Height, occupation, and intergenerational mobility: an instrumental variable analysis of Dutch men, birth years 1850-1900

ABSTRACT

Height and labor market outcomes appear to be related to one another. The taller people are, the more likely they are to have better jobs and to earn more money. This is especially the case for men. However, whether height is causally related to labor market outcomes is an open question, which instrumental variable (IV) analysis may help to answer. To our knowledge, no study has yet used IV analysis to test these relationships in a historical context. The present study addressed this gap, by examining height’s relationship to occupational status and intergenerational mobility in a sample of Dutch men, birth years 1850 through 1900. Data were drawn from: the Historical Sample of the Netherlands, providing life course information on the research person; the Heights and Life Courses Database, providing information on the research person’s height at conscription; and the Male Kin Height Database, providing information on the average height of the research person’s full brothers. This combination of data sources yielded a sample of 1,465 men. Height z-score’s relationships to occupational status (characterized as HISCAM score), and to intergenerational mobility (characterized as the difference between research person’s HISCAM score and father’s HISCAM score) were examined. The average of brothers’ heights z-score was used as an instrumental variable. In terms of results, one standard deviation increase in height was associated with a 1.370 increase in HISCAM score (95% CI: 0.310–2.429), and a 1.127 increase in intergenerational mobility score (95% CI: −0.114–2.368). As Dutch men were growing taller and had greater abilities to choose their occupations, it appeared that tallness was associated with a better job, and increased intergenerational occupational mobility. This study thus offered preliminary evidence that height and labor market outcomes were perhaps causally related during the late nineteenth and early twentieth centuries.

KEYWORDS:

• Height
• occupation
• labor market
• instrumental variable analysis
• occupational mobility
• The Netherlands

1. Introduction

Height and labor market outcomes appear to be related to one another. Taller individuals are more likely to have better jobs, and to earn more money. This appears to be particularly the case for men. Already, this topic has attracted a great deal of attention from researchers. Two literature reviews, Hübler (2016) and Thompson et al. (2022), have studied height’s relationship to wages, and have found evidence of positive, significant associations.

However, whether height is causally related to labor market outcomes is an open question. Evidence of this relationship is largely based on observational data, and non-causal analyses. These estimates generally suffer from omitted variable suspicion. In all likelihood, height is endogenous, and is correlated with unobserved factors. For example, height is probably confounded by genetics, information that most observational datasets lack, and early-life environmental conditions, which are also difficult to sufficiently correct for with observational data. Instrumental variable (IV) analysis offers a solution to this issue: by virtue of accounting for unobserved confounders, IV analysis can yield more causal estimates than other analysis types (Iwashnya & Kennedy, 2013). Using IV analysis would thus facilitate better understanding of height’s relationship to labor market outcomes.

To date, several studies have examined height’s relationship to labor market outcomes, but have arrived at conflicting conclusions. For instance, Wang et al. (2020) and Böckerman et al. (2017), with data from China and Finland, respectively, found that height was not related to labor market outcomes, based on the results of IV regressions. However, Tyrrell et al. (2016), exploiting data from the U.K., found that height and labor market outcomes were indeed related, although IV estimates were smaller than ordinary least squares (OLS) estimates. When examining data from the Ivory Coast and Ghana, Schultz (2003) found significant relationships between height and wages, with larger IV estimates, compared to OLS estimates. It is not yet clear why the strength of height’s relationship to wages might vary across studies.

However, initial evidence suggests that labor market structure, as well as the broader contexts in which studies are set, may play a role in determining the strength of height’s relationship to labor market outcomes. Based on the results of a systematic review, Thompson (2022) found that height’s relationship to wages was stronger in lower-resource settings than in higher-resource ones. Although height’s relationship to labor market outcomes has been studied in relatively diverse contemporary settings, it has not yet been studied when using a historical sample: all existing studies using IV methods examined modern samples, born in the mid- to late-twentieth century. This may represent a knowledge gap, because height’s relationship to wages appears to vary over time (Bleakley et al., 2014). To understand whether this relationship is present in historical populations, it must specifically be studied. Doing so may help to ultimately identify the mechanisms underlying height’s relationship to labor market outcomes, by highlighting moments of convergence and divergence across different contexts. To that end, a sample of Dutch men born in the nineteenth century was examined in the present study, and analyzed using IV methods.

The present study also differed from existing studies on the topic in several other important ways. Instead of wages, labor market outcomes were characterized as occupational status based on job titles and intergenerational occupational mobility (the difference between an individual’s occupational status score and that of his father). Further, the average height of a research person (RP)’s brothers was used as an instrumental variable, in contrast to genetic variants (e.g. Böckerman et al., 2017; Böckerman & Vainiomäki, 2013; Tyrrell et al., 2016; Wang et al., 2020) and economic indicators and parental education level (Schultz, 2002, 2003). Using the average height of an RP’s brothers may help to account for shared family inheritance, such as genetic similarities among same-sex siblings, and shared household environments during development. This study ultimately examined the question: to what extent was height related to occupational status and intergenerational mobility among Dutch men, birth years 1850–1900, when using brothers’ heights as an instrument?

1.1. Theoretical framework

Why might height be related to labor market outcomes? Height is reflective of beneficial conditions during development, from conception until maturity. In particular, adult height reflects net nutrition, or the quantity and quality of food an individual consumes, minus different stressors (Floud et al., 2011). These stressors include: basal metabolism, disease, physical labor, and ‘fight or flight’ responses (Floud et al., 2011). All else being equal, the taller someone is, the more beneficial their early-life conditions tend to have been. Of course, genes also play a role. Silventoinen (2003) estimated that, in modern, high-income contexts, 80% of human height variation is due to heritability, with the remaining 20% due to environmental conditions. In poorer and/or historical contexts, environmental conditions likely played a larger role in adult height variation (Silventoinen, 2003).

Further, taller height may be associated with one or several beneficial characteristics in adulthood that are valuable on labor markets, because height and these characteristics may be determined by similar environmental and genetic influences. Improved health is thought to be one of these characteristics, and has been shown on a population level (e.g. Fogel & Costa, 1997), although some studies have found weak or non-linear relationships between height and measures of health, namely mortality (e.g. Alter et al., 2004). Increased health (and height) would enable individuals to work harder, and to be more productive (Floud et al., 2011). Researchers have argued that other related characteristics that are rewarded on labor markets, including greater intelligence (Case & Paxson, 2008); social skills improved social skills (Persico et al., 2004); greater interpersonal dominance (Stulp et al., 2015); and self-esteem (Booth, 1990; Prieto & Robbins, 1975). Perhaps because it is very difficult to fully account for these beneficial characteristics, as well as the early-life determinants of height, correlations between height and labor market outcomes are often found. If these factors were more fully accounted-for, the relationship between height and labor market outcomes would perhaps be weaker.

However, even if these beneficial characteristics had been taken into account, height might be related to labor market outcomes. If a study exploited ‘perfect’ observational data, including full life-course information on identical twins (accounting for both genetics and shared early-life conditions) and their characteristics in adulthood, a relationship between height and occupation could still be found. Here, height would perhaps function as a signal of a characteristic, for instance, intelligence, strength, health or dominance, even if the person in question does not actually possess this trait (Batres et al., 2015). If height was related to labor market outcomes after sufficiently accounting for these beneficial characteristics, and conditions in early-life, we could say that these relationships were causal.

These beneficial characteristics, whether real or perceived, may also help to explain why height’s relationship to wages may vary by context. For instance, strength may be more strongly valued in labor markets with a greater share of manual occupations (Schultz, 2002; Thomas & Strauss, 1997). Alternatively, this may mean that, tallness is less strongly associated with occupational success in these contexts: taller individuals may be more likely to sort into more manual occupations, because the opportunity cost of education was too high (Costa, 2015). Bleakley et al. (2014) found evidence of this, as the relationship between height and income was stronger in the earlier twentieth century United States than in the nineteenth century. However, when comparing the United States with Ghana and Brazil, Schultz (2002) found that height’s relationship to wages was smaller in the United States and much larger in Ghana and Brazil. Schultz (2002) offers evidence that the strength of height’s relationship to occupational outcomes is not simply a matter of how much a labor market values ‘brawn’ over ‘brains’.

