In the Name of the Father? Fertility, Religion, and Child Naming in the Demographic Transition

Abstract

This article shows that parents reveal information about their fertility behavior through how they name their children. I arrive at this finding from a detailed examination of the net fertility of 130,000 married couples in Ireland, a country known for its historically high fertility rate, circa 1911. After stringently accounting for couples' occupation, religion, and location, I find higher fertility rates among couples who chose distinctly Catholic names and traditional names for their children, with the latter being particularly important. Exposure to towns and cities lowered net fertility and weakened preferences for traditional and Catholic names. Cumulatively, these findings highlight the role of traditional rural norms over explicitly religious influences in driving high fertility rates in Ireland. The impact of towns and cities in reducing net fertility suggests that Ireland's sluggish urbanization was a key factor in its high historical fertility rate.

Introduction

The decline of global fertility rates reflects one of the most profound behavioral shifts in modern history. From 1860 to 1940, the average number of live births per married woman in Britain fell from six to two (Szreter 1996). Despite decades of research across many countries, vibrant investigations into the fertility decline continue (Beach and Hanlon 2019; Hacker 2020; Hacker and Roberts 2017; Jaadla et al. 2020; Klüsener et al. 2019). This continued work reflects the many unanswered questions of how and why fertility rates decline, aided by the novel insights enabled by improving data sources (e.g., Ruggles et al. 2020).

The resistance of Ireland to the fertility decline is one of the mysteries of European demography. Despite being a close neighbor to Britain, Ireland had fertility rates that remained strikingly high across most of the twentieth century. Estimates from the 1960s show that Catholic women in Ireland, North and South, averaged three more children across their reproductive lifespans than women in England and Wales (Kennedy 1973). A lively, albeit still unresolved, debate has centered on the role of structural forces (Fernihough 2017; Guinnane 1997) relative to Catholic pronatalism (Compton 1976; Day 1968) in driving Ireland's peculiar demographic history.

This study provides new insight into this old discussion by examining more than 130,000 Irish couples recorded in the recently digitized 1911 Census of Ireland. The 1911 census was the last full enumeration of the population of the entire island, before the partition of Northern Ireland from what is now the Republic of Ireland. The 1911 census provides significant detail on Ireland's 4.4 million inhabitants, 73% of whom were Catholic. With 65% of the population living in rural areas, Ireland's rural population share was three times that of its neighbors, England and Wales. This study leverages these data from this important period in Irish history to provide new insight into Irish fertility patterns.

My analysis focuses on how fertility patterns relate to the signals revealed in the names that parents chose for their children. Using parents' distinctly Catholic naming preferences as a signal of religious behavior and their selection of broadly established names as a measure of traditionalism, I examine how these two connected ideational signals relate to fertility. I find that preferences for both religious names and traditional names are associated with higher fertility, but traditional naming is a particularly powerful fertility indicator. Insofar as high Irish fertility has cultural roots, my findings suggest that these roots were more broadly conservative rather than anchored in personally held religious values or attachment to Catholicism.

My examination of Irish Catholicism and demography casts light on the broader interlinkage of religion and fertility. Religiosity and pronatalist values are often correlated with fertility rates (Fernandez and Fogli 2009; Hayford and Morgan 2008; Okun 2017), but it is difficult to determine whether these associations are rooted in religious values and practice or other correlated influences (Lehrer 1996; McQuillan 2004). Given that religiosity and religious practice tend to be stronger in more conservative contexts, particularly those that place high importance on family values (Vogl and Freese 2020), it is challenging to directly distinguish religious from conservative influences. The findings from this analysis place greater weight on the role of conservatism over religiosity in shaping fertility outcomes. The enduring influence of conservatism on Irish fertility rates had roots in Ireland's impotent urbanization.

This study is novel in leveraging names to understand the ideational roots of demographic behavior. Naming can provide insight on parents' ideation, identity, tastes, and intentionality with respect to their children (Goldstein and Stecklov 2016b; Lieberson and Bell 1992; Zelinsky 1970), but until now, researchers of fertility have largely used child naming as a proxy for religious culture (Haan 2005; Hacker 1999). However, name choices reveal far more than religious affinity or embeddedness, and they provide a much wider array of psychological and sociocultural signals of relevance to demographers. In Ireland in the early twentieth century, for example, 75% of Irish boys held 1 of only 18 names, and more than one in five newborn girls were named Mary.¹ I contend that this narrow spectrum of names reflects conservative behavioral norms and insularity from external influences, which manifested in the reluctance of Irish parents to experiment with new names and to control fertility.

