Comparison Guidance: Income and Poverty

Bookmark and Share

By Victoria Velkoff 

Last year, the U.S. Census Bureau implemented methodological changes to the 2014 Current Population Survey’s Annual Social and Economic Supplement (ASEC). Based on substantial research to improve the quality of the data we collect, we redesigned the way we asked the survey questions about income in 2014. Over the course of the past year, we evaluated the effects of the redesign and sought out experts to review and provide feedback on our efforts. Based on the results of our evaluation, we fully implemented the redesign in 2015.

Maintaining the time series means two estimates for 2013 income and poverty. 

In the 2014 ASEC, we introduced the redesigned income questions using a probability split panel design. Of the 98,000 addresses selected to participate in the 2014 ASEC, approximately 68,000 addresses received the traditional set of income questions. The remaining 30,000 addresses received the redesigned income questions.

The split design functions as a bridge for year-to-year comparisons of the data. Last year, we used the traditional income questions to look at changes between 2012 and 2013. This year, we will use the estimates from the sample eligible to receive the redesigned income questions to look at changes between 2013 and 2014.

Making time series comparisons across the 2013 bridge. 

For income, there were statistically significant differences for many key measures between the redesigned and traditional questions. For example, median household income calculated by using the redesigned questions was 3.2 percent higher than the median income found using the traditional questions. As a result, it is difficult to assess whether an apparent change in median household income relative to previous years was a “real” change in income or an artifact of changes to survey questions.

The earnings questions were not changed in the redesign and the difference in earnings across the two subsamples was not statistically significant. Therefore it is appropriate to compare earnings over the entire length of the time series.

On the other hand, there were major changes to the questions concerning retirement income, interest income, dividends, disability income and public assistance. Estimates for these income sources should be compared with caution to pre-2013 estimates.

For poverty, our research found that for the total population and for most major demographic subgroups, the differences in poverty rates across the two subsamples were not statistically significant. Therefore, it is possible, with caution, to make comparisons to earlier years. For specific demographic groups for whom the differences were statistically significant, such as blacks and children, comparisons may not be appropriate.

To assist users in gaging the impact the questionnaire changes had on income and poverty estimates, the new report, Income and Poverty in the United States: 2014, will include an appendix showing estimates for 2013 by selected demographic characteristics for both the traditional and redesigned questions. In addition, the full set of detailed tables showing estimates for 2013 using the redesigned questions are available on-line. When making historical income and poverty comparisons, users should consult these tables.

The Census Bureau is constantly researching new methods and working to improve its surveys and we have made methodological changes to the Current Population Survey in the past. We publish tables with income estimates back to 1947 and poverty estimates back to 1959 but include extensive footnotes with these tables noting many of these changes (see Figure 1). Data users should take into account all of these changes, including the 2014 redesign when making historical comparisons for income and poverty.

Figure 1 Examples of Footnotes Showing Historical ASEC Changes



Posted in Income, Poverty | Leave a comment

Health Insurance Coverage Measurement in Two Surveys

Bookmark and Share

Written by: Marina Vornovitsky

Next week, the U.S. Census Bureau is releasing two sources for health insurance statistics in the United States: the Annual Social and Economic Supplement to the Current Population Survey and the American Community Survey.

Many people ask us which estimate they should use. Well, it depends. The benefit of the Current Population Survey is the combination of detailed employment and detailed income information, along with the health insurance coverage statistics, which provides an excellent overall picture of the well-being of our nation. With the new Current Population Survey baseline in 2013, annual comparisons between 2013 and 2014 are rich with detail. For detailed analysis of subnational geographies, we recommend using the American Community Survey statistics because of its larger sample size and smaller sampling errors. Also, the American Community Survey can provide historic comparisons.

Both surveys have questions that attempt to measure the same phenomena. They obtain a person’s health insurance status by asking if they have insurance through a number of different sources, such as an employer, directly through an insurance company, Medicare, Medicaid, Veterans Administration and other public sector insurance, and the military. So, how are the two surveys different? They differ in the timing of data collection, the reference period, the time frame of the resulting health insurance coverage estimates, and, ultimately, in uses of the data.

In particular, the Current Population Survey has produced health insurance statistics every year since 1987, making it one of the most widely used sources of statistics on health insurance coverage in the United States. The survey provides information on health insurance status (insured or not insured) for the nation and by demographic groups. Detailed income data available from the Current Population Survey also make it possible to view changes in health insurance coverage in relation to changes in the overall economic well-being of the nation.

