Katherine Ryan

Max 201

Final Paper: Incomes of Americans

12/3/2019

1
 Introduction

Many claim that money can’t buy happiness, however, money can put people in a better place to

have a more comfortable lifestyle ultimately leading to happier lives. Income inequality in

America affects millions who just happen to be the wrong race, gender or even age. The

consequences of income inequality go beyond the discomfort of individuals. Cultures that

experience high variations of income inequality will also experience an increase in crime,

increase in political inequality, decrease in quality of health, and a decrease in education

attainment levels. All these consequences can also be cyclical, meaning as these problems get

worse income inequality will also increase. This is why we must monitor incomes among various

sub-groups, so we are aware of income inequality and then the policies we need to implement in

order to resolve the problem.

Prior research on this subject has been done by Benjamin J. Newman and his colleagues as they

researched class awareness in local income inequalities and its impact on policymaking and

implications for the American political system. This is documented in his journal article “False

Consciousness or Class Awareness? Local Income Inequality, Personal Economic Position, and

Belief in American Meritocracy”. Newman found that despite growing income inequality

Americans do not tend value economic redistribution (Newman, et al., 2015). For this credit is

given to the principles of the American Dream that through hard work and doing the right things

Americans can still reach high levels of income (Newman, et al., 2015). This research was

conducted by merging four surveys conducted by the Pew Research Center: the 2005 News

Interest Index Poll, the 2006 Immigration Survey, the 2007 Values Survey and the 2009 Values

2
Survey (Newman, et al., 2015). They looked to find values of meritocracy among the

respondents in these surveys.

To further understand the issue of income inequality I was curious what demographics, if any,

contribute to an individual’s income level in the United States. This became the foundation of

research for this paper. Using the 2016 SPSS data, information was gathered to see what effects

different independent variables would have on income16 or the income of the respondents. These

independent variables included age, sex, race, birth status, educational attainment, and marital

status. Hypotheses are structured around individuals who will have higher incomes, to see a

complete list of null and alternative hypotheses please see the appendix. For age, those older will

have a higher income. For sex, males will have a higher income. For race, white individuals will

have a higher income. For birth status, those born in America will have a higher income. For

educational attainment those with higher levels of educational attainment will have a higher

income. For marital status, those who are married will have a higher income.

Methods1

In the SPSS codebook the variable age corresponds to the respondent’s age. It is coded by

sampling using the respondents age as the response. To better fit the purpose of my research this

variable was recoded by creating a new variable which condensed the answers. The new variable

for age has 3 groups the first ranging from 18-39, the second 40-59, and the third 60+. These

three groups were created to encompass younger respondents, middle age respondents and older

or eldered respondents.

1
All variables first included a declaration of missing values in the GSS data file.

3
The next variable sex has two responses: male and female. No recoding was necessary as it

encompasses what is needed for the purpose of research.

The next independent variable race includes three responses: white, black, and other. This

required no recoding because of the simple structure of the question in the SPSS.

The variable of birth status is referred to as thee variable BORN in the codebook. This measures

whether or not a respondent was born in the United States. Answers include “yes”, “no”, “don’t

know” and “no answer”. For the purposes of this research “don’t know” and “no answer” were

declared as missing values so only “yes” and “no” would be recorded.

The variable educational attainment in the codebook is referred to as DEGREE. The question

asks the respondent’s degree. Answers include “less than high school”, “high school”,

“associates or junior college”, “bachelors”, “graduate”, “don’t know” and “no answer”. For the

purposes of this research “don’t know” and “no answer” were declared as missing values. No

further recoding was necessary.

The variable marital status is referred to as MARITAL in the codebook. It asks if a respondent is

currently--married, widowed, divorced, separated, or have you never been married? Answers

include “married”, “widowed”, “divorced”, “separated”, “never married” and “no answer”. To

simplify “no answer” was declared as a missing value. The answers were then condensed into

two categories: married and single/separated. Married includes only the response married.

Single/separated includes widowed, divorced, separated and never married.

4
The dependent variable is income, referred to as INCOME16 in the SPSS codebook. There are

27 numerical responses coded in the book starting at under $1,000 to $170,000 and over. Using

Investopedia’s qualifications income was recoded into a new variable with three responses being

low income, middle income and upper income (Investopedia). Low income refers to an answer

below $25,000, middle income refers to, 25,000 and above and below 90,000, and upper income

refers to $90,000 and above. This created a much simpler yet effective foundation to conduct

research.

To discover significance level for all of the data sets chi-squared tests were run. This is because

of the categorical nature of the recoded version of INCOME16. The chi-square test allowed for

more accurate results regarding income levels.

Results2

To first understand the distribution of income levels, I examined the frequency of each income

level. As presented by the pie graph low income accounted for 26.9%, middle accounted for

55.8%, and upper income accounted for 17.3%. Results that are as expected with middle income

taking up the majority of the responses.