Theoretical Contributions on Fertility Decline

Mass declines in fertility rates are typically conceptualized in terms of adjustment and innovation (Carlsson 1966). The adjustment perspective conceives of fertility behavior in terms of the supply and demand for children, and it views fertility as principally influenced by shifts in child mortality rates, the costs of raising children, and changes in child labor laws and schooling (Easterlin and Crimmins 1985; Guinnane 2011). For industrializing economies, incentives are understood to increasingly favor child quality over quantity, thereby reducing aggregate fertility rates (Becker et al. 1990; Fernihough 2017). These perspectives imply that fertility patterns are heavily shaped by economic considerations.

The innovation perspective, in contrast, emphasizes the destabilization of long-standing customs related to marriage, intercourse, and fertility control. These changes are shaped by geographical and cultural boundaries and are spurred by the diffusion of new knowledge, ideas, and aspirations (Bongaarts and Watkins 1996; Coale 1986; Lesthaeghe 1983). The adoption of new norms and behaviors is aided by psychological shifts toward greater self-determination and away from more traditional outlooks (Cleland and Wilson 1987). In the innovation view, fertility decline is dependent on cultural transformation and diffusion, and these changes are implicitly spatial and potentially detectable in how parents name their children.²

The innovation and adjustment frameworks are helpful in understanding Irish fertility patterns. The view that the cultural influence of Catholicism can account for the high fertility rate of the Irish-born is consistent with the innovation framework. At its extreme, this perspective would view high Irish fertility rates as a product of the persisting influence of religiosity and the Catholic Church on the sexual and contraceptive behavior of Irish couples.

Economic historians, in contrast, have rejected the view that the Irish did not limit fertility in the past, arguing instead that the Irish experience is less atypical than has previously been suggested. In addition to showing substantial evidence of fertility adjustment, Ireland's late entrance to the fertility decline may be consistent with the country's modest industrialization alongside higher fertility enabled and incentivized by family farming and mass rural emigration (Guinnane 1997; Ó Gráda and Walsh 1995). Admittedly, however, structural forces fall short in explaining the particularly high fertility rates of regions in Ireland that were majority Catholic or of Irish Americans relative to other ethnicities in the United States (Fernihough 2017; Guinnane et al. 2001; Ó Gráda 1991; Ó Gráda and Duffy 1995).

The Harvard Irish Mission

Arensberg and Kimball's (1940) groundbreaking Family and Community in Ireland provides unique insight on the community forces that shaped Irish fertility. This work, emerging from the Harvard Irish Mission (1932–1936), dealt with many aspects of rural Irish life, including family, land, work, and sexual norms. The discussion of sexual norms was controversial to the extent that the Taoiseach of Ireland, Eamon de Valera, petitioned Harvard University Press to halt its publication (Byrne and O'Sullivan 2019).

When discussing sexual norms, Arensberg and Kimball (1940) emphasized the cultural structures of rural communities over religious prescription. They specifically cast marriage and childbearing in terms of the needs and pressures of the community, where young people were raised in an “atmosphere of sex and breeding” (p. 197), and “marriages are for the purpose of producing children and assuring continuity of descent and ownership” (pp. 200–201). Marriage was not built on love or desire but occurred through matchmaking, involving family negotiations and financial obligations. In sociological terms, the reproductive value of a woman was “completely integrated with (her) role in social life” (p. 208). Arensberg and Kimball therefore painted a picture of a traditional rural system in which social roles were deeply intertwined with reproduction.

Arensberg and Kimball also emphasized the role of the Church in regulating behavior. Local priests enforced pronatalist marital and sexual norms, equating “any departure from the accepted norm as a sin, a lack of religion” (p. 203). The costs of deviating from these norms were prohibitively high: anyone breaking “the taboos surrounding sex and family, or (who) comes into conflict with the church, lays himself open not only to community condemnation but also to heavy punishment” (p. 376). Furthermore, knowledge of women's reproductive cycles or self-determination in childbearing was sharply curtailed. In the account of Arensberg and Kimball, the Church reinforced prevailing traditional rural norms, social structures, and behaviors, but the Church was not the primary driving force behind them.

Data and Methods

Census Extraction

I extracted the subpopulation used for this analysis from the complete-count, nonanonymized 1901 and 1911 Censuses of Ireland, which are arguably the richest surviving data sources on the population of Ireland in the past.³ The 1901 census asked questions on residential address, name, age, sex, household relationships, occupation, marital status, and birthplace. The 1911 census is of a similar structure but asked married women additional questions, including the duration and the total number of children born alive and still living from her current marriage. My analysis relies on these retrospective fertility reports from the 1911 census. I used the 1901 census only to measure the religious distinctiveness and the traditionality of names based on earlier birth cohorts.

I applied a stringent criterion to extract my subpopulation of interest from the 1911 census. I restricted the population to women who, by 1911, were married for fewer than 15 years, married between ages 16 and 40, were coresident with their spouse in the census, and living outside group quarters. For these women, I appended their husbands' characteristics (e.g., occupation) and removed mothers who had birthed more than 14 children. When I later compared the net fertility levels of couples in the 1911 Census of Ireland with their counterparts in the United States, I imposed the same restrictions on these populations.

