{"id":"d6f9290e-aa6f-4749-b39d-436a5dd505d5","arxiv_id":"2412.08801","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"Agent-based simulations of nine cities show that ride-sharing raises service rates and lowers emissions while reducing total fare revenue by about 2.3% at a 20% discount.","lead":"Ride-sharing can get more passengers served with fewer cars and lower emissions, but it can also shrink the money a ride-hailing company collects compared with selling solo trips. This paper simulates nine Chinese cities to measure that trade-off and to identify when pooling loses revenue.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Fixed demand assumption omits induced demand, so the revenue-loss dilemma may be an artifact of holding trip volume constant while ride-sharing's economies of scale are unmodeled.","rationale":"The reader's weakest assumption is exactly the fixed-demand issue: the simulation models only the choice among existing orders, so any induced-demand effect from cheaper shared rides is absent and the revenue loss could shrink or reverse if lower prices attract additional riders. I agree that this is the most load-bearing concern because the paper's central 'development dilemma' is a revenue-social welfare trade-off whose revenue side is computed at a constant demand level, and because the paper itself invokes economies of scale in the introduction. The fixed-demand assumption is not merely an external validity caveat; it suppresses the feedback loop that the paper argues is key to ride-sharing's success. The reader's CONDITIONAL verdict is appropriate: the simulation is transparent and the fixed-demand assumption is clearly stated, but the policy relevance of the dilemma is uncertain until induced demand is incorporated. My stress-test does not change this verdict; it reinforces it. I have not identified an internal inconsistency in the model as specified, and the paper deserves credit for a clear methodology, real-world data from nine cities, and a transparent accounting of revenue and social welfare metrics. The proposed test directly probes the sensitivity of the headline revenue estimate to demand scale, which is the most direct way to determine whether the dilemma persists under a plausible extension of the model.","tokens_in":45530,"tokens_out":9677,"duration_ms":108840,"concrete_test":"Re-run the C2-D30 scenario for a representative city (e.g., Chengdu) with demand multiplier m in {1.2, 1.5, 2.0} applied to the order arrival rate for both the pure solo and mixed systems, keeping the fleet fixed at 500. Compute total revenue, service rate, SSR, and SDR for each m. If the mixed-system revenue minus solo revenue crosses zero or becomes positive at some m, or if SSR and SDR rise monotonically with m, then the revenue-loss finding is not robust to demand scale and induced demand could eliminate the dilemma. Also repeat in a low-demand city (e.g., Chongqing) to check whether the effect interacts with baseline density.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that ride-sharing reduces revenue while increasing social welfare is computed under a fixed demand stream. Sections 3.2 and 3.3 specify that customers only choose between solo and shared service for an exogenously given set of orders; there is no price-induced trip generation or mode shift. This matters because the revenue loss is modest (2.3% at a 20% discount and 30% detour guarantee, Figure 11), and the paper's own spatiotemporal analysis (Section 4.3) shows that the successful sharing ratio (SSR) and shared distance ratio (SDR) are substantially higher in hot, high-demand areas. The introduction itself argues that ride-sharing has economies of scale ('higher adoption' leads to 'better service performance' and then 'higher adoption'), yet the simulation holds demand fixed and thereby precludes the very mechanism that could offset the discount. If cheaper shared rides attract additional riders, the larger pool of shareable trips improves SSR and saved distance, so the 2.3% revenue loss could shrink or reverse. The paper acknowledges this gap in Section 4.2 ('customers may shift their choices to ride-sharing due to the shorter waiting time, inducing a higher demand... we leave it for future research'), but the limitation is load-bearing because the policy recommendation to subsidize ride-sharing depends on the generality of the dilemma. The 'development dilemma' may therefore be an artifact of the constant-demand assumption rather than an inherent property of ride-sharing.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper develops an agent-based simulation of an online ride-hailing platform to compare a solo-hailing-only system with a mixed solo/ride-sharing system. Demand comes from downsampled order data in nine Chinese cities, and customer willingness to accept ride-sharing is calibrated from a 402-response survey on discount and detour trade-offs. The platform uses upfront pricing (Eqs. (1)-(2)), batch customer-vehicle assignment (Eqs. (3)-(7)), and repositioning. The headline experiment, at a 20% ride-sharing discount and a 30% maximum detour guarantee, finds that ride-sharing increases the service rate by 15.7%, reduces CO2 per km by 7.1%, increases vehicle occupancy by 16.7%, and reduces waiting time by 2.6%, while reducing total fare revenue by 2.3% (Figure 11). The paper attributes the revenue loss to low successful sharing ratios in low-demand areas, limited saved