## Abstract

Although difficulty conceiving a child has long been a major medical and social preoccupation, it has not been considered as a predictor of long-term outcomes in children ultimately conceived. This is consistent with a broader gap in knowledge regarding the consequences of parental health for educational performance in offspring. Here we address that omission, asking how resolved parental infertility relates to children’s academic achievement. In a sample of all Swedish births between 1988 and 1995, we find that involuntary childlessness prior to either a first or a second birth is associated with lower academic achievement (both test scores and GPA) in children at age 16, even if the period of infertility was prior to a sibling’s birth rather than the child’s own. Our results support a conceptualization of infertility as a cumulative physical and social experience with effects extending well beyond the point at which a child is born, and emphasize the need to better understand how specific parental health conditions constrain children’s educational outcomes.

## Introduction

Difficulty conceiving a child has been a major medical and social preoccupation for thousands of years. Ancient Egyptians believed that such problems reflected a disruption between the genitals and the digestive tract, and prayed to the gods for reprieve (Morice et al. 1995). European women in the Middle Ages attempted cures in the form of pagan-derived rituals, such as drinking from sacred fountains or sucking on eggs that had been buried near a chapel (Morice et al. 1995). Despite millennia of clinical interest, modern techniques by the beginning of the twentieth century still provided little hope for a medical resolution, and infertility remained largely a private, personal struggle (Scritchfield 2009).1

The first baby conceived via in vitro fertilization (IVF) was born in England in 1978, marking a watershed event in the treatment of infertility (Sullivan 1981). Popular media was quickly lamenting an “epidemic” of infertility curable by new medical procedures, despite little evidence of any change in actual infertility prevalence (Scritchfield 2009). From a sociological perspective, one line of the limited research on infertility has thus taken a “social problems” approach, theorizing the social consequences of a pressing but largely hidden medical issue suddenly meeting the technology able to address it (Scritchfield 2009). Another line of research has suggested infertility as a useful case study for investigating traditional topics in the sociology of health and illness, such as disparities in health care; the relationships between health, gender, and identity; and ways in which variation in patriarchal social norms shapes the relationship between society and the body (Greil et al. 2011). However, infertility—and particularly infertility that is ultimately resolved by childbirth—has not thus far been considered from a sociodemographic perspective as a potential predictor of long-term variation in social outcomes, let alone for the children whose births constitute the “cure” for their parents’ condition.

This omission is consistent with a broader gap in knowledge regarding the consequences of parental health for educational outcomes in offspring. Indeed, in the two decades since Haveman and Wolfe (1995) named the scarcity of data on health status among parents as a notable omission in models of children’s attainment, few studies have addressed this topic (Bratti and Mendola 2013). A larger body of research has linked parental illness to variation in more micro-level indicators of child functioning: children of parents living with chronic pain, for example, have been found to be less generally well-adjusted, with behavior problems, poorer social skills, and higher levels of delinquency (Mikail and von Baeyer 1990; Rickard 1988). Parental mental illness is also considered a risk factor for child cognitive and socioemotional development (Canadian Paediatric Society 2004; Cummings and Davies 1994) associated with adjustment problems and increased likelihood of eventual development of depressive disorders (Downey and Coyne 1990). Few studies, though, have attempted to link specific parental health conditions with longer-term measures of children’s educational performance, despite the vast breadth of research aimed at identifying predictors of academic attainment and achievement.

To the extent that the effect of an extended period of infertility continues past childbirth and into parenting years, similar consequences to those observed in children whose parents experience other categories of illness provide a plausible pathway via which resolved infertility might be hypothesized to produce disparities in an array of longer-term child outcomes, such as academic achievement. In the present analysis, we begin with that question: whether parental infertility that has been formally resolved by childbirth is associated with persistent educational consequences for children. We then consider how three measurable sources of variation in the experience of infertility may affect the relationship between infertility and children’s educational achievement.

First, we ask how the duration of infertility matters for child academic performance. Second, we ask when infertility matters: are years of infertility in a couple without children (primary infertility) a greater detriment to child educational outcomes than infertility before later births (secondary infertility)? Third, we ask for whom infertility matters: is any effect on achievement limited to children whose births resolved a spell of infertility, or do effects extend to all children in a household?

In a sample of all births in Sweden between 1988 and 1995, we find years of infertility preceding either a first or a second birth to be associated with grade point average (GPA) and math test scores at the completion of primary school (typically age 16) in both firstborn and second-born children ultimately conceived, with differences in the magnitude of the effect by all three sources of variation considered (duration of infertility, timing of infertility, and birth order of the child). Although future research is needed to identify the more micro-level mechanisms driving these associations, the finding that infertility can have long-term negative consequences on children’s academic achievement suggests that infertility might be best conceptualized as a cumulative and intergenerational physical and social experience, extending well beyond the simple binary of whether a child is ultimately born. It further speaks to the need to better understand the role of parental health in constraining children’s educational outcomes, both as a contribution to models of achievement and attainment, as well as to our understanding of the relationship between the physical body and the social world.

## Background

The lack of prior research on the sociodemographic consequences of infertility leaves open a wide range of competing hypotheses regarding how and why an experience of resolved parental infertility might be expected to matter for long-term social outcomes in children. From a parental investment perspective, for example, one might characterize efforts to conceive through a period of infertility as demonstrating parents’ willingness and ability to invest time and energy in a child even before that child’s birth. Given that parental investments are associated with better educational outcomes (Yeung et al. 2002), this framework could suggest a positive effect of infertility on children’s educational performance later in life.

