Through the Wire: Understanding Inequality in Household Telephone Access, 1960-2010

Introduction

From its invention in the late 1870s, the telephone has revolutionized the economic and social lives of countless Americans. Today, almost everyone in the United States has access to a telephone, whether in the form of a home phone or personal cell phone. Access to this technology, however, has not always been so widespread. After its invention, the telephone spread from the realms of the government and large corporations to the households of everyday Americans very slowly, only becoming widely adopted in the mid-20th century (Fischer Carroll 1153). The aim of this project is to analyze the distribution of access to telephones between 1960 and 2010. It focuses especially whether or not there existed (or exists) inequality in telephone distribution based on four dimensions of analysis: birthplace, primary language, race, sex, and income. We live in an era in which our lives are defined by technology, access to which can determine how we communicate, the quality of the education and healthcare we receive, and whether or not we are able to succeed in a rapidly evolving economic landscape. This project seeks to shed light on inequality as it relates to the adoption of new technology and access to information, perhaps enabling better understanding of contemporary issues related to the internet and freedom of information.
In order to complete this analysis, I use a combination of ACS and Census data (1% samples for all years). A question related to telephone availability has been part of every Census (long form) and ACS form since 1960, as have questions related to my various demographic dimensions. This project makes liberal use of visualizations, especially faceted bar charts to make comparisons between demographic groups and scatterplots/line graphs to make sense of broader, long-term trends and relationships. It also makes use of regression analysis to study the relationship between two variables: income and race. This analysis makes used of fixed effects for each year of data, hopefully helping clarify the relationships being analyzed.
Available academic literature related to trends in access to and ownership of telephones is quite expansive and surprisingly interdisciplinary. The topic has been studied by scholars from a wide variety of fields, including business, economics, sociology, anthropology, and history. Because my analysis of telephone access and distribution will focus primarily on quantitative measures of inequality in telephone access and ownership, the majority of the literature I chose to focus on is quantitative in nature and focuses on demographics. More specifically, the literature primarily focuses on variation in telephone access and ownership as it relates to race, sex, and age. While the literature available seems to be quite broad in subject matter, many of the studies I found (particularly quantitative ones) were published in the last two decades of the twentieth century, leaving me completely in the dark with regards to the ways distribution of access to and ownership of telephones has changed over time. I hope my project is able to shed some light on that question.
Studies differed dramatically in their conclusions related to racial differences in telephone ownership and access. A study conducted using 1970 census data found the results one might predict when considering racial inequalities in telephone ownership and access: “households with telephones were more likely to have white, male heads of higher average
income, education and age” (Wolfle 421). This result is roughly in line with the most obvious logical explanation for inequality in telephone access. First, it makes sense that income might be related to telephone access because telephones were fairly expensive to install, especially before they became completely ubiquitous. Second, because telephone infrastructure requires significant investment, it is possible primarily African-American communities were passed over when it came time to make these investments, possibly explaining part of the racial disparity in telephone access. More recent studies have found dramatically different results. A 2009 study of low-income individuals living in rural Maryland found “individuals without landline telephones were more likely than those with landline telephones to be white and to have an income of $20000 to <35000” (Shebl et al. 499). The most obvious possible explanation for this difference in result is, of course, the fact that the studies were conducted nearly thirty years apart, with the first study using data nearly forty years older than that of the second study. It seems reasonable to assume that by 2009 cell phones were well on their way to replacing landline telephones in most American homes, perhaps preferred over their household counterparts. The results of the second study also suggest that geography is likely worth taking into consideration or at least including as a possible control.
Much of the literature regarding telephone distribution and access also considers sex as a factor worth considering. One such study, published in 1981 by Bell laboratories, found that women tended to use phones more extensively than men. Although this study isn’t directly related to household telephone access or ownership, disparities in use might certainly be worthwhile to consider in my analysis. Conversely, another study (the Wolfle study cited earlier) found that households with telephones were more likely to be headed my men, while a third (the Shebl study) found no significant difference in access and ownership between men and women. Unlike these studies, my paper analyzes whether these disparities persist over time and whether they might be related to other variables.
While the literature on the topic of telephone ownership and access distribution is quite extensive, I believe my broader analysis of the topic adds significant value. I hope that my project is able to shed some light on the subject from a broader, more data-driven perspective.

Methods

The analysis conducted in this project makes use of the following ipums datasets:
1960 1%, 1970 met1%, 1980 metro1%, 1990 1%, 2000 1%, and the 2010 ACS. My primary analysis will make use of the variables PHONE (whether or not a household has access to a phone), BPL (birthplace), RACE (race), RELATE (relationship to head of household used to identify household heads), SPEAKENG (whether or not an individual speaks English), and SEX (whether a household head is male or female). My secondary analysis will also include CPI99 (income multiplier based on the 1999 Consumer Price Index), FTOTINC (total family income), and AGE (an individual’s age). For all income-related analysis I calculate income using CPI99 to ensure incomes are standardized for reasonable comparison. In my linguistic analysis I divide those who speak no English or describe themselves as speaking English “not well” from those who claim to be native speakers or speak English “well” in order to create my two categories. My racial analyses exclude a Hispanic category due to the dramatic changes in its use and definition over the course of the last half-century. I also aggregate every Asian census category (Vietnamese, Japanese, Chinese, etc.) into a single “Asian” category. All variables are at the household level (code: https://gist.github.com/jocrug/dbb3e0837f7801be9dd972ae7755338e)
This paper’s primary analysis focuses on the relationships between various demographic characteristics and the availability of telephones. Central to this analysis are visualizations. I make use of faceted bar charts to highlight inequalities in telephone distribution over time. These charts facilitate qualitative comparisons between individuals with the various characteristics included in my analysis. I have created one set of faceted bar charts for each dimension of analysis I include, faceting by year to facilitate comparisons in inequality over time. Line graphs and histograms are also used to highlight income inequality and illustrate the relationship between income and average access to telephone use amongst various cohorts. Furthermore, this project’s secondary analysis makes use of more rigorous statistical methods, developing and visualizing multivariate regression models for the spread of telephone availability in the latter half of the twentieth century.
For my secondary analysis I converted several race and demographic variables into dummies for use in regression. I then exported the resulting dataset into STATA, a popular statistical software tool used for a wide variety of social science analysis. I conduct two types of statistical analysis of this data; first, I created a linear regression model of the effect of race and income on household phone access that included year fixed effects and duplicated observations based on weight. While linear models can be used for the analysis of dummy dependent variables, they often fail to fit the data as well as a non-linear model might (Von Hippel). Thus, I also attempted to analyze the data using logistic regression, again weighting by duplicating observations and using fixed effects. However, I was not able to complete this analysis due to apparent computational constraints.

Results

These plots reflect the trends I expected to see based on my review of literature on the topic. Many of the older papers I covered generally found that “households with telephones were more likely to have white, male heads of higher average income, education, and wage” (Wolfle 421). The plots I created observing different trends in telephone ownership by race and sex clearly reflect this disparity, particularly further in the past. Especially notable are the racial disparities in home telephone access, particularly those between Native American and White-headed households. These plots suggest racial and sex-related disparities in telephone ownership decrease as time goes by, consistent with the findings of the newer literature on the topic, such as the Shebl et al. I was unable to find previous literature on my two other dimensions of analysis, English fluency and birthplace. The trends, however, match what one would logically expect.

Screen Shot 2016-11-25 at 11.04.33 PM

My analysis of race and telephone access (fig. 1) focused on disparities between white-headed households and households headed by Native Americans and black Americans, two groups subjected to extreme discrimination throughout the 19th century. The results line up approximately with my expectations; between 1960 and 1990, white-headed households enjoyed significantly greater telephone access than their black and Native American counterparts. This disparity persisted to a much lesser degree after 1990 as telephones became more or less ubiquitous. Although the literature doesn’t offer a possible explanation for this phenomenon, it can likely be attributed to one of two possible causes. The first is related to income. Based on an analysis of IPUMS census data, the average incomes of households headed by black Americans and Native Americans are significantly lower than those of white Americans (fig. 2), perhaps negatively impacting their ability to invest in telephone installation.

Screen Shot 2016-11-25 at 11.05.05 PM

Alternatively, the racial disparity in telephone access could have resulted from differences in the concentration of telephone-related infrastructure between 1960 and 1990. For example, the disproportionate number of Native Americans living in rural areas like reservations could have restricted their access to telecomm infrastructure situated in more urban areas. Similarly, because telephone infrastructure requires significant investment, it is possible governments and corporations discriminated against poorer African-American communities when choosing where to build, further contributing to inequality in household telephone access between black and white Americans. As is reflected in fig. 1, these inequalities decrease as time passes. This decrease in inequality may be a result of decreases in income inequality over the course of the latter half of the twentieth century (fig. 2), paired with decreases in the price of telephone infrastructure and equipment as the technology becomes more widespread. Alternatively, this fall in inequality could possibly be attributed to the decrease in the level of segregation over the course of the last half-century.

Screen Shot 2016-11-25 at 11.05.17 PM

My analysis of birthplace and telephone access focused on disparities (fig. 3) in telephone access between households headed by foreign-born individuals and households headed by native-born individuals. This analysis suggests only a very small (if any) disparity in household telephone access between these two cohorts, with native-born Americans enjoying a slight access advantage over immigrants in the 60s and 70s but immigrants overtaking them slightly every later year but 2000. There are a variety of possible explanations for these trends. The most intuitive possible explanation for this trend is demographic: between 1960 and 2000, immigrants were more likely to live in central cities than their native-born counterparts, possibly allowing them to take advantage of more established technological infrastructure (Boustan et al 133). This inequality may have evened out as telephones became almost ubiquitous even in rural areas.

Screen Shot 2016-11-25 at 11.05.29 PM

My analysis of telephone access by whether or not a household head is fluent in English (fig. 4) similarly found much smaller disparities than my race-based analysis. This plot suggests fluent speakers enjoyed slightly more access to household telephones in 1980 and 1990, but that this disparity practically disappeared starting in 2000. There are a variety of possible explanations for the disparity that existed earlier on. The first possible explanation is income-related; perhaps those without the ability to speak English (and thus less able to participate in the economy) were less likely to be able to invest in expensive telephone equipment and installation. Because household heads not fluent in English were more likely to have lower household incomes, it is possible they had less to invest in telephone installation before the technology became completely ubiquitous (and cheaper), accounting for the disparity that existed before 1990. Alternative explanations could include systematic geographic and demand-related differences between fluent and non-fluent English speakers.

Screen Shot 2016-11-25 at 11.05.41 PM

My analysis of household head sex and telephone access (fig. 5) found that households headed by women were far less likely to have access to a household telephone than their male-headed counterparts. This disparity, like the ones described above, can also easily be attributed to income disparities. Women leading households, especially during the mid-twentieth century, were more likely to be, young, employed and, part of low-income households than their male counterparts. Interestingly, this disparity decreases each decade until 2010, when female-headed households outstrip their male counterparts in telephone access. This reversal is possibly related to the dramatic rise in the popularity of cell phones during the first decade of the twentieth century, which may have decreased demand for household (read: landline) telephones.

Screen Shot 2016-11-25 at 11.06.26 PM

I chose to take a different approach to my analysis of income and household telephone access (fig. 6). Because income is a continuous measure, I chose to create a scatterplot instead of a faceted bar graph to fully capture the relationship. Instead of arbitrarily designating income categories, I chose to analyze the relationship between income and telephone access through the lens of race, a demographic characteristic intimately tied to income (fig. 2). This race-income plot clearly seems to illustrate a positive relationship between a race-year cohort’s average income and the percent of its members with access to household telephones, suggesting income might at least partially explain racial inequality in telephone access. This analysis, however is far from complete. My secondary analysis seeks to untangle this relationship between income and race by making use of a modified ipums dataset and mathematical regression analysis. The two regression outputs below (tables 1 and 2) summarize the two linear models developed to describe the relationship between household phone access and race.

Screen Shot 2016-11-25 at 11.07.40 PM

Screen Shot 2016-11-25 at 11.07.54 PM

These models suggest significant relationships between race and an individual’s likelihood of owning a phone (v24) that remain approximately in line with the disparities observed in fig. 1 even when income is controlled for (table 2), suggesting other factors may be at play. The extremely weak fit of these models, however, suggests linear regression fails to fully capture the relationship.

Conclusion

            This analysis found inequalities in household telephone access and distribution over several demographic, racial, and linguistic groups. It builds on previous literature by analyzing these trends as they change over time, generally finding that these disparities decreased dramatically as telephones became completely ubiquitous in the twenty-first century. It also illustrated reversals in some of these trends in 2010, potentially associated with demand-side effects related to the rise in the popularity of cellular phones. This analysis is valuable because it may help shed light on how the adoption of technologies relates to inequality, something that is important to understand as our rate of technological progress increases due to the enormous impact technology has on health, education, economic outcomes, and communication. Although inequality in telephone access may not be directly analogous to inequality in modern technology like access to high speed internet, the examples of the past may still be valuable to the policymakers of today when it comes to ensuring all their citizens benefit equally from investments in technology and infrastructure. To build on this study, I would seek to further understand how the various possible causes of inequality analyzed here interact, possibly through more advanced statistical analysis. I would also seek to better understand these trends as they change over time in an effort to understand what happens to inequality in access to specific technologies over time as those technologies are replaced. Finally, I would also seek to analyze similar trends as they relate to other types of technology, especially access to the internet.

 

Sources Cited:

Boustan, Leah, and Allison Shertzer. “Population Trends as a Counterweight to Central City Decline, 1950-2000.” Demography 50.1 (2013): 125-47. ProQuest.Web. 16 Nov. 2016.

 

Brandon, Belinda B. The Effect of the Demographics of Individual Households on Their Telephone Usage. Cambridge, Mass: Ballinger Pub. Co, 1981. Print.