distance from pooling, and fare disparities between pooled customers, and it argues that carbon-price benefits are too small to offset the loss. It recommends dynamic pricing for TNCs and public subsidies for ride-sharing.","tokens_in":45747,"tokens_out":7838,"duration_ms":80130,"significance":"If the results are taken at face value, the paper offers a concrete, data-grounded demonstration that ride-sharing can create a social-welfare/revenue trade-off, and its three-factor decomposition gives operators and regulators testable levers. The use of nine cities with different spatial structures, a survey-based elasticity calibration, and a relatively transparent simulation architecture are genuine strengths. However, the magnitude of the headline revenue change is small, and the model's fixed-demand assumption, fare-based revenue measure, and lack of uncertainty quantification mean that the generality of the 'development dilemma' is not yet established. I do not see a circularity problem: the scenario identity in Eq. (11) follows from the pricing formulas and is not used to fit the simulations. The paper's contribution is conditional on additional robustness work.","major_comments":[{"comment":"The demand stream is exogenous in the simulation: customers only choose between solo and shared service for a fixed set of orders, so cheaper shared fares cannot generate new trips or attract riders from other modes. Since the headline loss is only 2.3% (Figure 11), an induced-demand channel could plausibly offset or reverse the result: more riders would improve the successful sharing ratio and shared distance, which the paper's own Section 4.3 shows are higher in hot, high-demand areas. The paper explicitly defers this to future work in Section 4.2, but the policy conclusion that ride-sharing needs subsidies to overcome a structural revenue loss depends on the dilemma holding under demand feedback. Please add a sensitivity analysis with induced demand (or mode substitution) or clearly scope the claim to a fixed-demand market.","section":"Sections 3.2-3.3 and 4.2, Figure 11"},{"comment":"The revenue metric is total trip fare charged to served customers, with operating costs, driver payments, and the TNC-driver revenue split all excluded. Therefore the statement that ride-sharing leads to 'a loss of revenue for TNCs or drivers' (abstract and Conclusions) is a statement about gross fare revenue, not profit. Because ride-sharing reduces vehicle distance traveled and idle time in this model (e.g., Figures 9 and 10), drivers or the platform may face lower operating costs that partially offset the 2.3% fare loss. Please either incorporate a cost model, or consistently describe the result as a fare-revenue loss rather than a profit or motivation loss.","section":"Section 3.4.1"},{"comment":"All experimental results are reported as single point estimates with no standard deviations, confidence intervals, or repeated runs. The 2.3% revenue decline is small relative to the city-to-city variation visible in Figure 8, where Jinan shows the opposite pattern. Without a measure of variability, the reader cannot assess whether the sign of the revenue effect is robust or merely an artifact of a particular downsampled dataset and simulation seed. Please report variability across cities and, if feasible, across random realizations or subsamples.","section":"Figures 8-13, especially Figure 11"},{"comment":"The fleet size is fixed at N=500 for all nine cities although arrival rates range from 1,598 to 2,274 orders/hour and average trip distances from 6.0 to 8.0 km (Table 1). Eq. (8) then implies that cities operate at very different normalized loads, so the pooled average effects may be dominated by supply-constrained cities. A robustness check that varies fleet size (or calibrates it to a common service rate) is needed before the average revenue and welfare effects can be generalized across cities.","section":"Section 4.1, Table 1, Eq. (8)"}],"minor_comments":[{"comment":"The simulation horizon is specified as '6 am to 12 pm,' but the text later refers to 18-hour simulations; this should be 6 am to midnight (12 am) if the horizon is 18 hours.","section":"Section 4.1"},{"comment":"There are typographical errors: 'Hence, The revenue' should be 'Hence, the revenue,' and the bullet 'Average number of scheduled requestsrefers' is missing a space.","section":"Section 3.4.1"},{"comment":"The claim that the solo baseline 'should match the idle vehicle with C1' assumes C1 maximizes revenue per unit time; because pickup distances can differ, this is not guaranteed by L1 >= L2 alone. Please state the required condition or restrict the analysis to cases where it holds.","section":"Section 5.1, Eqs. (9)-(11)"},{"comment":"The shortest-path algorithm is consistently spelled 'Dijksta' instead of 'Dijkstra' in the text and the reference list.","section":"Section 3.1.1 and References"},{"comment":"The panels in Figure 12 are not individually labeled with the metric they show, making the figure hard to read; consider adding panel titles.","section":"Figure 12"}],"recommendation":"major_revision","confidential_remarks":"The paper is within the journal's scope and the core simulation is clearly specified, but the headline claim is currently overgeneralized relative to the model's assumptions. I would ask for a sensitivity analysis on induced demand, a clarification of revenue versus profit, and uncertainty quantification; these are fixable within the manuscript's scope. No concerns about citation practices beyond the normal expectation that future versions compare more carefully with prior simulation studies."