I'm running a bilateral trade panel data of 42 reporter with 186 partner for the year 2015 and 2017. I do have a lot of missing datas in my dataset. I am trying to assess the impact of trade facilitation measures on trade flows. lgeneraltf_ij is the geometric average of trade facilitation score of country i and j (UN global survey on trade facilitation)
I grouped the country pair so I can use the command xtreg on Stata 15.1. The Hausman test indicates that I should use a fixed effect model.
However, when I run the FE model, lots of my control variables are not significant and the R2 within is very low. I am wondering if it is because my time dimension is too restrictive (only 2 years ) and not having much variation across years. Could anyone tell me why the R2 is that low ?
In this case, should I just use the reg command and adding the fixed effect manually with dummy variables or run the between estimator with the command xtreg, y x1 x2 xn, be without taking into account the time dimension ?
Code:
estout REmodel FEmodel, cells(b(star)) stats(r2 N) -------------------------------------------- REmodel FEmodel b b -------------------------------------------- lgdp_ij .1399031*** .0211473 lpop_ij .0411576*** .1443081 ldist -.965924*** 0 lstariff .0713956 .1930599 contig 1.556638*** 0 comlang_off .5072642** 0 comcol .2396629 0 landlocke~ij -2.526593*** 0 llsci_ij .9406411*** .3432247 lgeneralt~ij 3.731172*** 1.029607 _cons 9.73321*** 5.629184*** -------------------------------------------- r2 within .0125 .0175668 r2 between .6705 .3961 r2 total .6720 .4009 N 4088 4088 --------------------------------------------
Best,
Anthonin
0 Response to Gravity model, low R2-within
Post a Comment