Saving Anthropometric History: A Solution to the “Estonian Antebellum Paradox”

Abstract

First of all, I claim that Jörg Baten and Mathias Blum (2015a) have provided inaccurate data on Estonian male height. I discuss in more detail the most broadly used international data sources on human height in the next chapter. At this point, I only state that the decadal average of 173.4 cm for adult Estonian men born between 1890 and 1899 is not supported by other sources. The most reliable reports are surveys of provinces, presented annually by Russian governors to the central government, as they are based on the largest number of observations (N). According to this source, the mean height of draftees (N = 1654) to the Russian army in 1913 from the Estland Province was 172.0 cm (Obzor Èstljandskoj gubernii na 1913 god: 44). I use this 172.0 cm value in my quantitative analysis, although this correction only makes Estonians appear as the second tallest males (following Swedes with 172.4 cm and sharing second place with New Zealanders).

First of all, I claim that Jörg Baten and Mathias Blum (2015a) have provided inaccurate data on Estonian male height. I discuss in more detail the most broadly used international data sources on human height in the next chapter. At this point, I only state that the decadal average of 173.4 cm for adult Estonian men born between 1890 and 1899 is not supported by other sources. The most reliable reports are surveys of provinces, presented annually by Russian governors to the central government, as they are based on the largest number of observations (N). According to this source, the mean height of draftees (N = 1654) to the Russian army in 1913 from the Estland Province was 172.0 cm (1914 Obzor Èstljandskoj gubernii na 1913 god: 44). I use this 172.0 cm value in my quantitative analysis, although this correction only makes Estonians appear as the second tallest males (following Swedes with 172.4 cm and sharing second place with New Zealanders).

For other countries, we accept data from Baten and Blum (2015a), extending them with NCD-RisC (2016) data for several countries (Belgium, Finland, Greece, Hungary, Romania and Slovakia) with no information in the first source. For Lithuania, I use the weighted mean value (166.4 cm) of conscript height in the Suvalkai Province in 1913 (165.47 cm; N = 1483) and the Kaunas Province in 1912 (166.76 cm; N = 4468, see Norkus, Ambrulevičiūtė et al. 2022). However, we were unable to attain height data for 1913 from the reports by governors of the Livland and Kurland provinces. Therefore, for Latvia, we use data from Baten and Blum (2015a). Conscript height data are also missing in the reports from the Vitebsk Province, including the three counties (u’ezd) of Eastern Latvia (Latgale). The same applies to the Suvalkai Province, where reports for various years (except 1913) have still not been located by researchers. Fortunately, height data for the Suvalkai Province for 1913 were found in the Lithuanian State Historical Archive (Podrobnyj otchiot 1913) (Table 12.1).

Table 12.1 Height of conscripts in the Kurland, Estland, Kaunas and Suvalkai provinces of the Russian Empire in 1907–1913

To control for the effects of differences in the genetic stock, only European countries and their so-called Western offshoots – Australia, Canada, New Zealand and the USA populated mainly by people of European descent (N = 30) are included. Comparing the decadal averages of 1890–1899 cohort height and 1890–1899 GDPpc data would perhaps allow us to make a stronger point. However, for the decade 1890–1899, GDPpc data are available only for a significantly smaller number of countries than in 1913. In fact, data for the Baltic countries in 1913 is absent in the MPD too, but we can use estimates from Norkus and Markevičiūtė (2021), presented in Chap. 8, Table 8.4: 3834 int $ in Latvia, 3341 in Estonia and 2650 in Lithuania.

The regression results confirm the general hypothesis that there is a positive correlation between GDP per capita and mean height. For the year 1913–1914, our estimated regression equation for men was: Height_male = 149.4375 + 2.2697 × ln(GDPpc). In the analysis, the natural logarithm of GDPpc was used to account for the non-linear relation (obvious from the scatterplot) between GDPpc and mean height. Moreover, this type of relationship is theoretically expected because otherwise it would predict human beings becoming giants as GDPpc grows. There is an upper limit for the increase in body height, set by genetic inheritance. As societies become affluent, this potential becomes exhausted, as the absolute majority of children then grow up in a favourable environment (Fig. 12.1).

