“There is Only One Mother”: The Importance of Mother in Rural Spain,1750–1959

The Importance of the Mother on the Probability of Death of her Children

A mother's intervention in breastfeeding, feeding her children and caring of them had a great influence on their survival. Table 2 shows the probabilities of death of children according to the survival status of their mothers and the age of children at death. These results clearly point to the important role of the mother in the survival of her offspring. For the entire period, the chances of survival of children whose mother was alive until at least 10 years of age were between 1.4 and 3.9 times higher than those of children who had lost their mother. Losing a mother in the first one or two years of life doubled the chances of dying in almost all age groups, especially the youngest. The ratios obtained in our analysis confirm that the presence of the mother is a determining factor in survival.

OPEN IN VIEWER

Table 2.

Probabilities of Death [q type rates] during Early Life by Mother's Survival Status, Middle Huerva, Aragón (Spain), 1750–1959.

	Estimated q type rates by survival status of mothers
	Mother survives for at least 10 years after the child's birth(1)	Mother died before her child was 10 years old(2)	Mother dies, child < 24 months(3)	Mother dies, child < 12 months(4)	Ratio 1(2)/(1)	Ratio 2(3)/(1)	Ratio 3(4)/(1)
1750–1959	n = 11.610	n = 4.929	n = 1.177	n = 694
<1 month	0.0532	0.0998	0.1606	0.2061	1.87	3.01	3.87
1–6 months	0.0544	0.0790	0.1093	0.1340	1.45	2.00	2.46
6–12 months	0.0509	0.0804	0.0871	0.0794	1.57	1.71	1.55
<1 year	0.1388	0.2420	0.3314	0.3890	1.74	2.38	2.80
1–2 years	0.0801	0.1251	0.1630	0.1086	1.56	2.03	1.35
<2 years	0.2141	0.3532	0.4715	0.4870	1.64	2.20	2.27
<10 years	0.3944	0.4800	0.5616	0.5620	1.21	1.42	1.42
1750–1869	n = 4.371	n = 2.998	n = 695	n = 390
<1 month	0.0688	0.0991	0.1712	0.2282	1.44	2.48	3.31
1–6 months	0.0542	0.0829	0.1250	0.1561	1.52	2.30	2.88
6–12 months	0.0565	0.0767	0.0635	0.0512	1.35	1.12	0.90
<1 year	0.1700	0.2372	0.3209	0.3821	1.39	1.88	2.24
1–2 years	0.1100	0.1557	0.2373	0.1950	1.41	2.15	1.77
<2 years	0.2512	0.3559	0.4820	0.5026	1.41	1.91	2.00
<10 years	0.3642	0.4993	0.5928	0.6000	1.37	1.62	1.64
1870–1909	n = 3.244	n = 1.378	n = 338	n = 214
<1 month	0.0540	0.1132	0.1598	0.1916	2.09	2.95	3.54
1–6 months	0.0682	0.0810	0.1056	0.1329	1.18	1.54	1.94
6–12 months	0.0675	0.1006	0.1654	0.1733	1.49	2.45	2.56
<1 year	0.1809	0.2671	0.3728	0.4206	1.47	2.06	2.32
1–2 years	0.1225	0.1614	0.2075	0.1371	1.31	1.69	1.11
<2 years	0.2699	0.3853	0.5030	0.5000	1.42	1.86	1.85
<10 years	0.3809	0.5065	0.5621	0.5421	1.32	1.47	1.42
1910–1959	n = 3.995	n = 553	n = 144	n = 90
<1 month	0.0355	0.0705	0.1111	0.1444	1.98	3.12	4.06
1–6 months	0.0451	0.0739	0.1094	0.1558	1.63	2.42	3.45
6–12 m	0.0456	0.0777	0.0965	0.0923	1.70	2.11	2.02
<1 year	0.1055	0.2061	0.2847	0.3444	1.95	2.69	3.26
1–2 years	0.0396	0.0661	0.0874	0.0678	1.66	2.20	1.71
<2 years	0.1397	0.2586	0.3472	0.3889	1.85	2.48	2.78
<10 years	0.1741	0.3092	0.4097	0.4444	1.77	2.35	2.55

Note: Date of birth defines periods. “n” refers to the number of births to women of each category.