 

Fischer, Claude S., and Carroll Glenn R. “Telephone and Automobile Diffusion in the United States, 1902-1937.” American Journal of Sociology 93.5 (1988): 1153-178. Web.

 

Ruggles, Steven, Katie Genadek, and Ronald Goeken, Josiah Grover, and Matthew Sobek. Integrated   Public Use Microdata Series: Version 6.0 [Machine-readable database]. Minneapolis: University of Minnesota, 2015.

 

Shebl, Fatma et al. “Measuring Health Behaviors and Landline Telephones: Potential Coverage Bias in a Low-Income, Rural Population.” Public Health Reports (1974-), vol. 124, no. 4, 2009, pp. 495–502. www.jstor.org/stable/25682264.

 

Von Hippel, Paul. “Linear vs. Logistic Probability Models: Which Is Better, and When? | Statistical Horizons.” Statistical Horizons. Statistical Horizons, 5 July 2015. Web. 13 Nov. 2016.

 

Wolfle, Lee M. “Characteristics of Persons with and without Home Telephones.” Journal of Marketing Research 16.3 (1979): 421-25. Web.

Race and Income in Higher Education

In order to see the post with charts, please click here.

Abstract: Within the American education system, a radical reshuffling of student demographics is underway. While broad demographic shifts in higher education are covered fairly often in academia and media, the literature seems to overlook important distinctions between different categorizations of institutions, such as online/traditional, private/public, and two-year/four-year universities.  It is the purpose of this study to address this hole in the literature by showing how these shifting demographics—specifically race and socioeconomic status—manifest across Ivy League institutions, with a special focus on Dartmouth College, as compared to non-Ivy League American Universities. Beyond identifying the demographic disparities across institutions, I also analyze some of literature which explores some of the underlying causes of these trends.

Introduction: Within the American education system, a radical reshuffling of student demographics is underway. As of 2014, the National Center for Education Statistics (NCES) reported that, for the first time, the total percentage of minority students – Latinos, African-Americans, Asian-Americans, Pacific Islanders and Native Americans combined – was larger than the percentage of whites in public grade-school classrooms (NCES 2014).  While lagging behind grade-school classrooms, these demographic shifts are reflected in higher education as well with the percentage of minority students growing from 29% in 2000 to 42% in 2014 (Hussar 2016, 4). Not only limited to race, the demographic landscape in higher education has made notable shifts in many categories including age, gender, and socioeconomic status.

Though these broad demographic shifts in higher education are covered fairly often in academia and media (Keller 2001) (Williams 2014), the literature seems to overlook important distinctions between different categorizations of institutions, such as online/traditional, private/public, and two-year/four-year universities.  It is the purpose of this study to fill this hole in the literature by showing how these shifting demographics—specifically race, age, socioeconomic status and geography—manifest across these different institutional categories with a special focus on Ivy League institutions. Beyond identifying the demographic disparities across institutions, I will also analyze some of the underlying causes of these trends.

Income: Before exploring the ways in which the demographics within higher education have changed over the past half-century, namely in terms of race, it is important to look into one defining demographic that has remained constant over this period–the socioeconomic backgrounds of students. As the chart below demonstrates, over the past 50 years, the income stratification among recipients of bachelor degree has remained largely the same since the 1970s.

[Figure 1]

This chart is based on the data collected from the Pell Grant Institute’s 2015 report “Indicators of Higher Education Equity in the United States” (Cahalan, 2015). The data source used in the report is from the October Education Supplement of the Current Population survey conducted by the U.S. Census Bureau. Data from 1970 to 1986 consider unmarried 18- to 24-year-olds, and data from 1987 to 2013 are based on dependent 18- to 24-year-old. The income quartiles have been converted to be read in terms of constant 2014 real dollars in order to adjust for inflation. In 2014, the family income quartiles identified by the Census Bureau were as follows: $0-$34,933, $34,933-$65,496, $65,496-$116,466, and $116,466+. The major limitation of this dataset is that it omits independent students, who constitute roughly 35% of undergraduates (Wei and Neville 2005). As a result, we can expect that the data for the top quartile for dependent students are especially likely to overestimate degree attainment relative to entire population of individuals from the top family quartile.

What is most notable about this chart is that in 2014, 54% of degree recipients were in the top quartile of family income and 10% were in the bottom quartile, marking a 2% decrease in students from the bottom quartile and a 2% increase in students from the top since 1970. Additionally, over the 44-year period, the percentage of bachelor degree recipients coming from the bottom two quartiles  fluctuated between a high of 28.5% in 1977 and a low of 19% in 1985.

Beyond graduation rates, the low-income student experience is very different from that of high-income students on various levels, including major concentration, the control of the institution, and the selectivity of the institution. For example, one of the most salient divisions between selective and nonselective schools can be found in the income backgrounds of the enrolled students. This divide is so striking that the New York Times reported that for every student from the entire bottom half of the nation’s income distribution at Penn, Princeton, Yale, Brown and more than a few other colleges, there appear to be roughly two students from just the top 5 percent (Leonhardt 2016). However, at Dartmouth College, the ratio of low-income to high-income students is even lower than that.

[Figure 2A,  Figure 2B]

These charts compare the family incomes of the US population and of Dartmouth students in 2013. The data is drawn from three sources: (1) Current Population Survey (CPS) Annual Social and Economic (ASEC) Supplement for the 2013 census year, (2) the Office of Financial Aid Data for 2013 (Asch 2014), and (3) data collected through Dartmouth Pulse.  This chart reveals that for every one student in the bottom 45% of the nation’s income distribution, there are nearly ten students in the top 6.5% percent. In other words, 59% of Dartmouth students come from families in the top 6.5% of households by income in America. This lack of economic diversity is by no means unique to Dartmouth or even the lowest compared to peer institutions. Among 179 universities with a five-year graduation rate of 75 percent or higher, Dartmouth ranks only 55th in terms of economic diversity, according to the Upshot’s College Access Index.  However, it is important to note that this index doesn’t use family income, but instead relies on the share of students who receive Pell Grants, the graduation rates of those students, and the net tuition price for low- and middle-income students to determine the accessibility of the institution.

Similar to the Upshot, we will not be looking at the family incomes of all Ivy League institutions and national Institutions as a measure of economic diversity because this information is largely unavailable, incomplete, and unreliable. Instead we will be looking at the proportion of students receiving federal Pell Grants, which are awarded to undergrads from low-income families or low-income independent undergrads to help pay for college. In 2013-2014, 2.3 million or 61% of the 3.8 million Grants were awarded to students with family incomes below 30,000$, and 87% were awarded to students. In addition to family income, Pell Grants are awarded based on family size and number of family members attending college. While this is by no means a perfect measure, several news organizations and scholars alike understand Pell figures as the best available proxy of socioeconomic diversity on campus (Kahlenberg et al 2014) (Muraskin 2004). This next chart gives us insight into how Pell Grants are distributed among the Ivy League as compared to all other universities in the US.

[Figure 3]

The data to make this chart (code) is drawn from the Department of Education’s College Score Card dataset which includes data from 1996 through 2016 for all undergraduate degree-granting institutions of higher education. The percent of undergraduates awarded Pell Grants is only available over the 2008-2014 period. This chart shows that students receiving Pell Grants are far less likely to attend an Ivy League institution than another University. In fact, in 2014, the average percentage of students who received Pell Grants among all national universities was 53% compared to just 14% at Dartmouth and other Ivy League institutions. It is also important to note that while the share of non Ivy League students who received Pell Grants increased from 43% in 2008 to 56% in 2011 in the wake of the Great Recession, we only see an increase of 2% among Ivy Leagues during the same period.

Confirming this trend, Bastedo, M. N., & Jaquette, O. (2011) were able to study the family income of students directly and found that as institutional selectivity increases, the percentage of lower income students decreases. Through an analysis of the US Department of Education statistics, Bastedo shows that in 2004, only 4% of students attending schools he classifies as “Most Competitive” were from the bottom socioeconomic quartile, while 69% of these students were from the top socioeconomic quartile. While the gap between the top and bottom percentiles has declined from 73% to 65% since 1972, this was largely due to the increase of students from the third quartile, who see a 10% increase. Meanwhile, over this 32-year span, the percent of students attending these schools from the bottom quartile only increased by a mere 1.2%.

According to the 2016 Pell Grant Institute report, we do not only see a disparity in Pell Grant recipient enrollment between the level of selectivity of an institution, but we can also see a disparity in the type of level of the institution. Specifically, in 2013, 56% of full-time undergraduates who received Pell or other Federal Grants attended 4-year private institutions rather than 2-year institutions, compared with 75% of students who did not receive any Federal Grants. (Callahan and Perna 2015). We can also see an enrollment difference within the type of control of institution attended by Federal Grant Recipients. According to the National Center for Education Statistics, Federal Grant recipients were more than 3 times as likely as non-Federal Grant recipients to attend private for-profit institutions in 2013 (U.S. Department of Education 2015).

Scholars, policymakers and higher ed institutions alike have all developed many theories trying to explain these various divides between low- and high- income students. Most scholars broadly agree that it is a confluence of factors inclusive different level of schooling, parental education and financial resources. When it comes to specifics, some scholars turn to the strong positive correlation between high SAT scores and high family income to explain why low income students flock to certain types of institutions. Scholars disagree over the causal factor of this relationship, with some arguing that it the SAT score disparity arises because wealthy students can afford prep courses (Buchmann et al 2010). However, other research suggests that test prep has fairly limited effect on scores (Zwick and Green 2007).

Moving beyond SAT scores, other scholars maintain that there is a significant pool of low-socioeconomic-status (SES) students who are attending colleges that are less selective than the ones they could have attended based on their academic preparation or SAT scores. This hypothesis is often referred to as the “under-matching” hypothesis. (Bowen, Chingos, & McPherson, 2009)  Supporting this hypothesis, one 2013 study found that many low income students do not apply to selective college even when their SAT scores are sufficiently high. Instead, these students are more likely to attend non-selective two or four year universities close to their  hometown. The dominant explanation for this trend offered by the authors is that low income students tend to be geographically isolated from other high achieving students and therefore lack the information or encouragement support that high-income students receive (Hoxby and Avery 2013).

In addition to enrollment disparity, the different levels of graduation completion further widens the gap between low- and high income students. According to a series of longitudinal study conducted by the National Center for Education Statistics (BPS: 1996, 2001, 2009), only 26% of first-time students enrolled in a postsecondary education institution obtained a bachelor’s degree within 6 years, compared to 59% of students from the top income quartile. This 33% gap between the two income quartiles have remained largely stable since the first longitudinal study began in 1990.

Race:

Unlike income which has been mostly static over the past few decades, the racial demographics of college students have undergone radical reshuffling paralleling that of the general US public.

[Figure 4]

The data used to make this chart is drawn from several sources including the U.S. Department of Education, National Center for Education Statistics, Higher Education General Information Survey (HEGIS), “Fall Enrollment in Colleges and Universities” surveys, 1976 and 1980; Integrated Postsecondary Education Data System (IPEDS), “Fall Enrollment Survey” (IPEDS-EF:90); and IPEDS Spring 2001 through Spring 2015, Fall Enrollment component.  As we can see, the share of White students enrolled in American Universities has declined from 82.6% in 1976 to 55.6% to 2014–a reflection of the decline of the White population in the US from 84.3% to 58.3% during the same time.

[Figure 5]

While in terms of college enrollment, students are more or less racially representative of the rest of the country, in terms of completion rates and receiving a Bachelor degree, the student population is not representative.

[Figure 6]

The data used to make this chart is also drawn from several sources including the U.S. Department of Education, National Center for Education Statistics, Higher Education General Information Survey (HEGIS), “Degrees and Other Formal Awards Conferred” surveys, 1976-77 and 1980-81; Integrated Postsecondary Education Data System (IPEDS), “Completions Survey” (IPEDS-C:90-99); and IPEDS Fall 2000 through Fall 2014, Completions component.

While White, non-Hispanics as a share of degree recipients continue to be overrepresented as compared to the US population, their degree share has decreased from 89% to 69% over the course of 34 years—reflecting the national decline from 80% to 62% in the same period. Meanwhile the percentage of Black degree and Hispanic recipients grew from 7% and 3% to 11% and 11% respectively. Comparing this to the overall shifts in the national population, this means that African Americans and Hispanics are now over 1.5 and 1.8 times as likely to be represented among bachelor’s degree recipients as in the population, respectively. Degree recipients of the remaining races including American Indian and Asian have grown roughly proportionately to the population of the US. We can see this disparity clearly in the following chart.

[Figure 7]

Not only can race be used as a predictor of the likelihood of university completion, but like income, it also can be used to predict the type of university that a student attends. Among Ivy League institutions and most selective colleges, Black and Hispanic students are dramatically underrepresented as compared to the undergraduate average, while Asian students tend to be overrepresented.

[Figure 8]

The data to make this chart (code) is also drawn from the Department of Education’s College Score Card dataset. As we can see the average share of Asians across all Universities has hovered consistently around 3.5% from 1996 to 2014. Meanwhile, at Dartmouth this number has grown from 8.68% to 14.6% over the same period. Inversely, the average share of Blacks and Hispanics across all Universities has increased from 14.9% and 9.34% to 18.9% and 16.2% respectively over this period, while the on Ivy League campuses these numbers have hovered around 7-9%.

While we might be tempted to attribute this racial disparity in the Ivy League to the aforementioned socioeconomic disparity, scholar have shown that the probability of enrolling in a highly selective college is five times greater for white students than black students, even after controlling for income (Hussar 2016). Building off of this finding, other studies have concluded that race and ethnicity have become one of the strongest predictors of SAT scores and in turn college admissions, as compared to family income and parental education levels (Geiser 2015).

