Objectives:

This report presents complete period life tables for each of the 50 states and the District of Columbia by sex based on age-specific death rates in 2021.

Methods:

Data used to prepare the 2021 state-specific life tables include: 2021 final mortality statistics; July 1, 2021, population estimates based on the Blended Base population estimates produced by the U.S. Census Bureau; and 2021 Medicare data for people ages 66–99. The methodology used to estimate the state-specific life tables is the same as that used to estimate the 2021 national life tables, with some modifications.

Results:

Among the 50 states and District of Columbia, Hawaii had the highest life expectancy at birth, 79.9 years in 2021, and Mississippi had the lowest, 70.9 years. From 2020 to 2021, life expectancy at birth declined for 39 states, increased for 11 states, and remained unchanged for the District of Columbia. In 2021, life expectancy at age 65 ranged from 16.1 years in Mississippi to 20.6 years in Hawaii. Life expectancy at birth was higher for females in all states and the District of Columbia. The difference in life expectancy between females and males ranged from 3.9 years in Utah to 7.6 years in New Mexico.

Explanation of life table columns

Column 1. Age (between x and x+1)—Shows the age interval between the two exact ages indicated. For instance, 20–21 means the 1-year interval between the 20th and 21st birthdays.

Column 2. Probability of dying (q_x)—Shows the probability of dying between ages x and x+1. For example, for males who reach age 20 in Massachusetts, the probability of dying before reaching their 21st birthday is 0.000756 (Table MA–2). This column forms the basis of the life table; all subsequent columns are calculated from it.

Column 3. Number surviving (l_x)—Shows the number of people from the original hypothetical cohort of 100,000 live births who survive to the beginning of each age interval. The l_x values are computed from the q_x values, which are successively applied to the remainder of the original 100,000 people still alive at the beginning of each age interval. For example, out of 100,000 male babies born alive in Massachusetts in 2021, 99,196 will survive to their 21st birthday (Table MA–2).

Column 4. Number dying (d_x)—Shows the number dying in each successive age interval out of the original 100,000 live births. For example, out of 100,000 males born alive in Massachusetts in 2021, 75 will die between ages 20 and 21 (Table MA–2). Each figure in column 4 is the difference between two successive figures in column 3.

Column 5. Person-years lived (L_x)—Shows the number of person-years lived by the hypothetical life table cohort within an age interval x to x+1. Each figure in column 5 represents the total time (in years) lived between two indicated birthdays by all those reaching the earlier birthday. Consequently, the figure 99,233 for males in the age interval 20–21 is the total number of years lived between the 20th and 21st birthdays by the 99,271 males in Massachusetts (column 3) who reached their 20th birthday out of 100,000 males born alive (Table MA–2).

Column 6. Total number of person-years lived (T_x)—Shows the total number of person-years that would be lived after the beginning of the age interval x to x+1 by the hypothetical life table cohort. For example, the figure 5,703,533 is the total number of years lived after reaching age 20 by the 99,271 males reaching that age in Massachusetts (Table MA–2).

Column 7. Expectation of life (e_x)—At any given age, shows the average number of years remaining to be lived by those surviving to that age, based on a given set of age-specific rates of dying. It is calculated by dividing the total person-years that would be lived beyond age x by the number of people who survived to that age interval (T_x/l_x). For example, the average remaining lifetime for males in Massachusetts who reach age 20 is 57.5 years (5,703,533 divided by 99,271) (Table MA–2).

Standard errors of probability of dying and life expectancy

Although based on complete counts of death, the life table functions presented in this report are subject to error. As a result, standard errors of the two most important functions, the probability of dying and life expectancy, are also presented. The mortality data on which state life tables are based are not affected by sampling error because they are based on complete counts of deaths and, as a result, standard errors reflect only stochastic (random) variation. While measurement errors such as age misreporting on death certificates or census data are known to affect mortality estimates, they are not considered in calculating the standard errors of the life table functions. In most cases, standard errors for life expectancy at birth and the probability of dying are small due to large numbers of deaths. However, for some states with small populations, particularly at the youngest ages, the standard errors presented are relatively large.

Complete life tables for 50 states and D.C.

