# RECORD 433778217c0b45ad4ea3ea67a7a55879a3671224 # TITLE: Indoor and outdoor health factors in the pricing of commercial real estate: A hedonic analysis of U.S. office buildings. # YEAR: 2025.0 VENUE: Journal of Environmental Management DOI: 10.1016/j.jenvman.2025.125312 # AUTHORS: Katharina Minkow; F. Fuerst # PAGES: 46 CANDIDATE TABLE PAGES (1-based): [12, 16, 17, 28, 30, 43, 44, 45, 46] # ABSTRACT: Research on the connection between occupant health and the financial performance of commercial buildings is scarce. This study tests the willingness to pay for health-promoting features and estimates the relative contribution of indoor and outdoor features. For this purpose, a hedonic framework is applied to assess the effect of different attributes on prices. Based on a proprietary, assembled dataset of physical and financial characteristics, a sample of health-certified U.S. office buildings is compared to a large sample of non-certified buildings in the same markets. Potential biases are mitigated by statistical methods such as panel data estimation and propensity score matching. Health-certified offices are found to achieve a rental premium of 4-6Β % on average, across model specifications. In addition, proximity to health services and an encouraging environment for active transport modes of commuting tend to increase office rents while outdoor air quality and hospital quality ratings yield mixed results. Comparing the individual price effects of indoor and outdoor health-related features shows that higher active commuting scores or walkability have the strongest positive association with office rents. Overall, both indoor and outdoor health-related features are found to drive rental rates, but it appears that neighborhood characteristics exert a larger cumulative impact on rent than building health certification. Our findings suggest that health aspects of the built environment, although not widely monitored or tracked by owners or tenants, are nonetheless valued and reflected in pricing within the analyzed office markets. ===== PAGE 1 (score 6, numeric-paren 0, table-words 0, var-words 6) ===== 1 Indoor and outdoor health factors in the pricing of commercial real estate: A hedonic analysis of U.S. office buildings Abstract Research on the connection between occupant health and the financial performance of commercial buildings is scarce. This study tests the willingness to pay for health-promoting features and estimates the relative contribution of indoor and outdoor features. For this purpose, a hedonic framework is applied to assess the effect of different attributes on prices. Based on a proprietary, assembled dataset of physical and financial characteristics, a sample of health-certified U.S. office buildings is compared to a large sample of non-certified buildings in the same markets. Potential biases are mitigated by statistical methods such as panel data estimation and propensity score matching. Health-certified offices are found to achieve a rental premium of 4-6% on average, across model specifications. In addition, proximity to health services and an encouraging environment for active transport modes of commuting tend to increase office rents while outdoor air quality and hospital quality ratings yield mixed results. Comparing the individual price effects of indoor and outdoor health-related features shows that higher active commuting scores or walkability have the strongest positive association with office rents. Overall, both indoor and outdoor health-related features are found to drive rental rates, but it appears that neighborhood characteristics exert a larger cumulative impact on rent than building health certification. Our findings suggest that health aspects of the built environment, although not widely monitored or tracked by owners or tenants, are nonetheless valued and reflected in pricing within the analyzed office markets. Keywords: healthy buildings, green buildings, voluntary certification schemes, office rents, air pollution, accessibility to health services, walkability ===== PAGE 12 (score 15, numeric-paren 0, table-words 2, var-words 5) ===== 12 Table 2: Summary Statistics of the Overall Sample in Q2 2022 Control buildings (N=4,435) Health-certified buildings (N=425) Mean Std. Dev. Mean