Poverty is determined by comparing the total income of the family to poverty thresholds for that size family. The thresholds account for annual changes in the Consumer Price Index. You can view the current poverty thresholds and discussion of poverty and its measurement under Poverty at the U.S. Census Bureau's web site.
The CPS ASEC sample does not include observations for most counties. The sample design provides consistent national-level estimates, but consideration is also given to providing reliable annual state-level estimates of unemployment. Of the more than 3,100 counties in the country, about 1,300 counties are in the sample.
To use the CPS ASEC data at the county level, we needed to adjust the sample "weights" (the number of people represented by each sample person). In the CPS ASEC design, some counties represent a group of counties, while others represent only themselves. For our purposes, each county must represent itself.
We derive inputs for the models from the CPS ASEC data by computing the key statistics for each state and each of the counties in the sample. In order to reduce the sampling variability of the CPS ASEC estimates for each county, we computed three-year weighted averages centered on the target estimation year, rather than basing the estimate on data about the target year alone. Any bias resulting from the inclusion of the data for adjacent years was felt to be less important than the reduction in variance gained by including data from the additional years. For example, to obtain the observations for the 1993 models, we averaged estimates for 1992, 1993, and 1994 obtained from the March 1993, March 1994, and March 1995 CPS surveys. State models, however, use only data from the target year (data from the March 1994 survey for income year 1993).
Further information about the characteristics of the CPS and the ASEC can be found on the CPS Homepage.