Official Poverty Rates Using the CPS Public-Use Microdata Series
Updated: September 2017
Every year, the Census Bureau produces official poverty estimates using the Current Population Survey Annual Social and Economic Supplement (see Census Poverty website). Poverty estimates have been produced since 1959 and are listed on Table B-1 - Poverty Status: 1959 to 2016 from the Current Population Report: Income, Poverty, and Health Insurance Coverage in the United States: 2016. IPUMS-CPS makes available an intervalled variable, POVERTY. Many users have noted that exact replication of the estimates published by the Census Bureau is not possible. This fact is highlighted in Table 1, below, which compares the official poverty estimates to the rates obtained using the POVERTY IPUMS variable.
The IPUMS-CPS team worked closely with the Census Bureau to reproduce official poverty estimates from the PUMS. Updated estimates that replicate the official poverty estimates of the Census Bureau can be found in the IPUMS variable OFFPOV and are also listed by year on Table 1.
- Methodology for Obtaining Official Poverty Rates
- Subfamily Income Recode
- Differences between OFFPOV and POVERTY
What is IPUMS-CPS releasing?
To facilitate analysis for researchers, IPUMS-CPS makes available 6 new variables that allow for the replication of official poverty estimates: OFFPOV, OFFPOVUNIV, OFFCUTOFF, OFFTOTVAL, OFFREASON, and FTYPE.
OFFPOV is the official poverty status of each person. OFFPOVUNIV is the official poverty universe classification of each person. OFFCUTOFF gives the official poverty threshold for each person, treating the total number of adults and children within primary and related subfamilies as one family. (Note that CUTOFF between 1975 and 1987 reflects the threshold for each individual family, treating related subfamilies separate from primary families). OFFTOTVAL gives the total family income, treating primary families and related subfamilies as one family unit. OFFREASON denotes the potential reason for discrepancies between OFFPOV and POVERTY. Finally, FTYPE gives the official family type for each person.
Methodology for Obtaining Official Poverty Rates
In general, the poverty classification of any given person is simply a comparison between their family's total income and the poverty threshold for a family of that size. The family unit is defined by the variable FTYPE. This variable consists of 5 family types: primary family, nonfamily householder, related subfamily, unrelated subfamily, and secondary individual. Total family income is the sum of the individual incomes of each family member. Total family income is available through FTOTVAL while total individual income is available through INCTOT. Each family's total income is then compared to a specific poverty threshold that depends on the size of the family and the ages of the family members. Specific thresholds are available through the Census Bureau CPS Poverty Thresholds webpage. Given this information, the computation of each family's poverty status is very straightforward. As the Census Bureau explains,
- If total family income is less than the threshold appropriate for that family:
- The family is in poverty.
- All family members have the same poverty status.
- For individuals who do not live with family members, their own income is compared with the appropriate threshold.
- If total family income equals or is greater than the threshold:
- The family (or unrelated individual) is not in poverty.
The official poverty estimates are then calculated by taking the ratio of the total number of people in poverty and the total number of people in the poverty universe. From 1969 through 1979, the universe for official poverty rates includes all persons except for unrelated individuals under age 14. In 1980 the universe changed to all persons except for secondary individuals under age 15. All official estimates are weighted using the ASEC weight WTSUPP.
See How the Census Bureau Measures Poverty for more information.
Table 1, below, compares the official poverty rates (as reported by the Census Bureau - Table B-1 - Poverty Status: 1959 to 2016) to the weighted poverty rates given by POVERTY and OFFPOV. Notice that without making further restrictions to POVERTY, the poverty rate given by POVERTY will not match the official poverty rate for most years. These restrictions are describe in the Differences between OFFPOV and POVERTY section below. On the other hand, OFFPOV replicates official poverty rates for all except 5 years (1970, 1975, 1980, 1993 and 2000). The most likely reason for this discrepancies in the 5 years is due to changes in the population controls that affected weighted estimates.
