Ambient temperature-related sex ratio at birth in historical urban populations: the example of the city of Poznań, 1848–1900

This study examines whether exposure to ambient temperature in nineteenth-century urban space affected the ratio of boys to girls at birth. Furthermore, we investigate the details of temperature effects timing upon sex ratio at birth. The research included 66,009 individual births, aggregated in subsequent months of births for the years 1847–1900, i.e. 33,922 boys and 32,087 girls. The statistical modelling of the probability of a girl being born is based on logistic GAM with penalized splines and automatically selected complexity. Our research emphasizes the significant effect of temperature in the year of conception: the higher the temperature was, the smaller probability of a girl being born was observed. There were also several significant temperature lags before conception and during pregnancy. Our findings indicate that in the past, ambient temperature, similar to psychological stress, hunger, malnutrition, and social and economic factors, influenced the viability of a foetus. Research on the effects of climate on the sex ratio in historical populations may allow for a better understanding of the relationship between environmental factors and reproduction, especially concerning historical populations since due to some cultural limitations, they were more prone to stronger environmental stressors than currently.


Results
In Model 0, which estimated the effect of birth year and birth month on the probability of giving birth to a girl, the seasonal effects (annual and month effects) were not significant.Next, we added the mean temperature in the year of conception to the components in Model (M0), obtaining (MT0).The temperature in the year of conception was statistically significant (Table 1).

Discussion
The impact of ambient temperature on the physiology of reproduction Temperature-dependent sex determination is well documented in vertebrates such as reptiles, fish, and some amphibians.It is related to the developmental growth rate 22,25 .A similar relationship is also observed in humans.This is supported by studies of modern populations, suggesting that in warmer months/higher temperatures more sons are produced than daughters [20][21][22][23] .In higher temperatures, male foetuses develop faster than female ones 22 .Therefore, males outnumber females in hot climates 26 .It is also suggested that the genome contains genes that determine the child's sex, the so-called "sex-determining region on the Y chromosome", SRY.These genes trigger the expression of several genes involved in the development of the male gonad 27 .It is suggested that these genes may be activated by temperature 22 .In stressful conditions, stress hormones produced by mother's body disrupt mechanisms related to the functioning of SRY protein that determines the embryogenesis of male foetuses 28 .Moreover, research conducted on the populations of economically developing countries, proves that the sex ratio is influenced by the temperature at the time of conception and in very early pregnancy: the higher Table 1.Parametric and non-parametric smooth effects for birth year, birth month, and mean annual temperature in the year of conception on the probability of giving birth to a girl.***p < 0.001; **p < 0.01. the temperatures at the time of conception, the higher the rate of masculinization of newborns 29 .Hajdu and Hajdu 30 reached similar conclusions when examining births in Hungary in the years 1980-2015.Additionally, researchers report that exposure to hot temperatures changes the timing of some conceptions, i.e., some pregnancies might experience a shift in conception 30 .
Our study verified whether ambient temperatures could have affected the proportion of boys to girls at birth.The results show that introducing the average temperature in the year of conception into the model has a significant impact on the proportion of boys to girls at birth: the higher the temperature, the lower the probability of giving birth to a girl, while the probability of giving birth to a boy increased, and the other way around.The situation is similar when average monthly temperatures for the months before conception and during foetal development (during pregnancy) are introduced into the model.Although higher air temperatures may lead to the increased likelihood of having a boy, such extremely high temperatures may have the opposite effect and shift the sex ratio towards females (lower sex ratio values).Research shows that ambient temperature may influence primary sex determination through variable fertilization success with sperm with X or Y chromosomes 22 .Temperature influences the course of spermatogenesis and oogenesis 31,32 .In mammals, high ambient temperature affects the concentration of steroids in ovarian follicles 33 .These endocrine changes reduce the activity of follicles  and change the mechanism of ovulation, which leads to a decrease in the quality of oocytes and embryos.The uterine environment also changes, which reduces the probability of embryo implantation and increases the risk of embryo loss 32 .The function of the uterine endometrium is also impaired and its secretory activity changes, which may lead to termination of pregnancy 33 .Hyperthermia during pregnancy can cause embryo's death, miscarriage, growth retardation, and developmental defects 2,34 .Extremely high temperatures affect placental blood flow and increase the risk of hypertensive crises and stillbirth by increasing vasoactive substance production, blood viscosity, and endothelial cell activity 35 .Environmental stresses may also impair sperm motility, potentially promoting transition to female sex at birth [36][37][38] .Studies conducted on experimental animals suggest that males exposed to high temperatures are less fertile 39,40 , which is due to poorer semen quality 39,40 .The effect of heat exposure on semen quality is transient, producing the strongest effects not immediately after heat stress, but several weeks later [41][42][43] .
In turn, exposure to low temperature increases plasma fibrinogen levels, increases blood viscosity, and may cause placental vasoconstriction, reducing uteroplacental blood flow and inhibiting foetal growth 18 .Several other processes may occur during winter that may have detrimental effects on foetal development, such as increased exposure to infections, decreased physical activity, increased exposure to pollutants, and increased pregnancyinduced hypertension 18 .
The high mortality of more sensitive male foetuses during temperature stress may have contributed to the reduction in the ratio of live-born boys to girls in conditions of extremely high/low temperatures 20,26 .According to estimates, environmental stresses can lead to the loss of at least 70% of conceptions (they are miscarried before delivery) [44][45][46] .In the light of the evolutionary theory, fewer males should be born during stressful periods because the weaker males would not survive and be unable to reproduce where the females could reproduce 47 .It is known that the male foetus is more susceptible to environmental insults, and indeed the male foetus is generally more fragile and prone to be stillborn 26 .

