# RECORD 291ca9688b933cfaf7d01e6f6552e428bb2432b5
# TITLE: Occupant Well-Being and House Values
# YEAR: 2018.0  VENUE: nan  DOI: 10.36934/wecon2019-023
# AUTHORS: R. Rijnks; S. Sheppard
# PAGES: 38  CANDIDATE TABLE PAGES (1-based): [17, 23, 24, 25, 26]
# ABSTRACT: A difficulty in identifying the contribution of structure and neighborhood attributes to the market value of residential property is the lack of data on subjective characteristics of the neighborhood (friendliness of neighbors, proximity to friends and acquaintances) or difficult-to-observe subjective attributes of the structure itself (such as "curb appeal" or the presence of unpleasant odors). Concern may also arise from the understanding that the observed market price of most residential property is the result of a process of bargaining. A buyer who is optimistic by nature may assume that the quality of the neighborhood will be wonderful, or that the unusual odor will eventually go away, and therefore be willing to bid a higher price for the structure than a prospective buyer who is more nervous about all the ways that a house purchase can generate disappointment. Estimates of the value of structure or neighborhood attributes may tell us as much about the emotional affect of the buyer as they do about the actual costs or benefits of the attributes (or of cleaning or mitigating them). These observations suggest that incorporating data on the levels of subjective well being (SWB) and emotional affect of the buyers might be usefully applied to improve hedonic analysis of housing markets. The goal of this paper is to undertake such analysis and to explore the potential for improved analysis of the value of residential property. We make use of unique data collected as part of a multi-year analysis of health outcomes, matched with data on market transactions of residential property in three provinces of the Netherlands. We employ the spatial model developed by Kelejian and Prucha (2010) which allows us to incorporate a spatial error specification, as well explicitly control for possible endogeneity between the measure of SWB and the transaction price. By examining aggregate measures of SWB at different spatial scales, we obtain insights into whether these measurements are capturing subjective characteristics of the community, the neighborhood or the structure and the buyer who negotiateover the eventual price.


===== PAGE 1 (score 8, numeric-paren 0, table-words 1, var-words 6) =====

Occupant Well-Being and House Values †
Richard H. Rijnks‡†
Stephen Sheppard §‡
August 29, 2020
Abstract
Recent research indicates that subjective evaluation of well-being increases when conditions of housing are improved.
This suggests that subjective well-being might serve as a useful proxy for characteristics of a home or neighbourhood
that are relevant to an occupant, but unobserved by the analyst. In this paper, we assess this idea through analysis
of residential property valuation, using a sample of 95,413 respondents matched to house sales for 2000 to 2012 in the
North of the Netherlands. Using a spatial econometric approach, we find a significant and positive association between
individual and regional subjective well-being and house prices. This suggests that house buyers are willing to pay more
for, or that house sellers require greater compensation to sell and move from, properties and areas in which the resident
experiences greater happiness. Our study provides the first estimates of the importance of these effects.
Keywords: Subjective well-being, house price, hedonic model
JEL Classification: D91, R2, R3
Author(s) Conflict-of-interest: None
†We would like to thank Erik Johnson for helpful comments, and to acknowledge funding support from the University of Groningen, Campus
Fryslân, and the Waddenacademie
‡Department of Economic Geography, University of Groningen, Groningen, Netherlands
§Department of Economics, Williams College, Williamstown, Massachusetts


