The (temporary) Covid-19 baby bust in Mexico

Abstract

In this paper, we investigate whether fertility and newborn health changed during the Covid-19 pandemic in Mexico. We use national administrative data and an event-study design to examine the impact of the Covid-19 pandemic on fertility and newborn health characteristics. Our findings suggest that Mexico’s fertility declined temporarily as measured by conceptions that likely occurred during the stay-at-home order. Initially, the general fertility rate fell by 11–12 per cent but quickly rebounded and returned close to its original levels by the end of 2021. Newborn health also deteriorated during the pandemic. Instances of low birthweight and prematurity substantially increased, with both remaining elevated over the entire pandemic period.

Keywords:

• Covid-19
• pandemic
• births
• fertility
• newborn health
• low birthweight
• Mexico

Introduction

Like previous extraordinary disasters, the Covid-19 pandemic has impacted fertility decisions around the world. Empirically speaking, the magnitude and duration of its fertility impact is unknown. The majority of the existing literature suggests that the Covid-19 pandemic lowered fertility and had a negative impact on newborn health outcomes (Luppi et al. 2020; Wilde et al. 2020; Aassve et al. 2021; De Rose et al. 2021; Emery and Koops 2022; Kearney and Levine 2022), at least in the short run.

Theoretically speaking, the vast literature on the fertility impacts of related types of crises, such as economic downturns, supports several hypotheses on the effect of the Covid-19 pandemic on fertility outcomes. In typical economic recessions, fertility acts as a leading indicator, with conceptions declining before the start of the recession (Sobotka et al. 2011; Buckles et al. 2021). Unlike in a typical recession, the onset of the pandemic was sudden, and so we would expect conceptions to decline after the start of the pandemic. However, some scholars have argued that the pandemic’s effects are temporary, with an initial decline followed by a baby boom and a subsequent return to the long-term trend (e.g. Ullah et al. 2020; Boberg-Fazlic et al. 2021).

Covid-19 continues to create uncertainty around the world. This uncertainty impedes individuals from having a clear picture of their future. Until recently, a narrative of the future was not considered a central determinant of individuals’ fertility. However, demographers are increasingly proposing uncertainty and narratives of the future as central to the theory of contemporary fertility plans (Esping-Andersen 2002; Vignoli et al. 2020). In countries with extensive social support for families and employment, individuals may experience only moderate uncertainty during a pandemic, but in countries with small or non-existent welfare systems, the pandemic outbreak will carry a great deal of uncertainty (Aassve et al. 2021). In the latter group of countries, economic hardship brought about by a pandemic may affect childbearing plans, as there is no safety net in place to compensate for loss of income. Conversely, in the former group, a pandemic may bring about restructuring of labour and household arrangements, allowing more flexibility, and this may ultimately lead to childbearing as a way of increasing wider family ties or overcoming a crisis psychologically (McDonald 2000; Comolli et al. 2021). A pandemic may also bring additional risks to pregnant mothers and foetuses, through elevated maternal stress and infection risks. These elevated risks may lead to fertility reductions, even more than in a typical economic recession (Chandra and Yu 2015; Chandra et al. 2018; Boberg-Fazlic et al. 2021). Further, a pandemic may worsen newborn health, leading to a generation of children disadvantaged from birth. Children conceived during a pandemic will face both excess maternal stress and a higher societal disease burden. Each of these in utero exposures may also worsen newborn health at birth (Coutinho et al. 2020; Berghella and Hughes 2021; Janevic et al. 2021; Villar et al. 2021), with potentially lifelong consequences (e.g. see Behrman and Rosenzweig 2004; Almond 2006; Bleakley 2007; Case et al. 2008; Bozzoli et al. 2009; Case and Paxson 2009; Currie 2009; Maluccio et al. 2009; Almond et al. 2011; Currie and Almond 2011; Beach et al. 2016; Hoynes et al. 2016; Bhalotra et al. 2017; Hjort et al. 2017; Bütikofer et al. 2019; Hoehn-Velasco 2021; Beach et al. 2022).

