About School Trips

School Trips models every SFUSD and charter student's trip to school, today and with each change. Today 22,184 of 51,769 students, 42.9%, come by single-family car, against the 30% goal in SF Safe Routes to School's 2024-25 report. That report's Travel Tally gives 48.6%, but it weights the grades by who answered, mostly kindergarten and 5th grade. We weight every grade by its CDE Census Day 2025-26 enrollment, and older grades drive less.

Every figure is our central estimate, with a low and a high beside it. The low and high are our assumption cases, not a statistical range. Student figures cover SFUSD and charter schools only, since private schools publish no enrollment by grade. Kids hurt counts every child in the city, private school students included.

Scenarios

Today has no changes. Best for Car Mode Shift, Best for Safety and Balanced are plans for one budget, $5M to $100M over 4 years, that pays for the school changes and the new bikeways together. A budget a year is the yearly cost plus the one-time cost spread over 4 years.

Each plan starts with nothing funded and keeps adding whichever change, at whichever school, does the most per dollar until the budget runs out. Best for Car Mode Shift counts kids out of single-family cars, Best for Safety kids hurt prevented, and Balanced adds the two with equal weight, each divided by what that goal's own plan reaches at the same budget, so 1 is the best we found for that goal. A change counts only if it gets at least 1 student out of a car or prevents 0.01 kids hurt a year, so a plan that runs out of such changes spends less than its budget.

The new bikeways are one citywide step with 6 sizes, the sets under New Bikeways below, and a plan moves to a bigger set when its gain per dollar beats the next school change's, so the plan itself splits the budget between bikeways and schools. Carpool matching and school zone speed limits are citywide steps too, at every school or none. A change only counts at a school if it takes at least one student out of a car or prevents at least a hundredth of a kid hurt a year, and a shuttle loop needs at least 25 riders. Each reach of the walking upgrades is its own choice, so a school can go straight to a quarter mile. Every reach is priced by the school's own intersections, and one near two schools is paid for at both. The plans are picked at our central estimates and a 20% bike switch, so the slider changes their figures but not what they fund. Every plan uses the Neighborhood + Citywide Programs assignment the Board adopted on 13 October 2026, as it stands in 2030-31 with 52% of TK-5 students under it. It costs nothing and on its own takes about 492 kids out of single-family cars and prevents 0.2 kids hurt a year, so part of each plan's figures is the assignment. Choosing another option under Advanced Controls moves a plan's figures by the two options' own difference.

A plan is picked from a table of every change at every school scored on its own, which comes within 2.5% of a slower pick that rescores every move with the rest of the plan in place. Under Advanced Controls, each change has a switch for whether the plans may use it. Turning one off picks the plan again in your browser from the same table, and moves its figures by the table's difference between the two plans, which we checked against exact scores for all 252 plans with one change off: within 5.3% on kids out of cars and 0.08 kids hurt a year. Those figures are scored at a 20% bike switch, so the slider moves a re-picked plan less exactly.

A plan's cars come from each school's own simulation with the changes funded there, added up. Its kids hurt come from each school's own crashes: safety upgrades act on the kids hurt nearest that school within their reach, speed limits on the kids hurt in its 500 ft zone, and fewer cars cut injuries at each change's citywide rate per car. The two maps show a plan approximately, scaling each change's shift on every block by the share of its citywide effect the plan's schools carry.

A change with no cost shows No cost per $1M, and one that costs money but saves nothing on that measure shows n/a. Changes overlap, so their own effects don't add up to a scenario's central figure, which comes from running them together. A scenario's low moves that central figure by each change's own gap between its low and central, added up, and its high the same way. Kids hurt are multiplied instead: if one change leaves 90% of injuries and another 80%, together they leave 72%.

Four changes are tools in SFMTA's Safe School Connections draft approach (September 2026), shown in blue in the each-change charts: School Street Activation, Safety Upgrades for Walking, 15/20 mph School Zone Speed Limits and School Connector Street. The rest are our additions, in yellow.

What the Model Does

Each rule below applies to every change. The ones we set ourselves are marked as our assumptions.

No study measures the bike switch for this exact group, so the rates are ours. They come from studies that measured travel after something was built. Where a route passed a California Safe Routes project, 15.4% of children walked or biked more, against 4.3% elsewhere. In London's mini-Hollands, adults who cycled in the past week rose from 15.2% to 22.4% after a year. Across 31 studies, cycling rose a median 22%. As a share of families who were driving, that's about 10% to 25%. Our rate applies only to families who now have a safe ride as quick as the car, so 20% sits near the top. What parents say runs far higher, 84% in Los Gatos, and we don't count it.

