Sample Selection Criteria and Modeling Choices
| Category | Criterion Type | Subcategory | Baseline Choice | Alternative |
|---|---|---|---|---|
| Water quality | Measurement | Water quality statistic | Summer mean | Summer minimum, summer maximum |
| Water quality | Measurement | Water quality timing | Closest within ± 5 years | Year of sale, year prior to sale, closest within previous 5 years |
| Water quality | Minimum threshold | Coefficient of variation for Secchi depth | 0.25 | N/A |
| Property | Measurement | Property attributes | Lot size; building size; building age | None, bedrooms and bathroomsa |
| Property | Maximum threshold | Lakefront property buffer | 150 m | 300 m |
| Property | Minimum threshold | Sales in buffer | 1 | 0, 25 |
| Property | Maximum threshold | Property distance to lake | 1,000 m | 2,500 m |
| Modeling Choices | ||||
| Functional form | Sale price, water quality | Log, log | Log, linear | |
| Functional form | Covariatesb | Log | N/A | |
| Fixed effects | Spatial unit | Census tract | Block group | |
| Fixed effects | Interactions | Spatial unit × sale year | Spatial unit + sale year | |
↵a All combinations of property attributes are considered (e.g., individual attributes, two at a time, three at a time, and all attributes). Bedrooms and bathrooms are always included together.
↵b In addition to the selected property attributes, model covariates include distance to lake, lake area, median income of block group, and slope and elevation of the parcel.