H. Spencer Banzhaf and V. Kerry Smith, "Meta Analysis in Model
Implementation: Choice Sets and the Value of Air Quality Improvements",
Journal of Applied Econometrics, Vol. 22, No. 6, 2007, pp. 1013-1031. 

There are two zip files. The file mergemaster.zip contains a single data
file called mergemaster_public.csv, which is an ASCII file in DOS format.
The file program.zip contains 132 Stata .do files, which are also ASCII
files in DOS format. In both cases, Unix users should use "unzip -a".

The file mergemaster_public.csv is organized as a comma delimited text file
with 321,344 records. Each record of this file pertains to one house sold in
the Los Angeles metro area. The fields for each record include:
 
  houseid [a unique identification code for each house], 
  county [The county in which the house is located],
  tract [the centsus tract in which the house is located],
  block [the census block in which the house is located],
  avginc [the mean household income for the 1990 US Census blockgroup in
          which the house is located]
  saledate [the date the house was sold],
  ozonefree [the number of days *without* an ozone alert at the ozone
             monitor nearest to the house],
  tchstud [average teachers/student ratio in the school district in which
           the house is located],
  mathscore [the average 10th grade math score in the school district in
             which the house is located],
  public safety [coded as 3000 minus the number of FBI crime index crimes
                 per 10000 people], 
  andcoast [a dummy variable equal to 1 if the house is within 1 mile of
            the Pacific Ocean],
 
The air quality data were obtained from the Califonia Air Resources Board
webpage (www.arb.ca.gov/aqd/aqd.htm). The education data (tchstud and
mathscore) were obtained from the National Center for Education Statistics.
The crime data were obtained from the California Department of Justice, and
include willful homicide, forcible rape, robbery, larceny-theft, burglary,
aggravated assault, motor vehicle theft, and arson. Finally, the income data
were obtained from the 1990 US Census, at the block-group level. These data
can be found at http://www.census.gov./main/www/cen1990.html.
 
As described in the paper, the analysis included individual house prices for
the transaction and housing characteristics (bedrooms, bathrooms, square
footage, lot size, fireplace, and the presence of a swimming pool).
Unfortunately, these data were purchased from a private for-profit company
and are proprietary. We are contractually forbidden from disseminating these
data. However, they are available for purchase. We purchased them from what
was then Transamerica Intellitech and is now Fidelity National Data Service.
See for example their webpage for the "siteX" data product:
http://www1.sitexdata.com. First American Real Estate Solutions is also a
good alternative, selling the same data with a different interface.  
 
However, note well that, at both firms, their pre-packaged products offer
only information on the *most recent sale* of each house. To replicate our
work with our original (1989-94) data, you will want to be sure to speak
with a sales representative about obtaining historical data.
 
As described in the paper, these data are used to create choice sets
for each sampled household. A zip file containing 128 STATA programs
generating these choice sets, using each of the 128 definitions
described in the paper, is also provided.

For further questions, please contact:

Spencer Banzhaf
Dept of Economics
Andrew Young School of Policy Studies
Georgia State University
PO Box 3992
Atlanta, GA 30302, USA
404-651-6981
hsbanzhaf [at] gsu.edu