Joseph V. Terza, "Alcohol Abuse and Employment: A Second Look", Journal of
Applied Econometrics, Vol. 17, No. 4, 2002, pp. 393-404.
The data are in an ascii file called terza.data, which is in DOS format and
is zipped in the file terza-data.zip. Unix users should use "unzip -a".
The data file contains 39288 rows. An observation comprises four rows, so
there are actually 9822 observations, each with 52 variables. The 52
variables include (in column-order of appearance in the dataset): d, y, Z
and X, where
d = alcohol abuse as defined in Table 1 of the paper.
y = 1 if out of the labor force, 2 if unemployed, and 3 if employed.
Z = 9822 x 29 matrix of variables included in the first-stage probit
(including the constant term). The first 20 columns of this matrix include
the variables in columns 2 through 21 of the matrix X (d is not included -
see below). NOTE THAT THEY ARE NOT, HOWEVER, IN THE SAME ORDER AS THE
VARIABLES IN X. THE ORDERING OF THIS SET OF VARIABLES CORRESPONDS TO THAT OF
THE FIRST 20 ROWS OF TABLE IV OF THE PAPER. THE IMPORTANT DIFFERENCE IS
THAT THE CONSTANT TERM IS LISTED FIRST AMONG THIS GROUP OF VARIABLES. The
remaining 9 columns correspond to the following variables in order (see
Table 1 of the paper for definitions):
Beer tax
Cigarette tax
App. ethanol
Beer tax squared
Cigarette tax squared
App. ethanol squared
Bio-mom alcoholic
Bio-dad alcoholic
Lived w/ alcoholic
X = 9822 x 21 matrix of variables included in the multinomial logit
estimations (including d and the constant term). The variable values are in
the same column order as listed in the rows of Table II of the paper.
Please contact the author directly to obtain access to the software used
in this paper.
Joseph V. Terza
Department of Economics
and Social Science Research Institute
Penn State University
University Park, PA 16801
jvt@psu.edu
http://econ.la.psu.edu/~terza