===== PAGE 17 (score 49, numeric-paren 0, table-words 13, var-words 18) =====

Table 3: Descriptive statistics for model variables
Variable name
Mean
Std Dev
Min
Max
Property sales price
200,311
96,523
18501
2450000
Ln(sales price)
12.118
0.412
9.826
14.712
Ln(positive affect)
1.511
0.095
0.693
1.792
Ln(well-being+1)
4.366
0.207
0
4.615
Ln(lag well-being+1)
4.381
0.057
3.219
4.615
Ln(urban distance)
2.443
1.154
-4.216
3.862
Ln(density)
5.904
0.942
4.007
7.797
Ln(foreign+1)
1.633
0.469
0.693
2.485
Ln(burglary+1)
3.294
0.508
1.902
4.055
Ln(floor area+1)
4.813
0.290
3.555
6.223
Ln(lotsize+1)†
5.399
1.742
0
11.195
Ln(rooms)
1.541
0.260
0
3.526
Dichotomous Variables
Two Bathrooms
0.056
0.0019
0
1
Three Bathrooms
0.001
0.0002
0
1
Balcony
0.126
0.332
0
1
Parking
0.605
0.489
0
1
Basement
0.936
0.245
0
1
Attic
0.344
0.475
0
1
Under Roof Area
0.143
0.350
0
1
Listed
0.004
0.065
0
1
Monument
0.007
0.082
0
1
The population density of the municipality containing the house is included as an additional control for effects of smaller
population centers (rural towns and large villages).
The official classification of “Non-western foreigners” in Dutch statistics is people who have a migration history (their
own or at least one of their parents) from Africa, Latin-America, Asia, or Turkey, excluding Indonesia and Japan, as described
in Dutch National Statistics (CBS) (2018a). We include 1 plus the percentage of population with a non-western background
in the municipality containing the house as a proxy for ethnic diversity in the neighbourhood.
From Table 4 we see that just over 60% of the properties in the sample were constructed in the 1970’s or later. The
median sales date of properties in the sample is mid-2006.
Operationalisation of main variables
The use of subjective measures of happiness has seen tremendous progress over recent decades, although the field has yet to
reach consensus on which measures are preferable, as noted by Frey and Stutzer (2002). Generally, three distinct constructs
of happiness are evaluated which capture separate but complementary information on overall subjective well-being. The first
component of overall subjective well-being is the general happiness or general life satisfaction component. This is measured
either through a single survey item (e.g. all things considered, how satisfied are you with your life these days?) or preferably
16


===== PAGE 18 (score 7, numeric-paren 0, table-words 1, var-words 1) =====

Table 4: Descriptive statistics for indicator variables
Variable
Mean
Variable
Mean
Interior maintenance
Exterior maintenance
Interior 1 (Poor)
0.10
Exterior 1 (Poor)
0.11
Interior 2
0.07
Exterior 2
0.02
Interior 3
0.66
Exterior 3
0.48
Interior 4
0.24
Exterior 4
0.26
Interior 5
5.18
Exterior 5
4.84
Interior 6
2.79
Exterior 6
3.06
Interior 7
77.24
Exterior 7
78.33
Interior 8
1.47
Exterior 8
1.41
Interior 9 (Excellent)
12.25
Exterior 9 (Excellent)
11.50
Construction period
Year property sold
Up to 1905
4.35
Y2000
5.47
1906-1930
9.65
Y2001
6.91
1931-1944
7.39
Y2002
7.98
1945-1959
5.18
Y2003
8.05
1960-1970
12.98
Y2004
8.79
1971-1980
22.51
Y2005
10.24
1981-1990
13.81
Y2006
11.05
1991-2000
18.82
Y2007
9.51
2000 and later
5.31
Y2008
8.64
Y2009
6.95
Insulation
Y2010
6.43
No insulation
9.99
Y2011
5.53
1 type of insulation
32.52
Y2012
4.45
2 types of insulation
13.58
3 types of insulation
12.48
4 types of insulation
10.94
Fully insulated
20.50
a compound measure using a set of questions as suggested in Kahneman and Krueger (2006). The general happiness or
general life satisfaction component is used to measure a cognitive evaluation of the quality of life by the respondent.
The measurement used for this study is drawn from the SF-36 item short form survey. The SF-36 survey is one of the
most widely used approaches to measuring health related quality of life, and is surveyed in Hays and Morales (2001). The
survey groups 36 items into eight separate constructs. Seven of these deal with a variety of issues relating to physical health,
subjective health, social functioning, and experienced limitations and pain. The construct from SF-36 we use in this study is
Emotional well-being. Emotional well-being is measured on a 0 to 100 scale by weighting and adding together the underlying
items, which are the extent an individual felt happy, nervous, depressed, calm, and downhearted over the past four weeks.
The responses then averaged for each individual. The measure is tested for internal reliability (Cronbach’s Alpha = 0.83),
showing that the underlying items are reliably correlated. Spatial lags of this variable were considered at 1, 2.5, 5 and 10
kilometers, using an inverse euclidean distance weighting.
17


