You are on page 1of 22

Journal of Happiness Studies

https://doi.org/10.1007/s10902-019-00081-2

RESEARCH PAPER

We Do Not Have the Same Tastes! Evaluating Individual


Heterogeneity in the Preferences for Amenities

Gustavo Ahumada1 · Victor Iturra1 · Mauricio Sarrias1

© Springer Nature B.V. 2019

Abstract
It is widely recognized by scholars that amenities affect the individual well-being. Ample
empirical evidence has been provided for developed countries, although this analysis for
developing economies is scant. The aim of this paper is to study the association between
locational specific characteristics and self-reported measures of subjective well-being.
We focus our analysis in Chile, a developing country located in South America endowed
with a rather heterogeneous set of amenities across cities. Using data from several sources
to account for both natural and urban city amenities along with individual traits, our first
results suggest that natural and urban amenities do affect the level of subjective well-being
across Chilean cities. Afterwards, we allow for the estimated parameters associated to
amenities vary to characterize the whole distribution rather than a single average param-
eter. This analysis uncovers the existence of unobserved individual heterogeneity, that is,
individuals display different tastes for amenities not captured by observed traits, and con-
sequently compensating variations associated to amenities differ across the sample. These
results provide valuable elements to policy makers and city planners to the design of poli-
cies that enhance the population well-being and to the understanding of the development of
cities in developing economies.

Keywords  Subjective well-being · Amenities · Heterogeneous preferences · Random


parameters models · Compensating variation

JEL Classification  C35 · I31 · Q51

* Gustavo Ahumada
gustavo.ahumada@alumnos.ucn.cl
Victor Iturra
viturra@ucn.cl
Mauricio Sarrias
mauricio.sarrias@ucn.cl
1
IDEAR, Department of Economics, Universidad Católica del Norte, Avenida Angamos,
0610 Antofagasta, Chile

13
Vol.:(0123456789)
G. Ahumada et al.

1 Introduction

In recent decades, the study of how site-specific characteristics affect individuals’ well-
being has received increasing attention (Winters and Li 2016; Sarrias 2018; Cazzuffi and
López-Moreno 2018; Ferreira and Moro 2010). When explaining how individuals’ well-
being varies across cities, the literature recognizes that several locational characteristics
might be important, such as local labor market conditions and city-specific attributes.
Among them, amenities have emerged as one of the key factors affecting individual well-
being (Frey et al. 2010; Glaeser et al. 2001). Amenities are defined as site characteristics
that affect household willingness to live in a place and can be classified as natural ameni-
ties and man-made amenities. While the first group encompasses natural characteristics
such as mild temperature, access to a lake view and open space, the second group includes
site characteristics that emerge from the urban development such as restaurants, theaters
and high-quality urban transportation (Partridge 2010). Because individuals’ well-being or
utility is not directly observable, the literature has focused on self-reported measures to
study the association between site characteristics and subjective well-being. In this vein,
the Life Satisfaction Approach (LSA) establishes that self-reported life satisfaction, com-
monly referred as subjective well-being (SWB), can be used as a proxy measure for indi-
vidual’s utility (Ferrer-i Carbonell and Frijters 2004; Frey et al. 2009; Goetzke and Islam
2017).1
Although the research on the relationship SWB and site-specific characteristics is prom-
ising, the role played by unobserved individual heterogeneity has not fully explored. While
theoretical (Albouy et al. 2016) and empirical findings (Cazzuffi and López-Moreno 2018)
suggest that individuals’ tastes for amenities are heterogeneous, much of the previous work
has tried to model individual heterogeneous preferences using observed socio-demographic
characteristics such as age, sex, schooling, and marital status (Boxall and Adamowicz
2002; Krupka 2009). While it is possible to explain the observed heterogeneity by interact-
ing amenities (or city attributes) and individual characteristics, it has also been recognized
that an important part of this heterogeneity is not observed because individuals’ prefer-
ences to amenities are intrinsically idiosyncratic (Ardeshiri et al. 2018). Thus, limiting the
tastes for amenities to be a deterministic component might result in a loss of explanatory
power and misleading conclusions about the extent of the heterogeneity (Hess 2014).
In this context, when analyzing amenities and SBW a question that arises is: why is
it important to analyze unobserved individual heterogeneity? The answer can be found
going back to the origin of the spatial equilibrium model. Rosen (1979) and Roback

1
 The concept of SWB, defined as “a person’s cognitive and affective evaluations of his or her life”
(Diener et al. 2005), has been used in different fields such as economics, sociology and psychology to eval-
uate the key factors affecting individuals’ well-being. In general, SWB is measured by survey questions
about individuals’ self-perception about different domains such as life satisfaction, happiness and quality
of life. Although, some debate of the correct interpretation of each of these domains persists (see for exam-
ple, Veenhoven 2000, 2017), in economics—the field from which the theoretical framework used in this
article is derived—the measurement of individual welfare using data on reported subjective well-being has
become popular and has served as an empirically adequate and valid approximation for individually experi-
enced welfare (Frey et al. 2009). Furthermore, the empirical evidence seems to support the theoretical link
between SWB and the unobserved utility. For example, Benjamin et  al. (2014) and Perez-Truglia (2016)
using behavioral experiments and observational data, respectively, find that SWB is a reliable proxy for
individuals’ utility and preferences. Thus, in the economics literature, these concepts are often used inter-
changeably (see Di Tella and MacCulloch 2006; Ferreira and Moro 2010) and we adopt the same approach
in this study.

13
We Do Not Have the Same Tastes! Evaluating Individual…

(1982) pointed out that there are unobserved differences in tastes for amenities that might
either overstate or understate the estimates of willingness to pay (WTP) using the hedonic
approach. More recently, Bishop and Timmins (2011) highlight that unobserved workers’
heterogeneity hinders the identification of parameters using a hedonic price model, even
using instrumental variables.
This paper attempts to fill that void by quantifying the importance of unobserved indi-
viduals’ heterogeneity when analyzing their preferences for natural and urban amenities.
Using the LS approach, we estimate how heterogeneous preferences due to unobserved fac-
tors (unobserved for the analyst) result in different WTP for local amenities. To achieve
this goal, we employ a flexible methodology to estimate the whole distribution of tastes
for amenities across individuals to obtain a profile of their WTPs. This method not only
enables us to identify the mean utility/disutility associated to amenities, but also to assess
whether, and to what extent, the conditional relationship between SWB and amenities var-
ies between individuals. To our knowledge, this is the first work that makes use of a ran-
dom coefficient approach to disentangle individuals-specific tastes for amenities using a
national representative survey.2
We focus our empirical analysis on Chile, a long and narrow (4230  km long and, on
average, and 180 km wide) middle-income country located at South America. Chile rep-
resents an interesting case to this study because its cities, as the first three maps of Fig. 1
show, are rather heterogeneous in terms of natural amenities, e.g. temperature and precipi-
tation. Additionally, Chile has a population of 17.4 million inhabitants and it is character-
ized by an excessive spatial concentration of the population in and around the middle of the
country as shown in the fourth map, where the metropolitan region of Santiago is located.
These characteristics coupled with marked differences in our measure of SWB over space,
as the last two maps display, lead to the research question of this study: After controlling
for individual characteristics, are amenities significantly associated to city differences in
SBW? If so, is this association heterogeneous between individuals?
To answer these research questions, we use the Socio-Economic Characterization
National Survey 2013 (CASEN 2013), which is nationally representative sample and con-
tains individual measures of self-reported life satisfaction and demographic characteristics.
City natural amenities come from the 2009 WordClim Version 1 data sets. City man-made
amenities were obtained from the National System of Municipal Information (SINIM), and
National Institute of Statistics (INE). Our main findings suggest that SWB is higher in cit-
ies characterized by more amenable weather conditions, low population density and low
levels of robberies and pollution. In addition, we identify the existence of heterogeneity
in individuals’ preferences, resulting in substantial differences in monetary valuation for
some amenities. Thus, our study suggests that the analysis of amenities requires modeling
unobserved individual heterogeneity to draw more accurate conclusions about their impact
on SWB.
The rest of this paper is structured as follows. Section  2 provides a brief overview of
the literature about the relationship between locational characteristics and individual well-
being. Section 3 describes the data. Section 4 outlines the econometric strategy, along with
some issues related to the empirical estimation. Results are presented in Sect. 5. Finally,
Sect. 6 concludes and discusses some policy implications.

