History 90.01: Topics in Digital History — Portfolio Introduction

The following body of work was produced throughout the participation in Professor Emily Merchant’s course at Dartmouth College titled “History 90.01: Topics in Digital History”. Throughout the course we examined the general history and development of the US Census as well as the role the Census played in both making and recording history. The course was divided into four sections: “Race”, “Migration”, “Labor”, and “Families”. For each section of the course we were required to produce a detailed analysis, using data from the University of Minnesota’s Integrated Public Use Microdata Series, of a specific topic in history and its relation to the Census.

Each section of the course helped me to better understand the history of the topic as well as develop my skills and knowledge in data analysis and visualization. For the Race unit, I looked into the growing demographic trend of missing Black men in the city of Chicago and it’s metro-area, studying how the US Census Bureau has historically used the Census for both population control and policy development. For the Migration unit, I analyzed a trend with personal relevance; differing historical emigration trends out of Texas from the middle of the Great Migration to 2000. For the Labor section, I examined the effect of the institution of the 40-hour work week on wage-income trends across race and sex cohorts from 1940-1990. For the Families section, I looked at the differing relationships between marital status, child custody, and wage incomes across sexes.

This course developed both my knowledge of the history for each section as well as my critical thinking skills for analysis of history and data in general. I will use the frameworks through which we analyzed history to help better understand current events. Furthermore, the data analysis and visualization techniques learned here will be invaluable in my professional career.

Project Portfolio

For the class, U.S. History Through Census Data, we used IPUMS data to answer questions about various aspects of American history.  We used IPUMS data to research how a certain governmental position or cultural phenomenon affected the population by making graphs of the data.  These graphs allowed us to observe patterns in the data that helped answer our questions.  IPUMS data is particularly helpful in this regard because they have synthesized data from all censuses beginning in 1850.

We completed four project over the course of the term on the categories of race, immigration, work and family.  While there was no requirement to relate the projects, my projects on race and family both looked at the effects of slavery on the African American population in America post-emancipation.  My first project on race researched the population distribution of the ‘black’ race category following the civil war.  Ultimately I found that the large population of African Americans living in the south after the civil war primarily stayed in the south likely due to an inability to leave.  My fourth project looked at marriage rates in the ‘black’ population compared to the ‘white’ population in America following the civil war.  I found that the institution of slavery did not have a noticeable effect on marriage rates between the two races.  My projects on immigration and work both focus on the Immigration Act of 1924.  The second project on immigration looks at the overall immigrant population before and after the Act as well as the percent of different immigrant groups in the total immigrant population.  My third project looked at the female immigrant population’s participation in the labor force during the same time period as my second project.  It looks at which occupations female immigrants are drawn to as well as the percent of each immigrant group in those occupations.

Marital Status of the African American Population post-emancipation

Slavery is one of the most highly researched and talked about institutions in America.  The effects of slavery can still be felt today in many parts of American society.  This post focuses on how slavery has culturally affected the African American population in comparison to the white population.  Many abolitionists believed that slavery permanently damaged the African American population due to sexual abuse, slave breeding, and other horrible acts practiced by slave owners (Mintz and Kellogg, 1988, 67).  More importantly, abolitionists also claimed that slavery destroyed family life in the African American population due to the undermining of slave parents and frequent breakups of families because of the practice of buying and selling slaves/family members, and the illegitimacy of slave marriages (Mintz and Kellogg, 1988, 67).  Mintz and Kellogg prove in their book, Domestic Revolutions: A Social History of American Family Life, that many of these allegations made by abolitionists were not in fact true or too extreme.  While the effects of slavery have not been as extreme as many people think, I believe that there must have been some affect on the African American population culturally, especially compared to the white population.  In this post I focus on the marriage rates within the adult African American population compared to the white population from 1870-1970 in order to see if slavery has effected marriage practices in the African American community.


I gathered data from the Integrated Public-Use Microdata Series (IPUMS) for this project.  I included 1% samples from the years 1850-1960.  I used the RACE, AGE, REGION, PERWT and MARST variables from these samples for my research.  The PERWT variable represents the sample weight of each individual.  I separate the RACE variable into ‘white,’ ‘black,’ and ‘other.’  The ‘white’ category includes the white population of the United States which incorporates most European immigrants as well as Mexican immigrant until the early 20th century.  The ‘black’ category encompasses all persons that reports themselves as black as well as those who were classified as mulatto (In the census from 1850-1930), quadroon  or octaroon (In the census from 1890-1930).  The REGION variable indicates the region where each individual is currently living during the year of that census.  I have grouped the regions in the West, Midwest, South, and Northeast for this project.  The MARST variable denotes the marital status (married, divorced or separated, widowed, never married) of the individual.


I graphed the ‘black’ and ‘white’ populations from 1850-1970 by region using marital status as the filler for the bars.  I targeted the portion of the population of marrying age and excluded those too young to be married (less than 20 years of age) and the elderly (I believe they might skew the data with higher rates of widowing).  I created two graphs that illustrate this data in which each race and region has its own graph.  Each region and race category is laid out next to each other in order to see how they differ.  The first graph represents the total data as it is whereas the second graph displays that data as percents of the total in order to see which marital status is most common.  Both graphs include men and women.  The code used for this project can be found here.


Figure 1


Figure 1 shows the total populations of the United States from 1850-1970 by region and race with marital status as the fill for the bars.  Clearly the ‘white’ population is far greater the the ‘black’ population in each region with regards to total numbers.  It is difficult to discern wether or not marriage rates are higher in the ‘white’ or ‘black’ population from viewing the total populations.  There is not much to see regarding the ‘black’ population, but the ‘white’ population is large enough and grows enough each year to see changes in it.  The ‘Widowed’ and ‘Separated or Divorced’ categories do not have a significant impact upon the ‘white’ populations in this graph.  In the West and South, the proportion of ‘Married’ to ‘Never Married’ individuals remains constant over the time period.  In contrast, the ‘Never Married’ category becomes a larger portion of the population in the Northeast and Midwest in the ‘white’ population.

