Human Nature

pp 1–16 | Cite as

The Null Relation between Father Absence and Earlier Menarche

Article

Abstract

Researchers have claimed that the absence of a biological father accelerates the daughter’s menarche. This claim was assessed by employing a large and nationally representative sample of Indonesian women. We analyzed 11,138 ever-married women aged 15+ in the Indonesian Family Life Survey 2015. We regressed age at menarche on the interaction of father absence (vs. presence) and mother absence (vs. presence) at age 12 with or without childhood covariates. For robustness checks, we performed a power analysis, re-ran the same specification for various subgroups, and varied the independent variable of interest. All results produced a null relation between father absence and age at menarche. The power analysis suggests that a false negative was unlikely. Our review of the literature indicates that the claim of the relation between father absence and earlier menarche was based on weak statistical foundations. Other studies with higher-quality datasets tended to find no relation, and our results replicated this tendency. Therefore, the influence of father absence does not appear to be universal.

Keywords

Father absence Menarche Evolution Indonesia 

Menarche is a landmark in female sexual development, so it has important implications for evolution (Ellis 2004). Life history theory (Stearns 1989) implies that early menarche may lengthen the reproductive lifespan and consequently increase the number of offspring, but it may hinder growth, lead to weaker development of the reproductive organs, and consequently decrease the number of offspring. Menarche also has implications for health, albeit they are open to debate for some diseases (e.g., Sohn 2016f). Karapanou and Papadimitriou (2010) summarized that early menarche is related to abdominal-type obesity, insulin resistance, glucose intolerance, cardiovascular risk, coronary heart disease, increased bone mineral density, and increased cancer mortality; late menarche is related to osteoporosis, adolescent depression, and social anxiety symptoms.

Because of the importance of menarche in evolution and health, researchers have searched for the determinants of menarche. One putative determinant is the absence of a biological father. An oft-cited theoretical argument based on evolution is that if a girl’s experiences in and around her family of origin lead her to perceive others as untrustworthy, relationships as opportunistic or self-serving, and resources as scarce or unpredictable, she will choose “quantity” over “quality” in her reproductive strategies, thereby experiencing accelerated maturation and a longer reproductive period (Belsky et al. 1991)—father absence is a conspicuous experience for such development.

A number of studies have presented empirical evidence to support this theory. Their evidence, however, is based on questionable foundations. The sample was not randomly drawn, the sample size was small, or the sampling area was small. The first limitation is likely to cause selection bias, the second limitation poses the threat of a false negative, and the third limitation makes it difficult to generalize the results. Addressing one limitation does not automatically address another limitation. For example, increasing the sample size by itself does not address selection bias unless it is accompanied by random sampling. Alternatively, covering a large area by itself does not address the threat of a false negative unless accompanied by a large number of observations. These limitations should be addressed simultaneously to produce convincing results. In addition, covariates were often not controlled for, and when they were controlled for, they were linearized or their number was usually small. Therefore, it is uncertain whether the results were robust to nonlinear assumptions or whether the relation was confounded. An early example of such problematic research is provided by Jones et al. (1972), and some combinations of the limitations continued to plague subsequent studies, such as Surbey (1990), Moffitt et al. (1992), Wierson et al. (1993), Hoier (2003), Bogaert (2005), Chisholm et al. (2005), Vigil and Geary (2006), Matchock and Susman (2006), and other studies discussed below. Webster et al. (2014) meta-analyzed the literature and sided with this group, but an analysis with questionable inputs produces only questionable outputs.

Among these limitations, we choose one limitation, nonrandom sampling, and illustrate that even this single limitation can easily compromise the results. For example, Alvergne et al. (2008) collected information from about 700 female students at the University of Montpellier in France and employed a general linear model to determine whether age at menarche differed by family composition. Women who lived without their fathers at ages 0–5 experienced menarche almost one year earlier than women with both biological parents; the difference decreased as the timing of father absence considered became later. Common sense, however, suggests that father-absent women constitute a special group. Father-absent households are typically poorer than intact households (McLanahan 1985), and the female students from such households had to overcome many disadvantages arising from relative poverty to be enrolled in the university. A more accurate analysis would be comparing all women before some of them were enrolled in the university.

Culpin et al. (2014) adopted a promising strategy by following a large number of UK girls for a long time and adjusting for a set of covariates; they found that father absence before five years of age was related to earlier menarche with a beta coefficient of −0.237. Attrition, however, was severe and strongly related to socioeconomic disadvantage. They thought that imputation could address this concern, but imputation is ineffective when attrition is not random. When low socioeconomic status (SES) delays menarche (as is explained below) and low SES respondents tended to attrite (as they admitted), this biased attrition can explain their finding.

Even if the results of Alvergne et al. (2008) and Culpin et al. (2014) are accepted as they are, almost all studies have focused on developed Western countries. It is thus difficult to know whether the relation is the same in a country that is substantially different from developed Western countries. A comparative study helps one determine which mechanism is more plausible if the relation does exist. We attempted to address all of the limitations by analyzing a nationally representative sample of more than ten thousands Indonesian women in nonlinear multivariate settings. We considered Indonesia because it differs greatly from developed Western countries. For example, its main religion is Islam (vs. Christianity), its climate is tropical (vs. temperate), and its income level is low (vs. high). We examined the relation by using several methods, including power analysis and ordinary least squares (OLS), and checked the robustness of the results.

Methods

Our main dataset was the Indonesian Family Life Survey (IFLS).1 The IFLS began to collect data on more than 22,000 individuals in 7224 households from 13 provinces in 1993 (IFLS1); the population of these provinces represented 83% of the Indonesian population in 1993. The IFLS sampling scheme stratified by provinces and then randomly selected 321 enumeration areas in the provinces and then households within each of the enumeration areas. A representative member of each of the households provided household-level demographic and economic information, and interviewers randomly selected several household members and obtained detailed individual information. Five follow-ups ensued in 1997 (IFLS2), 1998 (IFLS2+), 2000 (IFLS3), 2007 (IFLS4), and 2014 (IFLS5). We analyzed IFLS5 because this follow-up is most up-do-date and contains a greater number of observations than the other follow-ups. More importantly, only this follow-up provides information on the absence or presence of a biological father at age 12.

