# Researcher ignored huge coefficients in research results. Why?

I apologize beforehand. This is a stupid question.

I'm doing article review on one research paper, and I'm bamboozled by its results. The dependent variable in this OLS is the log of ratio with results ranging from -2 to 2. The research is focused on the connection between military effectiveness and economic development analyzing hundreds of battles in the last century. The part that I don't understand is how can researcher ignore the highlighted values and talking only about the effect of variable GDP per capita (and others, but this one is the most important).

My knowledge of statistics is limited. I can understand OLS somehow, but I have never seen this scenario before. I can't understand why the highlighted variables with huge coefficients don't overshadow every other variable to the point where you can't reliably research their influence, because possible error in "Artillery" variable is bigger than Beta of almost any other variables.

In case of more information needed, write a comment or see the article (section Empirical Analysis).

Even though the number is big compared to other, here, you should also consider the standard error (presumably the number in parentheses), which is also quite big. This measure shows the variability of the estimate : the lower, the better. The ratio of the value, i.e. $$10.70/4.72 = 2.42$$, which is significant (there is indeed a grand prize star!), is most helpful to envision the magnitude of the effect. Compared to this value, consider also the result : $$.78/.21=3.71$$ , which is somewhat much better (two stars!).

Not sure why the researcher(s) omitted the result. Maybe it was not of interest. My idea is that, because the effect vanished in model 2, they did not want to point it out. But I might be wrong, I only base my statement on this very table.