Figure 2


Figure 2 displays the ‘white’ and ‘black’ populations in each region showing what percents of the populations fall into different marital status categories.  This graph more clearly shows the marriage rates of each racial population and how they differ.  The ‘Widowed’ and ‘Separated or Divorced’ categories remain a small fraction of the population.  In the ‘white’ population, these two categories maintain their insignificant percentage over the time period.  In contrast, the ‘black’ population has a sudden spike in the ‘Separated or Divorced’ category in all regions in 1950, placing it above the percent of the ‘white’ population that falls into that category, and maintains that level for the rest of the data.  The ‘black’ population also has a marginally greater ‘Widowed’ percent of their total population in comparison to the ‘white’ population in each region.  In opposition to the ‘Separated or Divorced’ category, the ‘Widowed’ category decreases in 1950 and remains at a lower percent through 1970.  It is much easier to read the trends of the ‘Never Married’ and ‘Married’ categories.  In each region, the ratio of ‘Married’ to ‘Never Married’ individuals is greater in 1970 than at the beginning of this period.  There is a steady decline of the ‘Never Married’ category in the ‘white’ population, with a slight resurgence in 1960 and 1970.  The ‘black’ population also shows a steady decline in the ‘Never Married’ category and an increase in its ‘Married’ group until 1940 when that trend reverses.  The ratio of ‘Married’ to ‘Never Married’ individuals in the ‘white’ population is noticeably greater by 1970, but was practically equal for the first half of the 20th century.  These patterns are found in each region.


These results do not follow the claims made by abolitionists.  The percentages of each marital status within the ‘white’ and ‘black’ populations are more or less equal from the emancipation of slavery up until 1950 when the ‘black’ population gains a greater percentage of people who have never married.  From this we can conclude that the institution of slavery did not culturally make the ‘black’ population differ from the ‘white’ population in America with regards to marriage.  The ‘black’ population diverges from the ‘white’ population with regards to marital status starting in 1950 and continuing through 1970.  I do not think that this change in the data can be attributed to a late result of slavery.  If slavery had made the ‘black’ population view marriage differently than the ‘white’ population it would have immediately created a noticeable disparity in marriage rates starting in the 19th century.  A contemporary cultural event must have affected the ‘black’ population in America to start such a change.  Seeing that this began in the 1950 census, it is possible that WWII could have affected the ‘black’ population in such a way.  Furthermore, the civil rights movement, while it began in earnest during the 60s, may have been starting since the end of WWII and could have affected the ‘black’ population with regards to marriage rates.  These are only two guesses at why the change in data I found occurred.  For the purposes of this post, the data has illustrated that the institution of slavery did not seem to affect how marriage was culturally viewed by the ‘black’ population compared to the ‘white’ population in America following the civil war.

Works Cited

Mintz, Steven, and Susan Kellogg. Domestic Revolutions: A Social History of American Family Life. New York: Free, 1988. Print.

Immigration and identity

My portfolio uses IPUMS data in order to explore trends that differentially affect the outcomes of black Americans and immigrant Americans. By exploring questions of racial declaration, occupational standing, and household and individual metrics, I hope to quantify many of the same trends that researchers have long observed, within the context of a specific historical period. My research largely relied on the age, sex, birthplace, and racial declaration variables, as well as counts of individuals and children fitting specific criteria.  I also drew on earning and education quantile score in order to assess the opportunities available to first-generation Americans. Other variables I used included citizenship status, year of immigration, and parental birthplace.

My first project considered the classification of black-white biracial Americans and how they were classified for census purposes when forced to choose one part of their identities. My second project considered first and second generation Slavic Americans changes in each cohort’s population by census year. Links to prominent events in the early-mid 20th century were also made. My third project utilized the R Shiny framework to interactively explore changes in occupational standing for immigrant Americans by decade and over years-since-immigration. Finally, my last project considered the number of children sharing a household and whether or not this statistic varies with the immigration status of the householder and his/her spouse.  This analysis showed highly similar results at the individual and household analysis levels.

Although error codes and comparability issues made these investigations frustrating at times, the clean and complete nature of the IPUMS-USA database made it an absolute pleasure to explore.






Intercensal children, by person and household

For this investigation, I decided to explore the number of intercensal births within a household by the number of immigrants leading that household. For my purposes, the leaders of a household are defined as the householder and his or her spouse.

Choices and trends in fertility have historically depended on both cultural attitudes and economic needs. Acknowledging that first-generation immigrants often differ from the average native-born American in these areas as a result of upbringing and the unique pressures they face in America, I decided to investigate the number of intercensal births within immigrant and non-immigrant households from 1900-1950, with the hope that some of these differences would be reflected in the resulting data.

I decided to use both the immigration status of both the householder and his or her spouse due to the problematic nature of the householder designation (Smith 1992, 422). The census has historically done a poor job defining this designation, or keeping it consistent from year to year. In some years, women were barred from being the household head. In others, enumerators were instructed to use a woman’s husband if she claimed to be the household head. Smith (1992, 430) speaks of a time in earlier history when the title had real meaning. Before women had the right of personhood and standing under the law, the head of her household (i.e. father, then husband)  would be accountable for her actions. With this archaic mechanism gone by the 1900s, it makes little sense to limit our analysis to just householders.

Data and methodology

I turned to the Integrated Public-Use Microdata Series, using census samples from 1900 through 1950. I used the 1% sample provided by IPUMS for each year studied. Before any further stratification, I filtered to only include individuals in regular households (that is, individuals not living in group quarters).

After this, I labelled household heads and their spouses by immigration status (foreign born or not) and created an indicator variable for all children born within the previous intercensal period (i.e. those less than ten years of age). This definition of child allows me to avoid counting the same child in two census years and allows me to draw more accurately proxy for fertility within a given intercensal period.  No attempt was made to determine biological linkage through the relate variable; this information is not relevant to the question at hand. Rephrasing the question to include only biological children would leave me with no way account for foster children, adopted children, stepchildren, and young relatives.

With that in mind, it should be noted that many of the households we will find with a large number of children will be in this scenario due to cohabitation of multiple families in a single house. Hareven (1974, 325) discusses some of the difficulties with distinguishing between family structure from habitation arrangement. In this respect, concerns about the increasing prevalence of “broken homes” do not always consider the reasons behind a parent living elsewhere (ibid.).

The immigration status dummy variable was aggregated by household as was the child dummy variable. As a result, I was able to stratify the population of children by the number of children that they lived with, and how many householder/spouses in that child’s household were first generation immigrants. This analysis was performed at the individual (i.e. how many children live with x-many other children) and household (i.e. how many households have x-many children) level.

The trends within and across years were clear and striking. Although the individual-level analysis is biased in proportion with the number of children in a household, this effect was not significant in my analysis. The reason is that much the bias it is cancelled out when one considers the proportion of children/households of a certain criteria rather than an absolute count (where the multiplicative nature of the bias would be apparent). The investigation confirms this by noting the general lack of distinction between Figures 1-3 and Figures 4-6.