Social psychological studies on the consequences of infertility paint a different picture. Parents struggling with extended periods of infertility often spend years prior to a birth managing the associated anxiety (Biringer et al. 2015) and depressive symptoms (Cousineau and Domar 2007), while simultaneously idealizing the experience of parenting (Hammarberg et al. 2008; Möller and Fällström 1991). After a child arrives, parents may have difficulty forming a secure parental identity after having doubted their likelihood of ever becoming parents, and may have difficulty reconciling the realities of raising a child with their idealized image of what parenting would be like (Hammarberg et al. 2008). The disjuncture between the parenting experience imagined during the period of infertility and the subsequent experience of actual parenting—combined with increased defensiveness about one’s parenting abilities (McMahon and Gibson 2002), and a sense that mothers with resolved infertility “got what they wanted” and must hide any negative emotions associated with pregnancy and parenting— may contribute to the higher risk of postpartum depression among previously infertile mothers (Olshansky and Sereika 2005). This literature is largely based on small samples in which group differences between previously infertile and never infertile parents may be difficult to detect, but those studies that do detect meaningful effects are generally consistent in finding infertility to be a negative experience for parents (Greil et al. 2010).

It is from this perspective that we hypothesize infertility may have negative long-term consequences for children, similar to the negative child outcomes associated with other parental health problems. This hypothesis is generally consistent with findings from the handful of social psychological studies investigating the consequences of resolved parental infertility on child adjustment, although such studies have frequently left questions open regarding whether observed differences are truly among parents or children. For example, infertility-treated parents express more concern about their child’s behavior through age 5 (Gibson et al. 2002; McMahon and Gibson 2002), although whether the observed differences are in child behavior itself or instead in parental perceptions of child behavior is debated. At least one study found differences in child behavior by early grade school, with 9- and 10-year-olds conceived via infertility treatments found to be less well-adjusted and reporting poorer emotional well-being (Levy-Shiff et al. 1998). To our knowledge, no research to date has investigated the relationship between parental infertility and child outcomes past that age.

Our analysis additionally differs from prior literature in considering episodes of infertility even among couples that have already given birth to a first child. Although a majority of research on infertility has focused on couples who struggled to conceive any children at all, difficulty in conceiving a second child may also be traumatic, particularly in a country with a strong two-child norm such as Sweden (Bernhardt and Goldscheider 2014). Social narratives surrounding childbearing do not stop with the binary question of parenthood or not; rather, pressure to provide a sibling as well as concerns about stigma against only children remain salient pressures among couples experiencing secondary infertility (Falbo and Polit 1986). Parents suffering from secondary infertility often face the additional challenge of having their struggle taken less seriously than in the case of primary infertility (Muthler 2013).

As in any quantitative field of analysis, the data collected on involuntary childlessness both inform, and are informed by, the questions that researchers seek to answer (Espeland and Stevens 2008). That little is known about the long-term socioeconomic consequences of parental infertility on offspring is due in part to limitations in available data; for example, in the United States, no large longitudinal study currently includes reliable measures of both years infertile prior to childbirth as well as later child academic achievement. National registry data, such as those available in Sweden, are thus a unique resource for considering the association between resolved infertility and intergenerational social outcomes. That infertility treatments are available as part of publicly provided medical services in Sweden also greatly reduces the concern that would be present in a U.S. sample regarding infertility being more frequently resolved among those with sufficient private resources to afford medical intervention.

Within sociological research on health and illness, the limited availability of data on physical conditions such as infertility reflects a disciplinary tradition of portioning off illness into the biological elements (such as symptoms, treatments, and other medical specifics) versus the social elements of interest (such as how doctors and patients interact, how the experience of illness affects relations in families, school, and work, and so on) (Timmermans and Haas 2008). Defining illness as strictly a social phenomenon means that remarkably little is known about how variation within particular diseases affects social outcomes and vice versa, despite the potential benefit of such research for better understanding the relationship between the physical body and the social world. This study additionally contributes at that juncture, asking not only whether infertility as a blanket experience is associated with child academic achievement, but also how variation within the experience of infertility may differentially affect educational performance.

## Methods and Analysis

Our study population is extracted from the Swedish Interdisciplinary Panel (SIP), administered at the Centre for Economic Demography, Lund University, Sweden. SIP covers the entire population living in Sweden born from 1973 to 1995 and is also linked to parents and siblings born outside this period. Through the linking of multiple administrative registers, SIP allows for the examination of individuals’ socioeconomic, demographic, and health outcomes over time, with the period of data availability restricted to 1968–2012. As noted earlier, to our knowledge, no U.S. data source currently includes measures of both years of infertility as well as long-term outcomes in parents and in children ultimately conceived.

For this study, key information is obtained from the medical birth register (MBR) and from the Grade 9 register. The MBR is provided by the Swedish Board of Health and Welfare and contains a range of pertinent information on both the mother and her child, collected during pregnancy and at delivery. Our independent variable of interest is taken from the question in which an expecting mother was asked whether her current pregnancy was preceded by a period of involuntary childlessness, reported as integer years. We use the term “infertility” as per the medical definition, describing an inability to attain a live birth despite consistent exposure to pregnancy for a period of time, typically one year or more (Mascarenhas et al. 2013; World Health Organization (WHO) 1991). Medical infertility is thus largely synonymous with infecundity in that it does not describe an absence of children but rather an inability to produce children. Primary infertility describes a childless couple struggling to produce children, and secondary infertility describes difficulty attaining a live birth among parents who have already had a child (WHO 1991). A couple that ultimately has a child following a period of involuntary childlessness is considered to have resolved infertility.2

The Grade 9 register provides our dependent variables of interest: individuals’ GPA when finishing elementary school (grundskola) and national standardized math test scores from the same year. Unfortunately, no earlier academic performance measures are available. Elementary school in Sweden is typically completed during the year in which a child turns 16, and the GPA obtained at its completion has implications for the program or study track that a student enters at the high school (gymnasium) level. In Sweden, GPAs range from 0 to 320; the GPA measure for the full population is approximately normally distributed around a mean of 204 and a standard deviation of 55. A student’s GPA is obtained by summing grades from 16 completed courses in which each letter grade is assigned a point value.3 Although national tests are also administered in Swedish and in English, math is the only subject in which test scores are graded on a continuous scale; other subjects are graded on a pass/fail system.4 Using data on GPA, test scores, and from the MBR restricts the analytical sample to individuals born between 1988 and 1995, who thus completed elementary school between 2004 and 2012.

