Good morning and thank all of you for being here today. (Slide 1)
I would particularly like to thank the many Census staff who are involved in collecting, processing, compiling, preparing and disseminating the data for these important statistics – many of whom are here today. THANKS for your hard work and patience. It is only through the work of these dedicated staff that the Census Bureau can produce timely, accurate and relevant information about the nation’s economy.
I would also like to thank the millions of you – the respondents – who talked to us on the phone, invited us into your homes, or sent us your completed paper questionnaires. It is only because of you that we have these results today.
Most of the figures we are presenting today are included in two separate publications, which are included in your packet or can be found online. (Slide 2) One report presents income, poverty, and health insurance coverage data from the Annual Social and Economic Supplement of the Current Population Survey (or CPS), see the image of that report cover on the left. The other report presents the income, earnings, and poverty data from the American Community Survey (or ACS), see the image of the report cover on the right.
As in past years, the CPS is the survey used for national income and the official poverty estimates. The CPS is also the best source of data for trends for various demographic groups. The ACS, however, is the data source for examining subnational estimates and differences by geography.
For the third year, the Census Bureau collected information in every county in the nation for the ACS. Data collected in 2007 allow us to provide annual information for about 7,000 places, including all States, congressional districts, and counties with populations over 65,000. The Census Bureau also collects information about Puerto Rico in the Puerto Rico Community Survey (PRCS). While my remarks today focus on the 50 states and the District of Columbia, selected maps of the United States include Puerto Rico where the data are available.
On our website (and in your packet1), we have guidelines for using each of these data sets. Let me summarize this guidance for you. If you are interested in state estimates in comparison to each other and relative to national estimates we urge you to use ACS estimates. If you are interested in the relationship between local areas and states, you should also use the ACS. You should not compare the national estimates from the CPS to local estimates from the ACS, or annual changes from one survey compared to changes from the other survey.
We have structured this talk to show you how one might integrate the income and poverty results from both these data sets. The national data from the CPS are presented primarily in charts showing trends over time. The subnational data from the ACS are presented primarily in the form of maps depicting differences between states or smaller geographic areas. Since the Census Bureau discourages people from making direct comparisons of the estimates from one survey with the other, we hope this presentation style helps you interpret and use the data appropriately. The 2007 health insurance data presented today are all from the CPS.
Let me first summarize the national findings about income, poverty, and health insurance coverage from the Current Population Survey. (Slide 3)
Turning to a time series chart on real – inflation-adjusted – median household income, we can see that median household income has experienced three consecutive annual increases returning household income to a level not statistically different from the 1999 pre-recession income peak. (Slide 4)
The time series for the poverty rate on Slide 5 generally reflects the time series for median income. From the most recent trough in 2000, the poverty rate had 4 consecutive annual increases, from 11.3 percent in 2000 to 12.7 percent in 2004. And at 12.5 percent in 2007, it is not statistically different from the levels in the past 4 years. Similarly, the number of people in poverty is not different than the number in 2005 (and 2004).
The CPS data provides us with a consistent time series of data on median incomes and poverty rates. It also provides data that describes the disparities among households and people with different demographic characteristics.
This figure on Slide 6 shows historical real median household income by race and Hispanic origin.2 The median incomes of Black households and non-Hispanic White households rose between 2006 and 2007 – the first real increases in annual income for these households since 1999. The real median incomes of Asian households and Hispanic households remained statistically unchanged.
Among the race groups and Hispanics, Black households had the lowest median income in 2007, $33,900, which was 62 percent of the median for non-Hispanic White households, $54,900. Asian households had the highest median income, $66,100, about 120 percent of the median for non-Hispanic White households. The median income for Hispanic households was $38,700 in 2007, which was 70 percent of the median for non-Hispanic White households.
The beauty of the ACS data for income is that it allows us to examine the disparity of income across the country. Using the ACS, we can estimate household income in each state in every county and metropolitan or micropolitan statistical area with a population of at least 65,000.
The map on Slide 7 shows the median household incomes of states from the 2007 ACS. The median household income estimates in the 2007 ACS varied from state to state, ranging from a median of $68,100 for Maryland to $36,300 for Mississippi.3 Incomes were generally higher on the East and West coasts than they were in the rest of the country. In fact, 13 out of the 18 states with median household incomes higher than the ACS U.S. median were East and West coast states.