1.2. Context

It is therefore relevant to examine height’s relationship to occupational outcomes in a context when both heights and the occupational structure were changing. This was the case in the nineteenth-century Netherlands. First examining heights, the Dutch were growing remarkably fast in this period. While they were relatively short compared to their European peers in the mid-nineteenth century, the Dutch were the tallest nation in the world by the mid-twentieth century. Figure 1 puts the Netherlands’ secular growth trend in comparative perspective, by using data from Baten and Blum (2012). These men’s heights were measured at conscription. While conscription age varied by country, conscription generally took place around twenty years of age. Baten and Blum (2012) also applied corrections to height measurements if conscription age varied within countries over time (Baten & Blum, 2012).

Figure 1. Heights of conscripts in comparative perspective.

The labor market also experienced several profound shifts during this study’s research period. The industrial revolution arrived rather belatedly to the Netherlands, after 1860 (Mokyr, 2000). Used as a rough indicator of the rate of industrialization, the number of steam engines purchased in the Netherlands increased 500% between 1865 and 1900 (Schulz, 2013). Industrialization was accompanied by several other trends, including mass communication and transportation, and increased migration, secularism and urbanization.

These processes have long been assumed to be associated with changing occupational structure and intergenerational mobility (Treiman, 1970). There is some evidence that this was the case in the Netherlands. Between 1807 and 1909, the share of individuals working in industrial jobs increased from 26.2% to 34.4% of the labor force, while the share of individuals working in agriculture declined from 43.1% to 30.4% (Smits et al., 2000). There is also evidence that educational access was expanding, with the number of students enrolled in secondary education increasing sixfold between 1865 and 1900 (Schulz, 2013). Schulz (2013) argued that ‘with the occupational and educational structure being “upgraded”, people increasingly got better jobs’ (p. 8). This likely impacted intergenerational occupational mobility: sons were less likely to inherit their father’s jobs, because their father’s jobs were increasingly out-of-step with the labor market (Knigge et al., 2014). It appears that the idea of an upwardly mobile career emerged in Europe in the latter half of the nineteenth century (Maas & van Leeuwen, 2019; van Leeuwen & Brown, 2004). This may be relevant for this study, in that an individual’s occupation was increasingly determined by his own capabilities (perhaps as reflected and/or signaled by his height), rather than by the occupation of his father.

What might this context mean for the relationship between height and labor market outcomes? While the general consensus in the literature is that height and labor market outcomes are positively related, it is not immediately clear what the strength of this relationship would be in the late nineteenth- and early twentieth-century Netherlands (Thompson et al., 2021). If, as Bleakley et al. (2014) hypothesized, height’s relationship to labor market outcomes strengthens as the share of occupations shifts from being largely manual to largely non-manual applies to other settings, we should expect to see a weaker relationship between height and labor market outcomes, relative to studies conducted in modern, higher-resource settings. If, as Schultz (2002) hypothesized, health capital, as indicated by height, is more valuable in lower-resource settings than in higher-resource ones, we might expect to find a stronger relationship between height and labor market outcomes.

2. Methods

2.1. Data

The starting point of this study’s dataset was the Historical Sample of the Netherlands (HSN). The HSN is a representative sample of Dutch people born between the years of 1812 and 1922, and contains, at a minimum, birth certificates for these individuals (Mandemakers, 2000). In most cases, the HSN also includes death certificates, and marriage certificates if an RP was married. The HSN in full contains approximately 85,500 RPs. From 1850 onward, population registers, which recorded much more detailed information on household composition, were implemented. Between 1910 and 1920, this was followed by family and personal cards, which recorded similar information to the population registers, and ended in the late 1930s (Mandemakers, 2019).

Next, a sample of male RPs was linked to their conscription records, in the Heights and Life Courses database, which is unique to the Giants of the Modern World project (Kok et al., 2016; Mandemakers, 2019). This database contains height information from nine of the eleven nineteenth-century Dutch provinces. For this study, only RPs who survived until age 40 were included, so that RPs would have sufficient time in the workforce to develop their careers. Because the final observation year is around 1940, and in order to observe all RPs until age 40, this study’s research period ended at birth year 1900. For birth years 1850 to 1900, there were 3,396 unique RPs with height information who were observed until at least age 40. Examining the representativeness of this dataset, Quanjer and Kok (2020) found evidence of some minor selection biases (e.g. the under-representation of the sons of elites). Overall, however, this dataset represents a relatively unbiased anthropometric source that appears to be reflective of the overall Dutch population.

Finally, the Male Kin Height database, also unique to the Giants of the Modern World project, contains information about RPs’ male kin (Kok et al., 2016; Mandemakers, 2019). More specifically, this dataset includes information from a sample of RPs’ male kin’s conscription records (including their heights), as well as marital and death information from civil certificates (Kok et al., 2016; Mandemakers, 2019). It is worth emphasizing that not all RPs’ male kin were searched for, or found. The Male Kin Height database contains a sample of RPs’ male kin, and excludes several provinces that were included in the Heights and Life Courses database. As mentioned, information specifically from full brothers, who shared a mother and a father with the RP, were used. There were 3,189 brothers who were linked to RPs in this dataset. However, many RPs have multiple brothers in the dataset, with 1,465 unique brother groups with complete covariate information in this dataset. Because brothers’ average heights were applied to their corresponding RPs, ultimately a sample of 1,465 RPs was used in this study.

2.2. Variables

Two labor market outcomes were tested: occupational status and intergenerational occupational mobility. First, to characterize occupational status, the RP’s highest occupational status recorded in the HSN’s population registers was used. One of the drawbacks of the population registers is that the exact timing of a change in occupation is unknown, but it is known what occupation an RP had at varying points in his life (Schulz, 2013). For example, we could not examine all RPs’ occupational status at age 45. To characterize occupation, the Dutch-specific HISCAM scale, version 1.3.1, was used (Lambert et al., 2013). HISCAM, or the historical CAMSIS scale, is an occupational stratification scale that is suitable for use for the period from 1800 through 1938 (Lambert et al., 2013). The scale is based on the premise that people who are closer in social position interact more often (Lambert et al., 2013). Social connections were identified primarily through marriage records, particularly intergenerational occupational comparisons at the moment of marriage (Lambert et al., 2013). HISCAM scores are continuous (ranging from 40.24 to 99.00), with a higher score indicating a higher occupational status.

To characterize intergenerational occupational mobility, the RP’s highest HISCAM score and his father’s highest HISCAM score were differenced. For this variable, we only had information for 1,319 RPs (instead of the full sample of 1,465 RPs), and therefore only included these 1,319 RPs in the intergenerational mobility analyses.

Height was the key independent variable in all analyses. Height was collected at conscription to the nearest millimeter, when the majority of RPs were 20 years old. In this study, height was characterized as z-scores and weighted by ten-year birth cohorts, because there is a time trend in height (Carslake et al., 2013).

As mentioned, brothers’ height was treated as the exogenous instrument in the IV analyses. Because, in some cases, RPs have multiple brothers in our dataset, the average height of brothers was used. A priori, we expected the average height of brothers to be a suitable instrument. We anticipated that brothers’ average height was correlated with the height of the RP (e.g. Alter & Oris, 2008). The Pearson’s correlation coefficient between RPs’ and brothers’ heights was 0.51. We also expected that brothers’ height was not directly associated with the RP’s HISCAM or intergenerational mobility score, so that the exclusion restriction was satisfied. While the exclusion restriction is empirically untestable (e.g. Öberg, 2021), we nonetheless sought to assess more generally whether the exogenous variable, the average of the RPs’ brothers’ heights, was related to the outcomes. We regressed the HISCAM and intergenerational mobility scores on the average RP’s brothers’ height, while excluding the RP’s height, the endogenous explanatory variable. We found small, insignificant relationships between the average height of RPs’ brothers and the outcomes, indicating that there was little predictive power of the instrumental variable on the outcomes (Wang et al., 2020).

Further, in adjusted analyses, a number of factors that likely confounded the relationship between height and occupational status were included. For instance, birth period, characterized as ten-year birth cohorts, and region of birth were included. The eleven nineteenth-century Dutch provinces were categorized into four regions, based on geography and economy type: north (Friesland, Groningen, Drenthe); middle (Utrecht; Overijssel; Gelderland); coastal (Noord and Zuid Holland); and south (Noord-Brabant, Zeeland, Limburg).

Father’s occupational status was also included in adjusted analyses. The highest occupation available for fathers in the HSN was used, and was characterized as HISCLASS-5 score. When the father’s occupational status was unknown, the mother’s occupational status was used instead. HISCLASS is a widely-used and validated historical occupational classification system, with its five-category, condensed version most appropriate for inclusion in quantitative analyses of relatively small samples (van Leeuwen & Maas, 2011). Six categories were included: elite; middle class; skilled workers; farmers; unskilled workers; and unknown.

