Not starting at zero is a common practice in science and data processing. The difference between bad and good data visualization is in relevance. Good data visualization starts an axis at non-zero numbers because the fluctuation is more relevant than the zero. Bad data visualization hides relevant data to present an alternate takeaway.

Here, a change in birth rate of even 0.1 or 0.2 is a major societal change, and showing that change is more relevant than showing the zero (how would it even be possible that there were zero births in a year, anyways?)