Crowdsourced genealogies

Download the data

Download europe.dta.zip (1.37 GB)

The release details and SHA-256 checksum are available on the v1.0 release page.

Overview

This release accompanies:

Guillaume Blanc (2024), “Demographic transitions, rural flight, and intergenerational persistence: evidence from crowdsourced genealogies,” manuscript dated April 10, 2024.

The release contains an individual-level dataset derived from the public FamiLinx genealogies assembled by Kaplanis et al. (2018) from public Geni.com profiles. The unit of observation is an anonymized individual profile. This is a crowdsourced genealogy sample rather than a random sample of the historical population. The accompanying paper evaluates its representativeness.

europe.dta contains 9,426,965 observations and 119 variables, in Stata format 118 (Stata 14 or later). It covers individuals born and died between 1400 and 2018.

Geographic Coverage and Sample Selection

Selection is based on three events: the individual's birth, the birth of the individual's first linked child, and the individual's death.

Coverage extends to Austria, Belgium, Bulgaria, Croatia, Czechia, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Iceland, Ireland, Italy, Latvia, Lithuania, Luxembourg, North Macedonia, the Netherlands, Norway, Poland, Portugal, Romania, Slovakia, Slovenia, Spain, Sweden, Switzerland, and the United Kingdom.

An individual is included in europe.dta when at least one selection event maps to one of these countries. Birth in Europe is not required. For example, an individual born outside Europe who had a first linked child in Europe or died in Europe is included. After inclusion, all retained event fields are preserved, including fields for events outside Europe.

In the string country variables, North Macedonia and the United Kingdom are recorded as Macedonia and UK.

Non-European life events are not discarded for people selected into europe.dta. When such an event maps to the United States or Canada, its coordinates and place fields remain in the observation. america and the event-specific *_america variables describe those retained event locations; they do not define an additional North American sample.

Data Source and Construction

The FamiLinx source is reported to contain 86,124,645 individual profiles plus direct parent-child links. The released dataset was constructed in four stages.

Profiles, dates, and plausibility screens

Family links and fertility

fert is the number of retained parent-child link rows for an individual; it is an observed linked-child count, not a verified count of all biological children. Parent-child links for a parent with more than 50 rows are discarded rather than top-coded. If more than two parents are linked to a child, only the first two after numeric-ID ordering enter the parent and co-parent structure. A one-parent record receives a synthetic missing-co-parent placeholder during construction; the released co-parent identifier is missing.

The variables named spouse, couple_id, and n_spouses are constructed from shared children and should be interpreted as co-parent measures, not verified marriages:

First- and last-child years, monthly dates, months, and locations are derived from linked children whose profiles could be matched. Relation counts can therefore be observed even when dated child profiles are not.

The fertility-sample indicator is constructed from the individual's own linked fertility and a construction-only ancestral measure, fert_bef. The latter is the row mean of the nonmissing linked-child counts in 30 pedigree positions: 2 parents, 4 grandparents, 8 great-grandparents, and 16 great-great-grandparents. Missing positions are ignored, and fert_bef is missing only when fertility is unavailable for all 30 positions. Pedigree collapse can cause the same ancestor to occupy more than one position.

The indicator is defined only when both the individual's fert and fert_bef are observed:

Every nonmissing linked-child count is an integer of at least one. Consequently, fert_bef>1 is equivalent to at least one observed pedigree position having more than one linked child. Likewise, s_fert==0 means that every observed pedigree-position count equals one. Missing fert or s_fert must not be interpreted as zero fertility: a profile without a retained linked child is assigned missing fert, not fert=0.

Observed fertility coverage in europe.dta: of the 9,426,965 total observations, fert is observed for 2,070,330. Among those, s_fert is missing for 8,313,251 (ancestral fertility unavailable), s_fert==0 for 353,890, and s_fert==1 for 759,824 — the analytic fertility sample. s_fert is nonmissing for 1,113,714 observations overall.