I begin by considering the earliest census in the dataset. As is clear (in Figure 1), the proportion of children in each cohort that lives in a one-immigrant household is fixed (at about ~15%) for most of household size configurations. This number is identical to the proportion of children without any compatriots in two-immigrant households. Nevertheless, I find that the proportion of children in two immigrant households steadily increases to ~30% when I consider the 8 child household configuration. This concentration comes at the expense of the proportion of no-immigrant households with a large number of children. One might speculate that additional children are more desirable to two-immigrant households as a result of attitudes from their home countries. However, I find that at the extreme (9 children or more), only no-immigrant households are represented. Given the low total number of children living in these configurations, it may be possible that this phenomenon is the result of fundamentalist religious doctrine.

Figure 1.

Figure 1.

By 1950, I see that the landscape has shifted dramatically. Through the 8 child configuration, native born households dominate, increasing their share at every tier (Figure 2). Unsurprisingly, I can recall the effects of the baby boom. Having children the post-war days was more convenient, fashionable, and encouraged. This economically prosperous time meant that soldiers returning from WWII could worry less about the cost of child rearing and weigh its benefits more heavily.

Figure 2.

Figure 2.

In searching for the point of inflection between there two contrasting trends, I can consider the children born in the 1920’s (Figure 3). In the first five configurations, the proportion barely change by immigration status. Children living in immigrant households account for approximately 10% of children within their configuration cohort. The uptick I see in the proportion of children living in non-immigrant households in the higher configurations does not appear to conflict with the overarching trend.

Figure 3.

Figure 3.

Household-level results

As mentioned, I can compare the above results to an analysis at the household level. That is to say, I may choose to consider the proportion of households in any configuration cohort. My use of proportion means that I will did not expect to see any bias in my initial analyses and thus any subsequent analyses at the household level should be fairly consistent. Doing this analysis at the household level also allows me to consider homes without any children. In my previous charts, this configuration is unaccounted for because there is no child present to experience it.

Turning to Figure 4, I find few notable differences between the comparable categories in this figure and Figure 1. One discontinuity worth noting is the higher proportion of one-immigrant  households with no children. Perhaps this group includes a significant portion of young and recent immigrants beginning their families. Nevertheless, this anomaly does not effect the overall trend observed.

Figure 4.

Figure 4.

The second chart (Figure 5) is remarkably similar. Admittedly, many subtle differences between 2 and 4 are likely lost to the overwhelming visual effect of the baby boom. Nevertheless, there is little worth noting here.

Figure 5.

Figure 5.

Figure 6 leaves me in a very similar situation. Controlling for absolute size by using proportions allowed me to eliminate virtually all distinctions between my personal and individual level analyses. 1930 still marks the inflection point in trends, with great similarity across child count configuration cohorts.

Figure 6.

Figure 6.

In order to obtain a better sense of the size of each cohort, and the relative dearth of households with more than two children, I chose to visualize the data in through a mosaic plot. The area of each rectangle in a mosaic plot is proportional to the number of observations represented by that cohort. Looking at the year 1900, I limited the dataset to households with no more than seven children and plotted the information in Figure 7.

Figure 7.

Figure 7.

This figure really helps put many of the trends I had observed earlier into perspective. Although there were patterns worth observing in families with many children, these patterns only really describe what is going on in a small minority of households.


As per predictions, there were notable trends in the number of children born in immigrant households between censuses. The fact that these trends shifted from decade to decade implied that the attitude towards children in America, among non-immigrants changes quickly and dramatically. The rise of the nuclear family, which coincided with the Baby Boom, is one example of a recent trend largely resulting from  example of this from recent US history, and is likely driving the inversion in trend that we see between 1900 and 1950.


Source code for visualizations

Hareven, Tamara K. “The Family as Process: The Historical Study of the Family Cycle”, Journal of Social History, Spring 1974; Vol. 7, No. 3: 322-329. Print.

Ruggles, Steven. “Family Interrelationships”, Historical Methods, Winter 1995; 28(1): 52-58. Print.

Smith, Daniel Scott. “The Meanings of Family and Household: Change and Continuity in the Mirror of the American Census”, Population and Development Review, Sep. 1992; Vol. 18, No. 3: 421-456. Print.

The professional mobility of immigrants

In this investigation, I examine whether or not immigrant groups who had arrived in America before 1930 are able to benefit from professional mobility. I do so by considering data on occupational earnings and education quantile scores from the 1900 to 1930 censuses. I address whether or not mobility existed in a specific rregion-time combinations for certain groups while also considering how access to professional mobility might have changed over time, as measured by these indicators.


Given America’s reputation as a country of immigrants, it should not come as a surprise that factors associated with America’s economic greatness, including hard work, thrift, unrelenting progress, have long been associated with the journey experienced by many of the newest Americans.

Adelman and Tolnay (2003, 182) speak of the preference that certain groups of immigrant workers received over minority Americans for certain jobs. This was largely the result of an employer’s earnest belief that immigrants (often those from particular countries) are uniquely qualified for a job on the basis of their country of origin (ibid.).  Be that as it may, the American Dream relies on more than just the aforementioned leg up; it requires the possibility of social mobility across and within generations. Those who work hardest should be able to ascend the proverbial ranks.

The question becomes: can an individual who betters himself truly make it? This investigation explores this in further detail, particularly in the context of first-generation American immigrants. Nevertheless, it is difficult to do so objectively and quantitatively. As unfortunate as this may sound, one way to take stock of the professional and social mobility of a group is to compare it to a reference group that is doing more poorly in these areas, noting the differences. Adelson and Tolay (2003, 179), find that the greater white collar employment rate and average SEI score for immigrants (when compared to black Americans) immediately following the Great Migration points to a greater access to opportunity and professional mobility. The disparities indeed narrow by the  end of the century.

I will be using the population of workers in 1950 (who would have uniformly distributed scores for the 1950-basis occupational standing by earnings and education) as my reference group, knowing that this group better approximates the total population for a given year, as we move forward in time. This choice was made as a result of IPUMS’ choice of base year for generating these metrics.

Data and methodology

I turned to the Integrated Public-Use Microdata Series, using census samples from 1900 through 1930. I used the 1% sample provided by IPUMS for each year studied. Before any further stratification, I filtered the dataset to only include individuals with birthplaces outside of the United States and any other outlying territories or possessions.

