Interaction between 2 dummies - interpretation

[Guest]

Hi guys,

In short: does it matter if sustainability reports are audited etc. I measure this with the bid-ask spread and a range of control variables.
Independent variables of interest: dummy variable for assurance and for big 4 auditor.

Scheme:
if assured: 1, if not 0.
if assured by big4: 1, if not 0.

Of course, one cannot be audited by a big4, if not being assured. So there is no case where big4 is 1 and assurance is 0.


Question to you: how to interpret/measure the additional effect of a big4 assurance provider?

Regression: Spread = DummyAssurance##DummyBig4 + Control variables (stock price etc)

Output:
I get the coefficient for assurance, big4, but not the interaction. That one is omitted because of collinearity...

Do you have any ideas? (sorry for the output)

______________________lnspready | Coef. Std. Err. t P>|t| [95% Conf. Interval]
____________1.c_externalassurance | -.0535894 .0520514 -1.03 0.303 -.1556631 .0484844
________________1.c_provider_big4 | .0493626 .054771 0.90 0.368 -.0580444 .1567696
_______________________________|
c_externalassurance#c_provider_big4 |
____________________________0 1 | 0 (empty)
____________________________1 1 | 0 (omitted)
_______________________________|
_____________________c_wlnpricey | -.4889269 .017056 -28.67 0.000 -.5223742 -.4554797
___________________c_wlnturnover | -.3369853 .0070716 -47.65 0.000 -.3508528 -.3231179
_________________c_lnstdevdailyret | .441063 .0479018 9.21 0.000 .3471266 .5349993
___________________________cons | -1.022986 .1685667 -6.07 0.000 -1.353549 -.6924236

Kind regards,

Guest

[Clyde Schechter]

The way to do this is to set up a three-level variable, call it assurance, that is 0 if not audited, 1 if audited by other than big4, and 2 if audited by big4. Then run your regression with ib1.assurance (no interaction term). Then the coefficient of 2.assurance will be the incremental effect of a big4 audit over other auditing, and the coefficient of 0.assurance will be the effect of being unaudited compared to a non-big4 audit.

[Guest]

Hi Clyde,

Thank you! It is a very simple solution, but I was thinking in whole other directions as you might have noticed.
I have tried it and this was exactly what I needed.

The thread can be closed, if necessary.

Kind regards,

Guest