Malthusian Remittances:

The cultural diffusion of the fertility transition –

Evidence from internal migration in 19th century France

Guillaume Daudin

Paris-Dauphine & Sciences Po (OFCE)

Raphaël Franck

Bar-Ilan University & Marie Curie Fellow at Brown University

Hillel Rapoport

Paris School of Economics and Migration Policy Center, EUI

Migration Forum, EUI, 5 February 2016

1. Introduction

The decline in fertility rates started everywhere in Europe in the second half of the 19th c. (rather, after 1880)

– Shift from the Malthusian to the Modern demographic and growth regime (Lee, 2003, Galor, 2011)

L’exception culturelle française – the French fertility puzzle

– France was the first country where the demographic transition occurred (started after 1750, continued during the 19th c) (Guinnane, 2011, Weir, 1994).

Coale fertility rates

Fertility rate across French départements

Source: (Bonneuil 1997) and authors' computations for 1911.

The French case

The decline in fertility is seen as a convergence between départements

– The mean Coale Fertility Index declined from 0.4 in 1811 to 0.24 in 1911.

– Its standard deviation declined from 0.11 to 0.04 (from 27% of the mean to 17% of the mean).

Specificities of France in 19th c.

– France was not fully integrated from a cultural perspective at the time of the revolution and until the middle of the 19th century (Weber, 1976).

– Importance of internal migration, unlike in countries (e.g., Italy, Sweden) where migrants moved to the New World (Hatton and Williamson, 1998). Yearly annual gross emigration rates, 1860-1913 (Hatton and Williamson, 1998), per thousand inhabitants: France 0.18, Germany 1.5, Great Britain 4.61, Italy 9.25.

Why is this important? 1. Different self-selection patterns into migration according to fertility

behavior: “International rural-to-rural migration … required stable families with large numbers of children. Families of that sort … were well suited to the destination countries where land was abundant and so a large family of workers an advantage. Similarly advantageous were the traditional social and family values of those migrants. [On the contrary] Migration from the countryside to cities and industrial regions, where workers were employed primarily as wage workers in manufacturing and construction, favored instead a different profile, namely, … individuals whose family ties were looser, … nuclear families able to carefully plan births”. Livi Bacci, M. (2012): “A Short History of Migration”, pp. 54-55.

2. Different fertility norms prevail at destination: the “New World” is a high land-to-labor ratio destination with high fertility, while industrial, urban internal destinations are low-fertility regions; different adaptative strategies for immigrants, different “social remittances” sent back home.

The role of internal migration

The transmission of cultural norms through migration

– Immigrants transmit their cultural norms to their children (Fernandez and Fogli, 2006, Blau et al., 2011) but in general “assimilation” is taking place;

– Emigrants have a cultural influence on their country of origin: “Fargues’ conjecture” (Fargues, 2007, comparative analysis of Morocco v. Egypt; Beine, Docquier and Schiff 2013 for cross-country comparisons, Bertoli and Marchetta 2015, econometric analysis for Egyptian households).

The convergence in the French fertility rate had a strong spatial dimension (Murphy, 2015)

– Spatial regressions show that factors linked to travel costs (rather than to distance per se) contributed to the decline in fertility.

Empirical strategy

We seek to estimate the effect of internal migration on fertility

Identification strategy: exogenous variations in the bilateral travel costs between French départements entailed a time-varying decrease in travel costs and had a positive effect on migration.

Substantial anecdotal evidence, confirmed by our falsification tests, suggests the railroad network was developed independently from fertility patterns and migration choices (Caron, 1997).

Summary of Results

2. Data

Panel of 87 départements every ten years between 1861 and 1911.

Dependent variable:

Coale fertility index f of each département (Bonneuil, 1997)

� = �∑ ���.�� ��� �∑ �� .����� ⁄ (1)

��� is the number of women in age group k

��� is the rate of childbearing among women in the kth age interval

Hk represents the fertility rates observed for the Hutterites

The Coale Fertility Index f is the ratio of the number of observed

births to the number of births if all women had Hutterite fertility.