Other covariates were included in adjusted analyses, namely: religion; migrant status, or whether an RP migrated to Amsterdam or Rotterdam, the major urban centers of the period, before conscription age; the number of siblings an RP had at age ten; municipality size at birth, based on the 1889 census (Centraal Bureau voor de Statistiek, 2011); and the infant mortality rate in the RP’s municipality of birth, using Ekamper and van Poppel (2008)’s dataset. Municipality size and infant mortality were characterized as quintiles. These categories were aggregated for all HLC RPs, and not only the RPs used in the present study. These categories are therefore not evenly distributed among this study’s 1,465 RPs. We also tested whether other covariates, including sibling sex ratios and birth order, were covariates, but found they were not. These covariates were ultimately excluded from analyses.

2.3. Analyses

All analyses were performed in Stata version 16. Descriptive statistics were first generated. Second, height’s relationships to both occupational status, measured by HISCAM score, and intergenerational mobility, or the difference in HISCAM scores between the RP and his father, were analyzed using ordinary least squares analyses.

Third, height’s relationships to both outcomes were tested with IV analyses. The two-stage least squares method was used. As mentioned, the RP’s height was treated as the endogenous variable, and brothers’ average height was treated as the exogenous variable/instrument.

In terms of the adjusted model specifications, the RP’s height was regressed on the average height of his brothers, the instrument, as well as other covariates in the first stage. In the second stage, either HISCAM score or the intergenerational mobility score was regressed on the predicted value of RP’s height, along with other covariates. The two stages of the analyses are specified in Equations (1)(1) $\begin{array}{rl}& \hat{x}={\gamma }_0+{\gamma }_{1}BrotherHeigh{t}_1+{\gamma }_{2}BirthCohor{t_2}^{+\dots +}{\gamma }_{5}BirthCohor{t}_5+\\ & {\gamma }_{6}Regio{n_6}^{+\dots +}{\gamma }_{9}Regio{n}_9+{\gamma }_{10}FatherHISCLAS{S_{10}}^{+\dots +}{\gamma }_{14}FatherHISCLAS{S}_{14}+\\ & {\gamma }_{15}Religio{n_{15}}^{+\dots +}{\gamma }_{18}Religio{n}_{18}+{\gamma }_{19}Migran{t}_{19}+{\gamma }_{20}NumberSibling{s_{20}}^{+\dots +}\\ & {\gamma }_{22}NumberSibling{s_{22}}^{+\dots +}+{\gamma }_{23}MunicipalitySiz{e_{23}}^{+\dots +}{\gamma }_{26}MunicipalitySiz{e}_{26}+\\ & {\gamma }_{27}IM{R_{27}}^{+\dots +}{\gamma }_{31}IM{R}_{31}+\epsilon ,\end{array}$(1) and (2(2) $\begin{array}{rl}& y={\beta }_0+{\beta }_1{\hat{x}}_1+{\beta }_{2}BirthCohor{t_2}^{+\dots +}{\beta }_{5}BirthCohor{t}_5+{\beta }_{6}Regio{n_6}^{+\dots }\\ & {}^{+}{\beta }_{9}Regio{n}_9+{\beta }_{10}FatherHISCLAS{S_{10}}^{+\dots +}{\beta }_{14}FatherHISCLAS{S}_{14}+{\beta }_{15}Religio{n}_{15}\\ & {}^{+\dots +}{\beta }_{18}Religio{n}_{18}+{\beta }_{19}Migran{t}_{19}+{\beta }_{20}NumberSibling{s_{20}}^{+\dots +}\\ & {\beta }_{22}NumberSibling{s_{22}}^{+\dots +}+{\beta }_{23}MunicipalitySiz{e_{23}}^{+\dots +}{\beta }_{26}MunicipalitySiz{e}_{26}+\\ & {\beta }_{27}IM{R_{27}}^{+\dots +}{\beta }_{31}IM{R}_{31}+\nu ,\end{array}$(2) ).First stage:

(1) $\begin{array}{rl}& \hat{x}={\gamma }_0+{\gamma }_{1}BrotherHeigh{t}_1+{\gamma }_{2}BirthCohor{t_2}^{+\dots +}{\gamma }_{5}BirthCohor{t}_5+\\ & {\gamma }_{6}Regio{n_6}^{+\dots +}{\gamma }_{9}Regio{n}_9+{\gamma }_{10}FatherHISCLAS{S_{10}}^{+\dots +}{\gamma }_{14}FatherHISCLAS{S}_{14}+\\ & {\gamma }_{15}Religio{n_{15}}^{+\dots +}{\gamma }_{18}Religio{n}_{18}+{\gamma }_{19}Migran{t}_{19}+{\gamma }_{20}NumberSibling{s_{20}}^{+\dots +}\\ & {\gamma }_{22}NumberSibling{s_{22}}^{+\dots +}+{\gamma }_{23}MunicipalitySiz{e_{23}}^{+\dots +}{\gamma }_{26}MunicipalitySiz{e}_{26}+\\ & {\gamma }_{27}IM{R_{27}}^{+\dots +}{\gamma }_{31}IM{R}_{31}+\epsilon ,\end{array}$(1)

where $\hat{\mathrm{x}}$ = the predicted value of RP’s height; γ₁ = the exogenous instrument; γ₂−γ₃₁ = exogenous covariates; and ε = the error term.Second stage:

(2) $\begin{array}{rl}& y={\beta }_0+{\beta }_1{\hat{x}}_1+{\beta }_{2}BirthCohor{t_2}^{+\dots +}{\beta }_{5}BirthCohor{t}_5+{\beta }_{6}Regio{n_6}^{+\dots }\\ & {}^{+}{\beta }_{9}Regio{n}_9+{\beta }_{10}FatherHISCLAS{S_{10}}^{+\dots +}{\beta }_{14}FatherHISCLAS{S}_{14}+{\beta }_{15}Religio{n}_{15}\\ & {}^{+\dots +}{\beta }_{18}Religio{n}_{18}+{\beta }_{19}Migran{t}_{19}+{\beta }_{20}NumberSibling{s_{20}}^{+\dots +}\\ & {\beta }_{22}NumberSibling{s_{22}}^{+\dots +}+{\beta }_{23}MunicipalitySiz{e_{23}}^{+\dots +}{\beta }_{26}MunicipalitySiz{e}_{26}+\\ & {\beta }_{27}IM{R_{27}}^{+\dots +}{\beta }_{31}IM{R}_{31}+\nu ,\end{array}$(2)

where y = the value of HISCAM scores or intergenerational mobility scores; $\hat{\mathrm{x}}$ = the predicted value of RP’s height from the first stage equation; β_2- β₃₁ = exogenous covariates; and ν = the error term uncorrelated with the other variables in the equation.

The results of the second-stage analyses were presented in the results section. The results of the first-stage analyses were presented in Appendix A. For all analyses, beta coefficients were reported. For IV analyses, Cragg-Donald Wald F-statistics were reported. F-statistics of 10 and lower were considered to indicate weak instruments (Cragg & Donald, 1993).

2.4. Robustness checks

Several robustness checks were performed in order to increase confidence in this study’s findings, and were reported in Appendix B. First, the distribution of HISCAM scores may have been a source of bias. Figure 2 illustrates that most HISCAM scores were between 42 and 70, with a positive skew. To identify whether this impacted this study’s results, an IV regression was performed, examining height’s relationship to maximum HISCAM score, when only including RPs with HISCAM scores less than 70 (n = 1,329). Results were similar to those of the main analyses.  

Figure 2. Histogram of maximum HISCAM score.

Figure 3 illustrates the distribution of the intergenerational change in HISCAM scores, and makes clear that the largest share of RPs experienced no change in HISCAM score relative to their fathers. It may be that the significant findings in these analyses were generated by a few RPs making large jumps in HISCAM scores. To assess whether this was the case, height’s relationship to intergenerational mobility as a binary variable (1 = a change in intergenerational mobility greater than 1; 0 = a change in intergenerational mobility equal to or less than 1) was analyzed with logistic regression. Again, results were similar to those of the main analyses.

Figure 3. Histogram of intergenerational change in HISCAM score.

