1 1. Analytical Framework for Identifying Ride Provision

The 2017 Yamagata Regional Person Trip Survey provides detailed daily travel records for all members of sampled households on both weekdays and holidays. However, as with comparable household travel surveys in Japan, although each respondent’s trip record indicates whether they drove or rode as a passenger, it does not identify who was driving for passenger trips, nor who, if anyone, was riding along for driver trips. As a result, it is not immediately evident from the dataset who provided ride provision—the central focus of this study—to elderly household members. To address this limitation, I match driver and passenger records on the basis of spatiotemporal consistency, identifying pairs of trips made simultaneously in the same vehicle.

The matched trip pairs, however, encompass not only ride provision but also joint travel—trips in which the driver and passenger share a common destination and purpose. For example, when a driver and passenger go shopping together, the trip is classified as joint travel and does not constitute ride provision. Conversely, when a driver takes a passenger to a hospital appointment, the trip qualifies as ride provision. Identifying ride provision therefore requires a two-step process: first, identifying pairs of shared rides within the same household (Section 2), and then excluding those that qualify as joint travel (Section 3). Unless otherwise noted, passengers throughout this process are defined as household members aged 75 or older.

2 2. Identification of Driver–Passenger Trip Pairs

2.1 2.1 Identification Framework

The 2017 Yamagata Regional Person Trip Survey records each household member’s trips individually but does not indicate who shared a vehicle with whom. Driver–passenger pairs are therefore identified indirectly, by matching trip records within the same household on the basis of two criteria: (1) the passenger’s modal structure—whether the trip involves car travel alone or includes a transfer to or from public transit—and (2) the spatial and temporal consistency between the passenger’s trip and the driver’s recorded origin and destination.

2.2 2.2 Classification of Driver and Passenger Patterns

Driver and passenger travel behavior is classified into distinct types before the matching algorithm is designed. Drivers may pick up or drop off passengers at intermediate points along their route; when such boarding or alighting does not coincide with the driver’s recorded origin or destination, it goes unrecorded in the data. For passengers, boarding or alighting a car at an intermediate point during a trip is treated as a multimodal case involving a transfer between modes—for example, being dropped off by car at a train station and continuing by rail to the final destination.

Under these assumptions, both driver and passenger behavior can each be classified into four types, as illustrated in Table 2.1.

Table 2.1: Driver and Passenger Pattern Types
Type Name Travel Pattern Description
Driver patterns
D1 Dedicated driver Travels from the same origin to the same destination as the passenger
D2 En-route pick-up driver Travels alone first, then picks up the passenger at an intermediate stop
D3 En-route drop-off driver Drops off the passenger at an intermediate stop, then continues alone
D4 En-route two-stop driver Picks up and drops off the passenger at two separate intermediate stops
Passenger patterns
P1 Unimodal passenger Travels entirely by car with no mode transfer
P2 Pick-up passenger Uses public transit first, then transfers to a car at a station or bus stop
P3 Drop-off passenger Travels by car first, then transfers to public transit at a station or bus stop
P4 Transfer passenger Uses public transit, transfers to a car, then transfers back to public transit

Legend: → Independent segment;  ⎯ Shared ride segment (by car);  ○ Recorded origin/destination;  ● Unrecorded stop (driver only);  ◇ Transfer point (passenger only)

2.3 2.3 Definition of Target Trips

Driver trips and passenger trips are extracted from the person trip survey microdata according to the following criteria.

  • Driver trips: The primary mode is a privately owned car, the respondent was the driver, and the vehicle carried two or more occupants. Drivers may be of any age.
  • P1 passenger trips: The respondent traveled as a car passenger driven by a family member, is aged 75 or older, used at least one car mode, and used no public transit mode. Combinations with walking or moped are permitted.
  • P2 passenger trips: Same basic conditions as P1 (car passenger, driven by a family member, aged 75 or older), with the additional requirement that the final segment of the trip was made by car and at least one public transit mode was used, representing a public-transit-to-car transfer. The zone of the transfer point is identified from the recorded car boarding information.
  • P3 passenger trips: Same basic conditions as P1, with the additional requirement that the first segment of the trip was made by car and at least one subsequent mode was public transit, representing a car-to-public-transit transfer. The zone of the transfer point is identified from the recorded information on where the passenger transfers from car to public transit.

