Also available for viewing at https://rpubs.com/sheffm8/620230
This research utilises the self-completed questionnaire as part of Wave 2 (2010-11) of the UKHLS, specifiically three questions focused on gender-role attitudes. Recent research has shown that gender-role attitudes are relevant in preventing female labour market participation in the UK - particularly for certain ethnic minorities. (Wang, 2019) Wang (2019) also established a statistical link between more egalitarian gender role attitudes and female labour force participation (LFP) using Wave 2 of the UKHLS. Other research has demonstrated a higher level of education is linked with more egalitarian gender views. (Thompson et al, 1983) Higher LFP among women has been shown to lead to positive effects such as improved empowerment and independence, as well as better quality of life. (Khoudja & Fleischman 2018) Therefore, this research aims to explore the link between education and gender-role attitudes using the opportunity presented by the UKHLS, in order to contribute to discourse about preventing gender discrimination in the labour market.
The starting point for this research was ‘The Role of Gender Role Attitudes and Immigrant Generation in Ethnic Minority Women’s Labour Force Participation in Britain’ - Senhu Wang (2019). This paper used Wave 2 of the UKHLS, specifically three questions from the self-completed questionnaire which were measured on a five-point Likert scale;
“A pre-school child is likely to suffer if his or her mother works”
“All in all, family life suffers when the woman has a full-time job”
“A husband’s job is to earn money; a wife’s job is to look after the home and family.” (Understanding Society: Innovation Panel, Waves 1-11. 2019)
Wang (2019) uses polychoric factor analysis on responses to these three questions, which aim to assess whether the respondent holds traditional gender role attitudes, to extract one factor which explained 71.7% of the total variance in answers between questions. They then ran logistic regression models examining the relationship between these answers, a binary LFP variable (whether the respondent was part of the labour force) and age, generation cohort and education level. The key findings from this paper were that gender-role attitudes can explain part of the differing LFP rates among different ethnic groups in the UK; ‘roughly half’ of the significantly lower LFP rates amongst Pakistani and Bangladeshi women can be explained by ‘traditional gender role attitudes’. (Wang, 2019) A clear relationship between education and LFP rates are also found in this paper.
However, Wang’s modelling fails to properly consider the potential impact that education can have in this relationship. Both hypotheses presented in the paper involved the assumption of a relationship between education and LFP, but this relationship is not justified with reference to relevant literature. As demonstrated below, a wealth of literature has, however, established that there is a relationship between education and gender-role attitudes. Therefore, there is the potential for an interaction between these two variables or even reverse causality in the modelling from Wang.
Research into the relationship between education and gender-role attitudes was carried out. Theoretical backing for this relationship can be found from Fortin (2005), who found women’s gender-role attitudes are formed in youth due, due to level of parental education among other things. Thornton et al (1983) linked more egalitarian gender-role attitudes with education, specifically that education results in more liberal attitudes among women (and found a husband’s education level can cause change in the gender-role attitudes of women). Boehnke & Mandy (2011 clearly demonstrated the relationship between higher levels of educational attainment and more liberal gender roles, while this link was also made specifically with ‘higher education’. (Thornton et al 1983)
Data
Wave 2 (2010-2011) of UKHLS - Stratified and clustered General Population sample of around 40,000 households. Includes ethnic minority boost (EMB) to provide higher sample size in the five largest ethnic minorities.
Measures
Education - recoded to 3 way variable; Below secondary school, secondary school or equivalent or Higher.
Age - under 21’s and over 65’s removed as the measures of education used considering cannot all be achieved before this age.
Likert scale measures of the three ‘gender-role’ attitude questions;
“A pre-school child is likely to suffer if his or her mother works”
“All in all, family life suffers when the woman has a full time job”
“A husband’s job is to earn money; a wife’s job is to look after the home and family.”
Labour force participation - binary variable coded as ‘employed’, ’ready to start work or searching for work. (Wang, 2019)
Gender - binary Male/Female
Ethnicity - ethnicity coded to match Wang (2019) - smaller ethnic minorities not included in the EMB removed due to the
Sample Sample of ~30,000 21-65 year olds who responded to the questions regarding gender role attitudes and were either white or part of the 5 EMB groups.This was reduced to ~28,000 for the regressions due to removal of non-responses.
Three ordinal logistic regression models were run, the results of which are shown below. These modelled the effect of education level (along with gender, age and LFP) on the likelihood of a positive/negative response to gender-role attitudes questions from Wave 2 of the UKHLS.