We model the association between infertility and both test scores and GPA based on the understanding that these quantities capture overlapping but distinct constructs: math test scores are a single-time-point measure of academic performance, whereas GPA is a cumulative measure of achievement that will also reflect aspects of socialization (Kelly 2008). Although very few studies have considered the effects of infertility on school-aged children, Levy-Shiff et al. (1998) found no cognitive, physical, or neurological differences between 9- and 10-year-olds who were conceived via reproductive technology versus those who were not—but they did find significant differences in socioemotional adjustment. Socioemotional differences as early as age 9 might well translate into differences in test-based measures of achievement by age 16. Following Levy-Shiff et al. (1998), we also hypothesize an additional negative effect of infertility on GPA even net of test scores, as might be expected in the presence of persistent differences in noncognitive skills.

Because of the strong two-child norm in Sweden (Bernhardt and Goldscheider 2014), we focus on the influence of resolved infertility prior to the birth of firstborn or second-born children whose parents have no prior children from previous relationships. This decision was partially driven by concern that infertility for parents who already have a family of normative size may be a markedly different social and emotional experience from infertility prior to that point; furthermore, approximately 87 % of parents in our sample had two children or fewer.5 We define our analytical sample as the 264,013 first and 201,653 second births for whom year 9 GPA and test scores as well as physical health measurements at birth were recorded. Respondents with missing data on key control variables were dropped from the sample out of concern that missingness from government registries on standard demographic measures could well be associated with outlying true values of the missing data themselves, but this was a rare occurrence in any case, amounting to <2 % of the sample.

Table 1 shows the sample means from the first-birth and second-birth samples, together and separately.6 To investigate whether couples that experienced infertility systematically differ from those that did not, sample means are also presented conditional on infertility, with statistically significant differences noted. Approximately 7 % of firstborn children in our sample have mothers who reported difficulty conceiving for one or more years, corresponding to nearly 20,000 first births that were preceded by a period of resolved infertility. Approximately 3 % of second-born children have mothers who reported difficulty conceiving for one or more years, corresponding to about 6,000 second births that were preceded by a period of resolved infertility. Among women experiencing a spell of infertility prior to a second child, 23 % also reported experiencing infertility prior to the first child. In general, the second-born sample largely resembles the firstborn sample, both overall as well as separated by whether the child’s birth was preceded by a period of infertility.

Despite the common assumption that infertility is the province of older mothers, mean age at first birth among mothers who experienced at least one year of primary infertility is 29.7 years, relative to 26.4 years among first-time mothers reporting no infertility. Mean age at second birth among mothers who experienced at least one year of secondary infertility is 31.1 years, relative to 28.7 years among second-time mothers reporting no infertility. Mean duration of resolved primary infertility is 3.2 years, and mean duration of resolved secondary infertility is 2.1 years; the distributions of these variables are predictably left-skewed, with only 5 % of ever-infertile mothers reporting more than 10 years of primary infertility.

Our outcomes of interest are child’s GPA and national math test scores, each standardized to have a mean of 0 and a standard deviation of 1. Children whose parents have experienced involuntary childlessness as well as those who have not differ in GPA by a statistically significant but substantively meager 3 % of a standard deviation. Differences in math test scores are similarly small. The health of the child at birth is captured via measures of gestational age, birth weight, head circumference, and the five-minute APGAR score (ranging from 0 to 10). Children born subsequent to a spell of involuntary childlessness are characterized by statistically significant but surprisingly small differences in these measures relative to children born to parents reporting no involuntary childlessness. For example, whereas 2.4 % of firstborn children whose parents did not experience infertility had a measured head circumference of more than 2 standard deviations below the mean, this figure was 2.6 % among children born following a spell of primary infertility. Indicators are included for child’s year of birth, age at graduation, and sex.

Maternal health is captured via indicators of whether a mother had ever experienced any physical or mental health episode requiring hospitalization, as well as smoking behavior and age at childbirth.7 Although our data are constrained to information recorded in official government records and thus do not capture more minor health concerns, our measure will control for any major illnesses that might be expected to predict infertility. Because foreign-born mothers will not have complete medical data recorded in the Swedish register, we limit our sample to include only children of mothers born in Sweden. Parental socioeconomic and demographic controls include marital and cohabitation status when the child is 14 years old, educational attainment and lifetime earnings for both parents, the highest recorded number of children living in the household during the child’s upbringing, and whether the father is foreign-born.8 As might be expected, parents who experienced infertility are more likely to still be married or cohabitating when the child is 14 years of age relative to the general population, given that reported years of involuntary childlessness assumes a relationship of at least that duration prior to childbirth. By comparison, the non-infertile population is expected to also include children born to individuals who were in less-committed relationships of shorter duration prior to conception. Future research might investigate whether involuntary childlessness is associated with increased likelihood of relationship dissolution among couples whose relationships were of similar duration prior to childbirth.