Public transit modes include subway, conventional rail (JR), shinkansen, fixed-route bus, community bus, expressway bus, chartered bus, and demand-responsive transport.

2.4 2.4 Spatiotemporal Matching Criteria

Matching relies on both temporal and spatial criteria. For the temporal dimension, a tolerance of 30 minutes is applied to account for reporting variability among respondents. For the spatial dimension, consistency is assessed using the zone codes of the 2017 Yamagata Regional Person Trip Survey, which are structured in three hierarchical levels—large, medium, and small zones. Matching is performed at the most detailed level (small zone); where the small zone is not recorded, the medium zone is used instead.

2.5 2.5 Matching Algorithm

As shown in Table 2.1, drivers and passengers each have four types, yielding 16 possible D×P combinations. Matching proceeds stepwise from the most to the least stringent conditions, as summarized in Table 2.2. Pairs identified at each step are excluded from subsequent steps to prevent double-counting. When multiple candidate drivers or passengers satisfy the matching criteria, the pair with the smallest difference in departure time (or arrival time) is selected.

Step 1 (D1+P1, two-point match: origin and destination): Driver trips and P1 passenger trips within the same household are matched when both the origin zone and the destination zone are consistent and departure times fall within the tolerance window.

Step 2 (D1+P2, two-point match: transfer point and destination): For P2 passengers whose transfer point zone has been successfully identified, matching is performed when the driver’s origin zone is consistent with the passenger’s transfer point zone, the destination zones are consistent, and arrival times fall within the tolerance window. P2 passengers whose transfer point zone could not be identified are deferred to Step 4.

Step 3 (D1+P3, two-point match: origin and transfer point): For P3 passengers whose transfer point zone has been successfully identified, matching is performed when the origin zones are consistent, the driver’s destination zone is consistent with the passenger’s transfer point zone, and departure times fall within the tolerance window. P3 passengers whose transfer point zone could not be identified are deferred to Step 5.

Step 4 (D2+P1/P2, one-point match: destination only): Unmatched P1 and P2 passenger trips are matched with driver trips based solely on destination zone consistency and arrival time. Inconsistency in origin zones is accepted as a structural feature of the D2 pattern, in which the driver picks up the passenger at an intermediate point.

Step 5 (D3+P1/P3, one-point match: origin only): Unmatched P1 and P3 passenger trips are matched with driver trips based solely on origin zone consistency and departure time. Inconsistency in destination zones is accepted as a structural feature of the D3 pattern, in which the driver drops off the passenger at an intermediate point.

Supplementary steps: To handle specific reporting conventions in the survey, the following supplementary matching procedures are applied.

  • Handling zone discrepancies in consecutive shared rides: When a driver and passenger share multiple consecutive trips, zone records may contain discrepancies or omissions due to reporting variation. Matching is permitted despite zone inconsistency or missing values, provided that shared travel is confirmed in the immediately preceding or following trip and times are consistent.
  • Handling omitted times in home-return trips: Departure and arrival times are frequently omitted for home-return trips. For pairs where the preceding trip has already been matched and the current trip purpose is “return home,” matching is performed even when time records are missing.
  • Handling secondary passengers in multi-occupant driver trips: For driver trips carrying three or more occupants, unmatched passengers who share the same origin zone, destination zone, and times as an already-matched passenger on the same driver trip are added as secondary passengers.

The correspondence between each step and the D×P pattern combinations is shown in Table 2.2.

Table 2.2: D × P Pattern Classification Matrix
P1
P2
P3
P4
D1 Step 1 Step 2 Step 3 Not observed
D2 Step 4 Step 4 Undetectable Not observed
D3 Step 5 Undetectable Step 5 Not observed
D4 Undetectable Undetectable Undetectable Undetectable

As Table 2.2 shows, certain D×P combinations cannot be captured due to structural limitations of the algorithm. D2+P3 and D3+P2 are undetectable because the driver’s intermediate stop is unrecorded and the passenger’s transfer point is also incompletely recorded, leaving neither origin nor destination as a viable matching anchor. D4 is inherently undetectable, as both the boarding and alighting points lie at unrecorded intermediate locations. Finally, the P4 pattern—trips involving a public-transit–car–public-transit sequence—was not observed in this survey. These undetected cases represent a potential source of downward bias in the estimates and are discussed accordingly in the analysis.