2
  In general, studies using a random coefficient approaches use stated preferences surveys.

13
G. Ahumada et al.

Fig. 1  Amenities and SWB across Chilean cities: Quartile map. Note: (1) Max. of warmest month: the
maximum monthly temperature occurrence over a given year. The category Extremely low indicates tem-
peratures between 13 and 19 °C, while Extremely High indicates temperatures between 28 and 31 °C. (2)
Min. of coldest month: the minimum monthly temperature occurrence over a given year. The category
Extremely low indicates temperatures between − 7.9 and − 0.5  °C, while Extremely High indicates tem-
peratures between 7.1 and 13.4  °C. (3) Annual precipitation: it is the sum of total monthly precipitation
values. The category Extremely low indicates precipitations between 0 and 338 mm, while Extremely High
indicates precipitations between 1647 and 2578 mm. (4) Population density: The category Extremely low
indicates inhabitants/km2 between 0.2 and 11.4, while Extremely high indicates inhabitants/km2 between
117.1 and 17,093. (5) Completely satisfied indicates the proportion of individuals who manifest to be com-
pletely satisfied. The category Extremely low indicates a proportion between 11.1% and 26.2%, while cat-
egory Extremely High indicates a proportion between 45.2% and 68.1%

2 Literature Review: Individual Well‑Being and Amenities

As mentioned above, amenities represent the key set of variables for our study. For this rea-
son, this section begins with a deeper revision of what amenities mean followed by a brief
review of the empirical association between amenities and SWB.
In general, amenities comprise any characteristic associated to a specific location that
affect individuals’ well-being. As emphasized by Roback (1982), the literature does not
provide us a comprehensive list of what should be considered as amenities, but we can
observe how well-being is affected by them. Because of this, the empirical literature usu-
ally uses a large group of site characteristics or spatial fixed effects to control for amenities
(Gabriel and Rosenthal 1999).
With respect to amenity categories, Brueckner et al. (1999) propose a classification of
urban amenities into three groups: natural (climate and topographic characteristics), histor-
ical (monuments and well-preserved buildings) and modern amenities or man-made amen-
ities (private and public goods). In this work, we focus on two broad groups of amenities:
natural as in Brueckner et al. (1999) and man-made amenities, which include both histori-
cal and modern amenities. In addition to this, as cities grow, some congestion costs arise,

13
We Do Not Have the Same Tastes! Evaluating Individual…

such as pollution, crime and traffic congestion. When available, these city traits should also
be incorporated into an empirical analysis because they represent important elements that
shape individual differences in well-being across cities.
Another important element is the spatial scale of analysis because when studying intra-
city amenities, characteristics associated to neighborhoods are usually the focus of the
analysis, while when studying inter-city amenities (as in our study), analysts place a greater
emphasis on attributes related to cities rather than neighborhoods.
The empirical literature has usually focused on both wages and housing prices to esti-
mate the effect of natural amenities on individual well-being. According to Maddison
(2001), British household value higher temperature as an amenity and higher precipita-
tion a disamenity. However, Maddison and Bigano (2003), for an Italian sample, find that
winter temperature does not display a significant association with welfare. In contrast,
Rehdanz (2006) finds that some households in Great Britain would be willing to pay for
higher mean temperatures in July. For USA, Rappaport (2007) finds that housing price
growth increases with warmer winters and decreases with hotter summers. In other words,
the housing market compensates households for differences in natural amenities. In Chile,
Iturra (2018) finds that Chilean regions with higher level of precipitation represent more
attractive places to households.
In focusing on LSA, some scholars have analyzed the relationship between individual
well-being and specific environmental attributes. For example, Rehdanz and Maddison
(2005) analyze the effect of climate characteristics on well-being in a panel of 67 countries.
Their findings show that well-being is positively related to higher temperature in winter,
and lower temperature in summer, and displays a negative association to the rainfall. Fer-
reira and Moro (2010) using both LSA and hedonic pricing approaches find that individual
SWB is affected positively for warmer temperature and air quality in Ireland.
Moving on to man-made amenities, the literature presents consistent evidence of the
relationship between these amenities and well-being using the spatial equilibrium model.
Roback (1982) and Blomquist et al. (1988) find that crime and air pollution are negatively
correlated with individual well-being. Similarly, Berger et  al. (2008) find that crime rate
and air pollution are negatively correlated with the quality of life. Colombo et al. (2014)
state that an average household in Italy would be willing to pay for living in cities with
green areas, less crime, low unemployment rate and better transportation system. In Chile,
households require a positive compensation to live in cities with higher crime levels, but
there is no a significant association to green areas (Iturra 2018).
With respect to the analysis of man-made amenities using LSA, Levinson (2012) and
Luechinger (2009) find a negative association between air pollution and individual well-
being. More recently, Ambrey and Fleming (2014) find a positive relationship between
green areas and self-reported life satisfaction; moreover, individuals display a positive will-
ingness to pay for an increase in public green space. It is worth mentioning that while
larger cities are characterized by a whole set of consumption possibilities (Glaeser and
Gottlieb 2009; Glaeser et al. 2001), more dense places also have higher levels of pollution,
crime, and commuting cost, therefore it is difficult to identify the specific effect of these
elements on SWB. Because of this, the literature usually focuses on estimating the overall
effect of urbanization of SWB, as stressed by Winters and Li (2016).
From the foregoing, it can be argued that the empirical studies are mainly focused on
developed countries. As Welsch and Ferreira (2014) claim, the lack of data in developing
countries precludes us from performing an adequate analysis of amenities and well-being.
Despite this fact, it is possible to find some studies in Latin America that analyze how
city attributes affect the individual well-being. Among these studies, the main findings are

13
G. Ahumada et al.

similar to those in developed economies, that is, city attributes are significantly associated
to the well-being of individuals. More specifically, better quality public goods emerge as
one the key factors that improve the quality of life in Latin American inhabitants, as sug-
gested by the empirical evidence in cities of Uruguay, Colombia and Argentina (Gandel-
man et  al. 2012; Medina et  al. 2010; Cruces et  al. 2010). Interestingly, as in developed
economies, city attributes associated to pollution and crime are also important when Latin
American households assess their subjective well-being and, as expected, these factors
decrease the reported well-being of individuals in Latin American cities.

3 Data

We use the Chilean National Socioeconomic Characterization Survey (CASEN 2013)


which contains individuals’ characteristics such as monthly wage, gender, marital status,
age, years of schooling and race and questions related to life satisfaction.3 The initial sam-
ple corresponds to 29,695 individuals aged between 15 and 65  years that answered the
question related to SWB and declare a positive monthly wage. After dropping workers with
missing values in the covariates and merging the individual data set with city characteris-
tics, our final sample includes 18,774 workers located in 320 cities.
The dependent variable corresponds to the response to the question “How satisfied are
you with life?” The answer to this question was coded to create five categories: “Com-
pletely unsatisfied” (1), “Unsatisfied” (2), “Moderately unsatisfied” (3), “Satisfied” (4), and
“Completely satisfied” (5). We use this domain of the SWB because it is considered as a
better proxy of individuals’ welfare and utility in the economic literature compared to other
SWB questions. For example, Benjamin et al. (2012) using an experimental approach show
that SWB measures can be considered as a reliable proxy of the real preferences (utility) of
individuals. They also argue that life satisfaction performs better than happiness for most
of the hypothetical scenarios across respondents. Similarly, Perez-Truglia (2015) find that
life satisfaction correlates significantly better with objective measures of well-being (such
as suicide rates and frequency of smiling) than other subjective measures of well-being.
Finally, Frey et al. (2009) point out the necessary conditions that must be met to consider
life satisfaction data as a valid measure to be used in the LS approach.
Table 1, which reports summary statistics, shows that the average life satisfaction across
workers is approximately 4 revealing that the distribution is mostly skewed to the right.
It is important to note that the fact that individuals tend to report life satisfaction near the
top end of the scale is a common result in the literature (see for example, Mentzakis 2011;
Schurer and Yong 2016).
We use both natural and man-made amenities at the city level. The first set of vari-
ables are related to climate variables: (1) temperature seasonality measures the amount
of temperature variation over one year based on the standard deviation of monthly tem-
perature averages; (2) maximum (minimum) temperature of warmest (coldest) month is
the maximum (minimum) monthly temperature occurrence over a year (expressed in Cel-
sius degrees); (3) annual precipitation is the sum of all total monthly precipitation values
(expressed in millimeters). All these variables come from the 2009 WorldClim Version

3
  Although the CASEN has been conducted every 2 or 3 years since 1987, we only use the 2013 version
because it is the only wave containing questions about individuals’ life satisfaction.