Source: AMHDB.

In analysing the probabilities of death for the three sub-periods mentioned above, we found that the mortality patterns by age were similar. The chances of dying for the youngest children, especially those younger than one month old, increased between 1.9 and 3.9 times if they lost their mother. Children from zero to nine years old continue to have high ratios. We not only confirm the importance of the mother, but also find that they were even more pertinent in the late stages of the demographic transition. However, the number of deaths (and thus the probability of death) decreased during the demographic transition, although without significant changes in the ratios. For example, the chances of dying in the group of children under age two are 0.5 (very high) in the period 1750–1869 among children who had lost their mother in the first year of life, but it decreased to 0.39 in the period 1910–1959. This represents a mere 22-percent decrease in the probability of dying. However, among children whose mother was alive until at least 10 years of age, the odds of dying before age two fell from 0.25 in 1750–1869, to 0.14 in 1910–1959, a decrease of 44 percent. In other words, the improvements in mortality were more beneficial for children whose mothers lived, although these improvements were felt to a lesser extent by all groups (we will discuss this in the Discussion section and propose an explanation for these results).

In Table A1 of the Appendix, we compare the age group with the most extreme results (the children who were younger than one month old when they lost their mother) separately. That is, we are taking only children whose mother died in their first month of life and they survived at least one day after her death. We have compared them with the children whose mothers survived at least until they turned 10 years old. The results show that, in the early stages of life, their chances of dying skyrocketed to 6.2 times higher than those whose mothers survived. After six months of age, however, the differences disappeared for those whose mother died in their first month of life, presenting an even lower probability of death. Therefore, if children survived past six months despite the death of their mother in the first month of life, they were more likely to survive than others who had not lost their mother. High mortality in the first months of life possibly conditioned children in such a way that only the strongest children reached six months of life and, therefore, their subsequent mortality rates were relatively low. Then there could have been a process of selection of the strongest. In addition, the mortality of children at an age close to that of their mothers has also been largely affected by perinatal causes that affect the mortality of the mother and the children, as noted for the majority of British cases by the Royal College of Pediatrics and Child Health. ¹⁷

Table A1 also shows that the odds of dying after the first- and second-year increase for children of surviving mothers but decrease for motherless children. It is possible that this is the consequence of selection biases. By multiplying the mortality of motherless in the first months of life several-fold, the strongest survived, and this may mean that they are less impacted by mortality problems later in life.

From the day of the mother's death, we have calculated how many more days they lived. The results confirm that the children who were the youngest at the time of their mother's death had the shortest life span. The mortality of children whose mother died during childbirth or very close to it (possibly due to perinatal causes) was especially high, and they usually died in the days following delivery. The effect of the mother is clearly very prominent, with almost 10 percent of the children who lost their mother between zero and five months dying in the week after their mother's death, and 17.5 percent in the following month. In the case of children between 6 and 23 months, these percentages drop to 3.9 and 7.6 respectively.

Table A2 confirms that the death of the mother had a strong impact on the chances of the survival of her children in the short- and medium-term. This was especially true among children who were in the early stages of breastfeeding, so their surviving family may have been forced to find a surrogate nurse. Likewise, it is possible to indirectly observe the influence of breastfeeding on the nutrition of children through the evolution of mortality. The results present some differences with those obtained for the Spanish town of Aranjuez with a similar analysis for a shorter period. ¹⁸ In our study area, if the mother died in the first week of life, 9.5 percent of the children died in the first five months (12.5 percent in the case of Aranjuez). If the mother died in the first month of life, 17.5 percent of the children in our study area died in the first five months (37.5 percent in Aranjuez). In this sample, we require that the child survives for at least one day after the death of the mother. This factor alone does not explain the gap between Aranjuez and our area. Since it is a traditional rural environment, the majority of fathers who had just lost their wives could have had more relatives in the same locality (such as grandparents, siblings, brothers, and sister-in-law or nephews) who could have helped the child's upbringing in the early stages after the death. Maternal grandmothers, for instance, have been found to improve the survival of children, but paternal grandmothers worsened their chances and grandfathers had no effect. ¹⁹

The Importance of Other Family Members Residing in the Same Household of the Mother on Probability of Death