To assess whether choice of instrument impacted this study’s findings, IV analyses were performed when using the height of the RP’s father as an instrument. In the robustness check for height’s relationship to HISCAM score, results were somewhat similar to those reported in the main results. However, these results were not significant. We also found highly insignificant results for height’s relationship to intergenerational mobility. This may be because of type II error, as a result of using a smaller sample.

More generally, this study used a relatively small sample (n = 1,465), and only included individuals with brothers present in the dataset. To assess whether this was biasing our results, we performed OLS analyses with RPs born between 1850 and 1900, who did not have a male sibling recorded in the Male Kin Height database (n = 1,572). We obtained similar results when using this sample, and the sample of RPs who had at least one brother in the Male Kin height database, and who were included in this study’s main analyses.

3. Results

3.1. Sample characteristics

Table 1 presents this study’s sample characteristics. The HISCAM scores of RPs ranged between 40.2 (working class) and 99.0 (elite), the minimum and maximum, respectively, on the Dutch HISCAM scale, with an average HISCAM score of 57.6. The intergenerational change in HISCAM score ranged between a decrease of 48.7, and an increase of 50.3, with an average of a 4.0 increase between fathers and sons. RPs measured on average 168.1 cm, and their brothers measured on average 168.2 cm.

Table 1. Sample characteristics.

Outcome variable	Obs.	Mean	SD	Min.	Max.
Highest HISCAM score	1,465	57.64	10.30	40.24	99.00
Intergenerational change in HISCAM score	1,319	4.03	10.34	−48.66	50.30
Endogenous variable and instrument
RP height (cm)	1,465	168.06	7.03	131.70	198.90
Average height of brothers (cm)	1,465	168.17	6.51	135.6	189.50
Covariates	Obs.	%
Birth cohort
1850–1859	151	10.31
1860–1869	343	23.41
1870–1879	387	26.42
1880–1889	440	30.03
1890–1900	144	9.83
Birth region
North (Groningen, Friesland, Drenthe)	424	28.94
Middle (Utrecht, Overijssel, Gelderland)	278	18.98
Coastal (Noord-Holland, Zuid-Holland)	674	46.01
South (Limburg, Noord-Brabant, Zeeland)	89	6.08
Father’s highest occupational class
Elite	27	1.84
Middle class	339	23.14
Skilled workers	491	33.52
Farmers	210	14.33
Unskilled workers	390	26.62
Unknown/no occupation	8	0.55
Religion
Catholic	356	24.30
Liberal Protestant	838	57.20
Neo-Calvinist	124	8.46
Jewish	50	3.41
No/unknown religion	97	6.62
Migrant to a Dutch city during development
Did not migrate to Amsterdam or Rotterdam	1,208	82.46
Migrated to Amsterdam or Rotterdam	257	17.54
Number of living siblings at age ten
1–2 siblings	131	8.94
3–5 siblings	618	42.18
6–8 siblings	538	36.72
9 or more siblings	178	12.15
Municipality quintile (based on 1889 census)
First	283	19.32
Second	256	17.47
Third	315	21.50
Fourth	270	18.43
Fifth	341	23.28
Infant mortality rate per municipality
First	265	18.09
Second	282	19.25
Third	306	20.89
Fourth	327	22.32
Fifth	278	18.98
Unknown	7	0.48

Next, height’s relationship to HISCAM score is depicted in Figure 4. Those in the shortest two quintiles had the lowest median HISCAM scores (both 52.5). RPs in the fourth height quintile had a median HISCAM score of 53.4. Those in the third and fifth quintiles both had median HISCAM scores of 53.9.

Figure 4. Maximum HISCAM score, by height z-score.

Figure 5 illustrates height’s relationship to intergenerational mobility. Those in the shortest three quintiles had the lowest median intergenerational mobility score, relative to their fathers, all with scores of 1.1 points. RPs in the fourth quintile had a median increase of 1.9 points relative to their fathers, while those in the fifth quintile had a median increase of 1.4 points.

Figure 5. Intergenerational change in HISCAM score, by height quintile.

3.2. Height’s relationship to HISCAM score

Table 2 presents the results of height’s relationship to HISCAM score. One standard deviation increase in height was associated with a 0.636 increase in HISCAM score (95% CI: 0.134–1.138) when adjusted for covariates.

Table 2. Height’s relationship to HISCAM score, OLS and IV, n = 1,465.

	OLS						IV
	β	p-val.	95% CI	β	p-val.	95% CI	β	p-val.	95% CI	β	p-val.	95% CI
Height z-score	0.851	0.000	0.328–1.374	0.636	0.013	0.134–1.138	1.138	0.001	0.720–2.862	1.370	0.011	0.310–2.429
Birth cohort
1850–1859				−0.404	0.674	−2.286–1.478				−0.359	0.707	−2.225–1.508
1860–1869				0.974	0.172	−0.424–2.373				0.987	0.163	−0.400–2.373
1870–1879				Ref.	Ref.	Ref.				Ref.	Ref.	Ref.
1880–1889				0.897	0.193	−0.453–2.247				0.832	0.224	−0.509–2.173
1890–1900				2.546	0.010	0.610–4.481				2.487	0.011	0.567–4.407
Birth region
North				0.650	0.490	−1.196–2.495				0.545	0.560	−1.290–2.380
Middle				−0.709	0.380	−2.295–0.876				−0.730	0.363	−2.302–0.842
Coastal				Ref.	Ref.	Ref.				Ref.	Ref.	Ref.
South				−1.328	0.276	−3.720–1.064				−1.089	0.372	−3.480–1.302
Parental HISCLASS
Elite				19.906	0.000	16.153–23.659				19.520	0.000	15.767–23.273
Middle class				6.040	0.000	4.597–7.484				5.861	0.000	4.412–7.310
Skilled workers				3.518	0.000	2.184–4.852				3.340	0.000	1.998–4.682
Farmer				2.430	0.004	0.785–4.075				2.151	0.012	0.482–3.820
Unskilled workers				Ref.	Ref.	Ref.				Ref.	Ref.	Ref.
Unknown/no occupation				3.219	0.349	−3.517–9.955				3.095	0.364	−3.585–9.775
Religion
Catholic				0.124	0.854	−1.200–1.449				0.109	0.871	−1.205–1.422
Liberal Protestant				Ref.	Ref.	Ref.				Ref.	Ref.	Ref.
Neo-Calvinist				−0.305	0.757	−2.235–1.626				−0.280	0.774	−2.195–1.634
Jewish				6.974	0.000	4.094–9.854				7.525	0.000	4.584–10.465
No religion				−0.524	0.634	−2.683–1.635				−0.515	0.637	−2.655–1.626
Migrated to a Dutch city during development?
No				Ref.	Ref.	Ref.				Ref.	Ref.	Ref.
Yes				2.472	0.000	1.140–3.805				2.388	0.000	1.063–3.714
Number of living siblings at age ten
1–2 siblings				1.700	0.066	−0.115–3.516				1.590	0.084	−0.215–3.396
3–5 siblings				Ref.	Ref.	Ref.				Ref.	Ref.	Ref.
6–8 siblings				−0.957	0.091	−2.066–0.153				−0.916	0.103	−2.018–0.185
9 or more siblings				0.193	0.816	−1.432–1.818				0.240	0.770	−1.372–1.852
Population size quintile
First				0.454	0.635	−1.422–2.330				0.260	0.786	−1.616–2.137
Second				1.043	0.233	−0.673–2.759				0.913	0.295	−0.796–2.622
Third				Ref.	Ref.	Ref.				Ref.	Ref.	Ref.
Fourth				2.021	0.020	0.320–3.722				1.905	0.027	0.212–3.598
Fifth				1.060	0.322	−1.040–3.160				1.035	0.330	−1.047–3.117
Infant mortality rate quintile
First				−0.461	0.615	−2.258–1.336				−0.303	0.741	−2.096–1.490
Second				0.550	0.525	−1.146–2.246				0.622	0.469	−1.061–2.306
Third				Ref.	Ref.	Ref.				Ref.	Ref.	Ref.
Fourth				−0.213	0.792	−1.800–1.373				−0.153	0.849	−1.728–1.422
Fifth				−0.405	0.650	−2.160–1.349				−0.443	0.618	−2.183–1.298
Unknown				0.418	0.910	−6.870–7.706				0.861	0.816	−6.386–8.108
Cragg-Donald F-statistic	n/a			n/a			46.137			40.560
Adjusted R-squared	0.007			0.140			0.114			0.139

Based on the results of the adjusted IV regression, one standard deviation increase in height was associated with a 1.370 increase in HISCAM score (95% CI: 0.310–2.429). The F-statistics, 46.137 in the unadjusted model and 40.560 in the adjusted model, indicated that the exogenous variable, brother’s height, was a suitable instrument.

