Understanding the Concept of Latent Demand in Traffic Prof. Patricia L. Mokhtarian Civil & Environmental Engineering, UC Davis [email protected].

Understanding the Concept of Understanding the Concept of Latent Demand in TrafficLatent Demand in Traffic

Understanding the Concept of Understanding the Concept of Latent Demand in TrafficLatent Demand in Traffic

Prof. Patricia L. MokhtarianProf. Patricia L. Mokhtarian

Civil & Environmental Engineering, UC DavisCivil & Environmental Engineering, UC Davis

[email protected]@ucdavis.edu

www.its.ucdavis.edu/telecom/www.its.ucdavis.edu/telecom/

(530) 752-7062(530) 752-7062

Outline of this TalkOutline of this TalkOutline of this TalkOutline of this Talk

What are latent and induced demand, and their What are latent and induced demand, and their implications?implications?

Empirical approaches to assessing induced demandEmpirical approaches to assessing induced demand– Typical resultsTypical results– LimitationsLimitations

UC Davis study using matched pairsUC Davis study using matched pairs More recent work: Cervero/Hansen & Choo/Mokh.More recent work: Cervero/Hansen & Choo/Mokh. SummarySummary Concluding thoughtsConcluding thoughts

What is Latent Demand?What is Latent Demand?What is Latent Demand?What is Latent Demand?

Often used interchangeably with “induced Often used interchangeably with “induced demand”, but the two concepts can be demand”, but the two concepts can be technically distinguished as follows:technically distinguished as follows:– Latent demand:Latent demand: Pent-up (dormant) demand for Pent-up (dormant) demand for

travel, travel that is desired but travel, travel that is desired but unrealizedunrealized because of constraintsbecause of constraints

– Induced demand:Induced demand: RealizedRealized demand that is demand that is generated (induced, “drawn out”) because of generated (induced, “drawn out”) because of improvements to the transportation systemimprovements to the transportation system

Induced DemandInduced DemandInduced DemandInduced Demand

The increment of new vehicle traffic that The increment of new vehicle traffic that would not have occurred at all without the would not have occurred at all without the capacity improvement.capacity improvement.

Clear in theory, but difficult in practice!Clear in theory, but difficult in practice! Observed increases in traffic on a capacity-Observed increases in traffic on a capacity-

enhanced network link can arise from a enhanced network link can arise from a variety of sources:variety of sources:

When is Traffic Growth When is Traffic Growth Induced Demand?Induced Demand?

When is Traffic Growth When is Traffic Growth Induced Demand?Induced Demand?

Shifts in departure timeShifts in departure time? Changes in route or destination (no for Changes in route or destination (no for

vehicle trips but maybe for VMT)vehicle trips but maybe for VMT) Shifts from shared modes to drive aloneShifts from shared modes to drive alone New or longer trips to existing locationsNew or longer trips to existing locations Background demographic growth (WHOA)Background demographic growth (WHOA)? Trips generated by new development Trips generated by new development

attracted to the improved corridorattracted to the improved corridor

Why do we Care about Why do we Care about Induced Demand?Induced Demand?

Why do we Care about Why do we Care about Induced Demand?Induced Demand?

Need to be able to forecast newly-created Need to be able to forecast newly-created travel (that WNHOA): travel (that WNHOA): – Affects the cost-benefit calculation for the improvementAffects the cost-benefit calculation for the improvement

– Affects the assessment of environmental impactsAffects the assessment of environmental impacts

Legal/political ramifications:Legal/political ramifications:– Sierra Club v. MTC, 1989Sierra Club v. MTC, 1989

– UK abandoned “predict and provide” policy in mid ’90sUK abandoned “predict and provide” policy in mid ’90s

Empirical ApproachesEmpirical ApproachesEmpirical ApproachesEmpirical Approaches

Case studiesCase studies Cross-sectional disaggregate modelingCross-sectional disaggregate modeling Cross-sectional aggregate modelingCross-sectional aggregate modeling Time series aggregate modelingTime series aggregate modeling Cross-sectional/time series aggregate Cross-sectional/time series aggregate

modelingmodeling Time series link/facility level analysis with Time series link/facility level analysis with

controlscontrols

Case StudiesCase StudiesCase StudiesCase Studies

Change in traffic on single facility measuredChange in traffic on single facility measured Results mixed, but have generally found Results mixed, but have generally found

observed volumes higher than forecastsobserved volumes higher than forecasts May highlight idiosyncratic circumstancesMay highlight idiosyncratic circumstances Often short-term; difficult to distinguish Often short-term; difficult to distinguish

induced demand from shifted demand or induced demand from shifted demand or background growthbackground growth