These coastal states tend to have higher costs of living, which is reflected in higher median incomes. Household income of $50,000 in California may not represent the same level of economic well-being as $50,000 in Minnesota due to cost-of-living differences. An advantage of the ACS is that we also collect information on the rents and housing costs for each area. These other pieces of information can be used when analyzing income and poverty across areas (which I will discuss in a minute with respect to poverty).
Slide 8 presents an example of how the ACS allows us to explore smaller geographic areas by showing the variation of median household income across Metropolitan and Micropolitan statistical areas with populations of 65,000 or more. The 2007 ACS provides data for about 500 such areas.
Similar to the map of state median household income, this map shows a higher concentration of higher income areas (dark green areas with median incomes in the range from $65,000 to $83,793) along the West and East coasts. Nationally, the median income is higher for households located in metropolitan statistical areas than for those living outside metropolitan or micropolitan areas.4
In the 2007 ACS, the total U.S. median household income in metropolitan or micropolitan statistical areas was $51,700. For households not in metropolitan or micropolitan statistical areas, their median income was $37,800.
There is also variation between the incomes for households inside and outside of principal cities within metropolitan statistical areas. In the 2007 ACS, the median income for households in principal cities within metropolitan statistical areas was 78 percent of the income for households living in the suburbs.5 In fact, median incomes were lower for households in principal cities of metropolitan statistical areas than households not in principal cities in all states except Alaska where they were not statistically different.
While the median represents just one point on the distribution of household income – the point at which half of the households have income below it and half above it — other points along the distribution provide additional information about the nation’s household income disparity. For example, the 10th percentile is the income level at which 10 percent of the households have income below it. The 90th percentile is at the other end of the distribution, the level at which 90 percent of the households have income below it.
Turning back to the CPS, Slide 9 shows that the 10th percentile was $12,200 in 2007. The 90th percentile was $136,000 in 2007. Comparing the relationship of these two values over time indicates growing income disparity. Over the past four decades, the ratio of the 90th percentile to the 10th percentile has grown 21.2 percent, suggesting that inequality has increased. However, in recent years (since 2003), there were no statistically significant differences.
Using the information about the distribution of household income from the CPS, we can produce a Gini index – the most widely used measure of inequality. The Gini index indicates higher inequality as the index approaches one. The figure in Slide 10 shows increasing inequality over time. While the Gini index decreased by 1.5 percent between 2006 and 2007, the 2007 index was not statistically different from any of the annual Gini indexes over the 1997 to 2005 period.
One aspect of the household income distribution is that it treats all households the same, regardless of the number of people residing there. For example, a single-person household with income of $30,000 is the same for the purposes of the income distribution as a household with two adults with combined income of $30,000. However, these households may have different needs (and hence, the use of different poverty thresholds).
Income inequality statistics, like the Gini index, can also be calculated using family and individual incomes that are family size adjusted, a method that considers the sharing of resources and economies of scale. Slide 11 presents this equivalence-adjusted income approach to measuring income inequality. The equivalence-adjusted income for a single person living alone would be more than twice as much as a 4-person family with the same income. As shown here, the 2007 Gini index for equivalence-adjusted income is 3.9 percent lower than the Gini index when using the conventional measure of money income; however, while the two Gini measures themselves were significantly different in each year, the patterns for the two Gini measures weren't significantly different since 1993.6
Returning to poverty, Slide 12 demonstrates that they also show disparities by demographic groups. Both the poverty rate and the number in poverty for people aged 18 to 64 were not statistically different in 2007 than in 2006, at 10.9 percent and 20.4 million in 2007. The poverty rate for people 65 and older remained statistically unchanged at 9.7 percent, while the number in poverty increased to 3.6 million in 2007 from 3.4 million in 2006. However, in 2007, both the poverty rate and the number in poverty increased for children under 18 years old (18.0 percent and 13.3 million in 2007, up from 17.4 percent and 12.8 million in 2006).
For many years, the poverty rate for children has been higher than the rates for people 18 to 64 years old and those 65 and older. In contrast, the 2007 poverty rate for people 65 and older (9.7 percent) was – for fifth consecutive year – lower than the poverty rate for adults aged 18-64 and was not statistically different from its pre-recessionary low in the 1999-2000 period.7 However, the poverty rates for children and adults are both higher in 2007 than their most recent pre-recessionary lows.
