Studies were eligible if they empirically evaluated an income-based or means-tested fare discount, a reduced-fare program targeted at low-income riders, or a fare capping policy, and reported at least one outcome on ridership, travel frequency, access to opportunities, or the affordability of travel for low-income riders. This criterion corresponds to the second and third research questions. To answer the first research question, descriptive syntheses and evaluations of program design were retained when they documented the eligibility thresholds, discount levels and capping structures of implemented programs, even when they reported no behavioral estimate.

Eligible designs were randomized and quasi-experimental studies, observational and modelling studies, cross-sectional surveys, and mixed-methods evaluations; reviews, editorials and commentaries were excluded. Records published from 2013 to 2026 in English or Spanish were considered, matching the bibliographic window of the commission. Universal fare-free policies and flat fare reductions that were not tied to income were treated as a distinct intervention and were excluded from the core set, although they were retained for comparison in the discussion because they are the most frequent alternative to means-tested discounts.

No restriction was placed on geography, because the aim was to capture evaluations from both high-income systems and cities of the Global South where the subsidy context differs. Expressed as a PICO question, the population was low-income public transport riders, the intervention was an income-based discount or fare cap, the comparison was the regular fare or the pre-policy period, and the outcomes were ridership, travel frequency, access and affordability.

The Scopus strategy combined controlled field terms with a year filter. One core query read TITLE-ABS-KEY(("reduced fare" OR "discounted fare" OR "fare discount" OR "concessionary fare" OR "income-based fare" OR "means-tested fare") AND ("low-income" OR "low income" OR "equity" OR "affordability")) AND PUBYEAR > 2012 AND PUBYEAR < 2027, and returned 28 records. A second query captured fare capping with TITLE-ABS-KEY("fare capping" OR "fare cap" OR "capped fare") AND PUBYEAR > 2012 AND PUBYEAR < 2027, and returned 7 records.

A third query combined transit fare terms with affordability, equity, ridership and access, TITLE-ABS-KEY(("transit fare" OR "public transport fare" OR "transit pass") AND ("low-income" OR "affordability" OR "equity") AND ("ridership" OR "access" OR "travel")) AND PUBYEAR > 2012 AND PUBYEAR < 2027, and returned 49 records. Equivalent free-text queries were run in OpenAlex, Crossref and Semantic Scholar, which returned 100, 100 and 47 records respectively. Targeted queries for named programs, including Fair Fares, ORCA LIFT, Muni Lifeline and the Low Income Fare is Easy program, supplemented the core searches and returned a further set of candidate records.

Reference lists of the included studies and of the two recent reviews were also inspected, which added records that the database queries had missed.( 5,6)

For each included study one reviewer extracted the following items: study design, country, sample and population, the fare instrument and its eligibility or capping structure, the outcome construct, the principal result with its reported estimate, and the risk-of-bias judgment, and these fields were checked against the source text.( 10)

Risk of bias was assessed with ROBINS-I for the non-randomized studies, and the randomized trial was appraised against the same domains of confounding, selection, classification of interventions, deviations from intended interventions, missing data, measurement of outcomes and selective reporting.( 11)

The extraction form captured the unit of analysis, the fare instrument, the eligibility threshold or cap level where reported, the outcome definition, and the reported estimate with its measure of uncertainty when available. For the behavioral question P2, the effect measures of interest were changes in trips, card taps or ridership, price elasticities, and reported trip frequency.

For P3, the constructs of interest were access to jobs and services, perceived accessibility, fare burden expressed as a share of income, and the distribution of subsidy benefits across income groups. Because the studies reported these constructs in different units, no common effect size could be computed, and the synthesis relied on the direction and magnitude of each estimate as reported by its authors.

Effect measures were not comparable across studies because designs, fare instruments and outcome definitions differed, so results were synthesized narratively and organized by research question rather than pooled in a meta-analysis. For the first question, program characteristics were tabulated and described. For the second question, the direction and magnitude of reported ridership effects were grouped by design strength, with causal designs given more interpretive weight than descriptive or modelling work. For the third question, reported access and affordability outcomes were grouped by construct. Heterogeneity was explored narratively by instrument type, population and time horizon, and no statistical pooling was undertaken because the studies did not report compatible effect sizes.

Where a study reported multiple outcomes, the primary behavioral outcome was used. Vote counting was avoided in favor of a structured narrative, because simple counts of positive and negative findings would obscure the large differences in design strength; instead, causal designs were reported first and descriptive and modelling results were treated as corroborating or contextual rather than as direct evidence of effect.

