Wednesday 27 January 2016

Agglomeration

Via @thomasforth, from Centre for Cities:

When it comes to productivity, size isn’t everything – at least not in the UK. Unlike the US and Germany, there is no clear relationship between city size (as measured by the number of residents) and productivity (as measured by the wage premium of each city once workforce characteristics are taken into account).

The three charts in the article plot relative productivity against major cities ranked by population in the UK, USA and Germany. The difference is huge in the USA (the trend line is steep), quite big in Germany and smaller in the UK (the trend line is nearly flat), if you ignore London which is well above the line.

Well duh. We could have guessed that...

1. The population of a city is just one indicator or factor. Just as important are links to other cities or indeed the rest of the world. The more other places you can get to, and the quicker/cheaper you can get there, then better. So the smaller the country and the quicker/cheaper the journeys, the lower the differential between smaller and larger cities. Or consider two similar sized cities, but one has an airport with flights to other business centres - the one with the airport will be more productive.

2. How do you define a city? Are Birmingham and Coventry separate cities for these purposes? No of course not. Bracknell and Wokingham, smallish towns in themselves, are well above the line (i.e. very productive). But they are just part of the M4 corridor, which is everything from Reading to Slough to Heathrow. Once you are on the M4, all these towns and the airport are within easy reach, in economic terms, they are one city.

3. Similarly, London and the M4 corridor and large chunks of the south east reinforce each other; good transport links, close together and three major airports etc. London is shown as having a population of 12.5 million on that chart, which is probably accurate if you include the whole hinterland and commuter belt.

4. This is why Berlin is so far below the line. It's the largest city in Germany with 3.4 million people, but it's stuck in the middle of nowhere (surrounded by ex-DDR). A couple of my cousins live and work there and they confirm that it is no great shakes, there are just lots of small businesses with negligible agglomeration effects.

5. The charts do not show absolute productivity, they show relative productivity between small and large cities. So the charts do not necessarily indicate that large UK cities fail to tap in to agglomeration benefits; they could just as easily be taken to indicate that small cities in the USA (or Germany) do particularly badly (especially if they are miles from anywhere).

6. If you look closely, they have messed with the horizontal axis. The horizontal axis on the UK chart is from 50,000 to 12.8 million. The one for Germany is from 50,000 to 3.2 million. This means that the trend line for the UK is actually even flatter than it appears, and if you removed London, probably would be completely horizontal.

And so on and so forth.

8 comments:

Dinero said...

Also the horizontal axis are logarithmic, so you can't use the angle of the slopes as a direct visualisation of the degree to which how the effect being studied varies between the cases.
The log scale makes the angle steeper.

Dinero said...


Sinkhole news.

http://www.bbc.co.uk/news/uk-england-kent-35353331

Pablo said...

Why innovation thrives in cities
http://news.mit.edu/2013/why-innovation-thrives-in-cities-0604

"Double a city’s population and its economic productivity goes up 130 percent. MIT researchers think they know why."

Pablo said...

and, A Physicist Solves the City
http://tinyurl.com/hbgrhuv

Mark Wadsworth said...

Din, the scaling effect is also logarithmic, look up Geoffrey West on YouTube who observes the same basic ratio whether it is animals, forests or cities. Thanks for sinkhole.

Stands to reason - go from 100,000 to 200,000, makes a measurable difference. Go from 10 million to 10.1 million, the effect is barely measurable.

Din, thanks, so double in size = plus 15%, which is what we thought. And a Geoffrey West link, thanks!

Dinero said...

There are two things with the way data is represented

1. The x axis for London is compressed by 80% compared to Germany which make the London graph steeper

2. The logarithmic scale makes the angle more steep for low ratio than it does for high ratios, and so the graphs can not be compared.

Mark Wadsworth said...

Din, yes, but that was my point 6 already.

Mark Wadsworth said...

Pab, thanks for links (I messed up my comment above).