Fig. 12.1

GDPpc in 1913 and adult male (born in 1890–1899) height. Data sources for height: Baten and Blum (2015a); NCD-RisC (2016); Norkus, Ambrulevičiūtė et al. 2022. Data source for GDPpc: MPD (2020), 2011 benchmark, author’s own estimations (for the Baltic countries)

Thus, a deceleration is seen in the increase of human height as GDPpc continues to grow. By 1914, Estonian men emerged as the most conspicuous outliers, distinguished by the largest positive residual (4.15 cm), that is, the difference between the predicted (167.85 cm) and observed height values (172.0 cm). How then can we explain the disparity between Estonia’s international ranking in the early twentieth century in terms of anthropometric measurements and the economic indicators of well-being?

There are many well-researched disparities between economic output and height. Hence, the search for an explanation should start with checking whether the Estonian male height paradox can be (dis)solved by regressing height on other variables, which are reputed to be the “usual suspects” in cross-country quantitative comparative research on environmental causes of height variation (Grasgruber et al. 2014, 2016; Koepke and Baten 2005; Steckel 2009, 2013). Economic output may be the most important background environmental cause accounting for variations in height. However, this cause can be substituted, reinforced or suppressed by other environmental causal factors, which can be described as proximate positive or negative causes of human growth.

Among the proximate positive causes, availability of animal protein-rich food is considered the most important in recent research. Consumption of milk and dairy products has a most beneficial impact on growth: ‘in conclusion, the most likely effect of dairy products supplementation is 0.4 cm per annum additional growth per 245 ml of milk’ (Beer de 2012: 307). However, this effect is conditional on lactose tolerance, which is a genomic feature unevenly distributed among world populations (Cook 2014).

For the availability of animal protein, we use the number of pigs per capita as the proxy, while availability of milk is measured by the number of cattle per capita. For both variables, we use Clio-Infra Dataverse data for 1900 (Klein 2015a, 2015b), supplementing it with data from 50 European provinces of Russia for the same year: 0.3346 cattle and 0.116 pigs per capita.¹ The data regarding lactose intolerance are from Ingram et al. (2009). According to this source, 57% of Estonia’s population is lactose-tolerant. What is missing in this source is information about the two other Baltic countries; hence, data were taken from Storhaug et al. (2017) about Latvia and from Urbonas (2016) about Lithuania. According to these sources, 67% of Lithuanians and 75% of Latvians are lactose-tolerant.

Inequality in the distribution of wealth and income is another factor, explaining repeatedly occurring positive correlation failures between GDP growth and mean height changes (Blum and Baten 2012; Blum 2013a, 2013b). Greater inequality deprives members of the lower socio-economic strata and their children of good nutrition and medical care, reducing the longevity of adults and stunting the growth of their children. The standard measure of income inequality is the Gini income distribution index. However, it cannot be calculated from taxation data for countries or historical periods when only a small part of the population paid income taxes, while the main sources of state revenue were indirect taxes or state monopolies. Alternatively, its values can be derived from the height variation coefficient CV (Baten and Blum 2014; Moradi and Baten 2005). For Estonia and Latvia, we used the Gini income index data for 1910 derived in this way from Moatsos et al. (2015).

To calculate the Gini income index value for Lithuania, which is missing in this data set, the archives of the former Kaunas Province were searched, perusing 251 files of 20–21-year-old Lithuanian men, who were drafted in 1913 (Prizyvnyje spiski po g. Kovno na 1913 g). The mean value for our sample was 166.13 cm (95 percent confidence interval: 165.31–166.95 cm), and the standard deviation was 6.68. From this data, we derived Gini_i1913 = 51.3, using formulas CV_it = σ_it /μ_it × 100 and Gini_it = −33.5 + 20.5 × CV_it. and Gini_income= = 9.25 + 10.46 × Gini_it, as recommended by the compilers of the income inequality database, where gaps in the historical income statistics data are filled by deriving them from height distribution data (Moatsos et al. 2015).

Infant mortality estimates for Estonia were derived by using the mean of the infant mortality values in the Estland and Livland provinces, and in Latvia, the mean of the Kurland and Livland provinces and the Latvian districts of the Vitebsk Province. The infant mortality mean value for Lithuania in 1897 was taken as the average of those for the Vilnius and Kaunas provinces (no data is available for the Suvalkai Province). Infant mortality is used as a proxy for the disease environment, which is shaped by the density of population, the incidence of infectious diseases and the availability of basic medical services (Baten and Blum 2015b; Baten and Blum 2014; Blum and Baten 2012).

