By Stephen H. Graham

Over the last two decades, public-transportation expansion and ridership in one of the world’s largest cities, São Paulo, has surged, taking several points of mode share from automobiles. Despite that, and despite unusually tight restrictions on certain cars and trucks during business hours, congestion, as conventionally understood, continues nearly unabated. The pattern suggests that while mass-transit availability on its own solves many or most people’s transportation problems, it does not necessarily lower the congestion rate of private vehicles. That may be possible only by tolling congestion. 

São Paulo (SP) is the single largest municipality in the Americas, with 12 million inhabitants, anchoring a metro area of 22 million, the tenth largest globally.1  Figure 1 gives a sense of its scale and economy, traditionally industrial but now mostly in services.  It is representative of a new class of developing-country mega-cities whose populations grew fast in the 20th century but have stabilized in the 21st and which are building transit infrastructure to catch up after the fact.  It constitutes a test case of mobility modes and traffic congestion before and after transit networks become available.  

Figure 1: São Paulo. Source: 

In the early 1960s SP already had a reputation for bad traffic, and by the 1990s rush-hour jams were reported in hundreds of kilometers. Paulistas across the metro area travel from place to place 42 million times a day on average for work, school, errands, and socializing.2  The city connects to the rest of São Paulo state over the most extensive network of highways in Latin America, shown in Figure 2, and upwards of a million cars pour into town each day. Municipal traffic engineers estimate that 3.8 million motor vehicles3 overall circulate actively within city limits, while the state puts the number at 4.4 million.4

Vehicle congestion at first was tackled by adding and widening roads. As in other cases around the world, however, that encouraged even more use of cars, possibly worsening the problem. Real-time traffic information became so critical that one major radio station broadcast only traffic reports, keeping a helicopter in the air much of the day. When the real-time traffic app Waze (now part of Google) emerged as a better tool, SP became a major early market for it.5 

The first subway line opened in 1974, but the buildout was slow, and traditional suburban commuter trains meanwhile deteriorated, logging less than 770 thousand daily rides as recently as 1998.6 In 1996, authorities began barring circulation by one of every five cars each weekday at rush hour, based on license-plate numbers, in a large area known as the expanded center. This had surprisingly little impact on congestion, however, for reasons discussed below.  

Only in the 21st century,  with Brazil’s economy more organized, São Paulo no longer swelling with migrants, and new approaches such as public/private partnerships for transit funding, did conditions begin to improve.  The four main thrusts were:

  • 1) Sustained buildout of subways and regional commuter rail, with upgrades and integration of the two systems; 
  • 2) Priority to buses, with hundreds of new miles of dedicated lanes, often configured as physically separate Bus Rapid Transit; 
  • 3) Tighter traffic management, with camera and radar systems and removal of most heavy trucks; and 
  • 4) Rezoning to favor dense residential construction along transit corridors, discourage parking supply in new buildings, and encourage light mobility.
Figure 2: São Paulo state highway map. Source:
Figure 3: São Paulo location within Brazil. Source:

These days São Paulo’s slow traffic means mainly slow movement of cars and trucks, not necessarily of buses or commuters overall.  Growing numbers of people get around by public transit more quickly than ever. The combined rail system (Figure 4) now includes 13 routes totaling 377 kilometers (236 miles), not counting an entire new subway line under construction.7  Two state companies and a private one operate the trains.  In 2019, before Covid-19 disrupted work and transportation patterns, the trains topped 6.5 million rides daily on average.8  That made São Paulo the first city in the Americas to surpass the combined subway and rail transit volumes of New York City and its metro area, by 14%.9   

Subway and train station openings are now routine in SP. Four more opened in 2021, and work on new underground lines, monorails, inter-urban trains, airport connections, and station overhauls also continued, even under Covid. With so much new infrastructure, there are features sometimes missing in older systems or richer countries, such as communications-based train control (CBTC) and in-train broadband. Glass platform barriers in new stations improve safety and passenger experience. Where subway lines meet commuter rail there are combined stations, free transfers, easy access, and unified signage. The fare from anywhere to anywhere on the system — maximum distance around 115 km or 72 miles — is about US$0.90.  

