Reengineered In-Office Address Canvassing Methods to Increase Efficiency and Reduce Costs of 2020 Census

Bookmark and Share

Written by: Shonin Anacker, Address and Spatial Analysis Branch, Geography Division

In-office address canvassing allows the Census Bureau to update its address list to ensure a complete and accurate census. These assessments will increase the efficiency of the census and reduce the expensive in-field workload.

The U.S. Census Bureau updates its address list before each census to ensure that each person in the United States and is correctly counted where they live on Census Day.   For the 2010 Census, field workers traveled to roads and places where people were thought to live across the U.S. in order to update the Master Address List in an operation called address canvassing.

Following the 2010 Census, research indicated that the housing inventory, and therefore addresses, in many neighborhoods did not change from one census to the next, and that physically walking each area to make updates to the address list is not necessary. Research also indicated that the Census Bureau can reliably detect and classify such stable areas. Therefore, for the 2020 Census, the Census Bureau is taking a different approach, one that would target fieldwork only where it is needed. The use of reengineered in-office and in-field canvassing methods is estimated to save the American taxpayer approximately $900 million in costs during the 2020 Census.

The Census Bureau has divided the nation into more than 11 million small geographic areas called “blocks.” In-office address canvassing will assess and classify each block to determine whether in-field canvassing is necessary or whether an in-office review can produce a complete and accurate address list. The first major step of the operation is the “interactive review.” This began in the fall of 2015 and, to date, reviewers have assessed and verified more than 1.7 million blocks (about 15 percent of the national total).

The interactive review includes three major steps that together take an average of 80 seconds per block to complete. In the first step, reviewers compare satellite imagery from the 2009 timeframe (when 2010 Census address canvassing occurred) and the current timeframe to detect changes in the number of residential buildings. In the second step, reviewers assess whether the current census address list accurately reflects the reality on the ground based on the number and type of structures visible in the imagery. In the final step, reviewers assess the structures in the imagery and the property lines (parcels) to determine whether the block is “built-out” or has additional room to grow.

Address Canvassing Fig 1

Taken together the three steps provide a rich and detailed profile for each block, indicating what housing changes have occurred since the last census, how well the Census Bureau’s address list is keeping up with the changes and how likely changes are to occur in the future. Blocks with detected changes or deficiencies will move on to a second process in which a staff member will attempt to update the address list in the office. If the second step is not able to resolve the issue, the block will go to the field for canvassing.

To date, the interactive review has determined that about 69 percent of housing units are located within passive blocks that do not require action, approximately 21 percent of housing units are located in active blocks that do require action, and about 10 percent of housing units are located within blocks that are on hold.

Additionally, reviewers have classified about 38 percent of the blocks reviewed (containing approximately 53 percent of the housing units in the reviewed blocks) as being not only stable and correct but also “built-out” (completely covered by buildings or undevelopable land like parks) and, thus, relatively unlikely to change in the near future. These figures, while preliminary in nature, support the initial research findings that the majority of blocks in the nation do not require costly in-field address updating.

With the interactive review off to a successful start, we have embarked on two other major research projects critical to the overall success of in-office address canvassing. We must develop methods of assessing and capturing the housing change that we cannot identify via imagery analysis, including apartment splits and redevelopment within existing buildings. And to make sure that we do not miss changes that occur later in the decade, we must also test and implement a set of “triggers” that will flag blocks for re-review.

In-office address canvassing allows the Census Bureau to update its address list to ensure a complete and accurate census. These assessments will increase the efficiency of the census and reduce the expensive in-field workload.

Address Canvassing Fig 2

To view the 2020 Census Operational Plan, click here.

To view the 2020 Census Detailed Operational Plan for the Address Canvassing Operation, click here.

Posted in Uncategorized | Leave a comment

Understanding Census Bureau Address Ranges

Bookmark and Share

Written by: Michael Orris, Geographer

The Master Address File/Topologically Integrated Geographic Encoding and Referencing (MAF/TIGER) system is a national geospatial database of address ranges, linear features, addresses, address location points, geographic areas and boundaries.

Address ranges describe a label given to a unique collection of addresses that fall along a road or path. Address ranges provide a way of locating homes and businesses based on their street addresses when no other location information is available.

