Dear Forum,

I hope you are all doing well. I am really struggling with some content of my econometrics course. I am currently working through example data sets in Stata to get a more 'hands-on' understanding of panel data and prepare for my thesis which I have to write next year. I still massively struggle with choosing the right estimation strategy, and I hope someone can help me a bit with my thinking.

I have a 2 period panel data set (years 2011, 2014) with community data and how much residents in these communities are satissfied with waste disposal services. There are 2 variables which I am interested in:

- 1. How satisfied are you with the waste disposal services in your community (reported on a percentage scale, from 0% to 100%)
- 2. How clean do you think your community is ( 1=dirty; 2=not clean; 3=clean; 4=very clean)

I also know the budget of each community in years 2011-2014.
Between 2012-2013, the communities all receive an exogenous sum of money (meaning more budget in 2014) which differs among each community. I now want to estimate the effect of the amount of money granted on the two dependent variables.

1. So first, when I read 'treatment' in the exercise combined with panel data, I was instantly thinking about a Diff-in-Diff strategy. However, as the treatment is 'continuous' (let's say we have 10 communities which receive 1,2,3,4,5,6,7,8,9,10 euros, and communities started with different amounts of money in 2011) it is kind of hard to justify where to draw the line between treated and untreated. Diff-in-diff therefore does not work --> Is this thinking correct?

2. I then thought that using a Fixed Effects estimator should be the right strategy. I have community budget and both dependent variables for both years 2011 and 2014, so this should work right? I then thought again about it and realised that only variable 1. is continuous, whereas 2. is not. So ultimately I thought that Fixed Effects would be the correct strategy in order to estimate the effect on variable 1, but I need something else for variable 2 --> Is this thinking correct?

3. I then thought about how to best estimate the effect on variable 2. We have briefly discussed logit/probit in the lectures but I do not fully understand it. From what I got so far, could you use a logit (logistic regression) for variable 2? Does this work even though we are dealing with balanced panel data?

I hope those are not too many/too stupid questions. I really try to 'understand' the content and not just memorize it.
I would appreciate any help so much!! I feel like most pages on the internet are really really complicated.

Many thanks in advance.
Will