Code
library(tidyverse)
library(fixest)
library(did)
library(etwfe)
library(tidyverse)
library(modelsummary)
library(gt)
options(scipen = 999)This page provides model tables for the additional Q1 robustness checks discussed in dissertation Section 5.4.9 (printed pages 61-62). These checks examine whether estimated price changes associated with FEMA map updates depend on geographic proximity or on controls for local housing-market conditions.
| Check discussed in the dissertation | Results on this page |
|---|---|
| Nearby comparison properties | Sample restricted to the 300-meter proximity list |
| Possible spatial spillovers | Exclusion of nearby properties without an SFHA transition |
| Year-by-municipality fixed effects | Nearby sample |
| Year-by-township fixed effects | Nearby sample and full prepared sample |
| Year-by-FIRM-panel fixed effects | Nearby sample |
The outcome is the logarithm of sale price. The models include property (pin) fixed effects and cluster standard errors at the 10-digit parcel (pin10) level. The geographic time controls vary by section. Preliminary and effective map designations are estimated separately; within each section, the preliminary-map table precedes the effective-map table.
In the tables labeled I and II, I estimates the added-to-SFHA indicator and II estimates the removed-from-SFHA indicator. The corresponding Subset columns exclude the opposite transition and the always-SFHA group (for additions) or never-SFHA group (for removals). Where shown, Symmetric uses the in-SFHA indicator, while column 0 includes addition and removal indicators together. Positive coefficients indicate higher estimated sale prices and negative coefficients indicate lower estimated prices; a log coefficient b converts to a percentage change as 100 * (exp(b) - 1). Stars and standard errors should be read alongside the sample and fixed-effects specification.
These tables use the existing prepared dataset outputs/q2_marginaleffects/bldg_sfha_q2_dissertation_prepped.RDS; despite its folder name, the models below use Q1 FEMA map-transition indicators. The 300-meter indicator comes from data/processed/bldgswithin300mof_SFHA_arcgis.xlsx, matched on PIN10. This page documents the implemented samples. It does not establish that the current tables reproduce the exact 15.9% and 18.1% estimates reported in the dissertation; that requires a separate check of the original data vintage, model, and coefficient transformation.
The first robustness check asks whether comparisons among geographically close properties produce similar results to the broader Q1 analysis. All observations in these models are restricted to nearby_300 == TRUE, not just control observations. Membership is defined by the existing GIS proximity list; the buffer geometry is not reconstructed here. The first table shows separate preliminary/effective addition and removal models with property and year fixed effects. The following tables also show the restricted comparison groups described above.
# way too big of buffer zone
#close_comps <- readxl::read_xlsx("data/processed/parcelswithin600mof2026SFHA_ExportTable_TableToExcel.xlsx")
#300 meter buffer
close_comps <- readxl::read_xlsx("data/processed/bldgswithin300mof_SFHA_arcgis.xlsx")
lake100ft <- readxl::read_xlsx("data/processed/parcels_2025_lakeMI_within100ft_ExportTable_TableToExcel.xlsx")
lake500ft <- readxl::read_xlsx("data/processed/parcels_2025_lakeMI_within500ft_ExportTable_TableToExcel.xlsx")
# already has eff_sfha_category in it
data_path <- "outputs/q2_marginaleffects/bldg_sfha_q2_dissertation_prepped.RDS"
df <- readRDS(data_path)
df <- df |>
mutate(coastal_100 = pin10 %in% lake100ft$PIN10,
coastal_500 = pin10 %in% lake500ft$PIN10,
nearby_300 = pin10 %in% close_comps$PIN10)
df_nocoast <- df |> filter(coastal_500==FALSE)nearby_props_model_pre_add <- feols(
log_price ~ addedto_prelim_sfha | pin + sale_year ,
cluster = ~ pin10,
data = (df |> filter(nearby_300 ==
TRUE))
)
nearby_props_model_pre_remove <- feols(
log_price ~ removedfrom_prelim_sfha | pin + sale_year ,
cluster = ~ pin10,
data = (df |> filter(nearby_300 ==
TRUE))
)
nearby_props_model_eff_add <- feols(
log_price ~ addedto_eff_sfha | pin + sale_year ,
cluster = ~ pin10,
data = (df |> filter(nearby_300 ==
TRUE))
)
nearby_props_model_eff_remove <- feols(
log_price ~ removedfrom_eff_sfha| pin + sale_year ,
cluster = ~ pin10,
data = (df |> filter(nearby_300 ==
TRUE))
)
modelsummary(list(
nearby_props_model_pre_add, nearby_props_model_pre_remove, nearby_props_model_eff_add, nearby_props_model_eff_remove), output = "gt", stars = TRUE, estimate = "{estimate}{stars}",
statistic = "({std.error})", gof_omit = "IC|Log|Adj|AIC|BIC",
# coef_map = coef_map2_short,
title = "Q1 FEMA map transitions: nearby sample, property and year fixed effects") |>
tab_source_note(md("FE: pin, sale_year. SE clustered at 10-digit parcel level."))| (1) | (2) | (3) | (4) | |
|---|---|---|---|---|
| addedto_prelim_sfhaTRUE | 0.099 | |||
| (0.107) | ||||
| removedfrom_prelim_sfhaTRUE | 0.040 | |||
| (0.039) | ||||
| addedto_eff_sfhaTRUE | 0.032 | |||
| (0.098) | ||||
| removedfrom_eff_sfhaTRUE | 0.159*** | |||
| (0.036) | ||||
| Num.Obs. | 88540 | 88540 | 88540 | 88540 |
| R2 | 0.889 | 0.889 | 0.889 | 0.889 |
| R2 Within | 0.000 | 0.000 | 0.000 | 0.000 |
| RMSE | 0.27 | 0.27 | 0.27 | 0.27 |
| Std.Errors | by: pin10 | by: pin10 | by: pin10 | by: pin10 |
| FE: pin | X | X | X | X |
| FE: sale_year | X | X | X | X |
| FE: pin, sale_year. SE clustered at 10-digit parcel level. | ||||
prelim_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_prelim_sfha | pin + sale_year,
cluster = ~pin10,
data = (df |> filter(nearby_300 ==
TRUE))
),
"I Subset" = feols(
log(sale_price) ~ addedto_prelim_sfha | pin + sale_year,
cluster = ~ pin10,
data = (df |> filter(change_type_prelim != "Always SFHA" & treat_year_remove_prelim == 10000 & nearby_300 == TRUE))),
"II" = feols(
log(sale_price) ~ removedfrom_prelim_sfha | pin + sale_year,
cluster = ~ pin10,
data =(df |> filter(nearby_300 == TRUE))
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_prelim_sfha | pin + sale_year,
cluster = ~ pin10,
data = (df |> filter(change_type_prelim != "Never SFHA" & treat_year_add_prelim == 10000& nearby_300 == TRUE))
),
# Symmetric effect of location in risk zone
"Symmetric" = feols(
log(sale_price) ~ in_prelim_sfha | pin + sale_year,
cluster = ~ pin10,
data = (df |> filter(nearby_300 ==
TRUE))
),
# Asymmetric: added vs removed in one equation
"0" = feols(
log(sale_price) ~ addedto_prelim_sfha + removedfrom_prelim_sfha | pin + sale_year,
cluster = ~ pin10,
data = (df |> filter(nearby_300 ==
TRUE))
)
)
all_models_formatted <- modelsummary::modelsummary(prelim_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
| I | I Subset | II | II Subset | Symmetric | 0 |
|---|---|---|---|---|---|---|
addedto_prelim_sfhaTRUE | 0.1 | 0.1 | 0.1 | |||
(0.11) | (0.11) | (0.11) | ||||
removedfrom_prelim_sfhaTRUE | 0.04 | 0 | 0.04 | |||
(0.04) | (0.05) | (0.04) | ||||
in_prelim_sfhaTRUE | -0.03 | |||||
(0.04) | ||||||
Num.Obs. | 88540 | 78203 | 88540 | 10337 | 88540 | 88540 |
R2 | 0.889 | 0.888 | 0.889 | 0.890 | 0.889 | 0.889 |
R2 Adj. | 0.802 | 0.800 | 0.802 | 0.805 | 0.802 | 0.802 |
R2 Within | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
R2 Within Adj. | -0.000 | -0.000 | -0.000 | -0.000 | -0.000 | -0.000 |
AIC | 96345.8 | 85657.8 | 96345.4 | 10628.2 | 96345.8 | 96347.0 |
BIC | 462095.9 | 404871.1 | 462095.5 | 43339.7 | 462095.8 | 462106.5 |
RMSE | 0.27 | 0.27 | 0.27 | 0.26 | 0.27 | 0.27 |
Std.Errors | by: pin10 | by: pin10 | by: pin10 | by: pin10 | by: pin10 | by: pin10 |
FE: pin | X | X | X | X | X | X |
FE: sale_year | X | X | X | X | X | X |
+ p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 | ||||||
eff_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_eff_sfha | pin + sale_year,
cluster = ~pin10,
data = (df |> filter(nearby_300 ==
TRUE))
),
"I Subset" = feols(
log(sale_price) ~ addedto_eff_sfha | pin + sale_year,
cluster = ~ pin10,
data = (df |> filter(change_type != "Always SFHA" & treat_year_remove_eff == 10000 & nearby_300 == TRUE))),
"II" = feols(
log(sale_price) ~ removedfrom_eff_sfha | pin + sale_year,
cluster = ~ pin10,
data =(df |> filter(nearby_300 == TRUE))
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_eff_sfha | pin + sale_year,
cluster = ~ pin10,
data = (df |> filter(change_type != "Never SFHA" & treat_year_add_eff == 10000& nearby_300 == TRUE))
),
# Symmetric effect of location in risk zone
"Symmetric" = feols(
log(sale_price) ~ in_eff_sfha | pin + sale_year,
cluster = ~ pin10,
data = (df |> filter(nearby_300 ==
TRUE))
),
# Asymmetric: added vs removed in one equation
"0" = feols(
log(sale_price) ~ addedto_eff_sfha + removedfrom_eff_sfha | pin + sale_year,
cluster = ~ pin10,
data = (df |> filter(nearby_300 ==
TRUE))
)
)
all_models_formatted <- modelsummary::modelsummary(eff_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
| I | I Subset | II | II Subset | Symmetric | 0 |
|---|---|---|---|---|---|---|
addedto_eff_sfhaTRUE | 0.03 | 0.04 | 0.03 | |||
(0.1) | (0.1) | (0.1) | ||||
removedfrom_eff_sfhaTRUE | 0.16*** | 0.13** | 0.16*** | |||
(0.04) | (0.04) | (0.04) | ||||
in_eff_sfhaTRUE | -0.15*** | |||||
(0.04) | ||||||
Num.Obs. | 88540 | 78073 | 88540 | 10467 | 88540 | 88540 |