13
We Do Not Have the Same Tastes! Evaluating Individual…

Table 1  Individual and city characteristics: definition and summary statistics


Individual characteristics Mean SD

Life satisfaction (1 = completely unsatisfied to 5 = completely satisfied) 4.020 0.960


Female (1 = female) 0.569 0.495
Age (year old) 42.107 11.182
Married (1 = married/cohabiting couple) 0.625 0.484
Schooling (total years of schooling) 12.078 3.660
Good health (1 = if self-perception of health is good) 0.958 0.198
Indigenous (1 = if individual self-identifies as in one of the eight indigenous group) 0.102 0.303
Size household (number of household members) 3.454 1.591
Log of monthly wage (monthly wage earned) 12.649 0.778
Log of monthly rent (monthly housing rent) 11.978 0.417
Monthly hours worked 43.438 13.427
Total sample (N) 18,774
Perception of neighborhood attributes
Noise pollution (1 = if respondent identifies noise pollution) 0.265 0.441
Waste pollution (1 = if respondent identifies waste pollution) 0.215 0.410
Robberies (1 = if respondent identifies robberies) 0.424 0.494
Total sample (N) 18,774
City amenities
Temperature seasonality (standard deviation * 100) 338.836 85.140
Max. temperature of warmest month (°C) 25.716 3.954
Min. temperature of coldest month (°C) 4.862 3.360
Annual precipitation (mm) 708.914 626.095
Log(distance to sea) (km) 9.793 2.037
Parks (number of parks) 5.417 6.223
Log of the population by city 11.607 1.082
Log of the area by city 6.495 2.106
Number of cities 320

Own elaboration based on the sample used in the main regressions

1 data set, which is the latest version available for Chile. The second set of variables are
intended to capture urban amenities: logarithm of distance to sea (kms.), number of parks
by city (Park) obtained from National Municipal Information System (SINIM), logarithm
of the population by city from the National Institute of Statistics (INE) of Chile, and the
logarithm of the area by city.
In addition, a set of dummy variables indicating the perception of the neighborhood are
included. The first two variables are related to self-perception of pollution. Specifically, the
dummy variables are obtained from the response to the question: ‘What problem related
to pollution do you identify in your neighborhood or location?’ The variables “noise pol-
lution” and “waste pollution” were created depending on whether the interviewees per-
ceived these problems. The proportion of workers reporting these problems are 26% and
22%, respectively (see Table 1). Similarly, individuals were asked ‘What problem related
to public security do you identify in your neighborhood or location?’ The dummy variable
“robberies” was created if workers perceived robberies/assaults on people, home and/or
vehicles in their neighborhood.

13
G. Ahumada et al.

We control for several individual characteristics commonly used in the literature of


SWB. Table  1 shows that females account for almost 57% of the sample, whereas the
average age is about 42 years old. Workers had an average of 12.1 years of education and
almost 63% of them are married or cohabiting. About 10% of the workers identify them-
selves as indigenous, while 96% consider that they have good health. The mean of monthly
wage is about $311,451 Chilean pesos.

4 Methodology and Individual Heterogeneity

The analysis of the relationship between amenities and SWB is usually done using lin-
ear regression models or standard ordered choice models. These models assume a single
(fixed) coefficient to represent the whole sample. This assumption is not flexible enough
to incorporate heterogeneous preferences for amenities. For example, consider the effect
of warmer month on individual SWB. Cuñado and de Gracia (2013) and Winters and Li
(2016) found, on average, a negative correlation. However, Rehdanz and Maddison (2005)
suggest that people would prefer higher mean temperature in the coldest month and lower
temperature in the hottest month. These results can be explained by the implicit recogni-
tion of individual heterogeneity in preferences. An average negative (positive) coefficient
for temperature might mask the fact that for some share of the population high temperature
creates disutility (utility). Likewise, even if the relationship between some disamenity and
well-being were negative, one would expect that the magnitude would not be the same for
all individuals. Finally, the estimated parameters and WTPs might be inconsistent if we do
not control for presence of heterogeneous preferences (Hess 2014).
Our empirical strategy is based on random parameter ordered probit (RPOP) model.4
The RPOP assumes that coefficients vary randomly across individual according to some
continuous distribution, such as f (𝜃) , where 𝜃 represents the first and second moments of
the coefficient distribution. At this stage, the probability density function (pdf) f (⋅) is as
generic as possible. In Sect.  5.2, we give more detail about this issue. Such distribution
may also allow observed heterogeneity in the relationship between individual SWB and
amenities. However, the tastes for amenities depend on other factors that are unobserved
for the researcher (e.g. idiosyncratic preferences, personal traits, etc.). These unobserved
factors are captured by the second moment of the distribution.
As in Welsch and Kühling (2009), we assume that the true latent well-being
( ) of individ-( )
ual i located in city j depends
( ) on a vector of natural and urban amenities a(j  , )income yij  ,
average housing rent rj  , a vector of socio-demography characteristics zij  , and unob-
served factors 𝜀ij . The latent well- being function can be stated as:
SWB∗ij = a�j 𝜶 + 𝛿yij + 𝛾rj + z�ij 𝛽 + 𝜀ij , (1)
where is a latent (unobserved) continuous variable. We adjust Eq. (1) allowing 𝜶 to
SWB∗ij
vary across individuals in the population in the following way:
SWB∗ij = a�j 𝜶 i + 𝛿yij + 𝛾rj + z�ij 𝜷 + 𝜀ij , i = 1, … , N; j = 1, … , J (2)

𝜶 i ∼ f (𝜶|𝜽)

4
  For some applications, see Falco et al. (2015) and Greene et al. (2014).

13
We Do Not Have the Same Tastes! Evaluating Individual…

Note that now we allow the vector of amenities to vary across individuals assuming
some parametric distribution f (⋅) . Given that 𝜶 i is unobserved, it must be integrated out.
For simplicity, let 𝚯 = (𝛿, 𝛾, 𝛽, 𝜽)� . Thus, the unconditional pdf will be the weighted aver-
age of the conditional probability evaluated over all possible values of 𝜶 , which depends on
the parameters of the distribution of 𝜶 i:
( )
∫ (3)
Pij (𝚯) = Pr(SWBi = s|𝜶 i )f 𝜶 i d𝜶 i

This probability has no-close form solution. That is, it is difficult to integrate out the
random parameters and to perform ML estimation.5 To overcome this( problem, ) we simu-
late the probability in (3) using random draws from the distribution f 𝜶 i  . The Simulated
Maximum Likelihood (SML) estimation procedure is given by:
N

log L(𝚯) = log P̃ i (𝚯),
n=1

where P̃ i (𝚯) is the simulated probability for individual i  . Since our dependent variable is
an ordered variable, and assuming normally distributed error terms, P̃ i (𝚯) is computed as:
R S { ( ) ( )}dis
1 ∑∏
P̃ i (𝚯) = 𝚽 ks − a�j 𝜶 ir + 𝛿yij + 𝛾rj + z�ij 𝜷 − 𝚽 ks−1 − a�j 𝜶 ir + 𝛿yij + 𝛾rj + z�ij 𝜷 ,
R r=1 s=1

where 𝚽(.) denotes the cumulative density function of the standard normal distribution,
and dis = 1 if individual i’s life satisfaction equals category s , and 0 otherwise; 𝛼ir is the r
-th draw of 𝛼i , and R is the total number of draws used to integrate out the random com-
ponents. Lee (1992) and Hajivassiliou and Ruud (1994) show that under regularity condi-
tions, the SML estimator is consistent and asymptotically normal. When the number of
draws, R , rises faster than the square root of the number of observations, the estimator
is asymptotically equivalent to the maximum likelihood estimator. In this study, we use
Rchoice package in R for the SML procedure and Halton draws for the simulation (Sarrias
2016).