Explanatory variables: "Fertility Norms"

1. Mean fertility of immigrants’ birth département (weighted by stocks in the single host département)

2. Mean fertility of emigrants’ host département (weighted by stocks in the various host départements)

To compute the norms, we need bilateral migration data:

How many people born in Rhône live in Seine?

(and vice-versa)

“TRA” dataset (Enquête des 3000 familles -Survey of the 3000 Families)

– This dataset allows us to compute migrations every ten years from 1861 to 1911

– Whole population of people whose family name starts by “TRA”

• We know where they were born and where they died

- This may not be representative of the département-level

• We apply a marginal standardization algorithm (identical results with the cross-entropy method) to compute census-compatible bilateral flows

For comparison purposes, the French Census which provides

– bilateral data for 1891, 1901 and 1911

– the number of domestic immigrants in each département in 1861, 1872 and 1881 (and the number of emigrants can be computed from the difference between natural and actual population growth), but no disaggregation by department of origin

Bilateral migrant stocks, 1891 (census)

Bilateral migrant stocks, 1891 TRA

Explanatory variables: control variables

(source: successive French censuses, except where noted)

Life Expectancy at Age 15 (Source: Bonneuil, 1997)

Infant Mortality (under age 1, in %) (Source: Bonneuil, 1997)

Urban (% residents living in towns with more than 2,000 inhabitants)

Industries (% of the working population in the industrial sector)

Professionals (% of lawyers, doctors, etc… in the working population)

Explanatory variables: control variables

(source: French Census)

Female Education (% 5-19 year old females in school)

Male Education (% 5-19 year old males in school)

Share of girls in Catholic primary schools (in %, out of the number of girls in Catholic and secular primary schools)

Share of boys in Catholic primary schools (in %, out of the number of boys in Catholic and secular primary schools)

"Bad controls"

(source: successive French censuses, except where noted)

Variables of informational transmission

Total Number of Periodicals (Source: Annuaire de la Presse Française)

Share of Children Born out of Wedlock out of the Total Number of Births

Share of "Not Legitimized" Children out of Children Born out of Wedlock

Variables of cultural transmission

Share of Married Men Age 20-24 (t)

Share of Married Men Age 25-29 (t)

Share of Married Men Age 30-34 (t)

Share of Married Men Age 35-39 (t)

Share of Married Women Age 20-24 (t)

Share of Married Women Age 25-29 (t)

Share of Married Women Age 30-34 (t)

Share of Married Women Age 35-39 (t)

3. Econometric methodology

The baseline model estimates the relation between the fertility of the

inhabitants of a département and the fertility norms of immigrants and

emigrants, controlled by socio-economic variables.

log���, � =�1. log������,�� + �2. � !�"#�,�� + �3. � !�"#�,��. log������,�� +�4. log�&'���,�� + �5. � !��#�,��+�6. � !��#�,��. log�&'���,�� +*. log�,i,t + αi + αt + 0�,� (1)

�1,2 fertility rate in département i in year t, i.e., the inhabitants' residence norm

ERFN is the emigrants’ residence fertility norm

����1,2 = 34516,2 . �6,2671 8 34516,2671 89

IBRN is the immigrants' birthplace fertility norm.

&'��1,2 = 34561,2. �6,2671 8 34561,2671 89

• 516,2 people born in départment i living in département j in year t,

• 561,2 people born in départment j living in département i in year t,

• �6,2 the fertility in département j at time t

The share of emigrants "#1,2 in proportion of the population of département i

"#1,2 = :4 516,2671 ; <1,2=

The share of immigrants �#1,2 among inhabitants of département i

�#1,2 = :4 561,2671 ; <1,2=

516,2 people born in départment i living in département j in year t,

561,2 people born in départment j living in département i in year t,

<1,2 the total population of département i in year t.

Some remarks on the specification

The Princeton Project vs. Brown and Guinnane (2007).