Using a house number, street name, street side and ZIP code, address ranges can locate the address to the geographic area associated to that side of the street. Once geocoded, the U.S. Census Bureau can assign the address to a field assignment area or tabulate the data for that address. In addition, academics, researchers, professionals and government agencies outside of the Census Bureau use MAF/TIGER address ranges to transform tabular addresses into geographical datasets for decision-making and analytical purposes.

Address ranges must be unique to geocode addresses to the correct location and avoid geocoding conflicts. Multiple elements in MAF/TIGER are required to make an address range unique including street names, address house numbers and street feature geometries, such as street centerlines. The address range data model is designed to maximize geocoding matches with their correct geographic areas in MAF/TIGER by allowing an unlimited number of address range-to-street feature relationships.

The Census Bureau’s Geography Division devises numerous operations and processes to build and maintain high quality address ranges so that:

  • Address ranges accurately describe the location of addresses on the ground.
  • Address All possible city-style addresses are geocoded.
  • Address ranges can handle all known address and street name variations.
  • Address ranges conform with current U.S. Postal Service ZIP codes.
  • Address ranges are reliable and free from conflicts.

Automated software continually updates existing address ranges, builds new address ranges and corrects errors. An automated operation links address location points and tabular address information to street feature edges with matching street names in the same block to build and modify address ranges.

The following graphic displays how the automated operation associates address house numbers (103 Main St) and address location points (blue dots) to a street feature (Main Street) and builds new address ranges (bold red text) that contain a group of the address house numbers.

Orris figure

Building Address Ranges

Many business rules and legal value checks ensure quality address range data in MAF/TIGER. For example, business rules prevent adding or modifying address ranges that overlap another house number range with the same street name and ZIP code. Legal value checks verify that address ranges include mandatory attribute information, valid data types and valid character values.

Some of the TIGER/Line products for the public include address ranges and give the public the ability to geocode addresses to MAF/TIGER address ranges for the user’s own purpose. The address range files are available for the nation, Puerto Rico and the U.S. Island Areas at the county level. TIGER/Line files require geographic information system (GIS) software to use.

The Census Bureau Geocoder Service is a web service provided to the public. The service accepts up to 1,000 input addresses and, based on Census address ranges, returns the interpolated geocoded location and census geographies. Users can access the service a web interface or a representational state transfer (REST) application program interface (API) web service.

Posted in Uncategorized | Leave a comment

Evaluating 2013-2014 Trends in County-Level Health Insurance Coverage for Low-Income Working-Age Adults

Bookmark and Share

Written by: Lauren Bowers and David Powers, Statisticians, Social, Economic and Housing Statistics Division

In 2014, many major provisions of the Patient Protection and Affordable Care Act went into effect. One key provision of the Affordable Care Act was the option for states to expand their Medicaid eligibility to most working-age adults living at or below 138 percent of poverty. During 2014, 26 states and Washington, D.C., chose to do so (See Figure 1).

Bowers FIg 1

Evaluating 2012-2014 Trends in Health Insurance Coverage for All U.S. Counties asks how health insurance coverage changed at the county level for low-income, working-age adults from 2013 to 2014. Did states that expanded their Medicaid eligibility criteria have more counties with decreasing uninsured rates?

Using Small Area Health Insurance Estimates (SAHIE) from the U.S. Census Bureau, we find that from 2013 to 2014, almost 60 percent of U.S. counties (1,877) had a decrease in their estimated uninsured rate for low-income, working-age adults (See Figure 2). In states that expanded their Medicaid eligibility, 96 percent of counties had a decrease in the SAHIE uninsured rate compared with 38 percent of counties in states that did not expand.

Bowers Fig 2

Large decreases in the estimated uninsured rate, nine percentage points or more, for low-income working-age adults, occurred in 51 percent of counties in expansion states compared with 4 percent of counties in nonexpansion states (See Figure 2, dark blue counties). Further, these declines in the uninsured rate for low-income, working-age adults were more pronounced among states that chose to expand their Medicaid eligibility than among states that chose not to expand eligibility.

The source data for our research are the published 2013 SAHIE and a preliminary version of 2014 SAHIE. The official 2014 SAHIE will not be published until late spring 2016. SAHIE publishes single-year health insurance coverage estimates for 3,141 U.S. counties annually by sex, with detail for five age groups and  six income groups. The income groups include 0-138 percent of poverty, making SAHIE a vital source for evaluating county-level health insurance coverage for low-income individuals.