Figure 4: São Paulo’s growing rail network. Source: Companhia do Metropolitano de São Paulo

Separately, São Paulo overhauled its urban Master Plan in 2014.  The city upzoned, replaced parking minimum requirements with maximums, and offered builders tax discounts for building densely along rail and bus corridors. This has already begun to reshape the skyline, with some 124 thousand compact apartments (up to 45 sq. meters or 484 sq. feet) built within 500 meters of subway or train stations or 300 meters of a dedicated bus corridor, between 2014 and 2020.10

The travel-mode impact of all this, shown in Figure 5, is that subway and train ridership in SP doubled between 1997 and 2017, according to the state’s decennial Origin and Destination survey. Combined rail gained five percentage points in modal share of motorized trips over the period, rising to 16.3%. Private cars lost 6.7 points of share.  

Figure 5: Mass transit gains.

BRT and other bus lanes

Meanwhile, the city’s main attempts to improve the surface situation involve bus lanes and Bus Rapid Transit. Public and school buses now provide 15.8 million rides on average daily, three times the volume in greater London, for example.11 Many of these are on Bus Rapid Transit (BRT) lines, a system used around the world but invented in Curitiba, Brazil in the 1970s.  BRT buses run not merely on specially marked lanes but on physically separated ones, including their own elevated roadways and bridges, so blockage by unauthorized vehicles is rare. There are bypasses at many stops so a bus halted for passengers doesn’t obstruct others coming behind, and some busways are doubled for long stretches. The vehicles have extra-wide doors and all-door boarding at platform level, without steps. Passengers pay their fares at stations or have electronic passes, so there is no delay taking fares to board and dozens of people can get on and off in seconds.   Complementing the BRT corridors are traditional bus lanes (see Figure 6), some operating all day, and others only during rush hours. These were expanded substantially starting in 2013. The lanes are not as fast as BRT routes, but São Paulo has no compunction about using camera enforcement, so for the most part car drivers stay out of the way. The CET estimated in 2019 that at rush hours on major arterials, only 24% of people passing by were in cars, including taxi/rideshare, while 64% of people were in buses, 5% on motorcycles and 6% on foot.12

Figure 6: SP exclusive bus lanes (blue) and physically separate bus corridors (orange) Source: CET, Mobilidade no Sistema Viário Principal, Volumes e Velocidades 2019

Smoothing car traffic and barring trucks

One other move to speed traffic flow has been, counterintuitively, to lower speed limits.  Traffic enforcement in SP is by the municipal traffic agency, not police, and is largely automated. Most drivers are familiar with a letter in the mail with a photo of their car, perhaps moving as little as 10% over the speed limit, or in a prohibited lane, or caught on the street just after the rush-hour prohibition begins for their plate.  These infractions cost not just money but four to seven points on a license. Forty points in one year of this type (or less if there are more-severe infractions) means a six-month license suspension. 

With so many cameras, drivers tend to drive at the limit in many areas of the city. They may use apps or cruise-control settings to warn of drift over the limit. The impact was visible when São Paulo speed limits dropped in 2014. Peak speeds fell immediately, but average speeds actually rose, because traffic flowed more smoothly with less clumping, less braking and reacceleration, and fewer collisions. City traffic deaths dropped by a fifth between 2014 and 2016.13   

As for cargo traffic, SP in recent years completed most of a new super-highway around the metro area. With Santos, Latin America’s busiest port, located just 20 miles away but half a mile down an escarpment, lines of trucks once crowded SP avenues and inner ring roads. They blocked avenues and added diesel air pollution. The new outer ring road solves this problem by connecting eight of the nine highways into the city to each other, with limited local access. The ninth is soon to be included via a northern section. Trucks now loop around instead of through the city to continue down the mountains or to points north and south.  Municipal rules also now ban cargo vehicles — not just 18-wheelers but any truck over 7.2 meters (24 feet) long —  from the expanded central city from 5 am to 9 pm weekdays.