R2 | 0.889 | 0.888 | 0.889 | 0.891 | 0.889 | 0.889 |
R2 Adj. | 0.802 | 0.800 | 0.802 | 0.807 | 0.802 | 0.802 |
R2 Within | 0.000 | 0.000 | 0.000 | 0.001 | 0.000 | 0.000 |
R2 Within Adj. | -0.000 | -0.000 | 0.000 | 0.001 | 0.000 | 0.000 |
AIC | 96346.2 | 85574.3 | 96330.9 | 10695.0 | 96331.6 | 96332.9 |
BIC | 462096.3 | 404220.6 | 462081.0 | 43862.1 | 462081.7 | 462092.3 |
RMSE | 0.27 | 0.27 | 0.27 | 0.26 | 0.27 | 0.27 |
Std.Errors | by: pin10 | by: pin10 | by: pin10 | by: pin10 | by: pin10 | by: pin10 |
FE: pin | X | X | X | X | X | X |
FE: sale_year | X | X | X | X | X | X |
+ p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 | ||||||
prelim_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_prelim_sfha*factor(pre_date) | pin + sale_year,
cluster = ~pin10,
data = (df |> filter(nearby_300 ==
TRUE))
),
"I Subset" = feols(
log(sale_price) ~ addedto_prelim_sfha*factor(pre_date) | pin + sale_year,
cluster = ~ pin10,
data = (df |> filter(change_type_prelim != "Always SFHA" & treat_year_remove_prelim == 10000 & nearby_300 == TRUE))),
"II" = feols(
log(sale_price) ~ removedfrom_prelim_sfha | pin + sale_year,
cluster = ~ pin10,
data =(df |> filter(nearby_300 == TRUE))
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_prelim_sfha*factor(pre_date) | pin + sale_year,
cluster = ~ pin10,
data = (df |> filter(change_type_prelim != "Never SFHA" & treat_year_add_prelim == 10000& nearby_300 == TRUE))
)
)
all_models_formatted <- modelsummary::modelsummary(prelim_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
| I | I Subset | II | II Subset |
|---|---|---|---|---|
addedto_prelim_sfhaTRUE | 0.11 | 0.11 | ||
(0.13) | (0.13) | |||
factor(pre_date)2015-02-12 | -0.06 | |||
(21276.26) | ||||
addedto_prelim_sfhaTRUE × factor(pre_date)2015-02-12 | -0.07 | -0.06 | ||
(0.13) | (0.13) | |||
removedfrom_prelim_sfhaTRUE | 0.04 | 0.12 | ||
(0.04) | (0.13) | |||
removedfrom_prelim_sfhaTRUE × factor(pre_date)2015-02-12 | -0.15 | |||
(0.13) | ||||
Num.Obs. | 88540 | 78203 | 88540 | 10337 |
R2 | 0.889 | 0.888 | 0.889 | 0.890 |
R2 Adj. | 0.802 | 0.800 | 0.802 | 0.805 |
R2 Within | 0.000 | 0.000 | 0.000 | 0.000 |
R2 Within Adj. | -0.000 | -0.000 | -0.000 | -0.000 |
AIC | 96349.8 | 85659.8 | 96345.4 | 10628.4 |
BIC | 462118.6 | 404882.3 | 462095.5 | 43347.2 |
RMSE | 0.27 | 0.27 | 0.27 | 0.26 |
Std.Errors | by: pin10 | by: pin10 | by: pin10 | by: pin10 |
FE: pin | X | X | X | X |
FE: sale_year | X | X | X | X |
+ p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 | ||||
eff_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_eff_sfha*factor(eff_date) | pin + sale_year,
cluster = ~pin10,
data = (df |> filter(nearby_300 ==
TRUE))
),
"I Subset" = feols(
log(sale_price) ~ addedto_eff_sfha*factor(eff_date) | pin + sale_year,
cluster = ~ pin10,
data = (df |> filter(change_type != "Always SFHA" & treat_year_remove_eff == 10000 & nearby_300 == TRUE))),
"II" = feols(
log(sale_price) ~ removedfrom_eff_sfha*factor(eff_date) | pin + sale_year,
cluster = ~ pin10,
data =(df |> filter(nearby_300 == TRUE))
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_eff_sfha*factor(eff_date) | pin + sale_year,
cluster = ~ pin10,
data = (df |> filter(change_type != "Never SFHA" & treat_year_add_eff == 10000& nearby_300 == TRUE))
)
)
all_models_formatted <- modelsummary::modelsummary(eff_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
| I | I Subset | II | II Subset |
|---|---|---|---|---|
addedto_eff_sfhaTRUE | 0.03 | 0.04 | ||
(0.1) | (0.1) | |||
factor(eff_date)2019-11-01 | -0.09 | -0.08 | ||
(21276.64) | (21260.46) | |||
removedfrom_eff_sfhaTRUE | 0.16*** | 0.13** | ||
(0.04) | (0.04) | |||
Num.Obs. | 88540 | 78073 | 88540 | 10467 |
R2 | 0.889 | 0.888 | 0.889 | 0.891 |
R2 Adj. | 0.802 | 0.800 | 0.802 | 0.807 |
R2 Within | 0.000 | 0.000 | 0.000 | 0.001 |
R2 Within Adj. | -0.000 | -0.000 | 0.000 | 0.001 |
AIC | 96348.2 | 85574.3 | 96332.9 | 10695.0 |
BIC | 462107.7 | 404220.6 | 462092.4 | 43862.1 |
RMSE | 0.27 | 0.27 | 0.27 | 0.26 |
Std.Errors | by: pin10 | by: pin10 | by: pin10 | by: pin10 |
FE: pin | X | X | X | X |
FE: sale_year | X | X | X | X |
+ p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 | ||||
This implements the municipality-year check described on dissertation page 62. The models use property fixed effects plus sale_year^clean_name, allowing each municipality to have its own effect in each year. They retain the 300-meter sample restriction. Differences from the property-plus-year models therefore reflect the additional geographic time controls within that restricted sample.
prelim_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_prelim_sfha | pin + sale_year^clean_name,
cluster = ~pin10,
data = (df |> filter(nearby_300 ==
TRUE))
),
"I Subset" = feols(
log(sale_price) ~ addedto_prelim_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df |> filter(change_type_prelim != "Always SFHA" & treat_year_remove_prelim == 10000 & nearby_300 == TRUE))),
"II" = feols(
log(sale_price) ~ removedfrom_prelim_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data =(df |> filter(nearby_300 == TRUE))
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_prelim_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df |> filter(change_type_prelim != "Never SFHA" & treat_year_add_prelim == 10000& nearby_300 == TRUE))
),
# Symmetric effect of location in risk zone
"Symmetric" = feols(
log(sale_price) ~ in_prelim_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df |> filter(nearby_300 ==
TRUE))
),
# Asymmetric: added vs removed in one equation
"0" = feols(
log(sale_price) ~ addedto_prelim_sfha + removedfrom_prelim_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df |> filter(nearby_300 ==
TRUE))
)
)
all_models_formatted <- modelsummary::modelsummary(prelim_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
| I | I Subset | II | II Subset | Symmetric | 0 |
|---|---|---|---|---|---|---|
addedto_prelim_sfhaTRUE | 0.05 | 0.08 | 0.05 | |||
(0.09) | (0.09) | (0.09) | ||||
removedfrom_prelim_sfhaTRUE | 0.05 | -0.05 | 0.05 | |||
(0.04) | (0.1) | (0.04) | ||||
in_prelim_sfhaTRUE | -0.04 | |||||
(0.04) | ||||||
Num.Obs. | 88437 | 78105 | 88437 | 10026 | 88437 | 88437 |
R2 | 0.903 | 0.903 | 0.903 | 0.918 | 0.903 | 0.903 |
R2 Adj. | 0.822 | 0.820 | 0.822 | 0.834 | 0.822 | 0.822 |
R2 Within | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
R2 Within Adj. | -0.000 | -0.000 | -0.000 | -0.000 | -0.000 | -0.000 |
AIC | 86948.6 | 77580.4 | 86947.4 | 8654.1 | 86947.8 | 86949.3 |
BIC | 467489.6 | 411325.6 | 467488.4 | 45108.3 | 467488.8 | 467499.6 |
RMSE | 0.25 | 0.25 | 0.25 | 0.23 | 0.25 | 0.25 |
Std.Errors | by: pin10 | by: pin10 | by: pin10 | by: pin10 | by: pin10 | by: pin10 |
FE: pin | X | X | X | X | X | X |
FE: sale_year^clean_name | X | X | X | X | X | X |
+ p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 | ||||||
eff_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_eff_sfha | pin + sale_year^clean_name,
cluster = ~pin10,
data = (df |> filter(nearby_300 ==
TRUE))
),
"I Subset" = feols(
log(sale_price) ~ addedto_eff_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df |> filter(change_type != "Always SFHA" & treat_year_remove_eff == 10000 & nearby_300 == TRUE))),
"II" = feols(
log(sale_price) ~ removedfrom_eff_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data =(df |> filter(nearby_300 == TRUE))
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_eff_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df |> filter(change_type != "Never SFHA" & treat_year_add_eff == 10000& nearby_300 == TRUE))
),
# Symmetric effect of location in risk zone
"Symmetric" = feols(
log(sale_price) ~ in_eff_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df |> filter(nearby_300 ==
TRUE))
),
# Asymmetric: added vs removed in one equation
"0" = feols(
log(sale_price) ~ addedto_eff_sfha + removedfrom_eff_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df |> filter(nearby_300 ==
TRUE))
)
)
all_models_formatted <- modelsummary::modelsummary(eff_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
| I | I Subset | II | II Subset | Symmetric | 0 |
|---|---|---|---|---|---|---|
addedto_eff_sfhaTRUE | 0.04 | 0.03 | 0.04 | |||
(0.09) | (0.09) | (0.09) | ||||
removedfrom_eff_sfhaTRUE | 0.07 | -0.06 | 0.07 | |||
(0.05) | (0.08) | (0.05) | ||||
in_eff_sfhaTRUE | -0.06 | |||||
(0.05) | ||||||
Num.Obs. | 88437 | 77975 | 88437 | 10162 | 88437 | 88437 |
R2 | 0.903 | 0.903 | 0.903 | 0.918 | 0.903 | 0.903 |
R2 Adj. | 0.822 | 0.820 | 0.822 | 0.836 | 0.822 | 0.822 |
R2 Within | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
R2 Within Adj. | -0.000 | -0.000 | 0.000 | -0.000 | 0.000 | -0.000 |
AIC | 86948.7 | 77518.8 | 86946.6 | 8691.3 | 86946.8 | 86948.6 |
BIC | 467489.7 | 410694.4 | 467487.6 | 45632.7 | 467487.8 | 467499.0 |
RMSE | 0.25 | 0.25 | 0.25 | 0.22 | 0.25 | 0.25 |
Std.Errors | by: pin10 | by: pin10 | by: pin10 | by: pin10 | by: pin10 | by: pin10 |
FE: pin | X | X | X | X | X | X |
FE: sale_year^clean_name | X | X | X | X | X | X |
+ p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 | ||||||
These tables repeat the nearby-sample municipality-by-year specifications above. They are retained for continuity with the existing output and should not be interpreted as a distinct robustness check.