Correction of unobserved heterogeneity

• We use panel equations with time and département fixed effects, using intra-département variability

Functional form (Brown and Guinnane, 2007)

• Does the effect of migrants depend on their relative numbers?

o We use interaction variables

• Should the regression be run in levels or in difference?

o In levels

Identification strategy

To overcome potential endogeneity in the estimation of Eq. (2), we use travel costs as an instrumental variable.

A decrease in travel costs should have a positive effect on migration and should, therefore, increase the stocks of migrants.

Travel costs were time-varying during the 19th century because of the development of the railroad network throughout the country.

• The development of the railroad was not linked to cultural proximity (L'étoile de Legrand).

• Reverse causality and self-selection?

L'étoile de Legrand

Identification strategy

These transport costs are computed in a four-step procedure.

1) we use Caron (1997)'s rail network map to determine the available travel (railroad, road, sea) links between adjacent départements.

2) we compute the great-circle distance between the administrative centres (chef-lieu) of adjacent départements.

3) since rail prices were regulated by the State (Toutain, 1967, p.277), there was a constant road or rail price per kilometer throughout France and this strategy provides the travel cost between adjacent départements.

4) we apply a short-route finding algorithm taken from the UCINET network analysis program (Borgatti et al., 2006) to compute the cheapest route and hence the travel costs between each département.

First stage regression

We use 20 and 30-year lagged transport costs because the mean migrant age was (in the TRA dataset) 38 years old in 1861, 40 in 1872, 41 in 1881, 43 in 1891, 45 in 1901 and 50 in 1911, i.e. between 20 and 30 years after migration.

We estimate a panel theoretical gravity model with the standard Poisson Pseudo Maximum Likelihood that solves for heteroskedasticity and the existence of zero migrant stocks (Silva and Tenreyro, 2006) log�516,2� = � + * log��>�?@A >�B @�@2CDE + B log��>�?@A >�B @�@2CFE +βH + βI + βJ + 0 (5)

where 516,2 are the migrant stocks, while βt, βo and βd are fixed effects for

year, origin- départements and destination- départements.

Bilateral predicted migrant stocks 1891

4.Results

Table 1: Travel costs and migration: first stage regressions

All

migrants

Male

migrants

Female

migrants #16,2 the stock of migrants born in départment i living in départment j at time t

20-Year Lagged Travel Costs -0.9***

(0.06)

-0.8***

(0.08)

-0.9***

(0.07)

30-Year Lagged Travel Costs -0.6***

(0.05)

-0.7***

(0.06)

-0.5***

(0.06)

Year fixed effects Yes Yes Yes

Origin-département & destination-département fixed effects Yes Yes Yes

Pseudo R2 0.60 0.53 0.55

Number of observations 43,690 43,690 43,690

Number of clusters 7310 7310 7310

Note: Robust standard errors clustered at the origin- département & destination-département are reported in brackets. *** indicates significance at the 1%-level, ** indicates significance at the 5%-level, * indicates significance at the 10%-level.

Table 2: Determinants of the fertility decline in France, 1861-1911

(1) (2) (3)

OLS OLS w/ control

variables IV w/ control

variables

Fertility (t)

Emigrants' Residence Norm (t) 0.539*** 0.397*** 0.788**

[0.136] [0.107] [0.302]

Immigrants' Birthplace Norm (t) 0.0631 -0.149 0.378

[0.126] [0.0937] [0.287]

Emigrants' Residence Norm (t) * Share of Emigrants(t) -1.655*** -1.300** -2.990***

[0.620] [0.530] [0.788]

Immigrants' Birthplace Norm (t)* Share of Immigrants (t) 2.288** 3.113*** 1.104

[0.981] [0.721] [0.806]

Share of Emigrants (t) -2.800*** -2.025** -3.973***

[1.045] [0.819] [1.200]

Share of Immigrants (t) 4.273*** 4.911*** 1.818*

[1.305] [0.981] [1.000]

Life Expectancy Age 15 (t) -0.00871 -0.00855

[0.00912] [0.0106]

Infant Mortality (t) 0.684** 0.649*

[0.299] [0.346]

Urban (t) -0.0926 0.152

[0.320] [0.298]

Industries (t) -0.0122 -0.00283

[0.00745] [0.00697]

Professionals (t) -0.0137 -0.00641

[0.0134] [0.0126]

Female Education (t) -0.0472 -0.0209

[0.0414] [0.0401]

Male Education (t) 0.00908 0.00840

[0.0484] [0.0500]

Share of Girls in Primary Catholic Schools (t) 0.0111 0.0172

[0.0181] [0.0213]

Share of Boys in Primary Catholic Schools (t) -0.000696 0.00441

[0.0156] [0.0154]

Observations 486 486 486

Child mortality significantly lowers fertility.

