IPUMS NHIS

Before beginning research on retirement or retirees, data users should have a clear definition of retirement and their study population. Some definitions of retirement require that an individual not be in the labor force. Some researchers may be interested in semi-retirement, e.g., individuals who receive a Social Security pension but continue to work, or have returned to work after retiring.

Researchers studying retirement or retirees will need information on one or more of the following factors:

• Retirement status: Some data collections and samples include variables that clearly indicate that an individual is retired.
• Labor force/employment status: If retirement status is not explicitly reported, users can make assumptions about who is and is not retired based on employment or labor force status.
• Social Security income receipt and amount: Many U.S. retirees receive a Social Security pension—but some individuals who are not retired (i.e., survivors of Social Security-eligible individuals, and people with disabilities) also receive Social Security income.
• Other retirement income receipt and amount: Other sources of retirement income to consider include income from IRAs, 401Ks, other private savings accounts, and employee pensions.

The table below summarizes the information available on retirement in four IPUMS data collections: IPUMS USA, IPUMS CPS, IPUMS Health Surveys (IPUMS MEPS and IPUMS NHIS), and IPUMS ATUS. 

Table 1. Information on retirement in IPUMS data collections

Example studies of retirement using IPUMS data

• Coile, C. C., & Levine, P. B. (2011). Recessions, retirement, and social security. American Economic Review, 101(3), 23-28.
• Flood, S., Moen, P., & Wang, J. (2019). SHORT-TERM US PATHS TOWARD RETIREMENT: SIMILARITIES AND DIFFERENCES ACROSS AGE-GENDER-CLASS INTERSECTIONS. Innovation in Aging, 3(Supplement_1), S805-S805.

Suggested citation

Pastoor, Isabel. 2026. “Studying Retirement with IPUMS Data.” University of Minnesota Life Course Center, November 3, 2026. https://lcc.umn.edu/node/2251.

The principal challenge to studying grandparents and their relationships with their adult children and grandchildren is the need to link related individuals who live in different households. In most public use census and survey data, a grandparent cannot be linked to family members who are not part of the same household. The relationships between individuals who do not live together are unknown. Names or family relationship links are not provided in public data.
 

This means that researchers interested in studying grandparents using IPUMS data will usually need to study grandparents who live with their child(ren) and grandchild(ren). The household roster in most IPUMS data collections allows users to observe the age and sex of each household member. The family interrelationship variables (sometimes present in the original data, and sometimes constructed by IPUMS using linking rules) provide further information about how household members are related to each other. For example, a pointer variable reports the person number of each individual’s mother, if their mother is in the household. Users can leverage these sets of variables to determine the number of generations in a household and how each individual is related to one another. Read more about how IPUMS has constructed family interrelationship variables for IPUMS USA and IPUMS CPS.

Limiting a study to only grandparents who are co-resident with their child(ren) and grandchild(ren) has drawbacks. Sample sizes will be relatively small, and may be unusable depending on the data collection used and other restrictions placed on the study population. Grandparents who live with these family members are also a selected group; they are likely systematically different from the general population of grandparents. 
 

Researchers who wish to study co-resident grandparents can do so using most IPUMS data collections that provide microdata (i.e., not IPUMS NHIGS or IPUMS IHGIS). Some data collections are more suitable than others. See a few relevant collection-specific notes below:
 

• In recent NHIS samples, studying the relationships between household members is nearly impossible, as even basic demographic information (e.g., age and sex) is unavailable for household members other than the sample adult and sample child.
• Three variables in IPUMS USA provide information on living with and caring for grandchildren: GCHOUSE, GCRESPON, and GCMONTHS. 
• In IPUMS ATUS, users can study time spent with grandchildren, even if they do not live in the same household. Users can also study time spent on caring for grandparents, even if they do not live in the same household.
• IPUMS USA provides historical linked data from the U.S. decennial census, allowing researchers to observe individuals and households across decades. These links make it possible to identify grandparents who do not live with their grandchildren in some cases.
• A variable called MULTGEN, which reports the number of generations in each household, is available from IPUMS USA and from IPUMS CPS.

Example studies of grandparents and grandchildren using IPUMS data

• Glaser, K., Stuchbury, R., Price, D., Di Gessa, G., Ribe, E., & Tinker, A. (2018). Trends in the prevalence of grandparents living with grandchild(ren) in selected European countries and the United States. European journal of ageing, 15(3), 237-250.
• Kwon, S. J. (2024). Grandparents and parental labor supply during the COVID-19 pandemic. Review of Economics of the Household, 22(3), 935-964.