In this study, we analyse changes in the general fertility rate (GFR) and newborn health during the Covid-19 pandemic in Mexico. We use national administrative data from Mexico’s Ministry of Health. The data contain monthly state-level information on births from both public and private hospitals. In our main analysis, we focus on births that occurred from January 2016 to December 2021 and consider the impact of the pandemic on pregnancies conceived after the start of the pandemic, for deliveries occurring from December 2020 to December 2021 (term conceptions from March 2020). For the main outcomes of interest, we focus on GFR, birthweight (high and low), and gestational age (very premature, premature, and term), all at the state level. We also provide additional results that test for socio-economic differentials in Covid-19’s impact on fertility by looking at changes in maternal age and whether delivery location changed during the pandemic.

Our results from using an event-study method, suggest that there was a corresponding 11–12 per cent reduction in GFR in January and February 2021. Deliveries over these months would have been conceived during the stay-at-home order, which suggests that the initial phases of the pandemic can be linked to a clear reduction in fertility. Then, beginning in April 2021, the GFR rose again and approached baseline levels by the end of 2021. Overall, the GFR shows a short-term U-shaped pattern, declining substantially then gradually approaching its original level but failing to fully recover.

Next, we turn to newborn health. Instances of low and high birthweight increased substantially for conceptions that occurred during the pandemic. Instances of high birthweight started to decline again by the end of 2021 but did not fully return to baseline. By contrast, instances of low birthweight increased and stayed elevated over the pandemic period. The increased incidence of low birthweight is also confirmed by increases in prematurity (measured by gestational age). These findings suggest that the effects of the pandemic on fertility may have been temporary but that there are ongoing permanent effects on infant health.

Our results for newborn health align with the existing literature. So far, the majority of hospital-based studies have demonstrated an increase in prematurity and low birthweight during the pandemic (Adhikari et al. 2020; Allotey et al. 2020; Khalil et al. 2020; Woodworth et al. 2020; Chinn et al. 2021; Elsaddig and Khalil 2021; Gurol-Urganci et al. 2021; Norman et al. 2021; Wei et al. 2021). However, a caveat to these findings is that not all studies have found an increase in preterm births (Been et al. 2020; Philip et al. 2020; Kim et al. 2021; Shah et al. 2021; Son et al. 2021). Our findings add to this literature by documenting a (national) increase in cases of low birthweight and prematurity after the start of the pandemic.

Finally, we also consider type of delivery attendant and delivery location using maternity data available from Mexico’s National Institute of Statistics, Geography and Informatics (INEGI) (see INEGI 2020). We find a significant increase in home deliveries and deliveries by a midwife (as opposed to a physician). We also show that maternal age did not change substantially during the pandemic, suggesting that the main deterioration in newborn health was likely not due to changes in maternal characteristics.

Overall, this paper adds to the literature by clearly documenting the short-run decline in the GFR as well as the impact on newborn health at the onset of the Covid-19 pandemic. Our study also contributes to the literature by showing the impact of the pandemic on fertility using administrative data from a large middle-income country. Most importantly, our study considers the case of a country (Mexico) in which social spending is very low—the lowest among Organisation for Economic Co-operation and Development countries (OECD 2020)—and, at the same time, there was no additional economic aid from the government during the peak of the pandemic. This is, perhaps, our main contribution to the existing literature, as most of the countries that have been studied implemented additional welfare support during the pandemic or already had enough social spending in place. In summary, the lack of a safety net is what makes Mexico an interesting case study for the impact of the Covid-19 pandemic on fertility outcomes, as uncertainty and individuals’ narratives of the future were greatly modified during this time period.

Background

The Covid-19 pandemic in Mexico

The initial onset of the Covid-19 pandemic occurred from March to May 2020. More specifically, it began when the World Health Organization officially declared Covid-19 to be a pandemic on 11 March 2020. Almost two weeks later, Mexico’s official health authority, the General Health Council, announced an immediate countrywide lockdown on 23 March 2020 (CSG 2020a). This stay-at-home order lasted until 30 May 2020, throughout Mexico (CSG 2020b). Beginning in June, every state transitioned into a local traffic-light system, which allowed state-by-state differences in movement restrictions. This continued until December 2020, when the second wave of Covid-19 hit Mexico and states began to implement more severe restrictions on mobility.