Among profiles with observed own fertility, 956,616 have missing s_fert because ancestral fertility is unavailable. The file retains main-sample profiles regardless of whether s_fert is observed, which is why its total observation count is much larger than its fertility-sample count.

Geographic mapping and historical urban status

European coordinate keys are matched using 2018 LAU geography and 2013 NUTS 3, NUTS 2, and country-level geography. For retained events that map to the United States or Canada (see above), coordinate keys are matched to 2019 Census counties and states, or 2016 Canadian census divisions and provinces, respectively. These modern units are applied to historical event coordinates. The released *_dep, *_region, and *_country values are dataset-specific integer categories with Stata value labels generated from area names; they are not official geographic codes. For purposes of matching historical city populations, selected region-level consolidations are applied for London, Brussels, Antwerp, Liège, Dublin, and Belfast.

The broad European groups in *_group are:

Historical urban status is based on Bairoch, Batou, and Chèvre (1988) municipality populations for 1500, 1600, 1700, 1750, 1800, and 1850. Population values are expressed in thousands, and a mapped place is urban when population is at least 5 (5,000 inhabitants). Selected missing intermediate populations are geometrically interpolated between surrounding benchmarks. Once a place is observed as urban, a missing status at a later benchmark is carried forward as urban; other missing statuses are coded zero in covered countries. Helsinki is manually added as an urban place in 1850. Urban fields are set missing for Estonia, Iceland, Latvia, Lithuania, and Macedonia, as well as for events outside Europe.

For all three mapped events, the applicable urban benchmark is selected using the individual's year of birth. Thus first_child_urban describes the first-child location at the parent's birth-year benchmark, and death_urban describes the death location at the deceased person's birth-year benchmark. These are the measures used in the paper to classify origin and destination places relative to urban status at the individual's birth.

Derived variables, parents, and weights

Great-circle birth-to-death distance is generated in kilometres. Ages at first child, last child, and death are differences between Stata monthly dates divided by 12.

Parent 1 and parent 2 are ordered by numeric identifier, not by sex. Parent identifiers and linked-child counts can be present even when the parent's own profile is absent from the final individual file. Parent sex, event dates, mapped places, ages, and distance are populated only when that linked parent is also found in the retained intermediate individual sample.

Event weights use estimated country population in 1820 from the Maddison Project Database 2020. Missing 1820 values are manually supplied for Croatia, Estonia, Iceland, Latvia, Lithuania, Luxembourg, Macedonia, Slovakia, and Slovenia, using the assumptions recorded in the construction. For each European country and event, the weight equals 1820 population in thousands divided by the number of genealogy observations in the event-specific denominator window: [1675,1900) for birth, [1700,1925) for first child, and [1750,1975) for death. Weights are not normalized to a mean of one. The same country-event factor is attached to observations outside its denominator window, so it should normally be used only with the corresponding recommended window.

Key Variables

Variable names are case-sensitive.

Individual and sample variables

Fertility and family links

The fertility sample retains an individual when own fertility is observed, at least one of the 30 preceding-generation pedigree positions has observed fertility, and at least one observed position has more than one linked child. Childless individuals cannot be identified reliably. The measure is completed gross linked fertility; child mortality is under-recorded outside direct vertical lineages.

Parent variables

Parent 1 and parent 2 are ordered by numeric FamiLinx identifier, not by sex. Use parent1_male and parent2_male; do not interpret parent 1 as the father or parent 2 as the mother.

For each parent, the file can include:

Dates and ages

For event prefix birth, first_child, last_child, or death:

The same date-construction caveats apply to parent event fields. The first- and last-child year, month, and date are calculated separately and can refer to different linked children. last_child_m is the minimum reported month among children in the latest observed child-birth year, rather than the maximum. When the chronologically first child's location is missing, the earliest dated linked child with a nonmissing location can be substituted. A first-child location therefore need not correspond exactly to first_child_date.

