A five-minute political economy

In which we map shared beliefs from the unsurveyable issue space into the one-plus-dimensional policy space.

A recurring theme in the newsletters so far was that most of the observations I wrote about were not restricted to what we might call “economics proper”.

Social organization follows a few universal patterns, no matter which level or purpose. Profit-seeking companies, political parties, educational institutes, sports clubs, religious congregations, even rock bands share a bunch of commonalities when it comes to bundling forces to pursue their goals, which is why it makes sense to work out universal patterns first and restrict ourselves to “economics proper” afterwards.

But there are also some fundamental differences, which is why we see different aims and resolution mechanisms between these types of organizations.

The “textbook” way in economics is to establish foundational rules about exchange between economic organizations: markets, firms, households, the economic branch of government, and then to impose these rules to realms outside the purview of economics proper, giving rise to the popular accusation of “economic imperialism”.

One of the fundamental things about patterns I like to stress is that they all have a scope: a realm where they quickly and intuitively apply, surrounded by a grey area where the fit becomes uncomfortable, and finally the hinterland where they obviously don't help us at all.

This scope should always be front and center in any discussion about a pattern, and a key ordering principle I'm trying to pursue is that I go from universal patterns to domain specific patterns.

In this newsletter I will hop into one of those domains: the political realm, or even more specifically the political realm of modern nation states where collective decisions come about as the outcome of a “political” deliberation process: the fine balance between top-down and bottom-up that involves elections, voters, parliaments, governments, and so on.

For now this is all we need to hold fixed, because as it turns out, few things I discuss will be restricted to that realm. The fundamental question is still how we take the preferences, beliefs, dispositions, moral frameworks of a diverse group of people and aggregate them in a way that we achieve something we can fuzzily enough describe as “common progress”.

Or, as I described it in the newsletter on organization: how do we resolve the conflict between moving forward and staying together if everyone has a different idea what “forward” is.

The last newsletter contained a high-minded discussion about organizations as belief structures, and I promised to make up for it by offering a step-by-step, pattern-based exposé about how individual world views align in a “five-minute political economy”.

From issue space to policy space

The starting point is something everyone who has ever been badgered by a survey research institute should know: the issue.

An “issue” is anything we might have an opinion on that can be compressed into a single dimension: do you prefer A or B, or maybe something in-between? Or, to make it both exhaustive and mutually exclusive: do you prefer A or its opposite, “Not A”?

To take it one step further, take A to mean “let's do something”, and Not A “let's not do it”, aka let's stick with the status quo. Which brings us to the second fundamental conflict in organization: the conflict between moving forward and staying put.

All the questions that can be expressed in this “let's do it” vs “let's not do it” dichotomy form our issue space, the dimension of which is (I used this word before) unsurveyable.

The reason for this is not only that there are lots and lots of issues that shape the public discourse in the political realm, but also that any single issue can be expressed in many different ways, leading to wildly varying answers to what should be the same question.

This is a hotly contested topic in survey design — a discussion I'll sidestep here. The important takeaway for us is that we start from a bunch of expressions of agreement or disagreement on a bunch of topics, all of which are one-dimensional and all of which are anchored in the status quo as the default, “let's not do anything” option.

Depending on how the issue questions were framed, we might have a simple set of binary yes vs no answers, a discrete five step (strong yes, weak yes, neutral, weak no, strong no) scale, a 100%-in-favor to 100%-against sliding scale, or even a mix of answers. At this point, this is a minor concern.

What's more of a concern is how these issues relate to each other, or, to express it with a pattern rooted in a quote by Philip Converse, of ideology as the science of “what goes with what”.

What “unsurveyable” means is that if we treat every issue as a separate dimension, our policy space will become so high-dimensional that we can extract very little information from it, much less any kind of prediction.

But what Converse posited is that there is a lot of predictability between issues, that we can often take someone's known positions on a bunch of issues and use them to predict their position on a different, seemingly completely unrelated issue with a high level of accuracy. The more we can predict across issues, the more “ideological” this person behaves.

In other words, we can take the “high-dimensional” issue space and fold it into a much lower-dimensional policy space without losing too much information or predictability.

If this sounds strange, we already do it intuitively every time we speak of “left vs. right” (either as a binary variable or a continuum) in politics. Indeed we're so used to mapping everything in politics onto this unidimensional line that we forget how atypical this is — we pretty much encounter it only in politics and nowhere else.

The labels “left” and “right” derive from the seating arrangement in the National Assembly after the French Revolution of 1789, but terms we usually consider synonymous are “progressive” and “conservative”, connecting left and right to our own dichotomy of “let’s move forward” vs “let's stay put”.

Indeed we can take this one step further, split our population in five subgroups, and connect them to the famous technology adoption curve, except the “technology” under discussion could be any political issue. It simply represents individual risk preference (a form of belief) for this particular issue, from adventurous to cautious.

We're so used to matching the labels left as progressive and right as conservative that we rarely ever notice that they're fundamentally wrong.

No matter which group we're looking at, we'll find that those on the left are progressive on some issues and conservative on others, similar to those on the right, except there's a high similarity about issues within and a high divergence between the subgroups — conservatives tend to be progressive about things progressives are conservative about, and vice versa — which is what makes them belief clusters.

Converse's recognition was that this is a common characteristic in politics, and it's the first step to realize that this is a universal pattern which explains all ordering principles based upon it, from political parties to legislative procedures to the fundamental functioning of a representative democracy.

From policy space to politics

Which leads us to the question “how many dimensions does the policy space really have?” Beyond the much invoked left-right dimension, some models also like to add a second “for or against authority” dimension or even go multi-dimensional (economic, social, civic, technological).