Again the ACS provides the information on the difference in poverty rates for subnational areas. The map in Slide 13 illustrates the variation in state poverty rates using the 2007 ACS. Poverty rates among the 50 states and the District of Columbia varied from a low of 7.1 percent for New Hampshire to a high of 20.6 percent for Mississippi. Similar to median income, states in the south tended to have higher poverty rates, while states in the northeast tended to have lower poverty rates. As shown in the map, fifteen states have lighter shades of which twelve had poverty rates below 11 percent, and five of those are in the Northeast.8
Slide 14 shows that the same geographic disparity in poverty exists for children, with poverty ranging from 8.8 percent in New Hampshire to 29.3 percent in Mississippi.9 Child poverty rates are higher than the overall poverty rates in almost all states.10
Official poverty, as defined by the Office of Management and Budget’s Statistical Policy Directive 14, uses a set of money income thresholds that vary by family size and composition but do not vary geographically. The Census Bureau uses the same threshold regardless of where a person or family resides. For example the 2007 poverty threshold was $21,027 for a family with two adults and two children regardless of whether the family lived in Los Angeles or Milwaukee.
The National Academy of Sciences (NAS) Panel on Poverty and Family Assistance stated that the cost of housing varied across geographic areas and they encouraged researchers to examine adjustments to poverty thresholds based on differences in housing costs.
The ACS data provides us with an opportunity to examine these differences in cost. We can use the actual rental costs in metropolitan statistical areas and non-metropolitan statistical areas within state to adjust the poverty thresholds (money income cutoff) to account for these differences. The map in Slide 16 presents these alternative poverty rates for the 50 states and the District of Columbia based on adjustments to the poverty thresholds for differences in housing costs using the same scale as the map in Slide 13. Comparing the two maps illustrates that poverty rates in many of the southern states, like Texas and Alabama, move to a lower bracket than without the adjustment, while the poverty rate in California moves to a higher bracket. In fact, the disparity between states is lower using the geographically-adjusted thresholds, with the range now from 7.9 percent to 18.4 percent.
In addition to geographic differences in cost of living, the Census Bureau recognizes that measuring money income may not completely capture the economic well-being of individuals and families. Families and individuals also derive economic well-being from noncash benefits, such as food stamps and housing subsidies, and they have reductions in disposable income due to taxes. While the income and poverty estimates shown in these reports are based solely on money income before taxes and do not include the value of noncash benefits, the Census Bureau computes a number of other measures of income and poverty that do attempt to account for those factors.
These alternatives fall into two categories: poverty measures based on the recommendations of the National Academy of Sciences and income and poverty estimates that use various formulas to add or subtract from resources to examine the incremental impact of these changes.
The Census Bureau has a web-based tool for you to explore alternative income and poverty measures.11 With this calculator, one can examine the sensitivity of the poverty rate to changes in the resource definition and the poverty threshold (money income cutoff) definition; examine relative poverty; and calculate poverty rates using a measure suggested by the National Academy of Sciences. Slide 17 presents estimates made using the web-based tool. Be aware that these data are 2006 income-year estimates (using last year’s CPS ASEC) because the additional input required to calculate these alternatives are not yet available for 2007.
Hence, using data from 2006, this web-based tool allows one to see the incremental impact of the addition or subtraction of a single resource element. If the cash-value of food stamps were added to the resources of families, this would move 1.8 million people above the poverty line (from 36.5 million in poverty using money income only to 34.7 million in poverty including cash-value of food stamps with income).
If income and payroll taxes were subtracted from money income (including refundable credits like the Earned Income Tax Credit), then 1.7 million children would move above the poverty line (from 12.8 million in poverty using money income to 11.1 million in poverty using after-tax income).
Finally, one can examine the effectiveness of Social Security income on the poverty of people age 65 and older. In 2006, the number of poor elderly people would be higher by 12.6 million people if Social Security payments were excluded from money income, more than quadrupling the number of poor elderly (from 3.4 million people using money income to 16.0 million people when subtracting Social Security income from money income).
The Census Bureau will release the 2007 data on alternative measures of income and poverty later this year.
Turning back to the official calcuation of poverty and to the disparities across subnational areas, the map in Slide 18 provides us with poverty estimates by metropolitan and micropolitan statistical areas with population of 65,000 or more.
Similar to the differences in income within metropolitan statistical areas there is additional variation for people living in principal cities within metropolitan statistical areas as compared to those living in suburbs. The poverty rates are about 80 percent higher for those living in principal cities.
