SC208 Lab 3 – Week 8

0 Comments

SC208 Lab 3 – Week 8
Please supplement these instructions with Field Chapter 8
1. Data and Start-Up
The goal of this week is to look at the link between major sources of social stratification – class, ethnicity, and gender – and inequality in wages, family type, housework, and attitudes towards gender. To do this, we are using a subset of the first two waves of Understanding Society gathered in 2010-2012. Data from wave 1 has the prefix a_ and data from wave 2 has the prefix b_. Understanding Society is a large multipurpose survey containing information on work, health, attitudes, education, and family, among other topics. More information on the dataset is found here:
https://www.understandingsociety.ac.uk/documentation/mainstage/dataset-documentation
I have restricted the sample to include only prime age adults ages 25 to 55 in wave 1, since our current topic considers issues of work-family balance which are most important for people in these age ranges.
The data has been cleaned for you – most of the observations with missing data have been removed. However it is important to note a few things:
1. Some questions ask about division of labour among couples. This information is missing and denoted with a -1 in the data for those who are not part of a couple.
2. Some questions only make sense for people who work – for instance monthly pay =0 if you are not working.
3. For some questions, for instance father’s education, “other” was chosen and in denoted with a value like 97
Basically, it is very important to check the frequencies and distributions of your variables to make sure they are coded in a way that makes sense. For instance, you might not want to do a regression that includes those who are not in work if you are interested in wage inequality among workers; you also want to make sure to exclude those who are single if you are interested in looking at division of labour among couples.
Finally, I have also created two variables for you that are not currently in the survey documentation.
These are:
a) Gender attitudes [tradideas] : this is a sum of the level of agreement (1 highest -5 lowest) to the following statements:
a. Pre-school child suffers if mother works
b. Family suffers if mother works full time
c. Husband and wife should contribute to household income (reverse coded)
d. Husband should earn, wife should stay at home
Higher scores denote less traditional gender role ideology.
b) Household Labour[hhtasks]: This is a sum of the number of tasks that are “mostly” done by the respondent (rather than shared or mostly by their partner):
a. Grocery shopping
b. Cooking
c. Cleaning
d. Washing/Ironing
e. Gardening
f. DIY
Higher numbers mean more tasks.

To start SPSS, go to the Start icon and find SPSS 23 under ‘All programs’. You can also start SPSS by double-clicking on an SPSS file. The data file Assignment 3 Data.SAV will be used today. Open a web browser and go to the sc208 Moodle website. Under Week 8 you’ll see the data ready for download. When it’s downloaded, simply click on the SPSS file and SPSS will open up.
2. Think about relationships between wages, housework, and attitudes towards gender roles and stratification variables such as social class, gender, ethnicity/immigration status and age. The readings and lectures from the past two weeks should be helpful for this. How would you measure this with these data?
Examples of topics:
a. SES background of respondent and wages
b. Qualification of respondent and attitudes towards gender
c. Number of hours worked and gender division in household for women
d. Qualification of respondent and gender division
e. Number of children in the home and attitudes towards gender
f. Marital status and wages
g. Gender attitudes and well-being
h. Wages and well-being
i. Relationship status and well-being
However there are many more you can look at with this data!
After exploring the data, it is helpful to state an explicit hypothesis, related to your topic of interest, that is a relationship between two variables that are actually in the data. Generally you will think of an independent variable that you think predicts/has an effect on a dependent variable.
Because we are doing regressions in this lab, your dependent variable needs to be continuous.
An example of a hypothesis:
H1: Because traditional gender roles are generally more constraining for women than men, I anticipate that women [a_sex_dv] will be have less traditional gender role ideas [tradideas] than men.
Write your hypothesis in a word document.
Just like last time, we are also going to want to think of a variable which may mediate or confound this relationship. However, will not look at moderators this time, because that involves interaction effects. We may address interaction effects in the next assignment.
An example:
H1: Because traditional gender roles are generally more constraining for women than men, I anticipate that women [sex] will be have less traditional gender role ideas [tradideas] than men.

H2: However, women also may be more highly educated than men, and more highly educated individuals [a_hiqual_dv] may have less traditional gender role ideas. Thus, educational attainment may be an important mediator in gender differences in gender role ideas.

3. Check out the frequencies and distributions of your variables

A) Continuous variables are best examined with histograms, and descriptive statistics
B) Dichotomous or categorical variables are best examined with frequencies

For instance, I go to Analyze, then Descriptive Statistics, then Descriptives. I choose tradideas and press OK.