I chose this time frame because all four of the censuses conducted within it asked immigrant respondents for their year of immigration. Although its accuracy may be effected by prevalent attitudes on immigration (or rather the effect of these attitudes on self-reporting), this variable is consistent in its definition across all years, with the relevant questionnaire text never changing in a material way. From this variable, I was able to construct a years-since-immigration estimate for every immigrant. I factored the resulting estimate by rounding the values it produced to the nearest five year interval. This was done to make the cohort-level spineplot analysis more understandable and meaningful.

The other two variables worth noting are the 1950-basis occupational education score and the 1950-basis occupational earnings score, which serve as my outcome variables. For a given occupation in any given year, these occupational standing variables represent the quantile rank of the mean outcome among workers of that occupation in 1950 (i.e. the quantile rank of the mean earnings of plumbers). Although use of these scores has garnered some academic controversy, Sobeck (1995, 49) highlights some misconceptions about them. A major claim is that they are distorted by changes in job definitions over time. Sobeck (1995, 51) refers to an earlier article that showed this concern to be empirically insignificant. Indeed, Sobeck highlights that this concern (which is at the root of many others) is far less significant for censuses before 1950 than for those conducted in 1960 or later.

This class of metric is valuable for a number of reasons. It allows us to proxy for these outcomes when I cannot access this information for all years in a comparable way (i.e. when considering the 1900-1930 time frame). It is a quantile measure, eliminating the need to account for factors such as monetary or academic degree inflation. Most importantly, it is resistant to the effect of any discrimination in the actual wages paid to immigrants or any level of educational over-qualification. In this sense, these occupational standing variables help highlight the class of opportunities immigrants are receiving rather than simple raw outcomes. Sobeck (1995, 49),  touches on this, stating that “although the income score is derived from individual-level data, it should not be interpreted as actual income […] an occupation with a good score is well-rewarded and probably high-status”. Even with its limitations , this metric of occupational standing is often seen as a superior alternative to SEI and other “subjective” scoring models (Sobeck 1995, 50).

The stratifying variables not worth mentioning in great detail consist of geographic region of residence, gender, citizenship status, and year. I considered adding continent of birth, but the effect that stratifying by another variable would have on computation times led me to eliminate that option.

The filtered data was then used to generate a spineplot on the basis of years-since-immigration and the chosen outcome measure. As mentioned previously, individuals were grouped into cohorts on the basis of these metrics. Outcome scores were split into quartiles and years-since-immigration was rounded to the nearest five year interval. The proportions of individuals represented by each outcome quartile (of all individuals within a given years-since-immigration cohort) were plotted. Doing so controlled for disparities in the number of immigrants across these cohorts.


Link to interactive heat map of professional mobility

Given the nature of occupational standing scores, it is not particularly meaningful to consider absolute quantities. After all, what numerical bearing does the quantile earnings rank of a job in 1950 have on workers in that job in 1910? With all that said, I still comment on observable trends, patterns, and under or over-representation based on what I see in the various spineplots. Indeed, in order to infer over or under-representation, I must make the plausible assumption that the score quartile cutoffs from 1950 roughly correspond with the quartile cutoffs in each year of analysis. That is to say, I assume a fairly uniform distribution of outcome scores (for each outcome) within every year being considered.

Moving onto my chart, I find that the starting position for immigrants (that is, their outcome within 7.5 years of immigration) becomes more difficult from 1910 to 1930. In all four years, nearly three quarters of those with less than 2.5 years of residency fell into the bottom two earning score quartiles. However, the proportion of those within the bottom quartile increases by almost a half between 1910 and 1930.

Overall, it appears that the prospects of male immigrant workers have improved over time, when I focus on earnings score. Whereas men in 1900 were underrepresented in each of the top two quartiles for all residency estimate cohorts, these proportions improved moderately in 1910. By 1920, those within the top two quartiles exceeded 50% for men who had spent approximately 25-35 years in America. By 1930, I find a fairly mobile picture of the employment market place; when an individual has spent approximately 30 years in America, the odds that they land in any quartile are about 1 in 4.  This observation seems to hold true for immigrants when I include women as well. Perhaps if I extended this analysis further into the future, I would find stagnation in improvements, which would support Adelson and Tolnay (2003).

For women only, there are also trends of drastic improvement, although the general deficit in female earnings means that women are generally in jobs in the lower quartiles.  Nevertheless, I find that although half of immigrant women in 1900 and 1910 fell in lowest earning score quartile, this number dipped below a half for women who had spent approximately 5-30 years in America in 1920 and 1930.

Other notable observations include the fact that the distributions change (and begin) less dramatically for citizens than for non-citizens. For most region and time combinations citizens also terns to see better outcomes (i.e. higher representation in better quartiles) than non-citizens. The differences are starkest in the top and bottom quartiles. Non-citizens appeared to have the best outcomes in the Northeast and the poorest outcomes in the South. Conversely, citizens appeared to do the best (i.e. proportion in top two quartiles after two decades in America) in these regions.

When I looked to perform an analogous analysis on education, the results were surprising. For most region, year, gender, and citizenship combinations, upwards of 75% of persons represented were in the lowest quartile, with this number slimming down slightly for every census year. The exceptions to this rule included female citizens in the South and West. While it may be the case that many of the jobs requiring education in 1950 simply did not exist in large numbers during the early 1900’s, I did not find this line of reasoning compelling.

The best conjecture I could come up with after looking at the literature, was that immigrants at this time generally did not begin their careers in jobs demanding a high degree of education. Even if they later shifted into more or less demanding jobs (and demonstrated mobility in earnings score) its not likely that they would move into jobs were education was a priority. This finding is sobering in light of Sobeck’s (1995, 51) caution against over-reliance on the earnings score. He notes that it does a poor job in distinguishing white and blue collar workers, which is an area in which the education score might plausibly excel.


Although the results were fairly consistent with the initial hypothesis, it was surprising to see the stark contrast  between the earnings score and educational score distributions. Immigrants in all decades appeared to have the opportunity to change their lot in life.

While starting position and the degree of mobility varied by year and region, having social mobility as a male immigrant was as good as a guaranteed. Results indicated that the same could be said (with less enthusiasm) of women; particularly of those with citizenship.

The lack of parallelism with mobility in educational score indicates that perhaps the opportunity granted to immigrants is more one-dimensional than I, as a country, as willing to collectively admit.


