Separating Contextual from Endogenous Effects in Automobile Ownership Models

Abstract

Using the 1997/98 New York Metropolitan Transportation Council household survey and United States Census, we estimate an instrumental variable probit model to test the impact of contextual and endogenous social interaction effects on auto ownership and determine that the probability of car ownership is affected by both types of social interaction effects. Previous research focused only either on contextual effects, or, increasingly, on endogenous effects using contextual effects variables as instruments. Therefore we were unable to find studies looking at both social interaction effects simultaneously. Consistent with earlier results, we find that households have a higher probability of possessing a vehicle if they are surrounded by other automobile-owning households (endogenous effect). However, we find that contextual effects are correctly measured only when the endogenous effect is included. In our case, everything else being equal, households in poorer neighborhoods are more likely to own vehicles, and households in neighborhoods with higher proportions of people with graduate degrees are less likely to own vehicles. This suggests that car ownership in New York City is a status symbol for poorer households and that non-car-ownership is a status symbol for people with post baccalaureate education. The results are important in two policy contexts: as auto ownership is a precursor to trip generation and mode choice, auto ownership estimation is important to effective travel forecasting; as vehicle miles traveled (VMT) is tied to auto ownership, VMT reduction strategies, as a way to improve air quality, reduce congestion, and reduce greenhouse gas emissions, may depend on strategies to reduce auto ownership. In either case, correct modeling of auto ownership will lead to more effective policy outcomes.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 58 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.