3 3. Classification of Ride Provision and Joint Travel

3.1 3.1 Classification Framework

Driver–passenger pairs identified in Section 2 are classified into two categories: joint travel, in which the driver and passenger share a common destination and purpose, and ride provision, in which the driver transports a passenger solely in service of that passenger’s travel needs. The classification draws on the driver and passenger trip patterns identified in Section 2 (D1–D3, P1–P3) and applies a stepwise algorithm that incorporates behavioral context and trip purpose.

3.2 3.2 Assignment of Driver and Passenger Patterns

Driver and passenger patterns (D1–D3 and P1–P3) are assigned to each pair as identified in Section 2. As shown in Table 2.2, the resulting D×P combinations yield seven types: D1+P1, D1+P2, D1+P3, D2+P1, D2+P2, D3+P1, and D3+P2.

3.3 3.3 Classification Algorithm

The algorithm is designed to process D×P combinations in order of classificatory certainty—from those that can be definitively assigned to those requiring behavioral judgment.

Step 1: Definitive Classification of Multimodal Passengers and En-route Drivers

Pairs involving a multimodal passenger (P2 or P3) are classified as ride provision regardless of driver pattern. Boarding or alighting a car at a station or bus stop is itself direct structural evidence of ride provision: the driver serves as the car leg of a journey that the passenger completes partly by transit. These pairs are assigned to RP3 (drop-off at station or bus stop) or RP4 (pick-up at station or bus stop) accordingly.

Pairs in which an en-route driver (D2 or D3) carries a unimodal passenger (P1) are likewise classified as ride provision. A driver who picks up a passenger at an intermediate point along the route (D2+P1) or drops one off mid-route (D3+P1) exhibits a behavioral structure that cannot arise from joint travel toward a shared destination.

Step 2: Dedicated Driver with Unimodal Passenger (D1+P1)

Because a dedicated driver (D1) and a unimodal passenger (P1) share the same recorded origin and destination, a single matched trip alone is insufficient to distinguish ride provision from joint travel. A two-stage procedure is therefore applied.

Step 2-1: Trip Chain Comparison

Joint travel typically involves the driver and passenger traveling together throughout the trip chain—defined here as the sequence of trips made between leaving home and returning home on the survey day. Exploiting this characteristic, the full trip chains of the matched driver and passenger are compared. If the driver and passenger have chains of equal length and if the origin, destination, and trip purpose are consistent across every trip in the chain, the pair is classified as J1 (pure joint travel). Any discrepancy in chain length or in origin, destination, or trip purpose at any point in the chain triggers progression to Step 2-2. Where the destination of the final trip in a chain is not recorded, the respondent’s home zone is substituted.

Step 2-2: Behavioral Classification of Individual Trips

D1+P1 pairs for which the full trip chains do not match are assessed at the level of the matched trip, using the driver’s travel behavior immediately before and after the shared trip. A threshold of 10 minutes is applied: dwell times at or below this threshold are interpreted as brief stops, while those exceeding it indicate activity at the location.

Driver travels alone immediately after the shared trip (drop-off): When the driver travels alone immediately after the matched trip, the structure corresponds to dropping off a passenger and departing—a defining feature of ride provision. If the dwell time at the destination is at or below the threshold, or if the driver records the trip purpose as drop-off/pick-up, the pair is classified as RP1 (dedicated drop-off). If the dwell time exceeds the threshold and the driver and passenger share the same recorded trip purpose, the pair is classified as J2 (partial joint travel), interpreted as shared activity followed by the driver departing independently. In all other cases, the pair is classified as Indeterminate.

