The problem is not the elevation; it's the angle you are looking at it.
Looking from top down, you'll see the real proportions, while looking at 30° you'll see half the size (directed towards the line of sight). At 3° it's only a bit more than 5%.
Technically it's the sinus function.
The size perpendicular to the line of sight is influenced by the distance only.
Here is a sketch illustrating the principle in 1D using a line instead of a 2D rectangle (note: I just used the center of the line; actually beginning and end have different angles and distances):

I used three sample points at the same height, but with varying distance, and a sample point at same distance (well: almost), but different altitude.
You'll see that the line will be shorter when reducing the altitude much more than increasing the distance would.
(I know: The sample should have been below the "Distance 92" point for better comparison)
Also not I used rather arbitrary scaling for reasonable numbers to compare: Viewing down at a distance of 100 would yield 100% virtual size.
Alternative Illustration
The following illustration has a black element in the lower right that has equal horizontal (H) and vertical (V) size.
Then there are three reference points:
- The blue starting point looking down at 45°
- The closer (but same height) red point looking mostly down
- The lower (but same horizontal distance) green point looking mostly forward
Colored lines illustrating the center (corner) of the black object, as well as the borders (front and top).

The rest is a bit complicated, so take the time to understand:
The thicker more intensely colored lines attached to at least one corner of the black object are perpendicular on the corresponding view direction, thus a projection of the black object's sizes.
The colored letters H and V mark the corresponding proportions of the projected horizontal and vertical size, respectively.
So for the blue view H and V are equal size, while for the closer red view the vertical element seems much smaller than the horizontal one.
Finally for the lower green view the vertical component of the black element seems much larger than the (projected) horizontal one.
"3D" Illustration
Finally a "3D" illustration of the effect of lowering the altitude shortening the horizontal size (as explained above).
The object looked at is a square with a "center cross", and the relative sizes should be correct (while the absolute size was "zoomed in a bit" to let you see some details):

I hope I got the mathematics right.
The numbers right to each image are:
lateral offset (X),elevation (Y),distance (Z) (negative)@viewing angle (A)
The first three numbers are integers anyway, and the angle is rounded to whole degrees.
The distance is constant the elevation increases (top to bottom, left to right).