Figure 7: Down the mountains to the port of Santos: Rodovia dos Imigrantes. Source:

Why congestion remains

Puzzlingly, despite all the rail and bus improvements and their success at shifting millions of people onto public transit, despite keeping out even mid-sized trucks during the business day and a fifth of cars during every rush hour, and despite major highway expansions, São Paulo still ranks among the worst cities for congestion globally. Traffic-tech company Inrix ranked it fifth in the world for traffic jams in 2019,14 while GPS-mapping company TomTom ranked it 24th of 416 cities.15

Weekday rush-hour vehicle volumes and speeds on 41 stretches of arterial routes and local lanes of highways are measured by São Paulo’s municipal traffic agency CET, or Traffic Engineering Company. The samples historically were taken only during a few days each year (timed to avoid vacation periods and holidays) and are affected by numerous factors besides mode share, such as general economic activity.  Nevertheless, the graph in Figure 8 makes clear that rush-hour non-bus inbound speeds in CET’s sample in 2019 were at an all-time low — a painful 15.1 kph or 9.4 mph — while the outbound pace was the second slowest ever.  

The drag is not likely due to the lowered speed limits in 2014, since the recorded averages are far below the 50-to-70 kph (31-to-44 mph) legal limits on arterials and non-express lanes of highways and 40 kph (25 mph) on lesser streets.  As noted earlier and shown in the graph, average speeds actually rose just after speed limits were lowered, although an economic downturn also helped clear the roads. 

Figure 8: Rush-hour traffic as slow as ever (outside bus lanes). Source: CET via Secretaria Municipal de Urbanismo e Licenciamento

The CET studies aren’t necessarily the definitive description of reality, since they are a snapshot of a limited sample of a large, complex system. There is some survey and anecdotal evidence that traffic is not quite as maddening in recent years as in previous decades. The agency releases data on the sum of traffic jams across the city, and the worst ever was reported as 344 km (215 miles) on a day in 2014 with rain and a bus strike.16 The normalized post-Covid pattern is meanwhile not known yet.  

Still,  in São Paulo today a car trip from the international airport that takes 30 minutes without traffic may easily take an hour and 45 minutes at rush hour,17 while CET rush-hour reports of traffic-jam totals can still top 200 km (125 miles).18  

Cars don’t fit

How can there be so little progress on congestion when so many anti-congestion tools have already been deployed?  The answers are in geometry, demographics, and economics, particularly the concept of induced demand for road space.  

Beginning with geometry, the picture in Figure 9 shows the packed seven lanes of just one side of the widest highway inside city limits, the Marginal Pinheiros.  Problems on this route routinely slow ambulances, delay deliveries, and complicate business appointments.

Figure 9: Marginal Pinheiros jammed. Photo: MeuCarro, Aloisio Mauricio / Estadão Conteúdo

It’s surprising just how few individuals are actually causing the obstruction. Using the far right lane (from the reader’s point of view) as a base line, there are only about 20 vehicles from the very front of the picture all the way to the array of cameras on a gantry in the mid-background. All but four of those appear to be passenger cars, most of them presumably carrying a single person, the driver.  Other lanes have a greater share of trucks and buses, but if there are 1.43 people on average per car (SP’s rush-hour average according to CET)19 and 84 cars visible, then about 60% of the entire space is being taken up by only about 120 people.  

Meanwhile the two buses in the foreground each can carry about 52 passengers,20 104 total, comfortably seated, nearly the equivalent of all the private cars in the foreground. Standard SP mainline urban buses (Figure 10) have twice that capacity, 99 passengers seated and standing, according to bus-transit agency SPTrans. The common double articulated models on SP streets can carry twice that again, 198 people.21 One of those could replace all the foreground cars in Figure 9 and make the road look half empty but still be less than 2/3 full itself.

