The stratified effect of extreme temperatures on birth weight: the role of energy prices

Abstract

This paper fills a notable gap in the existing literature on birth outcomes and extreme temperatures by examining the causal moderating effect of energy prices on the impact of extreme temperatures during pregnancy on birth weight. It uses a sharp increase in energy prices that occurred in Spain in March 2021 as an identification strategy and incorporates a new weather dataset that increases the number of monitors from which temperature information is derived in previous research by more than sixfold. The results show that the negative effects of extreme temperatures on birth weight are amplified at higher energy prices, especially for mothers of lower socioeconomic status. By focusing on how energy prices moderate the impact of extreme temperatures on birth outcomes, opportunities arise for policy interventions aimed at reducing health inequalities at birth. Moreover, given the critical role that events in utero play in individuals’ later development, such policies have the potential not only to reduce health inequalities at birth, but also to address broader inequalities in long-term outcomes.

Introduction

A large body of literature has highlighted the importance of studying the effects of ambient temperature during pregnancy on birth outcomes (Andriano, 2023; Auger et al., 2017; Chen et al., 2020; Cho, 2020; Dadvand et al., 2011; Davenport et al., 2020; Hajdu & Hajdu, 2021). However, it is surprising how little importance is attached in this literature to a crucial moderating factor in this relationship: energy prices. In the context of an affluent country, I argue that energy prices play a moderating role by influencing the affordability of heating and air conditioning, allowing mothers to protect themselves from the negative effects of extreme temperatures during pregnancy. Furthermore, as changes in energy prices are more prone to impact vulnerable mothers, it is likely that an increase in energy prices will exacerbate existing disparities in children’s health at birth linked to maternal socioeconomic status, thereby increasing socioeconomic inequalities.

This study fills a gap in the literature on birth outcomes and extreme temperatures. In contrast to previous research, this study focuses on the context in which extreme temperatures play a more important role in birth health. In particular, this research draws on the concept of energy poverty and focuses on the moderating role that energy prices may have on the relationship between extreme temperatures and birth outcomes. As this energy price-dependent protective mechanism has not yet been investigated, previous literature also lacks evidence on how energy prices affect protection from extreme temperatures differently depending on the socioeconomic status (SES) of the mother during pregnancy. That is, high SES mothers are more able to cope with extreme temperatures by heating in winter and air conditioning in summer than mothers with low SES, who are more likely to experience energy poverty, reflecting differences in mothers’ purchasing power. In addition, low SES mothers are more likely to live in areas with poorly insulated buildings, increasing their vulnerability to extreme temperatures. Furthermore, rising energy prices put additional strain on their budget, leading to increased stress and a lower quality of goods and services consumed during pregnancy, potentially affecting the health of newborns.

Despite the theoretical appeal of energy prices moderating the impact of extreme temperatures, the lack of suitable variation of these prices within countries has prevented a comprehensive study of this critical factor. To overcome this problem of lack of variation, I focus on Spain as a case study, where I take advantage of a sharp increase in energy prices in March 2021. Specifically, electricity prices increased by about 250% and gas prices by about 530% in just 9 months in 2021. Following (Duflo, 2001), I analyze how this increase affected health at birth, depending on whether mothers faced extreme temperatures during pregnancy and their SES. In addition, I create a new weather dataset by taking into account the weather monitors owned by autonomous communities (NUTS 2), which significantly increases the number of monitors used to calculate the weather experienced by the mothers compared to previous studies (Achebak et al., 2019; Conte Keivabu, 2022; Follos et al., 2021). Hence, Spain offers a unique opportunity to study this effect, for three main reasons: first, the significant increase in energy prices in recent years. Second, the large temperature variations between regions and seasons allow for a precise identification of the target effect. Third, the availability of high-quality data on birth outcomes and weather allows a comprehensive analysis for the entire population of newborns.

This study has important policy implications. While the global nature of climate change means that governments cannot directly reduce temperatures to tackle health inequalities at birth, they can focus on mitigating the effects of extreme temperatures. Policies could include cash transfers to low SES mothers during pregnancy to assist with energy bills, as well as education campaigns that emphasize the importance of avoiding extreme temperatures for maternal and newborn health. These measures can help reduce the negative impact of extreme temperatures on birth outcomes. Recognizing the critical role of events in utero in shaping future individual development suggests that such interventions can not only reduce health inequalities at birth, but also address disparities in long-term development (Almond & Currie, 2011; Behrman & Rosenzweig, 2004; Gluckman & Hanson, 2006).

In summary, the objective of this study is to analyze the causal impact of energy prices on health at birth as a function of maternal exposure to extreme temperatures and SES. The analysis carried out allows to better understand whether and for which socioeconomic group energy prices have the strongest effect on birth outcomes through their moderating effect on extreme temperatures. This is particularly important in a process of global warming associated with rising energy prices and inequalities.

Background and hypotheses

Extreme temperatures and birth outcomes

While there is no clear conclusion regarding the biological mechanisms by which extreme temperatures impact birth outcomes, previous studies have suggested various factors such as heat stress due to gestational weight gain (e.g., Dadvand et al., 2011; Lin et al., 2017; Wells, 2002), disturbed sleep (Lin et al., 2017; Strand et al., 2011), and fetal cell death (Auger et al., 2017).

