Abstract

Economic and social development are closely linked with fertility. Several studies have shown that the relationship follows an inverse J-shape: the association is negative at low and intermediate levels of development and reverses to become positive at high development levels. However, more recent research building on subnational and U.S. data found only mixed evidence for the inverse J-shape. In this article, we draw on subnational data on development and fertility in the U.S. states between 1969 and 2018 to examine the relationship between development and fertility. Using a longitudinal approach and addressing several criticisms of the fertility reversal hypothesis, our results support the inverse J-shaped pattern under most model specifications. However, this pattern might have vanished since the 2007–2008 financial crisis. Our findings provide insights into the mechanisms that link development and fertility, showing that gender equality and economic uncertainty mediate the relationship between development and fertility.

Introduction

Are economic and social development and fertility negatively or positively associated? From a theoretical perspective, proponents of the demographic transition model have long argued that development increases the costs of having children, improves the means to control childbearing, and gives rise to life goals that conflict with fertility (Davis 1945; Notestein 1945; Lesthaeghe and van de Kaa 1986; van de Kaa 1987). Thus, fertility and development should have a negative association. This theory accurately describes the lowest-low fertility observed in high-income countries, including the drop in fertility below the replacement level in the United States (Kohler et al. 2002; Ruggles 2015). However, this model was challenged when a study found reversals of fertility declines (Myrskylä et al. 2009). The association between fertility and development was shown to follow an inverse J-shaped pattern, with a negative association at low and medium levels of development and a positive association at higher levels of development.

The initial evidence on the inverse J-shape and reversals of fertility declines spawned a rich, partly critical body of literature that generated mixed evidence. Several studies replicated the original findings at the national and subnational levels and argued that gender attitudes, late childbearing, and family policies have been key contributors to recent fertility increases (Anderson and Kohler 2015; Fox et al. 2019; Kolk 2019; Luci-Greulich and Thévenon 2014; Mavropoulos and Panagiotidis 2021; Myrskylä et al. 2011). However, other studies failed to find an inverse J-shaped relationship between fertility and development (Gaddy 2021; Harknett et al. 2014; Ryabov 2015). For instance, Ryabov (2015) did not find a J-shaped relationship in an analysis of cross-sectional data for U.S. counties. One potential explanation for why the inverse J-shape could be a spurious finding cites measurement errors. Fertility is often measured using the total fertility rate (TFR), which suffers from tempo distortions (Bongaarts and Sobotka 2012), and development is captured through the Human Development Index (HDI), which is known to be imprecise (Ghislandi et al. 2019).

In this study, we use data on the U.S. states and the District of Columbia for 1969–2018 to reexamine the relationship between development and fertility. To address the criticisms raised in the literature, we use several measures of fertility and development and apply several panel regression approaches. For instance, we test the inverse J-shape with three measures of fertility: the TFR, a tempo-adjusted TFR, and the TFR for men. The tempo-adjusted TFR removes distortions of fertility levels caused by postponement (Bongaarts and Watkins 1996), whereas the TFR for men can differ substantially from the TFR for women because of birth squeezes caused by migration and cohort size (Dudel and Klüsener 2021). We also provide insights into the potential mechanisms behind the association between development and fertility, including gender relations and economic uncertainty. The results are fully reproducible, and all code is available online.1 The data can be obtained from the National Bureau of Economic Research (NBER), the United States Mortality DataBase (USMDB), and the Global Data Lab.

Studying the relationship between development and fertility at the subnational level is crucial for the discussion of fertility decline reversals. Income, living standards, and well-being vary considerably within countries (for the United States, see Porter and Purser 2008; Scherbov and Gietel-Basten 2020). Moreover, within-country research designs are promising because they are robust to common sources of error in cross-country research. Empirical investigations at the country level can be biased by unobserved heterogeneity due to cultural and institutional differences that are difficult to control for. In addition, cultural differences tend to be less pronounced, and the institutional setup shows less variation within countries than between countries.

The United States is an interesting case for studying the relationship between development and fertility for several reasons. Evidence suggests that the United States has experienced a pronounced reversal of the fertility decline at a comparatively high level of fertility (Luci-Greulich and Thévenon 2014), making it a somewhat special case among high-income countries. Furthermore, because of considerable variation in development and fertility trends at the subnational level and over time, the United States is ideally suited for conducting subnational analyses (Scherbov and Gietel-Basten 2020). Finally, given the inconsistent results from previous research on fertility decline reversals in the United States, studying the country in more depth might help explain these inconsistencies (Porter 2017; Ryabov 2015).

This article contributes to the literature in several ways. We provide the first longitudinal analysis of U.S. fertility decline reversals using data at the subnational level for all U.S. states across 50 years. Moreover, we address general criticisms of the reversal hypothesis raised in the literature by using several indicators of development and fertility and by conducting several robustness checks. Our analyses reconcile inconsistent findings regarding fertility reversals at the subnational level and provide new insights into the potential drivers of the reversals. Furthermore, we examine how the association between development and fertility has changed in recent years, during a period not covered by most research. We find that development and fertility were not associated during the post-recession period.

Background

Fertility and Development at the National Level

Several theoretical approaches argue that development and fertility are linked. Here, we broadly consider development to be the material conditions, wealth, technological progress, social equality, and public support in a spatially bounded area that have an impact on individuals’ well-being (Sen 1998). Thus, the concept highlights the importance of contextual characteristics for individuals’ lives.

Demographic transition theory hypothesizes a negative connection between fertility and development, starting with the observation that for much of the twentieth century, fertility declined with increasing development. The first demographic transition theory asserts that modernization and associated increases in wealth, expanded education, and improved survival are linked to fertility reductions (Bryant 2007; Davis 1945; Notestein 1945). Once the demographic transition is completed, fertility stabilizes in long-term equilibrium, with mortality around the replacement level (Casterline 2003:211).2

van de Kaa (1987) and Lesthaeghe and van de Kaa (1987) suggested that the first demographic transition is followed by a second demographic transition, which is characterized by increasing nonmarital cohabitation and the emergence of lowest-low fertility. The underlying driver of the second demographic transition is individualization, which is itself related to development because increases in wealth and changes in the occupational structure are assumed to spur value change (Beck 1986/1992; Inglehart 1977). Individualization leads to the emergence of competing life goals and the weakening of traditional institutions, which lead to increased nonmarital cohabitation rates, fertility reductions, and high childlessness levels. Whether individualization enables individuals to reach their life goals is a subject of discussion (Mills 2007; Smart and Shipman 2004; Worts et al. 2013). In its original formulation, the theory predicts that individualization entrenches fertility at low levels (Lesthaeghe and van de Kaa 1986; van de Kaa 1987).

Beyond the macro-level theories, Becker's (1981) household economics framework provides a micro-level foundation that offers an explanation for the negative relationship between development and fertility. This perspective assumes that development changes the structure of society by expanding educational participation and increasing wages. In response to increasing education and wage levels, individuals shift their orientation from the quantity of children to the quality of children because they can invest more in each child (Becker and Lewis 1973; Becker and Tomes 1976). Thus, women are inclined to have fewer children because of monetary constraints. Beyond describing the quality and quantity trade-off, Becker (1981) showed that as women's educational and employment levels increase, elevated opportunity costs reduce their fertility. Therefore, as educational levels and wages increase, more individuals are expected to opt out of forming a family or of having children altogether.

