Census Estimates for 2009 -> 2010 Apportionment

[muon2]

The Census Bureau released its new estimates for the population of the states as of July 1, 2009. As in past years I have used that data to project the April 1, 2010 apportionment populations. This requires finding the population growth in the resident population for each state, then applying that to the apportionment population. In this method I assume that the growth until April 1, 2010 will match the uniform growth from April 1, 2000 to July 1, 2009.

One special circumstance is to account for the effect of hurricane Katrina. LA saw an estimated drop of 250 K in the 10 months following the hurricane. If I used the normal methodology, that would project a continued decline through 2010. Instead, for LA I took the percentage growth from July 1, 2006 to July 1, 2009 then applied that to the estimate for July 1, 2006 to get April 1, 2000. To this I added the difference between the resident and apportionment populations in 2000 to reach a projected apportionment population for 2010.

Based on this projection, the following adjustments would be required to reapportion the seats in 2010:

AZ +2
FL +1
GA +1
IL -1
IA -1
LA -1
MA -1
MI -1
MN -1
NV +1
NJ -1
NY -1
OH -2
PA -1
SC +1
TX +3
UT +1
WA +1

Compared to last year's projection this is a shift of two seats from FL and TX to MO and WA. The last states awarded seats were NY 28 (431), WA 11 (432), AZ 10 (433), MO 11 (434) and CA 53 (435). These seats are on the bubble and most at risk to fluctuations in growth during the time from the estimate last summer until Census Day. The effects the the recession damping growth in states like AZ are particularly important. I'll post an alternate analysis to look at that effect later in this thread.

The next five seats would go to FL 27 (436), TX 36 (437), MN 8 (438), OR 6 (439), and NC 14 (440). These are the bubble states on the other side, and can benefit from fluctuations in their favor.

[Devilman88]

What type of growth would NC have to see to gain an EV?

[muon2]

The recession has fallen late in the decade, and it has had a particularly hard impact on housing and relocation. Many states that had seen robust growth for most of the decade have seen that growth slow down substantially. I have made a simple model for the effect of the recession on projections for apportionment. These can be compared to my decade-long baseline noted at the top of the thread.

For this model I have taken the estimated populations for July 1, 2007 and July 1, 2009 as released today and used that two-year period to calculate an annual rate of growth. I then applied that rate for 3/4 of a year on the July 1, 2009 data to get a base projection. To account for military and other overseas residents counted for apportionment I added the difference from 2000 to the 2010 projection.

Based on this projection, the following adjustments would be required to reapportion the seats in 2010:

AZ +1
FL +1
GA +1
IL -1
IA -1
LA -1
MA -1
MI -1
MN -1
NV +1
NJ -1
NY -1
OH -2
PA -1
SC +1
TX +4
UT +1
WA +1

Compared to my conventional model, this projection shifts a seat from AZ to TX. In this model the last states awarded seats were NY 28 (431), WA 11 (432), CA 53 (433), MO 11 (434) and TX 36 (435). The next five seats would go to MN 8 (436), OR 6 (437), AZ 10 (438), NC 14 (439), and FL 27 (440). They are actually fairly close in their outcomes with the same states on the bubble under both models.

[muon2]

What type of growth would NC have to see to gain an EV?

Depending on the growth model for the other states, NC would need between 40 K to 60 K more than is currently projected.

[Devilman88]

What type of growth would NC have to see to gain an EV?

Depending on the growth model for the other states, NC would need between 40 K to 60 K more than is currently projected.

So, we still have a chance at getting our 14th seat...

[Tender Branson]

One thing that is to note:

Between 1990 and 2000, the US population grew by 33 Mio., of which 17 Mio. was natural increase (births minus deaths) and 16 Mio. was migration gain from other countries.

In the last year from July 1, 2008 to July 1, 2009 the net gain was 2.63 Mio. with 1.78 Mio. natural increase and 0.85 Mio. migration gain from other countries.

This means that the international migration gain dropped by more or less 50% during this decade, barring any over- and undercounts in 1990 and 2000.

[Brittain33]

Thank you for doing this and sharing it.

[muon2]

Here's the map from my first set of projections in Atlas form. Feel free to use it for discussions about 2012, 2016 or 2020.

[CARLHAYDEN]

Lets compare that to the 2008 vote for President:

                              McCain          Obama

Growth                      5                    3

Loss                          1                    9

[Torie]

Well by my count:

   McCain   McCain   Obama   Obama
AZ +1   1         
FL +1                 1   
GA +1   1         
IL -1                          -1
IA -1                          -1
LA -1      -1      
MA -1                          -1
MI -1                          -1
MN -1                          -1
NV +1                 1   
NJ -1                         -1
NY -1                         -1
OH -2                         -2
PA -1                         -1
SC +1   1         
TX +4   4         
UT +1   1         
WA +1                 1   

McCain   8   Obama   3   
McCain   -1   Obama   -10   

[CJK]

Any particular insight on why Ohio is the only one losing 2 seats?

