Betta coefficient changing

[Pedro Martinez]

Dear all as you can see below the beta coefficients if leasehold change when adding a variable. How can I solve this? I have tried everything and changed my model 3 times + the VIF is of only 1.39 and r is of -.534


reg log_price leasehold

Source | SS df MS Number of obs = 113,665
-------------+---------------------------------- F(1, 113663) = 5583.49
Model | 1826.53664 1 1826.53664 Prob > F = 0.0000
Residual | 37182.7847 113,663 .327131826 R-squared = 0.0468
-------------+---------------------------------- Adj R-squared = 0.0468
Total | 39009.3214 113,664 .343198562 Root MSE = .57195

------------------------------------------------------------------------------
log_price | Coefficient Std. err. t P>|t| [95% conf. interval]
-------------+----------------------------------------------------------------
leasehold | -.2541154 .0034008 -74.72 0.000 -.2607808 -.2474499
_cons | 12.71848 .0024849 5118.35 0.000 12.71361 12.72335
------------------------------------------------------------------------------

. reg log_price leasehold log_size

Source | SS df MS Number of obs = 113,665
-------------+---------------------------------- F(2, 113662) = 30505.22
Model | 13625.365 2 6812.68248 Prob > F = 0.0000
Residual | 25383.9564 113,662 .223328434 R-squared = 0.3493
-------------+---------------------------------- Adj R-squared = 0.3493
Total | 39009.3214 113,664 .343198562 Root MSE = .47258

------------------------------------------------------------------------------
log_price | Coefficient Std. err. t P>|t| [95% conf. interval]
-------------+----------------------------------------------------------------
leasehold | .1969873 .0034274 57.47 0.000 .1902696 .203705
log_size | .885197 .0038512 229.85 0.000 .8776488 .8927453
_cons | 8.644154 .0178444 484.42 0.000 8.609179 8.679129
------------------------------------------------------------------------------

[Fei Wang]

Why would you "solve" this? The first coefficient is an overall effect of leasehold on log_price, and the second is a partial effect of leasehold on log_price, log_size being constant. It's very likely that they are different.