# When should I use a dummy variable?

I have data for neighbourhoods with median income. I also have a standard low-income cutoff. (Just using regression analysis) I feel it would lead to a simpler and cleaner result if I just use the cutoff to code the income variable into a dummy (1 for low-income neighbourhood).

Is this okay to do? Or is it better to just use the average where it's available. I'm planning to do this for other quantities too. And I don't want to run into trouble with hypothesis testing.

• If you use dummy, you’re saying it’s perfectly fine for you if the model makes completely different predictions for cutoff + 1\$income vs cutoff - 1\$.
– Tim
Jun 18 at 17:48
• As @Tim alludes to, binning destroys information and, usually, is poor methodology.
– Dave
Jun 18 at 17:53
• Alright! I understand. I hadn't thought of it this way before
– Dee
Jun 19 at 17:21