leaders and laggards in life expectancy among eu ropean ...demography doi the authors this is an...

Demography (2021) 58(1):111–135DOI 10.1215/00703370-8938107 © 2021 The AuthorsThis is an open ac cess ar ti cle dis trib uted un der the terms of a Creative Commons license (CC BY-NC-ND 4.0).

Published online: 22 January 2021

Leaders and Laggards in Life Expectancy Among Eu ro pean Scholars From the Sixteenth to the Early Twentieth Century

Robert Stelter, Da vid de la Croix, and Mikko Myrskylä

ABSTRACT When did mor tal ity first start to de cline, and among whom? We build a large, new data set with more than 30,000 schol ars cov er ing the six teenth to the early twen ti eth cen tury to an a lyze the tim ing of the mor tal ity de cline and the het ero ge ne ity in life ex pec tancy gains among schol ars in the Holy Ro man Em pire. The large sam ple size, well-de fined en try into the risk group, and het ero ge ne ity in so cial sta tus are among the key ad van tages of the new da ta base. After re cov er ing from a se vere mor tal ity cri sis in the sev en teenth cen tu ry, life ex pec tancy among schol ars started to in crease as early as in the eigh teenth cen tu ry, well be fore the Industrial Revolution. Our find ing that mem bers of sci en tific acad e mies—an elite group among schol ars—were the first to ex pe ri ence mor tal ity im prove ments sug gests that 300 years ago, in di vid u als with higher so cial sta tus al ready enjoyed lower mor tal i ty. We also show, how ev er, that the on set of mor tal ity im prove ments among schol ars in med i cine was delayed, pos si bly be cause these schol ars were ex posed to path o gens and did not have germ the ory knowl edge that might have protected them. The dis ad van tage among med i cal pro fes sion als de creased to ward the end of the nineteenth cen tu ry. Our re sults pro vide a new per spec tive on the his tor i cal tim ing of mor tal ity im prove ments, and the da ta base ac com pa ny ing our study fa cil i tates rep li ca tion and ex ten sions.

KEYWORDS Mortality dy nam ics • Differential mor tal i ty • Holy Ro man Em pire • Thirty Years’ War

Introduction

When the first mor tal ity im prove ments oc curred and who benefited first from these gains are among the key ques tions that arise in dis cus sions of the first de mo graphic tran si tion. Understanding where and among whom mor tal ity prog ress started is im por-tant for un der stand ing the long-run dy nam ics of hu man well-be ing.

In this ar ti cle, we fo cus on the Eu ro pean sci en tific elite: schol ars ac tive at uni ver-si ties or acad e mies of sci ences. Observing each schol ar’s first ap point ment or nom i-na tion to a sci en tific in sti tu tion helps to over come com mon meth od o log i cal is sues in his tor i cal pop u la tions given that the ap point ment can be used to de fine the en try into the pop u la tion at risk. More im por tant ly, tak ing into ac count each schol ar’s sci en-

ELECTRONIC SUP PLE MEN TARY MA TE RI AL The online ver sion of this ar ti cle (https://doi.org/10.1215/00703370 -8938107) contains supplementary material.

Downloaded from http://read.dukeupress.edu/demography/article-pdf/58/1/111/850330/111stelter.pdfby EBSCO SUBS useron 15 February 2021

https://doi.org/10.1215/00703370-8938107

https://doi.org/10.1215/00703370-8938107

112 R. Stelter et al.

tific field and po ten tial mem ber ship in an acad emy of sci ences pro vi des new in sights into the role of med i cine and so cial sta tus in the pro cess of mor tal ity im prove ments. Finally, in a world where face-to-face com mu ni ca tion was es sen tial for both knowl-edge trans mis sion and en hance ment, the length of pro duc tive life among the elite was an im por tant de ter mi nant of the ex tent to which mem bers of the elite were able to in flu ence their cul tural and eco nomic en vi ron ments (de la Croix 2017; Lucas 2009).

Drawing on lo cal ev i dence and data on spe cific so cial groups, his to ri ans and de mog-ra phers have al ready shown that mor tal ity gains were made in the sev en teenth and eigh teenth cen tu ries. Hollingsworth (1977), for in stance, constructed mor tal ity ta bles for the Brit ish no bil ity sam pled from ge ne a log i cal da ta. Vandenbroucke (1985) pro-vided vi tal sta tis tics for the Knights of the Golden Fleece, an or der started in 1430 by the Dukes of Burgundy and maintained un der the Habs burg rul ers, the kings of Spain, and the Austrian em per ors. Andreev et al. (2011) com pared life ex pec tan cies of mem-bers of the Brit ish and the Rus sian acad e mies of sci ences.1 Cummins (2017) greatly ex tended de mo graphic re search on Europe’s ar is toc racy by an a lyz ing the lon gev ity of the Eu ro pean no bil ity over a long pe riod that cov ered sev eral crit i cal events, such as the Black Death and the Industrial Revolution. Longevity started ris ing as early as 1400 and con tin ued to in crease over the fifteenth cen tu ry. However, this first phase has been ob served in Ireland and the United Kingdom only and these find ings are sub ject to con sid er able un cer tain ty. Even though the to tal sam ple size is large, when stretched over sev eral cen tu ries, the un cer tainty re gard ing any spe cific time point be comes large. This phase of lon gev ity im prove ments was followed by an other af ter 1650 that has been ob served through out Europe in other stud ies as well (Hollingsworth 1977; van Poppel et al. 2016; van Poppel et al. 2013). Building a da ta base drawn from the Index Bio-bibliographicus Notorum Hominum, which con tains entries on fa mous peo-ple from about 3,000 dic tio nar ies and en cy clo pe dias, de la Croix and Licandro (2015) found no trend in adult lon gev ity among in di vid u als born be fore the sec ond half of the sev en teenth cen tu ry. Their find ings also sug gest that per ma nent im prove ments in lon gev ity pre ceded the Industrial Revolution by at least a cen tu ry. The lon gev ity of fa mous peo ple in creased steadily starting with the gen er a tions born in the 1640–1649 pe riod and grew by a to tal of roughly nine years in the fol low ing two cen tu ries.

Although the stud ies of de la Croix and Licandro (2015) and Cummins (2017) are im por tant, they are not with out weaknesses. In the pop u la tions they stud ied, who belonged to the sam ple and when peo ple en tered the pop u la tion at risk could not be pre cisely de fined. Some of the in di vid u als in these pop u la tions, such as fa mous mar tyrs, might have en tered at death; oth ers, such as art ists, may have en tered post mortem; and still oth ers, such as mem bers of royal fam i lies, en tered at birth. In this study, we pres ent data that over come such weaknesses and use these data to reanalyze the tim ing of mor tal ity im prove ments among the Eu ro pean elite. Furthermore, us ing in for ma tion about rel a tive sta tus within the elite, we in ves ti gate whether dif fer ences in so cio eco nomic po si tion were al ready influ enc ing mor tal ity when sec u lar changes

1 Edwards (2008), Winkler-Dworak (2008), and van de Kaa and de Roo (2007) also stud ied the lon gev ity of mem bers of acad e mies of sci ences but with a more re cent fo cus and much smaller sam ple sizes. Carrieri and Serraino (2005) and Hanley et al. (2006) com pared the life ex pec tan cies of popes and art ists, whereas van Poppel et al. (2013) ex am ined the lon gev ity of art ists us ing in for ma tion from the RKDartists da ta base.


113Leaders and Laggards in Life Expectancy

in mor tal ity first started, or whether this pat tern is more re cent. Finally, we ex ploit in for ma tion about the sci en tific fields in which the schol ars in our da ta base were work ing to ex am ine whether there were lead ers or lag gards by dis ci pline. A par tic u lar fo cus of our anal y sis is on med i cine, which may have had both pos i tive and neg a tive ef fects on lon gev i ty, depending on whether the ben e fits of med i cal knowl edge off set the added haz ards resulting from ex po sure to path o gens.

Our data set, which is mainly based on uni ver sity pro fes sor cat a logs and lists of mem ber ships in sci en tific acad e mies, is constructed to clearly in di cate who belonged to the pop u la tion and when each in di vid ual en tered the risk pop u la tion. An in di vid ual en tered our pop u la tion at risk as soon as he or she was appointed for the first time to any of the for mal sci en tific in sti tu tions cov ered in our da ta base. These in sti tu tions in clude all uni ver si ties and tech ni cal uni ver si ties established be fore 1800 as well as sci en tific acad e mies lo cated in the Netherlands and the ter ri to ries of the Holy Ro man Em pire (HRE) as of 1648. Although its bor ders changed over time, the HRE oc cu pied a large area of cen tral Europe from its founding in the Middle Ages un til its dis so lu tion in 1806. Furthermore, for most ac a demic in sti tu tions in these ter ri to ries, there are data sources that pro vide in for ma tion on each schol ar’s dates of ap point ment, ex it, birth, and death.

Our sam ple, which cov ers more than 30,000 schol ars born be tween the four teenth and the nineteenth cen tu ry, has sev eral ad van tages. The sam ple al lows us to con sider left-trun ca tion and right-cen sor ing. Moreover, be cause the sam ple is large and fo cused on a well-de fined pop u la tion, we can use it to make pre cise es ti ma tes at the to tal pop-u la tion level while also performing sub pop u la tion- and age group-spe cific an a ly ses.