Conclusion: While the causes underlying these trends in Higher Education are not yet fully understood, this analysis has shown that in order to begin to explain these trends, it is important to look beyond the aggregate levels of fall enrollment across a given demographic. First, it is important to also look at graduation rates as well, as in the case of race, some races are far more likely to graduate than others. Secondly, it is critical to look at the enrollment and graduation rates by selectivity of institution. By using the Ivy League as a proxy for selectivity, I have demonstrated that there are stark differences in both race and socioeconomic status between Ivy League students and students from other universities. For future research, I would suggest looking at the breakdown in enrollment across other level of institutions including private vs public, two-year vs four-year, and online versus traditional.

Works Cited:

Bastedo, Michael N., and Ozan Jaquette. “Running in place: Low-income students and the dynamics of higher education stratification.” Educational Evaluation and Policy Analysis 33.3 (2011): 318-339.

 

Buchmann, Claudia, Dennis J. Condron, and Vincent J. Roscigno. “Shadow education, American style: Test preparation, the SAT and college enrollment.”Social Forces 89.2 (2010): 435-461.

 

Cahalan, Margaret, and Laura Perna. “Indicators of Higher Education Equity in the United States: 45 Year Trend Report.” Pell Institute for the Study of Opportunity in Higher Education (2015).

 

Dreier, Peter, and Richard D. KAHLENBERG. “Making Top Colleges Less Aristocratic and More Meritocratic.” Web log post. The Upshot. New York times, 14 Sept. 2014. Web.

 

Geiser, Saul. “THE GROWING CORRELATION BETWEEN RACE AND SAT SCORES: NEW FINDINGS FROM CALIFORNIA.” Center for Studies in Higher Education 10.15 (2015): n. pag. Web.

 

Hoxby, Caroline, and Christopher Avery. “The missing” one-offs”: The hidden supply of high-achieving, low-income students.” Brookings Papers on Economic Activity 2013.1 (2013): 1-65.

 

Hussar, William J. “Projections of Education Statistics to 2023.” Projections of Education Statistics to 2023. National Center for Education Statistics, 4 Apr. 2016. Web. 11 Oct. 2016.

 

Keller, George. “The new demographics of higher education.” The Review of Higher Education 24.3 (2001): 219-235.

 

Leonhardt, David. “California’s Upward-Mobility Machine.” New York Times. N.p., Sept.-Oct. 2016.

 

Muraskin, Lana, and John Lee. “Raising the Graduation Rates of Low-Income College Students.” Pell Institute for the Study of Opportunity in Higher Education (2004).

 

Williams, Joseph P. “College of Tomorrow: The Changing Demographics of the Student Body.” US News. N.p., Sept.-Oct. 2014.

 

US Department of Education, ed. “The Condition of Education.” National Center for Education Statistics (2014).

 

U.S. Department of Education, National Center for Education Statistics, Integrated Postsecondary Education Data System (IPEDS), Digest of Education Statistics, 2015. Table 331.20

 

Wei, Christina Chang Chang, and Stephanie Nevill. “Independent Undergraduates: 1999-2000.” National Center for Education Statistics. N.p., Oct.-Nov. 2005. Web.

 

Zwick, Rebecca, and Jennifer Greif Green. “New perspectives on the correlation of SAT scores, high school grades, and socioeconomic factors.”Journal of Educational Measurement 44.1 (2007): 23-45.

 

 

Writing The United States Census

By Colby Gardner

Introduction

The United States Census is written by men and women who’s job it is to accurately and effectively represent the people of the country. This article delves into the details behind the writing of the Census. I seek to explore the actions taken by the Bureau and the actual changes made to the Census form in an attempt to analyze the statistical integrity as well as usefulness of the Census from 1900 to 1980 in reference to the variables for women’s occupation, race, and ancestry.  In existing literature, authors have made reference to important adaptations in the census from year to year such as categorization of race or occupation. However, these articles seldom go beyond a short blurb and fail to take a harder look to explore the reasons behind such significant and statistically relevant change. This project investigates the thoughts and ideas of the people who were responsible for writing the census as well as the general surrounding political and pragmatic pressures that impacted their decisions.

The writing of the United States Census takes place behind closed doors in hearings and conversations about what to put into an upcoming census. These hearings provide a explanation of specific processes that went on during the deliberations before the census form for the upcoming decade comes out. While writing the questions that would appear on the next decade’s form, the Census Bureau and governing bodies had hearings describing and debating precisely how to carry out the survey. The parts of these hearings that are of special interest to this project are the conversations about which questions or responses to include in the next form. In The Hearings before the subcommittee on Census and Government Statistics of the Committee on Post Office and Civil Service House of Representatives in 1959, there is a transcript of a conversation in the hearing with Dr. Burgess and Mr. Cunningham about the elimination of income as a question on the 1960 Census and its replacement, mother tongue. In a similar setting, The Hearings Before the Subcommittee on Census and Statistics of the Committee on Post Office and Civil Service House of Representatives in 1967, the speaker talks about new categories for the 1970 Census including mental and physical disability. More specifically to this project,The Hearings Before the Subcommittee on Census and Populations of the Committee on Post Office and Civil Service House of Representatives in 1978 contains deliberations over how to ask a new question of ancestry without overlapping with other questions on the census. In addition to hearings, there have also been sources about the selection of enumerators and surveyors for the Census between the years 1880 and 1940, such as Diana Magnuson’s “The Making of the Modern Census”(1995). It references key steps one had to take on the way to enumerator status, including tests and interviews to insure statistical accuracy (Magnuson, 1995). These enumerators were responsible for the gathering of the data that stood to represent the people, and the directions taught to them effected data all over the country, as seen in women’s occupation in the early 20th century. These individual glimpses into the inner workings of the Census Bureau yield a full image of how the Bureau operated, yet do not supply a statistical source to show the fallout of each of the Bureau’s decisions.

On the flip side of the coin, Census officials are also conscious of who they are polling and the difficulties associated with them. In order to truly understand and critique Census officials, a general picture of the pressures facing the Bureau in this time period is important to understand, as well as the statistical and political change stemming from their decisions. As an essay in Tourangeau’s Hard-to-survey Populations, the writer explains that immigrants are one of the most difficult groups of people to survey as a result of inconsistent housing, uncommon family structures, below average education and income, and a general distrust of society (Massey, 2014). A separate article in the same collection goes on to add that an immigration population is often a linguistic and cultural minority at the bottom of the racial hierarchy of the time (Harkness, 2014).  These hard to survey populations match with many of the races the Census Bureau was trying to gather more information about in their racial reorganization from 1900 to 1940 (Hochschild and Powell,2008). Also these variables that contribute to hard to survey populations also share characteristics of the population of Mexican Americans that check “Non Hispanic” in response to the hispanic origin question yet fill in “Mexican” under ancestry (Duncan and Trejo, 2008). ThMassey also shines a light on the repercussions of eliminating parental birthplace from the census in favor of the ancestry variable, describing the time as one which immigration and therefore second generation immigrant populations grew feverishly yet went undocumented (Massey, 2014).

Often times, especially with more recent censuses, there are understandable and logical reasons for dropping important questions. From budgeting the length and expense of a census to just replacing questions with more valuable ones, there is not always a biased or racially fueled reason to change a census that will affect the statistical integrity of the survey (Congress, 1959). Although there is a plethora of material on the creation of the United States Census and how the questions and the process were conducted, there are seldom raw data and analysis to convey some sort of statistical impact that each decision had on the census. In this article I further explore the changes made to each year’s census form as well as the reactions in the data that followed.

Method

To create each of the graphs used in this article, there were a number of methods that were used for all of the visualizations. First, 1% samples from the IPUMS website were taken from every decade of interest in order to form an accurate portrayal of the American population. Second, in order to gather a more interesting data set, Alaska and Hawaii were included in the data sets despite the unavailability of this data in 1940 for the diversity of their populations. For this reason, only the year 1940 in data representations will exclude these states. Lastly, because no sample line variables were used in this data section, the data was weighted with the IPUMS variable PERWT.

In the first figure exhibits a bar graph of women’s occupation by percentage and race. Using IPUMS variable OCC1950, which gives a person’s occupation under the definitions supplied prior to 1950. Lastly, the range of this graph is from census years 1900 to 1940 and are separated into four races; White, Black, Native American, and Asian. No other populations were considered. This visualization is used in this article to show the vast variation in results between the years of 1900 and 1940 when the instructions for enumerators were changing. In this graph, women’s occupation shifts drastically, creating a cause for concern in census taking.

The second figure is another bar graph, this time showing populations of each race recorded by the census at any given year between 1900 and 1940. If a race did not appear on the census in any decade’s form, that race was given a population value of zero for that year. This graph shows the sporatic changes that went on in the categorization of race between 1900 and 1940, bringing attention to these revisions to the census form.

The final figure shows a choropleth map representing the first response for ancestry given in the 1980 and 1990 censuses. Using the IPUMS variable ANCESTR1, the map contains each state’s the most common place of ancestry amongst the state’s population. The Ancestry variable was a new installment in 1980 and data collected from these two years fluctuated greatly, raising concerns about why something that should remain relatively constant should waver so significantly.

Results

In the early twentieth century, the United States was a primarily agriculturally based economy. This meant that many American citizens owned their own farms and worked crops to survive. In this way, it came as no surprise in 1900 when the census category “Farmers and Farm Laborers” made up a significant portion of the summation of all occupations in America. However, more specifically, the share of women farm laborers compared to men was dramatically lower.

Finalwomensocc

Figure 1 shows a percentage breakdown of women’s occupation

As seen in figure 1, the percentage of women with any occupation at all is quite low throughout the time period in question, but especially prevalent in 1900. This is in part a result of the way enumerators were asking questions about employment. In 1910, a dramatic shift in how the US Census Bureau classified occupation occurred.  Prior to this year, questions dubbed as “sorter” or “filter”  questions, such as asking whether a person was economically active, caused many people, especially women, to be disqualified from answering any follow up questions about occupation (IPUMS,1999). From 1910 on to 1940, all sorter questions were eliminated from the census and women who were not economically active could answer questions about their occupations (IPUMS,1999). Seen most notably in the graphs for Back and Asian women, a significant spike in percentage of women working as farmers and farm laborers was recorded in the years after 1900. The reason behind this spike specifically in farmers and farm laborers is that, despite not having an employer nor earning a salary, women working on family farms were now considered as having an occupation under this category.

Screen Shot 2016-11-16 at 1.07.32 AM

Figure 2 is a bar graph of the number of female farm laborers from 1900 to 1940

In the tabulated form of the bar graph in figure 2 above it is clear that in White, Black, and Asian women, total farm laborers increased at a clip of almost 90% from 1900 to 1910 most commonly resulting from how the US census recorded and understood data entries.

Growing at an alarming rate, the Census Office in Washington D.C. tripled in size every ten years from 1860 to 1880 to reach almost 1,500 employees. 62 marshals and 6,530 assistant marshals took the census in 1870, whereas in 1900 the process of recording each American’s census data had grown so large that 300 supervisors and 53,000 enumerators were required to keep up with the growing population (Anderson, 1988). With the number of employees growing at such an enormous pace, it became increasingly important to train the new enumerators, giving specific instructions as to how to represent the population. In 1910, enumerators around the country were taught specifically in regards to the question of occupation, “An entry should be made in this column for every person enumerated,” with the entry falling in one of three categories, (1) the occupation pursued, (2) an entry of ‘own income’ for persons following no specific occupation but having an independent income upon which they live, or (3) an entry of “none” for persons not falling in either of the two previous categories. Columns 19 and 20 were to be left blank when the entry in 18 was either “own income” or “none” (Enumerator Instructions, 1910). This rhetoric created a monumental shift in women’s occupation, seen in the column graph and table under year 1910, as the Census Bureau replaced the questions from previous years based on economic participation.  This shift allowed family farm workers, especially women who worked at home in the fields, to be considered as having occupations in the census form.

Occupation was not the only category in the early twentieth century that was going through an experimental phase. In this time, the political and social landscape of the United States was built upon the hierarchy of race. In this way, the categorization and divvying up of racial and ethnic categories was a source of great interest and effort for Census officials of the time.  In the course of 5 census forms and 41 years, the race question saw ten different race categorizations. The category of Black became more complicated, sometimes containing a “Mullato” option in the check boxes. Nationality became increasingly relevant in specifying Asian race, as these categories became more and more detailed. This period stood as a time for experimentation in this section of the census and left no race or ethnicity unaffected. As a result of the constant fluctuation in options, the definitions and demographics of every racial category changed from year to year.

FinalRaceGraph

Figure 3 shows every race that appeared on the census between 1900 and 1940

This constant shifting of categories caused individuals answering the census each decade to swing from racial group to group, often never filling out the same box two census forms in a row. This is because for example a person of Mexican heritage would have checked white in census of 1920, but as a result of the addition of Mexican to the census of 1930, would have checked that box in the coming decennial form. The category “Mulatto” was added to the census form in 1850 as a way to separate and count mixed race African Americans, created by men like Josiah Nott, who used the Census to try and prove people of mixed races had lower fertility and shorter lives in order to create racist insurance tables set to overcharge Mulattos (Hochschild and Powell, 2008).  In 1900, when Mulatto was omitted from the form, not only did the census lose a category, but the definition of Black now changed from one that used to exclude Mulatto to one that now included the classification.  As seen in figure 2, the population of people who checked Black on the census actually went down considerably in 1910 and 1920, as Mulatto reappeared as an option. In 1930, once Mulatto was removed for the last time from the census, the black population jumped to over ten million people. A very similar situation is evident in the category Hawaiian.  Between Hawaiian, Hawaiian and Asian, Other Asian or Pacific Islander, and Hawaiian Mixed, which all sporadically appear and reappear between 1900 and 1940, an accurate population trend over the course of more than thirty or forty years in any of these five categories is statistically impossible. With similar experimentation happening throughout the early twentieth century, Census results became inconsistent and increasingly useless in comparing statistics from year to year. As seen most clearly in figure 3 above, most categories that appeared in the census between the years of 1900 and 1940 did not appear on every census for in this time frame, causing several gaps in the bar graphs above.  As race categories change, it becomes increasingly difficult to track and compare specific races over time as data is missing whenever the census decided to recategorize. The census set out the experiment with the way they collected data on race in the United States to create a more accurate representation of the population. However, by doing this the Census Bureau simultaneously created a larger problem in the difficulty that came with understanding such widely varying data.

