How to improve the Elizabeth Line

Learning from how Japan runs the most reliable trains in the world

The Elizabeth Line has not yet been open three years, but it is already hard to imagine London without it. It has relieved the Central Line and opened up new fast, direct journeys across London. At both its eastern and western end, the stations it took over have seen their passenger numbers skyrocket.

All of this has brought substantial challenges. The Elizabeth Line is already overcrowded: it operates 16 trains per hour (tph) in the off-peak, and 24 tph in the peak. This gives it a maximum capacity of only 36,000 passengers per hour per direction (pphpd), which is less than the Victoria Line’s 40,608 pphpd, even though Victoria Line trains are smaller. Anecdotally, the Elizabeth Line is close to being full in the peak.

The crowding is worsened by the fact the Elizabeth Line awkwardly splits in the east, so that half the trains go to Canary Wharf and half to Stratford. Even though both of these are very important destinations, each only gets a half service.

There is, however, another problem with the Elizabeth Line – reliability. The most recent figures for October–December 2024 show that only 78% of station stops were on time. This is better than the national average of 62%, but it is concerningly low given that the Elizabeth Line is so new. Of Britain’s modern electric rolling stock, the Elizabeth Line trains are some of the most prone to breaking down. There have been problems with the fact the trains have to work with three different signalling systems, and particular issues with the overhead wires on the western branch to Reading.

All of these technical problems are tractable, and they will get fixed. But there remains a fundamental problem with the Elizabeth Line: it is trying to operate a service pattern much more complex than most metro lines, and is trying to do so at high train frequencies. This will inevitably constrain the line’s ability to run a reliable service.

This is a particular issue when we think about extending the Elizabeth Line. At the moment, half of the westbound trains terminate at Paddington: it would be useful instead to send them out to Hertfordshire, terminating at Tring to support housing development. An extension from Abbey Wood down to Dartford and Gravesend in the Thames Estuary would also enable development.

If the Elizabeth Line cannot run trains reliably today, then adding more services will only serve to compound the problem. We need to find a way to make it more reliable.

The Elizabeth Line’s trilemma

All railways face a trilemma between three desirable things: reliability, frequency, and complexity. ‘Complexity’ means express services, branches, and sharing tracks with other services. Railways can, essentially, pick two of these things.

• The Victoria Line runs 36 tph very reliably, because it is not a complex line to operate. Every Victoria Line train stops at every station, and there are no branches.

• The Northern Line’s branching is quite complex for a metro line, but it is still pretty reliable, at the expense of frequency: it is restricted to 24 tph.

• Thameslink is much less reliable than it ought to be, partly because it tries to squeeze a very complicated service pattern into a central core that operates 20 tph, which is very high for a mainline railway.

The reason for this is straightforward. A more complex line has more points of failure, so it is easier for delays to cascade from one service to another. This is less of a problem if the line is not busy, as there is recovery time built into the timetable. But operating a complex service pattern at high frequencies will lead to delays.

The eastern side of the Elizabeth Line is not especially complex: it has two branches, and every train stops at every station. The western half, however, is much more complex. In the peak, the 24 tph break down as follows:

• 14 tph terminating at Paddington.

• 4 tph to Heathrow Terminal 4, stopping at all stations.

• 2 tph to Heathrow Terminal 5, skipping some stations.

• 4 tph to Reading, skipping a different group of stations from the Terminal 5 trains.

Furthermore, west of Slough it has to share tracks with trains to Didcot, and on much of the line there are freight trains as well, which can lead to cascading delays.

The central core of the Elizabeth Line is highlighted in dark purple, while the pre-existing lines it took over are in a lighter shade.

Reading is 63 km by rail from Farringdon, so it needs fast services to ensure competitive journey times. On any railway, there are two ways to run fast services: using a separate pair of tracks for the expresses, and skip-stop service patterns. Although there are express tracks between London and Reading, they are already taken by the services from London to Bristol and Cardiff. Skip-stop operation is therefore the only way to ensure competitive journey times on the western branch.

Skip-stop operation would probably also be necessary if the Elizabeth Line is extended to the north west or south east. I think the most sensible north western extension would be to Tring in Hertfordshire, which is 54 km by rail from Farringdon. The south eastern extension to Gravesend would be 37 km from Farringdon, and if it carried on down to Rochester it would be 50 km. Finally, skip-stop operation would also be useful on the Shenfield branch: it currently takes 45 minutes to travel 35 km from Shenfield to Farringdon.