Although the 2013-2014 declines we observe in the uninsured rate among counties correlate with the implementation of state Medicaid expansions, we have not studied any policy causation, and so we cannot attribute changes in the uninsured rates to the implementation of the Affordable Care Act. However, since the SAHIE data uniquely capture uninsured rates at the county level for single years, they are an important resource for evaluating changes in health insurance coverage by detailed characteristics from 2013-2014 and in future years.

Also see this Research Matters blog: Health Insurance Disparities are Closing Among Working-Age Adults.



Posted in Uncategorized | 1 Comment

Poverty by Age and Sex: An Examination of the Distribution in Poverty Between 1966 and 2014

Bookmark and Share

Written by: Kayla R. Fontenot, Bernadette D. Proctor, and Trudi J. Renwick, Social Economic and Housing Statistics Division

Researchers, analysts and policymakers have long noted that women are more likely to fall below the poverty line than men. However, the poverty differential between men and women is not consistent throughout life.

In this project, Poverty by Age and Sex: An Examination of the Distribution in Poverty Between 1966 and 2014, we use data from the U.S. Census Bureau’s Current Population Survey Annual Social and Economic Supplement (CPS ASEC) and the American Community Survey (ACS) to examine the gender differences in poverty rates for all ages and to explore how those differences have changed over time.

One way to measure the difference in poverty rates between men and women is to calculate a female/male poverty ratio. If the poverty rates for men and women are equal, the ratio will have a value of one. If the ratio is greater than one, the poverty rate for women is higher than the poverty rate for men. For example, in 2014 the poverty rate for women 95 years of age and older was 14.25 percent while the poverty rate for men in this age category was 10.46 percent. The female/male poverty ratio, therefore, was 14.25 divided by 10.46 or 1.36 (see Figure 1).

Results from the 2014 ACS show that for the majority of age categories, females had higher poverty rates than males. For example, at each single year of age from 16 to 94 women experienced higher levels of impoverishment than their male counterparts (see Figure 1).

Renwick Fig 1

Data from the CPS ASEC show that individuals aged 75 and over not only had the largest differences in poverty rates between men and women in 2014, but the differential for this group was up from 1966. At the same time, the gap for individuals aged 45 to 54 and 55 to 64 was down from 1966 (see Figure 2).

Renwick Fig 2

In general, the results of this project showed that women continue to face a disproportionately high risk of living in poverty compared with men. In particular, women age 75 and older and age 25 to 34 remain considerably worse off than their male counterparts.

Posted in Uncategorized | Leave a comment

Separate but Unequal: The Nature of Income Inequality in U.S. Metropolitan Statistical Areas

Bookmark and Share

Written by: Brian Glassman, Social, Economic and Housing Statistics Division

Income inequality as a concept is straightforward. Are there many haves and have-nots or do most people earn close to the same amount of money? However, the measurement of income inequality is not so straightforward.

There are a number of different income inequality metrics available and the choice of a particular metric alters the conclusions about the level of income inequality and the relationship between income inequality and economic well-being. In this poster, Separate but Unequal: The nature of income inequality in U.S. Metropolitan Statistical Areas, eight inequality metrics are calculated and compared with one another in order to examine how the use of a particular metric affects the understanding of economic well-being in 381 U.S. metropolitan statistical areas (MSAs).

We used data from the U.S. Census Bureau’s 2014 American Community Survey to calculate seven after-tax household income inequality metrics. Household after-tax income includes wages and salary income, self-employment income, retirement income, interest and dividends, as well as transfer payments (Supplemental Security Income, Social Security and public assistance) and subtracts federal and state taxes using the National Bureau of Economic Research’s TAXSIM program.

We adjust the income variable by the size of the household using a three-parameter equivalence scale and then adjust by MSA cost of living using factors developed by the Bureau of Economic Analysis (BEA).  This cost of living adjustment is only necessary when looking at overall income inequality in the United States and not by individual MSAs.