===== PAGE 23 (score 24, numeric-paren 9, table-words 6, var-words 0) =====

Table 6: Subjective well-being and time in residence
Coefficients:
Years
Days
Constant
2.8497***
2.8530***
σ
0.0522
0.0547
Ln(Affect + 1)
0.7642***
0.7710***
σ
0.0167
0.0173
Ln(Price)
0.0294***
0.0284***
σ
0.0039
0.0041
TiR
−1.4867 × 10−3
−5.5870 × 10−6
σ
1.5609 × 10−3
5.7080 × 10−6
TiR2
3.5530 × 10−4**
3.0070 × 10−9**
σ
1.4390 × 10−4
1.3180 × 10−9
F(4, 14852)
562.7***
527.6***
RSE
0.1931
0.1943
R2
0.1316
0.1943
¯R2
0.1314
0.1324
*** - p < 0.001, ** - p<0.01, * - p<0.05, † - p<0.10
spatial econometric approaches to account for spatially correlated errors, spatial autoregressive structures, and spatially
lagged effects of subjective well-being. The dependent variable in all models is the log of the transaction price.
22


===== PAGE 24 (score 102, numeric-paren 64, table-words 9, var-words 6) =====

Table 7: Hedonic model estimates
Variable
OLS Model 1
OLS Model 2
OLS Model 3
IV Health
IV Health/Pain
GNS
SAC
Behavioural and Subjective Characteristics
Well-being
0.0385***
0.181***
0.1744***
0.0893***
0.1149***
σ
0.009
0.029
0.026
0.018
0.023
Lag well-being (θ)
0.1007**
σ
0.035
Positive affect
0.1018***
-0.0077
-0.0027
0.0321
0.0094
σ
0.02
0.029
0.028
0.021
0.024
ρ (spatial error dependence)
0.4282***
0.4238***
σ
0.012
0.012
λ (spatial autocorrelation)
0.2409***
0.2445***
σ
0.012
0.012
neighbourhood and Environmental Characteristics
Ln(Urban distance)
-0.0053*
-0.0055*
-0.0058*
-0.0058*
-0.0119***
-0.0116***
σ
0.002
0.002
0.002
0.002
0.003
0.003
Ln(Density)
0.1008***
0.1006***
0.1003***
0.1003***
0.0859***
0.0859***
σ
0.005
0.005
0.005
0.005
0.007
0.007
Ln(Foreign)
-0.1536***
-0.1535***
-0.1531***
-0.1531***
-0.1164***
-0.1159***
σ
0.009
0.009
0.009
0.009
0.013
0.013
Burglary
-0.0185***
-0.0182***
-0.0171**
-0.0172**
-0.0118
-0.012
σ
0.005
0.005
0.005
0.005
0.008
0.008
Ln(Water distance)
-0.0087***
-0.0087***
-0.0088***
-0.0088***
-0.0116***
-0.0115***
σ
0.002
0.002
0.002
0.002
0.002
0.002
Ln(Forest distance)
-0.0045*
-0.0043*
-0.0038*
-0.0039*
-0.0039
-0.004†
σ
0.002
0.002
0.002
0.002
0.002
0.002
Ln(Nature distance)
-0.0553***
-0.0549***
-0.0546***
-0.0547***
-0.0357***
-0.0359***
σ
0.003
0.003
0.003
0.003
0.004
0.004
Structure Characteristics
Ln(Floor Area)
0.6616***
0.6508***
0.6485***
0.6471***
0.6471***
0.5932***
0.5931***
σ
0.009
0.009
0.009
0.009
0.009
0.008
0.008
Ln(Lotsize)
0.0461***
0.0514***
0.0514***
0.0514***
0.0514***
0.0553***
0.0552***
σ
0.001
0.001
0.001
0.001
0.001
0.001
0.001
*** - p < 0.001, ** - p<0.01, * - p<0.05, † - p<0.10
Continued on next page
23