A set of complete period life tables for each state and D.C. is available online from “U.S. State Life Tables, 2021” at: ftp.cdc.gov/pub/Health_Statistics/NCHS/Publications/NVSR/73-07/. Table I lists table titles for each of these tables. Table numbering is based on the federal information processing standards, or FIPS, alphabetical code for the state combined with a table code. The table codes are 1 for the total population, 2 for males, 3 for females, and 4 for the standard errors of the probability of dying and life expectancy. For example, Table FL–2 refers to the complete period life table for males in Florida.

Life expectancy in 50 states and D.C.

Table A shows life expectancy at birth for the total, male, and female populations for each state, D.C., and the United States. In 2021, among the 50 states and D.C., Hawaii ranked first for the total, male, and female populations, with life expectancies at birth of 79.9, 77.0, and 83.1 years, respectively. Mississippi ranked 51st among the 50 states and D.C. for the total and male populations, with life expectancies at birth of 70.9 and 67.7, respectively. West Virginia ranked 51st for females with a life expectancy of 74.2. In comparison, life expectancy at birth for the entire United States was 76.4, 73.5, and 79.3 for the total, male, and female populations, respectively. Figure 1 presents a U.S. map with state-specific life expectancy at birth grouped into quartiles. It shows that states with the lowest life expectancy at birth were mostly Southern states (Oklahoma, Louisiana, Mississippi, Alabama, Georgia, South Carolina, Tennessee, Arkansas, Kentucky, and West Virginia) but also included New Mexico, Ohio, and Alaska. States with the highest life expectancy at birth were predominantly Western (Hawaii, California, and Washington) and Northeastern states (New York, Vermont, New Hampshire, Connecticut, Massachusetts, Rhode Island, and New Jersey) but also included Utah and Minnesota.

Table A

Life expectancy at birth, rank, and standard error, by sex: Each state, District of Columbia, and United States, 2021.

Figure 1

Life expectancy at birth: Each state, District of Columbia, and United States, 2021. SOURCE: National Center for Health Statistics, National Vital Statistics System, mortality data file.

The difference in life expectancy between the sexes in the United States was 5.8 years in 2021, ranging from a high of 7.6 years in New Mexico to a low of 3.9 years in Utah (Figure 2). With a few exceptions, the states with the largest differences by sex are those with lower life expectancy at birth, while the smallest sex differences are found mostly among states with higher life expectancy.

Figure 2

Difference between male and female life expectancy at birth: Each state, District of Columbia, and United States, 2021. NOTE: The color key reflects age ranges of life expectancy at birth for the total population for each area. SOURCE: National Center (more...)

Table B shows life expectancy at age 65 for the total, male, and female populations for the 50 states, D.C., and United States. In 2021, Hawaii ranked first for the total, male, and female populations, with life expectancy at age 65 of 20.6, 18.9, and 22.2 years, respectively. Mississippi ranked 51st, with the lowest life expectancy among the 50 states and D.C. for the total and male populations, with life expectancy at age 65 of 16.1 and 14.6, respectively. West Virginia ranked 51st for females, with life expectancy at age 65 of 17.3 years. In comparison, life expectancy at age 65 for the entire United States was 18.4, 17.0, and 19.7 for the total, male, and female populations, respectively. Figure 3 shows that states with the lowest life expectancies at age 65 are mostly concentrated in the South, with Florida being a noted exception, and those with the highest life expectancies are mostly in the West and Northeast.

Table B

Life expectancy at age 65, rank, and standard error, by sex: Each state, District of Columbia, and United States, 2021.

Figure 3

Life expectancy at age 65: Each state, District of Columbia, and United States, 2021. SOURCE: National Center for Health Statistics, National Vital Statistics System, mortality data file.

From 2020 to 2021, life expectancy at birth declined for 39 states (Table C, Figure 4) (5). The declines ranged from 0.1 to 2.1 years. Life expectancy increased for 11 states, with increases ranging from 0.1 to 1.5 years. The states with the greatest decreases in life expectancy at birth from 2020 to 2021 included Alaska, West Virginia, New Mexico, Florida, Oklahoma, Oregon, and Tennessee. The states that experienced increases in life expectancy included New Jersey, New York, Connecticut, North Dakota, Massachusetts, Maryland, Rhode Island, Illinois, Iowa, Wisconsin, and Nebraska. Life expectancy did not change for D.C. Overall, life expectancy in the United States declined by 0.6 years from 2020 to 2021, mostly due to the COVID-19 pandemic and increases in unintentional injuries (mainly drug overdose deaths) (1).