Containment and closure policies harmed economic activity (Hoehn-Velasco et al. 2022, 2021a). Namely, Mexico’s gross domestic product (GDP) declined by −0.9 per cent during the first quarter of 2020, contracting even more (−17.9 per cent) in the second quarter (during the national lockdown), recovering by 13.5 per cent during the third quarter (after the end of the national lockdown), and expanding by 3.5 per cent during the last quarter of that year (OECD 2022). This economic recuperation stagnated for the next four quarters of 2021, in which GDP changed by 0.8 per cent, 1.0 per cent, −0.7 per cent, and 0.0 per cent, respectively. This meant that by the end of 2021, Mexico’s GDP was still 3.8 per cent below its pre-pandemic level (OECD 2022). This translated into severe economic consequences for Mexican households, including sharp reductions in employment and income (Hoehn-Velasco et al. 2022). In contrast to most middle- and high-income countries, Mexico’s government provided minimal assistance in the form of aid during the pandemic: this totalled less than 0.5 per cent of GDP (Hale et al. 2020). Other than frontloading payments of old-age and disability pensions by four months and issuing a few loans that charged interest to firms and workers, there was no additional welfare support during that period.

Some relief came in 2021 with the vaccination campaign rollout. The campaign began with the older population and transitioned to younger individuals. Mexico set a goal of vaccinating all adults aged 18+ before the end of 2021. Nevertheless, the vaccination campaign arrived too late, as cumulative confirmed Covid-19 deaths came close to 300,000 by the end of 2021 (Mathieu et al. 2022). As testing was limited in Mexico, that estimate was likely biased. In fact, excess mortality in Mexico during the 2020–21 period was over 610,000 deaths (Mathieu et al. 2022). This made Mexico one of the countries hardest hit globally by the Covid-19 pandemic.

Fertility and newborn health in Mexico

Fertility in Mexico has been on a downward trend for the past few decades. The TFR dropped from 3.1 births per woman in 1994 to 2.1 births per woman in 2019 (World Bank 2020). According to INEGI, there were close to 800,000 fewer births recorded in 1994 (2.9 million) than in 2019 (2.1 million), despite the population continuing to increase from 90 million to 127 million (INEGI 2020).

The fertility decline has come as a consequence of several factors: most notably, a lower teenage birth rate (e.g. Murillo-Zamora et al. 2019). In 1994, the share of mothers aged 19 years or younger was 16.9 per cent (INEGI 2020). By contrast, in 2019, said share had fallen to 15.1 per cent, a drop of 1.8 percentage points (INEGI 2020). Women’s higher use of contraceptives has been another important factor, although in 2018 there was a significant drop in contraceptive use. Because births have continued a downward trend, the factors contributing to the fertility decline have most likely been multifactorial. For example, increases in violent crime have also been associated with reduced instances of teenage pregnancy (Tsaneva and Gunes 2020). In all, this fertility decline has followed a similar trend to that of the United States (US) (Kearney and Levine 2015).

In terms of newborn health, like the majority of countries, Mexico has seen neonatal mortality drop substantially during recent decades. Specifically, the neonatal mortality rate fell from 19.82 newborn deaths per 1,000 live births in 1994 to 8.59 in 2019 (UNICEF 2020). This improvement in newborn health has been accompanied by an equal amelioration in the maternal mortality ratio, which declined from 55 to 36 maternal deaths per 100,000 live births in a matter of 15 years, between 2000 and 2015 (UNICEF 2020). In the following section, we describe the data and methods used to investigate whether fertility and newborn health changed during the Covid-19 pandemic in Mexico.

Data and methods

Data

We use administrative data on all births occurring in Mexico, as collected by the Ministry of Health. These data contain monthly statewide information on deliveries, birthweight, and gestation. Using this information, we generate our outcomes: the GFR and rates of low birthweight, high birthweight, term gestation, very premature birth, and premature birth. An unfortunate limitation of these data is that they do not include information on maternal age or other risk factors.