The 1911 census records provide three main windows into Irish fertility outcomes: the two retrospective questions on children ever born and children surviving, and the composition of the 1911 household. Given that couples may increase parity in response to the death of infants and children, I focus on the number of surviving children as my primary dependent variable (net fertility). As in other recent examinations of net fertility, I restricted the sample to women whose total number of living children matched their total number of coresident children under age 14 (Dribe et al. 2017; Dribe and Scalone 2014; Reid et al. 2020). For each couple under study, the oldest child was born within the time frame of the current marriage. Although this measure of net fertility is my preferred outcome variable, I show that my results are robust to a range of alternatives.

Unlike most complete-count census data from the United States (Ruggles et al. 2020), the Irish censuses have not been standardized and coded. Building on recent efforts (Connor 2019; Connor et al. 2011; Fernihough et al. 2015), I manually prepared the variables used in this analysis. Husband's occupation was the only variable for which I relied on a published coding scheme. To classify occupations, I built on earlier work by Fernihough et al. (2015) to map the occupational reports in the Irish censuses to the Historical International Standard of Classification of Occupations (HISCO), which I then cross-walked to its related 12-class occupational scheme (HISCLASS) (van Leeuwen et al. 2002; van Leeuwen and Maas 2011).

In addition to these couple-level characteristics, I measured geographical context using Ireland's 3,588 electoral divisions (DEDs). At the DED scale, I measured the Catholic population share, the share of workers employed as clergymen and agricultural workers, the adult literacy rate, the share of children deceased to mothers married fewer than 15 years, and the male-to-female ratio. I use these characteristics to describe the local religious, occupational, and demographic context for couples.

Name-Based Metrics

I characterized names along two dimensions: whether a name is distinctly favored by Catholics, measured with the Catholic Index and whether a name is traditional or commonly held by earlier generations, assessed via a Traditional Name Score. The Catholic Index captures the tendency for Catholics to pick names at higher rates than non-Catholics, whereas the Traditional Name Score measures the wider popularity of names in earlier generations. The reports on religion from these censuses have spurred several other recent demographic studies (Connor 2017; Fernihough et al. 2015; Henderson 2017; Reid et al. 2016).

To construct these two name-based measures, I extracted a large subpopulation of households from the 1901 Census of Ireland. I restricted the 1901 census to males and females aged older than 12 in 1901, or likely born between 1801 and 1889. Because the children born in these years are not included in the analysis of net fertility, this restriction alleviates endogeneity issues between the name-based metrics and couples' fertility outcomes.

where the numerator is the proportion of Catholics holding a given name, and the denominator is the sum of the proportion of Catholics holding a given name with the proportion of non-Catholics holding that same name. The Catholic Index ranges from 0 to 100, with 0 reflecting that no Catholics held the name in question and 100 indicating that only Catholics possessed that name. Because this is a relative measure, the index is not highly skewed by the fact that the population was majority Catholic.

where the numerator is the total number of individuals holding name i, and the denominator is the average number of people per name in the population (439). Thus, the Traditional Name Score is a direct measure of how overrepresented a name is in the population relative to the average name.

With these measures, I characterized the names of children who were aged 14 or younger in 1911 (likely born between 1897 and 1911) and whose parents meet the criteria just described. Although one might be concerned that the Catholic Index and the Traditional Name Score capture the same phenomena, the correlation coefficient is actually quite modest (.15). The variation between these two measures is supported by Table 1, which lists the 30 most common names given to sons and daughters in this analysis. The two most common names—Mary and John—score 64 and 57, respectively, on the Catholic Index. These index values imply that although Catholics disproportionately named their children Mary and John, these names were also commonly held by non-Catholics. Furthermore, the two most distinctly Catholic names—Bridget and Patrick—have Traditional Name Scores that are similar to those of distinctly non-Catholic names like William and Sarah.

A comparison of four popular girls' names further elucidates these differences. The names Bridget and Sarah have Traditional Name Scores above 100, suggesting that they had been popular over an extended period. In both cases, the average age of women named Bridget or Sarah is 34. However, the name Bridget scores substantially higher on the Catholic Index than Sarah (99 vs. 23), reflecting that 99% of women named Bridget were Catholic, but only 54% of women named Sarah were Catholic.

By contrast, Eileen and Nora are examples of names that were growing in popularity in the early twentieth century. The average age for women named Nora was 26, and the average age for those named Eileen was only 11. The concentration of these names among more recent birth cohorts explains their relatively low value on the Traditional Name Score. Notably, however, 97% of women with name Nora were Catholic (Catholic Index  =  90), whereas only 74% of women named Eileen were Catholic (Catholic Index  =  28). From these four examples, then, Bridget and Nora were more common names among Catholics, with the latter increasing in popularity; the names Sarah and Eileen were more widely spread by religion, but Eileen was a particularly up-and-coming name.

