There’s controversy in Germany over building U- and S-Bahn extensions, in which environmentalists argue against them on the grounds that people can just take trams and the environmental benefits of urban rail are not high. For example, the Ariadne project prefers push factors (green regulations and taxes) to pull factors (building better alternatives), BUND opposes U- and S-Bahn expansion, and a report endorsed by Green politicians argued based on shoddy analysis leading to retraction that the embodied carbon emissions of tunneling exceeded any savings, which it estimated at only 714 t-CO2 per underground km built. Against this, it’s important to sanity-check car and public transport ridership to arrive at more solid figures.

To start with, virtually everyone travels by car or by public transport. There’s a notable exception for cycling, but cycling is typically done at short ranges, and the metro expansions under discussion here (all outside the Ring) are beyond that range. In Berlin, the modal split for cycling peaks in the 1-3 km range and is small past 10 km. Beyond the scale of a neighborhood or maybe a college town, cars and mass transit are substitutes for each other.

Nor does public transport expansion lead to hypermobility, in which overall trips grow longer as people commute from farther away and car use doesn’t decline or only weakly declines. If anything, the ratio of substitution for passenger-km rather than trips is that a p-km by metro substitutes for more than one p-km by cars, because metro-oriented cities can be denser and allow for shorter commute trips. Berliners average 3.3 4.6-km trips per day, or 15 km/day; Germany-wide, it’s 35.5 km/day (see table 11 of MiD). If anything, the presence of a large city core also shortens the average car trip by reducing exurb-to-exurb driving at low density.

Nor does polycentricity solve the problem. Indeed, ridership in polycentric regions is weaker than in monocentric ones. MiD has data by state and Verkehrsverbund in Germany, with modal splits by trips (all trips, not just work trips) and passenger-km, the latter measure having far less in the way of cycling and walking. From this, we have the following table:

Geography	Transit % (trips)	Car % (trips)	Transit % (p-km)	Car % (p-km)
Berlin	27	24	47	40
Brandenburg	9	51	22	71
Mecklenburg-Vorpommern	7	52	14	80
Saxony-Anhalt	9	48	12	81
Saxony	11	56	16	77
Thuringia	9	55	16	79
Hamburg	22	29	43	47
Bremen	14	33	30	60
Schleswig-Holstein	8	56	14	79
Lower Saxony	8	53	15	77
Nordrhein-Westfalen	10	55	18	74
Rheinland-Pfalz	9	57	19	75
Saarland	10	65	15	79
Hesse	12	52	19	74
Baden-Württemberg	9	53	17	75
Bavaria	10	56	18	75
Munich (city)	25	29	39	51
Frankfurt (city)	24	29	38	52
Stuttgart (city)	23	36	33	60
Munich (MVV)	19	41	31	61
Hamburg (HVV)	16	33	29	63
Hanover (Region)	15	42	25	66
Rhine-Main (RMV)	13	49	22	71
Rhine-Neckar (VRN)	10	51	17	75
Rhine-Ruhr (VRR)	12	53	20	73
Rhine-Sieg (VRS)	12	49	22	70

Berlin is by all measures the most public transport-oriented and least car-oriented part of Germany. The source doesn’t explicitly break out VBB, but VBB comprises Berlin and Brandenburg, whose population ratio is 59:41, so we get a modal split by trips of 20% transit, 35% car; a similar computation for p-km is less certain since Brandenburgers, many of whom commute to Berlin, have longer trip lengths, but it’s likely Berlin and Brandenburg’s combined modal split is slightly better than those of MVV and HVV, both monocentric. Brandenburg, notably, has the highest modal split by p-km outside the city-states, owing to the Berlin commuters.

In contrast, the polycentric regions – Rhine-Neckar (Mannheim), Rhine-Ruhr (excluding Cologne), Rhine-Sieg (Cologne-Bonn), and to a large extent also Rhine-Main (Frankfurt) – all have weak modal splits. The cities themselves have healthy usage of public transport, judging by the data that’s available and by ridership on their Stadtbahn systems, but most of the Rhine-Ruhr’s population doesn’t live in Düsseldorf, Duisburg, Essen, Dortmund, or Wuppertal, and this population drives.

The upshot is that rail development that strengthens city center at the expense of suburban job clusters should be considered a positive development for transitioning from cars to public transport. Job clusters outside city center do not reduce commuting, but instead make commuting more auto-oriented.

This, in turn, creates serious estimation problems for the diversion rate, which is why environmental benefit-cost analyses underrate the effect of urban rail construction. An expansion of a north-south line like U8 would not just increase the residential connectivity of Märkisches Viertel but also, on the margins, increase the commercial connectivity of Alexanderplatz and other central stations served by the line. This, in turn, should induce additional ridership on lines nowhere near Märkisches Viertel, for example, east-west lines like U5 and U2. At the neighborhood level, the construction of the line would create a lot of induced trips and not have a high diversion rate from cars. But at the city level, little examples of diversion as more work and non-work destinations cluster in Mitte would multiply, never enough for an easy comparison, and yet enough that, as we see, more people would be living and working here without driving, where otherwise they’d be driving between two Kreise elsewhere in Germany.

Taken all together, the diversion rate at the level of trips should be considered 100%: at large enough scale, every trip by public transport is a trip not done by car, perhaps in the city, perhaps elsewhere in the country. Every p-km by public transport is multiple p-km not done by car, since dense cities allow for shorter trips without the traffic congestion problems caused by trying to fit high density and also a high modal split for driving.

With that in mind, a calculation of a first-order diversion rate is in order. A daily trip by rail is a daily trip not done by car. The average trip length in Germany by all modes is 12 km, but this is weighed down by short walking and biking trips; the average daily driving rate per car is 26 km (see table 21 of MiD) when the car is in use, and is 10,000 km/year per car. If we take 10,000 v-km to be the diversion rate per 3 public transport trips, we get that, at the emissions intensity of 2017, a daily public transport trip represents an annual emissions reduction of 0.43 t-CO2. The Märkisches Viertel extension of U8, estimated to get 25,000 trips/day, would reduce Germany-wide emissions by 10,000 t/year, which is nearly an order of magnitude more than the carbon critique of Berlin U-Bahn expansion got. At the current 670€/t cost used in German benefit-cost analyses this is around a 2% rate of return on cost purely from the carbon savings, never mind anything else – and usually green policy uses a low discount rate due to the long-term effects of greenhouse gas emissions, 1.4% in the Stern review.