Immigrants have no consistent effect on fertility, either in terms of sign and statistical significance.

• Immigrants, who moved from high- to low-fertility areas, did not prevent the decline of fertility in their destination départements.

Emigrants had a consistent impact on the decline of fertility in the origin département (counter-intuitive as goes against self-selection), and more so if they went to low-fertility destinations.

• Our interpretation: why would destination matter? Transmission of information and cultural transmission of low-fertility norms.

Tests of the interpretation: placebo tests, plausibility checks (Paris)

Robustness Checks

Table 3. Migration in 1861-1911 and fertility decline 1811-1861 (placebo) OLS OLS w/ controls

variables IV w/ controls

variables Fertility (t-50) Emigrants' Residence Norm (t) 0.0061 -0.164 -0.335

[0.147] [0.142] [0.304] Immigrants' Birthplace Norm (t) -0.0680 -0.101 -0.134

[0.146] [0.142] [0.352] Emigrants' Residence Norm (t) * Share of Emigrants(t) 0.416 0.827 -0.0261

[0.930] [0.865] [1.129] Immigrants' Birthplace Norm (t)* Share of Immigrants (t) 0.0496 0.0918 0.885

[1.100] [1.111] [1.009] Share of Emigrants (t) 1.028 1.696 -0.0403

[1.514] [1.373] [1.590] Share of Immigrants (t) 0.292 0.00920 0.159

[1.409] [1.379] [1.356]

Within R2 0.5 0.57 0.59

Control variables No Yes Yes

Year-fixed effects Yes Yes Yes

Département-fixed effects Yes Yes Yes

Observations 486 486 486

Table 4. Newspapers, out-of-wedlock births, age at marriage and the fertility decline, 1861-1911 (exclusion restriction)

OLS w/ controls

OLS w/ controls

OLS w/ controls

IV w/ controls

IV w/ controls

IV w/ controls

Fertility(t)

Emigrants' Residence Norm (t) 0.388*** 0.386*** 0.395*** 0.783** 0.775** 0.796**

[0.108] [0.112] [0.102] [0.301] [0.299] [0.304]

Immigrants' Birthplace Norm (t) -0.138 -0.148 -0.153 0.391 0.384 0.334

[0.0940] [0.0956] [0.0992] [0.286] [0.284] [0.298]

Emigrants' Residence Norm (t) * Share of Emigrants(t) -1.277** -1.266** -1.311** -2.981*** -2.945*** -3.054***

[0.543] [0.542] [0.516] [0.788] [0.792] [0.771]

Immigrants' Birthplace Norm(t)* Share of Immigrants(t) 2.967*** 3.115*** 3.185*** 0.930 1.076 1.180

[0.718] [0.720] [0.812] [0.760] [0.821] [0.773]

Share of Emigrants (t) -1.993** -1.978** -2.014** -3.978*** -3.905*** -4.064***

[0.835] [0.846] [0.795] [1.199] [1.204] [1.178]

Share of Immigrants (t) 4.725*** 4.916*** 5.041*** 1.612* 1.786* 1.929**

[0.977] [0.979] [1.109] [0.921] [1.016] [0.942]

Continuing next page

Total Number of Periodicals (t) -0.0188 -0.0236 [0.0163] [0.0147]

Share of Children Born out of Wedlock -0.0599 -0.0951* out of the Total Number of Births [0.0593] [0.0502]

Share of Not Legitimized Children -0.00257 -0.0206 out of Those who were Born out of Wedlock [0.0473] [0.0400]