Cross-sectional Disaggregate Cross-sectional Disaggregate ModelingModeling

Cross-sectional Disaggregate Cross-sectional Disaggregate ModelingModeling

Using 1995 NPTS (travel diary data), Using 1995 NPTS (travel diary data), analyze association of VMT with speedanalyze association of VMT with speed

Higher speeds associated with greater VMTHigher speeds associated with greater VMT Speed is a more behaviorally-sound Speed is a more behaviorally-sound

influence on VMT than capacityinfluence on VMT than capacity Association doesn’t guarantee causality; Association doesn’t guarantee causality;

can’t identify long-term impactscan’t identify long-term impacts

Cross-sectional Aggregate Cross-sectional Aggregate ModelingModeling

Cross-sectional Aggregate Cross-sectional Aggregate ModelingModeling

Models impact of lane-miles on VMT for Models impact of lane-miles on VMT for metro areas in USmetro areas in US

Increase of 1% in lane-mi leads to ~0.8% Increase of 1% in lane-mi leads to ~0.8% increase in VMTincrease in VMT

Potentially represents long-term equilibriumPotentially represents long-term equilibrium Bi-directional causality impossible to Bi-directional causality impossible to

untangle with single equation, no dynamic untangle with single equation, no dynamic elementelement

Time Series Aggregate Time Series Aggregate ModelingModeling

Time Series Aggregate Time Series Aggregate ModelingModeling

Decomposed VMT growth (Milwaukee, Decomposed VMT growth (Milwaukee, 1963-1991) into sources based on assumed 1963-1991) into sources based on assumed relationshipsrelationships

6-22% of total VMT growth attributable to 6-22% of total VMT growth attributable to new capacitynew capacity

Regional focus; decomposition approach Regional focus; decomposition approach usefuluseful

Still only one direction of causality permittedStill only one direction of causality permitted

Cross-sectional/Time Series Cross-sectional/Time Series Aggregate ModelingAggregate Modeling

Cross-sectional/Time Series Cross-sectional/Time Series Aggregate ModelingAggregate Modeling

Models VMT as function of lane-mi among Models VMT as function of lane-mi among other variables, for multiple areas over timeother variables, for multiple areas over time

1% increase in ln-mi 1% increase in ln-mi → 0.2 – 0.9% increase → 0.2 – 0.9% increase in VMT (long-run > short-run)in VMT (long-run > short-run)

Advantages:Advantages:– Covariates help capture background influencesCovariates help capture background influences– If area large enough, demand shifts accounted forIf area large enough, demand shifts accounted for– Temporal precedence can be establishedTemporal precedence can be established

Cross-sectional/Time Series Cross-sectional/Time Series Aggregate Modeling Aggregate Modeling (cont’d)(cont’d)

Cross-sectional/Time Series Cross-sectional/Time Series Aggregate Modeling Aggregate Modeling (cont’d)(cont’d)

Disadvantages:Disadvantages:– Not all background influences capturedNot all background influences captured– Facility/metro-level analyses subject to Facility/metro-level analyses subject to

confounding with changes in classification and confounding with changes in classification and urban boundary over timeurban boundary over time

– Even temporal precedence doesn’t guarantee Even temporal precedence doesn’t guarantee causalitycausality

– Effectiveness of lagged variables depends on Effectiveness of lagged variables depends on whether planning horizon is longer than the lagwhether planning horizon is longer than the lag

Time Series Link/Facility Time Series Link/Facility Level Analysis with ControlsLevel Analysis with Controls

Time Series Link/Facility Time Series Link/Facility Level Analysis with ControlsLevel Analysis with Controls

Compares growth in ADT on improved links, Compares growth in ADT on improved links, to that on matched set of unimproved linksto that on matched set of unimproved links

Study of 18 matched prs in CA (UCD faculty) Study of 18 matched prs in CA (UCD faculty) found no difference in growth ratesfound no difference in growth rates

Controls for causes of growth common to Controls for causes of growth common to improved and comparison segmentsimproved and comparison segments

Several disadvantages:Several disadvantages:

Time Series Link/Facility Time Series Link/Facility Level Analysis Level Analysis (cont’d)(cont’d)

