1. The driver’s real objective
The naive model is:
“Take any fare above cost.”
That is wrong.
The better model is:
The driver maximizes expected daily net income, subject to fatigue, risk, time loss, and a cash target.
So the driver is not evaluating a ride in isolation. He is evaluating:
- money from this ride
- time consumed by this ride
- probability of getting the next ride after this one
- risk of dead mileage after drop-off
- mental/physical hassle
- how close he already is to his daily target
That means every passenger request is really a portfolio decision, not just a yes/no to immediate revenue.
2. The core unit: value of one ride request
For any offered trip , define:
Immediate economics
- = fare earned from the trip
- = direct operating cost of the trip
(fuel, wear and tear, tiny incremental costs) - = time spent on the trip
(pickup + driving + traffic + unloading/waiting)
Post-trip position
- = expected repositioning cost after drop-off
(empty return distance, wasted fuel, time to next fare) - = expected quality of destination for the next fare
(station good, dead-end colony bad)
Friction / hassle
- = non-monetary hassle cost
(difficult passenger, luggage, rain, bad roads, police zone, dispute risk)
Then the expected net value of accepting ride is:
V_i = F_i - C_i - R_i - H_i + Q_i
But that still misses the time value.
So divide by total expected time commitment:
\Pi_i = \frac{F_i - C_i - R_i - H_i + Q_i}{T_i}
This is the driver’s effective expected earnings per minute from taking that ride.
That is the real decision variable.
3. Acceptance rule
The driver accepts a ride if:
\Pi_i \geq \theta
Where is the driver’s reservation earnings rate: the minimum expected earnings per minute needed to make the ride worth it.
This threshold is not fixed. It changes during the day.
4. What determines the threshold
A. Daily income target
Let:
- = target income for the day
- = income already earned by time
If the driver is far below target, he may lower his threshold to keep cash flowing.
If he is close to target, he becomes choosier.
So:
\theta_t = \theta_0 + a(\text{fatigue}) + b(\text{time scarcity}) + c(\text{distance from target, with sign})
A more intuitive form:
- early in the day: moderate selectivity
- midday slump: take more rides if demand is weak
- near target: reject annoying or low-yield rides
- very bad day: threshold may fall sharply because liquidity matters
This helps explain why the same driver may accept a ride at 11 a.m. and reject it at 8 p.m.
B. Fatigue
As the day goes on:
- willingness to tolerate hassle falls
- willingness to drive to awkward areas falls
- the subjective cost of traffic rises
So fatigue raises and often raises .
C. Demand conditions
If demand is strong, the opportunity cost of a bad ride is high.
If demand is weak, waiting is dangerous, so even mediocre rides become acceptable.
So the driver compares:
\Pi_i \text{ vs expected value of waiting}
Call the value of waiting .
Then the real rule is:
\text{Accept ride if } \Pi_i \geq W_t
This is even better than the fixed-threshold version.
5. The value of waiting
Drivers are constantly making this comparison:
Option 1: take current ride
earn something now, but get tied up
Option 2: wait
hope for a better passenger very soon
Let:
- = probability of getting another fare soon
- = expected value of the next fare
- = waiting time cost
Then:
W_t = p \cdot E[\Pi_{next}] - w
If the current passenger offers less than this, the driver rejects.
This explains several common behaviors.
6. Why short rides get refused
A short ride often has:
- low fare
- almost fixed boarding/search time
- possible poor drop location
- high chance of empty repositioning
So even if the passenger thinks:
“It’s nearby, easy money.”
The driver may calculate:
“This blocks me for 12 minutes, pays too little, and leaves me in a bad spot.”
In model terms:
\Pi_{short} < W_t
So refusal is rational.
7. Why destination matters so much
Destination affects three things:
- future passenger density
- empty return probability
- traffic escape difficulty
So a ride to a railway station, market, hospital, or main road may have high .
A ride deep into an interior area may have low or negative .
That means two identical fares are not equal.
Same fare, different future value:
- Trip A to station:
- Trip B to dead-end colony:
So drivers are not just pricing the ride.
They are pricing the next 30 minutes of their day.
8. Why bargaining happens
Suppose the official metered fare gives:
\Pi_i < W_t
Then the driver has two choices:
- reject
- quote a higher fare so that the trip crosses the acceptance threshold
Let required fare be . Then:
F_i^* = \theta T_i + C_i + R_i + H_i - Q_i
This is the minimum acceptable fare.
If the official fare is below that, bargaining appears.