Relying on the new data we col lect ed, we aim to make two con tri bu tions to the lit-er a ture. First, our anal y sis of schol ars’ life ex pec tancy pro vi des im por tant new in for-ma tion about the dy nam ics and the tim ing of mor tal ity changes be fore and dur ing the Industrial Revolution. Our new es ti ma tions con firm that life ex pec tancy started to im prove in the mid dle of the eigh teenth cen tu ry—and, hence, be fore in dus tri al i-za tion. Most of the de vi a tions of our find ings from existing es ti ma tes of mor tal ity dy nam ics can be explained by dif fer ences ei ther in the meth ods used or in how the role of so cial sta tus is taken into ac count. In ad di tion, our long time se ries on mor tal-ity pro vide a novel find ing on a no ta ble mor tal ity cri sis around 1620–1650, which was likely driven by the Thirty Years’ War. Studies with out a long ob ser va tion win dow such as ours, which opens be fore the Thirty Years’ War, could mis tak enly con clude that the re cov ery from the cri sis marked the start of sec u lar mor tal ity im prove ments.

Second, we shed light on mor tal ity dif fer ences be tween groups by com par ing mem-bers and non mem bers of sci en tific acad e mies as well as schol ars in the med i cal field with schol ars in other sci en tific fields. Members of sci en tific acad e mies rep re sent an elite within the elite. Although it may be as sumed that higher so cial sta tus trans lates into mor tal ity ad van tages, the ev i dence on the as so ci a tion be tween so cial sta tus and mor tal ity is mixed. Hollingsworth (1977) and Vandenbroucke (1985) found that mor-tal ity re duc tions oc curred as early as in the sev en teenth cen tury among the no bil i ty, and thus showed that lon gev ity im prove ments among the up per so cial clas ses an tic i-pated the over all rise in life ex pec tancy by at least a cen tu ry. By con trast, de la Croix and Licandro (2015) found mor tal ity re duc tions that in the sev en teenth and eigh-teenth cen tu ries took place not just in the lead ing countries but al most ev ery where in Europe and that these mor tal ity im prove ments were not dom i nated by any par tic u lar



oc cu pa tion.2 In a lit er a ture re view, Bengtsson and van Poppel (2011) con cluded that the im pact of so cial sta tus var ied across ar eas, time pe ri ods, and con texts. Mean-while, we ob serve that among mem bers of acad e mies of sci ences, mor tal ity gains ac cel er ated around the time when life ex pec tancy started to in crease sustainably. This find ing sug gests that as early as 300 years ago, schol ars with higher so cial sta tus enjoyed lower mor tal i ty.

The role of the med i cal pro fes sion in these trends is less clear. Given that the germ the ory of dis ease was not well de vel oped be fore the sec ond half of the nineteenth cen tu ry, it may be as sumed that in di vid u als work ing in the med i cal pro fes sion prior to that time re ceived lit tle pro tec tion from their med i cal knowl edge while also be ing ex posed to el e vated in fec tion risks (de la Croix and Sommacal 2009). It is, there fore, pos si ble that med i cal pro fes sion als had a net dis ad van tage. For ex am ple, to pro tect against the bu bonic plague, peo ple used beak-like masks that did lit tle more than pro-tect against the smells that were be lieved to be the main dis ease vec tor. Thus, for med-i cal pro fes sion als, the com bi na tion of in creased ex po sure to sick peo ple and the lack of med i cal knowl edge may have been life-threat en ing. Our re sults par tially sup port this rea son ing: al though we find only weak ev i dence of a sys tem atic dis ad van tage among med i cal pro fes sion als be fore sustained im prove ments in lon gev ity be gan, we show that once mor tal ity im prove ments started, med i cal pro fes sion als ex pe ri enced life ex pec tancy gains later than the rest of the sci en tific elite.

Scholars in the Holy Ro man Em pire

Universities and Scientific Academies

Our data set con tains in for ma tion on schol ars who were ac tive in the HRE. The Em pire was founded around 962 as Otto I sought to re vive the Ro man Em pire by lay ing claim to the im pe rial cult of Rome. Thus, the HRE existed long be fore the first uni ver si-ties appeared in this ar ea. Although the bor ders of the Em pire changed over its al-most 850 years of ex is tence, its elec tive mon ar chy uni fied the Ger manic pop u la tion and other peo ples over this long pe riod through a unique set of cul tural and po lit i cal ar range ments. In the nineteenth and twen ti eth cen tu ries, fol low ing the dis so lu tion of the Em pire in 1806, the ter ri tory of the Ger man state de clined sub stan tial ly, and the pop u la tions of cen tral Europe shifted. To take ad van tage of the rel a tively sta ble in sti tu-tional setup pro vided by the HRE, we fo cus on pop u la tions liv ing within the Em pire’s 1648 bor ders and in the Netherlands. As shown in gray in Figure 1, the ter ri to ries of the Em pire cor re spond to the cur rent ter ri to ries of Austria, Belgium, Germany, Lichten-stein, Luxembourg, Slovenia, and the Czech Republic as well as small parts of Croatia, France, Italy, Poland, and Switzerland. The ter ri to ries that made up the Netherlands are depicted in light gray.

Scholars might have been ac tive in uni ver si ties, acad e mies of sci ences, or courts. Because the first two types of in sti tu tions are quan ti ta tively the most im por tant, we de fine schol ars as in di vid u als who were ac tive in one of these two types of sci en tific

2 Furthermore, Bengtsson and Dribe (2011) found ev i dence of the late emer gence of a mor tal ity ad van tage in Sweden. Using data from Geneva, Schumacher and Oris (2011) documented an ad van tage among the higher clas ses that ended in the sev en teenth and eigh teenth cen tu ry.



in sti tu tions, and we re strict our sam ple to schol ars who were born be fore 1900. This pre cise de mar ca tion of our pop u la tion should not mask the fact that many con di tions changed since the cre a tion of Prague University in 1348. Rules for granting de grees, such as a doc toral de gree, or the so cial sta tus of uni ver sity pro fes sors changed, for in stance, af ter the Humboldt re form (Schimank and Winnes 2000). Salaries prior to the mod ern pe riod were of ten low, paid in kind, or not paid at all , for ex am ple, dur ing the Thirty Years’ War (Klinge 2004; Langer 2011; Vandermeersch 2003). Still, con-sid er ing a chang ing en vi ron ment, we are as close as pos si ble to a well-de fined pop u-la tion. Any other pro fes sions, even the no bil i ty, would face sim i lar changes.

We in clude in our da ta base schol ars who were appointed to uni ver si ties that were founded be fore 1800. Based on Frijhoff (1996) and Steiger (1981), we iden tify 63 such uni ver si ties. Universities founded later than 1800 are ex cluded be cause the HRE ended in 1806 and be cause the Eu ro pean uni ver sity un der went rad i cal changes around this time. A large num ber of uni ver si ties disappeared at the be gin ning of the nine-teenth cen tu ry. In Germany, for in stance, 18 of 34 uni ver si ties were closed (Charle

45

3

1

2

9

12

11

13

7

10

15

8

14

16

6

21

23

24

30

18

31

28

29 22

27

26

19

25

20

17

38

44

50

46

32

34

51

45

49

42

48

36

47

40

35

37

39

33

43

65

70

67 71

66

69

68

53

56

61

58

62

52

57

55

60

6374

72

75

73

59

54

6441

Berlin

Prague

Vienna

Brussels

Amsterdam

Germany

Poland

France

Italy

Czech Republic

Austria Institution founded in

<14001400−14991500−15991600−16991700−1799>1799

Fig. 1 Universities and academies in the territories of the Holy Roman Empire and the Netherlands. Uni-versities and scientific academies located in the 1648 territories of the Holy Roman Empire (gray) and of the Netherlands (light gray). The area of the circle is proportional to the number of observations of the institution, and the color indicates the century of foundation. Numbers 1–63 mark universities sorted by year of foundation, and numbers 64–75 mark academies of sciences. For an entire list of the corresponding institutions, see Table A1 in the online appendix.



2004). More im por tant ly, uni ver si ties founded in the nineteenth cen tury followed a new uni ver sity model shaped by the Humboldt re form (Schimank and Winnes 2000). Thus, be cause our fo cus is on older uni ver si ties, our re sults are not af fected by the sig nifi cant changes that oc curred around 1800.

Still, the 63 older uni ver si ties were rather het ero ge neous along sev eral di men sions. The map of Figure 1 il lus trates the spa tial dis tri bu tion of these in sti tu tions as well as the num ber of schol ars re lated to each in sti tu tion by the area of the cir cle. Established in 1348, the University of Prague (marked 1 on the map) was the oldest uni ver sity in the HRE, followed by the University of Vienna (2) founded in 1365 and the University of Heidelberg (3) founded in 1386. The University of Bonn (62) and Karl’s High School (63) were the two youn gest uni ver si ties. In ad di tion, four more ap plied uni ver si ties, such as the Freiberg University of Mining and Technology (58), were among the in sti-tu tions established in the eigh teenth cen tu ry. In the Dutch ter ri to ries, uni ver sity ed u ca-tion started in 1575 with the es tab lish ment of Leiden University (27). Although seven uni ver si ties were founded in the Netherlands be fore 1800, the University of Nijmegen (46) ed u cated stu dents for only a very lim ited num ber of years, and the uni ver si ties in Franeker (31) and Harderwijk (45) closed near the end of the Napoleonic era. Several uni ver si ties in the HRE met the same fate. These uni ver si ties were closed ei ther in re-sponse to geo po lit i cal move ments, as was the case in Cologne (4) and Erfurt (5), or as a re sult of sec u lar i za tion, as was the case in Bamberg (44) and Dillingen (21).

