Transcript The Consumer Price Index Housing Sample - DC

The Consumer Price Index
Housing Sample
An address based sample
Robert Poole
Office of Prices and Living Conditions
Bureau of Labor Statistics
CPI Housing Sample Overview


Used to estimate inflation for Rent and
Owners’ Equivalent Rent – the value of
the flow of services one receives from
an owned home
One sample of renters is used to move
both rent and OER
Some Select Sample Details



Cluster sampling is used
Sample segments (neighborhoods)
based on total housing expenditure
Sample households within each
segment with equal probability
Ooooh! More Details



Renters are implicitly matched to
owners geographically by neighborhood
Assumption that the Rent change in a
neighborhood is the same as the
unobservable Owners’ Equivalent Rent
change
No assumption is made about price
levels
Sampling Details: Minutiae

Just kidding!
1998 Revision Sample



Our segments were census block based
Used a manual address listing process
for each segment – very costly
Renters difficult to find in high owner
density neighborhoods
“Current” Sample



Our segments are now census block
group based
Expect 4.5 times the number of
addresses per segment
Generalissimo Francisco Franco is still
dead – renters in high owner density
areas still hard to find
Cost



Hmmm. How can we make the listing
process manageable and cheaper?
ADDRESS LISTS!!!!
But wait …
Address List Issues


Coverage – are all the households in an
area on the lists?
Accuracy – are the addresses geocoded
accurately?
Coverage



Multi-unit structures can have but one
delivery point in the Delivery Sequence
File
DSF augmented with marketing lists
Tested for bias in price change
associated with “missing” addresses –
none found
Geocoding Accuracy



Compared list geocoding with census
geocoding
Rob’s bout of swine flu prevented him
from digging up numbers about the
accuracy in time for this presentation.
It was good. Trust him.
Accuracy on par with manual listing
What about them pesky, hard
to find renters?




Hey! We’re using marketing lists! Can
they help?
Binary renter/owner coding not very
helpful
Sliding scale renter/owner coding was
very helpful in identifying “certainty”
owners – over 98% helpful
Oh yeah! (knock on wood)
Address List Advantages




Cheaper than listing
Owner/Renter information allowed us to
effectively double our sample size
Phone Numbers – Too early to tell
Names – A possibility for the future
Questions?