1. Introduction

Life history theory predicts that individuals optimize the allocation of their available resources to maximize fitness. In doing so, they experience a tradeoff between energy allocation for growth and survival versus reproduction as well as that between the quantity and quality of offspring (Mace, 2000, Mace, 2007). These two tradeoffs are unveiled by an association between increasing offspring number and reduced parental and offspring survival (Penn & Smith, 2007). Such an association has been demonstrated, for instance, in preindustrial societies in which high-fertility mothers experienced higher infant mortality (Mace, 2007, Strassmann & Gillespie, 2002). As a consequence, reducing offspring count together with investing more in each individual child may be a means to increase individual fitness (Kaplan et al., 1995, Penn & Smith, 2007); this would hold true particularly for women because of their high reproductive investment during gestation and lactation.

Life history theory further predicts that resource availability influences resource allocation. Accordingly, in light of fitness maximization, women are expected to choose mates of high socioeconomic status: they offer greater access to resources as compared with those of lower socioeconomic background (Buss, 1999). Consequently, men of higher socioeconomic status should have higher access to reproductive women, who, in turn, should adjust their reproductive decisions accordingly. The result is a positive association between male socioeconomic status and reproductive success.

Such a positive association among male status, resources, and reproduction has been demonstrated in a wide range of preindustrial societies (Borgerhoff-Mulder, 1988, Chagnon, 1988, Cronk, 1989, Irons, 1979, Mealy, 1985, Voland, 1990). In industrial societies since the end of the 19th century, however, a null or even a negative relationship has emerged (Wrong, 1980). This challenges evolutionary explanations of human behavior (Borgerhoff-Mulder, 1998). What might explain this null or negative relationship? A study on modern French Canadians (Pérusse, 1993), for instance, may have been biased toward individuals of higher income and education. Moreover, the used composite measure of socioeconomic status may have led to ambiguous results. Straightforward and separated analyses of single factors could potentially lead to more consistent findings. Only when particular society subsets are analyzed (Fieder et al., 2005, Weeden et al., 2006) might a combined indicator of socioeconomic status be reasonable: within any society subset, income, formal position, and education are more clearly interrelated than they are within whole societies. The study by Vining (1986) on social versus reproductive success, which also reported on an overall negative relationship between fertility and status, has several limitations. It compared the reproductive output of men with that of women, the study participants were still in their reproductive years, and only indirect measures were used as a proxy for status. The study by Kanazawa (2003), based on the General Social Survey from the late 1980s to the mid 1990s, also reported on a negative association between male status and reproduction but failed to consider that highly educated individuals may delay their reproduction. Finally, the offspring count of men is very rarely sampled in censuses and other surveys. Thus, many of the assumptions made on how socioeconomic status influences reproductive output rely solely on data from women.

Controlling for confounding variables, on the other hand, may yield results that meet evolutionary expectations even in modern societies. Kaplan and Lancaster (2000), for instance, reported on a positive relationship between income and offspring count in modern men. A similar positive association—among men but not women—was also found by Hopcroft (2006), who analyzed General Social Survey data as well. By analyzing a subset of society, Fieder et al. (2005) demonstrated a positive association between formal status and reproduction in men as well as a negative association in women. These findings support the importance of society subsets (Mace, 2000b) and homogeneous data sets (Voland, 2007) when investigating evolutionary predictions in human behavior. Accordingly, even Vining (1986), who reported on an overall negative relationship between male status and reproductive output, described a positive association within particular cohorts of his sample.

Overall, these studies on the association between status and reproductive output underline the following key issues: (1) sex differences; (2) the type of measure of socioeconomic status; (3) the reproductive age of the study population; and (4) the homogeneity of the data set. Further potentially important issues concern (5) the effects of childlessness and marital status: both parameters may determine fertility patterns (Fieder et al., 2005, Forsberg & Tullberg, 1995, Oppenheimer, 2003, Weeden et al., 2006) but have not been sufficiently considered in previous studies on the association between status and reproduction.

This study therefore investigated separately for postreproductive men and women (1) the association among income, education, and reproductive output (both including and excluding childless individuals) and (2) the association among income, education, marital status, and childlessness.