Figure 1 also includes the 12 acad e mies of sci ences (some times com bined with arts) for which we have da ta. Although other acad e mies existed dur ing the pe riod we cov er, these in sti tu tions were, on av er age, of lesser im por tance.3 Over the course of his to ry, large num bers of sci en tific acad e mies have appeared, and some have disappeared. By far the most im por tant of the sci en tific acad e mies we cover is the Collegium Naturae Curiosorum, established in 1652 and bet ter known as Leopoldina (64). The Ba var ian Academy of Sciences and Humanities (68) and the Royal Nether-lands Academy of Arts and Sciences (72) were also well-known. The lat ter acad emy was founded in 1808. We in clude this in sti tu tion be cause acad e mies did not un dergo the same struc tural changes as uni ver si ties. Thus, there was no rea son to ex clude these more re cently established in sti tu tions.

Data and Sources of the Data Set

We use a range of sources to com pile our sam ple of schol ars. We as sign the in sti tu tions to four qual ity groups based on data avail abil ity and the data sources used. In the first qual ity group are the in sti tu tions for which we have (al most) com plete da ta. In this op ti-mal case, we rely on two types of high-qual ity sources: online pro fes sor cat a logs, such as the Catalogus Professorum Lipsiensium; or books that pro vide biographical in for-ma tion on pro fes sors, such as Drüll (2009, 2012) and Drüll-Zimmermann (1991, 2012) on the University of Heidelberg. Overall, the first qual ity group includes 23 uni ver si ties and 10 acad e mies of sci ences. Our sources of data for these acad e mies in clude of fi cial lists of mem bers pro vided ei ther di rectly by the acad emy or by their pub li ca tions.

3 For an over view on acad e mies, see the Scholarly Societies Project (http: / /www .references .net /societies /).


http://www.references.net/societies/


For other uni ver si ties, the cat a logs do not cap ture ei ther the whole time span or all fac ul ties. Because these sources still pro vide highly re li able in for ma tion, they are in cluded in the sec ond qual ity group, which is made up of in sti tu tions with par tially com plete da ta. For in stance, Günther (1858) pro vided in for ma tion on the University of Jena pro fes sors only up to the uni ver si ty’s 300th an ni ver sary in 1858, whereas Flessa (1969) pro vided in for ma tion on only med i cal fac ulty pro fes sors in Altdorf. In ad di tion to 15 uni ver si ties, the Royal Academy of Sciences, Letters and Fine Arts of Belgium is in cluded in this sec ond qual ity group.

The sources we use for the in sti tu tions in the third qual ity group en able us to fur-ther com plete our list of schol ars. The avail able data for the in sti tu tions in this third qual ity group are less com plete (noncomplete). For 12 uni ver si ties, we re con struct as many ob ser va tions as pos si ble from a va ri ety of sources, in clud ing lists on Wiki-pedia, which are when ever pos si ble backed-up with ad di tional sources, such as the Deutsche Biographie (2014–2020).4 For ex am ple, we col lect data on schol ars from the University of Erfurt (5) and Brandenburg University in Frankfurt (18) us ing this strat e gy. The remaining sci en tific acad e my, the Palatinate Academy of Sciences in Mannheim (68), is also assigned to the qual ity group noncomplete da ta.

The remaining 14 uni ver si ties are in the last qual ity group, scattered da ta. Their mem bers are cap tured ei ther via other uni ver si ties of a higher class or us ing data col lec tions, such as Fischer (1978). The oldest uni ver sity in this qual ity group is the University of Trier (14).5

By com bin ing the data from the sources in all four qual ity groups and re mov ing du pli cates, we gather in for ma tion on 33,498 schol ars. This pop u la tion forms our to tal sam ple of schol ars who were born be fore 1900 and who were ac tive in the de fined uni verse of uni ver si ties and sci en tific acad e mies. Individuals en tered the pop u la tion of schol ars at the time when they were first appointed. Individuals exited the pop u-la tion at death if that event is ob served. If death is not ob served, they are cen sored at the last exit from one of our in sti tu tions.

Data Quality

Given that our his tor i cal data cover more than four cen tu ries, some un cer tainty about these data is un der stand able, par tic u larly at the very be gin ning of our time span. In this sec tion, we dis cuss two po ten tial ca ve ats: miss ing val ues and heaping in the years of birth and death.

Table 1 sum ma rizes im por tant de scrip tive sta tis tics. The to tal sam ple of schol ars de clines from 33,498 to 31,176 if we con sider only those in di vid u als for whom the year of ap point ment and of death or exit are known. The sam ple fur ther shrinks to 27,842 if we in clude only those in di vid u als for whom their ages at the events are known; we use this as the base line sam ple for our mor tal ity es ti ma tions. Most cases for which the ages are un known suf fer from mul ti ple miss ing val ues. In 96.2% of

4 The re li abil ity of Wikipedia in for ma tion is not al ways easy to val i date. Less than 2% of our ob ser va tions are solely based on Wikipedia. We checked that all our re sults are ro bust to the ex clu sion of all in sti tu tions with out tra di tional sources; for de tails, see the Robustness sec tion and the online ap pen dix.5 Table A1 in the online ap pen dix pro vi des com plete over views of the qual ity groups and sources of all 75 in sti tu tions.



Tabl

e 1 Observations by year of ap point ment

Appointments W

ith Known Age

b

All Appointments

aMean Age at:

University Appointments

Academy Appointments

Year

NN

Appointment

Death

Death Observed (%

)Total

Medicine (%

)Total

Medicine (%

)

<1400

217

4637.5

63.3

97.8

462.2

0—

1400–1449

616

8732.5

64.2

98.9

878.0

0—

1450–1499

607

162

33.0

62.5

97.5

162

6.2

0—

1500–1549

770

349

32.3

60.5

98.6

349

8.0

0—

1550–1599

998

719

33.3

60.6

97.4

719

11.7

0—

1600–1649

1,085

893

34.0

58.8

96.9

891

10.7

2100.0

1650–1699

1,662

1,458

33.9

60.2

96.9

1,213

7.9

245

84.5

1700–1749

2,401

2,194

34.8

62.0

97.9

1,497

7.5

697

42.0

1750–1799

3,468

3,228

35.7

64.1

97.1

1,556

11.6

1,672

28.9

1800–1849

4,506

4,306

37.6

67.0

98.6

1,516

17.2

2,790

20.0

1850–1899

5,920

5,802

38.2

69.2

98.2

2,187

22.2

3,615

19.6

1900–1929

5,845

5,730

40.3

72.0

92.5

2,962

26.9

2,768

18.6

≥1930

3,081

2,868

53.9

76.9

92.6

1,258

24.6

1,610

25.7

All

31,176

27,842

39.0

67.8

96.2

14,443

17.1

13,399

23.7

a Used in the an a ly ses presented in Figure 3.

b Used in the an a ly ses presented in Figures 4–7.



cases in which we ob serve age, we also ob serve death. The share of right-cen sored cases for which death is un known in creases slightly for the most re cent pe ri ods be-cause our data sources, such as Conrad (1960) for the University of Tübingen, are from a time when many of the schol ars were still alive. We know the age at death for schol ars who died af ter the date of pub li ca tion only if we find them in other sources.

In the 1400–1700 time span, the mean age at first ap point ment was rather sta ble, at ages 32 to 35. In the more re cent pe ri od, the mean age at first ap point ment in creased, reaching as high as age 40 in the early twen ti eth cen tu ry. By con trast, the mean age at death followed a U-shaped pat tern. Starting at around age 64 for schol ars appointed in the early fifteenth cen tu ry, it de clined to age 59 among those who en tered the data set in the first part of the sev en teenth cen tu ry. The mean age at death in creased to more than 70 years for schol ars appointed at the on set of the twen ti eth cen tu ry.

The pop u la tion of schol ars was het ero ge neous along sev eral di men sions, two of which we fo cus on here. First, to ex plore the role of med i cal knowl edge, we dis tin guish be tween schol ars with and with out a med i cal back ground. To iden tify these in di vid u-als, we as cer tain whether the schol ars in our sam ple stud ied med i cine, held a PhD in med i cine, were ac tive in a med i cal fac ul ty, held a chair in med i cine, were ac tive in a field of re search linked to med i cine, or belonged to a class of med i cine in an acad emy of sci ences. Second, we dis tin guish schol ars by the sci en tific in sti tu tions to which they belonged. Members of acad e mies of sci ences rep re sent a sort of elite within the knowl-edge elite. Because these schol ars had more sci en tific achieve ments and bet ter ac cess to net works than non mem bers, they likely had higher so cial sta tus. Thus, we use mem-ber ships in acad e mies as so cial sta tus in di ca tors that we can link to mor tal ity dy nam ics.

Num bers for the four groups of schol ars are shown in Table 1: schol ars ac tive only in uni ver si ties, and among them, those linked to the med i cal pro fes sion; and schol-ars af fil i ated with an acad e my, and among them those linked to med i cine. In to tal, our data cover 14,443 schol ars af fil i ated with a uni ver sity only and 13,399 schol ars with an acad emy af fil i a tion. Although only a small per cent age of schol ars appointed be fore the sev en teenth cen tury were ac tive in the field of med i cine, this share had in creased to one-quar ter by the end of the fol low-up.

Uncertain or un known years of birth and death may be ap prox i mated by years end-ing with 0 or 5. The use of this ap proach might re duce the share of miss ing val ues, but it could also cause birth and death year heaping in our da ta. To check the scale of the heaping, we fol low the con cept of the Whipple-Index (Hobbs 2004) and com pare the ob served num ber of birth years end ing with 0 or 5 with the over all num ber of births in the same pe ri od.