The introduction of Mexican as a category on the 1930 census was met with great opposition from a Mexican population wary of racial discrimination based on the new census separation between whites and Hispanics. The deputy director of the census in 1930 once wrote that “if the Mexicans in this country could be convinced of the value of the census work and of the impossibility of the information they give being used against them, I believe we could secure their hearty cooperation” (Hochschild and Powell, 2008).  Unfortunately, many people did not recognize the change for its scientific accuracy and instead saw it as a way to discriminate.  Mauro Machado, an associate of the Mexican civil rights activist Alonso S. Perales, described the attempt to document the Mexican American population as “the cowardly way in which [Box Committee witnesses] try to make us possessors of negro blood” (Graton and Merchant, 2016). As a result of these political and social pressures, the Census Bureau never again listed Mexican as a race after 1930, despite its statistical relevance.

map_1980

Figure 4 shows the most common ancestry response in each of the 50 states in 1980

In addition to research on race and ethnicity, the Census Bureau added an open ended entry in the 1980 census form to collect information on ancestry. This addition came as an effort to grasp a better understanding of the nationality and background of people answering the census.In previous years, asking for the country of birth for an individual’s mother and father in order  to find a person’s lineage only worked if the individual was a second generation immigrant. The question of ancestry replaced these questions and gathered data no matter how many generations removed from their nation of origin a person was (Rosenwaike, 2002). The Census Bureau asked the question of ancestry in the hopes that it would replace the parents birthplace question, identify people of Hispanic origin by gathering information about their ethnic origin, and gain information on the ethnicity of the other part of the population (Farley, 1990).

map_1980

Figure 4 shows the most common ancestry response in each of the 50 states in 1980

map_1990

Figure 5 shows the most common ancestry response in each of the 50 states in 1990

 

Figure 4 and 5 above are choropleth maps of the United States showing the most common ancestry for each of the 50 states. In 1980, the most common ancestry throughout the United states was English, spreading throughout every corner of the country from Oregon and the entire west coast to Florida up to Maine. The next most common ancestry spanning the northern Midwest was German, ranging from Idaho all the way East to Ohio and Pennsylvania. The other most common ancestries were Italian, Afro-American, Irish, and Japanese. In 1990 however, English as an ancestry shrank to the most common entry in just 4 states as German became far and above the most overwhelming ancestry in America. Afro-American also became much more common as many southern states top entry. Other entries in this census included Mexican, which showed up in most states bordering with Mexico including California, New Mexico, and Texas; French; American; Italian; Irish; and Japanese.

imgres

Figure 6 shows the ancestry question in the 1980 census

imgres

figure 7 shows the ancestry question in 1990

A reason for such a dramatic shift in Ancestry in just 10 years could have been the way the census asked the questions from year to year. Above in Figures 6 and 7 are the census questions about ancestry in each form from 1980 to 1990.  In 1980, 16 ancestries appeared on the census as suggestions. In 1990, 21 ethnic groups for ancestry were specified as options for the open ended question. Out of the 16 suggestions in 1980, three groups, English, French, and Hungarian, did not appear on the list in 1990. All three of these groups experienced significant declines in the number of persons reported in each group. This decline in English ancestry is the primary reason behind the stark contrast between maps from 1980 and 1990 above. From 1980 to 1990, English ancestry dropped 32%, French ancestry dropped 28%, and Hungarian ancestry dropped 20%, all because of their omission from the suggested categories under the open ended question (Rosenwaike, 2002). The two groups that moved up the farthest on the list between the two census forms were German and Italian. Between the two censuses, the number of people who answered having German ancestry increased 6% while Italian ancestry increased 8.5% (Rosenwaike, 2002). When writing the census in the 1970s, the Census Bureau actually had trouble deciding how to include the new question on ancestry. In a hearing about the upcoming census in 1978 Mr. Lehman said, “I am still bothered by the double question on the 100 percent form that deals with when you want to talk about ethnic origin, heritage, ancestry, and race. I think if you are not awfully careful with that question, it would be so counterproductive” (Hearing before the Subcommittee on Census and Population, 1978). Although they were able to create a productive question for the census of 1980, it was the slight change in wording in the next decennial census that created a counterproductive swing in the data.

Conclusion

This Article explores the process of creating the United States Census. We looked at women’s occupation in the early 20th century and how enumeration techniques varying year to year could influence the results of each year’s occupation count. In 1910, when enumerators were directed not to ask filter questions ensuring that all workers were economically active, women with occupations skyrocketed, especially those working on family farms as farm laborers. Race from 1900 to 1940 was also studied, as race categories like Asian or Black where split into finer and finer subcategories, each time making the statistics less and less consistent and therefore harder to analyze. Finally, the new ancestry category of 1980 and 1990 was put under the spotlight. Although English ancestry was the most common ancestry in America in 1980, its removal from the suggested entries in 1990 form caused a massive amount of people to write something that was suggested, and English ancestry was replaced my German ancestry as the most repeated answer. In each of these cases, whether it be the way a question on the census form was asked or the way enumerators recorded their data, they all prove that the Census Bureau wields tremendous power over how the United States population is represented. All it takes is changing the order of a list or changing the name of one category and the entire country’s makeup looks different. The job of the United States Census is to create data that represents the people. In this way, the men and women responsible for the administration and interpretation of the census serve as a bridge between people and statistics and even the smallest adjustment of words could lead to the misrepresentation of the American people.


Works Cited

Hearings before the subcommittee on Census and Government Statistics of the Committee on Post Office and Civil Service House of Representatives. April 8, 1959 http://congressional.proquest.com/congressional/result/pqpresultpage.gispdfhitspanel.pdflink/$2fapp-bin$2fgis-hearing$2f4$2f3$2fb$2f4$2fhrg-1959-poh-0007_from_1_to_49.pdf/entitlementkeys=1234%7Capp-gis%7Chearing%7Chrg-1959-poh-0007

Hearings Before the Subcommittee on Census and Statistics of the Committee on Post Office and Civil Service House of Representatives. May 23, 1967 http://congressional.proquest.com/congressional/result/pqpresultpage.gispdfhitspanel.pdflink/$2fapp-bin$2fgis-hearing$2f3$2ff$2f7$2fa$2fhrg-1967-poh-0015_from_1_to_179.pdf/entitlementkeys=1234%7Capp-gis%7Chearing%7Chrg-1967-poh-0015

Hearings before the subcommittee on Census and Population of the Committee on Post Office and Civil Service House of Representatives. March 21, 1978  http://congressional.proquest.com/congressional/result/pqpresultpage.gispdfhitspanel.pdflink/$2fapp-bin$2fgis-hearing$2ff$2fb$2fd$2fc$2fhrg-1978-poh-0020_from_1_to_68.pdf/entitlementkeys=1234%7Capp-gis%7Chearing%7Chrg-1978-poh-0020

“1910 Census: Instructions to Enumerators”: Extract from Enumeration Forms on IPUMS https://usa.ipums.org/usa/voliii/inst1910.shtml

Duncan and Trejo. “Ancestry versus Ethnicity: The Complexity and Selectivity of Mexican Identification in the United States”, 2008.

Farley, Reynolds. “Race and ethnicity in the US census: An evaluation of the 1980 ancestry question.” Report prepared from research conducted as American Statistical Association National Science Foundation/Bureau of the Census Fellow. University of Michigan (1990).

Hochschild and Powell. “Racial Reorganization and the United States Census 1850–1930: Mulattoes, Half-Breeds, Mixed Parentage, Hindoos, and the Mexican Race”, 2008.

Magnuson, Diana L. https://usa.ipums.org/usa/voliii/enumproc1.shtml, “The Making of a Modern Census: the United States Census of Population, 1790-1940,” Ph.D. dissertation, University of Minnesota, 1995.

Merchant and Gratton. “La Raza: Mexicans in the United States Census”, 2016

Rosenwaike, Ira. https://gateway.dartmouth.edu/science/article/pii/,DanaInfo “Ancestry in the United States Census, 1980-1990″, 2002.

Tourangeau, Roger. 2014. Hard-to-survey populations. Harkness. 2014. Surveying Cultural Linguistic Minorities; Hard-to-survey populations. Massey. . 2014. Challenges to Surveying Immigrants; Hard-to-survey populations.

 

Code for my graphs can be found at

https://github.com/wcolbygardner/QSS-30.50/blob/master/Final%20Project/finaldsf.R

 

Mixed Race in America: How the Social and Political Landscape of the U.S Shaped Race in the 20th Century

Introduction

Mixed heritage people have often gone unrepresented in history. Though the reality is that most, if not all, people come from a background of mixed heritage, interracial marriages were not legalized across the whole United States until 1967 (Aldridge 1978, 356). I want to use U.S. Census Data to help me formulate a better understanding of what social and political changes in regard to race meant for the mixed race community. This is a hard task to accomplish because the census did not allow for people to mark themselves with more than one race until the 2000 census. Since the change, the census has grown to accommodate 63 race categories, and 57 of those are composed of mixed-race combinations (Allen and Turner  2001, 513). This change shows how difficult it is to categorize humans through race as there are inevitably infinite possibilities in the racial composition of individuals.

However, to understand the history of mixed race people and the relationships between different races, through the analyzation of mixed peoples, it is necessary to use the census data from before the 2000 census, which for many reasons proves to be a difficult task. Before the 2000 census allowed for the selection of multiple races, people were guided to choose one of several discrete races with which they identified to the most. For this reason I want to use data on the children of interracial marriages to approximate the presence of mixed race persons in the US. The research previously conducted on the matter is incomplete, and furthermore, it focuses primarily on specific race mixtures – most often that of white and black couples.

Despite my ability to draw on IPUMS data to create data visualizations for the mixed race population in America, the data is probably lacking in accuracy. There are a lot of constraints on my research that make it hard to gather accurate data. One flaw in my method is that it is likely that a large population of interracial couples did not legally get married, so they may have been less likely to have lived together, or to respond that they did (Monahan 1977,66). Additionally, attempts have been made in the past to destroy racial categorization – leaving the data available to me possibly incomplete. Thomas Monahan claims that activists in the Civil Rights era aimed to erase racial identification from public records in attempts to reduce the societal discrimination based on color or race. This is particularly damaging as it led to the erasure of racial data in certain large states. Despite the efforts of some individuals, the Census Bureau itself is also responsible for both reflecting and participating in the social construction of race, which influences what we  consider mixed-race to mean. Their role in creating racial categories can be seen as beneficial in many ways, but it is also important to understand that by creating such categories they have influences the perception of certain individuals in this country. So, the changing attitudes and opinions regarding race have had direct impacts on the data collected on the topic.

Additionally, not all those who were enumerated by the census were enumerated correctly. 20th Century America had ever-changing definitions and standards regarding the categorization of race. Factors such as “skin color, the primary objectivation of racial group membership, play[ed] an important role in determining the degree of assimilation” of colored people in America (Lewis and Ford-Robertson 2010, 409). To this extent, individuals with lighter skin tones, who may not have been white, could have been categorized, by their own volition or against it, as white. This is especially true for Hispanic Americans, who were not identified as their own ethnic/racial category until 1980 (Lewis and Ford-Robertson 2010, 411). When the US Census Bureau began sending enumerators door-to-door to enumerate those who did not mail-in the census, more accurate accounts of the population were likely taken by the Bureau. Mail-in censuses disproportionately undercounted the more socio-economically troubled segment of the population. Nonetheless, the many problems with categorizing race, a social construct, are apparent in any data analysis conducted using census data. For this reason the 2000 Census, the only set of data that allows for a somewhat accurate count of the mixed race population in America, has to be included in studies of mixed race people prior to the turn of the 20th century.

Methods

Figures 1 and 2 show the percentage of children (people under the age of 18) with parents of different races in America. Figure 3 shows the population of people who responded to the 2000 census as appertaining to two or more races.  Raw data is gathered from IPUMS 1% samples, asides from 1970 where 1% state form is used and 1% metro in 1980. Data is not available for Hawaii and Alaska in 1940 and1950. Data was weighted by PERWT in all years. A method to define parents of different races needed to be created to generate the code that counts for children with different race parents. IPUMS classifies census race categories into values of RACE, which is what I used to define parents of different races. Using the values of RACE is instrumental because it easily allows for one to pull race data from the census, but, unfortunately, the values of RACE are not consistent.

The type/identity of people that were categorized into different RACE values changed with the changing social constructions of race. Certain races are only available in the RACE variable for certain years. Additionally, asian nationalities, such as Chinese, Japanese, and ‘Other Asian’, are marked as different races on the census. This necessitates a methodology for understanding race within this research. I have chosen to leave the asian nationalities as different races to reflect the viewpoints of the census, and in this vain, the viewpoints of society of the time.

Children are defined as people under the age of 18, and since their race is not being analyzed directly, the results rely on the race(s) of their parents. This is problematic because the analyzation will not account for children with different race parents unless the parents both live with the child. Furthermore, mixed parents are not accounted for because they were forced to choose one race to represent them in the census up until 2000, so some mixed race children will not be accounted for this reason.

Figure 1 is a map; it shows the percent of children in each American state who have parents of different races from 1900 to 1990. Figure 2 is a line graph that shows the percent of children with parents of different races in the US as a whole in those same years. Figure 3 is a population pyramid which shows the distribution of sex and age in people who marked themselves as appertaining to two or more races in the year 2000.