The authors of the fertility-trap hypothesis (Lutz et al. 2006) argued that once fertility has fallen to lowest-low levels, fertility and development might be decoupled, and fertility would then remain at low levels. The key mechanism of this trap is the decreasing cohort size: if individuals grow up in an environment with low fertility and relatively few children, their fertility aspirations will be affected accordingly. Moreover, low fertility can put pressure on welfare states through the accelerated aging of the population. As income levels and welfare protections of younger cohorts decrease, fertility might become entrenched at low levels. In addition, these authors have argued that the detrimental effects of decreasing net income on fertility are reinforced by increasing economic aspirations resulting from past economic growth and from small sibling numbers, given that siblings can limit the attention and resources each child receives.

Whereas the fertility-trap hypothesis postulates that fertility could become entrenched at low levels, McDonald (2000), Goldscheider et al. (2015), and Esping-Andersen and Billari (2015) argued that progress in gender equality might lead to increases in fertility. They observed that the later stages of the gender revolution might offset the suppressing effect of work–family conflict, thereby removing a mechanism underlying the negative association between development and fertility. In the first stage of the gender revolution, women's increasing participation in education and paid work reinforced the demographic transition because it empowered women to make individual fertility decisions while intensifying career–family conflict. However, in the second stage of the gender revolution, gender equity spread to the individual sphere, leading to a more equal distribution of power and roles within the household and facilitating the reconciliation of work and family. This framework essentially used Becker's opportunity-cost argument (1981) to explain the first stage and argued that the institutional context could mitigate these constraints.

Beyond the institutional context, development might spur fertility increases by transforming the economy into a more childbearing-friendly environment. First, the modernization of the economy shifts employment away from routine and manual tasks and toward service jobs, perhaps providing more flexible work arrangements that can help ease work–family conflicts. For instance, having flexible working hours might enable parents to align their working schedule with childcare opening hours, and the option to work from home might save parents commuting time (Fox et al. 2019). Second, economic development plays a crucial role in the globalized market because it improves individuals’ competitive positions and thus their future prospects (Mills et al. 2006). Working in a competitive sector might provide individuals with the economic stability they require for making long-term commitments, such as raising children (Adserà 2004; Hofmann and Hohmeyer 2013). However, the claim that economic restructuring and flexibility have positive effects on fertility has been contested. Labor market flexibility might lead to higher levels of employment uncertainty, which could inhibit childbearing because couples could be inclined to postpone life-changing commitments if they see the future as unpredictable (Comolli 2017, 2021; Vignoli et al. 2020). Moreover, the positive effects of globalized markets have been questioned, given that the decline in manufacturing jobs has been associated with decreases in fertility in the United States (Seltzer 2019).

Providing empirical evidence on the reversal of this association, Myrskylä et al. (2009) found that fertility declines reverse at high development levels. They examined the relationship between fertility and development using data from 140 countries on the TFR and the HDI and uncovered an inverted J-shaped relationship between development and fertility. In line with the aforementioned theoretical perspectives, they observed that fertility fell steadily from high levels at low development stages to historical lows. However, they also observed that recent increases in development have been accompanied by increases in fertility. In line with gender revolution theory (Esping-Andersen and Billari 2015; Goldscheider et al. 2015; McDonald 2000), they argued that this reversal is attributable to gender and social equality, the introduction of more effective family policies, and increases in living standards and labor market flexibility. These trends, which are associated with economic and societal development, have facilitated childbearing, making it easier for couples to achieve their childbearing intentions.

Several studies that further examined the mechanisms behind the reversal of fertility declines have reproduced the Myrskylä et al. (2009) findings (Luci-Greulich and Thévenon 2014; Mavropoulos and Panagiotidis 2021; Myrskylä et al. 2011). These studies found that changes in gender attitudes and family policies can lead to higher fertility at the highest-high development levels (Myrskylä et al. 2011). However, they also showed that whether fertility declines reverse depends on women's labor market participation, which points to the importance of policies that support the reconciliation of work and family (Luci-Greulich and Thévenon 2014). One study found that fertility decline reversals and the conditions under which reversals occur vary across countries, thus highlighting the role of contextual factors in fertility, including women's employment and culture (Lacalle-Calderon et al. 2017).

Fertility and Development at the Subnational Level in the United States

The research findings on fertility decline reversals spurred a debate about the mechanisms contributing to fertility increases and whether they are limited to nation-states. For example, development levels can vary considerably within countries, which might affect subnational fertility levels (for the United States, see Porter and Purser 2008; Scherbov and Gietel-Basten 2020). We discuss several mechanisms that might cause fertility levels to increase at high development levels in some subnational units while remaining low in others. Given that regional differences in development levels can be large, this relationship is likely to be of interest to both policymakers and academics.

In regions with lower development levels, relatively large shares of the population are still employed in routine task-intensive activities, which face pressure from globalization forces and technological change (Acemoglu and Autor 2010; Mills et al. 2006). As a consequence, the working population might experience economic uncertainty, which could lead them to postpone or forgo childbearing, given that economic uncertainty is negatively related to fertility (Adserà 2004; Hofmann and Hohmeyer 2013). For instance, in the United States, state-level economic performance, as measured by the unemployment rate, is negatively related to nonmarital childbearing among groups with low socioeconomic status (Schneider and Hastings 2015).

The aforementioned unequal spatial distribution of industries might contribute to fertility increases in highly developed regions dependent on the extent to which these industries allow for reconciling family and employment. Althoff et al. (2022) showed that in the United States, progress in workplace flexibility is not universal. Instead, they found that the share of remote work in each region depends on the region's economic structure and population density and is particularly high in urban regions with a high proportion of jobs in the service sector. Workplace arrangements play an important role in U.S. fertility: having flexible working hours and the option to work from home might facilitate childbearing among working women, given the high costs of childcare (Fox et al. 2019). It thus appears that in contexts with high development levels, eliminating an obstacle to childbearing has the potential to increase fertility.

Beyond these direct mechanisms, development might interact with migration in producing fertility increases. More developed areas are often urban and technological centers that attract large numbers of international migrants seeking employment opportunities (de Haas et al. 2020). For instance, U.S. states along the East and West Coasts and in the South, which are also among the leaders in terms of development levels, have larger shares of migrants than other states (Alexander et al. 2022). In the period immediately after their arrival, international migrants from high-fertility sending countries tend to have higher fertility. Moreover, migrants often postpone childbearing until they have settled in the host country (Lichter et al. 2012; Milewski 2010). Thus, the arrival of migrants might boost fertility levels in more developed areas.

Empirical evidence suggests fertility decline reversals at the subnational level in Europe and the United States. Fox et al. (2019) analyzed data at the NUTS 2 level (subnational areas with a population between 800,000 and 3 million according to the European Union's territorial classification) for 20 European countries subdivided into 256 regions for 1990–2012. They measured development using employee compensation—an indicator of household income—and measured fertility using the TFR and the tempo-adjusted TFR. They concluded that fertility declines have reversed at the subnational level: between 1990 and 2012, the relationship between fertility and development became less negative or even positive in most of the 20 countries studied. The exceptions were Finland, West Germany, the United Kingdom, and France, where the relationship became more negative. These findings held even after the researchers accounted for tempo distortions by using the tempo-adjusted fertility rate.

Empirical studies investigating this relationship at the U.S. subnational level produced mixed results, perhaps because of their cross-sectional approaches. Ryabov (2015) found no evidence of a fertility reversal among U.S. counties with very high development levels, concluding that the combination of the second demographic transition and high human development levels resulted in persistent low fertility. By contrast, a study by Porter (2017) also used county-level data and reproduced the inverse J-shaped association. A potential explanation for these discrepancies is that they explored fertility and development with different measures and models. Moreover, the cross-sectional approaches applied in these two studies relied on strong assumptions for assessing the causal relationship related to unobserved heterogeneity (Firebaugh 2018; Wooldridge 2002). Hence, using longitudinal data might help resolve the inconsistencies in earlier research findings.