Share of Married Men Age 20-24 0.101 -0.0155 [0.163] [0.155]

Share of Married Women Age 20-24 0.102 0.125 [0.144] [0.131]

Share of Married Men Age 25-29 -0.118 -0.131 [0.129] [0.135]

Share of Married Women Age 25-29 0.0546 0.0828 [0.187] [0.174]

Share of Married Men Age 30-34 -0.253* -0.276** [0.133] [0.133]

Share of Married Women Age 30-34 -0.242 -0.191 [0.208] [0.191]

Share of Married Men Age 35-39 0.384* 0.427* [0.214] [0.226]

Share of Married Women Age 35-39 -0.0552 -0.0755 [0.276] [0.260]

Within R2 0.8 0.75 0.76 0.8 0.76 0.77

Year-fixed effects Yes Yes Yes Yes Yes Yes

Département-fixed effects Yes Yes Yes Yes Yes Yes

Observations 486 486 486 486 486 486

Table 5: Determinants of the fertility decline in France – spatial regressions

OLS OLS w/ controls IV w/ controls Fertility (t) Emigrants' Residence Norm (t) 0.468*** 0.358*** 0.646***

[0.0964] [0.0996] [0.168] Immigrants' Birthplace Norm (t) 0.0227 -0.134 0.233

[0.115] [0.0960] [0.196] Emigrants' Residence Norm (t) * Share of -1.892*** -1.376** -2.859***

[0.492] [0.535] [0.728] Immigrants' Birthplace Norm (t)* Share of 2.110** 2.572*** 0.850

[0.904] [0.760] [0.693] Share of Emigrants (t) -3.098*** -1.826** -3.685***

[0.793] [0.803] [1.086] Share of Immigrants (t) 3.993*** 4.482*** 1.772**

[1.223] [1.019] [0.876] ρ 2.630*** 2.492*** 1.703***

[0.302] [0.335] [0.536] σ

2 0.00474*** 0.00337*** 0.00330*** [0.000544] [0.000343] [0.000330]

Within R2 0.6 0.7 0.8 Log-pseudolikelihood 607.6 690.6 698.0 Control variables No Yes Yes Year-fixed effects Yes Yes Yes Département-fixed effects Yes Yes Yes Number of clusters 81 81 81 Observations 486 486 486

5 Channels of the fertility decline

Which migrants channeled information that contributed to the fertility decline?

• Female and male migrants;

• Migration from and to Paris.

Two strategies:

1) We run regressions on separate samples (only male migrants, only female migrants, excluding migration from and to Paris)

2) We compute the counterfactual values of the fertility rate in each département under the assumption that no changes in fertility norms and in the shares of migrants had occurred after 1861.

Table 6: The fertility decline in France, 1861-1911: male norms and migration

OLS OLS w/ controls IV w/ controls Fertility (t) Emigrants' Residence Norm (t) 0.478** 0.365*** 0.822***

[0.118] [0.0920] [0.293] Immigrants' Birthplace Norm (t) 0.0500 -0.136* 0.329

[0.109] [0.0793] [0.276] Emigrants' Residence Norm (t) * Share of -1.248** -1.043** -2.851***

[0.534] [0.450] [0.724] Immigrants' Birthplace Norm (t)* Share of 1.239 2.108*** 0.830

[0.854] [0.604] [0.847] Share of Emigrants (t) -2.094** -1.588** -3.809***

[0.830] [0.628] [1.109] Share of Immigrants (t) 2.688** 3.520*** 1.518

[1.156] [0.808] [1.084] Control variables No Yes Yes Year-fixed effects Yes Yes Yes Département-fixed effects Yes Yes Yes Number of clusters 81 81 81 Observations 486 486 486

Table 7: The fertility decline in France, 1861-1911: female norms and migration

(1) (2) (4) OLS OLS w/ controls IV w/ controls

Fertility (t) Emigrants' Residence Norm (t) 0.493*** 0.294*** 0.686**

[0.136] [0.103] [0.304] Immigrants' Birthplace Norm (t) 0.00404 -0.151* 0.474

[0.120] [0.0834] [0.299] Emigrants' Residence Norm (t) * Share of Emigrants(t) -1.980*** -1.187* -2.891***