The dependent variable in OLS was age at menarche. Age at menarche was recalled in whole years by ever-married women aged 15–49 and women who completed the same module in IFLS4, irrespective of age. Some women provided unrealistic ages at menarche, so we restricted the range of ages at menarche to 8–19. Considering the maximum of this range, we set the minimum current age at 20. The independent variable of interest was whether the respondent lived with her biological father at age 12. If the respondent said no to the question, “When you were 12, did you live with your biological father?” we assigned a value of one, and zero otherwise. The mean age at menarche in this study was 13.6 years (Table 1), so the reference age should closely reflect the influence of father absence around menarche. Although mother absence was not our main interest, it would be interesting to compare its relation with age at menarche to that of father absence. By interacting these two dummy variables, we could observe the influence of four combinations of parental presence or absence on age at menarche, with the reference group being both parents present. However, many researchers simply considered father absence (vs. presence) whether the mother was present or absent (for a recent example, Culpin et al. 2014). Hence, for comparative purposes, we began and ended analysis with this dichotomous case.
Table 1

Descriptive statistics (N = 11,138)

A. Continuous Variables

 

Mean (SD)

Age at menarche

 

13.6 (1.7)

Current age

 

36.5 (9.7)

B. Discrete Variables

 

%

Presence/Absence

Father present and mother present

85.0

 

Father present and mother absent

2.2

 

Father absent and mother present

7.7

 

Father absent and mother absent

5.1

Breadwinner’s job

  
 

Self-employment alone

23.1

 

Self-employment with public or private-sector work

36.7

 

Government worker

10.8

 

Private worker

14.6

 

Others

14.8

N of older brothers

  
 

0

52.3

 

1

28.4

 

2

12.4

 

3

4.4

 

4 or more

2.5

N of older sisters

  
 

0

55.0

 

1

27.4

 

2

11.2

 

3

4.0

 

4 or more

2.5

N of younger brothers

  
 

0

47.8

 

1

31.7

 

2

13.9

 

3

4.7

 

4 or more

2.0

N of younger sisters

  
 

0

51.1

 

1

29.7

 

2

12.8

 

3

4.4

 

4 or more

2.1

Electricity

  
 

No

41.0

 

Yes

59.0

Water

  
 

Well or other

86.5

 

Piped water

13.5

Sewer

  
 

Own toilet without a septic tank, shared toilet, public toilet, or other

60.0

 

Own toilet with a septic tank

40.0

Books

  
 

None or few

82.2

 

Enough to fill at least one shelf

17.8

Health in childhood

  
 

Poor or fair

37.6

 

Good

41.4

 

Very good or excellent

21.1

Since evidence suggests that SES in childhood is an important determinant of age at menarche (e.g., Tanner 1962), we included the following covariates experienced at age 12: the breadwinner’s employment status; the number of older or younger brothers or sisters; whether the household utilized electricity (vs. no electricity); whether the main water source for drinking in the household was piped water (vs. a well or other); whether the majority of householders had their own toilet with a septic tank (vs. own toilet without a septic tank, shared toilet, public toilet, or other); and whether there were no or very few books in the household (vs. enough books to fill at least one shelf). We created five categories for the employment status: self-employment alone, self-employment with employees, government worker, private-sector worker, and other. We also created five categories for the number of siblings: zero, one, two, three, and four or more. The rest were dichotomous. We included self-reported health status in childhood with the following three categories: excellent or very good, good, and fair or poor. Finally, we included current age in dummy form (38 categories) to capture all time-invariant factors in each birth year—namely, age (or birth year) fixed effects. When we excluded observations with missing values, we were left with 11,138 observations.

Our OLS specification is as follows:
$$ {\mathrm{menarche}}_{ia}={\upbeta}_1\mathrm{FP}/{\mathrm{MA}}_{ia}+{\upbeta}_2\mathrm{FA}/{\mathrm{MP}}_{ia}+{\upbeta}_3\mathrm{FA}/{\mathrm{MA}}_{ia}+{X}_{ia}{\upbeta}_4+{\mathrm{age}}_a+{u}_{ia} $$
where menarcheia refers to the age at menarche of individual i at the current age a; FP/MA, father present and mother absent; FA/MP, father absent and mother present; FA/MA, father absent and mother absent; X, all covariates; age, age fixed effects; u, the random error; and β, the coefficients to be estimated. β2 is the coefficient of interest. Since all independent variables were nonlinear, our specification was robust and flexible.

Results

Before proceeding to the relation between father absence and age at menarche, we report descriptive statistics to illustrate the poor background of Indonesian women (for more, see Sohn 2014b, 2014c, 2015a, 2015e, 2016b, 2016c, 2016g, 2017a, 2017c). The mean age at menarche was 13.6 years, and the current mean age was 36.5 years. Of all the women in our sample, 85.0% lived with the two parents, but 7.7% experienced father absence and mother presence at age 12. Of the breadwinners, 59.8% were self-employed, and 14.8% were engaged in casual jobs (classified as “other”). Self-employment is not something to aspire to in Indonesia (Kwon and Sohn 2017; Sohn 2013, 2015a, 2015b, 2016h; Sohn and Kwon 2016). About half of our sample had no older or younger brothers or sisters, and it is uncommon to have three or more older or younger brothers or sisters. Therefore, most women grew up without many siblings. At age 12, 41.0% grew up in households without electricity, and only 13.5% with piped water for drinking. In addition, only 40.0% grew up in households that owned toilets with septic tanks, and 82.2% with no or few books at home. Given the poor environment, it is not surprising that 37.6% recalled that their health was poor or fair in childhood.

