Earlier quoted context omitted.
In some cases, abstraction can be done without loss of accuracy. For example, in physics we can assume that charges bounded inside a sphere can be modeled as a point charge at the center of the sphere. It's not just a simpler model; it's mathematically equivalent, and provably so.
There are a lot of simplifications that get used in physics and engineering calculations. Ideal electrical circuits, ideal gasses, thin cylinder walls, lossless collisions, etc. etc. Whether or not a given simplification is appropriate depends on the situation (how much accuracy is needed? how close the simplification is to reality? what's the cost of not simplifying?)
And in some respects, the simplifications are even more justified in economics than physics. Friction between atoms doesn't tend to reduce when people start writing papers finding interesting results contingent on friction being significantly above zero; January increases in stock prices do shrink to the point where they don't represent a viable trading opportunity when traders read papers by Thaler et al on lack of any informational or tax reason for "January effects"