Geography

For event prefix birth, first_child, or death:

No other last-child geography fields are included.

Event-level fields Event in Europe Event in United States or Canada
*_id, *_dep, *_region, *_country, *_COUNTRY, coordinates Available when mapped Available when mapped
*_group Available Missing
*_urban Available where historical urban data are covered Missing
*_weight Available for mapped European countries Missing

Urban status, migration, and weights

Use the weight corresponding to the event-time outcome being analyzed.

Recommended Analysis Windows

For its European analyses, the paper focuses on periods where comparisons with representative sources suggest selection is most limited:

The file retains observations outside these windows.

Variable Dictionary

The dataset has the following 119-variable schema.

# Variable Definition
1 id Anonymized FamiLinx individual identifier.
2 male Male indicator: 1=male, 0=female; missing if source sex was not identified.
3 distance Great-circle distance in kilometres between mapped birth and death locations.
4 death_age Age at death in years, calculated from monthly dates and divided by 12.
5 europe 1 if birth, first-child, or death location maps to Europe; otherwise missing.
6 america Auxiliary indicator equal to 1 if any retained birth, first-child, or death location maps to the United States or Canada; missing otherwise. It does not determine inclusion in this file.
7 fert Number of retained parent-child link rows; use s_fert==1 for fertility analysis.
8 s_fert Fertility-sample indicator: missing if own fertility or all ancestral fertility values are missing; 0 if both are observed and every observed ancestral linked-child count equals 1; 1 if both are observed and at least one observed ancestral count exceeds 1.
9 s_main Main-sample indicator; equals 1 for every observation in both released files.
10 spouse Identifier of the selected co-parent sharing the most linked children with the individual; not a verified spouse.
11 spouse_fert Retained linked-child count of the selected co-parent.
12 couple_id Identifier for the individual-selected-co-parent pair.
13 couple_fert Number of linked children shared by the individual and selected co-parent.
14 n_spouses Number of candidate co-parent-pair records before selection; fertility-family fields are absent above 10.
15 fert_parents Mean observed fertility of recorded parents.
16 s_fert_parents 1 if at least one linked parent is itself retained with s_fert==1; missing otherwise.
17 parent1_id Identifier of parent 1; parent order is not sex-specific.
18 parent1_fert Observed number of linked children of parent 1.
19 parent1_male Male indicator for parent 1; parent order is not sex-specific.
20 parent1_first_child_age Age of parent 1 at birth of first linked child.
21 parent1_last_child_age Age of parent 1 at birth of last linked child.
22 parent1_death_age Age at death of parent 1, in years.
23 parent1_distance Great-circle distance in kilometres between parent 1 birth and death locations.
24 parent1_birth_y Year of parent 1 birth.
25 parent1_birth_date Stata monthly date of parent 1 birth.
26 parent1_birth_id Mapped LAU/place identifier for parent 1 birth.
27 parent1_birth_latlon_ID Raw FamiLinx latitude-longitude key for parent 1 birth.
28 parent1_birth_COUNTRY String country name for parent 1 birth.
29 parent1_first_child_y Year of parent 1 first linked child's birth.
30 parent1_first_child_date Stata monthly date of parent 1 first linked child's birth.
31 parent1_first_child_id Mapped LAU/place identifier for parent 1 first linked child's birth.
32 parent1_first_child_latlon_ID Raw FamiLinx latitude-longitude key for parent 1 first linked child's birth.
33 parent1_first_child_COUNTRY String country name for parent 1 first linked child's birth.
34 parent1_death_y Year of parent 1 death.
35 parent1_death_date Stata monthly date of parent 1 death.
36 parent1_death_id Mapped LAU/place identifier for parent 1 death.
37 parent1_death_latlon_ID Raw FamiLinx latitude-longitude key for parent 1 death.
38 parent1_death_COUNTRY String country name for parent 1 death.
39 parent2_id Identifier of parent 2; parent order is not sex-specific.
40 parent2_fert Observed number of linked children of parent 2.
41 parent2_male Male indicator for parent 2; parent order is not sex-specific.