Figure 1 presents two overlapping histograms of age at menarche. For comparison and illustrative purposes, we compared father-absent women and father-present women, regardless of mother presence or absence. Filled, unbordered bars refer to father-present women, and empty, bordered bars to father-absent women. The two histograms are almost identical.
Fig. 1

Histogram of age at menarche by father presence overlain on that for father absence (n = 9757 for father-present women and 1436 for father-absent women)

We tested the null hypothesis that the two distributions were equal by using the Kolmogorov-Smirnov test and found that the test failed to reject the null hypothesis with a p value of 1.0. Since the distributions were the same, the mean age at menarche was the same at 13.64 for both groups. Not surprisingly, a t test failed to reject the null hypothesis that the means were equal, with a p value of 0.97; the alternative hypothesis was that the means were not equal. When we changed the alternative hypothesis to the one that father-absent women experienced menarche earlier, the test still failed to reject the null hypothesis.

We performed power analysis by taking into account the number of observations for each group (9712 and 1426), the SD of age at menarche for all observations at 1.67, the mean of age at menarche for the father-present group at 13.64, and the 99% significance level. The power was 0.32 for a difference of 0.1 years but jumped to 0.95 for a difference of 0.2 years; thereafter, the power reached one. We were thus able to detect a small difference with a very low probability of a false positive or negative.

Column 1 in Table 2 delivers the same message with the four family structures. The coefficient on father absence–mother presence was small at 0.034 and statistically nonsignificant. Column 2 confirms that the null relation was robust to controlling for covariates. For that matter, no family structure dummy was related to age at menarche.
Table 2

Relationship between Father Absence and Age at Menarche: OLS

 

1

2

Father present and mother present

reference

reference

Father present and mother absent

−0.194

(0.109)

−0.183

(0.108)

Father absent and mother present

0.034

(0.060)

0.027

(0.060)

Father absent and mother absent

0.101

(0.141)

0.179

(0.140)

Breadwinner’s Job: Self-employment alone

 

reference

Self-employment with workers

 

0.013

(0.042)

Government worker

 

0.002

(0.061)

Private worker

 

−0.117

(0.053)*

Others

 

−0.058

(0.052)

N of older brothers: 0

 

reference

1

 

0.014

(0.037)

2

 

−0.049

(0.051)

3

 

−0.056

(0.079)

4 or more

 

−0.120

(0.102)

N of older sisters: 0

 

reference

1

 

0.011

(0.037)

2

 

0.086

(0.052)

3

 

−0.035

(0.082)

4 or more

 

−0.040

(0.103)

N of younger brothers: 0

 

reference

1

 

0.071

(0.037)

2

 

0.150

(0.049)*

3

 

0.150

(0.077)

4 or more

 

0.393

(0.116)*

N of younger sisters: 0

 

reference

1

 

0.010

(0.037)

2

 

0.115

(0.050)*

3

 

0.178

(0.079)*

4 or more

 

0.360

(0.111)*

No electricity, well water, no septic or shared toilet, one shelf of books

 

reference

Used electricity

 

−0.229

(0.041)*

Piped water

 

−0.170

(0.048)*

Own toilet with a septic tank

 

−0.094

(0.037)*

None or few books

 

0.013

(0.043)

Health in childhood: poor or fair

 

reference

Good

 

−0.046

(0.035)

Very good or excellent

 

−0.097

(0.043)*

Constant

13.65

(0.02)*

13.80

(0.06)*

Age fixed effects

No

Yes

Adjusted R2

<0.000

0.029

Sample size = 11,138. Standard errors are in parentheses. * p < 0.05

On the other hand, variables of childhood SES exhibited the expected relations with age at menarche—namely, better SES with earlier menarche (for more, see Sohn 2014a, 2015c, 2016d, 2016e, 2016f, 2017b). Compared with a woman whose breadwinner at age 12 was self-employed alone, a woman whose breadwinner was a private-sector worker experienced menarche 0.12 years earlier. The number of older siblings was not related to age at menarche whereas having more younger brothers or sisters was generally related to later menarche. Compared with a woman who had no younger brother, a woman with four or more younger brothers experienced menarche 0.39 years later. The corresponding figure for younger sisters was 0.36 years. Compared with a women who grew up in a house without electricity, the counterpart with electricity experienced menarche 0.23 years earlier. The relation between better SES in childhood and earlier menarche was found regarding the water source for drinking and the condition of the toilet. Better health in childhood was also related to earlier menarche. The size of each relation was modest, but each relation was estimated after controlling for the rest. Since an affluent family typically possesses more than one amenity, multiple sources of higher SES probably induced the girls to experience menarche much earlier than each coefficient indicates.

Although we controlled for fixed effects of current age, there could have been some interaction effects between father absence and birth cohort. For example, economic development in Indonesia might have attenuated the influence of father absence, and older women could exhibit a greater negative relation. We thus divided the pooled sample into three age cohorts (20–29, 30–39, and 40 or above) and re-ran the specification with the same covariates in Column 2 of Table 2. We continued to estimate a small and statistically nonsignificant relation between father absence and age at menarche (Table A-1 in the ESM). It could be that extremely impoverished girls do not have extra energy to divert to earlier maturation; then, father absence would exert little influence on their maturation. We thus divided the sample into two subgroups by using some childhood covariates, such as using electricity vs. no electricity, and separately re-ran the specification for the two groups. In none of the cases did we find a statistically significant relation between father absence and age at menarche (Table A-2 in the ESM).

We believe that using the four family structures is a more accurate way to investigate the relation, but many studies often considered father presence or absence, regardless of mother presence or absence. We thus checked whether the null relation would change if we followed this method. When we re-ran the specification after replacing the four family structures with father absence (vs. presence) and mother absence (vs. presence), the coefficient on father absence continued to be small and statistically nonsignificant (Table A-3 in the ESM).