2
All Crosstabs and data sets can be located in the Appendix.

5
 Chart 1: Income Distribution

Upper
IncomeLow
17% Income
27%

Middle
Income
56%

Low Income Middle Income Upper Income

The first crosstab is between the independent variable AGE and the dependent variable

INCOME16. Overall 17.3% of respondents have what is considered an upper income. This data

shows the middle-aged group of 40 to 59 having the highest percentage (22.9) for having the

highest income. Compared to the group 18-39 who only have 13.9% of having an upper income.

This data indicates a higher chance of being upper income if the respondent is middle-aged. The

null hypothesis that there is no relationship between age and income level must be rejected. This

means the difference can be considered statistically significant. This can be proven if the

asymptotic value is less than .05 than the difference is statistically significant. For this case the

asymptotic value is 0.0, which is much less than .05. In the case of the rejected null hypothesis

we do not support the alternative hypothesis that the older the respondent the higher the income

instead the data suggests if a respondent is middle-aged, they are more likely to have an upper

income.

6
For the variable SEX, 17.3% of the total population are considered to have an upper income.

However, men (19.4%) have a higher chance than women (15.6%) to have an upper income. The

asymptotic value for this data is 0.0, which is under .05. This shows there is a significant

difference in the data among the sexes. Therefor we must reject the null hypothesis that there is

no relationship between sex and income level and instead accept the alternative hypothesis that

men are more likely to have a higher income.

Chart 2: Race on Income Level

25.00%
Percantage of Upper Income

20.00%

15.00%

10.00%

5.00%

0.00%
Total White Black Other
Race

For the variable, RACE, 17.3% of the total population are considered to have an upper income.

However white respondents (19.7%) have the highest percentage of upper income respondents

compared to black (6.5%) and other respondents (18.0%). The asymptotic value for this data is

0.0, which is under .05. This shows there is a significant difference in the data among the races.

Therefor we must reject the null hypothesis that there is no relationship between race and income

level. Instead we support the alternative hypothesis that white respondents are more likely to

have a higher income level.

7
For the variable BORN 17.3% of the total population are considered to have an upper income.

There is a slight favor of those who were born in the United States (17.1%) compared to those

who were not (19.1%) to have a higher income. Despite these differences the asymptotic value is

.093 value above the .05 threshold. This shows the difference between being born in the United

States or not, in terms of income, are not statistically significant. In this case we support the null

hypothesis that there is no relationship between whether a respondent was born in the United

States or not.

For the variable BORN 17.4% of the total population are considered to have an upper income.

The highest percentage of respondents who have a high income is seen at the graduate degree

level (45.1%) compared to those with less than a high school degree (5.0%). The asymptotic

value for this data is 0.0, which is under .05. This shows there is a significant difference in the

data among degree levels. In this case we must reject the null hypothesis that there is no

relationship between degree level and income level. Instead we support the alternative

hypothesis that the higher a degree level the more likely an individual is to have a higher income.

For the variable MARITAL 17.3% of the total population are considered to have an upper

income. The highest percentage of respondents who have a high income is seen at the married

level (30.6%) compared to those single (7.5%) The asymptotic value for this data is 0.0, which is

under .05. This shows there is a significant difference in the data among marital statuses. In this

case we must reject the null hypothesis that there is no relationship between marital status.

Instead we support the alternative hypothesis that married respondent is more likely an

individual is to have a higher income.

8
To determine if there were any factors that should be considered more powerful than others a

three-way cross tab was run. This cross tab held race constant while determining the relationship

between sex and income level. Overall 17.3% of respondents recorded having a high income. In

general, males (19.4%) have a higher percentage of high incomes than females (15.6%). The

asymptotic value for this data is 0.006, which is under .05. This shows there is a significant

difference in the data among the sexes, while holding race constant. However, to take a greater

look at this, white men (21.7%) are more likely to have a higher income than white women

(17.9). The asymptotic value for this data is 0.038, which is under .05. This shows there is a

significant difference in the data among the sexes, while holding race constant. Black men

(10.3%) are less likely to have a higher income than white men, but still much more likely than

black women (3.6%). The asymptotic value for this data is 0.019, which is under .05. This

shows there is a significant difference in the data among the sexes, while holding race constant.

However, among other races men (14.7%) are less likely to have a higher income than women

(20.4%). The asymptotic value for this data is 0.192, which is above .05. This shows there is no

significant difference in the data among the sexes, while holding race constant. Despite the lack

of a significant difference it must be noted that when holding race constant there was a deviation

from the typical pattern for other races.

Conclusion

In summation, based on the data presented here, factors that contribute to income level include

age, sex, race, educational attainment, and marital status. Each of these variables showed

significant differences between their responses leading to the conclusion that they have an effect

on income level in some variation. The only variable that did not have a significant difference

9
was nativity status or the variable BORN. This leads to the conclusion that is has no effect on

income level.