prelim_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_prelim_sfha | pin + sale_year^clean_name,
cluster = ~pin10,
data = (df |> filter(nearby_300 ==
TRUE))
),
"I Subset" = feols(
log(sale_price) ~ addedto_prelim_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df |> filter(change_type_prelim != "Always SFHA" & treat_year_remove_prelim == 10000 & nearby_300 == TRUE))),
"II" = feols(
log(sale_price) ~ removedfrom_prelim_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data =(df |> filter(nearby_300 == TRUE))
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_prelim_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df |> filter(change_type_prelim != "Never SFHA" & treat_year_add_prelim == 10000& nearby_300 == TRUE))
),
# Symmetric effect of location in risk zone
"Symmetric" = feols(
log(sale_price) ~ in_prelim_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df |> filter(nearby_300 ==
TRUE))
),
# Asymmetric: added vs removed in one equation
"0" = feols(
log(sale_price) ~ addedto_prelim_sfha + removedfrom_prelim_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df |> filter(nearby_300 ==
TRUE))
)
)
all_models_formatted <- modelsummary::modelsummary(prelim_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
| I | I Subset | II | II Subset | Symmetric | 0 |
|---|---|---|---|---|---|---|
addedto_prelim_sfhaTRUE | 0.05 | 0.08 | 0.05 | |||
(0.09) | (0.09) | (0.09) | ||||
removedfrom_prelim_sfhaTRUE | 0.05 | -0.05 | 0.05 | |||
(0.04) | (0.1) | (0.04) | ||||
in_prelim_sfhaTRUE | -0.04 | |||||
(0.04) | ||||||
Num.Obs. | 88437 | 78105 | 88437 | 10026 | 88437 | 88437 |
R2 | 0.903 | 0.903 | 0.903 | 0.918 | 0.903 | 0.903 |
R2 Adj. | 0.822 | 0.820 | 0.822 | 0.834 | 0.822 | 0.822 |
R2 Within | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
R2 Within Adj. | -0.000 | -0.000 | -0.000 | -0.000 | -0.000 | -0.000 |
AIC | 86948.6 | 77580.4 | 86947.4 | 8654.1 | 86947.8 | 86949.3 |
BIC | 467489.6 | 411325.6 | 467488.4 | 45108.3 | 467488.8 | 467499.6 |
RMSE | 0.25 | 0.25 | 0.25 | 0.23 | 0.25 | 0.25 |
Std.Errors | by: pin10 | by: pin10 | by: pin10 | by: pin10 | by: pin10 | by: pin10 |
FE: pin | X | X | X | X | X | X |
FE: sale_year^clean_name | X | X | X | X | X | X |
+ p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 | ||||||
eff_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_eff_sfha | pin + sale_year^clean_name,
cluster = ~pin10,
data = (df |> filter(nearby_300 ==
TRUE))
),
"I Subset" = feols(
log(sale_price) ~ addedto_eff_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df |> filter(change_type != "Always SFHA" & treat_year_remove_eff == 10000 & nearby_300 == TRUE))),
"II" = feols(
log(sale_price) ~ removedfrom_eff_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data =(df |> filter(nearby_300 == TRUE))
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_eff_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df |> filter(change_type != "Never SFHA" & treat_year_add_eff == 10000& nearby_300 == TRUE))
),
# Symmetric effect of location in risk zone
"Symmetric" = feols(
log(sale_price) ~ in_eff_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df |> filter(nearby_300 ==
TRUE))
),
# Asymmetric: added vs removed in one equation
"0" = feols(
log(sale_price) ~ addedto_eff_sfha + removedfrom_eff_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df |> filter(nearby_300 ==
TRUE))
)
)
all_models_formatted <- modelsummary::modelsummary(eff_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
| I | I Subset | II | II Subset | Symmetric | 0 |
|---|---|---|---|---|---|---|
addedto_eff_sfhaTRUE | 0.04 | 0.03 | 0.04 | |||
(0.09) | (0.09) | (0.09) | ||||
removedfrom_eff_sfhaTRUE | 0.07 | -0.06 | 0.07 | |||
(0.05) | (0.08) | (0.05) | ||||
in_eff_sfhaTRUE | -0.06 | |||||
(0.05) | ||||||
Num.Obs. | 88437 | 77975 | 88437 | 10162 | 88437 | 88437 |
R2 | 0.903 | 0.903 | 0.903 | 0.918 | 0.903 | 0.903 |
R2 Adj. | 0.822 | 0.820 | 0.822 | 0.836 | 0.822 | 0.822 |
R2 Within | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
R2 Within Adj. | -0.000 | -0.000 | 0.000 | -0.000 | 0.000 | -0.000 |
AIC | 86948.7 | 77518.8 | 86946.6 | 8691.3 | 86946.8 | 86948.6 |
BIC | 467489.7 | 410694.4 | 467487.6 | 45632.7 | 467487.8 | 467499.0 |
RMSE | 0.25 | 0.25 | 0.25 | 0.22 | 0.25 | 0.25 |
Std.Errors | by: pin10 | by: pin10 | by: pin10 | by: pin10 | by: pin10 | by: pin10 |
FE: pin | X | X | X | X | X | X |
FE: sale_year^clean_name | X | X | X | X | X | X |
+ p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 | ||||||
This implements the FIRM-panel-year check described on dissertation page 62. Property fixed effects are combined with sale_year^FIRM_PAN, so common annual changes can differ across map panels. Both tables restrict observations to the 300-meter sample. These models differ from the earlier FIRM-date interaction models: here, the panel-by-year terms are fixed effects. Read any omitted coefficients in light of the variation remaining after these controls.
prelim_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_prelim_sfha | pin + sale_year^FIRM_PAN,
cluster = ~pin10,
data = (df |> filter(nearby_300 ==
TRUE))
),
"I Subset" = feols(
log(sale_price) ~ addedto_prelim_sfha | pin + sale_year^FIRM_PAN,
cluster = ~ pin10,
data = (df |> filter(change_type_prelim != "Always SFHA" & treat_year_remove_prelim == 10000 & nearby_300 == TRUE))),
"II" = feols(
log(sale_price) ~ removedfrom_prelim_sfha | pin + sale_year^FIRM_PAN,
cluster = ~ pin10,
data =(df |> filter(nearby_300 == TRUE))
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_prelim_sfha | pin + sale_year^FIRM_PAN,
cluster = ~ pin10,
data = (df |> filter(change_type_prelim != "Never SFHA" & treat_year_add_prelim == 10000& nearby_300 == TRUE))
),
# Symmetric effect of location in risk zone
"Symmetric" = feols(
log(sale_price) ~ in_prelim_sfha | pin + sale_year^FIRM_PAN,
cluster = ~ pin10,
data = (df |> filter(nearby_300 ==
TRUE))
),
# Asymmetric: added vs removed in one equation
"0" = feols(
log(sale_price) ~ addedto_prelim_sfha + removedfrom_prelim_sfha | pin + sale_year^FIRM_PAN,
cluster = ~ pin10,
data = (df |> filter(nearby_300 ==
TRUE))
)
)
all_models_formatted <- modelsummary::modelsummary(prelim_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
| I | I Subset | II | II Subset | Symmetric | 0 |
|---|---|---|---|---|---|---|
addedto_prelim_sfhaTRUE | -0.07 | -0.06 | -0.07 | |||
(0.09) | (0.09) | (0.09) | ||||
removedfrom_prelim_sfhaTRUE | 0.01 | -0.03 | 0.01 | |||
(0.04) | (0.09) | (0.04) | ||||
in_prelim_sfhaTRUE | -0.02 | |||||
(0.04) | ||||||
Num.Obs. | 88179 | 77830 | 88179 | 9752 | 88179 | 88179 |
R2 | 0.909 | 0.909 | 0.909 | 0.924 | 0.909 | 0.909 |
R2 Adj. | 0.826 | 0.825 | 0.826 | 0.836 | 0.826 | 0.826 |
R2 Within | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
R2 Within Adj. | -0.000 | -0.000 | -0.000 | -0.000 | -0.000 | -0.000 |
AIC | 84645.2 | 75297.4 | 84645.4 | 8241.4 | 84645.3 | 84647.1 |
BIC | 479692.9 | 423170.2 | 479693.1 | 45985.4 | 479693.0 | 479704.3 |
RMSE | 0.24 | 0.24 | 0.24 | 0.22 | 0.24 | 0.24 |
Std.Errors | by: pin10 | by: pin10 | by: pin10 | by: pin10 | by: pin10 | by: pin10 |
FE: pin | X | X | X | X | X | X |
FE: sale_year^FIRM_PAN | X | X | X | X | X | X |
+ p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 | ||||||
eff_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_eff_sfha | pin + sale_year^FIRM_PAN,
cluster = ~pin10,
data = (df |> filter(nearby_300 ==
TRUE))
),
"I Subset" = feols(
log(sale_price) ~ addedto_eff_sfha | pin + sale_year^FIRM_PAN,
cluster = ~ pin10,
data = (df |> filter(change_type != "Always SFHA" & treat_year_remove_eff == 10000 & nearby_300 == TRUE))),
"II" = feols(
log(sale_price) ~ removedfrom_eff_sfha | pin + sale_year^FIRM_PAN,
cluster = ~ pin10,
data =(df |> filter(nearby_300 == TRUE))
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_eff_sfha | pin + sale_year^FIRM_PAN,
cluster = ~ pin10,
data = (df |> filter(change_type != "Never SFHA" & treat_year_add_eff == 10000& nearby_300 == TRUE))
)
)
all_models_formatted <- modelsummary::modelsummary(eff_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
| I | I Subset | II | II Subset |
|---|---|---|---|---|
addedto_eff_sfhaTRUE | 0.08 | 0.08 | ||
(0.12) | (0.11) | |||
removedfrom_eff_sfhaTRUE | 0.05 | 0.09 | ||
(0.05) | (0.09) | |||
Num.Obs. | 88179 | 77700 | 88179 | 9875 |
R2 | 0.909 | 0.909 | 0.909 | 0.925 |
R2 Adj. | 0.826 | 0.824 | 0.826 | 0.837 |
R2 Within | 0.000 | 0.000 | 0.000 | 0.000 |
R2 Within Adj. | -0.000 | -0.000 | -0.000 | -0.000 |
AIC | 84645.3 | 75235.1 | 84644.2 | 8283.0 |
BIC | 479693.0 | 422535.8 | 479691.9 | 46467.1 |
RMSE | 0.24 | 0.24 | 0.24 | 0.22 |
Std.Errors | by: pin10 | by: pin10 | by: pin10 | by: pin10 |
FE: pin | X | X | X | X |
FE: sale_year^FIRM_PAN | X | X | X | X |
+ p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 | ||||
This implements the township-year check described on dissertation page 62, using property fixed effects and sale_year^Township. The sample is restricted using the same SFHA proximity list as the preceding models. The preliminary and effective tables show whether results persist after allowing each township its own annual conditions.
