Document Type


Degree Name

Doctor of Philosophy (PhD)


Geography & Environmental Studies


Faculty of Arts

First Advisor

Barry Boots

Advisor Role

Thesis Supervisor


This study examined the relationships between road access and the fishing site choices of northern Ontario recreational anglers. A revealed preference choice model (random utility model) was estimated with fishing trip data from an angling diary with resident anglers from the Thunder Bay and Wawa areas. The results showed that poor quality gravel roads and trails heavily and negatively impacted fishing site choices by Thunder Bay anglers who fished only during the open water season. Poorer quality roads and trails had much less impact on the fishing site choices of other Thunder Bay anglers. Wawa area anglers were, on average, less impacted by poor quality roads and trails than were Thunder Bay area anglers. Several methods of incorporating spatial complexities into the fishing site choice models were also investigated. First, an accessibility attribute was included in the models to account for potential spatial cognitive limitations of anglers when choosing fishing sites. While this attribute had a significant effect in the models, the effect was different for Thunder Bay and Wawa area anglers. A second spatial measure focused on whether anglers took fishing trips near their previously chosen fishing sites. Anglers often took fishing trips back to the fishing sites they previously chose. Thunder Bay area anglers also tended to take fishing trips that were close to their previously chosen fishing site. Finally, various generalized extreme value models were used to determine if nearby sites have correlated unobserved utilities. Results from a cross-nested logit model, which permit researchers to allocate fishing alternatives into more than one nest, showed that spatially near fishing alternatives shared some unobserved utility. Therefore, nearby fishing sites were better substitutes than were far away fishing sites. Generalized nested logit models were estimated to assess whether one global parameter could capture the correlation pattern among the unobserved utilities for the fishing sites. A global parameter was rejected in favour of nest specific parameters. While not truly a local level analysis, the generalized nested logit model was sufficient to capture some spatial heterogeneity present in the correlations among the unobserved utilities.

Convocation Year


Convocation Season