2. Methods

We used a representative data set of 7000 Swedish men and 7000 Swedish women from the Total Population Register of the year 2000 obtained from Statistics Sweden. The sample has been matched by Statistics Sweden with the Multigeneration Register, Register of Population Changes, Register of Income and Wealth, and Register of Education. The sample contains only those individuals born between 1945 and 1955 and thus aged between 45 and 55 years. This yields data about lifetime reproductive success because, in this sample, more than 99.7% of women and more than 96.5% of men have finished reproduction by age 45 years. The main advantage of this data set is that it is one of the very rare data sets that include highly accurate data about the offspring count of men and women and that cover society as a whole. The created file has been depersonalized and contains the following variables: sex; age; number of biological children born up to 2003; individual-level income in 2000 (classified by Statistics Sweden into four income categories: INC1, 0–156,002 Skr; INC2, 156,003–211,444 Skr; INC3, 211,445–272,471 Skr; and INC4, ≥272,472 Skr); highest education achieved by 2000 (classified into three categories: EDUC1, primary and lower secondary education; EDUC2, upper secondary education; and EDUC3, tertiary education); and marital status (never married vs. married at some point).

We used the SPSS 12.1 statistical package for Windows (SPSS, Chicago, IL). We performed each analysis separately for men and women. Offspring count was not normally distributed. We therefore performed a square root transformation of offspring count before the analyses. We analyzed the relationship among income, education, and offspring count including and excluding childless individuals with the general linear model (calculating main statistics and repeated contrasts). Offspring count was used as the dependent variable; income and education were used as fixed factors; and age was used as a covariate. The association among income, education, marital status, age, and childlessness—as well as income, education, age, and marital status—was analyzed using logistic regression.

3. Results

We found that in the male sample including childless individuals, men of higher income categories have, on average, more children as compared with those of lower income categories (Fig. 1A). The average offspring count increased with increasing income category and was, in each income category, significantly lower than that in the next higher category (Table 1, contrasts). In the female sample including childless individuals, an inverse pattern emerged. Wealthier individuals have, on average, fewer children as compared with those of lower income categories, with the average offspring count decreasing as income increased (Fig. 1A). The overall differences in average offspring count among income categories were less pronounced in the female sample as compared with the male sample. In addition, in the female sample, contrasts between adjacent income categories were nonsignificant, except for significantly more offspring among women of Income Category 2 as compared with those of Income Category 3 (Table 2).

Fig. 1. Offspring count (mean±SE) in men (solid circles) and that in women (open triangles) of different income categories: (A) including childless individuals and (C) excluding childless individuals. Offspring count (mean±SE) in men (solid circles) and that in women (open triangles) with different educational levels attained: (B) including childless individuals and (D) excluding childless individuals. Percentage of childless individuals: among men (solid bars) and women (open bars) of different income categories (E) and different educational levels attained (F). Sample sizes: for Panels (A) and (E), men=1131/1176/2021/2648 and women=1990/2347/1669/987; for Panels (B) and (F), men=1894/3064/2018 and women=1422/3317/2254; for Panel (C), men=753/912/1646/2313 and women=1726/2088/1469/839; and for Panel (D), men=1474/2470/1680 and women=1250/2948/1924.

Table 1. General linear model (main statistics and repeated contrasts) using offspring count as the dependent variable, income and education as fixed factors, and age as a covariate for men including childless individuals

	Main statistics	Corrected model	Age	Income	Education	Income×Education
	Adjusted R2	0.027
	df	12/6975	1	3	2	6
	F	17.235	0.495	43.783	0.621	2.031
	p	<.001	.482	<.001	.538	.058
		Income			Education
	Repeated contrasts	INC1 versus INC2	INC2 versus INC3	INC3 versus INC4	EDUC1 versus EDUC2	EDUC2 versus EDUC3
	Contrast estimate	−0.117	−0.068	−0.098	−0.016	−0.011
	95% confidence interval	−0.177 to −0.057	−0.121 to −0.016	−0.140 to −0.057	−0.054 to 0.022	−0.057 to 0.035
	p	<.001	.011	<.001	.414	.648

Table 2. General linear model (main statistics and repeated contrasts) using offspring count as the dependent variable, income and education as fixed factors, and age as a covariate for women including childless individuals

	Main statistics	Corrected model	Age	Income	Education	Income×Education
	Adjusted R2	0.002
	df	12/6992	1	3	2	6
	F	2.357	0.035	2.812	0.860	0.975
	p	.005	.851	.038	.423	.441
		Income			Education
	Repeated contrasts	INC1 versus INC2	INC2 versus INC3	INC3 versus INC4	EDUC1 versus EDUC2	EDUC2 versus EDUC3
	Contrast estimate	−0.018	0.041	0.031	−0.009	0.023
	95% confidence interval	−0.055 to 0.02	0.001 to 0.082	−0.028 to 0.089	−0.055 to 0.037	−0.011 to 0.057
	p	.363	.042	.301	.689	.190

The highest educational level attained and average offspring count were also positively associated in men including childless individuals (Fig. 1B), although contrasts between adjacent education categories were nonsignificant (Table 1). In the female sample including childless individuals, however, moderately educated women have, on average, the most children and the best educated women have, on average, the least children (Fig. 1B). Again, contrasts between adjacent education categories were nonsignificant (Table 2). When examining the effects of income and education on offspring count and correcting for age, only income was significantly associated with offspring count in the male and female samples including childless individuals. The interaction between income and education in the male sample was marginally significant (Tables 1 and 2).