On demographics, returning to the statistics cited earlier, there are 3.8–4.4 million cars active in São Paulo. If drivers choose to drive even 5% more or less on a given day, they add or subtract 190–220 thousand cars from circulation.  Considering the limitation seen above, where it takes only 84 cars to occupy a substantial stretch of the city’s widest freeway, numbers of that magnitude overwhelm the system.  At full speed rather than in a traffic jam, those same cars of course take up several times more space.

Figure 10: Standard SP mainline urban bus. Source:

Free is a bad price

On economics, ever since Robert Moses built new bridges in New York City to ease congestion on older ones but found that the new ones soon congested too — and even seemed to worsen congestion on the old ones22 — traffic engineers have had to deal with “induced demand.” Road space is a valuable resource, particularly in crowded areas.  Any scarce resource given away for free or below market price tends to induce excess demand. 

That’s why barring some vehicles at certain times in São Paulo has never had the congestion impact expected. The bans open up space each day that those cars would have occupied, but that space becomes available — for free — to others who otherwise wouldn’t have driven. 

Free road space is rationed by queue rather than price. When streets congest, some potential drivers drop out of the queue. They choose to shop closer to home, eat lunch closer to the office, share rides to drop kids at school, or chain errands together to be more efficient.  They may walk or bike for a short trip, or take the subway or a bus corridor at peak traffic hours, even if they use their car weekends and evenings.  One 2019 survey showed only 3% of São Paulo residents say they “never” use public transportation.23

When new space opens up and traffic flows, however, the decisions start to reverse. People rejoin the queue until its length (i.e. their delay) again discourages the marginal choice to drive.  These daily decisions trip by trip are complemented by long-term decisions to choose more distant residential or business locations or own more vehicles as roads expand. 

Going back to events in the 1990s, this dynamic can be seen in the remarkable ban of 20% of all cars from rush hours starting in August 1996. That instrument was so severe that few doubted it would depress car usage, but in fact it barely showed up in traffic volumes, as shown in Figure 11. Evening growth flattened (the return from work tends to be more flexible in timing than morning arrival) but quickly resumed and never showed anything near a 20% drop.  Morning volumes actually rose substantially the year after the new restrictions. The subsequent dip in 1998 looks like any other economic-output graph for Brazil that year, as a recession hit, but car volumes again hit all-time highs in 1999.

Figure 11: 20% ban in 1990s fails to stop rush-hour car volumes from rising.

The overall picture is of a quasi-natural equilibrium. Volumes rise (or fall) to fill just the space available, at least for a city like SP with a large pool of automobiles constrained by congestion. 

This is also apparent in the more recent period.  Over the past two decades, significant space on arterials was removed from car access to prioritize buses.  Figure 8 earlier showed that vehicle speeds reached their lowest ever around 2019, even though Figure 12 below reveals that far fewer cars are using the major arterials than in 1999.  Putting those two observations together suggests that driver usage of those routes dropped only to the exact point of bearable congestion.

Figure 12: Fewer cars in 2019 than in 1999 (but in less space, for same congestion.

Considering these numbers and the mode-switch to mass transit under way, it’s tempting to conclude that automobile congestion is perhaps not really much of a problem, since it affects a shrinking proportion of the population.  Most Paulistas today simply move around the blockage. That may in fact be a healthy way of thinking about congestion in general, shifting away from the “windshield perspective” that presumes moving cars is the primary goal, as opposed to moving people by any mode.

However, many car trips are essential, and one critical category of vehicle can’t simply change modes: commercial trucks. Although larger trucks are banned in central São Paulo during the day, the market has adapted with fleets of smaller ones. A huge array of services is also carried out with vans, pickup trucks, SUVs, and hatchbacks, from package delivery to window cleaning.  São Paulo is Brazil’s business capital, and while the first comprehensive Origin/Destination study of truck traffic in the city is still under way, the National Surface Transportation Agency counts nearly a quarter of a million businesses in the municipality alone that likely depend on surface cargo movement.24 Congestion remains a major problem for them.