Suggested citation

Pastoor, Isabel. 2026. “Under the Same Roof: Identifying Grandparents in IPUMS Data.” University of Minnesota Life Course Center, October 28, 2026. https://lcc.umn.edu/node/2246.

Using NHIS for the Study of Aging, Health, and the Life Course

The NHIS is well-suited for a broad range of aging, health, and life course research applications, including on topics such as mental health, chronic conditions, functional limitations, health behaviors, public program participation, mortality, and health care access and utilization. The NHIS data are unique among long-running, population-representative data resources for the inclusion of multiple validated mental health measures, information on sexual orientation, topical supplements and items designed to capture emerging public health content, and a publicly available linkage with mortality information from the National Death Index (NDI). A sampling of  recent publications include studies using the NHIS data from IPUMS to investigate whether engaging in moderate to vigorous activity can mitigate the mortality risks of fall injuries for older adults (Adeyemi et al., 2025); the association between health care cost burdens and the frequency of depressive/anxious feelings in older adults (Choi, et al. 2024); and the link between housing assistance and health care access for older adults with and without Alzheimer’s Disease and Related Dementias (Fenelon and Chen 2025).

Analytic Considerations for Researchers

The NHIS data are a rich data resource suitable for the study of many aging and health-related topics, but interested researchers should keep a few analytic considerations in mind when they use these data. First, the NHIS is administered to a sample of the community-dwelling (i.e., civilian, non-institutionalized) population, making it unsuitable for the study of institutionalized populations such as nursing home residents. Second, geographic detail in the NHIS public use data about where NHIS survey participants live is limited to region (northeast, midwest, south, and west), potentially limiting the utility of the public use NHIS data for research questions requiring more detailed information about where people live. Third, the NHIS uses a stratified design to select its sample, requiring users to account for the complex survey design to correctly produce population-representative estimates using the data. Please see the IPUMS NHIS user note on variance estimation for more information and sample code. 

Example Studies Using IPUMS Data

Adeyemi, Oluwaseun, Tracy Chippendale, Olugbenga Ogedegbe, Dowin Boatright, and Joshua Chodosh. "Activity Intensity and All-Cause Mortality Following Fall Injury Among Older Adults: Results from a 12-Year National Survey." In Healthcare, vol. 13, no. 19, p. 2530. MDPI, 2025. https://doi.org/10.3390/healthcare13192530

Choi, Namkee G., C. Nathan Marti, Bryan Y. Choi, and Mark M. Kunik. "Healthcare cost burden and self-reported frequency of depressive/anxious feelings in older adults." Journal of gerontological social work 67, no. 3 (2024): 349-368. https://doi.org/10.1080/01634372.2024.2326683

Fenelon, Andrew, and Jie Chen. "Housing assistance and health care access among older adults with Alzheimer’s Disease and related dementias: evidence from Multisource linked Survey-administrative data." The American Journal of Geriatric Psychiatry (2025). https://doi.org/10.1016/j.jagp.2025.11.007

The NHIS is a nationally representative survey that collects data on health, healthcare access, and health behaviors. IPUMS NHIS provides data from the survey from 1963 to present. In 1974, and 1976-2024 (the most recent year of data available), the survey included a height variable. Children’s height has only rarely been collected (see the universe of HEIGHT in IPUMS NHIS), so this blog post focuses on the height of adults age 18 and older. Height is respondent-reported, rather than measured directly at the time of the survey, and may be reported by someone other than the person themselves (i.e., another household member responding to the NHIS). Extreme values of height are censored in the public use data file, and individuals whose height was censored are not included in the analysis in this blog. See the codes section of HEIGHT for more detail.

Using cross-sectional survey data, it is not possible to observe within-individual changes in height over time. That is, among respondents to the National Health Interview Survey, if the average height declines in age, it does not necessarily mean that people become shorter throughout the life course; rather, it could reflect features of the survey design or methodology, mortality selection, migration selection, survey non-response, or other factors. For example, men are on average taller than women, and also have a shorter life expectancy than women. As a cohort becomes more female with age, average height will become more reflective of women’s heights than men’s. Additionally, improvements in childhood health or nutrition over several decades could mean that older cohorts are on average shorter than younger ones. 

