The answer here is: it depends on the kind of decision we’re making.
FPTP works well for binary decisions: Yea vs Nay, or one vs another. Do we adopt the motion, or not? Do we adjourn, or not? Do we choose this one candidate, vs that one candidate? In cases like these, FPTP is hard to beat.
When it comes down to making a single choice out of more than two options, however, FPTP is highly problematic, for, as noted, in such cases we tend to get the decision of the largest minority, not a decision of the majority.
There are other, better, systems for such cases: the preferential ballot, for instance, and in such cases, as proposed here, we’ll get an optimum result with Condorcet/Ranked-Pairs (or any Condorcet method), or at least a moderate improvement with something like IRV / AV.
Choosing at which restaurant to hold our annual dinner, for example, is in this category, as is choosing a Mayor, perhaps, or a single member to represent one’s district in the legislature or parliament.
Choosing multiple options out of a suite of them introduces additional possibilities. Here, we get multiple-representation options, such as: Multi-Member Plurality (MMP), the Single Transferable Vote (STV), or even Ranked Pairs, as well as proportional-representation systems, such as: Open or Closed List PR, Mixed-Member Proportional (MMPR) (particularly when upgraded to Condorcet-MMPR), and such.
What food shall we order-in for the meeting? – Chinese, vs Thai, vs Italian, vs Indian, vs… whatever – would be in this category. Ranked Pairs would tell us which of these the majority of the given group prefers, and even give us an ordered list of such preferences, but it wouldn’t assist us in determining appropriate portions.
But if we can order-in from more than one place, we might indeed want a proportional decision, so 45% of the order might be Indian, and maybe 30%, Italian, and maybe the rest Chinese, according to the varied tastes of the group. Why not?
Perhaps we want to pick our representatives in the legislature this way, too, so that the various ideological or philosophical perspectives of the group can be better represented as well. Not a problem, providing that’s what we want to do.
Each voting system has its own pluses and minuses, for various purposes. They’re different tools for different jobs, and should be employed as appropriate for the job at hand.
I submit that for electing single-representatives to legislatures, when we typically have more than two candidates (and, indeed, I offer the value judgement that we should, in fact, encourage multiple candidacies), Condorcet/Ranked-Pairs, or at least some other Condorcet method, is the best way to go.
The decision about whether or not we want multiple- or proportional-representation should not be made because FPTP is so bad at electing single-representatives, or for making “all votes count” (a contentious argument, in my view), or railing on about the disconnect between the FPTP so-called popular-vote vs the final aggregate seat count (another contentious argument), but should, instead, be made because we explicitly want more varied voices and viewpoints within our legislatures, or not – while at the same time recognizing the benefits and pitfalls, including minority governments, and possible log-jams and accountability issues that can also ensue.
Let’s have that discussion, by all means, but meanwhile let’s not let it pin us down to an egregiously flawed FPTP status quo.
Let’s get the upgrade from FPTP underway, now, and leverage that, perhaps, into a later adoption of a Condorcet-MMPR, if it turns out that proportional representation is really what we want.