Below I have linked my code for each figure which I have uploaded to GitHub:

Figure 1 (animated map) code can be found here

Figure 2 (line graph) code can be found here

Figure 3 (population pyramid) code can be found here

Results and Interpretation

 

anmap1

Figure 1 – Animated map of children with parents of different races by state, 1900-1990.

Figure 1 demonstrates percentages of children (those under 18) with parents of two different races in each state of the U.S between 1900 and 1990. This map shows the changes in the population of mixed race children, and where they occurred, over the majority of the 20th century. The changes are small, but they reflect the increasing acceptance of racial mixing throughout the nation. In 1940 and 1950 it will appear as if Hawaii, where the greatest percentage of people of mixed race reside, lost all their mixed race children. However, this is only a reflection of the fact that U.S. census data is not available for Hawaii and Alaska in the 1940 and 1950 censuses.

The maps show a small, but steady, general increase in the presence of children with mixed race parents. This increase goes hand in hand with what one would expect studying American history. As the US became more populated, and more immigrants entered (Hochschild and Powell 2008, 63), new racial categorizations had to be created. The increased immigration subsequently resulted in the higher presence of mixed race children over the years. However, the low figures for children of interracial relationships goes to show how it was nonetheless difficult for racial integration to occur in America. Because the results are very low, they may be more useful in showing the lack of representation/presence of mixed race children as a result of race tensions in America.

In fact, the data categorizes Japanese, Chinese, and Koreans as different races, which may promote a slightly larger percentage of mixed race children than if those nationalities were considered as one Asian race. This reflects the fact that race is a racial construct that is highly dependent on historical context. Using a map gives the reader an easily interpretable geographic understanding of the presence of mixed race children in America. Seeing the changes in population of each state over 90 years is enough of a time frame to give one a sense of the how the changes happening were affecting the nation as a whole. The maps do not suggest a very specific trend in the location of mixed race people, though certain states on the West Coast, as well as the non-continental state Hawaii, can be seen as places where mixed people, and therefore interracial relationships, are more frequently found. Looking at the map of 1990 one can see that the whole West Coast is composed of 3% or more mixed race children, which is more than the East Coast. This is perhaps a result of the large Asian immigration to the West as the West is geographically closer to Asia. Nonetheless, it can be inferred that inter-racial relationships were more frequent in areas in with greater immigration.    

One area in which my research led me to a different conclusion from my research was in the percentage of mixed race people in the South compared to the North (Lewis and Ford-Robertson 2010). My data analysis, which is visually represented by the maps, did not firmly conclude that the South had significantly lower rates of mixed children than the rest of the nation, especially compared to the North.  Perhaps, the inference was made that states and regions frequently considered more racist would be less conducive to interracial relationships. Because I did not find this to be the case, the reality might be that, despite lower levels of overt racism in the North, race was still a difficult social boundary to break.

 

lineproj

Figure 2 – Line graph depicting children with parents of different races 1900-1990.

     Figure 2 utilizes a line graph to demonstrate the overall trend of the population of children with parents of different races in the U.S over most of the 20th century (1900-1990). The line graph promotes the notion that children with mixed race parents were not a common occurrence, but that their population saw a statistically significant increase in the years following the Civil Rights movement and Court rulings such as Loving vs. Virginia which allowed for interracial marriages across the 50 states (Aldridge 1978). Aldridge, in his article “Interracial Marriages: Empirical and Theoretical Considerations”, also highlights the reality that interracial relationships remained a tiny fraction of relationships in America, but that they did see an increase over the years as race relations changed. In 1970, a study by David Heer using Census Bureau data showed a 26% increase in Black/White marriages between 1960 and 1970 compared to same-race marriages for the two races (Lewis and Ford-Robertson 2010, 410). However, since race in the 20th century was often dichotomized between white and non-white, results may be unreflective of the true presence of mixed race relations.

Seeing how the percentage of children with parents of different races reacted to the social movements of the 60s promotes the notion that shifting political and social views in regards to race have had a measurable effect on the population of mixed race children. The line of the graph is useful in visually interpreting the surge that occurred after the Civil Rights era. With this surge in mind, the inclusion of visible percentages to the data points on the graph keeps the visualization in relation to the overall population. The data point percentages show that despite what could be interpreted as a significant change in the population after the civil rights movement, the de facto percentage of mixed race children in relation to the nationwide population remained very low, peaking at 2.65 percent in 1990 – helping prove Aldridge’s findings. This leads to my next point, which aims to understand the difference between perception and reality.

One major theme that I came across in my research was that there are racial boundaries that have made mixed relationships less prevalent than the media would suggest. As racial boundaries become less and less rigid, popular culture reflects this reality through increased inclusion of minorities in media. Today, this can be seen in our television programs, tabloid magazines, and a variety of other media outlets. However, despite the progressive movement of the Civil Rights era, which from the line graph can be seen as having had a measurable affect on the mixed race population, and the increased inclusion of minority races in popular culture, research suggests that the union between differing cultures is still rather rare. Lewis and Ford-Robertson’s research promotes the idea that interracial relationships are prominently between people of similar backgrounds. So, perhaps, as racial boundaries differentiate people less and less it is likely that the mixed race population will continue to grow. Ultimately, the line graph helps us understand that social movements do in fact have resonance in population demographics, but also, that these reactions have to be interpreted on a larger scale to reveal their true impact, or lack thereof, on our population as a whole.

 

Population Pyramid

Figure 3 – Population pyramid depicting distribution by age and sex of U.S. population that identified as having two or more races in the 2000 census.

Figure 3, the population pyramid, demonstrates the age and sex demographics of people who marked themselves as appertaining to two or more races on the 2000 census data. Allen and Turner acknowledge that the 2000 census is the first census which can provide somewhat accurate counts of the mixed race population, and for this reason I have decided to include it in my project. This graph can help provide more accurate numbers in regard to the mixed race population, and it can also be useful by helping visualize the demographic patterns of the preceding decades.

This population pyramid works to show that the youth, people of ages 0 to 9, are by far the largest group of people with mixed race parents. My research supported this finding, and the data helps demonstrate how the improved race relations of the late 20th century led to increased interracial relationships and mixed race children. This was discussed to some depth by Lewis and Ford-Robertson in their article “Understanding the Occurrence of Interracial Marriage in the United States Through Differential Assimilation”. In the article they take interest in how the Civil Rights movement of the 60s led to a cultural movement intent on giving minorities equal rights. It can be deducted that the social and political action of 60s, and beyond, led to some degree of destigmatization towards interracial mingling. Naturally, with better race relations, the taboo of interracial relationships was pushed back upon, slightly, but still significantly enough to be traceable in the population pyramid. This push towards the acceptance of interracial relationships is made evident through the proportionally large number of mixed race youth in the year 2000.

The use of a visualization for the year 2000’s census is important in determining whether the research done on the trends and patterns of the preceding decades was somewhat correct and relevant. James Allen and Eugene Tuner’s 2010 article “Bridging 1990 and 2000 Census Race Data: Fractional Assignment of Multiracial Populations” was in part responsible for giving me an idea of why it might be important to include data from when the census accommodated for mixed race people. The article highlights how between 1990 and 1998 there was a 41% increase in the birth of mixed children. Since my other data visualizations stopped in 1990 I thought it would be important to include at least one visualization that included 2000. Furthermore, by using a population pyramid, I can look at more than just the year 2000 since the pyramid shows people across all ages.

Further Analysis

Government laws, just like the Census Bureau’s racial categorizations, had an instrumental role in promoting racial boundaries within society. It is more likely than not that laws before the 1954 Supreme Court decision Brown v. Board of Ed, that made segregating school unconstitutional, and the 1967 ruling Loving v. Virginia created impediments to racial mingling. This had the effect of making inter-racial relationships rare (Aldridge 1978). The Supreme Court decision to erase segregation in schools most likely lead to greater mingling between students of different races in high school and in college. In the 1970s, people were entering relationships at earlier ages, and colleges provided a safe space away from parents for young adults to interact. This, in addition to young peoples’ revolts on “traditional institutions and values” that led them to “reject the taboos on dating across racial lines” created an atmosphere that was more accepting of mixed race partnerships(Aldridge 1978, 357). So, in the post Civil Rights era years, there was a surge in interracial relationships that is shown in my data visualizations. Of course, this was only possible in areas in which racial segregation was low. Regardless, inter-racial relationships remained very low, and became even less frequent in the South after Brown v. Board of Ed according to Aldridge. Possibly as a result of white backlash to the government’s measures to reduce discrimination.

Conclusion

       Mixed race people have been a small demographic of the United States of America, however, this is a rapidly growing demographic. This growth is indicative of a changing nation and of a changing world. Globalization, immigration, and technology have all led to greater interactions between people of different cultures. Because of this change, I believe it is important to add the experience of mixed race peoples to the general conversation about racial and cultural identity. Unfortunately, an overarching theme presented in the sources I collected is that more research on the topic of mixed heritage persons needs to be conducted, but that the limitations on collecting this data have presented a very difficult challenge for past researchers. This lack of relevant data in past years shows that it is important that the census keeps adapting and evolving over the years. As America shifts towards being a more racially heterogeneous country, going forwards it will be important that mixed race people are included in the decisions and conversations that have traditionally excluded and/or ignored mixed race peoples. Even though the data I have compiled suggests that mixed race people have been much less prevalent in our society than one might perceive, this is a reality that is rapidly changing. Perhaps, this change may even lead to a renewed perception of race and racial boundaries. This research hopes to communicate that measuring something as fluid, and socially constructed, as race is important, but that it is important to look at the bigger picture as well. The very definition of mixed race is complicated for this reason. Regardless, if this information can be used to better understand the state of our nation, and of the world, and if it can be used for good, then it is important we continue to conduct this sort of research.


 

Works Cited

1. Lewis, Richard, and Ford-Robertson Joanne. “Understanding the Occurrence of Interracial Marriage in the   United States Through Differential Assimilation.” Journal of Black Studies 41.2 (2010): 405-20. Web.

2. Aldridge, Delores P. “Interracial Marriages: Empirical and Theoretical Considerations.” Journal of   Black Studies 8.3 (1978): 355-68. Web.

3. Monahan, Thomas P. “Interracial Parentage as Revealed by Birth Records in the United States, 1970.” Journal of Comparative Family Studies 8, no. 1 (1977): 65-77. http://www.jstor.org/stable/41600991.

4. Jennifer L. Hochschild and Brenna Marea Powell, “Racial Reorganization and the United States Census 1850-1930: Mulattoes, Half-Breeds, Mixed Parentage, Hindoos, and the Mexican Race,” Studies in American Political Development 22(2008): 59-96.

5. Allen, James P., and Turner Eugene. “Bridging 1990 and 2000 Census Race Data: Fractional Assignment of Multiracial Populations.” Population Research and Policy Review 20.6 (2001): 513-33. Web.

6. Monahan, Thomas P. “An Overview of Statistics on Interracial Marriage in the United States, with Data on Its Extent from 1963-1970.” Journal of Marriage and Family 38.2 (1976): 223-31. Web.

Working Mothers and Fathers: 20th Century in the United States

I) Introduction

Scholars have differing opinions on the causes of and outcomes from the rising labor force participation of wives in the United States during the 20th century. Some attribute the rise to pressure on married women to increase their families standard of living. Others attribute the rise to shifting gender norms encouraging women to work. Using IPUMS data, I will test scholars arguments regarding the causes of, and the outcomes that came from the rising labor force participation of married women in the United States.

Continue reading

The Intersectional Landscape of Sex, Race and Immigration in the United States (1930 – 2000)

Introduction

Recent fanfare on what has been deemed “the feminization of migration” by the United Nations starkly contrasts the abysmal invisibility of female immigrants in statistical and scholarly representation.  Where publicized data–even official datasets such as public US Census data from the Integrated Public Use Microdata Series (IPUMS) project–have erased and discounted their existence, academic narrative is limited to one of two reductive narratives—the default of the male immigrant, or the stereotypical immigrant female who has meekly followed the lead of her immigrating husband (Pearce, Clifford and Tandon 2011). For the immigrant female, the effect of being bastardized by census classifications is as multivariate as it is decimating of a total identity. The recent rise in discussion about immigrant females by mass media and academics alike has found itself resting on feeble attempts to interpolate the narrative for the male immigrant to the female counterpart by considering isolated identities of the female immigrant within the masculinized landscape. As Hondagneu-Sotelo (2003) notes, such reductive analyses bear severe consequences for the identity of the immigrant female as well as the humanized understanding, policy formation and sociological adaptations surrounding such a highly heterogeneous demographic.

Motivated by the emerging interdisciplinary literature that attempts to provide a more honest and humanized picture of the immigrant female and based on evidential analysis drawn from IPUMS data for 1930 through 2000, I aim to augment the existing portrayal of the immigration landscape with snapshots of the realities of this particular demographic through my research. I argue that the cumulative effect on the identity and lived experiences of immigrant females in the United States of sex, race and immigrant status is profoundly intersectional, rather than additive. To that end, an empirical approach and mixed-methods analysis will be employed to intercalate statistical data with the corpus of existing literature and arrive to a quantification, demarcation and visualization of the locus of intersection of these three variables that is both cogent and cohesive.

Methods

For all graphs, raw data were obtained from IPUMS 1% samples (5% scaled down to 1% for 1980“; 1960 is excluded for all the visualizations due to unavailability of several relevant variables for that year. All measures are given as a proportion of total subpopulation identity. In the following section, I refer to each of these identity subgroups as “subpopulation”. For the purposes of this post, I define as immigrants only those who have not naturalized (though they may have received first papers). Conversely, non-immigrants were identified as those born abroad of American parents and naturalized citizens; those with unavailable citizenship data were excluded. The intent behind including naturalized citizens was to reflect the hierarchy of privilege existing within the American immigration infrastructure regarding ease of naturalization, with the concurrent acknowledgement that naturalized citizens often still face unique and identical issues of immigrant identity as non-naturalized immigrants.