Poverty Year refers to the reference year of the survey. For example, in 2013, the reference period of all income questions is the past 12 months. Survey Year refers to the year that the survey was taken. The Official Rate comes from the Census Bureau report on poverty (linked above). POVERTY lists the weighted poverty rates (weight=WTSUPP) using the POVERTY variable. POVERTY - Official Rate is the difference between the POVERTY rates and the official rates. OFFPOV refers to the weighted poverty rates (weight=WTSUPP) using OFFPOV.
|Poverty Year||Survey Year||Official Rate||POVERTY||POVERTY - Official Rate||OFFPOV|
Subfamily Income Recode
According to the Census Bureau poverty guidelines for defining families, a family includes the primary family and all related subfamilies living within the same household. This definition implies that the incomes of primary families and related subfamilies (based on INCTOT) are treated as one family income. Similarly, the relevant threshold must reflect the size of the primary family and related subfamilies added together.
To clarify, consider the following example: a household record lists 8 different persons living within the same household. The primary family consists of 5 people (2 adults and 3 children) and the total family income is 60,000 dollars. The related subfamily consists of 3 people (1 adult and 2 children) and the total family income is 10,000 dollars. The table below shows that, without the proper family income recode, it is possible for a family to erroneously be considered as in poverty.
|Poverty Year||Survey Year||AGE||FTYPE||FTOTVAL||CUTOFF||POVERTY||Corrected FTOTVAL||Corrected CUTOFF||Corrected POVERTY|
|2011||2012||30||1 - Primary Family||60,000||26,844||Above Poverty Line||70,000||38,247||Above Poverty Line|
|2011||2012||32||1 - Primary Family||60,000||26,844||Above Poverty Line||70,000||38,247||Above Poverty Line|
|2011||2012||6||1 - Primary Family||60,000||26,844||Above Poverty Line||70,000||38,247||Above Poverty Line|
|2011||2012||4||1 - Primary Family||60,000||26,844||Above Poverty Line||70,000||38,247||Above Poverty Line|
|2011||2012||2||1 - Primary Family||60,000||26,844||Above Poverty Line||70,000||38,247||Above Poverty Line|
|2011||2012||25||3 - Related Subfamily||10,000||18,123||Below Poverty Line||70,000||38,247||Above Poverty Line|
|2011||2012||5||3 - Related Subfamily||10,000||18,123||Below Poverty Line||70,000||38,247||Above Poverty Line|
|2011||2012||4||3 - Related Subfamily||10,000||18,123||Below Poverty Line||70,000||38,247||Above Poverty Line|
In this example, FTOTVAL does not make the proper recode to give all people within the same primary and related subfamily the same total family income. All people within this household should have a total family income of 70,000 dollars (60,000 + 10,000). Meanwhile, the relevant poverty threshold should be that of 3 adults and 5 children. The corrected family income, cutoff and poverty status show that the related subfamily would erroneously be considered as below the poverty line when, by definition, all the people in this family are above the poverty line. Users may verify using INCTOT that OFFTOTVAL includes the income of related subfamilies.
Differences between OFFPOV and POVERTY
The main difference between OFFPOV and POVERTY is a universe issue. IPUMS-CPS attaches family-level variables to each person within that family. In following the Census Bureau data dictionary, IPUMS-CPS does not make any universe adjustment for POVERTY since the data dictionary does not specify a universe. However, since official poverty rates do have a universe specification, this process explains about half of the differences between the OFFPOV and POVERTY classification between each person. If one restricts the poverty universe correctly, POVERTY will replicate official poverty rates for most years except for survey years 2000, 1993, 1987-1975 and 1970. In order to facilitate analysis of the differences between OFFPOV and POVERTY, OFFREASON classifies the differences into three main categories.
First, OFFREASON equal to 01 distinguishes people who have no family income issue but do have either a universe issue or a coding issue. Note that if one restricts the universe to include only those in the poverty universe, then OFFREASON equal to 01 will lists all those individuals who have a coding issue. That is, cases where a comparison of total family income to the poverty cutoff shows that family income is greater than the cutoff but POVERTY classifies them as poor or where family income is less than the cutoff but does not classify them as poor.
Second, OFFREASON equal to 02 codes cases where POVERTY classifies people as poor but these are families with missing family incomes. These people should not have a poverty classification. Again, most of these differences are not an issue if POVERTY is restricted to the proper poverty universe. However, even if this is done, there are some cases where this is an issue.
Third, OFFREASON equal to 03 codes the discrepancies where family incomes do not match. All of these cases are subfamily recode issues. Once subfamily incomes are added to the primary family incomes, the resulting incomes are correct.
Each of these differences are coded with their particular discrepancy explanation through the variable OFFREASON (frequencies are available under the codes tab). Note that the discrepancies are based on a comparison using POVERTY without any universe restrictions.