First, thanks for this wonderful platform to ask queries and clear doubts. I have read a few threads and based on some of those threads, I am posting my doubts
Code:
* Example generated by -dataex-. For more info, type help dataex clear input long co_code int year float(lever_w size_w dummy trade_credit_w) 11 2011 .5504084 7.692159 0 .0919173 11 2012 .5697871 7.931967 0 .08887213 11 2013 .5391156 8.025386 0 .06992412 11 2014 .51597077 8.138857 0 .027474463 11 2015 .4759782 8.160204 0 -.05730373 11 2016 .4625662 8.169874 0 .008463308 11 2017 .39045715 8.268808 1 .023023874 11 2018 .33070305 8.310906 1 .018141583 11 2019 .2839154 8.385649 1 .0127974 289 2011 .54094803 7.413488 0 .3342782 289 2012 .5555683 7.465369 0 .28388202 289 2013 .51333773 7.327781 0 .3884363 289 2014 .48577145 7.207416 0 .4057359 363 2011 .4761355 9.358657 0 .04327172 363 2012 .4731195 9.282261 0 .065523885 363 2013 .53195494 9.338382 0 .04692944 363 2014 .4994195 9.323365 0 -.00004465878 363 2015 .51551414 9.20133 0 .008859474 363 2016 .5284721 9.21132 0 -.02343703 363 2017 .5849603 9.158089 1 -.02696266 363 2018 .4416465 9.171402 1 -.1686716 363 2019 .689352 9.14435 1 .033798363 414 2016 .4542443 6.180017 0 .40248445 414 2017 .006420134 5.964607 1 .0811505 414 2018 .005577936 5.776723 1 .213511 415 2018 . 7.591811 1 .2690852 415 2019 . 7.536204 1 .11372487 771 2013 .4651715 6.630683 0 .18166228 771 2014 .5203515 6.897806 0 .24957076 771 2015 .461695 7.113387 0 .2509973 771 2016 .5473848 7.495042 0 .2136068 771 2017 .5459776 7.675732 1 .5205994 771 2018 .4951338 7.772121 1 .5433114 771 2019 .4645513 7.871731 1 .53449523 783 2011 .19831736 7.452112 0 .1440673 783 2012 .2092255 7.504777 0 .1442726 783 2013 .19775394 7.699978 0 .1591269 783 2014 .13007422 7.688272 0 .10547055 783 2015 .08677534 7.688822 0 .13824525 783 2016 .033318035 7.686621 0 .13262965 783 2017 .025697127 7.672339 1 .13933243 783 2018 .014560171 7.673084 1 .14108947 783 2019 .0009790963 7.622028 1 .16375385 1120 2011 .016338103 9.46848 0 .17673732 1120 2012 .032134537 9.61161 0 .18890423 1120 2013 .08688986 9.826855 0 .12449393 1120 2014 .05394961 10.014935 0 .13960168 1120 2015 .03865692 10.14689 0 .12071372 1120 2016 .06567325 10.239388 0 .11970769 1120 2017 .043705 10.37602 1 .1161778 1120 2018 .03458266 10.482662 1 .12359595 1120 2019 .031668555 10.609154 1 .13287362 2248 2017 .1482503 7.934478 1 .05025253 2248 2018 .15626974 8.00933 1 .04204195 2717 2011 .3987688 10.212482 0 -.1650117 2717 2012 .5396248 10.179546 0 -.06461293 2717 2013 .521483 10.359493 0 -.11464672 2717 2014 .5139446 10.229534 0 -.05649667 2717 2015 .4747386 10.36415 0 -.08115133 2717 2016 .31001255 10.589538 0 -.11957822 2717 2017 .26234335 10.676764 1 -.10998888 2717 2018 .2209677 11.94411 1 -.032753333 2717 2019 .19191967 11.893708 1 -.008160293 2842 2011 .24950735 6.491785 0 .3027133 2842 2012 .3842184 6.770675 0 .3133387 2842 2013 .3622974 6.984253 0 .21834183 2842 2014 .3780394 7.066552 0 .2918693 2842 2015 .3651952 7.071573 0 .27504244 2842 2016 .19154836 6.804171 0 .3693434 2842 2017 .012137886 6.426327 1 .2519825 2842 2018 .11468551 6.541318 1 .29500863 2842 2019 .1701168 6.762383 1 .2371921 3335 2011 .3992996 7.243656 0 .3574185 3335 2012 .1974824 7.581821 0 .1959026 3335 2013 .15542907 7.733684 0 .14128721 3335 2014 .02286336 7.778254 0 .07554123 3335 2015 .028543843 7.775822 0 .12504724 3335 2016 .03890818 7.826403 0 .1643721 3335 2017 .05525101 7.927613 1 .14483556 3335 2018 .10858244 8.108293 1 .0895572 3335 2019 .2077677 8.2735405 1 .14501888 3990 2011 .5109358 8.563141 0 .10884224 3990 2015 .4456229 9.215437 0 .1399847 3990 2016 .43831205 9.308274 0 .12745605 3990 2017 .4056455 9.385167 1 .0745315 3990 2018 .3955588 9.53976 1 .07392676 3990 2019 .3502277 9.585972 1 .14768772 3998 2011 .37776425 9.546169 0 .15586448 3998 2012 .3780212 9.717851 0 .1495952 3998 2013 .39086115 9.987824 0 .10335066 3998 2014 .3920296 10.192468 0 .03823784 3998 2015 .409129 10.28826 0 .07513884 3998 2016 .4354037 10.297828 0 .07631767 3998 2017 .4468606 10.463384 1 .06948911 3998 2018 .4743695 10.69 1 .07081716 3998 2019 .4099103 10.978138 1 .08838603 4024 2018 .4125189 9.126492 1 .09601225 4024 2019 .4504824 9.173573 1 .09431474 4030 2016 .17944816 6.26435 0 .3450047 4030 2017 .1738609 6.726233 1 .27482015 end
