Hi All,
I am new to the forum, so please let me know if I am not formatting my request properly, and I'll be happy to make the appropriate changes. I am having problems running an event-study in Stata with two-dimensional fixed-effects.
I have a panel of firms and year, with a shock (i.e., a change in a law) occurring in a staggered fashion to several groups of firms, so that by the end of the sample period all the firms are eventually treated. I am trying to run an event study, where the dependent variable "y" is regressed on firm and year dummies, plus a set of dummies to check for the absence of a pre-trend. For example, dumm5 is equal to 1 form firm i in year t if the shock to firm i occurred in year t-5. dumm_1 is omitted, so that all the remaining dummies should all be identified. Indeed, if I use "areg", this is my output:
If I use instead "reghdfe", I get the following output, with one dummy omitted:
If I cluster by firm, the problem seems to appear also with "areg". I would like to use "reghdfe", as I am planning to use multi-dimensional fixed-effects, but I would really like to plot all the coefficients, with just dumm_1 normalized to 0.
Vincenzo
I am new to the forum, so please let me know if I am not formatting my request properly, and I'll be happy to make the appropriate changes. I am having problems running an event-study in Stata with two-dimensional fixed-effects.
I have a panel of firms and year, with a shock (i.e., a change in a law) occurring in a staggered fashion to several groups of firms, so that by the end of the sample period all the firms are eventually treated. I am trying to run an event study, where the dependent variable "y" is regressed on firm and year dummies, plus a set of dummies to check for the absence of a pre-trend. For example, dumm5 is equal to 1 form firm i in year t if the shock to firm i occurred in year t-5. dumm_1 is omitted, so that all the remaining dummies should all be identified. Indeed, if I use "areg", this is my output:
Code:
. areg y dumm* i.year,absorb(id)
note: 151.year omitted because of collinearity
Linear regression, absorbing indicators Number of obs = 42928
F( 42, 40192) = 33.09
Prob > F = 0.0000
R-squared = 0.4961
Adj R-squared = 0.4618
Root MSE = 4.2764
y Coef. Std. Err. t P>t [95% Conf. Interval]
dumm_13 -.1311646 .2493519 -0.53 0.599 -.6199 .3575709
dumm_12 -.0774617 .2734669 -0.28 0.777 -.6134632 .4585397
dumm_11 .4890132 .254919 1.92 0.055 -.0106338 .9886602
dumm_10 .3322305 .2209377 1.50 0.133 -.1008124 .7652734
dumm_9 -.5587808 .1927939 -2.90 0.004 -.9366612 -.1809004
dumm_8 -.4640176 .1665719 -2.79 0.005 -.7905024 -.1375329
dumm_7 -.3920493 .1746904 -2.24 0.025 -.7344464 -.0496522
dumm_6 -.1567121 .157182 -1.00 0.319 -.4647924 .1513682
dumm_5 -.5046396 .1608119 -3.14 0.002 -.8198345 -.1894446
dumm_4 -.368474 .1597363 -2.31 0.021 -.6815608 -.0553872
dumm_3 -.0716386 .156353 -0.46 0.647 -.378094 .2348167
dumm_2 .0561017 .1523417 0.37 0.713 -.2424917 .354695
dumm0 .4193639 .1537777 2.73 0.006 .117956 .7207718
dumm1 .5340008 .1590671 3.36 0.001 .2222257 .845776
dumm2 .7855772 .1642925 4.78 0.000 .46356 1.107594
dumm3 .7899929 .1661612 4.75 0.000 .4643131 1.115673
dumm4 1.059333 .1614825 6.56 0.000 .7428234 1.375842
dumm5 1.072899 .177329 6.05 0.000 .7253299 1.420468