Critiques of the Reversal Hypothesis

The reversal hypothesis has stimulated a debate among scholars, some of whom have criticized its claims. In particular, some have noted the impact of tempo effects on fertility decline reversal. Bongaarts and Sobotka (2012) suggested that recent increases in fertility are attributable to cohort tempo fertility recuperation rather than an increase in the quantum of fertility caused by increasing development. Empirical support for this critique comes from two studies that aimed to replicate the inverse J-shaped relationship but found no evidence of fertility decline reversal in contexts with highest-high development levels (Gaddy 2021; Harttgen and Vollmer 2014). They argued that the J-shape hypothesis held for only a short period when fertility postponement ended.

Another critique of the reversal hypothesis related the fertility decline reversals to Simpson's paradox (Lesthaeghe 2020). Several studies have shown that even when an inverse J-shaped relationship between fertility and development is observed at the national level, it might vanish when studied within country groups (Lesthaeghe 2020; Lesthaeghe and Permanyer 2014; Rindfuss et al. 2016). These authors suggested that national idiosyncrasies of the Nordic and Anglo-Saxon countries—with the former having supportive social policies and the latter having flexible labor markets—entirely explain the inverse J-shaped relationship and that the positive association is therefore a data artifact rather than a causal relationship.

Moreover, some have criticized the HDI, which several studies used as an indicator for development. The United Nations–provided HDI is based on four indicators: mean years of schooling, expected years of schooling, life expectancy at birth, and gross national income per capita. The main criticisms of the HDI are that it is only a crude indicator of development and thus ignores many aspects relevant to development, its components suffer from measurement error, the estimation method has been revised repetitively, and the weight of each component is not well justified (Gaddy 2021; Ghislandi et al. 2019; Harttgen and Vollmer 2014; Scherbov and Gietel-Basten 2020).

Data and Methods

Overview

In this study, we aim to test the J-shape hypothesis in the U.S. states for 1969–2018. We examine the relationship between development and fertility using longitudinal data. Because we seek to contribute to the ongoing debate, we consider the critiques outlined earlier and run several robustness checks. We control for tempo distortions. We also account for measurement error by running several analyses using alternative indicators of fertility and development. The indicators, their annual coverage, and their data sources are summarized in Table 1. Moreover, we apply several regression techniques in our robustness checks to assess the findings’ model dependency.

Fertility Indicators

Our main fertility indicator is the TFR for 1969–2018 for all 50 U.S. states and the District of Columbia. For 1969–2004, we calculate the state-level TFR from vital statistics data provided by the National Center for Health Statistics (2022) and from population counts provided by the NBER (2023). For 2005–2018, we derive the state-level TFR from the annual birth collection published by the Centers for Disease Control and Prevention (2023).

We perform robustness checks using two alternative fertility indicators that will show the extent to which the results depend on the fertility measurement. First, we calculate the tempo-adjusted TFR following Bongaarts and Feeney (1998:278) to account for fertility postponement, which could distort the results, as argued in the literature (Bongaarts and Sobotka 2012). The adjusted TFR is calculated from the female perspective for 1969–2004 using the data provided by the NBER (2023). The time series is shorter because we lack access to state-level birth counts by parity for later years. We use the female midyear population as exposure for each parity. We combine all parities above five into one category. Second, we use the male TFR, counting births by the father's age in the numerator and using the midyear population for men instead of the exposure counterpart for women as the denominator. Although the male TFR is closely linked to the female TFR, the two can differ3 (Dudel and Klüsener 2021; Schoumaker 2019). These differences might be attributable to imbalances in the population size for men relative to that for women, which can be caused by gender-selective migration or changing cohort sizes. The male TFR is also based on the data provided by the NBER (also see Dudel and Klüsener 2019).

Development Indicators

Our main indicator for human development is the Human Life Indicator (HLI) (Ghislandi et al. 2019). It captures the average length of life as well as the lifespan distribution and is defined by the geometric average of the age-at-death distribution. Life expectancy is central to development because “mortality information has (1) intrinsic importance (since a longer life is valued in itself), (2) enabling significance (since being alive is a necessary condition for our capabilities), and (3) associative relevance (since many other valuable achievements relate-negatively-to mortality rates)” (Sen 1998:22). In addition, the HLI captures lifespan inequality, which is related to development because it reflects how societies organize health care, insurance, pensions, and other social policies and programs (van Raalte et al. 2018:1002). Furthermore, lifespan inequality has an inherent dimension of social equality. In contrast to the HDI (see below), which is available only from 1990 onward, the HLI can be calculated at the state level from 1969 through 2018. This availability is our primary reason for using the HLI in our main analysis. Moreover, in contrast to the data used for the HDI components, the life table data used as input for the HLI are very reliable and robust (Ghislandi et al. 2019). We obtain the life tables from the USMDB (2022).

We use three alternative measures of development— the HDI, life expectancy at birth (e0), and the Gini coefficient for the lifespan distribution—to show whether the measurement of development affects the substantive findings. We present the results for the HDI and life expectancy alongside those for our main indicator. The HLI and e0 are estimated from female life tables when the outcome is the female (tempo-adjusted) TFR, and life tables for men are used when the outcome is the male TFR. Results for the Gini coefficient are briefly covered in the Discussion and are fully reported in the online appendix.

Further Control Variables

In our main analysis, we include only development and fertility in the regression models. However, trends in fertility might be driven by factors other than development or by factors that mediate the impact of development. To account for these mechanisms, we conduct robustness checks that include additional control variables. First, in some analyses, we include the proportion of jobs in the service sector, which accounts for structural economic change (Ruggles 2015; Seltzer 2019). Furthermore, in some instances, we include the annual state unemployment rate to account for economic conditions and shocks, such as the 2007–2008 financial crisis, which might have driven some of the observed fertility trends (Comolli 2021; Schneider and Hastings 2015). The results might also be confounded by heterogeneous trends in gender equality, as suggested by the gender revolution theory (Esping-Andersen and Billari 2015; McDonald 2000). Hence, we include a proxy for gender equality in norms and household roles: the mean age difference between parents. The parental age difference is a good indicator for gender equality in the domestic sphere because it affects the bargaining power within the relationship (Carmichael 2011; Dudel et al. 2023; Presser 1975).

As stated earlier, tempo distortions of the TFR are a major threat to the identification of the effect of development on fertility. Beyond using the tempo-adjusted TFR, which is available only for 1969–2004, we expand the time series to 2018 by using the mean age at childbearing as a control variable in robustness checks. Guided by the models presented in the literature, we use the Myrskylä et al. (2011) specification and include the first and second differences of the mean age at childbearing as controls. We also estimate an alternative specification suggested by Luci-Greulich and Thévenon (2014), including the (undifferentiated) mean age at childbearing as a linear and a squared term because postponement might have nonlinear effects on the TFR.

A description of the data is displayed in Table 1. For each variable, the table shows the total number of state-year observations, the years covered, and the minimum and maximum values, with the latter indicating the state to which the value refers.

Methods

For our main analysis, we use a fixed-effects individual slope (FEIS) regression model, which accounts for unobserved heterogeneous trends across states in addition to unobserved time-constant and state-specific heterogeneity (Rüttenauer and Ludwig 2020; Wooldridge 2002). This approach is more flexible than the fixed-effects or two-way fixed-effects approach and makes less-restrictive assumptions. However, because the model effectively reduces the number of observations, it imposes greater demands on the data and often produces larger standard errors. Our dependent variable is the TFR in state i in year t. Our explanatory variables are a development indicator in state i in the previous year t − 1 and the square of the development indicator. For the HLI, the regression equation is as follows:

where µi is the state-specific slope, λi is the individual fixed effect, γt is the year fixed effect, and εi,t is the idiosyncratic error. We estimate the coefficients in the equation after taking the first differences and then de-meaning. Thus, µi, λi, and γt are not estimated explicitly, as would be the case for λi and γt in a standard one-way or two-way fixed-effects model.