Discussion

Our finding of null relation in Indonesia challenges many studies that asserted that father absence accelerates menarche. In fact, our finding is not an exception in the literature. For example, Ellis et al. (1999) did not directly use age at menarche but included menarcheal status in a composite measure of pubertal timing in their study of 87 female students in a metropolitan public school district in the US. They found a correlation coefficient of −0.37 (statistically significant) between the daughter’s pubertal timing and her age when an unrelated father figure first came into her life, but a correlation coefficient of −0.13 (statistically nonsignificant) between her pubertal timing and her age when her father moved out. They thus concluded that pubertal timing is affected not by father absence but by stepfather presence. Graber et al. (1995) recruited 75 girls who were from white, well-educated, middle- to upper-middle-class families and attended private schools in a major northeastern urban area in the US. They reported a null correlation coefficient between father absence and age at menarche. Campbell and Udry (1995) followed girls born in a hospital in Oakland, California, at age 5 (N = 310), 9–11 (N = 456), and in late adolescence (N = 518, with a mean age of 17). They regressed age at menarche on father absence and covariates and found that the coefficient on father absence was statistically nonsignificant for age 5 or ages 9–11 but significant for late adolescence. This pattern is opposite to that of Alvergne et al. (2008) and implies a spurious relation of father absence to menarche; during late adolescence, most of the girls had already experienced menarche. Anderson (2015) examined three ethnic groups (black, white, and mixed; N = 944, 300, 926, respectively) in Metropolitan Cape Town, South Africa by employing a Cox proportional regression. The relation between “never lived with father” and age at menarche was not statistically significant, and its sign was contrasting among the three groups. A null relation was even found by Belsky, a strong supporter of the role of father absence in menarche (Belsky et al. 2007). We can go on with other similar studies, such as Romans et al. (2003), Jorm et al. (2004), Maestripieri et al. (2004), Boothroyd and Perrett (2006), and James-Todd et al. (2010), but that would be redundant. We cite these studies not to support their arguments; these studies are not free of the same limitations as the previous studies that supported the role of father absence in menarche. We only illustrate that crude data and methods can yield anything. That said, it is worth highlighting that we found the null relation by analyzing high-quality data with sophisticated methods.

In this context, it is of interest that as the other researchers’ samples covered a wider area, their sample size was larger, and the method was more sophisticated, the results were closer to ours. For example, Hulanicka (1999) cited a study that examined 30,003 girls of the Upper Silesia region in 1991. When postmenarcheal status was related to father absence and covariates in a logit model, the coefficient on father absence was not statistically significant. Instead, all covariates of SES were statistically significantly related to postmenarcheal status in the expected direction: higher SES was related to earlier menarche. Quinlan (2003) analyzed a US representative sample (the 1995 National Survey of Family Growth, N = 10,097) by employing Cox regression. They found no difference in age at menarche among women from two-parent, mother-only, and father-only households. A positive finding with high-quality data but with crude methods can reinforce our position. Doughty and Rodgers (2000) analyzed US nationally representative data (the National Longitudinal Survey of Youth) and found that the mean age at menarche was younger for girls who lived with their mothers only than those living with both parents (12.84 vs. 12.95). We believe that this simple comparison of means produced a spurious relation because, compared with white girls, black girls experience menarche earlier (Herman-Giddens 2006) and more frequently live with their mothers only (Quinlan 2003). We suspect that living with the mothers only captured the effect of being black. Note that Quinlan (2003) controlled for a dummy indicating being black when he estimated no relation between father absence and age at menarche.

Sheppard et al. (2014a) turned attention away from developed Western countries to Malaysia. Their data were representative of Peninsular Malaysia and found no relation between father absence and menarche however they defined father absence. This null relation could be a false negative because of their small sample size (N = 249), but our results support theirs with a large sample size. Note that Indonesia and Peninsular Malaysia are very similar regarding ethnicity, climate, and religion—the territorial demarcation between Malaysia and Indonesia is historically new, established by the Anglo-Dutch Treaty of 1824. Furthermore, Sheppard et al. (2014b) employed the Original Kinsey Survey collected from 1938 to 1963 in the US. Although the survey was not nationally representative, Alfred Kinsey and his team endeavored to amass a large enough sample to draw general conclusions about human sexual behavior. They included menarche to construct a composite measure of puberty and found no relation between father absence and age at puberty (N = 4624).

Randomized experimental data can be of high quality, and such data can overcome concerns stemming from a small sampling area. Since it is out of the question to randomly assign girls to father presence or absence, it is necessary to find natural conditions closer to a randomized experiment. Mendle et al. (2006) examined the female offspring of monozygotic and dizygotic female twins in Australia to determine whether stepfather presence was related to earlier menarche. Their sample of twins allowed them to control for genetic and shared environmental influences on age at menarche. As mentioned above, Ellis et al. (1999) argued that the influence of stepfather presence was stronger than that of father absence. Therefore, if this is true and Mendle et al. (2006) found no influence of stepfather presence, father absence is likely to have no influence on menarche—that is what they found.

Tither and Ellis (2008) presented an alternative setting. They distributed circulars advertising their study to about 65,000 mailboxes in urban areas in New Zealand (primarily Christchurch) and recruited 93 pairs of sisters from biologically intact families and 68 pairs of sisters from biologically disrupted families. The former was the control group, and the latter the treatment group. After the marital dissolution, the younger sister lived either primarily with her mother or in joint custody between her mother and father. Their idea was that since younger sisters were exposed longer to father absence in biologically disrupted families than their older sisters, the difference in age at menarche between younger and older sisters would be more pronounced in the treatment group. They conducted a 2-by-2 mixed ANOVA and found that in biologically intact families, younger sisters experienced menarche later than older sisters (12.66 vs. 12.52, statistically nonsignificant), but the opposite was the case in biologically disrupted families (12.34 vs. 12.65, statistically nonsignificant). Tither and Ellis (2008), however, failed to achieve what they attempted to do—namely, show the causal effect of father absence on age at menarche. To estimate the causal effect, the control group had to be the same in every way but father absence. Father absence, however, is nonrandom, and therefore, parents in the two groups were different in many unobservable aspects. The implausible control group opens the possibility for some third factor to cause the earlier menarche of younger sisters in biologically disrupted families.

