search

TylerRayRogers / ECMT-674-Project

Develop forecasting models of County level Sales Tax collections in the State of Texas
2 stars 0 forks source link

Question #3 #3

Open TylerRayRogers opened 3 years ago

TylerRayRogers commented 3 years ago

a. Present some type of multivariate forecasting model based on either: i. A set of “leading indicators.” ii. A VAR model of the Sales variable plus 2 or more series of your choice. b. Your presentation should include a description of why you chose certain variables for your multivariate model. c. Based on your final multivariate model, present a 12 month ahead path forecast of the Sales Tax variable. How does it compare to your Univariate model forecasts?

TylerRayRogers commented 3 years ago

Q3 a) We have elected to build a leading indicators model. The variables we will use are the number of real estate sales in Cameron County and the Consumer Confidence Index for the State of Texas, each lagged by 12 months for forecasting purposes.

TylerRayRogers commented 3 years ago

b) We elected to examine the real estate sales volume because it will indicate the number of households/businesses coming to Cameron County. It should especially be apt as a leading indicator because real estate does not tend to be a short-term investment. New families or businesses which are buying into the county this month can be reasonably expected to be purchasing and consuming within the county a year from now. The statewide Consumer Confidence Index is being used because the CCI is typically seen as a useful measure of consumer willingness to spend, and this feeling ebbs slowly over time, making it useful for forecasting. Both variables show significant values of correlation with our outcome variable (the volume of sales tax receipts) so it was decided they were adequate.

TylerRayRogers commented 3 years ago

x1_x2 corr_tests

TylerRayRogers commented 3 years ago

c) Leading Indicators model summary Leading Indicators vs Actual Data Leading Indicators Forecast summary Leading Indicators Forecast plot

TylerRayRogers commented 3 years ago

[COMPARE TO UNIVARIATE FORECASTS] Off the top of my head, the reason there is such a disconnect in the forecast plot is that the predictions are relying on variable inputs from right at the beginning of the COVID pandemic. Leading indicator models would obviously not respond well to major unforeseen shocks, and the economy in this case is no different. Consumer Confidence and real estate transactions both dropped overnight and have been rebounding to previous levels.