I think it's a bit more nuanced than this. It's like the example given in the Mythical Man Month. If you want to grow corn, you can pretty much be assured that no matter what you do, you can harvest the corn more quickly if you put more people towards harvesting it. This is because harvesting corn is a partitionable task (in other words, you can easily divide the work up amongst an arbitrary number of workers). Of course, productivity per person will go down, but the overall rate at which you're harvesting corn will still grow.

Programming isn't easily partitionable. Programmers need to communicate with each other to do their task. Thus, adding more people makes you progress faster up to a point. At a certain point, adding more people starts to slow you down. Therefore, individual worker productivity will go down more quickly faster than it would for the corn harvesters.

Lumping all the different positions from all these different kinds of companies ignores these differences and really makes it difficult to apply the idea in general. For technology companies in particular, I'd say that the productivity loss can be even higher than 3/2.