I am really not an expert of this opening, but maybe 3.c3 is better in a sense.
After 1.d4 d5 2.Bf4 c5 3.e3, according to you, black should "refute" this inaccuracy by taking immediately on d4 3..cxd4 4.exd4 and now we have the Caro-Kann Exchange Variaton, with white having played the suboptimal Bf4 (1. e4 c6 2.d4 d5 3.exd5 cxd5 4.Bf4). After 4...Nc6 5.c3 Bf5(!) black has solved the problem of how to develop its white squared bishop and should have no further problems. Of course it is nothing for black in the winning sense.
[FEN " "]
1.e4 c6 2.d4 d5 3.exd5 cxd5 4.Bf4 Nc6 5.c3 Bf5
In the line 1.d4 d5 2.Bf4 c5 3.c3 black can also take on d4: 3...cxd4 4.cxd4, but then we have reached the Slav Exchange Variation (1.d4 d5 2.c4 c6 3.cxd5 cxd5 4.Bf4) and here 4.Bf4 is found among the top moves. Furthermore the Slav Exchange Variation poses black more problems to consider, than the Caro-Kann Exchange Variaton.
[FEN " "]
1.d4 d5 2.c4 c6 3.cxd5 cxd5 4.Bf4
Considering values of pawns alone has led us to a good idea in this case.
Of course it should go without saying that in general there are more factors to consider in a chess postion, Space, Material and Activity among the top candidates. Only a good engine and a lot of experience can help us out here.
One example to illustrate the idea:
1.e4 c5 2.Nf3 e6 3.d4! cxd4
White voluntarily allows black to exchange his c-pawn against a valuable center pawn. This is a positional problem, but feasable because white gets something in return, more space and more active pieces.
White can simply take back (4.Nxd4) or play the sly 4.c3, the Morra Gambit. Now it is white posing the positional threat of promoting his c-pawn to a center pawn with 5.cxd4. Black can accept the challenge, after 4..dxc3 5.Nxc3 white has traded material for space and time, it is playable AFAIK. Black can also defend with 4..d3.
[FEN " "]
1.e4 c5 2.Nf3 e6 3.d4! cxd4 4.Nxd4 ( 4.c3 4...dxc3 ( 4...d3 ) 5.Nxc3)