Information about the total fertility rate (children per woman) in 1897 was available in the Gapminder data (Ferenc et al. 2017) for all our cases. Crude birth rates (CBR) per 1000 of population in 1897 are from Mitchell (2007), supplementing it with the estimates for Estonia, Latvia and Lithuania. The only difference between this and the estimates of infant deaths is that the CBR for Lithuania is the mean of the rates for the Kaunas, Vilnius and Suvalkai provinces. The fertility rate and CBR are considered proxies (imperfect, but the only ones available) for family household size.

Table 12.2 provides the descriptive statistics of the cases in the regression analysis in identifying the determinants for the variation in the height of males born in 1890–1899. Table 12.3 presents the results of the regression analysis.

Table 12.2 Descriptive statistics for male height and its determinants in 30 countries with populations of European descent. For sources, see references in this chapter

Table 12.3 Linear regression models for separate predictors of the mean height of men born in 1890–1899: standardised beta coefficients

Inspecting the regressions, no residuals were detected with Estonia’s residual ranking less than eight by its size. For GDPpc and cattle per capita, Estonia’s residual is the largest, while for infant mortality rate and lactose tolerance, its residual is the third largest. Instead of resolving our paradox, exploring the effects of GDPpc on height by other putatively relevant environmental variables only heightened the puzzlement. However, Estonia’s residuals are much lower in terms of absolute size and rank order position for height regressions on total women’s fertility rate and birth rate, pointing to the direction in which an explanation can be found for the Estonian male height exceptionality. The regression on total women’s fertility is significant only at the p < 0.1 level. However, the sample is heterogeneous, encompassing countries that were in different phases of (first) demographic transition around the juncture of the nineteenth and twentieth centuries (see Bogin 2001: 172–180; Chesnais 1993, Caldwell 2009).

Human height may have a correlation with decreasing fertility and birth rates (Grasgruber et al. 2016: 116) because of parental decisions to invest in the quality of the upbringing of their children (Becker 1960; Becker and Lewis 1973; Hatton 2017). However, even when demographic change has a paramount impact on height variation at the country level, the correlation between demographic variables and stature can still remain weak at the sample level if demographic transition pioneers are surrounded by countries in a similar stage of transition, either awaiting the transition or already in the early phases of transition.

We surmise such an effect in our sample. To recall, in the early phases of demographic transition, the increase in GDPpc is accompanied by an increase in natural population growth because of the sharp and rapid drop in mortality rates, while fertility and CBR remain at the former levels or decrease slower. However, there is broad variation among countries in the relative timing of mortality and fertility decrease. In a few cases (including the Baltic provinces, see Chap. 11), fertility transition may antedate the mortality transition. These complexities prevent us from identifying the demographic situation of specific countries unless a case study analysis is done, which is presented in the next section.

We argue that family household size is one of the most important factors accounting for the “height gap” between Estonia and Lithuania as well as between Latgale and mainland Latvia (see below in this chapter). Importantly, the Russian census of 1897 provides data about the mean size of rural and urban households (see Table 12.4). We consider as key evidence the data that rural households in the Baltic provinces (the Estland, Livland and Kurland provinces) were smaller than those in the neighbouring provinces of Lithuania, Russia and Belorussia. The most relevant point in the data is that the share of households with at least six persons was much smaller (26.3–29.3 percent) in the Baltic provinces than in the neighbouring ones (up to 54.8 percent). These differences across the provinces are in close correspondence with the variation of mean conscript height, as documented in the database of Russian historian Boris Mironov (2012).

Table 12.4 Family household size (in 1897) and conscript height in the Baltic provinces, Lithuania, and the neighbouring governorates (provinces) of the Russian Empire

Other environmental variables, discussed in the previous section, are less promising in explaining the regional cross-country differences. At the sample level, access to animal proteins (proxied by cattle per capita) is the most important powerful variable. However, according to our data, among populations of the late nineteenth–early twentieth centuries in lesser socio-economically advanced regions, Lithuania had better access to animal proteins (due to the dominance of subsistence agriculture), milk (with a lower share of lactose-intolerant persons) and a less stressful disease environment (due to lower levels of urbanisation) than the Baltic provinces.