prelim_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_prelim_sfha | pin + sale_year^Township,
cluster = ~pin10,
data = (df |> filter(nearby_300 ==
TRUE))
),
"I Subset" = feols(
log(sale_price) ~ addedto_prelim_sfha | pin + sale_year^Township,
cluster = ~ pin10,
data = (df |> filter(change_type_prelim != "Always SFHA" & treat_year_remove_prelim == 10000 & nearby_300 == TRUE))),
"II" = feols(
log(sale_price) ~ removedfrom_prelim_sfha | pin + sale_year^Township,
cluster = ~ pin10,
data =(df |> filter(nearby_300 == TRUE))
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_prelim_sfha | pin + sale_year^Township,
cluster = ~ pin10,
data = (df |> filter(change_type_prelim != "Never SFHA" & treat_year_add_prelim == 10000& nearby_300 == TRUE))
)
)
all_models_formatted <- modelsummary::modelsummary(prelim_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
| I | I Subset | II | II Subset |
|---|---|---|---|---|
addedto_prelim_sfhaTRUE | 0.12 | 0.13 | ||
(0.09) | (0.09) | |||
removedfrom_prelim_sfhaTRUE | 0.02 | 0 | ||
(0.04) | (0.05) | |||
Num.Obs. | 88343 | 78019 | 88343 | 10293 |
R2 | 0.897 | 0.896 | 0.897 | 0.905 |
R2 Adj. | 0.814 | 0.813 | 0.814 | 0.823 |
R2 Within | 0.000 | 0.000 | 0.000 | 0.000 |
R2 Within Adj. | -0.000 | -0.000 | -0.000 | -0.000 |
AIC | 90312.1 | 80538.3 | 90312.6 | 9623.5 |
BIC | 458895.3 | 402607.3 | 458895.9 | 44299.4 |
RMSE | 0.26 | 0.26 | 0.26 | 0.24 |
Std.Errors | by: pin10 | by: pin10 | by: pin10 | by: pin10 |
FE: pin | X | X | X | X |
FE: sale_year^Township | X | X | X | X |
+ p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 | ||||
eff_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_eff_sfha | pin + sale_year^Township,
cluster = ~pin10,
data = (df |> filter(nearby_300 ==
TRUE))
),
"I Subset" = feols(
log(sale_price) ~ addedto_eff_sfha | pin + sale_year^Township,
cluster = ~ pin10,
data = (df |> filter(change_type != "Always SFHA" & treat_year_remove_eff == 10000 & nearby_300 == TRUE))),
"II" = feols(
log(sale_price) ~ removedfrom_eff_sfha | pin + sale_year^Township,
cluster = ~ pin10,
data =(df |> filter(nearby_300 == TRUE))
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_eff_sfha | pin + sale_year^Township,
cluster = ~ pin10,
data = (df |> filter(change_type != "Never SFHA" & treat_year_add_eff == 10000& nearby_300 == TRUE))
)
)
all_models_formatted <- modelsummary::modelsummary(eff_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
| I | I Subset | II | II Subset |
|---|---|---|---|---|
addedto_eff_sfhaTRUE | 0.03 | 0.03 | ||
(0.09) | (0.09) | |||
removedfrom_eff_sfhaTRUE | 0.04 | 0.02 | ||
(0.04) | (0.05) | |||
Num.Obs. | 88343 | 77889 | 88343 | 10425 |
R2 | 0.897 | 0.896 | 0.897 | 0.906 |
R2 Adj. | 0.814 | 0.812 | 0.814 | 0.824 |
R2 Within | 0.000 | 0.000 | 0.000 | 0.000 |
R2 Within Adj. | -0.000 | -0.000 | -0.000 | -0.000 |
AIC | 90312.7 | 80476.5 | 90311.8 | 9685.2 |
BIC | 458896.0 | 401978.1 | 458895.1 | 44828.2 |
RMSE | 0.26 | 0.26 | 0.26 | 0.24 |
Std.Errors | by: pin10 | by: pin10 | by: pin10 | by: pin10 |
FE: pin | X | X | X | X |
FE: sale_year^Township | X | X | X | X |
+ p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 | ||||
These models use the same township-by-year controls without the 300-meter restriction. The full prepared dataset remains subject to its upstream selection rules and is not every Cook County property. Comparing these tables with the nearby-sample township models helps distinguish the proximity restriction from the choice of time controls.
prelim_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_prelim_sfha | pin + sale_year^Township,
cluster = ~pin10,
data = (df )
),
"I Subset" = feols(
log(sale_price) ~ addedto_prelim_sfha | pin + sale_year^Township,
cluster = ~ pin10,
data = (df |> filter(change_type_prelim != "Always SFHA" & treat_year_remove_prelim == 10000 ))),
"II" = feols(
log(sale_price) ~ removedfrom_prelim_sfha | pin + sale_year^Township,
cluster = ~ pin10,
data =(df)
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_prelim_sfha | pin + sale_year^Township,
cluster = ~ pin10,
data = (df |> filter(change_type_prelim != "Never SFHA" & treat_year_add_prelim == 10000))
)
)
all_models_formatted <- modelsummary::modelsummary(prelim_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
| I | I Subset | II | II Subset |
|---|---|---|---|---|
addedto_prelim_sfhaTRUE | 0.14 | 0.14 | ||
(0.09) | (0.09) | |||
removedfrom_prelim_sfhaTRUE | 0.02 | 0 | ||
(0.04) | (0.05) | |||
Num.Obs. | 665738 | 650852 | 665738 | 14856 |
R2 | 0.863 | 0.862 | 0.863 | 0.915 |
R2 Adj. | 0.758 | 0.756 | 0.758 | 0.843 |
R2 Within | 0.000 | 0.000 | 0.000 | 0.000 |
R2 Within Adj. | -0.000 | -0.000 | -0.000 | -0.000 |
AIC | 946239.4 | 930115.5 | 946239.8 | 13348.3 |
BIC | 4252660.6 | 4156087.4 | 4252661.0 | 64971.3 |
RMSE | 0.32 | 0.32 | 0.32 | 0.24 |
Std.Errors | by: pin10 | by: pin10 | by: pin10 | by: pin10 |
FE: pin | X | X | X | X |
FE: sale_year^Township | X | X | X | X |
+ p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 | ||||
eff_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_eff_sfha | pin + sale_year^Township,
cluster = ~pin10,
data = (df)
),
"I Subset" = feols(
log(sale_price) ~ addedto_eff_sfha | pin + sale_year^Township,
cluster = ~ pin10,
data = (df |> filter(change_type != "Always SFHA" & treat_year_remove_eff == 10000))),
"II" = feols(
log(sale_price) ~ removedfrom_eff_sfha | pin + sale_year^Township,
cluster = ~ pin10,
data =(df )
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_eff_sfha | pin + sale_year^Township,
cluster = ~ pin10,
data = (df |> filter(change_type != "Never SFHA" & treat_year_add_eff == 10000))
)
)
all_models_formatted <- modelsummary::modelsummary(eff_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
| I | I Subset | II | II Subset |
|---|---|---|---|---|
addedto_eff_sfhaTRUE | 0.05 | 0.05 | ||
(0.08) | (0.08) | |||
removedfrom_eff_sfhaTRUE | 0.04 | 0.02 | ||
(0.03) | (0.05) | |||
Num.Obs. | 665738 | 650713 | 665738 | 14997 |
R2 | 0.863 | 0.862 | 0.863 | 0.915 |
R2 Adj. | 0.758 | 0.756 | 0.758 | 0.844 |
R2 Within | 0.000 | 0.000 | 0.000 | 0.000 |
R2 Within Adj. | -0.000 | -0.000 | -0.000 | -0.000 |
AIC | 946240.0 | 930016.3 | 946239.4 | 13409.6 |
BIC | 4252661.1 | 4155255.9 | 4252660.6 | 65553.6 |
RMSE | 0.32 | 0.32 | 0.32 | 0.24 |
Std.Errors | by: pin10 | by: pin10 | by: pin10 | by: pin10 |
FE: pin | X | X | X | X |
FE: sale_year^Township | X | X | X | X |
+ p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 | ||||
prelim_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_prelim_sfha | pin + sale_year^Township,
cluster = ~pin10,
data = (df_nocoast |> filter(nearby_300 ==
TRUE))
),
"I Subset" = feols(
log(sale_price) ~ addedto_prelim_sfha | pin + sale_year^Township,
cluster = ~ pin10,
data = (df_nocoast |> filter(change_type_prelim != "Always SFHA" & treat_year_remove_prelim == 10000 & nearby_300 == TRUE))),
"II" = feols(
log(sale_price) ~ removedfrom_prelim_sfha | pin + sale_year^Township,
cluster = ~ pin10,
data =(df_nocoast |> filter(nearby_300 == TRUE))
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_prelim_sfha | pin + sale_year^Township,
cluster = ~ pin10,
data = (df_nocoast |> filter(change_type_prelim != "Never SFHA" & treat_year_add_prelim == 10000& nearby_300 == TRUE))
)
)
all_models_formatted <- modelsummary::modelsummary(prelim_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
| I | I Subset | II | II Subset |
|---|---|---|---|---|
addedto_prelim_sfhaTRUE | 0.12 | 0.13 | ||
(0.09) | (0.09) | |||
removedfrom_prelim_sfhaTRUE | 0.02 | 0 | ||
(0.04) | (0.05) | |||
Num.Obs. | 84014 | 73692 | 84014 | 10291 |
R2 | 0.894 | 0.893 | 0.894 | 0.905 |
R2 Adj. | 0.810 | 0.807 | 0.810 | 0.822 |
R2 Within | 0.000 | 0.000 | 0.000 | 0.000 |
R2 Within Adj. | -0.000 | -0.000 | -0.000 | -0.000 |
AIC | 85928.0 | 76110.1 | 85928.6 | 9622.5 |
BIC | 434767.3 | 378682.6 | 434767.8 | 44290.2 |
RMSE | 0.26 | 0.26 | 0.26 | 0.24 |
Std.Errors | by: pin10 | by: pin10 | by: pin10 | by: pin10 |
FE: pin | X | X | X | X |
FE: sale_year^Township | X | X | X | X |
+ p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 | ||||
eff_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_eff_sfha | pin + sale_year^Township,
cluster = ~pin10,
data = (df_nocoast |> filter(nearby_300 ==
TRUE))
),
"I Subset" = feols(
log(sale_price) ~ addedto_eff_sfha | pin + sale_year^Township,
cluster = ~ pin10,
data = (df_nocoast |> filter(change_type != "Always SFHA" & treat_year_remove_eff == 10000 & nearby_300 == TRUE))),
"II" = feols(
log(sale_price) ~ removedfrom_eff_sfha | pin + sale_year^Township,
cluster = ~ pin10,
data =(df_nocoast |> filter(nearby_300 == TRUE))
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_eff_sfha | pin + sale_year^Township,
cluster = ~ pin10,
data = (df_nocoast |> filter(change_type != "Never SFHA" & treat_year_add_eff == 10000& nearby_300 == TRUE))
)
)
all_models_formatted <- modelsummary::modelsummary(eff_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
| I | I Subset | II | II Subset |
|---|---|---|---|---|
addedto_eff_sfhaTRUE | 0.03 | 0.03 | ||
(0.09) | (0.09) | |||
removedfrom_eff_sfhaTRUE | 0.04 | 0.02 | ||
(0.04) | (0.05) | |||
Num.Obs. | 84014 | 73562 | 84014 | 10423 |
R2 | 0.894 | 0.893 | 0.894 | 0.906 |
R2 Adj. | 0.810 | 0.807 | 0.810 | 0.824 |
R2 Within | 0.000 | 0.000 | 0.000 | 0.000 |
R2 Within Adj. | -0.000 | -0.000 | -0.000 | -0.000 |
AIC | 85928.7 | 76048.3 | 85927.8 | 9684.2 |
BIC | 434767.9 | 378056.6 | 434767.0 | 44819.0 |
RMSE | 0.26 | 0.26 | 0.26 | 0.24 |
Std.Errors | by: pin10 | by: pin10 | by: pin10 | by: pin10 |
FE: pin | X | X | X | X |
FE: sale_year^Township | X | X | X | X |
+ p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 | ||||
prelim_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_prelim_sfha | pin + sale_year^clean_name,
cluster = ~pin10,
data = (df_nocoast )
),
"I Subset" = feols(
log(sale_price) ~ addedto_prelim_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df_nocoast |> filter(change_type_prelim != "Always SFHA" & treat_year_remove_prelim == 10000 ))),
"II" = feols(
log(sale_price) ~ removedfrom_prelim_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data =(df_nocoast)
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_prelim_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df_nocoast |> filter(change_type_prelim != "Never SFHA" & treat_year_add_prelim == 10000))
)
)
all_models_formatted <- modelsummary::modelsummary(prelim_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
| I | I Subset | II | II Subset |
|---|---|---|---|---|
addedto_prelim_sfhaTRUE | 0.12 | 0.13 | ||
(0.1) | (0.1) | |||
removedfrom_prelim_sfhaTRUE | 0.05 | -0.05 | ||
(0.04) | (0.11) | |||
Num.Obs. | 656716 | 641937 | 656716 | 14479 |
R2 | 0.864 | 0.863 | 0.864 | 0.924 |
R2 Adj. | 0.758 | 0.756 | 0.758 | 0.851 |
R2 Within | 0.000 | 0.000 | 0.000 | 0.000 |
R2 Within Adj. | -0.000 | -0.000 | -0.000 | -0.000 |
AIC | 928380.2 | 912500.6 | 928380.0 | 12316.4 |
BIC | 4202758.2 | 4106998.4 | 4202758.0 | 65652.5 |
RMSE | 0.32 | 0.32 | 0.32 | 0.23 |
Std.Errors | by: pin10 | by: pin10 | by: pin10 | by: pin10 |
FE: pin | X | X | X | X |
FE: sale_year^clean_name | X | X | X | X |
+ p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 | ||||
eff_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_eff_sfha | pin + sale_year^clean_name,
cluster = ~pin10,
data = (df_nocoast)
),
"I Subset" = feols(
log(sale_price) ~ addedto_eff_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df_nocoast |> filter(change_type != "Always SFHA" & treat_year_remove_eff == 10000 ))),
"II" = feols(
log(sale_price) ~ removedfrom_eff_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data =(df_nocoast )
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_eff_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df_nocoast |> filter(change_type != "Never SFHA" & treat_year_add_eff == 10000))
)
)
all_models_formatted <- modelsummary::modelsummary(eff_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
| I | I Subset | II | II Subset |
|---|---|---|---|---|
addedto_eff_sfhaTRUE | 0 | -0.01 | ||
(0.11) | (0.1) | |||
removedfrom_eff_sfhaTRUE | 0.06 | -0.07 | ||
(0.04) | (0.08) | |||
Num.Obs. | 656716 | 641798 | 656716 | 14628 |
R2 | 0.864 | 0.863 | 0.864 | 0.924 |
R2 Adj. | 0.758 | 0.756 | 0.758 | 0.853 |
R2 Within | 0.000 | 0.000 | 0.000 | 0.000 |
R2 Within Adj. | -0.000 | -0.000 | -0.000 | -0.000 |
AIC | 928380.7 | 912401.8 | 928379.3 | 12360.0 |
BIC | 4202758.7 | 4106167.9 | 4202757.3 | 66276.7 |
RMSE | 0.32 | 0.32 | 0.32 | 0.23 |
Std.Errors | by: pin10 | by: pin10 | by: pin10 | by: pin10 |
FE: pin | X | X | X | X |
FE: sale_year^clean_name | X | X | X | X |
+ p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 | ||||
The dissertation’s second spatial check asks whether nearby comparison properties might also respond to flood-risk information. The code here retains observations outside the 300-meter list, plus properties ever added to or removed from an SFHA under either preliminary or effective maps. It therefore excludes nearby properties with no recorded transition. This is a transition-based rule, rather than a direct filter for current non-SFHA status; it can also exclude nearby always-SFHA properties. That distinction matters when comparing these results with the dissertation’s description.
