, where the tempo effect forecast using the ARIMA
(2,1,0) model and the alternative ARIMA (1,1,0) model are presented. The latter forecast did not generate a stationary sequence. Table 8 – Regression Model – TFR (Dependent) – Tempo Independent Variable
Model Summary
Model R R Square Adjusted R Square
Estimate 1 .895a .800 .793 .306896186000 000
- a. Predictors: (Constant), Tempo
Coefficientsa
Model Unstandardized Coefficients Standardized Coefficients t Sig. B
Beta 1 (Constant) 2.120 .061 34.519 .000 Tempo
.009
.000
- a. Dependent Variable: TFR
SLIDE 17 Table 9: Forecasted Tempo Effect based on ARIMA (2,1,0) and (1,1,0) and Model Fertility Simulation Figure 14:
PERIOD TEMPO TFR Adj-TFR MODEL (2,1,0) PERIOD Tempo (2,1,0) MODEL (2,1,0) Tempo (1,1,0) MODEL (1,1,0) 1982
3.80 4.62 3.35 2011
2.35 1.89 2.47
1.88 1983
3.60 4.38 3.35 2012
2.26 1.90 2.59
1.87 1984
3.42 4.17 3.33 2013
2.15 1.91 2.72
1.86 1985
3.27 3.98 3.30 2014
2.08 1.92 2.86
1.85 1986
3.15 3.83 3.24 2015
2.09 1.92 3.01
1.83 1987
3.04 3.70 3.17 2016
2.20 1.91 3.16
1.82 1988
2.92 3.56 3.07 2017
2.44 1.89 3.33
1.80 1989
2.80 3.41 2.96 2018
2.83 1.85 3.50
1.78 1990
2.68 3.26 2.83 2019
3.34 1.80 3.69
1.77 1991
2.58 3.14 2.66 2020
3.97 1.74 3.87
1.75 1992
2.52 3.07 2.48 2021
4.70 1.67 4.07
1.73 1993
2.48 3.02 2.27 2022
5.48 1.59 4.27
1.71 1994 0.38 2.45 2.98 2.08 2023
6.30 1.52 4.48
1.69 1995 2.01 2.41 2.94 1.93 2024
7.10 1.44 4.69
1.67 1996 3.16 2.33 2.84 1.82 2025
7.86 1.37 4.91
1.65 1997 3.81 2.24 2.73 1.75 2026
8.55 1.30 5.13
1.63 1998 3.98 2.14 2.61 1.74 2027
9.15 1.24 5.36
1.61 1999 3.74 2.06 2.50 1.76 2028
9.66 1.19 5.59
1.58 2000 3.21 1.97 2.40 1.81 2029
10.07 1.15 5.82
1.56 2001 2.56 1.87 2.27 1.87 2030
10.40 1.12 6.06
1.54 2002 1.93 1.77 2.16 1.93 2003 1.47 1.70 2.07 1.98 2004 1.23 1.66 2.02 2.00 2005 1.22 1.64 2.00 2.00 2006 1.39 1.61 1.96 1.99 2007 1.70 1.59 1.93 1.96 2008 2.03 1.57 1.91 1.93 2009 2.28 1.56 1.90 1.90 2010 2.37 1.56 1.90 1.89
0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 2020 2022 2024 2026 2028 2030
TFR YEAR
Observed-Forecasted- Total Fertility Rate versus Simulated Fertility based on Tempo Effect's Forecast - Brazil - 1982-2030
TFR-OBS-ARIMA TFR-Reg-Est-Tempo
SLIDE 18 Notice that in Table 9, both models simulating TFR based on tempo effect forecasts do not predict a strong decline in TFR until 2020, but the decline in period fertility is rather strong between 2020 and 2030. This pattern is depicted in Figure 14, based on TFR forecasts ARIMA (1,1,0) and simulated fertility based on model (2,1,0). VI – Forecasting TFR using and the Role of Age at First Birth on TFR The singulate mean age at first birth is the age at having the first birth for all women that went through this transition. This variable is clearly associated with the operation of the tempo effect and also with the “postponement transition” concept. The literature suggests that the onset of the postponement transition takes place when the mean age at first birth increases two years from the previous stable values. Table 11 suggests that this variable was rather stable during the Brazilian fertility transition, it started to rise in the beginning
- f this century, going from around 25 to around 27 in 2010. Thus, the year 2010 marks the onset
- f the postponement transition in Brazil.
The forecast of mean age at first birth using ARIMA (1,1,0) is displayed in Figure 15 and Table 11, based on the regression model displayed in Table 10. Based on the forecast of rising mean age at first birth, the fertility decline simulated in Table 11 is quite strong and beyond reasonable values.