For fertility, we calculate the monthly GFR as the number of births per woman of reproductive age, on an annualized basis:. (1) $\begin{array}{rl}& \mathrm{Monthly}\mathrm{GFR}\\ & ={\displaystyle \frac{\mathrm{Monthly}\mathrm{live}\mathrm{births}}{\mathrm{Female}\mathrm{population}\mathrm{of}\mathrm{reproductive}\mathrm{age}\left(15\text{–}44\right)}}\\ & \times 1,000\times 12\end{array}$(1) For the remainder of the newborn health characteristics, we take the rate of each outcome per 1,000 births. We focus on low birthweight (<2,500 g, capturing prematurity), high birthweight (4,000 g or more, a proxy for maternal conditions, such as gestational diabetes, worsening over time), term gestation or longer (37+ weeks), premature birth (32–36 weeks), and very premature birth (<32 weeks).

We compare the effects of the pandemic over six years of data: 2016, 2017, 2018, 2019, 2020, and 2021. The months from January 2016 to November 2020 (the pre-pandemic period) capture fertility and newborn health outcomes from conceptions prior to the onset of the pandemic. Our pandemic period is made up of months from December 2020 to December 2021, which represent outcomes from conceptions that occurred during the pandemic (as babies born at term between December 2020 and December 2021 correspond to conceptions from March 2020 to March 2021). To address seasonality in fertility and in newborn health outcomes, we consider the entire period from January 2016 to December 2021, which allows us to account for month fixed effects.

We present the evolution of newborn health characteristics and fertility in Figure 1. Figure 1(a) descriptively illustrates our main finding, where the general fertility rate fell but then slightly recovered. By contrast, rates of low birthweight and high birthweight increased during that period (Figure 1(b)–(c)).

Figure 1 Monthly evolution of fertility and newborn health characteristics, Mexico, January 2016 to December 2021

Notes: The GFR is per 1,000 females of reproductive age (15–44). Other rates are per 1,000 births. Shaded area represents conceptions during the stay-at-home order. Source: Ministry of Health, Mexico.

We also quantify this decline in the descriptive statistics in Table 1. Fertility dropped from 64.1 per 1,000 females of reproductive age in the January 2016 to November 2020 period to 51.9 in the December 2020 to December 2021 period, a 19 per cent decline. However, fertility was already declining before the pandemic, which suggests that accounting for these pre-existing changes may be essential in isolating the direct impact of the pandemic itself. Newborn health characteristics also changed during the pandemic: low birthweight instances increased by 44 per cent, high birthweight instances rose by 16 per cent, cases of prematurity (by gestational age) increased by 12 per cent, and very premature births increased by 14 per cent.

Table 1 Descriptive statistics: fertility and newborn health characteristics in Mexico, pre-Covid and Covid periods

Rate	Mean		Difference (Covid minus pre-Covid value)
	Pre-Covid period	Covid period	Absolute	Percentage change
General fertility rate	64.11	51.89	−12.22***	−19.1
Low birthweight	64.19	92.58	28.39***	44.2
High birthweight	31.12	35.95	4.83***	15.5
Term gestation	925.84	916.94	−8.90***	−1.0
Premature birth	74.16	83.06	8.90***	12.0
Very premature birth	8.82	10.08	1.26***	14.2
N	1,888	416	2,304

*p < 0.1, * p < 0.05, ***p < 0.01.

Notes: The GFR is per 1,000 females of reproductive age (15–44). Other rates are per 1,000 births. Unweighted means are reported. The pre-Covid period includes January 2016 to November 2020. The Covid period includes December 2020 to December 2021.

Source: Ministry of Health, Mexico.

Empirical strategy: Event-study design

We rely on an event-study specification to consider the dynamic effect during the pandemic. Using an event study has two advantages. First, it allows us to test the parallel trends assumption directly using the months leading up to the pandemic. This assumption is necessary to establish causality. Second, using a difference-in-differences approach would give us only the average of the effect over the entire pandemic period, while the event study yields the month-by-month impact from the beginning of the pandemic. It is essential to consider this dynamic impact of the pandemic, as many outcomes likely follow a U-shaped pattern (Silverio-Murillo et al. 2020; De la Miyar et al. 2021; Hoehn-Velasco et al. 2021b). If this U-shaped pattern occurred for the GFR, the difference-in-differences estimator could give us an average impact close to zero (Wolfers 2006; Goodman-Bacon and Marcus 2020).

