On Mike Berkowitz's 80k Podcast

post by Timothy_Liptrot · 2021-04-21T01:53:33.363Z · EA · GW · 10 comments


  On assessing pundits
  Some contextual disagreements
  A few examples

A few criticisms of Mike Berkowitz's performance on 80k

On assessing pundits

The core problem here is how to choose which pundits to trust. In most sciences, the response is to check if their statements are provably true or false. Causal inference is really hard in politics and many events are one-off, like Trumps second presidential election. Even worse, if we grade pundits on provable falsehoods we select for pundits that make vague, unfalsifiable predictions. This is even worse.

The alternative is to ask two more subjective questions

  1. Are the persons statements consistent with the causal statements that we can prove?

  2. When asked about an area were we know provably true and important information, does the person share that information?

For example, if you ask a pundit to explain why Obama won the 2008 election by such large margins they might list many plausible conjectures. But we can prove that reductions in real income have a large effect on incumbent vote share, and if the pundit doesn't mention this fact we should update down on credibility.

Some contextual disagreements

The general disagreement is between people who study democracy globally, called comparativists, and people who study just the USA, called Americanists. Ziblatt and Levitsky wrote a book "How democracies die" which argues that the United States is backlsiding toward autocracy. The problem is that Z&L describe backsliding as an erosion of the local norms of democracy, so the book is very scary to Americanists. But to comparativists, the erosion of any particular US democracy norm is simply movement within the democratic category. Globally democracies have a very diverse internal institutional rules and norms, and the US has highly unusual norms like inter-party cooperation. Since most democracies have no inter-party cooperation comparativists would not view a decline in IPC as backsliding, but Z&L would.

A few examples

  1. Why were Trump voters able to influence the Republican party so much?

In any democratic debate, your prior should be that the wining faction will adopt the position of the median voter. Basically, you need 50% of the voters, so if 51% of the voters prefer some position k on a one-dimensional axis, candidates will adopt position k to win. Berkowitz never mentions that the median voter in most Republican primaries is currently "pro-Trump" so he leaves out the single sentence explanation.

  1. The description of Trumps attempted coup was vague and did not address the critical details that determine coup success and failure; Organization, expectation and coalitions.

Coups require an organized body of people to deter rivals with violence and operate enough of the state to compensate their violence specialists. A coup must either deter or coopt the existing bodies of violence specialists: the federal army and each states armies. Recall that the national guard of Virginia was massing on the DC border within hours of the capital attack. Trump needed an organization to pay and coordinate the violence specialists so at least a few hundred bureaucrats. Bodies with enough organization to launch a coup include: the federal army, either party or a coalition of state governors, the FBI.

In theory you could convince and organize enough non-specialists to keep occupying the capital and being punished until all the above violence specialists are deterred. Usually this requires enough protestors to fill the jails, and regimes respond by first releasing all criminals and eventually executing overflow protestors. This path requires the greatest organizing capacity because the costs on each protestor must be redistributed. You need a lot of bureaucrats if you want people to accept being beaten and jailed for the cause.

No president will ever achieve that without their own party behind them. Individual politicians are often irrational, especially in their lame duck periods. But parties stay rational. We have no evidence that Mitch McConnell, Kevin McCarthy or Governor Abbott of Texas considered joining Trump. Likewise, generals and populists have sharply divergent interests because generals can safely and happily return to the ballots but populists cannot.

"the other thing those bills are doing is shifting power toward more partisan bodies. There is a bill in Georgia punishing Brad Raffesnberger. It also clearly will mean that next time around Trump or someone else trying to overthrow the election they will get much further than Trump."

The professional politicians of the Republican party were not close to siding with Trump. Will the Republican speaker (elected by the median house Republican) see higher expected value in supporting a coup or rejecting it? The party loses massive membership if they support, and gains defacto political power if they win. But Republicans just want to veto bills, so why transition to a populist regime. It will never be a good choice for the party. The Democrats have had many opportunities to grant themselves more capacity at the price of alienating their bourgeoise by reforming the national labor relations act, but refused.

"People in positions like [Raffensbergers] risked their lives or careers and took on tremendous abuse to ensure politicians they didn't like got elected"

This is not really true. People like Raffensberger risked those things to ensure their copartisans got reelected.

  1. He is agnostic on electoral reform?

Minute 36:00. He's like "yeah I don't understand the problem, but I'll keep talking". But I do know that "all these solutions are not silver bullets to these challenges (populism)".

How can he not understand how electoral reform shapes party behavior, but also know that no electoral reform removes the populist threat? Empirically, there are democratic structures which are robust to populism. Japan and New Zealand have shown that sovereign parliamentary democracies do not manifest even nascent electoral movements.