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a careful simulation study, not a discovery. The new thing is a clean quantitative decomposition of why pooled ride-sharing can cut revenue while improving welfare, run on nine cities with real order data and a small elasticity survey. The headline numbers — service rate +15.7%, CO2/km -7.1%, occupancy +16.7%, fare revenue -2.3% at a 20% discount with a 30% detour guarantee — are honest outputs of a transparent model, and the paper deserves a serious referee.\n\nWhat it does well: the model is clearly specified (pricing, batch matching, repositioning), baseline and mixed systems are compared on the same demand stream, and the scenario analysis in Section 5.1 gives an analytical identity for the revenue ratio. The authors state their main limitation plainly: in Section 4.2 they note that shorter waiting times could induce more demand, and leave that for future work. That is the right way to frame a simulation result.\n\nThe soft spots are the ones the Pith Report names. Fixed demand is the biggest: because the revenue loss is modest, the dilemma could shrink or reverse if lower shared fares attract more riders, as the stress-test says. The paper cannot rule that out, and a referee should push for a sensitivity analysis with induced demand or at least a careful argument about the likely sign. Revenue also excludes operating costs, so the profit implication for TNCs and drivers is undetermined. I would like to see error bars or multiple seeds; none are reported. The 402-person elasticity survey is a thin basis for customer behavior, though it is at least transparent.\n\nNone of this undermines the paper for what it is: a well-documented simulation with a clear, bounded claim. I'd send it to review with a request for robustness checks, not a desk reject. The right audience is anyone working on ride-sourcing economics or mobility simulation; they will cite it for the quantified trade-off.","headline":"A transparent, well-specified simulation that quantifies a modest revenue loss from ride-sharing under fixed demand; the induced-demand gap is real but acknowledged.","tokens_in":46351,"tokens_out":4437,"would_cite":false,"duration_ms":44795,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Adding ride-sharing to a solo-hailing system lifts social welfare but cuts fare revenue, creating a development dilemma for ride-hailing platforms.","keywords":["ride-sharing","revenue","social welfare","agent-based simulation","carbon emissions","pricing","detour elasticity","ride-hailing"],"falsifier":"Run a live A/B test on a ride-hailing platform, randomly assigning matched cities or time windows to solo-only and mixed services at the same 20% discount and 30% detour guarantee, and compare total fare revenue and service rate; if the revenue loss does not appear outside simulation, the dilemma is an artifact of the fixed-demand assumption. A companion demand model with induced trips—where the number of requests grows as shared fares fall—would directly test whether the 2.3% loss persists.","tokens_in":45270,"feed_emoji":"🚗","tokens_out":4838,"duration_ms":47479,"temperature":0.7,"pith_summary":"This paper asks whether ride-sharing can be good for society and good for the companies that operate it. Using trip-order data from nine Chinese cities and a survey of customers' willingness to accept detours in exchange for fare discounts, it simulates a platform running solo-hailing alone or mixing solo-hailing with ride-sharing. It finds that adding ride-sharing raises the share of requests served, cuts carbon emissions per kilometer, and increases vehicle occupancy, but also lowers total fare revenue. The paper identifies three concrete reasons for the revenue loss and argues that carbon benefits are too small to compensate, so operators need better pricing or public subsidies.","feed_headline":"Ride-sharing lifts welfare, trims platform revenue 2.3%","feed_subtitle":"Simulations across nine cities show better service and lower emissions, but less income for operators.","key_machinery":"The load-bearing object is an agent-based simulation of a platform serving a fixed stream of trip orders, with customers choosing solo or shared service based on upfront prices and a detour guarantee, and with matching done by a batch integer linear program that can assign vehicles to single or pooled trips. The scenario analysis reduces the platform's choice to a revenue ratio $\\mu_{\\text{share}}/\\mu_{\\text{solo}} = \\frac{(1-\\theta)(1+w_2/w_1)}{(x_1+x_2+x_3+x_4)/(x_1+L_1)}$, which shows when pooling two customers pays relative to serving only the higher-fare customer solo. This ratio encodes the three failure modes: low fare ratio $w_2/w_1$, high discount $\\theta$, and long shared-route distance relative to solo distance.","core_discovery":"The central claim is a dilemma: under the model, introducing ride-sharing into a solo-hailing system improves social welfare while reducing revenue. In the baseline scenario with a 20% fare discount and a 30% maximum detour guarantee, the mixed service raises the customer service rate by 15.7%, lowers CO2 emissions per km by 7.1%, raises vehicle occupancy by 16.7%, and cuts waiting times by 2.6%, but total fare revenue falls by 