Help Needed - Basic Problem

[tomtomm]

Hey guys, am completely new to this forum so if this is in the wrong place then I apologise!

Basically I'm an absolute newbie to Eviews and need some help.

I'm currently doing a dissertation studying the effectiveness of foreign aid in a country in Africa. I have undertaken a pretty thorough literature review and studied recognised growth models as well as papers that created new models to try and measure the effectiveness (for anyone familiar with these I have studied Chenery-Strout, Boone, Burnside-Dollar among others).

From this Lit. Review I've got a fairly good idea of the variables I want to be using however when I first put them into the Eviews equation, the output suggested that they were all statistically insignificant, which I find pretty hard to believe. I have read through both help guides but couldnt find anything that would help explain why and it was just getting rather annoying.

If anyone could help me then that'd be brilliant, either through this topic conversation or I could PM you a bit more in depth of what I am doing (what the output says etc). Sorry if this seems like a waste of time for you guys as you all seem pretty clued up on Eviews and this is an extremely novice question..I just wasnt really sure where to turn for help!

Thanks in advance if any of you can help.

[startz]

You might post a sample regression so folks can have a more specific idea what you've done, how many observations you have, etc.

[tomtomm]

Thanks for your reply.

I've returned home from the library so unable to access Eviews right now but I'll post my output tomorrow if it'll help. I'm only observing over 10 years, which means that I only have 11 observations which I know is pretty low. Had planned to do more years however the presence of data, or lack of as it were, meant that I had to do a shortened time span. I'm using 10 variables, though this is pretty flexible for now - my first regression was just a tester but I was surprised at the outcome..P-values being pretty high etc. suggesting insignificance.

[startz]

Getting good results with 11 observations and 10 variables is pretty hopeless. You're likely doing everything right, but your data just doesn't contain enough information.

[tomtomm]

I was thinking that myself, I feel I may have limited myself by sacrificing observations just so I could include more of the variables I wanted. Will try again using more observations..I may even have to use another country to add to the observations and change my dissertation title slightly.

[tomtomm]

After adding observations and reducing variables I have achieved this regression output, which is significantly better than before:

Dependent Variable: GDPCHANGE
Method: Least Squares
Date: 04/13/10 Time: 14:09
Sample: 1988 2008
Included observations: 21

Variable Coefficient Std. Error t-Statistic Prob.

INFLATION -0.000939 0.034348 -0.027335 0.9786
LACTIVEPOP -34.78800 24.30556 -1.431277 0.1743
LAID -2.658923 2.934423 -0.906114 0.3802
LBUDGEXP 1.069556 3.649803 0.293045 0.7738
LFDI 1.200523 0.556212 2.158390 0.0487
LIMPORTS 7.478531 3.346070 2.235019 0.0422
C 258.5749 156.6515 1.650637 0.1211

R-squared 0.519460 Mean dependent var 6.923810
Adjusted R-squared 0.313514 S.D. dependent var 2.401646
S.E. of regression 1.989871 Akaike info criterion 4.475218
Sum squared resid 55.43419 Schwarz criterion 4.823392
Log likelihood -39.98979 Hannan-Quinn criter. 4.550781
F-statistic 2.522314 Durbin-Watson stat 2.415494
Prob(F-statistic) 0.072139

The number of observations is still low however as I am only studying one country and data is only available annually then I can only stretch back most of my data to the late 1980s/early 90s.

From anyone's experience should I be doing any further tests from this? I'm unsure as to where to go from here, my knowledge on Eviews is pretty basic as you can tell! Should I do an ADF test?

[startz]

I think there are really two answers. The first is that you still have very little information. You can't expect much in the way of statistical results. The second point is that you have what are probably endogenous variables on the right. Least squares doesn't give the correct answers in this situation, but with so little data there probably isn't much else you can do.

[tomtomm]

The only method I should be using is Least Squares. I appreciate there are very little observations however if I'm trying to study one country then I realistically cannot obtain widespread data from any years before 1988, and all this data is annual. Can I really not obtain answers from so few observations? Should I consider studying other countries also to increase the observations?