Regression on unbalanced panel

[Grace Eneel]

Hi,

I'm currently working on impact of risk - taking behavior on firm growth, and here is my panel: (the number wasn't real)

NAME	YEAR	GROWTH	INCOME	AGE	SIZE	RISK	PERFORMANCE
A	2000	12	8	6	6	3	4
A	2001	14	6	7	6	2	3
A	2002	15	9	7	6	2	2
A	2003	16	5	6	4	3	3
B	2000	14	3	4	3
B	2001	17	2	3	4
B	2002	13	5	2	2
B	2003	12	3	5	9
C	2000	22	2	6	3
C	2001	17	7	7	4
C	2002	22	4	4	5
C	2003	34	4	3	7

My panel is unbalanced.

According to XU Peng's paper, Risk taking and firm growth
RISK: the standard deviation of EBITDA(t)/Assets(t) over 4 years.
Performance : sum of EBITDA(t)/Assets(t) over 4 2000-2003.

So I calculated EBITDA/Assets of each firm each year.
Performance of 2000= sum(Ebitda/assets firm A 2000, ebitda/assets firm B 2000 and so on)
RISK 2000= stdev.p(Ebitda/assets firm A 2000, ebitda/assets firm B 2000 and so on)

4 rows, 4 year of RISK and PERFORMANCE

So my questions are:
1. Did I calculate RISK and PERFORMANCE right?
2. How can I regress those, with
growth= risk + control variables
performance=risk +control variable
risk = age + size + ownership+ leverage + income
3. How about auto-correlation Wooldridge test, White test and Variance inflation factor (VIF) test, are they can be run normally?

I'm a beginner and I do this research for requirement, so I don't know much about this. Thanks for your time. I'm appreciated any help.

[Phil Bromiley]

You didn't get a quick answer. You'll increase your chances of a useful answer by following the FAQ on asking questions - provide Stata code in code delimiters, readable Stata output, and sample data using dataex. It is best if you give full cites for papers

Since you don't provide actual code, it is hard to know what you've done. There are some serious problems in using multi year data to calculate risk when you have models that suggest the expected value of risk is not stable. It is also creating possible endogeneity since performance appears in the risk measure.

We don't normally help with homework.