Announcement

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

  • Year effect in a Fixed Effects model

    Hello,

    I've got some questions on including year effects in my regression model.
    My data runs from 1985 to 2012.

    When including year effect the first 2 years (1985 and 1986) are dropped. Can someone explain why these are dropped and what this implies?
    Furthermore, the first few years are highly insignificant, but from 1992 everything is significant. What would be the reason?
    Lastly, the first few years have a positive coefficient, but from 1990 it becomes negative. It seems like almost every year is becoming more negative. Can someone explain to me how I should interpret this.

    P.S. My independent variable is 'Real GDP per Capita'. Dependent variables are related to infrastructure, trade, health, etc.

    Thanks in advance!

    Regards,
    Pvunderink

  • #2
    Hi Pvunderink,
    I think would be better if you could post your results to give you any advice
    Fernando

    Comment


    • #3
      I hope the files are visible.
      Attached Files

      Comment


      • #4
        Maybe the code will help too.
        Code:
        xtreg GDPpercapitaconstant2005US Capitalpercapita  LaborforceWorkingofTotal  Schoolenrollmenttertiaryg lagInternetusersper100people lagMobilecellularsubscriptionsp lagAgriculturalirrigatedlando lagImprovedwatersourceofpopu lagImprovedsanitationfacilities lagEnergyproductionper10000peop lagRoadsper10000people lagRaillinesper10000people lagLinershippingconnectivityinde lagAirper10000people lagHealthIndexFinal  lagGINIindex lagUrbanpopulationoftotal lagNetTradeValue i. Year, fe

        Comment


        • #5
          Hi Pvunderink,
          I think the reason why you are missing the first two years is a: you loose one dummy (from the full set) to avoid multicolinearity with the constant. Second, you loose one because you are using lags, thus already dropping one year of observations.
          Now for why are the signs turning neative and more significant, that is something related to the nature of your data, so there isnt much one can say to explain that.
          Fernando





          Comment


          • #6
            Originally posted by FernandoRios View Post
            Hi Pvunderink,
            I think the reason why you are missing the first two years is a: you loose one dummy (from the full set) to avoid multicolinearity with the constant. Second, you loose one because you are using lags, thus already dropping one year of observations.
            Now for why are the signs turning neative and more significant, that is something related to the nature of your data, so there isnt much one can say to explain that.
            Fernando




            Thanks Fernando!
            Now I know why the two years are not reported.
            So, maybe I should rephrase my second question. What is the interpretation of the coefficient of the year dummies? What do they tell me?

            Comment


            • #7
              Is the following statement true about year effects?
              "The year fixed effects can be interpreted as unmeasured national or international factors that affect GDP per capita in all states to the same degree."

              Maybe this will help? Are these coefficients compared to a base year?

              Comment


              • #8
                Yes, the coefficients on the year dummies are relative to 1986, which is the base year because 1985 is dropped. These capture aggregate conditions in the world economy. Nothing says aggregate conditions have any particular pattern. Rarely does one care about these coefficients: the year effects act as controls.

                Comment


                • #9
                  Originally posted by Jeff Wooldridge View Post
                  Yes, the coefficients on the year dummies are relative to 1986, which is the base year because 1985 is dropped. These capture aggregate conditions in the world economy. Nothing says aggregate conditions have any particular pattern. Rarely does one care about these coefficients: the year effects act as controls.
                  Thanks for the help!

                  I still don't really get why the coefficient is negative. Don't we expect an upward trend. GDP per capita is rising over time, so why would the year dummies become more negative over time?
                  Should I report all these dummy values? Or only the 'corrected' independent variables?

                  Comment


                  • #10
                    Keep in mind that the trend you are estimating using the dummies, do not capture the trend of GDP alone. For instance, if you were to estimate the model only using GDP and year dummies, you will be observing the upward trend. Now, since you are using other explanatory variables, the year dummy are capturing something equivalent to the residual trend.
                    Fernando

                    Comment

                    Working...
                    X