The extremism emerges from a citizenry whose primary notion of civic duty is voting, and trying to get out of jury duty. The fastest way to reform a political party is swamp it with majority newcomers who vote in rough lockstep for about 10-20 years. This is surprisingly easy to do in the US, where for example, if you want to start moving the Republicans away from a war posture, show up with enough of your friends around all the precincts in all the states that key Republican party members are in (like parliamentarians and committee chairs), and you squeeze them out before they can even get into the regular annual/bi-annual/quadrennial machinery, and vote out the remainder who get into leadership from convention votes. Neither the Republican nor Democratic parties get into the cray cray like loyalty tests at most levels of party machinery, so showing up is all it takes. Loyalty oath demands are common, but as politicians lie all the time, I don't feel sorry for those who administer those oaths when the oaths don't count in the voting turnout. Actions, not words count.
Currently, attendance at most precincts/wards/etc. in both Democratic and Republican parties is fractions of a percent of eligible voters. This yields the emergent behavior you see today: the ones controlling the levers of political parties are the "true believers" who simply showed up, and as an aggregate tend towards more extreme positions. It is Yeats' "the centre cannot hold" and "The best lack all conviction, while the worst / Are full of passionate intensity" come to life.
Not everyone who shows up even needs to spend all that much time participating at the precinct level. Just enough to send onwards to the next level of the party machinery those you trust to vote similarly-enough to your own interests. It works out to about 2-3 hours per year at most: you mount an effective DDOS upon the party establishment, then go home if you want. Say you spend about another couple day's worth of due diligence each year to suss out who you vote to send onwards to the next level of participation.
Participating at the higher levels of both parties' machinery takes more time and adds money to the requirements, but not all that much in the grand scheme. To represent a county-level/senatorial-district-level/etc. at a state convention, generally about the highest most people will get within a handful of years of starting participation, is once every two years spend 4 combined days in a couple major cities in your state, split into a one-day and a 3-day meeting spread across several months. If you pack your own meals or similar (grocery store run and cook your own), drink only water, drive to the cities, and stay with a bunch of like-minded citizens who split the tab on a house-share, usually about a day's worth of accumulated logistics organizing, about $120-200 total expenditures. If you eat out, order drinks out, stay at a hotel and do some tourist activities, about $600+. The state conventions do fall on a weekday, so you'll give up a vacation day once every other year. Say you spend another couple hours a week discussing and reading about issues of the day, so call it about 17 8-hour days every year of participation.
However, not many people really know how this all works, and give up within 2-3 years of participating. They think that showing up, organizing and kibbutzing with like-minded citizens for a year or two should yield perceivable results. It realistically takes closer to 10-20 years of sustained effort, starting only from the year where the numbers are on your side. The Ron Paul and Bernie Sanders supporters learned this lesson to their disappointment. Most who attend for the first time do not cotton on that they need to show up, find enough like-minded to effectively caucus, study how the current leadership maintains and loses their power in their internecine knife fights, devise their own strategy and tactics, then make their bid to seize power of the leadership posts, a process that takes many years, failures, and setbacks that teach along the way.
Politicians are money-driven because most voters' civic duty starts and stops at voting (when in an active civic civilization, casting a vote is actually the endpoint of an involved process), and most voters are highly susceptible to media-based messaging in such a low-information decision-making environment. Politicians pay more attention to giant majority voting blocs over money; no amount of money can win against overwhelming votes of aware citizens.
We simply don't have enough aware citizens.