To affirm that any association between infertility and achievement was not simply channeling differences by infertility in marital stability, we ran supplemental analyses constraining our sample to only those children whose parents were still cohabiting or married when the child was age 14. Results were substantively unchanged from models run on the full sample. We ran a second supplemental analysis constraining our sample to singletons only in order to affirm that any association between infertility and achievement was not simply capturing higher likelihood of being a singleton among children of parents experiencing involuntary childlessness; results were again substantively unchanged from models run on the full sample.

### Analytic Strategy

To answer the question of whether parental infertility is associated with lower educational achievement, we use the ordinary least squares (OLS) regression equation,
$yi=α+βcCi+βRRi+βTTi+βDDi+βSSi+βSSi×Ci+βFFj+εi,$

in which yi is a measure of achievement for child i, via either test scores or GPA. The independent variable of interest—a measure of involuntary childlessness prior to child i’s birth—is denoted by C. To address nonlinearity in C, we assess the effects of two categories of infertility: one to four years, and five or more years.9R is a vector of child characteristics, including a full set of indicators for child year of birth and child age of graduation from elementary school, and measures of child’s health at birth (gestational age, birth weight, head circumference, and five-minute APGAR score). In models where the outcome is GPA, T denotes child i’s math test scores. D is a vector of socioeconomic and demographic characteristics of child i’s parents and household, including marital and cohabitation status, educational attainment for both parents, lifetime earnings for both parents, highest recorded number of children living in the household during the child’s upbringing, whether the father is foreign-born, maternal age at birth, and maternal health (hospitalizations for physical or mental health problems, and smoking habits). S denotes years of infertility before the birth of a child’s sibling, again as two categories spanning one to four years and five or more years of infertility. F is a fixed effect on each unique combination of the hospital where a child was born and the school where a child’s GPA at age 16 was recorded, denoted j. Parental earnings and math test scores are continuous measures; all remaining variables in the model are included as a full set of indicators for each categorical value.

Although we include separate measures of primary and secondary infertility, we also consider that a second bout of infertility may be a meaningfully different experience for a family that already endured primary infertility relative to a family that did not. Although such parents enter into secondary infertility having already experienced the trauma of prior infertility, they also have the firsthand understanding that their infertility can be resolved, are less likely to be surprised by difficulty conceiving a second time, and may have an understanding of the specific processes through which their first bout of infertility ultimately resulted in a pregnancy. To the extent that a second experience of infertility in a single family thus has a smaller effect on child achievement than does secondary infertility in parents that were not previously infertile, imposing an assumption of linearity on separate parameter estimates for primary and secondary infertility risks artificially underestimating both parameters. To account for the possibility that the effect of experiencing more than one bout of infertility is not additive, we thus include interaction terms between C and S. Given the rarity of conceiving a child after five or more years of infertility, we include only the interaction between having experienced between one and four years of infertility prior to both a first and a second birth.10

Because the question on involuntary childlessness was asked of mothers at the time of childbirth, we observe only infertility that was ultimately resolved by the birth of a child. As such, our measure of years of infertility before a sibling’s birth (S) is complete for second-born children, but not for firstborn children, among whom we will not capture families with unresolved secondary infertility. Although prevalence rates of resolved infertility are difficult to estimate (Irvine 1998), 3 % of our second-born children have mothers reporting resolved secondary infertility, and prior research suggests rates of unresolved secondary infertility ranging between approximately 3 % (Irvine 1998) and approximately 10 % (Mascarenhas et al. 2013). As such, mothers with unobserved unresolved secondary infertility in our sample may even outnumber the mothers with resolved secondary infertility. Our estimates of the effects of infertility prior to a second birth on firstborn children should thus be interpreted cautiously as a lower-bound estimate on the true parameter value for the population. That said, sensitivity analyses limiting the sample to couples with exactly two children and to couples with at least two children yielded no substantive difference in the pattern of results, suggesting that the bias in our models from unobserved unresolved secondary infertility may be minimal.

We run our models separately by birth order, first using test scores and then GPA as the outcome. Only estimates on measures of involuntary childlessness (Ci and Si) are listed in the main tables; estimates on all other variables are provided in Tables S1S3 in Online Resource 1.

## Results

Results from the regression analyses on the association between parents’ years of involuntary childlessness and children’s math test scores are presented in Table 2. Among firstborn children (Model 1), longer durations of primary infertility are indeed associated with increasingly adverse test score outcomes, even net of our battery of sociodemographic controls, measures of child and maternal health, and school and hospital fixed effects. Relative to peers whose parents had no difficulty conceiving a first birth, firstborn children have a 4.4 % standard deviation lower math test score if their parents experienced primary infertility lasting from one to four years, and an 8.2 % standard deviation lower math test score if primary infertility lasted five or more years. Infertility prior to a second birth (Model 2), on the other hand, is not associated with any additional test score penalty on firstborn children; this holds even in supplemental analyses restricting the sample to families with two children or more, suggesting that the lack of any meaningful effect of secondary infertility on firstborn children’s test scores does not simply result from unobserved unresolved secondary infertility.

Like for firstborn children, we also find an association between primary infertility and test score performance in second-born children (Table 2). Relative to peers whose parents had no difficulty conceiving a first birth, second children have a 3.6 % standard deviation lower math test score if their parents experienced primary infertility lasting from one to four years, and a 6.0 % standard deviation lower math test score if primary infertility lasted five or more years. Unlike for firstborn children, secondary infertility among second-born children (Model 2) is associated with an additional test score penalty—one that is of similar magnitude to the penalty associated with primary infertility. Although this effect is statistically significant for only one to four years of secondary infertility, that may be due to the exceedingly small percentage of parents who conceived a second child after being infertile for a half-decade or more.