42 parent2_first_child_age Age of parent 2 at birth of first linked child.
43 parent2_last_child_age Age of parent 2 at birth of last linked child.
44 parent2_death_age Age at death of parent 2, in years.
45 parent2_distance Great-circle distance in kilometres between parent 2 birth and death locations.
46 parent2_birth_y Year of parent 2 birth.
47 parent2_birth_date Stata monthly date of parent 2 birth.
48 parent2_birth_id Mapped LAU/place identifier for parent 2 birth.
49 parent2_birth_latlon_ID Raw FamiLinx latitude-longitude key for parent 2 birth.
50 parent2_birth_COUNTRY String country name for parent 2 birth.
51 parent2_first_child_y Year of parent 2 first linked child's birth.
52 parent2_first_child_date Stata monthly date of parent 2 first linked child's birth.
53 parent2_first_child_id Mapped LAU/place identifier for parent 2 first linked child's birth.
54 parent2_first_child_latlon_ID Raw FamiLinx latitude-longitude key for parent 2 first linked child's birth.
55 parent2_first_child_COUNTRY String country name for parent 2 first linked child's birth.
56 parent2_death_y Year of parent 2 death.
57 parent2_death_date Stata monthly date of parent 2 death.
58 parent2_death_id Mapped LAU/place identifier for parent 2 death.
59 parent2_death_latlon_ID Raw FamiLinx latitude-longitude key for parent 2 death.
60 parent2_death_COUNTRY String country name for parent 2 death.
61 birth_y Reported birth year, with baptism year substituted when birth year is missing.
62 birth_m Reported birth month, with baptism month substituted when birth month is missing; may remain missing or invalid.
63 birth_date Stata monthly date constructed before baptism fallback; January is used when the original birth month is missing or invalid, and the value can disagree with birth_y or birth_m.
64 birth_id 2018 LAU identifier in Europe or source coordinate-key identifier in the United States/Canada for the birth place.
65 birth_dep Dataset-specific value-labelled category for the birth-place NUTS 3, county, or census-division name; not an official code.
66 birth_region Dataset-specific value-labelled category for the birth-place NUTS 2, state, or province name; not an official code.
67 birth_country Dataset-specific numeric country category with value labels for the birth location.
68 birth_COUNTRY String country name for individual's birth location.
69 birth_group Broad European country-group name for individual's birth location; missing outside Europe.
70 birth_europe 1 if individual's birth location maps to Europe; missing otherwise.
71 birth_america Birth-location indicator: 1=United States/Canada, 0=Europe, missing if unmapped.
72 birth_longitude Longitude of mapped birth place; a representative LAU point in Europe or matched source-coordinate point in the United States/Canada.
73 birth_latitude Latitude of mapped birth place; a representative LAU point in Europe or matched source-coordinate point in the United States/Canada.
74 birth_latlon_ID Raw FamiLinx latitude-longitude key for individual's birth.
75 birth_weight 1820 European population in thousands divided by the country's birth-record count in [1675,1900); generally missing outside Europe.
76 birth_urban Urban status of the mapped birth place at the latest benchmark at or before the individual's birth year; missing where urban data are not covered.
77 first_child_y Minimum nonmissing birth year among matched linked-child profiles.
78 first_child_m Minimum reported or fallback month among children in the earliest observed child-birth year; may be missing or invalid and need not match first_child_date.
79 first_child_date Minimum available Stata monthly birth date among matched linked-child profiles; it can refer to a different child from first_child_y or first_child_m.
80 first_child_age Age at birth of first linked child, in years.
81 first_child_id 2018 LAU identifier in Europe or source coordinate-key identifier in the United States/Canada for the mapped first-child place; a fallback may use the earliest dated child with a location.
82 first_child_dep Dataset-specific value-labelled category for the first-child-place NUTS 3, county, or census-division name; not an official code.
83 first_child_region Dataset-specific value-labelled category for the first-child-place NUTS 2, state, or province name; not an official code.