Various studies have shown that other relatives could partially substitute the role of the mother, although not as efficiently. ²⁰ In Table A3 of the Appendix, we analyze the role of fathers and mothers as a function of the number of siblings. The results are diverse, but we must keep in mind that this is an approximation including relatives but with a relatively small sample that could be skewing the results. The lowest mortality is found in households with both living spouses and a large number of children who could help with parenting. The probabilities of death were very low in these households (among the lowest found in the entire paper). If there were no other children who could collaborate for the rearing of the newborn, the probabilities of death increased between 1.7 and 2.4 times. However, this can also be linked to inexperienced parents (or “unhealthy families”). The number of living siblings, in this case, would not only be additional help for caring for the newborn, but would also be a sign of households with families with better hygiene behaviours.

More interesting are the results in Table A3 Columns (3) and (4), where there is a widowed father who takes over the household. The results confirm that no member of the family can replace the mother, especially during lactation. We found ratios between 4.6 and 0.6 times the probabilities of death with respect to families with both parents and several siblings. Siblings perhaps required a mother's directions in order to improve the chances of a newborn's survival. However, the most extreme result occurred when only the father was alive and there were no other children available (or just one) to care for the newborn. In that case, mortality rates for children under two years increased between 5 and 11.4 times, with the majority of children who had lost their mother dying in the first months of life.

These results have two implications: (1) a significant number of widowed fathers seem to have been unable to take a correct care of the newborn, and (2) that the number of living siblings could have affected the probabilities of the death of children, even when the mother was alive.

The limited effect of fathers on the survival of their children could perhaps be explained by the gender roles in households. Traditionally in rural Spain the role of childminding was not assigned to fathers; they were not prepared for the responsibilities associated with childcare. A young child could be a daily burden, it could disrupt work life, reduce socialization activities, and reduce the chance of remarriage. This result could also be attested to the misleading effect of the small sample size or other factors that we cannot control, so the explanation of gender roles should be taken with caution. In fact, the small sample size in Table A3 may have affected the results (which is why the results in Table 3 regarding siblings are more interesting). More data are needed to delve into these possible explanations and to understand the role of the father after the death of the mother. In any case, across studies on the role of fathers in child survival, only a third find any positive impact of the presence of fathers. ²¹

OPEN IN VIEWER

Table 3.

Cox Proportional-hazards Regression Results on Dying before Two Years of Age Controlling by Mother's Survival Status, Birth Cohorts 1750–1959.

		Model1750–1959	Model1750–1869	Model1870–1909	Model1910–1959	ModelFemales	ModelMales
Variable	Category	(1)	(2)	(3)	(4)	(5)	(6)
Motherless (before 24 months)	No	(ref.)
	Yes	1.49***	1.62***	1.47***	1.36**	1.55***	1.45***
Father's occupation	Farmer	(ref.)
	Low-skilled worker	0.95	0.86	1.02	1.10	0.92	0.99
	Artisan	1.08	1.10	1.09	1.05	1.06	1.11
	Upper class	0.93	0.99	0.87	0.91	0.94	0.91
	Other or unknown	0.82*	0.88	0.76	0.69	0.77**	0.87*
Parity (birth order)	1	(ref.)
	2	0.97	1.02	1.07	0.89	1.06*	0.94*
	3	1.01	1.01	1.14	1.05	1.19	1.02
	4	1.24	0.99	0.98	1.22*	1.22	1.03
	5	1.17	0.82	0.99	1.18**	1.17	1.02
	6 or more	1.34***	0.97	1.10	1.41**	1.39***	1.31***
Number of alive siblings at birth	0–1	(ref.)
	2–4	0.92**	0.98	0.80	0.69**	0.95**	0.90*
	5 or more	0.84***	1.06	0.95	0.75***	0.82***	0.87**
Dying in a high mortality year	No	(ref.)
	Yes	1.12***	1.34***	1.01***	1.03***	1.04***	1.44***
Village -fixed effects		Yes	Yes	Yes	Yes	Yes	Yes
Decade of birth-fixed effects		Yes	Yes	Yes	Yes	Yes	Yes
Number of episodes		11,216	4,091	3,160	3,965	5,454	5,762
Chi2		189.2	162.1	149.9	135.6	178.7	171.5
Prob. > Chi2		0.0000	0.0000	0.0000	0.0000	0.0000	0.0000

Source: Parish and municipal registers.

