Question

Status:Closed    Asked:Apr 20, 2015 - 12:58 PM

Feedback on data extraction for multi-year work/travel time data


Hi folks,
First time iPUMS user so pardon if this is a basic question.

We're trying to analyze trends in travel time to work (TRANTIME) and transportation mode (TRANWORK) by using IPUMS-USA data from the 1990 and 2000 decennials, along with 2010 ACS data. We've created a data extract from 1% samples of the full data file across these years. In addition to the work-related variables, we've added the below geographic-oriented variables to support analyses at varying geography levels:

STATEFIP (State (FIPS code))
COUNTY (County)
METAREA (Metropolitan area [general version])
METAREAD (Metropolitan area [detailed version])

I wanted to inquire if the data extract, as described above, is appropriate for this type of analysis or if there are any land mines we should be aware (e.g. comparing decennial data to ACS data). In particular, we're interested in trends in travel time/mode by MSA, so would be interested in feedback on that front.

Also, does anyone know of third-party analyses (perhaps pre-baked Census summaries) that cover the same variables? I'd like to sanity check any custom analysis we perform against a third party's summary data to ensure our summaries are accurate.

Any advice on the above is greatly appreciated.

 
Do you have the same question? Follow this Question
 

Staff Answer

avatar

Tim_Moreland

Staff

There are no inherent problems with comparing data from the samples you have listed; however, you should take caution in comparing MSA-level data across time. Since the most-detailed level of geography available for 1990-2010 is the PUMA, which do not necessarily align with official metro area boundaries, several metro areas are incompletely identified. Metro areas that are incompletely identified are unlikely to be a representative sample of the entire metro area and the weighted population of the identified portion will be lower than the actual population of the metro area by 1% to 74%. Additionally, METAREA designations for the 1990 sample are based on 1990 OMB metropolitan area delineations, while the 2000 and 2010 samples are based on the 1999 delineations.


The COUNTY variable is more comparable over time, although borders and names can change. As well, some counties are not identifiable at all, while others may not be identifiable in all three of your samples. Unlike with METAREA, if a COUNTY is identified, then it is completely identified.STATEFIP is the most stable of the geographic variables, since any changes in state borders should be minor.


As for your transportation to work variables, TRANWORK is completely comparable from 1990-2010. TRANTIME is essentially completely comparable, with one minor issue: in 1990, respondents were instructed to include waiting time for public transportation or for other passengers in carpools.


You can find Census tables for means of transportation and time of travel to work for several levels of geography at the NHGIS project website. While NHGIS is primarily focused on mapping data, the summary tables can be extracted independent of shape files as .csv files with optional descriptive headers.


Hope this helps.

 

Apr 21, 2015 - 04:47 PM

0
0
Report it

OTHER QUESTIONS NEEDING ANSWERS

when trying to declare the IPUMS data as panel on stata there are repeated ti...
2010 5-year PUMA definition
Where can I find city populations, by age, San Diego, New Orleans, Atlanta, C...
Hi, it has been over 4 hours for my data extract and I was wondering if there...
When I take several Nov Supplements, the weighted number of voters equals pub...
Are there aggregate categories for OCC/OCCLY in CPS?
Login   |   Register

Recently Active Members

View More »

Share |