Driver travels alone immediately before the shared trip (pick-up): When the driver travels alone immediately before the matched trip and the following trip is made with a passenger (or no following trip is recorded), the structure indicates that the driver traveled to collect the passenger—a defining feature of pick-up ride provision. If the dwell time at the origin is at or below the threshold, or if the driver records the trip purpose as drop-off/pick-up, the pair is classified as RP2 (dedicated pick-up). A further rule captures the common case in which the driver travels alone to collect the passenger and the two return home together: even when both driver and passenger record “return home” as the shared trip purpose, the pair is classified as RP2 if the driver’s immediately preceding trip—made without a passenger—was recorded as drop-off/pick-up. If the dwell time at the origin exceeds the threshold and the driver and passenger share the same trip purpose, J2 is assigned; in all other cases, the pair is classified as Indeterminate.

All other cases (catch-all): This branch covers cases in which the driver is never observed traveling alone before or after the shared trip, as well as cases where preceding or following trip information is missing. A characteristic example is a driver who accompanies a passenger to a hospital, waits during the appointment, and returns home with the passenger: no trip made without a passenger appears in the trip chain, yet the episode constitutes accompaniment—a form of ride provision (RP5). In such cases, if the driver’s trip purpose is recorded as drop-off/pick-up, or if the shared trip is a home-return trip immediately following a driver trip recorded as drop-off/pick-up, the pair is classified as RP5. If the driver and passenger share the same recorded trip purpose, J2 is assigned. In all remaining cases, the pair is classified as Indeterminate.

The resulting pattern types are summarized in Table 3.1. The correspondence between D×P combinations and their classification outcomes is summarized in Table 3.2.

Table 3.1: Ride Provision and Joint Trip Pattern Types
Pattern Driver Passenger Description
Pick-up/Drop-off
RP1: Dedicated drop-off D1 P1 Driver delivers the passenger to the destination and then departs alone
RP2: Dedicated pick-up D1 P1 Driver travels alone to collect the passenger and then delivers the passenger to the destination
RP3: Multimodal drop-off D1, D3 P3 Driver drops the passenger at a station or bus stop, where the passenger continues by public transit
RP4: Multimodal pick-up D1, D2 P2 Driver collects the passenger at a station or bus stop, where the passenger has arrived by public transit
RP5: Accompaniment D1 P1 Driver accompanies the passenger throughout, with no solo leg observed in the immediately preceding or following trips
RP6: En-route pick-up D2 P1 Driver picks up the passenger at an intermediate point along the route
RP7: En-route drop-off D3 P1 Driver drops off the passenger at an intermediate point along the route
Joint trip
J1: Pure joint trip D1 P1 Driver and passenger share all origins, destinations, and trip purposes throughout the trip chain
J2: Partial joint trip D1 P1 Driver and passenger share the main destination, but the driver makes at least one independent trip
Indeterminate
Indeterminate D1 P1 Cannot be classified as either ride provision or joint trip with available information
Table 3.2: Classification Outcomes by D × P Pattern
P1 P2 P3 P4
D1 J1 (Step 2-1);
RP1, RP2, RP5, J2, Indet. (Step 2-2) 		RP4 (Step 1) RP3 (Step 1) Not observed
D2 RP6 (Step 1) RP4 (Step 1) Undetectable Not observed
D3 RP7 (Step 1) Undetectable RP3 (Step 1) Not observed
D4 Undetectable Undetectable Undetectable Undetectable

4 4. Results

4.1 4.1 Identification of Driver–Passenger Pairs

Table 4.1 presents the number of matched pairs by D×P pattern combination. The total number of passenger trip candidates aged 75 or older was 2,220: 2,212 P1 trips, 3 P2 trips, 5 P3 trips, and 0 P4 trips. These candidates are limited to trips in which the respondent reported being driven by a family member. An additional 580 trips were reported as driven by someone other than a family member; their treatment is discussed below.

Of the 2,220 passenger candidates, 1,820 (82.0%) were successfully matched to a driver trip within the same household. P1 passengers—those traveling by car only—account for 99.7% of all matched pairs (1,815 pairs). The D1+P1 combination, in which a dedicated driver shares the same origin and destination as the passenger, is the most frequent at 1,785 pairs, with the majority of pairs identified in Step 1. D2+P1 (en-route pick-up) and D3+P1 (en-route drop-off) account for 21 and 9 pairs respectively, indicating that patterns in which the driver picks up or drops off the passenger at an intermediate point along the route are relatively uncommon. P2 and P3 pairs combined account for only 5 matched pairs, reflecting the auto-dependent nature of the Yamagata regional area, where elderly residents have limited opportunities to combine public transit and car travel.