Despite the lack of consensus on the underlying mechanisms, the existing literature consistently shows that extreme temperatures during pregnancy have adverse effects on birth outcomes. These outcomes include congenital heart defects (Agay-Shay et al., 2013; Auger et al., 2017), length of gestation and preterm birth (Conte Keivabu & Cozzani, 2022; Dadvand et al., 2011; Strand et al., 2012), and birth weight and the likelihood of low birth weight (Andriano, 2023; Chen et al., 2020; Cho, 2020; Conte Keivabu & Cozzani, 2022; Davenport et al., 2020; Deschênes et al., 2009; Hajdu & Hajdu, 2021).

This study focuses on the effects of extreme temperatures on birth weight, a key variable in the literature. Researchers such as Andriano (2023); Chen et al. (2020); Deschênes et al. (2009); Hajdu and Hajdu (2021) have investigated this relationship. Overall, their findings suggest that exposure to extreme temperatures during pregnancy is associated with negative effects on birth weight. Deschênes et al. (2009); Hajdu and Hajdu (2021) particularly emphasize that the greatest effects are observed during the second and third trimesters of pregnancy. Andriano (2023); Chen et al. (2020) similarly point out that the strongest effects of extreme temperatures occur during the third trimester of pregnancy.

While previous research has examined the effects of extreme temperatures on birth outcomes, there is a notable gap in understanding how energy poverty interacts with this relationship. In particular, little attention has been paid to how energy poverty, which is influenced by rising energy prices, might exacerbate the impact of extreme temperatures on the health of the newborn.

Poverty and energy poverty

Poverty is a prominent topic in the social sciences. Recently, a new aspect of poverty has emerged: energy poverty, which refers to the minimum energy consumption required to meet basic needs. (González-Eguino, 2015; Reddy et al., 2000). Energy poverty is critical here because when energy prices increase, mothers struggle to meet their basic energy needs, which can exacerbate the impact of extreme temperatures on the health of their newborns.

Previous studies have repeatedly shown the significant negative impact of energy poverty on individual health (e.g., Oliveras et al., 2021; Pan et al., 2021; Thomson et al., 2017). Narrowing the perspective, the impact of energy poverty on the population as a whole includes pregnant mothers, whose health may suffer, potentially affecting their newborns (Bhutta et al., 2010). However, although researchers from the fields of sociology, medicine, and economics have studied the relationship between energy poverty and population health, the impact on health at birth has been overlooked. That is, to the best of my knowledge, there is no research on how energy affordability may affect the health of the mother during pregnancy, and therefore the health of the newborn.

Extreme temperatures and birth outcomes: hypothesizing the role of energy prices and SES

The main objective of this study is to fill a gap in the literature by examining how energy prices moderate the impact of extreme temperatures on newborns’ health, particularly birth weight—the main dependent variable in this study and a widely used measure in newborns health research (Almond & Currie, 2011; Conley & Bennett, 2000; Gluckman & Hanson, 2006). I argue that energy prices have a significant impact on the affordability of heating and cooling, affecting the ability of mothers to maintain an adequate temperature at home. In particular, the importance of energy prices is likely to be more pronounced when mothers suffer from adverse weather conditions during pregnancy, that is, the increase in energy prices could have a greater impact on home thermoregulation when weather conditions worsen. Consequently, I argue that the impact of extreme temperatures on birth outcomes depends on the energy prices faced by mothers.

${\textbf {Hypothesis 1:}}$:

Exposure to adverse weather conditions during pregnancy affects newborns’ health more negatively when energy prices are high.

I also examine how energy prices affect mothers with different SES. I expect a greater impact on newborns born to more vulnerable mothers for several reasons. First and foremost, low SES families are likely to have greater difficulty regulating the temperature in their homes when energy prices rise than high SES families. Consequently, low SES mothers may suffer more from extreme temperatures during pregnancy than their high SES counterparts. Second, rising energy prices pose a financial challenge for mothers with limited economic resources, as a greater proportion of their funds must be spent on energy bills, which may affect the quality of other products and services that low SES mothers consume during pregnancy. For example, this could affect the quality of their diet, a factor known to influence birth outcomes (Günther et al., 2019) or their choice of transportation, such as switching from taxis to public transport. Finally, the financial burden could increase stress levels during pregnancy, which could have a negative impact on the health of the newborn (Cozzani et al., 2022; Torche, 2011).

${\textbf {Hypothesis 2:}}$:

Higher energy prices exacerbate the negative effects of adverse weather conditions during pregnancy on the health of newborns especially for those born to mothers of low socioeconomic status.

The case of Spain

Spain is an ideal case study for testing the hypotheses outlined for several reasons. First, the country has experienced an unprecedented huge increase in energy prices in recent years, providing a valuable opportunity to examine the impact of energy prices as a moderating factor in the relationship between extreme temperatures and birth outcomes. Second, there is considerable temperature variation between regions and seasons in Spain, which provides a recognizable basis for identifying the targeted effects. For example, the average January temperature from 1981 to 2010 in Leon (a city in the north) is 3.2 °C, while in Malaga (a city in the south), it rises to 12.1 °C. Similarly, the average temperature in August is 19.6 °C in Leon and 26 °C in Malaga. Further evidence of the temperature variations can be found in Appendix A. Third, the availability of high-quality data on birth outcomes and weather allows for a reliable analysis of the effects of interest on the entire newborn population.

For these reasons, Spain is ideal for examining the relationship under investigation. In this section, I discuss the energy context in Spain, looking at energy contracts, energy poverty, and the evolution of energy prices.