How We Model Each Change

Each figure below is the change alone at every school it reaches, against today. Seven changes use a study's measured effect: the speed limits, School Street Activation, Safety Upgrades for Walking, the Walking School Bus, 20 mph business districts, speed humps and speed cameras. Eight are our estimates, since no study measures them: Carpool Matching, Free Muni for Caregivers, School Shuttle Loops, E-Bike Help, the Daily Travel Tracker, the E-Cargo Bike Month, the Bell-Time Curb Restriction and the Bike Bus. The new streets and bikeways use the bike switch, and the assignment options use SFUSD's own distances.

15/20 mph School Zone Speed Limits

• What it is: a lower posted limit on the streets within 500 feet of each school while children are present, which AB 382 lets the city set. It's a tool in SFMTA's draft.
• Who it reaches: all 113 schools, and the kids hurt inside those zones while the limit applies, at any grade.
• How we model it: kids hurt within 500 feet of the school grounds while the limit applies fall 10%, range 5% to 25%. That's 24 kids a year in the zone, 45% of them on weekdays from 7 to 9 am or 2 to 5 pm, August to June. It moves nobody out of a car, since a lower limit doesn't make a street safe to bike. At our low, central and high, it takes nobody out of a single-family car, and there are 0.5, 1.1 or 2.7 fewer kids hurt a year.
• Why: speed sets how hard a car hits. Across studies of urban roads, cutting mean speed from 25 to 20 mph predicts about 23% fewer injury crashes (Elvik 2009, reproduced by Transport Scotland, Table F-1).
• Evidence: Edinburgh's 20 mph limit cut casualties 10% on the changed streets against unchanged ones, with mean speed down 1.3 mph. In Wales, casualties fell 26.2% on roads that went to a 20 mph default, which sets our high. Both were limits on whole areas at every hour, and a school zone's limit applies only part of the day, so we apply the cut only inside the zone while it does.
• Cost: about 12 signs a school, one at each street entering a zone that reaches a quarter mile, at $650 a sign: $7,500 a school, once, range $2,000, the 4.4 signs a school SFMTA posted for its 2012 school zones (SFMTA, 2012), to $16,000. That's $848,000 once at 113 schools. Cameras aren't priced.

School Street Activation

• What it is: the block in front of an elementary school closes to cars at drop-off and pickup. It's a tool in SFMTA's draft.
• Who it reaches: TK-5 students at the 79 schools with TK-5 grades, since its study covered primary schools.
• How we model it: each school's TK-5 single-family car share falls 2.7 points, range 0 to 5.3, and those trips move to the other ways of getting there in proportion. Fewer cars cut kids hurt near schools at the bells. The car-free block gets no extra credit yet, since no study has counted injuries on one. At our low, central and high, it takes 0, 670 or 1,315 kids out of single-family cars, and kids hurt a year change by 0 fewer, 0.3 more or 0.1 more.
• Why: it's the one controlled study. Of its 5.40-point drop in car trips, 2.68 points were park-and-stride, where the family still drives and walks the last block or two. The Travel Tally has no answer for that, so those kids would still count as driven here, and we take the other 2.7.
• Evidence: Patterson et al. 2025 compared 166 UK primary schools, 125 of them in London, with similar schools nearby. Car share fell 5.3 points (2.5 to 8.2) in its abstract, three years before against one year after. Our TK-5 car share is about 55%, so the same points are a smaller relative drop here.
• Cost: $30,000 a school a year, range $13,500 to $60,000: two staff, one at each end of the block, 1.5 hours a day for 180 days, at the $56 an hour SFMTA's crossing guards cost. Set-up is $17,000 a school, once, from Hampshire's School Streets' £12,000 to £15,000. That's $2.4M a year and $1.3M once.

Safety Upgrades for Walking

• What it is: bulbouts, raised crosswalks and flashing beacons at the intersections near schools where at least one street is a main street, a tool in SFMTA's draft. A main street is one OpenStreetMap ranks tertiary or above, the same test the crash chart sorts kids by. Corners where both streets are residential are left to calming, our addition: humps and diverters on the residential streets within 660 feet of the grounds of each school a $10M calming budget reaches.
• Who it reaches: TK-8 students at the 93 schools with TK-8 grades, the grades its study covered. Calming reaches 29 schools, taken in order of kids hurt within 660 feet over the last 10 school years.
• How we model it: walking and biking rise 3.3 points among each school's TK-8 students, range 0 to 6.3, at any reach, since no study measured more for a bigger package. Calmed streets count as safe to bike, so the bike switch moves more kids at those schools. The upgrades reach 1 block (350 feet), 2 blocks (660 feet) or a quarter mile from the school grounds, and cut kids hurt within that reach and within 150 feet of a main street intersection they treat, by how they were travelling: 40% (20% to 55%) of kids walking, 25% (0% to 40%) of kids biking or on a scooter, and 12% (0% to 25%) of kids in a car, counted at TK-8 schools. Those cuts apply in full from 7 to 9 am and 2 to 5 pm in term, the hours the studies measured, and at 50% the rest of the school day, our judgment, since the upgrades stay in place. Kids in a car include 16 to 18 year olds driving.