The first tables use property and year fixed effects on this spillover-exclusion sample. The subsequent tables add geographic time controls. Compare like-for-like transition indicators and map timing across the nearby-sample and spillover-exclusion tables; sample composition differs between the checks.
Could exclude untreated properties within the 300 m buffer around SFHA areas or I could exclude properties that were in the 1 in 500 SFHA zone but not the SFHA zone since that is essentially a bufferzone in itself.
# way too big of buffer zone
#close_comps <- readxl::read_xlsx("data/processed/parcelswithin600mof2026SFHA_ExportTable_TableToExcel.xlsx")
#300 meter buffer
close_comps <- readxl::read_xlsx("data/processed/bldgswithin300mof_SFHA_arcgis.xlsx")
# already has eff_sfha_category in it
data_path <- "outputs/q2_marginaleffects/bldg_sfha_q2_dissertation_prepped.RDS"
df <- readRDS(data_path)
#excludes untreated properties within 300 meters of treated properties.
df_spill <- df |>
mutate(nearby_300 = pin10 %in% close_comps$PIN10) |>
filter(nearby_300 == FALSE |
(ever_added_prelim == TRUE | ever_removed_prelim == TRUE |
ever_added_eff == TRUE | ever_removed_eff)
)
nearby_props_model_pre_add <- feols(
log_price ~ addedto_prelim_sfha | pin + sale_year ,
cluster = ~ pin10,
data = df_spill
)
nearby_props_model_pre_remove <- feols(
log_price ~ removedfrom_prelim_sfha | pin + sale_year ,
cluster = ~ pin10,
data = df_spill
)
nearby_props_model_eff_add <- feols(
log_price ~ addedto_eff_sfha | pin + sale_year ,
cluster = ~ pin10,
data = df_spill
)
nearby_props_model_eff_remove <- feols(
log_price ~ removedfrom_eff_sfha| pin + sale_year ,
cluster = ~ pin10,
data = df_spill
)
modelsummary(list(
nearby_props_model_pre_add, nearby_props_model_pre_remove, nearby_props_model_eff_add, nearby_props_model_eff_remove), output = "gt", stars = TRUE, estimate = "{estimate}{stars}",
statistic = "({std.error})", gof_omit = "IC|Log|Adj|AIC|BIC",
# coef_map = coef_map2_short,
title = "Q1 FEMA map transitions: spillover-exclusion sample, property and year fixed effects") |>
tab_source_note(md("FE: pin, sale_year. SE clustered at 10-digit parcel level."))| (1) | (2) | (3) | (4) | |
|---|---|---|---|---|
| addedto_prelim_sfhaTRUE | 0.133 | |||
| (0.101) | ||||
| removedfrom_prelim_sfhaTRUE | 0.042 | |||
| (0.041) | ||||
| addedto_eff_sfhaTRUE | 0.073 | |||
| (0.085) | ||||
| removedfrom_eff_sfhaTRUE | 0.181*** | |||
| (0.034) | ||||
| Num.Obs. | 579030 | 579030 | 579030 | 579030 |
| R2 | 0.854 | 0.854 | 0.854 | 0.854 |
| R2 Within | 0.000 | 0.000 | 0.000 | 0.000 |
| RMSE | 0.33 | 0.33 | 0.33 | 0.33 |
| Std.Errors | by: pin10 | by: pin10 | by: pin10 | by: pin10 |
| FE: pin | X | X | X | X |
| FE: sale_year | X | X | X | X |
| FE: pin, sale_year. SE clustered at 10-digit parcel level. | ||||
prelim_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_prelim_sfha | pin + sale_year,
cluster = ~pin10,
data = df_spill
),
"I Subset" = feols(
log(sale_price) ~ addedto_prelim_sfha | pin + sale_year,
cluster = ~ pin10,
data = (df_spill |> filter(change_type_prelim != "Always SFHA" & treat_year_remove_prelim == 10000))),
"II" = feols(
log(sale_price) ~ removedfrom_prelim_sfha | pin + sale_year,
cluster = ~ pin10,
data =(df_spill)
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_prelim_sfha | pin + sale_year,
cluster = ~ pin10,
data = (df_spill |> filter(change_type_prelim != "Never SFHA" & treat_year_add_prelim == 10000))
)
)
all_models_formatted <- modelsummary::modelsummary(prelim_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
| I | I Subset | II | II Subset |
|---|---|---|---|---|
addedto_prelim_sfhaTRUE | 0.13 | 0.13 | ||
(0.1) | (0.1) | |||
removedfrom_prelim_sfhaTRUE | 0.04 | 0.26*** | ||
(0.04) | (0.06) | |||
Num.Obs. | 579030 | 574303 | 579030 | 4727 |
R2 | 0.854 | 0.854 | 0.854 | 0.893 |
R2 Adj. | 0.742 | 0.741 | 0.742 | 0.808 |
R2 Within | 0.000 | 0.000 | 0.000 | 0.007 |
R2 Within Adj. | -0.000 | -0.000 | -0.000 | 0.007 |
AIC | 864280.9 | 858542.8 | 864280.8 | 4676.2 |
BIC | 3695295.3 | 3664221.6 | 3695295.2 | 18134.5 |
RMSE | 0.33 | 0.33 | 0.33 | 0.26 |
Std.Errors | by: pin10 | by: pin10 | by: pin10 | by: pin10 |
FE: pin | X | X | X | X |
FE: sale_year | X | X | X | X |
+ p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 | ||||
eff_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_eff_sfha | pin + sale_year,
cluster = ~pin10,
data = (df_spill)
),
"I Subset" = feols(
log(sale_price) ~ addedto_eff_sfha | pin + sale_year,
cluster = ~ pin10,
data = (df_spill |> filter(change_type != "Always SFHA" & treat_year_remove_eff == 10000))),
"II" = feols(
log(sale_price) ~ removedfrom_eff_sfha | pin + sale_year,
cluster = ~ pin10,
data =(df_spill )
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_eff_sfha | pin + sale_year,
cluster = ~ pin10,
data = (df_spill |> filter(change_type != "Never SFHA" & treat_year_add_eff == 10000))
)
)
all_models_formatted <- modelsummary::modelsummary(eff_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
| I | I Subset | II | II Subset |
|---|---|---|---|---|
addedto_eff_sfhaTRUE | 0.07 | 0.07 | ||
(0.09) | (0.09) | |||
removedfrom_eff_sfhaTRUE | 0.18*** | 0.43*** | ||
(0.03) | (0.05) | |||
Num.Obs. | 579030 | 574187 | 579030 | 4843 |
R2 | 0.854 | 0.853 | 0.854 | 0.898 |
R2 Adj. | 0.742 | 0.741 | 0.742 | 0.818 |
R2 Within | 0.000 | 0.000 | 0.000 | 0.025 |
R2 Within Adj. | -0.000 | -0.000 | 0.000 | 0.024 |
AIC | 864281.3 | 858446.0 | 864267.8 | 4723.1 |
BIC | 3695295.7 | 3663534.0 | 3695282.2 | 18543.3 |
RMSE | 0.33 | 0.33 | 0.33 | 0.25 |
Std.Errors | by: pin10 | by: pin10 | by: pin10 | by: pin10 |
FE: pin | X | X | X | X |
FE: sale_year | X | X | X | X |
+ p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 | ||||
For this spillover-exclusion sample, the preliminary-map addition models use municipality-by-year fixed effects, while the preliminary-map removal models use township-by-year fixed effects. The existing analysis note attributes this substitution to collinearity. All effective-map models in this section use municipality-by-year fixed effects. These columns therefore do not all use identical geographic controls.
