This dissertation is devoted to the study of environmental valuation with three independent topics. The first topic investigates the consistency of consumer preferences over time and revealed versus stated preference data. This study draws on data from the Iowa Lakes Project, which provides information on recreational usage patterns over several years and for approximately 130 lakes, along with detailed information on the water quality for each lake. This allows examination of the extent of convergence in how individuals respond to changing site characteristics. In addition, because of the nature of the data, this study was able to investigate the consistency of consumer preferences over time and between actual versus anticipated visitation patterns. The second topic examines how housing prices were impacted by this unexpected event, while controlling for pre-existing flood risk (as captured by 100- and 500-year floodplains). Both difference-in-differences (DID) and triple differences (DDD) techniques are used to isolate the impact of the 2008 flood. The results show prices for houses within the 100-year floodplain were discounted prior to the flood and no significant changes occurred in prices for those houses inundated by the 2008 flood. However, the results find a significant rebounding in post-flood prices in areas not actually inundated by the flood. On the other hand, the prices of properties in both the 500-year floodplains and outside the floodplains were not significantly discounted before the flood. There was a significant decrease in price after the flood, if the area was inundated. These findings imply as new information on flooding occurs, the housing market updates the risk perception for properties, as indicated by a change in housing prices. Finally, the third topic analyzes state dependence in recreational demand using panel data on lake visitation patterns from the Iowa Lakes Project. When calculating the welfare effect of an environmental policy, the estimation can be misleading--either by ignoring state dependence or by dealing with state dependence incorrectly. To avoid this problem, this topic adopts the approach proposed by Wooldridge. This approach starts with a single site case and then extends the analysis to a multiple site setting. For the single site case, a dynamic random effect (RE) logit model is utilized. In the multiple site setting, a RE two-step nesting structure model is used, capturing state dependence in terms of overall trip taking, although not in terms of the specific sites selected. For both the single and multiple site cases, a RE Poisson model is also estimated as an alternative approach to compare the results and as a robustness check. Also, a Monte Carlo simulation exercise is used to show the biases that can arise either from neglecting state dependence entirely or from treating it incorrectly.