3.3. Height’s relationship to intergenerational mobility

Next, height’s relationship to the intergenerational change in HISCAM score was examined in Table 3. In the OLS results, height’s relationships to intergenerational occupational mobility were not significant.

Table 3. Height’s relationship to intergenerational mobility, OLS and IV, n = 1,319.

	OLS						IV
	β	p-val.	95% CI	β	p-val.	95% CI	β	p-val.	95% CI	β	p-val.	95% CI
Height z-score	0.328	0.252	−0.233–0.889	0.338	0.250	−0.238–0.914	1.015	0.090	−0.158–2.189	1.127	0.075	−0.114–2.368
Birth cohort
1850–1859				0.338	0.757	−1.803–2.478				0.379	0.726	−1.742–2.499
1860–1869				0.750	0.361	−0.858–2.358				0.786	0.333	−0.807–2.379
1870–1879				Ref.	Ref.	Ref.				Ref.	Ref.	Ref.
1880–1889				1.472	0.059	−0.057–3.001				1.407	0.069	−0.110–2.923
1890–1900				3.138	0.004	0.981–5.294				3.093	0.005	0.958–5.229
Birth region
North				1.670	0.126	−0.472–3.812
Middle				−0.292	0.752	−2.108–1.524
Coastal				Ref.	Ref.	Ref.				Ref.	Ref.	Ref.
South				−0.135	0.921	−2.809–2.540
Parental HISCLASS
Elite				−3.642	0.125	−8.293–1.008				−3.907	0.097	−8.526–0.711
Middle class				−2.585	0.002	−4.210 – −0.960				−2.757	0.001	−4.383–1.131
Skilled workers				0.221	0.772	−1.280–1.722				0.042	0.956	−1.464–1.549
Farmer				−0.652	0.491	−2.509–1.206				−0.958	0.320	−2.846–0.930
Unskilled workers				Ref.	Ref.	Ref.				Ref.	Ref.	Ref.
Unknown/no occupation				−13.573	0.189	−33.828–6.682				−12.976	0.205	−33.046–7.094
Religion
Catholic				−0.100	0.896	−1.609–1.408				−0.123	0.872	−1.617–1.371
Liberal Protestant				Ref.	Ref.	Ref.				Ref.	Ref.	Ref.
Neo-Calvinist				−0.309	0.785	−2.519–1.904				−0.277	0.804	−2.470–1.913
Jewish				4.390	0.011	1.011–7.769				4.955	0.005	1.518–8.392
No religion				−1.169	0.359	−3.666–1.329				−1.243	0.325	−3.718–1.231
Migrated to a Dutch city during development?
No				Ref.	Ref.	Ref.				Ref.	Ref.	Ref.
Yes				0.962	0.222	−0.581–2.505				0.860	0.272	−0.675–2.394
Number of living siblings at age ten
1–2 siblings				0.856	0.414	−1.198–2.910				0.750	0.471	−1.288–2.789
3–5 siblings				Ref.	Ref.	Ref.				Ref.	Ref.	Ref.
6–8 siblings				−1.395	0.030	−2.656–0.133				−1.369	0.032	−2.618 – −0.120
9 or more siblings				−0.675	0.473	−2.522–1.170				−0.603	0.518	−2.434–1.227
Population size quintile
First				1.405	0.199	−0.741–3.551				1.178	0.283	−0.970–3.326
Second				0.902	0.368	−1.063–2.868				0.739	0.459	−1.220–2.698
Third				Ref.	Ref.	Ref.				Ref.	Ref.	Ref.
Fourth				2.048	0.038	0.117–3.980				1.907	0.052	−0.015–3.829
Fifth				−0.398	0.745	−2.797–2.002				−0.455	0.708	−2.831–1.922
Infant mortality rate quintile
First				−1.644	0.114	−3.680–0.393				−1.482	0.152	−3.511–0.547
Second				−0.360	0.713	−2.277–1.557				−0.267	0.783	−2.170–1.635
Third				Ref.	Ref.	Ref.				Ref.	Ref.	Ref.
Fourth				0.059	0.948	−1.719–1.836				0.147	0.870	−1.617–1911
Fifth				−0.493	0.630	−2.501–1.515				−0.508	0.616	−2.497–1.480
Unknown				0.923	0.818	−6.946–8.792				1.417	0.723	−6.404–9.237
Cragg-Donald F-statistic	n/a			n/a			39.069			34.845
Adjusted R-squared	0.000			0.015			0.000			0.035

Based on the results of the adjusted IV regression, one standard deviation increase in height was associated with a 1.127 increase in HISCAM score (95% CI: −0.114–2.368) when adjusting for covariates. The Cragg-Donald F-statistic were of 39.069 in the unadjusted model, and 34.845 in the adjusted model, again indicating that brother’s height was a suitable instrument.

4. Discussion

In this study, height’s relationships to occupational status and intergenerational mobility were examined. This was done with IV analyses, and by using the average height of an RP’s brothers as an instrument. While a number of other studies have examined height’s relationship to occupational outcomes with IV analyses, all did so with modern samples. By using a pre-twentieth century sample, this study helped to more fully understanding height’s relationship to labor market outcomes in historical contexts.

This study provided evidence that height was perhaps causally related to labor market outcomes among a sample of Dutch men born in the late nineteenth century. Based on the results of the IV analyses, height was positively related to both occupational status, as measured by HISCAM score, and intergenerational occupational mobility, as measured by the change in HISCAM score between RPs and their fathers. It is worth bearing in mind that the sizes of these relationships were very small: one standard deviation increase in height was associated with small increases in HISCAM scores and intergenerational mobility scores. Height alone does not seem to have played a sizeable role in determining labor market outcomes, but it does appear to have given taller men an edge over shorter peers. Height may have been one small piece of the puzzle that explained labor market outcomes in this study’s research period.

Even so, this study’s findings were somewhat surprising: in several studies using IV analysis, height's relationship to wages was insignificant (e.g. Böckerman et al., 2017; Wang et al., 2020). Why might we have found significant relationships between height and labor market outcomes, when others did not? First and perhaps obviously, we examined occupational status and intergenerational mobility, not wages. While both are widely-used indicators of labor market outcomes more broadly, height may differently influence occupational status and intergenerational mobility on the one hand, and wages on the other. For example, several studies have found that the height premium is explained by occupational sorting (e.g. Case et al., 2009; Cinnirella & Winter, 2009). That is, taller individuals, particularly men, are more likely to be selected into higher-earning occupations. Once in those occupations, height often does not play a determining role in wages. Perhaps height acted as a signal during the hiring process: taller job applicants were deemed to be more competent, or were otherwise considered to be the most suitable for jobs. However, in terms of actual job performance, on which salary would be based, height was not as significant a factor. It may be that by focusing on occupational status, versus wages, we have found significant, albeit very small, relationships.

Our choice of instrument may also explain this discrepancy. We are isolating the variation in an RP’s height as captured in his brothers’ average height, which includes shared genetics among same-sex siblings, and shared conditions in the household. Using this instrument, some genetic and environmental exposures were not captured. As mentioned, a number of other studies used genetic information, so that they are capturing the share of variation in height related to study subjects’ genes, and entirely excluding environmental factors. Indeed, studies that use non-genetic information as instruments, such as wages and parental education tend to find larger effects in IV analyses than in OLS analyses (e.g. Schultz, 2002, 2003).

It is also possible that brothers’ average height violated the exclusion restriction, in that it was not sufficiently uncorrelated with occupational outcomes. Using family information as instruments has been criticized for this reason (e.g. Trostel et al., 2002). Perhaps brothers’ average height had a direct effect on RPs’ occupational status and intergenerational mobility. However, Hoogerheide et al. (2012) argued that family information may make suitable instruments: while using family information as instruments yielded slightly more biased estimates, it did not result in significantly different conclusions. Moreover, the present study found low correlations between brother’s height and occupational status and intergenerational mobility. This indicated that brother’s height was a suitable instrument. Nonetheless, the results of this study should be interpreted with caution, as it is possible that causal relationships were not ultimately identified.