Ambient temperature values in historical Poznań and their link to sex ratio at birth
In Poland, the temperature trend has been systematically increasing since the middle of the nineteenth century.The last two decades of the twentieth century and the first decade of the twenty-first century are the warmest in the 230-year history of meteorological observations in the country.The extreme weather phenomena, which earlier occurred sporadically, have recently become more frequent and intense 48 .It is worth remembering that Poles have slightly lower chances in the clash with heat waves in comparison with the inhabitants of southern Europe.They usually have slightly brighter skin, are more susceptible to burns, and do not have air-conditioning systems in homes 49 .Hence, a rapid response of an organism to extremely high temperatures often leads to physiological disorders and/or also to a direct impact on health status in the form of asthma, allergies, and death-related complications 50 .
Let us recall once again that in our research, introducing the average temperature in the year of conception into the model significantly affects the probability of giving birth to a boy.The situation is similar when average monthly temperatures for the months before conception and during foetal development (during pregnancy) are introduced into the model.It suggests a positive impact of warm months on conception and foetal development and the increase in the ratio of live-born boys to girls, as discussed in the previous chapter.Climatological studies of nineteenth-century Poznań (see: "Results"; Fig. 3) indicate that in the years 1848-1900 the average annual temperature categorized as "normal" occurred 11 times, the average annual temperatures categorized as "light warm" was recorded in the following years: 1880, 1885, 1890, 1891 and 1899 (Fig. 3), and the average annual temperatures categorized as "warm" occurred in the years: 1852, 1869, 1874, 1877 and 1894 (Fig. 3).These temperatures may have favoured male conceptions through mechanisms described in the previous chapter of this work.McLaren and Wilde 29 drew attention to the relationship between temperatures in the months around conception and the child's sex.At the same time, researchers emphasized that the results of these studies apply to economically developing countries.In developed countries where it is possible to regulate indoor temperature, for example, by air conditioning, researchers have found no effect of ambient temperature at conception and during pregnancy on the sex ratio 29 .Poznań in the nineteenth century, in terms of infrastructure development, corresponded to some extent, to some modern economically developing countries.
In the period under study, Poznań experienced a rapid demographic development, resulting in the increase in population from 44,000 inhabitants in the mid-nineteenth century to 117,000 at the beginning of the twentieth century.At the end of the nineteenth century, Poznań was the sixth most densely populated city in the Prussian Empire 51 .On the outskirts of the city, tenement houses were built for poorer tenants, who were mainly immigrants coming to the city for work 52 .The small, stuffy, and humid apartment was occupied by several to a dozen people 52 .In 1871 in Poznań, one residential building was populated on average by almost 40 people, and the average household had over five people 53 .Therefore, it is not surprising that the living conditions in nineteenthcentury Poznań are described as terrible; Poznań residents struggled with the problem of high rents, overcrowded, cramped, damp, unheated, and poorly lit apartments 51,52 .At the beginning of the twentieth century, the fee for renting one room consumed from 40 to 56% of the income of the lowest earners 54 .There was little left to cover the costs of food, fuel, light, or clothing.The lowest rent was equal to the two-week earnings of an unskilled worker, and the highest rent exceeded their salary.Usually, one room was rented, due to prices-in the attic or basement.In 1900, over 10,000 people lived in one-room basements and attics, i.e., 1/10 of the city inhabitants 54 .It is worth adding that in the same year, almost 45% of Poznań inhabitants lived in one-room apartments: overcrowded, damp, and unheated 54 .Everyday life in such conditions was particularly burdensome in autumn, winter, and early spring when residents were bothered by cold and damp rooms, especially when the temperature was low or extremely low.Summer months were also difficult with high and extremely high-temperature values, when the rooms became stuffy, and the lack of hygiene and sewage system resulted in an epidemiological threat.
Working conditions also left much to be desired.A 12-h workday was common, sometimes extended to 16 h, without additional remuneration 54 .The workers stayed on the factory premises for an additional hour to over 2 h longer 54 .A significant percentage of the professionally active population in Poznań in the second half of the nineteenth century were low-skilled workers whose working conditions were often physically and mentally exhausting 55 .Work was often done at home as cottage industry, which, given the poor sanitary standard of Poznań apartments, did not have a positive effect on health.Workplaces often did not meet sanitary standards 55 .Many of the labourers worked in cramped and low rooms, which were damp and lacked ventilation.The lack of adequate airflow was especially bothersome for workers in craft workshops and factories, where there was also the problem of dust, and the factory halls were crowded with equipment, waste, and people 55 .Labourers working in mineral, food, and metallurgy industries especially had to struggle with high temperatures.This was particularly hard during hot weather, when employers, against legal regulations, did not provide adequate access to clean drinking water 55 .Women could enjoy some improvement in working conditions, as Prussian legislation introduced solutions in the last decade of the nineteenth century that were aimed at improving their working conditions, mainly by shortening the working day.Mandatory breaks were also introduced, and women's work was prohibited, among others, during night shifts and on Sundays and holidays 55 .
Our findings do not directly contribute to estimates of the effects of ambient temperature values on the proportion of boys to girls at birth.However, they indicate that in the past, ambient temperature, as well as psychological stress, hunger, malnutrition, and social and economic factors, influenced the viability of the foetus.The results, however, suggest a heretofore unrecognized impact of temperature on reproduction (who survives gestation) in past societies.As research shows that temperatures have been rising in Poland since the nineteenth century 49 , our results begin to track the impact of climate change on the sex ratio at birth.Our findings are based on data from Poznań, a city located in Central Europe.Therefore, the results can be generalized to other countries from this geographical and climatic zone from a similar period because the climatic conditions in many countries of Central and Eastern Europe are very similar.Moreover, it will be interesting to use the same deeper temperature-lag-effect analysis (based on the Almon lags) in different parts of the world to compare the profile of the temperature-effect trace.Our results show that in past and present-day populations, ambient temperature had implications for public health.