The answer lies in finding out how aligned, aka how correlated these purported axes really are: the “what goes with what” translates as “If a large majority of those who choose A (over Not A) also choose B (over Not B), the dimensions are highly aligned” and we can treat them as (almost) the same.

But can we? For the first two dimensions, that test is straightforward.

If all we want to consider is the one-dimensional (left-right) policy line, the question becomes “Can we place everyone somewhere on the policy line so that we can fit all issues without making too many mistakes?”

In other words, if we deem A a “left” issue and B a “right” issue, someone who favors A should be placed to the left of someone favoring Not A, and conversely, someone who chooses B should be to the right of someone choosing Not B, with few crossovers.

Moving up one dimension, the question becomes if I can draw each single issue as an angled line (aka a radial vector) into the two-dimensional “blackboard” coordinate system, and if lines that capture highly correlated issues are highly aligned (differ only by a few degrees).

The more mistakes we make (even if we choose the best individual placement into this one- or two-dimensional space), the less ideological, and in turn the less predictable, our population becomes. At the extreme end, “nothing goes with nothing”: we can make no predictions at all because no ideologies align and all voters make choices fully disconnected from each other.

This is even intuitively unrealistic, even informally we use the left-right model a lot to explain political phenomena to ourselves. And indeed, Keith Poole and Howard Rosenthal, the inventors of voteview, formalized this mapping of issue space into policy space in a bunch of settings where structured choice data is available, like the U.S. Congress or the U.S. Supreme Court, and concluded that the first dimension explains a lot, but not all of the observed behavior. A second dimension adds a bit, the third and any higher dimension almost nothing at all.

Politics is then typically “one-plus-dimensional”, except when we encounter upheavals, when existing coalitions (belief clusters) break up and form anew. That's when the second dimension becomes more prominent.

Again, this should surprise us more than it does, because outside of politics this is exceedingly rare. If we look at collective preferences for things like music or films, we get high-dimensional preference spaces and very little predictability, which is why most Silicon Valley companies spend a lot of effort trying to improve their recommender systems.

This triggers the question if there's something unique about politics that makes it one-plus-dimensional? I believe I have an answer to this, but I keep it for a future newsletter, because we're already pushing the five minute mark.

From politics to the design of political institutions

As we're crossing the five-minute mark, we can start mapping out departure points from this simple model of politics.

The first departure point is of course the focal point of this newsletter: How do we design politics using this spatial map of political preferences? This could be either as an analyst for a political party trying to maximize its vote share while keeping its core message intact (i.e. move forward while staying together) or in a public-minded role trying to come up with an optimal electoral system that achieves the same thing: enabling long-term progress for the whole commonwealth while keeping inner peace, keeping in mind that in the interplay between competing parties only a fraction of the electorate will be happy at any given time.

Up to this stage, I've held the canonical setup of a representative democracy with a full-time parliament (with parties as belief clusters) as the first line of defense and popular elections as the second line as given. All of this we could of course put to the test. Could we design a more direct democracy where the roles of parliament and parties are renegotiated? Even if we accept the need for a legislature, do we actually need general elections or can we organize an ongoing supervision model where representatives can be voted off any time?

Can we untangle the unseemly interaction of votes and money in politics? After all the parliamentary system stems from an era where representatives took many days riding on horseback from the capital to their constituency, and news travelled just as slowly.

Today we could, if so desired, aggregate the will of the people in real time, and disaggregate it down to each individual issue, so any gap between popular opinion and its reflection in legislative outcomes could — at least in theory — be bridged.

I don't necessarily want to kickstart an institutional revolution here. My hunch is that most of the institutions we built a few centuries ago to connect politicians with the popular sentiment will survive the onslaught of the information age, in part because they were not designed to bridge the gap in information flow, but the inevitably much longer timeframe of the collective deliberation process we might call “forming belief clusters”.

Another reason why the underlying currency of the political economy is “the vote” and not the dollar, the euro, or the yen is because it contains an integrity element that — even if it perennially falls short of its envisioned ideal — can't be replaced if we allow “make” (to have your voice heard, you need to get off the couch and line up at the poll station) with “buy” (trade your vote for some other, possibly more tangible benefit).

This seems obvious enough that we usually don't spend much time imagining a pure exchange political economy where policies are set according to the wishes of the highest bidder and sinecures are auctioned off by the ruling party.

Such a thing in its pure form would go against every conception of a functioning political system we have. But of course money matters in politics, and safeguarding the integrity of the democratic against well-financed interference is a critical design issue for which there are seemingly no good but only “less bad” solutions.

This triggers one more question, the last one for this newsletter because I have to cut it off somewhere, the one about scope. As I mentioned above, we will reencounter the social mechanics — the belief clustering — in other realms in a somewhat modified form, first and foremost in the economic realm.

The question is then what makes the economic realm so fundamentally different from the political realm that we find one conflict resolution mechanism — voluntary exchange supported by a monetary system — a perfect match for the former but wholly unacceptable for the latter?

There is a one-minute short answer because a full treatment deserves its own newsletter, namely that the modern nation state, the shell which holds the political realm, is the governance structure that channels all involuntary exchange, meaning all exchange that requires coercion or physical force.

For this it requires two things: a nation (a shared belief structure) and a state (a public bureaucracy that administers all involuntary exchanges within that belief structure). The dividing line between the involuntary and the voluntary realm is always in flux and subject to constant renegotiation, and for this we need the democratic process: a bottom-up control of top-down governance.

More on this and all other questions in future newsletters, because I'm already way past the five-minute mark.