Energy market

In the Spanish electricity market, households essentially have two tariff options: default rates and retail market rates. Default rates fluctuate hourly based on prices set daily by the government, which are linked to the wholesale electricity market. Retail market rates are fixed prices set by the companies for individual consumers and are usually reviewed annually. The gas market offers similar choices with default rates set by the government changing quarterly and retail rates varying with each contract review. Energy bills are issued monthly or bi-monthly depending on the contract type, with prices on the retail market generally higher than default rates.

Although the choice of tariff can influence energy consumption, according to a survey conducted by the competent agency (CNMC), only 30% and 22.7% of households were aware of the differences between default rate and retail tariffs for electricity and gas respectively, with no significant differences in household characteristics, suggesting a uniform response to electricity price increases regardless of the tariff type (Fabra et al., 2021).

Energy poverty

The Ministry of Ecological Transition stated that between 3.5 and 8.1 million people in Spain were affected by energy poverty in 2017, depending on the indicator used. This corresponds to 7.4 to 17.1% of the total population. Figure 1 shows the evolution of the percentage of households that have difficulty maintaining an adequate temperature in their homes from 2016 to 2021. Notably, between 2020 and 2021, a significant increase in energy poverty can be observed, as the number of households that have problems maintaining an adequate temperature in their homes increases by around 57%.

Fig. 1

Energy poverty. Notes: The graph shows the evolution in the proportion of households reporting to have difficulties keeping their home at an adequate temperature. Source: National Institute of Statistics: Living Conditions Survey

Fig. 2

Energy prices and Google Trends. Notes: Dark grey lines show the evolution over time of electricity prices (a) and gas prices (b) in Spain. Sources: Spanish Electricity Network (a) and Iberian Gas Market (b). The dashed lines indicate how frequently the terms “electricity price” (a) and “gas price” (b) are entered into the Google search engine, relative to the total Google search volume for that term from 2016 to the end of 2021 for Spain. A value of 100 indicates the maximum popularity of the term. Source: Google Trends

Energy prices and consumer price index evolution

Figure 2 shows the evolution of electricity and gas prices in Spain from 2016 to 2021. It is noteworthy that a significant price increase can be observed in March 2021, which serves as the basis for the identification strategy in this study. In the period from March to December 2021, electricity prices increased by around 250%, and gas prices by around 530%, which represents an unprecedented increase in Spain. This increase was mainly due to two reasons. Firstly, the increase in the price of CO2 emission allowances due to the goal of reducing greenhouse gas emissions in Europe, which had a direct impact on the cost of generating electricity using fossil fuel technologies. Secondly, the increased demand for gas on the Asian market, which is used to generate energy in combined-cycle power plants, led to higher prices on the international market, impacting both Asia and Europe. According to the Bank of Spain, half of the increase in electricity prices is due to the increase in the price of natural gas, while 20% is due to the increase in the price of CO2 emission allowances (Pacce et al., 2021).

Remarkably, this significant increase in energy prices coincides with the sharpest rise in energy poverty, as shown in Fig. 1. In addition, Appendix B shows the evolution of the annual change in the consumer price index by the ECOICOP group. It shows that the prices for electricity and gas increased the most in 2021, while the other prices remained relatively stable. This indicates that the increase in energy prices was not linked to the increase in other goods and services and suggests that the results of this study are not influenced by the increase in other prices.

A potential concern is that citizens may not be aware of energy prices, so they may delay their adjustments in consumption behavior at least until they receive subsequent bills or contractual updates. The dashed lines in Fig. 2 represent Google Trends for “electricity price” and “gas price.” They reflect public interest and show a positive correlation between energy prices and the popularity of these terms in Google searches suggesting that there is an awareness of energy price trends, which facilitates informed decision-making. In addition, Appendix C shows the annual number of articles published in El Mundo, one of the main Spanish newspapers, containing the terms “electricity price” and “gas price.” This pattern mirrors the trends observed in Fig. 2.

Data and variables

This section describes the data and variables used in this study. The primary dataset covers the entire population of births in Spain from 2016 to 2021 and provides information on the outcome variable (birth weight) and various characteristics of the newborns and their parents. In addition, a new weather dataset is created at the daily level to calculate the extreme temperatures experienced by mothers during pregnancy. Information on the characteristics of the municipalities in which the mothers live during pregnancy is also included.

Sources and definitions

This section explains the different data sources used for the analysis. The time span for all analyzes conducted in this study ranges from 2016 to 2021, as the classification of parents’ education in the birth database has changed from 2016 and no data is available beyond 2021. The data on births come from the Spanish National Institute of Statistics(INE). It is a comprehensive dataset with information on all births that took place in Spain between 2016 and 2021, organized by month and municipality.¹ Spain comprises 8131 municipalities, but the information on place of residence is only available for mothers living in municipalities with more than 10,000 inhabitants. For mothers living in smaller municipalities, which account for 16% of the sample, the place of residence is imputed to be either the municipality of registration of the newborn (if it has more than 10,000 inhabitants) or the municipality of delivery (if the two previous municipalities have less than 10,000 inhabitants).² Consequently, the dataset contains information on births from 768 municipalities, which corresponds to 98% of births in the sample years. However, as a robustness check, I re-estimate the main analysis, focusing only on mothers in municipalities with more than 10,000 inhabitants. The results, presented in Appendix N.1, align with the main findings.