prelim_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_prelim_sfha | pin + sale_year^clean_name,
cluster = ~pin10,
data = (df_spill )
),
"I Subset" = feols(
log(sale_price) ~ addedto_prelim_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df_spill |> filter(change_type_prelim != "Always SFHA" & treat_year_remove_prelim == 10000 ))),
"II" = feols(
log(sale_price) ~ removedfrom_prelim_sfha | pin + sale_year^Township,
cluster = ~ pin10,
data =(df_spill )
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_prelim_sfha | pin + sale_year^Township,
cluster = ~ pin10,
data = (df_spill |> filter(change_type_prelim != "Never SFHA" & treat_year_add_prelim == 10000))
)
)
all_models_formatted <- modelsummary::modelsummary(prelim_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
| I | I Subset | II | II Subset |
|---|---|---|---|---|
addedto_prelim_sfhaTRUE | 0.14 | 0.14 | ||
(0.1) | (0.1) | |||
removedfrom_prelim_sfhaTRUE | 0.03 | 0.09 | ||
(0.04) | (0.06) | |||
Num.Obs. | 578983 | 574254 | 577675 | 4612 |
R2 | 0.859 | 0.859 | 0.857 | 0.916 |
R2 Adj. | 0.750 | 0.749 | 0.748 | 0.841 |
R2 Within | 0.000 | 0.000 | 0.000 | 0.000 |
R2 Within Adj. | -0.000 | -0.000 | -0.000 | -0.000 |
AIC | 845864.4 | 840443.7 | 848360.7 | 3670.8 |
BIC | 3698697.8 | 3667891.0 | 3677400.7 | 17682.9 |
RMSE | 0.32 | 0.32 | 0.33 | 0.22 |
Std.Errors | by: pin10 | by: pin10 | by: pin10 | by: pin10 |
FE: pin | X | X | X | X |
FE: sale_year^Township | X | X | ||
FE: sale_year^clean_name | X | X | ||
+ p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 | ||||
eff_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_eff_sfha | pin + sale_year^clean_name,
cluster = ~pin10,
data = (df_spill )
),
"I Subset" = feols(
log(sale_price) ~ addedto_eff_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df_spill |> filter(change_type != "Always SFHA" & treat_year_remove_eff == 10000 ))),
"II" = feols(
log(sale_price) ~ removedfrom_eff_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data =(df_spill )
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_eff_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df_spill |> filter(change_type != "Never SFHA" & treat_year_add_eff == 10000))
)
)
all_models_formatted <- modelsummary::modelsummary(eff_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
| I | I Subset | II | II Subset |
|---|---|---|---|---|
addedto_eff_sfhaTRUE | -0.06 | -0.06 | ||
(0.14) | (0.14) | |||
removedfrom_eff_sfhaTRUE | 0.05 | -0.01 | ||
(0.04) | (0.14) | |||
Num.Obs. | 578983 | 574138 | 578983 | 4552 |
R2 | 0.859 | 0.859 | 0.859 | 0.922 |
R2 Adj. | 0.750 | 0.749 | 0.750 | 0.850 |
R2 Within | 0.000 | 0.000 | 0.000 | 0.000 |
R2 Within Adj. | -0.000 | -0.000 | -0.000 | -0.000 |
AIC | 845864.9 | 840363.9 | 845863.9 | 3448.3 |
BIC | 3698698.3 | 3667219.9 | 3698697.4 | 17502.5 |
RMSE | 0.32 | 0.32 | 0.32 | 0.22 |
Std.Errors | by: pin10 | by: pin10 | by: pin10 | by: pin10 |
FE: pin | X | X | X | X |
FE: sale_year^clean_name | X | X | X | X |
+ p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 | ||||
---
title: "Q1 Nearby Property Controls"
format:
html:
df-print: kable
code-fold: true
tbl-cap-location: margin
# error: true
toc: true
warning: false
message: false
---
```{r setup}
library(tidyverse)
library(fixest)
library(did)
library(etwfe)
library(tidyverse)
library(modelsummary)
library(gt)
options(scipen = 999)
```
# Supporting dissertation page 62 {#dissertation-page-62}
This page provides model tables for the additional Q1 robustness checks discussed in dissertation Section 5.4.9 (printed pages 61-62). These checks examine whether estimated price changes associated with FEMA map updates depend on geographic proximity or on controls for local housing-market conditions.
| Check discussed in the dissertation | Results on this page |
|---|---|
| Nearby comparison properties | [Sample restricted to the 300-meter proximity list](#nearby-control-properties-only) |
| Possible spatial spillovers | [Exclusion of nearby properties without an SFHA transition](#spatial-spillover-concerns) |
| Year-by-municipality fixed effects | [Nearby sample](#yearmuni-fixed-effects) |
| Year-by-township fixed effects | [Nearby sample](#yeartownship-fixed-effects-within-300-meters-of-treated-properties) and [full prepared sample](#yeartownship-fixed-effects-all-cook-county) |
| Year-by-FIRM-panel fixed effects | [Nearby sample](#yearfirm_pan-fixed-effects) |
## Reading the model tables
The outcome is the logarithm of sale price. The models include property (`pin`) fixed effects and cluster standard errors at the 10-digit parcel (`pin10`) level. The geographic time controls vary by section. Preliminary and effective map designations are estimated separately; within each section, the preliminary-map table precedes the effective-map table.
In the tables labeled I and II, I estimates the added-to-SFHA indicator and II estimates the removed-from-SFHA indicator. The corresponding Subset columns exclude the opposite transition and the always-SFHA group (for additions) or never-SFHA group (for removals). Where shown, Symmetric uses the in-SFHA indicator, while column 0 includes addition and removal indicators together. Positive coefficients indicate higher estimated sale prices and negative coefficients indicate lower estimated prices; a log coefficient b converts to a percentage change as `100 * (exp(b) - 1)`. Stars and standard errors should be read alongside the sample and fixed-effects specification.
These tables use the existing prepared dataset `outputs/q2_marginaleffects/bldg_sfha_q2_dissertation_prepped.RDS`; despite its folder name, the models below use Q1 FEMA map-transition indicators. The 300-meter indicator comes from `data/processed/bldgswithin300mof_SFHA_arcgis.xlsx`, matched on `PIN10`. This page documents the implemented samples. It does not establish that the current tables reproduce the exact 15.9% and 18.1% estimates reported in the dissertation; that requires a separate check of the original data vintage, model, and coefficient transformation.
[Return to the dissertation supporting-material index](index.qmd#page-62-question-1-additional-robustness-checks).
# Nearby control properties only {#nearby-control-properties-only}
The first robustness check asks whether comparisons among geographically close properties produce similar results to the broader Q1 analysis. All observations in these models are restricted to `nearby_300 == TRUE`, not just control observations. Membership is defined by the existing GIS proximity list; the buffer geometry is not reconstructed here. The first table shows separate preliminary/effective addition and removal models with property and year fixed effects. The following tables also show the restricted comparison groups described above.
```{r}
# way too big of buffer zone
#close_comps <- readxl::read_xlsx("data/processed/parcelswithin600mof2026SFHA_ExportTable_TableToExcel.xlsx")
#300 meter buffer
close_comps <- readxl::read_xlsx("data/processed/bldgswithin300mof_SFHA_arcgis.xlsx")
lake100ft <- readxl::read_xlsx("data/processed/parcels_2025_lakeMI_within100ft_ExportTable_TableToExcel.xlsx")
lake500ft <- readxl::read_xlsx("data/processed/parcels_2025_lakeMI_within500ft_ExportTable_TableToExcel.xlsx")
# already has eff_sfha_category in it
data_path <- "outputs/q2_marginaleffects/bldg_sfha_q2_dissertation_prepped.RDS"
df <- readRDS(data_path)
df <- df |>
mutate(coastal_100 = pin10 %in% lake100ft$PIN10,
coastal_500 = pin10 %in% lake500ft$PIN10,
nearby_300 = pin10 %in% close_comps$PIN10)
df_nocoast <- df |> filter(coastal_500==FALSE)
```
```{r}
nearby_props_model_pre_add <- feols(
log_price ~ addedto_prelim_sfha | pin + sale_year ,
cluster = ~ pin10,
data = (df |> filter(nearby_300 ==
TRUE))
)
nearby_props_model_pre_remove <- feols(
log_price ~ removedfrom_prelim_sfha | pin + sale_year ,
cluster = ~ pin10,
data = (df |> filter(nearby_300 ==
TRUE))
)
nearby_props_model_eff_add <- feols(
log_price ~ addedto_eff_sfha | pin + sale_year ,
cluster = ~ pin10,
data = (df |> filter(nearby_300 ==
TRUE))
)
nearby_props_model_eff_remove <- feols(
log_price ~ removedfrom_eff_sfha| pin + sale_year ,
cluster = ~ pin10,
data = (df |> filter(nearby_300 ==
TRUE))
)
modelsummary(list(
nearby_props_model_pre_add, nearby_props_model_pre_remove, nearby_props_model_eff_add, nearby_props_model_eff_remove), output = "gt", stars = TRUE, estimate = "{estimate}{stars}",
statistic = "({std.error})", gof_omit = "IC|Log|Adj|AIC|BIC",
# coef_map = coef_map2_short,
title = "Q1 FEMA map transitions: nearby sample, property and year fixed effects") |>
tab_source_note(md("FE: pin, sale_year. SE clustered at 10-digit parcel level."))
```
```{r}
prelim_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_prelim_sfha | pin + sale_year,
cluster = ~pin10,
data = (df |> filter(nearby_300 ==
TRUE))
),
"I Subset" = feols(
log(sale_price) ~ addedto_prelim_sfha | pin + sale_year,
cluster = ~ pin10,
data = (df |> filter(change_type_prelim != "Always SFHA" & treat_year_remove_prelim == 10000 & nearby_300 == TRUE))),
"II" = feols(
log(sale_price) ~ removedfrom_prelim_sfha | pin + sale_year,
cluster = ~ pin10,
data =(df |> filter(nearby_300 == TRUE))
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_prelim_sfha | pin + sale_year,
cluster = ~ pin10,
data = (df |> filter(change_type_prelim != "Never SFHA" & treat_year_add_prelim == 10000& nearby_300 == TRUE))
),
# Symmetric effect of location in risk zone
"Symmetric" = feols(
log(sale_price) ~ in_prelim_sfha | pin + sale_year,
cluster = ~ pin10,
data = (df |> filter(nearby_300 ==
TRUE))
),
# Asymmetric: added vs removed in one equation
"0" = feols(
log(sale_price) ~ addedto_prelim_sfha + removedfrom_prelim_sfha | pin + sale_year,
cluster = ~ pin10,
data = (df |> filter(nearby_300 ==
TRUE))
)
)
all_models_formatted <- modelsummary::modelsummary(prelim_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
```
```{r}
eff_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_eff_sfha | pin + sale_year,
cluster = ~pin10,
data = (df |> filter(nearby_300 ==
TRUE))
),
"I Subset" = feols(
log(sale_price) ~ addedto_eff_sfha | pin + sale_year,
cluster = ~ pin10,
data = (df |> filter(change_type != "Always SFHA" & treat_year_remove_eff == 10000 & nearby_300 == TRUE))),
"II" = feols(
log(sale_price) ~ removedfrom_eff_sfha | pin + sale_year,
cluster = ~ pin10,
data =(df |> filter(nearby_300 == TRUE))
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_eff_sfha | pin + sale_year,
cluster = ~ pin10,
data = (df |> filter(change_type != "Never SFHA" & treat_year_add_eff == 10000& nearby_300 == TRUE))
),
# Symmetric effect of location in risk zone
"Symmetric" = feols(
log(sale_price) ~ in_eff_sfha | pin + sale_year,
cluster = ~ pin10,
data = (df |> filter(nearby_300 ==
TRUE))
),
# Asymmetric: added vs removed in one equation
"0" = feols(
log(sale_price) ~ addedto_eff_sfha + removedfrom_eff_sfha | pin + sale_year,
cluster = ~ pin10,
data = (df |> filter(nearby_300 ==
TRUE))
)
)
all_models_formatted <- modelsummary::modelsummary(eff_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
```
## PIN + Year Fixed Effects by FIRM Date
```{r}
prelim_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_prelim_sfha*factor(pre_date) | pin + sale_year,
cluster = ~pin10,
data = (df |> filter(nearby_300 ==
TRUE))
),
"I Subset" = feols(
log(sale_price) ~ addedto_prelim_sfha*factor(pre_date) | pin + sale_year,
cluster = ~ pin10,
data = (df |> filter(change_type_prelim != "Always SFHA" & treat_year_remove_prelim == 10000 & nearby_300 == TRUE))),
"II" = feols(
log(sale_price) ~ removedfrom_prelim_sfha | pin + sale_year,
cluster = ~ pin10,
data =(df |> filter(nearby_300 == TRUE))
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_prelim_sfha*factor(pre_date) | pin + sale_year,
cluster = ~ pin10,
data = (df |> filter(change_type_prelim != "Never SFHA" & treat_year_add_prelim == 10000& nearby_300 == TRUE))
)
)
all_models_formatted <- modelsummary::modelsummary(prelim_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
```
```{r}
eff_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_eff_sfha*factor(eff_date) | pin + sale_year,
cluster = ~pin10,
data = (df |> filter(nearby_300 ==
TRUE))
),
"I Subset" = feols(
log(sale_price) ~ addedto_eff_sfha*factor(eff_date) | pin + sale_year,
cluster = ~ pin10,
data = (df |> filter(change_type != "Always SFHA" & treat_year_remove_eff == 10000 & nearby_300 == TRUE))),
"II" = feols(
log(sale_price) ~ removedfrom_eff_sfha*factor(eff_date) | pin + sale_year,
cluster = ~ pin10,
data =(df |> filter(nearby_300 == TRUE))
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_eff_sfha*factor(eff_date) | pin + sale_year,
cluster = ~ pin10,
data = (df |> filter(change_type != "Never SFHA" & treat_year_add_eff == 10000& nearby_300 == TRUE))
)
)
all_models_formatted <- modelsummary::modelsummary(eff_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
```
## Year-by-municipality fixed effects {#yearmuni-fixed-effects}
This implements the municipality-year check described on dissertation page 62. The models use property fixed effects plus `sale_year^clean_name`, allowing each municipality to have its own effect in each year. They retain the 300-meter sample restriction. Differences from the property-plus-year models therefore reflect the additional geographic time controls within that restricted sample.
