# Correct choice of analytical statistic method for time series problem

I want to make a statement about corruption (y) influenced by the ratification of the UN-convention-Against-Corruption "contract" (x). Luckily, most of them signed it within 3 years.

So, I have corruption data of the observed countries (y) 10 years before and 10 years after the ratification (x0) for those countries who signed the contract and those who did NOT sign.

• Thus, there are 2 control groups.
• It's a time series analysis.
• variable x is binary (ratified/not ratified; changing x to years of observations)

My questions:

1. What method could I employ to prove that the ratification had a negative effect on degree of corruption. (that is the hypothesis)
2. With what method could I analyse the assumed turning-point after x0 (the year of ratification)? I assume with some delay corruption will decrease. And I want to prove it with this turning point.
3. How do I compare the 2 time series of the control groups (ratified, not ratified)?
4. What are the correct names of these methods?

A few thoughts of mine to tackle this.

1. I thought I could calculate the mean of the corruption index over all countries per year. This is my new y-value for each year (x).
2. Then I would draw the regression line (what the correct name in time series?) for both control groups.
3. and then?

Any solid and clear hints are highly appreciated!

• It seems that the two types of methodology you want are an "Event Study" and "Difference in Differences". A couple of additional points. You do not have 2 control groups, you have a treated group (where the treaty was ratified) and a control group (where the treaty was not ratified). You're also going to have to make some strong assumptions on the comparability of these two groups. I suggest comparing their characteristics and any pre-ratification trends. – Greg Dec 2 '14 at 16:03
• thanks Greg. That ist a start! (why can I not upvote your contribution?) – feder Dec 5 '14 at 11:24