The eight inequality metrics are:

  • Gini index — A statistical measure of income inequality ranging from zero, perfect equality, to one, perfect inequality.
  • Palma ratio — Ratio of top 10 percent share of total income to lowest 40 percent share of total income.
  • Household income ratios of selected percentiles:
    • 80-20 ratio — 80th percentile income limit divided by 20th percentile income limit
    • 90-10 ratio — 90th percentile income limit divided by 10th percentile income limit
    • 90-50 ratio — 90th percentile income limit divided by 50th percentile income limit
    • 50-10 ratio — 50th percentile income limit divided by 10th percentile income limit
    • 99-90 ratio — 99th percentile income limit divided by 90th percentile income limit
  • The eighth metric measures educational inequality, instead of income inequality, using a Gini index based on educational attainment of heads of households who are at least 25 years of age.

Figure 1 and Table 1 present a comparison of two metro areas in order to illustrate the complexities of measuring inequality. The Gini indexes are not statistically different from one another for Washington, D.C. and Cincinnati, Ohio, while all the other metrics do show statistically significant differences.  Furthermore, different metrics lead to different conclusions about which MSA has higher income inequality.

Glassman Fig 1

Glassman Table 1

This research allows us to examine the relationship between income inequality and measures of economic well-being across the 381 MSAs. Figure 2 shows the correlation between three income inequality measures and the poverty rate and the unemployment rate across the MSAs. Using the Gini index would lead one to conclude that MSAs with higher income inequality have higher poverty and unemployment. However, income inequality measures focused on just specific parts of the income distribution may lead to different conclusions. For example, the 90-10 ratio also shows correlation between inequality (across the majority of the distribution) and both poverty and unemployment, while a measure focused on the top of the income distribution, the 99-90 ratio, does not.

Glassman Fig 2

The main take away from this research is that the choice of and use of a particular income inequality metric has important consequences for the analysis of the level of income inequality and the connection between income inequality and economic well-being among metro areas.

Posted in Uncategorized | 1 Comment

Health Insurance Disparities are Closing Among Working-Age Adults

Bookmark and Share

Written by: Carla Medalia, Marina Vornovitsky and Jennifer Cheeseman Day, U.S. Census Bureau

Note: This blog is based on ongoing research.

Between 2013 and 2014, the share of people without health insurance, or uninsured rate, decreased by 2.9 percentage points overall to 10.4 percent. This unprecedented decline in the uninsured rate occurred at the same time as many provisions of the Affordable Care Act went into effect. Many of these provisions targeted working-age adults, who experienced the largest changes in health insurance coverage. Two presentations at the Population Association of America meetings this week take a closer look at changes experienced by the working-age population.

Health Insurance Disparities and the Affordable Care Act: How Did Inequality Decline? finds that as the uninsured rate fell, most disparities in health insurance coverage within the working-age adult population also decreased. Notably, the gap in the uninsured rate between those at the lowest and highest ends of the income spectrum narrowed considerably.

On Jan. 1, 2014, 24 states and the District of Columbia expanded Medicaid eligibility to low-income working-age adults, while 26 states did not. Did adults in expansion states experience a greater increase in health insurance coverage than in non-expansion states?

This working paper finds that while decreases in the uninsured rate among working-age adults in Medicaid expansion states were sometimes greater than in non-expansion states, most disparities in health insurance coverage declined in parallel between expansion and non-expansion states. Decreases in the uninsured rate among working-age adults were primarily due to increases in Medicaid and direct-purchase coverage in expansion states; in non-expansion states, increases in direct-purchase coverage accounted for almost all of the total change in the uninsured rate, while the increase in employment-based health insurance coverage mattered for key groups, such as adults in poverty (See figure 1).

Medalia Fig 1

How Did the Affordable Care Act Affect Workers in 2014? A Closer Look at Detailed Occupations and Health Insurance Coverage investigates the disparity of health insurance coverage among workers in different occupations. This analysis finds that most occupations showed a significant decline in the percentage uninsured between 2013 and 2014, with no occupation experiencing an increase in the uninsured rate. In general, occupations with higher uninsured rates in 2013 experienced larger improvements (See figure 2).

Medalia Fig 2

Within occupations, the changes were most notable for people who worked less than full time, year round. Across occupations, those with lower median earnings experienced more change than occupations with higher median earnings. Findings were mixed for employment-based coverage: while some occupations showed a decline in the rate of employer-provided health insurance, many others showed a significant increase in this kind of insurance.

To learn more about these analyses, visit here.