  Reach	Kids hurt a year, walking	Biking or scooter	In a car	Fewer walking	Fewer biking	Fewer in a car	Fewer in all, with calming and fewer cars	One-time cost
  1 block	3.8	2.3	5.3	1.1	0.4	0.4	3.2	$21.6M
  2 blocks	6.5	2.5	10.5	1.9	0.4	0.9	4.5	$37.8M
  1/4 mile	10.8	6.3	26.0	3.1	1.0	2.3	7.6	$84.0M

  Calming cuts 40% (20% to 55%) of kids hurt where both streets are residential within 660 feet of a calmed school's grounds, 3.8 kids a year. At our low, central and high, it takes 14, 817 or 1,549 kids out of single-family cars, and there are 1.1, 3.2 or 5.5 fewer kids hurt a year.
• Why: it's the one study we found that measured walking and biking after engineering work at schools. Three studies measured the injury cut near schools, and crash studies of the treatments themselves measured kids in cars.
• Evidence: McDonald et al. 2014, p. 160, found walking and biking up 3.3 points at schools with an engineering improvement, across 801 schools in four states. It printed no interval, and its p-value puts the top near 6.3. Austin's STREETS study found no gain, so the low is 0. In Barcelona, child pedestrian injury collisions fell to 0.57 of before within 200 m of 64 Safe Routes schools, against 1.05 at matched schools. Barcelona's program was mostly road safety teaching, so it supports the size of the cut more than the treatments. That 200 m, about two blocks, is the only reach any study states. DiMaggio and Li 2013 found 44% fewer pedestrian injuries in New York City tracts picked for high injuries, and an 18-state study found 23% fewer kids hurt walking and biking together. No study measured these treatments for kids biking alone, so biking's figure is our judgment from the pooled one. For kids in cars, leading pedestrian intervals carry 0.86 of injury crashes for all road users, and NYC DOT found 10% to 13.5% fewer car occupants hurt after curb extensions and leading intervals. Flashing beacons may add rear-end crashes (FHWA 2025), so the low is 0, and the 12% is our judgment. London's engineered 20 mph zones cut casualties 42% (Grundy et al. 2009).
• Cost: $200,000 a school, once buys every crossing within 1 block of the average TK-8 school's grounds, range $60,000 to $290,000, between SF's ~$60,000 walk audits (School Access Plan, p. 46) and the $290,000 Monterey Boulevard corridor. That's 574 intersections citywide, and the upgrades go to the 359 of them on a main street, so a reach costs that price in proportion to its main street intersections: 359 within 1 block, 858 within 2 blocks and 2,285 within a quarter mile. Intersections come from OpenStreetMap, counting only named streets that cross, with a divided street's two sides counted once. A school's own price in the school view uses its own intersections. Calming is $10M once, citywide, range $5M to $20M, at $0.25M a mile, about 39 miles. With calming, that's $21.6M once at 1 block, $37.8M at 2 blocks and $84.0M at a quarter mile.

School Connector Street

• What it is: a link from a school's door to the nearest all-ages bikeway, separated on busier streets and calmed on residential ones. It's a tool in SFMTA's draft.
• Who it reaches: the 23 schools within 2 blocks of today's all-ages network, and the driving students at every grade whose bike trip it makes safe.
• How we model it: the shortest walk from the door to a safe run at least half a mile long counts as safe. A driving student it newly gives a safe ride within 5 minutes of the car switches to biking at 10%, 20% or 30%. We count it with School Streets on, so the block in front of the school isn't built twice. A child biking on a calmed street carries about 1.5 times a car trip's risk, and few kids switch, so it changes kids hurt by almost nothing. At our low, central and high, it takes 21, 42 or 63 kids out of single-family cars, and kids hurt a year change by 0.1 more, 0.01 more or 0.02 fewer.
• Why: it's how SFMTA's draft describes the connector, and the bike switch test is the same one every bike change uses.
• Evidence: the bike switch studies under What the Model Does. No study measures a school connector on its own.
• Cost: each block at its street type's rate a mile: $1.5M a mile on a main street, $0.25M a mile on a residential street. That's $6.4M once.