Panel a of Figure 2 dis plays the shares of births per 50-year pe riod end ing with 0 or 5. If there was no heaping, we would ex pect 20% of the birth years to end with 0 or 5. We ob serve that more than one-half of birth years ended with 0 or 5 in the fifteenth cen tu ry. However, data qual ity im proves rather quickly there af ter: by the be gin ning of the sev en teenth cen tu ry, the share was al ready close to 0.2. If we re strict the sam ple to schol ars with out an un cer tain year of birth, the data qual ity im proves sig nifi cantly and be comes ac cept able starting in the sixteenth cen tu ry.6

We ob serve less year heaping for mor tal ity data than for birth da ta. In pre vi ous cen-tu ries, birth data was re li ably recorded for some fam i lies, such as those of schol ars,

6 For more de tailed in for ma tion, see Figures A1–A5 in the online ap pen dix.



1400 1500 1600 1700 1800

Year

Pro

po

rtio

n o

f B

irth

Yea

rs E

nd

ing

in 0

or

5

.0

.1

.2

.3

.4

.5

.6

.7

AllBaseline mortality estimationOnly in quality group 1 or 2Only in quality group 1 or 2 with known birth yearPerfect quality

a. Proportion of Birth Years Ending in 0 or 5

1400 1500 1600 1700 1800 1900

Year

Pro

po

rtio

n o

f D

eath

Yea

r s E

nd

ing

in 0

or

5

.0

.1

.2

.3

.4

AllBaseline mortality estimationOnly in quality group 1 or 2Only in quality group 1 or 2 with known death yearPerfect quality

b. Proportion of Death Years Ending in 0 or 5

Fig. 2 Birth and death year heaping. All includes all scholars with either a known year of birth or a known year of death. Baseline mortality estimation restricts the sample to the 27,769 scholars with known ages at appointment and at death or exit used in the estimations of life expectancy (column 2 in Table 1). Only in quality group 1 or 2 further limits the sample to scholars with data from quality groups 1 (almost complete data) and 2 (partially complete data). Finally, Only in quality group 1 or 2 with known birth or death year excludes all observations with the year of birth or death marked as approximated in the sources.



cler gy man, or no bles, whereas the birth dates of chil dren in or di nary fam i lies were less well-documented or were ap prox i mated in the data sources. However, when an in di vid-ual was appointed to one of our in sti tu tions, he or she likely be came suf fi ciently im por-tant to have his or her death date recorded. Hence, the share of death years end ing with 0 or 5 is gen er ally much closer to 0.2 (Figure 2, panel b). Furthermore, the ob served heap-ing is not al ways due to poor data qual i ty; for ex am ple, the peak in 1900–1949 (Figure 2, panel b) is driven by the ex cep tion ally high mor tal ity at the end of World War II.

Significant birth year heaping and some death year heaping is found at the be gin ning of our ob ser va tion win dow. Whether this heaping biases any of our re sults de pends on whether the years are sys tem at i cally ad justed up ward or down ward. This is dif fi cult to eval u ate di rect ly. As an in di rect ro bust ness check, we rep li cate our an a ly ses while ex clud ing all in di vid u als with a year of birth end ing with 0 or 5. The re sults are ro bust to this check, as discussed lat er.

The Population of Scholars

The schol ars’ years of first ap point ment and of death—or of last exit if the year of death is miss ing—en ables us to cal cu late the dy nam ics of the to tal num ber of schol-ars. Panel a of Figure 3 plots the 25-year mov ing av er ages of the first ap point ments, deaths, and deaths or ex its due to cen sor ing. The fig ure shows a gen eral in creas ing ten dency in ap point ments that was followed by a com pa ra ble in crease in deaths.

The flows displayed in panel a of Figure 3 are turned into stocks of schol ars in panel b. This fig ure shows three pe ri ods marked by the trend line im posed on the to tal stock of schol ars. First, be fore 1618, ap point ments exceeded out flows ex cept for very short pe ri ods in the mid dle of the fifteenth and sixteenth cen tu ry. The num-ber of schol ars grew by an av er age of 0.5% per year, hence at the same pace as the to tal pop u la tion.7 Second, around 1618, ap point ments started to de cline and con tin-ued to de cline un til ap prox i ma tely 1648. In the 1618–1648 pe ri od, out flows exceeded ap point ments, and for roughly one-quar ter of a cen tu ry, the pop u la tion of schol ars de creased by ap prox i ma tely 0.6% an nu al ly. Third, the ap point ments bounced back starting around 1648. From that year un til around 1900, ap point ments strongly exceeded deaths, and the pop u la tion of schol ars grew at an an nual rate of 0.8%.

After the first acad emy was founded, the shares of the sam ple be long ing to uni ver-si ties and acad e mies were soon bal anced. After 1800, one-half of the sam ple were ac-tive in at least one acad emy of sci ences. However, to avoid small sam ple noise—and given that be fore the eigh teenth cen tu ry, the Leopoldina was the only acad emy to which schol ars could be appointed—we limit our in ves ti ga tion of the link be tween so cial sta tus and mor tal ity to the pe ri ods af ter 1700, when the sec ond acad emy was established. Before 1700, we can com pare only the Medicine and Universities and Nonmedicine and Universities groups. However, be cause panel b of Figure 3 shows that only a few schol ars were en gaged in med i cine be fore the sev en teenth cen tu ry, we limit our in ves ti ga tion of mor tal ity dy nam ics to the pe riod af ter 1600.

7 Pfister and Fertig (2010) documented an av er age growth rate of ap prox i ma tely 0.5% per an num for the Ger man pop u la tion.



The trends that we ob serve in the num bers of ap point ments and ex its/deaths in panel a of Figure 3—and hence, in our to tal pop u la tion—cap ture dif fer ent de vel op-ments. First, the ap point ment num bers are sen si tive to the size of each in sti tu tion. Second, the ap pear ance and dis ap pear ance of uni ver si ties and acad e mies—such as the clos ing of a num ber of in sti tu tions af ter the Napoleonic Wars—al ters the num ber of ap point ments and ex its. Because we lack ex haus tive sources for all the in sti tu tions in our sam ple, a cer tain num ber of schol ars within each in sti tu tion might be miss ing.

1400 1500 1600 1700 1800 1900

5

10

20

50

100

200

Year

Ap

po

intm

ents

, Dea

ths,

an

d E

xits

(lo

g)

Appointments

Deaths or exits

Deaths

a. 25-Year Moving Averages: Appointments, Deaths, and Deaths or Exits

1400 1500 1600 1700 1800 1900

1

510

50100

5001000

5000

Year

Sch

ola

rs (

log

)

ScholarsTrend

Scholars with known birth year

Nonmedicine and academy

Medicineand academy

Medicine and university

Nonmedicineand university

b. Number of Scholars

Fig. 3 The dynamics in the population of scholars



Thus, sam ple se lec tion is a third fac tor that could af fect our re sults. Missing or un cer-tain in for ma tion within the sam ple of schol ars on the years of events is a fourth fac-tor. Panel b of Figure 3 pro vi des some in sights into the role of miss ing events. If we limit our pop u la tion to schol ars whose year of birth is known, the ini tial pop u la tion is smaller and grows faster in the early pe ri od. Given that ages are re quired to es ti mate mor tal i ty, fur ther in ves ti ga tions rely on this smaller sam ple.

Methods

We es ti mate age-spe cific mor tal ity rates and life ex pec tancy at age 30 for the pop u la-tion of schol ars and for the sub pop u la tions. We use age 30 as the starting age for life ex pec tancy be cause for youn ger ages, the sam ple of schol ars be comes very small.8 Because of the lim ited sam ple size in the early years, we first smooth the death rates over age and time (Camarda 2012). We then com pute pe riod life ex pec tancy at age 30 over rolling 25-year in ter vals and ap ply Monte Carlo sim u la tions to es ti mate the cor-re spond ing con fi dence in ter vals (Andreev and Shkolnikov 2010; Chiang 1984). In the fol low ing, years mark the mid dle of the 25-year in ter vals. The con tri bu tions of each age to changes in life ex pec tancy are decomposed by the step wise re place ment al go rithm from Andreev and Shkolnikov (2012), which is de scribed in Andreev et al. (2002).

Findings

Life Expectancy

Three clear pat terns emerge in life ex pec tancy at age 30 for the pop u la tion of schol ars (Figure 4). First, we ob serve no sys tem atic im prove ment in life ex pec tancy among schol ars be fore the mid dle of the eigh teenth cen tu ry. Second, we find ev i dence of a sharp de cline in life ex pec tancy in the first half of the sev en teenth cen tu ry. Scholars’ life ex pec tancy at age 30 de clined from more than 30 to less than 27 years. Third, we see that there af ter, a phase of steady im prove ments in mor tal ity be gan. Between 1750 and 1900, con di tional life ex pec tancy in creased by roughly 7.5 years.

To shed light on life ex pec tancy dy nam ics, we add to Figure 4 two types of his tor-i cal events that might have influ enced life ex pec tan cy: wars and pan dem ics. We iden-tify four mil i tary con flicts that may have af fected mor tal ity dy nam ics: (1) the Ger man Peasants’ War, (2) the Thirty Years’ War, (3) the Seven Years’ War, and (4) the Rev-olutionary Wars. In ad di tion, we iden tify two waves of the plague (years 1547–1550 and 1625–1640) that may have influ enced mor tal ity dy nam ics.

In the 1500–1600 pe ri od, our point es ti ma tes for life ex pec tancy at age 30 fluc tu ate strongly around the smoothed av er age of ap prox i ma tely 30 years. This pe riod includes two im por tant his tor i cal events that may have re duced life ex pec tan cy: the Ger man Peasants’ War and the plague pan dem ic. Although it is pos si ble that these events influ-enced the schol ars’ life ex pec tan cy, the small-sam ple var i a tion and the width of the con-fi dence in ter vals un til around 1600 are so large that we can not draw clear con clu sions.