Std. Dev. Diff. in Means Std. Error ANOVA p-value HBS 0.00 0.00 0.43 0.50 0.43 0.02 - HBC 0.00 0.00 0.69 0.46 0.69 0.02 - Rent 41.82 20.08 50.37 23.53 8.55 1.18 0.00 Overall AQI (lagged) 41.15 4.45 40.98 4.63 -0.17 0.23 0.45 Distance to nearest hospital 1.28 1.05 0.99 0.61 -0.29 0.03 0.00 ACS 81.24 20.22 88.30 13.64 7.06 0.73 0.00 Walk Score 87.32 20.20 93.72 13.35 6.40 0.72 0.00 Size 232,275 328,187 660,762 579,563 428,487 28,542 0.00 Number of floors 11.85 11.46 25.52 17.69 13.67 0.87 0.00 GB 0.26 0.44 0.58 0.49 0.32 0.02 - N Pct. N Pct. Hospital rating Three 1,037 23.4 125 29.4 One/Two 918 20.7 96 22.6 Four/Five 1,587 35.8 172 40.5 Not Available 893 20.1 32 7.5 Age yrs<=5 114 2.6 26 6.1 Not Available 137 3.1 27 6.4 550 2,373 53.5 149 35.1 Renovation No Renovation 360 8.1 46 10.8 yrs<=5 317 7.1 26 6.1 520 1,055 23.8 85 20.0 Building class C 427 9.6 5 1.2 A 1,811 40.8 353 83.1 B 2,197 49.5 67 15.8 Note: 12% of the health-certified buildings hold both a standard and a Covid certification. ===== PAGE 13 (score 12, numeric-paren 0, table-words 1, var-words 5) ===== 13 Table 3: Summary Statistics of the Health-certified Building Sample by Certification Type in Q2 2022 WELL Standard (N=42) WELL Covid (N=227) Fitwel Standard (N=147) Fitwel Covid (N=69) Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Rent 52.53 30.93 49.64 23.01 55.26 23.64 51.74 23.32 Overall AQI (lagged) 41.62 4.81 41.10 4.18 40.27 5.10 40.90 4.78 Distance to nearest hospital 0.91 0.40 1.07 0.70 0.88 0.50 0.95 0.56 ACS 89.14 13.03 87.83 14.23 89.90 11.52 88.59 12.70 Walk Score 94.19 14.25 93.23 14.11 94.99 10.77 94.70 11.84 Size 658,998 513,383 725,638 637,405 703,822 662,024 559,333 464,262 Number of floors 27.40 19.13 27.05 18.31 26.59 18.40 22.84 15.43 GB 0.50 0.51 0.58 0.49 0.61 0.49 0.61 0.49 N Pct. N Pct. N Pct. N Pct. Hospital rating Three 15 35.7 63 27.8 51 34.7 17 24.6 One/Two 6 14.3 43 18.9 37 25.2 23 33.3 Four/Five 16 38.1 103 45.4 52 35.4 24 34.8 Not Available 5 11.9 18 7.9 7 4.8 5 7.2 Age yrs<=5 3 7.1 11 4.8 11 7.5 3 4.3 Not Available 3 7.1 13 5.7 9 6.1 4 5.8 550 12 28.6 88 38.8 54 36.7 20 29.0 Renovation No Renovation 2 4.8 28 12.3 14 9.5 6 8.7 yrs<=5 2 4.8 15 6.6 6 4.1 7 10.1 520 4 9.5 49 21.6 32 21.8 12 17.4 Building class C 0 0.0 3 1.3 2 1.4 0 0.0 A 35 83.3 184 81.1 128 87.1 59 85.5 B 7 16.7 40 17.6 17 11.6 10 14.5 Note: WELL Covid refers to the WELL Health-Safety Rating. Fitwel Covid refers to the Fitwel Viral Response Module. The sum of health-certified buildings split by certification type is greater than the number of buildings holding any health-related certification, because some buildings hold more than one certification. ===== PAGE 16 (score 22, numeric-paren 0, table-words 4, var-words 4) ===== 16 of non-health-certified buildings, it may be the case that we measure other effects than the treatment effect. Using the PSM approach, the heterogeneity between treatment and control group can be reduced for these observable characteristics. According to Rosenbaum & Rubin (1983) the propensity score equals the conditional probability that a subject receives treatment given a bundle of observable characteristics. Following Rosenbaum & Rubin (1983), propensity scores based on general building characteristics e(BC) in this study were estimated via the following logistic regression, in which the probability of receiving any type of health-related certification over the entire observed timeframe (HBX=1) is conditional on a vector of observed building characteristics (BC): 𝑒(𝐡𝐢) = π‘ƒπ‘Ÿ(𝐻𝐡𝑋= 1|𝐡𝐢) The building characteristics included are size, number of floors, building class, age, and renovation. Based on the estimated propensity scores, we used the nearest neighbor approach with a ratio of 1:2 so that for each healthy building observed at time t two non-health-certified building observations were selected into the control group. Table 4 shows the standardized mean differences of the selected building characteristics before and after PSM. A graphical representation of the PSM results is depicted in Appendix B. The propensity score is shown as distance for treatment and control group. As observed for the cross-section before, the large standardized mean differences for the full sample demonstrate that the groups of health-certified