5 Empirical Findings

5.1 The Relationship Between SWB and Amenities: Baseline Analysis

The main objective of this Section is to give some preliminary insights about the relation-
ship between amenities and SWB in Chile using standard ordered probit models. Table 2
displays the estimated marginal effects that an average individual manifest being com-
pletely satisfied with his life (SWB = 5) using different specifications of Eq. (2).6 We evalu-
ate the marginal effect at every observation and use their sample average (Scott 1997).7

5
  Note that for each random parameter we need an integral.
6
  For dummy variables, the marginal effects correspond to a discrete change from 0 to 1. The marginal
effects for the rest of the categories are presented in Table 5 in “Appendix”.
7
  Standard errors for the marginal effects are computing using Delta Method.

13
G. Ahumada et al.

Table 2  Marginal effects on the probability of reporting complete satisfaction with life (SWB = 5)
Model 1 Model 2 Model 3
ME SE ME SE ME SE

Log (wage) 1.02e−01*** 0.0056 1.02e−01*** 0.0054 9.97e−02*** 0.0055


Female 5.37e−03 0.0069 7.36e−03 0.0068 6.78e−03 0.0068
Age − 1.64e−03*** 0.0003 − 1.58e−03*** 0.0003 − 1.58e−03*** 0.0003
Married 8.63e−02*** 0.0067 8.46e−02*** 0.0067 8.48e−02*** 0.0067
Schooling 7.35e−03*** 0.0011 7.88e−03*** 0.0011 7.80e−03*** 0.0011
Good health 1.86e−01*** 0.0107 1.83e−01*** 0.0108 1.83e−01*** 0.0108
Indigenous − 3.20e−02*** 0.0101 − 3.23e−02*** 0.0097 − 3.27e−02*** 0.0097
Household size − 7.29e−03*** 0.0021 − 6.39e−03*** 0.0020 − 6.34e−03*** 0.0020
Temperature seasonality 8.57e−06 0.0001 1.18e−05 0.0001
Max. of warmest month − 4.57e−03* 0.0025 − 3.91e−03 0.0025
(°C)
Min. of coldest month (°C) − 6.54e−04 0.0028 − 5.99e−04 0.0028
Annual precipitation (mm) − 1.48e−05** 0.0000 − 1.08e−05* 0.0000
Log (distance to sea) − 1.27e−03 0.0026 − 1.64e−03 0.0026
Parks 8.55e-04 0.0005 7.22e-04 0.0005
Log (population) − 9.53e−03*** 0.0032 − 1.24e−02*** 0.0036
Log (area of the city) − 3.05e−04 0.0021 − 6.59e−04 0.0021
Noise pollution − 3.11e−02*** 0.0068 − 3.18e−02*** 0.0068
Waste pollution − 2.51e−02*** 0.0072 − 2.46e−02*** 0.0072
Robberies − 3.46e−02*** 0.0062 − 3.44e−02*** 0.0062
City dummies Yes No No
Log (rent) No No Yes
BIC 48,255.100 46,312.179 46,318.915
AIC 46,216.640 46,116.173 46,115.069
Pseudo R-squared 0.043 0.035 0.036
Observations 18,774 18,774 18,774
Wald test χ2 572.35 (0.000)
Compensating variation (USD)
Temperature seasonality − 7.4 77.67
Max. of warmest month 244 156.93
(°C)
Min. of coldest month (°C) 37 175.83
Annual precipitation (mm) 67* 40.27
Log (distance to sea) 10 16.45
Parks − 45 36.56
Log (population) 4.7*** 1.39
Noise pollution 200*** 44.72
Waste pollution 155*** 46.97
Robberies 216*** 40.94

The marginal effects (ME) are obtained through the estimation of an ordered probit with fixed parameters.
The dependent variable is life satisfaction. Standard errors (SE) were obtained using the Delta Method
p value for Wald test is in parenthesis. *p < 0.1, **p < 0.05, ***p < 0.001

13
We Do Not Have the Same Tastes! Evaluating Individual…

Model 1 includes only individual socio-economic variables and city-fixed effects to control
for relevant information common to individuals located in every city as suggested by Gabriel
and Rosenthal (1999). The variables at the individual level are stable across specification and
the estimated marginal effects show the expected signs (Brereton et  al. 2008; Ferreira and
Moro 2010; Ferreira et al. 2013; Cuñado and de Gracia 2013). Then, we test whether city-
fixed effects are jointly statistically significant. The 𝜒 2 statistic suggests that there is relevant
unobserved information related to cities that makes SWB to vary across space. The next ques-
tion is: how much of this unobserved city-specific variability is due to amenities?
Model 2 incorporates our set of amenities, where two out of the four city climate charac-
teristics are associated to individual well-being. Specifically, cities with higher maximum tem-
perature and higher precipitation represent unattractive places for people resulting in a nega-
tive probability of reporting completely satisfied with their life. These results are in line with
those found by Winters and Li (2016), who find a negative correlation between individual
SWB and rain in United States.
With respect to urban dis/amenities, it is common that scholars use population density as
a measure for urbanization (Brereton et al. 2008; Winters and Li 2016). However, in Chile,
there is no adequate information to identify the proportion of urban areas at the city level.
Therefore, city population is used as a proxy for urbanization while we control for the size
of the cities. The results indicate that highly populated cities represent less desirable places,
and therefore, the probability to report a complete satisfaction is lower. Although this result is
robust across the last two specifications, it should be taken with caution. As mentioned, large
urban areas offer benefits (Albouy et al. 2016; Glaeser et al. 2001) and costs (Winters and Li
2016), therefore the estimated coefficient of log (city population) represents a net effect of
urbanization on SWB.
Our measures related to neighborhood perception are highly significant and display the
expected signs. Specifically, if individuals identify noise pollution, waste pollution or robber-
ies in their neighborhood, the probability of reporting a complete life satisfaction decreases.
Importantly, these results are in line with those found by Van Praag and Baarsma (2005) with
respect to SWB and subjective perception of aircraft noise around Amsterdam Airport.
Finally, the Model 3 adds monthly housing rent as a proxy for the cost of living. As Graves
(1983) indicates, housing rents tend to capitalize unpriced amenity variation between cities,
therefore more attractive places should be characterized by a higher average housing rent com-
pared to less amenable cities. Interestingly, most of our coefficients are rather robust when
incorporating this measure, except the marginal effect associated to city population, which
decreases significantly, and maximum temperature of warmest month, which is no longer sta-
tistically significant.
We are now in position to push our analysis further by answering the following question:
what is the maximum amount of money that an average individual requires keeping his level
of well-being constant when facing a change of a specific amenity? To answer this question,
we use the compensating variation (CV) approach, which has been broadly used to evaluate
environmental attributes and other non-market goods (Frey et al. 2009; Welsch and Ferreira
2014). Considering Eq. (1), and since income is in logarithm, the implicit price for a specific
amenity evaluated at the mean of income ȳ can be computed as:
dy 𝛼
CV = = − ȳ . (4)
da 𝛿

13
G. Ahumada et al.

This ratio measures the marginal change in y to hold individuals’ SWB constant, given
a marginal change in amenity a . Thus, if the change in some amenity indicates a worsening
(improvement) of well-being we should observe a positive (negative) sign for CV.
As Table 2 specifies, the implicit prices of both noise pollution and city population are
positive and statistically significant. For city population, the result indicates that an average
individual requires a monthly compensation of $4.7 USD per 10,000 people increase in the
city population.8 This finding strongly suggests that higher population density represents
a disamenity. Similarly, for noise pollution, an individual is willing to accept a monthly
compensation of $200 USD to keep his well-being constant. Similar results were found for
waste pollution and robberies.
Perhaps surprisingly, CVs for natural amenities are not statistically different from zero.
The natural question is: Are natural amenities not important for individuals’ well-being
or is it rather that people’s tastes are so heterogeneous that they tend to compensate each
other, resulting in a coefficient not different from zero? It is important to recall that the esti-
mates in Table 2, and hence the CVs, do not allow for the existence of heterogeneity in the
relationship individual SWB and amenities. The aim of the following section is to account
for this potential unobserved individual heterogeneity.