New Bikeways: $5M to $100M

• What it is: for each budget, the set of street stretches that brings the most driving students within a safe ride. A stretch is about a mile of one street. It can be an unbuilt all-ages bikeway from the Biking and Rolling Plan, priced at its bikeway type's rate, except that on a collector or arterial a shared street the plan draws, or an existing plan bikeway that isn't all-ages, is priced and counted as a separated lane, or any other street block: a residential block priced as calming, a collector or arterial as a separated lane, with freeways and ramps left out. SFMTA's upcoming bikeways count as built in every scenario, at no cost.
• Who it reaches: driving students at every grade whose trip a chosen set makes safe. A budget plan buys one set out of the same budget as its school changes.
• How we model it: an off-plan block is dropped when an all-ages street, built, upcoming or on the plan, runs within 20 degrees of its direction and within 500 feet along most of its length, except a cross street that meets the plan street or a block that continues it past its end. Each budget's set is the better of two methods, whichever the full route search counts more driving students for: an integer program that solves for the whole set at once, and a ranking that adds the stretch with the most new driving students per dollar, one at a time. In the integer program, a student counts when a route to school passes the bike switch test, every 1,000 feet of streets off the network on it holding a built stretch, and the solver maximises the driving students counted within the budget. A stretch that brings nobody within a safe ride on its own but sits on a chosen route is a bridge, drawn with a dashed outline on the map. The same bike switch applies, at 10%, 20% or 30%. At our low, central and high, the kids out of single-family cars are 123, 246 or 368 at $5M, 197, 393 or 590 at $10M, 264, 528 or 792 at $20M, 475, 951 or 1,426 at $40M, 681, 1,363 or 2,044 at $80M, 767, 1,535 or 2,302 at $100M. Calming a set's residential blocks cuts kids hurt on them by 40% (20% to 55%), as safety upgrades' calming does, from Grundy et al. 2009. We leave out kids hurt within 660 feet of a school, where safety upgrades' calming may reach, and on the blocks picked for speed humps, so no kid is counted twice. New riders add risk of their own: protected lanes are 0% to 13% of a set's miles, and a kid biking on one carries a third of a car trip's risk, while on the calmed or shared rest it's 1.5 times. At our central the sets up to $20M change kids hurt by almost nothing, and the bigger sets spare about 1 a year or less. Few kids are hurt on these blocks, 13 or fewer in four years, so the figure moves when a budget swaps a few streets.
• Why: two stretches can be worth nothing alone and a lot together, and picking one at a time can't see that. Solving the set at once can, but on a city this size the solver doesn't finish in four minutes, so the ranking stays in as a floor.
• Evidence: the bike switch studies under What the Model Does.
• Cost: each stretch at its street type's rate a mile, the same rates as the connectors. The chosen sets cost $5.0M for 19.2 miles at $5M, $10.0M for 32.8 miles at $10M, $20.0M for 55.46 miles at $20M, $40.0M for 112.21 miles at $40M, $80.0M for 199.84 miles at $80M, $100.0M for 261.85 miles at $100M, once.

Walking School Bus

• What it is: kids walk to school as a group, led by an adult on a set route. Our addition.
• Who it reaches: TK-5 students living within a mile's walk, at the 79 schools with TK-5 grades.
• How we model it: walking and biking rise about 0.5 points school-wide, 0.37 of them from single-family cars, range 0 to 1 car point. That's the trial's 9.04 points among about 18 volunteers a school, spread over an SF TK-5 school of about 330. Fewer cars cut kids hurt near schools. A supervised walker counts as safe as the car trip they leave (range half a walker's extra risk to slightly safer), where a kid walking alone counts as riskier. One adult walks with about three kids, as Seattle's buses run, so two of three walkers need no parent of their own: 0, 82 or 223 kids. At our low, central and high, it takes 0, 92 or 248 kids out of single-family cars, and there are 0, 0 or 0.05 fewer kids hurt a year.
• Why: it's a randomized trial. It tested grades 3 to 5 among the families who signed up, so we spread its effect over the whole school.
• Evidence: Johnson et al. 2024 randomized 22 lower-income schools in Houston and the Seattle area, 2012 to 2016. Days walked or biked rose 9.04 points (1.10 to 16.98) for grades 3 to 5 within a mile. Its Intervention section says "The WSB program was led by trained research assistants", so volunteer-led buses may do less. The same team's Houston pilot found walking rose from 24% to 54% with buses.
• Cost: $25,000 a school a year, range $5,000 to $60,000: trained leaders on two routes every school day, as in the trial, about 540 leader-hours a year at $30 an hour, plus a part-time coordinator. Volunteer leaders set the low, and Córdoba's paid daily bus scaled to 180 days sets the high. That's $2.0M a year.

Carpool Matching

• What it is: a city tool that matches families to share school drives. Our addition.
• Who it reaches: driving families at every grade, citywide.
• How we model it: 0.5% of driven students move to a carpool shared by two families, range 0.2% to 1%, so half the parents are freed: 22, 55 or 111 kids. Each carpooling kid takes half a car off the street and stays in a car. We count kids hurt for TK-8 pools only, since teen-driven cars carry 4.7 times the injury rate per trip (TRB 2002). At our low, central and high, it takes 44, 111 or 222 kids out of single-family cars, and there are 0, 0 or 0 fewer kids hurt a year.
• Why: Contra Costa's SchoolPool drew match requests from 0.45% of its students, about 0.8% of driven students at SF's car share, and only 27% of its pools lasted a second year. 0.5% sits between those.
• Evidence: the one program with a stated denominator. Denver's SchoolPool reports 19,000 families on 150 campuses, method not stated, so we don't use it.
• Cost: $150,000 a year citywide, the middle of the School Access Plan's $100,000 to $250,000 for carpool coordination. Contra Costa's cost $95,000 in 2002-03 (FHWA).