Formally, the event-study specification for state s observed in month m and year y appears as: (2) $\mathrm{Log}(\mathrm{fertility}{)}_{\mathrm{smy}}=\sum _{\begin{array}{c}q=-11\\ q\ne -1\end{array}}^{12}{\text{β}}_q\mathrm{Covid}\text{-}19_{\mathrm{yq}}+{\alpha }_s+{\gamma }_m+{\eta }_y+{ϵ}_{\mathrm{smy}}$(2) where Log(fertility)_smy is the natural log of the fertility rate. We focus on the GFR as our primary outcome of interest but also study newborn health characteristics, consisting of rates of low birthweight, high birthweight, term gestation, premature birth, and very premature birth.

Covid-19_yq is a set of dummy variables that take the value of one for each month q before and after starting the lockdown. Months q = 0 and onward represent conceptions that likely occurred during the pandemic, where the parents would have had the information to respond to the pandemic. A limitation of our data is that they are presented in aggregates, thus we cannot derive the true conception month. Thus, December 2020 represents the first month of the pandemic period (q = 0) and captures births plausibly conceived in March 2020. The specification continues until q = 12, which represents 12 months after the pandemic began (i.e. December 2021, representing births conceived in March 2021). For the pre-pandemic periods, we extend the event study to conceptions that occurred throughout 2020, including 11 pre-Covid periods (back to q = −11). To avoid multicollinearity, we follow the event study literature and exclude the month before the first period in which the event occurred (q = −1). Further, it is worth noting that we have no staggered treatment timing, as Mexico’s national lockdown started on 23 March for the entire country. Thus, recent critiques of two-way fixed effects with variation in treatment timing do not apply in our setting (Goodman-Bacon 2021).

We also include several time and place fixed effects: ${\alpha }_s$ are state fixed effects; ${\gamma }_m$ are month-specific fixed effects; and ${\eta }_y$ captures the year fixed effects. However, to avoid perfect collinearity, we must combine the years 2020 and 2021 together as a single dummy. This avoids perfect collinearity because the event study is fully saturated across 2020 and 2021. Finally, ${ϵ}_{\mathrm{smy}}$ represents the regression error, which we cluster at the state level (we also provide the wild cluster bootstrapped p-values in the robustness checks, see Appendix Tables A.1 to A.3). In our preferred specification, we choose not to include state-specific linear trends. Borusyak et al. (2021) showed that including unit-specific trends contaminates the dynamic treatment effect in event-study models. However, we include results where we control for state-level annual trends over the pre-pandemic period in the Robustness checks subsection.

Results

Main findings

We show the findings from equation (2) in Figure 2. The results show a clear time-varying dynamic effect, emphasizing the importance of the event-study specification (as opposed to a difference-in-differences design). All considered outcomes show a clear change in the pandemic period.

Figure 2 Event-study results for fertility and newborn health characteristics, Mexico, 2020–21

Notes: Plotted coefficients are event-study dummy variables, β_q. The event study considers 2016–to 2021, with November 2020 as the omitted period: this period represents conceptions that likely occurred in February 2020 (if carried to term). The pre-pandemic period includes January 2016 to November 2020. The pandemic period includes December 2020 to December 2021. Solid diamonds and lines represent point estimates; dotted lines display 95 per cent confidence intervals. Baseline fixed effects are included at the state, month, and year levels. Outcomes include the natural log of the GFR (per 1,000 females of reproductive age, 15–44); the remaining outcomes are the natural log of each rate per 1,000 births. Robust standard errors are clustered at the state level. Shaded area represents conceptions during the stay-at-home order. See Tables A.1–A.3 for the wild cluster bootstrapped p-values.

Source: As for Figure 1.