8 For a more de tailed dis cus sion, see sec tion A.3.1 in the online ap pen dix.



The stron gest as so ci a tion be tween life ex pec tancy and the his tor i cal events was in the early part of the sev en teenth cen tu ry, where we ob serve a sharp de cline in life ex pec tancy co in cid ing with both the Thirty Years’ War (1618–1648) and the sec ond plague pan dem ic. It is plau si ble that the mor tal ity cri sis was caused by one or both of these his tor i cal events. For the Thirty Years’ War in par tic u lar, me di at ing ef fects may have been more im por tant than the di rect ef fects of mil i tary con flicts. Greifswald, for in stance, was forced to bil let more than 1,000 Imperial sol diers, along with their horses and ar ma ments, af ter the Duke of Pomerania sur ren dered in 1627. During this pe ri od, uni ver sity life was lim it ed, and pro fes sors did not re ceive salaries (Langer 2011). The pass ing sol diers spread in fec tious dis eases, and hy gienic stan dards de te-ri o rat ed. The seem ingly end less string of plague ep i dem ics that oc curred be tween 1625 and 1640 (sur face B in Figure 4) il lus trates the harsh con di tions surrounding the Thirty Years’ War and may ex plain the very high death rates. During the war, peo-ple suf fered not only from the plague but also from fam ine (Alfani and Gráda 2018). Hence, it is likely that the three fa mous Mal thu sian mech a nisms—fam ine, ep i dem ics, and war—com bined to lower life ex pec tancy (Flinn 1981).

At the end of the Thirty Years’ War, life ex pec tancy started to re cover and death rates de creased, al beit with some fluc tu a tions in the first part of the eigh teenth cen-tu ry. Still, it took al most a cen tury for life ex pec tancy to return to pre war and pre-plague lev els. Note that this pe riod of de clin ing mor tal ity could be misinterpreted as sig nal ing the on set of sys tem atic mor tal ity im prove ments if the time span does not in clude the pre-cri sis pe ri od.

1500 1600 1700 1800 1900

55

60

65

70

Year

Expe

cted

Age

at D

eath

: 30

+ e 3

0 1 − GermanPeasants’ War

2 − ThirtyYearsYY ’ War

3 − SevenYearsYY ’ War

4 − Revol.WaW rsrr

B − PlagueA − Plague

Fig. 4 The dynamics of scholars’ period life expectancy and historical events. The figure applies 25-year rolling intervals and two-dimensional smoothed data. The solid black line displays our estimated period life expectancy for scholars at age 30 in 25-year rolling intervals. For instance, the year 1550 covers the 1538–1562 period. The gray area marks the corresponding 95% confidence intervals. The brighter sur-rounding years on the left and the right of each event indicate estimated life expectancies that include years altered by the events due to the rolling intervals.



Figure 4 shows that sys tem atic mor tal ity im prove ments that can not be interpreted as re cov ery from a cri sis started around 1750. From that point on ward, life ex pec-tancy in creased with out in ter rup tions. Thus, no clear as so ci a tion be tween sub se quent wars and life ex pec tancy is ob serv able.

Panel a of Figure 5 shows which ages con trib uted to the life ex pec tancy changes across the three stages we iden ti fy: de cline from 1580 to 1630; re cov ery from 1630

30 40 50 60 70 80 90 100

Age

Year

s C

on

trib

ute

d t

o t

he

Ch

ang

es in

e30

−.3

−.2

−.1

.0

.1

.2

.3

1580 to 1630

1630 to 1750

1750 to 1900

a. Contribution to Decline (1580–1630), Recovery (1630–1750), and Increase (1750–1900) in Life Expectancy by Age

1500 1600 1700 1800 1900

2

5

10

20

50

100

200

Year

Dea

ths

per

1,0

00 S

cho

lars

(lo

g)

1580 1630 1750 1900

70+

60−69

50−59

40−49

30−39

1 − GermanPeasants’ War

rtyWar

3 − SSevenYears’sYY War

4 − Revol.Wars

B − PlP agueA − Plague

b. Age-Specific Death Rates

Fig. 5 Decomposition of gains in life expectancy. The figure applies 25-year rolling intervals and two- dimensional smoothed data.



to 1750; and in crease from 1750 to 1900. Panel b shows age-spe cific mor tal ity trends over the whole pe ri od.

In the first stage, not all ages were equally af fected by de clin ing life ex pec tan cy. Almost all of the life ex pec tancy losses were among peo ple un der age 60 (Figure 5, panel a). Older peo ple (ages 60–69 and 70+) also ex pe ri enced mor tal ity in creases dur-ing this pe riod (Figure 5, panel b), but their con tri bu tions to life ex pec tancy losses were small be cause so few in di vid u als sur vived to these ad vanced ages. The de com po si tion also shows that peo ple un der age 60 were the main con trib u tors to life ex pec tancy gains in the late sev en teenth and early eigh teenth cen tu ries: the in crease in life ex pec tancy was ef fec tively driven by the same ages that drove the sev en teenth-cen tury mor tal ity cri sis. When the sec u lar in crease in lon gev ity started in the mid-eigh teenth cen tu ry, peo ple at ages 60–80 joined peo ple un der age 60 in con trib ut ing to life ex pec tancy in creases (panel a). Panel b shows that af ter 1750, mor tal ity de clined fastest in the older age groups.

Comparison With Other Studies

We have presented our re sults by pe riod to fa cil i tate the anal y sis of when mor tal ity im prove ments started. Many other stud ies have used the co hort per spec tive. For com-par a tive pur poses, we con vert our data to the co hort per spec tive. The mes sage of our co hort re sults is sim i lar to that of our find ings from the pe riod per spec tive: co hort life ex pec tancy did not im prove be tween the 1400 and 1700 birth co horts; life ex pec tancy de clined for the co horts born in the late six teenth cen tury who were adults dur ing the Thirty Years’ War and the sec ond plague pan dem ic; and sec u lar im prove ments in life ex pec tancy be gan with the co horts born in the eigh teenth cen tu ry. Figure 6 com pares our co hort life ex pec tancy es ti ma tes at age 30 with oth ers found in the lit er a ture.

1400 1500 1600 1700 1800 1900

40

50

60

70

80

Birth Cohort

Expe

cted

Age

at D

eath

: x +

ex

Scholars, e30

Swedish Males(HMD 2019), e30North/N−Eastern European

nobles (Cummins 2017), e20

Central/EasternEuropean nobles

(Cummins 2017), e20

Famous people (De la Croix andLicandro 2015), e15

Cardinals (Fornasin et al. 2010),e60

Fig. 6 The dynamics of scholars’ cohort life expectancy in light of the literature. The figure applies 25-year rolling intervals and two-dimensional smoothed data of scholars.



We com pare our re sults with five other sets of find ings pertaining to no bles in 1400–1800 in north and north east ern Europe as well as cen tral and east ern Europe (Cum-mins 2017); fa mous peo ple in 1400–1875 (de la Croix and Licandro 2015); car di nals in 1400–1900 (Fornasin et al. 2010); and Swed ish life ex pec tancy in 1751–1899 (from the Human Mortality Database [HMD] 2019).

We high light two di men sions of this com par i son: trends and lev els. The trends show qual i ta tive sim i lar i ties: no bles (Cummins 2017), fa mous peo ple (de la Croix and Lican-dro 2015), and schol ars of the HRE (this man u script) all ex pe ri enced pe ri ods of stag-na tion in mor tal ity followed by sustained in creases in life ex pec tan cy. However, the mor tal ity im prove ments started ear lier among fa mous peo ple (1680s) and later among Sweden’s male birth co horts (1800) and car di nals (1850s) than among schol ars and no bles. More im por tant ly, none of the other pop u la tions un der went the mor tal ity cri ses ob served among co horts born at the end of the six teenth cen tu ry. Because of the dif fer-ences in ter ri to rial cov er age, the im pact of the Thirty Years’ War is likely less im por tant in this com par i son.

However, these stud ies reported dis pa rate lev els of life ex pec tan cy, pri mar ily be-cause they used dif fer ent starting ages. Cummins (2017) used age 20, and Fornasin et al. (2010) used age 60; de la Croix and Licandro (2015) ex cluded fa mous in di vid u-als who died be fore age 15. Only the Swed ish life ex pec tancy from the HMD is based on a starting age of 30. These dif fer ent starting ages lead to pre dict able dif fer ences in lev els. It is, for ex am ple, worth con sid er ing what dif fer ence it makes that Cummins used age 20 and there fore in cluded a larger share of young adult mor tal i ty. In the HMD life ta bles for Sweden, mean age at death con di tional to sur viv ing to age 20 for the 1800 male co hort is 56.1, and 59.2 con di tional on sur viv ing to age 30. The dif fer ences be tween our 1800 co hort and Cummins’ no bles are 9.2 years in north and north east ern Europe and 11.5 years in cen tral and east ern Europe. Hence, these dif-fer ences are larger than the 3.1 years in the HMD for Swed ish co horts. The remaining life ex pec tancy discrepancies be tween schol ars and no bles can be explained by the high shares of vi o lent deaths among no bles (Cummins 2017). The life ex pec tancy gap be tween schol ars and car di nals is 7.9 years in 1800, and is thus smaller than the dif fer ence based on the cal cu la tion from HMD’s life ta bles for Swed ish males (12 years). Finally, we do not ob serve any note wor thy sys tem atic de vi a tion from the un con di tional mean age at death es ti mated by de la Croix and Licandro (2015) for fa mous peo ple who sur vived un til age 15.

