Singapore, as a causal model

The problem

Singapore publishes an unusual amount of high-quality open data, and almost all of it arrives as isolated series. The resale price index lives on one page, the fertility rate on another, median household income on a third. Each is accurate, each is well presented, and none of them answers the question people actually have, which is what is pushing on what.

So the public conversation runs on asserted causation with nothing to inspect. Housing costs are blamed for the birth rate, immigration for housing, inflation for wages, in whatever direction the argument needs. A reader who wanted to check the shape of those claims would have to pull eight datasets, align them on a common year axis, and hold the relationships in their head. Nobody does this, so the claims go unexamined and the data goes unread.

Why now

Three things changed. Open data portals matured from file dumps into documented, stable APIs, so the underlying series are now reliable enough to build on rather than scrape. Browsers became capable of rendering and animating a few hundred data points as inline vector graphics with no library and no server, which puts a genuinely interactive model inside a single file.

And the public appetite arrived. Interest in national indicators is no longer confined to analysts — housing costs, fertility, and inflation are ordinary conversation, conducted with more conviction than evidence. The gap is not data availability and it is not technical capability. It is that nobody has connected the series into something a non-economist can interrogate in thirty seconds.

What it is

An interactive causal model of eight national indicators, borrowing its framing from the systems view in *Democracy 3*: each metric is a node, each asserted influence is a weighted edge, and selecting a node traces everything that pushes on it and everything it pushes in turn.

A year scrubber replays 2014 to 2024, so the system visibly moves — the 2020 collapse in transit ridership and GDP, the 2022 inflation spike, the long slide in fertility against the post-2020 climb in resale prices.

The honesty is load-bearing and stated in the interface. Link direction and weight are editorial judgements, labelled as such, not estimated coefficients. The product is a structured way to argue about the relationships, and being explicit about that is what separates it from a dashboard implying authority it has not earned.

Who it's for

The reader who follows the numbers without working in them: someone who noticed resale prices in the news, has a view about why, and has no way to check whether the view holds together. Journalists on deadline who need to see which indicators moved together before writing the sentence that connects them. Students meeting systems thinking for the first time, for whom a graph that responds is worth more than a chapter that asserts.

Explicitly not economists or civil servants with access to real models. They have better tools and will correctly object to editorial weights. The audience is the much larger group currently served by a headline figure and a chart with no context at all, for whom the alternative is not a better model but no model.

How it makes money

It does not directly, and that is the right call at this stage. The honest version is a free public artefact whose return is reputational: a piece of civic work that circulates, gets cited, and establishes that whoever built it can turn open data into something people actually use.

The routes that follow from that are commissioned rather than transactional. Newsrooms pay for bespoke versions around an election or a budget. Universities and think tanks license the engine for their own indicator sets. Government communications teams need exactly this and cannot build it in-house at speed.

What is deliberately rejected: advertising, which would undermine the civic framing, and a subscription, which fails because each visitor needs this once or twice a year rather than weekly.

Go to market

Launch as a single link, because the artefact is the argument. The first audience is the local subreddit and the policy-adjacent corner of LinkedIn, where discussion of housing and fertility already runs continuously and the usual contribution is an assertion. A tool that lets someone check an assertion gets forwarded in those threads without being promoted.

Second wave is the newsroom. Send it directly to data desks with an offer to build the same shape around whatever series they are covering that month, which converts reach into the commissioned work that actually pays.

The editorial-weights disclaimer is the growth mechanism rather than a liability. Publishing the weights invites disagreement about specific edges, and disagreement about a specific edge is the most useful thing that can happen to it: every correction improves the model and extends its reach.