The descriptive statistics presented in Table 2 provide further insight into how child naming differed by religion. At 65.38, the average Catholic Index for children's names is highest in Catholic families and roughly twice the level of the other four denominations. Thus, Catholics were more likely than others to choose names that were more common in earlier generations of Catholics. Despite some variation across other denominations—for example, Presbyterians were less likely to choose distinctly Catholic names than Anglicans—the differences are not substantial.

Group differences in traditional naming can be assessed from both the Traditional Name Score and the proportion of parents giving their own names to their sons (patronyms) or daughters (matronyms). With the exception of Jewish households, patronymic or matronymic behavior was quite widespread. Roughly 20% of non-Jewish first sons and 10% of non-Jewish daughters were given their parents' names. An interesting minor point is that not a single Jewish parent in the study chose their own name for their first child, a behavior that likely reflects the reluctance of Ashkenazi Jews to name their children after the living.

There is substantially more variation by religion in the Traditional Name Score. Catholics chose the most common names on average for their children, and Presbyterians chose the least common names. Anglicans, Jews, and others were similarly likely to choose traditional names. The fact that Jews appear to score high on the Traditional Name Score reflects their preference for well-known Biblical names. Jewish couples favored established names that were relatively uncommon names in Ireland (e.g., Abraham, Leah, and Isaac) and more popular names in Ireland (e.g., Sarah, Annie, Joseph, and Samuel). Withstanding these differences, the most notable feature of traditional naming is that Catholics preferred names that were disproportionately chosen by earlier generations of Catholics and names that were more common in the population as a whole.

In addition to these naming practices, Table 2 provides insight into broader denominational differences in net fertility. Despite marrying at older ages (Dixon 1978), Catholics had higher net fertility than Presbyterians, Anglicans, and other denominations. Only the small number of Jewish households in the data have higher average net fertility than Catholics, a pattern that has previously been examined in detail (see Ó Gráda 2006). These statistics also highlight that because of the very low rates of religious intermarriage—less than 2% for Catholics—mixed marriages are not likely to provide substantial insight into Irish fertility behavior (Fernihough et al. 2015).

Finally, Table 2 illustrates the class-based nature of religion in Ireland. Whereas 67% of Catholic households were headed by unskilled and agricultural workers, that number was only 41% for Anglicans, 45% for Presbyterians, and as low as 4% for Jews. Illiteracy was substantially higher for Catholic and Jewish couples, but higher illiteracy among Jewish couples may reflect a lack of English fluency rather than functional illiteracy. Each of these patterns is consistent with the place-level (DED) characteristics: Catholics disproportionately lived in places with more Catholics, higher illiteracy, and greater concentrations of unskilled and agricultural workers. Catholics' greater rurality and lower socioeconomic status underscore the necessity of using a regression framework to understand fertility differences in this context.

Results

Regression of Fertility on Child Naming

The main goal of this regression analysis is to determine the relationship between naming and fertility as well as the robustness of these relationships to economic forces. I do this by testing how fertility relates to religious behavior, as signified by the choosing of distinctly Catholic names, and to more general signals of traditionalism, as revealed in parents' choice of one of their own names or of very common names for their children. I adjust for potentially correlated influences, such as local labor market conditions or occupation, using a stringent set of control variables (described later).

Table 3 presents estimates of these relationships for net fertility (columns 1–3) and parity (columns 4–6). These coefficients are estimated from a series of ordinary least squares (OLS) regression models of the following basic form:

where fertility, referring to either net fertility (total surviving) or parity (children ever born), is measured for couples j who were married for fewer than 15 years and whose surviving children were all coresident with them in 1911. The primary variables of interest are the Catholic Index and the Traditional Name Score, which are calculated based on the names of children i observed inside the household of couple j. These variables are log-transformed to adjust for skewness. Table A1 in the online appendix shows that the main results hold irrespective of whether they are modeled within an OLS or a Poisson estimation framework.

To account for the structural relationship between net fertility and exposure to intercourse, the k control variables include single-year fixed effects for age at marriage and year of marriage. I stringently account for socioeconomic differences and other correlates with these k variables, which include parental religion, husband's occupation, and DED of residence. With more than 3,000 geographical divisions for a population of less than 5 million, the DED of residence is quite a strong control by the standards of even contemporary studies. I also adjust for possible differences by migrant status using indicators for mothers' county of birth and whether a mother lived outside her county of birth in 1911.

