This year saw a major, unexpected political upheaval in both the United States and Europe. Most commentators have reacted to these events by arguing that they just reinforce the validity of the pundits’ own previous views. Whether the commentator in question is on the left or the right, they tend to claim that Trump’s strengthens the case for their own longstanding commitments.

The fact that pundits double down on their existing views does not necessarily mean they are wrong. But if you consistently claim that even highly unexpected events prove that you were right all along, that suggests your judgment might be influenced by “confirmation bias,” the cognitive error of interpreting new evidence to conform to your preexisting views, regardless of whether it really supports them or not.

I am far from being a complete exception to these tendencies. For example, I have long argued that political ignorance is a serious problem. And I believe that ignorance was a major factor in this year’s election, and that it contributed to Trump’s victory.

Still there are some issues that this year’s events led me to change my mind about. Here are three of the most important.

I. The Perils of Polarization.

I have long argued that partisan bias is a serious problem. The body politic suffers when we reflexively support our own party and ignore its faults, while reflexively demonizing the opposition and dismissing ideas associated with it. But I also thought that political polarizationthe growing ideological gap between the two parties – is not a significant problem. So what if the two parties are becoming more extreme? There is no good reason to think that extreme positions are necessarily worse than moderate ones.

I still see little inherent virtue in moderation. But polarization is much more of a problem than I thought. The more polarized we are, the greater the partisan bias, and the greater the tendency to reject anything associated with the opposition. Polarization also makes voters and political activists more willing to tolerate bad behavior by their own party and its leaders. The greater the degree of polarization, the higher the stakes of political conflict. When your opponent wins in a highly polarized environment, policy will turn against your values in a big way, not just a small one. That makes it all the more imperative to avoid doing anything that might give ammunition to the opposing party.

For that reason, among others, partisans become even more willing than usual to turn a blind eye to the flaws of their own leaders, and tolerate behavior they would never accept if the other side did it. Also, it’s easy to assume that our party leaders can’t be all that bad so long as they are fighting against the really evil people in the other party.

This sort of dynamic is one key reason why the vast majority of Republican voters ultimately “came home” to Trump, despite the fact that many had severe reservations about him. But while the Trump phenomenon helped catalyze my rethinking on on this issue, the risks of polarization go well beyond it.

II. Should We Bring Back the Smoke-Filled Rooms?

Until this year, I tended to dismiss Jonathan Rauch’s argument that the political “establishment’s” loss of control over candidate nomination processes is a bad thing. As I saw it, presidential candidates selected under the more populist process adopted in 1972 do not seem to be, on average, worse than those previously chosen by party elites meeting in proverbial “smoke-filled rooms.” In addition, the influential “party decides” school of political science argued that party elites still retained effective control over the process. As a libertarian, I am no great fan of conventional major-party elites. Why give them more power?

Nonetheless, Rauch was closer to the truth than I was. Even some advocates of the “party decides” theory now recognize that the Trump experience shows that party elites’ power over the process is a lot more limited than was previously thought.

It may still be true that modern popularly elected presidential candidates are,generally speaking, no worse than those of the pre-1972 era. But even if the average quality is similar, the Trump phenomenon suggests that the variance is higher. Party elites support plenty of flawed candidates. But it is unlikely they would choose someone as authoritarian and poorly qualified as Trump, or for that matter a Castro-loving socialist like Bernie Sanders, who attracted strong support in the Democratic primary.

We need to give party elites a greater say in the process; not because they will select truly great candidates, but to reduce the danger of ending up with horrendously awful ones. Despite their many flaws, party elites have strong incentives to choose candidates who can appeal to a broad swathe of general election voters, and who are not likely to be dangerously incompetent or unhinged.

Admittedly, a more elite-driven process could screen out some really good candidates who might otherwise win the support of primary voters. But given widespread political ignorance and ideological bias in the electorate (the latter especially common among primary voters), I think it is more likely that anti-establishment populist candidates will be unusually bad than unusually good.

Canada has avoided the wave of dangerous xenophobic populism that has swept many other countries in part because their candidate-selection process is more elite-driven than in the US and some European countries. We would do well to learn from our neighbors on this score.

Perhaps we don’t need to go so far as to fully return to the “smoke filled room” era. I am not sure what the precise optimal mix between elite and voter control over nomination processes should be. But I do now think it should be more elite-driven than the status quo.

III. Rethinking the Unitary Executive.

Like many constitutional originalists, I have long supported the theory of the “unitary executive”: the idea that the president should have more or less complete control over the executive branch. Because Article I of the Constitution gives all “executive” power to the president, he should be free to hire, fire, and issue orders to executive branch officials as he sees fit, so long as he does not order them to do anything specifically banned by law or outside the scope of executive authority altogether.

I still think that the unitary executive is a sound interpretation of the text and original meaning of the Constitution. But, under modern conditions, a fully unitary executive would give a single person sweeping control over a vast federal bureaucracy that regulates and monitors nearly every aspect of our lives. Many of the functions of that bureaucracy are themselves unconstitutional under the original meaning. Thus, a unitary executive today means an executive who exercises a wide range of powers that the Founding Fathers never gave the federal government at all.

So long as the executive branch has such enormously broad power, it is dangerous to concentrate it in the hands of any one person. If it cannot be pared back, it should at least be dispersed. The prospect of Trump wielding such power has helped lead me to rethink this issue. But, in truth, no one person should be trusted with such enormous authority. Both Barack Obama and George Bush also pushed the limits of executive power in dangerous ways. Had Hillary Clinton won the election, it would not have been desirable to let her control a fully unitary executive either.

The theory of the unitary executive focuses on the distribution of executive power, not its scope. A unitary executive can, in theory, be a strictly limited one if the overall scope of executive authority is kept within tight bounds.

But unless and until we can substantially reduce the overall extent of executive power, Congress should have the option of at least partially insulating many executive branch agencies from complete presidential control. Before we can restore the unitariness of the executive, we should first cut back some of its substantive powers.

Obviously, this leaves open many hard questions about exactly which agencies Congress should be allowed to cordon off from presidential control and to what extent. But it does indicate that we should not try to institute a fully unitary executive – at least not for a long time to come.