Code:
**Here Dummmy takes the value 1 if year falls in range 2017-2019 and 0 otherwise
**Setting the panel
xtset co_code year
*1. Running regression with firm and year fixed effects (xtreg)
xtreg lever_w dummy i.year,fe vce(r)
/*. xtreg lever_w dummy i.year,fe vce(r)
note: 2019.year omitted because of collinearity.
Fixed-effects (within) regression Number of obs = 98
Group variable: co_code Number of groups = 15
R-squared: Obs per group:
Within = 0.2477 min = 2
Between = 0.0748 avg = 6.5
Overall = 0.0669 max = 9
F(8,14) = 1.78
corr(u_i, Xb) = 0.0080 Prob > F = 0.1639
(Std. err. adjusted for 15 clusters in co_code)
------------------------------------------------------------------------------
| Robust
lever_w | Coefficient std. err. t P>|t| [95% conf. interval]
-------------+----------------------------------------------------------------
dummy | -.0949773 .0485165 -1.96 0.071 -.1990348 .0090803
|
year |
2012 | .0128236 .0322462 0.40 0.697 -.0563376 .0819848
2013 | -.0024029 .0361445 -0.07 0.948 -.0799252 .0751194
2014 | -.027591 .0464378 -0.59 0.562 -.1271902 .0720083
2015 | -.0452929 .0456127 -0.99 0.338 -.1431224 .0525365
2016 | -.0478451 .0488754 -0.98 0.344 -.1526725 .0569822
2017 | -.0147827 .0322477 -0.46 0.654 -.0839471 .0543817
2018 | -.0287685 .0300549 -0.96 0.355 -.0932299 .035693
2019 | 0 (omitted)
|
_cons | .3690917 .0315912 11.68 0.000 .3013354 .436848
-------------+----------------------------------------------------------------
sigma_u | .16558787
sigma_e | .08819011
rho | .77902858 (fraction of variance due to u_i)
------------------------------------------------------------------------------
*/
*2. Running regression with firm and year fixed effects by using user written command reghdfe without absorbing
reghdfe lever_w dummy i.year i.co_code , noabsorb vce(r)
/*
. reghdfe lever_w dummy i.year i.co_code , noabsorb vce(r)
(MWFE estimator converged in 1 iterations)
note: 2019.year omitted because of collinearity
HDFE Linear regression Number of obs = 98
Absorbing 1 HDFE group F( 22, 75) = 96.26
Prob > F = 0.0000
R-squared = 0.8369
Adj R-squared = 0.7890
Within R-sq. = 0.8369
Root MSE = 0.0882
------------------------------------------------------------------------------
| Robust
lever_w | Coefficient std. err. t P>|t| [95% conf. interval]
-------------+----------------------------------------------------------------
dummy | -.0949773 .0484797 -1.96 0.054 -.1915539 .0015993
|
year |
2012 | .0128236 .0427197 0.30 0.765 -.0722784 .0979257
2013 | -.0024029 .0383865 -0.06 0.950 -.0788729 .074067
2014 | -.027591 .0412448 -0.67 0.506 -.1097548 .0545729
2015 | -.0452929 .0395357 -1.15 0.256 -.1240521 .0334662
2016 | -.0478451 .0405997 -1.18 0.242 -.1287239 .0330336
2017 | -.0147827 .0453062 -0.33 0.745 -.1050373 .0754719
2018 | -.0287685 .0404742 -0.71 0.479 -.1093971 .0518602
2019 | 0 (omitted)
|
co_code |
289 | .0217886 .0238414 0.91 0.364 -.025706 .0692832
363 | .0690747 .0424434 1.63 0.108 -.0154769 .1536264
414 | -.2572126 .1201764 -2.14 0.036 -.4966163 -.0178089
771 | .058144 .0336722 1.73 0.088 -.0089344 .1252224
783 | -.3580223 .0259388 -13.80 0.000 -.4096951 -.3063496
1120 | -.4128115 .027637 -14.94 0.000 -.4678672 -.3577558
2248 | -.2373975 .0281001 -8.45 0.000 -.2933757 -.1814193
2717 | -.0761221 .0406344 -1.87 0.065 -.1570699 .0048257
2842 | -.2101284 .0405275 -5.18 0.000 -.2908633 -.1293935
3335 | -.3227527 .0457242 -7.06 0.000 -.41384 -.2316654