### Data manipulation ###
df1 = read.spss("C:\\Uni\\Second Year\\Semester 2\\Quants\\Understanding society data\\UKDA-6614-spss\\spss\\spss24\\ukhls_w2\\b_indresp.sav", to.data.frame=TRUE)
#Coding education to 3 way
table(df1$b_qfhigh_dv)
df1$education <- ifelse(df1$b_qfhigh_dv == "missing", "missing",
ifelse(df1$b_qfhigh_dv == "inapplicable", "None",
ifelse(df1$b_qfhigh_dv == "Higher degree", "Higher",
ifelse(df1$b_qfhigh_dv == "1st degree or equivalent", "Higher",
ifelse(df1$b_qfhigh_dv == "Diploma in he", "Higher",
ifelse(df1$b_qfhigh_dv == "Teaching qual not pgce", "Higher",
ifelse(df1$b_qfhigh_dv == "Nursing/other med qual", "Higher",
ifelse(df1$b_qfhigh_dv == "Other higher degree", "Higher", ifelse(df1$b_qfhigh_dv == "A level", "High school or lower",
ifelse(df1$b_qfhigh_dv == "Welsh baccalaureate", "High school or lower",
ifelse(df1$b_qfhigh_dv == "I'national baccalaureate", "High school or lower",
ifelse(df1$b_qfhigh_dv == "AS level", "High school or lower",
ifelse(df1$b_qfhigh_dv == "Highers (scot)", "High school or lower",
ifelse(df1$b_qfhigh_dv == "Cert 6th year studies", "High school or lower",
ifelse(df1$b_qfhigh_dv == "GCSE/O level", "High school or lower",
ifelse(df1$b_qfhigh_dv == "CSE", "High school or lower",
ifelse(df1$b_qfhigh_dv == "Standard/o/lower", "High school or lower",
ifelse(df1$b_qfhigh_dv == "Other school cert", "High school or lower", "None"))))))))))))))))))
table(df1$education)
df1.1 <- subset(df1, education != "missing")
table(df1.1$education)
df1.1$education.1 <- ifelse(df1.1$education == "High school or lower", "Secondary School or equivalent", df1.1$education)
df1.1$education.1 <- factor(df1.1$education.1, levels = c("None", "Secondary School or equivalent", "Higher"), ordered = TRUE)
table(df1.1$education.1)
#Remove over 65 and under 18
df1.1$b_age_dv <- as.numeric(df1.1$b_age_dv)
df1.2 <- subset(df1.1, b_age_dv < 65)
df1.4 <- subset(df1.2, b_age_dv >= 21)
table(df1.4$b_age_dv)
#Remove none EMB minority ethnic groups
df1.4$b_ethn_dv <- ifelse(df1.4$b_ethn_dv == "british/english/scottish/welsh/northern irish", df1.4$b_ethn_dv,
ifelse(df1.4$b_ethn_dv == "indian", df1.4$b_ethn_dv,
ifelse(df1.4$b_ethn_dv == "pakistani", df1.4$b_ethn_dv,
ifelse(df1.4$b_ethn_dv == "bangladeshi", df1.4$b_ethn_dv,
ifelse(df1.4$b_ethn_dv == "caribbean", df1.4$b_ethn_dv,
ifelse(df1.4$b_ethn_dv == "african", df1.4$b_ethn_dv, NA ))))))
df1.4$ethnic <- df1.4$b_ethn_dv
df1.5 <- subset(df1.4, ethnic != 'NA')
table(df1.5$ethnic)
#Code LFP variable
summary(df1.5$b_employ)
df1.5$LFP <- ifelse(df1.5$b_employ == "yes", 1, 0)
summary(df1.5$b_julk4wk)
df1.5$LFP <- ifelse(df1.5$b_julk4wk == "yes", 1, df1.5$LFP)
summary(df1.5$b_jubgn)
df1.5$LFP <- ifelse(df1.5$b_jubgn == "yes", 1, df1.5$LFP)
table(df1.5$LFP)
#Coding sex
table(df1.5$b_scsex)
df1.5$sex <- ifelse(df1.5$b_scsex == "male", "0 - Male",
ifelse(df1.5$b_scsex == "inapplicable", "inapplicable",
ifelse(df1.5$b_scsex == "proxy", "inapplicable", "1 - Female")))
table(df1.5$sex)
df1.6 <- subset(df1.5, sex != "inapplicable")
### Making my Dataset for modelling ###
myvars <- c("education.1","LFP", "b_age_dv", "b_scsex", "b_jbstat", "sex",
"ethnic", "b_scopfamf", "b_scopfama", "b_scopfamb")
modeldf <- df1.6[myvars]
### Changing education to ordered ###
table(modeldf$education.1)
modeldf$education.2 <- ifelse(modeldf$education.1 == "None", "a",
ifelse(modeldf$education.1 =="Secondary School or equivalent", "b" , "c"))
table(modeldf$education.2)
##Coding 'Husband should earn, wife should stay at home' variable