After excluding childless individuals, the overall variance of average offspring count among different income categories and educational levels decreased in men and women (cf. Fig. 1A and C as well as Fig. 1B and D). Men with the lowest income have, on average, more children as compared with wealthier men if childless individuals were excluded, with the fewest offspring in the second highest income (Fig. 1C) and nonsignificant contrasts between adjacent income categories (Table 3). In the female sample, a similar pattern emerged regardless of whether childless individuals were included or excluded, with the average offspring count decreasing with wealth (cf. Fig. 1A and C). After excluding childless individuals, however, contrasts between adjacent income categories were significant (except between Categories 3 and 4; Table 4).

Table 3. General linear model (main statistics and repeated contrasts) using offspring count as the dependent variable, income and education as fixed factors, and age as a covariate for men excluding childless individuals

	Main statistics	Corrected model	Age	Income	Education	Income×Education
	Adjusted R2	0.001
	df	12/5623	1	3	2	6
	F	1.670	4.604	0.717	1.091	0.451
	p	.067	.032	.542	.336	.845
		Income			Education
	Repeated contrasts	INC1 versus INC2	INC2 versus INC3	INC3 versus INC4	EDUC1 versus EDUC2	EDUC2 versus EDUC3
	Contrast estimate	0.015	0.002	−0.013	0.009	−0.019
	95% confidence interval	−0.019 to 0.05	−0.027 to 0.031	−0.035 to 0.009	−0.012 to 0.031	−0.044 to 0.007
	p	.388	.893	.245	.394	.157

Table 4. General linear model (main statistics and repeated contrasts) using offspring count as the dependent variable, income and education as fixed factors, and age as a covariate for women excluding childless individuals

	Main statistics	Corrected model	Age	Income	Education	Income×Education
	Adjusted R2	0.01
	df	12/6121	1	3	2	6
	F	6.223	7.770	17.726	8.042	2.836
	p	<.001	.005	<.001	<.001	.009
		Income			Education
	Repeated contrasts	INC1 versus INC2	INC2 versus INC3	INC3 versus INC4	EDUC1 versus EDUC2	EDUC2 versus EDUC3
	Contrast estimate	0.029	0.045	0.017	−0.02	−0.029
	95% confidence interval	0.007 to 0.051	0.023 to 0.068	−0.016 to 0.05	−0.046 to 0.006	−0.049 to −0.01
	p	.009	<.001	.302	.126	.003

The highest educational level attained and average offspring count were not significantly associated in men excluding childless individuals (Fig. 1D; Table 3), whereas in women of this category, the best educated have significantly more children as compared with the moderately educated (Fig. 1D; Table 4). In men, only age was significantly associated with offspring count when examining the effects of income and education on offspring count, corrected for age and excluding childless individuals (Table 3). In women, however, each parameter as well as the interaction between income and education were significantly associated with offspring number (Table 4).

In the male sample, the proportion of childless individuals decreased with wealth and educational level (Fig. 1E and F). In the female sample, the lowest percentage of childless individuals was found within the second lowest income category and medium educational level; the highest percentage, within the highest income category and educational level (Fig. 1E and F).

Marital status was also associated with income and education. In the male and female samples including childless individuals, the percentage of individuals married at some point increased with wealth and educational level (Table 5). A logistic regression using marital status (encoded as 1=married at some point or 0=never married) as the dependent variable and income, education, and age as independent variables showed that income and education were significantly positively associated with marital status in men and women. Age was significantly negatively associated with marital status in women (Table 6). We claim that the age effect in the female sample could be attributed to a cohort phenomenon of changing martial patterns.

Table 5. Percentage of individuals who were married at some point among men and women of different income categories and educational levels attained, including childless individuals

		Men married at some point	Women married at some point
	INC1	1069 (40.1)	1866 (48.4)
	INC2	1110 (44.9)	2215 (50.7)
	INC3	1919 (54.8)	1591 (56.3)
	INC4	2524 (66.8)	939 (62.2)
	EDUC1	1782 (48.1)	1343 (39.2)
	EDUC2	2893 (53.4)	3117 (50.9)
	EDUC3	1923 (67.0)	2144 (64.7)

Data are expressed as n (%).

