Table 1.
General data description.
Fig 1.
a: Distribution of market spreads, as of 4/27/2022 for all non-sinkable bonds that were not sold via private placement and that do have a settle period > 2 months. 1 bp = 0.01%. b: Distribution of spreads at issue for all non-sinkable bonds issued after 2015 that were not sold via private placement and that do have a settle period > 2 months. 1 bp = 0.01%.
Fig 2.
a: Distribution of market spreads, as of 4/27/2022 for all non-sinkable water/sewer revenue bonds that were not sold via private placement and that do have a settle period > 2 months. 1 bp = 0.01%. b: Distribution of spreads at issue for all non-sinkable water/sewer revenue bonds issued after 2015 that were not sold via private placement and that do have a settle period > 2 months. 1 bp = 0.01%.
Table 2.
Dataset characteristics for the whole market data and water and sewer revenue bonds only, for both response variables: market spread and spread at issue.
Fig 3.
a: Count of bonds outstanding, grouped by climate risk (risQ Score) of the issuer, for the whole market. b: Count of bonds outstanding, grouped by climate risk (risQ Score) of the issuer, for water and sewer revenue bonds only.
Fig 4.
a: Count of bonds outstanding, grouped by percent Black of issuer, for the whole market. b: Count of bonds outstanding, grouped by percent Black of issuer, for water and sewer revenue bonds only.
Table 3.
Model variable definitions.
Table 4.
Results for the whole market, using market spread as the response variable.
Model 1 is the “hedonic” model, which includes bond structure only. Model 2 includes bond structure and non-race SES. Model 3 includes bond structure, non-race SES, and climate and race.
Table 5.
Results for water and sewer only, using market spread as the response variable.
Model 1 is the “hedonic” model, which includes bond structure only. Model 2 includes bond structure and non-race SES. Model 3 includes bond structure, non-race SES, and climate and race.
Table 6.
Results for the whole market, using spread at issue as the response variable.
Model 4 is the “hedonic” model, which includes bond structure and market condition. Model 5 includes bond structure, market condition, and non-race SES. Model 6 includes bond structure, market condition, non-race SES, and climate and race.
Table 7.
Results for water and sewer only, using spread at issue as the response variable.
Model 4 is the “hedonic” model, which includes bond structure and market condition. Model 5 includes bond structure, market condition, and non-race SES. Model 6 includes bond structure, market condition, non-race SES, and climate and race.
Fig 5.
Correlation matrix for continuous variables used in regression models.
Fig 6.
Racial credit spread penalties for a 100% Black community, for 1000 model simulations (histogram) versus actual data (vertical line).
The model simulations maintain all real data for each CUSIP, except for race, which is randomly shuffled at the CUSIP6 level.