STATA panel data

Tobyglen

Hi all,

I have a serious problem, I'm currently doing my thesis and the deadline is looming, I'm doing a regression analysis of Credit Ratings gainst Market risk of companies over a 5 year period. I'm using the data for roughly 80 companies.

I put the data in panel set using tsset but what I'm unsure of is how to regress the Ratings against the Market risk.
Below is the code I am using. I'm using Ratings as the dependent variable and Market risk (VaR) as the independent variable for each company. Basically that means I have 80 independent variable's and 80 dependent variables.


Is it a case of,
xtreg Ratings Market risk Ratings 1 Market risk 1 etc, fe

xtreg dependentvar independentvar1 independentvar2 independentvar3 . , fe

Thanks in advance,

David

Économiste Monétaire

Did you declare the panelvar when you used tsset?

Tobyglen

Hi
sort panelvar datevar
tsset panelvar datevar

I used the Rating as panelvar and date as datevar

I'm just unsure on how to set the regression up, do I go xtreg dependent dependent 1 etc independent independent 1 etc , fe

Link I'm using is here

http://dss.princeton.edu/online_help/analysis/panel.htm

Thanks

Économiste Monétaire

Why did you use rating? Why not

tsset company date

Économiste Monétaire

Your regression is this

[latex] \displaystyle Rating_{i,t} = \alpha_{i} + \beta_{1} X_{i,t} + \ldots + \epsilon_{i,t}[/latex]

for company i and year t, correct? Your dependent variable shouldn't be the panelvar.

Tobyglen

Économiste Monétaire wrote: »

Why did you use rating? Why not

tsset company date

The point of my thesis is a regression analysis between Credit Ratings of the top European companies (Which I have) and Market risk of the companies at the time.

I basically want to see what the relationship was between them during the years around the economic crash. My supervisor recommended panel data but he's not an expert on Stata.

Legs

Hi,

I also did my thesis relating to credit ratings, but looked at the impact of downgrades on European sovereign debt prices. I presume it's an event study that you're doing to assess the impact of a credit rating change on the market risk of each company around the event date?

I did mine in excel - it's quite time consuming but can be done once you're somewhat proficient in excel. There's quite a good paper explaining a common methodology which can be used:

Steiner, M. and Heinke, V. 2001 “Event Study Concerning International Bond Price

Effects of Credit Rating Actions,” International Journal of Finance and Economics,

Vol. 6, pp. 139-157

Best of Luck!

Tobyglen

Économiste Monétaire wrote: »

Your regression is this

[latex] \displaystyle Rating_{i,t} = \beta_{0} + \beta_{1} X_{i,t} + \ldots + \epsilon_{i,t}[/latex]

for company i and year t, correct? Your dependent variable shouldn't be the panelvar.

Could you break that down into code, I'm only learning STATA since last week, thanks.

Économiste Monétaire

Depending on what you've used as labels, you may need to alter the code. Try:

tsset company year

xtreg rating risk, fe

Tobyglen

Économiste Monétaire wrote: »

Depending on what you've used as labels, you may need to alter the code. Try:

tsset company year

xtreg rating risk, fe

Sorry but I'm unsure of what to do with company,
I have 1,200 (Daily) Risk and Ratings measurements for each company and presumed I just copied them into the data editor along with the date.
My headings were Date , Ratings, Risk....Ratings 1, Risk 1...etc
My dependent variable is Rating and independent variable is Risk

I then coded in
sort Ratings Date
tsset Ratings Date

I then ran a regression analysis by
xtreg Rating Date Rating1 Date 1 etc, fe

I'm a bit lost with the company variable you talk about, should I be entering the company names into the dataset?

Tobyglen

Legs wrote: »

Hi,

I also did my thesis relating to credit ratings, but looked at the impact of downgrades on European sovereign debt prices. I presume it's an event study that you're doing to assess the impact of a credit rating change on the market risk of each company around the event date?

I did mine in excel - it's quite time consuming but can be done once you're somewhat proficient in excel. There's quite a good paper explaining a common methodology which can be used:

Steiner, M. and Heinke, V. 2001 “Event Study Concerning International Bond Price

Effects of Credit Rating Actions,” International Journal of Finance and Economics,

Vol. 6, pp. 139-157

Best of Luck!

Thanks, I'll report back when I'm finished.

Économiste Monétaire

Based on this,

Tobyglen wrote: »

I'm doing a regression analysis of Credit Ratings gainst Market risk of companies over a 5 year period. I'm using the data for roughly 80 companies.

regress the Ratings against the Market risk.

Below is the code I am using. I'm using Ratings as the dependent variable and Market risk (VaR) as the independent variable for each company. Basically that means I have 80 independent variable's and 80 dependent variables.

I presumed your data looked liked this

Company|Year|Rating|Risk
1|1|blah|blah
1|2|blah|blah
1|3|blah|blah
1|4|blah|blah
1|5|blah|blah
2|1|blah|blah
2|2|blah|blah
2|3|blah|blah
2|4|blah|blah
2|5|blah|blah
3|1|blah|blah
3|2|blah|blah
3|3|blah|blah
3|4|blah|blah
3|5|blah|blah

and so on. Here, the panelvar is Company and datevar is Year.

Tobyglen

|Daily date|Rating|Risk
date|blah|blah
date|blah|blah
date|blah|blah
date|blah|blah
date|blah|blah
date|blah|blah
date|blah|blah
date|blah|blah
date|blah|blah
date|blah|blah
date|blah|blah
date|blah|blah
date|blah|blah
date|blah|blah
date|blah|blah

This is the way my data comes from the exel file, do you think I should look at doing it the way you mention, I have 80 companies and 1,300 dates for each so would it be possible?

Daily Date on left, Rating then Risk. Thanks

Économiste Monétaire

Try it my way. Label each company 1 through 80, and mimic the table I showed above, but using daily data instead of years. Mechanically, there shouldn't be a problem with doing this in Stata.

Tobyglen

Économiste Monétaire wrote: »

Try it my way. Label each company 1 through 80, and mimic the table I showed above, but using daily data instead of years. Mechanically, there shouldn't be a problem with doing this in Stata.

Thank you, I'll try that, I really appreciate your help, I'll report back tomorrow and let you know how it went.

Économiste Monétaire

You're welcome. Best of luck.