For fertility decline reversals, the coefficient β1 has to be negative, whereas β2 has to be positive. If either or both coefficients have the opposite sign, the data do not follow an inverse J-shape. Furthermore, we estimate the point at which the association between the TFR and development switches from negative to positive. The inversion point, I, can be calculated as I=β12β2. The standard error of I can be calculated from the standard errors of β1 and β2 (see section A of the online appendix).

Coefficient estimates consistent with the inverse J-shape are also consistent with other shapes of the empirical relationship between development and fertility—in particular, an L-shape and a U-shape. If the empirical relationship has an L-shape, a quadratic model can also provide a good fit, but fertility does not increase with development. If the empirical relationship has a U-shape, fertility increases much more after the turning point. However, these different shapes are not completely inconsistent. For example, an inverse J can turn into a U over time. To distinguish between these shapes, we conduct additional analyses. Employing observations after the turning point, we estimate TFR changes (∆TFR) relative to development changes (∆HLI) in the previous year. We classify the year-to-year changes as confirming a J-shape when the values are positive and as contradicting a J-shape when fertility declines following increases in development. Increasing TFR following development declines is a residual category, which we have removed from the data.

In our robustness checks, we use several other modeling approaches to account for model dependence. We apply two-way fixed-effects regression, which removes only the additive contribution of state and year effects (Imai and Kim 2021). We apply models with state fixed effects, which removes less variance from the outcome variable and which usually has lower standard errors than FEIS. In addition, we rerun the analysis using two-way random effects, which is statistically more efficient, albeit at the cost of making the additional assumption that unobserved heterogeneity is not correlated with the development indicator.

We also conducted further robustness checks. First, as discussed earlier, we are examining long-term development changes that ultimately determine fertility. To remove short-term fluctuations from the data, we smooth both the fertility and the development time series using locally weighted scatterplot smoothing (LOESS). After generating the smoothed time series, we proceed as described above. Second, to account for the spatial structure of the U.S. states, we replace the state-level fixed effects with fixed effects structured by census divisions. Doing so allows us to account for the fact that states that are close to each other are often relatively similar, leading to a high spatial correlation of state-level TFRs. Finally, we conduct a breakpoint analysis using data that are state de-meaned with different a priori specifications of the number of breakpoints. Although these models are less powerful than the main specification, they are less constrained in form and can therefore better distinguish between U-, L-, and J-shapes.

Results

Descriptive Results

Figure 1 shows the TFR trends in the top left panel for all states (semitransparent lines), as well as the TFR trend at the national level (thick line). From 1969 to 2018, the national-level TFR fell from 2.7 to a historical low of 1.8. Beyond indicating the overall trend, the graph reveals three phases: a strong decline around 1970 related to the baby bust that followed the baby boom, a gradual recovery between 1977 and 2008, and a strong decline following 2008. Although these phases roughly apply to most states, fertility levels also show considerable heterogeneity across states. For instance, in 2018, the state-level TFR ranged from 1.3 in the District of Columbia to 2.1 in South Dakota. Moreover, the trends in some states deviated from the country-level trends.

The time series of our three development indicators presented in Figure 1 (top right and bottom panels) show clear improvement in development in the United States over the analyzed period, with all development measures increasing between 1969 and 2018. However, this trend stalled somewhat in the most recent years: in the 2010s, both e0 and the HLI decreased, and the HDI increased more slowly. The declines in e0 and the HLI—measures based on life tables—reflect stagnating or even increasing mortality, which has been attributed to drug overdose crises and cardiovascular diseases (Jalal et al. 2018; Mehta et al. 2020).

Figure 2 provides a look at the subnational association between development and fertility. Each line represents a U.S. state, and each point indicates the average level of the development indicator and the average TFR during each decade. HLI and e0 show a similar pattern resembling a J-shape, with fertility decreasing at lower values and increasing again at a life expectancy of 75. Yet, the middle panel presents a different pattern: the relationship between the HDI and the TFR increases and decreases at different times and at different levels across regions.

Main Results

Table 2 presents the results of the main model using the TFR as the fertility measure for different development indicators. The signs of the coefficients for the HLI and e0 align with the fertility decline reversal hypothesis, whereas the results for the HDI (discussed later) contradict it. For the HLI and e0, the association between development and fertility is negative at lower levels of development, as indicated by the negative sign on the linear term. This association becomes positive at higher development levels, as the positive coefficient of the quadratic term starts to dominate. From these coefficients, we calculate that the female TFR starts to increase when a state has reached an HLI value above 74.2 (95% confidence interval [CI] = 72.9, 75.5) or an e0 of 116.71 years or higher (95% CI = 104.02, 129.4). The turning point for the HLI lies within the observed value range, whereas the turning point for e0 exceeds the maximal observed value, which contradicts the J-shape hypothesis.

The results based on the HDI as a development indicator, also shown in Table 2, do not provide evidence for fertility reversals. The linear and squared terms are positive, indicating a positive impact of development that is reinforced at higher levels. Further, these findings do not align with demographic transition theory. The descriptive results presented in Figure 2 point to a potential explanation for these results. The time series is shorter than that for the HLI and e0 and is observed only for relatively high development values, with little variance. In line with this explanation, our robustness checks for the HLI and e0 suggest that omitting several years at the beginning of the time series changes the results drastically. More generally, the results obtained using the HDI indicate that the conclusions are sensitive to the choice of development indicator and its availability over time, as our upcoming robustness checks confirm.

To evaluate the effect size of development on fertility, we estimate the marginal effect of a one-point increase in the HLI on the female TFR from the model parameters at the 25% quantiles (HLI = 56.99, 70.76, 75.79, and 80.73). At the lowest value of human development (HLI = 56.99), a one-unit increase in the HLI is expected to reduce the TFR by 0.0084 (CI = −0.0016, −0.0152). This value corresponds to a decrease of 0.8% in the TFR sample mean. At the other extreme, a one-unit increase at the HLI's maximum value (80.73) corresponds to an increase of 0.0016 (CI = −0.003, 0.006) in the TFR, which is equivalent to 0.16% of the total mean TFR of the sample. Overall, the pattern of the marginal effects is in line with the J-shape hypothesis.

Moreover, the observed data correspond with the fertility reversal hypothesis, which we evaluate using TFR–HLI slopes after the estimated turning point. We find that 52.15% of the slopes align with the J-shape hypothesis, given that they are positive or show at least fertility increases. Only 47.85% show negative slopes with fertility declines. When we restrict the time series further to the pre-recession period (before 2008), the share of confirmatory results increases to 72.45%. After the economic recession, the share shrinks substantially to 44.22%.

Robustness Checks

The results of our robustness checks are summarized in Figure 3, which displays the turning points across indicators and model specifications. Overall, the graph shows that the majority of points lie in the observed value range of the development indicator (shaded area), which supports the hypothesis that there have been fertility decline reversals in the United States. However, some results contradict the hypothesis. Despite variation across model specifications displayed on the y-axis, the two-way fixed-effects and random-effects models contradict the J-shape hypothesis, given that the turning points lie outside the observed value range. Moreover, the middle column shows fewer points than the other columns (owing to different signs in the regression coefficients for the HDI indicator), thus contradicting the J-shape hypothesis. Overall, fertility declines in the last decade are shifting the model-based turning points to higher values.