Table C

Change in life expectancy at birth: Each state, District of Columbia, and United States, from 2020 to 2021.

Figure 4

Change in life expectancy at birth from 2020 to 2021: Each state, District of Columbia, and United States, 2021. SOURCE: National Center for Health Statistics, National Vital Statistics System, mortality data file.

Data for calculating life table functions

The data used to prepare the U.S. state life tables (Table I) include state-specific final death counts from the National Vital Statistics System, state-specific population estimates from the U.S. Census Bureau, and state-specific death and population counts for Medicare beneficiaries ages 66–99 from the Centers for Medicare & Medicaid Services.

Table I

Complete period life tables: 50 states and District of Columbia, 2021. Available from: https://ftp.cdc.gov/pub/Health_Statistics/NCHS/Publications/NVSR/73-07/

Preliminary adjustment of data

Interpolation of P_x and D_x

Anomalies—both random and those associated with reporting age at death—can be problematic when using vital statistics and census data by single years of age to estimate the probability of death (2, 7). Graduation techniques are often used to eliminate these anomalies and to derive a smooth curve by age. Beers’ ordinary minimized fifth difference formula is used to obtain smoothed values of population counts (P_x) and death counts (D_x) from 5-year age groupings of _nP_x from ages 0–99 and _nD_x from ages 5–99, and where _n D_x has first been adjusted for not-reported age on the death certificate (see reference 8 for details on the application of Beers’ method). Beers’ interpolation is not applied to deaths at ages 0–4.

For states with zero death counts in the age range 1–4 years, those counts needed to be replaced using linear interpolation; otherwise, zero death counts would have resulted in discontinuation of the age-specific mortality distribution. In a few other cases, application of Beers’ interpolation of deaths in the age range 6–10 resulted in zero or negative death counts because of very small numbers of deaths, so linear interpolation was also applied. The assumption of linearity is warranted because mortality declines somewhat linearly between ages 1 and 10 or so, and the results led to smooth age patterns of mortality (see Table II for a list of states and ages where linear interpolation was used).

Table II

Application of linear interpolation by area, sex, and age.

Probabilities of dying at oldest ages

Medicare data are used to supplement vital statistics data for the estimation of q_x at the oldest ages because these data are more accurate, given that proof of age is required for enrollment in the Medicare program. Medicare data are used here to estimate the probability of dying for ages 66–99.

For this method, these steps are followed: First, vital statistics and Medicare death rates are blended in the age range 66–99. Second, a logistic model is used to smooth the blended death rates in the age range 85–99 and to predict death rates for ages 100–120. Third, final resulting death rates, M_x, are converted to probabilities of dying, q_x .

For the national life tables, vital statistics, $M_x^V$, and Medicare, $M_x^M$, death rates are blended in the age range 66–94 with a weighting process that gives gradually declining weight to vital statistics data and gradually increasing weight to Medicare data. For ages 95–99, $M_x^M$ is used exclusively. Due to the unavailability of census state population estimates for ages 85–100, calculating $M_x^V$ for this age span is not possible. As a result, the blending technique was modified such that $M_x^V$ and $M_x^M$ are blended in the age range 66–84, and $M_x^M$ is used exclusively in the age range 85–99. Blended M_x is obtained as:

when x = 85,…,99.

$M_x^M$ is estimated as:

The exclusive use of Medicare death rates beginning at age 85 for the state life tables is expected to have a negligible biasing effect on mortality at older ages in the life tables compared with the national life tables. As Figures I–III show, while large differences are found between Medicare and vital statistics death rates at ages 85 and older for the U.S. population, blended Medicare and vital statistics death rates are very similar to Medicare death rates for ages 85 and older.

Figure I

Age-specific vital statistics, Medicare, and blended death rates for total population: United States, 2021. ¹Rate (M_x) is defined as the number of deaths during the year divided by the midyear population (D_x/P_x). SOURCE: National Center for Health Statistics, (more...)