```{r}
prelim_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_prelim_sfha | pin + sale_year^clean_name,
cluster = ~pin10,
data = (df |> filter(nearby_300 ==
TRUE))
),
"I Subset" = feols(
log(sale_price) ~ addedto_prelim_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df |> filter(change_type_prelim != "Always SFHA" & treat_year_remove_prelim == 10000 & nearby_300 == TRUE))),
"II" = feols(
log(sale_price) ~ removedfrom_prelim_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data =(df |> filter(nearby_300 == TRUE))
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_prelim_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df |> filter(change_type_prelim != "Never SFHA" & treat_year_add_prelim == 10000& nearby_300 == TRUE))
),
# Symmetric effect of location in risk zone
"Symmetric" = feols(
log(sale_price) ~ in_prelim_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df |> filter(nearby_300 ==
TRUE))
),
# Asymmetric: added vs removed in one equation
"0" = feols(
log(sale_price) ~ addedto_prelim_sfha + removedfrom_prelim_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df |> filter(nearby_300 ==
TRUE))
)
)
all_models_formatted <- modelsummary::modelsummary(prelim_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
```
```{r}
eff_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_eff_sfha | pin + sale_year^clean_name,
cluster = ~pin10,
data = (df |> filter(nearby_300 ==
TRUE))
),
"I Subset" = feols(
log(sale_price) ~ addedto_eff_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df |> filter(change_type != "Always SFHA" & treat_year_remove_eff == 10000 & nearby_300 == TRUE))),
"II" = feols(
log(sale_price) ~ removedfrom_eff_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data =(df |> filter(nearby_300 == TRUE))
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_eff_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df |> filter(change_type != "Never SFHA" & treat_year_add_eff == 10000& nearby_300 == TRUE))
),
# Symmetric effect of location in risk zone
"Symmetric" = feols(
log(sale_price) ~ in_eff_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df |> filter(nearby_300 ==
TRUE))
),
# Asymmetric: added vs removed in one equation
"0" = feols(
log(sale_price) ~ addedto_eff_sfha + removedfrom_eff_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df |> filter(nearby_300 ==
TRUE))
)
)
all_models_formatted <- modelsummary::modelsummary(eff_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
```
## Year-by-municipality fixed effects: additional nearby tables {#yearmuni-fixed-effects-within-300-meters-of-sfha}
These tables repeat the nearby-sample municipality-by-year specifications above. They are retained for continuity with the existing output and should not be interpreted as a distinct robustness check.
```{r}
prelim_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_prelim_sfha | pin + sale_year^clean_name,
cluster = ~pin10,
data = (df |> filter(nearby_300 ==
TRUE))
),
"I Subset" = feols(
log(sale_price) ~ addedto_prelim_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df |> filter(change_type_prelim != "Always SFHA" & treat_year_remove_prelim == 10000 & nearby_300 == TRUE))),
"II" = feols(
log(sale_price) ~ removedfrom_prelim_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data =(df |> filter(nearby_300 == TRUE))
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_prelim_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df |> filter(change_type_prelim != "Never SFHA" & treat_year_add_prelim == 10000& nearby_300 == TRUE))
),
# Symmetric effect of location in risk zone
"Symmetric" = feols(
log(sale_price) ~ in_prelim_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df |> filter(nearby_300 ==
TRUE))
),
# Asymmetric: added vs removed in one equation
"0" = feols(
log(sale_price) ~ addedto_prelim_sfha + removedfrom_prelim_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df |> filter(nearby_300 ==
TRUE))
)
)
all_models_formatted <- modelsummary::modelsummary(prelim_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
```
```{r}
eff_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_eff_sfha | pin + sale_year^clean_name,
cluster = ~pin10,
data = (df |> filter(nearby_300 ==
TRUE))
),
"I Subset" = feols(
log(sale_price) ~ addedto_eff_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df |> filter(change_type != "Always SFHA" & treat_year_remove_eff == 10000 & nearby_300 == TRUE))),
"II" = feols(
log(sale_price) ~ removedfrom_eff_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data =(df |> filter(nearby_300 == TRUE))
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_eff_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df |> filter(change_type != "Never SFHA" & treat_year_add_eff == 10000& nearby_300 == TRUE))
),
# Symmetric effect of location in risk zone
"Symmetric" = feols(
log(sale_price) ~ in_eff_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df |> filter(nearby_300 ==
TRUE))
),
# Asymmetric: added vs removed in one equation
"0" = feols(
log(sale_price) ~ addedto_eff_sfha + removedfrom_eff_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df |> filter(nearby_300 ==
TRUE))
)
)
all_models_formatted <- modelsummary::modelsummary(eff_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
```
## Year-by-FIRM-panel fixed effects {#yearfirm_pan-fixed-effects}
This implements the FIRM-panel-year check described on dissertation page 62. Property fixed effects are combined with `sale_year^FIRM_PAN`, so common annual changes can differ across map panels. Both tables restrict observations to the 300-meter sample. These models differ from the earlier FIRM-date interaction models: here, the panel-by-year terms are fixed effects. Read any omitted coefficients in light of the variation remaining after these controls.
```{r}
prelim_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_prelim_sfha | pin + sale_year^FIRM_PAN,
cluster = ~pin10,
data = (df |> filter(nearby_300 ==
TRUE))
),
"I Subset" = feols(
log(sale_price) ~ addedto_prelim_sfha | pin + sale_year^FIRM_PAN,
cluster = ~ pin10,
data = (df |> filter(change_type_prelim != "Always SFHA" & treat_year_remove_prelim == 10000 & nearby_300 == TRUE))),
"II" = feols(
log(sale_price) ~ removedfrom_prelim_sfha | pin + sale_year^FIRM_PAN,
cluster = ~ pin10,
data =(df |> filter(nearby_300 == TRUE))
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_prelim_sfha | pin + sale_year^FIRM_PAN,
cluster = ~ pin10,
data = (df |> filter(change_type_prelim != "Never SFHA" & treat_year_add_prelim == 10000& nearby_300 == TRUE))
),
# Symmetric effect of location in risk zone
"Symmetric" = feols(
log(sale_price) ~ in_prelim_sfha | pin + sale_year^FIRM_PAN,
cluster = ~ pin10,
data = (df |> filter(nearby_300 ==
TRUE))
),
# Asymmetric: added vs removed in one equation
"0" = feols(
log(sale_price) ~ addedto_prelim_sfha + removedfrom_prelim_sfha | pin + sale_year^FIRM_PAN,
cluster = ~ pin10,
data = (df |> filter(nearby_300 ==
TRUE))
)
)
all_models_formatted <- modelsummary::modelsummary(prelim_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
```
```{r}
eff_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_eff_sfha | pin + sale_year^FIRM_PAN,
cluster = ~pin10,
data = (df |> filter(nearby_300 ==
TRUE))
),
"I Subset" = feols(
log(sale_price) ~ addedto_eff_sfha | pin + sale_year^FIRM_PAN,
cluster = ~ pin10,
data = (df |> filter(change_type != "Always SFHA" & treat_year_remove_eff == 10000 & nearby_300 == TRUE))),
"II" = feols(
log(sale_price) ~ removedfrom_eff_sfha | pin + sale_year^FIRM_PAN,
cluster = ~ pin10,
data =(df |> filter(nearby_300 == TRUE))
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_eff_sfha | pin + sale_year^FIRM_PAN,
cluster = ~ pin10,
data = (df |> filter(change_type != "Never SFHA" & treat_year_add_eff == 10000& nearby_300 == TRUE))
)
)
all_models_formatted <- modelsummary::modelsummary(eff_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
```
## Year-by-township fixed effects: nearby sample {#yeartownship-fixed-effects-within-300-meters-of-treated-properties}
This implements the township-year check described on dissertation page 62, using property fixed effects and `sale_year^Township`. The sample is restricted using the same SFHA proximity list as the preceding models. The preliminary and effective tables show whether results persist after allowing each township its own annual conditions.
```{r}
prelim_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_prelim_sfha | pin + sale_year^Township,
cluster = ~pin10,
data = (df |> filter(nearby_300 ==
TRUE))
),
"I Subset" = feols(
log(sale_price) ~ addedto_prelim_sfha | pin + sale_year^Township,
cluster = ~ pin10,
data = (df |> filter(change_type_prelim != "Always SFHA" & treat_year_remove_prelim == 10000 & nearby_300 == TRUE))),
"II" = feols(
log(sale_price) ~ removedfrom_prelim_sfha | pin + sale_year^Township,
cluster = ~ pin10,
data =(df |> filter(nearby_300 == TRUE))
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_prelim_sfha | pin + sale_year^Township,
cluster = ~ pin10,
data = (df |> filter(change_type_prelim != "Never SFHA" & treat_year_add_prelim == 10000& nearby_300 == TRUE))
)
)
all_models_formatted <- modelsummary::modelsummary(prelim_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
```
```{r}
eff_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_eff_sfha | pin + sale_year^Township,
cluster = ~pin10,
data = (df |> filter(nearby_300 ==
TRUE))
),
"I Subset" = feols(
log(sale_price) ~ addedto_eff_sfha | pin + sale_year^Township,
cluster = ~ pin10,
data = (df |> filter(change_type != "Always SFHA" & treat_year_remove_eff == 10000 & nearby_300 == TRUE))),
"II" = feols(
log(sale_price) ~ removedfrom_eff_sfha | pin + sale_year^Township,
cluster = ~ pin10,
data =(df |> filter(nearby_300 == TRUE))
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_eff_sfha | pin + sale_year^Township,
cluster = ~ pin10,
data = (df |> filter(change_type != "Never SFHA" & treat_year_add_eff == 10000& nearby_300 == TRUE))
)
)
all_models_formatted <- modelsummary::modelsummary(eff_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
```
## Year-by-township fixed effects: full prepared sample {#yeartownship-fixed-effects-all-cook-county}
These models use the same township-by-year controls without the 300-meter restriction. The full prepared dataset remains subject to its upstream selection rules and is not every Cook County property. Comparing these tables with the nearby-sample township models helps distinguish the proximity restriction from the choice of time controls.
