Advertisement
If you have a new account but are having problems posting or verifying your account, please email us on hello@boards.ie for help. Thanks :)
Hello all! Please ensure that you are posting a new thread or question in the appropriate forum. The Feedback forum is overwhelmed with questions that are having to be moved elsewhere. If you need help to verify your account contact hello@boards.ie

Omitted Variable Bias

  • 01-08-2010 2:27pm
    #1
    patriks
    Closed Accounts Posts: 62 ✭✭


    Hey guys,

    Basically, let's say one wanted to show the effect of technological innovation/diffusion on economic growth. Now, let's say one were to build a type of regression model, outlining economic growth as the dependent variable and spending on R&D, number of patents granted, high-tech export levels, human resources in technological industry, etc. as the independent variables. Would this be a valid regression?
    Obviously, to attribute all economic growth to technological innovation/diffusion wouldn't be accurate, as there are clearly other factors that affect economic growth. Given the above, the noise--or error term--would be quite large, resulting in omitted variable bias, right? Would this result in an invalid regression?

    Thanks!:)


Comments

  • #2
    Anonymous1987
    Closed Accounts Posts: 784 ✭✭✭Anonymous1987


    This is probably a little bit basic but have you tested for omitted variable bias. Just enter "ovtest" after you run the OLS regression in Stata and it will run a Ramsey RESET test for omitted variables. Probably more to it than that since like you said other things drive economic growth, maybe search the literature on repec to see how its normally managed.


  • #3
    Time Magazine
    Registered Users, Registered Users 2 Posts: 8,452 ✭✭✭Time Magazine


    Anonymous1987 wrote: »
    This is probably a little bit basic but have you tested for omitted variable bias. Just enter "ovtest" after you run the OLS regression in Stata and it will run a Ramsey RESET test for omitted variables.

    Just to clarify: this is a very weak OVB test. A RESET test analyzes nonlinear alternatives of your regression but cannot test to see if something else altogether is absent, obviously enough.

    OP the regression you outline is fraught with issues of endogeneity (e.g. does growth drive R&D or does R&D drive growth?) and multicollinearity (e.g. number of patents granted, export levels). It's a little harsh to say such a regression is "invalid" but you're unlikely to be able to stand over your coefficients as unbiased and you certainly can't claim to have estimated a causal relationship.


Advertisement