Note: coefficients in exp(β)-form. * Statistical significance at 10% level. ** Statistical significance at 5% level. *** Statistical significance at 1% level.

The Importance of Mother and Siblings From Multivariate Regression Models (Confirming the Previous Findings Using Event History Analysis)

Following the results of the previous subsections, in this subsection we want to confirm the results by using multivariate models. To do so, in Table 3 we have run six Cox proportional hazard regression models whose dependent variable is the death of a child before two years of age, controlling by the survival status of the mother. All models have the following independent variables: being motherless (the mother predeceased the child); the paternal occupation (as proxy for family socioeconomic status); the birth order of the children ²² (since parity might have affected mortality); the number of living siblings (to confirm their role in the survival of the newborns); whether the year of death was a year of particularly high mortality; and controls for village and decade of birth. The six models are distributed as follows: (1) Complete model, (2) Model for the birth cohorts 1750–1869, (3) Model for the birth cohorts 1870–1909, (4) Model for the birth cohorts 1910–1959, (5) Model for girls only, (6) Model for boys only. These models allow us to interact the presence of the mother with other variables such as household size or family socioeconomic status.

The results in Table 3 are clear and consistent. In all models being motherless is associated with an increased chance of dying of more than one-third and usually more than 50 percent. Another significant variable is birth order; children born in high parities increased their chances of dying by between 20 percent and 40 percent. This may be due to the fact that they were born to older mothers, which would be associated with greater problems during birth and even with problems in the development in the womb. In historical periods it is common to find poorer health status and higher mortality among children born to older mothers. ²³ As shown earlier, the number of siblings is strongly associated with a lower chance of sibling death (20 percent). Therefore, siblings are important to the survival of newborns, but not nearly as important as the mother, who is twice as important according to the coefficients of Table 3. In addition, it is interesting to note that increasing parity (related to increasing maternal age) was associated with higher odds of newborn death. However, a higher number of living siblings was linked to higher odds of survival, highlighting the importance of the presence of siblings independently of the parity. Also, as expected, epidemic years were associated with higher mortalities.

One point in which Tables 2 and 3 differ is in the intensity of the effect of the mother's presence. While in Table 2 a slight increase in the importance of this factor in mortality can be observed, in Table 3, controlling for various factors, a slight loss of importance is observed. On the one hand, we must take into account that they are different methodologies and that infant mortality in the last subperiod is very low. On the other hand, the result obtained seems to indicate that, controlling for other family and environmental factors, the loss of the mother can be softened in a context of medical and economic improvements and low mortality from contagious diseases.

The Sex Preferences and Their Influence on Probability of Death

A fundamental factor that has conditioned demographic behaviour in rural Spain is the preference for sons. ²⁴ The existence of preferential behaviour towards boys has been observed for the study area. ²⁵ This led to an excess in the mortality of girls during the period prior to the fertility transition, and even during the first decades of the twentieth century. ²⁶ However, little has been done to understand whether discriminatory behaviour is mainly conditioned by one parent or by both parents at the same time. Figure 2 shows the mortality ratios, following the methodology applied so far, according to the sex of the deceased child. The results must be explained from the biological (greater weakness of boys) and social (preference for sons) contexts. Despite the greater weakness of boys during the first months of life, it is motherless girls who experience the greatest increase in mortality in the first month of life. They suffered strong discrimination, and that ultimately affected their mortality during the weakest period of their lives. In the rest of the studied period, between the first month of life and the first year, boys and girls show similar excess mortality rates, although slightly higher in the case of boys. This is explained by the greater weakness of boys and because, during that period, children were breastfeeding (weaning started at around six to seven months of life, when mothers would start combining breastfeeding and solid food). This meant that newborn did not compete with other family members for family resources during breastfeeding. Some authors have pointed out the possibility that boys had biological advantages over girls during, at least, the second half of the first year of life. ²⁷ However, this is unclear according to the results of our analysis. The second period in which the mortality ratio of girls was lower than that of boys was between two and nine years. This is different from what Pavard et al. ²⁸ find in Canada in the seventeenth and eighteenth centuries where there was a threefold increase in mortality for motherless girls, compared to motherless boys, after the age of three. In this period, the strongest children survived and discrimination in food or care has less impact on mortality. Even so, in the total ratio of the number of boys/girls alive, the mortality of girls up to ten years of age continued to be higher than that of boys until the twentieth century.