The negative association between parental infertility and child achievement also holds when the outcome measure is GPA rather than test scores (Table 3). Relative to peers whose parents had no difficulty conceiving a first birth, firstborn children have a 6.4 % standard deviation lower GPA if their parents experienced primary infertility lasting from one to four years, and an 11.3 % standard deviation lower GPA if primary infertility lasted five or more years (Model 1.1). These effects are reduced in magnitude by approximately one-half but remain substantively meaningful and statistically significant after test scores are controlled for (Model 1.2); in the metric of grade points, the observed effect is equivalent to a decrease of approximately one letter grade in any of the 16 courses used to calculate cumulative GPA. This result is potentially consistent with Levy-Shiff et al.’s (1998) findings of poorer social adjustment in 9- and 10-year-old children conceived using infertility treatments, as a negative effect of infertility on GPA even net of test scores may suggest that the relationship is driven at least in part by persistent differences in noncognitive skills.

As discussed, a second bout of infertility may well be a meaningfully different psychosocial experience for a family that already resolved primary infertility, because such families are less likely to be surprised by their difficulty conceiving and more likely to understand infertility as a condition that can be resolved. If a second experience of infertility has a smaller effect on child achievement than does secondary infertility in parents who were not previously infertile, by failing to account for families that experienced infertility twice, we risk underestimating the parameters on both primary and secondary infertility. We observe this in Table 3: secondary infertility does not appear to affect firstborn children’s GPA (Models 2.1 and 2.2) until we account for repeated bouts of infertility within the same family (Model 2.3). Although our interaction between experiencing infertility before both a first and a second birth is not itself statistically significant, after accounting for repeated bouts of infertility, the parameter estimate on secondary infertility of one to four years is rendered both 30 % larger in magnitude and statistically significant. That the interaction term is in the opposite direction of the coefficients on both primary and secondary infertility of one to four years is consistent with the suggestion that secondary infertility is a greater detriment to children of parents that have not already resolved a bout of primary infertility.

The relationship between infertility and GPA among second-born children largely mirrors the findings for firstborn children (Table 4). Relative to peers whose parents had no difficulty conceiving a first birth, second children have a 4 % standard deviation lower GPA if their parents experienced primary infertility lasting from one to four years and a 7 % standard deviation lower GPA if primary infertility lasted five or more years. Both parameters are halved in magnitude after math test scores are controlled for (Model 1.2); in the metric of grade points, the observed effects are equivalent to a decrease of approximately two-thirds of a letter grade in any of the 16 courses used to calculate cumulative GPA. Infertility prior to a second child’s own birth is again nonsignificant net of test scores (Model 2.2) until we control for families that experienced infertility twice (Model 2.3), at which point the parameter estimate on secondary infertility of one to four years is rendered both 40 % larger and statistically significant. The estimate on the interaction term is nonsignificant but in the opposite direction of the coefficients on both primary and secondary infertility of one to four years in duration, suggesting that secondary infertility may be a greater detriment to children’s GPA in families where parents have not already experienced primary infertility.

## Discussion

Although recent decades have seen a shift toward more engagement with physical bodies in sociological research, studies of health and illness still typically endeavor to study social experience independent of specific physical details (Timmermans and Haas 2008). Few conditions better illustrate the limitations of such attempts to segment the biological from the social elements of illness than does infertility, wherein medical definitions classify parents as “cured” with the birth of a child regardless of whether psychological or social effects persist beyond that point. Because infertility is inherently a relationship between parents and a child, any sufficiently long-term negative effects on parents may well have long-term consequences for children as well, potentially extending to critical socioeconomic outcomes such as educational performance. Yet no study to date has considered whether persistent socioeconomic differences exist between parents who did and those who did not experience involuntary childlessness, let alone among the children whose births resolve their parents’ infertility.

We find that involuntary childlessness is indeed associated with lower academic achievement at age 16 among children ultimately conceived, both when measured as test scores and when measured as GPA net of test scores. Furthermore, we find variation in the association between parental infertility and child achievement by measurable differences in the experience of infertility: duration in years, timing (primary versus secondary infertility), and birth order (effects on firstborn versus second-born children). In general, longer bouts of infertility confer greater penalties on test scores and on GPA net of test scores for both firstborn and second-born children. Primary infertility is associated with a larger and more consistent penalty on achievement than is secondary infertility, although secondary infertility may have a stronger effect on the GPA of children whose parents had not already resolved a bout of primary infertility.

Given the dearth of studies investigating the effects of resolved infertility on school-aged children, a primary implication of this analysis is that more work is needed to unpack the micro-level processes linking infertility to educational achievement, as well as the macro-level factors that may mediate these effects. Although the results presented are clear in causal direction—a child’s achievement at age 16 clearly does not retroactively cause parental infertility—infertility is a complicated biological, psychological, and social experience that has the potential to affect child outcomes through a wide array of pathways.

Although our analysis does not allow for determination of specific causal mechanisms, we can speak to a limited set of plausible pathways that do not explain the association observed. First, the comprehensive set of controls included here to account for children’s physical health at birth suggests that infertility is not simply associated with having less healthy babies, which might otherwise easily translate into lower educational performance later in life. Second, that our results are robust across supplemental models constraining the sample to children whose parents remain married or cohabiting suggests that any marital instability associated with extended infertility does not account for the differences in child achievement observed. Third, that our results are robust across supplemental models constraining the sample to singletons suggests that infertility is not associated with academic achievement via a higher likelihood of being an only child.