One major limitation is that the GSS is a cross- sectional survey, but research was only

conducted from the 2016 data set. This gives a much smaller pool of data to work with and could

potentially be more or less representative of the American public depending on the year. This

could potentially lead to inaccuracies. Another limitation can be recoding data in which data gets

grouped into large categories. This can potentially drown out key data that will be missed. For

instance, for many reasons age was grouped into three categories. However, these categories still

include a very wide range of ages in which respondents would have drastically different

responses. Take the first age group of 18 to 39, the response of a college aged response is

typically going to be very different from responses from the 39-year-old respondent. This could

skew data and eventually lead to inconclusive results. Another limitation is the wording of the

variable INCOMEE16, which we base the majority of the results off of. In the GSS the question

asks about family income, rather than individual income. Though rare it can be the case that an

individual may have a low income, but another member of the family may make enough to

provide for the family. A more accurate representation of data might be seen had the GSS asked

about personal income rather than family income.

This is a good start to an education on factors that lead to income, which can ultimately impact

income inequality. By discovering factors that tend to make certain individuals more unequal

than others we can then create and implement policies to address these issues. Additional

research could be done yearly, as income inequality steepens each year. Another research option

10
could be to mimic the type of work Newman did. By using the GSS one could use income as the

independent variable then have dependent variables that relate to government intervention, fiscal

redistribution and even the American Dream.

11
 Appendix

Research Question: What demographics contribute to income level?

AGE
(H0) There is no relationship between age and income level in the United States
(HR) An older respondent is more likely to have a higher income level.

SEX
(H0) There is no relationship between sex and income level in the United States
(HR) A male respondent is more likely to have a higher income level.

RACE
(H0) There is no relationship between race and income level in the United States
(HR) A white respondent is more likely to have a higher income level.

NATIVITY STATUS
(H0) There is no relationship between whether a person was born in the United States and
income level in the United States
(HR) An American born respondent is more likely to have a higher income level.

EDUCATIONAL ATTAINMENT
(H0) There is no relationship between degree level and income level in the United States
(HR) A respondent with a higher level of degree is more likely to have a higher income level.

MARITAL STATUS
(H0) There is no relationship between marital status and income level in the United States
(HR) A married respondent is more likely to have a higher income level.

12
Original questions as they appear in the 2016 General Social Survey:

Note: Unless mentioned otherwise, any response option of “don’t know” or “no answer” was
dropped from analysis. These options are also not mentioned as response options below.

Respondents Age: Respondents were asked for their data of birth and the GSS recoded these
into the age of the respondent at the time they took the survey.

Response options ranged from 18 to 89, where 89 means anyone 89 years or older.

Recoding choices: This was recoded once to create three categories, one 18-39 for a young
category, two, 40-59 for a middle-aged category, and lastly 60+ for an elderly category. This was
done to simplify all the date presented and to protect the accurate results of a bivariate table.

Respondents Sex: Respondents were asked to identify their sex.

Response options were:

- male
- female

This variable did not need recoding.

Respondents Race: Respondents were asked to identify their race.

Response options were:

- White
- Black
- Other

This variable did not need to be recoded.

Nativity Status: Respondents were asked whether they were born inside the United States.

Response Options Were:

- yes
- no

This variable did not need recoding.

13
Educational Attainment: Respondents were asked to identify the highest degree they had
achieved in their education.

Response options were:

- Less than high school
- High school
- Associate/Junior college
- Bachelor’s
- Graduate
This variable did not need to be recoded.

Marital Status: Respondents were asked if they were currently married, widowed, divorced,
separated or never married.

Response options were:

- Married
- Widowed
- Divorced
- Separated
- Never Married
This was recoded to have two responses Married and Single/Separated. Married includes only he
response married. Single/separated includes widowed, divorced, separated and never married.
This was to simplify the data presented in the bivariate tables.

Respondents Income: Respondents were asked to group their family income, from all sources,
before taxes.

Response options ranged from less than $1000 to over $170,000.

Income was recoded into a new variable with three responses being low income, middle income
and upper income. Low income refers to an answer below $25,000, middle income refers to,
25,000 and above and below 90,000, and upper income refers to $90,000 and above. This created
a much simpler yet effective foundation to conduct research.

Appendix Table 1

14
Appendix Table 2

Appendix Table 3

Appendix Table 4

Appendix Table 5

Appendix Table 6

Appendix Table 7

15
 Appendix Table 8

Appendix Table 9

Appendix Table 10
16
Appendix Table 11

Appendix Table 12

17
Appendix Table 13

Appendix Table 14

18
 References

19
Investopedia. (2019, November 20). Income and Social Class Report | Course Researchers.

Retrieved December 2, 2019, from https://courseresearchers.com/income-and-social-class-

report/.

Newman, B. J., Johnston, C. D., & Lown, P. L. (2015). False Consciousness or Class Awareness?

Local Income Inequality, Personal Economic Position, and Belief in American

Meritocracy. American Journal of Political Science, 59(2), 326–340. doi: 10.1111/ajps.12153

20
21
22