Descriptive Statistics
N Minimum Maximum Mean Std. Deviation
Agreement with egalitarian attitudes 12104 4 20 13.50 3.040
Valid N (listwise) 12104

As you can see the scale runs from 4 to 20, with a mean of 13.5. The standard deviation is not particularly high, so most responses are clustered around the mean. You can also see this with a histogram.

You will want to look at all of your variables in this way. Some of them may have missing data, or as mentioned above, only apply to a subset of the data. If so you will need to use
Data  Select Cases to omit observations with missing data.
For example, household tasks (hhtasks) is only valid for people in couples:

Num HH tasks mostly by respondent
Frequency Percent Valid Percent Cumulative Percent
Valid -1 2984 24.7 24.7 24.7
0 874 7.2 7.2 31.9
1 1965 16.2 16.2 48.1
2 2414 19.9 19.9 68.1
3 1636 13.5 13.5 81.6
4 1482 12.2 12.2 93.8
5 567 4.7 4.7 98.5
6 182 1.5 1.5 100.0
Total 12104 100.0 100.0

You see that 2984 of the cases are -1 because the respondent is not in a couple.
To restrict the sample to couples only, you can go to DataSelect Cases Select cases if  hhtasks>=0

4) Going back to our original question, we can take a first exploratory look at my hypothesis about gender and gender role attitudes by using a comparison of means. I do this by going to Analyze – Compare Means. I then put gender in the independent list and tradideas in the dependent list and press OK. In general, there doesn’t seem to be a big difference: only about 0.35 on the traditional ideas scale. Another way to look at the difference is that it is over a tenth of a standard deviation for the whole sample (0.35 is a little more than 1/10th of 3.040) As expected, women have more egalitarian gender role ideas than men.

Report
Agreement with egalitarian attitudes
Sex, derived Mean N Std. Deviation
Male 13.30 5202 3.019
Female 13.65 6902 3.047
Total 13.50 12104 3.040

I then take a look at how education is related to gender role attitudes. As you can see below, there is a fairly steady positive relationship between level of qualification and egalitarian attitudes.

Report
Agreement with egalitarian attitudes
Highest qualification ever reported Mean N Std. Deviation
Degree 13.86 4046 3.068
Other higher 13.56 1736 3.054
A level etc 13.58 2237 2.979
GCSE etc 13.32 2417 2.987
Other qual 12.94 888 3.004
No qual 12.49 780 2.887
Total 13.50 12104 3.040

4. Recode your categorical variables if necessary.
Just like the last two labs, if you have categories with very few cases, you will want to collapse them. Look at the previous two labs for directions on how to do this.

Once you have decided on your final categories, you will now need to code them into dummy variables in order to put them into the regressions.

To do this, I go to Transform -> Create Dummy Variables
I put a_highqual_dv into the “Create dummy variables for” box. I then check the box “Create main effect dummies”.

In the box “root names” I put in “educ_cat”

In the lower left hand side there is a box “Measurement level usage” I select “create dummies for all variables”

I select OK

To make sure it came out right, I crosstabulate qfedhi and educ_cat_1. It came out like below, so I know it’s correct.

Highest qualification ever reported * a_hiqual_dv=Degree Crosstabulation
Count
a_hiqual_dv=Degree Total
.00 1.00
Highest qualification ever reported Degree 0 4046 4046
Other higher 1736 0 1736
A level etc 2237 0 2237
GCSE etc 2417 0 2417
Other qual 888 0 888
No qual 780 0 780
Total 8058 4046 12104

I also create a more intuitive variable for sex, naming it male, with code =1 if man, and =0 if woman.

Sex, derived * male Crosstabulation
Count
male Total
.00 1.00
Sex, derived Male 0 5202 5202
Female 6902 0 6902
Total 6902 5202 12104

5. Show descriptive statistics
You will want to display mean and standard deviation for continuous variables. You can get these by going to Analyze, Descriptive Statistics, and then Descriptives. For categorical and dichotomous variables, you can report the sample percentages for each category, available in Frequencies.