Scholars have argued that the intersection of these identities not only has social implications for the immigrant female, but also impacts policy framework, access to education and employment, and legislative rights uniquely relevant to them. Labour force participation as a data point fails to reflect the reality of the downward impact of intersecting identities on educational and occupational achievement. Through my first visualization, I will examine the complex influences of sex, race and immigrant status as they interact with each other on access to education and occupational status for the immigrant female. I use the IPUMS-constructed Nam-Powers-Boyd score as a cohesive indicator to account for the combined effect on both labour participation and the social capital granted by certain educational standards and occupations, instead of taking a one-dimensional approach and considering either education or labour force participation as isolated measures of success.

Static Screenshot of Figure 1

Static Screenshot of Figure 1

Figure 1: Click here for interactive plot

Figure 1 shows a sequence of graphs for weighted density distribution of total non-white male population, i.e. both immigrants and non-immigrants combined, as well as immigrant and non-immigrant non-white females separately in the United States per Nam-Powers-Boyd score for the census years 1930 through 2000. According to IPUMS, the Nam-Powers-Boyd score is a measure of occupational status based on both medians of earnings and educational attainment associated with each occupational category[1]. The scores were assigned based on different occupational categories on a 1950 and 1990 basis; thus, for consistent analysis, the dataset was assigned with the 1950-basis score for years 1950 and before (NPBOSS50), and with the 1990-basis score for years 1990 and later (NPBOSS90).  Analysis is limited to those identified as employed and not classified as white in the race variable. Density is represented in proportions of total subgroup population rather than absolute numbers, scaled up to a maximum standard value of 1 per curve. Data are weighted by PERWT. The x-axis shows discrete Nam-Powers-Boyd score, and the y-axis shows weighted density per score; panes are separated by year. The curves are stacked such that the zero-line for the top two starts at the upper line of the curve underneath it; blue represents data for non-white males, green for non-immigrant non-white females, and pink/peach for immigrant non-white females. Corresponding y-coordinate value along the curve represents proportion of subset population at the respective score.

The concept of “segmented assimilation” applies uniquely to female immigrants to dissimulate their existence within the fabric of American society.  Female immigrants not only negotiate identity within the overarching American society and the hierarchical structures that are nested in it (Gold 2003), but also within the tacit sub-societies that resemble remnants of the land they left behind and the unique structures of oppressive power embedded within that particular distinctly non-American society. These sub-societal structures are un-Americanized gendered institutions whose conflicts remain “uncontested when survey methods are used” (Hondagneu-Sotelo 2003). My second visualization examines how nested sub-societies (such as enclaves, immigrant networks and relatives) influence autonomy and decision-making for female immigrants in terms of employment, having to work without pay for related family, and ability to marry more than once against stigmatization; and whether or not they are obliged to conform to the role of the follower in the household if they are associated with a spouse in census data. To analyze the impact of these hierarchical structures, I use related subfamily and rule for linking spouse data for the female population in the United States. For the first three subgraphs, I use the subfamily type variable from IPUMS to identify only immigrant females who cohabit with subfamilies that are related to the household head by birth, marriage, or adoption. Propensity to work without pay for the family is measured using the ’class of worker’ variable from IPUMS. Since data on household head is not available for most years in my analysis, I use another variable—rule for linking spouse[2]–to examine hierarchy of power within a married couple where one half is an immigrant female, and compare data with that for their non-immigrant counterparts. Based on who was recorded as household head, this variable designates follower and precedent within a married couple.

(Code to generate Data Visualization 1 can be found here.)

Static Screenshot of Figure 2

Static Screenshot of Figure 2

Figure 2: Click here for interactive plots

Figure 2 shows a set of four column graphs for immigrant and non-immigrant females in the United States depicting the correlation of spousal linkage and subfamily influences on the variables mentioned above. Males were pre-excluded from the dataset, but white females are included. The x-axis of the first three subgraphs represents census year, while year is represented by the y-axis for the fourth graph (bottom right). For the first three subgraphs, the total subpopulation consists of non-immigrant or immigrant females with related subfamily; those with unavailable subfamily data, no subfamily and unrelated subfamily were excluded. For the fourth, the total subpopulation constitutes non-immigrant or immigrant females with spouse link data available (either adjacent or non-adjacent); those without a spouse linked to them or with previously allocated marital status (but not current) have been excluded. Weighting was performed on a PERWT basis as proportions of the respective subpopulation just described for each subgraph. Subgraph 1 (top left) shows separate columns for three different employment status classifications as per census data for the years 1930 through 2000 employed (in blue), unemployed (in orange), and not in labour force (in green). Bars on the top half of the graph (above the midline) represent data for immigrant females, while bars on the bottom half represent data for non-immigrant females. The y-axis represents percentage of total subgroup population, either immigrant or non-immigrant females with related subfamily. Subgraph 2 (top right) shows stacked columns depicting proportions of females with related subfamily who are also unpaid family workers; immigrants in red, non-immigrants in purple. Value along the y-axis for each stack represents the percentage proportion value with respect to the total number of people in the subpopulation for either immigrants or non-immigrants. Subgraph 3 (bottom left) shows column graphs with line plot overlay depicting number of spouses accounted for in the census data for females with related subfamily for the years 1940, 1950, 1970 and 1980. The y-axis represents percentage of total subpopulation, as defined for the first two subgraphs. The column graphs represent data for immigrant females–pink for those who have only been married once; brown for those who have been married multiple times. The line plot (in red) represents data for non-immigrant females who have been married multiple times only, for comparison. Those who have never been married, or with unavailable data for marital status or number of marriages in the census, have been excluded. This specific information was not collected after 1980; thus, years later than 1980 could not be visualized on the graph. Subgraph 4 (bottom right) shows a set of horizontal bar plots depicting percentage of total subpopulation who were either not linked to a spouse, or linked to a spouse as a follower or precedent of the spouse in census data (both via adjacent and non-adjacent linkage combined), for the years 1940 through 2000. Left half of the subgraph (i.e. left to the vertical axis) represents data for non-immigrant females, whereas bars on the right half represent data for data for immigrant females. The y-axis represents census year and the x-axis represents percentage of total subpopulation for either immigrant or non-immigrant females; length of bars along the horizontal axis is thus proportional to the subpopulation percentage value. Grey bars represent those with no spouse linked; green bars represent those who have been linked as following their spouse; and blue bars represent those who have been linked as preceding their spouse in the census form.

Racial and ethnic categorizations formed through the interplay of the census and American society reduces and otherizes the immigrant female by imposing confines of American womanhood and racial identities that are not endogenous to their individual narratives (Pearce, Clifford and Tandon 2011). Based on this hypothesis, my third visualization aims to examine how racial categories specific to American society and history act to further alienate immigrant females; specifically, I will study the correlation of race and birthplace for both immigrant and non-immigrant females of colour and visualize whether immigrant females find themselves identifying as and conforming to the single race and ethnic category boxes constructed by the American census as frequently and as comfortably as their non-immigrant counterparts.

(Code to generate Data Visualization 2 can be found here.)

Figure 3 consists of a pair of chord diagrams, produced using the R package circlize. The chord diagrams depict the relationships between racial identification in the US census for the years 1930 through 2000 and place of birth as indicated in census data for immigrant non-white females and non-immigrant non-white females, respectively. The diagram on top corresponds to data for immigrant non-white females, whereas the diagram on the bottom corresponds to their non-immigrant counterparts. The circumference of the circular plots represent both race and place of birth variables from IPUMS, with each variable denoted by a distinguishable coloured segment on the graph and marked with its corresponding clockwise label of identification. The length of each segment is corresponding to its occurrence in the dataset; thus, for example, a racial category has a longer segment associated with it than a birthplace since it is more likely to be a common variable among more individuals in the data than a particular birthplace. Race and birthplace variables were recoded using crosswalk between numeric values and corresponding IPUMS codebook identifications. Raw data was obtained from IPUMS 1% samples by pre-excluding males as well as anyone identified as white from the preliminary data extract. Data for each racial category and the corresponding birthplace of the individual was first aggregated on a PERWT basis, and then processed to produce an adjacency matrix–a data format commonly used to represent and quantify relations. In such a matrix, value in i-th row and j-th column represents the relation from element in the i-th row and the element in the j-th column. The relationship between the two data variables is bidirectional; as such, the chord diagrams constitute a form of directed graphing. The absolute value corresponding to the relationship between each i and j value–in this case, the number of incidences in which a race category and birthplace were incidental–measures the strength of the relation, which in turn is represented as area of coverage by the linking region between a particular race segment and a birth country segment. The width of the bezier curve for each connection, or more precisely, the area of each curve, represents the number of times an individual of the particular racial categorization has also been identified to have the place of birth linked to from the race category segment by the curve. Thus, the width or the area of the connecting linker between each segment of race and birthplace is proportional to the overall correlation or linkage of those two data points in the same individual’s data for the total population of either immigrant or non-immigrant non-white females in the United States–the wider the curve (the greater the area of coloured coverage between a race and a birthplace variable), the higher the correlation between the two connected variables.

(Code to generate Data Visualization 3 can be found here.)

Results

From the density distribution graphs in Figure 1, it is apparent that immigrant females have the lowest population density values at the tail end of the distribution curve, i.e. out of the three groups, they are the least likely to have been assigned a Nam-Powers-Boyd score on the highest end of the scale (i.e. far-right end along the y-axis). For example, for Nam-Powers-Boyd scores 90 and above, immigrant females of color have an average score range of 0.3  – 0.03 across the years, whereas their non-immigrant counterparts have a range of 0.4 – 0.15, and non-white males have a score range of 0.35 – 0.1. Fascinatingly enough, this figure provides a straightforward means of visualizing the relative magnitudes of impact by the isolated variables of race and sex on educational and occupational status achievement. Until the 1950s, non-white males and non-white immigrant females were represented in identical frequencies for Nam-Powers-Boyd scores 90 and above, i.e. an average of 0.01% for both. However, a gap starts emerging in the 1970s on the basis of immigrant status that drastically widens over the next three decades–the average gap in density for scores 90 and above consistently widens from approximately 0.05% at each score 90 and above to 0.07% at each score. The numbers appear miniscule in isolation; however, its cumulative effect when we consider integrating it over the entire control area while the difference accumulates is far from marginal even in the quantitative sense. Thus, there is strong evidence that while educational and occupational achievement are no doubt still dictated by sexism, the negative impact in terms of barriers to achievement due to immigrant vs non-immigrant status may consistently be several folds higher than due to sex differences. These results largely correspond with what scholarship has hypothesized in theory. Immigrant non-white females are at a particular disadvantage since they are entering an intersectionally segmented labour market foreign to them in more ways than one. Educational and occupational sectors in the United States are “structured largely along gendered lines”, where flux processes of globalization and immigration “muddy these binary distinctions in interesting ways”. Holding true to the theme of patriarchy, the higher the status associated with an occupation, the more the occupation is associated with masculinity, and the more a female who is also an immigrant “may be resented for infringing on male territory”. It is important to note that while traditional imagination of the docile immigrant female has shifted the burden of limited achievement on internalized submissive tendencies only; however, these scholars also conclude from analysis that since females “immigrating from societies with more rigid gender roles now move outside those boundaries [..it..] may make them more open to challenging the norms”, citing so as a driving reason for the significantly higher number of immigrant females than non-immigrant ones in “gender-atypical occupations”. Unfortunately, this puts them at a further disadvantage in a market that is subtly yet rigidly segmented along scripted gender lines. In a humanized analysis, Pearce et. al. conclude that being “one of many” as an immigrant and “one of few” as a woman in a gender-atypical occupation, an immigrant woman may be resented doubly—both for being a woman and for being an immigrant.

Non-immigrant females, surprisingly enough, consistently hold higher presence for scores 90 and above; however, it should be noted that since the score assigns equal weight to education level and occupational status, it does not account for sex differences in occupational status achieved between males and females for equivalent educational degrees. Hence, it is highly plausible that distribution on the higher end consists of both people who have achieved high occupational status with moderate levels of education, as well as those who have achieved extremely high levels of education and yet barred from the highest echelons of the workforce due to discrimination.

