Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: comparison of independent variables

From   Andrew Reed <>
Subject   st: comparison of independent variables
Date   Tue, 22 Oct 2013 13:53:43 +0200

Hi Statalisters,

I am trying to a project and in order to do so I am trying to add an additional regression in order to discern whether or not credit rating actions levied on banks have a greater effect than those levied on sovereigns vis-a-vis exchange rate movements. I have run regressions in which I include only spillover effects from banks and then those only from sovereigns, but in order to compare these and ascertain whether or not bank spillover effects are greater than sovereign spillover effects, I run a regression including both, as follows, where d.ex is the daily change in the natural log of entity-specific exchange rates, d.lccr_bank is the first difference in the logarithmic transformation of a comprehensive credit rating score for banks, and d.lccr_sovereign is the same as the previous measure but for sovereigns. 

d.ex.i = d.lccr_bank.j*x1 +d.lccr_sovereign.j*x2 + controls + error

As can be seen, exchange rates movements are followed by ''.i'', denoting the entity identity. ''.j'' denotes all entities that are not entity ''i''. j is a pooling of all entities not equaling i.

I want to compare these two variables but am not quite sure what test I should run to see if spillovers stemming from banks are greater than those stemming from sovereigns. I cannot include a dummy for banks (something I did when measuring how a rating action affected that same entity experiencing a rating action) because all this does is show how entities absorb spillover actions, based upon if they are a bank or not, thus not answering the question of whether the actual spillover effect from banks are greater or not. Therefore I believe I am left with comparing the independent variables within the aforementioned specification. What is the best test to use? A friend of mine and I have talked about Pearson Correlation Coefficients but I am not sure this is exactly what I need.

Any advice is welcome and I thank you in advance for your input!


*   For searches and help try:

© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index