Figure II

Age-specific vital statistics, Medicare, and blended death rates for male population: United States, 2021. ¹Rate (M_x) is defined as the number of deaths during the year divided by the midyear population (D_x/P_x). SOURCE: National Center for Health Statistics, (more...)

Figure III

Age-specific vital statistics, Medicare, and blended death rates for female population: United States, 2021. ¹Rate (M_x) is defined as the number of deaths during the year divided by the midyear population (D_x/P_x). SOURCE: National Center for Health Statistics, (more...)

A logistic model proposed by Kannisto is then used to smooth M_x in the age range 85–99 and to predict M_x in the age range 100–120 (8). The start of the modeled age range varies by sex because it is a function of the age at which the rate of change in the age-specific death rates peaks. In current times, the rate of change in the age-specific death rate rises steadily up to about ages 80–85 and then begins to decline. As a result, modeling a large age span such as 65–100 with one simple model is difficult without oversmoothing and consequently altering the underlying mortality pattern observed in the population of interest (9). Further, the observed data for the age range 65–85 or so is reliable and robust, as indicated by the very close similarity between vital statistics and Medicare death rates, making it unnecessary to model, or smooth, the entire age span (65–100).

The Kannisto model is a simple form of a logistic model in which the logit of u_x (or the natural log of the odds of u_x) is a linear function of age x ({8}). It is expressed as:

Because u_x is not directly observed but is closely approximated by m_x, and m_x=M _x, then the logit of M_x is modeled instead. A maximum-likelihood generalized linear model estimation procedure is used to fit the following model in the age range 85–99:

Finally, the predicted probability of death, ${\overline{q}}_x$, for ages 85–120 is estimated by converting ${\overline{M}}_x$ as:

The probability of death is extrapolated to age 120 to estimate the life table population until no survivors remain. This information is then used to estimate L_x for ages 100–120, which is used to close the table with the age category 100 and older, combined (see following discussion).

Figures IV–VI show the age-specific probability of dying, q_x, estimates for each of the 50 states and D.C. compared with the values for the United States in 2021. The observed probabilities for the states and D.C. are shown as circles, which appear as vertical bars where they overlap, and the U.S. probabilities are shown as an intersecting connected line. The state estimates fall about the U.S. values as expected, with a few outliers in the youngest childhood ages. These few cases are predominantly the result of a very small number of deaths, consistent with very low mortality in this age range, combined with very small populations in states such as Vermont, Wyoming, and North Dakota. Overall, age-specific estimates for the 50 states and D.C. follow the expected age pattern of mortality and are consistent with the mortality pattern observed for the entire United States.

Figure IV

Age patterns of mortality for states and District of Columbia compared with United States, 2021. SOURCE: National Center for Health Statistics, National Vital Statistics System, mortality data file.

Figure V

Male age patterns of mortality for states and District of Columbia compared with United States, 2021. SOURCE: National Center for Health Statistics, National Vital Statistics System, mortality data file.

Figure VI

Female age patterns of mortality for states and District of Columbia compared with United States, 2021. SOURCE: National Center for Health Statistics, National Vital Statistics System, mortality data file.

Calculation of remaining life table functions for all groups

Variances and standard errors of probability of dying and life expectancy

The mortality data on which the life tables are based are not affected by sampling error because the data are based on complete counts of deaths, and, as a result, variances and standard errors reflect only random variation. While measurement errors such as age misreporting are known to affect mortality estimates, they are not considered in the calculation of the variances or standard errors of the life table functions. Because the state life tables presented in this report are based on relatively large numbers of deaths, the variances and standard errors presented are rather small.

The methods used to estimate the variances of q_x and e_x are based on Chiang (10) with some necessary modifications due to the use of statistical modeling for smoothing and prediction of older-age death rates. Based on the assumption that deaths are binomially distributed, Chiang proposed the following equation for the variance of q_x :

where D_x is the age-specific death count. This equation is used to estimate Var  (q_x) throughout the age span with a modification where, for ages younger than age 66, D_x is the deaths from vital statistics data, smoothed by interpolation and adjusted for the number of deaths with age not stated. For ages 66 and older, D_x is obtained by treating the population as a cohort population and calculated from q_x because blended vital statistics and Medicare data were used for estimation (11):