What Is The States of Change Project?

Ronald Brownstein
Add to Briefcase
See more stories about...
Ronald Brownstein
Feb. 3, 2015, 7:56 p.m.

This series on the key states for the 2016 pres­id­en­tial elec­tion is based on his­tor­ic­al data and fu­ture pro­jec­tions provided to Next Amer­ica by States of Change: Demo­graph­ics and Demo­cracy. The States of Change pro­ject is a col­lab­or­a­tion sup­por­ted by the Hew­lett Found­a­tion that brings to­geth­er the Cen­ter for Amer­ic­an Pro­gress, the Amer­ic­an En­ter­prise In­sti­tute, and demo­graph­er Wil­li­am Frey of the Brook­ings In­sti­tu­tion. The pro­ject is dir­ec­ted by Ruy Teixeira of the Cen­ter for Amer­ic­an Pro­gress and Karlyn Bow­man of the Amer­ic­an En­ter­prise in­sti­tute.

In a state­ment to Next Amer­ica, the States of Change Pro­ject ex­plained its pro­gram as fol­lows:

The goals of the pro­ject are: (1) to doc­u­ment and ana­lyze the chal­lenges to demo­cracy posed by the rap­id demo­graph­ic evol­u­tion of the United States, from the 1970’s to the year 2060; and (2) to pro­mote a wide-ran­ging and bi­par­tis­an dis­cus­sion of Amer­ica’s demo­graph­ic fu­ture and what it por­tends for the polit­ic­al parties and the policy chal­lenges they (and the coun­try) face.

Res­ults from the first year of the pro­ject in­clude:

“¢ Trend ana­lys­is of 40 years of demo­graph­ic change in United States, 1974-2014, na­tion­ally and in every state, par­tic­u­larly as it has af­fected the pool of eli­gible voters

“¢ Pro­jec­tions of the ra­cial com­pos­i­tion of every state to the year 2060, both over­all and by eli­gible voters

These find­ings are be­ing re­leased in sev­er­al ways, first through this col­lab­or­a­tion with Na­tion­al Journ­al/Next Amer­ica. There will also be a de­tailed re­port dis­cuss­ing our na­tion­al and state res­ults over the en­tire 1974-2060 peri­od and an in­ter­act­ive web fea­ture that will al­low users to trace the demo­graph­ic evol­u­tion of any state’s elect­or­ate by a vari­ety of dif­fer­ent char­ac­ter­ist­ics. The re­port and in­ter­act­ive fea­ture will be re­leased on or around Feb­ru­ary 24, when a pub­lic con­fer­ence will be held at AEI to present and dis­cuss the pro­ject’s find­ings.

All the his­tor­ic­al data used in these are taken from the Census Bur­eau’s monthly Cur­rent Pop­u­la­tion Sur­vey. States of Change ana­lyzed the data for each rel­ev­ant year from the Voter Sup­ple­ment of the Novem­ber sur­vey. The stor­ies ex­am­ine two sets of res­ults. The first is eli­gible voters-that is, the share of the state pop­u­la­tion that is over 18 and a cit­izen in each demo­graph­ic cat­egory. The second is ac­tu­al voters, the share of the state pop­u­la­tion in each demo­graph­ic cat­egory that ac­tu­ally voted, ac­cord­ing to the Census sur­vey.

In ana­lyz­ing the his­tor­ic res­ults, the States of Change pro­ject sub­jec­ted the data to a stat­ist­ic­al pro­cess known as LOWESS. The pro­ject ex­plains that widely used stat­ist­ic­al pro­cess this way:

The es­tim­ates presen­ted here for eli­gible pop­u­la­tions were pro­duced us­ing LOWESS, a stat­ist­ic­al pro­ced­ure that ‘smooths’ our CPS and pro­jec­tion data in or­der to gen­er­ate more stable and ac­cur­ate res­ults. As with all sur­veys, res­ults from the CPS are sub­ject to ran­dom sampling er­ror and any par­tic­u­lar year’s es­tim­ates are likely to de­vi­ate from the true value. ‘Smooth­ing’ over these data points for char­ac­ter­ist­ics whose rates of change should be fairly con­sist­ent across elec­tions ul­ti­mately cre­ates bet­ter es­tim­ates for any par­tic­u­lar year.

The CPS sur­vey gen­er­ally paints a sim­il­ar pic­ture of the elect­or­ate each year as the exit polls con­duc­ted on Elec­tion Day for a con­sor­ti­um of me­dia or­gan­iz­a­tions. The CPS re­cently has found a slightly high­er white share of the vote than exit polls, though both have fol­lowed a sim­il­ar track of a de­clin­ing per­cent­age for whites. While the two data sources con­verged on the white share of the na­tion­al vote in 1996 (83 per­cent) and 2000 (81 per­cent), in each of the past three pres­id­en­tial elec­tions the Census has found the white share two per­cent­age points high­er than the exit polls did; in 2012, the Census put whites at 74 per­cent of the na­tion­al vote, com­pared to 72 per­cent for the exit.

One dif­fer­ence between the two data sets re­mains lar­ger. Com­pared to the CPS res­ults, the exit polls con­sist­ently find that col­lege-edu­cated voters rep­res­ent a great­er share of the elect­or­ate. For in­stance, in Ohio in 2008, the exit poll found that col­lege-edu­cated whites equaled 33 per­cent of the elect­or­ate; the CPS put their share at 26 per­cent. In Vir­gin­ia, that year the exit poll found that col­lege-edu­cated whites equaled 40 per­cent of voters; the CPS put their share at 33 per­cent.