Among the states, poverty rates for people living in principal cities within metropolitan statistical areas ranged from 6.5 percent to 23.7 percent, while the poverty rate for people within metropolitan statistical areas and not in principal cities ranged from 5.1 percent to 17.1 percent. Slide 19 shows this difference by assigning each state the difference between the two, literally the poverty rate within principal cities in the metropolitan areas minus the poverty rate in the suburbs. Higher numbers indicate relatively higher poverty in principal cities. Negative numbers indicate higher poverty in the suburbs.
This difference is higher in the Northeast (ratio of the poverty rate for people not in principal city to the rate for people in principal city was 0.38), and lower in the West (ratio of 0.72). (note: the difference is higher as the ratio is closer to 0.00 and lower as the ratio is closer to 1.0)
Among the states, the ratio of the poverty rate for people not in principal cities within a metropolitan statistical area to those in principal cities ranged from about 0.3 to about 1.1.12
The ACS provides the data necessary to drill down to areas within states. These areas can be counties or places with over 65,000, or special statistical areas with populations of 100,000 or more. The map on Slide 20 displays poverty rates for parts of southern California with special area boundaries based on population size. As this chart shows, poverty rates in these areas of southern California area range from 35 percent for inside Los Angeles (for example the Watts area) to 5 percent for the suburbs by the beach (for example the Palos Verdes area). This pattern is similar for the San Diego principal cities and suburbs.
Both the CPS and ACS can provide insight into the earnings of men and women and the disparities over time and by geography and occupation
Slide 21 presents the historical national CPS data on earnings from 1960 to 2007. The real median earnings of men who worked full-time, year-round rose from $43,500 in 2006 to $45,100 in 2007 and those of women rose from $33,400 in 2006 to $35,100 in 2007. Prior to 2007, both men and women had experienced three consecutive annual declines in real earnings. In 2007, the female-to-male earnings ratio was 0.78, which is an all-time-high.
While overall median household income in 2007 rose by 1.3 percent, the real median income of households with a householder that worked full-time, year-round rose 1.7 percent and the income of households with no earners declined 4.8 percent.13 This occurred while the median earnings of both men and women who worked full-time, year-round rose by 3.8 percent and 5.0 percent, respectively.14
Over the past ten years, the proportion of working women who reported being year-round, full-time workers increased 5.8 percentage points (from 55.6 percent to 61.4 percent), compared with a 2.9 percentage point increase for men (from 71.6 percent to 74.5 percent).
With the large sample sizes in the ACS, we can also produce estimates of median earnings for men and women by state and occupation group to give added insight to the national patterns shown by the CPS.
In the 2007 ACS, six states (Connecticut, New Jersey, Maryland, Massachusetts, New Hampshire, and Alaska) had median earnings of men over $50,000 while no state had median earnings for women above $50,000, and in each of the 50 states, women’s median earnings were less than men’s median earnings.
The table in Slide 22 shows the five occupation groups for men and women with the highest median earnings. Four out of five of the groups are the same for men and women; however the median earnings and the ranges are different for men and women. The median earnings for women in legal occupations was only 51.1 percent of men’s earnings, while for computer and mathematical occupations the percentage is 86.1 percent.15
There is more parity, however, between women’s and men’s earnings among the more detailed occupation sub groups. (Slide 23) For example, within the legal occupations, among lawyers, women’s earnings were 77.8 percent of men’s earnings, and among paralegals women’s earnings were 93.2 percent of men’s earnings. Hence, much of the divergence within these large groups may be due to the composition of men and women in these more detailed occupation sub-groups. In fact, 58 percent of women working in legal occupations were working as paralegals and in miscellaneous legal occupations, while only 12 percent of men in legal occupations were working in those specific areas.
This increase is parity in also demonstrated when comparing the ratios for the detailed occupations to the overall ratio. Of the 283 occupations with which we were able to produce statistically reliable women’s-to-men’s earnings ratios, 41 percent had a ratio above the average of 78 percent, while only 16 percent had a ratio below. This suggests that the average ratio is due in part by the percentage of women in low earnings occupations coupled with lower ratios.
Finally, the CPS provides statistics about health insurance coverage in the U.S. and the disparities across groups. Slide 24 presents the number and percent uninsured from 1987 to 2007. Both the percentage and the number of people without health insurance decreased in 2007. The percentage without health insurance was 15.3 percent in 2007, down from 15.8 percent in 2006, and the number of uninsured was 45.7 million, down from 47.0 million in 2006.
The number of people with health insurance increased to 253.4 million in 2007 (up from 249.8 million in 2006). Much of this increase was due to the increase in the number of people covered by government health insurance, 83.0 million people in 2007 (up from 80.3 million in 2006).
