“decessit sine prole” – childlessness, celibacy, and …...“decessit sine prole” –...

“Decessit sine prole” –Childlessness, Celibacy, and Survival of the Richest

in Pre-Industrial England

David de la Croix1 Eric Schneider2 Jacob Weisdorf3

1Université catholique de Louvain2London School of Economics3University of Southern Denmark

March 22, 2017

Introduction Theory Data Measuring the Four Margins Net reproduction Selection Implications Conclusion

What we do

There are two main demographic explanations of the Rise of the West

I European Marriage Pattern

I Evolutionary Advantage of the Rich

We assess the reality of their demographic features on Englishpre-industrial data (family reconstitution data)

Our contribution with respect to the literature: take all margins offertility into account, including celibacy and childlessness

2 / 40


European Marriage Pattern

Hajnal (1965) line from Saint Pe-tersburg to Trieste.

To the west of the line, late mar-riage, high celibacy rates

→ strong Malthusian preventivecheck

→ higher income per person

+ girls’ power (De Moor and Van Zanden 2010.)

Was Hajnal’s Western Europe marriage pattern instrumental to itseconomic success? disputed: Voigländer and Voth (AER 2013) againstDennison and Ogilvie (JEH 2014)

3 / 40


European Marriage Pattern and Preventive Check

from Clark’s book A Farewell to Alms, Princeton, 20074 / 40


Evolutionary Advantage of the Rich

In sum: the rich enjoyed higher fitness than the rest and their“capitalistic attitudes” spread as a result

Galor and Moav (QJE 2002): Take-off to modern growth can be explainedby an increase in the proportion of people in the population preferringquality (of children) over quantity

For their proportion to increase, need for an evolutionary advantage

Clark and Hamilton (JEH 2006): a rich father had 40% more kids than apoor one (2250 testators in England, c. 1600)same holds using Cambridge Group data (Boberg-Fazlic, Sharp, andWeisdorf. EREH, 2011)

→ the “capitalist” values spread through English society prior to theIndustrial Revolution

5 / 40


Theory

Purpose: analyze main components of net reproduction rate n(c).

Following Baudin, de la Croix and Gobbi (2016), we use the followingdecomposition:

n(c) = m(c) (1− z(c)) b(c) (1− d(c)) (1)

c: social class,m(·): marriage rate,z(·): fraction of childless married women,b(·): number of birth conditionally on having children,d(·): infant mortality rate.

(assumes singles do not have children, and fully homogamous marriage)

6 / 40


European Marriage Pattern with all Marginsage

45

15

% women0% 100%

child mortality

delaying marriage

celibacy

childlessness

birth b(c)

d(c)

z(c)

1−m(c)

adult mortality

7 / 40


New MarginsLiterature focuses on the intensive margin: b(c) (1− d(c))

and usually (1− d(c)) b′(c)− b(c) d′(c) > 0 i.e. evolutionaryadvantage to the high social class

Does it still hold when all margins are accounted for ?

n′(c) = (1− z(c))b(c)(1− d(c)) m′(c)︸︷︷︸marriage margin

−m(c)b(c)(1− d(c)) z′(c)︸︷︷︸childlessness margin︸︷︷︸

Extensive margins of fertility

+ m(c)(1− z(c)) ((1− d(c)) b′(c) − b(c) d′(c))︸︷︷︸intensive margin of fertility

.

one needs to estimate z(c), m(c), b(c) and d(c) from data.8 / 40


Parish registers

Parish registers were formally introduced in England on 5 September1538 following the split with Rome

Injunction requiring the registers of baptisms, marriages and burials tobe kept

Civil registration of births, marriages, and deaths for England and Walesbegan on 1 July 1837

9 / 40


Family reconstitution data

Collected by the CambridgeGroup for History of Populationand Social Structure, the full setof family reconstitution dataincludes over 300,000 individualsrecorded in registers comingfrom a total of 26 provincial,English parishes.

These parishes were chosen bythe Cambridge Group because ofthe high quality of data.

The full data cover 1541 to 1871

We take a subsample restricted to the reliable period (parish specific)

10 / 40


1733

bapt. wife 1 1766marriage

Inn Holder

1770

bur. wife 1

1766Elizabeth burial 1768

1770marriage

Victualler

burial 1808Printer

1802

bur. wife 2

1771

1773

1775

1777

1778

1782

1786

John

John

Elizabeth Ann

Sarah

Ann

Thomas

William

?

?

?

?

burial 1772

mar. 1794

burial 1777

burial 1780

mar 1812 burial 1820

burial 1786

InnKeeper

ElizabethNicholls

Elizabeth Treadwell

John Cheney

Parish of Banbury

In 1967 the renowned business family ofCheney decided to mark 200 years ofprinting in Banbury by publishing a bookto celebrate this remarkable milestone.