The main coefficients of interest can therefore be interpreted as the expected change in net fertility associated with parents' choice of more Catholic or more traditional names for children born to women who were married at the same age and in the same year, and with the same religion, occupation, location, and birthplace. If Catholic or traditional names are associated with higher net fertility, we would expect $B$ > 0.

I begin by estimating the association between the average Catholic Index of children's names and net fertility while adjusting for only martial duration and age at marriage (column 1). The coefficient in column 1 is large and positive, showing that a one-unit increase in the log of the Catholic Index of children's names is associated with a 0.245 increase in net fertility.⁴ This sizable significant coefficient unambiguously implies that Catholic naming is highly correlated with net fertility at the couple level, meaning that the association between net fertility and Catholic naming is not simply an artifact of Catholics being more likely to live in communities that otherwise had high fertility.

Column 2 introduces measures of traditional naming as both a control variable and for direct comparison to the coefficient on the Catholic Index. A one-unit increase in the natural log of the Traditional Name Score is associated with around a 0.09 increase in net fertility. Naming one's first child after their mother or father is associated with similar increases in net fertility. Interestingly, the coefficient associated with naming a first daughter after her mother is around one-third larger than is the coefficient for naming one's first son after his father, perhaps because matronyms were a particularly strong signal of traditionalism. Irrespective of this difference, these coefficients reveal that parents who relied on more traditional names also tended to have larger families.

The addition of the traditional naming variables to the model has implications for the already established link between Catholicism and fertility. After these variables are added, the coefficient on the Catholic Index attenuates by more than one-half to 0.109. This implies that much of the correlation between Catholic naming and fertility is driven by the fact that many distinctly Catholic names were also quite traditional, and traditional naming was also correlated with fertility.

Column 3 introduces the large battery of control variables, including the location (DED), religion, and occupational class of the couple. After I adjust for these differences, the coefficient on the Catholic Index attenuates by another 35% to 0.07, suggesting that a sizable portion of the association between Catholic naming and fertility is rooted in broader differences along these dimensions. Somewhat surprisingly, the coefficients for traditional naming remains largely unchanged, implying that although broader economic and sociocultural factors related to location, occupation, and religion can explain substantial shares of the link between Catholic naming and fertility, they account for much less variation with respect to the association between traditional naming and fertility.

Naming aside, there are still evident fertility differences by religious denomination that withstand these control variables. Anglicans and Presbyterians have significantly lower net fertility than Catholic and Jewish couples. Mixed-marriage couples also have lower fertility, on average. These estimates imply that even though the association between fertility and religion is not fully captured in the naming decisions of parents, religious and traditional naming are tightly interlinked with fertility patterns. Because the findings for net fertility largely mirror those for parity (columns 4–6), I show the coefficients for parity but do not interpret them in any detail.

Decomposition by Religion and Urban Status

What should we infer from these robust associations between net fertility and traditional and Catholic naming? Further insight into what these relationships signify can be gained by estimating these models separately by religious denomination. If Catholic naming is a genuine signal of Catholic values or of the influence of the Catholic Church and its followers, preferences for distinctly Catholic names might be correlated with fertility irrespective of the couple's religious denomination. The presence of such a relationship for non-Catholics would imply that these couples either had Catholic leanings or were interacting with Catholics and adopting Catholic behavioral norms by doing so. There is no obvious reason to think that the relationship between traditional naming and fertility would vary by religious denomination.

Results of tests of these hypotheses are shown in Table 4, which presents estimates from a set of models similar to those in Table 3 but in which the observations are split by religion. The first notable finding here is that the estimated effect of traditional naming on fertility is highly consistent by religious denomination: a log increase in the Traditional Name Score of children is generally associated with a 0.09 to 0.10 increase in net fertility. Despite some variation in the relative size of the coefficients, the association between patronymic and matronymic naming and fertility is also consistent in direction across the different splits, with more traditional naming being consistently and positively correlated with net fertility.

The consistency of the traditional naming coefficients across these models stands in stark relief to those of the Catholic Index. Among Catholics, a log increase in the average Catholic Index of children's names is associated with a 0.157 increase in net fertility. There is, however, no notable association between Catholic naming and net fertility for any other religious denomination: in every case, the coefficient is close to 0 and not statistically significant. That is, non-Catholics who preferred distinctly Catholic names for their children do not show any elevation in fertility. This finding implies that the wider link between Catholic name choices and fertility in the Irish population was driven exclusively by the behaviors of Catholics.

There are two potential explanations for why Catholic naming is associated with higher fertility for only Catholics. One possibility is that Catholics who revealed strong preferences for Catholic names were particularly religious and strongly adhered to the naming conventions and pronatalist teachings of the Church. Alternatively, the choice of Catholic names among Catholics could simply be another form of conservative behavior or traditionalism, wherein these names were well established and highly accessible for Catholics. In this case, non-Catholics who preferred distinctly Catholic names could, in fact, be revealing their openness and willingness to experiment by selecting names from outside their immediate social and cultural circles. For now, however, these hypotheses are purely speculative.