Fertility Indicators

To account for potential tempo distortions of the TFR, we use the tempo-adjusted TFR (Bongaarts and Feeney 1998:278). The results are displayed in Tables S1 and S2 in the online appendix. The coefficients have the same signs as in the analysis with the unadjusted TFR, and they are consistent with the J-shaped association and the fertility decline reversals. These findings give us further confidence in concluding that the J-shaped pattern was not caused by fertility recuperation, as Bongaarts and Sobotka (2012) argued.

The results presented in Table S3 (online appendix) based on the male TFR lead to very similar conclusions. These results reveal turning points similar to those in the main model. According to the FEIS regression, the male TFR starts increasing at a male HLI of 68.86 (CI = 68.09, 69.64) or at e0 for men of 74.72 (CI = 73.94, 75.5), both of which are within the observed data range. By contrast, the turning point of the HDI lies at 1.15, outside the range of observed and even possible values. However, the data for the HDI versus male TFR model contain only 14 years of observations and are thus less robust than the other estimates.

Adding Control Variables

Earlier, we discussed the role of gender equality and economic uncertainty in the development–fertility nexus. We introduce controls for the state-level unemployment rate, the state-level percentage of jobs in the service sector, and the average age gap between parents and test whether the results remain similar. Furthermore, including time-varying controls allows us to better account for heterogeneous trends in economic conditions and gender equality. The pattern of results (see Table S4 in the online appendix) corresponds to the arguments presented in the theoretical section. Controlling for unemployment or gender equality absorbs the J-shaped relationship between development and fertility. Thus, the association between development and fertility net of economic conditions and gender equality is likely small, and these two factors are key drivers of the overall association.

The results for the tempo-adjusted TFR are confirmed by the models with controls for the mean age at childbearing and the first- and second-difference time series of the indicator (see Table S5). The model estimates show a convex relationship and give reasonable turning points.

Different Regression Methods

To assess the extent to which our findings are dependent on model assumptions, we use several alternative regression models. We use two-way fixed-effects models to account for the fact that the FEIS models might absorb some of the variance in the outcomes that result from developmental processes. As hypothesized, the linear term is negative, and the quadratic term is positive, yielding a convex relationship. The results are similar to those presented earlier (see Table 3). However, the turning points occur at a higher value of the development variable and outside the observed value range. We conclude that the selection on trends might not be captured in the two-way fixed-effects model despite yielding significant results for the squared term. Using random-effect models, however, yields results in line with the J-shape hypothesis (see Tables S6 and S7, online appendix), providing additional support for the existence of a J-shaped relationship.

We model the relationship between development and fertility using smoothed time-series data. This approach should yield further evidence on the contribution of the impact of long-term effects of developments beyond short-term fluctuations, which are removed from the data using LOESS. The results displayed in Tables S8 and S9 (online appendix) point to the impact of long-term development on fertility. Therefore, we conclude that it is the overall trend in development, rather than short-term fluctuations in development, that has an effect on fertility.

We also account for the problem of spatial autocorrelation, revealed in Figure 2 from the within-census division similarities, by replacing state fixed effects with census division fixed effects. The results align with the hypothesis (see Table S10, online appendix). The inversion of fertility decline occurs within the range of observed values. Nevertheless, the statistical significance of the estimates and R2 is low, suggesting considerable variation around the expected values.

Finally, we conduct a breakpoint analysis using the data that are state de-meaned and report the results in Table S13 and Figure S9 (online appendix). The model with two breakpoints yields the best model fit. The slopes for the first, second, and third intervals are, respectively, −0.17 (CI = −0.175, −0.152), 0.2 (CI = 0.027, 0.035), and −0.08 (CI = −0.063, −0.046). The location of the breakpoints is estimated at an HLI of 68.1 (CI = 67.9, 68.3) and 74.8 (CI = 74.56, 75.0). These results indicate that (1) there might have been a J-shaped relationship in the past and (2) the relationship likely vanished in the most recent period.

Further Robustness Checks

To investigate the periodicity of the J-shaped pattern, we calculate two-way FEIS models for the TFR and the HLI for different time-series lengths, thus altering the start and end year of the time series, with the restriction of a minimum length of 10 years. The results show that most models confirm the J-shape pattern (see Figure S6, online appendix), with two important exceptions. First, the J-shape disappears when the time series starts after 1975 instead of 1969, indicating that the fertility decline between 1969 and 1975 (visible in Figure 2) drives the negative relationship between development and fertility at lower levels. Second, when only the last period (1995–2018) is considered, the relationship has an inverted U-shape or becomes fully negative, perhaps indicating that the J-shaped relationship might not exist for the most recent period. Results presented in Table S16 (online appendix) confirm that the relationship between the HLI and the TFR is negative for the 2008–2018 period. A similar analysis of state omissions indicates that the District of Columbia exerts the greatest influence, and omitting it shifts the estimated turning point to an HLI value of 82.2 (CI = 75.1, 89.3), which exceeds the observed value range.

Two competing explanations might account for the diverging results across indicators: (1) the time-series length and (2) the different dimensions of development. We reestimate the main regression model for the same time series (1990–2018), neutralizing the impact of the first explanation. The results change drastically, providing evidence of the relevance of the time-series length. The J-shape for e0 in the main model fades completely, whereas the estimated turning point increases to an HLI level of 180.04. The unexpected result for the HDI is mainly attributable to the availability of only more recent data. However, some differences remain across the three indicators, providing evidence that the indicators capture different dimensions of development.

Discussion

In this article, we examined whether the Myrskylä et al. (2009) J-shape hypothesis holds for the United States at the subnational level. Leveraging state-level data covering 1969–2018, we found that the association between development and fertility followed an inverse J-shape between 1969 and 2008 but not thereafter. This association was robust across many sensitivity checks. Thus, our findings support the Myrskylä et al. (2009) reversal hypothesis for the United States at the subnational level when applied to historical periods. Moreover, although development progress has led to some temporary fertility increases, it has not necessarily resulted in fertility increases beyond the replacement level. However, Myrskylä et al. (2009) did not expect that fertility increases would necessarily persist for a long time—that is, that the inverse J-shape would necessarily turn into a U-shape. Although adjusting for tempo effects indicated that these factors might have played some role in the fertility decline reversals, our main finding persisted. Moreover, we found that good economic prospects and high levels of gender equality were prerequisites for the reversal of the relationship. These results correspond to the findings of Myrskylä et al. (2011) and Luci-Greulich and Thévenon (2014). However, for the most recent decade, we observed an overall fertility decline despite some progress in development. The breakpoint analysis confirmed that the J-shaped relationship likely vanished in the most recent decade. Moreover, although the results were sensitive to the choice of indicator, further analysis showed that the time-series length accounted for most of the differences across indicators.

Our subnational findings align with a country-level analysis by Myrskylä et al. (2009, 2011), which found evidence of fertility decline reversals at high development levels. They are also consistent with a study by Gaddy (2021), which found no correlation between development and fertility in the recent decade. The high shares of confirmatory HLI–TFR slopes after the estimated turning point indicate that the relationship resembles a historical J-shape, rather than an L-shape, in the period before the Great Recession. Additional analyses corroborated this finding: since the recession, most year-to-year changes in development and fertility are inconsistent with the inverse J-shape. For 58% of these year-to-year changes, the TFR dropped despite increases in the HDI. In contrast, after the turning point estimated in our main model and before the recession, 67.7% of the year-to-year changes were consistent with the inverse J-shape hypothesis. Moreover, the temporal structure of the pattern might explain the mixed findings in subnational U.S. studies, which applied cross-sectional analysis to county-level data (Porter 2017; Ryabov 2015). Future research could seek to examine the longitudinal relationship using county-level data.