and non-health-certified buildings differ significantly from each other, especially in terms of building class, number of floors, and size. The PSM procedure reduces the standardized mean differences below 0.1 for most covariates and below 0.2 for the number of floors and size, which suggests that the treatment and control group are well- balanced after PSM (e.g. Cohen, 1977). That some variables show greater balancing improvements post-matching than others can be explained by greater differences in the variable distributions between treatment and control group pre-matching. For example, as presented in Table 2, the means and standard deviations of size are substantially different for the group of health-certified buildings and the control group, which can challenge balancing. Table 4: PSM Summary Balance All data Matched data Means HB Means Control Std. Mean Diff. Means HB Means Control Std. Mean Diff. Distance 0.21 0.07 0.83 0.21 0.19 0.12 Size 677,971 234,164 0.76 677,971 576,804 0.17 Number of floors 26.29 12.24 0.80 26.29 24.44 0.11 Building class C 0.01 0.10 -0.77 0.01 0.01 0.03 A 0.83 0.40 1.13 0.83 0.85 -0.05 B 0.16 0.50 -0.93 0.16 0.14 0.04 Age yrs<=5 0.04 0.03 0.07 0.04 0.05 -0.02 Not Available 0.06 0.03 0.12 0.06 0.06 0.01 550 0.28 0.51 -0.50 0.28 0.27 0.03 ===== PAGE 17 (score 24, numeric-paren 0, table-words 2, var-words 7) ===== 17 All data Matched data Means HB Means Control Std. Mean Diff. Means HB Means Control Std. Mean Diff. Renovation No Renovation 0.09 0.09 0.03 0.09 0.11 -0.04 yrs<=5 0.07 0.07 -0.01 0.07 0.07 0.01 520 0.12 0.16 -0.12 0.12 0.11 0.02 Sample sizes Control HB All 181,966 17,143 Matched 34,286 17,143 Unmatched 147,680 0 Note: The distance shows the propensity score for the treatment group of health-certified buildings and the control group of non-health- certified buildings. Table 5 presents the summary statistics for our main variables of interest for the PSM sample in Q2 2022, including properties without any healthy building certification, with any healthy building certification, with any standard certification, and with any Covid certification. As observed for the full sample, the health-certified buildings show a greater mean annual effective rent than the buildings without any healthy building label in the PSM sample. The descriptive statistics indicate that even after balancing the groups based on general building characteristics, health-certified buildings are situated at locations with on average closer access to health services and a better walkability. However, the summary statistics do not suggest a difference in the mean outdoor air quality between health-certified offices and non-health-certified offices. One-way ANOVA tests are performed for the cross-section of the PSM sample to investigate whether there is a significant difference in the mean values of each of the outdoor health-related variables for the groups of buildings with any HB standard certification versus buildings without any HB standard certification and for buildings with any HB Covid certification versus buildings without any HB Covid certification. Table 5: Summary Statistics of the PSM Sample by Certification Type in Q2 2022 Without any HB certificate (N=852) HB (N=425) HB Standard (N=184) HB Covid (N=294) Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Rent 43.87 21.63 50.37 23.53 54.54 25.31 50.07 22.98 Overall AQI (lagged) 41.52 3.71 40.98 4.63 40.57 5.12 41.04 4.32 O3 AQI (lagged) 41.28 3.91 40.74 4.91 40.30 5.50 40.80 4.57 SO2 AQI (lagged) 1.03 1.05 1.05 1.06 1.09 1.09 1.06 1.06 CO AQI (lagged) 5.21 2.06 4.47 1.66 4.29 1.58 4.58 1.68 NO2 AQI (lagged) 20.97 5.28 19.81 6.04 19.13 6.36 20.16 6.00 PM2.5 AQI (lagged) 33.12 5.33 32.64 5.04 32.66 4.81 32.60 5.00 PM10 AQI (lagged) 17.92 5.79 18.01 5.74 18.00 5.86 18.06 5.75 Distance to nearest hospital 1.23 0.98 0.99 0.61 0.89 0.48 1.04 0.67 ACS 83.05 18.64 88.30 13.64 89.61 11.98 88.03 13.90 Walk Score 89.51 18.69 93.72 13.35 94.70 11.73 93.55 13.65 Bike Score 75.71 17.92 80.15 13.81 80.96 13.15 80.21 13.58 N Pct. N Pct. N Pct. N Pct. Hospital rating Three 196 23.0 125 29.4 64 34.8 80 27.2 One/Two 157 18.4 96 22.6 43 23.4 66 22.4 Four/Five 373 43.8 172 40.5 66 35.9 125 42.5 Not Available 126 14.8 32 7.5 11 6.0 23 7.8 ===== PAGE 18 (score 9, numeric-paren 