84 first_child_country Dataset-specific numeric country category with value labels for the first-child location.
85 first_child_COUNTRY String country name for birth of first linked child location.
86 first_child_group Broad European country-group name for birth of first linked child location; missing outside Europe.
87 first_child_europe 1 if birth of first linked child location maps to Europe; missing otherwise.
88 first_child_america First-child-location indicator: 1=United States/Canada, 0=Europe, missing if unmapped.
89 first_child_longitude Longitude of mapped first-child place; a representative LAU point in Europe or matched source-coordinate point in the United States/Canada.
90 first_child_latitude Latitude of mapped first-child place; a representative LAU point in Europe or matched source-coordinate point in the United States/Canada.
91 first_child_latlon_ID Raw FamiLinx latitude-longitude key for birth of first linked child.
92 first_child_weight 1820 European population in thousands divided by the country's first-child-record count in [1700,1925); generally missing outside Europe.
93 first_child_urban Urban status of the mapped first-child place at the latest benchmark at or before the parent's birth year; missing where urban data are not covered.
94 last_child_y Maximum nonmissing birth year among matched linked-child profiles.
95 last_child_m Minimum reported or fallback month among children in the latest observed child-birth year; may be missing or invalid and need not match last_child_date.
96 last_child_date Maximum available Stata monthly birth date among matched linked-child profiles; it can refer to a different child from last_child_y or last_child_m.
97 last_child_age Age at birth of last linked child, in years.
98 death_y Reported death year, with burial year substituted when death year is missing.
99 death_m Reported death month, with burial month substituted when death month is missing; may remain missing or invalid.
100 death_date Stata monthly date constructed before burial fallback; January is used when the original death month is missing or invalid, and the value can disagree with death_y or death_m.
101 death_id 2018 LAU identifier in Europe or source coordinate-key identifier in the United States/Canada for the death place.
102 death_dep Dataset-specific value-labelled category for the death-place NUTS 3, county, or census-division name; not an official code.
103 death_region Dataset-specific value-labelled category for the death-place NUTS 2, state, or province name; not an official code.
104 death_country Dataset-specific numeric country category with value labels for the death location.
105 death_COUNTRY String country name for individual's death location.
106 death_group Broad European country-group name for individual's death location; missing outside Europe.
107 death_europe 1 if individual's death location maps to Europe; missing otherwise.
108 death_america Death-location indicator: 1=United States/Canada, 0=Europe, missing if unmapped.
109 death_longitude Longitude of mapped death place; a representative LAU point in Europe or matched source-coordinate point in the United States/Canada.
110 death_latitude Latitude of mapped death place; a representative LAU point in Europe or matched source-coordinate point in the United States/Canada.
111 death_latlon_ID Raw FamiLinx latitude-longitude key for individual's death.
112 death_weight 1820 European population in thousands divided by the country's death-record count in [1750,1975); generally missing outside Europe.
113 death_urban Urban status of the mapped death place at the latest benchmark at or before the individual's birth year; missing where urban data are not covered.
114 parent1_birth_europe 1 if parent 1 birth location maps to Europe; missing otherwise.
115 parent1_first_child_europe 1 if parent 1 first linked child's birth location maps to Europe; missing otherwise.
116 parent1_death_europe 1 if parent 1 death location maps to Europe; missing otherwise.
117 parent2_birth_europe 1 if parent 2 birth location maps to Europe; missing otherwise.
118 parent2_first_child_europe 1 if parent 2 first linked child's birth location maps to Europe; missing otherwise.
119 parent2_death_europe 1 if parent 2 death location maps to Europe; missing otherwise.

Citation

Please cite: