Please help: Fixed effect ----error: "near singular matrix"

[ummuaarifa]

Hi All,

my name is Intan. I am a newcomer to Eviews...
Currently i am doing a thesis to study the determinants of capital structure across 172 companies (from 8 industries) for period 2000-2011..
I want to see the effect of different industry classification on debt ratio, hence i create 7 dummy variables (DM1...DM7)..
my estimation model look like this:
LTD_RATIO C TANG PROF SIZE GROWTH BRISK NDTS INF_RATE GDP_RATE STOCK_MARKET_SIZE DM1 DM2 DM3 DM4 DM5 DM6 DM7

however, when i tried to estimate the fixed effect model using those equation, i failed and got an error message "Near Singular Matrix"
i suspect it is because of multicollinearity problem with my data, but i just couldnt find out why..is it because i have too many dummy variables?

can someone enlighten me on what the problem is with my data and how to correct it?
attach is my workfile..

thanks in advance..

[EViews Gareth]

If it fails without fixed effects, you've fallen in to the dummy variable trap.

[ummuaarifa]

hi gareth,

it runs OK with OLS..
but it gives me the "near singular matrix" when i tried the cross-section fixed effect...

i need to account for the industry effect, hence i include the 7 dummy variable (for 8 industries)..
how can i get thru this dummy variable trap while still using my 7 Dummy for cross-section fixed effect option?
btw, it doesnt show "near singular amtrix" error when i try period fixed effect option...

why it give different response between period and cross-section fixed effect?
in my model, which one is more appropriate? period or cross-section effect? i have those 7 industry dummies that dont vary with year...also, i have GDP and INF variables that dont vary across the 172 companies...

[startz]

The dummies plus constant are probably perfectly collinear with the cross-section fixed effects.

[crow1994]

Good day. I have problem with "near singular matrix" too. I want to make Chow Test (i want to know RSS for EQ with dummy = 1 and dummy = 0), but when I make an equtation and write in sample "collbarg = 0" ( collbarg is a dummy, 0 or 1), it tells me "near singular matrix". What can I do to improve the situation?

[startz]

You probably have some variable that is a constant when collbarg equals zero.