These three tables together form your “descriptive statistics” table, which is the first table you will need for your assignment. You can combine this data in excel to make a single table which is easier to read, like below.
Table 1. Descriptive Statistics of Men and Women ages 25-55, Understanding Society (N=12104)
Mean/Percent Standard Deviation
Traditional Ideas Scale 13.50 3.04
Degree or more .33
Other higher .14
A-levels .18
GCSE .20
Other Qual .07
No Qual .06
Male .43

6. Simple and Multiple Regression
Now, you are ready to do your regressions.
a) Start with a simple regression of your dependent variable on your independent variable. Go to Analyze, select Regression, then select Linear. Place your dependent variable (here, tradideas) in the dependent variable slot, and your independent variable (here, male) in the independent variable slot. Press OK.
The output viewer will give you a model summary and also a table with coefficients like the one below.

Model Summary
Model R R Square Adjusted R Square Std. Error of the Estimate
1 .057a .003 .003 3.035
a. Predictors: (Constant), male

Coefficientsa
Model Unstandardized Coefficients Standardized Coefficients t Sig.
B Std. Error Beta
1 (Constant) 13.651 .037 373.676 .000
male -.350 .056 -.057 -6.288 .000
a. Dependent Variable: Agreement with egalitarian attitudes

You will need to create a table in excel to summarise this information. You should include the unstandardized coefficients, standard errors, and significance level for each variable. You should also include the R2 for the model. I am including an example (table 2) below. The first column is the information for your “model 1”, your simple regression model.

b) Next, we will do a multiple regression. Go to Analyze, select Regression, then select Linear. Place your dependent variable (here, tradideas) in the dependent variable slot, and all your independent variables in the independent variable slot. Remember that if you have a categorical variable you will need to omit one of the categories as the comparison category (so for male, the omitted category is female. For education, I choose to omit those with a degree and include the dummy variables for all the other categories).

The output we get is as follows.

Model Summary
Model R R Square Adjusted R Square Std. Error of the Estimate
1 .136a .018 .018 3.012
a. Predictors: (Constant), a_hiqual_dv=No qual, male, a_hiqual_dv=Other qual, a_hiqual_dv=Other higher, a_hiqual_dv=A level etc, a_hiqual_dv=GCSE etc

Coefficientsa
Model Unstandardized Coefficients Standardized Coefficients t Sig.
B Std. Error Beta
1 (Constant) 14.012 .053 263.036 .000
male -.355 .055 -.058 -6.399 .000
a_hiqual_dv=Other higher -.319 .087 -.037 -3.691 .000
a_hiqual_dv=A level etc -.264 .079 -.034 -3.329 .001
a_hiqual_dv=GCSE etc -.554 .077 -.073 -7.152 .000
a_hiqual_dv=Other qual -.895 .112 -.077 -8.018 .000
a_hiqual_dv=No qual -1.375 .118 -.111 -11.671 .000
a. Dependent Variable: Agreement with egalitarian attitudes

c) Again, you will need to summarise this information in a table in excel. The second column in table 2 includes this information in “model 2”, the multiple regression model.

This is the second table that you need to provide for your assignment.
Table 2. Coefficients from Models of Agreement with Egalitarian Ideas, Men and Women ages 25-55, Understanding Society (N=12104)
B SE Sig. B SE Sig.
Male -.35 .06 .00 -.36 .06 .00
Educational attainment (Degree or more omitted)
Other higher .00 -.32 .09 .00
A-levels .00 -.26 .08 .00
GCSE .00 -.55 .08 .00
Other Qual .00 -.90 .11 .00
No Qual .00 -1.38 .12 .00
Constant 13.65 .04 .00 14.01 .05 .00
R2 .00 .02

Finally, describe the coefficients from both models in light of your hypotheses and discuss.
As you can see in the first model in table 2 above, my H1 is supported, men in general report more traditional gender role ideas than women. The difference in the scale is .350, same as we saw in the comparison of means.

In my H2, I had hypothesized that some of this gender difference might be mediated by educational attainment. If more highly educated respondents espouse more egalitarian gender attitudes, and women are more highly educated than men, than this may mediate the relationship between gender and gender role ideas.

In model 2, I test for this mediating relationship. As anticipated, respondents with a degree or more report, net of gender, more egalitarian attitudes than those with less education. However, educational attainment does not appear to mediate the relationship between gender and gender role ideas, since the coefficient for male remains essentially unchanged from model 1 to model 2. Thus, although educational attainment is independently associated with gender role attitudes, my H2 is not supported, as it does not mediate the relationship between gender and gender role attitudes.

Leave a Reply

Your email address will not be published.