Of the 2,220 passenger candidates, 400 (18.0%) could not be matched. Two main factors account for these failures. First, in some cases no qualifying driver trip exists within the household—either because the driver is not a household member, or because the driver is a household member whose travel was not reported. Regarding the latter, 141 of the 868 households (16.2%) containing an elderly passenger aged 75 or older had at least one household member with no reported travel, leaving open the possibility that one of those unreported members provided the ride. Second, technical failures arising from missing or inconsistent zone codes or time records account for a further share of unmatched cases.

Two sources of potential underestimation warrant discussion. First, the survey asks respondents whether the driver was a “family member” without specifying whether this refers to a household member. If respondents interpreted “family member” broadly to include non-household relatives—such as a child or grandchild living separately—then some of the 580 trips recorded as driven by a non-family member may in fact represent rides provided by non-household family members. Second, respondents who interpreted “family member” as restricted to household members may not have reported rides provided by non-household relatives at all, leaving such trips unrecorded in the candidate pool. These ambiguities suggest that the estimates presented here may undercount the true number of family-provided rides. In addition, the D4 pattern and the D2+P3 and D3+P2 combinations are structurally undetectable by the algorithm, constituting a further source of downward bias.

Table 4.1: Matching Results by D × P Pattern Combination
P1 P2 P3 P4
D1 1,785 3 2 Not observed
D2 21 0 Undetectable Not observed
D3 9 Undetectable 0 Not observed
D4 Undetectable Undetectable Undetectable Undetectable
Total 1,815 / 2,212 (82%) 3 / 3 (100%) 2 / 5 (40%) —

4.2 4.2 Classification of Ride Provision and Joint Trips

Table 4.2 presents the classification results. Of the 1,820 matched pairs, 287 (15.8%) are classified as ride provision, 1,494 (82.1%) as joint trips, and 39 (2.1%) as Indeterminate.

The most frequent type is pure joint trip (J1), accounting for 1,261 pairs (69.3%)—approximately 70% of all matched pairs. J1 is assigned when the driver and passenger share the same origin, destination, and trip purpose across every trip in the trip chain, indicating that the two travel together throughout the day. Partial joint trip (J2) accounts for a further 233 pairs (12.8%). The finding that joint trips account for more than 80% of all matched pairs indicates that the large majority of pairs identified in Section 2 represent family members traveling together toward a shared destination, rather than a driver serving the passenger’s mobility needs.

Within the 287 ride provision pairs, accompaniment (RP5) is the most frequent type at 121 pairs (6.6%), in which the driver accompanies the passenger throughout the trip—hospital visits and shopping trips being typical examples. Dedicated drop-off (RP1) accounts for 89 pairs (4.9%) and dedicated pick-up (RP2) for 42 pairs (2.3%). RP1 and RP2 represent cases in which the driver is observed traveling alone either after dropping off the passenger or before collecting the passenger, corresponding to a more strictly defined deliver-or-collect structure. En-route pick-up (RP6) and en-route drop-off (RP7) account for 21 and 9 pairs respectively (1.6% combined). Multimodal types (RP3 and RP4) total 5 pairs, again reflecting the auto-dependent nature of the study area discussed in Section 4.1.

The 39 Indeterminate pairs (2.1%) could not be classified because information on the driver’s preceding or following trips, or dwell times at the origin or destination, was missing. At approximately 2% of all matched pairs, the uncertainty introduced by unclassifiable cases is limited.

Table 4.2: Classification Results
Pattern n %
Pick-up/Drop-off
RP1: Dedicated drop-off 89 4.9
RP2: Dedicated pick-up 42 2.3
RP3: Multimodal drop-off 2 0.1
RP4: Multimodal pick-up 3 0.2
RP5: Accompaniment 121 6.6
RP6: En-route pick-up 21 1.2
RP7: En-route drop-off 9 0.5
Subtotal 287 15.8
Joint travel
J1: Pure joint trip 1,261 69.3
J2: Partial joint trip 233 12.8
Subtotal 1,494 82.1
Indeterminate
Indeterminate 39 2.1
Total 1,820 100.0