Documentation Deviations
March 3, 2010
Documents deviations program and data used The Diffusion of Wal-Mart and Economies of Density,” Econometrica, Vol. 79, No. 1 (January, 2011), 253-302
by Thomas Holmes
Date File pert_pv_complete.asc
800,353 x 24
Each row is a different pairwise deviation. Each column is a below:
re is:
Column |
Variable |
Description |
1 |
type_of_deviation (in program is type_tom) |
=1 then density decreasing =2 then density increasing =31 or 32 then a change within the same state, more then 2 years aport |
2 |
convert_cat |
=1 means store opening (general merchandise) is changed but not a conversion of super center timing =2 change order of supercenter conversion |
3 |
store1 |
store number of earliest store opened of the deviation pair |
4 |
store2 |
store number of later store opened of the deviation pair |
5 |
year_open1 |
year store1 opened |
6 |
year_super1 |
year store 1 became a supercenter (=9999 if never a supercenter) |
7 |
year_open2 |
year store 2 opened |
8 |
year_super2 |
year store 2 became a supercenter (=9999 if never a supercenter) |
9 |
pert_id |
an identification number |
10 |
n1 |
population density in 1990 (number of people in thousands within 5 miles of store location 1), Popden for 1990 in paper |
11 |
n2 |
population density in 1990 (number of people in thousands within 5 miles of store location 2) Popden for 1990 in paper |
12 |
ΔП (d_opprof) |
sum of general merchandise and food operating profit |
13 |
ΔD (d_dc_dist) |
sum of gen merchandise and food distribution mile differences |
14 |
ΔC1 (d_lnneig5) |
see paper, includes gen and food (food if applicable) |
15 |
ΔC2 (d_lnneig5_2) |
see paper, includes gen and food (food if applicable) |
16 |
weight_factor |
take beginning year of the perturbation. Then weight_factor=1/(.95^(year-1962)); |
17 |
d_opprof_gen |
Variables 17-24 are the same as 12-15, only with a breakdown by general merchandise and food |
18 |
d_opprof_groc |
see 17 |
19 |
d_dc_gen_dist |
see 17 |
20 |
d_dc_groc_dist |
see 17 |
21 |
d_lnneig5_gen |
see 17 |
22 |
d_lnneig5_groc |
see 17 |
23 |
d_lnneig5_2_gen |
see 17 |
24 |
d_lnneig5_2_groc |
see 17 |
Program bound_estimation_specification1.prg
First step is to read in a gauss matrix format version of this file.
Next it determines what sample to use.
For the type 2 deviations, I require that ΔD>=150.
These lines also allow me to pick year_start and year_end for the earlier opening year of the deviation.
In the example, year_start=0 and year_end=9999.
Note that the deviation differences in the file were calculated as follows:
I first ran the equivalent of program in model_calculations to calculate for the various opening policy the present value of sales (sal_gen for general merchandise, sal_groc for food), distance, F1 and F2., etc. These are prevent value as of 1962. Then the following were calculated: (noting that 0 means the policy actually chosen)
The code used to create the variables is as follows.
d_sal_gen= sal_gen0- sal_gen ; *i.e. difference
in present value of sales between actul and deviation
for general merchansise
d_sal_groc= sal_groc0-
sal_groc;
d_wage_gen = wage_gen0- wage_gen;
d_wage_groc = wage_groc0- wage_groc;
d_rent_gen = rent_gen0- rent_gen;
d_rent_groc = rent_groc0- rent_groc;
d_lnneig5_gen=
lnneig5_gen0- lnneig5_gen;
d_lnneig5_2_gen=lnneig5_2_gen0-
lnneig5_2_gen;
d_lnneig5_groc=
lnneig5_groc0- lnneig5_groc;
d_lnneig5_2_groc=lnneig5_2_groc0-
lnneig5_2_groc;
d_dc_gen_dist= dc_gen_dist0-
dc_gen_dist;
d_dc_groc_dist= dc_groc_dist0-
dc_groc_dist;
labor_requirement=3.6146;
d_opprof_gen=.17*d_sal_gen - labor_requirement*d_wage_gen -
.36118*(1/5)*d_rent_gen;
d_opprof_groc=.17*d_sal_groc - labor_requirement*d_wage_groc -
.36118*(1/5)*d_rent_groc;
weight_factor=1/(.95**(year_open1-1962));
d_opprof=d_opprof_gen+d_opprof_groc;
d_lnneig5=d_lnneig5_gen+d_lnneig5_groc;
d_lnneig5_2=d_lnneig5_2_gen+d_lnneig5_2_groc;
d_dc_dist=d_dc_gen_dist+d_dc_groc_dist;
Note the weight factor is used to convert present value to the year as of the earliest opening of the pair of stores.