The finding of a historical inverse J-shaped relationship has theoretical implications, given that it stands in contrast to the demographic transition theory and the low-fertility-trap hypothesis. The results point to the existence of factors that can relax and even reverse the negative association between development and fertility posited by demographic transition theory (Davis 1945; Lesthaeghe 2020; Lesthaeghe and van de Kaa 1986; Notestein 1945). As outlined earlier in the article, increasing development might reduce levels of gender inequality and economic uncertainty, which can lead to higher fertility. Furthermore, increasing fertility contradicts the low-fertility-trap hypothesis (Lutz et al. 2006). We found no sign of entrenchment at low levels, leading us to question whether low fertility is indeed self-reinforcing.

In the United States, the J-shaped relationship between development and fertility observed for 1969–2008 shifted notably around 2007–2008. We found a negative relationship between development and fertility for 2008–2018. Moreover, continuing fertility declines beyond 2018 lead us to speculate that the negative relationship holds today. This shift, which Gaddy (2021) also noted at the country level, might be attributable to a structural break. Questions about the potential role of economic uncertainty, value change, and contraceptive practices have been raised in the discourse on the recent fertility decline. First, given that the beginning of the fertility decline coincided with the economic recession, the decline might be explained by continuing economic uncertainty (Comolli 2021; Schneider and Gemmill 2016; Schneider and Hastings 2015; Vignoli et al. 2020). In addition, changes in values and family norms might be responsible for the structural break. For example, a recent study argued that an attitudinal shift might be a key driver of the recent fertility decline, given that it occurred across states and social classes (Kearney et al. 2022). Thus, the recent decline might be in line with the second demographic transition theory (Lesthaeghe 2020; Lesthaeghe and Neidert 2006). Finally, the development and distribution of modern contraceptive technologies, especially long-acting reversible contraceptives, might have reduced unintended births (Eeckhaut et al. 2021; Kavanaugh and Jerman 2018). Further research is needed for a full understanding of the structural factors underlying this recent decline in fertility.

An alternative explanation for the results for the period since the financial crisis is measurement error. Because our indicators do not perfectly capture development, some degree of measurement error is unavoidable. If measurement error increases over time because our indicators become less predictive of development, biased findings might result. There is, however, only modest evidence for increased measurement error. The association between different development indicators, health expenditures, GDP, and wages on the one hand and e0 on the other has been declining slightly. For instance, Table S14 (online appendix) shows that the correlation between the HLI, our main development measure, and the Gini coefficient has weakened. However, the association is still very strong.

Consistent with previous research, we showed that the results vary depending on the choice of the development indicator (Gaddy 2021; Harttgen and Vollmer 2014). We found no inverse J-shaped association when using the HDI but observed the inverse J-shape when using e0 (life expectancy at birth) and the HLI. These findings appear to support studies by Gaddy (2021) and Harttgen and Vollmer (2014), who criticized the indicators used in Myrskylä et al. (2009). However, additional analyses showed that these results were attributable to the only recent period of data availability of the HDI time series, which is more greatly affected by the years after the 2007–2008 financial crisis.

A key contribution of this study is that we addressed the major criticism of the J-shape hypothesis. First, we conducted several sensitivity checks that accounted for the potential impact of postponement, including the tempo-adjusted TFR and using the mean age at childbirth as a control variable. These robustness checks yielded an inverse J-shaped pattern. Thus, in contrast to findings from other studies (Bongaarts and Sobotka 2012), we conclude that postponement is not a major driver of the association between development and fertility in the United States. Second, changing the research design to a subnational longitudinal setup, which rules out unobserved cultural and institutional factors, allowed us to address Lesthaeghe's (2020) concern that national idiosyncrasies might be driving the J-shaped pattern. Finally, we addressed the critique regarding measurement error by using different indicators. We found that the general results depended more on the time-series length than on the indicator itself.

In line with previous studies, we found that economic and gender factors play a crucial role in the development–fertility nexus (Esping-Andersen and Billari 2015; Goldscheider et al. 2015; Kolk 2019; Luci-Greulich and Thévenon 2014). First, the fertility decline reversals were conditional on positive employment prospects. This evidence points to the role of economic uncertainty, as Schneider and Hastings (2015) and Comolli (2017, 2021) argued. Second, we found that the effect was moderated by household gender equality (Luci-Greulich and Thévenon 2014). Theoretical arguments emphasize the role of the reconciliation of family and work as well as women's opportunities to achieve their personal career goals (Goldscheider et al. 2015). Therefore, the gender dimension seems to play a crucial role in fertility in highly developed states, with fertility increasing as development progresses.

Methodological Considerations

When interpreting our results, it is important to remember that they show the association between development and fertility at the macro level and do not allow us to infer individual responses to development. Our results yield evidence only on contextual factors, which might also be specific to the United States in our analysis. Moreover, although we found that the association between development and fertility was strong, development was only one of many determinants of fertility, as some of our additional analyses highlighted. Some of these other determinants might be mediators or moderators of the impact of development, which calls for further research into the mechanisms linking development and fertility. Finally, our results refer to the average fertility behavior at the state level and do not capture the considerable heterogeneity within states (Daniels et al. 2018; Porter 2017; Porter and Purser 2008; Ryabov 2015).

Moreover, it is difficult to predict whether the associations we found will hold in the future. In recent years, U.S. fertility has been volatile and sensitive to external shocks. The TFR started falling between 2008 and 2010 following the Great Recession and did not rebound thereafter (Cherlin et al. 2013; Schneider and Hastings 2015). Thus, fertility decline reversals might be stalled or undone, as our results for 2008–2018 indicated. Further, compared with other countries, the United States stands out because its development increases have leveled off and it has had no large improvements in development in recent years.

This study makes a leap forward by using a wide array of robustness checks, most of which showed fertility reversals in the United States. Thus, our results confirm the existence of temporary fertility decline reversals at higher development levels but show a different pattern for the most recent period, beginning in 2008.

Acknowledgments

We thank Angela Greulich, Joshua Wilde, and conference participants in Berlin and Atlanta, GA, for their helpful comments. We also thank the International Max Planck Research School for Population, Health and Data Science and Nuffield College for supporting Henrik-Alexander Schubert. All errors remain the authors’ sole responsibility. Mikko Myrskylä was supported by the Strategic Research Council, FLUX consortium (decision numbers 345130 and 345131); by grants to the Max Planck–University of Helsinki Center from the Jane and Aatos Erkko Foundation, the Max Planck Society, Faculty of Social Sciences at the University of Helsinki, and Cities of Helsinki, Vantaa, and Espoo; and the European Union (ERC Synergy, BIOSFER, 101071773). Views and opinions expressed are, however, those of the author only and do not necessarily reflect those of the European Union or the European Research Council. Neither the European Union nor the granting authority can be held responsible for them.

Notes

2

Neither Notestein (1945) nor Davis (1945) claimed that fertility will remain close to replacement level. However, the widely adapted interpretation of the framework postulates the emergence of homeostasis before and after the first demographic transition, with replacement-level fertility.

3

Hereafter, TFR refers to the TFR for women unless explicitly stated otherwise.