One can criticize our study, too. The subjective and retrospective nature of age at menarche may introduce measurement error and bias. However, a long list of studies stretching from Livson and McNeill (1962) to Dorn et al. (2013) has demonstrated that reported age at menarche is quite accurate. To directly support our position, we extracted 7896 women who reported their ages at menarche (in the range of 8–19) in both IFLS4 and IFLS5 and found the mean and median differences in the two reported ages at menarche were merely 0.13 and 0 years, respectively. Furthermore, age at menarche in the IFLS behaved as expected; Table 2 has already demonstrated this. In addition, Sohn (2014a) found that age at menarche and height were unrelated, which makes sense in that the relation was negative in small-scale agrarian societies but positive in industrialized societies (McIntyre and Kacerosky 2011); Indonesia was in between these two groups in economic development. Sohn (2015c) reported that age at menarche decreased over time, which reflects the improvement of the living standard in Indonesia. Going beyond the general trend, Sohn (2016d) found that the trend in age at menarche more accurately reflected the vicissitudes of the national economy in childhood than did height. Therefore, age at menarche in this study is likely to be accurate.

There is a possibility of a false negative, but our results by power analysis discount this possibility. Selection might cause bias in age at menarche since our sample consisted of only ever-married women; never-married women were excluded. Sohn (2016e), however, analyzed another Indonesian survey (the Young Adult Productive Health Survey),2 which comprised unmarried women aged 15–24 years. When he compared this group with the IFLS group, the mean age at menarche for each birth year was similar. We thus believe that this type of bias is negligible. An indirect way to assess the degree of selection bias is to exploit the fact that almost all women marry by age 30 in Indonesia. If the results are similar whether we included or excluded ever-married women under 30, selection bias is likely to be small. Sohn (2015d, 2016a) applied this idea in Indonesian marriage studies and demonstrated that selection bias in the IFLS was negligible. We followed the same logic and confirmed the null relation (Table A-1 in the ESM).

We did not know whether the respondent lived with a stepfather at age 12, but this omission only reinforces our argument. Ellis and Garber (2000:495) asserted that “The younger the daughter at the time of the father figure’s arrival, the earlier her pubertal timing.” Recall that Ellis et al. (1999) took the same position. We have little confidence in their assertion because of the limitations mentioned in the introduction, but suppose that this assertion were true for the sake of argument. Then, we reason that some father-absent women lived with stepfathers and the coefficient on father absence in our specification captured the influence of stepfather presence, which would push the coefficient down to a negative value, not up to zero. Since our main claim is a zero coefficient, the omission of information regarding stepfather presence works against our argument. Furthermore, from this null relation, we can assess the influence of stepfather presence on menarche. The null relation implies that the influence of stepfather presence is small. Alternatively, only a small number of women lived with stepfathers at age 12. Even in this case, however, the influence of stepfather presence was not strong enough to compensate for the small number.

Another criticism would be that the timing of father absence was too late; if the timing had been earlier, we might have found a relation between father absence and earlier menarche. A related limitation is that father absence might have occurred after menarche for some women. However, only 6% of women experienced menarche before age 12 (Fig. 1). Therefore, the late timing was unlikely to severely bias our null relation, and almost all women experienced menarche after father absence. Quinlan (2003) reinforced this claim. As mentioned above, he analyzed a large nationally representative sample and did not find a relation between father absence and earlier menarche. However, when he limited the analysis to women whose parents separated when the daughters were between zero and six years of age, women who lived with their mother only or father only tended to experience menarche earlier than women from intact families. That is, if there is any influence of family composition on age at menarche, it is driven by the absence of one parent, not specifically of the father. What is important is not father absence but the number of parents or its correlates. Therefore, our argument remains valid that father absence per se is unrelated to age at menarche.

It is possible that our null relation is limited to Indonesia. This, by definition, reduces the generalizability of the putative influence of father absence on age at menarche. Furthermore, the discussion of the literature suggests that when the sample was representative of a greater area and the sample size was greater, the relation approached zero. One can find as many studies with a statistically nonsignificant relation as those with a significant one. Therefore, even if our null relation is limited to Indonesia, we can at least say that the influence of father absence does not appear to be universal. This argument, however, leaves the possibility open that the influence manifests itself when the environment is appropriate.

Given this tenuous evidence of the relation, it would be too hasty to attempt to explain the mechanism after assuming that the relation exists. That is what Kanazawa (2001) did by analyzing not individual but societal data. According to him, father absence is indicative of the degree of polygyny in society, and polygyny creates a shortage of women in reproductive age. Therefore, early puberty is advantageous because it allows girls to reproduce earlier and consequently to produce more offspring. To support this hypothesis, he employed 73 observations and regressed the mean age at menarche on a measure of polygyny (drawn from the Encyclopedia of World Cultures) and a few covariates. The coefficient on polygyny was −0.286. Interesting as it may be, we suspect that he committed an ecological fallacy. Many factors are related to polygyny. For example, Sohn (2016e) reported that the mean age at menarche was lower in the tropics than in temperate zones ceteris paribus. Since polygyny is more prevalent in the tropics than in temperate zones (see the Encyclopedia; Kanazawa 2001), polygyny might have captured the influence of tropical climates on menarche.

Our null relation does not deny the importance of family composition and its derivatives in determining age at menarche. For example, it could be that lower parental investment may cause earlier menarche (Belsky et al. 1991). We do, however, discount the possibility that father absence accelerates menarche via strictly biological mechanisms, such as pheromones. Ellis et al. (1999) considered this possibility by arguing that pubertal timing is accelerated by exposure to pheromones produced by unrelated adult male conspecifics and decelerated by related ones. Matchock and Susman (2006) supported this argument by tracing back even to evolutionary times. According to them, these phenomena evolved for promoting gene survival and preventing inbreeding. But pheromones are released whether in Indonesia or elsewhere. Our null relation implies that this mechanism is absent or negligible.