Based on the psychological literature on shorter-term parental experience of resolved infertility, combined with the meager research on parental health and child achievement, we began this analysis with the tentative hypothesis that infertility may itself constitute a sufficient psychosocial shock to parenting such as to have long-term effects on offspring. Although this explanation may account for the variation in test score outcomes that we observe between children whose parents did and did not experience infertility, our data do not contain early-life measures of cognition that would allow us to affirm that the effect on test scores operates through sociobehavioral pathways rather than reflecting underlying cognitive differences. Whereas Levy-Shiff et al. (1998) found sociobehavioral differences but no gap in cognitive functioning among 9- and 10-year-olds conceived via fertility treatments versus those conceived naturally, here we cannot rule out the possibility of at least a partial cognition effect. To the extent that GPA reflects both cognitive and noncognitive skills, however, the persistent penalty of infertility on GPA even net of test scores suggests that our results are at least in part a result of noncognitive differences between children whose parents did and did not experience infertility. This point is emphasized by the effect of secondary infertility on the GPA of firstborn children whose parents did not experience primary infertility, which will be necessarily conferred through lived experience in the family environment.

Even therein, the specific psychological dynamics linking parental experience of infertility with child achievement remain a black box and in need of further study. The benefits of registry data—including the ability to observe full birth cohorts with little missing information—come with the trade-off that we lack measures of more micro-level social and physical health experiences, as well as genetic, epigenetic, and other biological transmissions between parents and children (Lane et al. 2014). For example, our models contain comprehensive measures of children’s physical health at birth, but we observe only maternal physical and mental health problems if they were of sufficient severity to require hospitalization. Although infertility is not exclusively the province of parents with other observable medical problems, finer-grained health data would enable us to identify parents for whom both infertility and subsequent educational penalties among children might result from persistent maternal or paternal health conditions comorbid with infertility but too mild to require hospitalization. Examples might include hormonal disorders, sperm motility problems, history of sexually transmitted infections, and so on (e.g., Eisenberg et al. 2015).11 Such variables could be better accounted for in future research using more micro-level data.

One additional limitation of our data is the inability to distinguish bouts of infertility that were resolved spontaneously versus via infertility treatment, or via treatment obtained through a private clinic versus through the public system. That said, Sweden was a forerunner in uptake of IVF, and by the late 1980s, infertility treatments were widely available as part of government-provided medical care (Cohen et al. 2005). Even services at private infertility clinics remain far less expensive than equivalent treatments in the United States (Katz et al. 2011). These differences in the accessibility of treatment are evident in the use of such technology: in the year 2000, Sweden reported approximately 1,000 IVF cycles per 1 million inhabitants—approximately four times the rate of IVF treatment in the United States (Cohen et al. 2005). Future research might thus consider not only whether infertility serves as a risk factor with respect to other later-life child outcomes but also how such relationships differ as a function of variation by country in the accessibility of infertility treatment.

Even without identifying specific psychological mechanisms, the finding that infertility is associated with long-term negative consequences on children’s achievement should emphasize the need for both parents and medical practitioners to address all aspects of the infertility experience, rather than just the simple physical binary of whether a child is ultimately born. This holds for families suffering from primary infertility, given that the infertility experience has consequences for achievement not only in the child ultimately conceived, but even for that child’s siblings. It holds also for parents experiencing secondary infertility, who frequently describe the added pain of little support from friends, family, and medical professionals who view difficulty conceiving among individuals who are already parents as a lesser plight than among individuals without children (Muthler 2013). Because rates of unresolved secondary infertility have been estimated to be as high as 10.5 % (Mascarenhas et al. 2013), our finding of a possible association between achievement and resolved secondary infertility among firstborn children suggests a plausible effect on only children of parental secondary infertility that is never resolved. This point in particular merits further study.

Finally, that measurable differences in the experience of infertility (timing, duration, and birth order) would affect the association between infertility and social outcomes seems common sense, and yet exemplifies a standard omission in how health is conceptualized in sociological research. As per Timmermans and Haas (2008:665), “while sociologists may be attentive to differences in ethnicity, gender, communication styles, or social capital, disease is disease. Any stroke is equivalent to any other stroke. Severity, stages, or symptoms do not seem to matter.” Among diseases more frequently considered from a population research standpoint, obesity serves as a useful example of this critique: although numerous studies have considered the relationship between the binary experience of obesity relative to social outcomes such as educational performance (Crosnoe and Muller 2004; Kaestner and Grossman 2009; Kaestner et al. 2011), no analysis prior to Milesi et al. (2013) considered how the duration of obesity might matter for educational attainment. Their findings affirmed that any obesity is not equivalent to any other obesity, as our findings affirm the same for infertility. Considering a given disease as a complex physical experience rather than a blanket social state has the potential for practical application by clinicians and policy-makers, as well as for improving our theoretical understanding of the relationship between the physical body and the social world.

## Acknowledgments

The authors acknowledge funding from the Centre for Economic Demography and from the Cornell Population Center, and additionally thank participants at the seminars in the Centre for Economic Demography and the Cornell Population Center, who provided invaluable feedback on this manuscript.

## Notes

1

The term “infertility” is used here as per the medical definition, describing a failure to attain a live birth despite consistent exposure to pregnancy risk (Mascarenhas et al. 2013; WHO 1991). Medical infertility is thus largely synonymous with infecundity; find more on this point in the Methods and Analysis section.

2

Our measure of infertility is based on the time to a live birth. We unfortunately lack information on whether mothers miscarried during that period unless the miscarriage was late enough in a pregnancy to qualify as a stillbirth or otherwise require hospitalization. Cases of stillbirth were very few and were omitted from the sample.

3

Grades range from A to F, with E as the lowest passing grade. A top score (“pass with special distinction”) is equivalent to 20 points, and a minimal passing score is equivalent to 10 points.

