1. Migration and language gender marking
Migration and language gender marking
An analysis of gender wage gaps
Joanna Tyrowicz and Lucas van der Velde
Warsaw School of Economics
FAME|GRAPE
International Association for Feminist Economics,
June 2018
2. Migration and language gender marking
Introduction
Introduction
Gender language marking → higher gender inequality
Santacreu-Vasut et al. (2013, 2014), Tyrowicz et al. (2015), Hicks et al. (2015),
Mavisakalyan (2015)
but...
cross-sectional studies evidence confounds cultural and institutional
factors
Single country studies of migrants solve this issue
→ So far, only with US data
→ What about interactions? Representative?
3. Migration and language gender marking
Introduction
Our contribution
We test the role of language marking in explaining gender wage gaps
(GWG) among immigrants
Comparable measures of GWG from migrants in several destination
countries
“Epidemiological” approach
(Fern´andez and Fogli 2009, Blau 2015)
Multicountry-multilanguage analysis
4. Migration and language gender marking
Introduction
Gender wage gap of immigrants to Canada
0
.1
.2
.3
.4
Genderwagegapbetweenmigrants
No distinctions Strong distinction
Linguistic gender marking at home
Provinces: English speaking French speaking
5. Migration and language gender marking
Languages
Why would language affect gender inequality?
→ Indirect effects: language as a proxy for slow moving cultural
characteristics
(Gay et al. 2017, Mavisakalyan and Weber 2017)
Correlates with opinions on role of women WVS
→ Direct effects: language shapes how we perceive reality
(Chen 2013, Mavisakalyan et al. 2018)
E.g. Masculine and feminine nouns → bridges and keys
(Borodistky, Schmidt and Phillips 2002)
Language proximity can drive migration flows
(Alcia and Mariola 2015)
6. Migration and language gender marking
Languages
How to measure language gender marking?
We use data from the World Atlas of Language Structures
Santacreu-Vasut et al. (2013, 2014), Tyrowicz et al. (2015), Hicks et al. (2015)
1 Number of Genders (NG)
→ The language has 2 genders
2 Sex Based (SB)
→ Genders are connected to biological sex
3 Gender Assignment (GA)
→ Language has semantic and phonetic gender assignment
4 Gender Pronouns (GP)
→ Language has male and female personal pronouns
Descriptive statistics
7. Migration and language gender marking
Measuring gender wage gap
Data sources
Census data
Brazil (2015 – IPUMSI); Canada (2011 – IPUMSI); USA (2016 – ACS)
Labor Force Survey
France (2008-2013); UK (2010-2015)
Our database
∼ 5.7 million individual observations
of which ∼ 0.2 million are migrants
representing 95 unique origin countries
and 166 country pairs (origin+destination)
Details
8. Migration and language gender marking
Measuring gender wage gap
Method
We apply ˜Nopo (2008) non-parametric decomposition to hourly wages.
Raw gap = ∆X + ∆A + ∆M + ∆W
where
∆X : explained component
∆A: unexplained / adjusted component
∆M : component due to Men different than Women in the sample
∆W : component due to Women different than Men in the sample
Control variables: age group, education (3 levels) and marital status.
9. Migration and language gender marking
Measuring gender wage gap
The identification problem
Wages for individual k of gender g from country i living in j:
wagek,g,i,j = α(x) + θg,j (x) + θi,j + ηg,i,j + k,g,i,j
where
α(x): related to productivity of observed characteristics (x)
θg,j (x): gender penalty in j for individuals with characteristics (x)
θi,j (x): country specific penalty for migrants
ηg,i,j : is what we are interested in
→ a gender penalty specific for country of origin & destination
→ includes effects of culture
k,g,i,j is a random error.
10. Migration and language gender marking
Measuring gender wage gap
The identification problem
GWG for migrants from the same country equals:
E(WageM,i,j − WageW ,i,j ) = α(xM − xW )
Explained (∆X )
+ θg,j (x) + ηg,i,j
Unexplained (∆A)
If migrants from different countries had same (x), then
θg,j (x) = θj (x)
Variation in ∆A only due to ηg,i,j
This assumption is unrealistic Example
11. Migration and language gender marking
Measuring gender wage gap
The identification problem
How to recover θg,i,j (x)?