Our primary outcome, the log of the GFR, shows an evident decline in conceptions over the initial months of the pandemic (Figure 2(a)). For conceptions occurring in April and May 2020 (during the stay-at-home order), there was a corresponding 11–12 per cent reduction in fertility in January and February 2021. After the initial lockdown period ended (conceptions from June 2020), fertility gradually recovered, as seen between March and June 2021. By December 2021, fertility was near its original levels but not quite back to baseline. The plotted results also suggest a clear break in the log of the GFR with the onset of the pandemic, with pre-pandemic periods showing little evidence of pre-existing trends.

Turning to newborn health characteristics, we see that the pandemic is associated with a substantial change in birthweight (Figure 2(b)–(c)). Over the initial months of the pandemic, conceptions leading to births with high birthweight (4,000 g or more) and low birthweight (<2,500 g) both increased. Instances of low birthweight newborns remained elevated until the end of the series in December 2021. By contrast, instances of high birthweight gradually began declining again over the pandemic period.

Focusing on gestation, we note that premature deliveries (32–36 weeks; Figure 2(e)) follow a similar pattern to low birthweight, as expected. Cases of premature newborns started to increase for conceptions that occurred during the stay-at-home order. The rate of prematurity grew over the course of the pandemic, increasing by more than 20 per cent by the end of the analysis period. Finally, very premature deliveries (<32 weeks) also increased during the pandemic (Figure 2(f)), but the change is slightly less evident than for premature deliveries. Reflecting the increase in preterm deliveries, deliveries occurring at term or later (37+ weeks) also declined from February to December 2021 (Figure 2(d)).

Robustness checks

The initial results in Figure 2 show that the pandemic is associated with a stark break in the GFR trend and a change in newborn health characteristics. These findings also show little evidence of pre-existing trends. Still, we test whether our findings are robust across a battery of checks to ensure their validity. In particular, we perform the following checks: (1) considering whether the results are consistent across regions; (2) omitting Mexico City from the results, as Mexico City varies substantially from the remainder of Mexico in terms of population size; (3) presenting the event study with state-specific linear trends; (4) showing the event study with weights; (5) presenting the linear rates instead of the natural log of the rates; (6) displaying the results adding state-by-month fixed effects; (7) calculating the wild cluster bootstrapped p-values and upper/lower confidence intervals; (8) verifying that the coefficients are not statistically significant only by chance, using a false discovery test; and (9) using a bounding method to check for omitted variable bias.

First, we test whether the results are consistent across the different regions of Mexico. Figures A.1 and A.2 show our main outcomes across each of Mexico’s four INEGI-specified regions (‘Center’, ‘Center West’, ‘North’, and ‘Southeast’). The GFR fell across all regions, with the smallest declines in the Center and Center West regions and the largest decreases in fertility in the North and Southeast. Low birthweight increased similarly across regions, although high birthweight increased the most in the North of Mexico.

Second, we show several iterations of our event study in Figures A.3–A.7. These checks include: excluding Mexico City (Figure A.3); adding state-specific annual linear trends (Figure A.4); adding population weights (Figure A.5); using linear rates (Figure A.6); and adding state-by-month fixed effects (Figure A.7). Results from all the additional specifications largely reflect those from the baseline event study in Figure 2.

In Tables A.1, A.2, and A.3, we present the bootstrapped p-values and confidence intervals for the six outcomes. Since standard errors are downward biased with a low number of clusters (≤50), we conduct a wild cluster bootstrap procedure, as described in Cameron et al. (2008). For the most part, the results from the wild cluster bootstrap method confirm our main findings.

Next, we consider multiple hypothesis testing. In the case of the event studies, we estimate six equations. This multiple hypothesis testing considers whether some of the coefficients appear statistically significant by chance. To reduce the probability of false rejections of the null hypothesis, we compute sharpened false discovery rate q-values (Anderson 2008). The results using this method are presented in Table A.4. The p-values are presented in parentheses, and the sharpened q-values are in brackets. As expected, certain q-values are higher than the p-values. However, statistically significant coefficients (using p-values) remain generally statistically significant when using q-values.

