Estimating men’s fertility from vital registration data with missing values

Figures & data

Figure 1 Trends in the proportion of births with unknown paternal age, Sweden (1968–2014), US (1969–2015), Spain (1975–2014), and Estonia (1989–2013)

Source: Statistics Sweden, National Bureau of Economic Research, Spanish Statistical Office, Statistics Estonia, Human Mortality Database; own calculations.

Figure 2 Absolute bias of the men’s total fertility rate (MTFR) in MTFR points, based on the unconditional approach to estimating paternal age, by age shift and proportion missing; simulations for Sweden, 2014

Source: Own calculations.

Figure 3 Absolute bias of the men’s total fertility rate (MTFR) in MTFR points, based on the conditional approach to estimating paternal age, by age shift and proportion missing; simulations for Sweden, 2014

Source: Own calculations.

Figure 4 Absolute bias of the paternal mean age at childbirth (PMAC), based on the unconditional approach to estimating paternal age, by age shift and proportion missing; simulations for Sweden, 2014

Source: Own calculations.

Figure 5 Absolute bias of the paternal mean age at childbirth (PMAC), based on the conditional approach to estimating paternal age, by age shift and proportion missing; simulations for Sweden, 2014

Source: Own calculations.

Table 1 Proportion of simulations in which the conditional approach for estimating paternal age at childbirth outperforms the unconditional approach; by country and total (percentages)

	Men’s total fertility rate (MTFR)			Paternal mean age at childbirth (PMAC)
Column number	1	2	3	4	5	6
Criteria for inclusion
Age shift	−4 to +4	−2 to +2	−1 to +1	−4 to +4	−2 to +2	−1 to +1
Proportion missing	Any	>10 per cent	>10 per cent	Any	>10 per cent	>10 per cent
Percentage of simulations in which conditional approach outperforms unconditional approach
Sweden (1968–2014)	76	83	91	56	62	69
US (1969–2015)	75	92	97	73	94	100
Spain (1975–2014)	72	83	95	66	78	97
Estonia (1989–2013)	55	66	78	59	67	78
Average	71	83	92	64	76	86

Notes: Columns 1 and 4 take age shift values between −4 and +4 into account. Columns 2 and 5 are restricted to simulations with age shift values between −2 and +2, and also consider only simulations in which paternal age is missing for more than 10 per cent of births. Columns 3 and 6 cover age shift values between −1 and +1, and a proportion of missing values of at least 10 per cent.

Source: Own calculations.

Figure A1 Proportion of missing values for age of father conditional on age of mother, Sweden 2014

Notes: Empirical data (solid line) and upscaled distributions (dotted and dashed lines). For some of the youngest and oldest ages, the 90 per cent threshold was reached and the proportion of missing values was not further increased. Source: Statistics Sweden; own calculations.

Figure A2 Distribution of age of father conditional on mother being aged 30, Sweden 2014

Notes: Original distribution (solid line) and age-shifted distributions (dotted and dashed lines). Source: Statistics Sweden; own calculations.