As part of the November forecast update, five panel regression equations were respecified. This will drive a number of differences in both the baseline and scenario forecasts. The equations for farm proprietors' income, non-farm proprietors' income, and state residence income adjustment were adjusted to use differences instead of levels, helping to ensure a smoother transition between the last historical quarter and the first forecast quarter. In addition, where possible, more state- or concept-specific drivers were incorporated into these equations to better capture interactions within the state economy.
This change to incomes drove adjustments to non-wage income, which flow into total income. As a result, these changes have some downstream impact throughout the model.
In addition, the equation for home sales was respecified to incorporate a differences of logs structure. Further, a convergence term was added to ensure that home sales on a per household basis approach the U.S. ratio in the long run.
Lastly, the households equation was changed to a differences of logs specification as well. However, the resulting forecasts were "lined up" to last month's levels to ensure continuity, due in large part to the elevated degree to which that series flows into downstream variables.
For Puerto Rico, equations for both the unemployment and effective wage rates were respecified. The new structure of each is designed to eliminate the possibility of discontinuity between history and forecasted values and to make each equation more economically intuitive.
Note that all state-level changes will flow into relevant metro area models, but that the metro area equations for the concepts above, where relevant, have not been respecified. This is immaterial for the non-wage income components, as only total non-wage income is forecast for metro areas, but may be revisited for both households and home sales after the regional models are unlocked this spring.
|