We also implement a bounding method to check how sensitive the results are to omitted variable bias (Oster 2019). Some control variables not included in the analysis (e.g. maternal age) may affect our results. Unfortunately, variables such as maternal age are not available in our main data set but only in a separate data set that considers the health characteristics of newborns. In addition, when using an event-study method with covariates included, the event-study model can produce misleading calculations (Powell 2021). Instead, to ensure results are insensitive to omitted variable bias, we first verify that the assumption of no trend in the variable of interest prior to the event (the pandemic) is met. The results presented in Figure 2 suggest little evidence of a pre-existing trend leading up to the pandemic period.

Further, to confirm the previous results, we use a bounding method to generate a bound around the parameter of interest. If the bound excludes zero, then the results from the regression are robust to the problem of omitted variable bias. Tables A.1, A.2, and A.3 present the results using the bounding method proposed by Oster (2019); they show that the majority of the main findings continue to be statistically significant, particularly for January to December 2021. In the case of fertility, the months of January, March, August, September, and October pass this robustness test. For the remaining variables (low birthweight, high birthweight, term gestation, very premature birth, and premature birth), the majority of the results are robust to omitted variable bias (where the bounds exclude zero) for January to December 2021.

Additional results

We then consider whether delivery locations changed during the pandemic. Deteriorating conditions in hospitals and a limited supply of physicians may push individuals to deliver in settings other than hospitals (e.g. at home). We test whether this was the case in Figure 3.

Figure 3 Event-study results by attendance, location of delivery, and maternal age, Mexico

Notes: Plotted coefficients are event-study dummy variables, β_q. The event study considers 2016–to2020, with February 2020 as the omitted period (shown by the vertical line). The event study compares deliveries that occurred in 2020 against deliveries from 2016 to 2019. Solid diamonds and lines represent point estimates; dotted lines display the 95 per cent confidence intervals. Baseline fixed effects are included at the state, month, and year levels. Robust standard errors are clustered at the state level. The shaded area is the lockdown period, and the horizontal line represents zero.

Source: Ministry of Health, Mexico; INEGI (2020).

To calculate delivery attendant and location, we use maternity data available from INEGI (2020). Because these data are not available for 2021 at the time of writing, we compare deliveries that occurred in 2020 against a comparison group: deliveries in the four-year period 2016–19. We follow a similar event-study specification as equation (2). We include fixed effects by state, month, and year. In this case, Covid-19_yq is a set of dummy variables from January (q  =  −2) to December (q  =  9) 2020. March represents the month when the event started (q  =  0). To avoid multicollinearity, we exclude February 2020 (q  =  −1).

While our main data from the Ministry of Health are collected directly in hospitals, the INEGI data used here report births registered in civil registry offices. INEGI itself recommends that the data be used with caution as they may suffer from under-reporting. In particular, many families avoided going to civil registry offices during the pandemic due to fear of infection.

Figure 3 shows that deliveries by physicians declined during the pandemic. Simultaneously, midwife-attended births increased substantially (in percentage terms). However, we note that the proportion of midwife-attended deliveries is quite small in the data, representing only 1.5 per cent of deliveries before 2020.

For home deliveries, the model shows a similar-sized (percentage-change) increase during the pandemic. However, as with midwife-attended deliveries, this change represents only a small absolute change. Home deliveries composed 1.7 per cent of deliveries prior to 2020. Still, these findings align with expectations, suggesting that more births occurred in settings other than hospitals and were attended by non-physicians during the pandemic.

Finally, we also consider whether maternal age changed during the pandemic. A limitation of this approach is that the results do not reflect the same period as the main analysis. These results show only deliveries that occurred during the pandemic rather than those conceived during the pandemic. Still, the findings may be instructive for seeing any ongoing changes in maternal age. We focus on deliveries to the two groups at risk: women aged 35 + and teenagers. The former group captures mothers of advanced maternal age, which is associated with additional risks to the mother and newborn (Fretts and Simpson 2019).

The bottom two panels of Figure 3 show little clear change in maternal age. There was no increase in deliveries to women of advanced maternal age (35+). Teen deliveries show a slight increase, but teen deliveries were significantly higher before the pandemic began as well (2016–19). These findings suggest that the main changes in newborn health characteristics were likely not driven by changes in maternal risk by age.