Life Expectancy, Social Status, and Medical Knowledge

Figure 7 il lus trates the mor tal ity dy nam ics of schol ars sep a rated into four groups according to their field of sci ence and mem ber ship in an acad emy of sci ences. The fig-ure starts in 1600 for uni ver sity schol ars and in 1700 for acad emy mem bers be cause these strat i fied sam ples would be too small for the ear lier years. Panel a com pares uni ver sity pro fes sors who were and were not in the field of med i cine. We ob serve no sys tem atic mor tal ity dif fer ence be tween these groups of schol ars un til the early to mid-eigh teenth cen tu ry, when life ex pec tancy started to in crease among schol ars not in med i cine. Mortality im prove ments among schol ars linked to the med i cal field were delayed. In line with van Poppel et al. (2016), we find that life ex pec tancy was



lower for schol ars in med i cine than for other schol ars for ap prox i ma tely 100 years. However, this mor tal ity gap de creased to ward the end of the nineteenth cen tu ry.

Panel b com pares mor tal ity im prove ments among schol ars who were or were not in med i cine, and who were ac tive in acad e mies of sci ences. The pat tern is sim i lar to that ob served among uni ver sity schol ars: starting from around 1750, schol ars who were not in med i cine ex pe ri enced mor tal ity de clines, and those who were in med i-

1600 1650 1700 1750 1800 1850 1900

55

60

65

70

Year

Exp

ecte

d A

ge

at D

eath

: 30

+ e

30

Medicineand university

Nonmedicine anduniversity

a. Medicine and Nonmedicine in Universities

1600 1650 1700 1750 1800 1850 1900

55

60

65

70

Year

Exp

ecte

d A

ge

at D

eath

: 30

+ e

30

Medicineand academy

Nonmedicineand academy

b. Medicine and Nonmedicine in Academies

Fig. 7 Social status, medicine, and life expectancy. The figure applies 25-year rolling intervals and two- dimensional smoothed data.

(continued)



1600 1650 1700 1750 1800 1850 1900

55

60

65

70

Year

Exp

ecte

d A

ge

at D

eath

: 30

+ e

30Nonmedicineand academy

Nonmedicine anduniversity

c. Nonmedicine in Universities and Academies

1600 1650 1700 1750 1800 1850 1900

55

60

65

70

Year

Exp

ecte

d A

ge

at D

eath

: 30

+ e

30

Medicineand university

Medicineand academy

d. Medicine in Universities and Academies

Fig. 7 (continued)

cine ex pe ri enced these de clines with a de lay, with the gap narrowing to ward the end of the nineteenth cen tu ry.

Panels c and d com pare uni ver sity schol ars with those nom i nated to a sci en tific acad e my. Panel c re fers to schol ars out side med i cine, and panel d re fers to schol ars in med i cine. Both pan els show that from around 1750, mem bers of the acad e mies had a higher life ex pec tan cy. This mor tal ity ad van tage lasted for ap prox i ma tely 100 years, un til 1850. Thereafter the dif fer ences fluc tu ated so strongly that we re frain



from draw ing con clu sions other than that the ad van tage existed from ap prox i ma tely 1750 to 1850 and then per haps di min ished.

Robustness

We an a lyzed whether the fol low ing find ings remained af ter data lim i ta tions: life ex pec-tancy stag nated from 1500 to 1600; the early sev en teenth cen tury was marked by a mor tal ity cri sis; sec u lar in creases started around 1750; and schol ars who were out side med i cine and belonged to acad e mies of sci ences had a mor tal ity ad van tage com pared with schol ars who were in med i cine and were not appointed to acad e mies.

We first lim ited the sam ple to data from in sti tu tions for which we had (al most) com plete data or par tially com plete data (the first two qual ity groups presented in the data sec tion). It also ex cludes in sti tu tions with out tra di tional sources, such as books. Second, to test the ro bust ness with re spect to birth and death year heaping, we ex cluded all ob ser va tions with a year of birth end ing with 0 or 5. Third, we es ti mated life ex pec tancy with vary ing starting ages: 25, 35, 45, and 55. Our key re sults were ro bust to these tests (see Figures A7–A10 in the online ap pen dix).

Discussion

We gath ered data from thou sands of schol ars in the 1648 ter ri to ries of the Holy Ro man Em pire and the Netherlands. By com bin ing vi tal in for ma tion with ap point-ment and exit in for ma tion, we were able to com pute mor tal ity dy nam ics while tak ing into ac count left-trun ca tion and right-cen sor ing. Based on this novel data set, we pro-vide new in sights into when mor tal ity first started to de cline and which groups were the first to ex pe ri ence ris ing lon gev i ty. Our re sults show that adult life ex pec tancy was stag nant from 1500 to 1750; the early sev en teenth cen tury was marked by a mor-tal ity cri sis; sec u lar in creases started around 1750; and schol ars out side med i cine and within acad e mies of sci ences had a mor tal ity ad van tage com pared with those within med i cine and not appointed to acad e mies. These re sults par tially cor rob o rate and par tially chal lenge prior find ings. Furthermore, they persisted af ter sev eral ro bust ness checks were performed.

The het ero ge ne ity in our pop u la tion of schol ars en abled us to study dif fer ences in the tim ing of mor tal ity im prove ments. We showed that the higher so cial sta tus of mem-bers of sci en tific acad e mies rel a tive to that of or di nary schol ars at uni ver si ties was as so ci ated with ear lier im prove ments in mor tal i ty. These find ings are in line with the vast lit er a ture show ing that so cial sta tus and life ex pec tancy are pos i tively cor re lated (Andreev et al. 2011; Elo et al. 2014; Johansson 1999). Importantly, we showed that this dif fer ence al ready existed in the eigh teenth cen tu ry. However, our re sults sug gest that this dif fer ence may have been tem po rar ily di min ished to ward the end of the nineteenth cen tu ry. During this pe ri od, the com par a tively el e vated so cial sta tus as so ci ated with be ing a mem ber in a sci en tific acad e my, rather than an “or di nary” pro fes sor, might have weak ened. Universities changed from be ing vo ca tional schools to be ing re search in sti-tu tions fol low ing the Humboldt re form (Schimank and Winnes 2000).



We also found that mor tal ity im prove ments were delayed among med i cal pro fes sion-als. The lack of un der stand ing of germ the ory be fore the nineteenth cen tury sug gests that schol ars in the med i cal field would have faced a mor tal ity dis ad van tage. However, we found no sys tem atic dis ad van tage among med i cal pro fes sion als un til the be gin ning of the sec u lar lon gev ity im prove ments. The lim ited role of for mal med i cine in healing is a pos si ble ex pla na tion for this find ing. Although hav ing an ac a demic ca reer was cer-tainly use ful for obtaining of fi cial po si tions, such as a court or per sonal phy si cian, and was there fore linked to so cial sta tus, it was not nec es sar ily an ad van tage in com pet ing with prac ti tion ers on the med i cal mar ket place, such as sur geons, midwives, bar bers, apoth e car ies, and even folk heal ers and il le gal care pro vid ers (Broman 1995). It is even pos si ble that the high so cial sta tus of ac a demic med i cal pro fes sion als gave them a mor-tal ity ad van tage. However, when sys tem atic mor tal ity im prove ments be gan and the role of for mal med i cine in creased, gains for med i cal pro fes sion als were delayed. In line with van Poppel et al. (2016), we found that med i cal pro fes sion als had a mor tal ity dis ad van-tage for al most a cen tu ry. Rapidly in creas ing med i cal knowl edge and the dif fu sion of germ the ory might have quickly com pen sated for the higher in fec tion risks. As early as in the nineteenth cen tu ry, ex cess mor tal ity de clined among schol ars in the med i cal pro-fes sion re gard less of whether they belonged to acad e mies of sci ences.

Our es ti ma tion of schol ars’ lon gev ity also pro vi des us with in sights into the ca pac-ity for knowl edge ac cu mu la tion and dif fu sion. Our find ing that the rise in lon gev ity among the ed u cated seg ment of so ci ety pre ceded in dus tri al i za tion is con sis tent with the hy poth e sis that hu man cap i tal played a sig nifi cant role in the pro cess of in dus-tri al i za tion and the take off to mod ern growth (Galor 2011). The num ber of schol ars and the length of their pro duc tive lives may have af fected how much they were able to in flu ence their cul tural and eco nomic en vi ron ments—and, in turn, so cio eco nomic de vel op ment and eco nomic growth.9

The pop u la tion of schol ars we an a lyzed has the clear ad van tage of representing a dis tinct uni verse: in di vid u als ac tive at one of the de fined in sti tu tions. However, our data se lec tion also comes with lim i ta tions. First, our data re fer al most ex clu sively to elite men (only 89 women are in the sam ple).

Second, given that uni ver si ties and acad e mies of sci ences are ur ban in sti tu tions, our pop u la tion was also ur ban. Thus, these schol ars were ex posed to the ur ban mor tal-ity pen alty (Vögele 2000; Woods 2003). Because we were not able to an a lyze mor tal-ity among or di nary peo ple or wom en, we can only spec u late about the life ex pec tancy dy nam ics in the gen eral pop u la tion dur ing the era we an a lyzed. Our find ing that the elite lost sev eral years of life ex pec tancy dur ing the Thirty Years’ War might sug gest that the gen eral pop u la tion ex pe ri enced a sim i lar or greater mor tal ity cri sis: it is a rea-son able as sump tion that so cial sta tus had a pro tec tive ef fect dur ing wars and pan dem-ics. The dif fer ences in the tim ing of sec u lar mor tal ity gains among the up per tail of the

9 The out stand ing role of up per-tail hu man cap i tal in Europe’s his tor i cal de vel op ments—and, more pre-cise ly, in its knowl edge ac cu mu la tion, eco nomic growth, and in dus tri al i za tion—has been em pha sized in the re cent re search lit er a ture. For ex am ple, the num ber of peo ple in eigh teenth-cen tury France who sub-scribed to the Diderot’s and d’Alembert’s Grande Encyclopédie pre dicts sub se quent eco nomic de vel op-ment at both the city and the county level (Squicciarini and Voigtländer 2015). Moreover, Ger man cit ies that de vel oped bet ter in sti tu tions fol low ing the Reformation grew more quickly and had more res i dents who were reg is tered as fa mous in the Ger man bi og ra phy da ta base (Dittmar and Meisenzahl 2016).



elite (mem bers of the acad e mies) and the nor mal elite (pro fes sors) may also sug gest that in the gen eral pop u la tion, sec u lar mor tal ity im prove ments did not start be fore the mid dle or the end of the eigh teenth cen tu ry. However, this spec u la tion may be wrong given that most schol ars were liv ing in ur ban ar eas. Thus, the lev els and the tim ing of mor tal ity im prove ments among schol ars may have dif fered from those of or di nary peo ple liv ing in ru ral ar eas (Woods 2003).