We estimate ˆθg,j (x) from the native population
We obtain two counterfactual wage distributions:
1 Native men’s wage with same X’s as women from i → W c
m
2 Native women’s wage with same X’s as women from i → W c
w
ˆθg,j (x) = E(W c
m) − E(W c
w )
X’s correspond to migrant women in the common support. All estimates were
obtained via ˜Nopo (2008)
12. Migration and language gender marking
Measuring gender wage gap
Final specification
∆A,i,j = β0 + β1language gender marking + β2
ˆθg,j (x) + βZi,j + ε
where
∆A,i,j : adjusted gender wage gap for migrant from i living in j
β1: the parameter of interest
β2: deviations with respect to the average GWG in country j
Z: a vector of controls for origin and host countries’ characteristics
(education, fertility, GDP) and pair characteristics (distance,
common official language)
13. Migration and language gender marking
Results
A brief look into gender wage gap among migrants (I)
Adjusted gender wage gap among migrants by destination country
0
.2
.4
.6
.8
Not 2 Only 2
Number of genders
0
.2
.4
.6
.8
No Yes
Sex−based distinction
0
.2
.4
.6
.8
Semantic Sem. and phonetic
Gender assignment
0
.2
.4
.6
.8
No Yes, all persons
Gender pronouns
Adjustedgenderwagegap
Notes Estimates of the adjusted gender wage gap obtained via Nopo decomposition.
Controls include age, education and marital status. Observations with values higher than
1 (n=12) not reported. Characteristics refer to language of origin.
14. Migration and language gender marking
Results
A brief look into gender wage gap among migrants (II)
Adjusted gender wage gap among migrants by destination country
0
.2
.4
.6
.8
Adjustedgenderwagegapbetweenmigrants
BRA CAN FCN FRA GBR USA
Notes Estimates of the adjusted gender wage gap obtained via Nopo decomposition.
Controls include age, education and marital status. Observations with values higher than
1 (n=12) not reported. FCN stands for French Canada (Quebec).
15. Migration and language gender marking
Results
Does language matter?
Variables Adjusted gender wage gap
(1) (2) (3) (4)
Number of Genders -0.08 -0.15 -0.08 -0.16
(0.30) (0.19) (0.29) (0.20)
Sex Based -0.03 -0.02 -0.04 -0.02
(0.72) (0.87) (0.73) (0.93)
Gender Assignment 0.15* 0.15* 0.13 0.14
(0.06) (0.07) (0.22) (0.14)
Gender Pronouns 0.10 0.14 0.09 0.16
(0.59) (0.49) (0.63) (0.46)
Destination FE Y Y Y Y
Home charact. N Y N Y
Country pair vars N N Y Y
Observations 156 144 156 144
R-squared 0.12 0.18 0.13 0.18
F-test (p) 0.488 0.163
Notes: Robust p-values in parenthesis. F-test(p) is the p-value of a joint significance test for
country pair variables. In Column 3 (4) the restricted model is 1 (2).
16. Migration and language gender marking
Results
Home and destination interactions
Variables Adjusted gender wage gap
(1) (2) (3) (4)
Gender assignment (home) 0.15* 0.15** 0.16* 0.22**
(0.05) (0.04) (0.06) (0.01)
Gender Assignment (destination) -0.05 -0.05 -0.06 -0.07
(0.36) (0.37) (0.33) (0.29)
Lower G.A. in destination -0.06 -0.13 -0.09 -0.21*
(0.49) (0.16) (0.36) (0.06)
Home charact. N Y N Y
Country pair vars N N Y Y
Observations 155 143 155 143
R-squared 0.11 0.30 0.12 0.33
F-test (p) 0.845 0.164
Notes: Robust p-values in parenthesis.F-test(p) is the p-value of a joint significance test for
country pair variables. In Column 3 (4) the restricted model is 1 (2).
17. Migration and language gender marking
Conclusions
Concluding remarks
Linguistic gender marking offers a new light into gender disparities in
economics
Migrants from countries with higher distinction→ higher GWG in
destination
Lower marking destinations → lower migrants GWG
Effect is large, but noisy
18. Migration and language gender marking
Conclusions
Last slide
Questions or comments?
Lucas van der Velde
Contact: lvandervelde@grape.org.pl
19. Migration and language gender marking
References
Alcia, A. and Mariola, P.: 2015, The role of language in shaping international
migration, The Economic Journal 125(586), F49–F81.