Free Muni for Caregivers

• What it is: a caregiver rides free with a child going to school, extending Free Muni for Youth. Our addition.
• Who it reaches: the about 21,100 TK-8 students living over half a mile from school, in a straight line.
• How we model it: Muni's share among them rises 1 point, range 0 to 3. New riders come from single-family cars, carpools and bikes in the shares those students use today, so 88% leave a single-family car. That split is our assumption. A kid who switches to Muni sheds about 0.15 of a car trip's risk, since the walk to the stop is riskier for a child than the ride. The caregiver rides along, so no parent is freed. At our low, central and high, it takes 0, 187 or 560 kids out of single-family cars, and there are 0, 0.02 or 0.1 fewer kids hurt a year.
• Why: in Pittsburgh's randomized trial, free fares cut low-income adults' daily chance of a car trip 2.9 points, and trips with children didn't change. So 1 point is our judgment, with Pittsburgh's figure near the top.
• Evidence: Pittsburgh's trial is the one study of car trips under free fares. In Montreal, car occupants were hurt 3.7 times as often as bus occupants per km.
• Cost: $860 a rider a year, Muni's monthly pass for ten months, range $504 to $1,026. Every caregiver who rides with a child gets it, so today's riders count too: about 3,217 TK-5 students ride Muni at the citywide share, our assumption. That's $2.9M a year.

School Shuttle Loops

• What it is: yellow buses on fixed loops between neighborhood stops and several schools. Our addition, modelled on Denver's Success Express.
• Who it reaches: TK-8 students living over 2 miles' walk from school. About 6,500 students live that far, and SFUSD's general-education buses already carry about 2,000 of them (School Access Plan, p. 42). The Travel Tally has no school-bus bar, so we seed those 2,000 riders into today's mix, taken from every way of getting there in proportion.
• How we model it: 640 riders, range 580 to 700, at the take-up of Denver's Success Express loops, which carried 8.9% to 10.7% of eligible students, applied to SF's 6,500 TK-8 students over 2 miles. 48% of riders were driven before, from West Yorkshire's MyBus (64%) weighted by SF's grades, so about 310 fewer driven kids. A bus trip carries about a fifth of a car trip's injury risk (TRB 2002), and every rider needs no parent: 444, 490 or 536 kids. At our low, central and high, it takes 278, 307 or 336 kids out of single-family cars, and there are 0.1, 0.1 or 0.2 fewer kids hurt a year.
• Why: a loop shared by several schools costs less than a route per school, which is what made Denver's work.
• Evidence: Denver's loop cost about 30% less than its regular routes, $1.05M against $1.5M (Chalkbeat, 2011), and now carries almost 3,000 students to 35 schools (Denverite). SFUSD costed 1,520 more riders at $4.5M a year. A bus's lower risk is from Morency et al. 2018.
• Cost: $3,000 a rider a year before the state's share, from SFUSD's $2,960 a rider for its 19 new buses (School Access Plan, p. 42), range $2,050, today's fleet, to $3,400, SFUSD's 2023-24 cost (PPSSF). The state pays about 60%, so SFUSD's net is about $1,200. That's $1.9M a year at 640 riders.

E-Bike Help

• What it is: vouchers for drivers who'd bike only on an e-bike, at 4 schools.
• How we model it: 3%, 10% or 25% stop driving. At our low, central and high, it takes 0, 1 or 2 kids out of single-family cars, and there are 0, 0 or 0 fewer kids hurt a year.
• Evidence: UC Davis surveyed 617 rebate recipients: under 10% replaced a daily car trip.
• Cost: $1,500 a voucher, paid once to every family the e-bike makes eligible, whether or not they stop driving, above Denver's 2022 rebates of about $840 to $990 (BikePortland), plus $150,000 a year for the program.