Congestion pricing as the missing solution

But if even the São Paulo package of rail expansion, cleared bus lanes, zoning for transit, bans on cars at rush hour, removal of trucks, and automated traffic enforcement has not eased automobile congestion, does that mean there is no solution? Not necessarily. Traffic can be made to move smoothly via pricing, as shown in the few big cities globally that have put tolls on urban driving — Milan, London, Singapore, and Stockholm — and in tolled express highway lanes in the US.  This is especially true of dynamic pricing that rises and falls directly with the degree of congestion, in line with theory first laid out in 1963 by Nobel laureate William Vickrey.25

Urban tolls might in fact be more easily implemented in SP than in most places, because the city already monitors millions of specific cars each day to apply the 20% rotation rule, enforce speed limits, and keep the bus lanes open.  A mere change in the rules, with few physical changes on the street, could turn automated tickets into congestion charges.  

Congestion charging tends to be supported by the public once in place because it works, but introduction faces formidable political barriers. Car owners — who tend to be wealthier than average and have greater political voice, certainly in Brazil — naturally protest a system so different from what they’re accustomed to, and which amounts to a tax only on them, despite benefits also to them. 

If it can get past these politics, São Paulo could harvest the broad economic-efficiency gains of no longer pricing the scarce resource of urban street space so low that demand constantly exceeds supply. While most people in SP today solve their personal traffic problems by simply avoiding traffic, on public transit, congestion pricing could allow the entire city to flow more smoothly and efficiently. The most tangible benefits would go to those who most highly value open road space at a given moment, whether a student late for university entrance exams, a trucker making urgent deliveries, or an ambulance driver trying to reach crash victims.  

In his history of São Paulo,26  Roberto T. de Pompeu describes the grand opening in 1911 of the municipal opera house, still a landmark today. Hundreds of carriages and about 150 automobiles descended on the new theater, blocking the area so completely that the premiere production of Hamlet had to be delayed and the final act cancelled.  Couples in gowns and tuxedos refused to get out and walk in the mud, and at 1:00 in the morning the mess was still being untangled. Pompeu cites this as the city’s first modern traffic jam. 

Over a century later, São Paulo amounts to a test case of transit and traffic theory, with global implications. One question is whether a maturing mega-city can make traffic flow freely simply by offering drivers enough transit alternatives.  The São Paulo evidence suggests the answer is no.  It might be time for Americas’ biggest municipality to test the only approach that has worked elsewhere against urban congestion at scale, the pricing of road space.


Stephen H. Graham is a master’s degree candidate at NYU’s Wagner Graduate School of Public Service, focused on Urban Planning. He lived in Brazil for many years, first as researcher at the Business International division of The Economist magazine group, then stock analyst for investment bank UBS, and later Director of Research for Latin America for Goldman Sachs and Citigroup, where he also became Deputy Director of US Research. He teaches Emerging Markets Finance at NYU Stern, where he received his own MBA, has a BA in History from Occidental College in Los Angeles, and studied graduate Economics on a fellowship at the University of São Paulo. He is also responsible for the overconfident urbanism opinions on the Twitter account @ForwardBike, aka WorldBikeForward.



1. Brazilian Institute for Geography and Statistics, cited in

2. Companhia do Metropolitano de São Paulo, Pesquisa origem e destino: 2017, (2018), p. 15.

[Transl.: Metropolitan Company of São Paulo, Origin and destination survey: 2017] 

3. Carolina Giovanelli, et al. “Quantos veículos circulam por São Paulo ao certo?”, in  Veja São Paulo (Dec. 29, 2016).   [Transl.: Just how many vehicles circulate in São Paulo?].