Skip-stop services, however, have a big disadvantage: they make it easy for delays to cascade. Imagine a stopping train with a skip-stop service behind it. The stopping train is running late. Because the skip-stop train cannot overtake it, the skip-stop train has to go as slow as the stopping train. As a result, everything behind the skip-stop train gets delayed as well.

In shorty, skip-stopping is very useful for passengers, but it adds complexity and therefore makes the line more prone to delays, especially if the line is as intensively used as the Elizabeth Line.

To find a way around this problem, we need to learn from Japan.

How do the Japanese do it?

It is common among westerners to assume that Japan’s railways are punctual because of cultural values. The Japanese, the story goes, prize obedience and hierarchy, which means they dutifully run their trains punctually. There might be something in this, but I think it has at best a marginal impact. Japanese trains run reliably because they have better infrastructure and better operating practices.

Japan has a very wide variety of railway operators. There are six companies which are part of the JR Group, the privatised successors to Japan’s equivalent to British Railways. It also has the confusingly-named ‘private railways’, which have always been private, all of which focus on commuter services.¹ There is real competition between these companies and the JR Group: for instance, somebody travelling between Osaka and the neighbouring port city of Kobe has a choice between three different operators, all of whose lines are less than a kilometre from one another.

Amongst this diversity, there are several things that the lines have in common. One of those is an absence of four-tracking. Only six commuter lines in Japan have four tracks: the JR West line that goes Kyoto – Osaka – Kobe, and five lines radiating from Tokyo.

The rest of Japan’s lines have to make do with just two tracks, but owing to the size of Japanese cities, a mixture of stopping and semifast services is required so that everywhere gets quick journey times to the city centre. This is a very similar problem to one faced by the Elizabeth Line.

Japan manages to run these complex services reliably with passing loops. Rather than four-tracking an entire line, particular stations get four tracks, which merge back down to two shortly after the station. This allows the fast trains to overtake the slow ones while they wait in the platform at that station.

Express tracks do not have to go in the middle of the slow lines, although it is better operationally if they do.

A typical example is the Hanshin Main Line,² one of the lines between Osaka and Kobe. It runs for 32 km, with a station every kilometre.

Map adapted from Wikimedia Commons.

West of Amagasaki, a major shoulder station, eight stations have passing loops, with one roughly spaced every third station. (East of Amagasaki, the line splits to serve two different business districts in Osaka.)

The passing loops mean that Hanshin can operate a very complicated, intense service pattern reliably. In the evening peak, there are 21 tph heading west from Amagasaki, with five different stopping patterns. This is much more complicated than the Elizabeth Line, which only runs 8 tph west of Paddington with three stopping patterns.

The current timetable only uses the passing loops at two stations, Nishinomiya and Mikage. But of course the line has eight stations with passing loops rather than just these two. The big advantage of a high density of passing loops is flexibility and redundancy to the timetable: if a fast train is delayed, then a local train can wait around in a different passing loop. In turn, this means that trains can very quickly recover from delays.

(The downside is that the slow trains do have to wait around for the fast trains to pass. This is unavoidable: the only way to eliminate this trade-off entirely would be to quadruple the whole line, which would be very expensive.)

Passing loops are how Japan can operate services that are reliable, frequent and complex. To give an idea of the complexity of Japanese train services:

• At the Kobe end of the Hanshin Main Line, many services run onto a line owned by another railway company, the Kobe Kosoku Line.

• In turn many services run from that onto the Sanyo Electric Railway, which goes another 55 km west to the city of Himeji (home of the castle).

• Both the Kobe Kosoku Line and the Sanyo Electric Railway are also used by a fourth company, Hankyu, one of Hanshin’s competitors for the Kobe–Osaka traffic.

• At the Osaka end is complex too, the line connects with a tunnel underneath central Osaka, which goes through the city and out the eastern side to serve the 27 km-long line to Nara (home of the deer). This line is operated by a fifth company, Kintetsu.

In other words, the Hanshin Main Line is one component of a network of lines operated by five private companies that stretches over 100 km from one side of Japan’s second largest metropolitan region to the other. Passing loops are how this complicated network can be operated reliably and frequently.

Learning from Japan

I think Britain should learn from how Hanshin works: nearly every Japanese railway company uses passing loops, because they are the right way to operate a reliable, frequent service with skip-stop operation.

It is worth noting that no railways in Great Britain are run like Hanshin. As far as I am aware, the operating practice is unknown. We do sometimes use passing loops for freight trains,³ and there are one or two stations that are set up nicely for late-running expresses to overtake local trains,⁴ but this is certainly not a regularly timetabled practice.

