STATS Indiana

Indiana's Public Data Utility

LogoGet even more workforce data at Hoosiers by the Numbers

Methodology for Indiana State and County Population Projections, 2010 to 2050

March 2012

Prepared by Matt Kinghorn, Demographer
Indiana Business Research Center, Kelley School of Business, Indiana University

Overview

This latest version of the Indiana population projections were produced using the cohort-component method, with the 2010 Census results by age and sex for Indiana and its counties serving as the benchmark population figures. This method projects individual age-sex cohorts forward while applying specific mortality and migration rates. In addition, fertility rates are applied to the appropriate female cohorts to generate the number of births during each projection period. These projections follow the standard format of five-year age cohorts carried over five-year projection periods. The five-year cohorts begin with the 0-4 age group and extend through 80-84, with the final age group including all individuals age 85 and above.

Limitations

There is always uncertainty surrounding population trends. Future events can be difficult to predict. For instance, economic conditions or land use decisions can shift migration patterns or medical breakthroughs can increase life expectancy. Migration patterns are particularly subject to large and sudden swings. This instability provides one potential source of error.

General assumptions about the future course of demographic trends are introduced into the projections in order to deal with these inevitable shifts over time. Ultimately, the quality of these projections depend largely on the accuracy of these assumptions. Those that miss the mark will result in projections that deviate from future population totals. The various rates of change incorporated into these projections are based on trends observed in the recent past—2005 to 2007 in the case of fertility and mortality and 2000 to 2010 for migration. The assumptions for each component of population change are discussed below.

Population projections are susceptible to two other sources of error which should be noted. First, projections for less populous areas will be subject to a greater relative margin of error than more populous areas. Projections for Indiana, for instance, will be more reliable than projections for one of its smaller counties. Second, for reasons discussed earlier, long-term projections will be more unreliable than near-term projections. It’s advised that data users give greater weight to projections up to 2025 or 2030.

Finally, it is important to note that these projections are strictly demographic, meaning that they rely exclusively on the recent trends and the assumptions discussed above. No economic, land use, or environmental assumptions are introduced.

Fertility

Age-specific fertility rates (ASFR) for the age groups 10-14 through 45-49 were calculated for each county using data from the Indiana State Department of Health (ISDH). The 2005 to 2007 average annual ASFR was then computed for the state and for each county in order to smooth year-to-year fluctuations.

An assumption was made that Indiana’s fertility rates will remain constant over the next 15 years before beginning to decline. The change in these rates were based upon the changes in fertility projected at the national level in a recent set of population projections produced by the U.S. Census Bureau (www.census.gov/population/projections/). To implement this assumption, total fertility rates (TFR) for the nation were calculated for the population age 10 to 49 for each projection base year beyond 2010. The projected change in TFR from one period to the next at the national level was then applied to Indiana and its counties. The TFR for the state and each county was then disaggregated into ASFRs for each five-year age group, 10-14 to 45-49, based upon the proportion of the TFR that each age group held in the 2010 base year.

Mortality

State and county-specific survival rates for each age-sex group were calculated by constructing abridged life tables with five-year age groups except for the 0-1, 1-4, and 85+ population segments. Life tables were calculated for 2005, 2006 and 2007 using vital statistics data provided by the ISDH and the intercensal population estimates from the Census Bureau. An average of the survival rates from 2005 to 2007 were used in the projections.  

Life expectancy and survival rates improved dramatically throughout the last century and these improvements can be expected to continue. In order to project increased survival rates, we relied on the changes in survival rates projected at the national level in the middle series of the 1999 to 2100 national population projections produced by the U.S. Census Bureau (www.census.gov/population/projections/). Changes in age-sex survival rates from one projection base year to the target year in the national projections were applied to the corresponding projection cycles for the Indiana survival rates.

Migration

Age and sex-specific migration rate estimates were calculated using a residual method based on survival probabilities. The first step in this approach is to construct life tables with recent data and calculate age-specific 10-year survival rates. These survival rates were then applied to the Census 2000 population counts in the appropriate age-sex groups to estimate an expected population for that same cohort in 2010. These expected population numbers approximate what the 2010 population would be in a given cohort if there was no migration over the 10-year period and deaths were the only source of population change. The difference between the expected population and the actual 2010 Census count becomes the net migration estimate.

This process is commonly referred to as the forward survival rate method. There is a similar procedure called the reverse survival rate method. Under this approach, the 2010 population counts are divided by the 10-year survival rates in order to calculate an expected population in 2000. Again, the difference between the expected and actual population figures for a given cohort is the net migration estimate. These two methods yield similar results but the differences tend to widen for older age groups. For that reason, the migration rates used in these projections are an average of the forward and reverse survival rate methods. State and county-specific life table survival rates were calculated using mortality data from ISDH and population figures from the Census Bureau.

Since migration patterns can shift quickly, they are the most unstable component of population projections. For example, Indiana had a strong net out-migration in the 1980s yet posted a net in-migration of more than 200,000 residents during the 1990s. Indiana had a net inflow of residents last decade, too, yet—at roughly 80,000 residents—the net in-migration was well below the level experienced during the 1990s. Recently, in response to difficult economic conditions, migration appears to have slowed even more. In fact, the Census Bureau’s vintage 2011 population estimates show that Indiana may have posted slight a net out-migration in 2010.  

Given this uncertainty, migration rates were reduced slightly for the initial projection cycle. This means that areas expected to have a net in-migration will have somewhat less of an inflow while out-migration areas will have less of an outflow. Migration rates were further reduced in subsequent projection cycles. Migration rates were reduced in a non-linear trend that causes larger increase or decreases to occur in early projection cycles before leveling off in later cycles.

Adjustments to the Base Population

Adjustments to the base population were made in counties where college student or prison populations account for a sizeable share of the total. These adjustments were necessary because these populations do not tend to age in place. In these cases, student or prison populations were removed from the base population before applying the cohort-component method and then added back after the projections were complete. Student or prisoner populations were held constant throughout the projection period.

In most cases, group quarters data by age, sex and facility type from the 2010 census were used to adjust the base populations. However, the group quarters data do not sufficiently cover the student populations in Delaware, Monroe, St. Joseph, Tippecanoe, Vanderburgh and Vigo counties because many students live off campus. In these cases, college enrollment data by age from the U.S. Department of Education were used to estimate the student populations. The counties with adjusted base populations were:  

  • Delaware
  • Grant
  • Hendricks
  • Henry
  • Huntington
  • Jasper
  • Jefferson
  • Knox
  • LaPorte
  • Madison 
  • Miami 
  • Monroe
  • Montgomery
  • Perry
  • Parke
  • Porter 
  • Putnam
  • St. Joseph
  • Steuben
  • Sullivan
  • Tippecanoe
  • Vanderburgh
  • Vigo
  • Wabash
  • Wayne