Labor market conditions may also explain why the present study found significant relationships between height and labor market outcomes. As mentioned, height is associated with a number of beneficial characteristics in adulthood. It may be that one (or more) of these characteristics was more strongly valued in the nineteenth and early twentieth century Netherlands than in modern, higher-resource settings. Perhaps employers in agriculturally-oriented and/or lower-resource labor markets, where a greater share of jobs are manual, would value height more than modern, high-resource settings. This may be because taller height is associated with greater strength (Thomas & Strauss, 1997). This may also help to explain Schultz’s (2002) finding that height’s relationship to wages was stronger in the United States than in Ghana or Brazil. Similarly, studies conducted in Europe tended to find small and/or insignificant effects (e.g. Böckerman & Vainiomäki, 2013; Böckerman et al., 2017; Heineck, 2005; Tyrrell et al., 2016).

However, contemporary lower-resource settings are not perfect analogues of historical ones. Bleakley et al. (2014) argued that, in historical, manual labor markets, tallness was associated with an increased likelihood of having a manual occupation, and therefore much less strongly associated with wages. The authors provided evidence of this by examining American men born between 1810 and 1990. Bleakley et al. (2014) also noted that the absence of this tallness penalty in studies set in contemporary, lower-resource contexts is ‘a puzzle’ (p. 30). It is also somewhat of a puzzle why we in the present study found no evidence of this penalty. It may be because this study’s sample was born too late to experience it. The majority of men in this study’s sample, born between 1850 and 1900, would have entered the labor force in the late nineteenth or early twentieth century. This may be after the labor market began valuing ‘brain’ over ‘brawn’ (Bleakley et al., 2014). Perhaps if earlier cohorts were examined, a weaker relationship would have been found between height and occupational outcomes.

It is also possible that height was particularly useful as an indicator of intelligence in this study’s research period. During this time, academic credentials gradually supplanted paternal occupation as the main determinant of labor market outcomes (Schulz, 2013). In the intervening period, it may have been that employers sometimes had neither educational credentials nor family ties on which to base their hiring decisions. Instead, physical characteristics, visible to employers when meeting job-seekers, could have played a greater role. It may be the particularities of the nineteenth century Dutch labor market that underpinned this study’s findings.

4.1. Limitations

The present study had several limitations. First, interpreting HISCAM scores was not straightforward. In particular, some of the increases in HISCAM scores were perhaps not reflective of large changes in intergenerational mobility. If a civil servant’s son became a doctor, he would have had an intergenerational mobility score of 30. If the son of a day laborer became simply a laborer, he would have had an intergenerational mobility score of 7. We assessed whether this was an issue in robustness checks, by including individuals with a HISCAM of 70 or under, and using a binary indicator of intergenerational mobility. While we observed similar results in robustness checks relative to this study’s main results, this potential issue with HISCAM scores should be kept in mind when interpreting this study’s results.

Further, only RPs with brothers in the Male Kin Height database were included in this study’s sample. It is likely that, for any brother to be recorded in the database, his corresponding RP actually had multiple brothers. It may be that RPs with brothers in this study’s dataset were systematically different than those without brothers. For example, perhaps RPs with fewer brothers had more resources in early-life, and were ultimately taller (e.g. Quanjer & Kok, 2019). In a sensitivity analysis, we performed OLS analyses including RPs without brothers in the Male Kin Height database, and found similar results to those of this study’s main analyses. We were therefore again moderately confident that our results were robust.

Finally, terminal adult height was likely not used in this study. During this study’s research period, Dutch men were growing into their twenties (Thompson et al., 2020). This meant that, for many RPs, height measurements were reflective of late adolescent, not adult, height. Given that variations in height during development tend to be larger than variations in adult height, it may be that height’s relationships to labor market outcomes have been overestimated in this study. This is also complicated by the fact that RPs would have often begun working before they stopped growing (Bras et al., 2010). This may have resulted in this study overstating the size of height’s relationship to labor market outcomes.

5. Conclusions

In this study, whether height was related to labor market outcomes was examined with instrumental variable analyses among a sample of Dutch men born in the late nineteenth century. The average height of an RP’s brothers was used as the instrumental variable. This study found positive relationships between height and labor market outcomes. As Dutch men were growing taller and had greater abilities to choose their occupations, it appears that tallness was associated with a better job, and increased intergenerational occupational mobility. This study thus offered preliminary evidence that height and labor market outcomes were perhaps causally related during the late nineteenth and early twentieth centuries.

Acknowledgments

We would like to thank Jan Kok, Björn Quanjer, and the participants of the Height and Later Life Outcomes workshop for their helpful comments on earlier drafts of this paper. All remaining errors are our own. We are grateful to the Dutch Research Council (NWO) for funding the Giants of the Modern World project (project no. 360-53-190), of which this paper is a part.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by the Dutch Research Council (NWO), as part of the Giants of the Modern World project (file number: 360-53-190).

Appendix A.

First-stage regressions

Table A1. Brothers’ average height z-score’s relationship to RP’s height z-score (HISCAM analyses), n = 1,465.

	β	P-value	95% Confidence Interval
Brother’ height z-score	0.470	0.000	0.424	0.515
Birth cohort
1850–1859	0.008	0.925	−0.163	0.180
1860–1869	0.002	0.973	−0.125	0.130
1870–1879	Ref.	Ref.	Ref.	Ref.
1880–1889	0.094	0.133	−0.029	0.217
1890–1900	0.065	0.472	−0.112	0.241
Birth region
North	0.064	0.457	−0.104	0.232
Middle	0.030	0.681	−0.114	0.175
Coastal	Ref.	Ref.	Ref.	Ref.
South	−0.168	0.132	−.386	0.050
Father’s maximum HISCLASS score
Elite	0.257	0.141	−0.085	0.599
Lower Middle Class	0.157	0.019	0.026	0.288
Skilled Worker	0.184	0.003	0.063	0.306
Farmer	0.205	0.007	0.055	0.354
Unskilled Worker	Ref.	Ref.	Ref.	Ref.
Unknown	0.129	0.680	−0.485	0.743
Religion
Catholic	0.001	0.981	−0.119	0.122
Liberal Protestant	Ref.	Ref.	Ref.	Ref.
Orthodox Protestant	−0.079	0.381	−0.255	0.097
Jewish	−0.494	0.000	−0.756	−0.233
No religion	0.021	0.835	−0.018	0.218
Migrated to a Dutch city during development?
No	Ref.	Ref.	Ref.	Ref.
Yes	1.790	0.073	−0.010	0.232
Number of living siblings at age ten
1–2 siblings	0.106	0.217	−0.062	0.274
3–5 siblings	−0.013	0.804	−0.114	0.088
6–8 siblings	Ref.	Ref.	Ref.	Ref.
9 or more siblings	−0.005	0.948	−0.154	0.144
Population size quintile
First	0.251	0.004	0.080	0.421
Second	0.111	0.164	−0.045	0.267
Third	Ref.	Ref.	Ref.	Ref.
Fourth	0.091	0.251	−0.064	0.246
Fifth	0.026	0.788	−0.165	0.218
Infant mortality rate quintile
First	−0.137	0.101	−0.300	0.027
Second	−0.088	0.266	−0.242	0.067
Third	Ref.	Ref.	Ref.	Ref.
Fourth	−0.035	0.635	−0.180	0.110
Fifth	0.092	0.259	−0.068	0.252
Unknown	−0.186	0.583	−0.851	0.478
Adjusted R-squared	0.258

Table A2. Brothers’ average height z-score’s relationship to RP’s height z-score (Intergenerational mobility analyses), n = 1,319.