My visualization and analysis in the graphs for spousal and sub-society effects in Figure 2 sought to study the assertion by scholarship regarding the enmeshment of female immigrants in systemic influences that take shape within the smaller communities that they take part in. It has been argued that female immigrants not only negotiate identity within the overarching American society and the hierarchical structures that are nested in it (Gold 2003), but also within the tacit sub-societies that may resemble remnants of their birthplace. In the first subgraph, percentages of immigrant and non-immigrant females who are either employed or not in the labour force at all remain consistently near-identical throughout the timeline. However, observing the relative lengths of orange bars reveals a persistently higher percentage of immigrant females who are unemployed compared to their non-immigrant counterparts. This aligns and expands on analytical conclusion already confirmed by economists–labour force participation rate for females has a U-shaped curve across economic development–the faster the economic and technological growth in a nation, the lesser the participation rate for females (Golding 1995). While this would impact both categories of being unemployed and not in the labour force, it is highly probable that the factor has significant effect on unemployment rates for immigrant females. Particularly given that Figure 1 has already indicated exacerbation of employment barriers due to sex by immigrant status as well as the fact that the gap in unemployment increases visibly over the years with economic growth (and that the gap is the narrowest from 1930 through 1950, i.e. throughout the economic downturn due to the Great Depression and World War II), being an immigrant would likely deepen the trough of the curve further. The pressures of “segmented assimilation”, where faithful performance aligning with the national and domestic expectation of all three socially constructed images of sex, race and immigrant status work in favour of a combination of higher acceptance and integration into workplaces and society at large with domestic harmony, function as additional barriers of complexity to higher employment rates (Le Espiritu 2003; Menjívar 2003). Contrary to what would be expected, a much higher percentage of non-immigrant females living with related subfamily were identified as unpaid family workers than immigrant females for all years except 1980, where an anomalously larger proportion of immigrant females constituted this worker-class demographic than non-immigrants. The caveat here is that this graph is incapable of reflecting realities of worker class identification and how they may be influenced by cultural stigma and pressure enforced by the resident subfamily. For example, in many households with a subfamily–particularly in-laws–it may be taboo for the daughter-in-law to verbally identify as  “working WITHOUT PAY in family business or farm”, which is the exact wording of the census questionnaire for this category; moreover, the immigrant female may even be barred from holding the authority of a worker in family business at all. Since the 1950s, non-immigrant females have unambiguously married more than once in larger proportions than immigrant females. Marital liberty is often associated with female autonomy; thus, higher barriers to being able to marry once aligns with the analysis that female immigrants have to juggle negotiating identity within the hierarchical structures of American society as well as limitations set in place by the sub-societies and cultural communities (such as enclaves) which they are obliged to navigate (Le Espiritu, 2003). Interestingly, the fourth subgraph reflects a trend being recently observed by intersectional scholars–that the stereotypical image of the docile and domesticated immigrant female has rapidly shifted over the past few decades (Arnold 2011). While longer bars for having no spouse linked could either mean increased liberty or decreased opportunities to form long-term relationships in the United States, if we observe bars for the years after interracial marriage was deemed legal, the rate of decrease of immigrant females who were linked to a spouse in census data as following the spouse has been visibly higher than non-immigrants. Conversely, although females are drastically less likely to be linked on a census form with a spouse as their follower across the board, blue bars in the past three decades show an emerging trend of increasing in length for immigrants than for non-immigrants.

Perhaps the starkest contrast appears in the chord diagram visualizations in Figure 3. The purpose of this comparison was to test the hypothesis whether racial categories specific to American society act to further alienate immigrant females. It may be rather fitting that as I am about to segue into a figurative discussion about American racialization and its otherizing effect on immigrant females, visualization of the data confirms the hypothesis in a blatantly literal way; bezier curves connecting immigrant females with birthplaces all across the globe find their tail end at the racial segment for the census category “Other”. Indeed, the “other” category has held relatively volatile definition in census terms; however, noticeable itself is the comparison with data for non-white females who are not immigrants. What appeared to be wide expanses of area coverage between the “Other” category and immigrant females reduces to a faint, almost one-dimensional line for non-immigrant females of color. The second largest area for immigrants is covered by the Asian or Pacific Islander category; this may have several root causes ranging from the model minority dilemma to the nuanced ways in which immigrant networks from the region form and shape rigid community-based identities in the United States. To confirm the perplexity American racialization may pose for the immigrant female, I note here a general conclusive trend–the non-immigrant female of color, whether born here or abroad, is almost certainly likely to identify with a single pre-defined racial category in the census (thus, the predominance of bezier curve region connecting single race variables with immigrants from various birthplaces) while the reverse is true for the immigrant non-white female. The other plural-race category, i.e. multiple races, also sees a much larger total area of linkage between the variable and immigrants from different birthplaces than it does for non-immigrants. Sampled narratives in mixed-methods analysis is rather telling of how racial and ethnic categorizations of the American census act to alienate immigrants by imposing an expectation to participate in an idealized notion of community belonging–which, ironically, can not only invoke images of the trauma inflicted by community infrastructure (Villalon 2010), but also impose a tacit obligation of assimilating to the American imagination of the racial community and perform the racialized category ascribed to them collectively by the census and American society (Pearce, Clifford and Tandon 2011). An entirely different layer of complexity is added by the fact that census data has historically shaped the evolution of race and ethnicity by laying its roots in the establishment of colonialism (Kertzer and Arel 2001). Undoubtedly, racialization would serve as a significant source of trauma and additional imperialistic oppression to negotiate with for the immigrant female who may have once lived in a formerly colonized nation before transitioning to a country that has historically been the colonizer.

Conclusion

Mainstream narrative of female immigration has taken place around each of her multiple and heterogeneous identities as an isolated, homogenized subject of study under the conceptual assumption of ceteris paribus. In reality, as implicated by the work reviewed here, such an assumption is rendered null where variables and institutions such as sex, race and immigration not only coalesce but also function to symbiotically construct and mold each other within a globalized yet locally-situated social ecosystem in a multivariate manner. The results of my research, in conclusion, confirm a number of characteristics to be true for immigrant females–that their various sociological identities are deeply intersectional with nuanced impacts on other variables; that the identity of the female immigrant is both publicly created and privately enforced; and that their identities are far too individual–far too human–to be homogeneous or influenced solely by a single economic or sociopolitical institution.

[1] “Integrated Occupation and Industry Codes and Occupational Standing Variables in the IPUMS.” IPUMS-USA. User Guide, University of Minnesota, n.d. Web. https://usa.ipums.org/usa/chapter4/chapter4.shtml

[2]  “Family Interrelationships.” IPUMS-USA. User Guide, University of Minnesota, n.d. Web. https://usa.ipums.org/usa/chapter5/chapter5.shtml

Works Cited

Pearce, S., Clifford, E. & Tandon, R.. Immigration and Women: Understanding the American Experience. New York: NYU Press, 2011.

Villalon, R..Violence Against Latina Immigrants: Citizenship, Inequality, and Community. New York: NYU Press, 2010.

Hondagneu-Sotelo, Pierrette, and Pierrette Hondagneu-Sotelo, editors. “Gender and Immigration: A Retrospective and Introduction.” Gender and U.S. Immigration: Contemporary Trends, 1st ed., University of California Press, 2003, pp. 3–19, www.jstor.org/stable/10.1525/j.ctt1pnt0g.4

Gold, Steven J. “Israeli and Russian Jews: Gendered Perspectives on Settlement and Return Migration.” In Gender and U.S. Immigration: Contemporary Trends. Edited by Hondagneu-Sotelo, P. 127-148. University of California Press, 2003.

Kertzer, David I., and Dominique Arel. (2001), The politics of race, ethnicity, and language in national census. Cambridge: Cambridge University Press.

Goldin C. The U-Shaped Female Labor Force Function in Economic Development and Economic History. In: Schultz TP Investment in Women’s Human Capital and Economic Development. University of Chicago Press ; 1995. pp. 61-90.

Le Espiritu, Yen. “Gender and Labor in Asian Immigrant Families.” Gender and U.S. Immigration: Contemporary Trends, Edited by Pierrette Hondagneu-Sotelo, 1st ed., University of California Press, 2003, pp. 81–100, www.jstor.org/stable/10.1525/j.ctt1pnt0g.8.

Menjívar, Cecilia. “The Intersection of Work and Gender: Central American Immigrant Women and Employment in California.” Gender and U.S. Immigration: Contemporary Trends, Edited by Pierrette Hondagneu-Sotelo, 1st ed., University of California Press, 2003, pp. 101–126, www.jstor.org/stable/10.1525/j.ctt1pnt0g.9.

Le Espiritu, Yen. “‘We Don’t Sleep Around Like White Girls Do’: Family, Culture, and Gender in Filipina American Lives.” Gender and U.S. Immigration: Contemporary Trends, Edited by Pierrette Hondagneu-Sotelo, 1st ed., University of California Press, 2003, pp. 263–284, www.jstor.org/stable/10.1525/j.ctt1pnt0g.16.

Arnold, Kathleen R. American Immigration After 1996: The Shifting Ground of Political Inclusion. Pennsylvania State University Press, 2011, www.jstor.org/stable/10.5325/j.ctt7v1p6.

Women’s Employment and Earnings 1940-2000

 

 

Introduction

Historically, there has always been wage gap between men and women and it is still a controversial topic in contemporary era. Many scholars have argued that there are several critical factors attributed to the differences in wage between men and women. And the study shows how several factors, the marital status, the number of children in family, and highest level of education influence on the wage gap between men and women. Several scholars in women’s studies find that the trend of women’s labor force participation shows noticeable differences not only by sex but by race. White and Black women who have similar education levels and number of children sometimes had quite different qualities of lives with varied socioeconomic status. There are significantly many factors to determine an individual’s life. The trends in lives of different races written in history might be colored or slightly distorted depending on the authors of history references. So, through these activities, I would like to quantitatively prove or disprove that races, genders, education levels or other demographic factors affected an individual’s quality of life in terms of the amount of wages or groups of available occupations from 1940 to 2000. I also expect that although my analysis focuses on data collected between 1940 and 2000, I would be able to find the patterns defining the relations of various demographic factors recorded from the far past, and possibly guess how lifestyles of individuals in society of the 21st century would be changed. I would like to examine whether my initial assumption which indicates that low education level and the number of children in family would decline women’s income more than women with higher education level or with no children turns out to be true. So, with comparative analysis, First, I would like to show how different demographic factors such as age, genders, marital status, education levels, number of children in family and more variables could affect amount of wages and kinds of available occupations. And, second, I want to conduct a research on how strong or weak each of these demographic variables would be related, and how classification of occupations by various demographic factors could reflect the overall demographic interpretation from 1940 to 2000.

 

Methods

Data cover the years from 1940 to 2000 and the raw data for all graphs are obtained from IPUMS 1% samples (1% state fm1 for 1970 and 1% metro for 1980) and the data before 1960 from Alaska and Hawaii are excluded since data are not available for Hawaii and Alaska from 1940 to 1960. The marital status is classified as married, not married (widowed, separated, divorced) or never married status, and the education level is classified as the education level of nursery school to grade 4, grade 5 to grade 12, 1 to 4 years of college or higher, or no schooling. The data are weighted by PERWT, except for the data for Figure 3, which is weighted by HHWT.

Figure 1

This box plot shows the income of people of age 20 to 40 from 1940 to 2000. The figure shows 10th, 25th, 50th, 75th, and 90th percentile of income of white and non-white people sorted by same sex of each race. The currency is in 1999 dollars and income has been adjusted with IPUMS CPI99 and the Census top-coded income in all years from 1940 to 2000. For consistency of analysis, I have applied the lowest top code ($5001 in 1940 dollars; $59,941.99 in 1999 dollars) to incomes in all years. This variable is weighted by person weight (PERWT) for all years except 1950 for which income is weighted by sample line weight (SLWT). The Race categories of these visualization data include white and non-white people.  The x-axis shows year and the y-axis represents median income dollars of the same sex of people of different races.

Figure 2

The bar graph shows the number of children of women aged between 20 to 40 from 1940 to 2000. The figure shows 4-year college or higher educated and less than 4-year college educated white women. The x-axis shows year; y-axis represents proportion of the percentage of number of children women have in each year. This bar graph indicates the relation of the number of children, ages, educations levels by different times.

Figure 3

The line graph shows the median income of white women aged 20 to 40 by their marital status and number of children in 1940 to 2000. The currency is in 1999 dollars and income has been adjusted with IPUMS CPI99 and the Census top-coded income in all years from 1940 to 2000. For consistency of analysis, I have applied the lowest top code ($5001 in 1940 dollars; $59,941.99 in 1999 dollars) to incomes in all years. This variable is weighted by person weight (PERWT) for all years except 1950 for which income is weighted by sample line weight (SLWT). The race categories of these visualization data include white and non-white people. The x-axis shows year and the y-axis represents median income dollars of same sex in different race of people. Here, we can see that regardless of what marital status both parents have, the number of children in family impacts a lot on wage gap between different groups of couples by their family structure.

Figure 4

The bar graph shows the occupations of white people aged 15-65 by sex from 1940 to 2000. This graph portrays the labor division trends among men and women throughout this period.

Result

From the visualization data, I can see how women’s marital status, the number of children in family, and education level can differentiate women’s social roles and status in society by race and sex.

Code for figure 1 is available on Github

IncomeBb

Figure 1 shows the income difference between white male and female and non-white male and female by year

The figure 1 simply shows unequal wages earned by white and nonwhite men and women by sex and race. Each figure represents the amount of wages by white and non-white populations. We can find that in 1940s, the wages earned by white are much greater than those earned by non-white population, but in 2000, the gaps between two populations got smaller. As expected, regardless of their own sexual identity, white people typically get higher income than non-white people. The graphs indicate that for male population, white people consistently earned more money than non-white people while for female population, in 1950, white and non-white population earned very similar amount of wages and even in 1980s, the amount of wages for non-white population was slightly beyond that for white population. The increasing number of women’s involvement in labor movement played significant roles in bridging wage gap between women and men by their education level and marital status. In Great Depression, women’s traditional roles as mothers who predominantly manage housework could not be maintained as before. Many men lost their jobs and lack of usual male providers lead women to go out of house and work to pay. Before the World War I, increasing trend in female involvement in labor was considered as unfortunate and not feminine things, and as non-white women’s work in poorer class. However, in 1920, unmarried female officers were glamorized and a new ideology about female involvement in labor was formed (Sara, 1989). In labor population, one quarter of female workers of age 16 or more, worked, and in 1920s, 30 % of female workers were engaged in office jobs and sales marketing. Mainly, office jobs were highly positively considered as white collar class, and most of white women occupied these positions. Such office jobs gave opportunities to form a new ideology of society saying that women could work outside but, still, professional jobs which women quite positively evaluated from the end of 19th century to the beginning of 20th century. At the turning point of the 20th century, most of office jobs were occupied by male workers. Yet, overall change of life in 1920s was successful in leading women into individualistic atmosphere of consumer economy, and the separation of housework and outside work which formed traditional middle class ideology of gender was barely recognized.  This way, female social reform brought the foundation of welfare country in 1930s. The Great Depression in 1929 distressed many families whose socioeconomic status were determined and fixed by occupations and the amount of wages of male heads in families. In this period, male gender role could not be easily implemented by earning money to support family. Black women could not get jobs in manufacturing industry and, instead, started home working (Margaret, 1993). Therefore, between 1930 and 1940, home working increased by 25 % and most of workers engaged in home working were Black. In 1930s, 55 % of all home workers were non-white and in 1940s, this proportion reached 64% or two thirds of total workers (Sara 1989, 311)

After baby boom occurring from 1940s to 1950s, the fertility rate reached the lowest point at the end of 1960s. After World War II triggered many women to enter the labor market with economic necessity, some women aspired for upward mobility in economic world, but, most of their voices were unheard since women were considered as “disenfranchised appendage” to men (Pettigrew, 1974, 72). From the end of 19th century to the beginning of 20thcentury, discouraging working conditions and environments motivated many female human right activists to participate in labor campaigns. From the beginning of the 20thcentury, female activists’ movement for civil rights and equal rights in the work place got more and more encouraged. In 1940s, almost 0.8 million female workers joined in labor union, and fought against the longstanding opinions arguing that female workers should work at home rather than in factory. This historical background shows that the traditional gender norm that men are more specialized in working in the labor market while women is more specialized in working at home became one of the critical factors which perpetrated the discriminated wage income between men and women. Men are more likely to be employed in high-skilled required and hence high-paying jobs and many times employers believed that men work more productively than women since they thought women are more suitable to feed and support their families at home. Since men were more occupied with high-wage jobs, proportionally, regardless their marital status or level of education attainment, women were more likely to be hired in low-paying jobs in the past, either they were forced to do so or by their will. In the beginning of 20th century, gender identity, rather than other socioeconomic status or aspects, became a crucial part which widen the wage gap between men and women than in the end of the century.