Table 6. Logistic regression using marital status as the dependent variable and income, education, and age as independent variables for men and women including childless individuals

		n	Nagelkerke's R2	Income	Education	Age
	Men	6598	0.06	B=0.313	B=0.259	B=0.000
				p<.001	p<.001	p=.973
	Women	6604	0.088	B=0.076	B=0.454	B=−0.12
				p=.005	p<.001	p<.001

Marital status and childlessness were associated. In men, 62.5% of reproducing individuals but only 30% of childless individuals were married at some point. In women, the difference was less pronounced: 55.0% of reproducing individuals and 40.3% of childless individuals were married at some point. A logistic regression revealed that income and marital status were significantly negatively associated with childlessness (encoded as 1=childless or 0=reproducing) in the male sample. In the female sample, education was significantly positively but age and marital status were significantly negatively associated with childlessness (Table 7).

Table 7. Logistic regression using childlessness as the dependent variable and income, education, marital status, and age as independent variables for men and women including childless individuals

		n	Nagelkerke's R2	Income	Education	Martial status	Age
	Men	6598	0.135	B=−0.345	B=0.070	B=−1.237	B=−0.017
				p<.001	p=.126	p<.001	p=.091
	Women	6604	0.026	B=0.004	B=0.208	B=−0.689	B=−0.037
				p=.916	p<.001	p<.001	p=.001

4. Discussion

We show that the association between socioeconomic status—as indicated by income and education—and offspring count differs markedly between the sexes. Offspring number increased with wealth and educational level in men, whereas an inverse pattern emerged in women: fewer offspring as income and education increased. We attribute the positive association in the male sample mainly to the association between income/education and childlessness because the proportion of childless men increased as income and educational level fell. Accordingly, in men, the relationship between income/education and average offspring count differed markedly depending on whether childless individuals were included or excluded. Only if they were included in the analysis did a positive association emerge; when they were excluded, no significant association between income/education and offspring count was found.

We assume that female choice is a major reason for the reversed association when including and excluding childless men. To maximize fitness, women are expected to prefer men of high socioeconomic status as mates because they can offer more resources, protection, and earning potential as compared with those of lower status (Buss, 1999, Buss & Barnes, 1986). Accordingly, the latter group should have lower and less regular access to reproductive females and, consequently, a higher chance to remain childless. This view is supported by our findings of an increasing proportion of men who were married at some point as wealth and education increased as well as a higher proportion of individuals who never married among childless versus reproducing men. These data are consistent with marriage patterns reported thus far: Nakosteen and Zimmer (1997), for instance, showed that a higher percentage of lower-income men remained unmarried. In addition, the percentage of childless individuals and that of individuals without a partner are higher among unskilled male laborers than in men working in other occupations (Forsberg & Tullberg, 1995). Finally, Burgess, Propper, and Aassve (2003) reported that high earnings increased the probability of marriage and decreased the probability of divorce in young men.

In reproducing men (i.e., excluding childless individuals), we found the most offspring among the poorest and least educated individuals—a pattern similar to that in the female sample. Thus, it is reasonable to assume that this female association underlies the pattern found in men after excluding childless individuals.

In contrast to the male sample, including and excluding childless individuals in the female sample did not substantially change the relationship between income and offspring. Wealth remained to be associated with fewer children regardless of whether childless individuals were included or excluded. However, greater education was correlated with slightly more children only after childless women were excluded. On the one hand, these data point to the well-known difficulties of balancing motherhood and a career in modern societies as well as to the risk of missing the fertility window when reproduction is postponed during education (Goldstein & Kenney, 2001). From an evolutionary perspective, on the other hand, the negative association between income/education and offspring count in women may be attributed to a life history strategy. This predicts a tradeoff between offspring quality and quantity, yielding an “optimum number of offspring” (Mace, 2007). Accordingly, women increase their individual fitness by having fewer children yet investing more in each individual child via increased education and personal income. Note, however, that empirical evidence for such a quality–quantity tradeoff in fitness outcomes over generations is not clear in modern societies (Kaplan et al., 1995).

In addition, women may be victims of their evolutionary adaptations to prefer men of high socioeconomic status (Buss, 1999). Particularly, successful professional women are known to be very interested in the high income of a potential husband (Wiederman, 1993). This may complicate finding an adequate spouse because few high-status men are available. In fact, high income in women has been reported to decrease their probability of marriage (Burgess et al., 2003). Lack of an adequate partner, in turn, may limit a woman's reproductive output, although marriage is apparently becoming less important for reproductive decisions, particularly in women. Our study found the highest percentage of unmarried individuals among the poorest and least educated women, but these women had, on average, the most offspring.

This study has limitations as it lacked data on the spouses' income and education level. However, reproductive decisions involve two people—or at least the resources available to two people.