Name Similarity Effects

Report
Spurious?
Name Similarity Effects (Implicit Egotism)
in Marriage, Job, and Moving Decisions
Uri Simonsohn
forthcoming, in the
Journal of Personality and Social Psychology
“There are many hypotheses in science
which are wrong.
That's perfectly all right; they're the
aperture to finding out what's right.
Science is a self-correcting process.”
• Carl Sagan.
Background
• Name-letter-effect
– (Nuttin, 1985) + 10 replications;
– Names, birthday #s
• Three papers by Brett Pelham and
colleagues (JPSP 2002,2003):
– Location: e.g. George move to GEORGIA
– Profession: Dennis becomes a Dentist
– Marriage: e.g. Andrew marries Andrea
What did Pelham et al. do wrong?
• Analyses w/o chosen controls.
1) Confounds (Z)
– ZYour name
– Zyour decisions
– Corr(name,decision)>0
– Corr(name,decision|Z)=0
2) Reverse Causality
– Mr. Smith found Smithville
Analysis throughout
all new studies
• Count actual frequencies
• Estimate expected frequencies
• Analyze ratio(actual/expected)
• Null: R=1
• Name-similarity-effect  R>1
Eight Original Findings
1. Marriage: Last names
2. Marriage: First names
3. Occupation: first names
• Location
4.
5.
6.
7.
8.
State: first name
State: last name
Town: last name
Street: last name
Town: Birthday
Marriage & Last names
• S___ marries S___
• Smith marries Smith
Will show that:
1.
Same-last-name-initial effect
is actually Same-last-name
effect
2. Same-last-name effect is due to reverse
causality
Data: Marriages between Hispanics
Texas 2001; n=24,645
D.V. : R=ratio(actual/expected)
Why do same-last-names attract?
• Maybe implicit egotism needs more than a
letter?
• If so, then R(very similar last names)>1
• Does Gonzales marry Gonzalez too much?
• Looked at 20 Hispanic last names
– Next: 3
– Then: all 20
Same & very similar last name
Ready?
Figure with 20 last names
** ns ** ns
** ns
** ns
** ns
** ns ** ns
Alvarez : Alvarez (m=75)
Alvarez : Alvarado (m=23)
Espinosa : Espinosa (m=8)
Espinosa : Espinoza (m=5)
Espinoza : Espinoza (m=47)
Espinoza : Espinosa (m=6)
Gonzales : Gonzales (m=367)
Gonzales : Gonzalez (m=179)
** ns ** ns ** ns ** ns ** ns ** ns ** ** ** ** ** ns ** ns ** ns ** ns ** *
R(same) = 1.90 **
R(similar)= 0.97 n.s.
** ns
** ns
** ns ** na
Guerrero : Guerrero (m=74)
Guerrero : Guerra (m=26)
Mendez : Mendez (m=65)
Mendez : Mendoza (m=25)
Mendoza : Mendoza (m=96)
Mendoza : Mendez (m=26)
** ns ** ns
Salazar : Salazar (m=109)
Salazar : Salas (m=17)
** †
** na ** ns
Morales : Morales (m=118)
Morales : Mora (m=3)
** ns ** ns ** ns ** ns ** ns ** ns
** ns
** ns
Velazquez : Velazquez (m=11)
Velazquez : Velasquez (m=4)
** †
Velasquez : Velasquez (m=23)
Velasquez : Velazquez (m=2)
6.3
2
** ns
OVERALL SAME (m=2330)
OVERALLSIMILAR (m=680)
7.0
Vazquez : Vazquez (m=40)
Vazquez : Vasquez (m=27)
Vasquez : Vasquez (m=111)
Vasquez : Vazquez (m=18)
Salas : Salas (m=28)
Salas : Salazar (m=17)
Ratio of Actual/Expected frequencies
5
Mora : Mora (m=9)
Mora : Morales (m=3)
** ns
Guerra : Guerra (m=65)
Guerra : Guerrero (m=28)
Gonzalez : Gonzalez (m=940)
Gonzalez : Gonzales (m=211)
Alvarado : Alvarado (m=54)
Alvarado : Alvarez (m=24)
** ns
Aguirre : Aguirre (m=20)
Aguirre : Aguilar (m=14)
0
Aguilar : Aguilar (m=70)
Aguilar : Aguirre (m=22)
Lastnames
Groom : Bride
6
9.4
`
4
3
1.90
1
.97
** ns
Why only for exactly the
same last name?
• Reverse causality: bride changes name
before marriage
• Examples:
– Marriage abroad, immigration, marriage in US.
– Marriage, widow, marry brother-in-law
– Marriage, divorce, marry again (same guy)
Who does that?
Finding Meryl Streeps
- Start with:
- Meryl Baldwin & Alec Baldwin in Texas 2001
Search for all cases when
– Meryl ???? married/divorced Alec Baldwin.
– Meryl ???? Born same year as Meryl Baldwin.
– Alec Baldwyn born same year as this one.
Real Example:
Eight Original Findings
1. Marriage: last names
2. Marriage: First names
3. Occupation: first names
• Location
4.