The narrative synthesis therefore reports the direction of effect for each study rather than a pooled estimate, and the risk-of-bias ratings are used to order the presentation so that the most informative designs lead each subsection. The certainty of the evidence was not formally graded, but the risk-of-bias ratings were used to qualify the conclusions drawn from each study.

The searches identified 431 records, and after 92 duplicates were removed 339 records were screened by title and abstract. Of these, 303 were excluded at the title and abstract stage, 36 reports were sought for retrieval, and 21 could not be retrieved through open access because the publisher full text was behind a paywall without a legitimate open copy. The remaining 15 reports were assessed for eligibility and 15 studies were included in the review (Figure 1).

The title and abstract exclusions were dominated by studies of fare payment technology, fare evasion, and universal fare-free policies, together with records on unrelated topics that shared only the word fare. Scopus contributed 91 of the identified records, OpenAlex contributed 193, Crossref 100 and Semantic Scholar 47. The 21 reports that could not be retrieved were sought but were available only through paywalled publisher platforms.

Regarding P1, the included studies evaluated two broad families of instruments. Six studies examined income-based discounts or means-tested subsidies. The Bogota study evaluated a subsidy equal to 32% of the regular fare delivered to low-income users through a personalized travel card, and used the discontinuity in eligibility to isolate the subsidy effect from other differences between users.( 12)

The King County study randomized six months of free transit among low-income individuals and tracked card taps across the county fleet, which allowed the authors to separate genuine travel increases from changes in how fares were paid.( 13)

The reduced-fare synthesis documented that 17 of the 50 largest United States agencies operated low-income programs, that nine of the fourteen self-administered programs offered a 50% discount, and that eligibility thresholds ranged from 125% to 200% of the federal poverty level.( 14)

The Portland evaluation described a transportation wallet that combined a USD 308 prepaid card with free bike share membership and discounted services, delivered to residents of affordable housing.( 15)

The Christchurch study examined a half-price fare policy introduced in April 2022 as a national transport relief package, using a survey to capture experiences of the reduced fare.( 16)

The Madrid study quantified the distributional incidence of fare discounts and subsidies across income groups, connecting fare design to who actually received the benefit.( 17)

Two studies documented fare capping. A synthesis of the 101 largest United States agencies found at least 21 with capping policies, of which 20 used daily caps, four used weekly caps and 14 used monthly caps, and it described nested caps applied within reduced-fare programs as an emerging design.( 9)

A companion impact analysis used mobile ticket data to estimate how a capping program changed trips and fares paid, which moved the evidence beyond design description toward measured behavior.( 18)

The remaining studies examined fare structures whose equity consequences are directly relevant to low-income riders, including distance-based and flat fare schemes in Vancouver, Utah and Stockholm, rail subsidy incidence in San Francisco and Atlanta, a methodological audit of fare inequity in Costa Rica, a fare cut in Belo Horizonte, and a policy evaluation of targeted low-income fare programs in the United States.( 19-25)

Overall, the eligible studies described a mature set of instruments rather than a single homogeneous policy, and the differences between them, in eligibility, discount depth and cap horizon, are precisely the dimensions along which the behavioral results vary.

The variation in design is not accidental: means-tested programs dominate in systems with mature smart-card infrastructure, while modelling studies of fare geometry remain common where an agency is considering a structural reform of the fare table itself.

Geographically, the studies concentrated in the Americas, with eight from the United States, three from Latin America and one from Canada, while Europe contributed two and New Zealand one; no included study came from Africa or Asia, which limits the transferability of the findings to those contexts. Temporally, the included studies spanned 2014 to 2026, with the fare capping and quasi-experimental work appearing only after 2021. The designs were predominantly observational, with two quasi-experimental studies, one randomized trial and the remainder descriptive, survey or modelling work. Fifteen studies were included.

With respect to P2, the strongest designs reported positive effects. The Bogota regression discontinuity study quantified that the 32% subsidy significantly increased the total number of monthly public transport trips, with a more pronounced effect among economically active individuals and an effect that declined over the study window, a pattern the authors interpreted as novelty decay.( 12)

The King County randomized trial showed that free fares sharply increased transit card use, although part of the increase corresponded to payment method substitution and little effect persisted after the subsidy ended.( 13)

The fare capping impact analysis indicated that caps redistributed the fare burden toward frequent riders and could raise ridership among price-sensitive groups while moderating revenue losses.( 18)

The Portland wallet survey reported that participants made more trips and reached destinations they otherwise could not have reached.( 15)

The Belo Horizonte event study quantified a short-run price elasticity near -0.20 overall, with bus rapid transit responding at -0.42 and diametrical lines essentially inelastic at -0.03, and the parallel pre-trends assumption was supported by a pre-treatment test.( 24)

The Vancouver modelling study predicted that distance-based pricing would increase trips and reduce fares for low-income, captive and transit-dependent riders.( 19)

Taken together, every study that estimated a behavioral response reported an increase in trips, card use or predicted demand for low-income or price-sensitive riders, although the magnitude ranged from a near-zero elasticity on some corridors to a substantial subsidy effect in Bogota.