Advancing a demographic explanation, we draw on the rapidly growing body of research on the relationship between family size and height (Desai 1995; Lawson and Mace 2009; Öberg 2017), wherein researchers report a strong negative correlation between these variables (as family size grows, mean height decreases), with most explaining this finding as an outcome of the optimising choice of parents in the context of a ‘quality-quantity tradeoff’ (Becker 1960; Becker and Lewis 1973). As children in small families receive the resource shares of their unborn competitor siblings, they grow taller and are better educated. The thesis that agrarian reform was related (via changes in demographic behaviour) to the increase in mean height can be substantiated by a parallel tracing of demographic, anthropometric and economic changes during the decades following the agrarian reform.

In his doctoral dissertation defended at Tartu (Dorpat) University, Oscar Grube (1878: 33) calculated 164.28 cm as the mean height value for 17- to 60-year-old Estonian males (N = 100), measured in 1875–1878. According to Dmitrij Anučin (1889: 77), the mean height of conscripts from the Estland Province in 1874–1883 was 166.7 cm. Whereas Grube’s data is questionable because of its small N size, Anučin’s report is reliable because it is grounded in a very large sample.² In 1893, Aleksei Kharuzin (1893: 303) published his analysis of the anthropometric data of conscripts from the Estland Province, measured in November 1892. According to his calculation, the mean height of conscripts was 170.981 cm (N = 2523). Such a rapid height increase cannot be explained merely by economic growth, as Estonia remained an agrarian country suffering from the great European agrarian depression from 1870 to 1896 that was unleashed by the competition of imports from the USA. This was the time when the draftees measured in 1892 were growing up.

We argue that the protracted agrarian crisis merely accelerated the changes in demographic behaviour, leading to lower birth rates and smaller households. Indeed, the data of the first modern census in the Baltic provinces conducted in 1881 suggest a rapid and dramatic change in the family structure in the wake of agrarian reforms: ‘judging by the findings of the 1881 census, the researcher is tempted to hypothesize that the mid-century decades witnessed a major change in Baltic rural familial relationships. Judging by the tables, an era of complex rural family networks as depicted in the soul revisions seemed to come to an end abruptly, and an era of the simple family began equally abruptly’ (Plakans and Wetherell 2004: 58). Could such a change really transpire without displaying any demographic effects?

Our case-oriented arguments involve two comparisons of the available data on agrarian and demographic change on the one hand, and height variation on the other. First, we compare Estonia and its northern neighbour Finland, whose population shares a common genetic stock and speaks similar languages mutually understandable without previous instruction (in 1918–1919, plans and negotiations were underway for establishing a common state). According to NCD-RisC (2016) data, Finnish males born in 1896 were 167.74 cm tall on average, which means that they were 4.26 cm shorter than their Estonian peers (172.0 cm). However, after some 90 years (among men born in 1980–1989), the height gap between Finns (178.2 cm) and Estonians (179.1 cm) contracted to less than 1 cm (Baten and Blum 2015a).

Unfortunately, the Russian census of 1897 did not cover Finland, and hence, we cannot directly compare the mean size of the households in Finland and those in the Baltic provinces. However, the larger birth rates imply that Finnish boys grew up in larger families than their peers in the Baltic provinces. This may explain why Finnish male adults born in 1899–1900 were shorter than their Estonian peers with similar genetic stock, although their homeland was not poorer than the Baltic provinces (although it was much poorer than Sweden and the other Nordic countries at this time). The later start of demographic transition in Finland (see Chap. 10), related to institutional differences in land use compared to the Baltic provinces accounts for the emergence of the broad height gap of Finns behind the Estonians in the early twentieth century.

Another source that sheds light in tracing the connection between demographic transition and body height change is the cross-regional comparison of Eastern Latvia with its regions belonging to the Baltic provinces. Aggregate data in Baten and Blum (2015a) and NCD-RisC (2016) conceal cross-regional differences in height in Latvia. During the first independence years, the Latvian Statistical Office published data on Latvia’s conscript height means for all of Latvia and its regions (see Table 12.5). The data disclose mean height disparity between Eastern Latvia, encompassing three districts of the former Vitebsk Province, and populations of Latvia’s regions that were formerly part of the “Baltic provinces”. These are Vidzeme (the Latvian part of the former Livland Province), Kurzeme and Zemgale (together encompassing the territories of the former Kurland Province).