The story begins at the Unicorn wherethe first John Cheney was both printerand publican.

Johns time at the Unicorn ended in 1788when he moved the printing business to ashop in Red Lion Street.

11 / 40


Occupations

Use Clark and Cummins (Jpop 2015) classification

1 Labourers/Servants incl. seamen2 Husbandmen small farmers, weavers3 Craftsmen tailors, carpenters4 Traders innkeepers, butchers, bakers5 Farmers6 Merchants/Professionals clerks, medical, printers7 Gentry gentlemen, esquire0 Unknown

Observed from baptism registers (occupation of father), marriage, burial.

12 / 40


Occupations (2)9 per cent of the father’s occupations changed over life cycle.Assign maximum social group if reasonable. Assign unknown otherwise.

1 2 3 4 5 6 71 17666 1087 624 330 187 229 37

87.6% 5.4% 3.1% 1.6% 0.9% 1.1% 0.2%2 0 11827 463 168 230 139 28

92.0% 3.6% 1.3% 1.8% 1.1% 0.2%3 0 0 14440 368 38 236 43

95.5% 2.4% 0.3% 1.6% 0.3%4 0 0 0 7904 129 578 96

90.8% 1.5% 6.6% 1.1%5 0 0 0 0 3528 78 91

95.4% 2.1% 2.5%6 0 0 0 0 0 3516 232

93.8% 6.2%7 0 0 0 0 0 0 1581

100%

Max occupation× Min occupation13 / 40


Methodological contribution

The standard in the literature is to estimate fertility with a restrictedsample of women surviving past age 40 or 50

Problem is not only that many observations are lost, but also thecontributions to fertility of those with lower survival times are dismissed.

To overcome this, we propose Cox proportional hazards models forcensored data (early death) to compute the predicted fertility rates fromthe corresponding survival curves.

14 / 40


Marriage Rate m(c) - Sample

- Cox proportional hazard model estimating the risk of marriage ofthe female population aged 16+

- Right censoring of participants (death). Cox allows to use theinformation from those who died early.

- Sample of 8,611 individuals. Mean age at death is 46.09 years.Marriage rate: 58.6%

- Requirement: observe birth and death in the parish (no permanentmigration)

- Married = either married in the parish or child baptized in the parish

15 / 40


Marriage Rate m(c) - Method

- Period is divided in four subperiods of equal length→ timedummies (fct. marriage date of parents)

- Inclusion of 26 parish dummies

- Reference category: Labourers in parish 1, first subperiod

- Cox implies the baseline hazard rate (the risk of getting married) isshifted proportionally by the occupation of the father

16 / 40


Descriptive stat. - risk of getting married

20 25 30 35 40

0.02

0.04

0.06

0.08

0.10

age

haza

rd r

ate

Labourers, Servants & Husbandmen (solid line) vs Merchants,Professionals & Gentry (dashed line)

17 / 40


Marriage Rate m(c) - Estimation

Dependent variable: Proba. marryingcoef Haz. ratio ρ

Labourer & servants 1.00Husbandmen −0.085 0.92 0.01Craftsmen −0.193∗∗∗ 0.82 0.01Traders −0.181∗∗ 0.83 0.00Farmers 0.081 1.09 -0.02Merchants/Professionals −0.325∗∗∗ 0.72 -0.00Gentry −0.331∗∗ 0.72 0.01

UnknownOccupation −0.093∗ 0.91 0.00

Observations 8,6114 Period dummies yes26 Parish dummies yesNote: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01

18 / 40


Marriage Rate m(c) - ResultCompute the survival function (i.e. celibacy rate) at 45

m(c)

0.6

0.65

0.7

0.75

0.8

0.85

Unknown Labourers,Servants

Husbandmen Craftsmen Traders Farmers Merchants,Professionals

Gentry

19 / 40


Childlessness rate z(c) - Sample

Cox proportional hazard models to estimate the risk of a married couplehaving a first birth.

Couples become at risk of a first birth upon marriage and censoringoccurs when the mother dies.

We limit our sample to couples whose burial dates are known to ensurethat we do not attribute childlessness to couples that migrated out ofthe parish.