5.2 Unveiling the Unobserved Heterogeneity

The RPOP model assumes that coefficients vary randomly across individuals according
some continuous distribution, such as f (𝜽) . Thus, f (𝜽) can take several predefined func-
tional forms, being the most popular the normal, triangular, and log-normal (Train 2009).
However, as Hensher and Greene (2003) note, the choice of the underlying parameter dis-
tribution must be made a ‘priori’. A way to choose an appropriate distribution can be based
on some prior theoretical knowledge about the domain of the coefficients.
Under the normality assumption, the coefficient for amenity k can be written as:
𝛼ik = 𝛼k + 𝜎k vik ,
where 𝜎k is the standard deviation, which control for unobserved heterogeneity around
ak , and vik ∼ N(0, 1) . Given that the domain of the normal distribution is (− ∞, + ∞) , by
choosing this distribution we make the explicit assumption that there is a proportion of
workers with a positive and negative relationship between individual SWB and amenities.
This assumption seems to be reasonable if one considers, for example, that some individu-
als might like more rainy places, while others do not.
However, the normality assumption is not without drawbacks. First, the existence of a
positive relationship between noise pollution and well-being for a share of the sample is
theoretically counterintuitive. In cases like this, we would need a distribution f (⋅) whose
domain is restricted to the negative (or positive) domain. Another drawback of the normal
distribution is that it has infinite tails. As a result, some individuals could display extreme
values in their coefficients, and hence in their CVs.
To solve the first problem, we choose the log-normal distribution for those coefficients
for which heterogeneity is expected in the negative or positive domain only. Since the
domain of the log-normal distribution is (0, + ∞) , we create the negative of the variable for
those coefficients with heterogeneity in the negative domain. The specification is:

8
  This ratio was multiplied by 10,000 to get a meaningful measure.

13
We Do Not Have the Same Tastes! Evaluating Individual…

Table 3  Life satisfaction: random parameter ordered probit models


Fixed parameters All normal Triangular and log-
normal
Est SE Est SE % > 0 Est SE

a: mean parameter
Temperature seasonality 0.003 0.033 0.005 0.036 54 − 0.002 0.032
Max. of warmest month − 0.104 0.066 − 0.111 0.073 0 − 0.108 0.110
Min. of coldest month − 0.016 0.073 − 0.010 0.080 48 − 0.244* 0.066
Annual precipitation − 0.028* 0.017 − 0.027 0.019 41 − 0.083** 0.031
Log (distance to sea) − 0.004 0.007 − 0.004 0.008 44 0.003 0.012
Parks 0.019 0.014 0.025 0.018 57 − 0.051** 0.023
Log (population) − 0.033*** 0.009 − 0.035*** 0.010 5 − 0.043** 0.016
Noise pollution − 0.085*** 0.018 − 0.093*** 0.021 0 − 0.102 0.268
Waste pollution − 0.066*** 0.019 − 0.073*** 0.022 11 − 0.105*** 0.030
Robberies − 0.092*** 0.016 − 0.095*** 0.020 32 − 0.099 0.085
b: standard deviation
Temperature seasonality 0.041 0.070 0.101* 0.061
Max. of warmest month 0.100 0.35 0.013 0.111
Min. of coldest month 0.205 0.156 0.400** 0.163
Annual precipitation 0.122* 0.065 0.193** 0.086
Log (distance to sea) 0.023 0.017 0.036 0.024
Parks 0.155*** 0.042 0.221*** 0.049
Log (population) 0.021 0.017 0.02 0.035
Noise pollution 0.009 0.962 0.018 0.269
Waste pollution 0.059 0.077 0.020 0.036
Robbery 0.202*** 0.089 0.149*** 0.086
Log-likelihood − 23,031 − 23,025 − 23,022
AIC 46,115 46,122 46,117
BIC 46,319 46,405 46,399
N 18,774 18,774 18,774

Models were estimated using SML procedure with 500 Halton Draws. All the models include the same set
of controls as those included in Table 2
*p < 0.1, **p < 0.05, ***p < 0.001

( )
𝛼ik = exp 𝛼k + 𝜎k vik ,
( )
where vik ∼ N(0, 1) and ak and 𝜎k . Then aik ∼ log N ak , 𝜎k2 . . Like the normal distribution,
the log-normal distribution is characterized by a very long tail, which is a disadvantage
especially for willingness to pay calculations (Hensher and Greene 2003).
Assuming a triangular distribution is an appealing strategy to solve the second problem.
While it allows positive and negative coefficients, it has also shorter tails than the normal
distribution. In this case, the coefficient follows the next specification:
𝛼ik = 𝛼k + 𝜎k vik
� � � �
� ��√ � � � �
vik ∼ 1 𝜇ik < 0.5 2𝜇ik − 1 + 1 𝜇ik > 0.5 1 − 2 1 − 𝜇ik

13
G. Ahumada et al.

Fig. 2  Distributions for amenities’ coefficients. Notes The horizontal axis measures the coefficient of the
amenity. The densities correspond to kernel density plots of the distribution of the amenity coefficient
derived from models using different parametric distribution. The vertical line corresponds to the coefficient
of the specific amenity for the model without individual heterogeneity

𝜇ik ∼ U[0, 1].


Table  3 shows the estimations assuming a normal distribution (model 2), triangular and
log-normal distribution (model 3) for the coefficients of amenities while the other coef-
ficients are held fixed. The model with fixed coefficients is reported in column 1 for com-
parison purposes, and since our interest is the set of amenities, parameters associated to
individual characteristics are not reported.
In focusing on model 2 in Table 3, we can observe that the standard deviation for annual
precipitation, parks, and robberies are significant, indicating that these coefficients do
indeed vary across individuals; therefore, the central tendency does not give us complete
information about the relationship between these amenities and individual SWB. In terms
of average coefficients, the mean parameters are very close to the model without individual
heterogeneity (fixed parameters).
The normal distribution allows us to obtain the proportion of individuals with positive
tastes (see column 5).9 For example, precipitation increases the probability of reporting
higher well-being for a 41% of the sample, whereas none of the individuals report a posi-
tive correlation between maximum temperature of warmest month and SWB.
The distribution for each amenity coefficient in model 3 was chosen based on the fol-
lowing two criteria. First, all those coefficients that showed a low proportion of positive
(or negative) coefficients in model 2 were assumed to be log-normally distributed. Thus,
the coefficients for maximum temperature of warmest month, noise pollution, waste pol-
lution, and robberies follow a log-normal distribution. Second, for the rest of the vari-
ables for which is not clear a priori the sign of the coefficient we chose a triangular
distribution.
Since the choice of the distribution affects the interpretation of the estimated param-
eters, and consequently the computation of the CVs (Hensher and Greene 2003; Hess
et  al. 2005), we try to give some insights about the appropriateness of the shape of the

( )
9
  This is computed as 100 × 𝛷 𝜎̂ k , where 𝛷 is the cumulative standard normal distribution, 𝛼̂ k and 𝜎̂ k are
𝛼̂
k
the mean and the standard deviation of the coefficients estimated in column 3 of Table 3, respectively.