3990 | -.0117564 .0250256 -0.47 0.640 -.0616099 .0380971
3998 | -.0449503 .0357334 -1.26 0.212 -.1161349 .0262343
4024 | .0344518 .0316169 1.09 0.279 -.0285323 .0974359
4030 | -.2509533 .0336681 -7.45 0.000 -.3180236 -.183883
|
_cons | .5064104 .0359909 14.07 0.000 .4347129 .5781079
------------------------------------------------------------------------------
*/
*3. Running regression with firm and year fixed effects (reghdfe) with absorbing co_codes only
reghdfe lever_w dummy i.year, a(co_code ) vce(r)
/*
. reghdfe lever_w dummy i.year, a(co_code ) vce(r)
(MWFE estimator converged in 1 iterations)
note: 2019.year omitted because of collinearity
HDFE Linear regression Number of obs = 98
Absorbing 1 HDFE group F( 8, 75) = 3.00
Prob > F = 0.0057
R-squared = 0.8369
Adj R-squared = 0.7890
Within R-sq. = 0.2477
Root MSE = 0.0882
------------------------------------------------------------------------------
| Robust
lever_w | Coefficient std. err. t P>|t| [95% conf. interval]
-------------+----------------------------------------------------------------
dummy | -.0949773 .0484797 -1.96 0.054 -.1915539 .0015993
|
year |
2012 | .0128236 .0427197 0.30 0.765 -.0722784 .0979257
2013 | -.0024029 .0383865 -0.06 0.950 -.0788729 .074067
2014 | -.027591 .0412448 -0.67 0.506 -.1097548 .0545729
2015 | -.0452929 .0395357 -1.15 0.256 -.1240521 .0334662
2016 | -.0478451 .0405997 -1.18 0.242 -.1287239 .0330336
2017 | -.0147827 .0453062 -0.33 0.745 -.1050373 .0754719
2018 | -.0287685 .0404742 -0.71 0.479 -.1093971 .0518602
2019 | 0 (omitted)
|
_cons | .3690917 .0327578 11.27 0.000 .3038348 .4343486
------------------------------------------------------------------------------
Absorbed degrees of freedom:
-----------------------------------------------------+
Absorbed FE | Categories - Redundant = Num. Coefs |
-------------+---------------------------------------|
co_code | 15 0 15 |
-----------------------------------------------------+
*/
*4. Running regression with firm and year fixed effects (reghdfe) with absorbing both co_codes and year
reghdfe lever_w dummy , a(co_code year ) vce(r)
/*
reghdfe lever_w dummy , a(co_code year ) vce(r) //absorbing co_codes and year
note: dummy is probably collinear with the fixed effects (all partialled-out values are close to zero; tol = 1.0e-
> 09)
(MWFE estimator converged in 6 iterations)
note: dummy omitted because of collinearity
HDFE Linear regression Number of obs = 98
Absorbing 2 HDFE groups F( 0, 75) = .
Prob > F = .
R-squared = 0.8369
Adj R-squared = 0.7890
Within R-sq. = 0.0000
Root MSE = 0.0882
------------------------------------------------------------------------------
| Robust
lever_w | Coefficient std. err. t P>|t| [95% conf. interval]
-------------+----------------------------------------------------------------
dummy | 0 (omitted)
_cons | .3150925 .0089085 35.37 0.000 .2973458 .3328393
------------------------------------------------------------------------------
Absorbed degrees of freedom:
-----------------------------------------------------+
Absorbed FE | Categories - Redundant = Num. Coefs |
-------------+---------------------------------------|
co_code | 15 0 15 |
year | 9 1 8 |
-----------------------------------------------------+
.
*/1. Why the year 2019 got omitted in 1,2, &3? Is it to avoid the issue of dummy variable trap? Why 2019 and not some other year?
2. Point estimates of 1 & 2 are the same but robust standard errors are different, why is it so?
3. What is the difference between 3 & 4? Why 'dummy' variable got omitted in (4) which was there in 2 & 3.
4. When and under what conditions we should invoke the 'absorb' option
5. Finally, if it makes sense to say a 'good estimation method' in the above, which one is it? Apologies for such a mundane question
Can someone take the trouble to help me here?
No comments:
Post a Comment