Also see this Research Matters blog on Evaluating 2013-2014 Trends in County-Level Health Insurance Coverage for Low-Income Working-Age Adults.


Posted in Uncategorized | Leave a comment

How Does Growing Up in Assisted Housing Affect Adult Earnings?

Bookmark and Share

Written by: Mark J. Kutzbach, Center for Economic Studies, U.S. Census Bureau

In 2000, nearly 3 million children under age 18 lived in voucher-supported or public housing sponsored by the U.S. Department of Housing and Urban Development (HUD). Although assisted housing programs have been in place for some time, research on the long-term effects on resident children is scarce and hampered by methodological limitations. To shed light on this topic, my colleagues and I combined Census Bureau data with administrative data to track children through assisted housing and into the labor force as adults.

The Census Bureau often combines survey and administrative datasets to produce new statistics, but these data can also help answer complex research questions. For this project, we identify families with multiple teenage children counted in the 2000 Census and link them to HUD administrative records to observe how they move into and out of assisted housing between 1997 and 2005. We then match the children to their adult earnings from 2011 to 2013 using data from the Longitudinal Employer-Household Dynamics program.

To identify the impact of assisted housing on earnings, we compare adult earnings between siblings who experienced different amounts of assisted housing as teenagers. As siblings share many of the background characteristics that affect adult earnings — for example, household poverty and parental motivation — we are able to distinguish the effect of assisted housing participation from the effect of any other shared childhood experiences.

We also look at whether public and voucher housing might have different effects on boys and girls. In public housing, a household lives in a project run by the local housing authorities, whereas in voucher-based housing, the housing authority pays a large portion of a household’s rent and utilities in private housing chosen by the household.

We first observe that children growing up in assisted housing tend to have lower adult earnings compared with other children, even those from similarly low-earning households. However, observing a difference in adult earnings between children who participated in assisted housing and those who did not is not enough to conclude that the assisted housing participation caused the difference. For example, participating households are required to earn below specified thresholds in order to be eligible. Beneficiary children are therefore likely to come from impoverished backgrounds and — even in the absence of assisted housing participation — earn less as adults.

When we use only between-sibling differences, we find that assisted housing participation is associated with increases in adult earnings for girls and only modest, often statistically insignificant, decreases in earnings for boys. In other words, holding constant family characteristics, the negative effect of assisted housing disappears for most children.

The overall result that assisted housing raises earnings for girls more than boys might depend on the community under study. To shed additional light on this difference, we look at results separately for white, black and Hispanic households. The between-siblings effects are consistently positive only for black non-Hispanics, who represent roughly half of all HUD residents (and are mostly not distinguishable from zero effect for whites and Hispanics). The figure below shows both the between-siblings estimates and the naïve estimates, those that do not compare siblings, for black non-Hispanic households.

Kutzbach Fig

In the between-siblings model, girls in black non-Hispanic households earn 6.5 percent more for each year spent in voucher housing and 4.3 percent more for each year spent in public housing while a teenager. Boys in black non-Hispanic households earn 2.6 percent more for each year they spent in voucher housing and 3.8 percent more for each year spent in public housing. The difference in the effects of voucher housing for girls in black non-Hispanic households relative to boys is statistically significant.

How might housing assistance affect children? While housing assistance relieves families of a major expenditure, other studies have shown that it may also concentrate low-earning households together in large public housing buildings or low-income neighborhoods, exposing children to high-poverty settings. Thus, the net effects are not certain without an empirical analysis. When distinguishing between the effects of assisted housing programs and family characteristics, we found that more time spent in assisted housing participation for siblings led to increases in adult earnings, especially for black non-Hispanic girls.

We are doing more research to try to unravel what aspects of housing assistance might have the greatest effects and to explore other adult outcomes that may be influenced by housing. We are also trying to understand why we observe girls benefiting more than boys for any type of housing assistance.

For more information, see Childhood Housing and Adult Earnings: A Between-Siblings Analysis of Housing Vouchers and Public Housing, a joint paper written by Fredrik Andersson, Office of the Comptroller of the Currency; John C. Haltiwanger, University of Maryland and U.S. Census Bureau; Mark J. Kutzbach, U.S. Census Bureau; Giordano Palloni, International Food Policy Research Institute; Henry O. Pollakowski, Harvard University; and Daniel H. Weinberg, Virginia Tech.