	β	P-value	95% Confidence Interval
Brother’ height z-score	0.454	0.000	0.406	0.502
Birth cohort
1850–1859	0.026	0.784	−0.155	0.206
1860–1869	−0.028	0.681	−0.164	0.107
1870–1879	Ref.	Ref.	Ref.	Ref.
1880–1889	0.082	0.210	−0.046	0.211
1890–1900	0.039	0.673	−0.143	0.221
Birth region
North	0.038	0.681	−0.143	0.219
Middle	−0.000	0.995	−0.154	0.153
Coastal	Ref.	Ref.	Ref.	Ref.
South	−0.152	0.188	−0.377	0.074
Father’s maximum HISCLASS score
Elite	0.121	0.547	−0.272	0.513
Lower Middle Class	0.145	0.039	0.007	0.281
Skilled Worker	0.164	0.011	0.038	0.290
Farmer	0.206	0.010	0.049	0.363
Unskilled Worker	Ref.	Ref.	Ref.	Ref.
Unknown	−1.122	0.197	−2.830	0.585
Religion
Catholic	−0.008	0.898	−0.136	0.119
Liberal Protestant	Ref.	Ref.	Ref.	Ref.
Orthodox Protestant	−0.113	0.235	−0.300	0.074
Jewish	−.418	0.004	−.703	−0.134
No religion	0.121	0.260	−0.090	0.331
Migrated to a Dutch city during development?
No	Ref.	Ref.	Ref.	Ref.
Yes	1.870	0.061	−0.006	0.254
Number of living siblings at age ten
1–2 siblings	0.088	0.326	−0.088	0.264
3–5 siblings	−.020	0.717	−0.126	0.087
6–8 siblings	Ref.	Ref.	Ref.	Ref.
9 or more siblings	−0.021	0.789	−0.179	0.136
Population size quintile
First	0.272	0.003	0.091	0.452
Second	0.149	0.078	−0.017	0.314
Third	Ref.	Ref.	Ref.	Ref.
Fourth	0.114	0.169	−0.049	0.277
Fifth	0.043	0.674	−0.159	0.246
Infant mortality rate quintile
First	−0.123	0.160	−0.295	0.049
Second	−0.100	0.224	−0.262	0.061
Third	Ref.	Ref.	Ref.	Ref.
Fourth	−0.052	0.493	−0.202	0.097
Fifth	0.082	0.341	−0.087	0.252
Unknown	−0.209	0.536	−0.874	0.455
Adjusted R-squared	0.248

Appendix B.

Robustness checks

Table B1. Height z-score’s relationship to maximum HISCAM score, excluding HISCAM scores ≥70, IV regression, n = 1,329.

	β	P-value	95% Confidence Interval
Height z-score	1.220	0.073	−0.044	1.888
Birth cohort
1850–1859	−0.631	0.280	−1.775	0.513
1860–1869	0.142	0.742	−0.707	0.991
1870–1879	Ref.	Ref.	Ref.	Ref.
1880–1889	0.554	0.185	−0.266	1.374
1890–1900	1.024	0.094	−0.174	2.222
Birth region
North	0.711	0.220	−0.425	1.847
Middle	−0.246	0.616	−1.207	0.715
Coastal	Ref.	Ref.	Ref.	Ref.
South	−2.276	0.002	−3.741	−0.812
Father’s maximum HISCLASS score
Elite	5.254	0.001	2.178	8.330
Lower Middle Class	4.023	0.000	3.136	4.910
Skilled Worker	1.766	0.000	0.952	2.580
Farmer	0.846	0.101	−0.165	1.856
Unskilled Worker	Ref.	Ref.	Ref.	Ref
Unknown	0.332	0.876	−3.854	4.518
Religion
Catholic	0.307	0.458	−0.504	1.117
Liberal Protestant	Ref.	Ref.	Ref.	Ref.
Orthodox Protestant	0.386	0.515	−0.777	1.549
Jewish	6.445	0.000	4.573	8.318
No religion	0.287	0.667	−1.018	1.591
Migrated to a Dutch city during development?
No	Ref.	Ref.	Ref.	Ref.
Yes	0.945	0.028	0.102	1.787
Number of living siblings at age ten
1–2 siblings	0.712	0.228	−0.445	1.870
3–5 siblings	0.282	0.413	−0.393	0.957
6–8 siblings	Ref.	Ref.	Ref.	Ref.
9 or more siblings	1.000	0.047	0.014	1.986
Population size quintile
First	0.499	0.394	−0.649	1.648
Second	0.115	0.830	−0.929	1.158
Third	Ref.	Ref.	Ref.	Ref.
Fourth	1.165	0.029	0.117	2.213
Fifth	0.208	0.750	−1.069	1.484
Infant mortality rate quintile
First	−1.102	0.051	−2.207	0.004
Second	−0.561	0.294	−1.607	0.486
Third	Ref.	Ref.	Ref.	Ref.
Fourth	−0.010	0.984	−0.980	0.961
Fifth	−0.474	0.386	−1.546	0.597
Unknown	−2.954	0.206	−7.532	1.624
Adjusted R-squared	0.151

Table B2. Height z-score’s relationship to experiencing intergenerational occupational mobility (=1), IV probit regression, n = 1,319.

	Coefficient	P-value	95% Confidence Interval
Height z-score	0.198	0.048	.0014305	.3936939
Birth cohort
1850–1859	−0.002	0.993	−0.363	0.359
1860–1869	−0.050	0.702	−0.306	0.206
1870–1879	Ref.	Ref.	Ref.	Ref.
1880–1889	0.042	0.737	−0.203	0.287
1890–1900	0.235	0.185	−0.113	0.582
Birth region
North	0.201	0.251	−0.142	0.545
Middle	−0.160	0.278	−0.450	0.130
Coastal	Ref.	Ref.	Ref.	Ref.
South	−0.469	0.036	−0.907	−0.030
Father’s maximum HISCLASS score
Elite	−1.110	0.002	−1.826	−0.393
Lower Middle Class	−1.203	0.000	−1.507	−0.900
Skilled Worker	−1.020	0.000	−1.307	−0.734
Farmer	−0.739	0.000	−1.091	−0.388
Unskilled Worker	Ref.	Ref.	Ref.	Ref.
Unknown	−0.985	0.159	−2.358	0.387
Religion
Catholic	0.049	0.690	−0.193	0.291
Liberal Protestant	Ref.	Ref.	Ref.	Ref.
Orthodox Protestant	0.192	0.318	−0.185	0.569
Jewish	0.755	0.007	0.205	1.306
No religion	−0.182	0.348	−0.563	0.198
Migrated to a Dutch city during development?
No	Ref.	Ref.	Ref.	Ref.
Yes	−0.004	0.971	−0.2429	0.233
Number of living siblings at age ten
1–2 siblings	0.110	0.507	−0.214	0.434
3–5 siblings	0.234	0.022	0.034	0.435
6–8 siblings	Ref.	Ref.	Ref.	Ref.
9 or more siblings	0.229	0.148	−0.082	0.539
Population size quintile
First	0.376	0.037	0.022	0.729
Second	−0.068	0.670	−0.379	0.244
Third	Ref.	Ref.	Ref.	Ref.
Fourth	0.234	0.144	−0.080	0.548
Fifth	−0.309	0.104	−0.681	0.063
Infant mortality rate quintile
First	−0.282	0.102	−0.620	0.056
Second	−0.171	0.286	−0.486	0.143
Third	Ref.	Ref.	Ref.	Ref.
Fourth	0.179	0.217	−0.105	0.463
Fifth	0.080	0.625	−0.242	0.403
Unknown	0
N	1,319

Table B3. Height z-score’s relationship to HISCAM score, using father’s height as an IV, IV regression, n = 642.

	β	P-value	95% Confidence Interval
Height z-score	1.105	0.306	−1.011	3.222
Birth cohort
1850–1859	3.160	0.253	−2.254	8.573
1860–1869	1.497	0.356	−1.684	4.677
1870–1879	Ref.	Ref.	Ref.	Ref.
1880–1889	2.182	0.041	0.084	4.280
1890–1900	1.923	0.129	−0.562	4.409
Birth region
North	1.272	0.376	−1.543	4.087
Middle	−2.671	0.052	−5.368	0.025
Coastal	Ref.	Ref.	Ref.	Ref.
South	−2.050	0.301	−5.936	1.835
Father’s maximum HISCLASS score
Elite	15.470	0.000	9.705	21.234
Lower Middle Class	7.358	0.000	5.067	9.649
Skilled Worker	3.374	0.002	1.249	5.499
Farmer	2.110	0.109	−0.469	4.689
Unskilled Worker	Ref.	Ref.	Ref.	Ref.
Unknown	−1.165	0.867	−14.749	12.420
Religion
Catholic	0.451	0.663	−1.576	2.477
Liberal Protestant	Ref.	Ref.	Ref.	Ref.
Orthodox Protestant	−0.529	0.738	−3.634	2.576
Jewish	8.917	0.000	4.844	12.990
No religion	−0.707	0.724	−4.625	3.212
Migrated to a Dutch city during development?
No	3.653	0.000	1.609	5.697
Yes	Ref.	Ref.	Ref.	Ref.
Number of living siblings at age ten
1–2 siblings	2.768	0.030	0.272	5.263
3–5 siblings	1.683	0.069	−0.132	3.498
6–8 siblings	Ref.	Ref.	Ref.	Ref.
9 or more siblings	1.048	0.456	−1.709	3.806
Population size quintile
First	2.312	0.119	−0.597	5.221
Second	1.628	0.252	−1.156	4.412
Third	Ref.	Ref.	Ref.	Ref.
Fourth	3.426	0.014	0.694	6.158
Fifth	2.115	0.194	−1.076	5.306
Infant mortality rate quintile
First	−0.283	0.843	−3.094	2.527
Second	−2.319	0.084	−4.950	0.312
Third	Ref.	Ref.	Ref.	Ref.
Fourth	0.085	0.946	−2.370	2.539
Fifth	0.175	0.909	−2.810	3.159
Unknown	−6.633	0.502	−26.000	12.733
Cragg-Donald F-statistic	29.033
Adjusted R-squared	0.212

Table B4. Height z-score’s relationship to HISCAM score, using father’s height as an IV, IV regression, n = 639.