In addition to the outcome variable (birth weight), this database contains information on various characteristics of the newborns, including sex, and of the mothers, such as age, education, marital status, employment status, nationality, birth order, and whether it is a singleton birth. I use several indicators to measure SES, including maternal education, employment status, marital status, and nationality. This approach aligns with previous research arguing for a multidimensional perspective, claiming that measures that take into account different aspects of living conditions provide a more robust assessment of poverty than a unidimensional indicator (Bossert et al., 2013; Nolan & Whelan, 2010). Furthermore, these SES indicators are selected based on data from the INE, which shows that low-educated, unemployed, and foreign individuals and single-parent households are more likely to be affected by poverty.

Mothers’ education is categorized into three groups: low (up to first level of secondary education), medium (beyond first level of secondary education but less than a bachelor’s degree), and high (at least a bachelor’s degree). Marital status is either married (including civil partnership) or unmarried. Employment status includes employed, unemployed, and other situations. Nationality is categorized as native or foreign.

I have compiled a comprehensive new weather dataset by combining the daily average temperatures from the weather stations operated by the State Meteorological Agency (AEMET) and those provided by the autonomous communities (AC). Previous studies on extreme temperatures in Spain rely exclusively on the data provided by AEMET or on the E-OBS grid dataset, which draws its information from the AEMET monitors (Achebak et al., 2019; Conte Keivabu, 2022; Follos et al., 2021). Therefore, some of these studies have chosen to analyze the data at provincial level (NUTS 3) instead of municipal level or focus on large cities with a higher concentration of monitors. In 2021, AEMET operates 246 stations, and the latest version of the E-OBS dataset uses 221 of these AEMET stations for the grid calculation. Previous research has remarked on the low density of monitors used to compute the E-OBS grid data in Spain in comparison to other European countries (Cornes et al., 2018; Hofstra et al., 2009). Furthermore, previous research has also shown that low monitor density areas have a lower accuracy on their estimates of grid data and tend to report over-smoothed estimates, especially for extreme values (Hofstra et al., 2009, 2010). The new dataset in this study significantly expands the number of stations to 1594 in 2021, a more than sixfold increase over previous data sources. Figure 3 shows a map with the locations of all weather stations used for the analysis, with AEMET stations marked in red and AC stations in blue. Although the results when using E-OBS data are consistent with the ones in the main analysis (see Appendix N.6), an additional discussion on the benefits of using monitors data rather than E-OBS data can be found in Appendix F.³

Fig. 3

Map of weather monitors. Notes: Blue dots represent the location of autonomous communities (AC) weather monitors. Red dots represent the location of AEMET weather monitors

Fig. 4

Climatic areas in cold and hots seasons. Notes: In the cold season, the warmest area is indicated by an x, while the coldest is represented by an E. In the hot season, the warmest area is indicated by a 4 and the coldest by a 1

I additionally control for the minimum distance in road kilometers between the population center of each municipality and a hospital to account for the services available to each mother. Finally, I adjust for COVID hospitalizations at the province-month level using data from the Spanish National Epidemiological Surveillance Network to capture any effect from conception to birth. This accounts for delays in family planning and impacts on newborn health due to maternal stress and reduced prenatal care during the pandemic (Cozzani et al., 2023). Additionally, prolonged time at home during the COVID pandemic may alter exposure to extreme temperatures, impacting pregnant women differently depending on the regulation of their work environment.

Sample construction

As already mentioned, the unit of analysis is each of the births that took place in Spain between 2016 and 2021, by municipality and month. The data on energy prices apply to the whole country, but vary over time. Weather data, however, varies across time and by monitor. I transfer data from weather monitors to the municipal level, as described below.

Using the daily average temperature data, I operationalize temperature at municipality level following Neidell (2004) and Currie and Neidell (2005). First, I calculate the population center of each municipality using the 2018 GEOSTAT 1 km$^2$ population grid dataset. Second, I calculate the distances between each municipality center and each weather monitor. Third, for each municipality, I derive a weighted daily mean temperature that includes the weather stations within a 20 km radius. The weights are determined by the inverse of the distance of each monitor to the center of the municipality, favoring monitors closer to the population center over those farther away. A schematic example can be found in Appendix G.⁴

To perform the analysis outlined in the following section, I categorize municipalities based on monthly weather conditions as either favorable or unfavorable. That is, I calculate the extreme temperature variables in the following way. Following Agay-Shay et al. (2013); Dadvand et al. (2011), I first distinguish between cold and hot seasons. The hot season lasts from April to September, while the cold season lasts from October to March. Appendix H contains a map showing the average temperature during each season by municipality. Second, I classify the municipalities to the climate areas during these seasons according to temperature degrees and hours of sunshine. In Spain, there are six climate areas in the cold season and four in the hot season. Figure 4 illustrates the distribution of climate areas during both seasons. Third, I determine extreme temperature events. A municipality is considered extremely hot on a day in the hot season if the mean daily temperature exceeds the 90th percentile of the temperature in the sample period (2015–2021) for that climate area. Similarly, a municipality is considered extremely cold on a cold season day if the mean daily temperature falls below the 10th percentile of the temperature in the sample period in that climate area.⁵ Fourth, the number of days with extreme temperatures in the mother’s municipality of residence is calculated for each trimester of pregnancy. Finally, mothers are categorized according to whether they were exposed to extreme temperatures in each trimester. Exposure to extreme temperatures is defined as more than 10 days of extreme temperatures on average per month in each trimester of pregnancy, which affects about 5% of mothers.