Code for figure 2 is available on Github

ChildEdu

Figure 2 shows the number of children by women age and education level

Figure 2 shows the number of children by women age and education level

The figure 2 shows that differences in women’s education level and marital status not only by their race and sex influence on their fertility rates and eventually their working conditions. Women tend to receive more college and graduate degrees than men but female workers, on average, receive less wages than men (Chamie, 2014). The relation of ages and education levels resulting in the number of children. For women in age of 21 to 25, as the years go, the proportion of women with no children increase but women with no 4-year college or higher education have higher proportion of one or two children. As the ages go up, the proportion of one, two or more children consistently increase in both groups of women. For women of age 26 to 30, women without 4-year college or higher education have higher proportion of three or more number of children than those with 4-year college or higher education regardless of years. According to Pettigrew, women tend to believe that women are supposed to be economically reliable to their husbands and obey the traditional norm which says primary role of women is to take care of their family at home (Pettigrew, 1974). However, as women’s participation in labor market has increased during World War II with their economic necessity, they realize that they can also be significant contributors to economic growth and development of a country.  The Federal Civil Rights Act of 1964 enforced a law prohibiting discrimination in employment between men and women, however, still there is a large gap in terms of their earning based on their sex (Pettigrew, 73). However, the Civil Rights Act did not necessarily help to decline the income gap between men and women. Marriage and childrearing have always been critical factors which affect the employment and the wage rate, especially for women. Also, Sara’s study shows that women with more children were paid less money than women with less or no children and even mothers were less likely to work in the paid jobs. In contrast to decline of married women’s labor force participation, however, married men with children are more likely to earn and work in the paid labor force than men without children (Sara 1989, 323). Hence, this phenomenon may become a potential factor which leads more women of higher education earners to postpone having babies compared to less educated women (Hoffman, 1974).

In addition to this, Eileen Patten argues that there is an income gap across different race and ethnicities. Even though white and Asian women narrowed the wage gap from 1980 to 2015, black and Hispanic women only earned almost no higher fare than before. The situation among black and Hispanic men is not the same as with women. The black and Hispanic men have not made much progress in narrowing the income wage between white men since 1980 (PewResearch, 2016). Some researchers predict this result is because of the lower education attainment rate among blacks and Hispanics than whites, since the U.S. workers with a four-year college degree earn higher wage than those who do not (Patten, 2016). However, the main problem is that people at the same education level earn different wages regarding to their race, ethnicities, and gender. For example, according to 2015 census data analyzed by the Pew Research Center, college-educated black and Hispanic men earn about 80% of what white college-educated men. And Asian college-educated women earn about 80% of what white college-educated men (Patten, 2016). In addition, figure 2 shows the median income gap between men and women which demonstrates that during the 1980s to 2000s, women’s income has been increased while men’s income has decreased. The graph show that there is noticeable growth in women’s earning as the period reaches to the end of 20th century, whereas men’s earning slightly decreased and increased again in 1990s. According to Status of Women in the States, women’s labor participation rate increased largely from 1950 to 2014. The women’s labor participation rate has been increased from 33.9 percent in 1950 to 43.3 percent in 1970 (IWPR, 2015). Also, we can see that there is still gender based division of labor persists across the sector of employment. For example, employment in services such as health care, nongovernmental education, leisure, and other services are female-dominated, but only one in four men work in these industries. Also, the construction industry, manufacturing, and transforming and communications fields of employment are male-dominated even in the contemporary era (IWPR, 2015). The research addresses that there are many factors contributing to persistent wage gaps but they can be explained mainly by differences in education, labor force experience, occupation, or industry (Patten, 2016). According to a survey on racial discrimination by Pew Research, about 64% of blacks say black people are unequally treated in the workplace and especially less fairly than white people. On the other hand, only about 22% of white and 38% of Hispanic agree to this.

Code for figure 3 is available on Github

 

WhiteFamily

Figure 3 shows the income difference between women in various family structure

Figure 3 shows that women’s wage differences by their family structures. Single women aged 20 to 26 earn roughly 17 percent more than married women in 2015 (Chiodo and Owyang, 2003) From the studies, Chido and Owyang found that age and marital status influences on women’s wages. Several studies argue that there is not a direct link between wage and marital status but the timing of marriage and the wages are related to each other (Chiodo and Owyang, 2003). Timothy Chandler, Yoshinori Kamo, and James Werbel’s studies show that postponing marriage increases a women’s wage and it indicates that women’s early job experience before marriage affects the increase of their wage (Chandler, Kamo, and Werbel, 1994). In sum, the studies argue that employers tend to believe that young single women without children are more likely to dedicate to their companies than married women with children. Women tend to receive more college and graduate degrees than men but female workers, in general, receive lower wages than men. Women who married without children earn the highest income as compared to other groups of women who married with children, single with children or without. The proportion of married women without children tend to have better incomes than married women with children or single women with children or without children. The existing wage difference between married males and unmarried males can be partially explained by the result of a division of labor between husband wife. Historically, husbands were specialized in labor market work and wives were specialized in housework; married men are likely to be motivated to work harder to support their family (Blackburn and Korenman, 1994, 255). Many men and women attain higher education as time goes by, but still the income wage gap has not been bridged enough. And especially, women are hugely impacted by their fertility rates and marital status when they are employed.

Code for figure 4 is available on Github

Occup

Figure 4 shows the labor division of white people by year and sex

Figure 4 shows that more than half of women did not have any occupation until the end of 20th century, and farmers, craftsmen, operative, laborers were predominantly male-dominated jobs. Women’s participation in labor market largely increased by the end of the 20th century, however, even today, gender based division of employment persists in the United States. Many American men work in construction, manufacturing, and transportation and communications industry (IWPR, 2015). Moreover, the proportion of women who were employed in professional or managerial occupations, such as lawyers, doctors, nurses, and teachers has increased since 2004 (IWPR, 2015). However, in general, women earn substantially less income than men in the same professional and managerial occupation.  However, in general, women earn substantially less income than men in the same professional and managerial occupation. From this visualization data, we can assume that the occupational segregation between white and non-white people has been one of the potential factors that results in income inequality between men and women.

 

Conclusion

This project examined how some of critical socioeconomic features such as marital status, the number of children in family, or education achievement impact on differences in wage between men and women by race and sex throughout the 20th century. From the study, I could see that after the mid-20th century, the ratio of women’s labor market participation noticeably increased and women’s labor became more important. However, I also could see the certain pattern that came out with working women as she earned education degree, got married, and had babies. There were several literatures that support my initial assumption that married women with young children would have the low employment status and have low income jobs. Some literature explained the traditional gender norm and stereotypes in both men and women toward women caused more employers to hire men over women. Basically, women with young children barely had opportunities to have full-time jobs since they had to endure the burden came from working at workplace, doing housekeeping, and child-rearing. Hence, the literature I used support the idea that higher-educated and couples with no children were more likely to earn better wage than who were less-educated and with children. It made me see the pattern that traditional gender norm became even more strengthened as women got married and had babies. The graph of division of labor by sex and the noticeable division of male-dominated jobs and female-dominated jobs tell us that there were still stereotypical structures in industry where many industry sectors differentiate men’s jobs or women’s jobs. Even though making specific causal and effect connection between certain socioeconomic aspect and wage gap is still very complicated, the gender wage gap existed regardless of what race or ethnicity the person had. I think further research should focus on family structure by each year in 20th century and work on examining whether these patterns existing in the 20th century could be repeated in the 21st century in different formats of social structures.

 

 

 

 

 

Work Cited

  1. Pettigrew, L. Eudora, L. Thomas Keith, and Homer C. Hawkins. “Sex Discrimination and the American Labor Market: A Perspective.”Sociological Focus 1 (1974): 71-86. Web.
  1. “Changes in Women’s Labor Force Participation in the 20th Century: The Economics Daily U.S. Bureau of Labor Statistics.”S. Bureau of Labor Statistics. U.S. Bureau of Labor Statistics, n.d. Web. 01 Nov. 2016
  2. Evans, Sara M.Born for Liberty: A History of Women in America. New York: Free, 1989. Print.
  1. Hobbs, Margaret. “Equality and Difference: Feminism and the Defense of Women Workers during the Great Depression.” (1993). Web
  2. “Employment and Earnings – Women in the States.” Women in the States Employment and Earnings Comments. Web. 04 Nov. 2016.
  3. Patten, Eileen. “Racial, Gender Wage Gaps Persist in U.S. despite Some Progress.” Pew Research Center RSS. N.p., 01 July 2016. Web. 03 Nov. 2016.
  4. Chandler, Timothy; Kamo, Yoshinori; and Werbel, James. “Do Delays in Marriage and Childbirth Affect Earnings?” Social Science Quarterly, December 1994, Vol. 75, No. 4, pp. 838-53.
  1. Chiodo, Abbigail j. AND Owyang, Michael t. “Marriage, Motherhood and Money: How Do Womens Life Decisions Influence Their Wages?”, 2003.
  2. Blackburn, McKinley and Korenman, Sanders. “The Declining Marital-Status Earnings Differential.” Journal of Population Economics, 1994, Vol. 7, No. 3, pp. 249-70.
  3. Hoffman, Lois Wladis. “The Employment of Women, Education, and Fertility”, 1974.
  4. Chamie, Joseph. “Women More Educated Than Men But Still Paid Less.” YaleGlobal, 2014. Web.

Women, Marriage, Education, and Occupation in the United States from 1940-2000

Introduction:

Question: From 1940-2000, how did women’s involvement in higher education shift, and what influenced it?

Women’s participation in higher education increased in the 1960s and 1970s. Why? What factors influenced this trend from 1940 to 2000 and beyond? Starting in 1979, more women have been enrolled in higher education than men in the United States (Touchton, 50). In 2014, 30.2% of women had a bachelor’s degree, compared to 29.9% of men (Feeney). Women’s participation in higher education has been and continues to be influenced by many factors, including race, social norms, and marriage status.

Continue reading

A Woman’s Wage: Trends by Race and Family Structure 1920-1990

Introduction

This post investigates the social construction of the woman’s wage over the 20th century (1920-1990). The woman’s wage is a measure of the inclusion of women in society, and we gain important insights when we consider wage as a social construction rather than a theoretical outcome of supply and demand (Kessler-Harris 1991). Early in the 20th century, the minimum wage for women was defined in relation to the amount of money needed to support a family, termed the “family wage,” and the amount necessary to support the American standard of living termed the “living wage” (Kessler-Harris 1991). It was the social norm that the family wage should be earned solely by the breadwinner husband leaving the homemaker wife free to nurture her family (Folbre and Abel 1989, Kessler-Harris 1991). So, women’s wages were thought to be given at the expense of family, and employers paid women a minimum wage much lower than the standard living wage because society assumed a husband or a father supported them. In this way, men were paid on the value they created and women on their assumed need. This construction of a woman’s wage reproduced female dependence on husbands and fathers as well as a gendered workplace.

What is the legacy of this conception of the American family and how has the social construction of the woman’s wage changed over time? How has the woman’s wage been socially constructed differently for women of color? I analyze census data to look at trends in labor force participation and income of women by race and family structure. I argue that the social construction of the “family” and the women’s wage are interrelated: family status is an important factor in the social construction of a woman’s wage and woman’s wages may also affect family status and conceptions of family. I also argue that the woman’s wage is constructed differently for women of different races, as they have fundamentally different experiences than white women (Omi and Winant 2014). Increasing our understanding of the remarkable changes over the 20th century is still important, especially because in the United States women still earn only 80% of what men earn on average (Bailey and DiPrete 2016).

Continue reading

Working Women, Race, and Family

Introduction:

The changing roles and rights of women is an understandably well discussed topic.  Through 1920-1970 women gained more freedom and more independence, perhaps most clearly shown by increased numbers in the workforce, granting women their own incomes.  Slow increases in working population accompanied with big booms for the World Wars are well documented.  It’s also been well tracked that non-white women work more, and work for less.  A major topic with regards to women working in the 1900s was its effect on their family lives.  At the time there was much discussion on the morality and propriety of working mothers.  Overall, female employment has been well documented with respect to race and family values.  By use of census data, these topics can be more clearly numerically supported and more specifically analyzed with regard to family values.  Not just with what was considered correct at the time, but what was actually happening with regard to women’s employment, relationship status, and number of children.

Continue reading