0, table-words 2, var-words 5) ===== 18 Without any HB certificate (N=852) HB (N=425) HB Standard (N=184) HB Covid (N=294) Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Note: HB Covid refers to the WELL Health-Safety Rating and the Fitwel Viral Response Module. The sum of health-certified buildings split by certification type is greater than the number of buildings holding any health-related certification, because some buildings hold more than one certification. One drawback of the PSM approach is that it only accounts for observable characteristics. Therefore, we use panel data models in order to additionally address some of the unobservable heterogeneity. One option are fixed effects (FE) models. Including fixed effects allows to control for two types of unobservable attributes (Stock & Watson, 2019). Firstly, individual fixed effects allow to account for unobserved characteristics which are building-specific but time-consistent (Stock & Watson, 2019). Secondly, time fixed effects allow to control for unobservable attributes which are equal for all buildings but differ over time (Stock & Watson, 2019). One of the main drawbacks of fixed effects models is that time-invariant characteristics are neutralized by the use of fixed effects. This is particularly problematic in hedonic models as many time-constant or near-constant characteristics (e.g. location-related variables and stable occupancy rates) are important price determinants and we are particularly interested in observing and interpreting these coefficients. Healthy building certifications are not just issued for existing properties, but often for new developments so that for these buildings no change in the certification indicator variables can be observed over time. Another option are random effects (RE) models, which address the mentioned problems with fixed effects models. Random effects models can be estimated under the assumption that the unobserved effect is random and therefore not correlated with any of the regressor variables (Wooldridge, 2016). Model 1 is a random effects model and is used to investigate whether health-promoting offices achieve a rental premium compared to non-health-promoting offices, holding other characteristics constant. ln 𝑅𝑖𝑑= 𝛽0 + 𝛽1𝐻𝐡𝑖𝑑+ βˆ‘π›½2,𝑙𝐡𝐢𝑙,𝑖𝑑 𝐿 𝑙=1 + 𝛽3𝐺𝐡𝑖𝑑+ βˆ‘π›½4,π‘šπ‘†π‘š,𝑖 𝑀 π‘š=1 + βˆ‘π›½5,𝑛𝑀𝐼𝑛,𝑖𝑑+ 𝛼𝑖+ 𝛿𝑑+ πœ€π‘–π‘‘ 𝑁 𝑛=1 (1) The dependent variable is the natural logarithm of the annual effective rent per sqf in USD. The variable of interest is the indicator variable healthy building (HB), which takes value 1, if the building is WELL and/or Fitwel certified or pre-certified in quarter t. This takes into account all different versions of certifications, including Covid-related certifications. Further, this model includes several time-variant and time-consistent control variables covering general building characteristics (BC), such as the age based on the year of construction and the economic age based on the year of renovation, green building certifications (GB), submarkets (S), and macroeconomic indicators (MI). In this model specification the control for green buildings consists of a single indicator variable taking a value of 1, if the building holds a LEED and/or Energy Star certification in quarter t and 0, otherwise. As suggested by the Breusch- Pagan Lagrange Multiplier test (Breusch & Pagan, 1980) and an F test, time dummies (Ξ΄t) are included in addition to individual effects (Ξ±i). A full description of variables is shown in Table 1. The Hausman test indicates that the fixed effects model is more appropriate than the random effects model (Hausman, 1978). However, random effects is still preferable for this study as ===== PAGE 28 (score 92, numeric-paren 54, table-words 6, var-words 14) ===== 28 Table 6: Regression Results - Selected Variables Dependent variable: ln (Rent) (1) RE PSM sample, all HB (2a) RE full sample, HBS/HBC & OHBC (2b) RE PSM sample, HBS/HBC & OHBC (3) RE PSM sample, HBS by provider & OHBC (4a) RE full sample, HBS by provider & OHBC detailed (4b) RE PSM sample, HBS by provider & OHBC detailed HB 0.037*** HBS 0.041** 0.056*** HBC -0.002 0.012 WELL Standard 0.053+ 0.045 0.049 Fitwel Standard 0.048*** 0.046** 0.048** Overall AQI (lagged) 0.001*** 0.001*** 0.001*** O3 AQI (lagged) 0.002*** 0.001*** SO2 AQI (lagged) 0.009*** 0.010*** CO AQI (lagged) 0.012*** 0.012*** NO2 AQI (lagged) -0.005*** -0.004*** PM2.5 AQI (lagged) 0.002*** 0.001*** PM10 AQI (lagged) 0.000* 0.000 Distance to nearest hospital -0.025*** -0.009 -0.010 -0.006 -0.012 Hospital rating: 1-2 0.041* 0.019 0.019 0.064* -0.003 Hospital rating: 4-5 0.044** -0.014 -0.014 0.070** -0.017 Hospital rating: NA 0.059*** 0.013 0.013 0.120*** 0.070 Interaction: Distance to nearest hospital * Hospital rating: 1-2 -0.014 0.000 Interaction: Distance to nearest hospital * Hospital rating: 4-5 -0.021+ 0.014 Interaction: Distance to nearest hospital * Hospital rating: NA -0.044*** -0.027 ACS 0.003*** 0.003** 0.003** Walk Score 0.002*** 0.001 Bike Score 0.001 0.000 Age yes yes yes yes yes yes Renovation yes yes yes yes yes yes ln (Size) 0.024** -0.003 0.027** 0.027** -0.001 0.013 ln (Number of floors) -0.001 -0.004 -0.008 -0.008 0.001 0.011 Building class: A 0.039 0.023 0.042 0.042 0.025 0.145* Building class: B -0.011 0.006 -0.010 -0.010 0.003 0.089 GB -0.014 -0.028*** -0.013 -0.013 LEED -0.042*** -0.028** Energy Star -0.019* -0.012 Submarkets yes yes yes yes yes yes ln (Office space) 0.015 0.017*** 0.003 0.003 0.019*** 0.010 ln (UR) -0.141*** -0.134*** -0.139*** -0.138*** -0.113*** -0.136*** Time yes yes yes yes yes yes Constant 2.812*** 2.957*** 2.793*** 2.792*** 2.731*** 2.712*** Num.Obs. 51,429 199,109 51,429 51,429 168,759 42,471 R2 0.704 0.558 0.705 0.704 0.598 0.717 Adj. R2 0.703 0.558 0.704 0.703 0.597 0.716 AIC -81,588.2 -256,172.2 -81,763.9 -81,722.0 -226,393.5 -67,720.6 BIC -79,871.7 -253,938.1 -79,985.5 -79,943.6 -224,105.3 -65,885.4 RMSE 0.11 0.13 0.11 0.11 0.12 0.11 Std.Errors HC3 HC3 HC3 HC3 HC3 HC3 Note: *** significant at the 0.1% level, ** significant at the 1% level, * significant at the 5% level, + significant at the 10% level. The information shown in this table is based on data from CompStak, Inc. (financial and building characteristics), WELL (healthy building certifications), Fitwel (healthy building certifications), U.S. EPA (air quality), CMS (hospitals), Walk Score (walkability), LEED (green building certifications), Energy Star (green building certifications), and the U.S. Bureau of Labor Statistics (unemployment rate). ===== PAGE 29 (score 9, numeric-paren 0, table-words 3, var-words 3) ===== 29 The findings of Model 1 demonstrate that indoor health-promoting characteristics increase the WTP for office space without considering outdoor health-promoting characteristics. In addition, the results of Model 2, Model 3, and Model 4 indicate that certain outdoor health- related characteristics as the accessibility to health services and walkability are important factors in determining office rents. When the OHBC were included, the healthy building certifications remain to show a positive and statistically significant impact on rents across all model specifications for standard certifications in general and for the Fitwel standard certification in particular. Therefore, the results indicate that indoor and outdoor health- promoting features cannot be substituted fully by each other and that a combination of both adds the most financial value. On one hand, the healthy indoor environments are important to boost productivity, health and well-being of employees who spend most of their working day inside the office building. On the other hand, the health-related characteristics at the neighborhood level matter as they for example affect the way how employees commute to work or spend their lunch breaks and can also affect the indoor environmental quality. It is difficult to provide a clear ranking of which indoor and outdoor health-related features are most important due to the limitations of this study. Table 7 provides selected estimates for Models 2-4 following the standardization approach by Gelman (2008). For each model, the coefficients are ranked by absolute value. The rankings show mixed results across model specifications, but general trends are discernible. For example, the results indicate that the ACS (standardized coefficient: 0.107-0.113) or walkability (standardized coefficient: 0.023-0.064) have the strongest effect on office rents. Furthermore, if we directly compare the outdoor air quality and healthy building