Except that's not always been the case, either. SSNs have always been a terrible way to disambiguate people. There are weird, crazy edge cases in SSN history. Cases exactly such as JS Jr getting the same SSN as JS III in a podunk town because the local SSA administrator was feeling lazy that day and JS III was already deceased. The federal SSA website claims that that never happened, but if you have a big enough database (say, Equifax) you can spot all kinds of simple dumb human errors like that. For many, many years the SSA left local administrators in charge: the first 5 digits and the weird way they are hyphenated were local district numbers. For people born before 2011 (!), when the new randomization scheme was switched to, there is a 90% chance you can guess their first five numbers if you know their birth date and birth city (which is why it is so ridiculous that PII rules to keep SSNs safe ever considered it fine to show only the last four, those are only meaningful digits for still the majority of SSNs in the wild).
SSNs were never designed for what the credit bureaus and banks and insurance companies (and everybody else) use them for, and there are too many cracks and failure cases. Companies need to admit their failures and come up with a real solution; but companies have so much sunk cost in SSN-keyed databases they aren't likely to ever actually do that. (Maybe this Equifax breach pushes more companies to try. Cynicism says companies remain cheap and invested in their sunk costs.)