Explanatory Power of Naming

In assessing the importance of naming with respect to fertility patterns, I am interested in not only effect sizes but also the ability to account for variation in fertility in the population. Table 5 presents the R-squared values for seven regression models, where net fertility is the dependent variable, and the models differ by the inclusion of different variable blocks. The starting point in this comparison is to note that the full regression model (see column 3, Table 3) has an R-squared of roughly .49; jointly, the main variables of interest explain approximately 49% of the variation in net fertility. Most of this variation is explained by the exposure indicators of age at marriage and marital duration.

However, the most revealing aspect of this analysis is the contrast between traditional naming and religion. Whereas the religious denomination and religious naming patterns of the couple have an adjusted R-squared value of only .025, the three measures of traditional naming have an adjusted R-squared value of .049. Thus, compared with religion, the traditional naming measures explain almost twice as much variation in net fertility. Moreover, the traditional naming block has a higher adjusted R-squared value than either the location and birthplace variables or variables relating to socioeconomic status. Thus, these traditional naming measures have greater power in accounting for fertility differences than classically studied proximate determinants, such as location, socioeconomic status, or religion.

Robustness Checks on Regression Analyses

I demonstrate the robustness of these regression estimates through a series of alternate specifications. These checks are estimated through four regression models in Table A1 in the online appendix, where the specification largely mirrors that shown in Eq. (3).

First, my main analytical specification characterizes naming based on all valid children in the household. This decision was motivated by the fact that parent's preferences for particular types of names will be measured more precisely when observed across multiple children, particularly when first children are more likely to receive patronyms or matronyms. However, the use of all children in the household could introduce an endogeneity bias if parents' preferences for particular names change based on their total number of children (Goldstein and Stecklov 2016b). To ensure that such circularity is not driving my findings, I estimate the main specification but with the variables of interest measured only from the name of the firstborn child in each household. Even after imposing this restriction, I find significant positive relationships for the selection of Catholic and traditional names on net fertility (see panel A of Table A1). Although the effect sizes are smaller than before, the main relationships hold.

Second, net fertility is a count variable, and such measures can be sensitive to the estimation strategy. Although I relied on OLS models, the skewness and potential nonnormality of the residuals on net fertility may lead one to prefer a Poisson framework. Panel B of Table A1 displays results based on the Poisson framework, which are again very similar to the main regression models and would not lead to any difference in inference. Thus, I prefer the OLS to a Poisson model because of its more accessible interpretation and computational efficiency, particularly given the large number of fixed effects included in these models.

Third, I measure net fertility using children in the household up to the age of 14. This could introduce a bias if children selectively left home at younger ages, a process that I cannot observe. To ensure that such selectivity is not severely biasing my estimates, I estimate the main specification but restrict the population to recently married couples (married fewer than five years). Again, all relationships match those from the main analyses.

Finally, I constructed the Catholic Index by linking first names to the reported religion of adults in the 1901 census. The construction of the Catholic Index deviates from Hacker (1999), who instead measured religiosity using names from the Bible. I decided against using Biblical names because the religious norms of Catholics in Ireland differ from those of Catholics elsewhere, and the Irish census provides an opportunity to measure this distinctiveness more directly. Panel D of Table A1 presents estimates from a model in which the Catholic Index for the names of children is replaced with the proportion of children in the household possessing a Biblical name. This is the only model that produces a substantially different result: Biblical naming is negatively associated with net fertility in Ireland. The most likely explanation for this difference is that Biblical names diverge from common Irish Catholic names. For example, Mary is a Biblical name, but Patrick is not. Moreover, several higher-status names in Ireland (e.g., Sarah and Elizabeth) would be considered to be Biblical yet were not widely held by Irish Catholics. This finding raises further doubts over the role of formal theology as the driver of Irish fertility behavior.

Contextualizing Irish Fertility Patterns

The preceding regression analyses reveal strong relationships between ideational and cultural signals and fertility outcomes. To better understand the source of these relationships, I provide two final descriptive analyses on the spatial patterning of fertility and naming patterns within Ireland and a comparison of net fertility between Ireland and the United States. The link between naming and fertility could largely be idiosyncratic and attributable to random personality differences related to risk aversion, openness, or creativity. If this was the case, strong spatial patterns in these behaviors might not be expected. On the other hand, if contextual forces are acting on fertility and naming behaviors, these behaviors might be expected to cluster in particular places.

Figure 1 maps net fertility and naming decisions across Ireland in the early twentieth century. These surfaces, based on the centroids of DEDs, show the average values of children's names on the Catholic Index (panel a) and the Traditional Name Score (panel b), along with the average net fertility levels of couples (panel c). These estimates are derived from the subset of the population described in the aforementioned census extraction.