OPEN IN VIEWER

Figure 2.

Sex differences in excess mortality, Middle Huerva, Aragón (Spain), 1750–1959.

In addition, some work suggests that there could be epidemiological factors that could also play an important role in sex-specific differences in childhood mortality. ²⁹ Therefore, the explanation of gender preference to explain the mortality found is not the only possible one. However, what is notable is that we have found differences in mortality by sex after the loss of the mother at young ages.

The Effect of the Loss of the Mother on Child Development

To this point we have focused the analysis on the probabilities of death, but our interest goes further. We want to analyze the connection between the presence of the mother and the good biological and social development during childhood and youth. We measure this through two variables. We approach social and cultural development through the literacy rates of men and women, and biological wellbeing through the use of the heights of military conscriptions (at 21 years old).

The educational level of the mother and its impact on reducing childhood mortality during demographic modernization has been analyzed for different places. ³⁰ The educational level of mothers in urban Spain increased their children's chances of survival through better feeding patterns and greater attention to the advice of health authorities. ³¹ However, little is known about the impact of the death of a mother on her children's education. Mothers (especially literate ones) have played a fundamental role in the development of class attendance habits as well as transmitting the importance of education to their children. ³² At the same time, mothers, through their participation in the labour market, allow their children to attend school until later in life. ³³ Mothers thus prevent their children's possible early entrance into the labour market. ³⁴ According to Reher and González-Quiñones ³⁵ for the Spanish case: “A reduction of family income leads invariably to earlier entry of children into the labour market.”

However, in Table 4 we find that there were huge differences in literacy rates between motherless sons and daughters (who lost their mother in the first two years of life), and those with both parents alive. Looking 1850–1909 for this analysis (given that the first Spanish census with information on literacy is that of 1860), we can see that the difference in literacy between boys with both parents and boys without a mother is almost 12 percent, and the difference for girls is almost five percent. These differences by gender could be the consequence of the different economic role of boys and girls at early age. In the rural agrarian context that we are analysing, boys were more appreciated in agricultural tasks, so they had more job opportunities with which to substitute school attendance. At the same time, motherless children did not have a mother insisting on the importance of attending class. The role of father could not be a complete replacement of mother's role in this regard.

OPEN IN VIEWER

Table 4.

Percentage of Literate People (over 18 years old) by Mother's Survival Status, Middle Huerva, Aragón (Spain), 1850–1909.

	Literacy rate (percentage)		Educational gap (percentage point)
	Mother dies <24 months	Mother +10 years
Male	51.3%	62.9%	−11.6
Female	42.9%	47.4%	−4.5

Note: N(males) < 24m = 75; N(females) < 24m = 49; N(males) > 10y = 3,144; N(females) > 10y = 2,101. N(sex) < 24 means the number of cases where the mother died during the first 24 months of the child. N(sex) > 10y means the number of cases where the mother survived, at least, until her child reached 10 years old.

Source: AMHDB.

The results of Table 4 confirm that the mother not only played a fundamental role in the development of the child through nutrition: other cultural and social factors, such as educational level, were also affected by her absence. This influence could have played out in two ways. On the one hand, children might have needed to replace the mother's secondary breadwinner role through their own work. ³⁶ On the other hand, the greater importance of the mother in caring for the children and ensuring compliance in their educational habits (including attending classes and completing their homework) could have had negative effects on motherless children.

We also compare the biological wellbeing of motherless children and those with both parents alive by heights of military recruits. Heights are a cumulative indicator of the nutritional status of young soldiers from conception to the point of measurement. Parents play a fundamental role in height beyond the important genetic component. They are the ones who provide food and care for their children, especially when they are young. Biological wellbeing is determined by nutrients received, diseases experienced, and care received. In all these cases, parents play a fundamental role in: the distribution of family income; the type and quantity of food they give to their children; the care they give during childhood and adolescence; and the hygiene maintained to avoid contagious diseases.

