Because not every insurer wants to compete statewide.
Say I'm an insurer in Ohio. I have a network of doctors in the Cincinnati area, including a reasonable coverage of hospitals. Now, I sell a policy in Ashtabula, at the other end of the state. Do I tell that person to drive four hours every time that person needs to see a physician? Or do I end up adding an entirely new set of doctors in Ashtabula just for one customer? It's much easier just not to write policies in Ashtabula.
Prior to the Patient Protection and Affordable Care Act (PPACA, ACA, or Obamacare), insurers could participate in just one portion of a state. That portion might not necessarily be delineated by county boundaries. For example, the insurer might only write policies where the place of employment (employers usually purchase health insurance in the United States) was in certain zip codes. If the employee commutes into an area for employment, presumably the employee can travel the same distance for health care.
PPACA requires some easy way of determining what insurers are available in each area. Apparently they chose to do this by county. For the most part, insurers are available in adjacent counties. However, at some point, we get to the border of where an insurer wants to participate. Making that border the same as a county border is arbitrary, but it probably simplifies things relative to something like zip code. (Is 44444 close to Cincinnati? Not really, although it is closer than Ashtabula.) Or even worse, address. It's easy enough for an insurer to look at a map and pick out reasonable counties to cover.
Without PPACA, insurers weren't necessarily limited to county boundaries, but this also meant that they didn't have to take every applicant from a county either. Under PPACA, if an insurer wants to offer coverage through the exchange, the insurer has to offer it to every applicant in that county. They are preapproved. Without PPACA (or state legislation), the insurer could pick and choose.