File Organization
Because of the size of the RECS questionnaire, the variables were separated into groups by subject matter. These 10 smaller files make it easier to manipulate the data. Each file contains 6,229 records representing households in the 50 States and the District of Columbia. These households are weighted to represent 90.5 million households, as of November 1987, the midpoint of data collection activity. Included are energy-related characteristics of the housing structure, household demographic data, annualized household energy consumption and expenditures data, a weighting variable and weather data. Also added to the file are standard electricity prices for 1,000 kWh and end-use data. End-use data are estimated amounts of the household's annual fuel consumption and expenditures by six end uses for five fuels.
Several frequently used variables are included in each group of variables:
- HHID: household ID
- NWEIGHT: Final Household Weight
- REGIONC: Census Region
- DIVISION: Census Division
For each group of variables, there are two items: a codebook file and a data file. The codebook file is a text file which gives, for each variable on a file: the variable name, a description, and the meaning of any codes used. The data file is a comma separated value file.
Each of these 10 files can be used by itself or be merged with other files for more complex analyses. By merging files together, a new file can be created that contains, for each respondent, variables from two or more files. The variable HHID should be used to link the files.
In order to arrive at national estimates from the RECS sample, base sampling weights were calculated for each household (these are the reciprocal of the probability of that household being selected into the sample). Therefore, a household with a base weight of 1,000 represents itself and 999 similar, but unsampled households in the total building stock. The base weight is further adjusted to account for nonresponse bias. The variable NWEIGHT in the data file is the final weight. In order to obtain a national estimate, each sample household's value must be multiplied by the household's weight.
Confidentiality of Survey Respondents
In order to maintain the survey respondents' confidentiality, names and addresses of respondents are not provided to the EIA. To further prevent the possible identification of individual survey households, changes were made to the original data before creating the public use data. Potentially identifying geographic variables, such as sampling unit location variables and some of the weather- related variables were removed from the data files. The remaining weather variables such as heating and cooling degree-days, to various base temperatures, were introduced with random error.
Heating and cooling degree-days are continuous variables useful in general energy consumption analyses, and are provided on the RECS public use data file for a range of temperature bases. Degree-days relative to various base temperatures were computed using temperature records from the National Oceanic Atmospheric Administration (NOAA) weather station nearest to each sample household. These annual degree-day sums on the public use data have been inoculated with random errors from normal distributions of mean zero and small variance, to make it difficult to identify the source NOAA weather station.
These random errors are small enough so as not to adversely affect any statistical analyses performed by data users. The median absolute percentage change in degree-days following error inoculation ranged from 1.3% to 7.1%, when examined over categories defined by Census division and temperature threshold. For each household, the heating and cooling degree-days of different threshold temperatures were assigned normal error distributions with different variances, but all with mean zero. However, for each household, the same standard-normal random deviate was transformed to each of these different distributions to ensure consistency between degree-day inoculations for the different threshold temperatures for a given household. � |