Panel a reveals that couples living in the largely rural and Catholic communities throughout the West and Southwest of Ireland were most likely to choose Catholic names. By contrast, parents living near major cities, such as Dublin and Belfast, or in the counties to the North more generally were less likely to give their children Catholic names. These northern counties and major urban centers were not only more developed economically but also less Catholic, on average.

The selection of traditional names shown in panel b follows a somewhat similar pattern to those for Catholic naming. Less traditional name choices were common in the Northern counties of Ireland and in proximity to Dublin and Belfast. Furthermore, and again in line with Catholic naming, couples in the Western counties of Ireland were more likely to choose traditional names for children. One notable deviation is that there seems to have been a preference for less traditional names in the Southwest of the country, particularly around the counties of Cork and Kerry. Except for this discrepancy, however, there is substantial overlap in the geography of traditional and Catholic child naming.

Taking these panels in combination with panel c reveals strong spatial correlations between naming patterns and fertility. Couples in urban areas and the eastern regions of Ireland exhibit lower net fertility, on average. Net fertility is high, however, in the same Western counties that favored more Catholic and traditional names. Given that naming is primarily a product of culture, these patterns suggest that local contexts that fostered deviance from traditional naming also encouraged greater fertility control.

Perhaps these naming and fertility behaviors are a reflection of migrant self-selection or sorting across places. More adventurous people may have moved to cities, and the personalities and motivations of migrants may have also predisposed them to restrict their fertility. Recall, however, that the association between naming and net fertility withstood stringent controls for location, migration, and occupation, suggesting that the movement of people with particular personalities is not likely to be driving these patterns. The more plausible explanation is that local contexts reshaped fertility and naming behaviors.

What contextual influences drove these patterns? Some insight into this question can be gleaned from Figure A1, which shows the correlation coefficients of the characteristics of places with the naming and fertility measures. These correlation coefficients for place characteristics move in a similar direction irrespective of whether they refer to naming or fertility outcomes. The presence of farming, Catholic population shares, and male-to-female ratios are positively correlated with traditional naming, Catholic naming, and net fertility. By contrast, literacy and in-migration rates—indicators of urbanization—are negatively correlated with these behaviors. The salient contextual distinction therefore appears to be between urban areas and areas that were rural and majority Catholic.

Table 6 takes one final analytical view at this issue by leveraging the complete-count 1910 U.S. Census from the Integrated Public Use Microdata Series (Ruggles et al. 2020) to examine fertility and naming patterns across comparable couples in Ireland and abroad. This comparison offers a sense of how couples may have adjusted their fertility behavior with migration and exposure to the United States while also providing a more general benchmark for fertility patterns in Ireland.

Table 6 reveals several striking patterns in terms of both fertility and naming. The most notable finding is the progressive decline of net fertility as one figuratively moves farther from rural Catholic Ireland. Rural Catholic couples in Ireland who were married fewer than 15 years had averaged 3.13 children by 1911. Consistent with the maps shown in Figure 1, rural non-Catholics and urban dwellers in Ireland had substantially lower net fertility, in the range of 2.5 to 2.8 children. Thus, separately and in combination, being Catholic or living in an urban area was associated with substantial reductions in fertility.

The fertility outcomes of the Irish-born in the United States are also revealing. Given that emigrants from Ireland to the United States in this period were heavily drawn from Irish regions that were largely Catholic and rural and had high fertility rates (Connor 2019), their fertility outcomes in the United States are of particular relevance. Couples in the United States in which both husband and wife were Irish-born had an average net fertility level of 2.64. Average net fertility was lower again at 2.33 for Irish-born women who married men born outside Ireland. Although these reductions are sizable, Irish-born women in the United States still exhibited higher net fertility than their U.S.-born or foreign-born counterparts in the United States. High Irish fertility thus persisted in some measure in the United States, but the emigrants themselves exhibited lower fertility levels than their counterparts who stayed in Ireland.

Importantly, these fertility patterns roughly follow similar declines in couples' preference for traditional and distinctly Catholic Irish names. The Traditional Name Score of 367 is at its highest for rural-dwelling Catholic couples, compared with 320 for urban-dwelling Catholics. Although non-Catholic couples were generally less likely to choose traditional names, the average Traditional Name Score for non-Catholic couples is also larger for rural areas (220) than for urban areas (196). Moreover, Irish-born parents in the United States have a Traditional Name Score of 317, a very similar value to that of urban-dwelling Catholics in Ireland. There are similarly sized reductions in Catholic naming across these different cuts of the data. These differences underscore that the correlation between fertility and naming even holds when one considers the Irish in the United States. The fact that the Traditional Name Score is only 212 for Irish-born women in the United States who married non-Irish men could reflect a cultural effect of out-marriage, but we cannot be sure because out-marriage was also a highly selective process.