In 2007, the percentage and the number of children under 18 years old without health insurance (11.0 percent and 8.1 million) were lower than in 2006 (11.7 percent and 8.7 million). (Slide 25)
Slide 26 shows that the proportion of children not covered by health insurance varied by poverty status, age, race, and Hispanic origin. Children in poverty were more likely to be uninsured than the population of all children in 2007 — 17.6 percent compared with 11.0 percent. Children 12 to 17 years old had a higher uninsured rate than those under 12 years old — 12.0 percent compared with 10.4 percent. Of the 8.1 million children in the poverty universe and without health insurance, 80.5 percent are living with incomes below 300 percent of their poverty threshold.16
Finally, we can use the CPS to calculate 3-year average estimates of the uninsured rate for children by state to show the variation. Comparing 3-year-average uninsured rates for children for 2005-2007 across states shows that Texas (20.5 percent) had the highest percentage of uninsured. Among the states with the lowest point estimates for uninsured rates are Massachusetts, Michigan, Iowa, Hawaii, and Wisconsin. Many of the states with higher uninsured rates also had higher child poverty rates. (Slide 27 shows which states had a 3-year average uninsured rate higher than, lower than, or not statistically different from the national 3-year average uninsured rate.)
Later this year, the Census Bureau will release sample-based estimates for the child insurance rates for all counties. And next year, we will have ACS data on health insurance coverage for all areas with over 65000.
(Slide 29) We have shown you just some of the data available from the Annual Social and Economic Supplement of the CPS and the American Community Survey. Much more data are available in the reports and on the Census website, at www.census.gov and the links provided there.
That concludes my presentation. Thank you.
____________________________
1See www.census.gov/acs/www/UseData/compACS.htm.
2Federal surveys give respondents the option of reporting more than one race. Therefore, two basic ways of defining a race group are possible. A group such as Asian may be defined as those who reported Asian and no other race (the race-alone concept) or as those who reported Asian regardless of whether they also reported another race. The results presented here use the first approach (race alone). The report upon which the data are based includes estimates using both approaches. Hispanics may be of any race.
3The median household income for the state of Mississippi was not statistically different from the median household income for West Virginia.
4For the state of Wyoming the median household income in metropolitan statistical areas is not statistically different from the median household income outside metropolitan and micropolitan areas.
5The term suburbs refers to the territory within the metropolitan statistical area that is outside the principal city.
6The ACS data can also be used to summarize the disparity of income, and the report produces Gini indexes for all states. The Gini index calculated using 2007 ACS varied from state to state, ranging from 0.542 for the District of Columbia to 0.409 for Utah and Alaska.
7The rates were 9.7 percent in 1999, 9.9 percent in 2000, 9.8 percent in 2004, and 9.4 percent in 2006, all not statistically different from each other and the 2007 rate of 9.7 percent.
8The poverty rates for Delaware (10.5 percent), Nevada (10.7 percent), and Wisconsin (10.8 percent) are not statistically below 11 percent.
9The poverty rate for children in New Hampshire (8.8 percent) was not statistically different from the rates in Hawaii (9.8 percent) and Wyoming (11.6 percent).
10North Dakota and Wyoming are the only two states where child poverty rates are not statistically higher than the overall poverty rates.
11The tool, called CPS Table Creator II, is available in a link from the “Microdata Access” page on the poverty web site,
http://www.census.gov/hhes/www/poverty/poverty.html.
12Two states, Alaska and New Mexico, had ratios not statistically different from 1 – meaning the poverty rate for people living inside the principal cities was not statistically different from the poverty rate for people living not in principal cities within metropolitan statistical areas.
13The apparent difference between the increase in median household income of all households and that of households with a householder who worked full-time, year-round was not statistically significant.
14The number of working men age 15 and older increased by 0.6 million between 2006 and 2007 to 84.5 million. An estimated 74.5 percent worked full-time, year-round, a lower percentage than in 2006 (75.1 percent). The number of working women age 15 and older was 74.4 million, an increase of 0.6 million from 2006. About 61.4 percent of these women worked full-time, year-round in 2007, an all-time high, up from 60.6 percent in 2006.
15Men earned the most in the legal occupations ($105,200). Women who worked in computer and mathematical occupations had the highest median earnings ($62,000).
16Estimates for children below 300 percent of their poverty thresholds come from a special tabulation using CPS Table Creator, available online <http://www.census.gov/hhes/www/hlthins/data_access.html>.
-X-