Posted in Uncategorized | Leave a comment

Characteristics of Cohabiters and Their Generational Status

Bookmark and Share

Written by: Emily Schondelmyer, Fertility and Family Statistics Branch

Most adults in the United States have lived with a boyfriend, girlfriend or partner at some point in their lives. Specifically, we have seen an increase in the number of cohabiting couples over the last 20 years. However, much less is known about cohabiters who are either foreign-born or who have foreign-born parents.

I use the U.S. Census Bureau’s Current Population Survey data to examine how the place of birth of an individual and their parents are related to cohabitation and demographic differences of cohabiters by nativity. The Current Population Survey is a rich resource for data on the nativity of individuals and their parents since it includes not only data on nativity but also data on year of entry.

Here, generational status refers to three distinct groups shown in Table 1. A respondent who is foreign born with foreign-born parents is considered first generation. A native-born respondent with at least one foreign-born parent is second generation. Finally, a native-born respondent whose parents are also native-born is third generation.

Table 1

Emily Table 1

The demographics of cohabiters have been well established, but we see some interesting differences when we look at cohabiters by their nativity.

Figure 1

Emily Fig 1

Rates of cohabitation varied by generation (Figure 1). The third generation had the highest proportion of cohabiters (7.6 percent), followed by the second generation (6.2 percent), and ending with first generation having the lowest levels of cohabitation (5.7 percent).

When we look at characteristics like education, we see that nearly one-third of first-generation cohabiters have less than a high school degree, compared to just 9 percent of third-generation cohabiters (Figure 2).

Figure 2

Emily Fig 2

Additionally, most cohabiters have never been married. However, we see that second generation cohabiters had the highest proportion never married (72 percent of cohabiters, compared with 62 percent of the third generation, and 69 percent of the first generation) and the third generation had the greatest proportion of divorced individuals. Furthermore, we see that there are significant differences in marital status among those who have ever been married (Figure 3).

First generation cohabiters have a greater proportion in the categories of separated and spouse absent (4 percent and 6 percent, respectively), which could be related to immigrants leaving families behind while moving to the U.S. While third-generation cohabiters had the greatest proportion of divorced individuals (28.5 percent), there isn’t a significant difference in divorce between first- and second-generation cohabiters (19 percent and 21 percent, respectively).

 Figure 3

Emily Fig 3

As these differences show, it may be important to keep generational status in mind while exploring family and union formation of individuals. For more information on differences between cohabiters by their generational status, please see my poster on Nativity’s Influence on Cohabitation.


Posted in Uncategorized | 1 Comment

Over the Hill at 25? College Completion at Higher Ages

Bookmark and Share

Written by: Kurt Bauman, Education and Social Stratification Branch

For many years (since 1942 to be exact) when the U.S. Census Bureau has reported educational attainment, it has focused on the population age 25 and older. The reason for this practice was that, “The statistics made available today relate only to persons 25 years old and over, those who may generally be considered as having completed their formal education.”

Since then, few have questioned this practice and logic, and researchers have remained comfortable with the assumption that “most people get their schooling during their childhood, teens and 20s …”

Thinking about education as being “fixed” through most of adulthood makes things easier for researchers who want to examine trends in education. But in recent years, there have been hints that this traditional assumption is breaking down. To look more closely, I took advantage of the large collection of data from the Current Population Survey from 1967 to 2015. I was able to follow age groups across a large portion of their lives to see how education patterns evolved as they got older.

Figure 1 shows trends in the percentage of women who had a bachelor’s degree or higher for several birth cohorts (women who were born in the same group of years). Those born around 1935 are at the bottom, having lower education than women who were born later in the century. When the group was in their early 30s (around 1968), 9 percent of the group had a bachelor’s degree or higher. By 2010, when the group was around 75, 17 percent had a bachelor’s degree or higher.

Bauman Fig 1

A similar pattern can be observed for younger groups of women (born in later years). They have increasingly higher levels of college completion at age 25 and continue to gain as they move from 25 to older ages. As for men, the story is pretty much the same, although shifts from one cohort to the next are not as smooth as they are for women (not shown).

What can we make of the pattern of growth in college completion that takes place across the lives of a group of people? Are we seeing the result of lots of people who delayed college graduation until later in life? Are other forces at work?

