Re: Household Synthesis - Base Year

I basically agree with Paul. At SEMCOG, we are trying to create 2005 households as follows:

We have 2005 housing stock from parcel data, historical building permits, as well as everything else but households and their attributes, i.e. size, race, age, income, children, workers, and auto ownership. But we have regional control totals of households with these attributes (developed from REMI forecast outputs and PUMS) for every year from 2000 to 2035. Fortunately, we also have a 2005 household travel survey.

1) We estimated a household location choice model using the household travel survey and other 2005 data.
2) We synthesized 2000 households and created a usable 2000 base year dataset.
3) We created a development_events_exogenous table that includes all the new units added between 2000 and 2005, developed from building permits, year-built, etc..
4) We run UrbanSim, mainly the household location choice model, from 2000 to 2005 to "allocate" new households in the control totals table to the added housing units.

We are examining the results right now. Seems it is working reasonably well: Housing units in the development_events_exogenous table were added correctly; New households were added to these new housing units and controlled by the regional totals. We summed total households by community (city, village and township). Comparing the sum from 2005 model output and our own 2005 estimation by community, the differences are small - the "mean absolute percent error" is about 2%.

We are willing to talk about it more if anyone is interested.

Thanks.

Xuan Liu
Coordinator, Data Center
SEMCOG
535 Griswold St. Suite 300
Detroit, MI 48226
(313) 961-4266

>>> Paul Waddell <pwaddell_at_u.washington.edu> 3/12/2007 5:24 pm >>>
Jen

This is a big question for all of the household synthesis tools. One
approach is to use the 2005 housing stock from the parcel data or other
local sources like building permits or utility hookups, to adjust the
2000 marginals. Unfortunately this would imply that socio-demographic
characteristics remain stable from 2000 to 2005. I don't think there is
sufficient post-2000 census data accumulated to do very much better than
this, but would be interested in hearing suggestions.

Paul

Jennifer C. Duthie wrote:
> Hi,
>
> We'd like to use 2005 as our model base year, but we're worried about
> the problems this may cause during household synthesis. We have some
> thoughts on scaling the Census 2000 data to approximate 2005 levels..
> Has anyone else synthesized households for a non-Census year? Any advice?
>
> Thanks,
> Jen
>
> Graduate Research Assistant
> University of Texas, Austin
> _______________________________________________
> Users mailing list
> Users_at_urbansim.org
> http://www.urbansim.org/mailman/listinfo/users
>

_______________________________________________
Users mailing list
Users_at_urbansim.org
http://www.urbansim.org/mailman/listinfo/users
Received on Tue Mar 13 2007 - 05:19:17 PDT