Source code for shiny application

Adelman, Robert M. and Stewart E. Tolnay. “Occupational Status of Immigrants and African Americans at the Beginning and End of the Great Migration”, Sociological Perspectives, Summer 2003; Vol. 46, No. 2, 179-206. Print.

Sobeck, Matthew and Lisa Dillon. “Occupational Coding”, Historical Methods.Winter 1995; 28(1): 70-73. Print.

Sobek, Matthew. “Occupation and Income Scores”, Historical Methods. Winter 1995; 28(1): 47-51. Print

Wage Income and the 40 Hour Work Week: Sex, Race, and Urban Status

On August 20, 1866, the National Labor Union asked the United States Congress to pass a law mandating an eight-hour workday (Library of Congress, http://memory.loc.gov/ammem/today/aug20.html). After decades of battling with labor unions, on June 26, 1940, Congress finally ceded, officially amending the Fair Labor Standard Act to limiting the workweek to 40 hours (Fair Labor Standards Act, https://www.fas.org/sgp/crs/misc/R42713.pdf). This had a monumental effect on the trend of weekly hours worked throughout the US population.

Some cite that during the same period, the correlation between work hours and wage income has increased (Costa 1998, 1). Furthermore, some attribute the declining wage gap between men and women during the same period to longer work hours for women in the workforce (Mandel and Semyonov 2014, 1598). I am going to argue that the recent increase in correlation between weekly hours worked and wage income is confined to the cohort of White women.


My data have been collected from the Integrated Public Use Microdata Series (IPUMS) from the University of Minnesota. I am using 1% samples of the decennial US Census from 1940-1960 and 1990; for 1970 I am using the 1% State Form 1 sample; for 1980 I am using the 1% Metro sample.

IPUMS variables used in the analysis include URBAN to identify the urban/rural status of an individuals residence (residing in a township with a population of greater than 2,500), EMPSTAT to identify the employment status of an individual, HRSWORK2 to identify the hours worked per week for a sampled individual, and INCWAGE to identify the strictly wage income of an individual. IPUMS variables PERWT and SLWT were used to aggregately weight the sampled individuals. SLWT (sample line weight) is used strictly for the 1950 census.


I manipulated the data to examine the different relationships between weekly hours worked and wage income across urban/rural, and race divides between 1940 and 1990 in order to compare such prevailing trends to the growing connection between women’s working hours and wages during the same time period (Mandel and Semyonov 2014, 1598).

In order to look specifically to individuals in the labor force, I filtered the data to focus on individuals who were between the ages of 18-65, and only surveyed those who were reported by the EMPSTAT variable as currently employed and therefore in the labor force. Next, I separated the individuals in the data into cohorts following their defined urban or rural resident status. I used the Consumer Price Index in order to adjust the reported incomes for inflation to the 2000 level.

The IPUMS variable INCWAGE reports the total pre-tax wage and salary income for each respondent for the previous year. Using INCWAGE I was able to plot the median incomes for each race, sex, and urban status cohort. My R code for the analysis can be found here.


Figure 1:


Figure 2:


         Figure 1 shows the decennial median adjusted wage income for the White cohort between 1940 and 1990. Seen here is a steep increase in inflation adjusted wage income among urban White men in the labor force between the 1940 and 1970 censuses, and then decline in inflation adjusted incomes from 1970 to 1990. Rural white men experience a similar increase in inflation-adjusted incomes from 1940 to 1980, and then see a slight decline from 1980 to 1990. White women living in both urban and rural regions, however, do not experience a decline in inflation adjusted incomes at any point between 1940 and 1990. Rather, in both the urban and rural cohorts of White women, an increase in median inflation adjusted income is seen between 1980 and 1990, a period during which the median inflation adjusted incomes for both urban and rural White men had dropped.

Figure 2 represents the decennial percentages of White respondents in differing cohorts of weekly hours worked. Here we see the impact of the 1940 amendment to the Fair Labor Standards Act on the weekly hours worked throughout the White cohort of the US population. Although data for the 1950 census are not available for the categories, we see a precipitous increase in the percentages of both white men and white women working 40 hours per week. However, it is important to note that the Fair Labor Standards Act amendment had a larger impact on the ratio of white women working 40 hours per week than it did on the ratio of white men working 40 hours per week. This is especially important to note when considering the stated effect of hours devoted to paid work on the gender pay gap in recent decades. Mandel and Semyonov say, “…gender differences in average working hours accounted for 4% of the total gap in 1970 (less than sociodemographic characteristics and occupations). Only two decades later, in 1990, working hours accounted for one-fifth of the gross gap; and in 2010 working hours already explained one-third of the gross gender pay gaps.”(Mandel and Semyonov 2014, 1607). This explanation is consistent with my analysis of the data, as there is a continuous increase in the median wage income of white women between 1940 and 1970, despite a perceived decrease in the percentage of women working for greater than 40 hours per week. Furthermore, following Mandel and Semyonov’s explanation of the gradual increase in the correlation between women’s working hours and pay, the more recent increase in white women’s wage income shown in the 1990 census can be more so attributed to the perceived increase from 1980 to 1990 in the percentage of white women working more than 40 hours per week.

Figure 3:


Figure 4:


          Figure 3 shows the decennial median inflation adjusted income for non-white respondents from 1940 to 1990. For non-white men, the trend in this graph differs slightly from that of the same graph for white men. Inflation adjusted incomes for urban, non-white men did not begin to decline until between 1980 and 1990; inflation adjusted incomes for rural, non-white men did not decline at any time during the period analyzed. The observed trend in inflation adjusted wage incomes for both urban and rural, non-white female respondents also differs from that for both urban and rural, white female respondents. Although both urban and rural, non-white women experienced faster increases in inflation-adjusted incomes during the period examined, the decennial rate of increase between 1980 and 1990 is declining compared to the prior year, rather than increasing, as seen in the trend for both urban and rural, white women.

Figure 4 shows the decennial percentages of both urban and rural, non-white men and women in each of the various weekly hours worked cohorts. Although similar general trends exist here in comparison to Figure 2, it seems that the 1940 congressional amendment to the Fair Labor Standards Act had a magnified in both urban and rural, non-white men and women. Observed here is a greater increase in the percentages of non-white men and women working 40-hour workweek after the amendment than white men and women. Furthermore, when comparing this to Figure 2, there is a very similar increase in the percentages of both urban and rural non-white women working greater than 40 hours per week between the 1980 and 1990 censuses.