20 mph Business Districts

• What it is: a 20 mph limit on main street blocks lined with shops and restaurants within a quarter mile of a school, which Vehicle Code 22358.9 (AB 43) lets the city declare as a business activity district.
• Who it reaches: 43 corridors near 25 schools, everyone on those blocks at any hour. A block counts when it's at least 150 feet long and has at least one retail or dining business address per 100 feet in DataSF's registered business locations, on its own street, and isn't already at 20 mph or above 25 mph in DataSF's speed limits. One of the law's tests is half the frontage in shops, which the registry can't measure, so the floor is set where SF's own lowered blocks sit: about half the main street blocks SFMTA has posted at 20 mph since 2022 reach it. State highways and streets over 4 lanes are left out, from OpenStreetMap.
• How we model it: kids hurt on those blocks fall 10% (0% to 19%), at every mode, 2.3 kids a year. Blocks within 500 ft of a school are already its school zone, so they're left out. It moves nobody out of a car. At our low, central and high, it takes nobody out of a single-family car, and there are 0, 0.2 or 0.4 fewer kids hurt a year.
• Evidence: in Edinburgh, 20 mph signs cut casualties 10% (95% CI 0% to 19%) on changed streets against streets kept at 30 mph, with no split by mode (Kokka et al. 2024). That was a 10 mph cut, and SF's is 5 mph, from 25.
• Cost: $19,000 a corridor, once, range $14,000 to $27,000, from four SFMTA Board items. That's $817,000 once.

Speed Humps by Schools

• What it is: humps on the residential blocks within 1 block (350 feet) of each school's grounds, one every 300 feet, at least one a block.
• Who it reaches: 1,139 humps on residential blocks at 107 schools, from DataSF's street file, leaving out blocks under 150 feet or over 1,000 feet, tunnels and ramps, and blocks that already have a hump, cushion or table in DataSF's traffic calming list.
• How we model it: kids hurt on those blocks where both streets are residential fall 53% (5% to 76%) walking and 50% (25% to 75%) for everyone else, 4.0 kids a year. Safety upgrades' calming already humps the residential streets within 660 feet at the 29 schools its budget reaches. So with safety upgrades on, humps go only to the other schools, where they spare 0.4 kids a year for $8.1M, and nothing is counted twice. The budget plans offer humps only at those other schools too. It moves nobody out of a car. At our low, central and high, it takes nobody out of a single-family car, and there are 0.8, 2.0 or 3.0 fewer kids hurt a year.
• Evidence: FHWA's CMF 134 puts injury crashes on local streets at 0.5 after humps, standard error 0.13. In Oakland, children were 0.47 times as likely to be struck near home where there were humps (95% CI 0.24 to 0.95, Tester et al. 2004), the only study of children, so kids walking range from 5% to 76% fewer hurt.
• Cost: $10,000 a hump, once, range $5,000 to $21,000. That's $11.4M once at every school.

Speed Cameras by Schools

• What it is: a new speed camera on a main street within 500 feet of each school. This assumes the state raises the cap in AB 645, which lets SF run 33 cameras, all in use now.
• Who it reaches: 96 schools, each with a camera on the main street with the most length inside its 500 ft zone, at the point nearest the grounds. Schools with one of SF's cameras within 500 feet, in DataSF's camera citations, are left out, and so is a camera that would sit within a quarter mile of one on the same street, or within 660 feet of another school's new one. State highways are Caltrans's and left out.
• How we model it: kids hurt on the camera's street within a quarter mile of it fall 13% (0% to 25%) walking, 21% (0% to 30%) biking and 14% (0% to 25%) in a car, 14.8 kids a year at 35 of the cameras. Scooters get no cut, since NYC's other motorized riders were hurt more at camera and control corridors alike. It moves nobody out of a car. At our low, central and high, it takes nobody out of a single-family car, and there are 0, 1.9 or 3.5 fewer kids hurt a year.
• Evidence: NYC's cameras added in 2022 against corridors without one, 2021 to 2023, within a quarter mile of the camera on the same street (NYC DOT 2024 report, p. 20): pedestrians hurt 0.87 times the control's change, cyclists 0.79 and people in vehicles 0.86. NYC published no interval, so the low is 0 and the high is our judgment.
• Cost: $3,750 a camera a month, the flat fee in SFMTA's camera contract, $4.3M a year for all of them. Fines are expected to cover it.

Daily Travel Tracker

• What it is: daily class trip logs, as in WOW, at every TK-5 school, we assume.
• How we model it: car share falls 0, 1.8 or 3.5 points. At our low, central and high, it takes 0, 447 or 868 kids out of single-family cars, and kids hurt a year change by 0.5 more, 0.1 more or 0.02 fewer.
• Evidence: self-selected Scottish schools walked 5 points more after three years, 71% kept going.
• Cost: $5,000 a school a year.

E-Cargo Bike Month

• What it is: a free e-cargo bike month for 12 families at each of 4 schools yearly.
• How we model it: 3%, 10% or 20% of the 73% driving bike instead. At our low, central and high, it takes 1, 4 or 7 kids out of single-family cars, and kids hurt a year change by 0.02 more, 0 fewer or 0 fewer.
• Evidence: ELEVATE: 10 of 49 households bought one within a year.
• Cost: $400 a family, $150,000 yearly.