labels, we observe that the indoor health-related features tend to be a larger driver of office rents than outdoor air quality. In particular, a change in the HB certification variables from 0 to 1 is associated with a greater change in rents than a change in the air quality measures by two standard deviations in most model specifications. This finding may indicate that indoor features/certifications are more important for tenants than outdoor air quality. However, the fundamental question arises whether consumers are really aware of factors such as the outdoor air quality and consider this in their consumption decisions. As outlined in section 2 the outdoor air quality can affect the indoor air quality substantially (Allen & Macomber, 2020; Chen & Zhao, 2011), which may be considered when implementing features to improve the indoor air quality. However, our results indicate that the outdoor air quality may not influence the willingness to pay as much as healthy building labels, the accessibility to health services, and walkability. This could be attributable to the easier observability of the latter factors in contrast to outdoor air quality. Moreover, it should be considered that the healthy building certifications measure a bundle of many indoor features and entail a signaling effect, so that comparing the underlying individual features to the individual outdoor features would be a more suitable comparison. To better understand which exact indoor or outdoor health-related characteristics affect rents, future research should include the underlying indoor attributes in the regression instead of using certifications as a proxy when data hereto becomes available. Within the scope of this study, we can bundle the observed OHBC and compare their joint impact on rents in comparison to the HB certifications. Considering outdoor air quality, accessibility to health services, and walkability together, we conclude that health-related location parameters are a greater driver of office rents than indoor HB features as measured by HB certifications. ===== PAGE 30 (score 41, numeric-paren 24, table-words 2, var-words 10) ===== 30 Table 7: Ranked (Standardized) Health-related Coefficients Dependent variable: ln (Rent) Rank (2a) RE full sample, HBS/HBC & OHBC (2b) RE PSM sample, HBS/HBC & OHBC (3) RE PSM sample, HBS by provider & OHBC (4a) RE full sample, HBS by provider & OHBC detailed (4b) RE PSM sample, HBS by provider & OHBC detailed 1 ACS 0.113*** ACS 0.107** ACS 0.107** NO2 AQI (lagged) -0.068*** NO2 AQI (lagged) -0.055*** 2 Distance to nearest hospital -0.050*** HBS 0.056*** WELL Standard 0.053+ Walk Score 0.064*** CO AQI (lagged) 0.051*** 3 HBS 0.041** Distance to nearest hospital -0.016 Fitwel Standard 0.048*** CO AQI (lagged) 0.049*** WELL Standard 0.049 4 Overall AQI (lagged) 0.016*** HBC 0.012 Distance to nearest hospital -0.016 Fitwel Standard 0.046** Fitwel Standard 0.048** 5 HBC -0.002 Overall AQI (lagged) 0.011*** Overall AQI (lagged) 0.011*** WELL Standard 0.045 SO2 AQI (lagged) 0.048*** 6 SO2 AQI (lagged) 0.041*** O3 AQI (lagged) 0.026*** 7 O3 AQI (lagged) 0.038*** Walk Score 0.023 8 Bike Score 0.027 Distance to nearest hospital -0.022 9 PM2.5 AQI (lagged) 0.022*** PM2.5 AQI (lagged) 0.014*** 10 Distance to nearest hospital -0.012 Bike Score 0.006 11 PM10 AQI (lagged) -0.005* PM10 AQI (lagged) 0.002 Note: *** significant at the 0.1% level, ** significant at the 1% level, * significant at the 5% level, + significant at the 10% level. The information shown in this table is based on data from CompStak, Inc. (financial and building characteristics), WELL (healthy building certifications), Fitwel (healthy building certifications), U.S. EPA (air quality), CMS (hospitals), Walk Score (walkability), LEED (green building certifications), Energy Star (green building certifications), and the U.S. Bureau of Labor Statistics (unemployment rate). ===== PAGE 31 (score 10, numeric-paren 0, table-words 0, var-words 10) ===== 31 6. Conclusion It is well established in the extant literature that indoor environmental quality is a critical mediator and direct cause of a number of health conditions. The financial and economic dimensions of health-supporting building features are less well understood. As building owners require a