dumm6 1.053102 .1719503 6.12 0.000 .7160757 1.390129
dumm7 1.441002 .2046819 7.04 0.000 1.03982 1.842183
dumm8 1.874835 .208242 9.00 0.000 1.466676 2.282994
dumm9 1.725453 .2102388 8.21 0.000 1.31338 2.137526
dumm10 1.567202 .2219407 7.06 0.000 1.132193 2.002211
dumm11 1.597698 .2245723 7.11 0.000 1.157531 2.037865
year
133 .6652159 .1650878 4.03 0.000 .34164 .9887917
134 .9618773 .1680508 5.72 0.000 .6324939 1.291261
135 1.852587 .1598002 11.59 0.000 1.539375 2.1658
136 .2299018 .1518721 1.51 0.130 -.067771 .5275745
137 .675096 .1426957 4.73 0.000 .3954092 .9547828
138 1.024808 .1567313 6.54 0.000 .7176108 1.332005
139 1.730319 .1593536 10.86 0.000 1.417982 2.042656
140 -.3696904 .1602825 -2.31 0.021 -.6838479 -.055533
141 .1468877 .159575 0.92 0.357 -.165883 .4596584
142 .3556635 .1622465 2.19 0.028 .0376567 .6736703
143 .9430713 .1662872 5.67 0.000 .6171446 1.268998
144 -1.238555 .1682256 -7.36 0.000 -1.568281 -.9088292
145 -.6601735 .1676607 -3.94 0.000 -.9887924 -.3315547
146 -.035841 .1550907 -0.23 0.817 -.3398224 .2681404
147 .4189803 .1750731 2.39 0.017 .0758329 .7621277
148 -1.994363 .1807396 -11.03 0.000 -2.348617 -1.64011
149 -1.386854 .1809851 -7.66 0.000 -1.741589 -1.032119
150 -.6422426 .1771769 -3.62 0.000 -.9895135 -.2949718
151 0 (omitted)
_cons 3.724768 .1627825 22.88 0.000 3.405711 4.043826
id F(2693, 40192) = 14.192 0.000 (2694 categories)
If I use instead "reghdfe", I get the following output, with one dummy omitted:
Code:
. reghdfe y dumm*,absorb(id year)
(dropped 32 singleton observations)
(converged in 7 iterations)
note: dumm11 omitted because of collinearity
HDFE Linear regression Number of obs = 42,896
Absorbing 2 HDFE groups F( 23, 40192) = 6.36
Prob > F = 0.0000
R-squared = 0.4957
Adj R-squared = 0.4618
Within R-sq. = 0.0036
Root MSE = 4.2764
y Coef. Std. Err. t P>t [95% Conf. Interval]
dumm_13 1.466534 .4071696 3.60 0.000 .6684718 2.264596
dumm_12 1.387095 .3852358 3.60 0.000 .6320241 2.142166
dumm_11 1.820428 .3539591 5.14 0.000 1.12666 2.514196
dumm_10 1.530504 .3209338 4.77 0.000 .9014666 2.159542
dumm_9 .5063514 .289606 1.75 0.080 -.0612831 1.073986
dumm_8 .467973 .2621949 1.78 0.074 -.045935 .981881
dumm_7 .4067998 .241348 1.69 0.092 -.0662477 .8798474
dumm_6 .5089955 .2089094 2.44 0.015 .0995282 .9184628
dumm_5 .0279265 .2011369 0.14 0.890 -.3663064 .4221594
dumm_4 .0309506 .1886372 0.16 0.870 -.3387826 .4006837
dumm_3 .1946444 .1742638 1.12 0.264 -.1469167 .5362055
dumm_2 .1892432 .1597645 1.18 0.236 -.1238989 .5023852
dumm0 .2862224 .1483983 1.93 0.054 -.0046418 .5770865
dumm1 .2677178 .1493352 1.79 0.073 -.0249827 .5604183
dumm2 .3861526 .1534018 2.52 0.012 .0854816 .6868236
dumm3 .2574268 .15638 1.65 0.100 -.0490816 .5639352
dumm4 .3936252 .156271 2.52 0.012 .0873305 .6999199
dumm5 .2740497 .1686489 1.62 0.104 -.056506 .6046055
dumm6 .1211116 .1572343 0.77 0.441 -.1870713 .4292945
dumm7 .3758694 .2076184 1.81 0.070 -.0310674 .7828061
dumm8 .6765613 .2160116 3.13 0.002 .2531736 1.099949
dumm9 .3940382 .2207188 1.79 0.074 -.0385758 .8266522
dumm10 .1026452 .2301371 0.45 0.656 -.3484288 .5537192
dumm11 0 (omitted)
Absorbed F(2680, 40192) = 14.534 0.000 (Joint test)
Absorbed degrees of freedom:
Absorbed FE Num. Coefs. = Categories - Redundant
id 2662 2662 0
year 19 20 1
Vincenzo
Comment