The specific threshold of 10 days is chosen with the argument that the effects of extreme temperatures are not linear. For example, the effects of a change from 1 to 2 days of extreme temperatures are not necessarily the same as a change from 10 to 11 days. The threshold of 10 days serves as a benchmark. However, to thoroughly analyze these nonlinearities, I also examine different thresholds by varying the number of days with extreme temperatures in the main body of the results. This approach allows for a more nuanced understanding of how varying the duration of exposure affects the results. As a robustness check, I modify the computation of extreme temperatures in several ways. First, I use the number of days with extreme temperatures in each trimester as the main explanatory variable. Second, I analyze bins of absolute temperature and count the number of days categorized as very cold, cold, comfortable, hot, and very hot for each trimester of pregnancy. I also adjust the percentile thresholds used to define extremely hot or cold days. The results of these additional analyses are presented in Appendix N.2 and consistently align with those of the main analysis.

Summary statistics

Appendix J contains the summary statistics for all variables used in the analysis. The average birth weight, expressed in grams, was around 3200 g during the study period. Newborns below 500 g or over 5000 g were excluded from the sample to avoid unfeasible births or coding errors (Seri & Evans, 2008). Figure 5 shows the distribution of birth weight.

Fig. 5

Distribution of birth weight in grams

Fig. 6

Number of extreme temperature days. Notes: The graph on the left shows the distribution of the number of days of extreme temperature in the monthly average suffered by mothers in the first trimester of pregnancy. The graphs in the middle and on the left show the same, but for the second and third trimesters of pregnancy, respectively

The mean value of the variable “T1: ExtrT” indicates the percentage of mothers living in municipalities with extreme temperatures during the first trimester of pregnancy. This also applies to the second and third trimesters. About 5% of mothers faced extreme temperatures in each trimester, meaning that they faced such conditions on average at least 10 days per month in each trimester. Figure 6 illustrates the distribution of the monthly average number of days mothers faced extreme temperatures during each trimester of pregnancy. The right-skewed distributions show that the majority of mothers did not experience extreme temperatures during pregnancy.

The mean value of the variable “After Peak” during each pregnancy trimester indicates the percentage of pregnancies in which the respective trimester occurred after the increase in energy prices. In other words, the mean value of the variable “After Peak” for the first (second/third) trimester indicates the percentage of pregnancies where the first (second/third) trimester occurred after March 2021. The data shows that for around 4% of births, the first trimester occurred after the energy price increase, while these percentages for the second and third trimesters are 8% and 13%, respectively.

Methodology

To determine the causal moderating effect that energy price has on the relationship between extreme temperatures and birth outcomes, I follow the analysis conducted in Duflo (2001). In this paper, the author uses differences across regions and differences across birth cohorts due to a shock to find the causal effect analyzed. The proposed benchmark model is as follows:

$$\begin{aligned} Y_{ijt}=\alpha +\sum _{q=1}^3(W_{qjt}\times T_{qt}\beta _q+W_{qjt}\eta _q+T_{qt}\phi _q+{\textbf {U}}_{qjt}\pi _q)+{\textbf {SES}}_{it}\lambda +{\textbf {X}}_{it}\sigma +\gamma _j+\delta _t+\epsilon _{ijt} \end{aligned}$$

(1)

In this context, $Y_{ijt}$ refers to the birth weight of the newborn i, in the municipality j, and the date of birth t. The variable $W_{qjt}$ takes the value one if the mother lives in a municipality with adverse weather conditions during the qth trimester of pregnancy, and zero otherwise. For example, $W_{1jt}$ is equal to one if the mother’s municipality experienced an average of at least ten days of extreme weather conditions per month during the first trimester of pregnancy. The same is true for $W_{2jt}$ and $W_{3jt}$ for the second and third trimesters, respectively. The dummy variable $T_{1t}$ indicates whether the mother of the newborn i was in the first trimester of pregnancy after March 2021, the date of the energy price increase. Similarly, $T_{2t}$ and $T_{3t}$ stand for the second and third trimesters, respectively. ${\textbf {U}}_{qjt}$ is a vector containing the distance to the nearest hospital by municipality and the COVID hospitalizations by province in each trimester of pregnancy q and date of birth t.

${\textbf {SES}}_{it}$ is a vector of mothers’ characteristics used to capture their SES, including education, employment status, marital status, and nationality. Note that education level is only included when analyzing heterogeneous effects by maternal education, as this inclusion leads to a substantial reduction in sample size as it contains about 15% missing data.⁶${\textbf {X}}_{it}$ is a vector of various characteristics of both the newborn and the mother that serve as control variables, such as the sex of the newborn, the age of the mother, whether it is the mother’s first child, and whether it is a singleton pregnancy. $\gamma _j$ refers to municipality fixed effects (FE) and $\delta _t$ to month of conception FE. Finally, $\epsilon _{ijt}$ is the normally distributed error term. Thus, the coefficients of interest are $\beta _1, \beta _2$, and $\beta _3$, which indicate the estimated effects of exposure to adverse weather conditions during pregnancy on birth weight when energy prices are high in the first, second, and third trimesters of pregnancy, respectively. Intuitively, $\beta _1$ ($\beta _2$/$\beta _3$) captures the moderating effect of energy prices on the relationship between exposure to extreme temperatures in the first (second/third) trimester of pregnancy and birth weight.

