For CBR encodings, the bitrate is always kept at the bitrate specified regardless of if it is needed. For VBR encoding, the bitrate is an average target, however the media stream will use more or less data rate when it needs to but will try to average to the target.
This is why you see 1 pass and 2 pass VBR. 2 pass VBR first makes a pass to estimate where more or less data will be needed and then makes a second pass to make best use of the available data rate. 1 pass does a lot more guessing.
For an example of how this works, lets say we had a video that had 10 seconds of rapid action followed by 10 seconds of a still image.
With CBR, the same data rate would be used the entire time, the 10 seconds of still image would be crazy high quality since nothing is changing, but the action would then fall apart because it is changing too much.
With 1 pass VBR, the algorithm would adjust to use more data for the high level of action, but it might limit itself since the entire video might be like that. Then, when it got to the still shot, it would use almost nothing, but wouldn't hit the average rate because it didn't need the storage space.
With 2 pass VBR, the first pass would discover that there is no movement at the end, so then during the second pass, a lot of data rate would be used to handle the motion really well and then very little would be used for the still image, the average data rate would be hit and the high movement portion of the video would have the highest quality of any of the encodings while the still portion wouldn't really suffer.