References

Acemoglu, D., & Autor, D. (
2010
).
Skills, tasks and technologies: Implications for employment and earnings
(NBER Working Paper 16082).
Cambridge, MA
:
National Bureau of Economic Research
.
Adserà, A. (
2004
).
Changing fertility rates in developed countries. The impact of labor market institutions
.
Journal of Population Economics
,
17
,
17
43
.
Alexander, M., Polimis, K., & Zagheni, E. (
2022
).
Combining social media and survey data to nowcast migrant stocks in the United States
.
Population Research and Policy Review
,
41
,
1
28
.
Althoff, L., Eckert, F., Ganapati, S., & Walsh, C. (
2022
).
The geography of remote work
.
Regional Science and Urban Economics
,
93
,
103770
. https://doi.org/10.1016/j.regsciurbeco.2022.103770
Anderson, T., & Kohler, H.-P. (
2015
).
Low fertility, socioeconomic development, and gender equity
.
Population and Development Review
,
41
,
381
407
.
Beck, U. (
1992
).
Risk society: Towards a new modernity
(Ritter, M., Trans.).
London, UK
:
SAGE Publications
. (Original work published 1986)
Becker, G. S. (
1981
).
A treatise on the family
.
Chicago, IL
:
University of Chicago Press
.
Becker, G. S., & Lewis, H. G. (
1973
).
On the interaction between the quantity and quality of children
.
Journal of Political Economy
,
81
(
2, part 2
),
S279
S288
.
Becker, G. S., & Tomes, N. (
1976
).
Child endowments and the quantity and quality of children
.
Journal of Political Economy
,
84
(
4, part 2
),
S143
S162
.
Bongaarts, J., & Feeney, G. (
1998
).
On the quantum and tempo of fertility
.
Population and Development Review
,
24
,
271
291
.
Bongaarts, J., & Sobotka, T. (
2012
).
A demographic explanation for the recent rise in European fertility
.
Population and Development Review
,
38
,
83
120
.
Bongaarts, J., & Watkins, S. C. (
1996
).
Social interactions and contemporary fertility transitions
.
Population and Development Review
,
22
,
639
682
.
Bryant, J. (
2007
).
Theories of fertility decline and the evidence from development indicators
.
Population and Development Review
,
33
,
101
127
.
Carmichael, S. (
2011
).
Marriage and power: Age at first marriage and spousal age gap in lesser developed countries
.
History of the Family
,
16
,
416
436
.
Casterline, J. B. (
2003
).
Demographic transition
. In McNicholl, G. & Demeny, P. G. (Eds.),
Encyclopedia of population
(Vol.
1
, pp.
210
216
).
New York, NY
:
Macmillan Reference
.
Centers for Disease Control and Prevention
. (
2023
).
Natality public-use data 2003–2022
.
Washington, DC
:
Centers for Disease Control and Prevention
.
Cherlin, A., Cumberworth, E., Morgan, S. P., & Wimer, C. (
2013
).
The effects of the Great Recession on family structure and fertility
.
The Annals of the American Academy of Political and Social Science
,
650
,
214
231
.
Comolli, C. L. (
2017
).
The fertility response to the Great Recession in Europe and the United States: Structural economic conditions and perceived economic uncertainty
.
Demographic Research
,
36
,
1549
1600
. https://doi.org/10.4054/DemRes.2017.36.51
Comolli, C. L. (
2021
).
Resources, aspirations and first births during the Great Recession
.
Advances in Life Course Research
,
48
,
100405
. https://doi.org/10.1016/j.alcr.2021.100405
Daniels, K., Martinez, G. M., & Nugent, C. N. (
2018
).
Urban and rural variation in fertility-related behavior among U.S. women, 2011–2015
(NCHS Data Brief, No. 297).
Hyattsville, MD
:
National Center for Health Statistics
. Retrieved from https://www.cdc.gov/nchs/data/databriefs/db297.pdf
Davis, K. (
1945
).
The world demographic transition
.
The Annals of the American Academy of Political and Social Science
,
237
,
1
11
.
de Haas, H., Castles, S., & Miller, M. (
2020
).
The age of migration: International population movements in the modern world
(6th ed.).
New York, NY
:
Guilford Press
.
Dudel, C., Cheng, Y. A., & Klüsener, S. (
2023
).
Shifting parental age differences in high-income countries: Insights and implications
.
Population and Development Review
,
49
,
879
908
.
Dudel, C., & Klüsener, S. (
2019
).
Estimating men's fertility from vital registration data with missing values
.
Population Studies
,
73
,
439
449
.
Dudel, C., & Klüsener, S. (
2021
).
Male–female fertility differentials across 17 high-income countries: Insights from a new data resource
.
European Journal of Population
,
37
,
417
441
.
Eeckhaut, M. C. W., Rendall, M. S., & Zvavitch, P. (
2021
).
Women's use of long-acting reversible contraception for birth timing and birth stopping
.
Demography
,
58
,
1327
1346
. https://doi.org/10.1215/00703370-9386084
Esping-Andersen, G., & Billari, F. C. (
2015
).
Re-theorizing family demographics
.
Population and Development Review
,
41
,
1
31
.
Firebaugh, G. (
2018
).
Seven rules for social research
.
Princeton, NJ
:
Princeton University Press
.
Fox, J., Klüsener, S., & Myrskylä, M. (
2019
).
Is a positive relationship between fertility and economic development emerging at the sub-national regional level? Theoretical considerations and evidence from Europe
.
European Journal of Population
,
35
,
487
518
.
Gaddy, H. G. (
2021
).
A decade of TFR declines suggests no relationship between development and sub-replacement fertility rebounds
.
Demographic Research
,
44
,
125
142
. https://doi.org/10.4054/DemRes.2021.44.5
Ghislandi, S., Sanderson, W. C., & Scherbov, S. (
2019
).
A simple measure of human development: The Human Life Indicator
.
Population and Development Review
,
45
,
219
233
.
Goldscheider, F., Bernhardt, E., & Lappegård, T. (
2015
).
The gender revolution: A framework for understanding changing family and demographic behavior
.
Population and Development Review
,
41
,
207
239
.
Harknett, K., Billari, F. C., & Medalia, C. (
2014
).
Do family support environments influence fertility? Evidence from 20 European countries
.
European Journal of Population
,
30
,
1
33
.
Harttgen, K., & Vollmer, S. (
2014
).
A reversal in the relationship of human development with fertility?
Demography
,
51
,
173
184
.
Hofmann, B., & Hohmeyer, K. (
2013
).
Perceived economic uncertainty and fertility: Evidence from a labor market reform
.
Journal of Marriage and Family
,
75
,
503
521
.
Imai, K., & Kim, I. S. (
2021
).
On the use of two-way fixed effects regression models for causal inference with panel data
.
Political Analysis
,
29
,
405
415
.
Inglehart, R. (
1977
).
The silent revolution: Changing values and political styles among Western publics
.
Princeton, NJ
:
Princeton University Press
.
Jalal, H., Buchanich, J. M., Roberts, M. S., Balmert, L. C., Zhang, K., & Burke, D. S. (
2018
).
Changing dynamics of the drug overdose epidemic in the United States from 1979 through 2016
.
Science
,
361
,
eaau1184
. https://doi.org/10.1126/science.aau1184
Kavanaugh, M. L., & Jerman, J. (
2018
).
Contraceptive method use in the United States: Trends and characteristics between 2008, 2012 and 2014
.
Contraception
,
97
,
14
21
.
Kearney, M. S., Levine, P. B., & Pardue, L. (
2022
).
The puzzle of falling U.S. birth rates since the Great Recession
.
Journal of Economic Perspectives
,
36
(
1
),
151
176
.
Kohler, H.-P., Billari, F. C., & Ortega, J. A. (
2002
).
The emergence of lowest-low fertility in Europe during the 1990s
.
Population and Development Review
,
28
,
641
680
.
Kolk, M. (
2019
).
Weak support for a U-shaped pattern between societal gender equality and fertility when comparing societies across time
.
Demographic Research
,
40
,
27
48
. https://doi.org/10.4054/DemRes.2019.40.2