The similarities in the most recent (between 1980 and 1990) increase in the percentages of respondents working greater than 40-hour workweeks despite the differing trends in the rates of change in inflation adjusted incomes suggests that the increase in correlation between working hours and wage incomes is confined to urban and rural white women. This more recent trend is different from that observed from earlier in the century by Costa, suggesting an across-the-board increase in correlation.(Costa 1998, 1) Furthermore, Mandel and Semyonov seemingly overlook the difference in the connection of working hours and wage income between white and non-white women (Mandel and Semyonov 2014). The decrease, rather than increase, in the inter-census rate of change in the inflation adjusted wage incomes of non-white women, despite a comparable increase in the percentage of women working greater than 40-hour work weeks, suggests a smaller correlation between working hours and wage income of non-white women. Furthermore, the recent drop in the inflation-adjusted incomes of both white and non-white men despite a similar increase in the proportion of men working greater than 40-hour work weeks between 1980 and 1990 suggests a negative correlation between working hours and inflation adjusted wage income during the period examined.

Works Cited:

2014. “Gender Pay Gap and Employment Sector: Sources of Earnings Disparities in the United States, 1970-2010.” Demography 51: 1597-1618

  1. “The Wage and the Length of the Work Day: From the 1890s to 1991..” Journal of Labor Economics 18: 156-181

Mayer, Gerald, Benjamin Collins, and David H. Bradley. The Fair Labor Standards Act: A Resource Guide. Fairfax, VA: International Asoociation of Fire Chiefs, 1994. Fas.org. Congressional Research Service, 4 June 2014. Web. 6 Mar. 2016. <https://www.fas.org/sgp/crs/misc/R42713.pdf>.

“Today in History.” : August 20. Library of Congress, n.d. Web. 09 Mar. 2016 <http://memory.loc.gov/ammem/today/aug20.html>.


Child Custody, Divorce, and Wage Income Between Sexes


There has been considerable debate concerning the economic impacts of divorce on the livelihoods of women. Duncan and Hoffman (1988) state that, “the economic status of women fell an average of about 30 percent in the first year after divorce”(Duncan and Hoffman 1988, 641). However, others note that a negatively causal relationship between divorce and the economic well being of women is difficult to prove, as women who experience divorce often differ in ways that are difficult to measure (Holden and Smock 1991; Smock, Manning, and Gupta 1999). Important to note is the well-documented increase in the labor force participation of mothers throughout the 20th century (Shapiro and Mott, 1978; Shapiro and Mott, 1994). Considering this trend, this post will address the effects of divorce and childbirth on the economic well being of a subset of the population: those in the labor force. By examining the relationship between marital status, number of children, and wage income, I hope to achieve a greater understanding of the impact that divorce has on the wage incomes of this smaller subset.

I collected my data for this analysis from the Integrated Public Use Microdata Series (IPUMS). I used sample data from 1940 to 2010. I used 1% samples for 1940 to 1960, 1990, and 2000. I used the 1% State Form 1 sample for 1970; the 1% Metro sample for 1980, and the 1% American Community Study(ACS) for 2010. The US Census Bureau administers the ACS. I used the variables MARST, NCHILD, EMPSTAT, and INCWAGE for my analysis. The MARST variable gives describes each respondent’s marital status at the time of the census. The possible outcomes for the MARST variable are ‘Married, spouse present’, ‘Married, spouse absent’, ‘Separated’, ‘Divorced’, ‘Widowed’, and ‘Never Married/Single’. The NCHILD variable describes the number of own children living with each individual, including biological children, stepchildren, and adopted children. The EMPSTAT variable indicates whether the respondent is employed, unemployed, or not in the labor force. The INCWAGE variable indicates the total pre-tax wage and salary income for each individual for the previous year; for the census, this was the previous calendar year, for the ACS, this included the previous 12 months.

All of the following analyses are weighted by the PERWT variable for all years excluding 1950, and weighted by the SLWT variable for 1950.



I began my analysis by excluding respondents who were reported as ‘Never Married/Single’ (MARST==6). Next, in order to simplify my definition of married within the data, I decided to combine respondents who were recorded as ‘Married, spouse absent’ (MARST==2) with those who were recorded as ‘Married, spouse present’ (MARST==1) to make a single category labeled ‘Married’. I decided to limit my analysis to respondents with at least 1 child (NCHILD>=1), and top coded the number of children to 6. I created categories for the NCHILD variable simply labeled ‘1’, ‘2’, ‘3’, ‘4’, ‘5’, ‘6+’. I limited my analysis to individuals reported as employed in the labor force (EMPSTAT == 1). Finally I adjusted the wage incomes reported by the INCWAGE variable for inflation by using the Consumer Price Index (CPI) to find the multiple for each sample year and converting all dollars to the 2010 level. After manipulating the data in these ways, I created a plot comparing the inflation adjusted wage incomes across the relevant marital statuses and sex; a plot comparing the number of children in custody across sex and between divorced and married respondents; and a plot comparing the inflation adjusted wage incomes across sex and between divorced and married respondents. My R code for the analysis can be found here.



Figure 1:

           Figure 1 shows the median inflation adjusted wage incomes for fathers and mothers across the examined marital statuses. First, it is important to recognize the far gap between the median inflation adjusted incomes for fathers and mothers, representing the consistent gender pay gap present in the United States throughout history. Yet, there is a gradual lessening of the gap as time approaches the present day. Next, seen here is a steep increase in inflation adjusted wage incomes across all cohorts between 1940 and 1970. For fathers, this trend is followed by a drastic plateauing across all marital statuses. For mothers, the subsequent trend differs across the various marital statuses. For divorced mothers, there is a slight decline in the median wage income between 1970 and 1980, followed by a tapering rate of increase between 1980 and 2010. For married mothers, there is a similar slight decrease in median wage incomes between 1970 and 1980, followed by another period of high rates of increase between 1980 and 2010. Widowed mothers in the workforce also experience a slight decrease in median wage income between 1970 and 1980, followed by a lackluster increase between 1980 and 2010. Separated mothers experience a consistent increase in median wage income until 2000, followed by a decrease in 2010.

Possibly the most interesting trend examined in this plot is the difference in differences between divorced and married fathers and mothers. Across all years examined, married fathers have had consistently higher inflation adjusted wage incomes than divorced fathers. However, divorced mothers have experienced notable higher inflation adjusted wage incomes than married mothers, until only recently. In 2010, we see a convergence of the median wage incomes for both married and divorced mothers for the first time, suggesting a possible current reversal in this consistent historical trend.

