Data and Computer Code Appendix for "Entry and Exit Echoes" by Boyan
Jovanovic and Chung-Yi Tse

The Gort and Klepper dataset

The dataset (Table 1 of the paper) is constructed from the information in various tables in
Gort and Klepper (1982) as explained in notes 3, 9, 10 and section 3.0.1 of the paper.
The estimates in Figures 4 and 5 are computed by the gauss programs "est-T-3.e" and
"nr-15.e", respectively. In "est-T-3.e", r is set equal to 0.02. For other values of r, simply
amend line 8 accordingly.

The Aircraft dataset

The description of the data is in "codebook.pdf". The gauss program "P10.e" computes
the results shown in Figure 10 (for US airlines) and the program "P13.e" computes the
results shown in Figure 11 (for the world's airlines). The columns of the output of each
program are: (1) year, (2) no. of exiting airlines, (3) no. of aircraft operated by the exiting
airlines, (4) average age of planes operated by the exiting airlines in days, (5) no. of nonexiting
airlines, (6) no. of aircraft operated by the non-exiting airlines, (7) average age of
planes operated by the non-exiting airlines. The gauss data file "dease.dat" that the
programs take as input is omitted because it is converted from a proprietary data file
(available to us in stata data file format) compiled by Flightglobal.com
(www.fightglobal.com/StaticPages/Acas.html).

The Repeated Echoes dataset

Rajshree Agarwal (1998) collected annual information on the number of entrants and
exiting firms in 33 industries. The dataset is in the excel file "entry-exit.xls".
The first step in the analysis is to construct rates of entry and exit, defined as the numbers
of firms entering and exiting in a year divided by the number of firms producing in the
industry at the beginning of the given year. The rates are not defined for the first year of
the raw data series. Next we calculate the smoothed entry and exit rate series of each
product by the HP filter as explained in Section 4 of the paper. Finally we detect spike
dates. The computations are carried out in the gauss program "hp4.e". The program is to
be executed once for each of the 33 products. In each run, lines 416-9 of the program
should be amended according to which one of the 33 products is being considered. The
output of the program contains all the information necessary to construct Table 3 and
Figure 6.

Using the information in Table 3 and in "entry-exit.xls", the gauss program "est-T-4.e"
computes the estimates in Figures 7 & 8. Set x in line 30 of the program equal to dx for
the estimates in Figure 7 and equal to de for the estimates in Figure 8. The program is
essentially identical to est-T-3.e, except that the data embedded in it are different.

Using the information in Table 3, the frequencies in Figure 9 are computed by the gauss
program "p-value5.e".

The numbers reported in Table 4 are computed by hand from the information in Table 3 and 
the NBER recession dates (http://www.nber.org/cycles/cycles.main.html).

Using the information in Table 3, the gauss programs "c-t-2.e" and "c-t-3.e" compute the 
estimates in Tables 5a and 5b. The program "c-t-2.e" is for the first-halvbes of Tables 
5a and 5b, whereas the program "c-t-3.e" is for the second-halves of the two tables. The 
programs are to be executed once for each value of S (amend line 270 of "c-t-2.e" and 
line 292 of "c-t-3.e" accordingly.) The estimates in Table 5a are obtained by setting 
line 271 of "c-t-2.e" line 293 of "c-t-3.e" equal to d=cd0(S,1,2); the estimates in Table 
5b are obtained by setting the lines equal to d=cd0(S,2,1).