Table A4 of the Appendix shows the stature of young men at conscription (around 21 years of age) in relation to the survival status of their parents. The sample includes 4,262 men. The results show that men whose mother died in the first two years of life were on average 14.3 millimetres shorter than men whose parents were alive during their childhood and adolescence. In this case, we can only analyze boys, but the result seems to confirm the results found above. Another interesting possible finding of Table A4 is that the differences in height could have converged over time. Just as the probabilities of death were reduced by improvements in living conditions, the gap in levels of biological wellbeing could also have been reduced. According to our results, even in a male breadwinner society, when mother's roles were not necessarily linked to their economic contribution to the family, their presence still had an effect on the survival of their children.

However, as mentioned above, the results in Table A4 should be treated with caution due to several limitations of the data. The small sample size, especially in the first and last sub-periods, could be biasing the sample. In addition, there are several biological factors (such as the growth process and its stages) that we cannot take into account and that could explain the results. ³⁷

To confirm the relationship between being motherless before the age of two and height, in Table 5 we have conducted four models of OLS linear regressions with heteroskedasticity-robust estimation. Our aim is to confirm, as elsewhere in Europe, that there was a negative relationship between height and motherlessness. ³⁸ This would be independent of the improvements in height that occurred throughout the nineteenth and twentieth centuries with economic and social modernization. For this purpose, we have taken as a dependent variable the height of the individuals at 21 years of age in millimetres. In relation with independent variables, the first model only includes motherlessness before the age of two and controls for village and decade of birth. The second model replaces motherlessness with fatherlessness. The third model includes both variables along with controls. Finally, the fourth model builds on model 3 but also includes paternal occupation, literacy and appeals to avoid military service. The results, in all models, indicate that motherlessness was related to an average height at age 21 that is between 1.2 and 1.3 cm shorter. No (significant) relationship was found for fatherlessness in deaths before the age of two years. Regarding the rest of the variables, being the child of a low-skilled worker was related to an average height approximately 1 cm shorter than others and having physical problems at age 21 was associated with a strong penalty in height of more than 2.5 cm. In short, all the multivariate models confirm the descriptive statistics. Male children who had lost their mother before the age of two suffered a strong penalty in biological wellbeing during childhood.

OPEN IN VIEWER

Table 5.

OLS Regressions. Determinants of Height in the Aragonese Rural Area of Study, Birth Cohorts 1856–1974.

Dependent variable: height at 21 years in millimetres		Basic model Motherless	Basic model Fatherless	Basic model M + F	Complete model
		(1)	(2)	(3)	(4)
Motherless	No	(ref.)
	Yes	−12.978** (5.94)		−13.022** (5.94)	−12.888** (5.90)
Fatherless	No	(ref.)
	Yes		3.260 (6.18)	3.444 (6.17)	5.912 (6.16)
Father's occupation	Farmer	(ref.)
	Low-skilled worker				−11.270*** (3.75)
	Artisan				0.771 (7.53)
	Upper class				2.390 (16.12)
	Other or unknown				−14.328** (5.83)
Literacy	Illiterate	(ref.)
	Literate				5.147 (6.03)
	Unknown				3.585 (9.04)
Appeals for exemption	No appeal (fit to serve)	(ref.)
	Physical appeals				−27.462*** (6.17)
	Social appeals				4.035 (4.79)
	Intercept	1564.41*** (14.17)	1563.77*** (14.19)	1564.29*** (14.18)	1572.95*** (17.01)
Control village		Yes	Yes	Yes	Yes
Control decade of birth		Yes	Yes	Yes	Yes
	Sample size	1,488	1,488	1,488	1,488
	R-sq	0.117	0.114	0.117	0.139

Source: Parish and municipal registers.

Notes: se (the values in brackets) denotes robust standard error. * Statistical significance at 10% level, ** at 5% level. *** at 1% level.

The differences in height that result from the regression models (Table 5) are slightly larger than in the table with descriptive statistics (Table A4 of the Appendix). This is possibly telling us that, controlling for a set of factors such as socioeconomic status or literacy, the role of the mother is even more evident in the biological well-being of the children. Even though in-depth research on this issue is necessary given the small sample size.