Figure 2:


Figure 3:


           Figure 2 shows the distributions of number of children across divorced and married mothers and fathers. Seen here are similar trends in the number of children for married men and women as well as divorced men and women. Seen during the baby boom is a general increase in number of children across all cohorts, followed by a steady decrease between 1970 and 2010.

Figure 3 shows a magnified plot of the percentage of divorced and married mothers and fathers with 4, 5, and 6+ children. Again, similar trends are examined across all cohorts. However, it seems there may be a coding error in the percentage of divorced fathers with 5 children in 1950. Furthermore, divorced fathers are examined to have custody over a slightly higher percentage of the high-numbers groups of children.


Figure 4:


            Figure 4 shows the median inflation adjusted wage incomes for divorced and married mothers and fathers across cohorts of number of custodial children. Seen here is a sustained consistency in the difference in median wage incomes between divorced and married mothers and fathers, despite the increase in number of custodial children. There is a perceived gradual increase in the difference in median wage income for divorced men as the number of children increases, until the ‘6+’ category; the high variance in this category may point to a low number of male respondents in the samples who are divorced and have 6 or greater children in their custody.

Furthermore, there seems to be greater downward pressure on the median income of divorced fathers than divorced mothers as the number of children increases. This downward pressure is seen gradually as the number of children increases, but most drastically when the number of children in custody reaches 5. Although there may be slight downward pressure on difference in median wage income between divorced and married women, the magnitude seems to be less so than that for fathers.


The median wage incomes of divorced fathers with child custody are effected greater as the number of children in custody increases than the median wage incomes for divorced mothers with child custody. Furthermore, median inflation adjusted incomes for divorced mothers have been consistently greater than the median inflation adjusted incomes for married mothers. This trend is contrary to some of the literature, suggesting that, by the metric of wage income, divorced mothers may have an advantage over married mothers. However, this does not disprove a decrease in economic status post divorce (Duncan and Hoffman, 1988), as the median wage incomes for fathers are still far higher than those for mothers. Finally, there has been a recent convergence in the median inflation adjusted incomes for divorced and married mothers across all cohorts, suggesting a reversal of the consistent trend.


Works Cited:

Shapiro, David, and Frank L. Mott. “Long-Term Employment and Earnings of Womenin Relation to Employment Behavior Surrounding the First Birth.” The Journal of Human Resources 29.2 (1994): 248. Web.

Shapiro, David, and Frank L. Mott. “Labor Supply Behavior of Prospective andNew Mothers.” Demography 16.2 (1979): 199. Web.

Tienda, Marta, and Jennifer Glass. “Household Structure and Labor ForceParticipation of Black, Hispanic, and White Mothers.” Demography 22.3(1985): 381. Web.

Hoffman, Saul D., and Greg J. Duncan. “What Are the Economic Consequences of Divorce?” Demography 25.4 (1988): 641. Web.

Bedard, Kelly, and Olivier Deschênes. “Sex Preferences, Marital Dissolution, and the Economic Status of Women.” Journal of Human Resources J. Human Resources XL.2 (2005): 411-34. Web.

Female Immigrant Labor Force of the 20th century

This post looks into the female immigrant population during the 20th century in America, specifically their presence in the labor force.  I found in a previous post that the female immigrant population is not as affected by the Immigration Act of 1924 and becomes larger than the male immigrant population by 1960.  Female participation in the labor force has increased over the course of the 20th century for multiple reasons, narrowing the gender gap in the labor force.  In general, education levels, family income, and the presence and age of children have factored into female participation in the labor force (Cotter, 2001, 431).  Furthermore, the rising wage rates from the 20th century causes men to bring more leisure into their lives from the increased income and became an incentive for women to work more (Cotter, 2001, 432).  While there was an increase in the female labor force during this time, there was also a high level of scrutiny regarding immigration.  A large portion of this scrutiny centered on Americans feeling threatened by immigrants taking their jobs (Jacobson, 2008, 69).  This post looks into the participation of female immigrants in the labor force from 1900-1960 in order to relate the rise in female labor participation and the levels of female immigration during this time.


I gathered data from the Integrated Public-Use Microdata Series (IPUMS) for this project.  I used 1% data samples from the 1900-1960 censuses and took the BPL (Birthplace), SEX, PERWT, AGE and OCC1950 variables from these samples.  The PERWT variable represents the sample weight of each individual.  The SEX variable gives the gender of the individual.  The BPL variable indicates the birthplace of the individual.  I grouped the BPL variable into these categories: U.S. Born, Other NA, Central and South America, Northern EU, UK & Ireland, Western EU, Central/Eastern EU, East Asia, Other Asia, and Other.  I grouped the OCC1950 variable into these separate categories created by IPUMS: Professional, Farmers/Farm Labor, Managers/Proprietors, Clerical, Sales, Crafts, Operatives/Laborers, Private Service, Public Service and non-occupational.


I created multiple graphs in order to research this topic.  To narrow down the population to the working immigrant population I only included age cohorts of ten years starting at 20 years of age and ending at 70 years of age.  I took out the non-occupational category of OCCUPATION in order to select only the working population of female immigrants as well. I also made separate data sets for each census year to better analyze the female immigrant labor force.  I made multiple bar graphs that chart OCCUPATION on the x-axis and the number of people in each occupation on the y-axis.  The BIRTHPLACE variable is used to fill the bars in the graph, and it is faceted by the five created age cohorts and each individual census year.  The code for this project can be found here.


Figure 1:


Figure 1 illustrates the female immigrant labor force in 1900.  Most female immigrants work in either private service or the operatives/laborers categories with a more equal distribution over the other occupation groups.  Initially, private service is almost double the next category of operatives/laborers in the 20-29 years old age cohort, but this difference becomes almost nothing in all the other age cohorts.  Both the private service and operatives/laborers groups are significantly lower following the 20-29 age cohort, indicating both a lower participation from older immigrant women in the labor force.  The largest immigrant groups in the labor force in this graph are the UK & Ireland, and Central/Eastern and Northern Europe.

Figure 2:


Figure 2 displays the female immigrant labor force in 1910.  In this graph all occupational groups except for farmers/farm labor have increased, especially the operatives/laborers category in the 20-29 age cohort.  The operatives/laborers group, contrary to figure 1, is slightly greater than the private service group.  A large influx of European immigrants, especially from Central/Eastern Europe, boosted the operatives/laborers category and marginally increased the private service profession.  Like figure 1, there is a steep decrease in the labor force from the older cohorts, but they also show significantly increased participation in all categories.  This reflects the increased immigrant population as well as a higher participation of immigrants in the labor force.