5.
6.
7.
8.
State: first name
State: last name
Town: last name
Street: last name
Town: Birthday
Eight Original Findings
1. Marriage: last names
2. Marriage: First names
3. Occupation: first names
• Location
4.
5.
6.
7.
8.
State: first name
State: last name
Town: last name
Street: last name
Town: Birthday
First names & marriage
• 12 name pairs
– E.g. Eric / Erica; Paul / Paula
• Likely confound: cohort
• Does popularity of baby name
Andrew/Andrea move together over time?
BabyNameWizard.com
Searched for: Andre*
Searching Paul*
A more boring framing
of the first-name result
Do people born in the 60s
disproportionately marry
people born in the 60s?
New Analyses
Select better control names
Finding Better Name Controls
Approach:
Other names with similar spouse names.
Example
5% of Andrews marry a Katy, 7% a Deb, 4% Cynthia
5% of Mikes
marry a Katy, 7% a Deb, 4% Cynthia
Mike is a good control for Andrew.
New analyses
• Using all marriages Texas 1966-2007
• Find top-3 most correlated names in spouse
choice for each original name
• Ask:
“is Andrew more likely to marry Andrea,
rather than 3 women with names like Andrea,
than 3 men with names like Andrew are?
Andrea
CHRISTINA
RACHEL
MICHELLE
ANDREA
ROBERTA
PAULA
STEPHANIE
ANDREW
AA
ROBERT
PAUL
STEPHEN
Andrew JOSEPH
AA
PATRICK PETER
JOHN
THOMAS JAMES
MARGARET
ROBERTA
MARY
NANCY
ROBERT
RR
RR
PP
JOHN
SS
CYNTHIA
CINDY
PAULA
NANCY
ROBERT
PAUL
JOSEPH
PP
MICHALE DOUGLAS RUSSELL STEPHEN
ANGELA
MICHELLE
JENNIFER
STEPHANIE
SS
Overall (12 names)
Original: 1.08
New: 1.00
Eight Original Findings
1. Marriage: last names
2. Marriage: First names
3. Occupation: first names
• Location
4.
5.
6.
7.
8.
State: first name
State: last name
Town: last name
Street: last name
Town: Birthday
Occupation Studies
• Dennis the dentist
– Dennis is more likely than Walter and Jerry to be
a dentist
•Dennis is 40th most common name
•Jerry & Walter are 39Th and 41st
• Problem: census frequency a fairly imperfect
control.
• What if Dennis are younger than Walter and
Jerry?
Popularity of baby names
1880-2008
Figure 5. Popularity of first names used in dentists’ study (Study 6)
30,000
Dennis
20,000
Jerry
Dennis
Walter
15,000
Walter
Jerry
10,000
5,000
Birth year
2005
2000
1995
1990
1985
1980
1975
1970
1965
1960
1955
1950
1945
1940
1935
1930
1925
1920
1915
1910
1905
1900
1895
1890
1885
0
1880
Number of newborns
25,000
A more boring framing of the
occupation result
Are men who are still alive
more likely to work as dentists
than those no longer alive?
Cohort confound
Other samples with live men should overrepresent Dennis
• Parsimony: same lawyer dataset as JPSP1
• R(Dennis-dentist)=1.43
• R(Dennis-lawyer)=1.38
– (x2(1)=1.28, p=.26)
Eight Original Findings
1. Marriage: last names
2. Marriage: First names
3. Occupation: first names
• Location
4.
5.
6.
7.
8.
State: first name
State: last name
Town: last name
Street: last name
Town: Birthday
First name and states
• Original findings.
– With 8 first names
– Georgia GA ; Louis  Louisiana
• Reverse causality:
– Baby names more popular in matching state
• They did look at movers
– Get SSN in other state
• BUT:
1) Returning? (born in Georgia, got SSN in NY, then
moved back there, especially in earlier years)
2) Nearby states? (Georgia also more common in FL)
A more boring framing of the
first-name and states result
Do people disproportionately die near
where they were born?
Popularity of baby name Virginia
New Analyses
• Look where people are “born” (get SSN) as
function of their name
 Identify baby-naming confound
• Conditional on “born”, more likely to stay?
 Estimate implicit egotism net of confound
“Born”
Stayed|”Born”
Ratio of Actual/Expected frequencies
3
2.5
2
1.5
1.27
0.99
1
0.5
0
**
ns
Georgia
GEORGIA
n=126
N=3177
ns ns
** ns
**
ns
Louise
LOUISIANA
n=136
N=5306
Virginia
VIRGINIA
n=447
N=12692
Florence
FLORIDA
n=77
N=3381
**
George
GEORGIA
n=1469
N=44934
'Born in' (obtained SSN in state)
R(“born”) = 1.27 **
R(“stay”)= 0.99 n.s.