Last year, after more than a decade of research, we implemented redesigned health insurance coverage questions in the Current Population Survey to improve our measure of health insurance coverage. With this new strong baseline for calendar year 2013, we feel confident that we can measure the impact of the major provisions of the Affordable Care Act that went into effect in 2014, as well as any future changes in health insurance coverage.

Starting in 2008, the Census Bureau also began asking about health insurance coverage using the American Community Survey. With its much larger sample size, we can see health insurance statistics for subnational geographies, such as states, counties, metro areas, congressional districts and cities. This level of geographical detail is not available from any other survey that collects data on health insurance coverage.

The concept of “uninsured” is also slightly different between the two surveys. In the Current Population Survey Annual Social and Economic Supplement, which is conducted in February through April, respondents answer questions about whether they had health insurance coverage at any time in the previous calendar year. The survey, thus, measures if a person was insured on any day during the previous year. They are considered “uninsured” only if, for the entire year, they had no coverage under any type of health insurance.

In contrast, the American Community Survey is a rolling sample of households collected continuously all year long. We ask if a person is currently covered by any of the listed types of health insurance. So, the American Community Survey measures health insurance for the population based on whether people are insured at the point-in-time that they answered the survey during the year of collection.

There is also a variety of differences in the survey logistics. Census Bureau field representatives conduct the Current Population Survey by personal visit or telephone. For the American Community Survey, many respondents receive a paper form to complete and return in the mail or they can reply via the Internet. Because of space limits within a paper survey, the American Community Survey asks fewer and less detailed questions than the Current Population Survey. In addition, the American Community Survey asks about the insurance coverage of each household member specifically, while the Current Population Survey asks if anyone in the household is covered, and, if so, who that is.

With these variations and others, the two surveys produce consistent, though slightly different statistics on health insurance coverage. Estimates from these two surveys rose and fell in parallel between 2009 through 2012. This degree of consistency between the two surveys collected under such different conditions gives us confidence that these statistics are useful for those who need to understand the state of health insurance coverage in America.

Posted in Health Care | Leave a comment

How the Census Bureau Measures Income and Poverty

Bookmark and Share

Written by: Victoria Velkoff

Income, poverty and health insurance statistics for 2014 from the Current Population Survey Annual Social and Economic Supplement (CPS ASEC) will be released Wednesday, Sept. 16, 2015. One-year statistics from the 2014 American Community Survey (ACS) will be released on Thursday, Sept. 17, 2015.

In all likelihood, the national statistics from these two sources will not be identical. Why not? Which is correct? Well, it’s complicated.

There are several reasons why the statistics from the two surveys differ. One of the most notable ways is that one asks respondents about income in the previous calendar year while the other asks respondents about income in a rolling 12-month period throughout the year.

The Current Population Survey is conducted every month and serves as the nation’s primary source of statistics on labor force characteristics. Supplements are added in most months; the Annual Social and Economic Supplement to the survey provides the official annual statistics on the nation’s poverty levels as well as statistics on income, age, sex, race, marital status, educational attainment, employee benefits, work schedules, school enrollment, health insurance, noncash benefits and migration.

The American Community Survey, on the other hand, is the only source of small-area statistics available on a wide range of important social and economic characteristics for all communities in the country. In addition to income, poverty and health insurance, other topics include education, language ability, the foreign-born population, marital status, migration, homeownership, the cost and value of our homes and many more.

Statistics from these two surveys differ for multiple reasons. First, income questions on the CPS ASEC are much more detailed than the summary questions asked on the American Community Survey. For the CPS ASEC, trained interviewers administer the survey while people primarily respond to ACS questions over the Internet or by mail. (Trained interviewers follow up with households who do not respond to the ACS online or by mail.)

Second, the reference periods for the two surveys are very different. The CPS ASEC asks respondents to report on their income in the previous calendar year. The ACS asks about income in the prior 12 months. Since the ACS is a continuous survey administered throughout the year, some respondents to the 2014 ACS (those who fill out the survey in January 2014) are reporting income received between January 2013 and December 2013 while other respondents (those who fill out the survey in December 2014) are reporting income received between December 2013 and November 2014.