Restrict the sample to first marriages

Statistic N Mean St. D. Min Q(25) Med. Q(75) Max

At risk for 1st birth 14,730 0.850 0.357 0 1 1 1 1Spacing 12,517 1.37 1.35 0.00 0.72 0.94 1.52 18.01

20 / 40


Descriptive stat. - risk of having a first birth

0 1 2 3 4

0.0

0.5

1.0

1.5

2.0

age

haza

rd r

ate

Labourers, Servants & Husbandmen (solid line) vs Merchants,Professionals & Gentry (dashed line)

21 / 40


Childlessness rate z(c) - Estimation

Dependent variable: Proba. having first birthcoef Haz. ratio ρ

Labourer & servants 1.00Husbandmen −0.030 0.97 0.01Craftsmen −0.013 0.99 0.00Traders −0.038 0.96 0.01Farmers −0.203∗∗∗ 0.82 0.00Merchants/Professionals −0.185∗∗∗ 0.83 0.02Gentry −0.249∗∗∗ 0.78 0.02

UnknownOccupation −0.284∗∗∗ 0.75 0.00

Observations 14,7304 Period dummies yes26 Parish dummies yes

22 / 40


Childlessness rate z(c) - ResultCompute the survival function 10 years after marriage

z(c)

0.05

0.07

0.09

0.11

0.13

0.15

0.17



Gentry

23 / 40


Possible Reasons for High Childlessness

Mean age at marriage ? (entry sterility) No.

Bias in occupations as observed from baptism record

Inheritance practices among the wealthy (Gobbi and Goñi)

Syphillis more prevalent for Gentry. But would it affect P0 ?

Arranged/Loveless marriages among the wealthy

Husband was constantly traveling to oversee estates ?

Consanguinity leading to infertility?

Fear of deforming their bodies, of maternal mortality, of the burden of apregnant belly?

24 / 40


Mothers’ fertility b(c)

Parity progression ratios Pn =Proba to have n + 1 kids | having n

Mothers’ Completed Fertility = 1 + P1

(1 + P2

(1 + P3 . . .

(1 +

Pn1− Pn

)))But we are not only interested in completed families. Hence

b(c) = s0

(1 + s1P1

(1 + s2P2

(1 + s3P3 . . .

(1 +

snPn1− snPn

))))

where sn is the probability to survive after the nth kid.

→ need to estimate all the parity progression ratios Pn

25 / 40


Mothers’ fertility b(c) - Estimation

Dependent Proba. Proba. Proba. Proba. Proba.variable: 2sd child 3rd child 4th child 5th child 6th child

Labourers & ServantsHusbandmen 0.099∗∗ 0.070 0.051 0.087 0.049Craftsmen 0.126∗∗∗ 0.047 0.100∗∗ 0.166∗∗∗ 0.149∗∗

Traders 0.183∗∗∗ 0.166∗∗∗ 0.231∗∗∗ 0.280∗∗∗ −0.015Farmers 0.196∗∗∗ 0.117∗ 0.164∗∗ 0.043 −0.032Merchants/Professionals 0.231∗∗∗ 0.189∗∗∗ 0.207∗∗∗ 0.155∗ 0.132Gentry 0.119 0.157 0.385∗∗∗ 0.258∗∗ 0.385∗∗∗

UnknownOccupation −0.069∗∗ −0.102∗∗∗ −0.056 −0.019 −0.029Observations 12,519 10,738 8,993 7,280 5,679

26 / 40


Mothers’ fertility b(c) - Estimation

Dependent Proba. Proba. Proba. Proba.variable: 7th child 8th child 9th child 10th child

Labourers & ServantsHusbandmen −0.068 0.103 0.215 −0.050Craftsmen 0.022 0.007 0.269∗∗ 0.050Traders 0.233∗∗ −0.083 0.305∗∗ 0.327∗

Farmers 0.110 0.069 −0.055 −0.287Merchants/Professionals 0.219∗ 0.233∗ 0.481∗∗∗ 0.004Gentry 0.375∗∗ 0.462∗∗ 0.699∗∗∗ −0.224

UnknownOccupation −0.069 0.074 0.212∗∗ −0.116Observations 4,201 2,959 1,932 1,154

27 / 40


Mothers’ fertility b(c) - Results

b(c)

3.000

3.200

3.400

3.600

3.800

4.000

4.200

4.400

4.600

4.800



Gentry

28 / 40


Child mortality d(c) - Sample

Cox proportional hazard model.

Exposure for children begins from birth, events are closed if a child diesand right censoring occurs after age 15.

we must observe the child’s birth year and their mother’s year of death.

require that the mother’s year of death is 15 years after the child’s birthto ensure that the family did not move away from the parish before thechild reached 15.

Children who we do not observe dying are assumed to have survived toage 15.