We can similarly challenge the argument of genetic transmission from parents to their female offspring (see citations in Tither and Ellis 2008). According to this view, mothers who matured earlier tend to have sex earlier, to marry earlier, to have more marital conflicts, and to divorce or separate more often. Because menarche is heritable (Treloar and Martin 1990), their daughters repeat this cycle. Alternatively, fathers may carry genes that are associated with aggression, impulsivity, sexual promiscuity, and eventually, marital conflict and dissolution. If these genes are associated with earlier menarche, their daughters live without their fathers and experience menarche earlier (Comings et al. 2002). Parents, however, transmit their genes to their offspring in Indonesia and elsewhere; our null relation implies that this mechanism is absent or negligible. In addition, Jorm et al. (2004) considered the gene specified by Comings et al. (2002) and rejected their claim.

Our null relation and the inconsistent results in the literature suggest that family composition plays a minor role in determining age at menarche as long as family composition is unrelated to SES. The downward trend in age at menarche, which has been consistently found across the world (Sohn 2017b), suggests that the major force is an improvement in the standard of living via more nutrition, better healthcare, less disease burden, and less workload. This argument is consistent with life history theory because when the living standard improves, the burden of trade-off between growth, maintenance, and reproduction decreases. Earlier reproduction does not considerably compromise growth and maintenance while increasing the number of offspring (Sohn 2014a).

Belsky et al. (1991:659) relied on the same theory but interpreted the downward trend differently: “we must acknowledge the possibility that industrialized society may in certain respects be more stressful than its pre-industrialized counterpart. Consider in this regard not only that people live in closer proximity to nonrelatives than they ever did before, but also that the intuitive sense that parent-child relations and family life have become more harmonious in recent centuries remains unconfirmed.” This interpretation points to stress, pheromones from nonrelatives, and low parental investment as the main driving forces. History tells that life is considerably better now than before, whether it is measured in monetary or nonmonetary terms (Floud et al. 2011). The improvement in monetary terms is too obvious to discuss. The improvement in nonmonetary terms can be easily found in biological indicators such as height, weight, body mass, morbidity, and mortality. Given that stress increases morbidity and mortality (Sapolsky 2004), it is implausible to argue for greater stress now than before. We already mentioned the unlikely role of pheromones, whether they come from relatives or nonrelatives. The argument of low parental investment is too vague to be useful when the mechanisms are unspecified. That said, the great improvement in the biological standard of living suggests that parental investment is greater now than before. Therefore, their interpretation appears to be false.

Given this discussion, future research will be more fruitful when attention is turned to other determinants of age at menarche, such as SES. If one clings to the belief that father absence accelerates menarche at least in developed Western countries, an interesting question would be why there is discrepancy in the effect between developed and developing countries. If one insists that the effect is ubiquitous, the analysis should be repeated with data from developing countries and the hypothesis tested using our study as a benchmark.

Footnotes

Notes

Acknowledgements

I am grateful to the two anonymous reviewers for helpful comments and suggestions.

Supplementary material

12110_2017_9299_MOESM1_ESM.pdf (54 kb)
ESM 1(PDF 57 kb)