The three designs that identified causal effects each pointed in the same direction, which strengthens confidence in the direction despite the small number of studies, although the absence of confidence intervals in several reports prevents a formal assessment of precision. The elasticity estimate of -0.20 means that a 10% fare reduction was associated with roughly a 2% increase in ridership in the short run, a response that is economically meaningful but far from a proportional transfer of a fare cut into ridership gain.

The contrast between the bus rapid transit response and the near-inelastic diametrical lines also warns against applying a single systemwide elasticity to a policy that will affect corridors unevenly. No included study reported a negative effect of a discount or cap on ridership, which is itself informative given the concern that discounts mainly subsidize trips that would have occurred anyway. Six studies reported effects on ridership or travel frequency.

As for P3, access and affordability outcomes were reported in eight studies. The Christchurch analysis found that half-price fares widened access and reduced financial stress among riders experiencing transport poverty, indicating that even a temporary fare reduction can relieve the household budget constraint.( 16)

The Bogota subsidy was linked to improved access to local labour markets and recreational activities, which the authors framed as a pathway to well-being.( 12)

The Portland program increased access and enabled trips that participants could not otherwise make.( 15)

The reduced-fare synthesis estimated that participants at the eligibility threshold spent between 2% and 6% of annual income on transit, with much higher shares for very low income riders, which quantifies the residual burden that even discounted fares leave in place.( 14) The Madrid incidence analysis showed that existing subsidies were not always concentrated on lower-income riders and that fare adjustments could improve vertical equity.( 17)

The Costa Rica study identified routes whose observed fares exceeded expected levels and should be audited first because they charged higher operational costs into the fare, increasing inequity among dependent riders.( 23)

The rail cost-recovery analysis found that subsidy patterns were marginally progressive and correlated with disadvantaged socioeconomic groups.(22) The Wasatch Front analysis concluded that distance-based fares benefited low-income, elderly and non-white riders overall, although the effect was geographically uneven.( 20)

Across these studies the direction was consistent, in that lower fares for low-income riders were associated with wider access, a lower fare burden or a more progressive distribution of subsidy, and no included study reported a deterioration in access attributable to the discount itself. Eight studies reported effects on access or affordability.

Risk of bias was moderate in the two quasi-experimental studies and the randomized trial, which controlled the principal confounders but could not rule out substitution between fare instruments or short follow-up.( 12,13,24)

The descriptive, survey and modelling studies were rated at high risk of bias for the causal questions of P2 and P3, because most relied on self-reported travel, static demand models or cross-sectional comparisons, and the abstract-only records carried additional uncertainty about selective reporting.( 9,14-16)

For the ten studies assessed from abstracts, the domains of deviations from intended interventions and selective reporting could not be judged at all, and this incompleteness is itself a source of uncertainty that the ratings reflect. This distribution means the direction of the findings is more credible than any single magnitude.

Eligibility and extraction for ten of the fifteen included studies relied on structured abstracts because publisher full texts were not openly accessible, which may have introduced selective reporting of results and prevented a complete risk-of-bias judgment; future updates should use institutional subscriptions to retrieve these records.( 10)

For practice, the evidence supports targeting rather than flat reductions, capping as a complement for frequent riders, and outcome monitoring that separates new travel from payment substitution. Agencies planning an income-based discount should invest in enrollment partnerships with social service organizations, which the included syntheses identify as the difference between programs that reach low-income riders and programs that remain nominal.

Agencies planning a cap should monitor whether savings accrue to frequent low-income riders or to higher-income commuters who would have purchased a pass anyway, because the incidence of a cap depends on the distribution of trip frequencies in the rider population. For research, the field would benefit from pre-registered quasi-experimental evaluations of the same program over multiple years, from standardized access outcomes measured before and after fare changes, and from reporting effect estimates with confidence intervals so that future work can be pooled. What remains open is whether the initial ridership gains documented here persist once novelty fades and whether the access benefits translate into durable gains in employment, health and social participation.