Announcement

Collapse
No announcement yet.
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • Negative R squared in IV regression

    Good afternoon everyone,

    I am running an instrumental variable regression for panel data with the command xtivreg2 and everything looks fine: the Kleibergen-Paap and Hansen statistics indicate that the model is not underidentified and that the instruments are valid. My only concern is that the R squared is negative in some specifications (not in all of them).

    I have read that a negative R squared is not necessarily problematic in IV, but it might be a signal that the model is misspecified. My question is, can I just ignore the negative R squared since my Kleibergen-Paap and Hansen tests are OK?

    Thanks for your help

  • #2
    Whereas OLS minimizes squared error, 2SLS doesn't with respect to the original variables. A negative r-square indicates the model fits very poorly. Whether that is a problem for you is up to you.

    Comment

    Working...
    X