References

  1. Alvergne, A., Faurie, C., & Raymond, M. (2008). Developmental plasticity of human reproductive development: Effects of early family environment in modern-day France. Physiology & Behavior, 95(5), 625–632.CrossRefGoogle Scholar
  2. Anderson, K. G. (2015). Father absence, childhood stress, and reproductive maturation in South Africa. Human Nature, 26(4), 401–425.CrossRefGoogle Scholar
  3. Belsky, J., Steinberg, L., & Draper, P. (1991). Childhood experience, interpersonal development, and reproductive strategy: An evolutionary theory of socialization. Child Development, 62(4), 647–670.CrossRefGoogle Scholar
  4. Belsky, J., Steinberg, L. D., Houts, R. M., Friedman, S. L., DeHart, G., Cauffman, E., et al. (2007). Family rearing antecedents of pubertal timing. Child Development, 78(4), 1302–1321.CrossRefGoogle Scholar
  5. Bogaert, A. F. (2005). Age at puberty and father absence in a national probability sample. Journal of Adolescence, 28(4), 541–546.CrossRefGoogle Scholar
  6. Boothroyd, L. G., & Perrett, D. I. (2006). Facial and bodily correlates of family background. Proceedings of the Royal Society of London B: Biological Sciences, 273(1599), 2375–2380.CrossRefGoogle Scholar
  7. Campbell, B. C., & Udry, J. R. (1995). Stress and age at menarche of mothers and daughters. Journal of Biosocial Science, 27(2), 127–134.CrossRefGoogle Scholar
  8. Chisholm, J. S., Quinlivan, J. A., Petersen, R. W., & Coall, D. A. (2005). Early stress predicts age at menarche and first birth, adult attachment, and expected lifespan. Human Nature, 16(3), 233–265.CrossRefGoogle Scholar
  9. Comings, D. E., Muhleman, D., Johnson, J. P., & MacMurray, J. P. (2002). Parent–daughter transmission of the androgen receptor gene as an explanation of the effect of father absence on age of menarche. Child Development, 73(4), 1046–1051.CrossRefGoogle Scholar
  10. Culpin, I., Heron, J., Araya, R., Melotti, R., Lewis, G., & Joinson, C. (2014). Father absence and timing of menarche in adolescent girls from a UK cohort: The mediating role of maternal depression and major financial problems. Journal of Adolescence, 37(3), 291–301.CrossRefGoogle Scholar
  11. Dorn, L. D., Sontag-Padilla, L. M., Pabst, S., Tissot, A., & Susman, E. J. (2013). Longitudinal reliability of self-reported age at menarche in adolescent girls: Variability across time and setting. Developmental Psychology, 49(6), 1187–1193.CrossRefGoogle Scholar
  12. Doughty, D., & Rodgers, J. L. (2000). Behavior genetic modeling of menarche in US females. In J. Rodgers & D. C. Rowe (Eds.), Genetic influences on human fertility and sexuality (pp. 169–181). Norwell: Kluewer Academic.CrossRefGoogle Scholar
  13. Ellis, B. J. (2004). Timing of pubertal maturation in girls: An integrated life history approach. Psychological Bulletin, 130(6), 920–958.CrossRefGoogle Scholar
  14. Ellis, B. J., & Garber, J. (2000). Psychosocial antecedents of variation in girls' pubertal timing: Maternal depression, stepfather presence, and marital and family stress. Child Development, 71(2), 485–501.CrossRefGoogle Scholar
  15. Ellis, B. J., McFadyen-Ketchum, S., Dodge, K. A., Pettit, G. S., & Bates, J. E. (1999). Quality of early family relationships and individual differences in the timing of pubertal maturation in girls: A longitudinal test of an evolutionary model. Journal of Personality and Social Psychology, 77(2), 387–401.CrossRefGoogle Scholar
  16. Floud, R., Fogel, R. W., Harris, B., & Hong, S. C. (2011). The changing body: Health, nutrition, and human development in the western world since 1700. Cambridge: Cambridge University Press.CrossRefGoogle Scholar
  17. Graber, J. A., Brooks-Gunn, J., & Warren, M. P. (1995). The antecedents of menarcheal age: Heredity, family environment, and stressful life events. Child Development, 66(2), 346–359.CrossRefGoogle Scholar
  18. Herman-Giddens, M. E. (2006). Recent data on pubertal milestones in United States children: The secular trend toward earlier development. International Journal of Andrology, 29(1), 241–246.CrossRefGoogle Scholar
  19. Hoier, S. (2003). Father absence and age at menarche. Human Nature, 14(3), 209–233.CrossRefGoogle Scholar
  20. Hulanicka, B. (1999). Acceleration of menarcheal age of girls from dysfunctional families. Journal of Reproductive and Infant Psychology, 17(2), 119–132.CrossRefGoogle Scholar
  21. James-Todd, T., Tehranifar, P., Rich-Edwards, J., Titievsky, L., & Terry, M. B. (2010). The impact of socioeconomic status across early life on age at menarche among a racially diverse population of girls. Annals of Epidemiology, 20(11), 836–842.CrossRefGoogle Scholar
  22. Jones, B., Leeton, J., McLeod, I., & Wood, C. (1972). Factors influencing the age of menarche in a lower socio-economic group in Melbourne. Medical Journal of Australia, 2(10), 533–535.Google Scholar
  23. Jorm, A. F., Christensen, H., Rodgers, B., Jacomb, P. A., & Easteal, S. (2004). Association of adverse childhood experiences, age of menarche, and adult reproductive behavior: Does the androgen receptor gene play a role? American Journal of Medical Genetics Part B: Neuropsychiatric Genetics, 125(1), 105–111.CrossRefGoogle Scholar
  24. Kanazawa, S. (2001). Why father absence might precipitate early menarche: The role of polygyny. Evolution and Human Behavior, 22(5), 329–334.CrossRefGoogle Scholar
  25. Karapanou, O., & Papadimitriou, A. (2010). Determinants of menarche. Reproductive Biology and Endocrinology, 8(1), 115.CrossRefGoogle Scholar
  26. Kwon, I., & Sohn, K. (2017). Job dissatisfaction of the self-employed in Indonesia. Small Business Economics, 49(1), 233–249.CrossRefGoogle Scholar
  27. Livson, N., & McNeill, D. (1962). The accuracy of recalled age of menarche. Human Biology, 34(3), 218–221.Google Scholar
  28. Maestripieri, D., Roney, J. R., DeBias, N., Durante, K. M., & Spaepen, G. M. (2004). Father absence, menarche and interest in infants among adolescent girls. Developmental Science, 7(5), 560–566.CrossRefGoogle Scholar
  29. Matchock, R. L., & Susman, E. J. (2006). Family composition and menarcheal age: Anti-inbreeding strategies. American Journal of Human Biology, 18(4), 481–491.CrossRefGoogle Scholar
  30. McIntyre, M. H., & Kacerosky, P. M. (2011). Age and size at maturity in women: A norm of reaction? American Journal of Human Biology, 23(3), 305–312.CrossRefGoogle Scholar
