I am writing a paper on European poverty and I am using a treatment control setup similar to Difference-in-Difference methods. Now, since the data I am using is rather recent, I would like to condition on "austerity levels", i.e. I would like to make sure that countries in the control group and those in the treatment group are equal in terms of austerity that has been employed.
This is what I have so far:
First of all I defined austerity as umbrella term for the following:
- Public wage and employment cuts
- Cuts in welfare expenditure
- Rescaling labour protection institutions
- Changes in pension age
- Cutting the minimum wage
- Ad-hoc tax rises
First, I thought about using GDP change as a measure. Oviously, there might be a host of other reasons why GDP changes, especially when there has just been (or there still is) a large economic downturn (defined by declining GDP).
Then, I thought of social welfare spending as an indicator. However, welfare spending might also (and indeed does) increase in times of crisis, because more people might become unemployed and need support, even though the individual height of the help has decreased. Moreover, different countries might have chosen completely different approaches to austerity. E.g. Italy has chosen to increase the tax burden, to change the pensionary system and to fight inefficiencies in taxation and government, while Latvia has primarily chosen to cut down public salaries (sometimes by 50%), while Greece seems to try both desperately. That means that it is not so clear that austerity always reflects the same way and especially that social welfare spending is a bad proxy.
Thirdly, I thought of the height of government debt as a proxy, the logic being: countries which are highly indebted might be forced to cut down on the welfare state. Again, this is a rather weak predictor for austerity, simply looking at some examples: Latvia, with a public debt of 40% of GDP in 2012 had to go through a multilateral bailout, while Germany, with a public debt of about 80% was involved in many bailouts in the recent years.
What I came up with in the end is the following: the 10 year government bond interest rates show how much confidence financial markets have in the country in question, moreover high interest rates simply force the country to save, in order to stand a realistic chance of being able to service the debt in future. Therefore, conditioning on the 10-year government bond interest rate should be able to control for at least some of the austerity effects.
My problem is that I didn't find many recent year studies using a treatment control approach for aggregated data, and even less which would describe the choice of the right austerity indicator and I am not really trusting my logic. It seems like I am missing something or that I am making life somewhat too easy.
I would be really interested in hearing your thoughts and your critique, or in the best case someone knows a well-established austerity indicator. Thank you in advance!