# Proxy for financial crisis

I'm trying to do some empirical work and trying to find a good proxy for the financial shock of 2008. My first idea was to consider the change in real interest rate between 2007 and 2008 but I don't know if it's good or not.

Just for a general idea, I'm trying to find the relationship between GDP per capita, aging, and the shock of 2008.

If anyone has any good literature on where to start I'd be thankful.

• Are you trying to control for the impact of the financial crisis in a model of long-term (trend/potential) GDP growth? In that case there is no proxy that can be considered standard, because the identification of the impact itself depends on how you model potential growth.
– BrsG
Jul 14 at 8:47
• Yes, I'm trying to do that. From what I was reading volatility index was said to be a good proxy or bond spread. But I am not very knowledgeable in finance field so I wasn't sure if it's good to consider these or not Jul 14 at 11:32
• The VIX and indices like the St. Louis Fed [FSI][1] may be useful in many cases to control for the short-term impact. However it may not be helpful in controlling for damage to the economy that lasts well beyond the short-term impetus. [1]: fred.stlouisfed.org/series/STLFSI2
– BrsG
Jul 14 at 13:20
• I checked that and it would have been really helpful but it is only for some countries and I need world data. So that's why I'm stuck. Jul 14 at 13:24
• The ECB has an consistent indicator for the US, Europe, China. But I don't believe you won't find one for many more economies anywhere else. Beware of the longer-term issue. sdw.ecb.europa.eu/browse.do?node=9689686
– BrsG
Jul 14 at 13:27

I am not sure what you mean by proxy for the financial shock of 2008.

If you simply want to control for the 2008 recession, or estimate its effect why don't you simply include dummy following the NBER recession indicator.

Shocks are part of the error term. If you would run naive OLS model:

$$\ln Y_ = \beta_0 + \beta_1 \ln C_t + \beta_2 \ln I_t + \beta_3 \ln G_t + \beta_4 \ln NX_t + \epsilon_t$$

shocks to output are given by $$\epsilon_t$$, if you want you can impose some structure on that in various models, but one way or another the unexpected shocks (and financial crises are by definition unexpected) to the system will come from the error term.

• The notation suggests a regression of GDP over its spending components (in logs). Wouldn't the residual then capture a shock to the cointegration relationship (the composition of GDP), rather than the impact financial crisis itself?
– BrsG
Jul 14 at 8:26
• @BrsG you are right, that’s why I call it “naive” model eg pretending that there are no issues. Of course, you are right, and not just that here there is even simultaneity so using error is even more problematic, but I just wanted to have some quick equation to use as a backdrop for explanation
– 1muflon1
Jul 14 at 8:34
• That would be great I guess but from what I'm seeing (and correct me if I'm wrong) is that it only includes the US? Is there global data on this indicator? Jul 14 at 11:37