Modifications of the input variables between estimation of the 1993 and 1995 models improved the accuracy of the results. However, they reduced the comparability of 1993 and 1995 estimates. The changes are:

• At the state level, the food stamp variable now uses the average number of recipients of food stamps over a 12-month period rather than just one month. The average has a delay of six months relative to the model's income year. For example, the 1995 model uses the period July 1995 through June 1996. The following adjustments to the food stamp data at the state level are also new:
  • The number of food stamp recipients was adjusted to reflect natural disasters. Food stamps recipients added in response to a declared natural disaster were removed from the totals.
  • Outliers were smoothed on the basis of time series analysis of the period 1975 to 1997.
  • The numbers of recipients in Alaska and Hawaii were adjusted because the income eligibility thresholds for those states are higher than for the continental U.S.
  (See Food Stamp Recipients for more information.)

• At the county level, the values of the food stamp variable for July are raked to the improved state-level variable.
• The measurement of exemptions in data from federal income tax returns has been refined to more accurately reflect household dependents. Total exemptions now include children living away from home (previously, only children at home were counted), exclude parent exemptions, and exclude "other" exemptions. This changes all the variables that employ the number of exemptions in their calculation. In addition, tax returns claiming no exemptions -- i.e., those who are claimed as dependents on other returns -- are excluded from the calculation of all variables (except median income).