Next, I examine the different moderating effects of energy prices on the relationship between extreme temperatures and birth outcomes as a function of maternal SES. A common problem in previous studies is that individual and regional effects can be confounded. For example, if region A has less educated mothers and reports a lower treatment effect (the effect of the increase in energy prices at extreme temperatures) on birth weight due to unobserved regional factors, while region B with more educated mothers reports higher treatment effects on birth weight due to the same unobserved regional factors, an increase in energy prices may lead to estimates suggesting a greater impact of extreme temperatures on birth outcomes for highly educated mothers when energy prices are high. However, these results may be primarily influenced by regional differences. To overcome this problem, I apply a method proposed by Giesselmann and Schmidt-Catran (2022) that involves a double-demeaned interaction of SES with the treatment effect ($W_{qjt}\times T_{qt}$), that is, it implies demeaning the variables to be interacted before applying fixed effects. Their research shows that FE estimates can lead to bias, as they capture both within- and between-variation when examining interactive effects. By using the double-demeaned interaction estimator (dd-IE), I isolate the within-municipality variation and thus avoid the potential bias due to regional differences. Essentially, this approach compares the treatment effect on birth weight between high- and low-educated mothers within the same municipality. In contrast, the standard FE estimator compares this effect between low- and high-educated mothers across the country. Further details on the estimated model can be found in Appendix K.

Results

Figure 7a illustrates the predicted birth weights during each trimester of pregnancy, comparing situations with and without extreme temperatures, without taking into account the impact of energy prices. In other words, it illustrates the predicted birth weights derived from Eq. 1 without accounting for the impact of the energy prices variable ($T_{qt}$) and its interaction with extreme temperatures ($W_{qjt}\times T_{qt}$). The effects on birth weight are particularly large and significant in the third trimester. In particular, infants born to mothers who were exposed to extreme temperatures during this period show a negative difference in birth weight of about 100 g compared to infants born to mothers who were not exposed to such temperatures. These results are consistent with previous research (e.g., Chen et al., 2020; Deschênes et al., 2009).

Fig. 7

Birth weight predicted values. Notes: The figure on the left shows the predicted birth weight values if the mother suffered and did not suffer from extreme temperatures in each trimester of pregnancy. The figure on the right shows the predicted birth weight for children who suffered from extreme temperatures if they were in each trimester of pregnancy before or after the increase in energy prices. The estimates include as control variables: extreme temperatures in each trimester of pregnancy ($W_{qjt}$); distance to the nearest hospital by municipality; COVID hospitalizations by province in each trimester of pregnancy, whether the mother is employed, married, and foreign, the sex of the newborn, the mother’s age, whether it is the mother’s first child, whether it is a singleton pregnancy, municipality FE, and month of conception FE. The figure on the right also includes the energy price variable ($T_{qt}$) and its interaction with extreme temperatures. Standard errors are clustered by municipality and month of conception. Confidence intervals are given at the 95% level

Table 1 Effects of energy prices and extreme temperatures on birth weight by SES

Next, I analyze the impact of the moderating effect of energy prices on the influence of extreme temperatures on birth weight resulting from the interaction between extreme temperatures and energy prices. Model 1 in Table 1 shows the estimated results of Eq. 1. Extreme temperatures lead to a particularly pronounced significant decrease in birth weight in the third trimester of pregnancy. When the interaction term with the increase in energy prices (ExtrT$\times$After Peak) is taken into account, there is a considerable and statistically significant negative effect for the third trimester of pregnancy. This means that there is a moderating effect of energy prices on extreme temperatures, which exacerbates the impact of extreme temperatures during the third trimester of pregnancy on birth weight when energy prices are high. In terms of magnitude, high energy prices double the effect of extreme temperatures on birth weight. Thus, this result suggests that the negative effects of extreme temperatures in the third trimester of pregnancy increase with increasing energy prices, supporting Hypothesis 1.

Visually, Fig. 7b illustrates the predicted birth weight at extreme temperatures for mothers in each trimester of pregnancy before and after the increase in energy prices. When analyzing the changes before and after the energy price increase in infants exposed to extreme temperatures in each trimester (comparing the gray and blue bars), the greatest change is observed in the third trimester. Infants born to mothers exposed to extreme temperatures in the third trimester after the energy price increase tended to have lower birth weights than infants born to mothers exposed to extreme temperatures in the same trimester before the price increase.

Additionally, Fig. 8 replicates the estimated coefficients of Model 1 from Table 1, but modifies the extreme temperature variables based on different thresholds of days to classify a mother as experiencing extreme temperatures in each trimester of pregnancy. The number of days ranges from seven to 14, as the results become insignificant beyond this range. As shown, the effects of extreme temperatures at high energy prices follow a U-shaped pattern for the third trimester, with the most significant effect observed at a threshold of 12 days. These findings underscore the nonlinear relationship between extreme temperatures and birth weight when moderated by energy prices.

