Across the developing world, hundreds of millions of commuters rely on private minibuses as their main mode of transportation. A defining feature of these networks is their organizational structure: minibuses organize into associations that control specific subsets of routes. This is true of "trotros" in Accra, "matatus" in Nairobi, and “minibus taxis” in Kampala. While these associations may serve important functions – such as coordinating prices and managing routes – they also fragment the city into distinct territories.
This fragmentation creates a fundamental tension. Cities are inherently interconnected, and the demand for mobility is not similarly segmented. In Johannesburg, 25% of all commuter trips originate in one association’s territory and end in another’s. Operating these “between territory” routes requires coordination among associations – they must agree on how to share revenues, allocate vehicles, and set prices.
In my job market paper, we study how market segmentation and the resulting need for coordination affects the provision of public transportation and consequently, commuter mobility. In Johannesburg, minibuses serve as the primary mode of transportation for 40% of households. Households dependent on minibuses spend approximately two hours each day commuting to and from work. The importance of minibuses to daily mobility means that even small inefficiencies in service provision can impose substantial costs on residents.
Quantifying Coordination Frictions
To understand how territorial boundaries affect service provision, we analyze rich GPS data from a major South African minibus financier, covering over 19,000 minibuses and 40 million trips over a ten-month study period. We complement the supply-side data with anonymized smartphone GPS pings, which we translate into over 9 million commuter trips within Johannesburg.
In theory, associations could achieve efficient outcomes through coordination. In practice, enforcement challenges and lack of transparency might lead to inefficiently low or rigid supply on between-territory routes. In Johannesburg, associations form “joint ventures” to operate on these routes, creating informal agreements on revenue sharing and operations. Agreements either cap the number of minibuses each association can operate or restrict associations to picking up passengers only in their own territory. These agreements are easy to enforce and implement, but likely result in inefficient equilibrium outcomes.
To quantify how coordination between associations affects service provision, we compare supply on between-territory routes to supply on routes that require no inter-association coordination ("within-territory routes"). Through two main empirical exercises, we analyze how associations respond to supply and demand shocks, and how these responses differ by route.
How do associations react when their fleet size is reduced?
Associations assign minibuses to routes under their control, usually on a weekly rotation. When a minibus becomes inoperable, associations must decide how to allocate this loss amongst its routes. These decisions serve as a revealed preference measure of marginal profitability, as associations reduce service on their least valuable routes when forced to make cuts. To empirically estimate these fleet allocation decisions, we employ a differences-in-differences design that exploits quasi-exogenous shocks to fleet size from minibus accidents and repossessions.
The results reveal a stark asymmetry in how associations prioritize routes. When an association’s fleet decreases by 1%, they reduce the number of minibuses operating within territory by approximately 1% as well. In contrast, the number of minibuses allocated to between-territory routes decreases by only 0.3%.
Our interpretation of this asymmetric response is that profits are higher at the margin on between-territory routes, suggesting that there is inefficiently low service in equilibrium on these routes. This is supported by the fact that we observe higher fares on average on between territory routes compared to within-territory routes after controlling for distance.
How responsive is minibus supply to increases in demand?
We next examine how the industry responds to demand shifts, and how this varies by route type. The need for coordination on between-territory routes may make supply more rigid on these routes. To test this, we estimate the supply elasticity of minibuses using cyclical changes in mobility. Mobility in South Africa exhibits strong monthly cycles tied to paydays: using Google mobility data, we document that residents take 32% more trips to retail and recreation destinations on the 25th of each month – when most workers get paid – compared to the 24th.
We construct an instrument for mobility demand by interacting two indicators: whether the date falls within the "end of the month" period, and whether a route is "recreational" (defined by having a large mall at either its origin or destination). This generates a well-powered first stage – the instrument increases commuter trips, as measured by our smartphone GPS data, by 4% (SE 0.7).
Using this instrument, we estimate the impact of an increase in demand on the supply of minibuses. Within territory, associations adjust supply almost one-to-one with demand, with an elasticity of 0.93 (SE 0.02). However, on between-territory routes, the supply response is significantly muted, with an elasticity of just 0.40 (SE 0.05).
Our granular data allows us to trace how associations manage these demand fluctuations. Within territory, associations frictionlessly reallocate their fleet, reducing service on non-recreational routes by 3.4% while increasing service on recreational routes by 2.4%. In contrast, on between-territory routes, associations struggle to increase service – showing only a statistically insignificant 1% increase in supply.
The fact that these frictions persist even for predictable demand cycles highlights the depth of the coordination challenge. Even when associations can anticipate demand changes well in advance, the need for coordination creates significant frictions in service adjustment.
Costs to Commuters
We quantify the costs of segmentation to minibus passengers in one dimension – wait times. Specifically, we estimate how much wait times would reduce in the short-term if the supply elasticity on between-territory routes matched that on within-territory routes. We approximate wait times using the time between minibus departures (headway); assuming uniform passenger arrivals, average wait time equals half the headway.
The economic costs are significant. On recreational between-territory routes alone, passengers collectively lose 988,142 minutes each day due to reduced service flexibility. This translates to approximately four minutes of extra wait time on each trip (~10% of average trip duration). Using conservative time valuations (50-100% of average minibus user wages), this yields daily costs between 46,030 USD and 173,313 USD.
These figures likely understate the true economic impact, as our calculations only capture partial-equilibrium time costs to existing passengers. We do not account for potential longer-term effects, such as increased ridership that might result from more flexible service. Since minibuses typically only depart when full, such ridership increases would further reduce wait times, magnifying the economic benefits of improved coordination.
Policy Implications
In recent years, governments have shifted away from trying to replace minibus networks entirely, and are now focused on engaging with the industry. Current reforms include operator capacity building, with training programs covering scheduling, and vehicle management. For instance, the Johannesburg municipality recently funded a capacity-building program for 100 operators from 11 taxi associations.
Building on this momentum, our findings suggest targeted policy interventions could substantially improve service on between-territory routes with inefficiently low supply. Importantly, associations also have incentives to address these coordination frictions. There is precedent for successful intervention in route-level services; researchers in Kampala, Uganda subsidized a minibus association to operate on a new route. Our analysis provides a clear framework for identifying where such interventions would be most valuable. Policymakers could consider a range of tools, from targeted subsidies to encourage expanded service, to more direct regulatory measures like restructuring how between-territory routes are assigned and operated. For instance, between-territory routes could be assigned to single associations on a rotating basis, eliminating the need for coordination on service quantity decisions.
Join the Conversation