4

Test scores in Swedish and English are reported as an ordinal measure (Fail, Pass, Pass with distinction, Pass with high distinction). Running linear probability models using “Pass with high distinction” in Swedish language as the outcome, we obtain results similar to those presented for math. No results are found for English.

5

The mean number of children in the full registry sample over the time frame defined here was 1.72. Families with only one child constituted 43 % of the full sample; families with two children constituted 44 %; with three children, 11 %; with four or more children, 1.6 %.

6

Additional detail on measurement and timing of variables is presented in Table S4, Online Resource 1.

7

Although paternal health is also an important consideration in understanding causes of infertility, paternal health behaviors (such as smoking) were not recorded in the MBR.

8

The measurement of demographic controls at age 14 and test scores at age 16 reflects data availability rather than an effort to lag covariates. Lifetime earnings are calculated as per Lindahl et al. (2015).

9

Additional specifications included using indicators for each year of infertility, each two years of infertility, and including C as a continuous measure with higher-order terms. Results were substantively the same as those presented using the two-category specification of C. Frequencies of primary and secondary infertility for zero through five or more years are presented in Table S5, Online Resource 1.

10

Couples who experienced five or more years of infertility are rare, and couples who experienced five or more years of infertility before both a first and a second birth are extremely rare (<0.1 % of the sample). Coefficients on the remaining interactions between Ci and Si were in the same direction as for the interaction presented but with volatility in magnitudes, given the dramatically declining cell sizes.

11

Sexually transmitted infections (STIs) could be a particularly useful set of controls in this case, given the higher rates of STIs in Scandinavian countries relative to other Western European nations (European Centre for Disease Prevention and Control 2011). In addition to the medical consequences of STIs for fertility, STI prevalence may be indicative of country-specific cultural norms that also reflect other aspects of lifestyle relevant for parenting.

