Constant negative fitted I missing something?

Hello all,

I'm attempting to fit a model where y variable is a variety of $/ha/year figures (all positive) ranging between 0.01 and 2520. The output of this should always have positive fitted values (none of these values should ever be below 0), and yet I get constant negative fitted values. So far I have tried:
- multiplying y by 100 then log
-just log/log10 of y
- A Tobit model, but I couldn't figure out if I was doing it right, should I set lower=0 to mean nothing lower than 0?
-GLM Gaussian

The x variables are geographical variables such as Area (which is logged to improve dispersion), GDP and Population density, and methodological ones such as valuation type.

Any further suggestions on how to eliminate negative fitted values would be greatly appreciated, or on tests to confirm how to proceed. There is no values in x or y that are negative. Looking at the residuals, this definitely still looks linear (no cone shape etc)

Thank you for reading.
Hi, Thank you for replying. This is a multiple regression, but on the whole there is a positive relationship between X and Y. When I graphed my data pre-regression, there was limited positive relationships showing on the graph. Most of the variables have a long tail on either side of the distribution.