13
We Do Not Have the Same Tastes! Evaluating Individual…

Table 4  Distribution of compensating variations in USD


Variables Random parameters Fixed parameters
5th percentile Mean Median 95th percentile

Temperature seasonality − $140.91 $3.21 $2.98 $148.23 − $7.37


Max. of warmest month (°C) $185.01 $236.91 $225.30 $273.85 $243.89
Min. of coldest month (°C) − $64.20 $508.08 $507.65 $1078.59 $37.38
Annual precipitation (mm) − $102.96 $173.41 $173.27 $449.89 $67.15
Log (distance to sea) − $50.90 $0.05 − $0.01 $50.92 $10.23
Parks − $210.75 $105.19 $105.76 $419.27 − $45.05
Log (population) $3.73 $5.44 $5.44 $7.16 $4.66
Noise pollution $156.93 $212.58 $209.44 $279.34 $200.04
Waste pollution $157.08 $220.59 $216.44 $297.51 $154.69
Robberies $19.38 $207.15 $114.95 $688.20 $215.66

The compensating variations for random parameters were obtained from model 3 in Table 3. The distribu-
tions for each amenity were obtained using 10,000 draws from its specific distribution (triangular and log-
normal) and then dividing these draws by the income’s coefficient

distribution by plotting the kernel estimates for the unconditional distribution for each
coefficient in the Fig. 2.10 A quick look to Fig. 2 reveals that the triangular distribution
shows a more plausible shape than the normal distribution for all the amenities: it restricts
the domain of the coefficient avoiding extreme (infinite bounds) and unexpected values
for the CV. Furthermore, when comparing the normal and log-normal distribution it can
be observed that the latter provides more realistic interpretation. For example, it has been
documented that the log-normal distribution tends to overestimate standard deviation
(Hess et  al. 2005; Train 2009). However, our results show that—except robberies—the
spread of the log-normal distribution is lower than the normal distribution.
Assuming triangular and log-normal distributions does not only show more realistic
bounds for the coefficients, but they also detect unobserved heterogeneity unrevealed by
the normal distribution: results from model 3 show that the standard deviation for tem-
perature seasonality and the minimum temperature of coldest month become significant
when allowing greater flexibility and more reasonable distribution. However, it is also
important to highlight that the mean parameter for maximum temperature of warmest
month, log (distance to the sea) and noise pollution are not significant. Thus, we can-
not reject the hypothesis that these variables are not correlated with SWB, at least for
the sample used in this study. Furthermore, this lack of relationship does not seem to be
driven by heterogeneous preferences since their standard deviations are not significant.
Table 4 shows for all amenities the 5th and 95th percentile, the mean and median of the
distribution of CVs (in USD dollars), which is based on model 3 of Table 3.11 We focus
our analysis on those variables that present a significant standard deviation (see model 3
in Table 3). As expected, CVs for climate amenities differ significantly across individuals.
For example, some individuals view higher winter temperature as an amenity, while others
view it as a disamenity. In average, individuals should be compensated by approximately

10
  The unconditional distributions were computed using 10,000 draws from the estimated distribution.
11
  For more details about the CV distribution, please see Daly et al. (2012).

13
G. Ahumada et al.

$508 USD for one increase in 1 °C in the coldest month to keep the same level of SWB.
Similarly, the preferences for annual precipitation show a high degree of heterogeneity. For
example, individuals with strong preferences for rain should be compensated by approxi-
mately $103 for a 1-mm decrease in annual precipitation; however, for most workers rainy
days is considered as a disamenity.
With respect to man-made amenities, we find that all individuals in the sample consider
that the presence of robberies around the place where they reside is a disamenity, although
their monetary evaluations vary across the sample. Some individuals living in places with
robberies should be compensated with $19.4 monthly, while others in the right tail with
approximately $688.2. It is important to emphasize that these values are quite distant from
the median that approximately corresponds to $115. Finally, some individuals view the
presence of park as an amenity, while others view it as a disamenity. If the number of
parks in the city decreases by one unit, then an average individual should be compensated
by approximately $105 (ceteris paribus), which is very close to the median.
Important findings emerged from this Section. First, when the coefficients of city-spe-
cific characteristics (amenities) vary across individuals, we identify interesting patterns of
preferences for amenities. We have found that the same city-characteristic can be perceived
as an amenity, whereas for other it can represent a disamenity. Amenities that are perceived
by (dis)amenity for the whole sample, we find that there are some individuals that should
be compensated for much less, while for other individuals more compensation is required.
In other words, a given amenity might have different monetary evaluation across individ-
uals. The application of a standard discrete choice model (fixed coefficients) masks this
remarkable individual variation and led to broad generalizations ignoring these individual
differences in terms of CVs.

6 Discussion and Conclusions

The literature has established that locational factors affect individual well-being. However,
the empirical evidence for developing countries is mainly based on the analysis of housing
and labor markets. This paper tests directly the association between location-specific amen-
ities and individual SWB across Chilean cities using a life-satisfaction approach, which
enables us to obtain a preliminary approximation of workers’ WTP. Importantly, this study
goes further by analyzing how unobserved tastes for amenities can help us to reach a better
understanding of the relationship between urban and natural (dis)amenities and the level of
welfare that cities can provide to Chilean population.
Our first set of results suggests the existence of subjective well-being differences at
the city level. Specifically, we identify that in low populated cities with more amenable
weather conditions, and lower levels of robberies and pollution individuals enjoy from
higher levels of life satisfaction. Although these results are interesting, they represent aver-
age measures where unobserved heterogeneity is not considered.
Using the random parameter approach, we characterize the whole distribution of the
relationship between amenities and individual SWB across Chilean cities. In doing so,
our study uncovers several interesting results. The analysis supports the fact that there
exists substantial heterogeneity in the relationship between individual SWB and some
city-specific characteristics for the Chilean case. In this regard, we identified workers

13
We Do Not Have the Same Tastes! Evaluating Individual…

with different willingness to pay values for temperature seasonality, precipitation and
the number of parks. As an example, while some individuals require a positive mon-
etary compensation for an increase of precipitation (disamenity), others need to be com-
pensated for a decrease of precipitation (amenity). With respect to robberies, although
for the whole sample is considered a disamenity, the required compensation differs sig-
nificantly across individuals. In sum, that there are unobserved differences in tastes for
amenities that might either overstate or understate the estimates of willingness to pay
(WTP).
In our opinion, this work has relevance for several reasons. First, more than the esti-
mates per se, this study provides valuable empirical evidence suggesting that urban ameni-
ties represent an additional element to be considered in developing countries that affect
population well-being. This information can help scholars and policy makers to draw
more accurately conclusions about the impact of amenities on subjective well-being. For
example, policy makers regularly make decisions about the provision of public services
to improve living standards for urban populations. Making such decisions is a challeng-
ing task, because many of such amenities do not have a market price. Without appropriate
valuations, such decisions will be more difficult to implement. Despite the fact that some
amenities cannot be modified by policy makers (such as weather amenities), others such
as population, crime, the number of parks and different types of pollution can be modified
through policies aimed at reducing the negative externalities associated to urban growth
and fostering the benefits associated to natural amenities tied to cities. In this regard, we
show that at least for the Chilean population, local policies should be focused on reducing
robberies and noise and waste pollution, since they are the amenities that, on average, are
the most valued by individuals. Second, our results confirm that when analyzing compen-
sating variations for amenities explicit recognition of individual heterogeneity theoretical
models, applied research and policies is required. In this context, policy makers need to
consider that the development of cites as well as migration movements might be consid-
erably influenced by unobserved tastes for city amenities, which is information valuable
enough to understand the current pattern of urban development in both developed and
developing countries.

Acknowledgements  The authors acknowledge and appreciate the wise comments of two anonymous refer-
ees and the financial support of CONICYT/Chilean Fondecyt 11170018 “City differences in the return to
schooling”.

Appendix

See Table 5.