Discussion

The Covid-19 baby bust and subsequent rebound, as well as the deterioration in newborn health in Mexico during the pandemic, are likely to have a series of consequences that also apply to other countries. First, the case study of Mexico—a country with little social spending and no additional support during the pandemic—describes a clear situation in which a steep fertility decline occurred in the absence of policies to support households and employment during strenuous times. Existing evidence suggests tremendous heterogeneity, as there were important differences in socio-economic conditions at the time of the pandemic, all of which may transform the post-pandemic fertility trajectory (Aassve et al. 2020). This is already apparent in studies analysing the Covid-19 impact on fertility across countries (Aassve et al. 2020). Even single-country studies, such as those for the US, have shown regional heterogeneity in baby busts and rebounds, much of which can be explained by household income (e.g. financial distress) and unemployment rates (Kearney and Levine 2022). In fact, regional differences in Mexico also suggest an association between socio-economic conditions and fertility, as the Southeast states of Mexico experienced the largest decline in economic activity and formal employment (Banco de México 2022); however, we cannot infer causality as many of the economic variables are preliminary and subject to change.

Second, and most importantly, the newborn health consequences are likely to branch out—later in life—into a series of problems for this particular generation conceived during the pandemic. Researchers have explained the effects of previous pandemics, such as the 1918 influenza pandemic, with the foetal origins hypothesis, although this is not exclusive to pandemics. The foetal origin hypothesis suggests that the effects of foetal conditions are persistent, latent for many years (e.g. heart disease does not emerge until middle age), and reflective of biological mechanisms such as the epigenome (Almond and Currie 2011). One of the areas of research for scholars working on foetal origins is prenatal care as an ideal tool for human development (Almond and Currie 2011). The cost–benefit analysis of additional prenatal care during a pandemic is something that researchers should continue to study.

Finally, as more data become available, future research for other middle- and high-income countries should be able to assess whether fertility patterns look similar to or different from the ones in Mexico. Further, scholars should evaluate whether policy interventions can reverse the baby bust, like in the US, which ended up gaining additional births during the rebound phase (Kearney and Levine 2022). As we enter a post-Covid era in which fertility rates begin to recover in many countries around the world, researchers should pay special attention to the determinants of fertility plans that were particularly hit by the pandemic: for example, childcare, labour markets, and social norms (Doepke et al. 2022).

Conclusion

This paper uses Mexico’s national administrative data to consider whether fertility rates and newborn health outcomes changed during the Covid-19 pandemic. We find a short-term decline in the GFR, which initially declined by 11–12 per cent but nearly returned to baseline by the end of 2021. Newborn health characteristics also changed during the pandemic, indicating a decline in infant health at birth: the findings show a persistent rise in low birthweight and prematurity during the pandemic. Instances of high birthweight also increased briefly but then began declining again. Last, we find an increase in home deliveries and midwife-attended deliveries, but little clear change in maternal age. A caveat to this last statement is that our results for socio-economic differentials in Covid-19’s impact on fertility do not reflect the same period as the main fertility analysis; thus, they cannot be compared directly.

The findings of this study open up two critical areas of research. First, our findings confirm existing evidence suggesting that after a shock, such as a pandemic or economic crisis, the initial decline in fertility is followed by a subsequent recuperation in fertility (e.g. Ullah et al. 2020). However, our data series stops too early to determine whether a baby boom is on the horizon. Second, babies born during the Covid-19 pandemic were more likely to be born premature and with low birthweight, which will likely have long-term consequences (based on prior pandemics; see e.g. Almond 2006; Beach et al. 2022). Further studies tracking these affected children, who experienced the pandemic in utero, are warranted as the data progress.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 Adan Silverio-Murillo and Judith Senyancen Méndez Méndez are based in the School of Government, Tecnologico de Monterrey. Lauren Hoehn-Velasco is based in the Andrew Young School of Policy Studies, Georgia State University. Jose Roberto Balmori de la Miyar is based in the Business School, Universidad Anáhuac.

2 Please direct all correspondence to Adan Silverio-Murillo, School of Government, Tecnologico de Monterrey, Av. Revolución 756, CDMX, 03700, Mexico; or by E-mail: [email protected].