29 / 40


Child mortality d(c) - Estimation

Dependent variable: Proba. of death

Labourers & ServantsHusbandmen −0.006Craftsmen 0.027Traders 0.045Farmers −0.008Merchants/Professionals 0.054Gentry 0.043

Unknown Occupation −0.003Observations 48,8654 Period dummies yes26 Parish dummies yes

30 / 40


Net ReproductionCompare net reproduction with the intensive margins only:

n̄(c) = m̄ (1− z̄) b(c) (1− d(c))

to net reproduction with all margins:

n(c) = m(c) (1− z(c)) b(c) (1− d(c))

m 1− z b 1− d n n̄Labourers, Servants 0.80 0.92 3.91 0.75 2.14 1.95Husbandmen 0.77 0.91 4.00 0.75 2.09 1.99Craftsmen 0.73 0.91 4.20 0.74 2.08 2.07Traders 0.74 0.91 4.53 0.74 2.24 2.23Farmers 0.82 0.87 4.12 0.75 2.21 2.06Merchants/Professionals 0.68 0.87 4.57 0.74 2.01 2.24Gentry 0.68 0.86 4.60 0.74 1.98 2.26Unknown 0.77 0.85 3.50 0.75 1.70 1.74Mean 0.76 0.87 3.77 0.75 1.87 1.87

31 / 40


Net Reproduction

1.900

2.000

2.100

2.200

2.300

Labourers,Servants


Gentry

32 / 40


Selection issue

Three types of people

stayers (birth and death known)

in-migrants (birth certificate missing)

out-migrants (death certificate missing)

m(c) is based on stayers

b(c), z(c), and d(c) are based on stayers and in-migrants

33 / 40


Selection bias in m(c)

Potential problems:(i) different migration rates for women across the social groups(ii) women migrating from a particular group more likely to remain celibate

for (ii), we cannot do anything. For (i), compute emigration rate by group:

Emigration rates Linear Probability Model

Son of Daughter of Son of Daughter of

Labourers/Servants 58.5% 60.4% Reference groupHusbandmen 53.1% 58.1% −0.021∗∗ −0.001Craftsmen 52.8% 57.5% −0.043∗∗∗ −0.014∗Traders 56.1% 60.0% −0.033∗∗∗ −0.013Farmers 54.2% 61.6% −0.038∗∗∗ 0.001Merchants/Professionals 53.6% 57.0% −0.034∗∗∗ −0.019Gentry 46.7% 56.8% −0.054∗∗∗ 0.008UnknownOccupation 53.5% 57.0% −0.011 −0.011Observations 65,183 62,453

34 / 40


Selection bias in b(c), z(c), and d(c)

We can do more, assuming out-migrants are similar to in-migrants

We can compare in-migrants to stayers, and see if the social gradientdiffers

Example: To measure childlessness we used 14,730 couples with knownmarriage date and death date. Among them, 4,125 couples werein-migrants into the parish.

We introduce a dummy for in-migrant, and interact it with social groups

Conclusion: in-migrants are more likely to be childless, have lower parityprogressions, and have higher child mortality, but this is true for allsocial groups→ no bias in social gradient

35 / 40


Cox Model. Analysis of z(c): non-migrants vs in-migrants

Benchmark Model with in-migrant interactionsgroup in-migrant

dummies interactions

In-migrant Dummy -0.275∗∗∗

Husbandmen −0.030 −0.047 0.041Craftsmen −0.013 −0.050 0.103Traders −0.038 −0.069 0.076Farmers −0.203∗∗∗ −0.250∗∗∗ 0.130Merchants/Professionals −0.185∗∗∗ −0.238∗∗∗ 0.191Gentry −0.249∗∗∗ −0.337∗∗∗ 0.318Unknown Occupation −0.284∗∗∗ −0.305∗∗∗ 0.069

36 / 40


Implications for Growth

The different features of the West European marriage pattern were useddifferently even within the boundaries of a single economy.

Thomas Gainsborough – “Mr and Mrs Andrews”

Gentry: Highcelibacy &childlessnessrates

37 / 40


Labourers:Low birthrate withinmarriage

Ford Madox Brown – “Work”

Middle class: Traders, farmers: higher net reproduction rate

38 / 40


The evolutionary advantage of the middle class, growth promoting?

- Lower classes unable to invest in growth-enhancing education oftheir children

- Upper classes, especially the landed elites, relied on rental incomefrom their estates, did not participate in activities that fosteredeconomic growth (Doepke and Zillibotti QJE 2008)

- Less than three per cent of English industrialists were ofupper-class origin (Crouzet 1985).

- 85 per cent of industrialists came from a middle-class family.

39 / 40


Conclusion

Previous work: evolutionary advantage of the pre-industrial elites

This study: the middle classes outperformed the upper classes.

This reversal arises because of higher singleness and childlessness.

Given that an overwhelming majority of English industrialists came fromthe middle class, its evolutionary advantage might have been growthpromoting.

40 / 40

“decessit sine prole” – childlessness, celibacy, and …...“decessit sine prole” –...

Documents