Figure 3:


Figure 3 shows the female immigrant labor force in 1920.  There is a large decrease in both the private and public service categories in the 1920 census in the 20-29 age cohort with the most significant increase appearing in the clerical profession.  While the operatives/laborers category remains constant in the 20-29 age cohort, it has a large growth in the older age cohorts compared to the 1910 census.  The immigrant groups that most affected this increase in the operatives/laborers category were the Central/Eastern and Western European groups.  Overall, the total female immigrant population in the labor force increased from the 1910 census.

Figure 4:


Figure 4 illustrates the female immigrant labor force in 1930.  It is important to notice that the y-axis on this graph has a smaller scale than that of the first three figures.  This smaller scale can be attributed to decreased number of women in the 20-29 age cohort in the operatives/laborers profession therefore lowering the highest number need on the y-axis. Even with this change, there is still an increase in both the private service and clerical categories.  There is a steep drop in the clerical profession from the 20-29 age cohort to the 30-39 cohort.  In this figure, there is clear downward linear progression in most of the profession categories of participation going from the youngest to oldest age cohorts.  Like the previous figures, the most impactful immigrant groups on these categories are Central/Eastern and Western Europe.

Figure 5:


Figure 5 shows the female immigrant labor force in 1940.  This graph illustrates a completely different image of the female immigrant labor force than the four preceding figures.  Overall the number of female immigrants in the labor force has decreased from the 1930 census and the age cohort distribution has changed drastically as well.  Here the 40-49 cohort reports the most workers, with operatives/laborers holding more than double any other profession.  The 20-29 age cohort surprisingly has only the third most numbers, indicating either a lack of participation from the cohort or a decrease in the population of that cohort.  The Central/Eastern and Western European immigrant groups still maintain a majority on the total female immigrant population in the labor force with this decrease in population.

Figure 6:


Figure 6 displays the female immigrant labor force in 1950.  This graph continues to show that the working population of female immigrants in the U.S. ages and significantly decreases up to this point in the 20th century.  The 40-49 and 50-59 age cohorts illustrate more or less the same population and spread of professions as they did when they were the 20-29 cohorts in 1930 and 1920.  Those cohorts have mainly remained with the same profession for 20-30 years, with only the private service category declining.  The clerical profession sees the most participation out of the younger age cohorts in this figure.  Still, the Central/Eastern and Western European immigrant groups still dominate the 40-49 and 50-59 cohorts as well.  The 20-29 and 30-39 age cohorts in this figure have a significantly lower population and also have much more equal representation from all of the immigrant groups.

Figure 7:


Figure 7 represents the female immigrant labor force in 1960.  This graph depicts a large boom in the female immigrant labor force population.  The largest cohort is the 50-59 age group, but the levels of participation in other age cohorts and professions are much greater than in 1950.  The y-axis numbers returned to the same levels as the first three figures and there is also greater participation across all cohorts.  The clerical and operatives/laborers categories dominate the younger age cohorts whereas the operatives/laborers group increases markedly in the 50-59 cohort.  The public service profession also sees gradual increases each year compared to the clerical category which sharply declines after the 30-39 cohort and again after the 50-59 cohort.  The distribution of immigrant groups remains the same for the older cohorts, dominated by Central/Eastern and Western Europeans, but the younger cohorts become more balanced  and even show a larger percentage of immigrants from other parts of North America and Central America compared to European immigrants.


These graphs illustrate both changes in levels of immigration in the United States during the early to mid 20th century as well as professional biases that U.S. culture has towards female immigrants.  The female immigrant population most gravitated or more likely was pushed towards the profession categories of operatives/laborers, private service, public service, and clerical work.  These jobs, being a housemaid or secretary for example, are not as respected in American society but still essential to the economy.  These categories represent many of the positions that perform the grunt work of society.  Initially I am sure that the female immigrant population was pushed towards these jobs, but later on found comfort and unity in creating their own communities around these mostly service based industries based on the age cohorts that fell under the 20-29 groups in 1920 and 1930 having similar populations and distributions as they aged each decade.  Furthermore, these graphs clearly lay out the effects of the Immigration Act of 1924 on emigration to the U.S.  The drastic fall of the 20-29 age cohort of female immigrants in the labor force in 1940 and 1950 is a direct result of the restrictions placed on immigration by the Act.  By 1960, the flow of immigrants coming into the U.S. seems to have normalized to somewhere between the high of the early 20th century and the immediate decade after the Immigration Act of 1924.  With regards to the female labor force, the restrictions placed on certain countries, mainly in Europe, seem to have allowed larger populations from other countries like Canada and Mexico to become more of a presence in America and its labor force.

Works Cited

Cotter, David A., Joan M. Hermsen, and Reeve Vanneman. “Women’s Work and Working Women: The Demand for Female Labor”. Gender and Society 15.3 (2001): 429–452. Web..

Jacobson, Matthew Frye. Barbarian Virtues: The United States Encounters Foreign Peoples at Home and Abroad, 1876-1917. New York: Hill and Wang, 2000. Print.

Final Portfolio Overview

I utilize the University of Minnesota’s Integrated Public Microdata Series (IPUMS) census data for all my projects in History 90.01: U.S. History through Census Data (16W). IPUMS census data provides me, and researchers, both snapshot data and longitudinal data on American individuals (since 1850) and households. All my projects use IPUMS data to analyze important trends in American history. I also take advantage of other primary and secondary sources to explain my results. All my projects use R to visualize IPUMS data in the form of charts and graphs. IPUMS data allows researchers to analyze history at a more top-down, macro level.

My projects explore American history through census data. Native Americans and the Census studies how and why Native Americans were enumerated in the census from 1850 to 1950. It also looks at census enumerator instructions and the concepts of race in America. Japanese American Eastward Migration maps Japanese American immigration patterns from the West coast to East coast, Midwest and South. It also pays special attention to the internment and its effects on migration. Vietnam Veterans Wages and Education focuses on comparing income over time and education attainment of veterans and non-veterans by birth year cohorts. Finally, Divorce Rates among Women by Race and Income analyzes divorce rate trends from 1940 to 2010 by income and race and offers explanations on why divorce has rapidly risen.