So bargaining is not random greed. It is often an attempt to push the ride up to the driver’s internal reservation price.
9. Why drivers cluster at certain locations
Drivers prefer locations where:
- , probability of next fare, is high
- expected fares are decent
- trip distribution is favorable
- idle time is socially tolerable
- police/enforcement risk is manageable
So stands near stations, malls, hospitals, schools, and busy junctions become rational waiting nodes.
These are not just physical locations. They are high-option-value assets in the driver’s daily strategy.
10. Ownership model changes behavior
There are at least three broad driver types.
A. Owner-driver
Pays capital cost/EMI, but keeps full upside.
Behavior:
- more sensitive to maintenance
- more strategic long-term
- may tolerate lower fares if vehicle utilization matters
B. Rental driver
Pays daily fixed rental to owner.
Behavior:
- strong pressure to first “clear the rental”
- early-day desperation can be high
- after clearing rental, behavior may change sharply
C. Fleet/app-attached driver
Faces platform commissions, digital matching, less bargaining freedom.
Behavior:
- lower search cost
- weaker control over passenger selection
- more algorithm-driven utilization
So the same road behavior can only be understood after knowing the contract structure.
11. Add a simple daily target model
Let daily target be:
Y^* = D + M + S
Where:
- = household cash need for the day
- = vehicle/operating obligations
- = precautionary or aspirational surplus
The driver stops or becomes selective once expected additional work has low marginal utility.
Utility can be written as:
U(Y, E) = u(Y) - v(E)
Where:
- = income earned
- = effort/fatigue
And has diminishing marginal utility after subsistence and daily obligations are met.
So after target income is reached:
- one more awkward trip is much less attractive
- refusal rates rise
- preference for close/easy trips may rise
This is one reason evening behavior may feel inconsistent to passengers.
12. Behavioral overlays
The pure economic model is good, but incomplete. Real drivers are not calculators. They use shortcuts.
Heuristic 1: “Bad destination”
Certain neighborhoods acquire a reputation. Drivers may reject quickly without precise calculation.
Heuristic 2: “Short ride not worth it”
A rough rule evolved from repeated low-value outcomes.
Heuristic 3: “Rain surcharge mentality”
Even where illegal, drivers know rain raises both demand and hassle.
Heuristic 4: “Last ride home”
Near end of shift, drivers want rides that align with their own route.
So actual behavior is:
\text{Decision} = \text{economic threshold} + \text{habit} + \text{local norms} + \text{mood}
13. A compact decision tree
When a passenger approaches, the driver implicitly asks:
Step 1: Where are you going?
This estimates , , and .
Step 2: How much will this pay?
This gives .
Step 3: How annoying will this be?
This estimates .
Step 4: What happens if I wait instead?
This gives .
Step 5: Where am I relative to my daily target?
This changes threshold .
Then:
- if current ride beats waiting, accept
- if not, reject
- if close, bargain
That is the whole machine.
14. A very usable formula
For workshop or analytical purposes, use this:
A_i = F_i - (C_i + R_i + H_i) + Q_i - \theta T_i
Where:
- if , accept
- if , bargain
- if , reject
This is a very clean operational model.
15. What this model explains well
It explains:
- refusal of short rides
- destination discrimination
- preference for stands and transit nodes
- bargaining above meter
- time-of-day variation
- why some drivers are highly selective
- why drivers behave differently after hitting target income
- why passengers misread “easy trip” economics
16. What the model does not fully explain
It does not fully capture:
- union or stand politics
- police harassment or bribe risk
- city-specific permit distortions
- social identity and class conflict
- app-platform algorithmic manipulation
- extreme liquidity stress at household level
These can be added as external parameters.
For example:
- enforcement risk =
- political/stand control =
- app commission =
Then:
A_i = F_i - (C_i + R_i + H_i + P_i + K_i) + Q_i - \theta T_i
17. The deepest intuition
A rickshaw driver is not selling rides.
He is selling time slots of vehicle capacity under uncertainty.
Each passenger request competes with:
- the possibility of a better next passenger
- the possibility of a worse destination
- fatigue
- daily cash needs
- search time
- urban geography
So the right mental model is not:
“Why won’t he take my fare?”
It is:
“Why does this ride lose against his next-best alternative?”
Once you frame it that way, most driver behavior becomes intelligible.
18. One-line version
A rickshaw driver accepts a trip when its expected net earnings per minute, adjusted for destination quality and hassle, exceed the expected value of waiting for the next fare.
Posts
No comments:
Post a Comment