```{r}
prelim_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_prelim_sfha | pin + sale_year^Township,
cluster = ~pin10,
data = (df )
),
"I Subset" = feols(
log(sale_price) ~ addedto_prelim_sfha | pin + sale_year^Township,
cluster = ~ pin10,
data = (df |> filter(change_type_prelim != "Always SFHA" & treat_year_remove_prelim == 10000 ))),
"II" = feols(
log(sale_price) ~ removedfrom_prelim_sfha | pin + sale_year^Township,
cluster = ~ pin10,
data =(df)
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_prelim_sfha | pin + sale_year^Township,
cluster = ~ pin10,
data = (df |> filter(change_type_prelim != "Never SFHA" & treat_year_add_prelim == 10000))
)
)
all_models_formatted <- modelsummary::modelsummary(prelim_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
```
```{r}
eff_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_eff_sfha | pin + sale_year^Township,
cluster = ~pin10,
data = (df)
),
"I Subset" = feols(
log(sale_price) ~ addedto_eff_sfha | pin + sale_year^Township,
cluster = ~ pin10,
data = (df |> filter(change_type != "Always SFHA" & treat_year_remove_eff == 10000))),
"II" = feols(
log(sale_price) ~ removedfrom_eff_sfha | pin + sale_year^Township,
cluster = ~ pin10,
data =(df )
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_eff_sfha | pin + sale_year^Township,
cluster = ~ pin10,
data = (df |> filter(change_type != "Never SFHA" & treat_year_add_eff == 10000))
)
)
all_models_formatted <- modelsummary::modelsummary(eff_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
```
# Exclude Coastal Properties
## Only include properties within 300 meters of SFHA
```{r}
prelim_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_prelim_sfha | pin + sale_year^Township,
cluster = ~pin10,
data = (df_nocoast |> filter(nearby_300 ==
TRUE))
),
"I Subset" = feols(
log(sale_price) ~ addedto_prelim_sfha | pin + sale_year^Township,
cluster = ~ pin10,
data = (df_nocoast |> filter(change_type_prelim != "Always SFHA" & treat_year_remove_prelim == 10000 & nearby_300 == TRUE))),
"II" = feols(
log(sale_price) ~ removedfrom_prelim_sfha | pin + sale_year^Township,
cluster = ~ pin10,
data =(df_nocoast |> filter(nearby_300 == TRUE))
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_prelim_sfha | pin + sale_year^Township,
cluster = ~ pin10,
data = (df_nocoast |> filter(change_type_prelim != "Never SFHA" & treat_year_add_prelim == 10000& nearby_300 == TRUE))
)
)
all_models_formatted <- modelsummary::modelsummary(prelim_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
```
```{r}
eff_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_eff_sfha | pin + sale_year^Township,
cluster = ~pin10,
data = (df_nocoast |> filter(nearby_300 ==
TRUE))
),
"I Subset" = feols(
log(sale_price) ~ addedto_eff_sfha | pin + sale_year^Township,
cluster = ~ pin10,
data = (df_nocoast |> filter(change_type != "Always SFHA" & treat_year_remove_eff == 10000 & nearby_300 == TRUE))),
"II" = feols(
log(sale_price) ~ removedfrom_eff_sfha | pin + sale_year^Township,
cluster = ~ pin10,
data =(df_nocoast |> filter(nearby_300 == TRUE))
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_eff_sfha | pin + sale_year^Township,
cluster = ~ pin10,
data = (df_nocoast |> filter(change_type != "Never SFHA" & treat_year_add_eff == 10000& nearby_300 == TRUE))
)
)
all_models_formatted <- modelsummary::modelsummary(eff_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
```
## Year^Muni Fixed Effects
```{r}
prelim_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_prelim_sfha | pin + sale_year^clean_name,
cluster = ~pin10,
data = (df_nocoast )
),
"I Subset" = feols(
log(sale_price) ~ addedto_prelim_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df_nocoast |> filter(change_type_prelim != "Always SFHA" & treat_year_remove_prelim == 10000 ))),
"II" = feols(
log(sale_price) ~ removedfrom_prelim_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data =(df_nocoast)
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_prelim_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df_nocoast |> filter(change_type_prelim != "Never SFHA" & treat_year_add_prelim == 10000))
)
)
all_models_formatted <- modelsummary::modelsummary(prelim_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
```
```{r}
eff_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_eff_sfha | pin + sale_year^clean_name,
cluster = ~pin10,
data = (df_nocoast)
),
"I Subset" = feols(
log(sale_price) ~ addedto_eff_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df_nocoast |> filter(change_type != "Always SFHA" & treat_year_remove_eff == 10000 ))),
"II" = feols(
log(sale_price) ~ removedfrom_eff_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data =(df_nocoast )
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_eff_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df_nocoast |> filter(change_type != "Never SFHA" & treat_year_add_eff == 10000))
)
)
all_models_formatted <- modelsummary::modelsummary(eff_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
```
# Spatial spillover exclusion {#spatial-spillover-concerns}
The dissertation's second spatial check asks whether nearby comparison properties might also respond to flood-risk information. The code here retains observations outside the 300-meter list, plus properties ever added to or removed from an SFHA under either preliminary or effective maps. It therefore excludes nearby properties with no recorded transition. This is a transition-based rule, rather than a direct filter for current non-SFHA status; it can also exclude nearby always-SFHA properties. That distinction matters when comparing these results with the dissertation's description.
The first tables use property and year fixed effects on this spillover-exclusion sample. The subsequent tables add geographic time controls. Compare like-for-like transition indicators and map timing across the nearby-sample and spillover-exclusion tables; sample composition differs between the checks.
Could exclude untreated properties within the 300 m buffer around SFHA areas or I could exclude properties that were in the 1 in 500 SFHA zone but not the SFHA zone since that is essentially a bufferzone in itself.
```{r}
# way too big of buffer zone
#close_comps <- readxl::read_xlsx("data/processed/parcelswithin600mof2026SFHA_ExportTable_TableToExcel.xlsx")
#300 meter buffer
close_comps <- readxl::read_xlsx("data/processed/bldgswithin300mof_SFHA_arcgis.xlsx")
# already has eff_sfha_category in it
data_path <- "outputs/q2_marginaleffects/bldg_sfha_q2_dissertation_prepped.RDS"
df <- readRDS(data_path)
#excludes untreated properties within 300 meters of treated properties.
df_spill <- df |>
mutate(nearby_300 = pin10 %in% close_comps$PIN10) |>
filter(nearby_300 == FALSE |
(ever_added_prelim == TRUE | ever_removed_prelim == TRUE |
ever_added_eff == TRUE | ever_removed_eff)
)
nearby_props_model_pre_add <- feols(
log_price ~ addedto_prelim_sfha | pin + sale_year ,
cluster = ~ pin10,
data = df_spill
)
nearby_props_model_pre_remove <- feols(
log_price ~ removedfrom_prelim_sfha | pin + sale_year ,
cluster = ~ pin10,
data = df_spill
)
nearby_props_model_eff_add <- feols(
log_price ~ addedto_eff_sfha | pin + sale_year ,
cluster = ~ pin10,
data = df_spill
)
nearby_props_model_eff_remove <- feols(
log_price ~ removedfrom_eff_sfha| pin + sale_year ,
cluster = ~ pin10,
data = df_spill
)
modelsummary(list(
nearby_props_model_pre_add, nearby_props_model_pre_remove, nearby_props_model_eff_add, nearby_props_model_eff_remove), output = "gt", stars = TRUE, estimate = "{estimate}{stars}",
statistic = "({std.error})", gof_omit = "IC|Log|Adj|AIC|BIC",
# coef_map = coef_map2_short,
title = "Q1 FEMA map transitions: spillover-exclusion sample, property and year fixed effects") |>
tab_source_note(md("FE: pin, sale_year. SE clustered at 10-digit parcel level."))
```
```{r}
prelim_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_prelim_sfha | pin + sale_year,
cluster = ~pin10,
data = df_spill
),
"I Subset" = feols(
log(sale_price) ~ addedto_prelim_sfha | pin + sale_year,
cluster = ~ pin10,
data = (df_spill |> filter(change_type_prelim != "Always SFHA" & treat_year_remove_prelim == 10000))),
"II" = feols(
log(sale_price) ~ removedfrom_prelim_sfha | pin + sale_year,
cluster = ~ pin10,
data =(df_spill)
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_prelim_sfha | pin + sale_year,
cluster = ~ pin10,
data = (df_spill |> filter(change_type_prelim != "Never SFHA" & treat_year_add_prelim == 10000))
)
)
all_models_formatted <- modelsummary::modelsummary(prelim_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
```
```{r}
eff_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_eff_sfha | pin + sale_year,
cluster = ~pin10,
data = (df_spill)
),
"I Subset" = feols(
log(sale_price) ~ addedto_eff_sfha | pin + sale_year,
cluster = ~ pin10,
data = (df_spill |> filter(change_type != "Always SFHA" & treat_year_remove_eff == 10000))),
"II" = feols(
log(sale_price) ~ removedfrom_eff_sfha | pin + sale_year,
cluster = ~ pin10,
data =(df_spill )
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_eff_sfha | pin + sale_year,
cluster = ~ pin10,
data = (df_spill |> filter(change_type != "Never SFHA" & treat_year_add_eff == 10000))
)
)
all_models_formatted <- modelsummary::modelsummary(eff_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
```
## Geographic time controls after spillover exclusion {#yearmuni-fixed-effects-2}
For this spillover-exclusion sample, the preliminary-map addition models use municipality-by-year fixed effects, while the preliminary-map removal models use township-by-year fixed effects. The existing analysis note attributes this substitution to collinearity. All effective-map models in this section use municipality-by-year fixed effects. These columns therefore do not all use identical geographic controls.
```{r}
prelim_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_prelim_sfha | pin + sale_year^clean_name,
cluster = ~pin10,
data = (df_spill )
),
"I Subset" = feols(
log(sale_price) ~ addedto_prelim_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df_spill |> filter(change_type_prelim != "Always SFHA" & treat_year_remove_prelim == 10000 ))),
"II" = feols(
log(sale_price) ~ removedfrom_prelim_sfha | pin + sale_year^Township,
cluster = ~ pin10,
data =(df_spill )
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_prelim_sfha | pin + sale_year^Township,
cluster = ~ pin10,
data = (df_spill |> filter(change_type_prelim != "Never SFHA" & treat_year_add_prelim == 10000))
)
)
all_models_formatted <- modelsummary::modelsummary(prelim_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
```
```{r}
eff_twfe <- list(
"I" = feols(
log(sale_price) ~ addedto_eff_sfha | pin + sale_year^clean_name,
cluster = ~pin10,
data = (df_spill )
),
"I Subset" = feols(
log(sale_price) ~ addedto_eff_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df_spill |> filter(change_type != "Always SFHA" & treat_year_remove_eff == 10000 ))),
"II" = feols(
log(sale_price) ~ removedfrom_eff_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data =(df_spill )
),
"II Subset" = feols(
log(sale_price) ~ removedfrom_eff_sfha | pin + sale_year^clean_name,
cluster = ~ pin10,
data = (df_spill |> filter(change_type != "Never SFHA" & treat_year_add_eff == 10000))
)
)
all_models_formatted <- modelsummary::modelsummary(eff_twfe,
output = "flextable",
fmt = function(x) round(x, 2),
stars = TRUE)
all_models_formatted
```