SLIDE 19 Figure 15: Forecast – Age at First Birth – ARIMA (1,1,0) Table 10: Regression Model – TFR (Dependent) – Age at First Birth Independent Variable
Model Summaryb
Model R R Square Adjusted R Square
Estimate Change Statistics R Square Change F Change df1 d 1 .904a .817 .807 .16510893300000 .817 80.596 1
- a. Predictors: (Constant), FirstBirth
- b. Dependent Variable: TFR
Coefficientsa
Model Unstandardized Coefficients Standardized Coefficients T Sig. B
Beta 1 (Constant) 13.017 1.229 10.588 .000 FirstBirth
.048
.000
- a. Dependent Variable: TFR
SLIDE 20
Table 11: Forecasted Age at First Birth and Model Fertility Simulation VII – Final Remarks The first part of the paper deals with cohort fertility, built on the developed birth histories. Cohort fertility at late ages (in our case age 40) is less likely to be affected by the postponement of childbearing (tempo effect) and more likely to be a proxy of “quantum” fertility. Our data explorations suggested that cohort fertility is a little above replacement level for cohorts born in 1970 and 40 years old in 2010. Education and color segments differentiate cohort fertility, although cohort change through time is dominated by the composition effect in the case of education and by cohort fertility decline within segments in the case of color. Forecast exercises with cohort fertility suggest a reasonable decline in cohort fertility (quantum) for the birth cohorts born between 1970 and 1980, but it seems to exaggerate the cohort decline from that point on. Furthermore, the model simulation of period TFR based on cohort fertility predicts exceedingly low TFR with unreliable decline through time. The forecasting exercise with TFR using the ARIMA (1,1,0) model seems to predict a robust exercise compared to all other exercises in the paper, it could be a kind of baseline for all other
PERIOD Age1stB TFR Adj-TFR PERIOD Age1stB (1,1,0) MODEL (1,1,0) 1991 24.67 2.58 3.14 2011
27.17 1.22
1992 24.68 2.52 3.07 2012
27.36 1.14
1993 24.70 2.48 3.02 2013
27.53 1.07
1994 24.71 2.45 2.98 2014
27.68 1.00
1995 24.72 2.41 2.94 2015
27.83 0.94
1996 24.70 2.33 2.84 2016
27.96 0.88
1997 24.73 2.24 2.73 2017
28.10 0.82
1998 24.76 2.14 2.61 2018
28.23 0.77
1999 24.88 2.06 2.50 2019
28.35 0.71
2000 24.99 1.97 2.40 2020
28.48 0.66
2001 25.29 1.87 2.27 2021
28.61 0.60
2002 25.47 1.77 2.16 2022
28.73 0.55
2003 25.66 1.70 2.07 2023
28.85 0.49
2004 25.80 1.66 2.02 2024
28.98 0.44
2005 26.06 1.64 2.00 2025
29.10 0.39
2006 26.18 1.61 1.96 2026
29.22 0.33
2007 26.31 1.59 1.93 2027
29.35 0.28
2008 26.50 1.57 1.91 2028
29.47 0.23
2009 26.69 1.56 1.90 2029
29.59 0.17
2010 26.95 1.56 1.90 2030
29.72 0.12
SLIDE 21
- exercises. It indicates that the generation of a good TFR time series can be valuable for fertility
projections, even using a very simplistic model. The forecast of tempo effect and simulation of TFR based on a regression estimation is helpful to disentangle the role of the postponement transition in projected period TFR decline. The tempo effect time series is generated by the Kohler-Ortega methodology. We can not say if the trend is not affected by our birth history methodology, since we observe periods of increasing tempo effect followed by some plateau. Our ARIMA (2,1,0) forecast replicates that pattern in relation with the tempo effect. The simulation of TFR based on a regression predicts a slow fertility decline between 2010 and 2020, with a strong declining in the following decade. An alternative tempo effect forecast using ARIMA (1,1,0) predicts a faster decline path between 2010 and 2020 and a slower decline in the following decade. The first simulation seems more reliable. Finally, we tried to capture the tempo effect by forecasting the age at first birth. The forecast of rising age at first birth using ARIMA (1,1,0) seems reasonable. It confirms that Brazil will be going through a postponement transition between 2010 and 2030. The simulation of TFR based
- n a regression overpredicts the fertility decline in the next twenty years.
Balancing all exercises performed here, we conclude that the fertility decline pace seem to be lower between 2010 and 2020, but the momentum of the postponement transition and/or the operation
- f quantum fertility decline will most likely plunge TFR to lowest low levels in 2030, that is to
say, below the 1.4 mark. REFERENCES: Miranda-Ribeiro, Adriana de, Eduardo L.G. Rios-Neto, Ricardo Alexandrino Garcia (2016) Anticipación y postergación de los nacimientos en la transición de la fecundidad en Brasil. Notas de Población, v. 103, pp. 29-43. Rios-Neto, Eduardo L.G., Adriana de Miranda-Ribeiro (2015) Fertility Decline in Brazil: Tempo, Quantum and Parity Composition Effects. In: Population Association of America Annual Meeting, San Diego - Califórnia.