**
**
ns
Louis
LOUISIANA
n=616
N=10792
ns ns
Virgil
VIRGINIA
n=66
N=2385
**
*
Kenneth
KENTUCKY
n=1008
N=41798
'Stayed in' (died in state)
** ns
OVERALL
n=3945
N=124465
Reconciling differences in
findings
1. This paper: George does not stay in GA
2. Pelham et al: George more likely to move to GA
• I argue (2) is due to
– Some Georges who got SSN in other state returning
home to Georgia
– George being a more popular name in neighboring
states and hence in states with more migrants.
• Test my story in placebo states:
– More Georges “born” in state X
 More Georges “move” to X.
Men named George
(each dot is a state)
1.40%
Percentage of Immigrants
1.20%
State:Georgia
1.00%
0.80%
0.60%
0.40%
0.20%
0.00%
0.00%
0.20%
0.40%
0.60%
0.80%
1.00%
Percentage of Natives
1) More native-Georges more immigrant Georges
2) Not ’too many’ Georges move to GA
1.20%
1.40%
Women named Virginia
Eight Original Findings
1. Marriage: last names
2. Marriage: First names
3. Occupation: first names - skip
• Location
4.
5.
6.
7.
8.
State: first name
State: last name
Town: last name
Street: last name
Town: Birthday
Existing Finding
• Pelham et al. looked at 30 most popular last
names
• Smith, Johnson, etc.
• Look at towns containing those names
• Find higher % of Smiths in Smithville than in
US
• True for 27/30 names considered.
Reverse Causality
• Asked RA:
– “check who founded top 100 egotistical towns”
• Comes back with n=95
• X% founded by person with that last name
• X=72%
• Not in paper: 10 largest cities
no effect (e.g., Yorks don’t move to New york)
A more boring framing of the
towns results
• Do people founding small towns
disproportionately use their own rather than
someone else’s last name?
Eight Original Findings
1. Marriage: last names
2. Marriage: First names
3. Occupation: first names - skip
• Location
4.
5.
6.
7.
8.
State: first name
State: last name
Town: last name
Street: last name  Reverse causality
Town: Birthday
Who names their own street?
Does that really happen?
Lodi NY, very egotistical town.
• Asked RA
– “Check out how come Lodi is like that”
– Each block named after original owner (received
from military service in civil war)
A more boring framing of the
streets results
• Do people naming streets disproportionately
use their own rather than someone else’s
last name?
New Analysis
• If effect driven by reverse causality
• Should go away with very similar names
– Same idea as Gonzalez vs Gonzales
• Too few Smithers
• Similar first names instead
– E.g. William in Williams ave.
• Also, use West, East, South, North.
First names and streets
• William / Williams Ave
• John / Johnson St
• Jon / Jones Rd
• Jeff / Jefferson Ln
• David / Davis Pl
Why not:
• George / Washington Wy
• Thomas / Jefferson Blvd
Data
• NY Voter Registration N=12 Million
• Next:
– Are people named William more likely to live in
Williams Ave than other New Yorkers are?
Results
First Names
Ratio of Actual/Expected Frequency
2.00
Last Names
1.80
1.60
1.40
1.20
0.92
1.00
0.80
0.60
0.40
0.20
0.00
John
Johns on
n=104
N=138.5k
Wi l l i a m Jon/Jona tha n
Wi l l i a ms
Jones
n=48
n=6
N=89.7k
N=21.8k
Da vi d
Da vi s
n=58
N=96.2k
George
Jeff/Jeffery
Wa s hi ngton Jeffers on
n=171
n=57
N=34.8k
N=35.1k
Thoma s
Jeffers on
n=106
N=67.9k
Ea s t_
Ea s t
n=102
N=1.7k
Wes t_
Wes t
n=266
N=7.5k
North_
North
n=8
N=1.5k
South_
South
n=4
N=1.3k
Overa l l
n=930
N=496.2k
Eight Original Findings
1. Marriage: last names
2. Marriage: First names
3. Occupation: first names - skip
• Location
4.
5.
6.
7.
8.
State: first name
State: last name
Town: last name
Street: last name
Town: Birthday sampling error
• 0riginal:
– Feb 2nd birthdays move to Two Rivers
• Concern:
– Lab effect for #s is not significant (!)
– small n: 94 people
– Unusual test: look for sharing both day and
month, seems like the may have tried many
things and report the one that works.
• Try to replicate
– NY file
•Street #; Address #; apt #
Ratio of actual/expected frequency
Original (towns): n=485, m=94
R1 (address): n=12.8 million, m=1827
2.00
1.50
1.33
1.011.01
0.98
1.00
ns ns ns ns
** ns ns ns
ns ns ns ns
February - 2
March - 3
April - 4
ns ns ns ns
ns ns ns ns
ns ns ns ns
ns ns ns ns
** ns ns ns
0.50
May - 5
June - 6
July - 7
Birth Day
ORIGINAL: City name contains # - N=485, n=94
REPLICATION 1 - Address #
- N=12.8 million, n=1,827
REPLICATION 2 - Apartment #
- N=4.5 million, n=1,287
REPLICATION 3 - Street #
- N=12.8 million, n=632
August - 8
Overall

similar documents