financial mechanism for recouping voluntary investments into these features, higher rental income and, in turn, higher asset values are key features of a business case for healthy buildings. However, the outdoor and neighborhood environment, although largely beyond the control of building owners, is also a major driver of both financial success and health outcomes. This study investigated the financial performance of health-promoting indoor and outdoor building features by using hedonic pricing models on a sample of health-certified office properties and a control sample of non-health-certified offices in 12 U.S. metropolitan areas. The analysis was performed on a unique and proprietary dataset, which contains rental data and building characteristics from CompStak, public data from the certification schemes WELL und Fitwel, as well as data on the air quality, accessibility to health services, walkability, green building certification schemes, and macroeconomic indicators from various sources. In addition to the standard certifications from the named providers, the Covid-related certifications were also considered and analyzed. The results of the panel data and PSM analysis show that office buildings in the U.S. with a healthy building label achieve on average a higher effective rent than comparable offices without a healthy building certification. This rental premium is statistically significant across all model specifications and ranges from 4-6%. Moreover, the present findings indicate that some neighborhood health-related characteristics play at least an equally important role in rent formation as on-site features. For example, the proximity to health services has a significant positive effect on rents. Furthermore, active modes of transportation such as walking, biking, or partly travelling by public transport are associated with a significant increase in rents in these locations. However, results are mixed with regards to outdoor health-related characteristics for certain factors. For instance, a 1-point increase in measured air pollution is associated with a rent increase of 0.1% and hospital quality does not have a clear positive pricing impact. While office buildings close to a hospital with the highest quality rating of 4-5 tend to achieve higher rents than offices close to a hospital with a rating of 3, we do not observe the expected increase per score level. Regarding the question which health-related certification characteristics may be associated with a higher rental premium, the results suggest that these pertain mainly to standard certifications rather than specialized certifications targeting only specific aspects such as hygiene or infection risk. In particular, the Fitwel standard certification shows a statistically significant positive impact on office rents across all model specifications. Overall, we conclude that the impact of outdoor health-promoting features is additive to the value of indoor health-promoting features. Therefore, both aspects should be considered as important price determinants in the context of a health-supporting office space. However, the mentioned mixed results leave some questions regarding the outdoor air quality and quality of health services unanswered. Comparing the sum of the observable outdoor and indoor health- related characteristics, the findings indicate a stronger price effect of the location than the building. The greatest individual impact on rents was observed for the walkability of the neighborhood. ===== PAGE 43 (score 67, numeric-paren 56, table-words 1, var-words 3) ===== 43 Appendix C Regression Results - Extended Dependent variable: ln (Rent) (1) RE PSM sample, all HB (1b) FE PSM sample, all HB (2a) RE full sample, HBS/HBC & OHBC (2b) RE PSM sample, HBS/HBC & OHBC (3) RE PSM sample, HBS by provider & OHBC (4a) RE full sample, HBS by provider & OHBC detailed (4b) RE PSM sample, HBS by provider & OHBC detailed HB 0.037*** 0.037*** (0.009) (0.009) HBS 0.041** 0.056*** (0.013) (0.013) HBC -0.002 0.012 (0.010) (0.010) WELL Standard 0.053+ 0.045 0.049 (0.030) (0.031) (0.031) Fitwel Standard 0.048*** 0.046** 0.048** (0.014) (0.015) (0.015) Overall AQI (lagged) 0.001*** 0.001*** 0.001*** (0.000) (0.000) (0.000) O3 AQI (lagged) 0.002*** 0.001*** (0.000) (0.000) SO2 