Fig. 8

Marginal effect of extreme temperatures $\times$ high energy prices for different threshold days. Notes: The figure shows the estimated coefficients of the effect of extreme temperatures when energy prices are high on birth weight by trimester of pregnancy (ExtrT$\times$After Peak). The estimates use as additional control variables: whether the mother was in each trimester of pregnancy after the increase in energy prices, whether the mother faced extreme temperatures in each trimester of pregnancy, distance to the nearest hospital by municipality, COVID hospitalizations by province in each trimester of pregnancy, whether the mother is employed, married, and foreign, sex of the newborn, mother’s age, whether it is the mother’s first child, whether it is a singleton pregnancy, municipality FE, and month of conception FE. The standard errors are clustered by municipality and month of conception. The lines that appear next to the coefficients represent the confidence intervals at the 90% and 95% levels

Fig. 9

Marginal effect of extreme temp. $\times$ high energy prices on birth weight by SES. Notes: The figure shows the estimated coefficients in Table 1 (Models 2–5) for the effect of extreme temperatures at high energy prices by trimester of pregnancy in interaction with SES (ExtrT$\times$After Peak$\times$SES). The lines that appear next to the coefficients represent the confidence intervals at the 90% and 95% levels

Models 2 to 5 in Table 1 show the heterogeneous effects based on the mothers’ SES. In all models with heterogeneous effects, the impact of extreme temperatures is consistently negative and statistically significant for the first and third trimesters of pregnancy (T1: ExtrT, T3: ExtrT), but especially for the third trimester. Moreover, energy prices amplify the negative effects of extreme temperatures in the third trimester of pregnancy in all four heterogeneous models examined (T3: ExtrT$\times$After Peak). These results are consistent with those observed in the baseline estimation (Model 1).

Looking at the heterogeneous effects by SES (ExtrT$\times$After Peak$\times$SES), Model 2 shows that the moderating effect of energy prices on extreme temperatures increases significantly in the third trimester of pregnancy when the mother’s education level is low. In other words, if the mother has a low rather than a high level of education, the negative effects of extreme temperatures in the third trimester of pregnancy are about 65 g greater when energy prices are high. In the case of a heterogeneous effect due to the mother’s marital status and employment status, no significant effect is found for the triple interactions of interest. This means that the moderating effect of energy prices on extreme temperatures does not depend on whether the mother is married or not and whether she is employed or unemployed. Finally, the results presented in Model 4 show that being a foreigner negatively affects the moderating effect of energy prices on extreme temperatures in the first trimester of pregnancy. This means that energy prices play a greater role for foreign mothers than for native mothers when extreme temperatures occur in the first trimester of pregnancy.⁷

The results of these models are summarized in Fig. 9. Education significantly affects birth outcomes in the third trimester of pregnancy, with energy prices amplifying the impact of extreme temperatures, particularly for mothers with low levels of education. Marital status and employment status show no significant differences between SES groups. Finally, nationality shows significant differences during the first trimester, where exposure to high energy prices exacerbates the negative effects of extreme temperatures, especially for foreign mothers. In conclusion, increasing energy prices exacerbate the negative effects of extreme temperatures on birth weight, especially for mothers with low education levels and foreign mothers, supporting Hypothesis 2.⁸

Heterogeneity analysis and robustness checks

To ensure the reliability of the main results, additional tests are performed, which are described in detail in Appendix N. First, I re-estimate the main effects and their heterogeneity by SES using alternative dependent variables such as weeks of gestation, low birth weight, and preterm birth. The results, presented in Appendix N.4, are consistent with the primary analysis. I also analyze the impact of including preterm birth as a control. Although preterm birth is excluded from the main analysis to avoid post-treatment bias, the results presented in Appendix N.5 remain robust in terms of sign and significance when preterm birth is included. Furthermore, while this study uses a new comprehensive dataset combining the daily average temperatures of the weather stations of AEMET and the autonomous communities, I also perform the analysis using the E-OBS dataset in Appendix N.6. Furthermore, in the same section, I adjust the method for calculating temperatures by municipality using the monitors’ dataset to closely resemble the algorithm used in the E-OBS calculations. The results are similar to the main analysis. In Appendix N.7, I also examine how heatwaves interact with rising energy prices and affect birth weight. I find that there are significant effects, particularly in the third trimester of pregnancy and among mothers with low SES.

The variable “After Peak” in a trimester of pregnancy is assigned the value one if it occurs after the energy price increase originally set in March 2021. To account for possible discrepancies in the start date, I re-estimate the main equations with alternative start months (February and April 2021). The results, shown in Appendix N.8, are consistent with those of the main analysis. To avoid potential confounding bias, I consider interactions with all SES variables simultaneously. The results in Appendix N.9 show that low-educated and foreign mothers continue to be most affected by rising energy prices when all SES variables are considered together. In addition, I use principal component analysis (PCA) to create a composite SES measure from all the SES variables used in this study, as detailed in Appendix N.10. This analysis shows significant heterogeneous effects by SES in the first and third trimesters of pregnancy.

To account for the effects of the COVID-19 pandemic, the COVID-19 period is excluded in an additional analysis. The results, presented in Appendix N.11, confirm that the exclusion of this period does not change the main results of the study. I also test whether there are differences in the results when different levels of fixed effects are used. The results in Appendix N.12 show that the main results are independent of the selected FE. Finally, two further analyses examine differential effects based on the sex of the newborn and maternal urban/rural residence. However, both tests show insignificant effects, suggesting that extreme temperatures combined with rising energy prices do not affect birth weight differently by sex or maternal residence.

Discussion

The research presented in this paper fills a significant gap in the existing literature by examining the moderating role of energy prices in the relationship between extreme temperatures during pregnancy and health at birth. While numerous studies have investigated the impact of temperature on birth outcomes, this research uniquely highlights the role of energy prices as a critical factor influencing mothers’ capacity to mitigate the negative effects of extreme temperatures. The study contributes to understand the context in which extreme temperatures play a greater role in health at birth, particularly in terms of how SES affects mothers’ capacity to protect themselves from extreme temperatures during pregnancy.

