Announcement

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

  • I wonder why the OLS results and sys gmm results can be so different.

    Code:
    xtabond2 changeofEDTA_win l.EDTA_equity_asset_ratio_win weightdamageratio30d lnTA_win NLDTA_loan_asset_ratio_win  l.netincomeEratio_win customerdepositTAratio_win  realgdpgrowth lnrealGDPpercapita  i.newspecialization i.newcountry i.year i.newBvdAS                                   , gmm(changeofEDTA_win,lag(2 6)) iv(l.EDTA_equity_asset_ratio_win weightdamageratio30d lnTA_win NLDTA_loan_asset_ratio_win   l.netincomeEratio_win customerdepositTAratio_win  realgdpgrowth lnrealGDPpercapita  i.newspecialization i.newcountry i.year i.newBvdAS ,eq(both)) small twostep robust
    est store gmmEDTA
    
    reg changeofEDTA_win l.EDTA_equity_asset_ratio_win weightdamageratio30d lnTA_win NLDTA_loan_asset_ratio_win  l.netincomeEratio_win customerdepositTAratio_win  realgdpgrowth lnrealGDPpercapita  i.newspecialization i.newcountry i.year i.newBvdAS                                   , robust
    est store olsEDTA

  • #2
    (1) (2)
    olsEDTA gmmEDTA
    L.EDTA_equity_asset_ratio_win -0.228*** -0.613***
    (0.000) (0.000)
    L.T1R_tier_1_ratio_win
    weightdamageratio30d -20.099*** -1.3e+03***
    (0.009) (0.001)
    lnTA_win -0.003*** -0.009***
    (0.000) (0.000)
    NLDTA_loan_asset_ratio_win -0.020*** -0.014
    (0.000) (0.235)
    L.netincomeEratio_win 0.034*** 0.046***
    (0.000) (0.000)
    customerdepositTAratio_win -0.071*** -0.197***
    (0.000) (0.000)
    realgdpgrowth -0.004* 0.062***
    (0.058) (0.006)
    lnrealGDPpercapita 0.002** -0.046
    (0.040) (0.107)
    Specilization Yes Yes
    Country Yes Yes
    Year Yes Yes
    Accounting standard Yes Yes
    N 149894 149894
    adj. R-sq 0.219
    hansenp 0.306
    ar1p 0.000
    ar2p 0.341

    Comment


    • #3
      The coefficient of weightdamageratio30d in ols is about 65 times smaller than that in gmm sys dynamic.

      Comment


      • #4
        Thank you so much!

        Comment


        • #5
          Still need help here. Anyone have experience of this?

          Comment


          • #6
            Yunfei:
            with no data example/excerpt, the usual comment is: why do you expect similar results from such different regression models?
            Kind regards,
            Carlo
            (Stata 18.0 SE)

            Comment


            • #7
              Thank you so much, Carlo!

              Because my dataset is too big.... and I tried, but I feel only a small sample data won't be very helpful for my question here....I wonder whether I could sent to you my whole data to you, which is 11 MB....

              Although I know I am not supposed to have private contact with you......

              Many thanks!

              Best wishes,
              Yunfei

              Comment


              • #8
                Yunfei:
                I would not have the time to go in delve into your dataset.
                Sorry for this.
                Kind regards,
                Carlo
                (Stata 18.0 SE)

                Comment


                • #9
                  Thank you all the same, Carlo.

                  Best wishes,
                  Yunfei

                  Comment

                  Working...
                  X