Time Series Link/Facility Time Series Link/Facility Level Analysis Level Analysis (cont’d)(cont’d)

Disadvantages:Disadvantages:– Difficult to find suitable controlsDifficult to find suitable controls– Doesn’t control for spatial shifts from nearbyDoesn’t control for spatial shifts from nearby– Cannot establish a control for an entirely new linkCannot establish a control for an entirely new link

Another possible reason for difference: Another possible reason for difference: ADT v. VMT: new capacity may affect trip ADT v. VMT: new capacity may affect trip lengthlength more than more than frequencyfrequency

Recent Work: Cervero/HansenRecent Work: Cervero/HansenRecent Work: Cervero/HansenRecent Work: Cervero/Hansen

Cross-sectional/time series aggregateCross-sectional/time series aggregate– state hwys, 34 CA counties, 1976-97state hwys, 34 CA counties, 1976-97

Simultaneous equations:Simultaneous equations:– Lane-miles Lane-miles → VMT→ VMT– VMT → lane-milesVMT → lane-miles

Both directions of causality significant, Both directions of causality significant, lane-miles → VMT the stronger directionlane-miles → VMT the stronger direction

Cervero/Hansen Cervero/Hansen (cont’d)(cont’d)Cervero/Hansen Cervero/Hansen (cont’d)(cont’d)

Probably the most rigorous published study to dateProbably the most rigorous published study to date Issues:Issues:

– Did facility reclassification, metro area effects Did facility reclassification, metro area effects confound relationships?confound relationships?

– What happened to traffic on lower-classification What happened to traffic on lower-classification facilities?facilities?

– Are the instrumental variables appropriate?Are the instrumental variables appropriate?

– Is the high goodness-of-fit spurious?Is the high goodness-of-fit spurious?

– Were the lags long enough?Were the lags long enough?

Recent Work: Choo/MokhtarianRecent Work: Choo/MokhtarianRecent Work: Choo/MokhtarianRecent Work: Choo/Mokhtarian

Time series aggregate (USwide, 1951-2000)Time series aggregate (USwide, 1951-2000) Comprehensive structural modelComprehensive structural model

Sociodemo-graphics

Sociodemo-graphics

Travel Demand (VMT)

TelecomDemand

TelecomDemand

Transport.Sys.

Infrastructure

(lane-mi)

Telecom System

Infrastructure

Telecom System

Infrastructure

Travel Costs

Travel Costs

Telecom Costs

Telecom Costs

Economic Activity

Economic Activity

Land Use

Land Use

Endogenous Variable Category Exogenous Variable Category

Choo/Mokhtarian Choo/Mokhtarian (cont’d)(cont’d)Choo/Mokhtarian Choo/Mokhtarian (cont’d)(cont’d)

Time series aggregate (USwide, 1951-2000)Time series aggregate (USwide, 1951-2000) Comprehensive structural modelComprehensive structural model Corrected for high correlations due to Corrected for high correlations due to

similar temporal trendssimilar temporal trends Also found bAlso found both directions of causality oth directions of causality

significant, lane-miles → VMT the stronger significant, lane-miles → VMT the stronger directiondirection

SummarySummarySummarySummary It’s a complex issue!It’s a complex issue! Each approach has advantages and Each approach has advantages and

disadvantages, something to offer but not disadvantages, something to offer but not definitive answersdefinitive answers

To better understand extent to which To better understand extent to which answer depends on method, apply multiple answer depends on method, apply multiple methods to same regionmethods to same region

Nevertheless, the most sophisticated Nevertheless, the most sophisticated analyses find evidence for induced demandanalyses find evidence for induced demand

Concluding ThoughtsConcluding ThoughtsConcluding ThoughtsConcluding Thoughts

Transportation demand will continue to growTransportation demand will continue to grow Thus, can’t eliminate all system improvements Thus, can’t eliminate all system improvements

just because demand will increasejust because demand will increase Should rather weigh the costs (increased fuel Should rather weigh the costs (increased fuel

consumption, emissions) against the benefits consumption, emissions) against the benefits (increased mobility, economic gain)(increased mobility, economic gain)

Need to continue to improve our measurement and Need to continue to improve our measurement and modeling of both costs and benefitsmodeling of both costs and benefits

And continue efforts to more appropriately price And continue efforts to more appropriately price the provision of servicethe provision of service