  1. Why did so many people vote?

Berkowitz argues that Trump drives turnout. But 2016, Trump's first term, had moderate turnout. 2020 was Trump's second election, so why is he driving turnout in turn two.

The alternative explanation is the well established phenomenon of policy feedbacks. The more the government affects people's lives, the more they vote. The pandemic lead the government to actively disemploy millions of Americans while also implicating the government in each Covid-19 death. That effect will get you an 8 percentage point turnout bump; a reality tv star will not. At least Berkowitz should have mentioned Trumps modest 2016 turnout and policy feedbacks.

  1. "Why we didn't see intimidation on election day is a little hard to say"

People do not take on high personal costs for collective gains without a facilitating organization. Intimidating people on election day is a dangerous move for an individual activist. If you want individual activists to put themselves in harms way you need to convince them that others will do the same thing and the costs will be distributed see example. That requires a strong, on-the-ground organization. Trump did not have that. So intimidation did not occur. Literally exactly what you would expect.


Comments sorted by top scores.

comment by IanDavidMoss · 2021-04-21T14:30:45.028Z · EA(p) · GW(p)

The professional politicians of the Republican party were not close to siding with Trump. Will the Republican speaker (elected by the median house Republican) see higher expected value in supporting a coup or rejecting it? The party loses massive membership if they support, and gains defacto political power if they win. But Republicans just want to veto bills, so why transition to a populist regime. It will never be a good choice for the party.

The Republican House Minority Leader, Kevin McCarthy, was on Fox News November 6 saying, "Donald Trump won this election, so everyone who's listening: do not be quiet. Do not be silent about this. We cannot allow this to happen before our very eyes...Join together and let's stop this." He later  signed onto an amicus brief supporting a lawsuit that, if successful, would have overturned the election in four states after the results were already certified. He then voted to reject certification of the election results in Arizona and Pennsylvania after the insurrection, along with most of his caucus.

Replies from: Timothy_Liptrot
comment by Timothy_Liptrot · 2021-04-21T16:53:15.958Z · EA(p) · GW(p)

Interesting, that's a good point.

comment by Robert_Wiblin · 2021-04-21T13:09:06.964Z · EA(p) · GW(p)

Exciting to see a post about this episode 5 hours after we put it out (!).

A few quick thoughts:

"Berkowitz never mentions that the median voter in most Republican primaries is currently "pro-Trump" so he leaves out the single sentence explanation."

No but I say that. IIRC one of his responses also takes this background explanation as a given.

"Japan and New Zealand have shown that sovereign parliamentary democracies do not manifest even nascent electoral movements."

In general I'm with you on thinking some systems of government are less conducive to populist movements, but I'm not sure one can show that by choosing two cases without checking for counterexamples.

"Berkowitz argues that Trump drives turnout. But 2016, Trump's first term, had moderate turnout. 2020 was Trump's second election, so why is he driving turnout in turn two."

I mean there's multiple factors but I'm with Berkowitz here.

In 2020 polls more people had very strong views about Trump (strong approve or strong disapprove) than is typical of a president, while I don't think that was true in 2016. So I don't think there's anything strange about the idea that he raised turnout in 2020 but not 2016.

The other big factor I would think is easier postal voting.

"No it is not. People do not take on high personal costs for collective gains without a facilitating organization."

But that just raises the question, why weren't there better facilitating organizations? Folks expected them to be present but it seems they didn't organize effectively.

comment by Harrison D · 2021-04-21T03:37:52.836Z · EA(p) · GW(p)

I haven't had a chance to listen to the podcast yet, but I'll give a couple first-thought responses.

(0) I don't know why (or even whether) Berkowitz doesn't address the MVT, but my impression is that the assumptions baked into the MVT are out of touch with reality; consider for example that people may rationally vote "irrationally" (see e.g., https://en.wikipedia.org/wiki/The_Myth_of_the_Rational_Voter), since national democracy does not have reliable feedback mechanisms for making good choices with your vote. Of course, that's not to say that MVT is totally wrong; it could be decently "right for the wrong reasons," but I've increasingly heard people argue that the MVT is losing accuracy as politics is increasingly detached from policy effectiveness. Also, just casually glancing over the transcript it seems like Berkowitz indirectly touches on the notion by emphasizing how party primaries encourage selecting more-partisan candidates ("they really talk about the problem of partisan primaries"), although perhaps Berkowitz should have given some more-explicit (i.e., by-name) discussion to this.