These differences often result in different national statistics for such key indicators as poverty, median income and income inequality. Despite differences in the “levels” of these indicators, the trends over time tend to be very similar across the two surveys. The following graphs show median household income and poverty rates from the ACS compared with statistics from the CPS ASEC for previous years. The red line adjusts the CPS ASEC for the differences in reference periods.


Many people contact us each year asking which estimate to use for a particular purpose. For national statistics, we recommend the CPS ASEC because it provides a consistent historical time series at the national level back to, in some cases, more than half a century. Because of the larger sample size and smaller sampling errors, we recommend using the ACS for subnational geographies.

If you are interested in a longer time series than is available from the American Community Survey, generally we have recommended using two- or three-year averages from the CPS ASEC. However, because of the introduction of the redesigned income questions in a portion of the 2014 CPS ASEC sample and the entire 2015 CPS ASEC sample, we are not publishing these multiyear income and poverty estimates this year. We will resume production of these multiyear estimates next year when we have two years of consistent estimates.

Posted in Income, Poverty | Leave a comment

Economic Surveys Provide a Detailed Look at the Characteristics of Small Businesses

Bookmark and Share

Written By: Erika Becker-Medina and Patrice Norman

Today’s release of 2012 Survey of Business Owners (SBO) statistics provides a preliminary socio-economic picture of business owners across the nation with final, and more detailed statistics, to come later this year in December.

As part of the Economic Census, which is conducted every five years, the Survey of Business Owners uses a sample of 1.75 million employer and nonemployer businesses. The results provide estimates on firms, receipts, payroll and employment by gender, ethnicity, race and veteran status.

Today’s preliminary release provides a first look snapshot since the 2007 SBO data were released. Preliminary data are available by economic sector for the nation and individual states. We use the 2012 North American Industry Classification System (NAICS) to define the 18 sectors of the economy at the two-digit level. The final estimates in December will replace the preliminary estimates, as well as provide additional detail by geography and level of industry (two- through six-digit NAICS levels).

Additionally, more detailed levels of geography, including counties, metropolitan areas and places, will be published. December’s release will also include the Characteristics of Businesses and Characteristics of Business Owners data tables.

In between these two Survey of Business Owners releases, the Census Bureau plans to kick off data collection for the 2014 Annual Survey of Entrepreneurs. The Annual Survey of Entrepreneurs is a supplement to the Survey of Business Owners and will produce similar data more frequently.  This new collection is  a public-private partnership between the Ewing Marion Kauffman Foundation, the Minority Business Development Agency, and the Census Bureau. More information on this upcoming survey will be available soon.


Posted in Economy | Leave a comment

Majority of Americans Drive to Work, But Less so for Urban Millennials

Bookmark and Share

By: Brian McKenzie

If your drive to work feels a little lonely, you may be among the three-quarters of U.S. workers who drive to work alone. Driving alone reached its highest point in 2010, at 76.6 percent of workers, after a decades-long pattern of increase (see below). It remained the most common type of work travel in 2013.


About 86 percent of U.S. workers commuted by automobile in 2013, down from about 88 percent in 2000.The automobile has influenced the form of our communities and how we travel within them. It continues to dominate the work commute as the primary mode of transportation to work, although the rate of driving to work has declined in recent years.

Using data from the 2013 American Community Survey, the Census Bureau’s latest report on commuting, “Who Drives to Work? Commuting by Automobile in the United States: 2013” highlights differences in how people get to work and how that has changed.

Overall, automobile commuting has declined since 2000, in part, due to the continuation of a long-standing pattern of a decline in carpooling. One out of five U.S. workers carpooled in 1980, but in 2013, only one out of 10 carpooled. Hispanic workers showed the highest rate of carpooling in 2013, but also the largest declines in carpooling between 2006 and 2013, from 19 percent to 15 percent.


Some of the most notable recent changes in commuting have occurred among younger workers, particularly those living in cities. In 2013, about 74 percent of workers ages 16 to 24 living in principal cities commuted by automobile, compared with about 80 percent of the oldest workers. Workers between the ages of 25 and 29 living in principal cities showed the largest auto commuting declines since 2006, at about 4 percentage points. This is also the group that was least likely to have a vehicle at home in 2013.


For more information on commuting products, including special reports on topics such as working at home, daytime population change, and bicycling to work, visit the Census Bureau’s commuting home page.

Posted in Transportation | 2 Comments