Bell-Time Curb Restriction

• What it is: no stopping on the school block at the bells.
• How we model it: central case no change, high 1 TK-5 car point. At our low, central and high, it takes 0, 0 or 248 kids out of single-family cars, and kids hurt a year change by 0.2 more, 0 fewer or 0.01 fewer.
• Evidence: Dundee's voluntary cones cut driving 2.1 points, mostly to park and stride, which SF counts as driven.
• Cost: $5,000 a school, once.

Bike Bus

• What it is: an adult-led group ride, one route at each of 4 schools.
• How we model it: 0, 0.3 or 3 car points: volunteer, $4,500 weekly or $23,000 daily paid route. At our low, central and high, it takes 0, 4 or 44 kids out of single-family cars, and kids hurt a year change by 0.1 more, 0.01 more or 0 fewer.
• Evidence: no car outcome measured. 145 organizers report 10 kids a route, 60% weekly.

SFUSD's New Assignment

• What it is: Neighborhood + Citywide Programs, SFUSD's new elementary assignment, as it stands in 2030-31. The Board adopted it 5-2 on 13 October 2026 (The Standard, Mission Local), and the final policy is due 13 April 2027. Every scenario but Today uses it.
• Who it reaches: TK-5 students under the new rules: "Families enrolling in TK or kindergarten for the first time, and families new to SFUSD entering an elementary or K-8 school", from "School year 2028-29". It leaves out "Families currently enrolled, who will not be moved" (Study Session #1 deck, scope slide).
• How we model it: each TK-5 trip's distance shrinks by the option's ratio to today, and every student chooses again how to get there. We read the rollout as TK and K in 2028-29 and one more grade each year, so the new rules cover 20% of TK-5 students in 2028-29, 52% in 2030-31 and all of them in 2033-34, each grade weighted by its 2024-25 SFUSD enrollment. That schedule is our reading, not SFUSD's, and sibling priority could slow it: sibling is the first tiebreaker for TK and K today (SFUSD), and the 2026 decks don't say how siblings are treated. [inference — mine] The tool blends each school's mix 52% with the option and 48% without, as in 2030-31, the plans' fourth year. Then it takes 492 kids out of single-family cars, range 38 to 832. At full effect in 2033-34, Neighborhood + Citywide Programs takes 946. We put kids hurt at full effect at 0.4 a year, range 0 to 1, our estimate: kids still driven ride about 13.6% fewer car miles, and new walkers carry walking's higher risk. In 2030-31 it's 52% of that.
• Why: a shorter trip makes walking and biking quicker against the car.
• Evidence: Stanford's average elementary distance in SFUSD's Study Session #2: 1.40 miles today and 1.06 under Neighborhood + Citywide Programs, at full effect.
• Cost: none in the tool.

Kids Hurt

263 of the 427 kids hurt on SF weekdays were within a quarter mile of a school's grounds, and 213 of those on a main street.

That's about 107 kids aged 5 to 18 a year. 95 were inside the 500 ft school zone, where the 15/20 mph limits apply.

Kids aged 5 to 18 hurt in traffic on weekdays, 6 am to 6 pm, school years 2021-22 to 2024-25. Distance runs from the crash to the nearest grounds of the tool's 113 schools, measured from the edge of the grounds as California's school-zone law measures it (Vehicle Code §22352). Main means either street is tertiary or above in OpenStreetMap. 4 with no location are left out. Source: Traffic Crashes Resulting in Injury: Victims Involved, DataSF, read 2026-10-18.

246 of the 427 kids hurt on SF weekdays were riding in a car.

A kid who leaves a car for a school bus or Muni changes their own risk as well as the traffic around them, so the shuttle loops and free Muni cut injuries beyond the cars they remove.

The same kids. A car includes drivers aged 16 to 18. Other includes scooters and motorcycles. Source: Traffic Crashes Resulting in Injury: Victims Involved, DataSF, read 2026-10-18.

Near schools, kids walking and biking are hurt about 3 times as often in bell hours as at other daytime hours.

The city's crashes peak in the afternoon anyway, and for kids walking and biking, the bell hours still run 2.1 times what that clock predicts. That's why the speed limits and walking upgrades act near schools.

Crashes hurting a child walking or biking within 600 feet of a school's grounds, per 100,000 hours, at 207 schools on school days 2022-23 to 2025-26. Bell hours run from 45 minutes before the start bell to 15 after, and from 15 minutes before the end bell to 45 after, at each school's own SFUSD bell times (8:15 am and 3:15 pm where SFUSD lists none). Other daytime is 6 am to 8 pm outside them, and bell hours with no school are the same clock times on weekdays when school is out. A crash near two schools counts at both. Source: Traffic Crashes Resulting in Injury and its victims, DataSF.

The maps' heat layer, off until you turn it on, shows kids 5 to 18 hurt within a quarter mile of a school's grounds on weekdays, 7 to 9 am and 2 to 5 pm, August to June, school years 2005-06 to 2024-25, one point per kid. The maps outline each school's 500 ft zone round its grounds. Source: Traffic Crashes Resulting in Injury: Victims Involved, DataSF, read 2026-10-18.