===== PAGE 25 (score 99, numeric-paren 72, table-words 0, var-words 12) =====

... continued from previous page:
Variable
OLS Model 1
OLS Model 2
OLS Model 3
IV Health
IV Health/Pain
GNS
SAC
Rooms
0.0309***
0.0327***
0.0326***
0.0323***
0.0323***
0.0498***
0.05***
σ
0.009
0.009
0.009
0.009
0.009
0.008
0.008
Two Bathrooms
0.1438***
0.1395***
0.1386***
0.1374***
0.1375***
0.1164***
0.116***
σ
0.008
0.008
0.008
0.008
0.008
0.007
0.007
Three Bathrooms
0.2432***
0.2489***
0.2487***
0.2542***
0.254***
0.1801***
0.1806***
σ
0.063
0.061
0.061
0.061
0.061
0.053
0.053
Balcony
0.1196***
0.1158***
0.1153***
0.1153***
0.1153***
0.087***
0.0871***
σ
0.006
0.006
0.006
0.006
0.006
0.005
0.005
Parking
0.1244***
0.1342***
0.1338***
0.1327***
0.1327***
0.1203***
0.1205***
σ
0.004
0.004
0.004
0.004
0.004
0.004
0.004
Basement
-0.0192*
-0.0228**
-0.0227**
-0.0223**
-0.0223**
-0.032***
-0.0319***
σ
0.008
0.008
0.008
0.008
0.008
0.007
0.007
Attic
-0.0392***
-0.0371***
-0.0373***
-0.0389***
-0.0389***
-0.0316***
-0.032***
σ
0.004
0.004
0.004
0.004
0.004
0.004
0.004
Under roof
-0.0064
-0.0051
-0.0054
-0.0073
-0.0073
-0.0106*
-0.0112*
σ
0.006
0.006
0.005
0.006
0.006
0.005
0.005
Listed
0.2465***
0.2313***
0.2287***
0.2257***
0.2258***
0.1899***
0.1895***
σ
0.029
0.028
0.028
0.028
0.028
0.025
0.025
Monument
0.1416***
0.1425***
0.14***
0.1418***
0.1418***
0.1648***
0.1648***
σ
0.024
0.023
0.023
0.023
0.023
0.02
0.02
Constant
8.3239***
8.5819***
8.271***
7.8333***
7.8533***
4.8285***
5.1519***
σ
0.072
0.079
0.088
0.122
0.117
0.242
0.185
Other variables in models
Int/Exterior maintenance
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Insulation level
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Construction Period
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Year of sale
Yes
Yes
Yes
Yes
Yes
Yes
Yes
*** - p < 0.001, ** - p<0.01, * - p<0.05, † - p<0.10
Continued on next page
24