Third, the char ac ter is tics of the in sti tu tions and their mem bers evolved over time. The struc ture of uni ver si ties was quite dif fer ent in the late me di e val era than in more re cent pe ri ods.10 Prior to the mod ern era, schol ars’ salaries were of ten paid in kind (or not all). Thus, al though work ing at a uni ver sity was linked to higher so cial sta-tus, it of ten did not gen er ate sig nifi cant in come ad van tages. Furthermore, the pro-cess of ap point ment—and, hence, of se lec tion into our pop u la tion—changed. The role of kin ship in the ap point ment pro cess was grad u ally replaced by sci en tific cri te-ria (Klinge 2004; Vandermeersch 2003).

Fourth, how we de fined schol ars and which in sti tu tions we in cluded are po ten tial driv ers of the es ti mated pop u la tion size. For in stance, we may have underestimated the growth in the scholar pop u la tion in the nineteenth cen tury be cause we did not in clude uni ver si ties founded later than the eigh teenth cen tu ry. It is also pos si ble that schol ars were ac tive in a sci en tific in sti tu tion be fore the first ob served ap point ment. If so, we lost years at risk at ear lier ages, bi as ing our life ex pec tancy es ti ma tes down ward. The ac a demic po si tion that de fi nes the en try in the data set also varies across sources.

Fifth, be cause the schol ars were not appointed un til they reached young adult-hood, our in ves ti ga tion was lim ited to adult mor tal i ty. The ef fects of shifts in in fant and child mor tal ity on the evo lu tion of life ex pec tancy were neglected.

Finally, our new data set does not al low us to draw con clu sions about the im pact of so cial sta tus or the med i cal pro fes sion on mor tal i ty. We could only iden tify as so ci a tions.

In sum ma ry, our anal y sis of a new data set cov er ing more than 30,000 schol ars pro-vi des new in for ma tion on when mor tal ity first started to de cline and on the so cial dif fer-ences in the first mor tal ity gains. We find that adult mor tal ity was stag nant or de clin ing from the six teenth to the sev en teenth cen tury as wars and ep i dem ics led to a mor tal ity cri sis that resulted in the loss of sev eral years of adult life ex pec tan cy. However, sec u lar life ex pec tancy im prove ments started as early as in the eigh teenth cen tu ry, and the life ex pec tancy of schol ars as so ci ated with elite acad e mies (rather than only be ing af fil i ated with a uni ver si ty) in creased first, whereas mor tal ity im prove ments were delayed for schol ars in med i cine. Our re sults pro vide a new per spec tive on the his tor i cal tim ing of mor tal ity im prove ments and on the so cio eco nomic and oc cu pa tional dif fer en tials in this tim ing. A da ta base ac com pa ny ing this ar ti cle fa cil i tates rep li ca tion and ex ten sions. ■

Acknowledgments The data col lec tion greatly benefited from all the uni ver si ties that were will ing to share their pro fes sor cat a logues with us: the University of Amsterdam (Album Acaemicum), the University of Groningen (Catalogus Professorum Academeiae Groninganae), Leiden University (Leidse Hoogleraren Vanaf 1575), the University of Leipzig (Catalogus Professorum Lipsiensis), the University of Marburg (Catalogus Professorum Academiae Marburgensis), the University of Rostock (Catalogus Professorum

10 We sum ma rize some ma jor evo lu tions in the char ac ter is tics of the uni ver si ties and their schol ars in the online ap pen dix.



Rostochiensium), the University of Bamberg (Bamberger Professorinnen- und Professorenkatalog), and the University of Utrecht (Catalogus Professorum Academiae Rheno-traiectinae). We also grate-fully ac knowl edge the data on mem bers pro vided by the fol low ing acad e mies of sci ences: the Academy of Sciences and Literature in Mainz, the Ba var ian Academy of Sciences and Humanities, the Berlin-Brandenburg Academy of Sciences and Humanities, the Heidelberg Academy of Sciences and Humanities, the Leopoldina, and the Saxon Academy of Sciences. Furthermore, we are thank ful to Dagma Drüll and Tino Fröde for shar ing with us ad di tional ma te rial and their knowl edge in sup ple ments to their published works on the University of Heidelberg and the Upper Lusatian Academy of Sciences, re spec tive ly; to Tino Steyer for pro vid ing the data on the University of Helmstedt; to the mem bers of the work ing group on the Kiel pro fes sors for shar ing their work in prog ress; to the Göttingen Academy of Sciences and Humanities and the Royal Netherlands Academy of Arts and Sciences for send ing us books with the avail able da ta; and to all oth ers who pro vided us with valu able in for ma tion and in put in the pro cess of data col lec tion. Finally, we thank Julie Duchêne and Guillaume Catoire for their help in build ing the da ta base for Louvain, and we are deeply grate ful for the pa tient sup port of Annika Onemichl in the pro cess of gath er ing and cleaning the da ta. Our ar ti cle benefited from the ad vice we re ceived and dis cus sions at the Max Planck Institute for Demographic Research as well as from par tic i pants’ com ments dur ing the 2017 work shop “Elite Human Capital and the Road to Modernity: The East vs the West” in Mar seilles. Da vid de la Croix ac knowl edges the fi nan cial sup port of the pro ject ARC 15/19-063 of the Bel gian French-speak ing Community. He also thanks Imera (Marseille, France) for wel com ing him dur ing fall 2017 and for funding Robert Stelter’s visit there. Robert Stelter ac knowl edges fi nan cial sup port from the Max Geldner Foun da tion.

References

Alfani, G., & Gráda, C. Ó. (2018). The tim ing and causes of fam ines in Europe. Nature Sustainability, 1, 283–288. https: / /doi .org /10 .1038 /s41893 -018 -0078 -0

Andreev, E. M., Jdanov, D., Shkolnikov, V. M., & Leon, D. A. (2011). Long-term trends in the lon gev ity of sci en tific elites: Evidence from the Brit ish and the Rus sian acad e mies of sci ence. Population Studies, 65, 319–334.

Andreev, E. M., & Shkolnikov, V. M. (2010). Spreadsheet for cal cu la tion of con fi dence lim its for any life ta ble or healthy-life ta ble quan ti ty (MPIDR Technical Report TR-2010-005). Rostock, Germany: Max Planck Institute for Demographic Research.

Andreev, E. M., & Shkolnikov, V. M. (2012). An Excel spread sheet for the de com po si tion of a dif fer ence be tween two val ues of an ag gre gate de mo graphic mea sure by step wise re place ment run ning from young to old ages (MPIDR Technical Report TR–2012–002). Rostock, Germany: Max Planck Insti-tute for Demographic Research.

Andreev, E. M., Shkolnikov, V. M., & Begun, A. Z. (2002). Algorithm for de com po si tion of dif fer ences be tween ag gre gate de mo graphic mea sures and its ap pli ca tion to life ex pec tan cies, healthy life ex pec-tan cies, par i ty-pro gres sion ra tios and to tal fer til ity rates. Demographic Research, 7, 499–522. https: / /doi .org /10 .4054 /DemRes .2002 .7 .14

Bengtsson, T., & Dribe, M. (2011). The late emer gence of so cio eco nomic mor tal ity dif fer en tials: A mi cro-level study of adult mor tal ity in south ern Sweden 1815–1968. Explorations in Economic History, 48, 389–400.

Bengtsson, T., & van Poppel, F. (2011). Socioeconomic inequalities in death from past to pres ent: An in tro-duc tion. Explorations in Economic History, 48, 343–356.

Broman, T. (1995). Rethinking pro fes sion al i za tion: Theory, prac tice, and pro fes sional ideology in eigh-teenth-cen tury Ger man med i cine. Journal of Modern History, 67, 835–872.

Camarda, C. G. (2012). MortalitySmooth: An R pack age for smooth ing Poisson counts with P-splines. Journal of Statistical Software, 50(1), 1–24.

Carrieri, M. P., & Serraino, D. (2005). Longevity of popes and art ists be tween the 13th and the 19th cen-tu ry. International Journal of Epidemiology, 34, 1435–1436.

Charle, C. (2004). Patterns. In W. Rüegg (Ed.), A his tory of the uni ver sity in Europe: Universities in the nineteenth and early twen ti eth cen tu ries (1800–1945) (Vol. 3, pp. 33–82). Cambridge, UK: Cam-bridge University Press.


https://doi.org/10.1038/s41893-018-0078-0

https://doi.org/10.4054/DemRes.2002.7.14

https://doi.org/10.4054/DemRes.2002.7.14


Chiang, C. L. (1984). The life ta ble and its ap pli ca tions. Malabar, FL: Robert E. Krieger.Conrad, E. (1960). Die Lehrstühle der Universität Tübingen und ihre Inhaber (1477–1927) [The chairs

of the University of Tübingen and their hold ers (1477–1927)] [Doctoral dis ser ta tion, University of Tübingen, Germany].