4. Companhia do Metropolitano de São Paulo, Pesquisa origem e destino: 2017, p. 45.

5. Personal observation.  See also:

6. Companhia Paulista de Trens Metropolitanos, História da CPTM.  [Transl.: History of the Paulista Metropolitan Train Company].  (2009).  

7. Ricardo Meier, “Metrô de São Paulo chega a 104,4 km e rede sobre trilhos, a 377 km” [Transl.: São Paulo subway reaches 104.4 km and rail network 377 km], on MetrôCPTM blog, (Dec. 30, 2021).

8. Companhia do Metropolitano de São Paulo, Relatório integrado 2019 (2020), p. 20.   [Transl.: Metropolitan Company of São Paulo, Integrated report 2019]. (2019)

9. Metropolitan Transportation Authority – MTA, Annual report to the governor: 2019. (2020); 
Metropolitan Transportation Authority – MTA,; 
D. Hutter., “Port Authority sets passenger records in 2019;” in NJBiz. (Feb. 4, 2020); 
NJ Transit, 2019 annual report: independent auditors’ report. (2020)

10. Priscilla Mengue, “SP tem boom de microapartamentos em bairros nobres,” [Transl.: SP has boom of microapartments in upscale neighborhoods] in O Estado de São Paulo, May 14, 2022. 

11. Transport for London, “Number of journeys by TFL reporting period, by type of transport,” in Greater London Authority, (June 2022). 

12. CET Departamento de Pesquisa de Trânsito, Mobilidade no sistema viário principal: volumes e velocidades 2019. [Transl.: CET Traffic Research Department, Mobility on the arterial system: 2019 volumes and velocities]  (São Paulo, October 2020), p. 40.  

13. Estado Conteúdo, “Número de radares cresce 57,5% na gestão Haddad,” in Veja São Paulo (June 1, 2017). [Transl. “Number of radars rises 57.5% in Haddad administration”].

14. Inrix, Global traffic scorecard 2020 (2021),  

15., Traffic index 2019 (2020),

16. Amanda Gomes, “Trânsito chega a 344 km e bate recorde histórico em SP,” in Agora São Paulo (May 24, 2014).  [Transl.: Traffic reaches 344 km and hits historical record in SP]

17. Personal experience.

18. Notícias R7, “Saiba qual foi o dia que SP registrou recorde de lentidão em 2019,” (Dec. 31, 2019). [Learn which was the day SP registered the record for slowness in 2019.]

19. CET Departamento de Pesquisa do Trânsito, Mobilidade no sistema viário principal: volumes e velocidades 2019, p. 38.

20. Airport buses run by Azul Airlines, which observation shows to be Marcopolo-brand tour buses.  Capacity data available at

21. A.M. Andrade, Capacidade de transporte por modo (2014).  [Transl.: Transport capacity by mode.] (presentation)

22. Robert Caro, The power broker. (New York: Alfred A. Knopf, 1974), p. 518.

23. Rede Nossa São Paulo and Ibope Inteligência, Viver em São Paulo: Mobilidade urbana (presentation), [Transl.: Our São Paulo Network and Ibope Intelligence, Living in São Paulo: Mobility] (São Paulo: Rede Nossa São Paulo, 2019), p. 19.  

24. ANTP and CET, Série cadernos técnicos volume 22: Planejamento da pesquisa de origem/destino de cargas no município de São Paulo, [Transl.: Technical papers series volume 22: planning for the origin/destination survey in the municipality of São Paulo] (December 2015), p. 35.

25. William S. Vickrey, “Pricing in Urban and Suburban Transport,” in The American Economic Review,  Vol. 53, No. 2, (American Economic Association, May 1963), pp. 452-465.  

26. Roberto de Toledo Pompeu, Capital da vertigem: uma história de São Paulo de 1900 a 1954. [Vertiginous capital: a history of São Paulo from 1900 to 1954.] (Rio de Janeiro: Editora Objetiva, 2015), pp. 78-79.