The reason for this is probably historical. In the past, British railways have responded to the problem of running a mixture of fast and slow services by four-tracking a line for its entire length. As a result, while the entirety of Japan has six four-track lines, London alone has anywhere between ten and thirteen.⁵ We have never needed to use passing loops en masse because we had their big brother, express tracks.

The Elizabeth Line, however, is a Hanshin-shaped problem. It should rebuilt some of its stations so they can act as passing loops. This means the physical area of the station is expanded to have four Elizabeth Line tracks and two island platforms,⁶ in addition to two tracks on the adjoining fast line without platforms.

In some cases, there is room to do this: at Gidea Park, the fast lines could be moved a little bit to the south, to run where the station building currently is. I think this would require only demolishing one other building.

Image from Google Maps.

(To be clear, I am not advocating for putting a passing loop at Gidea Park specifically: it is merely illustrative of a station where it would be less complex to add one than in other places.)

In other places, it would be much more complicated: expanding Ilford, for instance, would require demolishing the shopping centre that takes up much of the town centre. This would probably need to be combined with redevelopment of the town centre.

To be clear, passing loops are expensive, disruptive and time-consuming to retrofit. But they are necessary to run an intensive skip-stop service reliably. The alternative would be either running the Elizabeth Line at lower frequencies, or withdrawing the skip-stop service altogether, both of which are non-starters.

Passing loops should be combined with stations every 1.0–1.5 km within the London urban area, which is roughly the optimal distance for suburban stations. At the moment, there are several long gaps between adjacent Elizabeth Line stations. There is, for instance, a gap of 3.8 km between Romford and Chadwell Heath, and 3.7 km between West Drayton and Hayes & Harlington.

In some cases, adding new stations would make it straightforward to put in a passing loop: it is easier to build a new station with four platforms from the word ‘go’ than to try to retrofit an existing one. For example, a good site for a station would be Justums Lane in Romford, and close to some greenfield and brownfield land suitable for housing development. The site is currently occupied by some sidings, which have space to build a station with a passing loop.

Justums Lane itself is the north-south road that crosses the railway line on the left of the image. Justums Lane station would be built to the east on the sidings. Image from Google Maps.

Choosing where to put in passing loops is essentially a pragmatic decision on the basis of where it is easiest to add them, and they do not need to be built all at once. Every new passing loop is useful, because it gives more options as to where slow trains can be overtaken. A programme of upgrading the Elizabeth Line through passing loops can and should be incremental.

A theme of this Substack is that there are a lot of underrated ideas to improve Britain’s railway services that come from abroad. The railway industry does not appear to be particularly curious about the rest of the world, which is a shame. We should be learning from Japanese passing loops, Swiss timetabling and German tram-metro systems. We need to stop pretenting we are still at the frontier of railways, as we have been at various points in our history: we are in many respects a developing country, and need to be a sponge for ideas from the rest of the world. Only then will we be able to get the railway services we deserve.

1. In Kanto the main ones are Tobu, Seibu, Keisei, Keio, Odakyu, Tokyu, Keikyu, and Sotetsu (which is focused on Yokohama). Nagoya is served by Meitetsu and Kintetsu. Kintetsu also operates in Kansai. The other four operators in Kansai are Nankai, Keihan, Hankyu and Hanshin. Finally, Nishitetsu operates around Fukuoka. There are also many other smaller private railways.

2. Hanshin 阪神 means ‘Osaka Kobe’. ‘Han’ is another reading of the character 阪 which can also be read as ‘saka’. ‘Shin’ is another reading of 神 which can be read as ‘ko’.

3. These are slightly different, as they are loops on the main line in between stations. In effect, the freight train is shunted into a siding so a fast passenger service can pass.

4. For instance, Retford on the East Coast Main Line. It is slightly frustrating that Newark Northgate and Grantham, the other two stations between Doncaster and Peterborough, are not also set up like this.

5. Undeniably, the SEML, BML, SWML, GWML, WCML, MML, ECML and GEML have four tracks. So do the Chiltern Main Line/Metropolitan Line, and the LTSR/District Line. The more arguable cases are the Chatham Main Line / Catford Loop and Waterloo–Reading Line / Hounslow Loop, and the District/Piccadilly Lines between Barons Court and Northfields.

6. In theory, it can be done with a side platform and an island platform. But island platforms have the very big advantage that if a late train needs to be routed onto another track, passengers just need to cross over to the other side of the platform rather than go across an overbridge or underpass to get to another platform. This in turn means trains can be rerouted at very short notice.