Mid-2011 Population Estimates to be released on Wednesday

[muon2]

I used the July 2011 estimates and the April 2010 Census base to get an annual growth rate. This correctly accounts for the 15 month period between the Census and the estimate. I then applied the annual growth rate to the 2010 reapportionment population to get the 2020 projection. This accounts for the extra overseas population used in reapportionment but not for redistricting. Ten years is a long stretch for a simple model like this, but here are the projected changes.

CA +1
CO +1
FL +1
IL -1
MI -1
MN -1
NY -1
NC +1
OH -1
PA -1
RI -1
TX +3
VA +1
WV -1

The bubble seats in this projection are based on the last five awarded and the next five in line.
The last five awarded are CO-8, AL-7, VA-12, CA-54, and FL-28 (#435).
The next five in line are WV-3, OR-6, NY-27, AZ-10, LA-7.

[muon2]

I wonder how this table (http://en.wikipedia.org/wiki/2010_United_States_Census#State_rankings) will change in the 2020 Census.

Here is the projected 2020 Census Population, based on current trends (last year's growth/loss x 8.75, added to the Mid-2011 population):

SORTED BY 2020 POPULATION:



SORTED BY NUMERICAL GROWTH/LOSS:



SORTED BY PERCENTAGE GROWTH/LOSS:

Your formula isn't very accurate since it assumes that the numerical growth stays fixed over the decade, when the projection should be based on a constant rate of growth. If you want to assume constant numerical growth the factor of 8.75 is not correct. Since the original period is 1.25 years the the remaining period is 8.75 but that is a factor of 8.75/1.25 = 7 times the amount of growth. By using 8.75 instead of 7 you overestimate the final population.

In reality a constant rate of growth compounds with each year. For example assume that a population of 1000 has an annual growth rate of 1% or 10. A constant numerical growth gives a population of 1100 at the end, while a constant rate of growth gives 1105 after 10 years. The effect is greater as the rate increases, so your method will have the fast growing states grow less compared to the slow growing states.

To get the correct projections including the effects of compounding you need to use the financial functions in the spreadsheet to get the rate and then the new future value. The functions and format depend on the spreadsheet syntax. The result can also be obtained directly from the math, using rate = (current/census)^(1/period)-1 where period = 1.25, and projection = census*(1+rate)^(decade) where decade = 10.

[muon2]

I wonder how this table (http://en.wikipedia.org/wiki/2010_United_States_Census#State_rankings) will change in the 2020 Census.

Here is the projected 2020 Census Population, based on current trends (last year's growth/loss x 8.75, added to the Mid-2011 population):

(charts)

Your formula isn't very accurate since it assumes that the numerical growth stays fixed over the decade, when the projection should be based on a constant rate of growth. If you want to assume constant numerical growth the factor of 8.75 is not correct. Since the original period is 1.25 years the the remaining period is 8.75 but that is a factor of 8.75/1.25 = 7 times the amount of growth. By using 8.75 instead of 7 you overestimate the final population.

In reality a constant rate of growth compounds with each year. For example assume that a population of 1000 has an annual growth rate of 1% or 10. A constant numerical growth gives a population of 1100 at the end, while a constant rate of growth gives 1105 after 10 years. The effect is greater as the rate increases, so your method will have the fast growing states grow less compared to the slow growing states.

To get the correct projections including the effects of compounding you need to use the financial functions in the spreadsheet to get the rate and then the new future value. The functions and format depend on the spreadsheet syntax. The result can also be obtained directly from the math, using rate = (current/census)^(1/period)-1 where period = 1.25, and projection = census*(1+rate)^(decade) where decade = 10.

I know very well what you want to say, but after only 1.25 years since the Census, I found it more convinient to just multiply the past year's growth by 8.75 to get the 2020 Census numbers. There are just too many variables that make it impossible to predict the population 8 years from now. Like I said above, the economy could grow at a faster pace in the next years, which could lead to higher immigration balances again from other parts of the world and/or the natural increase could be higher by 300K people by the end of the decade. I could create 10 different models right now, one based on numerical projections, one based on percentage growth etc. etc. It will only be in the years 2017/18/19 when these projections would get somewhat more accurate ...