	β	P-value	95% Confidence Interval
Height z-score	−0.286	0.819	−2.736	2.165
Birth cohort
1850–1859	2.687	0.414	−3.764	9.138
1860–1869	1.631	0.380	−2.009	5.271
1870–1879	Ref.	Ref.	Ref.	Ref.
1880–1889	2.254	0.058	−0.079	4.587
1890–1900	2.957	0.034	0.221	5.693
Birth region
North	3.377	0.038	0.194	6.559
Middle	−1.856	0.230	−4.886	1.173
Coastal	Ref.	Ref.	Ref.	Ref.
South	−0.096	0.967	−4.613	4.421
Father’s maximum HISCLASS score
Elite	−11.487	0.002	−18.699	−4.275
Lower Middle Class	−0.111	0.931	−2.631	2.409
Skilled Worker	0.348	0.771	−1.996	2.691
Farmer	−9.915	0.535	−3.805	1.975
Unskilled Worker	Ref.	Ref.	Ref.	Ref.
Unknown	0
Religion
Catholic	−0.002	0.998	−2.316	2.311
Liberal Protestant	Ref.	Ref.	Ref.	Ref.
Orthodox Protestant	−0.981	0.572	−4.383	2.421
Jewish	5.265	0.022	0.761	9.769
No religion	−2.383	0.288	−6.778	2.012
Migrated to a Dutch city during development?
No	Ref.	Ref.	Ref.	Ref.
Yes	2.112	0.072	−0.193	4.417
Number of living siblings at age ten
1–2 siblings	2.729	0.051	−0.018	5.477
3–5 siblings	3.256	0.002	1.231	5.281
6–8 siblings	Ref.	Ref.	Ref.	Ref.
9 or more siblings	1.564	0.316	−1.490	4.617
Population size quintile
First	2.743	0.101	−0.539	6.026
Second	1.008	0.528	−2.124779	4.141
Third	Ref.	Ref.	Ref.	Ref.
Fourth	2.970	0.056	−0.078	6.019
Fifth	0.350	0.847	−3.211	3.911
Infant mortality rate quintile
First	−1.774	0.281	−4.998	1.450
Second	−3.404	0.022	−6.313	−0.494
Third	Ref.	Ref.	Ref.	Ref.
Fourth	−0.708	0.684	−4.125	2.708
Fifth	−6.789	0.522	−27.588	14.009
Cragg-Donald F-statistic	27.08
Adjusted R-squared	0.082

Table B5. Height z-score’s relationship to maximum HISCAM score, RPs without siblings in database only, OLS, n = 1,572.

	β	P-value	95% Confidence Interval
Height z-score	1.094	0.000	0.619	1.570
Birth cohort
1850–1859	−0.038	0.975	−2.357	2.282
1860–1869	−0.873	0.250	−2.360	0.615
1870–1879	Ref.	Ref.	Ref.	Ref.
1880–1889	1.072	0.119	−0.275	2.418
1890–1900	0.878	0.229	−0.554	2.310
Birth region
North	−0.956	0.355	−2.983	1.071
Middle	−1.712	0.064	−3.525	0.101
Coastal	Ref.	Ref.	Ref.	Ref.
South	−1.881	0.056	−3.811	0.050
Father’s maximum HISCLASS score
Elite	14.996	0.000	12.038	17.955
Lower Middle Class	5.281	0.000	3.747	6.815
Skilled Worker	3.211	0.000	1.870	4.551
Farmer	0.826	0.262	−0.618	2.269
Unskilled Worker	Ref.	Ref.	Ref.	Ref.
Unknown	1.619	0.591	−4.294	7.532
Religion
Catholic	−0.060	0.931	−1.437	1.316
Liberal Protestant	Ref.	Ref.	Ref.	Ref.
Orthodox Protestant	−0.109	0.907	−1.937	1.719
Jewish	11.577	0.000	7.952	15.201
No religion	−0.367	0.691	−2.174	1.441
Migrated to a Dutch city during development?
No	Ref.	Ref.	Ref.	Ref.
Yes	2.047	0.011	0.466	3.629
Number of living siblings at age ten
Only child	−1.186	0.413	−4.029	1.658
1–2 siblings	1.465	0.038	0.079	2.852
3–5 siblings	0.463	0.432	−0.692	1.619
6–8 siblings	Ref.	Ref.	Ref.	Ref.
9 or more siblings	−0.403	0.650	−2.142	1.337
Population size quintile
First	−2.340	0.004	−3.914	−0.766
Second	−1.748	0.021	−3.232	−0.264
Third	Ref.	Ref.	Ref.	Ref.
Fourth	−1.057	0.188	−2.632	0.518
Fifth	−1.055	0.378	−3.402	1.293
Infant mortality rate quintile
First	−0.786	0.337	−2.393	0.820
Second	−0.969	0.231	−2.553	0.616
Third	Ref.	Ref.	Ref.	Ref.
Fourth	−0.345	0.683	−2.004	1.313
Fifth	−0.821	0.334	−2.486	0.844
Unknown	−0.591	0.929	−13.657	12.474

Table B6. Height z-score’s relationship to intergenerational mobility score, full sample, OLS, n=1,422.

	β	P-value	95% Confidence Interval
Height z-score	0.876	0.011	0.200	1.551
Birth cohort
1850–1859	−0.272	0.875	−3.669	3.125
1860–1869	−0.409	0.711	−2.570	1.753
1870–1879	Ref.	Ref.	Ref.	Ref.
1880–1889	−0.472	0.630	−2.396	1.451
1890–1900	0.269	0.795	−1.758	2.296
Birth region
North	2.000	0.173	−0.875	4.875
Middle	1.402	0.289	−1.189	3.993
Coastal	Ref.	Ref.	Ref.	Ref.
South	2.169	0.125	−0.605	4.942
Father’s maximum HISCLASS score
Elite	−8.720	0.000	−13.254	−4.186
Lower Middle Class	−3.319	0.003	−5.515	−1.124
Skilled Worker	1.454	0.141	−0.481	3.389
Farmer	−1.365	0.196	−3.434	−.704
Unskilled Worker	Ref.	Ref.	Ref.	Ref.
Unknown	−8.565	0.341	−26.222	9.092
Religion
Catholic	−0.598	0.552	−2.568	1.373
Liberal Protestant	Ref.	Ref.	Ref.	Ref.
Orthodox Protestant	−0.670	0.612	−3.261	1.921
Jewish	14.526	0.000	9.192	19.859
No religion	−1.074	0.418	−3.676	1.527
Migrated to a Dutch city during development?
No	Ref.	Ref.	Ref.	Ref.
Yes	2.659	0.023	0.372	4.946
Number of living siblings at age ten
Only child	0.341	0.879	−4.740	4.038
1–2 siblings	2.983	0.003	0.997	4.969
3–5 siblings	0.577	0.491	−1.067	2.221
6–8 siblings	Ref.	Ref.	Ref.	Ref.
9 or more siblings	−2.012	0.109	−4.476	0.452
Population size quintile
First	−1.799	0.116	−4.042	0.443
Second	−1.779	0.098	−3.890	0.331
Third	Ref.	Ref.	Ref.	Ref.
Fourth	−0.749	0.515	−3.006	1.508
Fifth	−11.529	0.000	−14.901	−8.156
Infant mortality rate quintile
First	3.592	0.003	1.262	5.923
Second	2.142	0.069	−.165	4.448
Third	Ref.	Ref.	Ref.	Ref.
Fourth	5.933	0.000	3.519	8.347
Fifth	3.118	0.011	0.716	5.521
Unknown	−3.995	0.754	−28.988	20.999