Besides graduation, there are at least two other things that could be happening. First, immigration could be contributing to college growth at later ages. While many immigrants have low levels of education, there is a large subset of immigrants to the United States with bachelor’s and higher degrees.

Another factor affecting the proportion of a cohort having a bachelor’s degree is mortality. Research has strongly established a relationship between education and mortality such that people with bachelor’s degrees survive to greater ages, and thus make up a greater part of the cohort as it gets older.

With the Current Population Survey data, I was able to measure the level of education, the rate of graduation and the effect of immigration for a set of cohorts of men and women such as those shown in Figure 1. In order to measure mortality, I made use of a dataset produced by the Census Bureau called the National Longitudinal Mortality Study. Using demographic life-table methods, I calculated what we might expect for each of these cohorts given the forces we could measure. The question was: Would I be able to come close to the observed patterns using just these factors?

Figure 2 shows the observed level of college completion for women who were born from 1926 to 1930, which moves in a slightly jagged pattern from 9 percent at age 38 (the youngest age for which I have data) to 12 percent at age 75. The “model” line is set up to start at the same point, 9 percent at age 38. From there, I allow graduation, immigration and mortality to change the percentage of the cohort with a college education, with no constraints. As it turns out, however, the model line runs along a very similar path to that which is observed from people reporting their levels of education. Both lines show 12 percent of the cohort having a bachelor’s degree at age 75.

Bauman Fig 2

When I looked at other cohorts of men and women born from 1926 to 1970, I obtained similar results. Although I was unable to conduct statistical tests because of the complexity of some of the calculations, the end result is satisfying. What appears at first to be a strange pattern of college attainment growing throughout the lifespan is indeed the plausible result of basic demographic factors at work. That is, the growth of education in this country continues across the lifespan because of the willingness of Americans to continue their education in adulthood and because of the contribution of immigrants who come to this country with higher degrees. A reward of higher education is greater longevity, which results in a greater share of college-educated people at higher ages. Taken together, these forces have had an important influence on the growth of education in the past century.

Check out my paper on College Completion by Cohort, Age and Gender, 1967-2015.

Posted in Uncategorized | Leave a comment

Does Doubling Up Improve Family Well-Being?

Bookmark and Share

Written by: Laryssa Mykyta, U.S. Census Bureau, and Natasha Pilkauskas, University of Michigan

Studies have shown that “doubling up,” or sharing housing, increased during the last economic recession and is most common in low-income families. Yet, we do not really know whether doubling up helps struggling families make ends meet.

On the one hand, doubling up might bring more resources into a household, reducing hardship and making it easier to make ends meet. Conversely, doubling up might increase hardship if it stretches limited resources or increases household costs. Doubling up may also result in household crowding or a loss of privacy and can affect whether a household is eligible for public programs.

In our paper Household Composition and Family Well-Being: Exploring the Relationship Between Doubling Up and Hardship, we use data from the adult well-being topical modules of the 2008 Survey of Income and Program Participation to examine the relationship between doubling up and the different kinds of hardship, such as not having enough money to pay for common life expenses. We also look at whether changes in sharing a household were associated with changes in these hardships.

Consistent with prior research, we found that people living in doubled-up households were more disadvantaged and reported more hardship than those who did not double up. For example, people who doubled up were more likely to report that they had trouble making ends meet, paying their rent or mortgage, getting to the doctor and having enough to eat than those who did not double up.

Moreover, we found the association between doubling up and hardship differs by type of hardship. Doubling up was associated with reduced housing and utility hardship. For example, these sharing households had less trouble paying housing and utility costs in models controlling for other individual and household characteristics (see figure below). The negative coefficients -0.01 and -0.02 illustrate the reduced likelihood of not being able to afford housing or utility costs while 0.03 and 0.025 represent the increased likelihood of not being able to get to the doctor when needed or not having enough food. This suggests that doubling up may be a strategy to alleviate hardship, such as insufficient income to pay for housing costs alone.

Mykyta fig

However, those who did double up were more likely to report difficulty getting to the doctor when needed and having enough food to eat. Although doubling up might reduce housing-related expenses, it may conversely reduce eligibility for public programs like the Supplemental Nutrition Assistance Program or force families to share limited food with more people, therefore increasing hardship.


Posted in Uncategorized | Leave a comment