Temperature observations in Poznań in the nineteenth century
The first temperature measurements began in Poznań in 1848, in the meteorological station at 1 Pocztowa Street, moved then to Pocztowa Street, and then to 1 Grobla Street.In the latter place, the thermometer was set at 2.5 m above ground level (AGL) 56 .In August 1885, the location of the station was changed from 1 Zielona Str. to 2 Zielona Str., where temperatures were measured with a thermometer set at 6.2 m AGL 56 .In September 1892 the station moved to 3 Długa Street and the thermometers were placed at 8.6 m AGL.Here the station operated until December 1911 56 .Until 1892, temperature measurements were taken with a thermometer without a cover and placed outside the window.Since September 1892, zinc sheet sheaths were used on thermometers to protect the measurements against atmospheric factors such as wind, sun, and humidity, which distorted the results 56 .Until 1884, measurements were taken three times a day: at 6 am, 2 pm, and 10 pm.Between December 1884 and May 1919, the hours of observation were changed for 7 am, 2 pm, and 9 pm.Meteorological observations in Poznań were carried out by professors at the Municipal Real School: first by Spiller, and since April 1862-by Magener.In 1889, meteorological measurements were taken by the Physical Institute of the Royal Academy 56 .Since the 1890s, a meteorological station operated also in the village of Jeżyce where the thermometers were set at 9.5 m AGL.A pharmacist Wildt was the manager of the station.Here the meteorological observations were taken three times a day: at 7 am, 2 pm, and 9 pm 56 .Initially, temperature was measured using the Reamur scale.The Celsius scale has been used since 1879.The daily temperature was the arithmetic mean of three daily observations 56 .