## References

Bernhardt, E., & Goldscheider, F. (
2014
).
Ambivalence about children in the family building process in Sweden
.
Finnish Yearbook of Population Research
,
49
,
57
71
.
Biringer, E., Howard, L. M., Kessler, U., Stewart, R., & Mykletun, A. (
2015
).
Is infertility really associated with higher levels of mental distress in the female population? Results from the North-Trøndelag Health Study and the Medical Birth Registry of Norway
.
Journal of Psychosomatic Obstetrics and Gynecology
,
36
,
38
45
. 10.3109/0167482X.2014.992411.
Bratti, M., & Mendola, M. (
2013
).
Parental health and child schooling
(GINI Discussion Paper No. 63).
Amsterdam, The Netherlands
:
Amsterdam Institute for Advanced Labour Studies
.
(
2004
).
Maternal depression and child development
.
Paediatrics & Child Health
,
9
,
575
583
. 10.1093/pch/9.8.575.
Cohen, J., Trounson, A., Dawson, K., Jones, H., Hazekamp, J., Nygren, K.-G., & Hamberger, L. (
2005
).
The early days of IVF outside the UK
.
Human Reproduction Update
,
11
,
439
460
. 10.1093/humupd/dmi016.
Cousineau, T. M., & Domar, A. D. (
2007
).
Psychological impact of infertility
.
Best Practice & Research: Clinical Obstetrics and Gynaecology
,
21
,
293
308
. 10.1016/j.bpobgyn.2006.12.003.
Crosnoe, R., & Muller, C. (
2004
).
Body mass index, academic achievement, and school context: Examining the educational experiences of adolescents at risk of obesity
.
Journal of Health and Social Behavior
,
45
,
393
407
. 10.1177/002214650404500403.
Cummings, E. M., & Davies, P. T. (
1994
).
Maternal depression and child development
.
Journal of Child Psychology and Psychiatry
,
35
,
73
122
. 10.1111/j.1469-7610.1994.tb01133.x.
Downey, G., & Coyne, J. C. (
1990
).
Children of depressed parents: An integrative review
.
Psychological Bulletin
,
108
,
50
76
. 10.1037/0033-2909.108.1.50.
Eisenberg, M. L., Li, S., Behr, B., Reijo Pera, R., & Cullen, M. R. (
2015
).
Relationship between semen production and medical comorbidity
.
Fertility and Sterility
,
103
,
66
71
. 10.1016/j.fertnstert.2014.10.017.
Espeland, W. N., & Stevens, M. L. (
2008
).
A sociology of quantification
.
European Journal of Sociology
,
49
,
401
436
. 10.1017/S0003975609000150.
European Centre for Disease Prevention and Control
. (
2011
).
Sexually transmitted infections in Europe, 1990–2009
(Surveillance report).
Stockholm, Sweden
:
ECDC
.
Falbo, T., & Polit, D. F. (
1986
).
Quantitative review of the only child literature: Research evidence and theory development
.
Psychological Bulletin
,
100
,
176
189
. 10.1037/0033-2909.100.2.176.
Gibson, F. L., McMahon, C. A., Cohen, J., Leslie, G. I., & Saunders, D. M. (
2002
).
Children conceived through ICSI and IVF at 5 years of age: Behavioral adjustment, parenting stress and attitudes: A comparative study
.
Fertility and Sterility
,
78
(
Suppl. 1
),
S28
S29
. 10.1016/S0015-0282(02)03455-6.
Greil, A. L., McQuillan, J., & Slauson-Blevins, K. (
2011
).
The social construction of infertility
.
Sociology Compass
,
5
,
736
746
. 10.1111/j.1751-9020.2011.00397.x.
Greil, A. L., Slauson-Blevins, K., & McQuillan, J. (
2010
).
The experience of infertility: A review of recent literature
.
Sociology of Health & Illness
,
32
,
140
162
. 10.1111/j.1467-9566.2009.01213.x.
Hammarberg, K., Fisher, J. R. W., & Wynter, K. H. (
2008
).
Psychological and social aspects of pregnancy, childbirth and early parenting after assisted conception: A systematic review
.
Human Reproduction Update
,
14
,
395
414
. 10.1093/humupd/dmn030.
Haveman, R., & Wolfe, B. (
1995
).
The determinants of children’s attainments: A review of methods and findings
.
Journal of Economic Literature
,
33
,
1829
1878
.
Irvine, D. S. (
1998
).
Epidemiology and aetiology of male infertility
.
Human Reproduction
,
13
(
Suppl. 1
),
33
44
. 10.1093/humrep/13.suppl_1.33.
Kaestner, R., & Grossman, M. (
2009
).
Effects of weight on children’s educational achievement
.
Economics of Education Review
,
28
,
651
661
. 10.1016/j.econedurev.2009.03.002.
Kaestner, R., Grossman, M., & Yarnoff, B. (
2011
).
Effects of weight on adolescent educational attainment
. In M. Grossman & N. H. Mocan (Eds.),
Economic aspects of obesity
(pp.
283
313
).
Chicago, IL
:
University of Chicago Press
.
Katz, P., Showstack, J., Smith, J. F., Nachtigall, R. D., Millstein, S. G., Wing, H., . . . Adler, N. (
2011
).
Costs of infertility treatment: Results from an 18-month prospective cohort study
.
Fertility and Sterility
,
95
,
915
921
.
Kelly, S. (
2008
).
What types of students’ effort are rewarded with high marks?
Sociology of Education
,
81
,
32
52
. 10.1177/003804070808100102.
Lane, M., Robker, R. L., & Robertson, S. A. (
2014
).
Parenting from before conception
.
Science
,
345
,
756
. doi:10.1126/science.1254400
Levy-Shiff, R., Vakil, E., Dimitrovsky, L., Abramovitz, M., Shahar, N., Har-Even, D., . . . Fish, B. (
1998
).
Medical, cognitive, emotional, and behavioral outcomes in school-age children conceived by in-vitro fertilization
.
Journal of Clinical Child Psychology
,
27
,
320
329
.
Lindahl, M., Palme, M., Sandgren Massih, S., & Sjögren, A. (
2015
).
Long-term intergenerational persistence of human capital: An empirical analysis of four generations
.
Journal of Human Resources
,
50
,
1
33
. 10.1353/jhr.2015.0000.
Mascarenhas, M. N., Flaxman, S. R., Boerma, T., Vanderpoel, S., Mathers, C. D., & Stevens, G. A. (
2013
).
Trends in primary and secondary infertility prevalence since 1990: A systematic analysis of demographic and reproductive health surveys
.
Lancet
,
381
(
Special Issue
),
S90
. doi:10.1016/S0140-6736(13)61344-6.
McMahon, C., & Gibson, F. (
2002
).
A special path to parenthood: Parent-child relationships in families giving birth to singleton infants through IVF
.
Reproductive Biomedicine Online
,
5
,
179
186
. doi:10.1016/S1472-6483(10)61622-7.
Mikail, S. F., & von Baeyer, C. L. (
1990
).
Pain, somatic focus, and emotional adjustment in children of chronic headache sufferers and controls
.
Social Science & Medicine
,
31
,
51
59
. 10.1016/0277-9536(90)90009-H.
Milesi, C., Palloni, A., & Bittman, R. (
2013
,
April
).
The effects of timing and duration of adolescent and early adulthood obesity on college enrollment and completion
.
Paper presented at the annual meetings of the Population Association of America
,
New Orleans, LA
.
Möller, A., & Fällström, K. (
1991
).
Psychological consequences of infertility: A longitudinal study
.
Journal of Psychosomatic Obstetrics and Gynecology
,
12
,
27
44
. 10.3109/01674829109077988.
Morice, P., Josset, P., Chapron, C., & Dubuisson, J. B. (
1995
).
History of infertility
.
Human Reproduction Update
,
1
,
497
504
. 10.1093/humupd/1.5.497.
Muthler, S. (
2013
,
July
24
).
Secondary infertility: I have one child. Why can’t I get pregnant again?
The New York Times
Olshansky, E., & Sereika, S. (
2005
).
The transition from pregnancy to postpartum in previously infertile women: A focus on depression
.
Archives of Psychiatric Nursing
,
19
,
273
280
. 10.1016/j.apnu.2005.08.003.
Rickard, K. (
1988
).
The occurrence of maladaptive health-related behaviors and teacher-rated conduct problems in children of chronic low back pain patients
.
Journal of Behavioral Medicine
,
11
,
107
116
. 10.1007/BF00848259.
Scritchfield, S. (
2009
).
The social construction of infertility: From private matter to social concern
. In J. Best (Ed.),
Images of issues: Typifying contemporary social problems
(pp.
131
146
). New
Brunswick, NJ
:
Transaction Publishers
.
Sullivan, W. (
1981
,
December
29
).
First “test-tube” baby born in U.S., Joining successes around world
.
The New York Times
, pp.
1A
.
Timmermans, S., & Haas, S. (
2008
).
Towards a sociology of disease
.
Sociology of Health & Illness
,
30
,
659
676
. 10.1111/j.1467-9566.2008.01097.x.
World Health Organization (WHO)
. (
1991
).
Infertility: A tabulation of available data on prevalence of primary and secondary infertility
[Database].
Geneva, Switzerland
:
World Health Organization
.
Yeung, W. J., Linver, M. R., & Brooks-Gunn, J. (
2002
).
How money matters for young children’s development: Parental investment and family processes
.
Child Development
,
73
,
1861
1879
. 10.1111/1467-8624.t01-1-00511.