Hypothesis testing in Stata

[Ana Ludwig]

Hi,

for my Thesis I investigate the effect of ESG scores on the financial performance of companies. My regression looks like: xtreg LN_TobinsQ total_esg_score Size EPS LN_leverage Age [Size EPS LN_leverage and Age are control variables].
Now my second Hypothesis is: Young companies have a more green footprint (higher ESG scores) than Old companies [in terms of Age]

I only do not know how to perform this in Stata (16.1). I tried to make a dummy variable of Age in such a way that AgeYoung = Age <= 25 years and AgeOld = Age > 25

Could someone help me try to figure this out. Thanks in advance!

[Carlo Lazzaro]

Ana:
welcome to this forum.
The best way is to have a two-level catagorical variable (let's call it -Age_threshold-):

g Age_Thresold=0 if Age<=25
replace Age_Thresold=1 if Age>25 & Age!=.

as reported in the following toy-example:

. set obs 5
number of observations (_N) was 0, now 5

. g Age=runiform()*100

. g Age_threshold=0 if Age<=50

. replace Age_threshold=1 if Age>50 & Age!=.

. label define Age_threshold 0 "young" 1 "old"

. label val Age_threshold Age_threshold

. list

     +---------------------+
     |      Age   Age_th~d |
     |---------------------|
  1. | 34.88717      young |
  2. | 26.68857      young |
  3. | 13.66463      young |
  4. | 2.855687      young |
  5. | 86.89333        old |
     +---------------------+

.

[Ana Ludwig]

Thank you for welcoming me and for responding so quickly!
Okay I understand that, and did that in Stata!

is it now possible to apply the following code

xtreg TotalESG Age_Thresold EPS Size LN_leverage

My main goal is to answer if younger REITS (=< 25 year) do have higher ESG scores than older REITS (> 25 years)

[Carlo Lazzaro]

Ana:
you should exploit the wondeful capabilities of -fvvarlist- notation:

xtreg TotalESG i.Age_Thresold EPS Size LN_leverage

In addition: have you carefully ruled out the need for cluster.robust standard errors?

[Ana Ludwig]

I first runned pooled OLS and performed the White test (code: imtest, white). Indeed there was homoskedasticitiy. However, since I ultimately run panel data regression I can cluster the standard errors in this way:
xtreg TotalESG i.Age_Thresold EPS Size LN_leverage, fe vce(robust) It is the same as I did with my first hypothesis which is: ESG rating postively affects Tobin's Q xtreg TobinsQ TotalESG EPS Size LN_leverage Age, fe vce(robust) However, I might be incorrect because I am still confused sometimes

[Carlo Lazzaro]

Ana:
the main difference with non-deafault standard errors in -regress- and -xtreg- is that under -regress- -robust- and -vce(cluster clusterid)- take heteroskedasticity and autocorrelation into account, respectively, whereas they do the very same job under -xtreg- (that is, they take both heteroskedastcity and/or autocorrelation of the systematic error into account), beacuse both options call the cluster-robust standard errors.

[Ana Ludwig]

Hi, I have another question. In order to conduct tables into world, I normally use the option asdoc. However, I was wondering how to get a outcome such as this table ?
What stata command to use? Normally I use asdoc summarize LN_Total_assets LN_Total_liabilities Cash FFO MV LN_MV NAV LN_NAV leverage LN_leverage TobinsQ Size, stat(mean median sd p25 p75 N) col(stat) save(Controlvariables.doc) label replace title(Table 1B Summary statistics Variables 2010-2019).