The usual reason is that negative values would have no meaning in the context for which the linear model was designed. This is particularly true in the field of economics, marketing, trends, etc. The equations used usually have the capability of processing negative amounts and producing negative results, but such values would have no meaning when applied practically.