Interpretation of variables in panel data set.

[Kane Griffiths]

I'm conducting some research for my dissertation on the relationship between military spending, arms trade and economic growth, however one of my variables is particularly difficult for me to interpret. The variable is a formula where net arms exports is calculated by: (Net Arms Exports - Net Arms Imports)/(Net Arms Exports + Net Arms Imports), hence creating a range of values from -1 to 1. I'm not too sure particularly of what units the interpretation would be in.

[Phil Bromiley]

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.

The interpretation of the meaning of your variables is a matter of substance, not statistics. You are the substance expert. You shouldn't be running models where you don't have a good reason for each of the measures. I don't understand exactly why the normalization, but the top of the ratio is net arms flows (export - import). So this is a measure of the proportion of international arms transactions by which exports exceed imports.

From a statistics standpoint, the parameter on this variable would be the expected change in the dv for a one unit (e.g., from 0 to 1) in this variable.

[Phil Bromiley]

Note that the interpretation will be a bit different depending on what panel estimator you run. A fe model will take out the mean effect and only look at variation within panels. That is, the fe will largely wash out whether a country is in general an exporter or importer of arms (or has large or small military spending), and only look at change over time around that base. A re estimate will combine within and across country effects. You might also look at xthybrid which estimates between and within at the same time.

[Kane Griffiths]

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.

The interpretation of the meaning of your variables is a matter of substance, not statistics. You are the substance expert. You shouldn't be running models where you don't have a good reason for each of the measures. I don't understand exactly why the normalization, but the top of the ratio is net arms flows (export - import). So this is a measure of the proportion of international arms transactions by which exports exceed imports.

From a statistics standpoint, the parameter on this variable would be the expected change in the dv for a one unit (e.g., from 0 to 1) in this variable.

Thank you for your help Phil, my decision to normalization was one based off the literature. I believe the reasoning behind it is because the data is taken from many different years, so the data from a long time ago is measured is measured in 1000's while the data more recently is measured in billions. So normalizing just removes the need to have to change the data I think. I don't know if this is a silly suggestion as well but could I multiply the ratio by 100 to obtain a range between -100 and 100, then interpret it as a percentage point increase in net arms exports.