Blau, F. D.: 2015, Immigrants and gender roles: assimilation vs. culture, IZA Journal
of Migration 4(1), 23.
Chen, M. K.: 2013, The effect of language on economic behavior: Evidence from
savings rates, health behaviors, and retirement assets, American Economic Review
103(2), 690–731.
Fern´andez, R. and Fogli, A.: 2009, Culture: An empirical investigation of beliefs,
work, and fertility, American Economic Journal: Macroeconomics 1(1), 146–77.
Gay, V., Hicks, D. L., Santacreu-Vasut, E. and Shoham, A.: 2017, Decomposing
culture: An analysis of gender, language, and labor supply in the household, Review
of Economics of the Household pp. 1–31.
Hicks, D. L., Santacreu-Vasut, E. and Shoham, A.: 2015, Does mother tongue make
for women’s work? Linguistics, household labor, and gender identity, Journal of
Economic Behavior & Organization 110, 19–44.
Mavisakalyan, A.: 2015, Gender in language and gender in employment, Oxford
Development Studies 43(4), 403–424.
Mavisakalyan, A., Tarverdi, Y. and Weber, C.: 2018, Talking in the present, caring for
the future: Language and environment, Journal of Comparative Economics .
Mavisakalyan, A. and Weber, C.: 2017, Linguistic structures and economic outcomes,
Journal of Economic Surveys .
˜Nopo, H.: 2008, Matching as a tool to decompose wage gaps, The review of
economics and statistics 90(2), 290–299.
20. Migration and language gender marking
Appendix
Santacreu-Vasut, E., Shenkar, O. and Shoham, A.: 2014, Linguistic gender marking
and its international business ramifications, Journal of International Business
Studies 45(9), 1170–1178.
Santacreu-Vasut, E., Shoham, A. and Gay, V.: 2013, Do female/male distinctions in
language matter? evidence from gender political quotas, Applied Economics Letters
20(5), 495–498.
Tyrowicz, J., van der Velde, L. and Siwinska, J.: 2015, Language and (the estimates
of) the gender wage gap, Economics Letters 136, 165–170.
21. Migration and language gender marking
Appendix
Language gender marking and opinions on women’s role
Women earns more creates conflict
Child suffers when mother works
University + important for boys
Work mom good relation w.child
Being a housewife is fulfilling
−.2 −.1 0 .1 .2
Notes: Data on poportion of people agreeing with statements come from World
Value Survey (all waves). Wave fixed effects are included in all regressions.
Back
22. Migration and language gender marking
Appendix
Linguistic gender marking in our sample
# % Common outcomes Example
NG SB GA GP (=0)
Number genders (NG) 62 1 Polish
Sex-based (SB) 131 0.58 1 Danish
Gender assignment (GA) 94 0.78 0.72 1 English
Gender pronouns (GP) 35 0.81 0.40 0.62 1 Italian
Notes
Spanish and Arabic are languges where all variables equal 1
Number of observations = 162. Languages might be counted several times
% common outcome = equal values
# of observations
Back
23. Migration and language gender marking
Appendix
Detailed sample selection
BRA CAN FCN FRA GBR USA
# observations 6405902 483752 154004 1236858 1010142 1994223
# migrants 12364 134665 24149 76681 89108 278465
Complete obs 7857 54874 3904 6894 11544 133177
Sufficient obs 6499 54874 3549 6632 9114 130857
# countries 14 15 11 11 27 86
Notes Complete obs. refers to observations from migrants with no missing data on age, ed-
ucation, marital status, hourly wages and country of origin. Sufficient obs. refers tomigrants
who, in addition to previous, are from countries with 48+ men and women. FCN stands for
French Canada (Quebec).
Back
24. Migration and language gender marking
Appendix
Comparison of Spanish and Hondurian migrants in US
Honduras Spain
Men Women Men Women
Differences in characteristics
Age 40.00 41.45 45.75 46.41
% secodary 0.48 0.44 0.37 0.34
% tertiary 0.21 0.32 0.52 0.58
% married 0.90 0.69 0.84 0.74
Differences in wages
Mean 16.51 13.70 30.13 20.94
Q1 9.80 7.50 11.76 9.57
Q2 13.24 10.99 18.82 14.26
Q3 19.61 16.32 31.75 22.88
Back