The changes in outdoor temperature in Poznań in the second half of the nineteenth century
Thermal conditions in Poznań in 1848-1900 were determined based on mean monthly air temperature values obtained from the work by Kolendowicz and a team 57 .The data made it possible to calculate a mean annual air temperature, as well as monthly means and the annual mean for the entire multiannual period.Then, the direction of changes and their statistical significance were analysed using a t-student test (significance level 0.05).In the next step, the classification of thermal conditions in particular months and years was conducted.The adopted classification is based on percentiles calculated for mean monthly (each separately) and annual values, with consideration of the following thresholds: 5, 10, 20, 30, 40, 60, 70, 80, 90, and 95 percentiles.The resulting 11 classes were calculated for the period 1848-1900 and adopted for the development of a heat map (Fig. 3).In the years 1848-1900 in Poznań, the mean annual air temperature was 7.5 °C.In the studied period, it varied from 5.5 °C in 1871 to 9.1 °C in 1872 (Fig. 4).These data suggest that in the analysed years, deviations from the multiannual mean reached −2.0 °C and 1.6 °C, respectively.In the analysed period, a high variability of mean air temperature was recorded.A relatively high standard deviation (0.8 °C) shows a broad distribution of mean values for particular years around the multiannual mean.A year with a high mean air temperature can occur right after a year with a low value, and vice versa.High variations in thermal conditions were observed in the 1860s and 1870s when extremely cold years were recorded (1855, 1864, 1871) following extremely warm years (1863, 1868, 1872) and vice versa (Fig. 3).The conducted research showed an increase in the mean annual air temperature at a level of 0.1 °C/10 years, although the changes were not statistically significant.In the analysed multiannual period, the mean monthly air temperature varied from −3.0 °C in January to 18.1 °C in July.Apart from January, mean monthly values below 0 °C were also recorded in February and December (−1.6 °C) (Fig. 4).The greatest range of variation of mean monthly air temperature was recorded in the winter months, and the lowest in the summer months.Standard deviation values ranged from 1.2 °C in August to 3.0 °C in January and February.Like in the case of the mean annual air temperature, high year-to-year variations were also observed from month to month in a year.The lowest mean monthly air temperature was −11.3 °C and occurred in January 1848.Air temperature below −10.0 °C was still recorded in January 1893 (−10.5 °C).The highest mean monthly air temperature was 20.5 °C, recorded in June 1889 and July 1865.Moreover, the value of 20.0 °C was exceeded in July 1874 and 1900.In the analysed multiannual period, except for June and October, an increase in mean warmer weather effects on a long-term (annual) scale only.Further, we use more detailed analyses focusing on temperature effects closer to the birth date month (before, at, and after).The approach is based on the Almon distributed lag 63 sub-model instead of the annual temperature average.This sub-model allows us to address the impact of differently distant pasts upon the PGB.Namely, we estimate coefficients (loosely interpretable as "weights") on monthly-averaged temperature for several months before the birth date simultaneously.Estimating these coefficients is notoriously problematic or even impossible to conduct directly (due to their heavy intercorrelations) and a regularization of this ill-posed problem is necessary.To this end, we use a classical approach developed in econometrics and now used in other fields 64,65 .Specifically, we use the model (MTA) with the Almon polynomial sub-model for addressing simultaneously the monthly temperature average effects with the lags of −9, −8, −7, −6, −5, −4, −3, −2, −1, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9 with respect to the conception.All lags are before the birth date.Negative lags correspond to the months before conception, while positive ones to the months after conception and before birth.
The annual trend was modelled nonparametrically (as a complexity penalized spline of disease-group specific shape).One of the main advantages of this approach is that it can capture complicated nonlinear shapes of trends, including periodic behaviour if the data support them.Computations were done in the R environment 66 .
All methods were carried out in accordance with relevant guidelines and regulations.

