Panel Data Fixed effects wrong signs

[Maria Sevillano]

Good morning,
I am carrying out a research to test whether oil and natural gas rents (measured as a % of GDP) increase or decrease military spending for a panel of 17 MENA (Middle East and North Africa) countries in the period 2008-2011. I am not sure if the model that I am using is the best, if there are problems with the data or if I am going in the wrong direction.
I have used the command :

xtreg milex oil gas conflict gdpgrowth unemployment corruption accountability, fe

It suggests that oil rents do significantly impact military spending but not in the direction expected. When oil rents increase, military spending decreases (-.095). The rest of the variables also seem to have the wrong sign because when GDP growth increases milex decreases, larger control of corruption decreases milex and higher accountability increases milex. Which is contrary to what previous research has found. Furthermore, R2 is not really high.
Additionally, if I run a simple OLS regression using "regress" the signs of the coefficients seem normal.

xtreg milex oil gas conflict gdpgrowth unemployment corruption accountability, fe
Regression results

Milex	Coef.		St.Err.	t-value		p-value	[95% Conf		Interval]		Sig
Oil	-.095		.015	-6.27		0	-.125		-.065		***
Gas	.219		.157	1.40		.165	-.091		.529
Conflict	.353		.388	0.91		.365	-.414		1.119
Gdpgrowth	-.03		.009	-3.55		.001	-.047		-.013		***
Unemployment	-.063		.066	-0.97		.335	-.193		.066
Corruption	-2.017		.558	-3.62		0	-3.118		-.915		***
Accountability	.904		.397	2.28		.024	.121		1.688		**
Constant	6.304		1.04	6.06		0	4.25		8.358		***
Mean dependent var		4.625			SD dependent var			2.540
R-squared		0.396			Number of obs			176
F-test		14.232			Prob > F			0.000
Akaike crit. (AIC)		494.786			Bayesian crit. (BIC)			520.150
*** p<.01, ** p<.05, * p<.1

I would appreciate any help Thanks!

[Jared Greathouse]

I don't do much theoretical work, but I happened to do a little, non-serious work on this exact subject in undergrad. Let's talk about it then. Explain to me why you think the sign is wrong.

[Maria Sevillano]

Jared Greathouse Hi Jared. What I have read from other scholars analysing the relationship between oil rents and military spending, they almost in all projects found that oil rents increase military spending. Additionally, GDP growth and corruption also usually increase spending in the military. Nevertheless, my results seem to show exactly the opposite and I don't know if it's because there's a problem with the data (some countries such as Qatar or Libya don't have much data for military spending in the period analysed) of a problem with the model. I would appreciate any help (or if you believe the output is okay, what could be an interpretation). Thanks!

[Andrew Musau]

Good morning,
I am carrying out a research to test whether oil and natural gas rents (measured as a % of GDP) increase or decrease military spending for a panel of 17 MENA (Middle East and North Africa) countries in the period 2008-2011.

Typical macro panels have at least \(T=30\) years as macroeconomic variables are slow-moving. Variables like GDP, unemployment and corruption hardly change over a short time period (unless there is some kind of shock) and 4 years would be considered as such. If "oil" represents oil revenues, then I am confident in saying that not all17 MENA countries are oil exporters. You would then be reporting 0 revenues for net importers, so perhaps you need to think about whether you should consider negative revenues (costs) for importers. You should also be controlling for country size (log population) as larger countries in general have higher military expenditures.