table(modeldf$b_scopfamf)
modeldf$h.earn.w.home <- ifelse(modeldf$b_scopfamf == "inapplicable", "none",
ifelse(modeldf$b_scopfamf == "proxy", "none",
ifelse(modeldf$b_scopfamf == "don't know", "none",
ifelse(modeldf$b_scopfamf == "strongly agree", "strongly agree",
ifelse(modeldf$b_scopfamf == "agree", "agree",
ifelse(modeldf$b_scopfamf == "neither agree/disagree", "neither agree/disagree", ifelse(modeldf$b_scopfamf == "disagree", "disagree", "strongly disagree")))))))
table(modeldf$h.earn.w.home)
modeldf.1 <- subset(modeldf, h.earn.w.home != "none")
modeldf.1$h.earn.w.home <- factor(modeldf.1$h.earn.w.home, levels=c("strongly disagree", "disagree", "neither agree/disagree", "agree", "strongly agree"), ordered=TRUE)
table(modeldf.1$h.earn.w.home)
#Coding 'Child suffers when mother works'
table(modeldf$b_scopfama)
modeldf$ch.suff.mo.wrk <- modeldf$b_scopfama
modeldf$ch.suff.mo.wrk <- ifelse(modeldf$b_scopfama == "inapplicable", "none",
ifelse(modeldf$b_scopfama == "proxy", "none",
ifelse(modeldf$b_scopfama == "don't know", "none",
ifelse(modeldf$b_scopfama == "strongly agree", "strongly agree",
ifelse(modeldf$b_scopfama == "agree", "agree",
ifelse(modeldf$b_scopfama == "neither agree/disagree", "neither agree/disagree", ifelse(modeldf$b_scopfama == "disagree", "disagree", "strongly disagree")))))))
modeldf.2 <- subset(modeldf, ch.suff.mo.wrk != "none")
modeldf.2$ch.suff.mo.wrk <- factor(modeldf.2$ch.suff.mo.wrk, levels=c("strongly disagree", "disagree", "neither agree/disagree", "agree", "strongly agree"), ordered=TRUE)
table(modeldf.2$ch.suff.mo.wrk)
#Coding family suffers if nother works full time
table(modeldf$b_scopfamb)
modeldf$f.suff.mo.work <- modeldf$b_scopfamb
modeldf$f.suff.mo.work <- ifelse(modeldf$b_scopfamb == "inapplicable", "none",
ifelse(modeldf$b_scopfamb == "proxy", "none",
ifelse(modeldf$b_scopfamb == "don't know", "none",
ifelse(modeldf$b_scopfamb == "strongly agree", "strongly agree",
ifelse(modeldf$b_scopfamb == "agree", "agree",
ifelse(modeldf$b_scopfamb == "neither agree/disagree", "neither agree/disagree", ifelse(modeldf$b_scopfamb == "disagree", "disagree", "strongly disagree")))))))
modeldf.3 <- subset(modeldf, f.suff.mo.work != "none")
modeldf.3$f.suff.mo.work <- factor(modeldf.3$f.suff.mo.work, levels=c("strongly disagree", "disagree", "neither agree/disagree", "agree", "strongly agree"), ordered=TRUE)
table(modeldf.3$f.suff.mo.work)### Modelling Dependent 1 ###
multi1 <- polr(h.earn.w.home ~ education.2 + sex + b_age_dv + LFP, data = modeldf.1, Hess = TRUE)
multi1.rrr <- exp(coef(multi1))| Dependent variable: | |
| A husband’s job is to earn money; a wife’s job is to look after the home and family | |
| Secondary school or equivalent | 0.904*** (0.026) |
| Higher education | 0.545*** (0.028) |
| Female (against Male) | 0.738*** (0.022) |
| Age | 1.010*** (0.001) |
| Labour Force Participation (Yes against No) | 0.530*** (0.028) |
| Observations | 28,132 |
| Note: | p<0.1; p<0.05; p<0.01 |
The first model shows that, for those with Secondary education or equivalent, the odds of favouring traditional gender roles (agree or strongly agree) is 0.904 compared to those with no education level; a higher level of education makes one less likely to favour traditional gender roles. This trend is also seen with Higher education; the odds for this is 0.545, meaning that having Higher education level is statistically linked with more liberal gender-role views. These values are all statistically significant.