Nonetheless, analyzing cross-sectional survey data to identify the relationship between age and height can still provide insights into how the demographic characteristics of a population are expected to be associated with the physical characteristics of the population. 

How does height vary by age?

The graph (Figure 1) below shows the average (mean) height in inches of U.S. adults by age, using pooled NHIS sample data from 2000-2024. The average height of adults declines in age, with the most dramatic reductions in average height occurring beginning around age 65. Since age is top-coded in the NHIS data at 85, the sharp drop in average height between ages 84 and 85 on the graph reflects the difference between mean age among 84-year-olds and mean age among adults age 85 and older. Reductions in mean height by age are modest between ages 18 and 55, amounting to a difference of less than 0.5 inches over the 37-year range. 

Differences by sex

Women are shorter than men, on average. But does the relationship between age and height differ by sex? Shown in the graph below (Figure 3), average height declines as age increases for both men and women. But the rate at which average height declines over age differs by sex.  For men, average height declines rather consistently and constantly with age between ages 18 and about 65. Then, after age 65, average height declines more rapidly, with a sharp dropoff after age 80. For women, average height declines more slowly at first, with more minor changes in average height between ages 18 and about 55, followed by a steeper decline after age 55. 

At age 18, men’s average height is 70.15 inches, or 5 feet and 10 inches. By age 84, the average height among men is 68.87 inches, or 5 feet and 8.9 inches. This is a decline of 1.28 inches. For women, average height declines from 64.47 inches (5 feet and 4.5 inches) at age 18 to 63.09 inches (5 feet and 3.1 inches) at age 84. This is a decline of 1.38 inches. Considering that women’s average height is lower than men’s, the decline in average height between ages 18 and 84 is larger for women than men in both absolute and relative terms. 

Height and mortality

The relationship between height, health, and longevity is highly debated, and differs across contexts (Deaton, 2007; Samaras, 2012; Perkins et al., 2016). Using mortality variables from the NHIS Linked Mortality File (LMF), it is possible to examine the relationship between height and age at death. The Linked Mortality File includes information on mortality status, and if deceased, year and month of death, for NHIS respondents eligible for mortality follow-up. Mortality status is recorded as of 2019. This means that in the 2000-2018 NHIS samples, which include mortality variables, deaths occurred between less than one year and 19 years after the NHIS survey was conducted.

Is height associated with age at death? For the purpose of descriptively answering this question, height is conceptualized as the difference between an individual’s height at the time of their participation in the NHIS survey, and the average height among individuals of the same age and sex. This measure is negative if the individual is shorter than average for their age and sex, and positive if the individual is taller than average. 

The graph below (Figure 3) shows this measure of height plotted by age at death, among people who were deceased by December 31, 2019 and eligible for mortality follow-up. The figure includes only people who died between age 33 and 98, as sample sizes of individuals who died before age 33 or after age 98 are very small.

Notably, among NHIS respondents eligible for mortality follow-up and who died before 2020, there is not a clear relationship between height and age at death. People who die at younger ages do not appear to be relatively shorter or taller (compared to their peers defined by age and sex) than people who die at older ages. 

How has the relationship between age and height changed throughout the 21st century?

Across the 21st century, there have been changes in how strongly age and height are related.

Many factors could influence the relationship between age and height, such as life expectancy, mortality selection, childhood nutrition, advances in medical technology and healthcare, and more. How has the correlation between age and height changed over time? The graph below (Figure 4) plots the correlation coefficient between age in years and height in inches over time, from 2000 through 2024. The correlation coefficient represents the relationship between a people’s age and their height. This coefficient is negative in all years, meaning that as age increases, average height decreases.

The correlation between age and height became weaker (got closer to zero) between 2000 and 2014. Then, the correlation grew stronger (became more negative) between 2015 and 2024. An important caveat underlying this trend is that the NHIS was redesigned in 2019, and data from 2018 and earlier are not comparable to data from 2019 and later. NHIS data users should acquaint themselves with the implications of the redesign for their work by reading the IPUMS NHIS page on the 2019 redesign.

Suggested citation

Pastoor, Isabel. 2026. “Age and Height: What's the Relationship?” University of Minnesota Life Course Center, July 23, 2026. https://lcc.umn.edu/node/2091.