Dear Statalist's,
I am trying to estimate the effect of a drop in captial gain taxes on the trading volume. Capital gains turned tax free in Germany after a one year holding-period for stocks bought before 01.01.2009. After 01.01.2009 this regulation was abolished.
Therefore I use a panel dataset with about 200 IPO's issued between 2000 and 2018 with daily trading volumes, on Stata v. 14. I am running a random-effects regression with tripple difference, as I have three dummies defining my effect:
(1) y_2009: dummy is 1 for every stock issued before 2009 (old regulation)
(2) e_m20_p20_p: dummy is 1 for a time range of 40 days around the IssueDate + 1 year (effect-period)
(3) cg_m20_p20_p: dummy is 1 for stocks with capitalgains which accured in the effect-period
(4)I also created dummy combi__m20_p20p which returns 1 if all dummies above are returning 1. This is a variable to run a t-test on whether the mean of the trading volume in the effect-period is different to the other trading volumes.
So I have two control groups: Stocks issued after 2009 and stocks with no accured capital gains in the effect-period.
This is my data, with id = id and time = DaysAfter, and I use the log(Trading Volume) = l_VC as my trading volume, as trading volume(VC) is not normally distributed.
Code:
* Example generated by -dataex-. To install: ssc install dataex
clear
input int(id DaysAfter) float(VC l_VC d_2009 e_m20_p20p cg_m20_p20p combi__m20_p20p)
1 1 20608.193 9.933444 0 0 0 0
1 2 4888.4 8.49462 0 0 0 0
1 3 2365.1 7.768576 0 0 0 0
1 4 1113.4003 7.015174 0 0 0 0
1 5 1038.7002 6.945725 0 0 0 0
end
And this is the regression result:
Code:
. xtreg l_VC i.e_m20_p20p##i.cg_m20_p20p##i.d_2009, re
Random-effects GLS regression Number of obs = 91,638
Group variable: id Number of groups = 160
R-sq: Obs per group:
within = 0.0003 min = 28
between = 0.0841 avg = 572.7
overall = 0.0573 max = 600
Wald chi2(7) = 41.57
corr(u_i, X) = 0 (assumed) Prob > chi2 = 0.0000
-----------------------------------------------------------------------------------------------
l_VC | Coef. Std. Err. z P>|z| [95% Conf. Interval]
------------------------------+----------------------------------------------------------------
1.e_m20_p20p | -.0619384 .0445417 -1.39 0.164 -.1492386 .0253618
1.cg_m20_p20p | 1.477089 .4878709 3.03 0.002 .5208795 2.433298
|
e_m20_p20p#cg_m20_p20p |
1 1 | .1816618 .0571906 3.18 0.001 .0695703 .2937532
|
1.d_2009 | .2777146 .4466912 0.62 0.534 -.597784 1.153213
|
e_m20_p20p#d_2009 |
1 1 | .1966056 .0574507 3.42 0.001 .0840042 .3092069
|
cg_m20_p20p#d_2009 |
1 1 | -.5789455 .6348865 -0.91 0.362 -1.8233 .6654092
|
e_m20_p20p#cg_m20_p20p#d_2009 |
1 1 1 | -.318946 .0768454 -4.15 0.000 -.4695602 -.1683318
|
_cons | 2.194597 .3709917 5.92 0.000 1.467466 2.921727
------------------------------+----------------------------------------------------------------
sigma_u | 1.9265326
sigma_e | 1.1342814
rho | .74258446 (fraction of variance due to u_i)
-----------------------------------------------------------------------------------------------
. ttest l_VC, by(combi__m20_p20p)
Two-sample t test with equal variances
------------------------------------------------------------------------------
Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval]
---------+--------------------------------------------------------------------
0 | 90,613 2.936599 .0075864 2.283643 2.92173 2.951468
1 | 1,025 3.356913 .0691598 2.214195 3.221202 3.492624
---------+--------------------------------------------------------------------
combined | 91,638 2.9413 .0075427 2.283294 2.926517 2.956084
---------+--------------------------------------------------------------------
diff | -.4203143 .0717073 -.5608599 -.2797687
------------------------------------------------------------------------------
diff = mean(0) - mean(1) t = -5.8615
Ho: diff = 0 degrees of freedom = 91636
Ha: diff < 0 Ha: diff != 0 Ha: diff > 0
Pr(T < t) = 0.0000 Pr(|T| > |t|) = 0.0000 Pr(T > t) = 1.0000
So, it would be SO great if you could help me or give me some suggestions with the following questions:
(a) I am trying to calculate the effect of the drop in capital gains on the trading volume: As it is a tripple Difference-in-Difference, I would suggest it is the sum of ( 1.e_m20_p20p + 1.cg_m20_p20p + 1.d_2009 + e_m20_p20p#cg_m20_p20p#d_2009), isn't it? I am trying to replicate the "margins, dydx()" command from
https://www.statalist.org/forums/for...ation-in-stata , but it is not possible for me, as I don't know what to put in the dydx() in a tripple Dif.
My interpretation so far is: The isolated effect of capital gains, accured during the effect-period in stocks issued before 2009 is a significant growth of (-6,1 % + 147,7 % + 27,7% - 31,8% ) (approximately) in trading volume
(b) Is it appropriate to use the logarithm of traiding volume (VC) in the regression and also in the ttest?
(c) Is the ttest an appropriate way to test, wether the mean of my target-group is significant different from the mean of my controll-group, as it is a paneldataset?
I am very very thankfull for everyone helping me and giving suggestions,
Best wishes,
Phill