[collegebookworm]

Any particular insight on why Ohio is the only one losing 2 seats?

Of the slowest-growing states from the 2000 to 2009 Census estimates (Ohio ranks 6th slowest), Ohio has the largest population.

Take note of the following data:

Slowest-growing states : 2000-2009 avg. growth rate (2009 pop) - last seat gained (projected change from current apportionment)
1. Michigan: 0.0161% (9,969,727) - 424 (-1)
2. Rhode Island: 0.0261% (1,053,209) - 418 (0)
3. Louisiana: 0.0573% (4,492,076) - 377 (-1)
4. West Virginia: 0.0786% (1,819,777) - 417 (0)
5. North Dakota: 0.0974% (646,844) - 34 [guaranteed seat] (0)
6. Ohio: 0.1736% (11,542,645) - 416 (-2)

Michigan nearly loses a second seat (its population is slightly smaller than Ohio's).  All the other states have less than half the population of Ohio, so it should be expected that they lose no more than 1/2 as many seats as Ohio.

[collegebookworm]

My projection:

• +4: Texas
• +1: Arizona, Florida, Georgia, Nevada, South Carolina, Utah, Washington
• -1: Illinois, Iowa, Louisiana, Massachusetts, Michigan, Minnesota, New Jersey, New York, Pennsylvania
• -2: Ohio

My Excel spreadsheet (Google Docs)

[muon2]

My projection:

• +4: Texas
• +1: Arizona, Florida, Georgia, Nevada, South Carolina, Utah, Washington
• -1: Illinois, Iowa, Louisiana, Massachusetts, Michigan, Minnesota, New Jersey, New York, Pennsylvania
• -2: Ohio

My Excel spreadsheet (Google Docs)

Your spreadsheet doesn't show the formulas for your 2010 estimates. It looks like you are only using the 2008-2009 rate, but even then I can't get the same value that you do. Are you using future value formulas to get the exponential growth for the 3/4 of a year that remain from 7/1/09 to 4/1/10?

The last seat is quite sensitive to the rate you apply, particularly for AZ. The single year rates fluctuate a lot for the states, even for years before the recession. The recession knocked the estimated rate in AZ from 2.7% to 2.2% to 1.5% over the last three years. The estimates had TX fairly close to 2.0% all three of those years. If AZ picks up to a rate like 2.5% for the final nine months as housing starts to pick up, then it passes TX to gain the seat that MN is losing. If MN regains its 0.8% rate from 2007 for the last nine months then MN keeps the seat instead.

[collegebookworm]

Your spreadsheet doesn't show the formulas for your 2010 estimates. It looks like you are only using the 2008-2009 rate, but even then I can't get the same value that you do. Are you using future value formulas to get the exponential growth for the 3/4 of a year that remain from 7/1/09 to 4/1/10?

The last seat is quite sensitive to the rate you apply, particularly for AZ. The single year rates fluctuate a lot for the states, even for years before the recession. The recession knocked the estimated rate in AZ from 2.7% to 2.2% to 1.5% over the last three years. The estimates had TX fairly close to 2.0% all three of those years. If AZ picks up to a rate like 2.5% for the final nine months as housing starts to pick up, then it passes TX to gain the seat that MN is losing. If MN regains its 0.8% rate from 2007 for the last nine months then MN keeps the seat instead.

Hmm...I can see the formula.  Perhaps the formulas will be visible with downloading?

Well, anyways, I took the 2000-2009 rate and multiplied by a one-year exponential-bias rate (2009 contributes 1/2 the rate, 2008 contributes 1/4, etc.).  After that, I applied the rate to the "Census 2000" values from the Census estimate data to correct from June 2000-2009 to April 2000-2009 (of course, this will always leave error, since 3-month incremental estimates aren't available).

Of course, I know that, like many estimates, mine are subject to even tiny population shifts, so feel free to believe that my estimates are bs, while I'll think they're perfect until the 2010 numbers come out and I'll feel like a fool.

[collegebookworm]

My Excel spreadsheet (Google Docs)

Your spreadsheet doesn't show the formulas for your 2010 estimates.

Ah, OK...go to the VIEW menu and check "Show formula bar"

[cinyc]

How close is RI to losing its second seat and MT and DE to gaining one?

[muon2]

My Excel spreadsheet (Google Docs)

Your spreadsheet doesn't show the formulas for your 2010 estimates.