AQI (lagged) 0.009*** 0.010*** (0.001) (0.001) CO AQI (lagged) 0.012*** 0.012*** (0.001) (0.002) NO2 AQI (lagged) -0.005*** -0.004*** (0.000) (0.001) PM2.5 AQI (lagged) 0.002*** 0.001*** (0.000) (0.000) PM10 AQI (lagged) 0.000* 0.000 (0.000) (0.000) Distance to nearest hospital -0.025*** -0.009 -0.010 -0.006 -0.012 (0.006) (0.012) (0.012) (0.010) (0.020) Hospital rating: 1-2 0.041* 0.019 0.019 0.064* -0.003 ===== PAGE 44 (score 138, numeric-paren 121, table-words 1, var-words 2) ===== 44 Dependent variable: ln (Rent) (1) RE PSM sample, all HB (1b) FE PSM sample, all HB (2a) RE full sample, HBS/HBC & OHBC (2b) RE PSM sample, HBS/HBC & OHBC (3) RE PSM sample, HBS by provider & OHBC (4a) RE full sample, HBS by provider & OHBC detailed (4b) RE PSM sample, HBS by provider & OHBC detailed (0.017) (0.034) (0.034) (0.026) (0.053) Hospital rating: 4-5 0.044** -0.014 -0.014 0.070** -0.017 (0.014) (0.024) (0.024) (0.023) (0.042) Hospital rating: NA 0.059*** 0.013 0.013 0.120*** 0.070 (0.017) (0.030) (0.030) (0.027) (0.050) Interaction: Distance to nearest hospital * Hospital rating: 1-2 -0.014 0.000 (0.013) (0.023) Interaction: Distance to nearest hospital * Hospital rating: 4-5 -0.021+ 0.014 (0.012) (0.021) Interaction: Distance to nearest hospital * Hospital rating: NA -0.044*** -0.027 (0.012) (0.023) ACS 0.003*** 0.003** 0.003** (0.001) (0.001) (0.001) Walk Score 0.002*** 0.001 (0.000) (0.001) Bike Score 0.001 0.000 (0.000) (0.001) Age: 550 -0.124*** -0.155** -0.325*** -0.127*** -0.127*** -0.341*** -0.161*** (0.032) (0.048) (0.020) (0.032) (0.032) (0.021) (0.033) Age: NA 0.032 0.035 0.041** 0.033 0.033 0.035** 0.018 (0.021) (0.022) (0.013) (0.020) (0.020) (0.013) (0.020) Renovation: yrs<=5 -0.005 -0.004 0.004 -0.004 -0.004 0.007 -0.007 (0.012) (0.013) (0.007) (0.012) (0.012) (0.007) (0.012) Renovation: 520 -0.001 0.000 0.005 -0.001 0.000 0.005 0.005 (0.018) (0.021) (0.011) (0.018) (0.018) (0.011) (0.020) ln (Size) 0.024** -0.003 0.027** 0.027** -0.001 0.013 (0.009) (0.005) (0.009) (0.009) (0.005) (0.010) ln (Number of floors) -0.001 -0.004 -0.008 -0.008 0.001 0.011 (0.013) (0.008) (0.013) (0.013) (0.008) (0.014) Building class: A 0.039 0.023 0.042 0.042 0.025 0.145* (0.052) (0.018) (0.052) (0.052) (0.018) (0.068) Building class: B -0.011 0.006 -0.010 -0.010 0.003 0.089 (0.048) (0.012) (0.048) (0.048) (0.013) (0.066) GB -0.014 -0.016 -0.028*** -0.013 -0.013 (0.009) (0.010) (0.007) (0.009) (0.009) LEED -0.042*** -0.028** (0.007) (0.009) ===== PAGE 46 (score 62, numeric-paren 37, table-words 5, var-words 9) ===== 46 Dependent variable: ln (Rent) (1) RE PSM sample, all HB (1b) FE PSM sample, all HB (2a) RE full sample, HBS/HBC & OHBC (2b) RE PSM sample, HBS/HBC & OHBC (3) RE PSM sample, HBS by provider & OHBC (4a) RE full sample, HBS by provider & OHBC detailed (4b) RE PSM sample, HBS by provider & OHBC detailed Energy Star -0.019* -0.012 (0.008) (0.009) Submarkets yes no yes yes yes yes yes ln (Office space) 0.015 0.017*** 0.003 0.003 0.019*** 0.010 (0.010) (0.005) (0.010) (0.010) (0.005) (0.011) ln (UR) -0.141*** -0.140*** -0.134*** -0.139*** -0.138*** -0.113*** -0.136*** (0.014) (0.014) (0.008) (0.014) (0.014) (0.008) (0.015) Time yes yes yes yes yes yes yes Constant 2.812*** 2.957*** 2.793*** 2.792*** 2.731*** 2.712*** (0.184) (0.092) (0.179) (0.179) (0.106) (0.213) Num.Obs. 51,429 51,429 199,109 51,429 51,429 168,759 42,471 R2 0.704 0.402 0.558 0.705 0.704 0.598 0.717 Adj. R2 0.703 0.375 0.558 0.704 0.703 0.597 0.716 AIC -81,588.2 -84,407.2 -256,172.2 -81,763.9 -81,722.0 -226,393.5 -67,720.6 BIC -79,871.7 -83,841.0 -253,938.1 -79,985.5 -79,943.6 -224,105.3 -65,885.4 RMSE 0.11 0.11 0.13 0.11 0.11 0.12 0.11 Std.Errors HC3 HC3 HC3 HC3 HC3 HC3 HC3 Note: *** significant at the 0.1% level, ** significant at the 1% level, * significant at the 5% level, + significant at the 10% level. The standard errors are shown in brackets. The information shown in this table is based on data from CompStak, Inc. (financial and building characteristics), WELL (healthy building certifications), Fitwel (healthy building certifications), U.S. EPA (air quality), CMS (hospitals), Walk Score (walkability), LEED (green building certifications), Energy Star (green building certifications), and the U.S. Bureau of Labor Statistics (unemployment rate).