### Modelling Dependent 2 ###
multi2 <- polr(ch.suff.mo.wrk ~ education.2 + sex + b_age_dv + LFP, data = modeldf.2, Hess = TRUE)
multi2.r <- exp(coef(multi2))| Dependent variable: | |
| A pre-school child is likely to suffer if his or her mother works | |
| Secondary school or equivalent | 1.271*** (0.026) |
| Higher education | 1.089*** (0.027) |
| Female (against Male) | 0.543*** (0.022) |
| Age | 1.014*** (0.001) |
| Labour Force Participation (Yes against No) | 0.636*** (0.027) |
| Observations | 28,104 |
| Note: | p<0.1; p<0.05; p<0.01 |
The second model however shows odds of 1.271 / 1.089 for Secondary / Higher education, both statistically significant, which shows that for this question having a higher education level is linked with being more likely to agree with the statement (albeit a relatively small odds).
## Modelling Dependent 3 ###
multi3 <- polr(f.suff.mo.work ~ education.2 + sex + b_age_dv + LFP, data = modeldf.3, Hess = TRUE)
multi3.r <- exp(coef(multi3))| Dependent variable: | |
| All in all, family life suffers when the woman has a full time job | |
| Secondary school or equivalent | 1.233*** (0.026) |
| Higher education | 1.037*** (0.027) |
| Female (against Male) | 0.849*** (0.022) |
| Age | 1.014*** (0.001) |
| Labour Force Participation (Yes against No) | 0.699*** (0.027) |
| Observations | 28,099 |
| Note: | p<0.1; p<0.05; p<0.01 |
The third model shows near very similar results to the second model; those with Secondary education are more likely that those with no education to agree with the statement expressing traditional gender-role attitudes.
In all three of these models, gender and LFP behave as expected; being in work and being Female make one more likely to disagree with traditional gender role attitudes (all statistically significant). The effect of Age is very small, however it is statistically significant and we see that with an increase in age, the odds of agreeing with traditional gender role attitudes increases.
The results of these regression analyses demonstrate that there is a statistical relationship between gender-role attitudes and education when using UK data. However, this relationship is complex and further investigation is required. While a relationship is shown using this data, it is unclear whether the link is caused by more a more complex combination of factors not investigated in this modelling. Using education as an explanatory variable when considering gender role attitudes, and Wang (2019) does, without further understanding of this relationship, could produce flawed analysis.
References
Wang, S. 2019. The Role of Gender Role Attitudes and Immigrant Generation in Ethnic Minority Women’s Labor Force Participation in Britain. Sex Roles 80, 234-245 (2019).
Yassine Khoudja and Fenella Fleischmann, Gender Ideology and Women’s Labor Market Transitions Within Couples in the Netherlands, Journal of Marriage and Family, 80, 5, (1087-1106), (2018)
Fortin, N. 2005. Gender Role Attitudes and the Labour-market Outcomes of Women across OECD Countries. Oxford Review of Economic Policy, Volume 21, Issue 3, 416-438.
Thornton, A , Alwin D & Camburn D. 1983. Causes and Consequences of Sex-Role Attitudes and Attitude Change. American Sociological Review, Vol. 48, No. 2 (Apr., 1983), pp. 211-227. American Sociological Association
Thornton, A & Freedman D. 1979. “Changes in the Sex-Role Attitudes of Women, 1962-1977: Evidence From a Panel Study.” American Sociological Review 44:831-42.
Boehnke, Mandy. 2011. Gender Role Attitudes around the Globe: Egalitarian vs. Traditional Views. Asian Journal of Social Science, vol. 39, no. 1, 2011, pp. 57-74. JSTOR, www.jstor.org/stable/43500538. Accessed 4 May 2020.
CASSIDY, M. L., & WARREN, B. O. 1996. FAMILY EMPLOYMENT STATUS AND GENDER ROLE ATTITUDES: A Comparison of Women and Men College Graduates. Gender & Society, 10(3), 312-329
University of Essex, Institute for Social and Economic Research, NatCen Social Research, Kantar Public. (2019). Understanding Society: Waves 1-9, 2009-2018 and Harmonised BHPS: Waves 1-18, 1991-2009. [data collection]. 12th Edition. UK Data Service. SN: 6614, http://doi.org/10.5255/UKDA-SN-6614-13.
University of Essex, Institute for Social and Economic Research. (2019). Understanding Society: Innovation Panel, Waves 1-11, 2008-2018. [data collection]. 9th Edition. UK Data Service. SN: 6849, http://doi.org/10.5255/UKDA-SN-6849-12.