[0.685] [0.603] [0.784] Immigrants' Birthplace Norm (t)* Share of Immigrants (t) 2.595*** 3.126*** 1.348*

[0.968] [0.684] [0.780] Share of Emigrants (t) -3.362*** -1.794* -3.804***

[1.201] [1.032] [1.192] Share of Immigrants (t) 4.655*** 4.803*** 2.068**

[1.267] [0.928] [0.952] Control variables No Yes Yes Year-fixed effects Yes Yes Yes Département-fixed effects Yes Yes Yes Number of clusters 81 81 81 Observations 486 486 486

Table 8: The fertility decline in France, 1861-1911, excluding migration to and from Paris

OLS OLS w/ controls IV w/ controls

Fertility (t)

Emigrants' Residence Norm (t) 0.399*** 0.293*** 0.467

[0.129] [0.105] [0.406]

Immigrants' Birthplace Norm (t) 0.0850 -0.162 0.499

[0.158] [0.112] [0.371]

Emigrants' Residence Norm (t) * Share of Emigrants(t) -0.557 -0.840 -2.634*

[0.837] [0.654] [1.520]

Immigrants' Birthplace Norm (t)* Share of Immigrants (t) 2.337* 3.339*** 2.055*

[1.283] [0.878] [1.156]

Share of Emigrants (t) -1.172 -1.390 -3.524

[1.321] [0.954] [2.332]

Share of Immigrants (t) 4.136** 5.102*** 3.518**

[1.780] [1.241] [1.737]

Control variables No Yes Yes Year-fixed effects Yes Yes Yes Département-fixed effects Yes Yes Yes Number of clusters 80 80 80 Observations 480 480 480

Panel 3a. Counterfactual fertility in France under no male and female migration

Panel 3b. Counterfactual fertility in France with only male migration

Panel 3c. Counterfactual fertility in France with only female migration

Panel 3d. Counterfactual fertility in France excluding migration from and to Paris

Summary

Actual fertility index in 1861=0.310; Actual fertility index in 1911=0.244

The standard deviation drops by 50% between 1861 and 1911.

Under no migration of men and women after 1861

Counterfactual fertility index in 1911 = 0.293

Under no migration of men and women to Seine after 1861

Counterfactual fertility index in 1911 = 0.276

Counterfactual analysis:

• Migration explains 75% of the fertility decline (at the mean)

• Without migrations, the standard deviation would have only decreased by 33% • Migrations to Paris explain 50% of the fertility decline (at the mean), that is, ¾ of

the effect of migration

Interpretation

Emigrants to Paris mattered more than the other emigrants, in line with the cultural, economic and political importance of Paris within France.

Would-be emigrants who sought to move to Paris, even if they eventually migrated to the closest regional industrial center, had few children because they learnt from emigrants from their regions that Parisians had few children.

Why Parisians had fewer children?

• a cultural element of Parisian life (the political and economic elites living in Paris already had few children by the 18th century, Livi-Bacci 1986)).

• an economic rationale: Parisians had few children because raising many children in Paris was costly and difficult (wet-nursing)

6. Conclusion

This study analyzes the impact of internal migration on the convergence in fertility rates within France between 1861 and 1911.

• The fertility decline in 19th century France is related to the transmission of information and cultural norms channeled mostly by emigrants (but not so much by immigrants).

• Our counterfactual analysis suggests that emigrants to Paris account for half of the total decline in fertility, even though only 5.25% of the total number of French internal emigrants moved to Paris between 1861 and 1911. Migrants to Paris mattered more, in line with the cultural, political and economic importance of the French capital;

Implications for current debates on migration and demography, especially in AFRICA

• Negative self-selection on fertility (BAD) • New element compared to historical contexts: more selective immigration

policies in Europe twist the quality-quantity tradeoff toward quality (GOOD),

and

• Migration generates demographic remittances of the Malthusian type (EXCELLENT).