This is the program file for "Casino Gaming and Local Employment Trends" by Thomas A. Garrett, Federal Reserve Bank of St. Louis Review (January/February 2004).

The following EViews program is used for calculating the forecasted values of household employment for each county under the assumption that no casinos were introduced.  The inputted data is household employment for the six counties, by month, from 1984 to 2001.

smpl 1984:01 2001:12

'Augmented Dickey Fuller tests (ADF) for stationarity are conducted on each pre-adoption employment series to determine the order of integration (d).  Employment for St. Clair and St. Louis counties is stationary in levels so those counties are not first-differenced.  Employment in Lee County is trend stationary.'
genr dtunica = tunicasa-tunicasa(-1)
genr dmassac = massacsa-massacsa(-1)
genr dwarren = warrensa-warrensa(-1)
genr dlee = leesa-leesa(-1)

'Outputted data is forecasted household employment for each county as if no casino had been introduced.  The end of each sample for the ARIMA models is based on the date the first casino opened in each county.'

'Akaike information criterion was used to determine that the appropriate ARIMA model for Tunica County is (2,1,1). ' 
smpl 1986:01 1992:08
equation ztunica.ls dtunica c ar(1) ar(2) ma(1)  

'Akaike information criterion was used to determine that the appropriate ARIMA model for Massac County is (1,1,0). A binary dummy variable is included which has a value of 1 for 1990:1 to account for a structural break in the data.'
smpl 1986:01 1993:01
equation zmassac.ls dmassac c ar(1) masdum

'Akaike information criterion was used to determine that the appropriate ARIMA model for Warren County is (2,1,0).'
smpl 1986:01 1993:01
equation zwarren.ls dwarren c ar(1) ar(2)

'Akaike information criterion was used to determine that the appropriate ARIMA model for St. Clair County is (1,0,4).  The coincident index for Illionois is included to capture the effect of business cycles.' 
smpl 1986:01 1993:06
equation zstclair.ls stclairsa c ar(1) ma(1) ma(2) ma(3) ma(4)  ilcoin 

'Akaike information criterion was used to determine that the appropriate ARIMA model for Lee County is (1,0,0).  The trend series is normalzed to 0 in the ARIMA model.' 
smpl 1986:01 1994:10
equation zlee.ls leesa c ar(1) @trend

'Akaike information criterion was used to determine that the appropriate ARIMA model for St. Louis County is (1,0,2).  A binary dummy variable is included which has a value of 1 over the period 1990:1 to 1993:1 to account for a structural break in the data.  The contemporaneous and one-lag values of the Missouri coincident index are included to capture the effect of business cycles.'
smpl 1986:01 1997:02
equation zlouis.ls louissa c ar(1) ma(1) ma(2) louisdum mocoin mocoin(-1)