  31. McLanahan, S. (1985). Family structure and the reproduction of poverty. American Journal of Sociology, 90(4), 873–901.CrossRefGoogle Scholar
  32. Mendle, J., Turkheimer, E., D'Onofrio, B. M., Lynch, S. K., Emery, R. E., Slutske, W. S., & Martin, N. G. (2006). Family structure and age at menarche: A children-of-twins approach. Developmental Psychology, 42(3), 533–542.CrossRefGoogle Scholar
  33. Moffitt, T. E., Caspi, A., Belsky, J., & Silva, P. A. (1992). Childhood experience and the onset of menarche: A test of a sociobiological model. Child Development, 63(1), 47–58.CrossRefGoogle Scholar
  34. Quinlan, R. J. (2003). Father absence, parental care, and female reproductive development. Evolution and Human Behavior, 24(6), 376–390.CrossRefGoogle Scholar
  35. Romans, S. E., Martin, J. M., Gendall, K., & Herbison, G. P. (2003). Age of menarche: The role of some psychosocial factors. Psychological Medicine, 33(5), 933–939.CrossRefGoogle Scholar
  36. Sapolsky, R. M. (2004). Why zebras don't get ulcers (3rd ed.). New York: Henry Holt.Google Scholar
  37. Sheppard, P., Snopkowski, K., & Sear, R. (2014a). Father absence and reproduction-related outcomes in Malaysia, a transitional fertility population. Human Nature, 25(2), 213–234.CrossRefGoogle Scholar
  38. Sheppard, P., Garcia, J. R., & Sear, R. (2014b). A not-so-grim tale: How childhood family structure influences reproductive and risk-taking outcomes in a historical US population. PloS One, 9(3), e89539.CrossRefGoogle Scholar
  39. Sohn, K. (2013). Monetary and nonmonetary returns to education in Indonesia. Developing Economies, 51(1), 34–59.CrossRefGoogle Scholar
  40. Sohn, K. (2014a). Age and size at maturity in Indonesian women: A norm of reaction? American Journal of Human Biology, 26(5), 713–715.CrossRefGoogle Scholar
  41. Sohn, K. (2014b). Job strenuousness and obesity: The case of a developing country. Journal of Development Studies, 50(9), 1289–1301.CrossRefGoogle Scholar
  42. Sohn, K. (2014c). Sufficiently good measures of obesity: The case of a developing country. Journal of Biosocial Science, 46(6), 797–817.CrossRefGoogle Scholar
  43. Sohn, K. (2015a). Gender discrimination in earnings in Indonesia: A fuller picture. Bulletin of Indonesian Economic Studies, 51(1), 95–121.CrossRefGoogle Scholar
  44. Sohn, K. (2015b). The height premium in Indonesia. Economics & Human Biology, 16, 1–15.CrossRefGoogle Scholar
  45. Sohn, K. (2015c). The trend in age at menarche in Indonesia: Birth cohorts 1944–1988. Journal of Biosocial Science, 47(3), 407–412.CrossRefGoogle Scholar
  46. Sohn, K. (2015d). The value of male height in the marriage market. Economics & Human Biology, 18, 110–124.CrossRefGoogle Scholar
  47. Sohn, K. (2015e). Sick but unaware: Hypertension in Indonesia. Biodemography and Social Biology, 61(3), 298–318.CrossRefGoogle Scholar
  48. Sohn, K. (2016a). Does a taller husband make his wife happier? Personality and Individual Differences, 91, 14–21.CrossRefGoogle Scholar
  49. Sohn, K. (2016b). Height and happiness in a developing country. Journal of Happiness Studies, 17(1), 1–23.CrossRefGoogle Scholar
  50. Sohn, K. (2016c). The role of spousal income in the wife’s happiness. Social Indicators Research, 126(3), 1007–1024.CrossRefGoogle Scholar
  51. Sohn, K. (2016d). Biological standards of living: Age at menarche vs. height. Annals of Human Biology, 44(1), 21–27.CrossRefGoogle Scholar
  52. Sohn, K. (2016e). The influence of climate on age at menarche: Augmented with the influence of ancestry. Homo, 67(4), 328–336.1.CrossRefGoogle Scholar
  53. Sohn, K. (2016f). Is age at menarche a good predictor of future body fat? The case of a developing country. Health Care for Women International, 37(11), 1239–1257.CrossRefGoogle Scholar
  54. Sohn, K. (2016g). Risk incomprehension and its economic consequences. Journal of Development Studies, 52(11), 1545–1560.CrossRefGoogle Scholar
  55. Sohn, K. (2016h). The risk preferences of entrepreneurs in Indonesia. Bulletin of Economic Research. doi:10.1111/boer.12088.
  56. Sohn, K. (2017a). The fatter are happier in Indonesia. Quality of Life Research, 26(2), 393–402.CrossRefGoogle Scholar
  57. Sohn, K. (2017b). Improvement in the biological standard of living in the 20th century Korea: Evidence from age at menarche. American Journal of Human Biology, 29(1), e22882.Google Scholar
  58. Sohn, K. (2017c). The association between height and hypertension in Indonesia. Economics and Human Biology, 27(Pt A), 74–83. doi:10.1016/j.ehb.2017.04.007.CrossRefGoogle Scholar
  59. Sohn, K., & Kwon, I. (2016). Does trust promote entrepreneurship in a developing country? Singapore Economic Review. doi:10.1142/S0217590816500144.
  60. Stearns, S. C. (1989). Trade-offs in life-history evolution. Functional Ecology, 3(3), 259–268.CrossRefGoogle Scholar
  61. Surbey, M. K. (1990). Family composition, stress, and the timing of human menarche. In T. E. Ziegler & F. B. Bercovitch (Eds.), Socioendocrinology of primate reproduction (pp. 11–32). New York: Wiley-Liss.Google Scholar
  62. Tanner, J. M. (1962). Growth and adolescence. Oxford: Blackwell.Google Scholar
  63. Tither, J. M., & Ellis, B. J. (2008). Impact of fathers on daughters' age at menarche: A genetically and environmentally controlled sibling study. Developmental Psychology, 44(5), 1409–1420.CrossRefGoogle Scholar
  64. Treloar, S. A., & Martin, N. G. (1990). Age at menarche as a fitness trait: Nonadditive genetic variance detected in a large twin sample. American Journal of Human Genetics, 47(1), 137–148.Google Scholar
  65. Vigil, J. M., & Geary, D. C. (2006). Parenting and community background and variation in women's life-history development. Journal of Family Psychology, 20(4), 597–604.CrossRefGoogle Scholar
  66. Webster, G. D., Graber, J. A., Gesselman, A. N., Crosier, B. S., & Schember, T. O. (2014). A life history theory of father absence and menarche: A meta-analysis. Evolutionary Psychology, 12(2), 273–294.CrossRefGoogle Scholar
  67. Wierson, M., Long, P. J., & Forehand, R. L. (1993). Toward a new understanding of early menarche: The role of environmental stress in pubertal timing. Adolescence, 28(112), 913–924.Google Scholar

Copyright information

© Springer Science+Business Media, LLC 2017

Authors and Affiliations

  1. 1.School of Economics and FinanceCurtin UniversityPerthAustralia
  2. 2.Department of EconomicsKonkuk UniversitySeoulSouth Korea

Personalised recommendations