2.3% compared with pure solo-hailing. The paper attributes the revenue loss to three mechanisms: low successful sharing ratios where demand is sparse, limited saved trip distance when two customers are pooled, and lost revenue when pooled customers have very different trip fares. It also finds that the monetary value of carbon savings is negligible—at most about 0.01 CNY per trip at current carbon prices—so environmental benefits do not offset the revenue loss and do not shift customer choices. The paper concludes that TNCs' profit motive alone will not sustain ride-sharing under these conditions, and that dynamic pricing or government subsidies are needed.","pith_inferences":["Because the model assumes a fixed demand stream, cheaper ride-sharing creates no new trips; if lower prices induce additional demand, the measured 2.3% revenue loss could shrink or even reverse in real markets.","A natural extension is dynamic pricing that raises shared fares in high-demand zones or disables sharing in cold zones; the paper's revenue-ratio formula suggests this could convert the loss into profit, though the paper does not test it.","Raising vehicle capacity beyond two customers would weaken the saved-distance and fare-ratio drawbacks, since more customers per route spreads the detour cost; this follows from the paper's mechanism but is not simulated here.","The elasticity calibration comes from stated survey preferences; field A/B tests on a real platform could reveal whether actual customers accept the modeled discounts and detours at the same rates."],"forward_implications":["At a 20% discount and 30% detour guarantee, mixed service raises the service rate by 15.7%, cuts CO2 per km by 7.1%, raises occupancy by 16.7%, and cuts waiting times by 2.6%, while total fare revenue drops by 2.3%.","Discounts above about 20% bring little extra service rate or emission benefit but deepen revenue loss, so high discounts are not attractive to operators.","In low-demand ('cold') areas, the successful sharing ratio stays at roughly 40% even at a 40% discount, meaning ride-sharing discounts many fares without actually pooling many trips.","Pooling customers with similar fares and itineraries can preserve revenue; pooling customers with very different fares can lose more than serving only the higher-fare customer solo.","Carbon savings from ride-sharing are worth at most about 0.01 CNY per trip at 2024 carbon prices, so carbon-credit schemes will not meaningfully drive customer adoption or compensate operators."],"supporting_citations":[{"why":"Supplies the batch-matching and trip-vehicle assignment algorithm that the simulation adapts, plus the high-capacity result that 98% of trips can be served with 23% of the fleet.","marker":"Alonso-Mora et al., 2017"},{"why":"Introduces the shareability network and the finding that about half of New York taxi trips can be pooled, providing the baseline for pooling feasibility.","marker":"Santi et al., 2014"},{"why":"Provides the agent-based simulation platform architecture, matching-process details, and CO2 calculation that this study extends with upfront pricing and joint solo/shared assignment.","marker":"Chen et al., 2024"},{"why":"Supplies the normalized-load threshold used to split zones into cold, normal, and hot demand areas.","marker":"Molkenthin et al., 2020"},{"why":"Defines the COPERT model used to compute CO2 emissions and the saved emissions from reduced travel distance.","marker":"Ntziachristos and Samaras, 2020"},{"why":"Provides theoretical pricing and equilibrium results for ride-pooling markets that frame the discount and detour trade-off.","marker":"Ke et al., 2020"}],"fun_headline_variants":["Ride-sharing: social welfare up, revenue down 2.3%","Ride-sharing's dilemma: welfare gains, revenue losses","Carbon savings too small to rescue ride-sharing revenue","Nine-city study: ride-sharing lifts welfare, trims profits"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The results assume the stream of trip requests is fixed: cheaper ride-sharing attracts no new trips and pulls no demand from other modes, so the platform only redistributes existing orders between solo and shared service.","fun_headline_variants_meta":{"raw":{"variants":["Ride-sharing: social welfare up, revenue down 2.3%","Ride-sharing's dilemma: welfare gains, revenue losses","Carbon savings too small to rescue ride-sharing revenue","Nine-city study: ride-sharing lifts welfare, trims profits"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000214,"raw_usage":{"total_tokens":1442,"prompt_tokens":982,"completion_tokens":460,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":598,"completion_tokens_details":{"reasoning_tokens":390}},"tokens_in":598,"tokens_out":460,"duration_ms":4902,"temperature":1.0,"reasoning_tokens":390,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T17:32:00.961186+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run a live A/B test on a ride-hailing platform, randomly assigning matched cities or time windows to solo-only and mixed services at the same 20% discount and 30% detour guarantee, and compare total fare revenue and service rate; if the revenue loss does not appear outside simulation, the dilemma is an artifact of the fixed-demand assumption. A companion demand model with induced trips—where the number of requests grows as shared fares fall—would directly test whether the 2.3% loss persists.","supporting_citations":[],"review_version":1}