But why 8.75? Is it based on any approximation, or just a guess? If a guess what motivates it?

Even if I take your approximation on your spreadsheet I can't duplicate the results. For example the Census report had IL grow by 38625 last year. If I multiply that by 8.75 I get 337969. If I add that to the 2011 midyear estimate of 12869257, I get 13207226 which is quite a bit larger than the number you show. If I derate the growth by a factor of 1.25 and then multiply by 8.75 and add to the midyear total I get 13139632. That still doesn't match the value in your chart.

Can you tell me where I'm off?

[muon2]

Would be very interesting to add a column to that sheet detailing the number of Congressional seats. Someone I know has an excel template that calculates it quickly.

I posted that already.

I used the July 2011 estimates and the April 2010 Census base to get an annual growth rate. This correctly accounts for the 15 month period between the Census and the estimate. I then applied the annual growth rate to the 2010 reapportionment population to get the 2020 projection. This accounts for the extra overseas population used in reapportionment but not for redistricting. Ten years is a long stretch for a simple model like this, but here are the projected changes.

CA +1
CO +1
FL +1
IL -1
MI -1
MN -1
NY -1
NC +1
OH -1
PA -1
RI -1
TX +3
VA +1
WV -1

The bubble seats in this projection are based on the last five awarded and the next five in line.
The last five awarded are CO-8, AL-7, VA-12, CA-54, and FL-28 (#435).
The next five in line are WV-3, OR-6, NY-27, AZ-10, LA-7.

[muon2]

Thanks, I was using the Census spreadsheet from their first link, and you were using the second link.

I still think that with the power and ease of a spreadsheet one ought to at least estimate the compounding effects. It matters for estimating seats in 2020, since adds more to the fast growing states.

[muon2]

2 other ways of projecting the 2020 population would be constant relative population growth rates of either the past year (Mid-2010 to Mid-2011) or the past 1.25 years, projected to 2020:

Scenario A: April 1, 2010 -> July 1, 2011 growth rate: 0.7368565%
Scenario B: July 1, 2010 -> July 1, 2011 growth rate: 0.7311597%

In scenario A, the April 1, 2020 population for the US would be 332.265.030
In scenario B, the April 1, 2020 population for the US would be 332.100.653

This has to do with the fact that the 1/4 year between April 1, 2010 to July 1, 2010 had slightly higher relative growth (0.7596469%) than the year that followed.

I prefer scenario A since it averages a longer period. If there are no seasonal effects in the estimate, it should provide a slightly better projection.

[muon2]

I used the July 2011 estimates and the April 2010 Census base to get an annual growth rate. This correctly accounts for the 15 month period between the Census and the estimate. I then applied the annual growth rate to the 2010 reapportionment population to get the 2020 projection. This accounts for the extra overseas population used in reapportionment but not for redistricting. Ten years is a long stretch for a simple model like this, but here are the projected changes.

CA +1
CO +1
FL +1
IL -1
MI -1
MN -1
NY -1
NC +1
OH -1
PA -1
RI -1
TX +3
VA +1
WV -1

The bubble seats in this projection are based on the last five awarded and the next five in line.
The last five awarded are CO-8, AL-7, VA-12, CA-54, and FL-28 (#435).
The next five in line are WV-3, OR-6, NY-27, AZ-10, LA-7.

I doubt the methodology here. Census 2010 had horrible undercount problems in some areas. Comparing July 2010 estimates to July 2011 estimates I think is the more apt comparison.

Obama politicizes everything. My assumption was that he would run Census 2010 to net Democrats more seats. What really happened, was that he hired enumerators, trained them, and laid them off, only to rehire them, and retrain them. He did that to increase the number of "jobs" he created. The problem was those enumerators were really needed in areas such as Queens. The Census was a mess in NYC, and, the final count was below estimates derived by statistical counting models.

If the Republicans hold the New York Senate this election, one of the reasons will be Obama's incompetence in adminsistering the Census.

July 1 estimates from the Census used here are based on the April 1, 2010 Census according to the Bureau. So if the April 2010 values are suspect so would any analysis based only on the July 2010 to July 2011 estimates. In any case, I want to model the Census methodology since that method, including any undercount, is the best way to project reapportionment counts.