(2) This analysis feels really shallow, and I'm not sure it's fair to Berkowitz based on what I'm seeing on the transcript. In fact, looking at a few quotes it even feels a bit misrepresentative of what Berkowitz said. From the transcript: "So I do think political reforms are important here. I’m a little agnostic personally about which ones, we could talk more about this, but I do think some reforms to the system are really key to get at the structures there as well." Further on this, Berkowitz gives counterexamples of where he claims parliamentary democracies have led to "populism", including the UK and Australia; you can't just say "well here are a few examples of where parliamentary democracy has worked very well; it clearly has to do with their structure." (In the case of Japan, I'd immediately suspect there are massive confounding variables, including having one of the most homogeneous populations in the world as well as having been highly economically successful)

(3) I haven't fully read/analyzed this section of Berkowitz' talk, but I think that once again this is some shallow/hasty dismissal: "turnout wasn't high in Trump 1.0, so Trump can't be responsible for higher turnout in Trump 2.0" is not a knock-down argument in itself. It's wholly possible that Trump was so polarizing after 4 additional years of hogging the limelight that he had an upward effect on turnout. However, I do agree it's probably not the most persuasive argument--but I don't think that you ought to be so confident in your own analysis, either. Having not done serious research on the matter, it still seems more likely that the expansion of mail-in/absentee voting also had a sizable upward effect. In the end, all three of our explanations could be right. Perhaps Berkowitz should have mentioned your point, perhaps it wasn't that crucial.

In summary, I'd say it's easy to pick apart any political pundit's analysis; I imagine when I listen to this podcast I'll have a number of criticisms. However, I think it's also important to apply similar scrutiny to our own criticisms.

Replies from: Timothy_Liptrot
comment by Timothy_Liptrot · 2021-04-21T16:46:11.771Z · EA(p) · GW(p)

That’s good pushback, thank you.

I'll need to edit the piece more deeply, but for now I've added an explanation at the top.

Here are some counter criticisms

  1. That is a great summary of the modern critique of the MVT, thanks for sharing. The problem is that it applies to policy outcomes, but the subject is candidate selection. Caplan’s criticism is that voters are irrational at connecting votes to policies and policies to outcomes, which is true. If voters have a simpler utility function like “I want to vote against any anti-Trump candidates”, the median voter will apply very well. If 51% of voters choose to vote against any anti-trump candidates and you are anti-trump, the MVT will still cook your goose. Compare with democratic primaries, where the Bernie faction mobilized 40%, but Bernie does not have 4/5s of Trumps influence at all.

The response concerns me because Berkowitz passes up a great opportunity to explain a fundamental dynamic of the system in question. Imagine asking an economist “why did the price of gasoline increase when supply contracted” and the economist didn’t mention supply and demand curves. Sure it could be a Giffen good with a weird speculative market, but your explanation should start with the theory that explains maximum variation, then move to edge cases.

  1. Saying there is no silver bullet is a strong positive claim that Berkowitz doesn’t back up. I’m not all claiming there is definitive evidence that a silver bullet exists. I am simply claiming that we have not seen evidence disproving silver bullets, which Berkowitz claims. I should have made that more clear.

Viz-a-viz populism, if we define populism as “did any populist party form” then yes, there are few patterns. Martin Gurri is probably right that populism emerges from the information environment. But if we define it as populists that cam close to ruling, by getting 20% of the vote share, then we have some patterns. The German Bundestag was stunningly successful at deescalating every issue that the populists could use after 2016, for example (they paid billions to keep refugees out of Europe). Keeping populists out of office can probably be done with electoral engineering, but keeping populist minorities from forming can’t, so depends on goal post definitions.

  1. Fair enough, I shouldn’t argue that policy feedback is the definitive reason and Trump not. I'm just concerned that Berkowitz doesn't mention other explanations. Generally I am suspicious that activists will overfit on data to make their issue seem really important.
Replies from: Harrison D
comment by Harrison D · 2021-04-21T20:42:09.647Z · EA(p) · GW(p)

Thanks for the responses. To go through the points you mention:

