"Simplicity is at the root of policy failures" > I enjoyed this piece by @pollymackenzie in @unherd and I basically agree except importantly on the semantics. Because I passionately believe that over-complexifying is the root of more policy failures. https://unherd.com/2020/05/why-simplicity-is-at-the-root-of-policy-failures/

How can that be? I agree that a one-size-fits-all central policy is a bad idea. There are two ways to fix it,
1/ Don't have a one-size-fits-all central policy and have lots of simpler local policies instead.
2/ Make the central policy complicated enough to handle more variation.

I'll give an example. It's a bit niche, but I know it well so I can be pretty confident in what I say. Bus Timetables!

Great Britain has a single national bus timetable standards. It's called TransXChange. It covers every single possibility for public transport. From a once a week special bus that boards a ferry in the Scottish Highlands to an every 90s tube service in Central London. Everything.

This is the ultimate UK complicated solution. And whenever we need it do something more, we make it more complicated. One size can fit all if you make it complicated enough. The standard is almost impossible to work with, it slows down innovation in public transport enormously.

But there are advantages. Once you've finally got this horrendous complicated mess to work, it can do everything. The result for us as normal people is that we can plan a trip that starts on a 90s frequency tube train, and ends via a bus on a ferry in Scotland. In Google Maps.

So what's the alternative? It's called GTFS. Developed by the local transit authority in Portland Oregon, adopted and thus established as the de facto world standard by Google. It is laughably simple compared to TransXChange.

Many transport experts genuinely laughed at it and did the usual "stupid tech bros, step away from our expert field with your toy, it's much more nuanced and complicated than that" dismissal. GTFS is now completely dominant. No other standard really matters. Why? Simplicity.

But how does this play out into policy? The GTFS standard is the tool is the simple localist. TransXChange is the tool of the complicated nationist. TransXChange lets the UK run all of its transport planning and transport analysis and transport investment nationally.

So we have national bodies like DfT, NIC, and ONS producing connectivity maps to a single standard, evaluating businesses cases within one process, justifying investment (or not) with one model, a model that is the most complicated in the world. One size fits all has to be.

Because it's so complicated, there is almost no competition. No government smaller than the UK can spend enough to enter the discussion. Few companies can invest enough to enter the market. We have an expensive system because it is so complicated. And it delivers poor results.

The alternative is GTFS, published (or not) by cities or by regions and solving 99% of problems for 10% of the effort. Around that simple standard has grown a huge ecosystem of software for planning and analysis. The downside is that there is no guaranteed national view any more.

There is a trade-off. In almost every field, if you take the advanced economies and line them up from "decentralised and simple" to "centralised and complicated" the UK is the most centralised. I think it leads to bad outcomes. Some people think the opposite. That's the choice.

ps. this is why almost no other country has a national free bus pass. They looked at it, said "wow, that introduces huge complexity into our system, I'd rather keep it simple", and said no. The UK chose a complicated one-size-fits-all policy. You may think it's worth it. I don't.

pps. Transport for London has consistently said "no thanks, that's far too complicated, we're doing our own much simpler thing" and that's part of why it's so good.

Oh and here's the open source converter from TransXChange to GTFS that I publish. It's a lovely demonstration of the relative simplicity of GTFS. The Class that defines a TransXChange file is 2514 lines of code long. The same for GTFS? 121. https://github.com/danbillingsley/TransXChange2GTFS/blob/master/TransXChange2GTFS_2/Program.cs