13

Table 5  Marginal effects on the probability of reporting a specific level of life satisfaction

SWB = 1 SWB = 2 SWB = 3 S4B = 4 SWB = 5

ME SE ME SE ME SE ME SE ME SE

13
Log (wage) − 9.67e−03*** 0.0007 − 1.88e−02*** 0.0012 − 6.03e−02*** 0.0034 − 1.09e−02*** 0.0012 9.97e−02*** 0.0055
Female − 6.59e−04 0.0007 − 7.32e−04 0.0013 − 4.11e−03 0.0041 − 7.32e−04 0.0007 6.78e−03 0.0068
Age 1.54e−04*** 0.0000 2.98e−04*** 0.0001 9.59e−04*** 0.0002 1.73e−04*** 0.0000 − 1.58e−03*** 0.0003
Married − 8.94e−03*** 0.0009 − 1.68e−02*** 0.0015 − 5.19e−02*** 0.0042 − 7.11e−03*** 0.0010 8.48e−02*** 0.0067
Schooling − 7.56e−04*** 0.0001 − 7.56e−04*** 0.0002 − 4.72e−03*** 0.0006 − 8.51e−04*** 0.0001 7.80e−03*** 0.0011
Good health − 3.57e−02*** 0.0041 − 5.30e−02*** 0.0049 − 1.19e−01*** 0.0074 2.51e−02*** 0.0053 1.83e−01*** 0.0108
Indigenous 3.47e−03*** 0.0011 6.54e−03*** 0.0021 2.01e−02*** 0.0061 2.56e−03*** 0.0006 − 3.27e−02*** 0.0097
Household size 6.15e−04*** 0.0002 1.19e−03*** 0.0004 3.84e−03*** 0.0012 6.92e−04*** 0.0002 − 6.34e−03*** 0.0020
Temperature seasonality − 1.15e−06 0.0000 − 2.22e−06 0.0000 − 7.15e−06 0.0001 − 1.29e−06 0.0000 1.18e−05 0.0001
Max. of warmest month (°C) 3.79e−04 0.0002 7.36e−04 0.0005 2.37e−03 0.0015 4.26e−04 0.0003 − 3.91e−03 0.0025
Min. of coldest month (°C) 5.81e−05 0.0003 1.13e−04 0.0005 3.63e−04 0.0017 6.54e−05 0.0003 − 5.99e−04 0.0028
Annual precipitation (mm) 1.04e−06* 0.0000 2.03e−06* 0.0000 6.51e−06* 0.0000 1.17e−06* 0.0000 − 1.08e−05* 0.0000
Log (distance to sea) 1.59e−04 0.0003 3.09e−04 0.0005 9.93e−04 0.0016 1.79e−04 0.0003 − 1.64e−03 0.0026
Parks − 7.00e−05 0.0001 − 1.36e−04 0.0001 − 4.37e−04 0.0003 − 7.88e−05 0.0001 7.22e−04 0.0005
Log (population) 1.20e−03*** 0.0004 2.33e−03*** 0.0007 7.48e−03*** 0.0022 1.35e−03*** 0.0004 − 1.24e−02*** 0.0036
Log (area of the city) 6.39e−05 0.0002 1.24e−04 0.0004 3.99e−04 0.0013 7.19e−05 0.0002 − 6.59e−04 0.0021
Noise pollution 3.25e−03*** 0.0008 6.20e−03*** 0.0014 1.94e−02*** 0.0042 2.92e−03*** 0.0006 − 3.18e−02*** 0.0068
Waste pollution 2.51e−03*** 0.0008 4.79e−03*** 0.0015 1.50e−02*** 0.0045 2.28e−03*** 0.0006 − 2.46e−02*** 0.0072
Robberies 3.41e−03*** 0.0000 6.57e−03*** 0.0000 2.09e−02*** 0.0001 3.55e−03*** 0.0000 − 3.44e−02*** 0.0062
City dummies No No No No No No No
Log (rent) Yes Yes Yes Yes Yes Yes Yes
BIC 48,255.100 46,312.179 46,318.915 46,318.915 46,318.915
AIC 46,216.640 46,116.173 46,115.069 46,115.069 46,115.069
Pseudo R-squared 0.043 0.035 0.036 0.036 0.036
G. Ahumada et al.
Table 5  (continued)

SWB = 1 SWB = 2 SWB = 3 S4B = 4 SWB = 5

ME SE ME SE ME SE ME SE ME SE

Observations 18,774 18,774 18,774 18,774 18,774

The marginal effects (ME) are obtained through the estimation of an ordered probit with fixed parameters. The dependent variable is life satisfaction. Standard errors (SE)
were obtained using the Delta Method
p value for Wald test is in parenthesis. *p < 0.1, **p < 0.05, ***p < 0.001
We Do Not Have the Same Tastes! Evaluating Individual…

13
G. Ahumada et al.

References
Albouy, D., Graf, W., Kellogg, R., & Wolff, H. (2016). Climate amenities, climate change, and American
quality of life. Journal of the Association of Environmental and Resource Economists, 3(1), 205–246.
Ambrey, C., & Fleming, C. (2014). Public greenspace and life satisfaction in urban Australia. Urban Stud-
ies, 51(6), 1290–1321.
Ardeshiri, A., Willis, K., & Ardeshiri, M. (2018). Exploring preference homogeneity and heterogeneity for
proximity to urban public services. Cities, 81, 190–2002. https​://doi.org/10.1016/j.citie​s.2018.04.008.
Benjamin, D. J., Heffetz, O., Kimball, M., & Szembrot, N. (2014). Beyond happiness and satisfac-
tion: Toward well-being indices based on stated preference. American Economic Review, 104(9),
2698–2735.
Benjamin, D. J., Kimball, M. S., Heffetz, O., & Rees-Jones, A. (2012). What do you think would make you
happier? What do you think you would choose? American Economic Review, 102(5), 2083–2110.
Berger, M. C., Blomquist, G. C., & Sabirianova Peter, K. (2008). Compensating differentials in emerging labor
and housing markets: Estimates of quality of life in Russian cities. Journal of Urban Economics, 63(1),
25–55.
Bishop, K., & Timmins, C. (2011). Hedonic prices and implicit markets: Estimating marginal willingness to pay
for large by. NBER working papers, 17611, pp. 1–38.
Blomquist, G. C., Berger, M. C., & Hoehn, J. P. (1988). New estimates of quality of life in urban areas. Ameri-
can Economic Review, 78(1), 89–107.
Boxall, P. C., & Adamowicz, W. (2002). Understanding heterogeneous preferences in random utility models: A
latent class approach. Environmental & Resource Economics, 23(4), 421–446.
Brereton, F., Clinch, J. P., & Ferreira, S. (2008). Happiness, geography and the environment. Ecological Eco-
nomics, 65(2), 386–396.
Brueckner, J., Thisse, J.-F., & Zenou, Y. (1999). Why is central Paris rich and downtown Detroit poor? An
amenity-based theory. European Economic Review, 43(1), 91–107.
Cazzuffi, C., & López-Moreno, D. (2018). Psychosocial wellbeing and place characteristics in Mexico. Health
and Place, 50, 52–64. https​://doi.org/10.1016/j.healt​hplac​e.2018.01.002.
Colombo, E., Michelangeli, A., & Stanca, L. (2014). La Dolce Vita: Hedonic estimates of quality of life in Ital-
ian cities. Regional Studies, 48(8), 1404–1418.
Cruces, G., Ham, A., & Tetaz, M. (2010). Well-being at the subcity level: The buenos aires neighborhood qual-
ity of life survey. In E. Lora, A. Powell, B. van Praag, & P. Sanguinetti (Eds.), The quality of life in Latin
America cities: Markets and perception (pp. 91–115). Washington, DC: Inter-American Development and
World Bank.
Cuñado, J., & de Gracia, F. P. (2013). Environment and happiness: New evidence for Spain. Social Indicators
Research, 112(3), 549–567.
Daly, A., Hess, S., & Train, K. (2012). Assuring finite moments for willingness to pay in random coefficient
models. Transportation, 39(1), 19–31.
Di Tella, R., & MacCulloch, R. (2006). Some uses of happiness data in economics. Journal of Economic Per-
spectives, 20(1), 25–46.
Diener, E., Oishi, S., & Lucas, R. E. (2005). Subjective well-being: The science of happiness and life satisfac-
tion. In S. Lopez & C. Snyder (Eds.), Handbook of positive psychology (2nd ed., pp. 63–73). New York:
Oxford University Press.
Falco, P., Maloney, W. F., Rijkers, B., & Sarrias, M. (2015). Heterogeneity in subjective wellbeing: An applica-
tion to occupational allocation in Africa. Journal of Economic Behavior & Organization, 111(1), 137–153.
Ferreira, S., Akay, A., Brereton, F., Cuñado, J., Martinsson, P., Moro, M., et al. (2013). Life satisfaction and air
quality in Europe. Ecological Economics, 88(1), 1–10.
Ferreira, S., & Moro, M. (2010). On the use of subjective well-being data for environmental valuation. Environ-
mental & Resource Economics, 46(3), 249–273.
Ferrer-i Carbonell, A., & Frijters, P. (2004). How important is methodology for the estimate of the determinants
of happiness? The Economic Journal, 114(1997), 641–659.
Frey, B. S., Luechinger, S., & Stutzer, A. (2009). The life satisfaction approach to valuing public goods: The
case of terrorism. Public Choice, 138(3), 317–345.
Frey, B. S., Luechinger, S., & Stutzer, A. (2010). The life satisfaction approach to environmental valuation.
Annual Review of Resource Economics, 2(1), 139–160.
Gabriel, S. A., & Rosenthal, S. S. (1999). Location and the effect of demographic traits on earnings. Regional
Science and Urban Economics, 29(4), 445–461.
Gandelman, N., Piani, G., & Ferre, Z. (2012). Neighborhood determinants of quality of life. Journal of Happi-
ness Studies, 13(3), 547–563.