Table 12.5 Mean height of conscripts (20-year-olds) to the Latvian army from different regions of Latvia

By 1939, conscripts from this part of the country still remained 2 or 3 cm shorter than their peers from Central and Western Latvia in 1939, as documented by Nikolajs Cauna (Kokare and Cauna 1999): Kurzeme and Zemgale 172.03 cm, Vidzeme 172.45 cm, Latgale 169.87 cm, and the all-Latvian mean as 171.70 cm. Contemporary Latvian biological anthropologist Inese Kokare (1998) compared these data with her own measurements of conscripts to the army of the restored independent Latvia in 1996. According to her findings, all the physical characteristics of draftees had improved since 1939, but those of draftees from Latgale had improved the most, including height. There was still a small height difference between the mean height of the enlisted draftees from Latgale (176.71 cm) and the all-Latvian mean (177.63 cm), but it contracted to some 9 mm, in comparison to the nearly 2 cm reported in 1927 and 1939 (Kokare and Cauna 1999: 378–379).

Like Finland and Estonia, Eastern and mainland Latvia are populated by people with the same genetic stock, so the existence of a height gap between Eastern and mainland Latvia can only be explained by differences in culture and institutions. Eastern Latvia is Catholic, while mainland Latvia is Lutheran Protestant (cp. Norkus 2022). The difference in religion certainly accounted for the endurance of the fertility gap between Eastern and mainland Latvia, where birth rates during the interwar period were at around the same level as in Estonia. However, the most important factor was the difference in the institutions of land property and land use, where Eastern Latvia was more similar to Lithuania (see Chap. 10) than to mainland Latvia.

As in Lithuania, serfdom was only abolished in Eastern Latvia in 1861. Complex extended family households survived until the early twentieth century. After the Stolypin reform in 1906 abolished obstacles to the legal division of allotments, the surviving extended households were divided, and Latgale became a region of small landholdings. In 1861, the mean size of a peasant landholding was 19.4 ha in the Dinaburg (Daugavpils), 19.0 ha in the Rezhica (Rēzekne) and 27.3 ha in the Lyucin (Ludza) districts of Latgale. By 1910, this size had reduced to 11.1 ha, 10.4 ha and 10.6 ha, respectively (Efremova 1982: 27), displaying the typical Malthusian dynamics of agrarian overpopulation. For comparison, on the eve of WWI, the mean size of a farm in the Latvian part of Livland (Vidzeme) was 47.0 ha and 41.5 ha in Kurland (Skujeneeks 1927: 402).

Despite the prohibition to divide a farm among successors if it was smaller than 10 ha, land property fragmentation continued in Latgale during the period of interwar independence (Efremova 1982: 225). Agrarian reform did little to reduce the extent of hunger for land here because there was less land to redistribute and more applicants. Similarly to Lithuania, its major impact was the dispersal of villages into farmsteads with ensuing liquidation of the open fields system (Malahovskis 2014). Suffering from agrarian overpopulation, Latgale sent its excess population to Saint Petersburg (only before World War I) and to Riga to seek employment in industry, construction or on farms in the more advanced regions of Latvia.

Under restored independence, Latgale still remains a less economically advanced region of Latvia. However, under the Soviet regime, this region went through forced collectivisation (making individual farmers members of large collective farms) and urbanisation. The side effect of these changes was the passing of particular features in the organisation of family, making Latgalian families similar to those in other regions of Latvia or the Baltic republics (Terent’eva and Šlygina Natalija 1962). Family farm households were replaced by conjugal family households with one or two children, both parents employed outside of the family and children spending a large part of their day in day care establishments or schools.

Most importantly, Latgale experienced the most dramatic decline in fertility, making it the region with the lowest birth rates in restored independent Latvia (Centrālā statistikas pārvalde. 2018: 126–127, 143). The final disappearance of the large traditional Latgalian family explains the fading out of many bodily features considered by some Soviet human biologists as characteristic expressions of the ‘Eastern Baltic’ anthropological type or (sub)race (e.g. Vitov 1959).