We obtained the age and sex of children from 301 families who attended secondary schools that recruited from a wide range of socioeconomic groups. The mean age difference Da (age of husband−age of wife) was 2.48 years ± 0.23 (s.e.m.) and there were 301 first-born and 260 second-born children. Among first-borns there was an excess of daughters from couples with low Da and an excess of sons from those with high Da (Da = −9 to −1 years: 14 sons and 29 daughters; Da = 0 to 5 years: 117 sons and 84 daughters; Da = 5 to 15 years: 37 sons and 20 daughters; x2 = 11.86, P = 0.0027). Among second-borns there was the opposite but non-significant tendency (Da = −9 to −1 years: 22 sons and 11 daughters; Da = 0 to 4 years: 93 sons and 89 daughters; Da = 5 to 17 years: 20 sons and 25 daughters; x2 = 3.93, P = 0.14).

The age of parents at the birth of the child has a weak effect on the child's sex3. However, multiple regression analyses with sex of child as the dependent variable and Da and age of mother or father at birth as independent variables showed that Da remained significantly associated with sex of child (Da/age of mother — Da: standardized partial regression coefficient b1 = −0.14, t = 2.35, P = 0.02; age of mother: b2 = 0.13, t = 0.22, P = 0.83; Da/age of father — Da: b1 = 0.14, t = 2.34, P = 0.02; age of father: b2 = 0.13, t = 0.21, P = 0.83).

Local and national patterns of Da during the period 1911-52 (ref. 4) are shown in Fig. 1a, c. If couples do not delay the birth of their first child, Da and sex ratio should be correlated and changes in the sex ratio should be preceded by changes in Da. This is seen in 1914-18 but not during the Second World War (Fig. 1b, c). Registration of second and subsequent births will weaken the relationship between Da and sex ratio so that an exact correlation is unlikely. Nevertheless a regression of sex ratio on Da shows that the latter explains 68% of the variance of the former (Fig. 1d). Age of woman at marriage was negatively related to the sex ratio (b = −0.003, r2 = 0.23, F = 12.19, P = 0.001). However a multiple regression analysis with sex ratio as the dependent variable and Da and bride's age as independent variables left Da as the only significant correlate of sex ratio (Da: b1 = 0.78, t = 8.26, P = 0.0001; age of bride: b2 = −0.14, t = 1.51, P = 0.14).

Figure 1: Parental age differences and sex-ratio statistics, 1911-52.
figure 1

a, The relationship between the mean (±s.e.m) of the difference in age between husbands and wives (Da) and year of marriage (1935-52) in the Woolton area of Liverpool. There is a significant curvilinear relationship with a peak value of Da in 1947 (second order polynomial, y = −42.15+2.024x−0.022x2, F = 5.88, P = 0.013, n = 469 marriages). b, Sex ratios of births registered in England and Wales from 1911-52; and c, Da for marriages in the same period. d, Linear regression of sex ratio of births in England and Wales against Da, 1911-52 (r2 = 0.68, F = 86.46, P = 0.0001).

Rank in many animals is related to the sex of their offspring5. In humans, the elite often form partnerships with high Da6 and have more sons than daughters7. It may be that during wartime women prefer to marry older men with high resources and this leads to an increase in Da. We do not know how the sex of first-borns is adjusted in relation to Da. Women could influence the motility of sperm bearing either X or Y chromosomes or they may invest differentially in males and females in utero leading to higher miscarriage rates of one or the other sex.