13
We Do Not Have the Same Tastes! Evaluating Individual…

Glaeser, E. L., & Gottlieb, J. D. (2009). The wealth of cities: Agglomeration economies and spatial equilibrium
in the United States. Journal of Economic Inequality, 47(4), 983–1028.
Glaeser, E. L., Kolko, J., & Saiz, A. (2001). Consumer city. Journal of Economic Geography, 1(1), 27–50.
Goetzke, F., & Islam, S. (2017). Testing for spatial equilibrium using happiness data. Journal of Regional Sci-
ence, 57(2), 199–217.
Graves, P. E. (1983). Migration with a composite amenity: The role of rents. Journal of Regional Science, 23(4),
541–546.
Greene, W., Harris, M. N., Hollingsworth, B., & Weterings, T. A. (2014). Heterogeneity in ordered choice mod-
els: A review with applications to self-assessed health. Journal of Economic Surveys, 28(1), 109–133.
Hajivassiliou, V. A., & Ruud, P. A. (1994). Classical estimation methods for LDV models using simulation. In
R. F. Engle & D. L. McFadden (Eds.), Handbook of econometrics (Vol. 4, pp. 2383–2441). Amsterdam:
Elsevier.
Hensher, D. A., & Greene, W. H. (2003). The mixed logit model: The state of practice. Transportation, 30(2),
133–176.
Hess, S. (2014). Impact of unimportant attributes in stated choice surveys. European Journal of Transport and
Infrastructure Research, 14(4), 349–361.
Hess, S., Bierlaire, M., & Polak, J. (2005). Estimating value-of-time using mixed logit models. In 84th Annual
meeting of the transportation research board, Washington DC.
Iturra, V. (2018). Amenity decomposition: The role played by firms and workers in explaining spatial wage
differences in Chile. Tijdschrift voor Economische en Sociale Geografie, 109(4), 542–560. https​://doi.
org/10.1111/tesg.12311​.
Krupka, D. J. (2009). Location-specific human capital, location choice and amenity demand. Journal of
Regional Science, 49(5), 833–854.
Lee, L. (1992). On efficiency of methods of simulated moments and maximum simulated likelihood estimation
of discrete response models. Econometric Theory, 8(4), 518–552.
Levinson, A. (2012). Valuing public goods using happiness data: The case of air quality. Journal of Public Eco-
nomics, 96(9–10), 869–880.
Luechinger, S. (2009). Valuing air quality using the life satisfaction approach. The Economic Journal, 119(536),
482–515.
Maddison, D. (2001). The amenity value of the global climate. New York: Taylor & Francis.
Maddison, D., & Bigano, A. (2003). The amenity value of the Italian climate. Journal of Environmental Eco-
nomics and Management, 45(2), 319–332.
Medina, C., Morales, L., & Núñez, J. (2010). Quality of Life in Urban Neighborhoods of Bogotá and Medellín,
Colombia. In E. Lora, A. Powell, B. van Praag, & P. Sanguinetti (Eds.), The quality of life in Latin Amer-
ica cities: Markets and perception (pp. 117–159). Washington, DC: Inter-American Development Bank
and World Bank.
Mentzakis, E. (2011). Allowing for heterogeneity in monetary subjective well-being valuations. Health Eco-
nomics, 20(3), 331–347.
Partridge, M. D. (2010). The duelling models: NEG vs amenity migration in explaining US engines of growth.
Papers in Regional Science, 89(3), 513–536.
Perez-Truglia, R. (2015). A Samuelsonian validation test for happiness data. Journal of Economic Psychology,
49, 74–83. https​://doi.org/10.1016/j.joep.2015.05.002.
Perez-Truglia, R. (2016). The effects of income transparency on well-being-evidence from a natural experiment.
Available at SSRN: https​://ssrn.com/abstr​act=26578​08.
Rappaport, J. (2007). Moving to nice weather. Regional Science and Urban Economics, 37(3), 375–398.
Rehdanz, K. (2006). Hedonic pricing of climate change impacts to households in Great Britain. Climatic
Change, 74(4), 413–434.
Rehdanz, K., & Maddison, D. (2005). Climate and happiness. Ecological Economics, 52(1), 111–125.
Roback, J. (1982). Wages, rents, and the quality of life. The Journal of Political Economy, 90(6), 1257–1278.
Rosen, S. (1979). Wage-based indexes of urban quality of life. In P. Mieskowski & M. Stratzhein (Eds.), Current
issues in urban economics (pp. 74–104). Baltimore: Johns Hopkins Press.
Sarrias, M. (2016). Discrete choice models with random parameters in R: The Rchoice package. Journal of
Statistical Software, 74(10), 1–31.
Sarrias, M. (2018). Do monetary subjective well-being evaluations vary across space? Comparing con-
tinuous and discrete spatial heterogeneity. Spatial Economic Analysis. https​://doi.org/10.1080/17421​
772.2018.14859​68.
Schurer, S., & Yong, J. (2016). Happiness, income and heterogeneity. Singapore Economic Review, 61(3), 1–23.
Scott, J. (1997). Regression models for categorical and limited dependent variables. Thousand Oaks: SAGE
Publications Inc.
Train, K. E. (2009). Discrete choice methods with simulation. New York: Cambridge University Press.

13
G. Ahumada et al.

van Praag, B. M., & Baarsma, B. E. (2005). Using happiness surveys to value intangibles: The case of airport
noise. Economic Journal, 115(500), 224–246.
Veenhoven, R. (2000). The four qualities of life. Journal of Happiness Studies, 1(1), 1–39.
Veenhoven, R. (2017). Greater happiness for a greater number: Did the promise of enlightenment come true?
Social Indicators Research, 130(1), 9–25.
Welsch, H., & Ferreira, S. (2014). Environment, well-being, and experienced preference. International Review
of Environmental and Resource Economics, 7(4), 205–239.
Welsch, H., & Kühling, J. (2009). Using happiness data for environmental valuation: Issues and applications.
Journal of Economic Surveys, 23(2), 385–406.
Winters, J. V., & Li, Y. (2016). Urbanisation, natural amenities and subjective well-being: Evidence from US
counties. Urban Studies, 54(8), 1956–1973.

Publisher’s Note  Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.

13

You might also like