Ah, OK...go to the VIEW menu and check "Show formula bar"

Thanks, that helps. You have a fairly complicated way to estimate the rate for the missing 12 months. In your spreadsheet those 12 months would be the 3 months of the decade from 4/1/2000 to 6/30/2000 and the last 9 months of the decade from 7/1/2009 to 4/1/2010. Your function gives only 1/256 of the weight to the actual rate from 2000-2001 yet gives 3/4 of the rate to the last two years which makes some sense. I would note that gives Texas an unusual boost due to the high rate from the year of Katrina immigration (05-06), which really has no effect in the first or last year of the decade.

I use a simple interest rate calculation. The rate between an estimate and the base value is (est/base)^(1/years)-1. The years can be a fraction such as 9.25 representing the time from 4/2000 to 7/2009. This what I use for my "traditional" analysis, with some adjustments in the Gulf states for Katrina. My alternative analysis uses the rate formula with the 2007 estimate as base and 2 years for the time.

To get the 2010 projection I use a compound interest function est2009*(1+rate)^0.75, which covers the 9 months from the final estimate to the present. I do adjust for overseas population that will be used in apportionment, though that only changes priority this year, not the seats assigned.

[CARLHAYDEN]

The problem with any formula is that it presumes pretty much a continuation of the recent past.

Consider the case of Louisiana, whose population radically decreased as a result of a hurricane.

What if California has a series of major earthquakes in January of 2010?

[muon2]

The problem with any formula is that it presumes pretty much a continuation of the recent past.

Consider the case of Louisiana, whose population radically decreased as a result of a hurricane.

What if California has a series of major earthquakes in January of 2010?

I completely agree. That's why I try to factor out Katrina from the Gulf and look at the recession years carefully. There's no question that there may be unforeseen events or trends that have been missed. At this point we are almost 10 years since the last complete count and errors will creep in. A formula only provides a rational way of making a projection in the face of incomplete information.

[CARLHAYDEN]

The problem with any formula is that it presumes pretty much a continuation of the recent past.

Consider the case of Louisiana, whose population radically decreased as a result of a hurricane.

What if California has a series of major earthquakes in January of 2010?

I completely agree. That's why I try to factor out Katrina from the Gulf and look at the recession years carefully. There's no question that there may be unforeseen events or trends that have been missed. At this point we are almost 10 years since the last complete count and errors will creep in. A formula only provides a rational way of making a projection in the face of incomplete information.

I think your estimates are very good.

Its just that a number of states are within a close margin of gaining or losing a seat that a small change could be important.

Also, it seems to me that redistricting within a state with more than one CD will have approximately as much impact on the composition of the house ad reapportionment.

[muon2]

How close is RI to losing its second seat and MT and DE to gaining one?

By my projections RI has only about 30 K people to spare this decade. Their growth rate is near 0, so it seems very likely that they will lose a seat in 2020.

MT is only 30 K short of a seat, but their growth rate is slightly less than the national average. Until their rate gets over the average for a while, they are going to stay just a little short of  a second seat.

DE is about 110 K short of a second seat, so it isn't possible this decade. However, they are growing faster than the national average, and if that continues they could pick up a seat in 2020 or 2030.

[Joe Cooper]

California Dept of Finance estimates 1.5 million people more than the Census estimates.

http://ballotbox.governing.com/2009/12/california-give-or-take-15-million-people.html

[CARLHAYDEN]

California Dept of Finance estimates 1.5 million people more than the Census estimates.

http://ballotbox.governing.com/2009/12/california-give-or-take-15-million-people.html

As to the actual number, well, we will have better date following the 2010 census.

However, I would bet that the federal estimate is probably closer to reality than the state guesstimate.

[jimrtex]

Any particular insight on why Ohio is the only one losing 2 seats?

It barely had the 18th seat in 2000.

Losers in terms of fractional seats:

NY -1.777, but it had about 29.25 to 27.50, so it had a little extra to lose, and barely holds on for 28, including some luck in population distribution.  Could lose 2 in 2020.

MI -1.332 had 15.3 so will be almost perfect 14.

OH -1.228 just barely had 18th, so lost a whole seat and lost its finger tip grip on second, but will only lose 1 in 2020.

PA -1.213 was a very solid 19, and is still well above 17.5 (but could lose 2 in 2020)

IL -0.979 Very solid 19 to very solid 18, as IL is starting to open up gap on PA and OH.

NJ -0.722 Solid 13, to very solid 12+, probably won't lose another in 2020.

MA -0.500 cross over 9.5 to lose 10th seat.  Won't lose another in 2020.