Although these data provide no direct evidence on the psychological processes governing fertility and naming decisions, Arensberg and Kimball (1940) did provide some speculative observational insight. They argued for the prominence of traditionalism with respect to marital and sexual norms in rural communities and the disruption of these norms through urbanization. Urban life was observed to produce substantial psychological and cultural shifts as rural values weakened with the “increasing acceptance of urban values and behavior” (p. 376). In particular, parents' aspirations were seen to shift from farming and land to the “future careers of sons and the marriage prospects of daughters” (p. 376). My findings suggest that this shift in parents' aspirations for their children may be observed in how they named their children. Specifically, names linked to continuity, descent, and tradition were increasingly replaced by names of a more modern and urban flavor, signaling the transition from the farm to the city.

With respect to Catholicism and fertility, my findings are consistent with demographic theory. Demographers emphasize that community influences regarding gender inequality, family, and sexual behavior, as well as the Church's ability to enforce behavior, play a greater role in shaping fertility than do theology or religious culture (Goldscheider and Mosher 1991; McQuillan 2004; Yeatman and Trinitapoli 2008). This view is consistent with the particularly pronounced fertility levels of closely knit rural Catholic communities that I have shown here and with the accounts of Arensberg and Kimball (1940). My findings also highlight the greater potential salience of traditionalism over Catholicism in explaining high historical fertility rates in Ireland.

Conclusion

The global decline of fertility reflects one of the most profound behavioral shifts in modern human history. Scholars have argued that these changes were rooted in mass shifts in values, attitudes, and deliberate birth control (Carlsson 1966), as well as cultural changes that shifted people away from traditional and religious constructs and toward self-determination (Bongaarts and Watkins 1996; Lesthaeghe and Surkyn 1988; Spolaore and Wacziarg 2019). I provide new and consistent evidence on this issue from Ireland, a historical outlier in its slow fertility decline. My analysis reveals a strong and highly robust relationship between traditionalism (inferred from naming) and fertility. Urban environments appear to have played a key role in attenuating net fertility and traditional naming patterns, perhaps because cities enabled new information and aspirations to spread, which had further implications for consequential behaviors like fertility control and less consequential behaviors like child naming.

These findings are notable because explanations of high fertility in Ireland are often cast between the influence of the Roman Catholic Church or the effect of more voluntaristic decisions guided by economic circumstances. My findings complement both views but do not support either one in isolation. My main contribution is in revealing strong links between cultural signals and fertility, but these cultural signals appear to have been rooted in traditionalism and conservatism rather than explicitly religious predilections. The shift away from traditional norms was tightly woven with forces of economic development and urbanization. Urbanization brought about economic pressures and sociocultural shifts that disrupted rural and religious norms and ultimately lowered fertility. Conversely, the coupling of Ireland's sluggish urbanization with persistently high fertility in more traditional rural communities should be at the center of any account of the country's reluctant participation in the European fertility decline.

Finally, this analysis was enabled by the growing availability of historical census data and new methods in the analysis of child naming. Complementing other work that leveraged historical microdata to study behavior and fertility (Guinnane et al. 2006; Jaadla et al. 2020; Klüsener et al. 2019), this study found that naming is a compelling and increasingly analyzable behavior with respect to demographic change. The value of using names to understand population patterns is underscored by prior analyses of how naming relates to fertility behavior (Goldstein and Stecklov 2016a; Hacker 1999), immigrant assimilation (Abramitzky, Boustan, and Connor 2020; Abramitzky, Boustan, and Eriksson 2020; Goldstein and Stecklov 2016a), and mortality (Cook et al. 2016), as well as how surnames relate to social mobility (Clark et al. 2015; Connor 2020). Given that demographers have not yet widely leveraged the insights revealed in child naming, this will continue to be a fruitful area for future research.

Acknowledgments

For their feedback and influential work, I express my sincere thanks to Connor Sheehan, Alan Fernihough, Cormac Ó Gráda, Kevin McHugh, David Hacker, Avril Brandon, Josh Goldstein, Guy Stecklov, Leah Boustan, Jessica Fraker, Amy Frazier, Angela Connor, Danny Connor, and participants at the 2019 meeting of the Social Science History Association, the 2021 meeting of the Population Association of America, the 2021 Conference of Irish Geographers, the Theory, History & Development group in the Economics Department at the University of California Irvine, and the editors and three excellent reviewers at Demography. The original idea for this article emerged from a series of blog posts curated by Phillip Lawton on the geography of Dun Laoghaire and was further enhanced by an exciting intellectual conversation with Sister Catherine from California, a fellow passenger on flight AA1971 to Chicago.

Notes

References