  1. I’m just not that convinced that the MVT is akin to the gas price situation you described, in that I don’t see it as that explanatory/fundamental/crucial to mention (in combination with the following remarks). Importantly, as part of this I’ll say that it seems like you’re watering down the MVT to increasingly become “politicians try to appeal to the majority of people,” which is arguably far more intuitive and thus less necessary to cover in name/detail. As I understood it, the MVT is more meant to explain why politicians converge to more-moderate policy preferences in order to win over the “median” voter. So if you’re just going to say (e.g.) that “candidates were less inclined to be anti-Trump because a majority of people wanted to vote against the anti-trump candidates,” you don’t need to mention the phrase “Median voter theorem” any more than you actually have to mention “supply and demand ** curves/graphs ** “ (as opposed to just “supply and demand”).
  2. I’m still not convinced there is a silver bullet, and my base rate for “situations where there is actually a silver bullet despite the dismissal by people who are more experienced than me, yet it’s just not getting used” is really low. Thus, I’m inclined to side with Berkowitz on this—including his observation that there are definitely multiple helpful reforms that could be taken.
  3. If the situation is as you described it, I definitely think that’s a fair concern—and it’s something that I’ve seen too many pundits do.
Replies from: Timothy_Liptrot
comment by Timothy_Liptrot · 2021-04-22T01:07:13.310Z · EA(p) · GW(p)
  1. The median voter theorem is just a mathematical conclusion from a set of simple assumptions, it can't be watered up or down. If some set of voters rank proposals on a single axis with single-peaked preferences then the only Condorcet winner will be the median voters ideal point. The key here is that the voters choose the axis on which to rank the proposals. The voters can rank the candidates on any axis, taxes or policies or height. Usually smarter actors rank proposals and candidates on an ideology axis by collapsing issues onto one axis, as a mental heuristic. But if they interpret the axis as "does he run ads where he shoots guns" instead of "does he vote for my tax rate", the gun ads determine the axis not the tax rate. In this case the assumption is that Republican primary voters care a lot about the Trump-support axis.

The two level game papers are cool, but not relevant here.

The MVT is important here because Trumps influence over represenatives is non-linear with his vote share. If Trump loses 20% of his primary influence and loses the median primary voter, he does not lose 20% of his influence, he loses most of his influence.

  1. There are better ways to select leaders, but our current leaders choose not to influence them. Our current leaders like the current leader-selection system because they win at it. (See paper)[https://onlinelibrary.wiley.com/doi/pdf/10.1111/j.1467-9248.2005.00514.x?casa_token=R0YSBl-4UKQAAAAA:pwa_9ENFHN8VU9PT50ymM6R1fy57NtdDSW1R7K5aeT8tGXMRvLzwHzkAoFh7di80Lp4mv8b4e8ZnFWg]
Replies from: Harrison D
comment by Harrison D · 2021-04-22T01:53:24.935Z · EA(p) · GW(p)

1, The “watering down” comment was really referring to the idea of expanding the “preference axis” assumption to include more than just policy, to the extent that MVT changes from “Politicians moderate their policies towards the center of policy axes” (which would be a perhaps unintuitive claim that doesn’t need explicit reference to “MVT”) to “Politicians appeal to the majority of the voting public” (which is almost “no-duh” except that it irons over potential wrinkles, like “someone who is very far right/left won’t even bother voting unless one of the candidates moves far enough towards them, rather than spending their time to go out and vote for the ‘less bad candidate’ in an election which they might deep down recognize they won’t actually have any impact on...”). Ultimately, I think the question of whether Berkowitz should have discussed MVT by name is less important that the question of MVT’s validity, but I’m not in an epistemic position to get deeper into the weeds on that. 😶 2. I still don’t see that as a true “silver bullet”; I imagine Berkowitz might consider it one of the potential positive reforms, though.

Replies from: Timothy_Liptrot
comment by Timothy_Liptrot · 2021-04-22T18:51:00.045Z · EA(p) · GW(p)

Original median voter theorem paper, Duncan Black in 1948

Let us suppose that a decision is to be determined by vote of a committee. The members of the committee may meet in a single room, or they may be scattered over an area of the country as are the electors in a parliamentary constituency. Proposals are advanced, we assume, in the form of motions on a particular topic or in favor of one of a number of candidates. We do not inquire into the genesis of the motions but simply assume that given motions have been put forward. In the case of the selection of candidates, we assume that determinate candidates have offered themselves for election and that one is to be chosen by means of voting. For convenience we shall speak as if one of a number of alternative mo- tions, and not candidates, was being selected.

Let there be n members in the committee, where n is odd. We suppose that an ordering of the points on the horizontal axis representing motions exists, rendering the preference curves of all members single-peaked. The points on the horizontal axis corresponding to the members' optimums are named O, 02, 03, . . . , in the order of their occurrence. The middle or median optimum will be the (n + I)/2th, and, in Figure 3, only this median optimum, the one im- mediately above it and the one immediately below it are shown

Anyway, this is really a pedagogic question. How best should we teach politics? Some people advocate that we should disregard the MVT because it is both "obvious" and "false". Setting that contradiction aside, I think the underlying assumption that only theories with perfect data fit should be taught is wrong. By the same logic, physics should not teach Newtownian mechanics because it is wrong relative to quantum mechanics. You can't just give the reader quantum mechanics, you need to start with a theory they can understand then update it.

comment by Harrison D · 2021-04-21T02:48:01.187Z · EA(p) · GW(p)

Quick, minor note: I'd recommend linking to the podcast you're referring to for ease of access.