The States of Change pro­ject has chosen to use the CPS data, so that is what Next Amer­ica is us­ing in these re­ports on the pro­ject’s work. (The Census data is the only avail­able source on the eli­gible voter pop­u­la­tion; the exit polls only pro­duce an as­sess­ment of the ac­tu­al voters.) Next Amer­ica has used both meas­ure­ments of the elect­or­ate’s com­pos­i­tion in the past, and will con­tin­ue to do so in the fu­ture.

While the Census does ask about vot­ing par­ti­cip­a­tion, it does not ask about vot­ing pref­er­ences. All the his­tor­ic data ref­er­enced in these stor­ies about the vot­ing pref­er­ences of the dif­fer­ent groups in the vari­ous states comes from the Elec­tion Day exit polls. The name of the in­sti­tu­tion con­duct­ing the exit poll has changed over time; in 2012 it was con­duc­ted by Edis­on Re­search.

In ex­amin­ing the res­ults by edu­ca­tion in these stor­ies, Next Amer­ica is re­port­ing the trends solely for whites with and without at least a four-year col­lege de­gree, be­cause that line of edu­ca­tion­al at­tain­ment has proven a sig­ni­fic­ant fault-line in vot­ing be­ha­vi­or among whites. The gap is not nearly as sig­ni­fic­ant in the vot­ing be­ha­vi­or of minor­it­ies with and without four-year de­grees.

The States of Change pro­ject is unique be­cause it has not only col­lec­ted ret­ro­spect­ive data but uses a demo­graph­ic mod­el con­struc­ted by Wil­li­am Frey of the Brook­ings In­sti­tu­tion to fore­cast the fu­ture shape of the elect­or­ate, by state, through 2060. The pro­ject ex­plained its meth­od­o­logy for mak­ing its pro­jec­tions of each state’s over­all pop­u­la­tion in this state­ment:

The pro­jec­tions em­ploy a multistate co­hort com­pon­ent meth­od­o­logy which be­gins with the 2010 census and pro­jects ahead in five year in­ter­vals, race and age-spe­cif­ic pop­u­la­tions for each state to 2060 based on the com­pon­ents of do­mest­ic mi­gra­tion, in­ter­na­tion­al mi­gra­tion, fer­til­ity and mor­tal­ity fol­low­ing from mod­el­ing put forth in An­drei Ro­gers,In­tro­duc­tion to Mul­tire­gion­al Math­em­at­ic­al Demo­graphy (New York: Wiley, 1975) These pro­jec­tions are per­formed sep­ar­ately for ra­cial groups wherein each state’s do­mest­ic mi­gra­tion flows are pro­jec­ted between that state and the re­mainder of the four census re­gions (North­east, Mid­w­est, South West). In­ter­na­tion­al mi­gra­tion to the US for each in­ter­val is al­loc­ated to states and re­gions. In both cases, these mi­gra­tion flows and im­mig­raton al­loc­a­tions are based on pat­terns re­cor­ded in the 2007-2012 mul­ti­year Amer­ic­an Com­munity Sur­vey. Race spe­cif­ic fer­til­ity and mor­tal­ity rates for each state as­sume na­tion­al rates spe­cif­ic to age and race. These as well as na­tion­al im­mig­ra­tion levels are broadly con­sist­ent with those used in US Census Bur­eau na­tion­al pro­jec­tions.

In ad­di­tion, the Pro­ject ex­plains its meth­od­o­logy for fore­cast­ing the eli­gible voter pop­u­la­tion this way:

Like all of the data presen­ted after 2014, the eli­gib­il­ity rates for these dif­fer­ent pop­u­la­tions are pro­jec­tions. These were gen­er­ated by tak­ing data from the Amer­ic­an Com­munity Sur­vey and di­vid­ing up the Amer­ic­an pop­u­la­tion in­to groups based on their state, race, and age. Mul­ti­level stat­ist­ic­al mod­els were then used to es­tim­ate the eli­gib­il­ity rates and the nat­ur­al­iz­a­tion rates for each state/race/age group. These groups were then tracked for­ward in time and had those unique nat­ur­al­iz­a­tion rates ap­plied to them. In ad­di­tion, these es­tim­ates ac­count for im­mig­ra­tion in­to the state and the res­ult­ing ef­fect it had on dif­fer­ent groups’ over­all eli­gib­il­ity rates. The end res­ult is a pro­ced­ure that is sens­it­ive to the dif­fer­ent rates of nat­ur­al­iz­a­tion ex­per­i­enced by each of these groups and the im­mig­ra­tion each state is pre­dicted to ex­per­i­ence in the fu­ture.

What We're Following See More »
Manafort Case Moves to Closing Arguments
2 days ago
Manafort Defense Rests
3 days ago
Judge Holds Witness in Contempt in Manafort Case
1 weeks ago

"A federal judge has found a witness in contempt for refusing to testify before the grand jury hearing evidence in special counsel Robert S. Mueller III’s investigation of Russian interference in the 2016 presidential election. U.S. District Chief Judge Beryl Howell made the ruling Friday after a sealed hearing to discuss Andrew Miller’s refusal to appear before the grand jury. Miller is a former aide to longtime Trump confidant Roger Stone."

Gates Says He Committed Crimes with Manafort
1 weeks ago

Paul Manafort's former business partner Rick Gates said in court today that "he conspired with Manafort to falsify Manafort’s tax returns. Gates said he and Manafort knowingly failed to report foreign bank accounts and had failed to register Manafort as a foreign agent."

Gates to Be Called Next in Manafort Case
1 weeks ago

Welcome to National Journal!

You are currently accessing National Journal from IP access. Please login to access this feature. If you have any questions, please contact your Dedicated Advisor.