Help with inteaction effects-both independ.variables dummies

[gishena]

http://postimg.org/image/fwlztnuwj/

This is my regression - both crisis and distres are dummies. The interacton effect is whether a firm in the crisis that was a client of a distressed bank had lower or higher target leverage. I know how to interpret it when one independable variable is a dummy but I do not know how to do it when both are. Please help.

Thanks in advance!

[popcorn]

Okay... firstly, I would warn that you should have other variables in your model as an "adequate model" should not ONLY include dummies. Maybe go back to the theory and include another variable that you could use.

In terms of interpretation, check on whether the dummies are statistically significant or not. If you re-look at your results, only one of the dummies is statistically significant. How to interpret this, you need to sum this value with the intercept coefficient. Woodridge gives a good guide on this.

If the dummies are statistically insignificant, this means that they do not have any impact on the dependent variable.

Your question relates to the interactive dummy. If the statistically significant dummy was just say Crisis, then you would interpret the impact of the crisis only. However, if it is debt crisis (interactive), AND it is statistically significant, then you would interpret the impact of BOTH crisis and debt TOGETHER on the dependent variable.

Hope that this clarifies things

[startz]

If the dummies are statistically insignificant, this means that they do not have any impact on the dependent variable.

This is wrong. If a coefficient is insignificant, it means it has no effect or that the evidence is too weak to determine whether there is an effect. Unless the standard errors are small, there is no reason to believe there is no effect.

[popcorn]

Okay okay,,, I stand corrected. What I meant was if the coefficient of the dummy is statistically insignificant. One can debate on the issue of no effect/ very weak effect. The main thing is that the impact does not differ. For example, if one is looking at whether being a male or female would affect wages, and the coefficient for the dummy for female was statistically insignificant, then this would show that whether one was a male or female would have little difference on the impact on wages.

One should focus on the question being asked... how does one interpret interactive dummies... which I think I answered or at least tried to answer above

[startz]

For example, if one is looking at whether being a male or female would affect wages, and the coefficient for the dummy for female was statistically insignificant, then this would show that whether one was a male or female would have little difference on the impact on wages.

Still no. It's the size of the coefficient that shows whether being male of female would have little difference on the impact on wages. If the coefficient is large but insignificant, then our best estimate is that gender makes a big difference-but there would be considerable uncertainty about whether that conclusion was right.