The model starts from 427 kids aged 5 to 18 hurt over four years on weekdays, 6 am to 6 pm, about 107 a year, from DataSF's crash data. Fewer cars cut those hurt within a quarter mile of a school's grounds, 263 of the 427, with the square root of traffic (Elvik and Bjørnskau 2017), with school cars at 0%, 5% or 25% of the traffic at the bell on residential streets. The low is 0 because, counting every injury crash and not only kids walking and biking, the bells show no excess: within 600 feet of a school's grounds, injury crashes at bell times were 290 against the 288 the city's clock predicts, a ratio of 1.01 (0.89 to 1.12), from our check of DataSF's crash data. The high of 25% is the most that check allows [inference — mine]. Main streets carry far more through traffic, so there we put school cars at a tenth of that, our judgment. 213 of the 263 were hurt on a main street. School cars are on the street only at the bells, which held 65 of 243 child walking and biking crashes within 600 feet of the grounds on school days, so the cut applies to that share. A kid leaving a car changes their own risk too. A school bus trip carries about a fifth of a car trip's risk (TRB 2002). Per km, a child walking is hurt 4.6 times as often as one in a car and a child biking 22 times (Lavoie et al. 2014), so we count a walking switcher at 1.5 times the car trip's risk and a biking one at 3 times, or 2 times on an e-bike voucher. A cycle track carries 0.11 of the risk of a major street with parked cars and no bike lane, and a local street 0.51 (Teschke et al. 2012), so a kid biking on a new protected bikeway counts at a third of the car trip's risk and on a calmed or shared street at 1.5 times. A bikeway set's riders split as its miles do. That leaves out the protected bikeways their trips already use, so it may count them riskier than they are, our judgment. We take our ordinary-street bike risk as that major street's [inference — mine]. The car passengers behind that risk include private school kids, so we scale them to SFUSD and charter students, 51,771 of SF's 80,919. We count injuries for the whole city, not school by school.

What the Changes Cost

This table gives each change's rate and where it comes from. A change's total adds the rate up at every school it reaches.

Staff counts assume 2 people a school for School Street Activation, one shuttle driver for every 50 riders and half a person for carpool matching. We don't estimate street design and construction staff.

Set to Match: Car Share by Grade

Driving peaks in elementary school and falls through high school, from 56% in 5th grade to 33% in 9th.

These are inputs the model is set to match, and it matches every grade. The 2024-25 Travel Tally measured four grades. We draw straight lines between them, hold TK at kindergarten's value, and hold 10th to 12th at 9th grade's.

Solid dots are the grades the Tally measured. Hollow dots are filled in between them. Single-family car share of SFUSD students, SF Safe Routes to School, 2024-25 school year.

Tested: Car Share by Distance

The model isn't tuned to these, so this is a real test, and it misses at both ends. Under a mile it has 10 points fewer K-5 kids driven than the 2016 Child Transportation Survey (p. 10), and over 2 miles 18 points more. So it likely understates how many nearby families drive, the ones walking changes reach, and overstates the far ones the shuttle loops reach. The survey asked K-5 parents how their child got to school and how far they live, and doesn't say how it measured distance. We use the walking distance.

Distance from home	Driven, 2016 survey	Driven, model	Model's K-5 students at this distance
Under 1 mile	46%	36%	45%
1 to 2 miles	61%	61%	21%
Over 2 miles	59%	77%	34%

Placed: Where Students Live

SFUSD's dashboard counts each school's students by home ZIP code. We matched 105 of its 116 school listings, a K-8 counting as two. For those, 78% of the model's students live in the same ZIPs as the dashboard's, and 3.5 of each school's top 5 ZIPs match. That's 44,633 students, fewer than our 51,769, since charter schools aren't on the dashboard and it hides ZIPs with under five students. At the 58 schools with an attendance area, a median 37% of students live in it, from 10% to 100%. These counts helped set the model, so this is a fit rather than a test.

The Maps and Your Trip

Each circle is a home block, sized by its students. The first map shows how students get to school, and the second the share with a safe trip. The mode map adds each change's own effect to today's, so it's close but not exact where changes overlap, while the figures beside it are exact. Your Trip routes one address on the same network and rules, in your browser, and Google sees only what you type in the box.

What's Left Out

• Aftercare. We count one trip a student, and 46% of K-5 families used aftercare daily in 2016 (Child Transportation Survey, p. 3).
• Real Muni times. Until each trip is timed from SFMTA's schedule, a Muni trip is 10 minutes to walk to the stop and wait, then the trip's length at 8 mph.
• 100 students, 0.2%, on blocks that send almost none to a school, which is why the tool counts 51,769 of the 51,869 enrolled.