Lacalle-Calderon, M., Perez-Trujillo, M., & Neira, I. (
2017
).
Fertility and economic development: Quantile regression evidence on the inverse J-shaped pattern
.
European Journal of Population
,
33
,
1
31
.
Lesthaeghe, R. (
2020
).
The second demographic transition, 1986–2020: Sub-replacement fertility and rising cohabitation—A global update
.
Genus
,
76
,
10
. https://doi.org/10.1186/s41118-020-00077-4
Lesthaeghe, R., & Permanyer, I. (
2014
).
European sub-replacement fertility: Trapped or recovering?
(PSC Research Report, No. 14-822).
Ann Arbor
:
Michigan Population Studies Center
.
Lesthaeghe, R., & van de Kaa, D. J. (
1986
).
Twee demografische transities? [Two demographic transitions?]
. In van de Kaa, D. J. & Lesthaeghe, R. (Eds.),
Bevolking: Groei en Krimp
[Population: Growth or decline] (pp.
9
24
).
Deventer, the Netherlands
:
Van Loghum Slaterus
.
Lesthaeghe, R. J., & Neidert, L. (
2006
).
The second demographic transition in the United States: Exception or textbook example?
Population and Development Review
,
32
,
669
698
.
Lichter, D. T., Johnson, K. M., Turner, R. N., & Churilla, A. (
2012
).
Hispanic assimilation and fertility in new U.S. destinations
.
International Migration Review
,
46
,
767
791
.
Luci-Greulich, A., & Thévenon, O. (
2014
).
Does economic advancement “cause” a re-increase in fertility? An empirical analysis for OECD countries (1960–2007)
.
European Journal of Population
,
30
,
187
221
.
Lutz, W., Skirbekk, V., & Testa, M. R. (
2006
).
The low-fertility trap hypothesis: Forces that may lead to further postponement and fewer births in Europe
.
Vienna Yearbook of Population Research
,
4
,
167
192
.
Mavropoulos, G., & Panagiotidis, T. (
2021
).
On the drivers of the fertility rebound
.
Economic Change and Restructuring
,
54
,
821
845
.
McDonald, P. (
2000
).
Gender equity in theories of fertility transition
.
Population and Development Review
,
26
,
427
439
.
Mehta, N. K., Abrams, L. R., & Myrskylä, M. (
2020
).
U.S. life expectancy stalls due to cardiovascular disease, not drug deaths
.
Proceedings of the National Academy of Sciences
,
117
,
6998
7000
.
Milewski, N. (
2010
).
Fertility of immigrants: A two-generational approach in Germany
.
Berlin, Germany
:
Springer-Verlag
.
Mills, M. (
2007
).
Individualization and the life course: Toward a theoretical model and empirical evidence
. In Howard, C. (Ed.),
Contested individualization: Debates about contemporary personhood
(pp.
61
79
).
New York, NY
:
Palgrave Macmillan
.
Mills, M., Johnston, A. D., & DiPrete, T. A. (
2006
).
Globalization and men's job mobility in the United States
. In Blossfeld, H.-P., Mills, M., & Bernardi, F. (Eds.),
Globalization, uncertainty and men's careers: An international comparison
(pp.
328
364
).
Edward Elgar Publishing
.
Myrskylä, M., Kohler, H.-P., & Billari, F. C. (
2009
).
Advances in development reverse fertility declines
.
Nature
,
460
,
741
743
.
Myrskylä, M., Kohler, H.-P., & Billari, F. C. (
2011
).
High development and fertility: Fertility at older reproductive ages and gender equality explain the positive link
(Technical Report, No. WP-2011-017).
Rostock, Germany
:
Max Planck Institute for Demographic Research
.
National Bureau of Economic Research (NBER)
. (
2023
).
SEER population data—United States files at state and/or county level
. Retrieved from https://www.nber.org/research/data/survey-epidemiology-and-end-results-seer-us-state-and-county-population-data-age-race-sex-hispanic
National Center for Health Statistics
. (
2022
).
Vital statistics natality birth data
. Retrieved from https://www.nber.org/research/data/vital-statistics-natality-birth-data
Notestein, F. W. (
1945
).
Population—The long view
. In Schultz, T. W. (Ed.),
Food for the world
(pp.
36
57
).
Chicago, IL
:
University of Chicago Press
.
Porter, J. R. (
2017
).
Human development and the fertility reversal: A spatially centered sub-national examination in the United States
.
Spatial Demography
,
5
,
43
72
.
Porter, J. R., & Purser, C. W. (
2008
).
Measuring relative sub-national human development: An application of the United Nation's Human Development Index using geographic information systems
.
Journal of Economic and Social Measurement
,
33
,
253
269
.
Presser, H. B. (
1975
).
Age differences between spouses: Trends, patterns, and social implications
.
American Behavioral Scientist
,
19
,
190
205
.
Rindfuss, R. R., Choe, M. K., & Brauner-Otto, S. R. (
2016
).
The emergence of two distinct fertility regimes in economically advanced countries
.
Population Research and Policy Review
,
35
,
287
304
.
Ruggles, S. (
2015
).
Patriarchy, power, and pay: The transformation of American families, 1800–2015
.
Demography
,
52
,
1797
1823
.
Rüttenauer, T., & Ludwig, V. (
2020
).
Fixed effects individual slopes: Accounting and testing for heterogeneous effects in panel data or other multilevel models
.
Sociological Methods & Research
,
52
,
43
84
.
Ryabov, I. (
2015
).
On the relationship between development and fertility: The case of the United States
.
Comparative Population Studies
,
40
,
465
489
.
Scherbov, S., & Gietel-Basten, S. (
2020
).
Measuring inequalities of development at the sub-national level: From the Human Development Index to the Human Life Indicator
.
PLoS One
,
15
,
e0232014
. https://doi.org/10.1371/journal.pone.0232014
Schneider, D., & Gemmill, A. (
2016
).
The surprising decline in the non-marital fertility rate in the United States
.
Population and Development Review
,
42
,
627
649
.
Schneider, D., & Hastings, O. P. (
2015
).
Socioeconomic variation in the effect of economic conditions on marriage and nonmarital fertility in the United States: Evidence from the Great Recession
.
Demography
,
52
,
1893
1915
.
Schoumaker, B. (
2019
).
Male fertility around the world and over time: How different is it from female fertility?
Population and Development Review
,
45
,
459
487
.
Seltzer, J. A. (
2019
).
Family change and changing family demography
.
Demography
,
56
,
405
426
.
Sen, A. (
1998
).
Mortality as an indicator of economic success and failure
.
Economic Journal
,
108
,
1
25
.
Smart, C., & Shipman, B. (
2004
).
Visions in monochrome: Families, marriage and the individualization thesis
.
British Journal of Sociology
,
55
,
491
509
.
United States Mortality DataBase (USMDB)
. (
2022
). [Dataset].
Berkeley
:
University of California, Berkeley
. Available from https://usa.mortality.org/
van de Kaa, D. (
1987
).
Europe's second demographic transition
(Population Bulletin, Vol.
42
No.
1
).
Washington, DC
:
Population Reference Bureau
.
van Raalte, A. A., Sasson, I., & Martikainen, P. (
2018
).
The case for monitoring life-span inequality
.
Science
,
362
,
1002
1004
.
Vignoli, D., Bazzani, G., Guetto, R., Minello, A., & Pirani, E. (
2020
).
Uncertainty and narrative of the future: A theoretical framework for contemporary fertility
. In Schoen, R. (Ed.),
The Springer series on demographic methods and population analysis: Vol. 51. Analyzing contemporary fertility
(pp.
25
48
).
Cham
:
Springer Nature Switzerland
.
Wooldridge, J. M. (
2002
).
Econometric analysis of cross section and panel data
.
Cambridge, MA
:
MIT Press
.
Worts, D., Sacker, A., McMunn, A., & McDonough, P. (
2013
).
Individualization, opportunity and jeopardy in American women's work and family lives: A multi-state sequence analysis
.
Advances in Life Course Research
,
18
,
296
318
.
Freely available online through the Demography open access option.

Supplementary data