The study conducts a comprehensive analysis using Spain as a case study, with the research design benefiting from a sharp increase in energy prices in March 2021. The study claims that the rise in energy prices, a crucial but previously unexplored moderating mechanism, could exacerbate health inequalities at birth between different social groups. Accounting for various SES factors deepens the understanding of how different groups of mothers and newborns may be affected by extreme temperatures. Notably, the study significantly improves the quality and coverage of weather data by integrating daily average temperatures from 1581 weather stations. This is a significant increase compared to previous data and represents a substantial improvement in granularity, allowing a more detailed analysis of the neonatal population as a whole.

The results show that extreme temperatures have a significant and negative effect on birth weight, with the third trimester showing the strongest influence suggesting a decrease of around 0.2 standard deviations (SD) of birth weight for mothers facing extreme temperatures during the third trimester of pregnancy in comparison to mothers that did not face extreme temperatures during this time. Looking at the moderating role of energy prices, the study shows that the negative effects of extreme temperatures during the third trimester of pregnancy on birth weight increase significantly by about another 0.2 SD when energy prices rise. In particular, high energy prices double the negative impact of extreme temperatures on birth weight, underlining the moderating role of energy prices in this regard. This indicates a decrease in birth weight of around 3%.

In addition, the study examines the heterogeneous effects by mothers’ SES. It suggests that the moderating effect of energy prices on extreme temperatures is more negative when mothers’ education is low and when they are foreign. This suggests that mothers with lower levels of education and foreign mothers are less likely to be able to protect themselves from extreme temperatures during pregnancy, reflecting differences in purchasing power and living conditions between different SES groups. These results are found in both the first and third trimesters of pregnancy.

As outlined in the hypotheses section, differential exposure to extreme temperatures during pregnancy is likely not the only mechanism contributing to the observed heterogeneous relationship as a function of SES. Other proposed mechanisms include the stress experienced by low SES mothers with limited economic resources when confronted with higher energy bills and the deterioration in the quality of goods and services consumed by low SES mothers resulting from their more constrained financial resources due to increased energy prices. All these mechanisms may influence heterogeneous effects by SES. The existing literature predominantly indicates that the primary effects of extreme temperatures on birth weight are typically observed in the third trimester of pregnancy (e.g., Andriano, 2023; Chen et al., 2020). Conversely, studies investigating maternal stress during pregnancy mainly indicate that the first trimester is the time when stress has the greatest impact on health at birth (e.g., Guantai and Kijima, 2020; Torche, 2011). As this study finds heterogeneous effects by SES in the first and third trimesters of pregnancy, I speculate that the heterogeneous effects by SES on the moderating role of energy prices in the relationship between extreme temperatures and birth weight are mainly influenced by stress in the first trimester and exposure to extreme temperatures in the third trimester of pregnancy. Future research should focus on disentangling the relative importance of each of these mechanisms across different socioeconomic groups to better understand their combined influence on birth outcomes.

The policy implications of the study are significant. While tackling extreme temperatures directly could be a global challenge, focusing on the moderating role of energy prices offers actionable recommendations. These measures could include cash transfers for low SES mothers during pregnancy and awareness campaigns. Given the profound impact of events during pregnancy on long-term individual development, such interventions may not only reduce health inequalities at birth, but also address broader inequalities in future outcomes.

Although this study provides valuable insights into how energy prices moderate the relationship between extreme temperatures and birth outcomes, it is important to recognize its limitations. First, the study focuses predominantly on energy prices as a moderating factor, leaving unexplored possible contributions of other moderating variables that could enrich the understanding of the relationship between extreme temperatures and birth health. In addition, the time frame of the study is limited due to the sharp increase in energy prices in Spain in March 2021, which limits the duration of the post-intervention data. An extended analysis of the subsequent years could provide an important understanding into the persistence of the observed moderating effects and thus contribute to a better understanding of the long-term impact. Finally, although the study categorizes socioeconomic status based on education, marital status, nationality, and work status, the multifaceted nature of SES suggests that additional dimensions may influence the moderating effect of energy prices on extreme temperatures. The proposed mechanisms suggest that income is critical as it is linked to poverty and the challenges of affording energy bills as energy prices rise. Therefore, examining heterogeneous differences by maternal income could improve the study’s conclusions by providing more depth and accuracy.

Conclusion

This study fills a critical gap in previous research by examining how energy prices moderate the impact of extreme temperatures on birth outcomes. By focusing on Spain as a study case and exploiting a significant increase in energy prices in March 2021, the research sheds light on how fluctuations in energy prices may exacerbate health inequalities at birth between different SES groups. The results show that while extreme temperatures independently affect birth weight, the moderating effect of rising energy prices significantly amplifies these negative effects, especially for low SES mothers.

This study has significant policy implications. Studying the impact of extreme temperatures on birth outcomes does not directly translate into policies to address health inequalities at birth, as governments cannot reduce temperatures in the medium term due to the global nature of climate change. However, by focusing on how energy prices moderate these effects, there is an opportunity to intervene. Possible interventions include cash transfers to low-income mothers during pregnancy to help with energy bills and conducting an education campaign on the importance of avoiding extreme temperatures during pregnancy for maternal and newborn health. Recognizing that prenatal events have a critical impact on future individual development suggests that such interventions can not only reduce health inequalities at birth, but also address inequalities in long-term development (Almond & Currie, 2011; Behrman & Rosenzweig, 2004; Gluckman & Hanson, 2006).