===== PAGE 26 (score 33, numeric-paren 21, table-words 4, var-words 0) =====

... continued from previous page:
Variable
OLS Model 1
OLS Model 2
OLS Model 3
IV Health
IV Health/Pain
GNS
SAC
Measures of model fit to data
Residual σ
0.2247
0.2175
0.217
0.2187
0.2186
R2
0.7035
0.7223
0.7235
0.7191
0.7195
0.7574
0.7568
¯R2
0.7024
0.7212
0.7224
0.7179
0.7183
AIC
-2150.58
-3109.98
-3172.68
Tests of performance of variables in model
F
662.7***
641.4***
624.5***
Wald test
615***
615.9***
Weak instruments
1744.71***
1057.375***
Wu-Hausman
28.05***
30.969***
Sargan
0.332
Tests for spatial dependence and autocorrelation
LM spatial error
4877***
3561.8***
3511.8***
LM spatial lag
177.84***
259.98***
266.8***
Moran’s I
55.294***
54.548***
Partial first stage results for well-being instruments
Ln(Health+1)
0.1723***
0.1723***
0.1723***
σ
0.006
0.006
0.006
Ln(Pain+1)
0.0807***
0.0807***
σ
0.005
0.005
*** - p < 0.001, ** - p<0.01, * - p<0.05, † - p<0.10
25


===== PAGE 27 (score 26, numeric-paren 0, table-words 1, var-words 24) =====

OLS Model 1 presents estimates that include only structure characteristics. The estimated parameters are largely as
would be expected. A one per cent change in floorspace is associated with a 0.65 to 0.66 percent increase in the transaction
price. A one percent change in lot size increases the transaction price by 0.05 per cent. The number of rooms, and availability
of bathrooms, balconies, and parking are all associated with an increase in the transaction price.
The presence of an attic, basement, and under-roof storage area are associated with lower transaction prices. The data
for usable interior surface area as described in NVM and VBO and VastgoedPRO and Vereniging Nederlandse Gemeenten
and Waarderingskamer (2018) specify that floorspace (floorm2) includes all usable area, which includes basements, attics,
and under-roof storage. The negative coefficients estimates arise from a portion of the interior area not being part of
the living area. Listed buildings (monuments) are more valuable than those that are not listed, and for non-monuments, a
monumental appearance leads to a smaller increase in price. Higher levels of interior and exterior maintenance are associated
with higher transaction prices, and similarly higher levels of insulation are associated with higher transaction prices, although
buildings with only one layer of insulation appear to sell for less than those with no insulation (reference group). Structures
from the 1980’s attract the lowest transaction prices.
OLS Model 2 adds the neighbourhood characteristics. Increases in density are associated with increases in transaction
price, which is consistent with the standard urban land use model presented in Alonso (1964). Controlling for that, larger
distances to an urban center (as a proxy for labor market access) are associated with lower transaction prices, burglaries are
associated with lower transaction prices, as is ethnic diversity.
OLS Model 3 includes subjective well-being and positive affect. Both coefficients are significant and positively related
to transaction price. This supports our hypothesis that higher transaction prices are associated with higher derived utility
from living in a house and location, which can be measured using subjective well-being. In addition, this model shows that
positive affect is associated with a higher transaction price, which supports the optimistic buyer hypothesis. The AIC’s
for these three models suggest that the fully specified model is the preferred one, decreasing from -2,151 for the property
characteristics model, to -3,110 for the model including neighbourhood characteristics, and finally -3,173 for the model
including subjective well-being.
The inclusion of subjective well-being in model 3 is associated with a number of small changes to the parameter estimates
for the usual variables included in a hedonic price model (model 2). The parameter estimate for floorspace, for example,
decreases by 0.0023, equivalent to 0.26 of the standard error of the parameter estimate. This suggests that the parameter
estimate in the standard hedonic is slightly biased upwards, meaning that some part of the increased willingness to pay
attributed to a larger floor area is actually due to the omitted attributes proxied through subjective well-being. We note a
similar change in the coefficients for the distance to forests and the distance to nature: the negative association in both
coefficients decreases when subjective well-being is included in the model. This indicates that part of the willingness to
pay attributed to proximity to nature is potentially due to some other unobserved characteristic associated with occupant
subjective well-being.
Similar to Anselin and Lozano-Gracia (2008) we test for spatial dependence in our data, and find robust Lagrange
Multipliers in excess of 3,511 for spatial dependence in the error term, and greater than 177 for spatial dependence in the
26