Methodological limitations
The methodological tools used in this work have certain limitations.The first is related to the use of aggregated data which contains information about the number of live births in particular months of the year only.This involved estimating the moment of conception, and based on which the months of gestation with exposure to ambient temperatures were established.Thus, the distinguished months of pre-conceptions/conceptions/ pregnancies could have been consistently affected by an error resulting from the previous estimates.Moreover, foetuses in gestation in one year could have been born in the next.The aggregated data did not enable a deeper insight into the studied phenomenon, which would have been possible if the information on individual births had been used.Individual data could have allowed the partitioning of gestation into trimesters and to capture the developmental stage which was the most sensitive to ambient temperatures.Secondly, the aggregated data on live births by sex in particular months of the year do not contain information on the duration of pregnancy, including pre-term and prolonged pregnancies.We assume that all pregnancies are terminated on time.In contemporary Poland, 92.4% of pregnancies end on time 67 , while only 7.6% of them are pre-term and post-term (the 7.6% of pre-and post-term pregnancies did not burden the results with an error).
Thirdly, lack of information on foetal deaths in the uterus, miscarriages/abortions, and stillbirths did not make it possible to capture the factors driving the decline/increase in SSR in some months or years in the period under study.The fourth limitation stems from the lack of data on daily temperatures.In this paper, we used average monthly and yearly temperatures.
Moreover, the sex ratio at birth may have been affected by biological factors.One of them is gestational age, GA.Weng and a team show that the sex ratio decreased proportionally with increasing GA 68 .Another factor influencing the sex ratio is birth order 69 : nulliparous increases the chances of delivering a boy 70 .Birth may be related to maternal age 70 .The mothers' socioeconomic factors, such as maternal marital status and type of job could also have been confounding factors.The exposure of unmarried pregnant women to stress resulting from financial problems, poor living conditions, unemployment, lack of psychosocial support social instability, and physical workload may have resulted in the elimination of weaker male fetuses more often than female ones 71,72 .

Figure 1 .
Figure 1.The effect of an average temperature in the year of conception on the probability of giving birth to a daughter.The figure based on data collected for the project supported by the National Science Centre, Poland.Grant: Long-term social costs of adverse conditions occurring in early life in the 19th-century populations on the example of the city of Poznań, No. UMO-2020/39/O/HS3/00524.

Figure 2 .
Figure 2. The effect of an average temperature in the months before and after conception on the probability of giving birth to a daughter.The figure based on the data collected for the project supported by the National Science Centre, Poland.Grant: Long-term social costs of adverse conditions occurring in early life in the 19th-century populations on the example of the city of Poznań, No. UMO-2020/39/O/HS3/00524.

Figure 3 .
Figure 3.The percentile classification of the annual mean temperature.Thresholds of 5, 10, 20, 30, 40, 60, 70, 80, 90, and 95 percentiles denote 11 classes (described in the legend).Classes were calculated for the reference period of 1848-1900.In the table, the criteria for thermal quantile classification of month and years.

Table 2 .
The effects of mean monthly temperatures lagged −9…to 0 and 0…to 9 concerning the month of conception on the probability of giving birth to a girl.L and U correspond to the lower and upper limits of (pointwise) 95% confidence intervals.*Statistically significant.