Cummins, N. (2017). Lifespans of the Eu ro pean elite, 800–1800. Journal of Economic History, 77, 406–439.

de la Croix, D. (2017). Did lon ger lives buy eco nomic growth? From Malthus to Lucas and Ben-Porath. In M. Cervellati & U. Sunde (Eds.), Demographic change and long-run de vel op ment (pp. 69–89). Cambridge, MA: MIT Press.

de la Croix, D., & Licandro, O. (2015). The lon gev ity of fa mous peo ple from Hammurabi to Einstein. Journal of Economic Growth, 20, 263–303.

de la Croix, D., & Sommacal, A. (2009). A the ory of med i cal ef fec tive ness, dif fer en tial mor tal i ty, in come in equal ity and growth for pre-in dus trial England. Mathematical Population Studies, 16, 2–35.

Deutsche Biographie. (2014–2020). Historische Kommission bei der Bayerische Akademie der Wissenschaften, Bayerische Staatsbibliothek [Historical Commission at the Ba var ian Academy of Sci-ences and Humanities, Ba var ian State Library] [Online source]. Retrieved from https: / /www .deutsche -biographie .de

Dittmar, J., & Meisenzahl, R. R. (2016). State ca pac ity and pub lic goods: Institutional change, hu man cap i tal and growth in early mod ern Germany (CEP Discussion Papers No.1418). London, UK: Centre for Economic Performance, London School of Economics and Political Science.

Drüll, D. (2009). Heidelberger Gelehrtenlexikon 1933–1986 [Encyclopedia of Heidelberg’s schol ars 1933–1986]. Berlin, Germany: Springer.

Drüll, D. (2012). Heidelberger Gelehrtenlexikon 1803–1932 [Encyclopedia of Heidelberg’s schol ars 1803–1932]. Berlin, Germany: Springer.

Drüll-Zimmermann, D. (1991). Heidelberger Gelehrtenlexikon 1652–1802 [Encyclopedia of Heidelberg’s schol ars 1652–1802]. Berlin, Germany: Springer.

Drüll-Zimmermann, D. (2002). Heidelberger Gelehrtenlexikon 1386–1651 [Encyclopedia of Heidelberg’s schol ars 1386–1651]. Berlin, Germany: Springer.

Edwards, R. D. (2008). Declining mor tal ity among Brit ish sci en tists dur ing the Age of Enlightenment. Population and Development Review, 34, 103–125.

Elo, I. T., Martikainen, P., & Myrskylä, M. (2014). Socioeconomic sta tus across the life course and all -cause and cause-spe cific mor tal ity in Finland. Social Science & Medicine, 119, 198–206.

Fischer, K. F. A. (1978). Jesuiten-Mathematiker in der deutschen Assistenz bis 1773 [Je suit math e ma ti-cians in the Ger man Assistancy]. Archivum Historicum Societatis Iesu, 47, 159–224.

Flessa, D. (1969). Die Professoren der Medizin zu Altdorf von 1580–1809 [Doctoral dis ser ta tion, Univer-sität Erlangen-Nürnberg, Erlangen, Germany].

Flinn, M. W. (1981). The Eu ro pean de mo graphic sys tem 1500–1820 (Johns Hopkins Symposia in Compar-ative History). Baltimore MD: Johns Hopkins University Press.

Fornasin, A., Breschi, M., & Manfredini, M. (2010). Mortality pat terns of car di nals (six teenth-twen ti eth cen tu ries). Population, 65, 631–652.

Frijhoff, W. (1996). Patterns. In W. Rüegg (Ed.), A his tory of the uni ver sity in Europe: Universities in the early mod ern Europe (1500–1800) (Vol. 2, pp. 43–113). Cambridge, UK: Cambridge University Press.

Galor, O. (2011). Unified growth the o ry. Princeton, NJ: Princeton University Press.Günther, J. (1858). Lebensskizzen der Professoren der Universität Jena seit 1558 bis 1858: Eine Festgabe

zur 300-jährigen Säkularfeier der Universität am 15., 16., und 17. Au gust 1858 [Short bi og ra phies of pro fes sors from the uni ver sity of Jena 1558–1858: Fest schrift for the cel e bra tion of the 300th an ni ver-sary of the uni ver si ty, the 15th, 16th and 17th of Au gust 1858]. Jena, Germany: Mauke.

Hanley, J. A., Carrieri, M. P., & Serraino, D. (2006). Statistical fal li bil ity and the lon gev ity of popes: Wil-liam Farr meets Wilhelm Lexis. International Journal of Epidemiology, 35, 802–805.

Hobbs, F. (2004). Age and sex com po si tion. In H. S. Shryock, J. S. Siegel, & E. A. Larmon (Eds.), The meth ods and ma te ri als of de mog ra phy (2nd ed., pp. 125–174). Suitland, MD: U.S. Census Bureau.

Hollingsworth, T. (1977). Mortality in the Brit ish peer age fam i lies since 1600. Population, 32, 323–352.Human Mortality Database. (2019). Berkeley: University of California; and Rostock, Germany: Max Planck

Institute for Demographic Research. Retrieved from www .mortality .org or www .humanmortality .deJohansson, S. R. (1999). Death and the doc tors: Medicine and elite mor tal ity in Britain from 1500 to 1800.

Cambridge, UK: Cambridge Group for the History of Population and Social Structure.


https://www.deutsche-biographie.de

https://www.deutsche-biographie.de

http://www.mortality.org

http://www.humanmortality.de


Klinge, M. (2004). Teachers. In W. Rüegg (Ed.), A his tory of the uni ver sity in Europe: Universities in the nineteenth and early twen ti eth cen tu ries 1800–1945 (Vol. 3, pp. 123–162). Cambridge, UK: Cam-bridge University Press.

Langer, H. (2011). Die Universität Greifswald im Dreißigjährigen Krieg [The University of Greifswald dur ing the Thirty Years’ War]. In T. Kossert (Ed.), Universitäten im Dreißigjährigen Krieg [Universi-ties dur ing the Thiry Years’ War]. Potsdam, Germany: Potsdam University Press.

Lucas, R. E., Jr. (2009). Ideas and growth. Economica, 76, 1–19.Pfister, U., & Fertig, G. (2010). The pop u la tion his tory of Germany: Research strat egy and pre lim i nary

re sults (MPIDR Working Paper 2010-035). Rostock, Germany: Max Planck Institute for Demographic Research.

Schimank, U., & Winnes, M. (2000). Beyond Humboldt? The re la tion ship be tween teach ing and re search in Eu ro pean uni ver sity sys tems. Science and Public Policy, 27, 397–408.

Schumacher, R., & Oris, M. (2011). Long-term changes in so cial mor tal ity dif fer en tials, Geneva, 1625–2004. Explorations in Economic History, 48, 357–365.

Squicciarini, M. P., & Voigtländer, N. (2015). Human cap i tal and in dus tri al i za tion: Evidence from the Age of Enlightenment. Quarterly Journal of Economics, 130, 1825–1883.

Steiger, G. (1981). Magister und Scholaren, Professoren und Studenten: Geschichte deutscher Universitäten und Hochschulen im Überblick [Magisters and schol ars, pro fes sors and stu dents: An over view on the his tory of Ger man uni ver si ties]. Leipzig, Germany: Ura nia-Verlag.

van de Kaa, D. J., & de Roo, Y. (2007). The mem bers of the Royal Netherlands Academy of Arts and Sciences: 1808 to 2000. A de mo graphic view (Working pa per No. 2007/2). The Hague, the Nether-lands: Netherlands Interdisciplinary Demographic Institute.

Vandenbroucke, J. (1985). Survival and ex pec ta tion of life from the 1400’s to the pres ent: A study of the knight hood or der of the Golden Fleece. Amer i can Journal of Epidemiology, 22, 1007–1015.

Vandermeersch, P. A. (2003). Teachers. In H. De Ridder-Symoens (Ed.), A his tory of the uni ver sity in Europe: Universities in early mod ern Europe 1500–1800 (Vol. II, pp. 210–255). Cambridge, UK: Cambridge University Press.

van Poppel, F., Bijwaard, G., van Lieburg, M., van Lieburg, F., Hoekstra, R., & Verkade, F. (2016). The life ex pec tancy of med i cal pro fes sion als in the Netherlands, six teenth to twen ti eth cen tu ries. Population, 71, 619–640.

van Poppel, F., van de Kaa, D. J., & Bijwaard, G. E. (2013). Life ex pec tancy of art ists in the Low Countries from the fifteenth to the twen ti eth cen tu ry. Population Studies, 67, 275–292.

Vögele, J. (2000). Urbanization and the ur ban mor tal ity change in Imperial Germany. Health & Place, 6, 41–55.

Winkler-Dworak, M. (2008). The low mor tal ity of a learned so ci e ty. Eu ro pean Journal of Population/Revue Européenne de Démographie, 24, 405–424.

Woods, R. (2003). Urban-ru ral mor tal ity dif fer en tials: An un re solved de bate. Population and Development Review, 29, 29–46.

Robert Stelter (cor re spond ing au thor)stelter@demogr .mpg .de

Stelter • Max Planck Institute for Demographic Research, Rostock, Germany; and Faculty of Business and Economics, University of Basel, Basel, Switzerland

de la Croix • IRES/LIDAM, UCLouvain, Belgium; and CEPR, London, UK

Myrskylä • Max Planck Institute for Demographic Research, Rostock, Germany; and Center for Social Data Science, University of Helsinki, Finland


mailto:[email protected]

leaders and laggards in life expectancy among eu ropean ...demography doi the authors this is an...

Documents